Hertfordshire Chalk PDF

The hydrogeology of bromate contamination
in the Hertfordshire Chalk: double-porosity

effects on catchment-scale evolution
Ciara Marie Fitzpatrick
A dissertation submitted in partial fulfilment

of the requirements for the degree of
Doctor of Engineering
of
University College London.
Department of Earth Sciences

University College London
2010
2
I, Ciara M. Fitzpatrick, confirm that the work presented in this thesis is my own. Where information
has been derived from other sources, I confirm that this has been indicated in the thesis.
3
Foreword
This Engineering Doctorate (EngD) was funded by an Engineering and Physical Sciences Research
Council (EPSRC) Studentship in association with Veolia Water Three Valleys Ltd1 (VWTV) and Thames
Water Utilities Ltd (TWUL).
As part of the research project two EngD studentships were awarded: (1) to Ciara Fitzpatrick, orig-
inator of this document, and (2) to Simon Cook. The research project was intended to be collaborative
between the water utilities and University College London, with neither individual student affiliated
preferentially to either company.
The work documented within this thesis constitutes independent original research, however the
research described is complementary to that by Simon Cook in:
The hydrogeology of Bromate Contamination in the Hertfordshire Chalk Aquifer: In-

corporating Karst in Predictive Models. EngD Thesis, University College London, 2010.
The views, opinions and conclusions expressed in the thesis are those of the researcher and do not
necessarily reflect the views of Veolia Water Three Valleys Ltd or Thames Water Utilities Ltd (TWUL).
1 formerly Three Valleys Water Ltd (TVW)

4
Abstract
Bromate contamination over an area of more than 40 km2 in the Hertfordshire Chalk aquifer was first
detected in 2000 and is the largest case of point-source groundwater contamination in the UK. Bromate
is a possible human carcinogen, and a regulatory limit for drinking water of 10 µg l−1 had been imple-
mented in the U.K. since 2003. Background concentrations of bromate in groundwater are believed to
be effectively zero. In the affected area, bromate at concentrations of several 100 µg l−1 have forced
the closure of a large public water supply source and restricted the use of seven other public supply
boreholes up to 20 km from the contamination source.
The source has been identified as a former industrial site which operated between 1955 and 1983.
Residual contamination at the site provides a continuing source of bromate to groundwater. A range
of conceptual scenarios for bromate mobilisation and release to groundwater have been developed and
quantified based on interpretation of the available data, and constrained by interpolation of the observed
concentrations.
Analysis and interpretation of all available monitoring and investigation data throughout the catch-
ment has revealed the influence of recharge, water level, and groundwater abstractions on bromate con-
centrations. These relationships, integrated with observations of the geology and hydrogeology of the
area, support a conceptualisation of transport of bromate by dominantly double-porosity processes within
the Vale of St. Albans area, which maintains a highly attenuated, stable contaminant distribution west of
Hatfield. An extensive karst network related to the position of the Palaeogene overlap of the Chalk influ-
ences bromate transport to the east of Hatfield, dispersing bromate rapidly over large distances toward the
Lea Valley. The revised conceptual understanding has enabled the development of a new interpretation
of bromate transport within the catchment between 2000 and 2008.
A new analytical network modelling approach has been developed to predict the long-term, large-
scale transport of bromate. The model simulates Fickian double-porosity diffusive exchange along in-
terconnecting flow-lines, linked to rapid karst flow. The model is parameterised on the basis of single
borehole dilution testing, catchment-scale natural gradient tracer testing, and literature derived values.
The network model, combined with quantified bromate source terms, simulates bromate and bromide
concentrations of the order of magnitude of those observed at locations within the Vale of St. Albans, and
predicts bromate concentrations to remain above regulatory limits for around 200 years. This highlights
the importance of double-porosity diffusion for the long-term evolution of contaminants at catchment-
scale in the Chalk aquifer.
5
Acknowledgements
I am sincerely grateful to my supervisors, Dr Willy Burgess and Prof. John Barker for providing inspi-
ration, support and guidance over the past four years, and above all, for believing in my capabilities. I
am indebted to Willy for his constant enthusiasm, and ability to renew mine.
This Engineering Doctorate was funded by an EngD studentship at UCL and I am grateful to EP-
SRC, Veolia Water Three Valleys Ltd, and Thames Water Utilities Ltd for sponsoring the project. In
particular, I would like to thank my supervisors Rob Sage, Lucy Lytton, and Philip Bishop for their
insights and information.
I owe my gratitude to Jon Newton at the Environment Agency, firstly for introducing me to the
Hertfordshire bromate problem, and subsequently for sharing his understanding, providing monitoring
data and a wealth of other relevant information.
The single borehole dilution tests were undertaken with assistance from a number of people: many
thanks to the Environment Agency for permission and funding, Adrian Sheriff for access to the Nashe’s
Farm borehole, Simon Cook and Gemma Russell for help with the fieldwork, Louise Maurice for guid-
ance on the method and use of the equipment, and Jessica Randle for associated geophysical testing of
the boreholes.
I am grateful to Simon Cook, for his contributions to understanding and progressing research into
the bromate contamination. Thanks also to Rakia Meister, Simon Quinn, Mohammed Abdul Hoque,
Qiong Li, Mike Davis, Bethan Hallett and Gemma Russell for putting up with sharing an office with me,
and for some welcome conversation.
‘Go raibh maith agat’ to my fiancé, Neal O’Grady, for his love, understanding and patience. Finally,
I thank my family; without their love and support over the last 29 years I would not be in a position to
be submitting this thesis. I dedicate this thesis to my grandparents, Mona and Chris Fitzpatrick.
6
Abbreviations
BH Borehole
BGS British Geological Survey
EA Environment Agency
MDL Method Detection Limit
NNR Northern New River
OBH Observation Borehole
P.S. Pumping Station
PWS Public Water Supply
STW Sewage Treatment Works
SMD Soil Moisture Deficit
SLC St. Leonard’s Court
TWUL Thames Water Utilities Limited
TVW Three Valleys Water Limited (renamed Veolia Water Three Valleys Limited in 2009)
Contents 7
Contents
1 Introduction 22
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.2 Bromate transport in the Hertfordshire Chalk aquifer . . . . . . . . . . . . . . . . . . . 24
1.3 The bromate source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.4 Research aims and objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.5 Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.6 Structure of thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.7 Environmental Hydrochemistry of Bromate and Bromide . . . . . . . . . . . . . . . . . 27
1.7.1 Occurrence of Bromate and Bromide in surface and groundwaters . . . . . . . . 27
1.7.2 Environmental Behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2 Groundwater flow and transport in the Chalk 30

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.2 The Chalk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.2.1 Chalk stratigraphy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.2.2 Lithology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.2.3 Tectonic History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2.4 Influence of periglaciation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.3 The Chalk as an aquifer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.4 Karstic behaviour of the Chalk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.4.1 Geomorphological Evidence of Chalk Karst . . . . . . . . . . . . . . . . . . . . 34
2.4.2 Evidence of rapid flow rates from tracer tests in the Chalk . . . . . . . . . . . . 35
2.4.3 Development of permeability within the Chalk . . . . . . . . . . . . . . . . . . 37
2.5 Hierarchy in the Chalk aquifer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.6 Porosity components of the Chalk aquifer . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.7 Permeability components of the Chalk aquifer . . . . . . . . . . . . . . . . . . . . . . . 42
2.7.1 Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.7.2 Bulk transmissivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.7.3 Permeability components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.8 Groundwater Flow in the Chalk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Contents 8
2.9 Solute transport in the saturated zone of the Chalk . . . . . . . . . . . . . . . . . . . . . 50

2.9.1 Advection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2.9.2 Adsorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2.9.3 Dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.9.4 Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
2.10 Flow and transport in the unsaturated zone of the Chalk . . . . . . . . . . . . . . . . . . 54
2.11 Modelling flow and transport in fissured rocks . . . . . . . . . . . . . . . . . . . . . . . 55
2.11.1 Equivalent Porous Medium (EPM) models . . . . . . . . . . . . . . . . . . . . 56
2.11.2 Double-porosity (DP) models . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.11.3 Double-permeability Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.11.4 Network Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.12 Diffusion exchange model for solute transport in fissured porous rocks . . . . . . . . . . 57
2.12.1 Block geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.12.2 Porosity Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.12.3 Characteristic Times . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.13 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3 A conceptual model for flow and transport of bromate in the Hertfordshire Chalk 62
3.1 Chapter Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.2 Geology and Hydrogeology of Hertfordshire . . . . . . . . . . . . . . . . . . . . . . . . 62
3.2.1 Topography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.2.2 Hydrology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.2.3 Geology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.2.4 Hydrogeology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.2.5 Karstic Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.2.6 Karst Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.2.7 Groundwater–surface water interactions . . . . . . . . . . . . . . . . . . . . . . 77
3.2.8 Chalk–Drift Groundwater interactions . . . . . . . . . . . . . . . . . . . . . . . 78
3.3 Piezometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.3.1 Groundwater flow in the Sandridge-St Leonard’s Court Area . . . . . . . . . . . 83
3.3.2 Abstractions and Discharges . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
3.4 Regional hydrochemisty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
3.5 Scavenge Pumping at Hatfield Pumping Station . . . . . . . . . . . . . . . . . . . . . . 85
3.5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.5.2 Data sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
3.5.3 Data handling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
3.5.4 Abstraction rates at Hatfield . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
3.5.5 Bromate and Bromide time series trends . . . . . . . . . . . . . . . . . . . . . . 86
3.5.6 Bromate and Bromide: Relationship to Hatfield abstraction rates . . . . . . . . . 100
Contents 9
3.5.7 Statistical relationships and bromate transport in Hertfordshire . . . . . . . . . . 107

3.6 Single borehole dilution testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
3.6.1 Calculation of horizontal specific discharge (Darcy Velocity) . . . . . . . . . . . 110
3.6.2 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
3.7 Conceptual Model for groundwater flow in Hertfordshire . . . . . . . . . . . . . . . . . 113
3.8 Summary and conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
4 The evolution of bromate contamination in the Hertfordshire Chalk 117

4.1 Chapter Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
4.2 Bromate Water Quality Monitoring Programme . . . . . . . . . . . . . . . . . . . . . . 117
4.3 Monitoring Data Quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
4.3.1 Sampling Locations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
4.3.2 Sampling Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
4.3.3 Analytical methods and detection limits . . . . . . . . . . . . . . . . . . . . . . 120
4.3.4 Sampling frequency and completeness . . . . . . . . . . . . . . . . . . . . . . . 121
4.4 Delineating the Bromate ‘Plume’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
4.4.1 Up-gradient of the source site . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
4.4.2 Source site and Sandridge area . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
4.4.3 Hatfield Quarry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
4.4.4 Hatfield area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
4.4.5 Lea Valley (east of Hatfield area) . . . . . . . . . . . . . . . . . . . . . . . . . . 150
4.4.6 Bromide spatial distribution in groundwaters . . . . . . . . . . . . . . . . . . . 158
4.5 Bromate-bromide ratios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
4.6 Bromate concentrations and water levels . . . . . . . . . . . . . . . . . . . . . . . . . . 162
4.7 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
5 The Bromate Source 167

5.1 Chapter objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
5.2 Chapter structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
5.3 Site History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
5.3.1 Sources of information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
5.3.2 General overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
5.3.3 Site investigation and remediation history . . . . . . . . . . . . . . . . . . . . . 169
5.3.4 Operational activities of the chemical works . . . . . . . . . . . . . . . . . . . . 169
5.4 Chronology and scope of investigations . . . . . . . . . . . . . . . . . . . . . . . . . . 169
5.5 Site Geology and Hydrogeology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
5.6 Contaminant Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
5.6.1 Spatial distribution of bromate and bromide within soil and soil porewater . . . . 174
5.6.2 Spatial distribution of bromate and bromide within groundwater . . . . . . . . . 194
Contents 10
5.6.3 Groundwater monitoring in the vicinity of the source site . . . . . . . . . . . . . 196

5.6.4 Relationships between contaminant concentrations and water levels . . . . . . . 196
5.6.5 Leachate results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
5.7 Mass of bromide and bromate at the source site . . . . . . . . . . . . . . . . . . . . . . 202
5.7.1 Previous estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
5.7.2 New estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
5.8 Mass flux of bromate in groundwater migrating from the source site . . . . . . . . . . . 207
5.9 Previous representations of the ‘Source Term’ . . . . . . . . . . . . . . . . . . . . . . . 208
5.9.1 Early assessments: 1984 and 1985 . . . . . . . . . . . . . . . . . . . . . . . . . 208
5.9.2 Recent assessments: 2002 to 2008 . . . . . . . . . . . . . . . . . . . . . . . . . 208
5.9.3 CONSIM modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
5.9.4 MT3D modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
5.10 New Conceptual Models for Contaminant Release . . . . . . . . . . . . . . . . . . . . . 210
5.10.1 Mechanisms of bromide and bromate release . . . . . . . . . . . . . . . . . . . 212
5.11 Source terms for bromide and bromate release from the source site . . . . . . . . . . . . 216
5.11.1 General equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
5.11.2 Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
5.11.3 Bromide mass between 1985 and 2001 . . . . . . . . . . . . . . . . . . . . . . 219
5.11.4 Scenario A: Catastrophic Leak + Recharge Pulse . . . . . . . . . . . . . . . . . 219
5.11.5 Scenario B: Steady Seepage + Recharge Pulse . . . . . . . . . . . . . . . . . . . 222
5.11.6 Scenario C - Late stage Seepage + Recharge Pulse . . . . . . . . . . . . . . . . 226
5.11.7 Bromate flux in 2001 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
5.12 Verifying source terms with observed down-gradient concentrations . . . . . . . . . . . 226
5.12.1 Fracture Spacing 2b . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
5.12.2 Fracture Aperture a . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
5.12.3 Fracture Porosity θm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
5.12.4 Matrix Porosity Φ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
5.12.5 Fracture Velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
5.12.6 Effective Diffusion Coefficient DE . . . . . . . . . . . . . . . . . . . . . . . . 230
5.12.7 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
5.13 Summary and conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
6 Multiple Analytical Pathways Approach 238

6.1 Chapter Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
6.2 Previous modelling approaches for Bromide and Bromate in the Chalk . . . . . . . . . . 238
6.2.1 Early model assessments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
6.2.2 Pollutant Linkage Assessment using CONSIM . . . . . . . . . . . . . . . . . . 239
6.2.3 One-dimensional analytical model DP1D . . . . . . . . . . . . . . . . . . . . . 239
6.2.4 Dispersion modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
Contents 11
6.2.5 Catchment-scale distributed flow modelling using MODFLOW and MT3D . . . 240
6.2.6 Weaknesses of MODFLOW and MT3D . . . . . . . . . . . . . . . . . . . . . . 242
6.3 Development of a Multiple Analytical Pathways Approach . . . . . . . . . . . . . . . . 245
6.3.1 DP1-D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
6.3.2 Multiple Analytical Pathways (MAP) model . . . . . . . . . . . . . . . . . . . 245
6.3.3 GoldSim Contaminant Transport Model . . . . . . . . . . . . . . . . . . . . . . 248
6.3.4 Comparison of DP1D, MAP and GoldSim . . . . . . . . . . . . . . . . . . . . . 248
6.4 Analytical Network Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
6.4.1 Mathematical Basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
6.4.2 Node Input - the Source Function . . . . . . . . . . . . . . . . . . . . . . . . . 249
6.4.3 Node and Branch description . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
6.4.4 Node and Branch Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
6.5 A Network Model for Hertfordshire . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
6.5.1 Selection of Nodes and Branches . . . . . . . . . . . . . . . . . . . . . . . . . 250
6.5.2 Parameters for ‘double-porosity’ branches . . . . . . . . . . . . . . . . . . . . . 250
6.5.3 Parameters for karst branches . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
6.6 Network Model for Hertfordshire - Results of initial simulations . . . . . . . . . . . . . 254
6.7 Discussion and conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
7 Conclusions 268
7.1 Fulfillment of research aims and objectives . . . . . . . . . . . . . . . . . . . . . . . . 268
7.1.1 Evolution of bromate contamination . . . . . . . . . . . . . . . . . . . . . . . . 268
7.1.2 The source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
7.1.3 Catchment-scale modelling of bromate transport . . . . . . . . . . . . . . . . . 270
7.2 Recommendations for further work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
Bibliography 272
Appendices 281
A Chronology of key events 281
B Business Case for the research 282
C Hatfield Pumping Trial statistical analyses 283
D Single Borehole Dilution Testing 284
E MAP, GoldSim and DP1D comparison 285
F Parameters for Hertfordshire Network Model 286

List of Tables 12
List of Tables
1.1 Bromide in UK groundwaters, summarised from Edmunds et al. (1989). r2 is the linear
correlation coefficient squared for a regression of Br vs. Cl . . . . . . . . . . . . . . . . 28
2.1 Molecular diffusion coefficients in fissured and unfissured chalk. After Hill (1984).
These values represent the mass flux through the saturated matrix per unit concentra-
tion gradient in the water. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.2 Characteristic times for infinite slab geometry, with slabs of thickness 2b separated by
fractures of aperture a. For this model, the ratio of volume to area for a matrix block (`)
is represented by b. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.1 Lithostratigraphy of the Hertford district. After Bloomfield et al. (2004). . . . . . . . . . 67

3.2 Lithostratigraphy of the Chalk of the Hertford district. Based on Woods (2003). . . . . . 68
3.3 Interpretation of lineaments. Based on Bloomfield et al. (2004). . . . . . . . . . . . . . 70
3.4 Chalk Group Aquifer Potential. After Mortimore et al. (1990). . . . . . . . . . . . . . . 72
3.5 Pearson correlation coefficients for Bromate concentration and Soil Moisture Deficit
(SMD) before the start of the Hatfield pumping trial on 29th July 2005. SMD-X cor-
responds to the SMD value X months previously. Shaded cells indicate the strongest
relationship. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
3.6 Summary of regression parameters for the ‘best-fit’ regressions for the response of bro-
mate concentration to Hatfield abstraction rate. . . . . . . . . . . . . . . . . . . . . . . 105
3.7 Coefficients determined by the ‘best-fit’ regressions for the response of bromate concen-
tration to Hatfield abstraction rate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
4.1 Typical analytical suite for water samples May 2000 to December 2008 . . . . . . . . . 120
4.2 Analytical methodology and detection limits for bromate analyses. . . . . . . . . . . . 120
4.3 Regression statistics for the response of bromide concentration to bromate concentration
and chloride concentration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
5.1 Chronology and scope of site investigations and monitoring at the source site . . . . . . 171
5.2 Geological strata encountered at the source site. Based on Komex (2000) and Atkins
(2002) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
List of Tables 13
5.3 Summary of mass estimates. Estimates for total mass in the unsaturated and saturated
zones refer to minimum, mean and maximum thicknesses defined in Figure 5.31. . . . . 207
5.4 Main areas of uncertainty in the history of bromide and bromate release to groundwater
beneath the source site. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
5.5 Mass predicted by source history scenarios A and B compared to observed mass con-
straints. Condition 4 is based on an estimate by Buckle (2002) of the mass removed at
Hatfield and Essendon between 1981 and 2000. . . . . . . . . . . . . . . . . . . . . . . 222
6.1 Parameter combinations for ‘best-case’ (lowest peak bromate concentrations) and
‘worst-case’ (highest peak bromate concentrations) scenarios. . . . . . . . . . . . . . . 253
6.2 Parameters derived from fitting the DP-1D model (Barker, 2005) to tracer breakthrough
curves from Water End injection (Cook, 2010). Characteristic times are in hours. . . . . 254
List of Figures 14
List of Figures
1.1 Extent of the bromate contamination in Hertfordshire. The regulatory limit for bromate
in drinking water is 10 µg l−1 . Background concentrations are effectively zero. . . . . . 23
1.2 Plot of Br vs. Cl concentrations in groundwaters from the Chalk of the Colne and Lee
River catchments. Points plotting in the ‘contamination’ box are from the Hatfield area.
From Shand et al. (2003). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.1 Relationship between the traditional and revised stratigraphy of the Chalk. After
Woods and Aldiss (2004) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.2 (a) An idealised double-porosity aquifer; (b) an idealised double-permeability aquifer.
From Price et al. (1993) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.3 Relationsip between fissure spacing, aperture, porosity and hydraulic conductivity for
a fissure system containing three plane, parallel, mutually perpendicular smooth-walled
fissures filled with pure water at 10 degC and porosity relationship (Price et al., 1993). . 44
2.4 Ranges of fracture spacings and fracture apertures for the Chalk. Results as cited in
Bloomfield (1996) and Watson (2004). . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.5 Double-porosity diffusive exchange of solutes. At an early stage, diffusion from the con-
taminated fracture water into the matrix water acts to retard the transport of contaminants
down-gradient. At a later stage, contaminated porewater acts as a persistent secondary
source of contamination. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.6 Governing equations and assumptions for a double-porosity mocel with slab geometry.
After Barker (1982). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.1 Location of study area, including topography and hydrology . . . . . . . . . . . . . . . 63

3.2 Solid Geology of the Study Area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.3 Solid and Drift Geology of the Study Area . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.4 Established tracer connections in Hertfordshire. After Cook (2010) . . . . . . . . . . . . 74
3.5 Environment Agency monitoring network long-term water level monitoring locations . . 79
3.6 Rainfall and Environment Agency monitoring network water level variations. Locations
of monitoring wells are shown in Figure 3.5 . . . . . . . . . . . . . . . . . . . . . . . . 80
3.7 Water Level, Soil Moisture Deficit and rainfall at Orchard Garage Monitoring Well . . . 81
List of Figures 15
3.8 Average piezometry 1998 to 2008. Contour levels are in m AOD. Arrows indicate
groundwater flow direction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
3.9 Abstraction rates at Hatfield PS between 31 June 2005 and 31 December 2008 . . . . . . 87
3.10 Time series of bromate and bromide concentrations at Hatfield PS, soil moisture deficit,
and monthly rainfall. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.11 Time series of bromate and bromide concentrations at Essendon PS, soil moisture deficit,
3.12 Time series of bromate and bromide concentrations at Chadwell Spring, soil moisture
deficit, and monthly rainfall. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
3.13 Time series of bromate and bromide concentrations at Amwell Hill PS, soil moisture
3.14 Time series of bromate and bromide concentrations at Amwell Marsh PS, soil moisture
3.15 Time series of bromate and bromide concentrations at Rye Common PS, soil moisture
3.16 Time series of bromate and bromide concentrations at Middlefield Road PS, soil mois-
ture deficit, and monthly rainfall. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.17 Time series of bromate and bromide concentrations at Hoddesdon PS, soil moisture
3.18 Time series of bromate and bromide concentrations at Broxbourne PS, soil moisture
3.19 Time series of bromate and bromide concentrations at Turnford PS, soil moisture deficit,
3.20 Assessment of residuals for each ’best-fit’ regression for the response of bromate con-
centration to Hatfield abstraction rate. (1) . . . . . . . . . . . . . . . . . . . . . . . . . 103
3.21 Assessment of residuals for each ’best-fit’ regression for the response of bromate con-
centration to Hatfield abstraction rate. (2) . . . . . . . . . . . . . . . . . . . . . . . . . 104
3.22 Regression coefficients: means and 95% confidence intervals . . . . . . . . . . . . . . . 108
3.23 Comparison of statistical response times for bromate concentration response to hatfield
abstraction and tracer travel times from Water End. Based on Cook (2010) . . . . . . . . 109
3.24 Methodology for determination of specific discharge (darcy velocity) from the results of
the Single Borehole Dilution Tests. Based on Ward et al. (1998) . . . . . . . . . . . . . 111
3.25 Specific discharge (darcy velocity) for each 0.5 m depth section at Nashes Farm. Es-
Ct −Cb
timated using the methodology in Figure 3.24. Plots of ln C 0 −Cb
are included in Ap-
pendix D. The value at each section is estimated based on Based on Single Borehole
Dilution test carried out at Nashes Farm 29 January 2008. . . . . . . . . . . . . . . . . 112
List of Figures 16
3.26 Specific discharge (darcy velocity) for each 0.5 m depth section at Comet Way BH.
Ct −Cb
Estimated using the methodology in Figure 3.24. Plots of ln C0 −Cb
are included in
Appendix D. The value at each section is estimated based on Based on Single Borehole
Dilution test carried out at Comet Way BH 4 February 2008. . . . . . . . . . . . . . . . 113
3.27 Specific discharge (darcy velocity) for each 0.5 m depth section at Harefield House BH.
Ct −Cb
are included in
Appendix D. The value at each section is estimated based on Based on Single Borehole
Dilution test carried out at Harefield House BH on 22 January 2008. . . . . . . . . . . . 114
3.28 Conceptual model for groundwater flow in the bromate affected area of Hertfordshire.
Position of conduits are based on the conceptual model developed by Cook (2010). Flow
rates and attenuation characteristics are inferred from the results of the single borehole
dilution testing presented in Section 3.6 and tracer tests undertaken by Cook (2010). . . . 115
4.1 All bromate monitoring locations 2000-2008. . . . . . . . . . . . . . . . . . . . . . . . 119

4.2 Sampling frequency for bromate at each monitoring location in 2000 . . . . . . . . . . . 122
4.11 Annual average bromate concentrations at groundwater sampling locations in 2000. . . . 132
4.20 Time series of bromate and bromide concentrations at selected locations between San-
dridge and Hatfield, soil moisture deficit, and monthly rainfall. . . . . . . . . . . . . . . 142
List of Figures 17
4.23 Time series of bromate and bromide concentrations at selected locations in the Hatfield
Quarry area, soil moisture deficit, and monthly rainfall. . . . . . . . . . . . . . . . . . . 146
area, soil moisture deficit, and monthly rainfall. . . . . . . . . . . . . . . . . . . . . . . 151
4.31 Time series of bromate and bromide concentrations at selected locations in the Lea Val-
ley, soil moisture deficit, and monthly rainfall. . . . . . . . . . . . . . . . . . . . . . . . 157
4.32 Annual average Bromide concentrations in groundwater 2000 to 2008. . . . . . . . . . . 159
4.33 Bromide concentrations at locations where bromate concentrations are less than MDL. . 160
4.34 Bromate/Bromide ratio variation with bromate concentration for groundwater and sur-
face water samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
4.35 Spatial distribution of mean annual bromate/bromide ratio 2000 to 2008. . . . . . . . . . 163
4.36 Percentage of samples of bromate concentrations for which there are accompanying wa-
ter level measurements for each location . . . . . . . . . . . . . . . . . . . . . . . . . . 164
4.37 Regression relationship for the response of bromate concentration to water level. Per-
centages refer to the amount of variation explained by the regression (R2 value) . . . . . 165
5.1 Location of the source site in Sandridge, Hertfordshire. Formerly the Steetly chemical
works, now the St Leonard’s Court residential development. . . . . . . . . . . . . . . . 168
5.2 Location of former process areas of the Steetly Chemical Works. Based on Atkins (2002)
interpretation of historical plans, aerial photographs, and the interview with a former
employee of the works. Aerial photograph taken in 1971. . . . . . . . . . . . . . . . . . 170
5.3 Borehole locations from investigations 1983-1985 (STATS, 1983a,b,c, 1984; Chemfix,
1985c) and 2000-2001 (Komex, 2000; Atkins, 2002). For locations from 1983-1985,
numbers in square brackets indicate date of drilling. . . . . . . . . . . . . . . . . . . . . 172
5.4 Trial hole locations from investigations in 1985 (Chemfix, 1985c) . . . . . . . . . . . . 173
5.5 Piezometry at the St Leonard’s Court site November 2001. From Atkins (2002) . . . . . 175
5.6 Cross-section parallel to groundwater flow direction. From Atkins (2002) . . . . . . . . 176
List of Figures 18
5.7 Spatial distribution of the bromide contamination based on investigations undertaken

between 1983-1985. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
5.8 Depth profiles of porewater bromide compared to pumped groundwater concentrations
for boreholes from investigations 1983-1985. . . . . . . . . . . . . . . . . . . . . . . . 179
5.9 spatial distribution of bromide (as mg kg−1 ) based on investigations undertaken between
2000 and 2001. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
5.10 spatial distribution of bromate (as mg kg−1 ) based on investigations undertaken between
2000 and 2001. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
5.11 Depth profiles of porewater bromate and bromide compared to pumped groundwater
concentrations for Borehole 214 from 2001 investigation (Atkins, 2002). . . . . . . . . . 182
5.22 Relationship between soil bromate and soil bromide concentrations based on soil sam-
ples from the 2001 site investigation (Atkins, 2002). . . . . . . . . . . . . . . . . . . . . 193
5.23 Groundwater bromate and bromide contours at the ‘source zone’ based on samples taken
in 2001 and 2002. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
5.24 Depth profiles of porewater bromide compared to pumped groundwater concentrations
for Borehole B1 from the 1985 investigation (Chemfix, 1985a) and Borehole 225 from
2001 investigation (Atkins, 2002) which are believed to have been located in similar
positions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
List of Figures 19
5.25 Relationship between bromide and bromate concentration in groundwater samples

from the monitoring data between 2000 and 2008 for locations 079 to 083 and loca-
tions 214 to 223. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
5.26 Groundwater monitoring locations in the vicinity of the source site that have been sam-
pled for bromide concentrations between 1983 and 1987 and between 2000 and 2008. . . 199
5.27 Groundwater bromide concentrations at monitoring locations in the vicinity of the source
site 1983 to 2008. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
5.28 Relationships between leachate concentration (mg l−1 ) and soil concentration
(mg kg−1 ) for samples from the 2001 investigation (Atkins, 2002). . . . . . . . . . . . . 201
5.29 Bromate soil concentration contours for 1.0 m thick grid slices based on investigation
data from 2000 and 2001 (Komex, 2000; Atkins, 2002). Estimates for total mass in
the unsaturated and saturated zones refer to minimum, mean and maximum thicknesses
defined in Figure 5.31. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
5.30 Bromide soil concentration contours for 1.0 m thick grid slices based on investigation
data from 2000 and 2001 (Komex, 2000; Atkins, 2002). Estimates for total mass in
the unsaturated and saturated zones refer to minimum, mean and maximum thicknesses
defined in Figure 5.31. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
5.31 Minimum and maximum saturated and unsaturated zone thicknesses. . . . . . . . . . . . 206
5.32 Estimates of bromate groundwater flux from the ‘source zone’ using equation 5.2 and
the area under a concentration profile taken across a flux plane through the source zone.
R x=B
The area under a graph represents the integral x=A C dx. The flux plane is shown in
Figure 5.23. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
5.33 The combined ‘source zone’ (centre figure) based on the locations of high concentrations
of bromate (left hand figure) and bromide (right hand figure) in groundwater . . . . . . . 212
5.34 Conceptual Model for bromate and bromide release from the source zone. . . . . . . . . 213
5.35 Derivation of equations for mass of bromide/bromate in the unsaturated zone and the
rate of input of bromide/bromate from the unsaturated zone to the saturated zone. . . . . 218
5.36 Equations for bromide mass, fit to observed values from 1985 and 2001. Parameters are
defined in Figure 5.11.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
5.37 Bromide and bromate concentrations for Scenario A and Scenario B from 1984 into the
future. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
5.38 Bromide source history for Scenarios A and B. . . . . . . . . . . . . . . . . . . . . . . 223
5.39 Bromide source history for Scenarios A and B. . . . . . . . . . . . . . . . . . . . . . . 224
5.40 Bromide and bromate concentrations for Scenario A and Scenario B between 1955 and
1984. After 1984 concentrations proceed as in Figure 5.37. . . . . . . . . . . . . . . . . 225
5.41 Bromate source history for Scenario C. . . . . . . . . . . . . . . . . . . . . . . . . . . . 227

5.42 Conceptual model for off-site verification simulations . . . . . . . . . . . . . . . . . . . 229
List of Figures 20
5.43 Comparison of simulated bromide concentrations for source history Scenario A and ob-
served concentrations at three monitoring locations. . . . . . . . . . . . . . . . . . . . . 232
5.44 Comparison of simulated bromate concentrations for source history Scenario A and ob-
5.45 Comparison of simulated bromide concentrations for source history Scenario B and ob-
5.46 Comparison of simulated bromate concentrations for source history Scenario B and ob-
5.47 Comparison of simulated bromate concentrations for source history Scenario C and ob-
5.48 Concurrent matrix and fissure concentrations are required to determine at which point
along the concentration-time graph a particular fissure concentration represents. . . . . . 237
6.1 Comparison of the superseded versions of the the source terms used by Cook (2010) to
the current versions in this thesis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
6.2 Conceptual and mathematical basis for the Multiple Analytical Pathways of Barker (2001).246
6.3 Conceptual and mathematical basis for the Multiple Analytical Pathways of Barker (2001).247
6.4 Nodes and branches represented in the Network Model for Hertfordshire. Note that
branches are shown schematically as straight-line connectors and are not intended to
indicate the precise geographical route. . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
6.5 Simulated bromate concentrations at Harefield House using source terms for Scenario
A, B and C (Section 5.11), and a constant concentration source term of 5000 µg l−1 . . . 257
6.6 Simulated bromate concentrations at Hatfield Quarry using source terms for Scenario A,
B and C (Section 5.11), and a constant concentration source term of 5000 µg l−1 . . . . . 258
6.7 Simulated bromate concentrations at Comet Way using source terms for Scenario A, B
and C (Section 5.11), and a constant concentration source term of 5000 µg l−1 . . . . . . 259
6.8 Simulated bromate concentrations at Arkley Hole Spring node, and at the end of con-
tributing branches, using source terms for Scenario A, B and C (Section 5.11), and a
constant concentration source term of 5000 µg l−1 . . . . . . . . . . . . . . . . . . . . . 260
6.9 Simulated bromate concentrations at Lynchmill Spring node, and at the end of contribut-
ing branches, using source terms for Scenario A, B and C (Section 5.11), and a constant
concentration source term of 5000 µg l−1 . . . . . . . . . . . . . . . . . . . . . . . . . . 261
6.10 Simulated bromide concentrations at Harefield House using source terms for Scenario A
and B. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
6.11 Simulated bromide concentrations at Hatfield Quarry using source terms for Scenario A
and B. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
6.12 Simulated bromide concentrations at Comet Way using source terms for Scenario A and B.264
6.13 Simulated bromate concentrations for Scenario C at Harefield House, Hatfield Quarry,
and Comet Way using ‘best-case’, ‘typical-case’ and ‘worst-case’ parameters. . . . . . . 265
List of Figures 21
6.14 Concurrent matrix and fissure concentrations are required to determine at which point
along the concentration-time graph a particular fissure concentration represents. . . . . . 266
22
Chapter 1
Introduction
1.1 Background
Changes to the Water Supply (Water Quality) Regulations (2000), effective from 2003, introduced bro-
mate (BrO3– ) as a new sampling parameter with a regulatory limit of 10 µg l−1 in drinking water. Bro-
mate is a possible human carcinogen based on extrapolation from rodent studies. Background concen-
trations of bromate in groundwater are believed to be effectively zero. In May 2000, during the course of
preliminary sampling in advance of the regulations, Three Valleys Water Ltd1 (TVW), detected bromate
concentrations of 135-140 µg l−1 , well in excess of this standard, at the Hatfield Bishop’s Rise Pumping
Station. The Environment Agency (EA) and the Drinking Water Inspectorate (DWI) were informed and
the Hatfield source was removed from public supply.
In June 2000 a joint water quality monitoring programme was initiated, involving the Environment
Agency, the local authorities, and the water companies to identify the source and extent of the bromate
contamination. The source has been identified as a former industrial site in Sandridge (Figure 1.1) which
operated between 1955 and 1983. The site is now the St Leonard’s Court residential development. This
site has been determined as ‘Contaminated Land’ and designated a ‘Special Site’ as defined under Part
IIA of the Environmental Protection Act 1990. The bromate contamination was found to extend up to
20 km to the east of the source site affecting an area of more than 40 km2 of the Hertfordshire Chalk
aquifer (Figure 1.1), and restricting the use of a further seven public water supply boreholes in the Lea
Valley. Bromide (Br – ) concentrations were also found to be elevated above background concentrations
across the catchment. The Environment Agency, TVW and TWUL have continued to monitor water
quality and water levels at a number of locations throughout the area affected by the bromate plume. A
chronology of key events is included in Appendix A.
The Chalk is one of the most important aquifers in the UK; according to the UK Groundwater
Forum2 it provides over half of all groundwater for public supply in the UK and, in the south-east of
England, up to 70 % of all water in public supply. The area of Chalk aquifer affected by the bromate
contamination has a licensed abstraction of approximately 200 Ml d−1 . Surface waters of the River Lea,
and tributaries of the River Colne are also affected. The bromate contamination is therefore a significant
1 renamed in 2009 as Veolia Water Three Valleys Limited
2 www.groundwateruk.org/
1.1. Background 23
Figure 1.1: Extent of the bromate contamination in Hertfordshire. The regulatory limit for bromate in
drinking water is 10 µg l−1 . Background concentrations are effectively zero.
1.2. Bromate transport in the Hertfordshire Chalk aquifer 24
threat to the long-term quality of a number of strategic public water supply sources, and also many private
supply sources. The financial cost to the water companies, which include costs incurred for additional
monitoring of sources, treatment of bromate and bromide contaminated water, operation of the interim
scavange pumping at Hatfield PWS, and drilling and investigatory work for replacement supply wells,
have been estimated to be of the order of in the region of £50,000,000 for TVW and TWUL (R. Sage &
P. Bishop, pers. comm.) for the period 2000 to 2006. A Business Case for the reserach is included in
Appendix B.
1.2 Bromate transport in the Hertfordshire Chalk aquifer

The behaviour of the Chalk as an aquifer is complex, and results from a combination of porosity and
permeability components that are a consequence of the Chalk lithology, tectonic history and weathering
and erosional processes. The Chalk is composed of very fine grained calcium carbonate micro-fossil
fragments which form a a highly porous, yet essentially impermeable, matrix. More than 95 % of water
in the Chalk is held in the interstices of the rock matrix, but the pore spaces are so small that this water
is effectively immobile. The mobile water (the remaining 5 % or less) is held within the fractures that
transect the chalk matrix. Some fractures have been enlarged by dissolution to become fissures or even
karstic conduits. The fissures and conduits provide the permeable pathways for flow. These multiple
components of porosity and permeability within the Chalk have long been recognised, and it has been
described as a double-porosity (dual-porosity) aquifer (Foster, 1975; Price, 1987; Barker, 1991; Price
et al., 1993), a double-permeability (dual-permeability) aquifer (Price et al., 1993), and a triple-porosity
and/or triple-permeability aquifer (Worthington, 2003; White, 2003).
The double-porosity nature of the chalk has important consequences for contaminant transport: dif-
fusive exchange of dissolved species can occur between mobile fissure water and the immobile matrix
water. In the early stages of a contamination incident, this double-porosity diffusive exchange acts to
retard the transport of contaminants relative to groundwater flow and attenuate contaminant concentra-
tions due to diffusion into the (initially uncontaminated) matrix porewater. At a later stage, when the
primary source of contamination input to the mobile fissure water has ceased, back-diffusion from the
matrix porewater acts as a secondary source of contamination, significantly prolonging the duration of
contamination. It is likely that the apparent stability of the bromate extent and concentrations over recent
years is a result of the attenuating effects of double-porosity diffusion in the Hertfordshire Chalk.
Double-porosity diffusion between mobile fissure water and immobile matrix water can be de-
scribed mathematically using Fick’s Laws of diffusion (e.g. Barker and Foster, 1981; Barker, 1982,
1985b). Previous modelling representations of bromate transport in Hertfordshire (e.g. Buckle, 2003;
Atkins, 2005) have used a first order mass-transfer coefficient (FOMT) to represent the effects of double-
porosity attenuation. The FOMT approach is a mathematical simplification of the Fickian approach, and
the approximation is likely to underestimate the mobile groundwater concentrations in the longer-term.
In the east of the study area, there is evidence (e.g. Harold, 1937; MacDonald et al., 1998) of rapid
groundwater flow occurring between swallow holes and stream sinks in the Water End area (Figure 1.1)
and springs and boreholes in the Lea Valley, which is indicative of a dispersive system of karstic conduits.
1.3. The bromate source 25
Such a system would allow rapid transport of bromate, with low-attenuation of concentrations, and is
likely to be the cause of the wide distribution of bromate across the east of the catchment. Previous
modelling exercises (e.g. Buckle, 2003; Atkins, 2005) have been unable to reproduce the migration of
bromate to the Lea Valley due to deficiencies in the representation of the karst system.
Therefore, in order to represent the distribution and temporal evolution of bromate contamina-
tion in the Hertfordshire Chalk, predictive models of bromate transport must integrate the effects of
double-porosity diffusive exchange between fissures and the chalk matrix with the effects of rapid, low-
attenuation transport within karstic conduits. The complex nature of these processes, and quantifying
their relative importance, presents particular difficulties for prediction of contaminant transport.
1.3 The bromate source

The site of a former chemical works in Sandridge, which is now a residential development, has been
identified as the source of the bromate and bromide contamination. Limited site data are available from
investigations undertaken during redevelopment of the site in the mid 1980s and subsequent to the dis-
covery of the bromate contamination at Hatfield in 2000. To date, relatively little attention has been
directed at quantifying the bromate source: models by Buckle (2003) and Atkins (2005) have used
a constant concentration source term which is unrealistic over the longer term. During calibration, the
bromate source term was shown to exert a significant control on the form and magnitude of the simulated
contaminant breakthrough. Therefore, the magnitude and dynamics of bromate release to groundwater
beneath the source site, and thus to locations down-gradient of the source site, have consequences for
the magnitude and duration of bromate contamination within the catchment. Predictive models of bro-
mate contamination within the catchment are dependent on a realistic and representative source term to
quantify the input of bromate from the source site.
1.4 Research aims and objectives

The overall aim of the research presented in this thesis is to develop greater understanding of the pro-
cesses controlling the spatial distribution and temporal evolution of bromate contamination within the
Hertfordshire Chalk aquifer, including bromate release from the source zone, and to use this under-
standing as the basis for predictive models incorporating the effects of double-porosity diffusion on the
long-term evolution of bromate.
The specific research objectives were:
Evolution of bromate contamination

• To develop a conceptual model for groundwater flow and contaminant transport in the Hertford-
shire Chalk aquifer system by review of existing data and interpretation of additional tracer testing
and geophysical testing;
• To use the available information and monitoring data to describe the spatial distribution and tem-
poral evolution of bromate across the catchment, and to interpret this in association with the con-
ceptual model of the flow and transport system.
1.5. Approach 26
The source
• To describe and quantify the distribution of bromate at the source site through collation and de-
scription of site investigation and monitoring data;
• To develop alternative conceptual scenarios for bromate release to groundwater and quantify these
as ‘source terms’;
• To use the available monitoring data to constrain the potential source terms.
Catchment-scale modelling of bromate transport

• To develop analytical network modelling of contaminant transport to allow representation of Fick-
ian double-porosity diffusion and to integrate karstic transport pathways within the network;
• To use this model to produce predictions for the likely bromate concentrations at key output loca-
tions over the long-term.
1.5 Approach
The objectives were approached by developing a conceptual model of groundwater flow within the
Hertfordshire Chalk, which was informed by a new interpretation of the Hertfordshire karst system
by Cook (2010) following catchment-scale tracer tests. All available monitoring and investigation data
were analysed and interpreted to describe the evolution of the bromate contamination within the aquifer.
The processes controlling bromate transport were investigated by examining relationships between bro-
mate concentrations and variables including bromide and chloride concentrations, piezometry, catchment
recharge, and groundwater abstraction rates.
The bromate pollution in Hertfordshire is of a similar scale to the extensive chloride pollution of the
Chalk of the Tilmanstone valley in Kent (Watson, 2004) from coalfield brines. Watson (2004) developed
an investigatory methodology - the ‘Tilmanstone Methodology’ - for characterising and parameterising
catchment-scale double-porosity flow and transport in the Chalk. The investigatory methodology com-
prised geophysical testing, single borehole dilution tracer testing, borehole-to-borehole natural gradient
tracer testing, and sampling of chalk porewater and chalk fissure water to produce vertical porewater
and fracture water profiles at locations along the flow line. A one-dimensional semi-analytical model
(DP1D), incorporating Fickian diffusion between matrix water and fracture water, was used to simulate
chloride migration and was able to reproduce the observed porewater and fracture water profiles. For-
ward modelling indicated that the double porosity diffusion extends the duration of contamination in the
catchment by several decades.
It was therefore considered that the ‘Tilmanstone Methodology’ could be adapted for application
to the Hertfordshire bromate contamination. However, initial plans for a number of cored boreholes to
provide porewater profiles in the catchment had to be reconsidered due to the financial constraints of the
Water Companies. The scope of the field testing was reduced to a number of single borehole dilution
tests to determine groundwater velocities in the Hertfordshire Chalk.
1.6. Structure of thesis 27
Representing Fickian double-porosity diffusion was still considered of great importance, despite the
lack of data for porewater concentrations to fully validate predictions of fissure concentrations. There-
fore, an analytical network model was developed which represents Fickian double-porosity diffusive
exchange between fissure water and matrix porewater along interconnecting flow-lines, while allowing
karstic branches to be incorporated into the network. The model was parameterised by a combination of
values found within the literature, and the results of the single borehole dilution testing and catchment-
scale natural gradient tracer testing. Results were compared to groundwater monitoring data at key
locations within the catchment.
The bromate source term for the Hertfordshire network model was quantified through analysis and
interpretation of site investigation data available for the source site to constrain a range of conceptual
scenarios for bromate mobilisation and release to groundwater.
1.6 Structure of thesis

Following this introductory chapter, Chapter 2 presents a review of the mechanisms of groundwater
flow and transport of solutes within the Chalk, and a summary of typical parameters related to flow and
transport in the Hertfordshire Chalk. Chapter 3 develops a conceptual model for groundwater flow and
contaminant transport in the Hertfordshire Chalk by review of existing data, interpretation of additional
tracer testing, and statistical analysis of the effects of scavenge pumping at Hatfield on bromate occur-
rence. Chapter 4 describes and interprets the distribution and temporal evolution of bromate contami-
nation in Hertfordshire in association with a new conceptual model developed in Chapter 3. Chapter 5
describes and quantifies the distribution of bromate and bromide at the source site, and uses the results to
constrain a number of conceptual scenarios for bromate and bromide release to generate source terms to
be used for predictive modelling. Chapter 6 critically reviews existing models that have been developed
for bromate transport in the Chalk, and then develops a novel analytical network model to represent
the Hertfordshire Chalk catchment which is used, along with the source terms developed within Chap-
ter 5, to simulate bromate and bromide transport at a number of key locations, to compare the simulated
results with observed data, and to predict bromate concentrations into the future. The importance of
double-porosity diffusion between the matrix and the fissures for the long-term evolution of bromate is
demonstrated and discussed. Chapter 7 discusses and summarises the finding of the work in the context
of the objectives outlined in Section 1.4 and makes recommendations for further work.
1.7 Environmental Hydrochemistry of Bromate and Bromide

1.7.1 Occurrence of Bromate and Bromide in surface and groundwaters
Bromine is a naturally occurring trace element in groundwater. Inputs to natural waters include atmo-
spheric aerosols, contact or mixing with saline environments, weathering of materials present within the
aquifer, and pollution (e.g. burning of fossil fuels). Under the field conditions of most natural waters,
bromine occurs almost exclusively as the unassociated bromide ion (Br – ) (Edmunds, 1996).
As such, bromide (Br – ) is found naturally in most water systems. The concentrations of Br in
1.7. Environmental Hydrochemistry of Bromate and Bromide 28
potable groundwaters in the United Kingdom range from 60 to 340 µg l−1 . Concentrations in the Chalk
are summarised in Table 1.1. Due to the similarity of the geochemical behaviour of Br and Cl, using the
Br/Cl ratio is necessary to identify significant anomalies in the natural environment (Edmunds, 1996).
Table 1.1: Bromide in UK groundwaters, summarised from Edmunds et al. (1989). r2 is the linear
correlation coefficient squared for a regression of Br vs. Cl
Aquifer Bromide (mg l−1 ) Chloride (mg l−1 ) Br/Cl r2

Min Median Max Median ×10−3 (Br/Cl)
Chalk (Berkshire) 0.020 0.067 1.140 17.7 3.78 0.95
Chalk (London) 0.045 0.183 0.620 42.0 4.35 0.98
In a survey of baseline quality of the Chalk of the Colne and Lee River catchments, Shand et al.
(2003) identified elevated concentrations of Br – : samples from the vicinity of Hatfield had (Br)/(Cl) ra-
tios which plotted above the background trend (Figure 1.2).
Figure 1.2: Plot of Br vs. Cl concentrations in groundwaters from the Chalk of the Colne and Lee River
catchments. Points plotting in the ‘contamination’ box are from the Hatfield area. From Shand et al.
(2003).
In contrast to bromide, bromate (BrO3– ) is not reported as occurring naturally in surface waters
and is not normally present in aquifers (Butler et al., 2005). However, bromate has been detected in the
surface water environment as a result of industrial oxidation/disinfection processes (Butler et al 2005)
1.7. Environmental Hydrochemistry of Bromate and Bromide 29
with one study of 36 river samples detecting bromate at levels from 4 to 8 µg l−1 (Kruithof and Meijers,
1995). The occurrence of bromate within aquifers other than the Hertfordshire Chalk could not be found
reported within the published literature.
1.7.2 Environmental Behaviour

The bromide ion (Br – ) is typically non-reactive under conditions typical of most groundwaters, it does
not readily form complexes, participate in redox reactions, or sorb onto mineral or organic matter. Bro-
mide has been used extensively as a tracer in aquifers due to it conservative nature (Ward et al., 1998).
The bromate ion (BrO3– ) exists as a number of salts, including the industrially important potassium
(KBrO3 ) and sodium (NaBrO3 ) bromates, which dissolve readily in water3 , and once in solution, bromate
is highly stable (Butler et al., 2005). There is no direct evidence for a specific bromate reduction pathway,
and it has been concluded by most authors that biological reduction of bromate to bromide is a side
reaction of the nitrate reduction pathway (Hijnen et al., 1995). However, given that the Hertfordshire
Chalk aquifer is largely unconfined, the reduction of bromate to bromide is likely to be negligible.
Bromate is therefore considered to behave conservatively under most conditions. The environmen-
tal behaviour of bromate is poorly understood, and there appears to have been little research to date into
this area. However Butler et al. (2005) suggest that nitrate (NO3– ) and perchlorate (ClO4– ) may be consid-
ered potentially analogous oxyanions; both of which tend to behave conservatively within groundwater
under the conditions encountered in natural, unconfined aquifers.
3 KBrO
3 solubility at 25◦ C is 75 g l−1
30
Chapter 2
Groundwater flow and transport in the Chalk
2.1 Introduction
The aim of this chapter is to describe the mechanisms of groundwater flow and transport of solutes
within the Chalk, with particular reference to behaviour of the Hertfordshire Chalk aquifer, to review
typical parameters related to flow and transport in the Hertfordshire Chalk, and to review approaches to
representation and modelling of flow and transport in the Chalk.
2.2 The Chalk

2.2.1 Chalk stratigraphy
Three subdivisions were traditionally recognised in the Chalk Group: Lower, Middle and Upper Chalk.
Over the last 20 years there has been much research into Chalk Group stratigraphy (e.g. Bristow et al.
(1997); Gale et al. (1999); Woods (2006)) and the British Geological Survey has now adopted a revised
classification that follows Rawson et al. (2001) (Figure 2.1).
2.2.2 Lithology
In general the Chalk is a very fine grained (less than 10 µm), pure (circa 98 % CaCO3 ), soft, white
limestone of high (∼40 %) interstitial porosity, containing some marl bands and flint (Hancock, 1975),
with pervasive fractures of a variety of styles. The Chalk matrix is composed of calcium carbonate
micro-fossil fragments: the major, finer fraction comprising coccoliths and coccolith debris, and the
minor, coarser fraction comprising foraminifera and other shell debris (Hancock, 1975).
The marl bands can be several centimetres thick, some of them being laterally continuous for several
hundreds of kilometres. Many were deposited on erosional surfaces and are thought to be of volcanic
origin as they contain Mg-rich smectite. Flint occurs predominantly in layers parallel to bedding, either in
tabular layers or as scattered discrete nodules. Locally it forms cross-cutting veins and vertical cylinders
with a burrow at the core. ‘Hardgrounds’ are horizons of hard, brittle chalk in which porosity is reduced
to 10-20 %. The formation of hardgrounds is thought to be a result of sea floor cementation associated
with a reduction or cessation in the supply of coccoliths.
2.2. The Chalk 31
Figure 2.1: Relationship between the traditional and revised stratigraphy of the Chalk. After Woods and
Aldiss (2004)
2.3. The Chalk as an aquifer 32
2.2.3 Tectonic History

Three basic types of fracture can be recognised in the Chalk (Bloomfield, 1996): faults, bedding plane
fractures and joints. Tectonic movements were responsible for the formation of faults in the Chalk and
controlled the orientation of joint sets. Joints generally form in three mutually perpendicular orienta-
tions: one parallel to bedding and two orthogonal set perpendicular to the bedding plane. The dominant
regional fracture set in the Chalk of southern England trends NW-SE with a minor cross-cutting NE-SW
trending set (Bloomfield, 1996), as a result of the stress regime initiated in the early Palaeogene. In
addition, in southern England a series of E-W flexures and fractures developed in Oligocene-Miocene
time.
2.2.4 Influence of periglaciation

The Chalk of southern England remained largely free of ice sheets throughout the Quaternary period,
but was significantly affected by ground ice (periglaciation). In particular, the ‘active’ surface zone of
the Chalk underwent cycles of freezing and thawing, which has produced a weathered mantle of broken
rubbly chalk, frequently 1 m to 2.5 m thick (Williams, 1987). In some places, just below the sub-soil, the
Chalk is highly pulverised to a ‘putty chalk’: structureless chalk with irregular sized blocks set in a soft
to firm putty matrix. Much of this is absent from the interfluves but is present within wide sections of the
valleys. In contrast, the permanently frozen sub-strata was probably not seriously affected by periglacial
activity.
Putty Chalk has a low permeability and can therefore impede the flow of groundwater at the top
of the Chalk, sometimes confining the aquifer below or restricting the flow of groundwater between the
gravels and the Chalk. Putty Chalk has been detected in sections of the Thames and the Colne valleys,
especially where the Chalk it is overlain by river gravels.
Periglacial activity is the most commonly invoked explanation for the origin of the dry valley net-
work of the Chalk (Goudie, 1990). Infiltration was reduced during cold climatic conditions so that rapid
melting of winter precipitation released a large volume of runoff which easily eroded chalk that had been
already weakened by frost action. Periglacial activity is also thought to have contributed to the devel-
opment of enhanced permeability in the Chalk, through the development and dissolution of fractures
(Section 2.4.3).
2.3 The Chalk as an aquifer

The Chalk is one of the most important aquifers in the UK; according to the UK Groundwater Forum1
it provides over half of all groundwater for public supply in the UK and, locally in the south-east of
England, up to 70 % of all water in public supply.
and erosional processes. The calcium-carbonate micro-fossil fragments that make up the chalk form a
highly porous, yet essentially impermeable, matrix. More than 95 % of water in the Chalk is held in the
1 www.groundwateruk.org/
2.4. Karstic behaviour of the Chalk 33
interstices of the rock matrix, but the pore spaces are so fine that this water is effectively immobile. The
mobile water (the remaining 5 % or less) is held within the fractures that transect the chalk matrix. Some
fractures have been enlarged by dissolution to become fissures or even karstic conduits. The fractures,
fissures and conduits provide the permeable pathways for flow.
Barker (1993) provided illustrative relative flow rates permitted by matrix, primary fissures (frac-
tures) and secondary fissures (fractures enlarged by solution): rates of movement of groundwater through
the Chalk’s secondary fissure system2 can often be several hundreds of metres per day, compared to rates
of the order of a few millimetres per day in the primary fissure system3 , and the order of just a millimetres
per year in the matrix4 . The matrix water is therefore considered to be essentially immobile.
2.4 Karstic behaviour of the Chalk

It is increasingly recognised (e.g. Atkinson and Smith (1974); Price et al. (1992); Banks et al. (1995);
MacDonald et al. (1998); Worthington (2003); Maurice et al. (2006)) that aspects of aquifer behaviour
of the Chalk of southern England are indicative of karstic development. The term karst is a geomor-
phological term associated with terrain having distinctive landforms and hydrology (Ford and Williams,
2007). Karst characteristics (particularly surficial features) are more developed in some terrains (e.g.
Carboniferous limestone) than others (e.g. Jurassic Limestone and Chalk). However, in hydrogeological
terms, the importance of karst is that groundwater flows rapidly through, a network of fractures, conduits
(significantly enlarged fractures) and caves (conduits that are large enough to be explored physically by
man). Following MacDonald et al. (1998), the term karst is used within this thesis to describe features
of any dimension that allow rapid groundwater flow and also, where appropriate, geomorphological
dissolution features.
Worthington et al. (2000) examined matrix, fracture and channel (conduit) flow in four contrast-
ing carbonate aquifers (including the Cretaceous Chalk) and found considerable similarities in hydraulic
functioning between them: in all four cases, more than 90 % of the aquifer storage is in the matrix
and more than 90 % of the flow is in the karst channels, with fractures playing an intermediate role.
Worthington (2003) suggests that the similarity between the matrix, fracture and and channel flow and
storage properites in the four contrasting carbonate aquifers points to a similarlity between all uncon-
fined carbonate aquifers, which by their nature are subject to fracturing and then dissolution, resulting in
channel networks which contribute minimally to enhancing aquifer porosity, but greatly enhance perme-
ability. Worthington (2003) argues that differences reported in the literature are often largely attributable
to sampling differences as a result of a priori assumptions on the behaviour of aquifers: tracer tests
from sinkholes to springs characterise channel flow in the aquifer but do not provide information about
non-channel flow, whereas studies using wells characterise fracture and matrix flow but give little of no
indication of the rapid solute transport than may be occuring in the channel network located between the
wells. Therefore, it has been suggested that carbonate aquifers should all be considered as possessing
2 basedon tracer tests

3 assuming K of the order of 10−7 to 10−5 m s−1 , a porosity of 1% and a hydraulic gradient of 10−3
4 assuming K of 10−8 m s−1 , a porosity of 35% and a hydraulic gradient of 10−3
varying degrees of karst, rather than being wholly karstic or wholly non-karstic (Atkinson and Smart,
1981), or that all carbonate aquifers should be considered as triple-porosity/permeability aquifers, with
matrix, fissure and channel/conduit components (Worthington, 2003).
MacDonald et al. (1998) presented various strands of evidence which point to rapid groundwater
flow being widespread throughout the Chalk of Southern England. This evidence was drawn from:
• Surface drainage patterns;
• Geomorphological features;
• The presence of explorable caves and cave systems;
• Rapid flow rates implied by tracer tests, including those associated with observed karstic features
(such as swallow holes, stream sinks, springs etc) and those carried out without direct association
to such features (such as boreholes, soakaways);
• Incidence of chalk boreholes pumping sand;
• Construction details and descriptions from adits;
• Water level response to pumping and recharge;
• Occurrence of indicator bacteria in abstraction boreholes.
The evidence suggests that rapid groundwater flow is generally more frequent close to Palaeogene
cover and may also be associated with other forms of cover and valley bottoms (Section 2.4.3).
2.4.1 Geomorphological Evidence of Chalk Karst

Geomorphological evidence of karst characteristics, e.g. dolines, solution pipes and swallow holes, are
widespread within the Chalk. The common features are defined below:
Swallow hole The point at which a stream sinks underground. A swallow hole generally implies nearly
instantaneous water loss into an opening at the bottom of a sinkhole or karst valley, whereas a
swallet may refer to gradual water loss into the gravel along a streambed, with no depression
apparent.
Sinkhole A closed depression (circumscribed by a closed topographic contour) which drains to the
subsurface.
Doline (dissolution sinkhole) A sinkhole resulting from gradual dissolution of the bedrock. A classic
bowl-shaped contour (gently sloping depression that is wider than it is deep). Dolines can be
considered as depressions that do not necessarily have water flowing into them.
Sinking stream any stream that disappears underground, typically into a swallow hole.
Spring Any natural discharge of water from rock or soil onto the surface of the land of into a body or
surface water. Springs which occur on the dip slope of the Chalk are nearly always in the bottoms
of valleys and reflect the emergence of the water table at the surface.
Bourne Seasonal streams arising from Chalk springs in response to seasonal water table fluctuations.
The relationship between surface karst features and hydrogeologically significant subsurface karstic
features in the Chalk is unclear. MacDonald et al. (1998) point out that there is an inherent bias in
the recorded solution features towards those that are easily visible from the ground surface such as
swallow holes or dolines, and not necessarily those that are hydrogeologically significant. Relatively
rare exposures of chalk faces at Quarries (e.g. Castle Lime Works Quarry, Hertfordshire) give some
indication of potentially more laterally persistent features. Solution enlarged fractures are observed
at Water End, Hertfordshire, which may indicate hydrogeolgical connections from the swallow holes
and dolines. The high concentration of swallow holes and dolines in the region of Water End may be
hydraulically significant as they are likely to provide a mechanism for directing a significant volume of
recharge into the chalk aquifer. In particular they may direct recharge rapidly to a depth at which more
laterally persistent solution features occur.
2.4.2 Evidence of rapid flow rates from tracer tests in the Chalk
Atkinson and Smith (1974) conducted a tracer test in the Hampshire Chalk, where swallow holes occur
near to the northern margins of the Eocene outcrop. Rhodamine WT dye was pumped into a swallow
hole for three days. The travel time to its emergence at the Bedhampton spring (a distance of 5.75 km)
was 62.5 hours, corresponding to a velocity5 of 2.2 km day−1 (peak concentration). Atkinson and Smith
(1974) concluded that turbulent flow in an open system of fissures widened by solution, or conduits, is
required to achieve the observed velocities.
Banks et al. (1995) conducted a tracer test in the Berkshire Chalk between the Holly Grove stream
sink and the Blue Pool spring 4.7 km away. Both the spring and the stream sink were less than 1 km
from the Chalk/Eocene boundary. The observed velocities were 5.8 km day−1 for peak concentration
and 6.8 km day−1 for breakthrough. The authors suggest that little attenuation occurred as the tracer
moved from the sinkhole to the spring. Maurice et al. (2006) carried out two injections of flourescein
tracer from the Smithcroft Copse (nearby to Holly Grove) to the Blue Pool. Maurice et al. (2006)
determined groundwater velocities of 5.12 and 4.71 km day−1 with tracer recoveries of 25.5 % and
21.7 % respectively.
In the Hertfordshire Chalk, there is abundant evidence of the existence of rapid preferential flow
routes within the Chalk. Water End is a well known site of swallow hole activity (Section 3.2.5). The
swallow holes are located close to boundary between the Chalk and Eocene cover. Three tracer tests,
carried out in 1927, 1928 and 1932 using fluorescein dye showed that water recharging the swallow holes
were detected in a series of springs and wells in the Lea Valley up to 16 km from the swallow holes
(Harold, 1937). The breakthrough times gave flow velocities up to 5.5 km day−1 , with tracers being
detected over a 30 ◦ arc from the swallow holes (Harold, 1937). The velocities between the swallow
holes and a particular detection point in the Lea Valley varied by up to 37 % between tests, and the
fastest route also varied between tests. A test carried out in 1935 in swallow holes at South Mimms
(3.5 km to the sourthwest of Water End) also showed flows towards the Lea Valley over distances up to
5 Velocities quoted are mimimum velocities calculated based on straight line distance between input and output locations
19 km, with velocities up to 3.5 km day−1 .

In addition, bacteriological examination of groundwater samples taken from the Lea Valley in 1935-
36 showed a strong correlation between rainfall and the level of bacteria (Harold, 1937). The increase
in bacteria numbers coincided with periods when streamflows known to contain bacteria entered the
swallow holes at North and South Mimms. Given the connection between the swallow holes and the
sampling points in the Lea Valley, this suggested that particles of bacterial size could travel rapidly
through fissures in this area of the Chalk.
Cook (2010) undertook an extensive programme of tracer testing to investigate the nature of karst
connections in the Hertfordshire Chalk. The tracing was conducted using three species of bacteriophage
(phage), injected at three locations in Hertfordshire:
• MS2 Coliphage were injected by borehole dilution in Harefield House Borehole, near Sandridge;
• Phi X174 phage were injected by borehole dilution at Comet Way Borehole, near Hatfield;
• Serratia Marcescens phage were added to a sinking stream close to a large swallow hole complex
at Water End near North Mymms.
Monitoring for the tracers was conducted at 21 locations throughout the study area, comprising abstrac-
tion wells, observation boreholes, springs and surface waters. The duration of the monitoring period was
approximately two months.
The connections indicated by the tracer tests are further described in Section 3.2.6. Results indicated
rapid groundwater flow between the Water End swallow holes and a wide spatial distribution of locations
in the Lea Valley. Rapid flow connections were also indicated between locations to the west of the main
karst system toward Hatfield, Sandridge and within the Vale of St. Albans.
Between Water End and the Lea Valley (Serratia Marcescens phage), travel times and distributions
were broadly consistent with the 1920s and 1930s tracing results. However, better data resolution al-
lowed for the observation of breakthrough curves and secondary tracer peaks suggesting multiple arrivals
of tracer. First detections indicated groundwater velocities of between 1.8 km day−1 and 3.9 km day−1 .
Overall tracer recovery was estimated at approximately 15 % of the injected tracer mass.
Identification of tracer breakthrough for the MS2 Coliphage and Phi X174 phage was complicated
by measured concentrations being close to background. However, observations suggest groundwater ve-
locities between Hatfield (Phi X174 phage, Comet Way BH) and Essendon PWS, Arkley Hole Spring and
Lynchmill Spring of between 0.8 km day−1 and 1.8 km day−1 . Groundwater velocities between San-
dridge (MS2 Coliphage, Harefield House BH) and Hatfield Quarry, Essendon PWS, Arkley Hole Spring
and Lynchmill Spring indicated groundwater velocities between 0.05 km day−1 and 0.9 km day−1 . In
both cases, attenuation was high: overall tracer recovery was estimated at less than 1 % of the injected
mass.
There are also examples of tracer tests indicative of karst flow in ares of the Chalk of Southern Eng-
land that are unconnected with obvious karstic features. Price et al. (1992) reported a tracer experiment
at the M1/M25 motorway intersection, an area close to the Chalk/Palaeogene boundary where karstic
features are common, but the studies were carried out in soakaways that were apparently unassociated
with karstic features. Some tracer traveled rapidly to a pumping station a distance of 3 km away, with
recorded velocities in excess of 2.4 km day−1 . The tracer recovery was very low, and it was thought that
a significant fraction of the flow was moving through fine fractures.
A number of borehole dilution tests and radial tests from an injection borehole to a pumped bore-
hole, were carried out in East Anglia (Kachi, 1987; Ward, 1989). The main conclusion of the extensive
testing and modelling was that flow was dominated by micro-fractures with a range of sizes rather than
by a few discrete high-permeability conduits.
2.4.3 Development of permeability within the Chalk

A number of factors contribute to the pattern of occurrence and development of fractures within the
Chalk, including the effects of periglaciation.
The lithology of the Chalk has an important effect on aquifer properties (Mortimore, 1993):
• Within the Newhaven, Lewes Nodular, New Pit, and Holywell Nodular Chalk members, conjugate
fractures are common. At the intersection of the fractures, cavities can develop which could give
rise to rapid groundwater flow and therefore high permeability.
• Marl horizons restrict vertical flow due to their high clay content. However, marl and flint bands
may act to ‘seed’ the development of larger solution features, e.g. by forcing groundwater moving
downwards from the Chalk above to dissipate horizontally.
• The presence of hardgrounds, where shallower than about 100 m below ground level, can sig-
nificantly increase the permeability of the Chalk. Hardgrounds (such as the Chalk Rock and the
Melbourne Rock) fracture more cleanly than other chalks due to their greater hardness (Price,
1987), and since the hardgrounds are better cemented, fractures within them tend to remain open
at greater depths than in softer chalk. The open fractures allow groundwater to flow through them
generating preferential flow paths which may then be enhanced by dissolution.
The structure of the Chalk can affect aquifer properties, although the relationship between structure
and permeability is complex. Generally, folding tends to increase the fracturing of the Chalk along the
axis of anticlines. Deep burial of the Chalk (e.g. beneath Palaeogene deposits of the London Basin)
tends to reduce permeability and storage . Rapid, anisotropic, groundwater flow is often associated with
fault zones in the Chalk (Allen et al., 1997): parallel to the fault groundwater flow is rapid, while it is
impeded perpendicular to the fault.
Price et al. (1993) reviewed the factors that may have contributed to the development high trans-
missivity along valleys. There may be a higher frequency of open fractures within valleys since valleys
often follow lines of structural weakness, and erosion along the valleys reduces effective stress, allowing
horizontal fractures to open. However, a higher frequency of fractures in not necessarily a prerequisite
to high permeability; it can develop by solution enhancement of a lower frequency of fractures. Price
(1987) illustrated how the increased flux of groundwater along a valley could give rise to high perme-
ability within the valleys and the development of a single zone of dissolution-enhanced fractures. The
2.5. Hierarchy in the Chalk aquifer 38
concentration of groundwater flux, and hence increase in velocity, occurs as groundwater flows from
recharge areas to discharge areas in the valleys. The mixing of groundwaters of different chemistry
towards discharge points can also increase the dissolution potential of groundwater.
Periglaciation is believed to have played an important role in the enhancement of permeability along
the valleys (Younger, 1989). Repeated freezing and thawing of the active layer would have broken down
the top few metres to a weathered chalk that would have been easily eroded. Furthermore, in the valleys,
the flow of surface water would have kept the ground unfrozen to a greater depth for large parts of the
year, forming a talik within the Chalk of the valley floor. Chalk is dissolved more easily under cold
conditions, therefore the concentrated flow of groundwater within these taliks and the low temperature
of the groundwater would combine to dissolve Chalk within factures. Within valleys, an increase of
fracturing is observed in the top 5–6 m of the Chalk. In some valleys, periglacial activity led to fractures
opening up to a depth of 20–30 m.
Geomorphological karstic features, and rapid groundwater flow, are generally more frequent close
to Tertiary cover (MacDonald et al., 1998). Based on a field survey of the Pang and Lambourn catchments
in Berkshire, Maurice et al. (2006) identified three distinctive geomorphic Zones characterised by the
density of surface karst features which was related to the proximity to the Palaeogene-Chalk contact.
Soils associated with Palaeogene deposits tend to be quite acidic. The soils associated with Quaternary
cover also tend to be clayey and therefore to concentrate runoff into discrete points. As recharge drains
through the cover, it remains unsaturated with respect to calcite until it reaches the Chalk surface, thus
allowing acidic recharge to be channelled to discrete points of the Chalk surface.
Solution pipes and swallow holes can allow acidic recharge to penetrate deep into the unsaturated
zone (and even the saturated zone). Therefore the acidic recharge can link up with fracture systems
within the aquifer and allow the enlargement of fractures to conduits. This is probably the mechanism
for the observed rapid groundwater flow near Quaternary cover (Allen et al., 1997). It is likely that zones
of slightly wider initial aperture occurring in fractures with asperities will be preferentially widened
by dissolution until discrete ‘conduits’ are formed (Younger and Elliot, 1995). Even once the cover
and geomorphological features have been moved by erosion, the deeper hydrogeological features will
remain, providing rapid groundwater flow and preferential flow paths.
Rapid groundwater flow and swallow holes are also often observed on valley bottoms, associated
with flowing streams. These ’karst’ type features may have a different origin to those observed on the
Chalk interfluves (Allen et al., 1997). Surface water flowing within streams has different chemistry
from the groundwater, therefore mixing can produce water that has a high dissolution potential. This
aggressive water, coupled with the high flux through the valleys and the dynamic of surface water flow
can produce sink holes and springs in stream beds, probably linked with conduits. Therefore, as in more
classic ‘karst’ environments, surface water can flow in discrete channels underground within valleys.
2.5 Hierarchy in the Chalk aquifer

Cushman (1990) introduced the concept of natural hierarchy in porous media. He recognised that many,
if not most, porous media, such as soil and geological formations, possess some sort of hierarchy: either
2.5. Hierarchy in the Chalk aquifer 39
structural (relating to successively nested, interacting, physical subunits) or functional (relating to a

hierarchy of transport processes). Furthermore, structural (functional) hierarchy may be discrete or
continuous. In a discrete hierarchical medium there are a finite number of nested structural subunits or
functional subprocesses, whereas in a continuous hierarchical medium there are an infinite number of
subunits (subprocesses) that cannot be clearly decomposed.
The Chalk has long been recognised as having multiple components of porosity and permeability,
being described as a double-porosity (dual-porosity) aquifer (Foster, 1975; Price, 1987; Barker, 1991;
Price et al., 1993), a double-permeability (dual-permeability) aquifer (Price et al., 1993), and a triple-
porosity and/or triple-permeability aquifer (Worthington, 2003; White, 2003).
A classic double-porosity (or dual-porosity) aquifer consists of a porous matrix divided into blocks
by a well-developed (orthogonal) fracture set (Price et al., 1993). The matrix pores provide storage and
the fractures/fissures provide the permeable pathways for flow. Dual-porosity aquifers are distinct from
double-permeability (or dual-permeability) aquifers, which are heterogeneous formations with perme-
able intervals alternating with layers or lower permeability (Figure 2.2). The permeable intervals may
have higher primary permeability than the others, or they may be beds displaying extensive fissuring.
Figure 2.2: (a) An idealised double-porosity aquifer; (b) an idealised double-permeability aquifer. From
Price et al. (1993)
The porosity and permeability components contributed by the fissures are referred to as the as the
fissure (or fracture) porosity and permeability. The blocks bounded by the fissures are usually described
as matrix blocks, and the non-fissured fraction of the porosity and permeability as the matix (or matric)
porosity and permeability.
The behaviour of the Chalk is more complex than described in the classical double-porosity model.
The majority of the water within the matrix pore space does not represent usable groundwater storage;
effective groundwater storage is primarily within the fracture network and the larger pores (Section 2.6).
2.6. Porosity components of the Chalk aquifer 40
In terms of groundwater flow, there is a hierarchy in permeability components (Section 2.7) as a result of
solution enhancement of primary fractures to form fissures and karstic conduits. The Chalk has therefore
been described as exhibiting dual-permeability behaviour (Price et al., 1993).
The Chalk is increasingly recognised as possessing karstic characteristics in places (Section 2.4.2).
This additional flow component is neglected by classic double-porosity models (Section 6.2.6). More re-
cently, it has been proposed that carbonate aquifers, including the Chalk, be considered as triple porosity
and triple permeability aquifers (Worthington, 2003; White, 2003), to account for flow and storage in
three porosity and permeability elements:
• the three-dimensional matrix;
• two-dimensional, or planar, fracture (fissure) elements; and
• one-dimensional, or linear, channel (conduit) elements.
2.6 Porosity components of the Chalk aquifer

Chalk matrix porosity and pore size distributions show both regional and stratigraphic variations. Bloom-
field et al. (1995) described regional trends in matrix porosity data, based on BGS Aquifer Properties
Laboratory core analysis database. Within the Thames and Chilterns area, mean porosity is 38.8 % for
the Upper Chalk, and 31.4 % for the Lower Chalk (standard deviations of 5.8 % and 6.6 % respectively).
Watson (2004) recorded slightly higher values for the Chalk in the Tilmanstone and Eastry areas of Kent:
43 % for the Upper Chalk (44.5 % for samples from the Seaford Chalk and 40.0 % for samples from the
Lewes Nodular Chalk) and 38 % for the Lower Chalk. Watson (2004) found a general decline in values
from approximately 45 % at 8 m below ground level (BGL) to 40 % at 60 m BGL. Price et al. (1976)
described regional and stratigraphical trends in pore size distributions in the Chalk. The median pore
size was 0.65 µm for the Upper Chalk of the Southern Province, 0.53 µm for the Middle Chalk, and
0.22 µm for the Lower Chalk. Atkins (2002) obtained values from 30.8 % to 46.8 %, with an average of
38.0 % from chalk samples taken during the 2001 site investigation at St. Leonard’s Court, Sandridge.
Although the matrix has high porosity, small pore throat diameters mean that the pores are not
readily drained under gravity. It has been estimated that for the Upper Chalk, about 3 % of the total
porosity (or 1 % of the bulk volume) of the Chalk represents useable storage (Price, 1987). Effective
(accessible) groundwater storage is probably primarily within the fracture network and the larger pores
(Price et al., 1993). The relative contribution to the storage coefficient from the matrix and fractures of
various sizes is difficult to ascertain. The total specific yield is in the range 1 % to 3 % (Allen et al.,
1997). Overlying deposits, where present, often make an important contribution to groundwater storage:
groundwater stored in the overlying Palaeogene deposits of the London Basin can be drawn down into
the Chalk facture system (Allen et al., 1997).
MacDonald and Allen (2001) collated data from pumping tests carried out on the Chalk (but see
Section 2.7.2 below for discussion of inherent data bias). One thousand two hundred pumping tests
had estimates of storage coefficient. The median value of storage coefficient from all the pumping tests
2.6. Porosity components of the Chalk aquifer 41
was 0.0023, and the 25th and 75th percentiles were 0.0004 and 0.01 respectively. Approximately one
order of magnitude difference was recorded between estimates of storage coefficient from confined and
unconfined tests: unconfined tests had a median of 0.008 and confined tests had a median of 0.0006. The
authors suggest that the relatively small difference between unconfined and confined measurements may
be explained by the limited gravity drainage of the Chalk matrix, and therefore the relative importance
of elastic storage in both confined and unconfined conditions. However, MacDonald and Allen (2001)
also found a direct relationship between transmissivity and the storage coefficient, which indicates that
gravity drainage from the fracture network is still a significant component of storage.
In fissured rocks such as the Chalk, where nearly all groundwater movement occurs in the fissure
network, the porosity that is of importance when groundwater flow velocities are being considered is the
porosity that represents the saturated pore space which contributes to the flow. This porosity, which will
be practically equivalent to the porosity of the fractures, is significantly lower than the total porosity: in
general, the porosity of the fractures is around 0.01 % Price et al. (1993). De Marsily (1986) defines the
ratio of the volume of water able to circulate to the total volume of rock as the kinematic porosity. The
term effective porosity is also frequently used in this context. However, Ward et al. (1998) encourages the
use of the term kinematic porosity over effective porosity for the purposes of solute transport to prevent
association with the specific yield, which is sometimes taken to be representative of the effective porosity,
and has been used for the purposes of solute transport in fractures (e.g. Bibby (1981)). However, it may
not be appropriate to use values of specific yield in this context: specific yield values are usually derived
during hydraulic testing and the fluid volumes released due to the stresses applied during testing may be
unrepresentative of the volume of water through which solute transport occurs.
The effective porosity is related to the aperture of the active fractures and the frequency of occur-
rence of those fractures. Atkins (2004) estimate an effective porosity range of 0.05 % to 0.50 % for
the Hertfordshire Chalk, assuming fracture apertures between 0.5 mm and 5.0 mm and an average of
one fracture per metre of borehole. However, they note that this range does not take into account the
contribution of the vertical fractures to the effective porosity (vertical boreholes only detect horizontal
or sub-horizontal fractures) and therefore state a more ‘reasonable’ effective porosity range of 0.5 % to
2.0 %.
Values for effective porosity in a fractured rock can be determined if the Darcy velocity or flux q
and average linear velocity v are known through the relationship
q
ne = . (2.1)
v
Watson (2004) calculated effective porosity for the Chalk of the Tilmanstone-Eastry Valley in Kent
based on the results of single borehole dilution tests and natural gradient tracer tests. Single borehole
dilution tests were undertaken to obtain darcy velocity, q, for specific depth intervals. The groundwater
(fissure) velocity, v, was estimated from the arrival times of the natural gradient tracer tests. The effective
porosity was then calculated using equation 2.1. The calculated effective porosities were in the range
0.1 % to 0.3 %.
2.7. Permeability components of the Chalk aquifer 42
2.7 Permeability components of the Chalk aquifer

2.7.1 Terminology
There is no clear consensus within the hydrogeological literature on the definition relating to openings or
partings in rock that carry water. Throughout this thesis, the term fracture is used to refer to any planar
discontinuity in the Chalk which has a finite aperture, and is used where the intention is to avoid any
generic or hydraulic inferences (e.g. Bloomfield (1996)). Fractures may originate as joints (tensional or
shear) or small faults (Section 2.2.3). The term fissure is applied to fractures that have been enhanced by
solution. By implication, fissures are hydraulically significant features of the rock.
Solution enhancement may take the form of small channels on the fracture surfaces or of a general
widening over a substantial proportion of the fracture area. Fissures are distinguished from conduits by
the fact that they retain the generally planar geometry of unmodified fractures. Following the terminol-
ogy used by Maurice et al. (2006), conduits are tubular voids or channels formed by solution, and are
distinguished from fissures by the aspect ratio they present in cross section: trace length to maximum
aperture ratios for conduits is approximately one in contrast to greatly in excess of ten for fissures.
In relation to groundwater flow within the Chalk aquifer, the terms primary fissures and secondary
fissures are often used (Price, 1987; Barker, 1993) to distinguish between fractures which are tectonic
in origin, and fractures that have been significantly enlarged, respectively. Price et al. (1993) point out
that in reality there is a more or less continuous range from tight fissures to greatly enlarged, karstic
openings.
2.7.2 Bulk transmissivity

Hydraulic properties determined by well tests (pumping tests) on the Chalk provide an averaged value
over the depth of the borehole. Similar aquifer properties might be estimated from two boreholes with
radically different flow regimes: a borehole that intersects one highly permeable fracture can give the
same transmissivity and storage coefficients as a network of small fractures with moderate permeability
(e.g. Foster and Milton (1974)).
MacDonald and Allen (2001) collated and analysed data from pumping tests carried out on the
Chalk, using both observation and production boreholes, at approximately 1300 locations throughout
England. The filtered dataset comprise 2100 measurements of transmissivity and 1200 estimates of
storage coefficient. Distributions give the appearance of log-normality, however are not truly log-normal.
The median transmissivity value was 540 m2 day−1 ; the 25th and 75th percentiles were 190 m2 day−1
and 1500 m2 day−1 respectively. However, MacDonald and Allen (2001) stress the inherent bias to this
data which has predominantly been calculated from pumping tests undertaken in boreholes that have
generally been drilled for production and are therefore in areas known to be high-yielding. Throughout
the Chalk outcrop, data are clustered within valleys, where the transmissivity and storage tends to be
high. The median value of 540 m2 day−1 is therefore unreasonably high for large areas of Chalk (Allen
et al., 1997).
It has long been observed that the transmissivity within the Chalk is greater in valleys than on
the interfluves (e.g. Ineson (1962)). However, the existence of this pattern has lead to the majority of
boreholes being drilled within the valleys, and the consequent lack of data available across the interfluves
hinders the understanding of the variations in transmissivity away from the valleys. However, Allen et al.
(1997) report that flow logging from boreholes drilled on interfluves of the Berkshire Downs indicated a
general thinning of the transmissive zone compared with valleys.
Transmissivity values were also found to be significantly higher in unconfined areas compared to
confined areas: unconfined sites had a median transmissivity of 920 m2 day−1 compared to a median
of 220 m2 day−1 at confined sites. It is thought that this reflects increased solution enhancement of
fractures under unconfined conditions.
2.7.3 Permeability components
2.7.3.1 Matrix Permeability

The Chalk matrix is generally isotropic with respect to permeability, and matrix hydraulic conductivity
is typically in the range of 10−4 to 10−2 m day−1 (Price, 1987). The distribution of horizontal hydraulic
conductivity values, derived from 977 gas permeability measurements on chalk core samples, was found
to approximate to a lognormal distribution with geometric mean of 6.3 × 10−4 m day−1 (Allen et al.,
1997). This low value indicates that the hydraulic conductivity of the Chalk matrix is negligible with
respect to the hydraulic conductivity of Chalk fracture systems (Section 2.7.3.2). For the same data, a
general correlation between bulk matrix porosity and hydraulic conductivity was observed and a linear
regression (r2 = 0.651) of the form:
log10 K (m day−1 ) = 0.0654 × Φ(%) − 5.3378
Allen et al. (1997) speculated that a correlation between some function of the matrix pore size distri-
bution and hydraulic conductivity is likely to be more significant, although such a correlation was not
established for the Chalk matrix.
2.7.3.2 Fissure Permeability

If groundwater flow withn the Chalk is primarily through a system of fissures, the transmissivity is
typically a function of the degree of fracturing. More specifically, the transmissivity of the Chalk is
dependent on the aperture and spacing of the fissures and the interconnectedness of the fissure network.
A single, perfectly planar, smooth-sided fissure with uniform aperture, a, can be ascribed a trans-
missivity, Tf , by the so-called cubic law (Snow, 1968; Barker, 1993):
ga3
Tf = (2.2)
12ν
where g is the acceleration due to gravity and ν is the kinematic viscosity. The transmissivity of a
fracture is therefore very sensitive to aperture size since the transmissivity is a function of the cube of
the aperture. Barker (1993) highlights this by a simple example: equation 2.2 gives a transmissivity of
about 10−3 m2 s−1 for a fissure with an aperture of only 1 mm, and an enormous 1 m2 s−1 for a 1 cm
aperture. Equation 2.2 is valid providing flow is laminar (Section 2.8).
2.7.3.3 Fracture aperture and spacing

Generally, there are three types of fractures (primary fissures) present in the Chalk: joints, bedding plane
fractures and faults. A combination of factors such as dissolution, weathering, and release of overbur-
den, can lead to significant enlargement of these initial fractures, either individually or in layers or zones,
(forming secondary fissures). The primary fissures typically have apertures of much less than 1 mm; So-
lution enlarged secondary fissures typically have apertures of greater than 10 mm (Worthington, 2003).
Price et al. (1993) calculated theoretical hydraulic conductivities for the Chalk related to aperture and
spacing (Figure 2.3).
Figure 2.3: Relationsip between fissure spacing, aperture, porosity and hydraulic conductivity for a
fissure system containing three plane, parallel, mutually perpendicular smooth-walled fissures filled with
pure water at 10 degC and porosity relationship (Price et al., 1993).
Bloomfield (1996) measured fracture orientation, trace length, spacing and aperture using section
and scan-line surveys on the Upper Chalk at Play Hatch Quarry, Berkshire. There were two dominant
joint sets: a set parallel to bedding and a set at a high angle to bedding. The trace length and spacing
distributions of the two joint sets approximated to log-normal distributions, with geometric mean trace
lengths of 0.15 m and 0.30 m, and spacings of 0.10 m and 0.12 m, respectively. Calculated fracture
interconnectivity indices suggest that the bedding parallel joint set is likely to be of greater hydraulic
importance than the high-angle joint set. Bloomfield (1996) used the fracture interconnectivity index
of Rouleau and Gale (1985) to demonstrate that the bedding parallel joint set fractures had a higher
interconnectivity, suggesting that the are likely to be of greater hydraulic importance than the high-angle
joint set. Bloomfield (1996) proposed that the results of the Play Hatch Quarry support a visualisation of
the Chalk consisting of “scale-invariant fault-bounded segments, where the internal fracture architecture
of each segment is dominated by continuous bedding plane fractures, and subordinate, scale-dependent,
arrays of joints”. The scale of jointing within a given fault-bounded segment is a function of bedding
thickness.
Aperture measurements obtained for a single bedding plane fracture ranged from less than 0.5 mm
to 23.5 mm. Apertures approximated to a negative exponential distribution (with a mean of 1.2 mm)
below 7 mm, and to a log-normal distribution (with a mean of 11.7 mm) above 7 mm. It was inferred
that the larger apertures have been affected by solution processes and that flow through bedding plane
fractures is channeled across 10–20 % of the fracture surface area.
Watson (2004) calculated block sizes (fracture spacing) for the Chalk of the Tilmanstone-Eastry
valley in Kent corresponding to fracture density predicted by the relationship established for the area:
Log10 [fracture density] = −1.685 × Log10 [depth] + 2.5365
This relationship was determined based on data from scanline surveys in the area and a fracture profiled
produced with televiewer an optiviewer logging.
Watson (2004) used data for block sizes and effective porosity (Section 2.6) to calculate fracture
aperture using the relationship
fracture aperture = block width × effective porosity
In general, fracture apertures were found to increase with depth, and ranged from 0.44 mm to 3.83 mm.
Watson (2004) also calculated fracture apertures from hydraulic conductivity data from packer testing
using an approximation to the cubic law:
Kef f ective ≈ 50a3 N
where N is the fracture density, a is the fracture aperture, and Kef f ective is the effective hydraulic con-
ductivity. Apertures determined by this method (cubic law or hydraulic apertures) ranged from 0.92 mm
to 1.39 mm.
Figure 2.4 summarises the range of fracture spacing and fracture aperture measurements from a
number of sources. It should be noted that fracture apertures are difficult to measure in-situ: at out-
crop, weathering effects may produce enlarged apertures, and accurate down-hole measurement is only
feasible for larger openings. As such, there are few direct observations for the Chalk, and apertures are
usually determined from cubic law approximations or in relation to fracture density and fracture porosity
measurements.
It has been shown that groundwater velocities measured from the tracer tests described in Sec-
tion 2.4.2 could result from flow through either small channel conduits or more laterally extensive fis-
sures. For the Hampshire study Atkinson and Smith (1974) calculated that, for the estimated head
gradient, the volume indicated by the tracer time to peak was equivalent to flow in a single circular pipe
Min
Fracture Spacing
Chalk (Kent) Max
Mean
Chalk (Kent, Sussex & Dorset)
Chalk (Yorkshire)
>10 m depth (Oxfordshire)
Near-surface (Oxfordshire)
Unweathered Chalk (Norfolk)
Weathered Chalk (Norfolk)
Normal to bedding (Berkshire, Yorkshire)
Bedding parallel (Berkshire, Yorkshire)
0.01 0.1 1 10
Fracture spacing (m)
Fracture Aperture
Kent Chalk Min
Max
Yorkshire Chalk Mean
Yorkshire Chalk
Yorkshire Chalk
Berkshire Chalk (solution enhanced)
Berkshire Chalk (joints)
0.1 1 10 100
Fracture aperture (mm)
REFERENCES
Figure 2.4: Ranges of fracture spacings and fracture apertures for the Chalk. Results as cited in Bloom-
field (1996) and Watson (2004).
740 mm) in diameter (although in reality several features would probably be involved). Price (1987) sug-
gested that in an ideal case (a plane parallel fracture with no roughness or chanelling) a fissure of only
4.5 mm width, with a transmissivity of 5000 m2 day−1 , provides an alternative hydraulic explanation
for the observed combination of displacement volume and velocity. By using the same method Banks
et al. (1995) calculated that a single fissure 5.4 mm in width could theoretically be sufficient to represent
the fracture system in Berkshire, and similarly, Maurice et al. (2006) calculated that the observed veloc-
ities would suggest a comparable aperture of 4.9 mm. However, they note that the hydraulic gradient
(0.004) is calculated using the elevation difference between the sink and the spring, and since the water
table is lower than the surface elevation of the stream sink, the actual hydraulic gradient must be smaller,
and therefore the calculation must underestimate aperture. Cook (2010) determined apertures around
2 mm to 4 mm using the cubic law approach for the results of the tracer tests in Hertfordshire. These
results were also interpreted as minimum equivalent conduit diameters between 0.6 m and 2.1 m, based
on maximum recorded discharges and flow velocities at locations including Water End and Catherine
Bourne swallow holes, and Arkley Hole and Lynchmill Springs.
2.7.3.4 Distribution of fractures

Small aperture fractures (primary fissures), present throughout the Chalk, are significant in increasing the
transmissivity of the Chalk aquifer compared to the permeability of the Chalk matrix. In general, flow
logs and packer tests indicate that permeability measured throughout the depth of a borehole is generally
at least an order of magnitude higher than matrix permeability (Allen et al., 1997), and Price et al.
(1982) found that packer test values for intervals in the Chalk are always higher than the intergranular
measurements (in contrast to the good agreement found in the Penrith Sandstone).
However, Price (1987) estimated the transmissivity of the Chalk without secondary fissuring to be
approximately 20 m2 day−1 , which is much lower than the typical values of the order of 2000 m2 day−1
reported in the Chalk (Section 2.7.2). Solution enlarged fissures, with high individual transmissivities,
are therefore necessary to account for the high (bulk) transmissivities that characterise the Chalk aquifer.
For example, in areas of high transmissivity, there is evidence that most of the thickness of the Chalk
may have relatively low hydraulic conductivities with the bulk of the water movement occurring at a
few horizons where fissures have been enlarged by solution (Headworth, 1978; Headworth et al., 1982).
Geophysical logging indicates that flow in the Chalk aquifer is often limited to discrete flow horizons,
and TV-logging reveals that these are often macroscopic solution pipes up to several centimetres across.
These conduits may be few in number, but will completely dominate the overall hydraulic conductivity
of the aquifer (Younger and Elliot, 1995).
These hydraulically significant fissures are not present through the full thickness of the Chalk, but
are concentrated in the upper sections of the Chalk, typically the upper few tens of metres, with little
flow deeper than 50 m below groundwater levels (Allen et al., 1997). For example, packer testing at
Trumpletts Farm on the Berkshire Chalk (Williams et al., 2006) indicated that hydraulic conductivity
varied by three orders of magnitude (from ∼50 m day−1 to ∼0.05 m day−1 ) over the 70 m of chalk
tested, with a strongly non-linear decrease with depth below ground level. Deeper within the Chalk the
frequency and aperture of fractures declines due to increasing overburden and a general reduction in
circulating groundwater and hence dissolution. The greatest permeability in the Chalk is observed in
the zone of water table fluctuation where the movement of groundwater can enhance the aperture of the
fractures by dissolution. A similar pattern of permeability variation with depth is observed where the
Chalk is confined by younger deposits within the London and the Hampshire Basin. In the London Basin,
flow logging has illustrated that the majority of inflows to boreholes, shown by geophysical logging,
occur within 20 m to 30 m of the upper surface of the Chalk.
Williams et al. (2006) compared the results of the packer testing at Trumpletts Farm with results
from borehole imaging and geophysical testing; while some of the highly permeable zones appeared to
be associated with obvious fractures, large fractures could be seen in zones which had much lower per-
meability, and some highly permeable zones appeared to be associated with poorly developed fractures.
Therefore, not all fractures are hydraulically active. Furthermore, Williams et al. (2006) found differ-
ences in flow velocity depth profiles (from single borehole dilution tests) in the same borehole which
was tested both before and during a pumping test at an abstraction borehole about 40 m from the site,
indicating that different fractures become active when the aquifer is stressed. Therefore, field studies
show that flow near individual boreholes is highly heterogeneous, and that there is uncertainty in the re-
lationship between the characteristics of a fracture observed in a borehole and the amount of flow which
that fracture contributes to the borehole.
Groundwater flow in the Chalk is highly heterogeneous and borehole yields may vary by orders
of magnitude over distances of less that 100 m reflecting the complexity of flow in the aquifer. The
single borehole dilution testing by Williams et al. (2006) showed differences in flow velocity depth
profiles between boreholes located within a few tens of metres across the site. These are inferred because
the different boreholes, although of similar depth and drilled in very close proximity, intersect slightly
different parts of the fracture network an hence groundwater flow system. In particular, a flowing feature
at the base of one borehole is not intersected by the second, which is drilled from a slightly higher
elevation.
When considering flow observations in boreholes, it is important to note that the presence of the
borehole connects flowing horizons and allows vertical flow. Therefore, the observed flow may not be
representative of flow in the aquifer in the absence of the borehole.
2.7.3.5 Connectivity of fissures and conduits

Worthington (2003) described a triple porosity/permeability model of carbonate aquifers (including the
Chalk). This visualises groundwater flow via an interconnected tributary network of channels, formed
by dissolution, which discharges to springs. The smaller (upgradient) channels are in the millimeter
to centimeter range (i.e. correspond to fissures). The larger (downgradient) correspond to conduits
(diameter >1 cm) and caves (diameter >1 m).
The nature and connectivity of fissures and conduits, and their relationship with surface karst fea-
tures is poorly understood. Fissures are the features most commonly encountered in boreholes and wells;
conduits are rarely intersected. However, tracer tests using surface karst features will directly encounter
2.8. Groundwater Flow in the Chalk 49
the conduit system. Connectivity between cave conduits and fissures intercepted by quarrying has been
demonstrated in the highly karstic Carboniferous Limestone (Edwards et al., 1991). The occurrence
of turbidity and bacteria in some Chalk boreholes suggest that there may also be interaction between
fissures and larger scale conduits in the Chalk (MacDonald et al., 1998).
During the Water End tracer tests in Hertfordshire (Section 2.4.2), visible colouration demonstrated
connections to seven spring and borehole abstraction sites up to 6 km apart. Visible colouration in the
Blue Pool complex during the tracer tests in Berkshire by Maurice et al. (2006) demonstrated that tracer
was discharged at a number of sites up to 100 m apart, suggesting flow through laterally extensive fissures
in the vicinity of the springs. The results of both these tracer studies are consistent with the downstream
sections of conduit systems being characterised by fissure flow resulting in lateral dispersion across a
large scale three-dimensional network.
Significant loss of tracer mass is indicated by the recovery data from the Hampshire and Berkshire
tracer tests: tracer recoveries were 70 % (Atkinson and Smith, 1974) and 25 % and 21 % (Maurice et al.,
2006) respectively. The breakthrough curves displayed a steep falling limb after the main peak, followed
by a long flat tail of low concentration. Such a pattern is indicative of exchange between mobile water
and immobile water by the mechanism of double porosity diffusion (Section 2.9.4), but the rapid travel
time implies that for a single conduit or fissure, tracer transport would be too rapid for double porosity
diffusion to account for the tracer loss. Maurice et al. (2006) invokes a complex flowpath, with sections
of the flowpath characterised by multiple pathways with dispersion from the main conduit into smaller
fissures and fractures, as a possible explanation for the loss of tracer.
The tracer test from soakaways at Bricket Wood in Hertfordshire (Price et al., 1992) showed very
low tracer recoveries of <0.01 % which the authors propose reflect a significant fraction of the flow
occurring in fractures and fissures with relatively small apertures, allowing attenuation of the tracer by
diffusion into the Chalk matrix.
2.8 Groundwater Flow in the Chalk

Conventional groundwater hydrogeology usually considers aquifers to be porous, granular media.
The quantity of water flowing through a granular medium is proportional to the hydraulic gradient,
a relationship expressed in what is termed Darcy’s Law:
dh
q = −K (2.3)
dl
where q is the flow per unit cross-sectional area of aquifer (the specific discharge), and K is the hydraulic
conductivity. Darcy’s Law is only valid if flow is laminar. Under laminar flow conditions individual ‘par-
ticles’ of water move in parallel paths in the direction of flow, with no mixing or transverse component
to their motion. As flow velocity increases, fluctuating eddies develop and transverse mixing occurs, and
the flow is termed turbulent.
The Reynold’s number, Re , identifies the critical velocity above which laminar flow gives way to
turbulent flow, and is expressed as:
ρνd
Re = (2.4)
µ
2.9. Solute transport in the saturated zone of the Chalk 50
where ν is the mean velocity of a fluid with density ρ and dynamic viscosity µ passing through a pipe of
diameter d. In a porous or fissured medium, d becomes a representative length dimension characterising
interstitial pore-space diameter or fissure width (Ford and Williams, 2007). Laminar flow generally
occurs for Re <2300.
For flow in pipes, under laminar conditions, specific discharge can be evaluated by what is termed
Poiseuille’s law:
πd4 ρg dh
q= . (2.5)
128µ dl
Increasing velocity, sinuosity and roughness may eventually result in flow through the tube be-
coming turbulent. For turbulent flow, the specific discharge can be calculated by the Darcy-Weisbach
equation:
2dg dh
q2 = . (2.6)
f dl
where f is a friction factor.
Turbulent flow conditions frequently arise in pipes and fissures in karst (Ford and Williams, 2007).
In karst, the range of conditions under which Darcy’s Law can be considered valid is very restricted: it
only applies in conditions that permit velocities to be low, and this usually involves some combination
of relatively small aperture and low hydraulic gradient (Ford and Williams, 2007).
For the range of karstic flow velocities observed in the Hertfordshire tracer tests (0.022–
0.068 m s−1 ), turbulent flows (where Re >2300) would occur above a conduit diameter of 0.04–0.13 m
(Cook, 2010). Therefore, the majority of karst flows within the Hertfordshire karst conduits are expected
to be turbulent.
2.9 Solute transport in the saturated zone of the Chalk

In a multiple-porosity aquifer such as the Chalk, solute transport is dominated by two processes: advec-
tion in fissures and diffusional exchange of solutes between fissures and matrix porewater. Adsorption
may also affect the transport of some solutes. At larger scales, the effects of dispersion across the network
of fissures may become important.
2.9.1 Advection
Advection is the term used to describe movement attributed to transport by the flowing groundwater:
advecting solutes travel at the same rate as the average linear velocity of the groundwater.
The groundwater velocity, v, is given by the volumetric flux, q, divided by the kinematic porosity,
ne :
q
v=
ne
The volumetric flux, q, can be determined from Darcy’s law (equation 2.3).
2.9.2 Adsorption
Solutes can be attenuated relative to advective transport by the process of adsorption. Most of the sur-
face area of chalk onto which adsorption can take place is within the rock matrix. Mineral or organic
deposits are quite often observed on the surfaces of fissures, and when present, must have a significant
impact on the amount of adsorption taking place, by providing adsorption sites and acting as a barrier
for diffusion into the matrix (Barker, 1993). Some models include a fracture skin to take some account
of the phenomenon.
2.9.3 Dispersion
Dispersion refers to the process of spreading (of solutes, particles or heat) during transportation. Ma-
trix diffusion and adsorption both have dispersive effects. The term mechanical dispersion refers more
specifically to the spreading caused by variations in groundwater flow velocity. Dispersion arises because
of the detailed variations in flow velocity in pores and fractures mainly due to the complex splitting and
joining of paths but also due to flow velocity variations in single paths. Strictly speaking, mechanical
dispersion is not a process in its own right; it is rather an expression of the fine (often random) detail
of the advection process. Therefore, if advective transport could be fully characterised throughout the
system, there would be no additional dispersion phenomenon to consider. Dispersion must therefore be
related to the advective model (conceptual or mathematical) in use, particularly to its scale of averaging
(Ward et al., 1998).
The normal approach to describing transport in a dispersive medium is via the convection-dispersion
equation (Equation 2.7 for flow in one-dimension), which contains a characteristic dispersion coefficient,
D.
∂C ∂2C ∂C
=D 2 −v (2.7)
∂t ∂x ∂x
The dispersion coefficient is normally considered to increase in proportion to the absolute value of
the velocity, v, so D = αv where α is known as the dispersivity. Dispersion is sometimes separated into
longitudinal and transverse dispersion to refer, respectively, to dispersion in the direction of flow and
dispersion perpendicular to that direction. Transverse dispersivity is typically much less than longitudinal
dispersivity. Dispersion coefficients for single fractures are more likely to show proportionality to the
square of velocity rather than to the velocity (Ward et al., 1998).
In practice, the processes of molecular diffusion and mechanical dispersivity cannot be separated in
flowing groundwater. Instead, a factor termed the coefficient of hydrodynamic dispersion is introduced
which takes into account both mechanical mixing and diffusion:
D = αv + D∗
where D is the coefficient of hydrodynamic dispersion, D∗ the effective molecular diffusion coefficient.
An analytical solution to Equation 2.7 was provided by Ogata and Banks (1961).
Dispersion in individual fissures is likely to be small and dependent on the fissure aperture, rough-
ness etc., and at a larger scale dispersion depends on the interconnecivity of fractures (Grisak and Pick-
ens, 1980, 1981). Barker (1993) demonstrated that dispersion, although not negligible, can normally be
ignored, in relation to matrix diffusion, when modelling solute transport in the Chalk.
2.9.4 Diffusion
Molecular diffusion represents the net movement of solute under a concentration gradient and can be
described by Fick’s first and second laws. Molecular diffusion of contaminants is not normally of prac-
tical consideration where advection and mechanical dispersion are dominant. However, within the water
in the chalk matrix, which is (effectively) immobile, transport of solutes can take place by molecular
diffusion. For any diffusive process, a characteristic time for diffusion over a distance x can be defined
x2
as D, where D is the diffusivity (Diffusion coefficient divided by porosity) (Barker, 1993).
For the double-porosity Chalk, and other comparable fractured porous media, Foster (1975) pro-
posed that a major component of solute movement was controlled by a mechanism involving solute
exchange, through lateral molecular diffusion, between mobile fissure water and (relatively) immobile
matrix water. This double-porosity diffusive exchange has been demonstrated to be of significance to
the interpretation of thermonuclear tritium and pollutants, such as nitrate, in the unsaturated zone of
the Chalk (Foster, 1975; Barker and Foster, 1981) and when predicting the rate of lateral migration of
pollutants in the saturated zone of the aquifer (Watson, 2004; Burgess et al., 2005).
Double-porosity diffusive exchange acts to attenuate contaminants and significantly prolongs the
duration of contamination (Figure 2.5). Considering an initially contaminated fracture water, and un-
contaminated matrix water, there will be a diffusive flux of the dissolved contaminant from the fracture
water into the matrix water. The movement of the contaminant down hydraulic gradient will therefore be
retarded compared to transport by advection. At a later stage, when the primary source of contamination
input to the fracture water has ceased, there will be a diffusive flux from the now contaminated matrix
water to the fracture water. The matrix water therefore acts as a secondary source of contaminant input,
which prolongs the duration of contamination detected down-gradient.
2.9.4.1 Diffusion Coefficients

Fick’s first law can be expressed as:
∂c
Jm = −DE (2.8)
∂x
where Jm is the mass flux (per unit area of water and rock) in the x direction in saturated rock, DE
is the effective diffusion coefficient, c is the mass per unit volume of water (not matrix and water), and t
is time.
Fick’s second law can be expressed as:
∂c ∂2c
= DA 2 (2.9)
∂t ∂x
where DA is the apparent diffusion coefficient.

The two diffusion coefficients are related through:
DE = φD DA (2.10)
where φD is the ‘diffusion porosity’ or ‘fictitious porosity’ (Section 2.12.2).

Oakes (1977) measured the effective diffusion coefficient for chloride in Chalk as 1.3 ×10−10 m2 s−1 .
Hill (1984) measured molecular diffusion coefficients of nitrate, chloride, sulphate and water (labelled
Figure 2.5: Double-porosity diffusive exchange of solutes. At an early stage, diffusion from the contam-
inated fracture water into the matrix water acts to retard the transport of contaminants down-gradient. At
a later stage, contaminated porewater acts as a persistent secondary source of contamination.
2.10. Flow and transport in the unsaturated zone of the Chalk 54
using tritium) in samples of both fissured and unfissured chalk (Table 2.1). These values represent the
mass flux through the saturated matrix per unit concentration gradient in the water. Hill’s values are
appropriate for use in Fick’s first law of diffusion. In a critical discussion of Hill’s results, Muller (1987
deduced that the ratio of the diffusion coefficient to the free water diffusion coefficients is about 0.25.
So the diffusion coefficient in chalk can be estimated as one quarter of the free water value, if this value
is known for a solute.
Table 2.1: Molecular diffusion coefficients in fissured and unfissured chalk. After Hill (1984). These
values represent the mass flux through the saturated matrix per unit concentration gradient in the water.
solute Diffusion coefficient range (m2 s−1 )

sulphate 0.28 ×10−10 1.47 ×10−10
nitrate 0.53 ×10−10 3.2 ×10−10
chloride 0.52 ×10−10 3.23 ×10−10
tritiated water 0.6 ×10−10 3.51 ×10−10
No values of diffusion coefficients for bromate or bromide have been reported in the literature. In
the absence of a specific value for bromide and bromate, it seems reasonable to expect the bromide and
bromate ions to behave similarly to chloride and nitrate ions.
2.10 Flow and transport in the unsaturated zone of the Chalk

A consequence of the small pore throat sizes of the Chalk matrix (Section 2.6) is that, even in the un-
saturated zone, the intergrannular pore space will be almost fully saturated (Foster, 1975; Price, 1987).
Thus the unsaturated matrix conductivity is very close to the saturated hydraulic conductivity. Within the
literature, there has been much debate as to the significance of fissure flow versus intergranular matrix)
flow in the unsaturated zone of the Chalk. The early consensus was that flow in the unsaturated zone of
the Chalk was entirely through the fissure network, supported by observations of rapid response of the
water table after high intensity rainfall events (Headworth, 1972). However, unsaturated zone porewater
tritium profiles from the Berkshire Chalk (Smith et al., 1970) indicated distinct tritium peaks, believed to
be associated with the high tritium levels found in rainfall during the periods 1963-1964 and 1958-1959,
at depths of 4 m and 7-9 m respectively. Smith et al. (1970) suggested that these profiles indicated that
85 % of the total flow through the unsaturated zone was by intergranular seepage through the matrix at
a mean rate of 0.88 m/yr , with some ‘bypass flow’ through fissures occurring to explain the presence of
tritium at depth. A ‘piston displacement’ mechanism , whereby recharge at the top of the column of ma-
trix displaces water at the bottom of the unsaturated zone , releasing it to the water table, was invoked to
explain the rapid water table response to rainfall (Price et al., 1993). Foster (1975) noted that Chalk out-
crop areas are characterised by an almost complete absence of surface runoff, which is incompatible with
UK rainfall rates of up to 100 mm d−1 , an order of magnitude greater than the unsaturated zone Chalk
matrix hydraulic conductivity (Section 2.7.3.1), unless substantial fracture flow occurs to transport the
excess during these rainfall events. Foster (1975) proposed a mechanism by which fracture-dominated
2.11. Modelling flow and transport in fissured rocks 55
flow could be reconciled with a much slower observed downward movement of tritium: fissures within
the Chalk would focus tritium input to the unsaturated zone and the concentration gradient between the
contaminated fracture water and the matrix water would cause lateral diffusion of the solute into the ma-
trix (i.e. double-porosity diffusive exchange discussed in Section 2.9.4), greatly retarding its downward
movement. Simulations using a double porosity diffusion exchange model (Barker and Foster, 1981)
indicated that this mechanism lead to preservation of the tritium profile in the unsaturated zone with only
minor dispersion. However, Mathias et al. (2005) showed that double porosity models which assume
matrix flow to be negligible (e.g. Barker and Foster 1981), require an unrealistically small fracture spac-
ing (<25 cm) to preserve peaks without ‘solute spreading’ through the profile. Mathias et al. (2005)
analysed the impact of flow in the matrix by comparing a double porosity model (based on Barker 1982)
with an equivalent double-permeability model with a portion of flow in the matrix and demonstrated that
solute spreading in such models can only be reduced (whilst using sensible estimates of fracture spacing
and diffusion coefficients) by allowing for a portion of flow in the matrix. Thus, Mathias et al. (2005)
argues that flow in the matrix of the Chalk of the unsaturated zone is significant.
However, Mathias et al. (2005) was not able to obtain a sensible estimate for the proportion of total
infiltration that enters the matrix as the model assumed steady-state flow conditions necessitating the use
of annual mean estimates of infiltration. The assumption of steady-state flow also forces fracture flow
to be either negligible or persistent whereas in reality it is likely to be intermittent (Price et al., 2000).
Subsequently, Mathias et al. (2006) developed a transient one-dimensional double permeability model
for the unsaturated zone of the Chalk. This model indicated that infiltration (as calculated by simple
two-store models) needs to be significantly attenuated to ensure that enough flow occurs in the matrix
such that solute spreading is reduced to a reasonable level. The justification for the attenuation was the
existence of soil and gravelly chalk layers. Mathias et al. (2006) demonstrated that such a model was
compatible with a fast water table response: there was a time lag of only three days between effective
precipitation input and water table flux.
2.11 Modelling flow and transport in fissured rocks

The Chalk is a fissured rock, and the interconnected fissures provide routes for water flow with the
intervening blocks of rock essentially impermeable. Domenico and Schwartz (1998) suggested that at
the scale of the field problem one of two approaches might be followed when trying to represent the flow
of water in fissured and/or karstified rocks:
1. the continuum approach, which assumes that the fractures mass is hydraulically equivalent to a
porous, granular medium, i.e. an equivalent porous medium (EPM) model; and
2. the discontinuum or discrete approach, which assumes that the rock cannot be characterised as a
granular medium, and so considers that flow is best dealt with in individual fractures or fracture
sets.
The appropriate model to simulate transport of water and/or solutes within fissured systems such as
the Chalk depends on a consideration of how the behaviour of a fissured system is related to the time-
2.11. Modelling flow and transport in fissured rocks 56
scales of the transport processes. For double-porosity systems, with advective transport in the fissures
and diffusive transport in the matrix, the suitable model representation depends on the time-scale of the
process under consideration in relation to the characteristic times for diffusion across a fissure or a matrix
block (Barker, 1993).
2.11.1 Equivalent Porous Medium (EPM) models

An EPM model is a homogeneous model with parameters chosen to be characteristic of the fissured rock.
Barker (1993) considers that an EPM model might be suitable under two regimes of double-porosity
behaviour:
1. When the fissures act independently of the matrix. If the time-scale of interest is small with
respect to the characteristic time for diffusion across the fissure width, then the effects of the
porous matrix can be ignored (because of both the restricted diffusion out of the fissure and of the
small volume of matrix accessed, in relation to fissure volume). Under these conditions (which
rarely exist outside a laboratory), the chalk can be modelled with an EPM model with a porosity
equal to the fissure porosity.
2. When the fissures and matrix act in unison. If the time taken for diffusive equilibrium between
fissures and matrix is small in relation to the time for any significant change in the fissure system,
then the chalk will behave as a (locally) homogeneous medium characterised by the total porosity.
A Quasi-steady-state (QSS) double-porosity model might also be adequate (Section 2.11.2.2).
EPM models are commonly used for regional water resources models, where the fissured system is
represented as homogeneous, with storage and permeability parameters characteristic of the matrix and
fractures combined.
2.11.2 Double-porosity (DP) models

DP models comprise two overlapping continuous media (the fracture and matrix phases) coupled by an
exchange mechanism. Such DP models can be divided into diffusive type and quasi-steady-state (QSS)
type models depending on the physical and mathematical description of the exchange mechanism:
2.11.2.1 Diffusive type

Diffusive-type models are those for which the transport in the matrix blocks can be described by a flux
law (i.e Darcy’s, Fourier’s or Fick’s Laws for water, heat and solute transport respectively). The potential
(head, temperature, or concentration) within a matrix block is controlled by the variations within the
fissure system, and the two potentials are normally assumed equal at the surfaces of the matrix blocks
(Barker, 1993). The shapes of the matrix blocks affect the behaviour (Barker, 1985a,b).
Diffusive-type double-porosity models, simplified by assuming an infinite matrix, are appropriate if
the time-scale of interest is a small fraction of the time for diffusion across a matrix block. Under these
conditions only the matrix/fissure surface area per unit volume of the rock is important, not the block
size or shape. If the time-scale is similar to the time for diffusion across a matrix block, then the sixes of
the matrix blocks become important, and the assumption of an infinite matrix is not valid. Under these
2.12. Diffusion exchange model for solute transport in fissured porous rocks 57
conditions, Barker (1993) considers that a general diffusive-type DP model should be used, (although a
QSS-type DP model may be adequate over some periods).
2.11.2.2 Quasi-steady-state (QSS) type

In QSS models, the matrix is characterised by a single potential (e.g. concentration) and the diffusive
flux between the matrix and the fissures is taken to be proportional to the difference between their (local)
potentials. QSS type models are only valid alternatives to the diffusive-type model if the time for any
significant change in the fissure system is slow in relation to the time for diffusive equilibrium across a
matrix block. Under these conditions the chalk will behave as a (locally) homogeneous medium charac-
terised by the total porosity (an EPM mode might also be suitable under these conditions).
2.11.2.3 Importance of matrix diffusion for water and solute transport in the Chalk
Characteristic times for hydraulic diffusion in the chalk matrix are around 5–500 seconds and for solute
diffusion in the chalk matrix are 50–500 years (Barker, 1993). Therefore, for water transport, matrix
diffusion will only be significant for very rapidly changing conditions, and it is reasonable to adopt a
QSS model for all but the most rapid transient pumping tests. In contrast, matrix diffusion will have an
important effect on solute transport over most time-scales of interest for contamination incidents, and
diffusive-type double-porosity models should be adopted.
2.11.3 Double-permeability Models

In contrast to double-porosity models, the advective velocity in the matrix, although much less than in the
fissures, is not regarded as negligible in double-permeability models. Barker (1993) considers double-
permeability concepts to be more applicable to the interaction between the primary and secondary fissure
systems in the Chalk than to the interaction between matrix and fissures, for which the permeability
contrast is much larger.
2.11.4 Network Models

Network models have been used for the study of groundwater in ‘hard rocks’, generally in relation to
radioactive waste disposal. Hard-rock systems generally have little intrinsic porosity, although they
do often contain a dense network of micro-fissures concentrated near the fractures which can impart a
double-porosity character. Barker (1993) considers that such models are valuable when considering flow
in the Chalk under karstic conditions, when matrix porosity may not be significant.
2.12 Diffusion exchange model for solute transport in fissured

porous rocks
In a series of papers, Barker (1982, 1985a,b) developed a general model of flow and transport in double-
porosity media. Barker (1982) considered a simple ‘slab geometry’ model, with parallel fissures sepa-
rated by finite slabs of matrix material (Figure 2.6). Fissured rocks have been represented in this way
by others e.g. Grisak and Pickens (1980). Because of the periodicity of the model only a single, semi-
infinite unit extending from the centre of a matrix slab to the centre of a neighbouring fissure need be
considered. The solution of the transport equations was developed as as far as Laplace transforms of the
solute concentrations in the fissure and matrix water. Numerical inversion of the transforms was then
used to investigate characteristic behaviour of the model for a number of special cases.
2.12.1 Block geometry

Barker (1985a) developed a Block Geometry Function (BGF) which characterises block shape based on
the diffusion equation. For simple geometries, such as the sphere, this function is quite simple. For
any well defined geometry, even mixtures of blocks of different shapes and sizes, this function can be
determined.
2.12.1.1 Block sizes

Barker (1985b) characterises the size of a block by the parameter, in this thesis referred to as `, which
represents the volume to surface area ratio. More precisely, ` is the root-mean-square distance of dif-
`2
fusion6 in time 2DA . For an infinite slab geometry, with slabs of thickness 2b, ` = b; for a spherical
geometry with blocks of radius r, ` = 3r .
2.12.2 Porosity Ratio

Barker (1985a) defines a parameter, σ, which represents the ratio of the matrix and fracture porosities,
and which equals the porosity of the matrix void volume per unit volume of rock matrix divided by the
fracture void volume per unit volume of total space within the rock. So the total porosity (matrix plus
fractures) is 1+ σ times the fracture porosity. More precisely, sigma is the ratio of the matrix volume to
fracture volume in a given total volume, times the ratio of DE and DA (diffusion coefficients defined in
Section 2.9.4.1) which is the ‘fictitious’ porosity.
The porosity ratio is also related to the diffusion porosity φD , for example for a simple slab model:
2bφD
σ= (2.11)
a
DE
The diffusion porosity is defined as φD = DA , which is the ratio of the ‘apparent diffusion coeffi-
cient’, DA , (as appears in Fick’s second law) and the ‘effective diffusion coefficient’, DE , (as appears
in Fick’s first law). This porosity has been referred to in the literature as a fictitious porosity and can be
somewhat less than the total porosity (Barker et al., 2000).
2.12.3 Characteristic Times

Characteristic times for the simple slab model are given in Table 2.2.
2.12.3.1 Fracture advection time

The fracture advection time is normally determined from the hydraulic gradient, (bulk) hydraulic con-
ductivity and kinematic porosity.
6 In molecular diffusion, the mean-square distance traversed by a particle in time t is given by 2Dt, where D is the diffusion
coefficient.
Figure 2.6: Governing equations and assumptions for a double-porosity mocel with slab geometry. After
Barker (1982).
2.13. Summary 60
2.12.3.2 Block diffusion time

Barker (1985b) defines a characteristic time for diffusion across a matrix block, in this thesis referred to
as tcb , which is a related to the block size (via the parameter `) and apparent diffusion coefficient DA by:
`2
tcb = (2.12)
DA
2.12.3.3 Fracture diffusion time

The behaviour of a double-porosity system can be fully described by σ, tcb , ` and the BGF. However,
Barker et al. (2000) introduces a further parameter which can be more characteristic of certain behaviour.
This is the characteristic time for diffusive equilibrium between fractures and matrix, in this thesis re-
ferred to as tcf , which is a defined as:
tcb
tcf = (2.13)
σ2
The time tcf can be thought of as the time for diffusion through a matrix volume equal to the fracture
volume. This parameter tends to be important when the interaction time between the fracture water and
the matrix water are less that the time for diffusion across a matrix block. Under those conditions the
fracture water concentration is determined mainly by the surface available for diffusion in relation to the
fracture size. The block size and geometry become unimportant.
2.12.3.4 Relative times

The relative values of the three times (tcb , tcf and ta ) can provide valuable insights into the behaviour
of a double-porosity system. When times are much less than tcb , then the only effective parameter is
tcf ; physically this represents conditions where fracture water concentration is determined mainly by
the surface area available for diffusion in relation to the fracture size.
Table 2.2: Characteristic times for infinite slab geometry, with slabs of thickness 2b separated by frac-
tures of aperture a. For this model, the ratio of volume to area for a matrix block (`) is represented by
b.
x
ta = v Advection time in mobile phase
2
b
tcb = Dim Characteristic time for diffusion across a matrix block
tcf = tσcb2 Characteristic time for diffusion from a fracture into an equal volume of matrix water
σ = θθim
m
The ratio of matrix to fracture porosity
2.13 Summary
and erosional processes. The Chalk is composed of very fine grained calcium carbonate micro-fossil
fragments which form a a highly porous, yet essentially impermeable, matrix. More than 95 % of
water in the Chalk is held in the interstices of the rock matrix, but the pore spaces are so small that
this water is effectively immobile. The mobile water (the remaining 5 % or less) is held within the
2.13. Summary 61
fractures that transect the chalk matrix. Some fractures have been enlarged by dissolution to become
fissures or even karstic conduits. The fissures and conduits provide the permeable pathways for flow.
Within the unsaturated zone of the Chalk, although the dominant flow pathways are via the fissures, a
small but significinat portion of flow is thought to occur within the matrix (Mathias et al., 2005, 2006).
These multiple components of porosity and permeability within the Chalk have long been recognised,
and it has been described as a double-porosity (dual-porosity) aquifer (Foster, 1975; Price, 1987; Barker,
1991; Price et al., 1993), a double-permeability (dual-permeability) aquifer (Price et al., 1993), and a
triple-porosity and/or triple-permeability aquifer (Worthington, 2003; White, 2003).
The Chalk is increasingly recognised as possessing karstic characteristics. In the Hertfordshire
Chalk, there is abundant evidence of the existence of rapid preferential flow routes within the Chalk.
Swallow holes and other dissolution features tend to be located close to boundary between the Chalk and
Eocene cover, and are particularly concentrated in the Water End area. Tracer tests have shown that rapid
groundwater flow occurs between swallow holes and stream sinks in the Water End area and springs and
boreholes in the Lea Valley, which is indicative of a dispersive system of karstic conduits.
In a multiple-porosity aquifer such as the Chalk, solute transport is dominated by two processes:
advection in fissures and diffusional exchange of solutes between fissures and matrix porewater. Adsorp-
tion may also affect the transport of some solutes. At larger scales, the effects of dispersion across the
network of fissures may become important.
Double-porosity diffusive exchange of of solutes between fissures and matrix porewater acts to
attenuate contaminants and significantly prolongs the duration of contamination. Considering an initially
contaminated fracture water, and uncontaminated matrix water, there will be a diffusive flux of the
dissolved contaminant from the fracture water into the matrix water. The movement of the contaminant
down hydraulic gradient will therefore be retarded compared to transport by advection. At a later stage,
when the primary source of contamination input to the fracture water has ceased, there will be a diffusive
flux from the now contaminated matrix water to the fracture water. The matrix water therefore acts as
a secondary source of contaminant input, which prolongs the duration of contamination detected down-
gradient. Double-porosity diffusion between mobile fissure water and immobile matrix water can be
described mathematically using Fick’s Laws of diffusion (e.g. Barker and Foster, 1981; Barker, 1982,
1985b), and has been demonstrated to be of significance when predicting the rate of lateral migration of
pollutants in the saturated zone of the aquifer (Watson, 2004; Burgess et al., 2005).
62
Chapter 3
A conceptual model for flow and transport of

bromate in the Hertfordshire Chalk
3.1 Chapter Objective

The objective of this chapter is to develop a conceptual model for groundwater flow and contaminant
transport in the Hertfordshire Chalk aquifer system by review of existing data, interpretation of addi-
tional tracer testing, and statistical analysis of the effects of scavenge pumping at Hatfield on bromate
occurrence.
3.2 Geology and Hydrogeology of Hertfordshire

3.2.1 Topography
The study area is located in Hertfordshire, south-east England, and covers an area of approximately
600 km2 bounded by eastings TL511000 and TL540000 and by northings TL200000 and TL219000
(Figure 3.1).
The highest elevations (+150 m OD) are found on the dip slope of the Chiltern Hills in the north-
west of the study area. The land slopes gently to the southeast towards the relatively flat lying Vale of St
Albans (elevations +60 to +75 m OD), which forms a broad valley trending northeast-southwest between
Colney Street and Hertford. To the east of Hatfield, the Vale of St Albans joins the Middle Lee Valley
(elevations +60 to +65 m OD). The land rises to the south-east of the Vale of St Albans to elevations
of approximately +100 m OD) over the Palaeogene escarpment (Tertiary Escapement), before falling to
the east and south-east towards the Lee Valley (elevations less than +50 m OD).
Superimposed on this overall topographic pattern are smaller river valleys. River Valleys (both
dry and flowing valleys) tend to be deep, incised valleys, which extend far up the dip slope. There are
a number of dry valleys in the Chalk uplands of the Chiltern Hills, including two dry valleys which
converge at Sandridge.
3.2.2 Hydrology
River flows, rainfall and potential evapotranspiration data are reviewed and analysed in Buckle (2002)
and Atkins (2004). Much of the data were collated in work by Entec (2000) in connection with the Upper
j
k
!
A
Ri
± h
WARE r As
Riv ve
River Beane
! Ri BROADMEADS
Ri
A er
! ve
v e r R ib
A rL Mi
mr ! !
HARPENDEN am Y
X Y
XA A AMWELL
ea WELWYN
(o GARDEN HERTFORD ChadwellN HILL Amwell
rL !
Spring e A X
w
ee CITY Y !Spring
AAMWELL
) Y
X
MARSH
R iv
er
k
j !
A
Arkley Hole RYE COMMON
Rye House
SANDRIDGE Spring j
k
ESSENDON YSpring
! Y MIDDLEFIELD RDX
HATFIELD A kj X ! !
QUARRY FORMER
A A
!
A j
k
j WATERHALL
k !
AERODROME HODDESDON A
k
jk j
k k
j j
k j
k YLynchmill
X
HATFIELDk j j
jjk
k jk
kkj jkj
j kj k
k
jk
jj Spring
k
!
A jk
k
j
k j
k
! k
j
Ri
s
ve A j
k !
A
ro o
rV ST ALBANS
an
BROXBOURNE
nb
er
!
A Y
X
lb
El l e
A
! !
A
t
A ROESTOCK
Ground Elevation
TYTTENHANGER
fS
m OD Water
j End
k kk
jjk
j
o j
kj
k !
10 - 20 j
k
NORTHk
j
j
k jkj A
k
jMYMMS Cuk
r
21 - 30
l eLONDON e
j j
k
j
k
j
k
A!k ffl
e j
k TURNFORD
Va COLNEY oln j
k j
jk
jk
k
y
31 - 40 C j
k
j
k j
k
j
k
Ri v e
r
41 - 50 Ri v e jk
jk j
Br o
j
k k k
j jk
kj k
j
j
k
Rive r Le a (or L
51 - 60
Ne w
ok
61 - 70 NETHERWILD k
jBourne k j
ee)
! k
j j j j
k
A jk
k i ne k jk
jk
71 - 80 er POTTERS
SOUTH
3.2. Geology and Hydrogeology of Hertfordshire
th
81 - 90 SHENLEY MIMMS BAR Legend
Ca
91 - 100 RADLETT Springs
Y
X
Mymmshall Brook
101 - 110
k
j Swallow Holes
111 - 120
!
A Public Water Supply
121 - 130
rivers and streams
131 - 140
141 - 150 lakes
151 - 160 urban region
0 1 2 4 6 8 10 !
© Crown Copyright/database right 2008. An Ordnance A
Survey/EDINA supplied service.
Kilometers Geological Map Data © NERC 2008.
63
Figure 3.1: Location of study area, including topography and hydrology

3.2. Geology and Hydrogeology of Hertfordshire 64
Mimram study, and along with additional data available from the Environment Agency.
In the east of the area, the upper River Lea and the River Ver flow south-east from the Chiltern
Hills. The surface water divide between the Lee and the Ver catchments is not well defined in terms
of topography, but it runs approximately south-east through Harpenden and Hatfield and also forms the
surface water divide between the catchments of the River Lee and the River Colne (Figure 3.1). The
River Colne flows southwest along the foot of the Palaeogene escarpment in the Vale of St Albans to join
the River Ver in the south-west of the study area. Significant groundwater–surface water interactions
occur in both the River Lee and River Colne catchments (Section 3.2.7).
The middle River Lee flows east through the central part of the study area along the northern foot
of the Palaeogene escarpment. Downstream of Hatfield, the Lee swings north-east and is joined near
Hertford by the Rivers Mimram, Beane and Rib flowing from the Chalk upland to the north, and further
downstream by the Rivers Ash and Stort. The River Lee then swings to flow south towards the River
Thames. Along this southerly flowing section, the middle and lower Lee is joined by a number of rivers
draining the Palaeogene escarpment. South of Hertford, the New River (an aqueduct constructed in the
17th Century) runs to the west of, and parallel to, the River Lea. The New River is fed from the River
Lee upstream of Ware and also accepts discharge from the Chadwell Spring when the spring is flowing.
The New River also takes pumped discharge from pumping stations of the Northern New River well
field.
In the west of the study area, the Catherine Bourne and the Mimmshall Brook drain to the River
Colne from the west of the Palaeogene escarpment. For most of the year, the Mimmshall Brook and
Catherine Bourne drain to a series of stream sinks and swallow holes near Water End. At times of
overflow, a spillway carries water north-west to the River Colne.
3.2.3 Geology
3.2.3.1 Regional geological context

The solid and drift geology of the study area is summarised in Figure 3.2 and Figure 3.3.
The study area lies on the north-western margin of the Thames Basin. The Chalk outcrops in the
north-west on the study area, where it forms part of the dip slope of the Chiltern Hills. The Chalk dips
gently (at <1 degree) south-east towards the main axis of London Basin syncline.
To the south and south-east of the area, the Chalk within the London Basin is overlain by Palaeogene
Deposits (Reading Beds, the London Clay and the Claygate Beds) which form the Tertiary escarpment.
There are Palaeogene outliers on the dip slope of the Chalk near St Albans and Welwyn Garden City.
In the Chiltern Hills, drift deposits overly the Chalk in the valleys (Alluvium and Valley Gravels)
and on the higher interfluves (Clay-with-Flints and associated Pebbly Clay and Sand). A band of Boulder
Clay and Glacial Gravel overly the Chalk in the Vale of St Albans and the Middle Lee Valley. Pebble
Gravel caps the Tertiary strata in places.
3.2.3.2 Lithostratigraphy
The generalised lithostratigraphy of the Hertford district is summarised in Table 3.1.
Ri
Solid Geology Riv
LONDON CLAY FORMATION
er
Mi
Chadwell h
Ri mr Spring r As
LAMBETH GROUP
ve am ve ±
v e r R ib
rL Ri
LEWES NODULAR CHALK Fm ea Y
X Y
X
SEAFORD CHALK Fm
River Beane
NEWHAVEN CHALK Fm
(o
rL
(UNDIFFERENTIATED) ee
) Y
X
Y
X Ne
w
R
SOURCE SITE Arkley Hole k

j
iver
Spring j
k
Y
X Y
X
k
j
j
k j
k
j
kk k j
k j
k j
k Y
X
j k
jjk
k k
jk
jjj k
j j
k
k
j j
k
s
k
k
jj
k
jk
jkj kj
j
k k
j
an
r
lb
Y
X
Ri v e
tA
S
Ne w
of
Water
l e End
Va j
k
j
k kk
jjk
j
e j
k
k
jj
k
j
k jk
k j
ln k
jk j
k
Co j j
r
j
k jk
k j
k
Ri v e j
k j
k j j
jk
kk
j
k j
k
j
k
jk
jk j
k k
j jk
k
Riv e r L ea
j
k j
kjkj
j
k j
k
j j
k ne k j j
k Legend
kk
j o ur jk
jk
B Springs
e
Y
X
rin
j
k Swallows
rivers
Cathe
Mymm shal l Brook
0 1 2 4 6 8 10 © Crown Copyright/database right 2008. An Ordnance Survey/EDINA supplied service.
65
Figure 3.2: Solid Geology of the Study Area

Ri
Riv
er
Mi
Chadwell h
Ri mr Spring r As
ve am ve ±
v e r R ib
rL Ri
ea Y
X Y
X
River Beane
(o
rL
ee
) Y
X
Y
X Ne
w
R
SOURCE Arkley Hole k

j
iver
SITE Spring j
k
Y
X Y
X
k
j
j
k j
k
j
kk k j
k j
k j
k Y
X
j k
jjk
k k
jk
jjj k
j j
k
k
j j
k
s
k
k
jj
k
jk
jkj kj
j
k k
j
an
r
lb
Y
X
Ri v e
tA
S
Ne w
of
Water
l e End
Va j
k
j
k kk
jjk
j
e j
k
k
jj
k
j
k jk
k j
ln k
jk j
k
Co j j
r
j
k jk
k j
k Drift Geology
Ri v e j
k j
k j j
jk
kk
j
k j
k
j
k Alluvium
jk
jk j Valley Gravel
k k
j jk
k
Riv e r L ea
j
k j
kjkj Brickearth
j
k j
k Solid Geology Sand & Gravel
j j
k
j k ur
ne kjkj
jk Legend Taplow Gravel
jk
k o LONDON CLAY FORMATION
B Glacial Gravel
e
Springs LAMBETH GROUP
Y
X Boulder Clay
rin
LEWES NODULAR CHALK Fm
j
k Swallows SEAFORD CHALK Fm Clay-with-flints
NEWHAVEN CHALK Fm
rivers (UNDIFFERENTIATED) Pebble Gravel
Cathe
Mymm shal l Brook
0 1 2 4 6 8 10 © Crown Copyright/database right 2008. An Ordnance Survey/EDINA supplied service.
66
Figure 3.3: Solid and Drift Geology of the Study Area

Table 3.1: Lithostratigraphy of the Hertford district. After Bloomfield et al. (2004).
Table 3.2: Lithostratigraphy of the Chalk of the Hertford district. Based on Woods (2003).
Woods (2003) provided a detailed description of the various Chalk formations in the Hertford Dis-
trict (BGS Sheet 239), subdividing the Chalk Group into 5 lithostratigraphical units on the basis of
borehole resistivity and gamma log interpretations (Table 3.2). The lithostratigraphical units correspond
to the revised Chalk Group stratigraphy following Rawson et al. (2001) (Section 2.2.1). Cross-sections
produced from the resistivity log interpretations show the general southeasterly dip of the individual
Chalk subgroups. The New Pit Chalk outcrops in the far west of the study area. Moving east, the Lewes
Nodular Chalk is at outcrop up to Hatfield, and then the Seaford Chalk outcrops to the east of Hatfield.
Borehole correlations suggest that there are no major lateral changes in the development of Chalk Group
lithostratigraphical units across the district, except for slight thinning of the New Pit Chalk and a slight
expansion of the Lewes Nodular Chalk (Woods, 2003).
3.2.3.3 Geological Structure

Bloomfield et al. (2004) present a structural interpretation of the Chalk in the study area based on slope
aspect mapping. Land surface aspect mapping of a digital terrain model (DTM) was used to identify
lineaments that may reflect structures within the Chalk aquifer in the Hertford District. These lineaments
were then used in conjunction with additional information to develop a model of the geological structure
of the Chalk in the study area (the area between St Albans in the west and the Lee Valley in the east).
A number of cross-sections across the study area have been produced by BGS, VWP and TWUL
which show the Chalk and Palaeogene formations dip gently southeast towards the centre of the London
Basin (forming the northern limb of the London Basin syncline). Superimposed on this surface are a
series of inferred buckles and/or faulted folds, and a number of faults, typically downthrown to the west.
The BGS slope aspect mapping study (Bloomfield et al., 2004) identified 78 lineaments within the
area of the Hertford district (BGS Sheet 239), which were distinguishable as four distinct sets (Table 3.3).
Sets L1 to L3 are consistent with the limited field observations on Chalk fracturing available and with
previous published regional and generic model of fracturing in the Chalk. Due to their presence in
Palaeogene outcrop areas only, Bloomfield et al. (2004) infer that the east-west lineaments may not be
associated with structures in the underlying Chalk.
3.2.4 Hydrogeology
3.2.4.1 The Aquifer System

The Chalk, which underlies the whole of the study area, forms the principal aquifer in the region and the
main source of public water supply. In the Chalk upland of the northern and western parts of the study
area the Chalk is mostly unconfined. In places, the Chalk is overlain by Clay-with-Flints, although this
unit is not considered to limit recharge to the Chalk (Klinck et al., 1998). In the valley areas, the Chalk is
in hydraulic continuity with gravels, where present, and elsewhere semi-confined by alluvium (Buckle,
2002).
Within the Vale of St Albans and Middle Lee Valley, the Chalk is overlain by an interbedded se-
quence of Boulder Clay and Glacial Sands and Gravels which forms a multi-layered hydrogeological
unit (Buckle, 2002). In general this unit comprises a lower sand and gravel aquifer (the Proto-Thames
Table 3.3: Interpretation of lineaments. Based on Bloomfield et al. (2004).

Gravels) in hydraulic continuity with the Chalk, and an upper perched sand and gravel aquifer above the
Boulder Clay. The degree of continuity between the upper sand and gravel aquifer and the Chalk aquifer
system is dependent on the extent and thickness of clay layers, which impede vertical flow. In places,
the Chalk-PTG aquifer system is only partially saturated and unconfined, elsewhere it is semi-confined
to confined by Boulder Clay.
In the southeast of the study area the Chalk is overlain by the Palaeogene Deposits, comprising the
Reading Formation of the Lambeth Group (formerly Woolwich and Reading Beds) and the London Clay.
The Thanet Sands are absent in the area. The Reading Fm sediments are unsaturated over the majority of
the study area, although in the far southeast of the study area they become saturated, and together with
the Chalk are confined by the London Clay aquiclude.
From a hydrogeological point of view, a unit known as the ‘Basal Sands’ is used to describe the
Palaeogene sediments that are in hydraulic continuity with the Chalk aquifer of the London Basin. In
general the ‘Basal Sands’ comprises the Thanet Sand and the lower part of the Lambeth Group. The top
of this unit is defined non-stratigraphically as the lowest clay greater than 3 m thick in the Palaeogene
succession (Board, 1972). In the study area, the London Clay aquiclude acts as the confining layer to the
Chalk-Basal Sands aquifer unit (Buckle, 2002).
3.2.4.2 Geomorphological controls

Geomorphology appears to be an important control on aquifer properties of the Chalk (Section 2.7.2):
transmissivity values appear to differ considerably between the valleys (dry and flowing) and the inter-
fluves (Allen et al., 1997). The origin of this pattern is discussed in Section 2.4.3.
Within the major valleys (Thames and Colne), high yields are gained from boreholes close to the
rivers. The high transmissivities recorded, (e.g. 25 000 m2 day−1 for the Chalk at Medmenham) are
in part attributable to high degree of leakage to from the River Thames (either directly or via the Grav-
els). Nevertheless, the high flux of groundwater flowing through the major valleys is likely to enhance
the aperture of existing fractures, and thus high transmissivity values are expected within the Chalk
in these areas. From the available data, Allen et al. (1997) state typical values in the range 1500 to
3000 m2 day−1 . Little quantitative data are available for dry valleys, although it is generally assumed
that the dry valleys show similar transmissivity values to flowing valleys. Allen et al. (1997) report that
transmissivity values of between 400 and 1000 m2 day−1 , and storage coefficient values of 10−3 to
10−2 were recorded for dry valleys near Chesham and High Wycombe. Within the valleys, putty chalk
can reduce the permeability of the aquifer. Putty Chalk is present beneath the wider sections of valley,
and has been detected in sections of the Thames and the Colne valleys. The formation and properties of
putty chalk are discussed further in Section 2.2.4.
Within the interfluve areas, in high ground away from flowing or dry valleys, transmissivities are
generally considered to be low. Although Allen et al. (1997) could not obtain any pumping test data for
the interfluve areas within the Chilterns, low transmissivities (<50 m2 day−1 ) and storage coefficients
(0.01) were obtained in interfluve area of the Kennet Valley in Berkshire.
Pumping tests conducted at the Northern New River wells indicate transmissivities in the range
1000 m2 day−1 to 4000 m2 day−1 . This equates to a hydraulic conductivity values between 3 m day−1
and 30 m day−1 .
3.2.4.3 Lithological controls

The individual Chalk formations vary in their fracture characteristics and hydraulic properties (Ta-
ble 3.4). It is likely that lithostratigraphic variation will affect the hydrogeology in the area. Indeed,
Atkins (2004) suggest that the change from New Pit Chalk to Seaford Chalk at outcrop/subcrop in the
Hatfield area may explain the observed difference in flow behaviour between the areas east of Hatfield
and the area to the west.
In general, the Seaford Chalk is thought to show the highest hydraulic conductivity (Table 3.4), al-
though there is little published information available on the variation between formations. Hardgrounds
are typically associated with higher flows. Marls are relatively impermeable and represent barriers to
vertical flow. However, preferential flow paths may develop above marl layers (Section 2.4.3). The de-
gree of hydraulic continuity between Chalk formations can be variable, with the potential for significant
head gradients.
Table 3.4: Chalk Group Aquifer Potential. After Mortimore et al. (1990).
3.2.5 Karstic Features

Evidence for geomorphological karstic features in the Chalk of Hertfordshire is abundant, and has been
well documented (e.g. Whittaker, 1921; Harold, 1937; Walsh and Ockenden, 1982. The distribution
of karstic features is related to the geology and hydrology of the area (Section 2.4.3), most notably the
Chalk-Palaeogene contact. The distribution of the karstic features in shown in Figures 3.2 and 3.3.
A number of stream sinks occur along the Chalk-Palaeogene contact between Radlett and Hert-
ford, and within the Lambeth Group inlier at Cuffley Brook. Bloomfield et al. (2004) report that the
density of stream sinks is highest where the Palaeogene is dominated by clay facies; where the strata
are more sandy, recharge may occur directly into the Chalk without the generation of surface streams
(although the density and size of dissolution pipes developed beneath the cover may be greater). Most
stream sinks are fed by small, usually ephemeral, streams, although larger perennial streams occur. The
sinkholes/swallow holes allow rapid percolation of surface water to the Chalk water table.
The greatest concentration of stream sinks is in the North Mimms–Water End area. The behaviour
of the swallow holes in the North Mimms/Water End area has been described by Whittaker (1921),
Walsh and Ockenden (1982), Harold (1937). The area is part of the surface water catchment of the
Colne, and includes the Mymmshall Brook, its tributary the Catherine Bourne, and the Welham Green
Brook, which meet at Water End. The upper streams rise in the Palaegene escarpment, where they are
underlain by London Clay, and flow northwards. The streams disappear underground in a number of
sinks within the lower parts of the main valley and tributary valleys, which are underlain by Lambeth
Group deposits (Reading Fm) and/or Chalk. The swallow holes that appear to receive the greatest volume
of water are located at Water End (Walsh and Ockenden, 1982). The actual sinks used by the streams
depend on the flow (Walsh and Ockenden, 1982). During dry periods, the feeder stream sinks at several
points before reaching the main sinkhole complex. During periods of wet weather, flows reach the Water
End swallow holes and a lake forms in the depression occupied by the swallow holes. When the capacity
is exceeded, the area overflows to the upper reaches of the River Colne via a (normally dry) channel
which flows beneath the A1(M) which runs north-south through Hatfield (Figure 1.1).
A second major set of swallow holes occurs in the Hatfield area (Whittaker, 1921). It is thought
that that it drains east to the large springs at Arkley Hole, and may be linked to the conduit system
draining the Water End sinks. Also, numerous swallow holes (sinks or dolines) were mapped by Price
et al. (1989) near the M25/M1 junction (Figure 1.1) at Bricket Wood in the far west of the study area
. Many of the sinks are associated with the margins of dry valleys and were interpreted by Price et al.
(1989) as having been caused by groundwater issuing from the Chalk during the Pleistocene period under
periglacial conditions.
The major springs, which may be fed by karstic conduits, are the Chadwell Spring, and the Arkley
Hole Spring. Whittaker (1921) described several other springs between Amwell and Rye House, and
between Hoddesdon and Broxbourne. Chadwell Spring, a large spring just west of Hertford, is thought
to act as an estavelle (Whittaker, 1921): it runs turbid after heavy rain indicating conduit flow, but in
drier weather, it ceases to flow and takes water from the New River. Bloomfield et al. (2004) report that
it is suspected that the Arkley Hole springs are fed by the sinks in the Hatfield area and possibly the
Palaeogene outliers to the north, although it is not clear what evidence this is based on.
3.2.6 Karst Flow

Section 2.4.2 described the tracer tests that have been undertaken in the Hertfordshire Chalk. The karstic
connections established by Harold (1937) and Cook (2010) are summarised in Figure 3.4. The Hertford-
shire tracer tests indicated rapid (several kilometres per day) flows from the Water End area to locations
in the Lea Valley 15 km to 20 km away. The dispersion of the detected tracer along a Section of the
Lea Valley several kilometres in length indictes that the system of conduits through which which the
74
Figure 3.4: Established tracer connections in Hertfordshire. After Cook (2010)

flow occurs is widespread. Furthermore, the series of tests show that the individual pathlines followed
by flow are not constant and may vary with water level. Cook (2010) considers that the major conduits
in Hertfordshire are most likely to be developed at or just below the zone of water table fluctuation. This
interpretation is supported by relatively stable groundwater elevations east of Hatfield (Section 3.3): the
water table may be controlled by the high transmissivity of such features.
Cook (2010) integrated the information from the historic and recent tracer tests, in combination
with a consideration of structural controls on groundwater flow, to form a new quantitative conceptual
understanding of the function of the karstic flow system in Hertfordshire. The tracer tests undertaken
in the study area suggest that there is a distributive karst flow system in Hertfordshire developed in
a broadly north-east direction between North Mymms and the Lee Valley (Figure 3.28). Tracing from
Water End and the Catherine Bourne has provided evidence that connectivity to the karstic features of the
Mymmshall Brook system extends along the entire Palaeocene feather edge between the Lee Valley as
far south as Turnford PS and at least as far west as south east Hatfield and also possibly to north-western
Hatfield. The overall system appears to comprise a recharge area comprising a convergent network of
conduits centered around the Mymmshall Brook Catchment and the North Mymms water table mound.
This then drains via a solution-enhanced pathway adjacent to the Chalk-Palaeocene boundary and via a
distributive network to springs in the Lee valley.
This distributive karst flow system is characterised by rapid, low attenuation transport. Cook (2010)
interprets the progression of tracer arrivals to suggest that flow paths could be coincident with the pattern
of surface karst and swallow holes between the feather edge of the Palaeocene outcrop and the River
Lee. This provides an alternative interpretation to earlier conceptual models, e.g. Buckle (2002), which
suggested a fan like series of major flow routes between Water End and the Lee Valley. Cook (2010)
proposes that the pattern of tracer breakthroughs could suggest that the Northern Loop of the Palaeocene
Feather Edge and River Lee is short-cut by the subsurface karst system, with a more direct karst com-
ponent to the central Lee Valley, and a dispersive component beyond Arkley Hole to the northern Lee
Valley.
Cook (2010) takes the consistency of connections for all three tracers to major springs at Arkley
Hole and Lynchmill Spring to suggest that these springs terminate karstic flow routes established prior
to the more recent development of abstraction wells. Furthermore, recovery of all three tracer species
at Essendon PWS as well as similarities with both bromate concentrations and turbidity at Arkley Hole
implies that this groundwater source is also probably directly connected to the same system. Therefore,
the implication is that a rapid approximately east-west aligned major flow pathway exists between Es-
sendon and the Southern Lee Valley, principally to the Lynchmill Spring, but also a likely distributary
connection to Amwell Marsh, the Rye House/Rye Common Area, Hoddesdon and further south to Turn-
ford. Cook (2010) notes that this flow pathway is approximately sub-parallel to the Hoddesdon Syncline
which points to a structural influence on the flow regime.
A number of public supply abstraction wells show connections to the karst network. These are
generally located relatively close to known springs (i.e. Broadmeads PWS and Chadwell Spring, Amwell
Marsh PWS and Emma’s Well, Rye Common and Rye House Spring, and Hoddesdon/Middlefield Road
and Lynchmill Spring) and the majority have extensive adit systems. It is likely that long-term operation
and aquifer development around the abstraction wells has encouraged the convergence of rapid flow
paths to these discharge points.
The tracing by Cook (2010) also indicated that rapid flow paths appear to extend further west
beyond the zone of main karst development into the Vale of St. Albans. Surface karst features are
not apparent within the Vale of St. Albans due to the extensive glacial deposits covering the Chalk.
The karst system in the Vale of St. Albans appears to be less continuous and less well developed than
the main karst network along the Palaeocene feather edge, perhaps restricted as a result of infilling,
weathering and/or erosion of karst features with increasing distance from the Palaeocene feather edge
(or within ‘Geomorphic Zone 2’ of Maurice et al. (2006)). Flow velocities in this area are more variable,
and generally lower. The breakthroughs suggest a general decline in velocity and mass recovery with
distance from the Palaeocene outcrop. Transport is significantly attenuated, perhaps suggesting multiple
flow paths reflecting a range of dissolution enhanced fissure sizes. This higher attenuation may also, at
least partly, be a consequence of the tracer injection locations being boreholes, and therefore not directly
connected to the conduit system, resulting in dilution and dispersion before the karst system is reached
(Worthington, 2003; White, 2003). In addition to the tracer evidence, small conduits have been observed
in boreholes in the northern part of St. Albans CL:AIRE (2002).
3.2.6.1 Karst flow in relation to bromate contamination

The injection locations used by Cook (2010) were approximately 1 km south east of the Bromate Con-
taminant Source (MS2 Coliphage, Harefield House Borehole), approximately 9 km from the source of
contamination (Phi X174 phage, Comet Way Borehole) and within a sinking stream close to the Water
End swallow hole complex (Serratia Marcescens phage, North Mymms) which was located south of the
bromate occurrence in the Chalk aquifer, but had been shown by the 1920s and 1930s tracer tests to be
connected to locations currently affected by bromate.
The interpretation of Cook (2010) is consistent with karst flow paths intersecting bromate affected
groundwater in the Hatfield area. However, tracers from Comet Way borehole suggest a by-pass of
Hatfield PWS by karst flow paths in this area. Bromate may be entering the karst flow system to the
North and East of the Comet Way borehole. In the area west of Hatfield, whilst some rapid flow pathways
exist, they are weathered and probably poorly connected and so do not dominate bromate transport as
they do east of Hatfield.
The observed spread of tracer from Water End matches closely the pattern of groundwater sources
affected by bromate contamination between Hatfield and Turnford and the inferred travel times show
close agreement with interpretation of the effects of abstraction at Hatfield PWS (Section 3.5) on bro-
mate concentrations at the Lee Valley sources. Therefore, between Hatfield and the Lee valley, bromate
transport appears to be in large part controlled by groundwater flow in Chalk karst which is influenced
to some extent by abstraction at Hatfield PWS which has been shown to be connected to the karst flow
system.
Tracer was not detected at Chadwell Spring in the recent tracer tests (Cook, 2010), despite being
detected in the 1920s and 1930s tests. Cook (2010) suggests that flow to this spring has been affected by
recent changes in abstraction patterns and it now derives water from further north, outside the catchment
area of the tracer. This is supported by evidence from water chemistry (Section 3.4). This could also
explain why bromate concentrations are typically lower at Chadwell Spring than at other locations to the
south along the Lea Valley.
Cook (2010) proposes that the pattern of tracer breakthroughs could suggest that the Northern Loop
of the Palaeocene Feather Edge and River Lee is short-cut by the subsurface karst system, with a more
direct karst component to the central Lea Valley, and a dispersive component beyond Arkley Hole to the
northern Lea Valley. This could provide the mechanism by which the bromate concentrations tend to
be higher in the more southerly Wells than the NNR Wells since they receive a more direct and a less
diluted karst component.
3.2.7 Groundwater–surface water interactions

Buckle (2002), Entec (2000) and Atkins (2004) describe the groundwater-surface water interactions in
more detail. A summary is provided below:
3.2.7.1 River Colne and Tributaries

As described in Section 3.2.6, the Mimmshall Brook swallow hole system allows rapid transfer of surface
water to the Chalk water table. Also, during wet periods, surface water overflow from the Water End
swallow holes flows to the River Colne (Section 3.2.2). The section of the Colne around the North
Mymms area, immediately downstream Water End overflow, is underlain by alluvial sediments over
sand and gravels which are in hydraulic continuity with the Chalk, and leakage from the river to the
surface water system is thought to occur.
In the area between North Mimms and Tyttenhanger Park, and the Ellenbrook, the Chalk-PTG
groundwater aquifer is separated from the river by a continuous (although <3 m in places) layer of
Boulder Clay. The shallow perched sand and gravel aquifer overlying the boulder clay is considered to
be in hydraulic continuity with the surface water system. This system is likely to have been influenced
by gravel extraction activities, which may have locally increased the connectivity between the Chalk
aquifer and the river.
Boulder Clay is absent in some areas over the section downstream of Tyttenhanger Park to the
confluence with the River Ver, which suggests potential for greater connection between the river and the
Chalk-PTG groundwater system. The river may be influent during periods of high groundwater levels
and effluent over some sections during periods of low water levels.
3.2.7.2 River Ver

Along most of its length the River Ver is underlain by shallow alluvium and valley gravels. The Ver has
high baseflow fed by the Chalk. However, locally near to abstractions there is leakage of river water to
the ground, especially in the vicinity of St Albans.
3.3. Piezometry 78
3.2.7.3 River Lee

The Upper Lee, like the Ver, flows within an incised valley underlain by shallow alluvium over Chalk. It
is a Chalk-fed stream, but has a lower base flow than the Ver. Downstream of Welwyn Garden City the
river is underlain by shallow alluvium and drift, which includes a layer of Boulder Clay that separates the
Lee from the Chalk-PTG aquifer system. Further downstream to Water Hall, there appears to be more
surface water–groundwater interaction, and at Water Hall the river is underlain by sands and gravels in
continuity with the Chalk.
Further downstream of Water Hall, the Lee is underlain by alluvium and drift. In some places the
alluvium lies directly over valley gravels and/or Chalk. It is expected that in this section the river is
in hydraulic continuity with the Chalk aquifer system, but the degree of hydraulic interaction may be
impeded by the vertical permeability of streambed sediments.
3.2.8 Chalk–Drift Groundwater interactions

3.2.8.1 Hatfield Quarry and Former British Aerodrome Site
A detailed review of water level data and borehole information from the RMC Hatfield quarry site was
undertaken by Buckle (2002) to assess the hydraulic relationships between the Chalk-lower sand and
gravel (PTG) aquifer and the upper shallow sand and gravel aquifer (UGD). The data indicated that the
Chalk-PTG aquifer experiences a range of conditions both seasonally and spatially across the site. In
some boreholes water levels vary from being below the top of the chalk, or within the PTG, to being
above the Boulder Clay; in others the lower aquifer is always confined with water levels observed to be
above the top of the PTG.
Water levels in the UGD aquifer are between 3 m and 7 m above those in the Chalk-PTG aquifer.
It appears therefore that under normal water level conditions there may be the potential for downward
movement of groundwater from the upper perched UGD to the Chalk-PTG, but not vice versa.
A similar situation is observed, with a difference in water levels of 6 m to 8 m at the former British
Aerodrome site. The degree of downward movement is dependent on the permeability of the Boulder
Clay and its lateral continuity (Buckle, 2002).
3.3 Piezometry
Long-term water level data is available for a number of observation boreholes across the catchment
(Figure 3.5). These boreholes form part of the Environment Agency monitoring network. The long-term
rainfall and water level variations are shown in Figure 3.6, and a detailed time series for Orchard Garage
monitoring well, located close to the source site in Sandridge is shown in Figure 3.7.
Details of the Chalk piezometry, groundwater movement and water level fluctuations in the area
have been collated and described by Entec (2000), Buckle (2002) and Atkins (2004). In general, water
levels in the Chalk respond to variations in rainfall. Locations in the western part of the study area, on
the upland interfluve between the River Lea and River Ver show the largest fluctuations (approximately
5 m to 8 m), and locations nearer to the river valleys (e.g. in the west of the study area near the River
Colne, or in the east of the study area, close to the River Lea and River Mimram) show less pronounced
±
TL11/6
_
^
TL31/1
_
^
TL31/98A
TL11/55 TL31/3 _
^
_
^ TL20/6D _
^
_
^
TL10/50 TL10/8A
_
^ TL10/14 _
^
_
^
TL20/14 TL20/49
_
^ _
^
TL10/63
_
^
3.3. Piezometry
_
^ Environment Agency long-term observation boreholes
OBH Ref Name First Year Last Year

TL10/14 THE CAMP, ST ALBANS 1974 1990
TL10/50 EXPRESS DAIRY 1959 2004
TL10/63 ALL SAINTS CENTRE 1986 current - 2008
TQ19/19C
_
^ TL10/8A SMALLFORD NURSERIES 1974 current - 2008
TL11/55 ORCHARD GARAGE 1988 current - 2008
TL11/6 AMWELL CORNER 1963 2001
TL20/14 NORTH MYMMS PARK 1986 current - 2008
TL20/49 GOBIONS WOOD OBH 1988 current - 2008
TL20/6D ROBINS NEST HILL 1994 current - 2008
TL31/1 BURY FARM 1964 current - 2008
TL31/3 BRICKENDONBURY 1990 current - 2008
TL31/98A CHELSEA COTTAGE 1976 1994
TQ19/19C RADLETT P.S. 1986 current - 2008
0 1 2 4 6 8 10
Kilometers
79
Figure 3.5: Environment Agency monitoring network long-term water level monitoring locations
200
100
160
120
80
40
0
Total Monthly Rainfall (mm)
(Rothamsted Station)
TL 11/6
80
TL 10/50
TL 11/55
TL 10/14
TL 20/49
TL 10/8A
TL 10/63
TL 19/19C
60
TL 20/14
TL_11_6
TL_10_50
Water Level (m OD)

3.3. Piezometry
TL_11_55
TL 20/6D TL_10_14
TL_10_8A
TL 31/1 TL_10_63
40
TL_20_49
TL 31/3 TL_19_19C
TL_20_14
TL 31/98A TL_20_6D
TL_31_1
TL_31_3
TL_31_98A
20
Jan-60 Jan-70 Jan-80 Jan-90 Jan-00 Jan-10

80
Figure 3.6: Rainfall and Environment Agency monitoring network water level variations. Locations of monitoring wells are shown in Figure 3.5
3.3. Piezometry 81
fluctuations (typically less than 3 m). The presence of sand and gravel drift deposits above the chalk is
also likely to reduce water level fluctuations as a result of greater storage capacity.
Notable fluctuations in water levels are summarised below:
• Between 1990 and 1992, water levels fell to some of their lowest recorded levels, following a
period of relatively high water levels between 1981 and 1988.
• Between 2000 and 2003 (particularly 2001), water levels rose to the maximum recorded levels.
• Between 2003 and 2006, water levels declined but started to rise again over 2007.
200
Orchard Garage TL 11/55
Rothamsted Station 6600 Lee Chalk

160
(Location 028)
120
SMD
80
40
Water Level (mAOD) Monthly Rainfall (mm)

0
160
120
80
40
0
82
80
78
76
74
Figure 3.7: Water Level, Soil Moisture Deficit and rainfall at Orchard Garage Monitoring Well
A series of piezometric contour maps have been produced by Entec (2000), Buckle (2002) and
Atkins (2004). The maps show that the overall pattern of piezometry has remained reasonably constant
over the 1990s and early 2000s (although data up until 1997 is less extensive than later dates, particularly
in the areas between Sandridge and Hatfield). There is some uncertainty in delineating contour lines
where insufficient data points were available, notably between Tyttenhanger PS and Netherwild PS. It
is thought likely that the most significant water balance shifts in the Upper Lea area probably occurred
prior to 1970 as a result of significant changes in abstraction regimes. The long-term average piezometry
is shown in Figure 3.8.
On the basis of the piezometric data, groundwater flow tends to follow topography away from
the Chalk upland in the north and north-west, where groundwater elevations can exceed +100 m OD,
towards the confined Chalk in the south and south-east, where groundwater elevations are less than
+20 m OD. Within the Lee Valley in the vicinity of the NNR sources, flow directions tend to follow the
course of the river such that groundwater generally flows from north to south (Atkins, 2004). Ground-
water elevations continue to fall south-east towards the Lee Valley and ultimately south towards Central
Broadmeads P.S.
!
A 80 !A!
± 78 !A
A
76
74 70 66 Chadwell Spring
72 A!
68 64 !
Amwell Hill P.S. A
62
60 !
A
58 56 Essendon P.S. Rye Common P.S.
50
! Waterhall P.S. Middlefield Road P.S.
A !
A !
A
A!
Stonecross P.S. Bishops Rise PS.
Hatfield P.S. Hoddesdon P.S.
!
A ! 45
Holywell PS Raw
A !
A
Broxbourne P.S.
! Roestock P.S. 40 30
A Tyttenhanger P.S.
56
!
! A
A
Turnford P.S.
3.3. Piezometry
North Mymms P.S. 35 25 !

A
!
A
20
64 62
Netherwild P.S.
50
!
A
! Public Water Supply BH

A
0 0.5 1 2 3 4 5 © Crown Copyright/database right 2008. An Ordnance Survey/EDINA supplied service.
82
Figure 3.8: Average piezometry 1998 to 2008. Contour levels are in m AOD. Arrows indicate groundwater flow direction
3.3. Piezometry 83
London. There is a groundwater divide between units contributing to the River Lee and those contribut-
ing to the River Colne. This divide runs through the central part of the study area between Sandridge
and North Mymms.
Superimposed on the general north-west to south-east pattern of groundwater movement, Buckle

(2002) note a number of features, in addition to anthropogenic abstractions and discharges, which exert a
significant local influence on groundwater flow. These features include the Radlett–North Mymms area
recharge mound, the Mimmshall Brook and the Water End swallow holes, the Colney Heath-Hoddesdon
syncline, and groundwater-surface water interactions.
The Radlett–North Mymms area recharge mound can be seen on piezometric maps as a groundwater
high aligned along the Palaeogene escarpment. The London Clay cover is absent in places, exposing the
Chalk and the sands and clays of the Reading Formation, thus allowing recharge to the Chalk aquifer to
occur more readily in this area compared to the surrounding areas. Recharge is also contributed to by
the presence of swallow holes in this area, notably the Water End complex. This feature is seen to be
present even during extended periods of low rainfall. Groundwater moves in all directions away from
the mound, although the greatest flow is towards the north-east. Piezometry and groundwater flow in this
area is not well defined, and is considered to be sensitive to changes in recharge Buckle (2002). Water
flow in this area is also influenced by the abstractions to the northwest and west of the recharge mound.
The swallow holes in the North Mymms-Water End area are discussed further in Section 3.2.5.
Walsh and Ockenden (1982) note that the there is an apparent barrier to flow coincident with the Shenley-
Broxbourne anticline, and a preferential flow direction along the line of the Colney Heath-Hoddesdon
syncline. This may indicate a high permeability conduit or zone linking the mound and the River Lee.
Buckle (2002) invokes the presence of such a zone, straddling the zone of water table fluctuation and
underlain by a lower permeability main body of Chalk, to explain the persistence of the mound. When
operative under high flow conditions, the zone allows removal of the groundwater from the recharge area.
However, under low recharge conditions, the recharge mound is maintained by the lower transmissivity
of the main body of Chalk in which flows are much reduced.
3.3.1 Groundwater flow in the Sandridge-St Leonard’s Court Area
Sandridge is situated at the confluence of two dry valleys, and the St Leonard’s Court (SLC) source site
itself is situated on the northern edge of the valley down from the confluence. The dry valley trends
south-east from Sandridge, following the general trend of a major joint/fracture set. Roberts (2001)
suggests that preferential flow paths may have developed within the upper levels of the Chalk in the
Sandridge area as the dry valley represents a local discharge area for the Chalk aquifer, and it was
probably also a tributary of the former course of the Thames through the Vale of St Albans. Piezometric
maps show that in the Sandridge area groundwater movement is in general from west-north-west to east-
south-east. However, information is not available in sufficient detail to allow more precise delineation
of flow direction along the line of the dry valley. The groundwater flow in the vicinity of the SLC site is
discussed further in the Chapter 5.
3.4. Regional hydrochemisty 84
3.3.2 Abstractions and Discharges

Abstractions within the central part of the study area exert a significant local effect on the piezometry,
causing a depression of water levels in immediate vicinity of the pumping borehole. The main TVW
abstractions are grouped at Roestock, Tyttenhanger and Hatfield and Netherwild PWS. TWUL ground-
water abstractions are concentrated along the River Lee and New River, including Amwell Marsh, Rye
Common, Hoddesdon, Broxbourne and Turnford PWS.
It is likely that groundwater pumping in the North Mymms area and the Lee valley influences the
natural flow of the karstic conduit system between and is likely to have a significant influence on the
natural flow regime. A reduction in springflows from Chadwell Spring can be directly correlated with an
increase in groundwater abstractions within the catchment (Hydrotechnica, 1988).
Significant dewatering activities are undertaken by Lafarge Aggregates at Tyttenhanger Quarry, and
subsequent discharges made to the River Colne. Additionally, gravel extraction activated in other areas
(e.g. Hatfield Quarry) may result in changes to the natural piezometry. Removal of superficial deposits
may in some instances lead to increased recharge to the Chalk aquifer system. Dewatering of working
areas may also be required during quarrying.
3.4 Regional hydrochemisty

The regional hydrochemistry, in relation to the NNR sources, was assessed by Hydrotechnica (1988),
and is summarised in Atkins (2004). On the basis of major ion chemistry, five major water types were
identified:
Type IA Background regional waters, characterised by calcium bicarbonate waters with low potassium
and magnesium concentrations. Occur extensively to the north of the River Lee and to the north
and west of Ware.
Type II Recharge/surface waters, with low alkalinity and high pH, nitrate, suplphate, chloride and
sodium concentrations. Grouped around the North Mimms area swallow holes and represent rapid
recharge waters.
Type III Mature groundwaters. Typically present in the confined section of the aquifer and have char-
acteristically very low to negligible nitrate. Found in the confined section of the Chalk.
Type IV Polluted groundwaters.
Type V Mixed groundwaters. These have characteristics intermediate between Type IA and Type II
waters. Wide spatial scatter and occur throughout the region, but found mainly at locations south
and west of the River Lee in the areas known to be affected by rapid recharge mechanisms.
The groundwater abstracted from the NNR sources appears to contain varying proportions of Type
IA and Type II groundwaters. Furthermore, the Type IA groundwaters can be further distinguished as
those derived from the north of the River Lee towards Ware and Stevenage (Type IA-N), and those de-
rived from the west of the area around Sandridge (Type IA-W). Atkins (2004) point out that determining
3.5. Scavenge Pumping at Hatfield Pumping Station 85
the proportion of Type 1A-W water (which contains bromate) contributing to each of the sources would
be an effective way of estimating likely bromate concentrations. To do this would require that Type
IA-N waters could be more readily distinguished from Type IA-W waters, for example on the basis of a
contaminant (other than bromate) present in one body of water but not the other.
The data were interpreted as suggesting that there is seasonal variation in dominant flow directions
between the Water End swallow holes, the Essendon area and the Lee valley in the vicinity of the NNR
wellfield. During the winter months flow directions away from the Essendon area appear to be predomi-
nantly northeast or east, whilst in the summer months they have less of a north-easterly component.
3.5 Scavenge Pumping at Hatfield Pumping Station

3.5.1 Introduction
In July 2005 Three Valleys Water (TVW) and Thames Water Utilities Limited (TWUL) began a scavenge
pumping trial at the TVW disused Public Water Supply (PWS) source at Hatfield. The source site
of the bromate contamination in Sandridge is hydraulically up-gradient of Hatfield PWS. The purpose
of the trial was to determine the influence of pumping Hatfield on bromate values elsewhere in the
contamination plume.
The Hatfield PWS source was taken out of supply in May 2000 due to bromate contamination. Prior
to this, the PWS source was used fairly continuously and on average abstracted close to its licence value
of 9 Ml day−1 . Between the cessation of pumping at Hatfield in 2000 and the start of the pumping trial in
2005, there was an increase in bromate concentrations recorded at Essendon. Also, the TWUL Northern
New River (NNR) sources, which have been monitored since mid-2001, appeared to be increasingly
affected by bromate contamination. It was therefore thought possible that pumping at Hatfield originally
acted to intercept some of the bromate released into the aquifer from the source site. Since cessation of
pumping, the sources down-gradient of Hatfield have been detrimentally affected by increased bromate
contamination. However, this hypothesis could not be verified owing to a lack of bromate data for all
sources prior to Hatfield ceasing abstraction.
The Hatfield Scavenge Pumping has provided the opportunity to make a quantitative statistical
analysis of connections between Hatfield and sites across the catchment to the Lea Valley. The objectives
of the statistical analysis are:
• to assess whether there is a statistically significant relationship between Hatfield abstraction rate
and bromate concentration at Essendon and the NNR sources;
• to assess if the relationship differs between sources.
The Hatfield source has been pumped at rates up to 9 Ml day−1 while bromate, bromide, sulphate
and chloride concentrations have been monitored at the TVW Hatfield and Essendon sources and the
TWUL Northern New River Sources (see Figure 3.1 for locations). Pumping has continued into 2009.
Initial analysis as an internal report to TWUL and TVW (Fitzpatrick, 2007) considered data up until the
end of December 2006 only. This thesis considers the data up until the end of December 2008.
3.5.2 Data sources

Data for bromate and bromide concentrations at the twelve monitored sources were provided by Three
Valleys Water (TVW) and Thames Water Utilities (TWUL), along with daily abstraction rates and water
level data. Water levels at a selection of Observation Boreholes (OBH) were provided by TVW.
Daily and monthly soil moisture deficit (SMD) data (MORECS) were obtained from the Envi-
ronment Agency for the Chilterns E-Colne and the Lee-Chalk areas. Daily rainfall was obtained for
four locations in the area: Mill Green, Darnicle Hill, Broadmeads and North Mymms. The Environ-
ment Agency also provided stage-discharge relationships and monitoring data for the Mymmshall Brook
catchment which had been derived from flow gauging work undertaken by Atkins.
3.5.3 Data handling

Raw data were managed using a Microsoft Access database (Section 4.2). Queries were used to select
data for statistical analysis and graphing. Where concentration results were recorded as below detection
limits, the value at the detection limit was used for data analysis purposes. Minitab software was used to
perform the statistical analysis (correlation and regression).
3.5.4 Abstraction rates at Hatfield

The pumping test of the Hatfield source commenced on 29 July 2005. This analysis considers data
up until the 31 December 2008. In general, continued periods of abstraction range from 3 Ml day−1 to
8 Ml day−1 , with brief periods up to 9 Ml day−1 (Figure 3.9). Abstraction has been intermittently halted
due to surcharging events after periods of heavy rainfall. The test was halted for prolonged periods of
time between January and May 2006, and October 2006.
In order to prevent surcharging events, abstraction rates have generally needed to be maintained
at lower rates during the winter and spring months. However, between November 2007 and June 2008
abstraction rates were maintained at relatively consistent high values of ∼ 6 Ml day−1 to ∼ 8 Ml day−1 .
3.5.5 Bromate and Bromide time series trends

Figures 3.10 to 3.19 show the time series for bromate and bromide concentrations at the monitoring loca-
tions, including raw data, monthly average data, and a ‘deseasonalised’ average trend (Section 3.5.5.1).
3.5.5.1 Data processing and Statistical Methodology

Bromate and bromide concentration data are available from May 2000 for the Hatfield and Essendon
sources, and from May 2001 or September 2001 for the Northern New River sources. Sampling inter-
vals are approximately weekly for Hatfield and Essendon. The sampling frequencies for the NNR wells
progressively increase from quarterly between May 2001 and July 2002, to monthly between July 2002
and January 2003, fortnightly between the end of January 2003 and the end of Aug 2003, and subse-
quently weekly. During the scavenge pumping trial, sampling frequencies increase to daily for all twelve
sources.
Monthly average concentrations were calculated from the raw concentration data. Where data were
not available, the monthly time series was completed by taking the mean of data for the previous and
160
for Rothamsted Station, Herts

120
Monthly Rainfall (mm)

160
80
Soil Moisture Deficit (SMD)

for Chilterns (East) - Colne
120
40
80
0
40
Hatfied Abstraction Rate (litres per day)
6
9.0x10
8.0x106 0
6
7.0x10
6
6.0x10
5.0x106
6
4.0x10
3.0x106
2.0x106
1.0x106
0
0.0x10
1-Jul-05 1-Jan-06 1-Jul-06 1-Jan-07 1-Jul-07 1-Jan-08 1-Jul-08 1-Jan-09
Figure 3.9: Abstraction rates at Hatfield PS between 31 June 2005 and 31 December 2008
following months. A ‘deseasonalised’ trend was estimated separately for data before and after the start
of the pumping trial. This was estimated following the method outlined in Chatfield (2004.):
1
2 x(t−6) + x(t−5) + . . . + x(t+5) + 21 x(t+6)
Sm (xt ) = (3.1)
12
where xt is the average monthly concentration at time t and Sm (xt ) is the deseasonalised component of
xt .
3.5.5.2 Seasonal Variations

The time series for Essendon bromate (Figure 3.11) exhibits distinct seasonal variation, with the lowest
concentrations on the cycle corresponding to the winter months (December, January, February) followed
by a gradual build up through Spring and Summer to concentration peaks in the early autumn months
(September/October) followed by a fall through November and December to winter concentration lows.
Annual differences between winter minimum and autumn maximum appears to be relatively constant at
approximately 10 µg l−1 . Seasonal variation is not as consistent in the Hatfield time series. However,
minimum seasonal bromate concentrations tend to occur between January and April after which concen-
trations peak in October and then fall back to lower concentrations during the winter/spring. Differences
in bromate concentrations are approximately 70 µg l−1 . The Hatfield pumping trial begins on 29 July
2005 when bromate concentrations are reaching their peak of the seasonal cycle. Concentrations are
rapidly decreased during pumping, rising in the spring of 2005 when continued abstraction was halted
at Hatfield between February and May 2006. Concentrations then generally decline through to October
once pumping is resumed. Distinct seasonality is not evident after the start of the pumping trial; peaks
appear to coincide with periods of Hatfield switch-off rather than regular seasonal variation.
Seasonality in bromate concentrations is also evident in the NNR time series. Amwell Hill (Fig-
200
6140 Chilterns - Colne

Rothamsted Station
160 160
120
120
SMD
80
80
40
40
0
0
Hatfield Bromate BrO3- Hatfield Abstraction

Hatfield switch off
Pump Test Start
Mean Monthly BrO3
600
Deseasonalised trend
Raw BrO3
Hatfield Abstraction Rate (x 106 litres day-1)

10
Concentration (µg l-1)
8
400
4
200
0 0

Hatfield Abstraction
Hatfield Bromide Br-
Hatfield switch off
Pump Test Start
1200 Average Monthly Br
Raw Br 10
8
800
4
400
0 0
Figure 3.10: Time series of bromate and bromide concentrations at Hatfield PS, soil moisture deficit, and
monthly rainfall.
200
6140 Chilterns - Colne

Rothamsted Station
160 160
120 120
SMD
80
80
40
40
0
0
Essendon Bromate BrO3- Hatfield Abstraction

60
Hatfield switch off
Pump Test Start
Mean Monthly BrO3
Raw BrO3

10
40
8
20
4
0 0
300 Hatfield Abstraction

Essendon Bromide Br-
Hatfield switch off
Pump Test Start
Average Monthly Br
Raw Br 10
200
8
100
4
0 0
Figure 3.11: Time series of bromate and bromide concentrations at Essendon PS, soil moisture deficit,
and monthly rainfall.
200
Rothamsted Station
160
6600 Lee Chalk

160
120 120
SMD
80
80
40
40
0
0
Chadwell Spring Bromate BrO3- Hatfield Abstraction

10
Hatfield switch off
Pump Test Start
Mean Monthly BrO3
8 Raw BrO3

10
6 8
6
4
2
2
0 0

Chadwell Spring Bromide Br-
Hatfield switch off
Pump Test Start
Average Monthly Br
Raw Br 10
200
8
100
4
0 0
Figure 3.12: Time series of bromate and bromide concentrations at Chadwell Spring, soil moisture
deficit, and monthly rainfall.
200
Rothamsted Station
160
6600 Lee Chalk

160
120 120
SMD
80
80
40
40
0
0
Amwell Hill BrO3- Hatfield Abstraction

50
Hatfield switch off
Pump Test Start
Mean Monthly BrO3
40 Raw BrO3

10
30 8
6
20
10
2
0 0

Amwell Hill Bromide Br-
Hatfield switch off
Pump Test Start
Average Monthly Br
Raw Br 10
200
8
100
4
0 0
Figure 3.13: Time series of bromate and bromide concentrations at Amwell Hill PS, soil moisture deficit,
200
Rothamsted Station
160
6600 Lee Chalk

160
120 120
SMD
80
80
40
40
0
0
Amwell Marsh BrO3- Hatfield Abstraction

50
Hatfield switch off
Pump Test Start
Mean Monthly BrO3
40 Raw BrO3

10
30 8
6
20
10
2
0 0

Amwell Marsh Bromide Br-
Hatfield switch off
Pump Test Start
Average Monthly Br
Raw Br 10
200
8
100
4
0 0
Figure 3.14: Time series of bromate and bromide concentrations at Amwell Marsh PS, soil moisture
200
Rothamsted Station
160
6600 Lee Chalk

160
120 120
SMD
80
80
40
40
0
0
Rye Common Bromate BrO3- Hatfield Abstraction

30
Hatfield switch off
Pump Test Start
Mean Monthly BrO3
Raw BrO3

10
20
8
10
4
0 0

Rye Common Bromide Br-
Hatfield switch off
Pump Test Start
Average Monthly Br
Raw Br 10
200
8
100
4
0 0
Figure 3.15: Time series of bromate and bromide concentrations at Rye Common PS, soil moisture
200
Rothamsted Station
160
6600 Lee Chalk

160
120 120
SMD
80
80
40
40
0
0
Middlefield Road Bromate BrO3- Hatfield Abstraction

50
Hatfield switch off
Pump Test Start
Mean Monthly BrO3
40 Raw BrO3

10
30 8
6
20
10
2
0 0

Middlefield Road Bromide Br-
Hatfield switch off
Pump Test Start
Average Monthly Br
Raw Br 10
200
8
100
4
0 0
Figure 3.16: Time series of bromate and bromide concentrations at Middlefield Road PS, soil moisture
200
Rothamsted Station
160
6600 Lee Chalk

160
120 120
SMD
80
80
40
40
0
0
Hoddesdon BrO3- Hatfield Abstraction

100
Hatfield switch off
Pump Test Start
Mean Monthly BrO3
80 Raw BrO3

10
60 8
6
40
20
2
0 0

Hoddesdon Bromide Br-
Hatfield switch off
Pump Test Start
Average Monthly Br
300 Raw Br 10
200
6
4
100
0 0
Figure 3.17: Time series of bromate and bromide concentrations at Hoddesdon PS, soil moisture deficit,
200
Rothamsted Station
160
6600 Lee Chalk

160
120 120
SMD
80
80
40
40
0
0
Broxbourne Bromate BrO3- Hatfield Abstraction

100
Hatfield switch off
Pump Test Start
Mean Monthly BrO3
80 Raw BrO3

10
60 8
6
40
20
2
0 0

Broxbourne Bromide Br-
Hatfield switch off
Pump Test Start
Average Monthly Br
300 Raw Br 10
200
6
4
100
0 0
Figure 3.18: Time series of bromate and bromide concentrations at Broxbourne PS, soil moisture deficit,
200
Rothamsted Station
6600 Lee Chalk

160 160
120 120
SMD
80
80
40
40
0
0
Turnford Bromate BrO3- Hatfield Abstraction

60
Hatfield switch off
Pump Test Start
Mean Monthly BrO3
Raw BrO3

10
40
8
20
4
0 0

Turnford Bromide Br-
Hatfield switch off
Pump Test Start
Average Monthly Br
300 Raw Br 10
200
6
4
100
0 0
Figure 3.19: Time series of bromate and bromide concentrations at Turnford PS, soil moisture deficit,
ure 3.13), Amwell Marsh (Figure 3.14) and Rye Common (Figure 3.15) sources follow a similar pattern.
Maximum concentrations for each cycle occur in the spring months (March, April, May) and lowest
concentrations occur in the autumn months between September and December. However, a second peak
occurs in 2004 during September, October and November followed by a trough in December or January.
The Hatfield pumping trial begins at the end of July when bromate concentrations are on the downward
limb of the seasonal cycle. The seasonal cycle appears to continue, with peak concentrations occurring
in April and concentrations declining to a trough in October, and appears to continue through 2007 and
2008.
The Hoddesdon (Figure 3.17), Broxbourne (Figure 3.18) and Turnford (Figure 3.19) sources show
a slightly different seasonal trend. Maximum concentrations occur slightly later in summer and early
autumn (June, July, August, September) with a more rapid decline to minimum concentrations in winter
(November, December, January). The Hatfield pumping trial begins at the end of July when bromate
concentrations are at their peaks. Concentrations appear to be rapidly decreased during pumping. The
seasonal cycle appears to be perturbed, with peaks occurring in April, earlier than expected and during
the period that pumping at Hatfield was suspended, and the falling limb occurring earlier over May and
June when pumping was resumed. However, the seasonal cycle is resumed during the prolonged periods
of pumping in 2007 and 2008.
3.5.5.3 General Trends
For Hatfield the deseasonalised trend indicates an overall increasing trend in bromate concentrations
between 2000 and July 2005, although mean concentrations appear to level off in 2006, and there is
an indication of an overall decrease in concentrations by December 2006. Since then the variation in
bromate concentrations about the average trend has increased dramatically. For Essendon, the seasonal
variation is superimposed on an overall rise in bromate concentration, which is clearly indicated by the
‘deseasonalised’ trend line. Between July 2005 and October 2007 there is a general decline in bromate
concentrations with a few peaks, corresponding to periods of non-pumping from Hatfield. Between
October 2007 and December 2008 concentrations appear to level off, with an indication of a slight rise
in bromate concentrations from January 2008.
For the NNR wells the overall trends in bromate concentrations since the beginning of 2002 indicate
a general increase from 2002 until mid 2003. Subsequently, concentrations appear to level off, albeit with
fluctuations. The trend lines then show an increasing trend through the late 2004 and into 2005 (with
the exception of Amwell Marsh where concentrations continue to decline into 2005). Concentrations
generally fall with the start of pumping from Hatfield at the end of July 2006. In February and March
2006 all NNR sources show a significant rise corresponding to the period when pumping was suspended
at Hatfield. Concentrations fall once again in May 2006 following the Hatfield switch on. The trend lines
for bromate concentrations show a general increase from late 2006-early 2007 until the end of 2008, with
concentrations approaching those seen prior to the start of the pumping trial (and higher in the case of
Middlefield Road).
3.5.5.4 Relationship to Soil Moisture Deficit (SMD)

Figures 3.10 to 3.19 indicate that prior to the start of the Hatfield pumping trial (29 July 2005) the
seasonal cycle of bromate concentrations follows the seasonal cycle of soil moisture deficit (SMD).
For Essendon the relationship is particularly apparent (L. Lytton, pers. comm.) . Seasonal peaks in
bromate concentrations correspond to peaks in SMD, i.e. higher bromate concentrations at Essendon
correspond to dry conditions (high SMD). Prior to the start of Hatfield pumping, the time series for
Turnford, Broxbourne, Hoddesdon, Rye Common and Middlefield Road show a similar trend, although
generally the peaks in SMD are offset slightly from the peaks in bromate concentrations and occur
approximately 2 months later. For the more northerly NNR wells Amwell Marsh and Amwell Hill,
SMD peaks generally occur 3-4 and 4-6 months later than bromate concentration peaks.
The strength of the correlations between seasonal bromate concentrations and SMD before the start
of the Hatfield Pumping trial were assessed by comparing departures from the ’deseasonalised’ trend
lines with monthly SMD data. Time lags were assessed by comparing monthly concentrations with
SMD for the previous month, two months previously etc. The correlation coefficients are summarised
in Table 3.5. Essendon and the southern NNR wells (Hoddesdon, Broxbourne, Turnford) show a similar
relationship to SMD. The northern NNR wells (Amwell Hill, Amwell Marsh, Rye Common) show a
more delayed response to SMD.
Table 3.5: Pearson correlation coefficients for Bromate concentration and Soil Moisture Deficit (SMD)
before the start of the Hatfield pumping trial on 29th July 2005. SMD-X corresponds to the SMD value
X months previously. Shaded cells indicate the strongest relationship.
SMD SMD-1 SMD-2 SMD-3 SMD-4 SMD-5 SMD-6

0.270 0.529 0.613 0.501 0.290 0.008 -0.255
Hatfield
0.055 0.000 0.000 0.000 0.039 0.956 0.070
0.818 0.607 0.246 -0.097 -0.353 -0.531 -0.618
Essendon
0.000 0.000 0.082 0.500 0.011 0.000 0.000
-0.460 -0.642 -0.677 -0.514 -0.199 0.118 0.336
Amwell Hill
0.005 0.000 0.000 0.002 0.253 0.499 0.048
0.119 -0.207 -0.471 -0.567 -0.513 -0.380 -0.199
Amwell
Marsh 0.497 0.233 0.004 0.000 0.002 0.024 0.253
0.569 0.682 0.456 -0.002 -0.254 -0.352 -0.431
Middlefield
Rd 0.086 0.030 0.186 0.995 0.479 0.319 0.214
0.225 0.054 -0.110 -0.273 -0.353 -0.368 -0.315
Rye
Common 0.195 0.759 0.530 0.112 0.037 0.030 0.065
0.787 0.547 0.171 -0.136 -0.321 -0.449 -0.557
Hoddesdon
0.000 0.001 0.326 0.437 0.060 0.007 0.001
0.700 0.396 0.070 -0.213 -0.445 -0.595 -0.536
Broxbourne
0.000 0.018 0.691 0.219 0.007 0.000 0.001
0.672 0.341 -0.056 -0.350 -0.505 -0.610 -0.595
Turnford
0.000 0.045 0.749 0.039 0.002 0.000 0.000
Pearson
key correlation
P-value
After the start of pumping at Hatfield, for the majority of the monitored source wells, the relation-
ship between bromate concentration and SMD is not as strong, and in many cases appears contrary to
previous years. This is especially noticeable at Essendon, Hoddesdon, Broxbourne and Turnford where
a peak in bromate concentration between March and May 2006 is associated with low SMDs. However,
as discussed further in Section 3.5.6.5, this is also associated with the period when Hatfield was switched
off and therefore likely to indicate a rebound in bromate concentrations which were lowered by pumping
at Hatfield.
3.5.6 Bromate and Bromide: Relationship to Hatfield abstraction rates

3.5.6.1 General Trends
Bromate (as BrO3– ) and bromide (as Br – ) concentrations for the Hatfield, Essendon and the NNR sources
are shown alongside abstraction rates from the Hatfield source in Figures 3.10 to 3.19. Examination of
the bromate concentrations at Hatfield over the period of the pumping trial indicates that the abstrac-
tion rate from Hatfield influences the concentration at the Hatfield source. In general, the highest rates
of pumping correspond to the lowest concentrations of bromate recorded. Following periods of non-
pumping, there are considerable rises in concentration to pre-pumping levels. Concentrations fall as
pumping is resumed. Bromide concentrations follow a similar pattern as bromate concentrations.
Concentrations of bromate and bromide at Essendon appear to be influenced by changes in Hatfield
abstraction rate in a similar way to concentrations at Hatfield. The pump test begins in July 2005,
on the rising limb of the usual seasonal trend. The expected rise to September/October is not seen
and concentrations are rapidly decreased. The seasonal component is much less consistent after the
start of the pumping test. A peak in bromate concentrations occurs between March and May 2006
associated with the period that pumping at Hatfield was suspended, and contrary to the trend before
the Hatfield pumping trial (Sections 3.5.5.2 and 3.5.5.4). Therefore, variations in pumping rate appear
to be the dominant influence on bromate concentration. The relationship between pumping rates and
bromate concentrations is discussed further in Section 3.5.6.5. The rising trend appears to be curtailed
by the pumping test. There is an indication of an overall decreasing trend between August 2005 and
December 2006. Maximum concentrations (which occur during times of non-pumping) have not risen
to the projected trend line if it is projected until December 2006. However, the general trend appears to
level off in January 2007, and even increase slightly from December 2007 through until the end of 2008,
despite relatively consistent high rates of abstraction from Hatfield during this period.
Concentrations of bromate measured at the Northern New River (NNR) sources appear to follow a
similar pattern to Hatfield bromate concentrations, and show a general decrease in response to increased
abstraction rates from Hatfield. There appears to be a time lag of 3-5 days between turning off of Hatfield
abstraction and a noticeable rise in bromate concentration, or between resuming abstraction at Hatfield
and a noticeable fall in bromate concentration. For a number of NNR sources, bromate concentrations
show a more marked decline during the second phase of the test from May to October 2006. This may
be attributable to the heavy rainfall in May 2006.
The highest bromate concentrations at the NNR wells are recorded at Hoddesdon, Turnford and
Broxbourne which show concentrations between 10 µg l−1 and 50 µg l−1 . Rye Common, Amwell Hill
and Amwell Marsh show concentrations between 5 µg l−1 and 20 µg l−1 . Amwell End, Broadmeads,
and Chadwell Spring show bromate concentrations less than 0.6 µg l−1 for much of the test. These three
sources are the most northerly of the NNR sources. Bromate concentrations at Amwell End remain less
than 0.6 µg l−1 throughout the period of the test. Bromate concentrations at Broadmeads rise above
0.6 µg l−1 on three occasions throughout the test to 1.1, 1.3 and 1.5 µg l−1 , although these rises occur at
separate times which do not appear to be associated with changes in pumping rates. Bromate concentra-
tions at Chadwell Spring drop below detection limits during the periods of high pumping rates, and rise
to maximum of 3.3 µg l−1 in March and April 2006. For these three sources bromide concentrations do
not appear to show a response to the changes in Hatfield abstraction rate.
As with Essendon, the seasonal relationship between bromate concentration and SMD at Hoddes-
don, Broxbourne and Turnford appears to be perturbed by the effect of abstraction at Hatfield. Peaks in
bromate concentration occur in April, earlier than in previous years and during the period when pump-
ing at Hatfield was suspended, and the falling limb occurs over May and June, earlier than in previous
years and when pumping was resumed. For Amwell Hill, Amwell Marsh and Rye Common, the ex-
pected peak concentration occurs during the spring months (March, April, May) which coincides with
the period when Hatfield was switched off.
3.5.6.2 Data processing and Statistical Methodology
The abstraction rate at Hatfield for the day corresponding to the bromate and bromide sample was se-
lected. Bromate and bromide concentrations for each source well were compared to abstraction rates
from Hatfield and also abstraction rates from that particular source. Linear regression was undertaken
using Hatfield abstraction rate as a predictor. Statistical summary tables are included in Appendix C.
For the regression analysis, bromate concentrations for the each of the source wells were first com-
pared to Hatfield abstraction rate for the same day as the sample was taken (day T ), and then sequentially
to abstraction rates for the previous days (day T − 1, T − 2, T − 3 etc.). At each stage the strength of the
relationship was assessed by examining the fitted line scatter plot, the F -statistic and associated P -value
to determine the significance of the relationship, the standard error of the regression as a measure of the
dispersion of the data around the regression line, and the R2 value as a measure of the proportion of
the variation in the response (y) variable that is accounted for the variation in the explanatory variable
(x). The combination of these parameters was used to select the time lag for Hatfield pumping rate that
produced the best predictor variable for bromate concentration at each source well. The residuals for this
‘best fit’ regression relationship were examined to assess whether the assumptions of linear regression
(Section 3.5.6.3) were adhered to and therefore if the regression relationship could be used for further
hypothesis testing.
Where the regression relationship was deemed to conform to the assumptions, the hypothesis that
the coefficient of the regression line was significantly different from zero was tested by examining the
t-statistic and associated P -value from the Minitab output.
3.5.6.3 Assumptions of linear regression

R2 is a measure of the percent of the variation in the response (y) variable that is accounted for by
the variation in the explanatory variables. The F -statistic determines if the regression relationship is
statistically significant, i.e. that the apparent relationship between y and x is not likely to arise due to
chance alone. However, a large R2 or a significant F -statistic does not guarantee that data have been
fitted well.
In order to predict y (and a variance for the prediction) for a given x, it is assumed that the y variable
is linearly related to x and that data used to fit the model are representative of data of interest. However,
in order to test hypotheses (such as whether the slope significantly differs from zero, or estimate con-
fidence or prediction intervals from the linear regression, the following assumptions are also necessary
(Helsel and Hirsch, 1993):
• the variance of the residuals is constant (homoscedastic), i.e. it does not depend on the x or on
anything else (e.g. on time);
• the residuals are independent;
• the residuals are normally distributed.
Plots of residuals versus predicted values can indicate heteroscedasticity of the residuals, and the
normality of residuals can be checked by a normal probability plot. A plot of residuals against time can
indicate correlations between residuals over time.
3.5.6.4 Source Abstraction Rate

Fitted line plots for bromate concentration at each source well versus abstraction rate at the same source
for the period of the pumping test are shown in Appendix C. Plots are not shown for Amwell End and
Broadmeads because concentrations remained below detection limits throughout the period of the pump-
ing trial. A number of the NNR wells showed apparently statistically significant regression relationships.
However, inspection of the fitted line plots indicate that the lines are strongly influenced by the uneven
spread of the data points, with points available for a limited range of abstraction rates. It is therefore
difficult to be confident in these apparent relationships. Data are available over a much wider range of
abstraction rates at the Essendon source. At Essendon the data points show large variability, and there is
no clear relationship between bromate concentrations and Essendon abstraction rates. Source abstraction
rate was not considered as a predictor in further regression analysis because of the likely influence of
uneven distribution of data points on the correlation.
3.5.6.5 Hatfield Abstraction Rate

Fitted line plots for the ‘best-fit’ relationship (Section 3.5.6.2) and plots of residuals are included in
Appendix C. The regression parameters are summarised in Table 3.6.
Analysis of the residuals of the regression lines (Figure 3.5.6.5 and 3.5.6.5) indicates that the re-
gression lines do not conform to the assumptions of linear regression for two main reasons:
1. variation in the residuals is larger for larger fitted values (lower pumping rates);
Residuals Adherence to
1
Source Time Lag Linearity of relationship 2 3 4 Assumptions of
Homoscedasticity Independence Normality
Linear Regression?
For equation Bromate = A + B(Hat abst)
Hatfield
Although there is a lot of scatter about the The variation of the
Seasonality is not Residuals follow the
0 days line, a linear relationship describes the residuals is reasonably
evident in the residuals. normal line well.
‘average’ data trend reasonably well. constant.
Including SMD
Weak seasonality may

be present in the

residuals: the most
Reasonable, but may
Essendon positive residuals are
need to add seasonal
generally associated
Reasonably constant, term to equation
Although there is a lot of scatter about the with higher SMD and Residuals follow the
although slightly less for
2 days line, a linear relationship describes the vice versa. normal line well.
smaller fitted values.
‘average’ data trend reasonably well. Residuals are still
correlated with date.

The most positive
Including SMD Reasonable, but still
residuals tend to occur
dependent on date.
between July 2005 and
June 2006.
Seasonality is evident
The variation of the
with positive residuals in Residuals show slight
residuals is reasonably
Amwell Marsh the winter/spring and departure from May need to add
Although there is a lot of scatter about the constant, although slightly
negative residuals in the normality. seasonal term
6 days line, a linear relationship describes the less for smaller fitted
summer/autumn.
‘average’ data trend reasonably well. values and more variation
Residuals are still
between July 2005 and
Including SMD associated with date.
June 2006. Still correlation with date
residuals is reasonably Slight seasonality Reasonable, but may
Hoddesdon
Although there is a lot of scatter about the constant, although slightly present. need to add seasonal
Residuals follow the
4 days line, a linear relationship describes the less for smaller fitted term to equation
normal line well.
‘average’ data trend reasonably well. values and more variation
Seasonality is not
Including SMD between July 2005 and Reasonable, but still
evident in the residuals.
June 2006. dependent on date.
residuals is reasonably Slight seasonality Reasonable, but may
Broxbourne
Although there is a lot of scatter about the constant, although slightly present. need to add seasonal
5 days line, a linear relationship describes the less for smaller fitted term to equation
normal line well.
‘average’ data trend reasonably well. values and more variation Residuals are still
Including SMD between July 2005 and associated with date. Reasonable, but still
June 2006. dependent on date.
3.5. Scavenge Pumping at Hatfield Pumping Station
The variation of the Seasonality is evident

residuals is reasonably with positive residuals in
Turnford 4 days constant, although slightly the winter/spring and May need to add
Although there is a lot of scatter about the Residuals show slight
less for smaller fitted negative residuals in the seasonal term
line, a linear relationship describes the departure from
values. summer/autumn.
‘average’ data trend reasonably well. normality.
The variation of the Residuals are still

Including SMD residuals is reasonably associated with date.
Still correlation with date
constant.
1
Inspection of fitted line scatter plots
2
Inspection of residuals versus fits plots
3
Inspection of residuals versus order plot
4
Inspection of normal probability plot
103
Figure 3.20: Assessment of residuals for each ’best-fit’ regression for the response of bromate concentration to Hatfield abstraction rate. (1)
Residuals Adherence to
1
Source Time Lag Linearity of relationship 2 3 4 Assumptions of
Homoscedasticity Independence Normality
Linear Regression?
For equation Log(Bromate) = A + B(Hat abst)
with positive residuals in
Chadwell Spring Residuals follow the
Variation is less for smaller the winter/spring and May need to add
There is more variation at low pumping normal line well.
fitted values and less negative residuals in the seasonal term
6 days rates, however the position of the line looks
variation between July summer/autumn.
reasonable.
2005 and June 2006. Residuals are still Residuals show slight

Including SMD associated with date. departure from
Still correlation with date
normality.
with positive residuals in Residuals show
residuals is reasonably
Amwell Hill the winter/spring and departure from May need to add
There is more variation at low pumping constant, although more
negative residuals in the normality. seasonal term
5 days rates, however the position of the line looks for positive residuals.
summer/autumn.
reasonable.
The variation of the Residuals are still
Including SMD residuals is reasonably associated with date.
normal line well. Still correlation with date
constant.

Residuals show slight
The variation of the Slight seasonality Reasonable, but may
Middlefield Road departure from
Although there is a lot of scatter about the residuals is reasonably present. need to add seasonal
normality.
8 days line, a linear relationship describes the constant, although slightly term to equation
‘average’ data trend reasonably well. less for smaller fitted Residuals are still
Including SMD values. associated with date. Reasonable, but still
dependent on date.
1
Inspection of fitted line scatter plots
2
Inspection of residuals versus fits plots
3
Inspection of residuals versus order plot
4
Inspection of normal probability plot
104
Figure 3.21: Assessment of residuals for each ’best-fit’ regression for the response of bromate concentration to Hatfield abstraction rate. (2)
Table 3.6: Summary of regression parameters for the ‘best-fit’ regressions for the response of bromate
concentration to Hatfield abstraction rate.
2 2
Time 2 b R R
Source a R Standard error P-value
lag Hat abst SMD
For equation Bromate = A + B(Hat abst) + C(SMD)

Hatfield 0 days 33.3% 39.44 0.000 29.0% 4.3%
Essendon 2 days 49.0% 5.00 0.000 38.1% 10.9%
Amwell Marsh 5 days 39.5% 3.45 0.000 21.7% 17.8%
Rye Common 7 days 4.9% 6.52 0.020 3.2% 1.7%
Hoddesdon 4 days 30.2% 9.17 0.000 24.6% 5.6%
Broxbourne 5 days 27.1% 7.20 0.000 26.1% 1.0%
Turnford 4 days 14.0% 8.71 0.000 13.8% 0.2%
For equation Log(Bromate) = A + B(Hat abst)
Chadwell 6 days 14.0% 0.311 0.000 11.9% 2.1%
Spring
Amwell Hill 5 days 55.1% 0.228 0.000 4.6% 50.5%
Middlefield Rd 8 days 7.5% 0.200 0.002 7.1% 0.4%
a
time lag refers to the number of days between the Hatfield abstraction rate and the strongest
response in source bromate concentration.
b
The P-value refers to the hypothesis that the regression relationship is statistically significant,
i.e. that the apparent relationship between y and x is not likely to arise due to chance alone.
2. seasonality is evident in the residuals.
Heteroscedasticity in the residuals was improved in the majority of cases by transforming the data
by taking logarithms of bromate concentrations (as recommended in Helsel and Hirsch 1993).
Seasonality was apparent in the majority of the NNR wells, and at Essendon. Therefore, SMD was
included as a predictor in the regression relationship. In general, including SMD as a predictor in the
regression increased the amount of variation explained by the regression (increased the value of R2 ).
The residuals showed correlation to SMD. Positive residuals, indicating that observed bromate con-
centrations are above the fitted regression line, generally occur when SMD is low (winter and spring),
and negative residuals, indicating that observed bromate concentrations are below the fitted regression
line, generally occur when SMD is high (summer and autumn). This appears to be contrary to the rela-
tionship observed between bromate concentration and SMD prior to the start of the Hatfield pumping test
(Section 3.5.5.2). Positive residuals occur earlier than the expected seasonal peak in bromate concentra-
tions. For Essendon, the normal relationship of positive residuals with low SMD and negative residuals
with high SMD occurred.
However, the residuals are also dependent on date, even after the inclusion of SMD. It is apparent
that the most positive residuals occur mainly during the period between January 2005 and May 2006 and
the period over November 2006 when the Hatfield pumping test was suspended for prolonged periods
and only sporadic abstraction at rates less than 3 Ml day−1 occurred for sampling purposes. The positive
residuals may therefore reflect a rebound in bromate concentrations after a reduction due to the effects
of pumping. The fact that the positive residuals become less positive and more negative when pumping
was resumed in May 2006, and are negative between August 2005 and December 2005 and between
May 2006 and December 2006 when prolonged periods of higher abstraction rates occurred, indicates
that the effect of Hatfield abstraction rate on bromate concentration appears to dominate the seasonal
relationship between SMD and bromate concentration. The apparently contrary relationship between
bromate concentrations and SMD may therefore be an artefact of the timing of the abstraction rate
variations at Hatfield.
The relative effect of SMD and Hatfield abstraction rate is difficult to measure due to the timing
of the abstraction rate variations at Hatfield. The SMD curves reach their seasonal trough between
January and April 2006. The Hatfield pumping test was suspended between January 2005 and May
2006 and only sporadic abstraction at rates less than 3 Ml day−1 occurred for sampling purposes. The
higher abstraction rates occurred from August 2005 to December 2005 and May 2006 to December 2006
when SMD was in the higher part of its seasonal cycle. This results in an apparent positive correlation
between Hatfield abstraction rate and SMD. In order to separate these two effects it would be necessary
to maintain more constant abstraction rates at Hatfield over the full cycle of SMD variations.
A more constant period of abstraction occurred between November 2007 and June 2008, when ab-
stratction rates were maintained at relatively consistent high values of ∼ 6 Ml day−1 to ∼ 8 Ml day−1 .
Residuals still tend to show seasonal variations, which suggests that other seasonal variables are impor-
tant in controlling bromate concentrations.
The effect of rainfall is also complicated as a result of its relationship to abstraction rate at Hatfield.
Sewer surcharging events in response to heavy rainfall caused Hatfield abstraction to cease. As a result,
the abstraction rate at Hatfield is heavily influenced by rainfall in the catchment.
3.5.6.6 Testing the hypothesis of direct control by pumping at Hatfield

On the basis of the analysis of residuals and the discussion of seasonal variables, Hatfield, Essendon,
Broxbourne, Hoddesdon, Turnford, Middlefield Road and Rye Common were considered to adhere to
the assumptions of linear regression reasonably well to allow an estimation of confidence intervals and
hypothesis testing. In this way, the statistical analysis was applied to test the controlling effect of Hatfield
pumping on bromate occurrence at particular sites, to quantify the time lag before the effect is observed,
and to assess if the gradient of the regression relationship differs between sources.
Mean coefficients along with upper and lower 95% confidence intervals are given in Table 3.7. For
the response of bromate concentration to Hatfield abstraction rate, the null hypothesis that the coefficient
of the regression is equal to zero can be rejected at a significance level of 0.001 for Hatfield, Essendon,
Hoddessdon, Broxbourne and Turnford. Therefore a significant linear correlation exists between bromate
concentration and Hatfield abstraction rate for these source wells. The regression equations indicate that
an increase in Hatfield pumping rate of 1 Ml day−1 would result in a predicted decrease in bromate
concentration of 1.2 µg l−1 to 2.2 µg l−1 at the abstracting sources (with the specified time lag). For the
response of bromate concentration to soil moisture deficit (SMD), the null hypothesis that the coefficient
of the regression is equal to zero can be rejected at a significance level of 0.001 for Hatfield, Essendon,
and Hoddessdon only. Therefore a significant linear correlation exists between bromate concentration
and SMD for these source wells.
Table 3.7: Coefficients determined by the ‘best-fit’ regressions for the response of bromate concentration
to Hatfield abstraction rate.
Mean Upper Lower Mean Upper Lower

Time b b
Source a Coefficient P-value 95.0% 95.0% Coefficient P-value 95.0% 95.0%
lag
B CI CI C CI CI
For equation Bromate = A + B(Hat abst) + C(SMD)
Hatfield 0 days -12.50 0.000 -14.06 -10.94 0.245 0.000 0.161 0.329
Essendon 2 days -1.66 0.000 -1.86 -1.46 0.054 0.000 0.042 0.067
Rye 7 days -0.35 0.076 -0.73 0.03 -0.021 0.098 -0.045 0.004
Common
Hoddesdon 4 days -2.19 0.000 -2.72 -1.65 0.060 0.001 0.027 0.094
Broxbourne 5 days -1.65 0.000 -2.11 -1.19 0.020 0.170 -0.009 0.049
Turnford 4 days -1.19 0.000 -1.69 -0.68 -0.009 0.576 -0.042 0.023
a
time lag refers to the number of days between the Hatfield abstraction rate and the strongest response in source bromate
concentration
b
The P-value refers to the hypothesis that the slope of the regression line is significantly different from zero. [The p-value
is the probability of obtaining a result at least as extreme as that obtained by chance alone, assuming the truth of the null
hypothesis (coefficient = 0).] Values in bold type indicate that the null hypothesis can be rejected at a significance level of
0.05 or less.
Figure 3.22 compares the mean and 95% confidence intervals for the coefficients determined by the
linear regression for Essendon, Broxbourne and Hoddesdon. The confidence interval for Essendon is
smaller than for Broxbourne and Hoddesdon and Turnford. The confidence intervals all overlap, indi-
cating that the differences in mean coefficients are not statistically significant at a 95% confidence level.
Therefore, the gradient of the relationships between bromate concentration and Hatfield abstraction rate
for the monitored sources apart from Hatfield do not appear to differ significantly between each source.
3.5.7 Statistical relationships and bromate transport in Hertfordshire

Statistical analysis of the time lag between a change in abstraction rate at Hatfield PWS and the ob-
served change in bromate concentrations at down-gradient locations (Section 3.5.6.5) indicates compa-
rable times to the peak response times determined by Cook (2010) from the tracer test travel times from
the Mymmshall Brook catchment (Figure 3.23).
The inferred arrangement of the conduit system (Section 3.2.6) connects the three regions Hatfield
PWS, Water End and the major abstraction wells and springs in the Lee Valley along the main route of
karst flows. Therefore, ‘scavenge pumping’ at Hatfield PWS appears to influence flows within the karst
system via a direct connection, which in turn influences the observed concentration of bromate observed
at the Lee Valley wells and springs. Karst flow paths are likely to be intersecting bromate affected
groundwater in the Hatfield area. Tracers from Comet Way borehole suggest a by-pass of Hatfield PWS
by karst flow paths in this area (Section 3.2.6). If, as seems likely, bromate is entering the karst flow
system to the North and East of the Comet Way borehole, the by-pass of Hatfield PWS by karst flow paths
implies that scavenge pumping at Hatfield PWS may only be having a partial influence on down-gradient
Figure 3.22: Regression coefficients: means and 95% confidence intervals

2
4
0
109
Figure 3.23: Comparison of statistical response times for bromate concentration response to hatfield abstraction and tracer travel times from Water End. Based on Cook
(2010)
3.6. Single borehole dilution testing 110
concentrations (Cook, 2010).

Essendon PWS responds fastest to both abstraction at Hatfield PWS and tracer arrival as might be
expected given its closest proximity. Next to respond after four days are Amwell Hill PWS and Turnford
PWS, situated at the apparent northern and southern boundaries of the bromate affected area. Changes
in concentration then propagate towards the central part of the Lee Valley, the area around Rye Common
responding slowest. Cook (2010) suggests that the spatial pattern of response could reflect a partial
de-watering of minor conduits and flow paths which diverge from the main transport routes. The more
marginally affected areas in the Lee valley respond fastest whilst the major flow paths, carrying the most
water and therefore greatest bromate load being least and slowest affected.
3.6 Single borehole dilution testing

A series of single borehole dilution tests (SBDT) within existing boreholes in the Hertfordshire Chalk
were undertaken during 2008. These single borehole dilution tests were undertaken in conjunction with
the programme of point-to-point Natural Gradient Tracer Testing using Bacteriophage which are de-
scribed by Cook (2010).
The objectives of the SBDTs were:
• To determine the hydraulically active horizons within the selected boreholes in order to guide
injection strategies for the natural gradient point-to-point tracer testing;
• To use uniform injection SBDTs to obtain a direct measurement of horizontal specific discharge
(Darcy velocity);
Detailed methodology and interpretation of the results is included in Appendix D. The determina-
tion of horizontal darcy velocities is described within this Section.
Horizontal specific discharge (Darcy Velocity) was determined (according the the methodology
outlined in Figure 3.24) for three locations within the study area (location numbers refer to Figure 4.1):
• Nashe’s Farm BH (location 019)
• Harefield House BH (location 226)
• Comet Way BH (location 402)
3.6.1 Calculation of horizontal specific discharge (Darcy Velocity)

3.6.2 Methodology
3.6.2.1 Nashe’s Farm BH
The estimated horizontal specific discharge (Darcy velocity) at each 0.5 m depth section ranges from
3.0 m d−1 to 0.3 m d−1 (Figure 3.25). The highest values occur in the top 1.0 m below the water table
(between 20.5 and 21.50 m bD). This corresponds to the horizon of increased borehole diameter (Ap-
pendix D). The high darcy velocities indicate that more rapid flow occurs in this area and the widening
is indicative of a fissure. The inferred outflow horizon at 24.5 to 26.0 m bD corresponds to velocities of
∼0.5 to 0.6 m d−1 , whereas the intervening sections show velocities of ∼0.4 to 0.5 m d−1 .
Figure 3.24: Methodology for determination of specific discharge (darcy velocity) from the results of
the Single Borehole Dilution Tests. Based on Ward et al. (1998)
Nashes Farm - Specific discharge

darcy velocity (m/d)
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50
-20.50
-21.50
-22.50
-23.50
-24.50
Depth (mbD)
-25.50
-26.50
-27.50
-28.50
-29.50
-30.50
Figure 3.25: Specific discharge (darcy velocity) for each 0.5 m depth section at Nashes Farm. Estimated
Ct −Cb
using the methodology in Figure 3.24. Plots of ln C0 −Cb
are included in Appendix D. The value at each
section is estimated based on Based on Single Borehole Dilution test carried out at Nashes Farm 29
January 2008.
3.7. Conceptual Model for groundwater flow in Hertfordshire 113
3.6.2.2 Comet Way BH

4.4 m d−1 to 14.7 m d−1 (Figure 3.26).
Comet Way - specific discharge

0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0
-16.0
-16.4
-16.8
-17.2
-17.6
Depth (mbD)
-18.0
-18.4
-18.8
-19.2
Figure 3.26: Specific discharge (darcy velocity) for each 0.5 m depth section at Comet Way BH. Esti-
Ct −Cb
mated using the methodology in Figure 3.24. Plots of ln C 0 −Cb
are included in Appendix D. The value
at each section is estimated based on Based on Single Borehole Dilution test carried out at Comet Way
BH 4 February 2008.
3.6.2.3 Harefield House BH

0.3 m d−1 to 1.3 m d−1 (Figure 3.27).
3.7 Conceptual Model for groundwater flow in Hertfordshire

The information and data reviewed, interpreted and analysed in this chapter have been used to develop
a conceptual model for groundwater flow and bromate transport in the Hertfordshire Chalk aquifer (Fig-
ure 3.28). Between the source site in Sandridge and Hatfield, the flow of bromate contaminated water is
in an east-south-easterly to south-easterly direction, following the hydraulic gradient. The nature of the
Chalk in this area indicates a dominance of double-porosity characteristics, and consequently, bromate
will be highly attenuated. There are some karstic rapid flow sections to the west of Hatfield, although
these appear to be less well developed and/or connected than those along the Palaeogene boundary to
the east of Hatfield, and consequently, flow rates are slower and attenuation of solutes is greater. To
the east of Hatfield, there is a well developed main karst network, which follows the Palaeogene bound-
ary. Flow in the karst network is likely to cause dispersion of the bromate contamination along the Lee
Valley, approximately following the path of the conduit network, as far east as the northern New River
3.8. Summary and conclusions 114
Harefield House - specific discharge

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
-16.0
-17.0
-18.0
-19.0
-20.0
Depth (mbD)
-21.0
-22.0
-23.0
-24.0
-25.0
-26.0
Figure 3.27: Specific discharge (darcy velocity) for each 0.5 m depth section at Harefield House BH.
Ct −Cb
are included in Appendix D. The
value at each section is estimated based on Based on Single Borehole Dilution test carried out at Harefield
House BH on 22 January 2008.
wellfield. Transport in the karst conduits is characterised by low attenuation and high flow rates. There
is also likely to be a background flow of bromate contaminated water east of Hatfield that is influenced
by double-porosity characteristics.
3.8 Summary and conclusions

The topography, geology and hydrology of the study area result in a predominantly south-easterly
groundwater flow direction within the Chalk from upland areas in the north and west, where the Chalk
aquifer is largely unconfined, toward the south and east where the Chalk is overlain by Palaeogene
Deposits. Within the Vale of St Albans and Middle Lee Valley, the Chalk aquifer is overlain by an in-
terbedded sequence of Boulder Clay and Glacial Sands and Gravels; the degree of continuity between the
upper sand and gravel aquifer and the Chalk aquifer system is dependent on the extent and thickness of
clay layers. A number of abstractions in the Vale of St Albans, Hatfield, and Lea Valley areas influence
local hydraulic gradients. A series of single borehole dilution tests were undertaken in the area between
the source site at Sandridge, and the Hatfield area, which Darcy velocities of the order of 10 m day−1 .
Superimposed on the south-easterly groundwater flow, an extensive karst network associated with
the Palaoecene boundary allows rapid flow from the Water End area to the northern Lea Valley to the
north east. Within this chapter, rigorous statistical analysis of the time lag between the relationships
between a change in abstraction rate at Hatfield PWS and the observed change in bromate concentrations
at down-gradient locations was undertaken, and the relationships indicate direct connections between
Double-porosity
characteristics
dominate Influence of karst system
increases towards
High attenuation Palaeocene boundary
Flow rates
~10 m day-1 Low attenuation
Rapid flows
Main Karst (1-4 km day-1)
network follows
Palaeocene
boundary
3.8. Summary and conclusions
115
Figure 3.28: Conceptual model for groundwater flow in the bromate affected area of Hertfordshire. Position of conduits are based on the conceptual model developed by
Cook (2010). Flow rates and attenuation characteristics are inferred from the results of the single borehole dilution testing presented in Section 3.6 and tracer tests undertaken
by Cook (2010).
the Hatfield area and springs and abstraction wells in the Lea Valley. In combination with tracer testing
undertaken by Cook (2010), this suggests a conduit system which connects three regions Water End,
Hatfield, and the northern and middle Lea Valley. The results of the tracer testing from Water End
indicate rapid flows of the order of 1000 m day−1 along the Palaeocene boundary.
The information assessed and interpreted in this chapter has been used to develop a conceptual
model for flow and transport within the Hertfordshire Chalk aquifer which considers that double-porosity
characteristics dominate close to the source site in Sandridge and within the Vale of St Albans, resulting
in high attenuation of bromate. The main karst network is developed along the Palaeocene boundary and
allows rapid transport of bromate, with low attenutation, toward the Lea Valley.
117
Chapter 4
The evolution of bromate contamination in the

Hertfordshire Chalk
4.1 Chapter Objectives

The objective of this chapter is to use the available monitoring data to describe the spatial distribution
and temporal evolution of bromate across the catchment, and to interpret the distribution and evolution
of bromate in association with the conceptual model of the flow and transport system developed in
Chapter 3.
4.2 Bromate Water Quality Monitoring Programme

New Drinking Water Regulations came into force in December 2003, which introduced a new standard
for bromate (BrO3– ) of 10 µg l−1 . In May 2000, during the course of preliminary sampling, Three Valleys
Water (TVW) detected bromate concentrations of 135-140 µg l−1 , well in excess of this standard, at the
Hatfield Bishop’s Rise Pumping Station. As a precaution the source was removed from public supply.
In June 2000 a joint water quality monitoring programme was initiated, involving the Environment
Agency, the local authorities, and TVW to identify the extent and source of the bromate contamination.
Bromate contamination was found to extend across the catchment from Sandridge in the west to the Lee
Valley in east where low levels were detected at an Thames Water Utilities (TWUL) Northern New River
(NNR) sources between Ware and Turnford. Following the initial phase of water quality monitoring,
management of the monitoring programme was essentially assumed by the Environment Agency (EA),
with TVW and TWUL undertaking monitoring of their own sources.
The EA, TVW and TWUL continue to monitor water quality and water levels at a number of loca-
tions throughout the bromate impacted area. A total of approximately 370 locations have been monitored
at some stage over the period 2000 to 2007 at various frequencies, although only approximately 50 lo-
cations continue to be monitored on a routine basis. Protection of the public water supply boreholes has
been the main objective of the monitoring programme, which pays particular attention to key ‘indicator’
boreholes located within the main body and at the margins of the bromate affected area (the ‘plume’) to
assess plume boundary movement.
The role of collating the monitoring data was initially undertaken by VWP within an Excel spread-
4.3. Monitoring Data Quality 118
sheet. (Initially the monitoring data was held by a number of different organisations undertaking separate
monitoring.) The Environment Agency assumed the role of data management and transfer of the data
occurred in September/October 2001. The data was entered in the Agency’s WIMS database, and from
this the data was subsequently exported into an Access database (the ‘Bromate Monitoring Database’
). Although at the time, the majority of samples taken were being analysed at the VWP laboratory it
was always the intention to eventually transfer sample analysis to the Agency laboratory (this occurred
in October 2002). Automatic transfer of data occurs from the laboratory to the WIMS database. The
EA continues to manage all publically available monitoring data (some public water supply data are
excluded) within the database, including both water quality results and water levels. The database also
includes a variety of other information including monitoring location reference information (e.g. NGR,
owner, borehole depth, topographic elevation etc) and sample schedules (‘runs’).
In September 2005, the Environment Agency commissioned Atkins Limited to produce a factual
and interpretative report (Atkins, 2006) of the data and associated information obtained through the five
years of monitoring. In addition, Atkins amended/upgraded the monitoring database. In order to form
the basis of the data presented in this thesis chapter, the Bromate Monitoring Database (including all
data up to the end of December 2008) was provided by the Environment Agency. Monitoring data (up to
the end of December 2008) provided by TVW and TWUL was included within a modified version of the
database, the ‘UCL Bromate Monitoring Database’. The UCL database was linked to a GIS database.
4.3 Monitoring Data Quality

The quality of data available within the UCL Bromate Database is reviewed in sections 4.3.1 to 4.3.4
4.3.1 Sampling Locations

The total number of locations monitored for bromate across the catchment is ∼380 locations (Figure 4.1).
The core ‘plume’ was found to extend from Sandridge to Hatfield and therefore the majority of the sam-
pling points were located in this area. Of these, ∼235 are are chalk groundwater locations (including
31 public water supply borehole locations), ∼15 are gravel groundwater locations (mainly abstraction
boreholes for quarrying operations), ∼90 are surface water locations and an additional 11 are floodwater
sampling locations (associated with the 2001 flooding in the area of House Lane, Sandridge). Approx-
imately 30 additional monitoring locations were also introduced in response to circumstances such as
groundwater flooding in Sandridge, discharge of bromate contaminated water at Hatfield Quarry as part
of the quarrying operations, the discovery of low concentrations of bromate in boreholes drilled at Ashley
Road, St Albans to investigate a hydrocarbon spillage. The chalk groundwater locations comprise public
and private supply boreholes, observation boreholes (mainly pre-exisitng, but some purpose-drilled), site
investigation boreholes, and landfill monitoring boreholes. The diameter, depth, construction (e.g. open
hole, screen, lithology sampled) therefore varies considerably between locations.
4.3.2 Sampling Methodology

The Environment Agency’s standard sampling protocol specifies purging of three well volumes where
possible prior to sampling. Where purging of three well volumes is not possible, purging should be
!
!
!
A!
! A ! 374
! 389 A 373 378
!
220A
A
!
A
A
!A
!
A
A! 225
! !
A !
A
221AA
222! 223 028 017
A 025 372
!
079219221
A !
A ! 018
! 214 !
A A !
065
A
A A
!A! 220222 226 ! 380
A
! A! ! 218 A 387 381 !
A 063
080 216
!A
A !
A 020 A!!007A 163 069
!
!
A
! 215217 A! A
223
A !
A ! 082
A !
083 081
A
!
A
!
A
227
019
!
A
A!
022
A ! A
!
385
386
A
! 071A
212 073
A 067 072180A !A
! 070 A
!!165 074
AA
!
! A
A
!
! 076 166
A
! !
A
! 075
±
!
A !
A ! ! 062
A !
A A164 066!
A
! !
0 20 40 60 80 100 Meters 225 A211 A
! 068 ! ! 064
A
A 061
A
!
A !
A
162
!
A
228
!
A !
A 060
198 199
059 !
A
0 0.5 1A! !
A
034 !
!
A Kilometers A
123
!
A 298
1000
! A!
035 087
A A
! 295 322323324
161 004 !
043 !
A A 307308 321 A
!!
!
A
044 !
A !
! A
A
A 320
!
A A! 306
!
A !
A 354
038
! 309311 ! 319
A
A 305 ! 296
A A! 318
A! 036037 303 !
A !
A
008 088 297
031A ! !033 !
!A
!A
! ! !
A 404 A 301
224
A A
A!!A ! A
167
370371
A!
! 090 !
039 029 A ! A
!225 374389! 378 379 !! 331
!A
A ! 369
! A! 089 304
! A ! ! !A
A A376 A !A A
!A
!
A ! 226 !
A 377
! !
A!
A 018 A !387
A!A A
A 375 !
365367 A
!A ! 227
222223 !A!! !A! 373 191 263264 143 A
! A
A AA!A!A AA
!386
! A
!!! 999
019A A!!385
A 212
! 068 !A
!!A!AA
!
A
381
!
!A
A!
A! 157
180 ! 158
A !
A ! 266
A AA A
! !360364
! ! 009 406 359
!A
A !!
A A! A A358
! !
A 300
AA A
A!
!!A
156 !! 402
400 265 !262
A 408
!A! !A
228A ! 164 A!A 211 A
!
A A !409 356357
A
199 ! A! A !
!
! 160
A A!
! A 154 A! !A
155
! !
A 405
A!A!407
198 A !
203 A!A!
159
A
4.3. Monitoring Data Quality
148 A!A
201 !202
A!A 152 A!A ! 005
A
153 001197
! !A
!A
! !
A!205 235 294 315 299
A A
!A
A! ! !
200204 A ! 078
!
A A A A!
312313314 !
A
348 A
349! A! 171195
196 345A!
! 010b ! 170 A A
!316
!!
!
A
A!347 A 057
A
A 346 021102 192 A!
141 193 142 !
A
A!
! A!
!
011A! A 049
012013 !
A 086
!! 015 302
A !
A
!
A 006 !
014 A
! ! 140
A A !
A
145
0 A
! 2 4 6 8 10
Kilometers
119
© Crown Copyright/database right 2008. An Ordnance Survey/EDINA supplied service.

Geological Map Data © NERC 2008.
Figure 4.1: All bromate monitoring locations 2000-2008.

undertaken until stablilisation of physicochemical parameters is achieved. In general, for groundwater

monitoring undertaken by the Environment Agency (EA), or by their appointed subcontractors, the EA’s
standard sampling protocol was followed to ensure, as far as possible, consistency of results. Instances
when the standard sampling protocol could not be applied (e.g. due to wellhead restrictions) have gen-
erally been noted as comments within the database.
4.3.3 Analytical methods and detection limits

The typical analytical determinands are summarised in Table 4.1. Analytical methodology and method
detection limits (MDL) for bromate and bromide are summarised in Table 4.2.
Table 4.1: Typical analytical suite for water samples May 2000 to December 2008
Early sampling (2000-2001) Routine sampling Physical parameters

Bromate Bromate pH
Bromide Bromide Temperature
cations and anions Chloride Electrical conductivity
Ammonium Sodium Total dissolved oxygen
Total oxidised nitrogen Total oxidised nitrogen
Organic compounds by GCMS phosphorous∗
total organic carbon∗
∗Determinand excluded from approx 2001 when it was concluded that
no significant down-gradient migration was occurring
Table 4.2: Analytical methodology and detection limits for bromate analyses.
Laboratory Method Method Detection Limit

(µg l−1 )
Up to end of Dec 2001 VWP ? 1.0
Jan 2002 end of Sep 2002 VWP ion-chromatography 0.5
EA (Starcross) ion-chromatography 0.5
Oct 2002 to end of dec 2009 VWP ion-chromatography 0.5
TWUL ion-chromatography 0.6
For the period May 2000 to late October 2002, all samples were analysed by the VWP laboratory.
During this time, from January 2001, new analytical methods for bromate and bromide were introduced.
A duplicate analysis was undertaken to compare the old and new analytical method. This involved the
collection and analysis of approximately 20 field samples. Initial testing showed that bromate results
were 20 % lower for concentrations above 150 µg l−1 .
According to Buckle (2002), the laboratory uses an ion-chromatographic technique, and has NA-
MAS accreditation for the parameters analysed. The occurrence of ‘isolated and unexpected’ bromate
sample results, particularly for TVW operated sources (e.g. Roestock and North Mymms), recorded
above the detection limit of 1 µg l−1 , but generally less than 5 µg l−1 , has shown that the technique
is not completely reliable at low concentrations. Under normal circumstances, bromate and bromide
concentrations over the detection limit are subject to an estimated measurement error of ±10 %.
From October 2002, all samples were collected by the EA were analysed by the Environment
Agency’s Starcross Laboratory in Exeter. During the period of changeover, duplicates (comprising some
20 field samples) were analysed by both the Starcross laboratory and VWP to check for consistency.
According to the EA, the results of these comparisons indicated that discrepancies resulting from dif-
ferent methods or laboratories were unlikely to exceed 20 % for bromate and bromide concentrations
greater than about 5 µg l−1 . (It should be noted that the comparisons were not intended to be rigorous
statistical exercises, but to provide reassurance that discrepancies resulting from different laboratories
or methods would not significantly distort the broad picture of the distribution of bromate and bromide,
whose concentrations ranged over there orders of magnitude within the plume.)
Further to the above, in June 2005 an inter-laboratory comparison exercise was undertaken to check
the consistency of results. Triplicate samples were analysed by VWP, TWUL and EA laboratories.
Eleven samples were taken for the NNR wells and from TVW sources at Hatfield, Bishops Rise and
Essendon. A comparison exercise was undertaken for the three laboratories, similar to that described
above. Over the concentration range 10 µg l−1 to 350 µg l−1 the maximum deviation of an individual
sample from the mean of three samples was 23 %. For one sample close to the MDL results varied from
<0.5 µg l−1 to 2.3 µg l−1 . Results for bromide showed less variation.
4.3.4 Sampling frequency and completeness

Broadly, the monitoring programme had two main phases, and the frequency of sampling at each location
(Figure 4.2 to Figure 4.10) has varied according to the main objectives:
• Phase 1 – An initial phase with the primary objective of identifying the source of contamination
and the extent of the affected area. This phase was effectively achieved by the end of 2000;
• Phase 2 – A subsequent on-going phase to monitor concentrations and assess boundary migration.
In addition, a number of ‘specialised investigations’, involving sampling at additional/alternative

locations and/or more detailed and/or more intense sampling, have occurred in response to specific issues
such as monitoring of bromate in groundwater floodwaters, investigation of the source site (St Leonard’s
Court) in relation to Part IIA assessment, investigation of a petrol spillage at Ashley Road, investigation
of the effects on dewatering at Hatfield quarry on bromate and bromide, depth sampling at Hatfield
Business Park, investigation of bromate in the River Lee and Ellenbrook/Colne system. Resolution of
anamolies in the results, suspected to have been related to sampling or laboratory analysis have also
generated specific additional sampling aside from the main sampling programme.
The main phase of monitoring underwent significant ‘rationalisation’ in early 2002. Locations
which provided no additional information were omitted, and additional boreholes specifically drilled
by the EA and TVW to assist in the delineation of the contamination distribution and extent were in-
F
G
!
A
!
A
!
A
!
±
A
!
A
!
A
!A! F
G
A !
!! !
A !
A A A ! A
A!A! !
A
A !
A
! !
A
!
A !
A
!
!
A
A ! !!
A A !
!
A
!! !! !A
A
!A!
A !AA! !
A F
G F
! ! G
A AAA
F
G A!A !
A F
G !
Source Site !
A A A
! A GG
F F !GA! !
A
A !
AF
!
A FA
!G
A
!
A
!
A
! !
A
G
F
A
F
G
!
F
G
!
A A
!
A
Bromate G
F !
F
G A
Monitoring !
A !
A F
G
Frequency
! F
G F
G
A PWS_sym G
F
F
G F
G
G
F !
A
F
G
Yr_2000 !
A
!
A F
G !
A
!
A
!
A 1 - 20
F
G !
A
!
A 21 - 75 !
!
A
A !
A
!
A
!
A 76 - 180
!
A !
A
Groundwater_sym !
A
Yr_2000 F
G
!
A 1-4 !
A
! 5 - 10
A
!
!
A
A
! 11 - 27
A
!
A
Surface water !
A
Yr_2000 !
A
F
G 1-5
Boundary of Source Site
F
G 6 - 10 !
A
Former Works
F 11 - 25
G Building Footprint
0 1 2 4 6 8 10
Kilometers
122
Figure 4.2: Sampling frequency for bromate at each monitoring location in 2000
F
G !
A
!
A
F
G
!
A
± !
A
!
A
!
A F
G
!
A
!
A
!
A
!
A
! !
!
!
!
!
A!
!
A ! AA
!
A
A
A
! ! !!
A A
Source Site
! AA
A! ! A
A
A
!!!A
!!
A ! ! !
A
!
G
FA
!A A !A
A
!! A
! A A F
G G
F
!G
AF !
A
! !
!GF F
G
!
A
A!AA
F
G !
A A
!
A
!F
GA A! !
A !
A
!
A! A
!
G
F
A! A
A ! !
A
F
G
! A!
A F
! G
A !
F
G !
! A
A F
G A
A
! !A
A F
G
!
!A
!
A !A G
Bromate A
! A
A! F
G !
A
!
F A!
! A !
A
Monitoring A
G !
Frequency A! FA
!
A !
A
F
G !
A
!
A PWS F
G
G G !
A
Yr_2001 ! !
A
A
FF
! G
A
F AA!!
F
G
A!
!
!
A 1 - 20
F
G !
A
! 21 - 75
A
!
A
!
A
A! 76 - 180
!
A
Groundwater !
A
Yr_2001 !
A
!
A 1-4 !
! A !
A
!
A !
A !
A !
A 5 - 10 A
!
A
!
A
!! 11 - 27 !
A!
A
AA
!
A
Surface water
!
A !
A
Yr_2001 !
A
F
G 1-5
F
G 6 - 10
Former Works !
A
Building Footprint
F 11 - 25
G
0 1 2 4 6 8 10
Kilometers
123
!
A
G
F
!
A
F
G
± !
A
!
A
F
G
F
G
!
A
A! A! !
A
A!
!
A G
FG
FG F
G
F
! A ! !! F
G G
F
!
!A
A ! AA! A
! A ! F
G F
GF
G F
G
!
! ! AA
A
Source Site A ! A !
A !
A !G
F !
A G
F
A
!G
F !
F
GA!
!A A
!AG !G
AG
FGF!A !
A
F!!
F
A A
! AF
G A!A
A !
A !
A
!
F
G
A ! ! !
A ! F
G F
A!A
GG
F
A !
! ! GF GFA
AA F
G
! A
F
G A
Bromate !
A ! G
F
A !
A !
A !
A
Monitoring !
Frequency !
A A
! F
G
A F
G
PWS
G
FGF
Yr_2002 ! G F !
A
A FA
G !
!
A 1 - 20
F
G !
! 21 - 75
F
G A
A !
A
!
A
!
A 76 - 180
!
A
Groundwater
Yr_2002
!
A
!
A 1-4
A
! A !
!A
!
A 5 - 10 A! A! !
! !
A!A !
A
A 11 - 27 AA! !
Surface water
!
A !
A
A
Yr_2002 A!
F
G 1-5
F
G 6 - 10
Former Works !
Building Footprint A
F 11 - 25
G F
G
0 1 2 4 6 8 10
Kilometers
124
G
F
!
A
AF
G
!
!
A
! !
A
!A
A F
G
!
± !A
A
!
A!
!
A
!
A
F
G A
!
A
!
A
!
A ! !
A A
!
A !
!
A ! A F
G
A !
A
!
!
A ! !
A
A
A !
A ! !
A !
A F
G
A !
A G
F
!
Source Site
A! ! A! !
A !
A F
G
!G
AF !
F
G
A
!
A F
G !
A A
A
!
A !
A
!
A !
A !
! F
G A
A
F
GG
F
Bromate ! !
A A !
Monitoring A !
A
!
A
Frequency !
A
F
G !
A !
!
A A
PWS !
A
Yr_2003 !
A
!
A
!
A 1 - 20
!
A 21 - 75 F
G
! 76 - 180 !
A
A
!
A
Groundwater
Yr_2003
!
A 1-4 !
A
! 5 - 10
A !
A
!
A 11 - 27
!
A
Surface water !
A
Yr_2003
F
G 1-5
F
G 6 - 10
Former Works
Building Footprint
F 11 - 25
G
0 1 2 4 6 8 10
Kilometers
125
F
G
F
G
!
A
!
A
± F
G
!
A
!
A !
A
!
A
!
A !
A
A!
!
!
!
AA !
A! !
A !
A
A F
G F
G
F
G F
G
Source Site ! ! ! G
A F !
A !G
F
A!
A !
A A A !
A! A
!
A
!
A
!
A
!
A !
! ! F
G
A
A A ! !
A A
F
G
!A! F
G
A F
G
Bromate !
A !
A !
A
Monitoring !
A
!
A !
A
Frequency !
! F
G
A
!
A
A! !
A! A
PWS
Yr_2004
! 1 - 20
GA
F
!
A
!
A 21 - 75
!
A
! 76 - 180 !
A
A !
A !
A
Groundwater
F
G Yr_2004
!
A 1-4 !
A
! 5 - 10
A !
A
!
A 11 - 27
!
A
Surface water !
A
Yr_2004
F
G 1-5
F
G 6 - 10
Former Works !
A
Building Footprint
F 11 - 25
G
0 1 2 4 6 8 10
Kilometers
126
!
! A
A
F
G
!
A
± !
A F
G
!
A
!
A ! !
!
A A A
! !
A !
A !
A
!
A A ! !
A A A !
! A
!
!
!
A
A! A !
! A ! ! !
A
A
!
A
!!! A A
A !
A
! !
A !
AA
! !
AA ! !
!A
! ! !
! AA A A
A
A A F
G !
A F
G
G ! ! ! !
Source Site FA A !
! A A FA
G !G
F
A! A
! A
F !
A ! !
G !A
A A A
A !
!A !
A
!
A ! F
G
A
! !
A !
A
!
A A !
A
! !
A F
G
A !
A
!
A
!
A !
F
G
Bromate A
! !
A
!
A A !
Monitoring !
A A
Frequency F
G
!
A
PWS
Yr_2005
!
A 1 - 20
!
A
!
A 21 - 75 F
G !
A
!
A
! 76 - 180 !
A
A !
A !
A
F
G
Groundwater
Yr_2005
!
A 1-4 !
A !
A
!
A
! 5 - 10
A !
A
!
A
!
A
! 11 - 27
A F
G
!
A
!
A
Surface water !
A
!
A
Yr_2005
F
G 1-5
F
G 6 - 10
Former Works !
A
Building Footprint
F 11 - 25
G
0 1 2 4 6 8 10
Kilometers
127
!
! A
A
!
A
!
± A !
A
!
A
F
G
!
A
!
A
!
A
A! ! !
!
A A A
! !
A !
A
!
A A ! !
A A !
A
!
!
A
A!
!
! !
A ! !
A
!
!A
! A
A
AA! F
G !
! !
A !
! !
! ! !
!
A A
A AA A AAA !
A
!
A ! F
G G
F G
F
Source Site G
FA!! !
A !
A !G
AF
A A G F A ! !
!
A !
A A
A
!A!
!
A F
!G
A
!
A !
! !
A
!
A F
G A !
A
! !
A
A G
F !
A
!
A G
FGF
!
A !
F
G F
G
Bromate A
! !
A
Monitoring A
! !
A A !
A
Frequency F
G
!
A
PWS
Yr_2006 !
A
!
A 1 - 20
!
A
!
A 21 - 75 F
G !
A
!
A
! 76 - 180 F
G !
A
A ! !
A
A F
G
Groundwater
Yr_2006
!
A 1-4 !
A
!
A
! 5 - 10
A !
A
! !
A
A 11 - 27 F
G !
A
!
A
Surface water !
A
Yr_2006 !
A
F
G 1-5
F
G 6 - 10
Former Works !
A
Building Footprint
F 11 - 25
G
0 1 2 4 6 8 10
Kilometers
128
!
! A
A
!
A
!
± A !
A
! G
A F !
A
!
A
!
A !
A
! !
A !
A !
A !
!
A A ! !
A A
A
!
!
A!
! A !
! A ! ! !
A
A
!
A
AA
!!
A A A
!
A !
! A F
G !
A !
A !A
!
A !
! A! A ! A
!
A A A A! F
G !
A F
G
! !
A
! !
A A ! !G
F
Source Site F ! A
! A !
A A
G !
GAA F !A
A !
A !
!
A
A
!A!
!
A
!
A ! F
G A !
A A
! !
A !
A !
A !
A
A
!
A F
G !
! !
A A
A ! !
A G
F
A F
G
!
A !
F
G F
G
A
Bromate ! !
A
A
! !
A A !
A
Monitoring
Frequency !
F
G
A
Surface water_sym
!
A
Yr_2007
F
G 1-5 !
A
F
G 6 - 10 !
A
!
A
!
A
F 11 - 25
G !
A !
A
Groundwater
Yr_2007
!
A 1-4 !
A
! !
A
A 5 - 10
!
A
!
A
!
A 11 - 27
!
A
!
A
PWS !
A
Yr_2007 !
A
!
A 1 - 20
! Boundary of Source Site
A 21 - 75
Former Works !
A
!
A 76 - 180 Building Footprint
0 1 2 4 6 8 10
Kilometers
129
!
! A
A
!
A
± !
A F
G !
A
!
A
!
A
! !
A !
A !
A ! !
A A
A
!
! ! ! !
A
!
A
A
A
A!! A A
A !
A !
A !
A
!
A !
A !
! A
A
!
A
! !
A !
A G
!
Source Site A !
A A !
! A
A !G
AF F G
F
!
F
G !
!
A
A !
A
A!A
!
A
! !
A !
A A
!
A !
A
! !
A
A F
G F
G
!
A ! G
F
Bromate A
Monitoring !
A
!
A !
A
Frequency F
G
!
A
Surface water
!
A
Yr_2008
F
G 1-5
!
A
F
G 6 - 10
!
A
F 11 - 25
G !
A
!
A
Groundwater
Yr_2008
!
A 1-4
! 5 - 10 !
A
A
!
A
!
A 11 - 27
!
A
!
A
PWS !
A
Yr_2008 !
A
!
A 1 - 20
! Boundary of Source Site
A 21 - 75
Former Works !
A
!
A 76 - 180 Building Footprint
0 1 2 4 6 8 10
Kilometers
130
4.4. Delineating the Bromate ‘Plume’ 131
cluded. Sampling frequencies were also reduced at most locations, although the aim was not to reduce
the frequencies until a full year of data at regular intervals had been obtained.
In June 2002, sampling frequencies were further amended with the aim to monitor at a frequency
of two-monthly or less with the exception of a small number of locations retained at monthly frequency
due to their strategic role in providing an early warning of change. Furthermore, in November 2002 the
sampling schedule was rationalised again due to funding constraints. Approximately 47 locations have
been retained for on-going and regular monitoring.
4.4 Delineating the Bromate ‘Plume’

The areal extent of the bromate contaminated groundwater is defined by a margin encompassing a num-
ber of locations that are typically, but not always, below the method detection limit (MDL). Beyond this
margin, no bromate has been detected at monitoring locations. The margins are defined by results from a
number of locations in mid-to-late 2000 and early 2001; during this time the emphasis of the monitoring
programme was to identify the source of the bromate contamination and its extent. Many of these loca-
tions have been sampled just a few times (Figure 4.2 and 4.3), although some key locations have been
sampled routinely as ‘indicator boreholes’ to monitor the potential migration of the contamination.
Broadly, the distribution of bromate contamination based on monitoring data from 2000 to 2008
(Figure 4.11) to 4.19, is a core ‘plume’ of concentrations in excess of 50 µg l−1 extending from the St
Leonard’s Court source site in Sandridge down-hydraulic gradient as far as Hatfield, some 5 km to the
south-east. To the east of the core ‘plume’, the bromate contamination extends as far as the Northern
New River wellfield, with concentrations between 1 and 50 µg l−1 affecting locations distributed along
the path of the karst system.
The extent of the contamination (the ‘plume’ margin), the distribution of contamination within the
core ‘plume’, and temporal variations between 2000 and 2008 are described in the sections 4.4.1 to 4.4.5.
Integration of the new conceptual model of bromate flow and transport in the Hertfordshire Chalk (Sec-
tion 3.7) has allowed an alternative interpretation of the distribution of bromate contamination across the
catchment. Previous interpretations have been a broad ‘envelope’ of bromate contamination extending
east of Hatfield to the northern New River. However, this distribution is not justified by the monitoring
data available.
4.4.1 Up-gradient of the source site

In 2000, a number of samples of Chalk groundwater locations up-gradient of Sandridge gave results
below the MDL. In 2002, a purpose-drilled monitoring borehole was installed at Location 224 (Pound
Farm, Sandridge). This location has been monitored regularly between 2000 and 2008. The time series
indicates that bromate concentrations are below the MDL, with the exception of two occurrences of
low bromate concentrations in 2002 (30-35 µg l−1 ). Bromide concentrations are detected within the
normal background range (50-80 µg l−1 ), with a few samples at around 100 µg l−1 corresponding to
the occurrences of bromate concentrations above MDL. Location 224 therefore defines the maximum
bromate extent margin up-gradient of the source site.
4.4. Delineating the Bromate ‘Plume’
132
Figure 4.11: Annual average bromate concentrations at groundwater sampling locations in 2000.
133
134
135
136
137
138
139
140
4.4.2 Source site and Sandridge area

Groundwater bromate distribution beneath the source site is discussed in detail in Chapter 5, Sec-
tion 5.6.2.
Down-gradient of the source site, in 2000, the 1000 µg l−1 contour encloses locations 028, and 019
and 062. This zone remains relatively stable. From 2001, location 166, from 2002 location 226, and
from 2004 location 385 are included in this zone. Location 067 shows variable concentrations and in
some years shows average concentrations above 1000 µg l−1 , and other years below this.
The locations 028, 019, 020 and 226 provide good time series data over the period 2000 to
2008 (Figures 4.20 to 4.22). In general, the highest bromate (and bromide) concentration peaks for
028 and 019 were seen in 2000, after which concentrations declined significantly in early and mid 2001.
At 028, concentrations rose again over 2002 and 2003, but since 2004 concentrations have generally
shown a declining trend, albeit with a slight increase in mid 2007. As a consequence, 028 falls outside
the 1000 µg l−1 contour in 2006 and 2008 (within 300-600 µg l−1 contour interval), and in 2007 (within
600-1000 µg l−1 contour interval). At 019, concentrations remain relatively stable from 2002 to 2008.
At 226, bromate concentrations show a rising trend between early 2002 and mid 2005; between mid
2005 and 2008, concentrations are relatively constant, with a slight decreasing trend. Interestingly, the
start of this declining trend coincides with the start of the Hatfield Scavenge Pumping Trial (Section 3.5).
However, a decline is not seen at 019, and at 028 concentrations had already started to decline over
2004 and 2005.
With the exception of the very low concentrations seen over 2001, the concentration trend
at 028 and 019 broadly matches the water level trend (water levels decline from highs in 2001 to lows
in late 2006, rising slightly over 2007. However, at location 226, bromate concentrations only start to
follow the water level trend from early 2005. For all three locations, fluctuations in concentration also
appear to follow fluctuations in rainfall. The relationship between bromate concentrations and water
level is discussed further in Section 4.6.
To the north, between the 1000 µg l−1 contour and the northern margin, locations 020 and 022
show bromate concentrations between 30 µg l−1 and 50 µg l−1 , although 020 does show annual
average concentrations below this (2003) and above this (2001 and 2002).
4.4.2.1 Northern Margin (west of Hatfield area)

The northern extent of the bromate contamination is defined by locations 016 and 167 (old and new
borehole at Old Cottage, Green Lane, Hatfield), location 017, location 025 and location 007. These
locations have been monitored fairly regularly between 2000 and 2008 and time series for these locations
show bromate concentrations below the MDL, with the exception of sporadic occurrences of low bromate
concentrations (<10 µg l−1 ). Location 025 does show bromate concentrations up to 45 µg l−1 . In 2000,
a number of locations were monitored to the north of the aforementioned locations, and these all returned
concentrations below the MDL.
Location 007, which has consistently yielded bromate concentrations below the MDL, causes
the maximum extent contour to deviate from an apparently smooth curve between 017 and 016. This
Monthly Rainfall (mm) Orchard Garage (Ref. 028) 200

Rothamsted Station
160
6600 Lee Chalk

160
120 120
SMD
80
80
40
40
0
0
Water Level (mAOD)

82
80
78
76
74
6000
Hatfield swith off
Pump Test Start
Mean Monthly Br
Raw Br
Mean Monthly BrO3
4000 Raw BrO3

2000
1
Bromate conc. / bromide conc.
0.8
0.6
0.4
0.2
Figure 4.20: Time series of bromate and bromide concentrations at selected locations between Sandridge
and Hatfield, soil moisture deficit, and monthly rainfall.
Monthly Rainfall (mm) Nashe's Farm (Ref. 019) 200

Rothamsted Station
160
6600 Lee Chalk

160
120 120
SMD
80
80
40
40
0
0
Water Level (mAOD)

82
80
78
76
74
8000
Hatfield swith off
Pump Test Start
Mean Monthly Br
6000 Raw Br
Mean Monthly BrO3
4000
2000

1
0.8
0.6
0.4
0.2
Monthly Rainfall (mm) Harefield House (Ref. 226) 200

Rothamsted Station
160
6600 Lee Chalk

160
120 120
SMD
80
80
40
40
0
0
Water Level (mAOD)

82
Jan-00 Jan-02 Jan-04 Jan-06 Jan-08 Jan-10 80
78
76
Orcharge Garage (Ref. 028) 74
Harefield House (Ref. 226)
72
10000
Hatfield swith off
Pump Test Start
Mean Monthly Br
8000
Raw Br
Mean Monthly BrO3
Raw BrO3
6000
4000
2000

1
0.8
0.6
0.4
0.2
deviation is not supported by concentrations above 50 µg l−1 which are consistently monitored at loca-
tion 065, location 372 and location 378.
4.4.2.2 Southern Margin (west of Hatfield area)

The southern margin is less clearly defined by the monitoring data. Close to the source site, location 227
(BH beside Jersey Farm Pond), drilled in 2002, shows concentrations below the MDL, except for low
concentrations in 2002. Locations 059, 060, and 199 to the south of Hatfield Quarry show bromate
concentrations generally below MDL, with some intermittent low bromate concentrations. Location 162
generally shows low concentrations of bromate (<10 µg l−1 ) but concentrations fall below the MDL
in 2007 and 2008. Additionally, location 228, drilled in 2002, yields bromate concentrations below,
and occasionally just above, the MDL. Location 010b (Glinwell’s Nursery) similarly shows isolated
incidences above the MDL (these occur in late 2001 and early 2002). Generally, it is not possible to
discern trends from the time series. Further south/southwest towards St Albans, there are groundwater
monitoring locations that have shown bromate concentrations above the MDL, although these appear to
be relatively isolated.
4.4.3 Hatfield Quarry

Further down-gradient, around the Hatfield Quarry area, the spatial distribution is defined by loca-
tions 067, 068, 061, 064 066, 062, 166, 063, 069, 065 and 163, and more recently (since 2005) lo-
cations 385, 386, 372 and 378. In general, the bromate distribution is indicated by concentrations of
approximately 500 µg l−1 to 1500 µg l−1 in the central part of the quarry, and lower concentrations
in the range 100 to 500 µg l−1 in the northern part of the Quarry (Sutton’s Farm area). A number of
boreholes within the quarry (locations 070 to 076) were sampled once in 2000 and all yielded bromate
concentrations below the MDL. There is no information in the database on the construction of these
boreholes, and it is therefore not clear whether they represent groundwater from the Chalk or from the
superficial deposits.
The locations 067, 065 and 166 provide good time series data over the period 2000 to 2008 (Fig-
ure 4.23 to 4.23). Locations 067 and 065 both show a rise in bromate (and bromide) concentrations in
2002 compared to 2000 and 2001 concentrations. Bromate concentrations then remain fairly stable until
the end of 2005, after which concentrations at location 067 fall to levels comparable to 2000 and 2001
(with a particularly low concentration in October 2007). Locations 166 shows large fluctuations in its
time series, However maximum concentrations remain at similar levels between 2000 and 2008, and no
clear trend is discernible.
At location 068, bromate concentrations have decreased from between 10 and 25 µg l−1 over 2001,
to around 5 µg l−1 from 2002 to 2008. Although there is a gap in sample results between October 2002
and October 2005, the following observations can be made regarding trends in bromate concentrations
at locations around the Quarry area:
Monthly Rainfall (mm) Hatfield Quarry WM9 (Ref. 065) 200

Rothamsted Station
160
6600 Lee Chalk

160
120 120
SMD
80
80
40
40
0
0
Water Level (mAOD)

84
80
76
72
Hatfield Quarry WM9 (Ref. 065)
Hatfield swith off
Pump Test Start
600 Mean Monthly Br
Raw Br
Mean Monthly BrO3
Raw BrO3
400
200

1
0.8
0.6
0.4
0.2
Figure 4.23: Time series of bromate and bromide concentrations at selected locations in the Hatfield
Quarry area, soil moisture deficit, and monthly rainfall.

Rothamsted Station
160
6600 Lee Chalk

160
120 120
SMD
80
80
40
40
0
0
Water Level (mAOD)

84
80
76
72
Hatfield Quarry WM12 (Ref. 067)
Hatfield swith off
Pump Test Start
5000 Mean Monthly Br
Raw Br
Mean Monthly BrO3
4000
Raw BrO3
3000
2000
1000

1
0.8
0.6
0.4
0.2

Rothamsted Station
160
6600 Lee Chalk

160
120 120
SMD
80
80
40
40
0
0
Water Level (mAOD)

84
80
76
72

Hatfield Quarry WM16 (Ref. 166) Hatfield swith off
Pump Test Start
Mean Monthly Br
3000 Raw Br
Mean Monthly BrO3
Raw BrO3
2000
1000

1
0.8
0.6
0.4
0.2
• Locations 061 and 064 show an apparent decreasing trend in bromate concentrations
• Locations 062 and 066 show generally stable trend in bromate concentrations
• Locations 211shows a general increase (although concentrations are quite variable).
Since 2005:
• Locations 381 and 372 show stable concentrations
• Locations 385 and 386 show stable, or slight decline.
• Location 378 shows a decline in concentrations.
Additionally, locations 059, 060, 162 and 199 to the south of Hatfield Quarry define the southern
margin of the bromate contamination, and locations 007, 380, 381 and 373 to the north of Hatfield
Quarry define the northern margin of the bromate contamination. Locations 060 and 199 show only
sporadic occurrences of bromate above the MDL. Location 059 shows concentrations below MDL or
just above (<1 µg l−1 ): the incidences of concentrations above the MDL become more frequent after
June 2003, and three successive results between 2 and 5 µg l−1 occur in the latter half of 2006. Bromide
concentrations show no obvious trend between 2001 and 2008. In contrast, location 162 shows variable
concentrations up to ∼ 12 µg l−1 , with slight rising trend observable until the end of 2006. Subsequent
concentrations have been below the MDL. Bromide concentrations have also decreased over 2007 and
2008.
Around the northern margin, at location 380, bromate concentrations have decreased from between
1 and 3 µg l−1 between late 2005 and early 2007, to below the MDL over 2007 and 2008. Bromate
concentrations are much more variable at location 381, fluctuating between 1 and 5 µg l−1 (although
one concentration of 9.3 µg l−1 in July 2008) with no discernible trend in bromate (or bromide) concen-
trations.
4.4.4 Hatfield area

In the Hatfield area, the spatial distribution of bromate is defined by locations 001, 002, 003, 191, 265,
160, 378, 379 and 402. Times series are given in Figure 4.26 to 4.30.
Location 001 (Hatfield P.S.) has a very good time series from 2000 to 2008. The trend had been
described in Section 3.5. Location 002 (Hatfield Business Park BH) shows a slight rising trend. There is a
gap between late 2002 and early 2006 when no samples were taken, and after this bromate concentrations
remain relatively stable at around 300 to 500 µg l−1 over 2006, 2007 and 2008. The slight rising
trend over 2000 and 2001 coincides with a decline in water levels. Also, Hatfield switch-off in May
2000 may have some influence on concentration trends, particularly as concentrations appear to level
off after the start of the Hatfield scavenge pumping trial. The times series for location 003 (Hatfield
Bus Garage BH) shows bromate concentrations around 600 µg l−1 from June 2000 until May 2001,
and between about 400 and 800 µg l−1 from May 2002 until July 2002. These elevated concentrations
are separated by an extended period of low bromate (and bromide) concentrations between June and
December 2001. The observations have been the subject of much discussion, and the Environment
Agency concluded that leakage from shallow depth (via the borehole annulus) was impacting on borehole
water quality (Atkins, 2006). Location 402 (Comet Way BH5) shows a rising trend of bromate (and
bromide) concentrations from January 2006 and December 2008, with concentrations increasing from
∼ 400 µg l−1 to ∼700 µg l−1 . Location 160 shows a declining trend in bromate concentrations, from
∼ 800 µg l−1 in September 2001 to ∼ 500 µg l−1 in July and August 2002, although concentrations
then increase to ∼700 µg l−1 in September and October 2002. No samples were taken again until May
2005, when concentrations were recorded below MDL until September 2005.
To the north, at location 378 bromate concentrations show a declining trend from ∼ 140 µg l−1 at
the end of 2005 to ∼90 µg l−1 at the end of 2008. At location 379 bromate concentrations are generally
around 40 to 60 µg l−1 , although a slight declining trend appears to occur over 2006, 2007 and 2008.
Bromide concentrations mirror bromate concentration trends for both of these boreholes. The northern
margin is defined by locations 016/162 (Old Cottage, Green Lane, Old BH/New BH), 375 and 376.
Location 016/162 shows only a few isolated results above the MDL. Location 375 yields concentrations
around 1 µg l−1 , with a peak of 3 to 4 µg l−1 in Autumn 2006.
The southern margin is defined by location 235 (Carter’s Pond BH), location 142 and location 195.
Locations 142 and 195 are almost always below the MDL. Time series for location 235 (Carter’s Pond)
shows bromate below the MDL, with the exception of intermittent concentrations of 1 to 2 µg l−1 in late
2001 and early 2002. Further south, locations 049 (Brand’s Nursery BH), location 006, 014 and 015 are
typically below MDL, but show intermittent concentrations above the MDL.
Slight rising trends in bromate concentrations are observed at locations 191 (Mill Green
BH) and 265 (Park Street BH) to the east towards the Lea Valley. Bromate concentrations rise from
∼2 to 4 µg l−1 in 2001 and 2002 to ∼8 to 10 µg l−1 in 2004 to 2008 (with lower concentrations of
∼ 2 to 4 µg l−1 in late 2006 and early 2008) for 191. However, there are a number of results below MDL
at 191. The time series for 265 is very variable, with concentrations fluctuating considerably. However,
there appear to be two parallel trend lines: one from ∼2 to 4 µg l−1 in 2002 to ∼18 to 22 µg l−1 in 2008,
and another from ∼40 to 45 µg l−1 in 2002 to ∼50 to 60 µg l−1 in 2007, and a fall in concentrations to
∼45 µg l−1 in 2008.
4.4.5 Lea Valley (east of Hatfield area)

The density of the monitoring locations decreases east of Hatfield. The lower density of monitoring
boreholes is at least partly due to the depths to groundwater being much greater where the chalk is
overlain by London Clay south-east of the Tertiary Escarpment. In general, locations are clustered along
the River Lea and New River. Consequently, it is difficult to define the full spatial extent of the bromate
contamination. The bromate contamination east of Hatfield is therefore interpreted as extending in an
arc following the main karst conduits (Section 3.7) along the River Lee and New River.
The bromate contamination continues from the Hatfield area along the course of the River Lea to
location 143 (Essendon P.S.). Arkley Hole Spring (location 287), located ∼1.1 km east of Essendon,
shows concentrations comparable to Essendon (143) (Figure 4.31).
Monthly Rainfall (mm) Hatfield Business Park (Ref. 002) 200

Rothamsted Station
160
6600 Lee Chalk

160
120 120
SMD
80
80
40
40
0
0
85
Water Level (mAOD)

80
75
Orcharge Garage (Ref. 028)
Business Park (Ref. 002) 70
65
60
Hatfield swith off
2000 Pump Test Start
Mean Monthly Br
Raw Br
1600 Mean Monthly BrO3
Raw BrO3
1200
800
400

1
0.8
0.6
0.4
0.2
area, soil moisture deficit, and monthly rainfall.
Monthly Rainfall (mm) Hatfield Bus Garage (Ref. 003) 200

Rothamsted Station
160
6600 Lee Chalk

160
120 120
SMD
80
80
40
40
0
0
84
Water Level (mAOD)

Jan-00 Jan-02 Jan-04 Jan-06 Jan-08 Jan-10 80
76
Bus Garage (Ref. 002)
68
64
60
Hatfield swith off
Mean Monthly Br
Raw Br
2000 Mean Monthly BrO3
Raw BrO3
1500
1000
500

1
0.8
0.6
0.4
0.2
200
Monthly Rainfall (mm) Comet Way BH5 (Ref. 402)
Rothamsted Station
6600 Lee Chalk

160 160
120 120
SMD
80
80
40
40
0
0
85
Water Level (mAOD)

80
75
Comet Way BH5 (Ref. 402) need to check datum level
65
60
55
Hatfield swith off
Mean Monthly Br
Raw Br
Mean Monthly BrO3
1200
Raw BrO3
800
400

1
0.8
0.6
0.4
0.2
M7 Mill Green (Ref. 191) 200

Rothamsted Station
160
6600 Lee Chalk

160
120 120
SMD
80
80
40
40
0
0
85
80
Water Level (mAOD)

75
70
M7 Mill Green (Ref. 191)
65
60
200 55
160 Hatfield swith off

Pump Test Start
Mean Monthly Br
120 Raw Br
Mean Monthly BrO3
Raw BrO3
80
40

1
0.8
0.6
0.4
0.2
200
Monthly Rainfall (mm) Park Street (Ref. 265)
Rothamsted Station
6600 Lee Chalk

160 160
120 120
SMD
80
80
40
40
0
0
90
Water Level (mAOD)

80
70
Park Street (Ref. 265) 60
50
Hatfield swith off 40

Pump Test Start
200 Mean Monthly Br
Raw Br
Mean Monthly BrO3
160 Raw BrO3
120
80
40

1
0.8
0.6
0.4
0.2
Between the Hatfield area and Essendon, at location 266 (Hill End Farm BH), there have been
two occurrences of bromate concentrations around 75 µg l−1 , but all other concentrations have been
below the MDL. These two high results lead to mean annual concentrations between 30 and 40 µg l−1
in 2003 and 2004. These results may be anomalous. In support of this, locations 262, 263 and 364,
in close proximity to 266, all show concentrations below MDL. This is despite these locations be-
ing directly between bromate in the Hatfield area and Location 143 (Essendon P.S.) where bromate
concentrations have consistently been above 10 µg l−1 . Interestingly, bromate is also not observed
at 144 (Water Hall P.S.), on the southern side of the River Lea, but is recorded at certain locations
to the north-east on the northern side of the River Lea at Southfield Wood Landfill boreholes (loca-
tions 329 to 331, 364 to 367 and 369 to 370). Bromate concentrations around 10 to 20 µg l−1 are
recorded at selected locations, generally in the central to north-eastern parts of this area. Additionally,
location 404 shows annual average bromate concentrations around 2 to 5 µg l−1 , although approximately
half of the samples are below MDL.
Bromate concentrations appear to have increased at location 089 (Holly Cottage BH), situated to the
north-east of Southfield Wood Landfill, but on the southern side of the River Lea. Bromate concentrations
generally rise from below MDL in 2000, to between 5 and 7 µg l−1 between October 2007 and July 2008.
This increase is accompanied by an increase in bromide concentrations. A bromate sample in October
2008 shows bromate concentrations return to 2 µg l−1 ; bromide concentrations remained on a rising
trend. Continuing on east along the Lea, locations 305 to 307 are generally below MDL, and always less
than 1 µg l−1 .
Location 005 (Hatfield London Country Club BH) appears to mark the southern margin; bromate
has been recorded on an intermittent basis (in approx two-thirds of the samples taken) throughout the
monitoring period, up to a maximum of 13.0 µg l−1 . There are a number of locations between 005 and
the River Lee that show concentrations below the MDL (Locations 405 to 409). These were monitored
once in 2007. Location 005 is at the end of an inferred conduit route, and locations 405 to 409 are to the
east of this.
The locations along the Northern New River define the easterly extent of the bromate contamination.
Chadwell Spring shows bromate concentrations intermittently above MDL. Chadwell Spring feeds into
the New River. In general, the most northerly of the Northern New River wells, locations 295 (Amwell
End P.S.) and 298 (Broadmeads P.S.), show bromate concentrations below the MDL, although sporadic
occurrences of bromate above the MDL have occurred, particularly since 2004. Annual average concen-
trations tend to be highest in the central and southern parts of the wellfield: Hoddesdon, Broxbourne,
Turnford and Amwell Marsh.
There are relatively few additional sampling locations to the west of the the River Lee - New River
Loop. Location 354 (Van Hage Nurseries BH) shows bromate concentrations below MDL. In 2003,
locations 304 and 309, and locations 293 and 394 and locations 312 to 316 were sampled once, and
showed bromate concentrations below MDL.
Monthly Rainfall (mm) 200

Arkley Hole Spring (Ref. 287)
Rothamsted Station 160
6600 Lee Chalk

160
120 120
SMD
80
80
40
40
0
0
Hatfield swith off

Pump Test Start
Mean Monthly Br
200 Raw Br
Mean Monthly BrO3
Raw BrO3
160
120
80
40

1
0.8
0.6
0.4
0.2
Figure 4.31: Time series of bromate and bromide concentrations at selected locations in the Lea Valley,
soil moisture deficit, and monthly rainfall.
4.5. Bromate-bromide ratios 158
4.4.6 Bromide spatial distribution in groundwaters

The distribution of elevated bromide concentrations in groundwater (Figure 4.32) is broadly similar to
the distribution of bromate contamination. It is more difficult to identify evidence of spatial variations
or to discern evidence of increasing time series trends above background levels owing to the natural
occurrence of bromide in groundwaters (typically in the range 50 to 150 µg l−1 ) (Section 1.7).
In order to assess the background concentration of bromide in the catchment using the data in the
UCL monitoring database, all samples from locations where bromate concentrations were consistently
below MDL were selected (Figure 4.33). There are a number of locations where bromide is elevated
above background concentrations despite these locations showing bromate concentrations below the
MDL: location 214 at the source site, locations 070 to 076 around the central part of Hatfield Quarry,
location 227 at the southern ‘plume’ margin between Sandridge and Hatfield Quarry. The background
concentration of bromide was estimated using the available samples from location up-gradient of SLC
and well beyond the bromate and bromide contamination extent (Figure 4.33). The mean bromide con-
centration was 65.8 µg l−1 .
The spatial distribution of elevated bromide concentrations is indicated in Figure 4.32. In general,
the distribution of bromide mirrors that of bromate. However, Atkins (2005) suggest that the bromide
‘plume’ appears to extend slightly further south than the bromate ‘plume’ with slightly higher bromide to
bromate ratio values in the southern section of Hatfield Quarry and in the Jersey Farm area. Interestingly
at the source site, highest bromate concentrations are in the north and bromide concentrations in the
south; it is possible that this variation in distributions is reflected at down-gradient locations.
Edmunds (1996) found that the bromide occurrence is best described in terms of the Br – /Cl – ratio.
Unfortunately all samples that were tested for bromide concentration do not have corresponding chloride
concentrations: approximately 22 % of samples have results for both chloride and bromide. Analysis
of the regression of bromide concentrations against bromate concentration and chloride concentration
indicate that more of the variation in bromide concentrations can be explained by a combination of
bromate and chloride concentration than either bromate or chloride independently (Table 4.3).
4.5 Bromate-bromide ratios

Bromate/bromide ratios vary from 0.00 to 1.60 (Figure 4.34). Ratios above 0.80 are associated with
Chalk groundwaters with bromate concentrations greater than 4000 µg l−1 . For samples with bromate
concentrations between 200 µg l−1 and 2000 µg l−1 , the bromate/bromide ratio is relatively consistent at
around 0.30 to 0.60 for both surface waters and groundwaters. For samples with bromate concentrations
between 40 µg l−1 and 200 µg l−1 , the bromate/bromide ratio is relatively consistent at around 0.20
to 0.40 for both surface waters and groundwaters. The ratio appears to increase approximately linearly
from 0.0 to around 0.20 with increasing bromate concentration for samples with bromate concentrations
below 40 µg l−1 . There does not appear to be a discernible difference in ratios between surface water
and groundwater ratio trends, or between chalk groundwater, gravel groundwater and groundwater from
public water supply (PWS).
±
0 20 40 60 80 100 Meters
Yearly Mean Bromide

Concentration (µg/l)
Groundwater Bromide 0 0.5 1
Average 2000-2008 Kilometers
<MDL
2 - 10
11 - 20
21 - 40
41 - 60
61 - 80
81 - 100
101 - 200
201 - 400
4.5. Bromate-bromide ratios
401 - 600
601 - 2000
2001 - 1000
1001 - 2000
2001 - 4000
4001 - 20000
20001 - 100000
100001 - 500000
0 1 2 3 4 5
Kilometers
159
Figure 4.32: Annual average Bromide concentrations in groundwater 2000 to 2008.

F
G
Bromide Concentration (µg/l) ±

(where Bromate conc. < MDL)
Groundwater
Bromide 00-08 F
G
0 20 40 60 80 100 Meters
25 - 50
51 - 100
101 - 150
151 - 200 F
F
G GG
F
201 + F
G
Surface water F
G
0 0.5 1GF
Bromide 00-08 F
F
G
Kilometers G
F 25 - 50
G
F 51 - 100
G
F 101 - 150
G
G 151 - 200
F
F 201+
G F
G
Sample used
to calculate
Bromide background
concentration F
G
F
G
F
G G F
F G
F
G
G
F F
GG
F F
G
F
G G
F G
F
4.5. Bromate-bromide ratios
F
G
F
G F
G
F
G
F
G
F
G F
G
F
G F
G
F
G
F
G
F
G
0 1 2 3 4 5
Kilometers
160
Figure 4.33: Bromide concentrations at locations where bromate concentrations are less than MDL.
4.5. Bromate-bromide ratios 161
2.0
Surface water
Groundwater - PWS
1.6 Groundwater - Chalk
Groundwater - Gravel
Bromate/bromide ratio
1.2
0.8
0.4
0.0
0 40000 80000 120000 160000 200000

1.2 Bromate concentration (µg l-1)
0.8
0.4
0 4000 8000 12000 16000 20000

Bromate concentration (µg l-1)
0.8
0.6
0.4
0.2
0 400 800 1200 1600 2000


0.6
0.4
0.2
0 40 80 120 160 200

Figure 4.34: Bromate/Bromide ratio variation with bromate concentration for groundwater and surface
water samples
4.6. Bromate concentrations and water levels 162
Table 4.3: Regression statistics for the response of bromide concentration to bromate concentration and
chloride concentration.
For equation Bromide = A + B(Bromate) + C(Chloride)

2 a
N = 557 R P-value
2
R overall 37.2% 0.000
2
R Bromate 31.5% 0.000
2
R Chloride 5.7% 0.000
a
The P-value refers to the hypothesis that the regression
relationship is statistically significant, i.e. that the apparent
relationship between y and x is not likely to arise due to
chance alone.
The spatial distribution of bromate-bromide ratio is shown in Figure 4.35. In general, the higher
ratios follow the distribution of bromate contamination. If locations with bromate concentrations less
than 1 µg l−1 ) are excluded from the plot, there is a noticeable cluster of locations with high ratios in the
northern part of the St Leonard’s Court source site, and following the path of the core bromate ‘plume’
to Hatfield Quarry and the Hatfield area.
4.6 Bromate concentrations and water levels

Figure 4.36 shows the percentage of samples of bromate concentrations for which there are accompany-
ing water level measurements for each location.
Fluctuations in bromate concentrations are dependent on water level at a number of locations. The
regression relationship for the response of bromate concentration to water level (m OD) was analysed
for all locations for which there were sufficient time series data with bromate concentrations above MDL
and associated water level measurements. Both positive and negative correlations were observed to be
statistically significant (p<0.05), and water level explains between 10 % and 90 % of the variation in
bromate concentration (Figure 4.37). Full statistical results are given in Appendix C. There does not
appear to be a pattern in the spatial distribution of negative and positive correlations, or any obvious
spatial pattern in those locations that show a statistically significant relationship and those that show no
relationship.
4.7 Summary and Conclusions

Detailed analysis of bromate and bromide monitoring data has been undertaken, which has revealed that
bromate concentrations are affected by influences including recharge (soil moisture deficit, rainfall), wa-
ter level, and abstractions. The distribution of these relationships supports the conceptual understanding
of an increasing influence of a karst system to the east of the Vale of St Albans area. The seasonal
influences are superimposed on a generally stable distribution of bromate and bromide concentrations,
which is likely to indicate the importance of the attenuating effect of the double-porosity Chalk. The
±
0 20 40 60 80 100 Meters
Bromate - Bromide ratio

Ratio Surface water
overall
0.00 - 0.10
0 0.5 1
0.11 - 0.30
Kilometers
0.31 - 0.50
0.51 - 0.80
0.81 - 1.60
Ratio PWS data

Overall
0.00 - 0.10
0.11 - 0.30
0.31 - 0.50
0.51 - 0.80
0.81 - 1.60
Ratio Groundwater - Gravel

Overall
4.7. Summary and Conclusions
0.00 - 0.10
0.11 - 0.30
0.31 - 0.50
0.51 - 0.80
0.81 - 1.60
Ratio Groundwater - Chalk

Overall
0.00 - 0.10
0.11 - 0.30
0.31 - 0.50
0.51 - 0.80 0 1 2 3 4 5
0.81 - 1.60 Kilometers
163
Figure 4.35: Spatial distribution of mean annual bromate/bromide ratio 2000 to 2008.
164
Figure 4.36: Percentage of samples of bromate concentrations for which there are accompanying water level measurements for each location
!
(
(
!
^
_ (
!
_^_
^ (
!
_
^ (
!
!
( !
(
^ _
^
_^_ _
^ !
(
(
! _
^
(
!
(
!
_
^
(
!
±
_
^ (
! (
!
_
^
(
!
0 20 40 60 80 100 Meters
(
!
_
^
_
^
(
!
Regression:
Bromate conc. versus
water level (m OD) 0 0.5 1 _
^
Negative Regression Kilometers
P-value < 0.05
_
^ 0% - 25%
_
^ 26% - 50%
_
^ 51% - 75%
_
^ 76% - 100%
Negative Regression
P-value > 0.05
(
!!
( (
!
(
! 0% - 25% _
^^_ _^ ^
_
^ (
! (
! _!
_!(^ (
!
( 26% - 50% (
! _
^ _
^
_
^ !
( !
(
(
_
^
!!
(
( 51% - 75%
! _
^ _
^ (
!
(
! (
!
_
^ _
^
(
! 76% - 100% _
^ (
!
Positive Regression
P-value < 0.05
_
^ 0% - 25%
_
^ 26% - 50%
_
^ 51% - 75%
_
^ 76% - 100%
Positive Regression
P-value > 0.05
(
! 0% - 25%
!
( 26% - 50%
( 51% - 75%
!
(
! 76% - 100%
0 1 2 3 4 5
Kilometers
165
Figure 4.37: Regression relationship for the response of bromate concentration to water level. Percentages refer to the amount of variation explained by the regression (R2
value)
4.7. Summary and Conclusions 166
revised conceptual understanding of flow and transport in the Hertfordshire Chalk has allowed a new
interpretation of the spatial distribution and evolution of bromate and bromide within the catchment to
be developed between 2000 and 2008. However, the interpretation of the spatial and temporal evolution
of bromate and bromide within the catchment is hampered by a number of inadequacies in the available
monitoring data:
• Monitoring data are available for a relatively short period of time (a maximum of 8 years continu-
ous data) in relation to the likely timescale of bromate contamination within the catchment;
• Monitoring frequency varies considerably between locations, and varies over time at individual
locations, which makes trends difficult to identify with confidence;
• The strong seasonal influences within the time series make trends difficult to discern;
• The data available are generally not depth-specific so that vertical distribution of bromate contam-
ination cannot be investigated;
• The sampling results refer to (mobile) fissure water and there are no data available for (immobile)
matrix porewater which is required to characterise the double-porosity behaviour and determine
the long-term evolution of bromate contamination within the catchment.
167
Chapter 5
The Bromate Source
5.1 Chapter objectives

The source of the bromate contamination has been identified as the site of a former chemical works
located in Sandridge, Hertfordshire. The site is now the St. Leonard’s Court residential development
(Figure 5.1). Limited site investigation and groundwater monitoring data are available for the source site
and the vicinity. The magnitude and dynamics of bromate release to groundwater beneath the source site,
and thus to locations down-gradient of the source site, has consequences for the magnitude and duration
of bromate contamination within the catchment. Predictive models of bromate contamination within the
catchment are dependent on a realistic and representative source term to quantify the input of bromate
from the source site. Uncertainty in the history of the source may be incorporated by development of a
range of source term scenarios which are constrained with the limited investigation data at the site.
The objectives of this chapter are:
5.2 Chapter structure

The chapter begins by describing the history of the site, then the distribution of bromate and bromide at
the source site. A number of conceptual scenarios for bromate release are developed, and the data are
used as a basis to constrain the source terms.
5.3 Site History

5.3.1 Sources of information
Information on the site history has been obtained from information and records held by the Environment
Agency comprising:
5.3. Site History 168
Figure 5.1: Location of the source site in Sandridge, Hertfordshire. Formerly the Steetly chemical works,
now the St Leonard’s Court residential development.
5.4. Chronology and scope of investigations 169
• Original records from St Albans District Council (the local authority); and
• Additional information collected by the Environment Agency in response to a public request for
information conducted in 2001.
An interview with a former worker at the chemical works, conducted by the Environment Agency in
August 2001, provided details on the operational activities (EA, 2005). Aerial photograph taken in 1971.
5.3.2 General overview

The Steetly Chemical Works occupied the site from 1955, and is thought to have remained operational
until around 1980. The buildings were demolished and the site cleared for redevelopment between 1983
and 1986. During this period the site was left uncovered, and free draining. The residential development,
St Leonard’s Court, has been present at the site since the end of 1987.
5.3.3 Site investigation and remediation history

A series of site investigations and assessments were undertaken during demolition and redevelopment
between 1983 and 1986 (Section 5.4). Some groundwater monitoring data are available from existing
boreholes in the vicinity of the site between 1983 and 1987.
Based on recommendations of the site investigations and assessments, the top layer of soil was
excavated and removed from selected areas where high levels of contamination had been identified. It is
understood this occurred between August 1986 and September 1986 (Roberts, 2001).
Subsequent to discovery of the bromate contamination at Hatfield, site investigations were under-
taken in 2000 and 2001 (Section 5.4), alongside sampling from boreholes in the vicinity of the site as
part of the bromate groundwater monitoring programme (Section 4.2).
5.3.4 Operational activities of the chemical works

The Steetly Chemical Works was a chemical manufacturing plant which specialised in the manufacture
of industrial and pharmaceutical intermediates including potassium bromate and organobromine com-
pounds. Raw materials including bromine, red and yellow phosphorus, and caustic soda were processed
into products including ceta-stearyl bromide, sodium and potassium bromate, and zinc bromide.
The former locations of the process areas, based on historical plans, aerial photographs, and the
interview are indictated in Figure 5.2. Bromine was stored as a raw material in two main areas of the
site: small glass bottles were stored in an open area in the centre of the site, and bulk storage was in the
north of the site, to the rear of the main building. Liquid bromate production and solid bromate handling
took place in a process room in the northern corner of the site. This building included a sump, which was
reported to be for collection of condensate from heating coils of reaction vessels, and spills of materials.
The sump is reported to have discharged to foul sewer.
5.4 Chronology and scope of investigations

The scope of the site investigations that have been undertaken at the source site is summarised in Ta-
ble 5.1 and Figures 5.3 and 5.4.
Figure 5.2: Location of former process areas of the Steetly Chemical Works. Based on Atkins (2002)
interpretation of historical plans, aerial photographs, and the interview with a former employee of the
works. Aerial photograph taken in 1971.
Table 5.1: Chronology and scope of site investigations and monitoring at the source site
Date of Date of
Company Scope of investigation Associated Reports Client
site work report
5 boreholes. Soil samples at depths of 0.5, 1.0 & 1.5m.

Aug 1983 STATS Interim report on site investigation, ref 83/3105. Aug 1983 Crest
Analysis for bromide and bromate.
Further tests on samples taken in August 1983. Analysis

None STATS Second report on site investigation, ref 83/3105A Sep 1983 Crest
for total bromine and water-extractable bromide.
Soil samples taken on grid pattern at depths of 0.75 and

Oct 1983 STATS Third report on site investigation, ref 83/3105C Dec 1983 Crest
1.5m. Tested for Bromide only
3 boreholes. 6-8m deep into putty chalk. Soil and

Mar 1984 STATS groundwater samples. Analysis for bromide only. Site was Report on further site investigations. ref AM/3554 May 1984 Crest
in process of being cleared by demolition contractors.
Borehole C1 drilled to 5.4m deep. Core samples tested for Field report for drilling of borehole C1 at House
Jan 1985 Chemfix Mar 1985 Crest
moisture content and bromide. Lane, Sandridge on 21/22 January 1985
Southern 51 shallow boreholes, depth generally approx 1.5m. Hand-augered borehole logs at Sandridge, Herts
Mar 1985 Mar 1985 Chemfix
Testing Analysis for bromide. for Chemfix International Ltd
Report of the second phase of the field

- Chemfix Report on Southern Testing trial holes from Mar 1985. Mar 1985 Crest
investigation at the Sandridge site
5.4. Chronology and scope of investigations
1 Borehole drilled 40m down-gradient of nearest part of

Evaluation of the results from the borehole
May 1985 Chemfix site. Soil and groundwater samples. Analysis for Jun 1985 Crest
situated 120m down dip from the Sandridge site
bromide.
St Albans
5 boreholes to depths of approx 12m. Soil and Site Investigation at St Leonards Court, City and
Aug 2000 Komex Oct 2000
groundwater samples. Analysis for bromide and bromate. Sandridge, St Albans. District
Council
St. Leonard's Court, Sandridge, St. Albans.

12 boreholes to depths between 6m and 20m. Soil and Environment
Nov 2001 Atkins Environmental Site Investigation and Quantitative Dec 2002
groundwater samples. Analysis for bromide and bromate. Agency
Pollutant Linkage Assessment.
171
Figure 5.3: Borehole locations from investigations 1983-1985 (STATS, 1983a,b,c, 1984; Chemfix,
1985c) and 2000-2001 (Komex, 2000; Atkins, 2002). For locations from 1983-1985, numbers in square
brackets indicate date of drilling.
Figure 5.4: Trial hole locations from investigations in 1985 (Chemfix, 1985c)
5.5. Site Geology and Hydrogeology 174
5.5 Site Geology and Hydrogeology

The geology and hydrogeology of the site have been described in detail by Atkins (2002). The geology
is summarised in Table 5.2.
Table 5.2: Geological strata encountered at the source site. Based on Komex (2000) and Atkins (2002)
Stratum Thickness Typical Description
Yellow to red brown clayey sandy gravely fill. In places

containing brick fragments, asbestos roofing
fragments, granite chippings, builders’ rubble, breeze
Average 1.4m blocks, ash, some putty chalk and flint.
Made Ground
(absent in places)
Mixed silt/sand/clay with flint gravel with brick
fragments. Occasional fly ash, clinker and metal
fragments.
Lenses of orange to brown, medium dense silty clay to

Fluvio-glacial sandy gravel.
4.25 to 1.35m
sand and gravel
Average 2.79
deposits Orange silt and sand with flint gravel. Some silty clay
and clay interbeds.
‘Putty’ Chalk
variable Structureless, weathered Chalk with occasional flints.
(Upper Chalk)
‘Blocky’ Chalk
>10 m Chalk with horizontal and vertical fractures.
(Upper Chalk)
Groundwater levels varied between +78.78 m OD and 81.15 m OD during the 2001 investigation
(Atkins, 2002). Borehole logs show that groundwater appears to be semi-confined by the low perme-
ability ’putty chalk’: there is generally a rise in groundwater rest levels compared to strike levels. Piezo-
metric contours (Figure 5.5) indicate a south-easterly flow direction, with a hydraulic gradient across
site of 0.0042 (Atkins, 2002). The local hydraulic gradient is lower: 0.0028 in south-easterly direction
(Atkins, 2002). Vertical groundwater elevation contours on section along flow direction indicate flow is
principally horizontal (Figure 5.6).
5.6 Contaminant Distribution

5.6.1 Spatial distribution of bromate and bromide within soil and soil porewater
The bromide and bromate soil concentration results are reported as mg kg−1 on a dry weight basis. As-
suming bromide and bromate to be completely soluble in the soil moisture (porewater) at the observed
concentrations, equivalent porewater concentrations for the unsaturated zone can be estimated by con-
verting the results from the soil analyses in mg kg−1 (dry weight) to a concentration in µg l−1 using the
soil moisture content1 .
Porewater analysis from the saturated zone was undertaken as part of the Atkins (2002) investiga-
tion. This is discussed in Section 5.6.2.
1 Soil moisture contents ranged from approximately 10 % to 15 % in the unsaturated zone (Atkins, 2002)
5.6. Contaminant Distribution
175
Figure 5.5: Piezometry at the St Leonard’s Court site November 2001. From Atkins (2002)
176
Figure 5.6: Cross-section parallel to groundwater flow direction. From Atkins (2002)
5.6. Contaminant Distribution 177
5.6.1.1 Pre-redevelopment as SLC: 1983 – 1985

The site investigations undertaken between 1983 and 1985 indicate that considerable bromide contami-
nation was present in soils beneath the site (Figure 5.7). Soil samples were not tested for bromate after the
initial investigation (STATS, 1983a) showed all samples to be below the detection limit of 20 mg kg−1 .
The exact locations of the boreholes from the STATS 1983 investigation is not known (the location plan
is missing from the available report). However, the approximate locations have be estimated based on
description within the text of the report and are shown in Figure 5.3 and Figure 5.7. Based on the oc-
currence of bromate contamination in the recent (2000 and 2001) investigations (Section 5.6.1.2), it is
surprising that bromate was not detected in some of these locations, particularly around BH-3[83] &
BH-4[83], which were reported to be located in the vicinity of the former bromate production area. The
samples tested were from relatively shallow depths (0.5 m to 1.5 m); it is possible that bromate was not
present in the shallower soils, but would have been encountered in the deeper strata.
BROMIDE distribution 1983 - 1985
Fluvio-glacial
deposits
Chalk
Figure 5.7: Spatial distribution of the bromide contamination based on investigations undertaken be-
tween 1983-1985.
The highest bromide concentrations (>1000 mg kg−1 ) occur in soil samples from boreholes in
the vicinity of the former ‘solid bromate handling’ and ‘bulk bromine storage’ areas, and close to the
sump in the ‘non-bromate production’ area (Figure 5.7). Locations in the southern and eastern areas of
the site, corresponding to the non-process areas of the site showed much lower bromide concentrations
(<200 mg kg−1 ).
Concentration-depth profiles for these boreholes (Figure 5.8) indicate highest bromide concentra-
tions in the Made Ground and in the Putty Chalk, with generally lower concentrations in the fluvio-glacial
deposits (clayey gravels).
The results of this sampling was used as a basis for the excavation of between 0.75 m and 1.5 m
of the top layer of soil over much of the site as part of remediation carried out between 1985 and 1986
(Roberts, 2001). It is unclear whether any verification samples were submitted, or whether there were
significant alterations to these proposals.
5.6.1.2 Post-redevelopment as SLC: 2000 – 2001

Figure 5.9 and Figure 5.10 illustrate the spatial distribution of the bromide and bromate contamination
based on the 2000 and 2001 site investigations.
The pattern of bromide contamination within soil is generally in agreement with the 1983-1987 dis-
tribution, although concentrations are considerably lower in 2000 and 2001. The site investigations un-
dertaken during 2000 and 2001 indicated generally low bromide concentrations (<0.01 to 8.0 mg kg−1 )
within the Made Ground and shallow soils (<1.5 m depth). This is presumably as a result of removal
of the contaminated top layer of soil during redevelopment. The highest bromide concentrations (100
to 300 mg kg−1 ) were encountered within the Chalk and fluvio-glacial deposits in soil samples from
boreholes in the vicinity of the former ‘solid bromate handling’ and ‘bulk bromine storage’ areas, and
close to the sump in the ‘non-bromate production’ area. Bromide concentrations between 10 mg kg−1
and 60 mg kg−1 were encountered in the fluvio-glacial deposits and Chalk in locations down-gradient
of the sump in the production areas. Many of the boreholes in the southern and eastern part of the site,
corresponding to the non-process areas, showed low concentrations of bromide (<10 mg kg−1 ) within
the fluvio-glacial deposits and (putty) Chalk.
It is difficult to discern a pattern from the bromide concentration-depth profiles (Figure 5.11 to Fig-
ure 5.21) as fewer depths were tested at each location than in the 1984 and 1985 investigations. However,
profiles from the Komex (2000) investigation indicate that the highest concentrations were found in the
top section of the putty chalk with lower concentrations in the fluvio-glacial deposits (including grav-
els, silty sands and silty clays). In the Atkins (2002) investigation, boreholes 219 and 222 showed this
pattern, boreholes 217 and 223 show higher concentrations within the fluvio-glacial deposits (gravelly
CLAY) than the Chalk, and borehole 218 shows relatively consistent concentrations.
The pattern of bromate contamination was found to be relatively similar to the pattern of bromide
contamination. The site investigations undertaken during 2000 (Komex, 2000) and 2001 (Atkins, 2002)
indicated generally low bromate concentrations (<0.010 mg kg−1 to 0.090 mg kg−1 ) within the Made
Ground and shallow soils (<1.5 m to <2.0 m depth). This is presumably as a result of removal of the
contaminated top layer of soil during redevelopment. Elevated bromate concentrations (195 mg kg−1
to 273 mg kg−1 ) were encountered within the Chalk (putty chalk) and bromate concentrations of 35 to
62 mg kg−1 in lower fluvio-glacial deposits (clay) from boreholes in the vicinity of the former bromate
production area (down-gradient of the sump) and bromate handling area, and bulk bromine storage.
The density of sampling locations around the former bromate production area is relatively sparse. It is
Boreholes 1983-1985
C1-[85] BH1-[84]
84 84 Porewater concentrations 84 Made Ground
Clay
83 83 83 Sand
Chalk
82 82 82 Monitoring well
screened interval
81 81 81
Calculated Porewater
80 C1-[85]
80 80 Porewater Bromide Conc.
Jan-85
BH1-[84]
79 79 79 Porewater Bromide Conc.
Mar-84
BH3-[84]
Porewater Bromide Conc.
78 78 78 Mar-84
BH2-[84]
Sample elevation (mAOD)

Mar-84
77 77 77
Pumped groundwater
76 76 76 BH2-[84]
BH1-[84]
75 BH3-[84]
75 75
0 10000 20000 30000
bromide concentration (mg/l)
BH2-[84] BH3-[84]
84 84 84 Pumped groundwater concentrations
10000 25
83 83 83
82 82 82
8000 20
81 81 81
80 80 80 6000 15
79 79 79
4000 10
78 78 78
Water Level (mAOD)

Bromide concentration (mg/l)
77 77 77 2000 5
76 76 76
75 0 0
75 75
0 4000 8000 12000 16000 Jan-84 Apr-84 Jul-84 Oct-84 Jan-85
bromide concentration (mg/l)
179
Figure 5.8: Depth profiles of porewater bromide compared to pumped groundwater concentrations for boreholes from investigations 1983-1985.
BROMIDE distribution 2000 - 2001
Fluvio-glacial
deposits
Chalk
Figure 5.9: spatial distribution of bromide (as mg kg−1 ) based on investigations undertaken between
2000 and 2001.
BROMATE distribution 2000 - 2001
Fluvio-glacial
deposits
Chalk
Figure 5.10: spatial distribution of bromate (as mg kg−1 ) based on investigations undertaken between
2000 and 2001.
Borehole 214 Porewater concentrations
214
84 Bromate Bromide
84 84
83
83 83
82 82 82
81 81 81
80 80 80
Made Ground
Clay
79 79 79
Sand

Chalk
78 78 78
Monitoring well
screened interval
77 77 77
Pumped Groundwater Conc.
Nov-01 0 5 10 15 20 25 0 5 10 15 20 25
Porewater Bromate Conc.
Nov-01 bromate concentration (mg/l) bromide concentration (mg/l)
Soil Bromate Conc.
Nov-01

Pumped groundwater concentrations Bromide
Nov-01
Bromate
Soil Bromide Conc. 0.0024 82 20 82
Nov-01
Groundwater Bromate
Groundwater Bromide
Water Level
80 80
0.0020 16
78 78
0.0016 12
76 76
0.0012 8
Water Level (mAOD)
Water Level (mAOD)
74 74
Bromate concentration (mg/l)

0.0008 4
72 72
0.0004 0
May-01 May-02 May-03 May-04 May-05 May-01 May-02 May-03 May-04 May-05
182
Figure 5.11: Depth profiles of porewater bromate and bromide compared to pumped groundwater concentrations for Borehole 214 from 2001 investigation (Atkins, 2002).
Borehole 215
Porewater concentrations
215
84
Bromate Bromide
85 85
83
84 84
82 83 83
82 82
81
81 81
80
Made Ground
80 80
Clay
Sand
79

79 79
Chalk
Monitoring well 78 78 78
screened interval
77 77 77
Nov-01 0 5 10 15 20 25 0 20 40 60 80
Soil Bromate Conc.
Nov-01

Nov-01
Soil Bromide Conc.
Pumped groundwater concentrations
Nov-01 Bromate Bromide
1 82 250 82
Groundwater Bromate
Groundwater Bromide
Water Level
80 200 80
1
78 78
1 150
76 76
0 100
Water Level (mAOD)
Water Level (mAOD)
74 74

0 50
72 72
0 0
183
Borehole 216
216
84 Bromate Bromide
84 84
83
83 83
82
82 82
81 81 81
80 80 80
79 79 79
78 78 78
77 77 77
Made Ground 76 76 76
Clay
75 75 75
Sand

74 74 74
Chalk
73 73 73
Monitoring well
screened interval
72 72 72
71 71 71
Nov-01 0 1 2 3 4 0 20 40 60 80
Soil Bromate Conc.
Nov-01

Nov-01
Soil Bromide Conc.
5 82 4 82
Groundwater Bromate
Groundwater Bromide
Water Level
80 80
4
3
78 78
3
2
76 76
2
Water Level (mAOD)
Water Level (mAOD)
74 1 74

1
72 72
0 0
May-01 May-02 May-03 May-04 May-05 May-06 May-01 May-02 May-03 May-04 May-05 May-06
184
Borehole 217
217 Porewater concentrations
85
Bromate Bromide
84 84 84
83 83 83
82 82 82
81 81 81
80 80 80
79 79 79
78 78 78
77 77 77
Made Ground 76 76
76
Clay
75 75 75
Sand

74 74 74
Chalk
73 73 73
Monitoring well
screened interval
72 72 72
71 71 71
Nov-01 0 2 4 6 8 10 0 20 40 60 80
Soil Bromate Conc.
Nov-01

Nov-01
Soil Bromide Conc.
1.6 82 2.0 82
Groundwater Bromate
Groundwater Bromide
Water Level
80 1.6 80
1.2
78 78
1.2
0.8
76 76
0.8
Water Level (mAOD)
Water Level (mAOD)
74 74
0.4

0.4
72 72
0.0 0.0
May-01 May-02 May-03 May-04 May-05 May-06 May-07 May-01 May-02 May-03 May-04 May-05 May-06 May-07
185
Borehole 218 218
85 Porewater concentrations
Bromate Bromide
84 84 84
83 83 83
82 82 82
81 81 81
80 80 80
Made Ground
Clay
Sand 79 79 79

Chalk
Monitoring well 78 78 78
screened interval
77 77 77
Nov-01 0 2 4 6 8 0 4 8 12 16
Soil Bromate Conc.
Nov-01

Nov-01
Soil Bromide Conc.
8 82 9 82
Groundwater Bromate
Groundwater Bromide
Water Level
8
6 81 81
4 80 80
6
Water Level (mAOD)
Water Level (mAOD)
2

5
0 4
186
Borehole 219 219 BH Log
85
Bromate Bromide
84 84
84
83 83 83
82 82 82
81 81 81
80 80 80
Made Ground
Clay
Sand 79 79 79

Chalk
Monitoring well
78 78 78
screened interval
77 77 77
Nov-01 0 100 200 300 400 500 0 100 200 300 400 500
Soil Bromate Conc.
Nov-01

Nov-01
Soil Bromide Conc.
20 82 22 82
Groundwater Bromate
Groundwater Bromide
Water Level 20
16
81 18 81
12
16
8
80 14 80
Water Level (mAOD)
Water Level (mAOD)


4
12
0 10
187
Borehole 220
220
85
Bromate Bromide
85 85
84 84 84
83 83 83
82 82 82
81 81 81
Made Ground 80 80 80
Clay
Sand

79 79 79
Chalk
Monitoring well
78 78 78
screened interval
77 77 77
Nov-01 0 2 4 6 8 10 0 20 40 60 80
Soil Bromate Conc.
Nov-01

Nov-01
Soil Bromide Conc.
8 82 22 82
Groundwater Bromate
Groundwater Bromide
Water Level 7 80 80
20
6 78
78
5 18
76 76
4
Water Level (mAOD)
Water Level (mAOD)
74 16 74

3
72 72
2 14
188
Borehole 221 221 BH Log
85
Bromate Bromide
84 84
84
83
83 83
82 82 82
81 81 81
80 80 80
Made Ground
Clay
Sand 79 79 79

Chalk
78 78 78
Monitoring well
screened interval
77 77 77
Nov-01 0 400 800 1200 1600 2000 0 200 400 600 800 1000
Soil Bromate Conc.
Nov-01

Nov-01
Soil Bromide Conc.
500 82 200 82
Groundwater Bromate
Groundwater Bromide
Water Level
400
160
81 81
300
120
200
80 80
Water Level (mAOD)
Water Level (mAOD)
80

100
0 40
189
Borehole 222 222 Porewater concentrations
85
Bromate Bromide
84 84
84
83 83
83
82 82 82
81 81 81
80 80 80
Made Ground
Clay
Sand 79 79 79

Chalk
78 78 78
Monitoring well
screened interval
77 77 77
Nov-01 0 50 100 150 200 250 0 100 200 300 400
Nov-01 76 bromate concentration (mg/l) bromide concentration (mg/l)
Soil Bromate Conc.
Nov-01

Nov-01
Soil Bromide Conc.
100 82 70 82
Groundwater Bromate
Groundwater Bromide
Water Level 81 81
80 60
81 81
60 80 50 80
80 80
Water Level (mAOD)
Water Level (mAOD)
40 40

20 30
190
Borehole 223 223 Porewater concentrations
85
84 Bromate Bromide
84 84
83 83 83
82 82 82
81 81 81
80 80 80
79 79 79
78 78 78
77 77 77
Made Ground
Clay 76 76 76
Sand

75 75 75
Chalk
74 74 74
Monitoring well
screened interval 73 73
73
72 72 72
Nov-01 0 100 200 300 400 500 0 100 200 300 400 500
Soil Bromate Conc.
Nov-01

Nov-01
Soil Bromide Conc.
12 82 12 82
Groundwater Bromate
Groundwater Bromide
Water Level
81 81
10
8
81 81
80 80
4
Water Level (mAOD)
Water Level (mAOD)
6

80 80
0 4
191
Borehole 225
225 BH Log
84 84 Bromate 84 Bromide
82 82
80 80 80
78 78
76 76 76
Made Ground 74 74
Clay

Sand 72 72 72
Chalk
Monitoring well 70 70
screened interval
68 68 68
Nov-01 0 0.05 0.1 0.15 0.2 0.25 0 100 200 300 400 500
Soil Bromate Conc.
Nov-01

Nov-01
Soil Bromide Conc.
0.008 82 300 82
Groundwater Bromate
Groundwater Bromide
Water Level
80 80
0.006
200
78 78
0.004
76 76
100
Water Level (mAOD)
Water Level (mAOD)
0.002

74 74
0.000 0
May-01 May-03 May-05 May-07 May-01 May-03 May-05 May-07
192
therefore possible that bromate and/or bromide contamination could extend further to the northern part
of the site. Low bromate concentrations (<0.5 mg kg−1 ) were encountered in the southern and eastern
part of the site. Bromate concentrations were below detection limits of 0.10 mg kg−1 in all but one
of the samples tested during the August 2000 investigation (Komex, 2000). The samples included the
fluvio-glacial deposits and Chalk strata, at a range of depths up to 11.7 m bgl. It is possible that bromate
was not detected because the locations were sited away from the bromate production and handling areas.
Clear concentration-depth patterns are difficult to discern for the bromate depth profiles. However, in
a number of boreholes where elevated bromate concentrations are encountered, higher concentrations
occur in the Chalk (putty chalk) compared to concentrations in the shallower fluvio-glacial deposits
(clays and gravels).
5.6.1.3 Water table elevations in relation to contamination profiles

The zone of water table fluctuation is within the top of the putty chalk and bottom of the fluvial-glacial
deposits, and incorporates the peak porewater concentrations (Figure 5.8 and Figures 5.11 to 5.21).
5.6.1.4 Bromide and Bromate relationships

Results of the 2000 and 2001 investigations indicate a statistically significant relationship between soil
bromate concentrations and soil bromide concentrations (Figure 5.22). The relationship indicated by the
regression is:
Bromate conc. (mg kg−1 ) = 0.36 × Bromide conc. (mg kg−1 ) + 32.1 (mg kg−1 )
2001 soil samples (Atkins, 2002)

160 Bromate conc. ≥ 10 mg kg-1
120
Soil Bromate (mg kg-1)
80 Fit Results
Fit 1: Linear
Equation Y = 0.3574023353 * X + 32.0602463
Number of data points used = 7
Average X = 90.2857
40
Average Y = 64.3286
Residual sum of squares = 1077.19
Regression sum of squares = 7955.88
Coef of determination, R-squared = 0.88075
Residual mean square, sigma-hat-sq'd = 215.439
0
0 100 200 300

Soil Bromide (mg kg-1)
Figure 5.22: Relationship between soil bromate and soil bromide concentrations based on soil samples
from the 2001 site investigation (Atkins, 2002).
5.6.2 Spatial distribution of bromate and bromide within groundwater

Bromate and bromide concentrations in groundwater are shown in Figure 5.23. The distribution of
groundwater concentrations broadly reflects the pattern of soil concentrations, i.e. concentrations gen-
erally highest around the bromate and bromide production and handling areas, particularly around the
sumps, and the bromine bulk storage areas.
Porewater analysis from the saturated zone was undertaken as part of the Atkins (2002) investi-
gation, and the results are included in Figure 5.11 to Figure 5.21. Porewater bromate (and bromide)
concentrations measured in November 2001 for location 221, which had a maximum groundwater bro-
mate concentration of 400 mg l−1 , are approximately the same concentration as (mobile) groundwater
concentrations sampled at that time (Figure 5.18). However, for 219, also around the sump of the bro-
mate production area, and showing high bromate concentrations, porewater bromate from the saturated
zone was up to eight times higher than bromate concentrations in groundwater measured in November
2001, although bromide concentrations were similar in porewater and groundwater (Figure 5.16). At
location 220, 222 and 215 bromate and bromide porewater concentrations are approximately the same
concentrations as bromate and bromide in groundwater sampled in November 2001. At the remaining
locations, porewater samples were not available from the saturated zone.
Relatively high (135 mg kg−1 and 465 mg kg−1 ) porewater bromide concentrations were encoun-
tered in location 225, which is located approximately 150 m down hydraulic gradient of the SLC site.
However, concentrations of bromate in porewater were low (3 µg l−1 to 14 µg l−1 ). This is consistent
with groundwater concentrations at location 225: Relatively high bromide groundwater concentrations
were encountered (192 mg l−1 ) but concentrations of bromate were low (0.002 mg l−1 ).
5.6.2.1 Trends in groundwater concentrations 2000 to 2008

The groundwater monitoring wells installed in 2000 by Komex and in 2001 by Atkins have been moni-
tored for bromate and bromide as part of the bromate monitoring programme (Section 4.2).
Time series of groundwater monitoring data for these locations are included in Figure 5.11 to Fig-
ure 5.21. Overall, since 2000, groundwater bromide concentrations at the SLC site typically show a
decline between 2000 and 2003 and where monitoring data are available, values decline or remain stable
up until 2007. Overall since 2000, bromate concentrations have shown a reduction at monitoring loca-
tions within the site area, although recently on the basis of the data from the five locations still sampled,
there is evidence of some slight increases in some of the locations (221 and 223).
5.6.2.2 Trends in groundwater bromide concentrations 1984 to 2008

Comparison of bromide concentrations in boreholes from the monitoring between 1983 and 1987 (Fig-
ure 5.8) with monitoring data between 2000 and 2008 (Figure 5.11 to Figure 5.21) indicate a decline in
bromide concentrations within groundwater: a range of 25 mg l−1 to 10000 mg l−1 in 1984/1985 to a
range of 10 mg l−1 to 1000 mg l−1 in 2000/2001 at similar locations.
Bromide concentrations in saturated zone porewater also appear to have decreased significantly
from 10000 to 15000 mg l−1 at BH2-[84] (based on soil samples from the 1984 investigation (STATS,
1984)) to approximately 200 mg l−1 at location 221. Estimated porewater concentrations in 1984
Bromate concentration Groundwater Bromate Groundwater Bromate Groundwater Bromate Groundwater Bromate
(microgram per litre) Oct 2001 - Dec 2001 Jan 2002 - Mar 2002 Apr 2002 - Jun 2002 Jul 2002 - Sep 2002
4.0E+005
220 210500 210500 210500
3.6E+005 210500
3.2E+005 079 221
219
2.8E+005 214 216
222
2.4E+005 217
218
2.0E+005 080
1.6E+005 210450 215 210450 210450 210450
223
1.2E+005 081 Flux Flux Flux
Flux 082
8.0E+004 plane plane plane
4.0E+004
plane
083
0.0E+000
210400 210400 210400 210400
517050 517100 517050 517100 517050 517100 517050 517100
Bromide concentration Groundwater Bromide Groundwater Bromide Groundwater Bromide Groundwater Bromide
(microgram per litre) Oct 2001 - Dec 2001 Jan 2002 - Mar 2002 Apr 2002 - Jun 2002 Jul 2002 - Sep 2002
4.0E+005
220 210500 210500 210500
3.6E+005 210500
3.2E+005 079 221

219
2.8E+005 214 216
222
2.4E+005 217
218
2.0E+005 080
1.6E+005 210450 215 210450 210450 210450
223
1.2E+005 081
8.0E+004 082
Br Flux Br Flux Br Flux Br Flux
4.0E+004
plane 083 plane plane
0.0E+000 plane
210400 210400 210400 210400
517050 517100 517050 517100 517050 517100 517050 517100
195
Figure 5.23: Groundwater bromate and bromide contours at the ‘source zone’ based on samples taken in 2001 and 2002.
were approximately five times greater than sampled groundwater concentrations at the same location
in BH2-[84] and BH3-[84], but they were similar in BH1-[84].
Bromide concentrations at borehole 225 compared to borehole B1-[85] (Figure 5.24) which was
located in approximately the same location, indicate slightly higher concentrations in 2002 compared to
1985/1986. However, concentrations decline substantially into 2004 and remain low into 2007.
5.6.2.3 Relationships Bromide and Bromate in groundwater

Bromide concentration in groundwater samples from the source site shows a strong positive correlation
to bromate concentration (Figure 5.25).
5.6.3 Groundwater monitoring in the vicinity of the source site

In connection with the on-site investigations, some regional monitoring of groundwater quality (includ-
ing analysis for bromide, but not bromate) was undertaken between 1983 and 1987 from existing bore-
holes in the vicinity of the source site (Figure 5.27). These boreholes were incorporated into the bromate
monitoring programme which commenced in 2000 (Section 4.2).
Figure 5.27 shows the time series of bromide in groundwater between 1983 and 1987 and between
2000 and 2001. Borehole locations 028 (Orchard Garage) and 019 (Nashe’s Farm) show potential in-
creases from concentrations between 1983 and 1987 to concentrations between 2000 and 2008. However,
a particularly high concentration at Orchard Garage confuses this pattern. A potentially anomalous high
concentration is also seen at location 020 (Capp’s Cottage). Boreholes 017, 018, 020 and 025 show
small decreases from concentrations between 1983 and 1987 to concentrations between 2000 and 2008.
Between 2000 and 2007 concentrations tend to show relatively constant concentrations, although a slight
decline is noticeable at locations 028 and 019.
Porewater analysis was also undertaken at the borehole locations 227 and 228. At locations
227 and 228, at distance of approximately 1 km and 2 km respectively down-gradient of the site along
House Lane, Chalk porewater bromate concentrations were low (<5 µg l−1 ) in both locations. Bromide
concentrations were relatively high in borehole 227 (1-4 mg kg−1 ) but low (<0.5 mg kg−1 ) in 228. In
each case, concentrations appear to be highest in the top of the Chalk and decrease with depth.
5.6.4 Relationships between contaminant concentrations and water levels

As described in Chapter 4, water levels were exceptionally high in 2001 (a very wet year in which
localised groundwater flooding occurred). Peak groundwater levels at Orchard Garage (location 028)
occurred in April 2001. Groundwater levels then declined into January 2002 before rising again to
follow the normal seasonal trends and amplitude.
Most locations at the source site show a substantial decrease in bromate and bromide concentration
from November/December 2001 to January 2002. It is possible that high groundwater levels in 2001
contributed to the high bromate concentrations recorded in November/December 2001: high groundwa-
ter levels are likely to have led to increased mobilisation of bromate and bromide (leaching from soil
and diffusion out of porewater held in matrix of the unsaturated zone). Higher groundwater levels may
also have activated different flow paths, transferring bromate to a wider area, which may explain why
225 Bromide B1 Bromide
300000 8 140000 25
120000 20
6
200000
100000 15
80000 10
Water Level (mAOD)

100000
Water Level (mAOD)
Bromide concentration (µg/l)

2
60000 5
225 Bromide
225 Water Level
0 0 40000 0
May-01 May-02 May-03 May-04 May-05 May-06 May-07 May-08 May-84 May-85 May-86 May-87
84 84 Porewater Bromide
225
84 Bromide
Groundwater Bromide conc.
82 Porewater Bromide conc.
80 80
80
78
76 76
76
74
72 72

72
70
68 68
68
0 100000 200000 300000 400000 500000 0 200000 400000 600000 800000
bromide concentration (µg/l) Bromide concentration (µg/l)
197
Figure 5.24: Depth profiles of porewater bromide compared to pumped groundwater concentrations for Borehole B1 from the 1985 investigation (Chemfix, 1985a) and
Borehole 225 from 2001 investigation (Atkins, 2002) which are believed to have been located in similar positions.
Figure 5.25: Relationship between bromide and bromate concentration in groundwater samples from the
monitoring data between 2000 and 2008 for locations 079 to 083 and locations 214 to 223.
bromate and bromide is seen over a wider area in 2001, followed by a rapid decline. Bromate relation-
ship to water levels have been assessed in Section 4.6. At the source site, the regression relationships for
the response of bromate concentration to water level are only statistically significant (P<0.05) for three
locations out of the six locations with sufficient data to undergo statistical analysis (Figure 4.37).
5.6.5 Leachate results

Leachate tests measure the amount of contaminant mass associated with the solid (soil) phase that
is mobile in the liquid (water) phase passing through the soil. Samples from the 2001 investigation
(Atkins, 2002) were submitted for leachate analysis using Environment Agency recommended method-
ology (Lewin et al., 1994). Concentrations of bromate and bromide in leachate samples from the 2001
investigation are strongly correlated with bromate and bromide concentration in the corresponding soil
sample (Figure 5.28). For bromide, there are no obvious differences in the regression relationships be-
tween samples from putty Chalk, clayey GRAVEL and gravelly CLAY samples. For bromate, samples
from the fluvio-glacial deposits appear to leach slightly more (steeper regression line) than samples from
the putty Chalk. As there is only one sample from clayey GRAVEL, it not possible to discern any
differences between leaching behaviour between the clayey GRAVEL and gravelly CLAY samples.
5.6.5.1 Partition coefficients, Kd

A partition (or distribution) coefficient, Kd , describes the distribution of a species between a solid and
aqueous matrix after equilibration. Sorption mechanisms include ion exchange (in particular cation
exchange) and surface complexation. In groundwater risk assessments, the Kd value describes the degree
of sorption of a particular species in the leachate and/or groundwater to the soil or rock that is in contact
with that liquid. Partition coefficients are expressed in units of L3 M−1 (e.g. l kg−1 or ml g−1 ).
224
Groundwater monitoring
(Bromide)
Data available:
1983-1987
079 1983-1987 and 2000-2006

2000-2006
225
B1[85]
028
017
025
018
226
020
220
079 221
219
BH1[84]214216 222
217
218
C1[85]080 227
215
BH3[84] 223
081 019
082
083
0 50
Meters 225
B1[85]
0 500 © Crown Copyright/database right 2008. An Ordnance Survey/EDINA supplied service.

Meters Geological Map Data © NERC 2008.
199
Figure 5.26: Groundwater monitoring locations in the vicinity of the source site that have been sampled for bromide concentrations between 1983 and 1987 and between 2000
and 2008.
300000 8000
225 028
019
224 200000 6000
100000 4000
Bromide Concentration (µg/l)

221222
223
216219

083 225 0 2000
May-01 May-02 May-03 May-04 May-05 May-06 May-07
028
017
025
0
KEY
Jan-83 Jan-86 Jan-89 Jan-92 Jan-95 Jan-98 Jan-01 Jan-04 Jan-07
Average Bromide 2000-2002 018
not present in 2000-02
0 - 10 mg/l 226
10 - 100 mg/l
100 - 500 mg/l 020
Boundary of
former factory site
227
6000 200 017
019
018
020
160
025
120
8000
228
227 4000
226 80
224
6000
40
0
4000
2000
Bromide concentration (µg/l) May-00 May-01 May-02 May-03 May-04 May-05 May-06 May-07
Bromide Concentration (µg/l)

2000
228
0
0
May-01 May-02 May-03 May-04 May-05 May-06 May-07 Jan-83 Jan-86 Jan-89 Jan-92 Jan-95 Jan-98 Jan-01 Jan-04 Jan-07
0 0.15 0.3 0.6 Kilometers

200
Figure 5.27: Groundwater bromide concentrations at monitoring locations in the vicinity of the source site 1983 to 2008.
20 BROMATE gravelly CLAY

clayey GRAVEL
putty CHALK
Trend line
16
Leachate Bromate (mg/l)
12
Fit Results
8
BrO3 Leachate-Soil
Equation Y = 0.07036647551 * X
4 Average X = 58.3195
Average Y = 4.70377
0
0 100 200 300

Soil Bromate (mg/kg)
12 BROMIDE
Leachate Bromide (mg/l)
Fit Results
4 Br Leachate-Soil
Equation Y = 0.08152457005 * X
Average X = 45.2182
Average Y = 3.68909
0
0 40 80 120 160
Soil Bromide (mg/kg)
Figure 5.28: Relationships between leachate concentration (mg l−1 ) and soil concentration (mg kg−1 )
for samples from the 2001 investigation (Atkins, 2002).
5.7. Mass of bromide and bromate at the source site 202
The partition (or distribution) coefficient, Kd , was calculated as described in Lewin et al. (1994) as
Csorbed
Kd =
Cdissolved
Where Csorbed is the sorbed contaminant concentration (mass of contaminant in mg ÷ mass of soil
in kg) and Cdissolved is the dissolved contaminant concentration (mass of contaminant in mg ÷ volume
of solution in litres).
The soil concentration (mg kg−1 ) Csoil was assumed to represent the total mass of contaminant
(sorbed plus dissolved). The dissolved contaminant concentration Cdissolved was assumed to be the
leachate concentration (mg l−1 ). Csorbed was calculated as the total soil concentration, Csoil , minus the
dissolved mass per kg of soil (Cleachate × L/S ratio2 ). Performing these calculations on the samples from
the the 2001 investigation (Atkins, 2002), gave a mean Kd of 11.8 l kg−1 for bromate and 12.7 l kg−1
for bromide.
5.7 Mass of bromide and bromate at the source site

5.7.1 Previous estimates
A review of available data by the Environment Agency (Roberts, 2001) estimated that between 7,535 kg
and 14,135 kg of bromide was present in the unsaturated zone (assumed to be 4 m thick); the estimated
mass in the unsaturated zone after excavation of 2,406 kg bromide in the top layer of soil in 1986 was
between 5,129 and 11,729 kg.
These estimates were based on the soil bromide distribution indicated by the results of the 1985
investigation (Chemfix, 1985c). The estimates were calculated by Roberts (2001) as follows:
• The site was divided up based on a 10 m × 10 m grid. Assuming a thickness of unsaturated zone
of 4 m, and a bulk density of 2000 kg m−3 , there is 8×105 kg of material in each block. The mass
of bromide in each block was then calculated as:
MBr = CBr × 10−6 × Ms
where MBr is the mass (in kg) of bromide in a block, CBr is the concentration (in mg kg−1 ) of
bromide in the block and Ms is the mass (in kg) of soil in block. Calculating this for each block
and summing the answers gives a total amount of bromide on site in the unsaturated zone.
• Two estimates were undertaken:
1. Using just the squares where samples were taken (a total of 20 squares) gave a total amount
of 7,535.2 kg bromide on site.
2. Assuming that the squares with no analysis had concentrations the average of those squares
surrounding them, and the contamination at the boundary is zero, gave a total of 14,135 kg
bromide on site.
2 Liquid to Solid ratio. For NRA leachate analyses (Lewin et al., 1994), the Liquid to Solid ratio is 10:1 (l kg−1 )
• The mass of bromide removed as part of the top layer was estimated by assuming that based on
the grid pattern, the remediation proposals encompassed the excavation of a total of 17 blocks
to a depth of 0.75 m, and 23.5 blocks to a depth of 1.5 m. Calculating the mass of bromide for
each block and summing the results, gave a total of 2406 kg (566 kg + 1840 kg). Subtracting this
amount from the estimates of mass above gave a residue of 5129 kg and 11729 kg of bromide
present in the unsaturated zone after contaminated soil removal.
5.7.2 New estimates

Additional estimates of bromide and bromate mass (Table 5.3) using additional data and a method of
3-D interpolation, were made as part of this thesis.
The Voxler software package (Golden Software) was used to plot data as (X,Y,Z) data points, with
bromate/bromide concentration as a fourth variable. Voxler was used to interpolate bromate/bromide
concentrations as a 3-D lattice (X,Y,Z,C). This provided a visualisation of the three-dimensional dis-
tribution of bromate and bromide contamination (Figure 5.7, Figure 5.9 and Figure 5.10). In order to
estimate the mass of bromate/bromide present in the soil, bromate/bromide concentration contours were
exported to Surfer as lattice slices of specified Z thickness (Figure 5.30 and Figure 5.29). The contours
were ‘blanked’ at the site boundary so that only bromide/bromate within the site boundary was included
in the mass estimates. Surfer was then used to calculate the ‘volume’ enclosed by each contoured sur-
face. This ‘volume’ (area × concentration), when multiplied by the thickness of the slice and the bulk
density of soil, represents the mass of bromate/bromide within the soil. The results from each slice (Fig-
ure 5.30 and Figure 5.29) were summed to calculate a total mass of bromate/bromide within the total
thickness of unsaturated zone.
Maximum, mean and minimum unsaturated and saturated zone thicknesses are defined as follows
(Figure 5.31):
• Based on the available samples in the vicinity of the source zone (i.e. from Boreholes 221 and 219),
bromate and bromide is present (above background concentrations) in porewater to a maximum
depth of approximately +77.5 m OD. The maximum depth of sampling was at +75 m OD in
Borehole 223, and samples indicated very low concentrations (below or close to the MDL). It is
therefore assumed that the depth of contamination extends to between approximately +75.5 m OD
and +77.5 m OD beneath the source site.
• Ground level in the vicinity of the source zone is approximately +83.5 m OD.
• Water levels in the source zone range from +81.5 m OD to +79.0 m OD.
• Therefore, the thickness of unsaturated zone ranges from 2.0 m to 4.5 m thick and the thickness
of saturated zone ranges from 1.5 m to 6.0 m thick. The mean unsaturated zone thickness (based
on 2001-2008 data) is 4.1 m, and the mean saturated zone thickness is 3.4 m.
+83.6 to +82.6 mOD +82.6 to +81.6 mOD

Bromate concentration 210520
250 mg/kg 220 220

210500
200 mg/kg 079 079
221 221
219 219
150 mg/kg
210480 214 216 214 216
125 mg/kg 222 222
217 217
100 mg/kg 218 218
210460 080 080
75 mg/kg
215 215
50 mg/kg 223 223
210440 081 081
10 mg/kg 082 082
0 mg/kg 210420
083 083
210400
517020 517040 517060 517080 517100 517120 517140
+81.6 to +80.6 mOD +80.6 to +79.6 mOD +79.6 to +78.6 mOD
220 220 220
079 221 079 221 079 221

219 219 219
214 216 214 216 214 216

222 222 222
217 217 217
218 218 218
080 080 080
215 215 215

223 223 223
081 081 081
082 082 082
083 083 083
220 220 220
079 221 079 221 079 221

219 219 219
214 216 214 216 214 216

222 222 222
217 217 217
218 218 218
080 080 080
215 215 215

223 223 223
081 081 081
082 082 082
083 083 083
Figure 5.29: Bromate soil concentration contours for 1.0 m thick grid slices based on investigation data
from 2000 and 2001 (Komex, 2000; Atkins, 2002). Estimates for total mass in the unsaturated and
saturated zones refer to minimum, mean and maximum thicknesses defined in Figure 5.31.
+83.6 to +82.6 mOD +82.6 to +81.6 mOD

Soil Bromide conc. 210520
250 mg/kg 220 220

210500
200 mg/kg 079 221 079 221
219 219
150 mg/kg 210480 214 216 214 216
222 222
125 mg/kg 217 217
218 218
100 mg/kg 210460 080 080
75 mg/kg 215 215

223 223
50 mg/kg 210440 081 081
10 mg/kg 082 082
0 mg/kg 210420
083 083
210400
517020 517040 517060 517080 517100 517120 517140
+81.6 to +80.6 mOD
+80.6 to +79.6 mOD +79.6 to +78.6 mOD
220 220 220
079 221 079 221 079 221

219 219 219
214 216 214 216 214 216

222 222 222
217 217 217
218 218 218
080 080 080
215 215 215

223 223 223
081 081 081
082 082 082
083 083 083
220 220 220
079 221 079 221 079 221

219 219 219
214 216 214 216 214 216

222 222 222
217 217 217
218 218 218
080 080 080
215 215 215

223 223 223
081 081 081
082 082 082
083 083 083
Figure 5.30: Bromide soil concentration contours for 1.0 m thick grid slices based on investigation data
from 2000 and 2001 (Komex, 2000; Atkins, 2002). Estimates for total mass in the unsaturated and
saturated zones refer to minimum, mean and maximum thicknesses defined in Figure 5.31.
Ground Level
Min. unsat. Zone

+83.5 mOD
Max. unsat. Zone

2.0 m
4.5 m
Max Water Level
+81.5 mOD
Min Water Level

Max. sat. Zone
Min. sat. Zone
Min. ‘plume’
+79.0 mOD
Max. ‘plume’
6.0 m
1.5 m
1.5 m
6.0 m
Max recorded
contamination
+77.5 mOD
Max depth of
sampling
+75.5 mOD
Figure 5.31: Minimum and maximum saturated and unsaturated zone thicknesses.
5.8. Mass flux of bromate in groundwater migrating from the source site 207
Table 5.3: Summary of mass estimates. Estimates for total mass in the unsaturated and saturated zones
refer to minimum, mean and maximum thicknesses defined in Figure 5.31.
1984-1985 1986 2000-2001

MIN MEAN MAX mass MIN MEAN MAX
excavated
BROMIDE unsat. zone (kg) 12745 24907 27617 -2800 402 812 916
BROMIDE sat. zone (kg) 7672 17936 32645 0 320 758 1283
TOTAL
BROMIDE (kg) – – – – – 1570 –
BROMATE unsat. zone (kg) – – – – 252 508 572
BROMATE sat. zone (kg) – – – – 191 445 764
TOTAL
BROMATE (kg) – – – – – 953 –
5.8 Mass flux of bromate in groundwater migrating from the

source site
The groundwater flux, F , across a plane perpendicular to groundwater flow direction taken directly
through the source area (Figure 5.23), is given by the equation:
Z x=B
F = Dvne C dx (5.1)
x=A
where D is the depth of the contaminated ‘plume’, v is the groundwater velocity and ne is the effective
porosity
The linear velocity, v, is related to the darcy velocity, q, by q = vne , so equation 5.1 can be written
as:
Z x=B
F = Dq C dx (5.2)
x=A
R x=B
The integral x=A
C dx is the area under a graph of concentration C against distance x along the
section line from A to B (Figure 5.32).
As described in Section 5.7.2, the bromate contamination is estimated to extend to depths of be-
tween +75.5 m OD and +77.5 m OD beneath the source site. Water levels in the source zone range from
+81.5 m OD to +79.0 m OD. Therefore, if the depth of the contaminated ‘plume’ is taken as the distance
between the top of the water column and the estimated deepest recored contaminated porewater sample,
then the thickness of of ‘plume’ could range from 1.5 m to 6.0 m (Figure 5.31).
Using a range of 1.5 m to 6.0 m for D, and a darcy velocity q of 0.9 m d−1 based on the results
of the single borehole dilution tests (Section 3.6.1) at the nearby Harefield House (location 226) and
Nashe’s Farm (location 019) boreholes, groundwater flux estimates for bromate (Figure 5.32 range from
5.9. Previous representations of the ‘Source Term’ 208
6.3 kg d−1 to 41 kg d−1 (572 to 3770 kg y−1 ) for a ‘thick plume’ and 1.6 kg d−1 to 10 kg d−1 (2289
to 15080 kg y−1 ) for a ‘thin plume’. Flux estimates for bromate apparently decrease with time between
2001 and 2003.
These flux estimates seem extremely high in relation to the amount of bromate present at source
site in the unsaturated and saturated zone (Section 5.7.2). This suggests that the dominant contributor to
the groundwater concentrations is bromide/bromate in the saturated zone porewater, and that the mass
of bromide/bromate in the porewater has not been fully accounted for in the mass calculations. Also, the
‘plume’ may extend deeper than accounted for in the mass calculations and/or concentrations may be
higher in locations where samples were not taken.
5.9 Previous representations of the ‘Source Term’

The representation of bromate (and bromide) release from the source site has received relatively little
attention in previous assessments of the Hertfordshire bromate contamination. The assessments which
have considered and quantified the ‘source term’ are discussed in the sections below.
5.9.1 Early assessments: 1984 and 1985

(Chemfix, 1984, 1985d) assumed an average porewater concentration of bromide of 5000 mg l−1 in
1984 over an area of 1200 m2 to a depth of 5 m, resulting in a total mass of bromide of 9400 kg.
The conceptualisation assumed that the contaminated porewater was replaced by infiltration. Therefore,
assuming an infiltration rate of 30 mm y−1 for the developed site, and 250 mm y−1 for the fallow
site, this results in a bromate loading to the aquifer beneath the site of 180 kg y−1 and 1500 kg y−1
respectively. This would have resulted in a source lifetime of 52 years and 584 years respectively.
The modelling studies of (Chemfix, 1984, 1985d) assumed that leaching of the contamination from
the soil from 1984 was the sole source of bromide into the aquifer. However, bromide concentrations in
groundwater of around 9000 mg l−1 at the time indicate that there was already a considerable ‘plume’
of contamination present at the site which may have represented the effects of up to 30 years of con-
tamination prior to 1984. The mass input to the aquifer is therefore likely to have been considerably
underestimated by the Chemfix modelling.
5.9.2 Recent assessments: 2002 to 2008
5.9.3 CONSIM modelling

Atkins (2002) formed a conceptual model for the source site comprising a bromate source zone within
the unsaturated zone soils. The pathway to the saturated zone was by leaching of bromate by infiltrating
recharge through the contaminated unsaturated zone.
The soil source was modelled based on the presence of two ‘hotspots’ located in the north-eastern
area of the site around locations 219 and221, where the highest soil concentrations were encountered.
The source was modelled by a 10 m by 10 m block around each location, corresponding to a 20 m long
by 10 m wide source zone. The source zone thickness was 2.0 m to 2.5 m based on the distance from
the base of the made ground to the rest groundwater levels (November 2001). The soil source bromate
Bromate Bromate
1.5 m thick plume 160000 Jan-Mar 2002 60000 Jan-Mar 2002
Area under graph
(ug/l * m) Flux (kg/d) Flux (kg/y)
Oct-Dec 2001 7645509 10.3 3770 120000
40000
Jan-Mar 2002 3253520 4.4 1604
Apr-Jun 2002 2863282 3.9 1412 80000
Jul-Sep 2002 2404950 3.2 1186 20000
Oct-Dec 2002 2143395 2.9 1057 40000
Jan-Mar 2003 1263853 1.7 623
Apr-Jun 2003 1735090 2.3 856 0
Bromate concentration (µg/l)

0

Jul-Sep 2003 1908070 2.6 941
Oct-Dec 2003 1160306 1.6 572 0 40 80 120 0 40 80 120
Bromate Bromate
160000 Apr-Jun 2002 80000 Apr-Jun 2003
Area under graph 6.0 m thick plume
(ug/l * m) Flux (kg/d) Flux (kg/y)
120000 60000
Oct-Dec 2001 7645509 41.3 15080
Jan-Mar 2002 3253520 17.6 6417
Apr-Jun 2002 2863282 15.5 5647 80000 40000
Jul-Sep 2002 2404950 13.0 4743
Oct-Dec 2002 2143395 11.6 4228 40000 20000
Jan-Mar 2003 1263853 6.8 2493
Apr-Jun 2003 1735090 9.4 3422 0

0

Jul-Sep 2003 1908070 10.3 3763
Oct-Dec 2003 1160306 6.3 2289 0 40 80 120 0 40 80 120
Bromate Bromate
120000 Jul-Sep 2002 100000 Jul-Sep 2003
80000
80000 60000
40000
40000
20000
0

0 40 80 120 0 40 80 120
Bromate
400000 Oct-Dec 2001 Bromate
5.9. Previous representations of the ‘Source Term’
Bromate
100000 Oct-Dec 2002 60000 Oct-Dec 2003
300000
80000
40000
200000 60000
100000 40000 20000

20000
0

0
0
0 40 80 120
Distance x along flux plane (m) 0 40 80 120 0 40 80 120
209
Figure 5.32: Estimates of bromate groundwater flux from the ‘source zone’ using equation 5.2 and the area under a concentration profile taken across a flux plane through the
R x=B
source zone. The area under a graph represents the integral x=A C dx. The flux plane is shown in Figure 5.23.
5.10. New Conceptual Models for Contaminant Release 210
concentration was assigned a range of 1.7 mg kg−1 to 273 mg kg−1 based on the analytical results of
samples collected from the unsaturated soil zone at these locations.
The probabilistic CONSIM modelling package was then applied to predict leaching of soil contam-
inants through the unsaturated zone and to predict concentrations arising in the Putty Chalk aquifer. The
subsequent dilution within the saturated Chalk aquifer (a 10 m thick mixing zone comprising saturated
Putty Chalk and Blocky Chalk) was also simulated. The results of the simulations indicated bromate con-
centrations close to observed groundwater concentrations at monitoring locations 219 and221. Atkins
(2002) concluded that this indicated that there was a ‘significant pollutant linkage’ (SOURCE → PATH-
WAY → RECEPTOR) between the bromate contamination source within the unsaturated zone soils and
groundwater within the Chalk aquifer receptor beneath the site, via a pathway of leaching through the
unsaturated zone.
5.9.4 MT3D modelling

Atkins (2005) used a constant source term of 5000 µg l−1 bromate at the source site from 1970 for the
duration of the modelling period until 2050. A justification for this source term was not given explicitly.
The 5000 µg l−1 concentration is equivalent to a constant input to the saturated zone of approximately
10 kg d−1 or 3650 kg y−1 . This corresponds well to the bromate flux estimated from the source site
(Section 5.8).
A constant source term is not realistic for the site: the bromate source will be depleted as mass is
transported down-gradient and a steady concentration would not be expected to be maintained into the
future. Also, monitoring data for bromide in groundwater at the site show that concentrations have de-
creased considerably between 1984/1985 and 2000/2001 (Section 5.6.2.2). There is also some indication
that concentrations have declined at locations in the vicinity of the source site (Section 5.6.3). If bromate
contamination has followed a similar history to bromide contamination, then it would be expected that
bromate concentrations at the source site would have also shown some decline.
5.10 New Conceptual Models for Contaminant Release

Based on the groundwater distribution outlined in Section 5.6.2, the ‘source zone’ can be taken as the
broad area encompassing the highest bromate and bromide concentrations, corresponding to the area
around the former bromine storage, solid bromate handling and liquid bromate production areas, and
also immediately downgradient of the sump for the non-bromate production area (Figure 5.33).
Based on the distribution of bromide from investigations in 1983, 1984 and 1985, it is apparent
that by the end of the operational lifetime of the chemical works, considerable bromide (and most likely
bromate), had accumulated in the unsaturated and saturated zone beneath the source area. Figure 5.34
presents a conceptual model for the release of bromate and bromide to groundwater beneath the source
zone. The mechanism by which the contamination was released and accumulated between 1955 and 1983
is unknown. The main areas of uncertainty in the history of bromate and bromide release to groundwater
beneath the site are summarised in Table 5.4.
Table 5.4: Main areas of uncertainty in the history of bromide and bromate release to groundwater
beneath the source site.
Area of uncertainty Observations and constraints
Timing of input The presence of high concentrations of bromide in

saturated and unsaturated zone porewater and
groundwater in locations beneath the source site, and at a
Release of contaminants could have begun distance of approximately 150 m down-gradient (borehole
at an early stage of the operational lifetime of B1-[85]), indicate that bromide input must have
the chemical works (c. 1955 - 1980), or could commenced prior to 1983 (shallow porewater), 1984
have occurred towards the end of operation, (porewater and groundwater at the site) and 1985
or during decommissioning and demolition of (porewater and groundwater 150 m down-gradient of the
the works (c. 1985-1987). source zone).
Form of input
The bromide concentrations in soil porewater and
groundwater recorded at the source site between 1983 and
The form of release could range from a
1987 are significantly higher than concentrations recorded
relatively constant input through continuous
in 2000 and 2001. The buildings were cleared and the site
leak/discharge, to an input over a short
was left to free-drain in 1984. The site clearance is likely to
period of time as a result of a catastrophic
have coincided with a substantial increase in infiltration
leak/discharge or a recharge pulse.
rate, and therefore a pulse of recharge. This recharge
pulse may have been important in transporting bromide
Release may have occurred as a focused
(and bromate) from unsaturated zone to the saturated
input over a small area of the site (e.g. via
zone.
sumps) or may have occurred as a more
diffuse input over a larger area.
Similarity of bromide and bromate release Based on the site investigation data for 2000 and 2001, the
history spatial distribution of elevated bromide and bromate in
porewater and groundwater shows similarities. Also,
The release history of bromate and bromide bromate and bromide concentrations are positively
may or may not be related and show correlated based on groundwater monitoring data from
similarities in the form and timing of bromate 2000 to 2007.
input.
However, the lack of data for bromate concentrations prior
If the predominant mobilisation of to 2000 means that it is uncertain whether or not bromate
contamination occurred via a pulsed concentrations were similarly elevated in the early 1980s.
recharge event, the form of bromide and The only concentrations of bromate available are for
bromate release from 1984 is likely to be shallow soil in 1983, and all samples were below detection
similar. limits. Therefore, it is possible that bromate contamination
release occurred later than bromide contamination (hence
However, if releases prior to 1983 have been bromate not as widespread in 1983), and/or at lower
more important in the mobilisation of bromide relative quantities.
to groundwater, then due to lateral
separation of bromide and bromate It is also possible that bromate was elevated in 1983 but
production areas, the relative magnitudes was not detected because release mechanism meant that
and timing of release may differ its distribution was deeper than 1.5 m (e.g. a more focused
considerably. input into drainage via sumps).
Figure 5.33: The combined ‘source zone’ (centre figure) based on the locations of high concentrations
of bromate (left hand figure) and bromide (right hand figure) in groundwater
5.10.1 Mechanisms of bromide and bromate release

It is likely that the original bromate and bromide contamination accumulated from spills or leaks of
substances containing bromide and/or bromate (and/or bromine) during the operation of the chemical
works. It is possible that bromide and bromate were discharged to groundwater via the drainage system,
allowing rapid introduction of these contaminants to deeper strata or direct to groundwater.
Due to the high solubility of bromide and bromate, it is assumed that bromide and bromate present
within the unsaturated zone soils are present as dissolved ions in porewater rather than associated with
the solid phase. The conceptual model considers that bromide and bromate ions are leached by infil-
trating water and transported through the unsaturated zone to groundwater within the saturated ‘putty
chalk’. Groundwater transports contaminants (vertically) through the ‘putty chalk’ to the underlying
‘blocky chalk’, where contaminants are transported laterally with groundwater flowing through the fis-
sures. Double-porosity diffusive exchange of contaminants occurs between groundwater within the fis-
sures and porewater within the saturated chalk. These processes and mechanisms are described in more
detail in the following sections.
5.10.1.1 Leaching of contaminants from unsaturated zone

The conceptual model considers that bromide and bromate ions are leached by infiltrating water and
transported through the unsaturated zone to groundwater within the saturated putty chalk. Before the
contaminated shallow soils were removed, there would have existed a plentiful source of bromide (and
possibly bromate) contamination which would have been readily leached by infiltrating water. During the
Figure 5.34: Conceptual Model for bromate and bromide release from the source zone.
operation of the factory, much of the site was covered with buildings or hard-surfaces which would have
restricted the amount of recharge. However, a much higher rate of recharge would have been possible
when the site was cleared and left to free-drain (end of 1983 to beginning of 1987). Leaching would have
the effect of transporting bromate and bromide ions through the unsaturated zone to groundwater within
the saturated putty chalk. The large decrease in bromide concentrations encountered in 2000 and 2001
compared to those recorded between 1983 and 1985 in comparable locations and depths, suggests that
high recharge rates during 1984 to 1987 may have played an important role in leaching out contamination
from the unsaturated zone profile.
Assuming recharge to occur via ‘piston flow’ (downward movement of water that has infiltrated at
the surface occurs via vertical drainage through unsaturated matrix porewater), a minimum time for the
complete leaching of the contaminant from the soil profile can be estimated by calculating the length of
time it would take to replace the water contained within the unsaturated zone soil profile:
VU
Time for replacement of water in unsat. zone = VR
Where
VU = Volume of water in unsaturated zone = Area x moisture content
VR = Volume of water recharged per year = Area x recharge rate
Following the estimates of Roberts (2001), assuming a 4 m unsaturated zone with moisture content
ranging from 11 % to 13 % and recharge rate 150 mm y−1 to 350 mm y−1 , the minimum time for
replacement is between 3.6 years and 1.3 years, with average values of approximately 2 years. These
estimates predict that all the water present in the unsaturated zone could be replaced over a period of
approximately 2 years, which implies that if the site was cleared at the end of 1983, and development
took place at the beginning of 1987, all the water present in the unsaturated zone at the time of site
clearance would have been replaced by infiltrated rainwater. Therefore, if the bromate and bromide ions
behave conservatively (i.e. they are transported at the same rate as the water in which they are dissolved),
then the ‘free’ bromide and bromate would have been almost completely removed from the unsaturated
zone over this period.
However, it is clear that although bromate concentrations are generally low in the upper part of
the unsaturated zone, relatively high concentrations still remain in the lower unsaturated zone (SUZ).
The presence of discontinuous clay horizons within the fluvial-glacial deposits may have the effect of
restricting recharge in some areas whilst concentrating recharge through other ‘windows’ in the clay.
Where clay horizons restrict recharge, this may account for the continued presence of bromide and
bromate within the unsaturated zone (present either as a result of transport from above, or introduction
during high groundwater levels), despite high potential rates of leaching.
The bromide ion is generally thought to be conservative (Section 1.7) and so is transported at the
same rate as the water in which it is dissolved. If the bromate ion behaves like the bromide ion, i.e.
conservatively, then bromate will also be transported at the same rate as the water in which it is dissolved.
However, the results of leachate tests carried out in 2001 (Section 5.6.5) at the site indicates that some
partitoning of bromate and bromide occurs between the soil phase and dissolved phase. The downward
movement of bromate and bromide is therefore inhibited by soil interaction, and the time needed for
complete removal from the soil is likely to be somewhat increased compared to the movement of soil
porewater.
5.10.1.2 Evidence for pulse release

The bromide concentrations in soil porewater and groundwater recorded at the source site between 1983
and 1987 are significantly higher than concentrations recorded in 2000 and 2001. The buildings were
cleared and the site was left to free-drain in 1984. The site clearance is likely to have coincided with a
substantial increase in infiltration rate 3 and therefore a pulse of recharge. This recharge pulse may have
been important in transporting bromide (and bromate) from unsaturated zone to the saturated zone. If
this was the case, a sudden rise in groundwater concentrations at the source site, and/or a decrease in
unsaturated zone concentrations between investigations in 1984 and 1985 might be expected.
The site was being cleared during the investigation in May 1984 (STATS, 1984). There are no re-
ported groundwater samples from locations at the source site prior to this. The time series of groundwater
concentrations from the three boreholes sampled in 1984 does indicate a rise in bromide concentrations
from May 1984 to September 1984. Comparing concentration-depth profiles from BH1-[84] and C1-
[85] which are reported to have been in approximately the same locations, bromide concentrations at
similar depths appear to be less (about 25 % to 30 %) in 1985 compared to 1984. However, the profile
has not clearly moved down as might be expected for piston-flow recharge, although it is difficult to be
conclusive on this matter since fewer depths were sampled in 1984.
5.10.1.3 Migration down-hydraulic gradient in mobile groundwater

Due to the low permeability of the ‘putty chalk’ layer, groundwater transport through the saturated putty
chalk is likely to occur predominantly in the form of vertical drainage. Diffusion into the matrix will
result in retardation of bromide and bromate contamination compared to advective transport. Within
the ’blocky chalk’ bromate and bromide contamination will be transported down hydraulic gradient
by advection, with dispersion and double-porosity diffusive exchange acting to retard the transport of
contamination relative to groundwater flow.
5.10.1.4 Diffusive exchange with immobile porewater within the saturated zone
The (blocky) Chalk is considered to behave as a double-porosity aquifer with groundwater flow occur-
ring predominantly through fissures, and the high porosity, low permeability matrix providing storage
(Section 2.5). Diffusive exchange between the mobile fissure water and the immobile matrix water will
occur during groundwater flow (Section 2.9.4).
Assuming initially contaminated fracture water and uncontaminated matrix water, bromate diffu-
sion into the matrix water would have acted to retard the transport of bromate down-gradient. At a later
stage, which may or may not have been reached at the SLC site, when the original source of contam-
ination within the fracture water has ceased, the direction of diffusion will be from the contaminated
3 Estimates of recharge (the portion of rainfall that percolates down to the water table) for free draining site are between
150 mm y−1 and 350 mm y−1 compared to approximately 60 mm y−1 for the site with predominant cover of hardstanding.
5.11. Source terms for bromide and bromate release from the source site 216
matrix water to the less contaminated fracture water, which will act to prolong the period of bromate
contamination.
5.10.1.5 Seasonal mobilisation of porewater within the zone of water table fluctuation
Within the zone of water table fluctuation, also referred to as the seasonally unsaturated zone (SUZ),
fractures are periodically filled and drained when the water table fluctuates. Due to the low permeabil-
ity of the Chalk matrix, the matrix remains saturated. Diffusive exchange between fracture water and
matrix water occurs when the fractures are saturated. This may provide a seasonal source (or sink) to
groundwater in the saturated putty chalk.
This process may explain the concentration of bromate and bromide within the SUZ. During periods
of high groundwater levels, diffusion is likely to have occurred between contaminated groundwater and
uncontaminated porewater in the SUZ. Periodic repetition of this process results in the accumulation
of contamination within porewater of the SUZ. At a later stage of groundwater plume evolution, or if
initially high concentrations are present in porewater (e.g. from leaching of contamination from above),
diffusive exchange between contaminated porewater and less contaminated groundwater would occur
seasonally, proving a seasonal source of bromate to groundwater.
5.11 Source terms for bromide and bromate release from the
source site
In order to encompass the range of possible scenarios for bromate release to groundwater beneath the
source site, three source term scenarios are developed within this chapter. They are summmarised below:
• Scenario A - ‘Catastrophic Release’ This scenario envisages a sudden, large, release of bromide
and bromate to the unsaturated zone at some point in the operational history of the factory.
• Scenario B - ‘Steady Seepage’ This scenario envisages a steady discharge of bromide and bro-
mate to the unsaturated zone during the operational lifetime of the factory.
• Scenario C - ‘Direct Release’ This scenario envisages a direct release of bromate to the saturated
zone (by-passing the unsaturated zone) for a period of time over the operational history of the
factory.
The three conceptual scenarios for bromate input to, and release from, the ‘source zone’ are quan-
tified in the following sections, using estimates of bromide and bromate mass in the unsaturated and
saturated zone of the source site (Section 5.7.2) to constrain the range of possible source histories.
5.11.1 General equations

Scenarios A and B are described by the following equations, which are derived in Figure 5.35. For all
scenarios, time t = 0 is taken as 1955, the start of the operation lifetime of the factory.
The distribution coefficient, Kd , relates the amount of solute sorbed to sediment (measured as
the soil concentration, CS , in mg kg−1 ) to the amount of solute in solution (measured as the leachate
concentration, CL , in mg l−1 ):
CS
Kd = (5.3)
CL
At time t, the total mass in the unsaturated zone, M (t), is represented by the equation:
S−R
M (t) = M0 exp−Kt + (1 − exp−Kt ) (5.4)
K
Where S is the rate of ‘seepage’ of mass to the unsaturated zone, R is the rate of mass removal from
the unsaturated zone (by excavation, remediation etc.), both of which are assumed to be constant over
defined time periods, and K is a constant defined as:
1000r
K= (5.5)
DρKd
where r is the infiltration rate or recharge rate, D is the thickness of the unsaturated zone, ρ is the bulk
density of soil within the unsaturated zone, and Kd is the distribution coefficient.
The rate of leaching of contaminant from the unsaturated zone, L(t), is represented as:
L(t) = KM (t) = KM0 exp−Kt +(S − R)(1 − exp−Kt ) (5.6)
The total mass of contaminant leached from the unsaturated zone between time T1 and T2 is then
given by:
Z T2
ML (t) = L(t)dt (5.7)
T1
The leached concentration, CL , in mg l−1 is then given by
L(t)
CL (t) = (5.8)
1000rAθm
where θm is the mobile porosity (the effective porosity) and A is the site area open to infiltration.
5.11.2 Constraints
The source term in bounded by the following constraints:
• Condition 1 – the mass in the unsaturated zone at the end of 1983 should correspond to the
observed mass estimate for the unsaturated zone in 1984/1985;
• Condition 2 – the total mass input up until 1983 should be at least the observed mass estimate for
the saturated zone in 1984/1985;
• Condition 3 – the mass in the unsaturated zone in 2000/2001 should correspond to the observed
mass estimate for the unsaturated zone in 2000/2001;
• Condition 4 – the total mass input up until 2008 should be at least the amount of mass estimated
to have been removed by Hatfield and other abstractions between 1981 and 2008.
Figure 5.35: Derivation of equations for mass of bromide/bromate in the unsaturated zone and the rate
of input of bromide/bromate from the unsaturated zone to the saturated zone.
5.11.3 Bromide mass between 1985 and 2001

Conditions 1 and 3 are used to constrain the general equations for bromide between 1985 and 2001,
where estimates for the total mass of bromide in the saturated zone (Section 5.7.2) are available based
on results of site investigation in 1984/1985 and 2000/2001. For the purposes of calculations, the mass
estimates are taken to be at the beginning of 1985 and the beginning of 2001 (t = 30 years and t =
46 years respectively).
Scenario A (Section 5.11.4) and Scenario B (Section 5.11.5), the recharge rate between 1984 and
1987 is assumed to be 500 mm year−1 corresponding to the ‘recharge pulse’ period when the site cover
was removed and the site left to drain freely. From 1987 onwards, when the site is covered with housing
and hardstanding, the recharge rate is assumed to be the lower rate of 290 mm year−1 . In 1986, a
quantity of bromate (for which there is an estimate in Section 5.7.1) was removed from the site by
excavation of material from the unsaturated zone.
The leachate test results (Section 5.6.5) suggest a Kd for bromide of approximately 12 l kg−1 .
However, this value suggests too much partitioning for reasonable values of infiltration rate, and equa-
tion 5.4 cannot be fit to the observed mass estimates in 1985 and 2001 (Figure 5.36). In order to fit
the observed mass results, a value of Kd of 0.20 l kg−1 is required (Figure 5.36). Therefore a Kd of
0.20 l kg−1 is used subsequently for calculating the source term for bromide.
For Scenario A (Section 5.11.4) and Scenario B (Section 5.11.5) assume that the pattern of bro-
mate release from the source zone corresponds to the pattern of bromide release. The bromate mass is
estimated from the bromide source term by the relationship observed in 2000 and 2001 (Figure 5.22):
Bromate conc. (mg kg−1 ) = 0.36 × Bromide conc. (mg kg−1 ) + 32.1 (mg kg−1 )
In order to fit the observed bromate mass results, a value of Kd of 0.23 l kg−1 is required (Fig-
ure 5.36). Therefore a Kd of 0.23 l kg−1 is used subsequently for calculating the source term for
bromate.
The source terms for Scenario A and Scenario B from 1984 onwards are illustrated in Figure 5.37
as bromide and bromate concentrations released to the saturated zone from 1984 onwards.
5.11.4 Scenario A: Catastrophic Leak + Recharge Pulse

The ‘catastrophic release’ scenario is represented by postulating a rapid, large release of bro-
mide/bromate approximately midway through the factory operational lifetime (1965). No additional
mass is released to the unsaturated zone. The pattern of bromate release is assumed to mirror the pattern
of bromide release from the source zone. The mass of bromide/bromate in the unsaturated zone is
subject to leaching by infiltrating water recharging the saturated zone. Recharge rates are assumed to be
290 mm year−1 , corresponding to the estimated value with a cover of buildings and hardstanding. From
1984, the scenario proceeds as described in Section 5.11.3 and Figure 5.37.
Equation 5.4 is used to calculate the mass of bromide in 1965 of 11,140,000 kg (Figure 5.38), and
the mass of bromate in 1965 of 221,600 kg (Figure 5.39). The ‘catastrophic’ release rate is therefore
Bromide - Unsat. Zone
40000
Equation of line:
M1984 = 34246 kg 1984 < t ≤ 1985

S = 0, R = 0 kg y-1
r = 500 mm y-1
30000
Kd = 0.20 litres kg-1
1985 < t ≤ 1986

M1985 = 24907 kg S = 0, R = 2800 kg y-1
r = 500 mm y-1
Equation fit to observed mass 20000
in 1985 and 2001 by adjusting Kd (Chart A)
1986 < t ≤ 1987
Bromide (kg)
M1986 = 18115 kg
Using the values for parameters r, S, R and M S = 0, R = 0 kg y-1
indicated on chart B r = 500 mm y-1
NOTE: linear scale

1987 < t ≤ 2001
S = 0, R = 0 kg y-1
10000 M1987 = 10777 kg r = 290 mm y-1
Max Br unsat
Mean Br unsat
Min Br unsat
Bromide - Unsat. Zone
0
59874
1981 1982 1983 1984 1986 1987 1988 1989
equations extrapolated for t < 1985 1980 1985 1990
M1985 = 24907 kg
22026
Kd = 12 litres kg-1 With Kd = 12 litres kg-1
too much partitioning occurs
8103 and equations cannot be fit to
observed mass of bromide
Equation fit to observed mass Kd = 0.20 litres kg-1
2981 in 1985 and 2001by adjusting Kd
Kd = 0.20 litres kg-1 allows equations to be fit to
observed mass of bromide
Bromide (kg)
1987 < t ≤ 2001
1097 S = 0, R = 0 kg y-1 M2001 = 812 kg
r = 290 mm y-1
NOTE: natural logartihmic scale

403
5.11. Source terms for bromide and bromate release from the source site
equations extrapolated as source term for t > 2001
148
1980 1985 1990 1995 2000 2005 2010 2015 2020

220
Figure 5.36: Equations for bromide mass, fit to observed values from 1985 and 2001. Parameters are defined in Figure 5.11.1
2
1x10
Scenario A and Scenario B 1984+ Bromide conc.
Bromate conc.
60
1
1x10
0
1x10
50
-1
1x10
40 -2
1x10
Concentration (mg litre-1)

-3
1x10
30
1x10-4
1980 1990 2000 2010 2020 2030 2040 2050 2060
Concentration (mg litre-1)

20
10
5.11. Source terms for bromide and bromate release from the source site
0
1980 1985 1990 1995 2000 2005 2010 2015 2020
221
Figure 5.37: Bromide and bromate concentrations for Scenario A and Scenario B from 1984 into the future.
taken as 11,140,000 kg of bromide and 221,600 kg of bromate input over the year of 1965. Table 5.5
summarises how the source term relates to the four conditions in Section 5.11.2. Scenario A is illustrated
as a source term in Figure 5.40.
5.11.5 Scenario B: Steady Seepage + Recharge Pulse

The scenario is represented by contaminant input (at a low rate) prior to the recharge pulse in 1984,
and then a rapid increase in input rate coinciding with recharge pulse. The pattern of bromate release is
assumed to mirror the pattern of bromide release from the source zone.
The ‘steady seepage + recharge pulse’ scenario is represented by postulating a constant rate of
contaminant release to the unsaturated zone during the operational lifetime of the factory. The mass
of bromide/bromate in the unsaturated zone is subject to leaching by infiltrating water recharging the
saturated zone. Recharge rates are assumed to be 290 mm year−1 , corresponding to the estimated
value with a cover of buildings and hardstanding. From 1984, the scenario proceeds as described in
Section 5.11.3 and Figure 5.37.
Equation 5.4 is used to calculate the rate of seepage of bromide between 1955 and 1984 as
6355 kg y−1 of bromide (Figure 5.38), 1837 kg y−1 of bromate (Figure 5.39). Table 5.5 summarises
how the source term relates to the four conditions in Section 5.11.2. Scenario B is illustrated as a source
term in Figure 5.40.
Table 5.5: Mass predicted by source history scenarios A and B compared to observed mass constraints.
Condition 4 is based on an estimate by Buckle (2002) of the mass removed at Hatfield and Essendon
between 1981 and 2000.
BROMIDE Observed Scenario A Scenario B

% of % of
kg kg observed kg observed
Mass in unsat. Zone 1985 =
Condition 1 2.49E+04 2.49E+04 100% 2.49E+04 100%
observed mass in unsat. Zone 1985
Mass leached between 1955 and 1984 >

Condition 2 1.79E+04 1.11E+06 6189% 1.50E+05 836%
observed mass in sat. Zone 1985

Condition 3 8.12E+02 8.12E+02 100% 8.12E+02 100%
Total mass leached up to 2008 >

Condition 4 2.23E+04 1.14E+06 5125% 1.84E+05 824%
mass removed by abstractions 1981-2008
BROMATE Observed* Scenario A Scenario B

% of % of
kg kg observed kg observed
Condition 1 9.00E+03 9.00E+03 100% 9.00E+03 100%
Mass leached between 1955 and 1984 >

Condition 2 6.49E+03 1.27E+05 1956% 3.28E+04 506%
observed mass in sat. Zone 1985

Condition 3 5.08E+02 5.08E+02 100% 5.08E+02 100%
Total mass leached up to 2008 >

Condition 4 8.84E+03 1.38E+05 2127% 4.39E+04 497%
mass removed by abstractions 1981-2008
* Bromate concentrations prior to 2000 are estimated

based on bromide concentrations.
Bromide - SCENARIO A
10000000 10000000
1000000 1000000
100000 100000
Rate of leaching (kg y-1)

Bromide (kg)
10000 10000
1000 1000
100 100
10 10
1 1
1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020
Bromide - SCENARIO B
40000 4000
30000 3000
Rate of leaching (kg y-1)

Bromide (kg)
20000 2000
10000 1000
0 0
1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020
Max Br unsat
Mean Br unsat
Min Br unsat
Rate of mass leaching from unsaturated zone (kg yr-1)
Mass in the unsaturated zone (kg)
Figure 5.38: Bromide source history for Scenarios A and B.

Bromate - SCENARIO A
1000000 1000000
100000 100000
Rate of Leaching (kg y-1)

10000 10000
Bromate (kg)
1000 1000
100 100
10 10
1 1
1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020
Bromate - SCENARIO B
16000 16000
12000 12000
Rate of Leaching (kg y-1)

Bromate (kg)
8000 8000
4000 4000
0 0
1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020
Max Br unsat
Mean Br unsat
Min Br unsat
Rate of mass leaching from unsaturated zone
Mass in unsaturated zone
Figure 5.39: Bromide source history for Scenarios A and B.

Bromide concentration
2
2.0x10
1.9x102
2
1.8x10
1.7x10
2 Scenario A
1.6x10 2 Scenario B
1.5x102
1.4x102
Bromide concentration (mg litre-1)
2
1.3x10
2
1.2x10
2
1.1x10
1.0x102
9.0x101
8.0x101
7.0x101
1
6.0x10
5.0x101
4.0x101
3.0x101
2.0x101
1.0x101
0.0x100
1955 1960 1965 1970 1975 1980 1985
Bromate concentration
3.5x101
1
3.0x10
2.5x101
Bromate concentration (mg litre-1)
2.0x101
1.5x101
1
1.0x10
5.0x100
0
0.0x10
1955 1960 1965 1970 1975 1980 1985
Figure 5.40: Bromide and bromate concentrations for Scenario A and Scenario B between 1955 and
1984. After 1984 concentrations proceed as in Figure 5.37.
5.12. Verifying source terms with observed down-gradient concentrations 226
5.11.6 Scenario C - Late stage Seepage + Recharge Pulse

Scenario C assumes that the pattern of bromate release differs from the pattern of bromide release from
the source zone. However, due to the absence of observed bromate concentrations in soil or groundwater
prior to 2001, it is difficult to constrain the source term; conditions 1 and 2 cannot be applied.
Scenario C assumes that the results from 1983 investigation that indicated bromate concentrations
below detection limits, were representative of bromate concentrations in the shallow unsaturated zone
of the site. Therefore, bromate input must have occurred directly to the saturated zone, by-passing the
shallow soils, to result in the elevated bromate concentrations observed from 2000 in the saturated zone
porewater and groundwater.
Scenario C is represented by postulating a constant flux of bromate to the saturated zone for a period
of the operational lifetime of the factory. The mass flux, and the time period over which this occurs, are
constrained by the observed porewater and groundwater concentrations at the source zone from 2000 to
2008. Bromate concentrations are simulated from the input at the top of the saturated putty chalk to the
base of the putty Chalk, at which point they are assumed to represent the groundwater and porewater
concentrations measured at the monitoring locations at the source zone (Locations 219, 221 and 222.
The simulation set-up is described in Section 5.12. The source term is varied until a reasonable fit to the
observed concentrations is achieved (Figure 5.41). The selected source term is a steady concentration of
3000 mg l−1 of bromate between 1955 and 1960.
5.11.7 Bromate flux in 2001

Using equation 5.6, and the observed mass of bromide in the unsaturated zone of 508 kg in 2001, the
bromate flux from the unsaturated zone to the saturated zone is calculated as 78 kg y−1 in 2001 for
Scenario A and Scenario B. This value is at least an order of magnitude lower than the estimated mass
flux based on interpolations of observed groundwater concentrations (Section 5.8). This suggests that
there is a source of bromate mass, additional to the bromate mass leaching from the unsaturated zone,
that is contributing to the mass transported off-site in groundwater. This could indicate that there were
inputs of bromate direct to the saturated zone (by-passing the unsaturated zone) that are not represented
by source history scenarios A and B. Source scenario C attempts to capture this possibility. Using
simulated concentrations at the base of the unsaturated zone for Scenario C in 2001, the bromate mass
flux in groundwater is approximately 3 kg d−1 to 13 kg d−1 (1095 kg y−1 to 4745 kg y−1 ), which is of
a similar magnitude to the estimates based on observed concentrations in 2001 to 2003 (Section 5.8).
5.12 Verifying source terms with observed down-gradient concen-

trations
The DP1-D (Dual Porosity in 1-Dimension) modelling code is a semi-analytical solute transport code de-
veloped by John Barker, based on Barker (1982). The code simulates one-dimensional transport through
a double-porosity medium with a specified input concentration and a constant flow regime. The outputs
are the solute concentration in the fractures and the average matrix porewater solute concentration for
a specified set of times and distance from the input point. The code is written in Fortran and has been
SOURCE TERM 500

Sim. Groundwater
input to the saturated Putty Chalk
3000 Sim. Porewater
Obs. Groundwater

400 Obs. Porewater
2000 300
200
1000
100
0
1950 1970 1990 2010 1950 1970 1990 2010 2030 2050
500
3000
400
2000 300
200
1000
100
0
1950 1970 1990 2010 1950 1970 1990 2010 2030 2050
500
3000
400
2000 300
200
1000
100
0
1950 1970 1990 2010
1950 1970 1990 2010 2030 2050
Figure 5.41: Bromate source history for Scenario C.

implanted in Excel. The spreadsheet accesses the code, held in a dynamic link library, via Visual Basic.
In order to evaluate the potential for the DP1-D code was used to simulate bromide and bromate
concentrations down-gradient along a one-dimensional flow line from the source zone. The simulated
results were compared to observed concentrations at three locations with bromide monitoring data avail-
able for 1983-1987 and 2000-2008, and bromate data for 2000-2008 only.
The parameters selected (Figure 5.42; Sections 5.12.1 to 5.12.6) were chosen as best available esti-
mates from investigation data (i.e. the model was not calibrated to produce the observed concentrations.)
The conceptual model for bromate and bromide release to groundwater beneath the site considers
the ‘putty chalk’ to represent a low permeability layer between the base of the unsaturated zone, and
groundwater migrating off-site in the ‘blocky chalk’. Groundwater flow is therefore assumed to be es-
sentially vertical within the putty chalk layer. Groundwater flow is simulated vertically through the putty
chalk layer to the saturated blocky chalk below the site, and then groundwater is simulated horizontally
along a groundwater flowline (Figure 5.42).
5.12.1 Fracture Spacing 2b

There is little available local information for fracture spacing in the region of the source site and 1 km
down-gradient. However, geophysical logging by Three Valleys Water in the Hatfield area indicates
flowing fracture separations of approximately 1.00 m to 1.40 m. This is comparable to the values re-
viewed in Section 2.7.3.3 for unweathered Chalk. Therefore, a fracture spacing (2b) of 1.0 m is used for
the ‘blocky chalk’.
For the weathered ‘putty chalk’, the fracture spacing is considered likely to be less than the un-
weathered ‘blocky chalk’. The values reviewed in Section 2.7.3.3 suggest a value of 0.10 m for fracture
spacing (2b) in the Putty Chalk.
5.12.2 Fracture Aperture a

Based on the review of data from the literature (Section 2.7.3.3), a representative fracture aperture of
10−3 m is used for the ‘blocky chalk’.
This is likely to be significantly less for the Putty Chalk. Atkins (2002) estimated fracture apertures
for the Putty Chalk of 10−6 m based on the hydraulic conductivity, K, values calculated from measured
putty chalk matrix porosities (2.7.3.1)and the relationship:
ρga2
K=
12µ
where ρ is the fluid density, µ is the fluid viscosity, and g is the acceleration due to gravity.
5.12.3 Fracture Porosity θm

The mobile (or fracture) porosity, θm , for slab geometry with fracture aperture a and block thickness 2b,
is given by equation 5.9:
a
θm = (5.9)
2b + a
Using the values for a and b in sections 5.12.2 and 5.12.1, the mobile porosity θm is estimated as 1.0 ×
10−5 for the putty chalk and 1.0 × 10−3 for the blocky chalk.
Figure 5.42: Conceptual model for off-site verification simulations

5.12.4 Matrix Porosity Φ

During the 2001 site investigation (Atkins, 2002), 15 samples of Chalk from the saturated zone were
tested. Results ranged between 0.308 and 0.457, with an average of 0.380. For the blocky chalk, a
representative value for the Chalk in the area is 0.388 (Section 2.6). This value for Φ is used for both the
saturated putty chalk and the saturated blocky chalk.
For a slab geometry, with fracture aperture a block thickness 2b, this matrix porosity, Φ, this equates
to an immobile porosity, θim , of 0.38, from equation 5.10:
2bΦ
θim = (5.10)
2b + a
Therefore, the immobile porosity θim is estimated as 0.38 for the putty chalk and 0.38 for the blocky
chalk.
5.12.5 Fracture Velocity

Groundwater velocity, v, was calculated from the darcy velocities, q, determined from the single bore-
hole dilution testing at Nashe’s Farm and Harefield House (Section 3.6) and the mobile porosity value
estimated above.
The darcy velocities ranged from 0.5 to 3.0 m day−1 , with average of 1.0 m day−1 (Nashe’s Farm),
and 0.3 to 1.3 m day−1 , with average of 0.8 m day−1 (Harefield House).
For the blocky chalk, with an effective porosity of 1.0 × 10−3 , this results in fracture velocities
ranging from 300 to 3000 m day−1 , with an average of 900 m day−1 .
For the putty chalk, hydraulic conductivity is estimated to be around 10−4 to 10−2 m day−1 ,
compared to 100 to 101 m day−1 for the blocky chalk aquifer. Therefore, the hydraulic conductivity for
the putty chalk is approximately 102 to 104 times less than for the ‘blocky Chalk’.
Additionally, the vertical hydraulic conductivity is likely to be around 10 % of the horizontal con-
ductivity. This results in an estimated vertical darcy velocity in the putty chalk of 104 times less than the
hydraulic conductivity in the blocky chalk. Therefore, the vertical darcy velocity in the putty chalk is
estimated as an average of 9.0 × 10−5 m day−1 . With an effective porosity of 1.0 × 10−5 , this results
in fracture velocities ranging from 3 to 30 m day−1 , with an average of 9 m day−1 .
5.12.6 Effective Diffusion Coefficient DE

The effective diffusion coefficient for bromide and bromate is taken as 8.64 × 10−6 m2 day−1 (Sec-
tion 2.9.4.1).
5.12.7 Simulation Results

5.12.7.1 Scenario A
For Scenario A, simulated bromide groundwater concentrations show good agreement with the observed
bromide concentrations at monitoring locations 225, 028 and 019 (Figure 5.43), although the simulated
concentrations for 1983 to 1985 are at the lower limit of the observed concentrations. Observed porewa-
ter bromide concentrations are only available for location 225; simulated porewater concentrations are
at the lower limit of observed porewater concentrations.
Simulated bromate concentrations groundwater show good agreement with the observed bromate
concentrations at monitoring location 225 and 028 (Figure 5.44), but simulated concentrations at loca-
tion 019 are at the lower limit of the the observed concentrations.
5.12.7.2 Scenario B
For Scenrio B, simulated bromide concentrations (Figure 5.45) and bromate concentrations (Figure 5.46)
are generally lower than observed concentrations by approximately one order of magnitude.
5.12.7.3 Scenario C
Simulated bromate concentrations for Scenario C (Figure 5.47) show good agreement with the observed
bromide concentrations at all three of the monitoring locations.
5.12.7.4 Discussion of simulation results

The simulated groundwater and porewater concentrations at locations 225, 028 and 019 for scenario A,
B and C indicate that observed concentrations (1983 to 1985 and 2000 to 2008) are on the rising limb of
bromide and bromate concentrations.
There is some indication in the bromide groundwater monitoring results (Section 5.6.3) that bro-
mide concentrations increased between 1985 and 2001 at locations 225, 028 and 019. However, the large
scatter in the observed monitoring data due to seasonal variations makes trends difficult to discern (See
Chapter 4). It is therefore very difficult to be certain about which point on the simulated concentration
versus time curve the current observations represent, and hence how successful the source terms and the
model are in simulating bromate evolution close to the site.
Saturated zone chalk porewater bromide concentrations are available from location 225 in 2001 and
in 1985 from location B1-[85], which was at a similar location. The large range of both porewater and
groundwater concentrations in both 1985 and 2001 make it difficult to decide on a representative figure
for each, and therefore to determine the relative magnitude of porewater and groundwater concentrations.
A mean porewater concentration and a mean groundwater concentration would indicate higher porewater
bromide concentrations than bromide groundwater concentrations in 2001. This would imply, contrary
to the simulated ‘rising limb’, that concentrations in 2001 were on the falling limb (Figure 5.48). How-
ever, the reverse is indicated in 1985; mean groundwater concentrations are higher than mean porewater
bromide concentrations.
If the results of these two boreholes can be considered representative of the same location, then
porewater bromate concentrations do show an increase between 1985 and 2001. The very large range
of the groundwater concentration in 1985 makes it difficult to determine whether (mobile) groundwater
concentrations have increased or decreased. An increase in bromide porewater concentrations between
1985 and 2001 would imply observed concentrations to be on the rising limb.
5.13 Summary and conclusions

Three source term scenarios have been developed which attempt to represent the range of possible bro-
mide and bromate source histories. Scenario A (‘catastrophic release’) and Scenario B (‘steady seepage’)
assume that bromide and bromate have had similar histories and that the main process contributing to
SOURCE TERM
10000 input to the saturated Putty Chalk
BH2-[84] ; BH3-[84]
1000 Observed
BH 221; BH 219
Observed
Monitoring
Source Term - Scenario A
100 Location
028 Concentrations simulated along
a pathline from the source zone
10 to locations 150 m, 850 m and
1200 m down hydraulic gradient
1 Source

Zone
Monitoring
Location
0 019
1950 1970 1990 2010 2030 2050 Monitoring
Location
225
1000 Location 225 / B1 8 Location 028 - Orchard Garage BH 8 Location 019 - Nashe's Farm BH
Obs. Groundwater
Obs. Porewater
Sim. Groundwater
800 Sim. Porewater
6 6
600
4 4
400
2 2
200

1950 1970 1990 2010 2030 2050 1950 1970 1990 2010 2030 2050 1950 1970 1990 2010 2030 2050
232
Figure 5.43: Comparison of simulated bromide concentrations for source history Scenario A and observed concentrations at three monitoring locations.
SOURCE TERM
BH 221; BH 219
1000 Observed
Source Term - Scenario A
Monitoring
100 Location
1 Source

Zone
Monitoring
Location
0 019
1950 1970 1990 2010 2030 2050 Monitoring
Location
225
6 Location 225 5 Location 028 - Orchard Garage BH 5 Location 019 - Nashe's Farm BH
Sim. Groundwater
Sim. Porewater
Obs. Groundwater
Obs. Porewater 4 4
4
3 3
2 2
2
1 1

1950 1970 1990 2010 2030 2050 1950 1970 1990 2010 2030 2050 1950 1970 1990 2010 2030 2050
233
Figure 5.44: Comparison of simulated bromate concentrations for source history Scenario A and observed concentrations at three monitoring locations.
SOURCE TERM
3 BH2-[84] ; BH3-[84]
10 Observed
BH 221; BH 219
2 Observed
10 Monitoring
Source Term - Scenario B
Location
1
10 a pathline from the source zone
to locations 150 m, 850 m and
100
Source
-1

10 Zone
Monitoring
-2
Location
10 019
1950 1970 1990 2010 2030 2050 Monitoring
Location
225
1000 Location 225 / B1 10 Location 028 - Orchard Garage BH 10 Location 019 - Nashe's Farm BH
Obs. Groundwater
Obs. Porewater
Sim. Groundwater
800 Sim. Porewater 8 8
600 6 6
400 4 4
200 2 2

1950 1970 1990 2010 2030 2050 1950 1970 1990 2010 2030 2050 1950 1970 1990 2010 2030 2050
234
Figure 5.45: Comparison of simulated bromide concentrations for source history Scenario B and observed concentrations at three monitoring locations.
SOURCE TERM
BH 221; BH 219
102 Observed
Source Term - Scenario B
Monitoring
101 Location
-1 Source
10

Zone
Monitoring
-2
Location
10 019
1950 1970 1990 2010 2030 2050 Monitoring
Location
225
Sim. Groundwater
Sim. Porewater
Obs. Groundwater
Obs. Porewater 4 4
4
3 3
2 2
2
1 1

1950 1970 1990 2010 2030 2050 1950 1970 1990 2010 2030 2050 1950 1970 1990 2010 2030 2050
235
Figure 5.46: Comparison of simulated bromate concentrations for source history Scenario B and observed concentrations at three monitoring locations.
SOURCE TERM
Source Term - Scenario C
2000 Monitoring
Location
to locations 150 m, 850 m and
1000
Source

Zone
Monitoring
Location
0 019
1950 1970 1990 2010 2030 2050 Monitoring
Location
225
Sim. Groundwater
Sim. Porewater
Obs. Groundwater
8 Obs. Porewater 4 4
6 3 3
4 2 2
2 1 1

1950 1970 1990 2010 2030 2050 1950 1970 1990 2010 2030 2050 1950 1970 1990 2010 2030 2050
236
Figure 5.47: Comparison of simulated bromate concentrations for source history Scenario C and observed concentrations at three monitoring locations.
Figure 5.48: Concurrent matrix and fissure concentrations are required to determine at which point along
the concentration-time graph a particular fissure concentration represents.
contaminant input to the saturated zone beneath the site is the leaching of bromide and bromate mass
from the unsaturated zone. The scenarios are constrained by estimates of the observed mass of bromide
and bromate in 1985 and 2001. Scenario C (‘direct release’) assumes that bromate is released direct to the
saturated zone, by-passing the unsaturated zone. Scenario C is constrained by the observed groundwater
concentrations at the source site in 2001.
The one-dimensional double-porosity transport code, DP1-D, has been used to simulate concentra-
tions in groundwater down-gradient of the source site. Simulated concentrations using the source term
scenarios A and C show relatively good agreement with observed groundwater concentrations at loca-
tions 150 m, 500 m and 1000 m down-gradient of the source site. However, the relatively short period of
time for which monitoring data are available, combined with the large seasonal variations in concentra-
tions, means that the trends are difficult to discern, and robust conclusions cannot be made as to whether
or not the simulations are representative of the observed data.
238
Chapter 6
Multiple Analytical Pathways Approach
6.1 Chapter Objectives

The objectives of this chapter are:
• To review the previous modelling approaches that have been implemented for bromate and/or
bromide contamination in Hertfordshire;
• To develop an analytical network modelling approach to allow representation of Fickian double-

porosity diffusion and to integrate karstic transport pathways within the network; and
• To use this model to provide predictions for the likely long-term evolution of bromate concentra-
tions at key output locations.
6.2 Previous modelling approaches for Bromide and Bromate in

the Chalk
6.2.1 Early model assessments
A series of modelling studies for the migration of inorganic bromide were undertaken in 1984 and 1985
prior to the redevelopment of the St Leonard’s Court site (Chemfix, 1984, 1985b,d). The studies resulted
in a significant underestimate of the extent of migration.
The modelling exercises were based on the two-dimensional advection-dispersion equation with
point-source contamination estimated from limited field data. Two conditions were considered: firstly
that the site was immediately redeveloped with predominantly impermeable hardcover and secondly that
the site was left fallow and open to infiltration. The reports concluded that, if the site was immedi-
ately redeveloped with predominantly impermeable cover, the bromide plume of concentrations above
100 µg l−1 would be limited to a plume some 250 m × 100 m in size in a direction 11 degrees south of
east, and was unlikely to affect abstraction boreholes. If the site remained fallow, there was some risk
of contamination of boreholes within the Sandridge area. The modelling was criticised by the Environ-
ment Agency (Thomas, 2000) for neglecting the influence of higher transmissivity along the dry valley
in Sandridge, which would have indicated a more southerly flow direction, and for not considering the
impacts of fissure flow within the Chalk. There were also some issues with the conceptualisation of the
6.2. Previous modelling approaches for Bromide and Bromate in the Chalk 239
source term which have been discussed in Section 5.9, and indicate that the mass input to the aquifer is
likely to have been considerably underestimated by the Chemfix modelling.
6.2.2 Pollutant Linkage Assessment using CONSIM

Atkins (2002) used the CONSIM modelling package to demonstrate that the bromate concentrations
measured in the unsaturated zone soils during the site investigation at the St Leonard’s Court site were,
through leaching, providing an on-going source of bromate pollution above the drinking water standards
to the saturated Chalk aquifer beneath the site. Their conceptualisation of the source has been discussed
in Chapter 5.
6.2.3 One-dimensional analytical model DP1D

Atkins (2004) used the DP1-D (Dual Porosity in 1-Dimension) code developed by John Barker (e.g.
Barker (2005)) to simulate bromate concentrations along three flow lines:
• Orchard Garage (Location 028) to Hatfield P.S. (Location 001)
• Orchard Garage (Location 028) to Essendon P.S. (Location 143)
• Hatfield P.S. (Location 001) to Hoddesdon P.S. (Location 300)
The model parameters were derived from literature data, and site-specific data where available,
and are reasonably consistent with the likely parameters for the Chalk aquifer in the region reviewed in
Chapter 2. A constant source term was assumed from 1970 to 2000 of 1000 µg l−1 at Orchard Garage
and 100 µg l−1 at Hatfield. Between 2000 and 2004, monitoring data at Orchard Garage and Hatfield
respectively were used to represent the source terms. From 2004 into the future constant source terms of
1000 µg l−1 at Orchard Garage and 244 µg l−1 at Hatfield were assumed.
The simulated bromate concentrations at Hatfield P.S. and Essendon P.S. corresponded well to the
‘average’ observed bromate concentration trends. Simulated concentrations at Hoddesdon corresponded
reasonably well to the ‘average’ observed bromate concentration trends. The simulations did not cap-
ture the seasonal variation evident in observed bromate concentrations due to the non-seasonality of the
model, the relatively constant source-term, and the smoothing effect of the double-porosity diffusive ex-
change. All scenarios indicated increasing concentrations over time. This rising trend was interpreted by
Atkins (2004) as suggesting that the ‘plume’ is not in steady state, i.e. fracture and matrix concentrations
had not yet reached equilibrium at any of these locations.
6.2.4 Dispersion modelling

Since the DP1-D modelling does not account for mechanical dispersion, Atkins (2004) used a steady-
state two-dimensional solution to the advection-dispersion equation derived by Bear (1972). A constant
source injection was assumed. The dispersion modelling assumed a direct travel path from the north
Hatfield area (e.g. Park Street, Location 265) to the NNR source at Hoddesdon, and calculated concen-
trations at NNR locations which would be expected assuming dispersion from a the direct travel path.
The calculated concentrations were compared to the observed concentrations at these locations. Atkins
(2004) concluded that concentrations at Amwell Hill, Amwell Marsh, Broxbourne and Turnford were
significantly higher than would be expected purely from the effects of dispersion, with the results for
Turnford and Amwell Hill in particular suggesting that a further mechanism of contaminant transport,
such as separate direct flow pathways leading to these wells, was important.
6.2.5 Catchment-scale distributed flow modelling using MODFLOW and MT3D

A series of modelling studies have been undertaken using the MODFLOW suite of codes (MacDon-
ald and Harbaugh, 1984; Harbaugh et al., 2000) coupled with the mass transport in three dimensions
(MT3D) code (Zheng, 1990; Zheng and Wang, 1999). The first major modelling study was initiated as
a joint project between the Environment Agency and Vivendi Water Partnership (now TVW). This sin-
gle layer model model, known as the Bromate Groundwater Flow Model (BGFM), described in Buckle
(2002, 2003), was developed as a subset of the Upper Lee and Mimram Model (ULMM) (Entec, 2002),
a regional scale groundwater resource model for the Chiltern Hills. The BGFM was unable to reproduce
observed groundwater flows within the Vale of St Albans immediately down-gradient of Sandridge, and
this resulted in inadequate representation of the ‘plume’ geometry within the Vale of St Albans lead-
ing to modelled breakthrough of bromate at Roestock P.S. (where it is not currently observed) and a
significant underestimate of concentrations at Essendon P.S. These deficiencies could not be improved
by additional calibration and the use of the model was subsequently suspended pending a review of the
conceptual understanding Buckle (2003).
The second major MODFLOW/MT3D catchment-scale flow and transport model, the Northern
New River (NNR) model (Atkins, 2005), was developed by Atkins for Thames Water (TWUL) with the
aim to model bromate transport to the Lee Valley. The NNR model was developed by extending the
ULMM of Entec (2002) to the south and west to include the Lee Valley. Calibration of the flow model
was improved relative to the BGFM and the relative spatial distributions of groundwater levels better
reflected that of observed data, although a number of deficiencies remained. In particular, calibration
of water levels close to source was poor, and as a result the simulated hydraulic gradient was steeper
than observed. Also, water levels at North Mymms were simulated to be higher, and the mound to
extend over a larger area, than observed. Bromate transport was modelled using MT3D-MS (Zheng,
1990; Zheng and Wang, 1999), a multiple species version of MT3D which is can be used to approxi-
mate double-porosity exchange though a first order exchange coefficient (but see the discussion of its
deficiencies in Section 6.2.6). Using a constant source term of 5000 µg l−1 , the shape and magnitude of
the apparent bromate ‘plume’ was well replicated by the model between Sandridge and Essendon P.S.,
but the model significantly underestimates the further migration of bromate to the NNR well field. Cal-
ibration attempts by varying the mass transfer coefficient, dispersivity and effective porosity and source
concentrations were able to increase migration toward the NNR well field, however this was at the ex-
pense of either the model stability or of the modelled plume geometry resulting in a much wider plume
than observed.
Therefore, the modelling exercises described above were not be able to duplicate the migration of
bromate to the NNR wells via the Hertfordshire karst system. Cook (2010) notes that although the role
of the karst in the transport of bromate was acknowledged in the conceptual understanding underpinning
the models (Buckle, 2002), the representation of the karst system within the model is limited, probably
due to deficiencies in the conceptual model and inadequate data to properly parameterise the system at
that time. Cook (2010) used his new conceptualisation of the Hertfordshire karst flow system, along
with hydrodynamic parameters determined by catchment-scale tracer testing, and the single borehole
dilution tests described in this thesis (Section 3.6.1), to develop a suite of new models incorporating
representations of the karst flow system.
Initially, Cook (2010) developed a steady-state subset MODFLOW model of the NNR model to
allow faster execution times, simpler initial calibration to long-term heads and flows and improved model
stability. The 200 m by 200 m grid of the NNR model was retained. The model uses an EPM approach,
and incorporates a karst zone along palaeogene boundary which is represented as a zone of high hydraulic
conductivities, between one and a few cells wide. The karst zone also extends westwards into the Vale of
St. Albans. The model was used to simulate flowlines (using MODPATH) for the 2008 bacteriophage
tracer tests, and parameters calibrated to achieve good representation of all three tracer breakthroughs
based on the advective transport routes indicated by MODPATH flowlines.
Cook (2010) then converted the calibrated steady-state flow and transport model to a transient flow
and transport model. The transient flow model was found, for the majority of locations, to replicate the
magnitude of head variation and also the seasonal behaviour and trends relatively well. The transport
model (using MT3D-MS) was run for three source scenarios (see discussion of source terms in Sec-
tion 6.2.5.1), as well as constant concentration source of 5000 µg l−1 for comparison with the NNR
model of Atkins (2005). Cook (2010) found that the simulations showed closer agreement with travel
time of the recharge pulse mass input and observed data at receptors if the westward extension of the
karst EPM into the Vale of St Albans was removed. With karst in the Vale of St. Albans, weakly pos-
tulated on the basis of the tracer tests, the timing of simulated peak concentrations precedes that of all
observations. However, with the karst zone removed, simulated concentrations are lower, and travel
times slower, due to additional dispersion and double-porosity attenuation.
Simulations using the constant concentration source of 5000 µg l−1 were able to closely replicate
observed bromate concentrations to Hatfield in magnitude and spatial distribution. Simulations using
the previous estimates (now superseded) of the three source term scenarios (Section 6.2.5.1) resulted in
concentrations that were universally lower than observed. Cook (2010) concluded that these source terms
were too low and increased them by a factor of three. The revised source term Scenario B was found
to result in relatively good spatial and temporal representation of bromate concentrations compared to
observed data. The situation as modelled by the revised source term of Cook (2010) suggests that current
levels of bromate in the aquifer are the result of the passage of the 1983-1987 high concentration recharge
pulse from the source zone.
The model produces a relatively stable ‘plume’ in the Vale of St Albans area after approximately 10
years of bromate input. Breakthrough to the Lee Valley occurs shortly afterwards due to the rapid trans-
port within the karst system, and is strongly seasonal. The model predicts that bromate concentrations
in excess of the drinking water standards persist within the Vale of St Albans for the modelling period
(i.e. at least up until 2050). For locations in the Lea Valley, dilution in the karst system acts to reduce
concentrations significantly to less than drinking water standards once the high mass flux associated with
the 1983-1987 recharge pulse has declined. However periodic seasonal pulses of a around 1 µg l−1 oc-
cur at locations in the Lee Valley for the remainder of the modelling period. These pulses are strongly
influenced by karstic seasonal dilution.
Cook (2010) lists the main areas of uncertainty that limit the effectiveness and confidence in the
MODFLOW/MT3DMS predictive modelling using the currently available data:
• The description of the source term and its implementation in the model;
• A limited period of observation data in relation to the likely duration of the bromate contamination
and with respect to the timing of the breakthroughs;
• Uncertainty with respect to the extent to which the chalk matrix has become contaminated and
detailed parameterisation of that process;
• Deficiencies in the model representation of both dual porosity and karstic transport and the trans-
fers between them.
6.2.5.1 Source Term comparison

The source terms used by Cook (2010) are superseded versions of the source terms developed in Chap-
ter 5. Figure 6.1 illustrates how these superseded source terms relate to the current versions in this thesis.
Between 1955 and 2008 superseded Scenario A represents a cumulative mass input to the saturated zone
beneath the source site of 22900 kg of bromate and superseded Scenario B a cumulative mass input
of 21000 kg of bromate compared to cumulative mass inputs of 138000 kg and 43900 kg for the cur-
rent Scenario A and Scenario B respectively. For comparison, a constant concentration source term of
5000 µg l−1 between 1970 and 2008 represents a cumulative mass input to the saturated zone beneath
the source site of approximately 139000 kg of bromate.
6.2.6 Weaknesses of MODFLOW and MT3D

The MODFLOW suite of codes (Harbaugh et al., 2000) were originally developed to model Darcian
flows in porous granular aquifers such as the extensive glacial sand and gravel aquifers of the United
States. In order to model fractures or karstified rocks, an EPM approach must be adopted.
The solute transport code MT3D-MS (Mass Transport in 3-Dimensions Multi-Species) (Zheng and
Wang, 1999) uses the groundwater flow files from MODFLOW as the basis for solute transport calcu-
lations. The multi-species version of MT3D has capability for dual-domain mass transfer to represent
double-porosity diffusive exchange. The code allows solute transfer between mobile and immobile do-
mains controlled by a first order mass transfer coefficient. The governing equations (without explicit
consideration of sorption) can be expressed as a statement of mass conservation for the mobile domain
∂Cm ∂Cm ∂ 2 Cm
θm = −q + θm Dim − ζ (Cm − Cim ) (6.1)
∂t ∂x ∂x2
Bromate - SCENARIO A
1000000
100000
10000
Bromate (kg)
1000
100
10
1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020
Bromate - SCENARIO B
16000
12000
Bromate (kg)
8000
4000
1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020
Max Br unsat
Mean Br unsat
Min Br unsat
Rate of mass leaching from unsaturated zone
Mass in unsaturated zone
Superceeded-input rate
Superceeded-mass
1984-1985
Superceeded-input ratex3
Figure 6.1: Comparison of the superseded versions of the the source terms used by Cook (2010) to the
current versions in this thesis.
and as a statement mass conservation for the immobile domain
∂Cim
θim = ζ (Cm − Cim ) (6.2)
∂t
Where Cm and Cim are the concentrations in the mobile and immobile domains respectively, θm
and θim are the porosity of the mobile and immobile domain respectively, q is the darcy flux, and ζ is
the first-order mass transfer coefficient.
The first order mass transfer coefficient approach is only an approximation to the diffusion pro-
cess, which is more accurately represented by a Fickian approach (as used in DP1D and MAP codes).
The predictions made by the mass transfer approach and Fickian diffusion converge as a steady-state
condition between fracture and matrix porewater solute concentration is attained (Barker, 1985b). At
early times the MT3D-MS approach will over-estimate solute concentration in fracture water by un-
derestimating the diffusive flux into the immobile matrix porewater domain. The rate of transfer from
the mobile to immobile domain is too slow as it does not account for the infinitely steep concentration
gradient that initially exists at the mobile/immobile interface. The mass transfer approach will therefore
predict a solute ‘plume’ to have moved further than that modelled using the Fickian exchange approach.
When the concentration gradient is reversed and the the movement of solute is from the immobile to
the mobile domain, the reverse will be the case such that the rate of transfer of solute back into the mobile
groundwater in the fracture system predicted by a first order mass transfer coefficient will be slower
than predicted by a Fickian approach. Consequently the mass transfer approach will underestimate the
fracture water solute concentrations.
Compounding this deficiency of a mass transfer approach, the MT3D-MS code of Zheng and Wang
(1999) does not allow back-diffusion from the immobile matrix domain to the mobile fissure water do-
main. Therefore the persistence of contamination will be considerably underestimated as the secondary
source of contaminant is effectively ‘lost’ into a matrix sink.
An indication of whether the system is likely to approach a steady-state within the time frame of
interest can be given by calculating the time for diffusion across a matrix block (Barker, 1993). If the
system is likely to achieve a steady state within the time frame for which predictions are required then
it is reasonable to model it using a mass transfer coefficient approach. However, if accurate predictions
are required for early or late times, a Fickian approach is recommended. Based on typical parameters for
the chalk west in the Vale of St Albans, tcb = 2.89 × 104 days ≈ 79 years (Appendix F). For the karst
system east of Hatfield, tcb ranges from 9 × 104 to 4 × 105 days, or approximately 250 to 1100 years
(Appendix F). Therefore, the mass transfer approach is not likely to be valid for bromate in Hertfordshire.
Furthermore, as noted by Cook (2010), an equivalent porous media model cannot represent simul-
taneously the conduit and fracture systems: model cells must either be an EPM of the karst conduits or
an EPM of the non-karstic fissured aquifer system. The best replication of Vale of St Albans transport in
MODFLOW and MT3D-MS is achieved using a non-karst double-porosity approximation. Conversely,
the majority of transport east of Hatfield is represented by an EPM representing karst transport. The role
of double-porosity diffusive exchange within the fracture system east of Hatfield is not simulated.
6.3. Development of a Multiple Analytical Pathways Approach 245
6.3 Development of a Multiple Analytical Pathways Approach

6.3.1 DP1-D
The DP1-D solute transport code has been introduced in Section 5.12. The DP1D model has been used
by Watson (2004) to model chloride contamination in the Chalk of the Tilmanstone Valley in Kent. The
simulations successfully reproduced the observed porewater and fracture water profiles at two times,
separated by 25 years. Forward modelling indicated that the double porosity diffusion extends the du-
ration of contamination in the catchment by several decades. The concentrations in the mobile zone
simulated by the DP1-D model were compared to concentrations modelled by a three-dimensional nu-
merical model using MODFLOW and MT3DMS and compared well, except for early and late times in
the contamination evolution. Karst conduit flow was not however considered to be a significant feature
at Tilmanstone.
6.3.2 Multiple Analytical Pathways (MAP) model

Barker (2001) developed the ‘Multiple Analytical Pathways’ (MAP) model to simulate long-term nitrate
concentrations in the saturated zone where the effects of double-porosity diffusion were important. The
model was used by Williams et al. (2003) to evaluate the effects of changes in soil water concentration
associated with changes in land use as a result of the Nitrate Sensitive Area Scheme on groundwater
nitrate concentrations at Public Water Supply boreholes and spring sources in the Oolitic Limestone
of Oxfordshire and the Sherwood Sandstone of North Yorkshire. Model output was compared with
observed concentrations, and the parameters (travel times in the unsaturated zone, aquifer kinematic
porosity and aquifer type) adjusted to give a calibrated model which was then used to predict future
concentrations over long time periods.
The MAP approach models flow and transport along a series of ‘streamtubes’ which represent flow
lines from recharge areas at the ground surface through to output features, such as wells, where the
concentration predictions are required. The flow system must be steady-state, and is modelled separately
from the transport problem to determine the flow lines. Williams et al. (2003) developed a steady-state
MODFLOW (Harbaugh et al., 2000) groundwater flow model and used the MODPATH (Pollock, 1989)
particle tracking program to determine the ‘streamtubes’, i.e. the travel distances and travel times from
the water table to various abstraction locations.
Contaminant transport along streamtubes is modelled analytically to provide solutions to a wide
range of transport processes, including Fickian double-porosity diffusive exchange. The form of the
analytical solutions that have been adopted is that of Laplace transforms (Barker, 1982), and implemen-
tation of the method is based on numerical inversion of the Laplace transform solutions. Figure 6.2
illustrates the approach, and Figure 6.3, the mathematical basis.
One advantage of this methodology is that the number of parameters which are varied during the
calibration process is small. This means that predictions of future concentrations are likely to be well
constrained (Barker, 2001). Another advantage is that only concentrations at the times and points of
interest need be computed, so that times can be increased on a logarithmic scale to permit accurate
prediction over very long time-scales typical of double-porosity systems, whilst keeping the method
Figure 6.2: Conceptual and mathematical basis for the Multiple Analytical Pathways of Barker (2001).
Mathematical basis
The approach relies on the concept of the Transfer function Cδ(t)

Complex functions remain relatively simple as Laplace transforms
A series of systems characterised by

separate transfer functions
The concentration at a point of interest is evaluated by numerical inversion of the Laplace

transform of the product of the cumulative transfer function and the input function.
A clear advantage is that only concentrations at points and times of interest need be
computed. E.g. times can be increased on a logarithmic scale when studying long-term
behaviour of dual-porosity systems.
Figure 6.3: Conceptual and mathematical basis for the Multiple Analytical Pathways of Barker (2001).
computationally fast. A disadvantage of the methodology is than no account can be made of variations
in abstraction volumes, because this is a steady-state model that specifies equilibrium flow pathways
from the water table to the abstractions. The necessary assumption that the transport equations are linear
restrict the type of systems that can be modelled. For example, when sorption is important, only the
linear isotherm is amenable to simulation with MAP.
The MAP model used by Williams et al. (2003) did not permit branching or joining streamtubes
(although in principle, branching or joining networks of streamtubes are possible provided each junction
is characterised by a simple additive relation between the streamtube concentrations and fluxes). In
addition, the model was not coded to give matrix concentrations as an output.
6.3.3 GoldSim Contaminant Transport Model

The simulation model GoldSim (developed by GoldSim Technology Group) uses an approach akin to
the MAP approach to simulate flow and transport along ‘pipe pathways’. GoldSim is a highly graphical,
object orientated computer program for carrying out dynamic, probabilistic simulations. The Contam-
inant Transport Module is an extension which allows simulation of the release, transport and fate of
mass within environmental systems (GoldsimTechnologyGroup, 2007). A contaminant transport model
is constructed by defining multiple transport pathways and linking them together into an interconnected
network. Transport pathways can be used to simulate horizontal transport in aquifers, and specifically
‘pipe pathways’ can be used to simulate a broad range of advectively-dominated transport processes
involving one-dimensional advection, longitudinal dispersion, retardation, decay and ingrowth, and ex-
changes with immobile storage zones (e.g. matrix diffusion). The transport equations are solved analyt-
ically using a Laplace transform approach based on Barten (1996) and Barker (1985a). Double-porosity
diffusion can be represented by matrix diffusion zones. Solutes diffuse from the mobile zone into a sur-
rounding porous immobile zone. The diffusive process is one-dimensional and orthogonal to the flow
direction, and according to Fick’s laws. Matrix diffusion zones can have one of three possible geome-
tries: slab, sphere or slot.
The GoldSim model provides simulation output as a time series of concentration or mass flux
leaving the pipe pathway within the mobile zone. The model is not currently coded to provide output for
concentration or mass within the immobile zone. However, a crude approximation can be made by using
a series of linked ‘cells’ (see explanation in Appendix E).
6.3.4 Comparison of DP1D, MAP and GoldSim

In order to explore the functionality of GoldSim and validate its predictions against the MAP model and
the DP1D model, the input and output files from one of the Nitrate Sensitive Areas (NSA) case studies
described in Williams et al. (2003) and Silgram et al. (2005) were obtained and the streamtubes from the
MAP model were reproduced in GoldSim. The case study was the Old Chalford NSA: a small (81 km2 )
catchment with a series of spring sources in the Oolitic Limestone in Oxfordshire.
The validation (Appendix E) indicated that mobile (fissure) concentrations simulated by the MAP
model and GoldSim CT model were almost identical both with and without simulation of double-
porosity diffusion, the apparent differences being a result of the mandatory 10 % minumum dispersion
6.4. Analytical Network Model 249
in GoldSim. However, the immobile (matrix) concentrations simulated by the ‘diffusion cells’ approxi-
mation in GoldSim did not compare well to the average immobile (matrix) concentrations simulated for
a similar scenario in DP1D.
The conclusion from this exercise was that neither MAP or GoldSim were currently appropriate
for the aims of the modelling for Hertfordshire bromate contamination as they are unable to provide
predictions for matrix porewater concentrations, which could be used in the future to validate the model
predictions for long-term fissure water concentrations.
6.4 Analytical Network Model

To allow the branching and joining necessary to represent the Hertfordshire karst system, and the double-
porosity diffusive exchange to be simulated as both mobile (fissure) concentrations and immobile (matrix
porewater) concentrations, Prof. John Barker developed a network model (J. A. Barker, pers. comm.).
The network is defined by ‘branches’ connecting up-stream and down-stream ‘nodes’. Branches
represent flow lines through the saturated zone. A number of branches may enter and/or leave each node,
allowing branching and joining to be represented. Flow and transport is simulated along the network.
The model is run in Microsoft Excel, with the code written in Visual Basic.
6.4.1 Mathematical Basis

As with the MAP model, the transport equations are modelled analytically using Laplace transforms,
and the solution implemented via numerical inversion of the Laplace transform solutions.
6.4.2 Node Input - the Source Function

The model allows considerable flexibility with the ‘source function’, i.e. the concentration input at a
node. The source function is specified by sets of:
• the node number;
• any two times;
• the concentrations at those times; and
• the type of behaviour (constant, linear or exponential) between those times.
6.4.3 Node and Branch description

Branches are defined between an up-stream node and down-stream node. The transport processes rep-
resented are double-porosity diffusive exchange between immobile matrix water and mobile fracture
water, and dispersion in the fracture network.
Branches are characterised by the following parameters:
• the volumetric groundwater flux along the branch, Q;
• the groundwater travel time along the branch, ta ;
• the characteristic time for diffusion across a matrix block, tcb ;

6.5. A Network Model for Hertfordshire 250
• the porosity ratio, σ;
• the geometry of the matrix blocks (currently only a slab geometry is modelled by the code);
• the dispersivity to path length ratio, α ÷ x.
6.4.4 Node and Branch Output

The following solute concentrations are output at specified times:
• cn - the solute concentration in the fractures at the node;
• cf - the solute concentration in the fractures at the downstream end of a branch;
• cm - the average solute concentration in the matrix at the downstream end of a branch;
• cf av - the average concentration in the fractures over whole branch;
• cmav - the average concentration in the matrix over whole branch;
• ctav - the porosity weighted average concentration for fractures and matrix.
Where branches converge at a node, the resulting solute concentration, cn , is given by the sum
of the concentrations, cf , at the end of each of these branches weighted according to the volumetric
groundwater fluxes through each.
6.5 A Network Model for Hertfordshire

6.5.1 Selection of Nodes and Branches
The network is chosen to connect key output locations within the study area that have been interpreted
to be connected by flowlines. The nodes and branches forming the network are illustrated in Figure 6.4.
Only the karstic connections that were indicated by the tracer tests are included as branches to Arkley
Hole Spring and Lynchmill Spring, although in reality there are likely to be additional pathways to these
output springs. The MODPATH flowlines from the model by Cook (2010) were used as an indication of
the existence of flowlines between the source site in Sandridge and locations down-hydraulic-gradient.
However, the travel times and fluxes indicated by the flow lines were not explicitly used because the
distributed model averages parameters over large grid squares, and as such it was considered that the pa-
rameters were not representative of the travel times along the branches, especially along karstic branches.
6.5.2 Parameters for ‘double-porosity’ branches

The parameter ranges for each branch are indicated in Appendix F.
6.5.2.1 Diffusion Coefficient, Dim

The effective diffusion coefficient for bromide and bromate is taken as 8.64 × 10−6 m2 day−1 which is
equivalent to 1.00 × 10−10 m2 s−1 (Section 2.9.4.1).
Ri
Legend Riv
er Chadwell
!
. Nodes Mi sh
Ri
ve mr
am Spring e rA
Karst pathway iv ±
v e r Rib
rL R
ea
Double-porosity pathway
River Beane
(o
Conduits
rL
ee
)
Ne
w
Ri v
Sandridge Arkley Hole

er
!
. Spring
!
. !
.
Harefield
!
.
House Comet Way Lynchmill
Hatfield
Quarry
!
. Spring
!
.
!
.
s
an
r
lb
Ri v e
tA
S
Ne w
of
le
Va
e !Water
.
ln
Co End
r
Ri v e
6.5. A Network Model for Hertfordshire
Riv e r L ea
rne
ou
B
e
rin
Cathe
Mymm shal l Brook
0 0.5 1 2 3 4 5 © Crown Copyright/database right 2008. An Ordnance Survey/EDINA supplied service.
Figure 6.4: Nodes and branches represented in the Network Model for Hertfordshire. Note that branches are shown schematically as straight-line connectors and are not
251
intended to indicate the precise geographical route.

6.5.2.2 Matrix Porosity Φ

For the blocky chalk, a representative value for the Chalk in the area is 0.388 (Section 2.6), with a
standard deviation of 0.058. Therefore, the range of values is taken to be 0.27 to 0.50, with a mean of
0.39.
6.5.2.3 Sandridge to Comet Way, via Harefield House and Hatfield Quarry
For Sandridge to Hatfield, the parameters were estimated from a combination of typical values from the
literature and values obtained from fieldwork in the area.
The parameters a (fracture aperture) and b (half block thickness) determine the value of σ, and along
with Dim , the value of tcb (sections 5.12.1 to 5.12.3). Based on the review of data from the literature
(Section 2.7.3.3), a representative fracture aperture a of 10−3 m is used for the Chalk between Sandridge
and the Hatfield area. It is considered that there is a large uncertainty associated with this parameter, and
therefore, for the purposes of uncertainty estimation, the range is taken to be from an order of magnitude
lower to an order of magnitude higher.
There is little available local information for fracture spacing in the region of the source site and
1 km down-gradient. However, geophysical logging undertaken by TVW in the Hatfield area indicates
flowing fracture separations of approximately 1.00 m to 1.40 m. This is comparable to the flowing
fracture spacing observed from the single borehole dilution testing (Appendix D). These values fall at
the upper end of the range of the literature values for unweathered Chalk reviewed in Section 2.7.3.3.
Although, as noted in the review, these values relate to observed discontinuities, and may not relate to
the presence of hydraulically significant (flowing) fissures. The values relating to the spacing of flowing
horizons tend to have larger spacings. Therefore, for the purposes of uncertainty analysis, the range of
block sizes (2b) is taken to be from 0.50 to 1.50 m, with a mean of 1.00 m.
The darcy flux, q, was taken from the results of the borehole dilution testing at Nashe’s Farm,
Harefield House, and Comet Way (Section 3.6.1). This was converted to a velocity, v, using the mobile
porosity determined from a and b. This was then converted to a travel time ta for the path length. The
darcy fluxes ranged from 0.5 to 3.0 m day−1 , with average of 1.0 m day−1 (Nashe’s Farm), and 0.3 to
1.3 m day−1 , with average of 0.8 m day−1 (Harefield House). At Comet Way, darcy flux was almost an
order of magnitude higher: ranging from 4.5 to 14.5 m day−1 , with average of 8.0 m day−1 . Therefore,
for the branch Sandridge to Harefield House, and from Harefield House to Hatfield Quarry, q was taken
to be 0.3 m day−1 to 3.0 m day−1 , with an average of 1.0 m day−1 . For the branch Hatfield Quarry to
Comet Way q was taken to be 4.5 m day−1 to 14.5 m day−1 , with an average of 8.0 m day−1 .
The volumetric groundwater flux, Q, was determined from the recharge model used for the MOD-
FLOW modelling by Atkins (2004) and Cook (2010), as indicated by the MODPATH flowline. For
Sandridge to Hatfield, the recharge flux was 0.008 m day−1 , which over the site area of 7600 m2 at
SLC, results in a volumetric flux of 6.08 m3 day−1 . The volumetric flux was assumed to remain con-
stant along the flowline.
The value for dispersivity α was taken as 10 % of the path length, x.

6.5.2.4 Comet Way to ‘conduit junction’

This branch was taken as a 1.59 km branch linking Comet Way BH to the main conduit network. The
parameters used were the same as for the branch linking Hatfield Quarry and Comet Way.
6.5.2.5 ‘Conduit junction’ to Arkley Hole Spring and Lynchmill Spring

It is postulated that flowlines characterised by double-porosity bromate transport continue from Comet
Way (via the ‘conduit junction’) to Arkley Hole Spring and Lynchmill Spring. There is very little in-
formation to parameterise these pathways. Therefore, the parameters used between Hatfield Quarry to
Comet Way and to the ’conduit junction’ are used. These branches were taken as 5.71 km and 15.71 km
respectively.
6.5.2.6 Worst case and best case scenarios

The range of parameters outlined in the sections above, are used to define a ‘typical case’, ‘worst case’
(highest peak bromate concentrations) and ‘best case’ (lowest peak bromate concentrations) by the com-
binations indicated in Table 6.1.
Table 6.1: Parameter combinations for ‘best-case’ (lowest peak bromate concentrations) and ‘worst-case’
(highest peak bromate concentrations) scenarios.
parameter ‘worst case’ ‘best case’

a min max
b max min
θm min max
Φ min max
θim min max
σ min max
q max min
v max min
6.5.3 Parameters for karst branches

The parameters for each branch are given in Appendix F.
The parameters tcb , σ, ta , and alpha ÷ x were obtained from Cook (2010), who fitted results from
the tracer test breakthroughs from Water End to Arkley Hole Spring, Lynchmill Spring, and Turnford
PS to the DP-1D model of Barker (2005). The DP-1D curves were fit to the observed data by adjusting
two parameters: the volume to area ratio, b, and the ratio of the inner and outer radius of a concentric
cylinder, ρ, using a least squares method. It should be noted that a number of combinations of b and
ρ produce comparable fits and thus the solutions are non-unique. However, Cook (2010) found that all
follow the same general form to give ta <tcf tcb .
The relative magnitudes of the characteristic times indicates that matrix diffusion is relatively in-
significant in terms of the conduit transport although it does play some role in extending the tracer tail at
6.6. Network Model for Hertfordshire - Results of initial simulations 254
low concentrations.
Characteristic times are given in Table 6.2. The data for the breakthroughs from Harfield House and
Comet Way boreholes were not considered robust enough to be fit to the DP-1D model. Therefore, the
parameters tcb and σ for the branches from Harefield House and Comet Way to Arkley Hole and Lynch-
mill spring were taken to be the same as those from Water End to Arkley Hole and Lynchmill springs.
The values for ta were taken from the travel times determined by Cook (2010) for these connections.
Table 6.2: Parameters derived from fitting the DP-1D model (Barker, 2005) to tracer breakthrough curves
from Water End injection (Cook, 2010). Characteristic times are in hours.
Pathway ta tcf tcb

1 5
Essendon PS 6.21 × 10 2.26 × 10 8.43 × 106
Arkley Hole Spring 6.75 × 101 2.90 × 104 1.05 × 107
Lynchmill Spring 1.05 × 102 6.00 × 103 2.17 × 106
Turnford PS 1.46 × 102 2.08 × 103 7.49 × 105
Groundwater fluxes along the pathways were estimated from the tracer mass recovery data from
the tracer tests. Cook (2010) estimated the pro rata flow rate to each location based on the average
gauged flow to Water End swallow holes of 10756.8 m3 day−1 . The recovery percentage compared the
recovered mass to the amount available to recover at that time assuming literature derived values for
phage inactivation. The groundwater fluxes for the branches from Comet Way and Harefield House were
estimated by multiplying the recovery percentage for these connections by the groundwater flux from
the boreholes as estimated by the results from the single borehole dilution testing (Section 3.6.1.
For the branch Water End to Arkley Hole and Water End to Lynchmill Spring, the relative flux was
estimated by reference to the water balance by Cook (2010). Allogenic recharge accounted for 23.7 % of
flow and groundwater inflow for 14.4 %. Therefore, the flux in the karst system coming from Water End
was taken to be 165 % (23.7 % ÷ 14.4 %) of the flux along the flow line from SLC (i.e. 9.8 m3 day−1 )
and based on the tracer recovery data, this is apportioned as 7.2 m3 day−1 and 2.6 m3 day−1 to Arkley
Hole and Lynchmill respectively.
The value for α÷x was taken as 0.1 % for the branches from Water End, and 1.0 % for the branches
to the west of the main conduit network based from Harefield House and Comet Way.
6.6 Network Model for Hertfordshire - Results of initial simula-

tions
Simulated bromate (Figure 6.5 to Figure 6.9) and bromide (Figures 6.10 to 6.12) concentrations using the
network model described above, with the source term scenarios defined in Section 5.11, show reasonable
agreement with observed concentrations. A constant concentration source term of 5000 µg l−1 bromate
between 1970 and 2050 was also simulated for comparison with Cook (2010) and
In general, Scenario B results in simulated bromate and bromide concentrations well below ob-
6.6. Network Model for Hertfordshire - Results of initial simulations 255
served concentrations between 2000 and 2008. For Harefield House, the observed bromate concentra-
tions span the simulated bromate concentrations for Scenario A and Scenario C. For Hatfield Quarry and
Comet Way, Scenario C passes through the cluster of observed bromate concentrations, and simulated
concentrations for Scenario A are at the lower limits of observed bromate concentrations. For bromide,
the simulated concentrations for Scenario A pass through the cluster of observations at Harefield House,
Hatfield Quarry and Comet Way (although the simulated concentration ‘front’ appears to occur 5-10
years late at Hatfield Quarry), and simulated concentrations for Scenario B are well below observed
concentrations at all locations.
The 5000 µg l−1 constant concentration source term takes longer to reach peak concentrations,
which are also higher than for the other three source scenarios. The constant concentration source term
predicts bromate concentrations lower than observed between 2000 and 2008. This is in contrast to
Atkins (2005) and Cook (2010) who found that the same constant concentration source term gave good
agreement with the observed concentrations between 2000 and 2008, and is probably due to the increased
travel time from the source site to the monitoring locations introduced by the representation of vertical
migration through a low permeability ‘putty chalk’ layer at the source site in the modelling in this thesis.
The general form of the predictions is a bell-shaped curve, with an extended ‘tail’ at later times
as a result of the diffusion from matrix water back into the mobile water. The predicted peak bromate
concentrations occur first for Scenario C, then for Scenario A, and last for Scenario B which shows a
much lower and broader peak.
The importance of double-porosity diffusion in maintaining an elevated ’tail’ can be clearly seen
by comparing the simulated fissure and matrix concentrations at the end of each branch. For Scenario
A, matrix and fracture concentrations attain diffusive equilibrium at 2060, 2080 and 2085 for Harefield
House, Hatfield Quarry and Comet Way respectively. For Scenario C, matrix and fracture concentrations
attain diffusive equilibrium at 2045, 2065 and 2070 for Harefield House, Hatfield Quarry and Comet Way
respectively. After this time, bromate within the matrix provides a secondary source of contamination
which acts to maintain elevated concentrations in the fissures for a prolonged period of time.
At Harefield House, simulated node concentrations remain above 10 µg l−1 until 2225 for Scenarios
A and C and 2175 for Scenario B. At Hatfield Quarry, simulated node concentrations remain above
10 µg l−1 until 2325 for Scenarios A and C and 2250 for Scenario B. At Comet Way, simulated node
concentrations remain above 10 µg l−1 until 2310 for Scenarios A and C and 2230 for Scenario B.
The ‘worst-case’ and ‘best-case’ scenarios (as illustrated for Scenario C in Figure 6.13) affect the
timing, magnitude and sharpness of the peak bromate concentration. Compared to the‘typical-case’
Scenario, ‘worst-case’ scenarios have a narrower, higher concentration peak which occurs earlier, and
‘best-case’ scenarios have a broader, lower concentration peak which occurs later. The relative difference
between the magnitude and timing of predicted concentrations for best, worst and typical scenarios
increases as the distance from the source site increases. The differences in the magnitude and duration of
the bromate concentration peaks are due to the extent of diffusive exchange that occurs between fissure
and matrix. The most bromate mass diffuses into the immobile matrix porewater from the mobile fissure
6.7. Discussion and conclusions 256
water for the ‘best-case’ set of parameters. This attenuates the rise of bromate concentrations in the
fissures, but the back-diffusion from the matrix porewater to the fissures slows the falling limb of the
concentration peak.
For the spring concentrations, simulated bromate and bromide concentrations are significantly lower
(by around an order of magnitude) than observed concentrations. The concentrations at the spring nodes,
reflect the contributions from the four branches joining the spring node (Figure 6.8 and Figure 6.9),
along with the appropriate dilution at the node to represent water flux from flow lines that do not contain
bromate. The double-porosity branch contributes bromate concentrations in a broad peak; maximum
concentrations occur later (between 2050 and 2100) and are maintained for a longer duration, than peak
bromate concentrations provided by the karstic branches. The karstic branches from Harefield House
and Comet Way transport bromate at high concentrations within the fissures to the Arkley Hole and
Lynchmill Spring, while the karstic branch from the ‘karst junction’ to the springs, transports lower
concentrations, although the relative flux is much higher. Simulated matrix concentrations at the end of
the karst branches shows that double-porosity diffusion between matrix and fissure concentrations does
have a significant effect in attenuating bromate concentrations. However, substantially more diffusion of
bromate occurs along the double-porosity branch.
6.7 Discussion and conclusions

The network modelling approach developed in this chapter has been successful in providing a repre-
sentation of double-porosity effects on the catchment-scale migration of bromate contamination. Model
simulations, using a selection of what are considered to be typical parameters for the Hertfordshire
Chalk, and the range of source terms developed earlier in this thesis, simulate bromate and bromide con-
centrations of the order of magnitude of those observed at locations within the Vale of St Albans, west
of the Palaeogene feather edge where the main karst system.
However, in comparison to the long time-scales predicted to be relevant to the evolution of bromate
within the double-porosity Chalk aquifer, observations are available over a very short period of time,
generally for a maximum of eight years between 2000 and 2008, and for significantly less than this at
a number of locations. The seasonal variations within the observed monitoring data also makes trends
difficult to discern (See Chapter 4). It is therefore very difficult to be certain about which point on the
simulated concentration versus time curve the current observations represent, and hence how successful
the model is in simulating bromate evolution. The only way of becoming more certain as to the simulated
curves are representing concentrations at the correct point in evolution, is to have a longer series of
monitoring data that would identify a clearer rising or falling trend, and to have data for porewater
bromate concentrations which would identify at which point along the curve a particular observation
represents.
No additional calibration was undertaken on the results because it is considered that the available
observed monitoring data are insufficient to be able to undertake a robust calibration. Over the timescale
for which elevated bromate concentrations are predicted by the model simulations, the observed data are
available over a very short timescale. It is therefore potentially possible to fit the curve to observations
Location 226 - Harefield House

1.0x101 3.0
Cnode LOG SCALE Cnode LINEAR SCALE
observed GW conc.
Bromate concentation (mg l-1)
Sim. conc. Scenario A

1.0x100
Sim. conc. Scenario B
Sim. conc. Scenario C 2.0
1.0x10-1 Sim. conc. Constant source
5 mg l-1 1965-2015
1.0x10-2
1.0
1.0x10-3
1.0x10-4 0.0
1950 2000 2050 2100 2150 2200 2250 2300 2350 1950 1970 1990 2010 2030 2050
End of Branch:
Harefield House SLC to Harefield House
3.0 3.0
Cnode Cf & Cm
observed GW conc.
Sim. conc. Scenario A - Cnode
2.0 2.0
Sim. conc. Scenario A - Cm
Sim. conc. Scenario A - Cf
1.0 1.0
0.0
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
End of Branch:
3.0 3.0
Cnode Cf & Cm
observed GW conc.
Sim. conc. Scenario B - Cnode
2.0 2.0
Sim. conc. Scenario B - Cm
Sim. conc. Scenario B - Cf
1.0 1.0
0.0
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
End of Branch:
3.0 3.0
Cnode Cf & Cm
observed GW conc.
Sim. conc. Scenario C - Cnode
2.0 2.0
Sim. conc. Scenario C - Cm
Sim. conc. Scenario C - Cf
1.0 1.0
0.0
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
Figure 6.5: Simulated bromate concentrations at Harefield House using source terms for Scenario A, B
and C (Section 5.11), and a constant concentration source term of 5000 µg l−1 .
Location 067 - Hatfield Quarry

1.0x101 2.0
observed GW conc.
0 Sim. conc. Scenario A

1.0x10 1.6
Sim. conc. Scenario C
-1
1.0x10 Sim. conc. Constant source 1.2
5 mg l-1 1965-2015
1.0x10-2 0.8
-3
1.0x10 0.4
-4
1.0x10 0.0
1950 2000 2050 2100 2150 2200 2250 2300 2350 1950 1970 1990 2010 2030 2050
End of Branch:
Hatfield Quarry Harefield House to Hatfield Quarry
2.0 2.0
Cnode Cf & Cm
1.6 observed GW conc. 1.6

1.2 Sim. conc. Scenario A - Cm 1.2

0.8 0.8
0.4 0.4
0.0
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
End of Branch:
2.0 2.0
Cnode Cf & Cm

1.2 Sim. conc. Scenario B - Cm 1.2

0.8 0.8
0.4 0.4
0.0
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
End of Branch:
2.0 2.0
Cnode Cf & Cm

1.2 Sim. conc. Scenario C - Cm 1.2

0.8 0.8
0.4 0.4
0.0
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
Figure 6.6: Simulated bromate concentrations at Hatfield Quarry using source terms for Scenario A, B
and C (Section 5.11), and a constant concentration source term of 5000 µg l−1 .
Location 402 - Comet Way

1.0x101 1.0
observed GW conc.

1.0x100 0.8
-1
1.0x10 Sim. conc. Constant source 0.6
5 mg l-1 1965-2015
1.0x10-2 0.4
1.0x10-3 0.2
-4
1.0x10 0.0
1950 2000 2050 2100 2150 2200 2250 2300 2350 1950 1970 1990 2010 2030 2050
End of Branch:
Comet Way Hatfield Quarry to Comet Way
1.0 1.0
Cnode Cf & Cm


0.4 0.4
0.2 0.2
0.0
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
End of Branch:
1.0 1.0
Cnode Cf & Cm


0.4 0.4
0.2 0.2
0.0
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
End of Branch:
2.0 2.0
Cnode Cf & Cm

1.2 Sim. conc. Scenario C - Cm 1.2

0.8 0.8
0.4 0.4
0.0
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
Figure 6.7: Simulated bromate concentrations at Comet Way using source terms for Scenario A, B and
C (Section 5.11), and a constant concentration source term of 5000 µg l−1 .
Location 287/288 - Arkley Hole Spring

0 Cnode - LOG SCALE Cnode - LINEAR SCALE
1.0x10 0.04
observed GW conc.
1.0x10-1 Sim. conc. Scenario B 0.03
Sim. conc. Constant source
5 mg l-1 1965-2015
1.0x10-2 0.02
1.0x10-3 0.01
1.0x10-4 0.00
1950 2000 2050 2100 2150 2200 2250 1950 2000 2050
End of Branch: End of Branch:
0.50 Double-porosity to Arkley Hole Double-porosity toArkley Hole

Cf Cm
0.40
0.30
0.20
0.10
0.00
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
0.50 End of Branch: Water End to Arkley Hole End of Branch: Water End to Arkley Hole
Cf Cm
0.40
0.30
0.20
0.10
0.00
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
2.50 End of Branch: Harefield House to Arkley Hole End of Branch: Harefield House to Arkley Hole
Cf Cm
2.00
1.50
1.00
0.50
0.00
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
2.50 End of Branch: Comet Way to Arkley Hole End of Branch: Comet Way to Arkley Hole
Cf Cm
2.00
1.50
1.00
0.50
0.00
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
Figure 6.8: Simulated bromate concentrations at Arkley Hole Spring node, and at the end of contributing
branches, using source terms for Scenario A, B and C (Section 5.11), and a constant concentration source
term of 5000 µg l−1 .
Location 382 - Lynchmill Spring

Cnode - LOG SCALE Cnode - LINEAR SCALE
1.0x100 0.05

-1
0.04
1.0x10 Sim. conc. Scenario C
observed GW conc.
0.03
-2
1.0x10
0.02
-3
1.0x10
0.01
1.0x10-4 0.00
1950 2000 2050 2100 2150 2200 2250 1950 2000 2050
End of Branch: End of Branch:

0.50 Double-porosity to Lynchmill Double-porosity to Lynchmill

Cf Cm
0.40
0.30
0.20
0.10
0.00
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
0.50 End of Branch: Water End to Lynchmill End of Branch: Water End to Lynchmill
Cf Cm
0.40
0.30
0.20
0.10
0.00
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
2.50 End of Branch: Harefield House to Lynchmill End of Branch: Harefield House to to Lynchmill
Cf Cm
2.00
1.50
1.00
0.50
0.00
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
2.50 End of Branch: Comet Way to Lynchmill End of Branch: Comet Way to Lynchmill
Cf Cm
2.00
1.50
1.00
0.50
0.00
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
Figure 6.9: Simulated bromate concentrations at Lynchmill Spring node, and at the end of contributing
branches, using source terms for Scenario A, B and C (Section 5.11), and a constant concentration source
term of 5000 µg l−1 .
Location 226 - Harefield House

1.0x101 10.0
observed GW conc.
Bromide concentation (mg l-1)

8.0
1.0x100 Sim. conc. Scenario B
6.0
1.0x10-1
4.0
-2
1.0x10
2.0
-3
1.0x10 0.0
1950 2000 2050 2100 2150 2200 2250 2300 2350 1950 1970 1990 2010 2030 2050
End of Branch:
10.0 10.0
Cnode Cf & Cm


4.0 4.0
2.0 2.0
0.0
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
End of Branch:
10.0 10.0
Cnode Cf & Cm


4.0 4.0
2.0 2.0
0.0
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
Figure 6.10: Simulated bromide concentrations at Harefield House using source terms for Scenario A
and B.
Location 067 - Hatfield Quarry

1.0x101 5.0
observed GW conc.

0
4.0
3.0
-1
1.0x10
2.0
1.0x10-2
1.0
-3
1.0x10 0.0
1950 2000 2050 2100 2150 2200 2250 2300 2350 1950 1970 1990 2010 2030 2050
End of Branch:
5.0 5.0
Cnode Cf & Cm


2.0 2.0
1.0 1.0
0.0
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
Hatfield Quarry End of Branch: Harefield House to Hatfield Quarry

5.0 5.0
Cnode Cf & Cm


2.0 2.0
1.0 1.0
0.0
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
Figure 6.11: Simulated bromide concentrations at Hatfield Quarry using source terms for Scenario A
and B.
Location 402 - Comet Way

1.0x101 5.0
observed GW conc.

0
4.0
3.0
-1
1.0x10
2.0
1.0x10-2
1.0
-3
1.0x10 0.0
1950 2000 2050 2100 2150 2200 2250 2300 2350 1950 1970 1990 2010 2030 2050
End of Branch:
5.0 5.0
Cnode Cf & Cm
observed GW conc.
4.0 Sim. conc. Scenario A - Cnode 4.0
Sim. conc. Scenario A - Cm
3.0 3.0
2.0 2.0
1.0 1.0
0.0
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
End of Branch:
5.0 5.0
Cnode Cf & Cm


2.0 2.0
1.0 1.0
0.0
1950 2000 2050 2100 2150 1950 2000 2050 2100 2150
Figure 6.12: Simulated bromide concentrations at Comet Way using source terms for Scenario A and B.
observed GW conc.
Sim. Cm. Scenario C - BEST
Sim. Cnode or Cf. Scenario C - BEST
Sim. Cm. Scenario C - WORST
Sim. Cnode or Cf. Scenario C - WORST
Sim. Cm. Scenario C - TYPICAL
Sim. Cnode or Cf. Scenario C - TYPICAL
Harefield House
3.0 3.0
Cnode End of Branch:
SLC to Harefield House
Cf & Cm
2.0 2.0
1.0 1.0
0.0
1950 2000 2050 2100 2150 2200 2250 1950 2000 2050 2100 2150 2200 2250
Hatfield Quarry End of Branch:

3.0 3.0
Cnode SLC to Hatfield Quarry
Cf & Cm
2.0 2.0
1.0 1.0
0.0 0.0
1950 2000 2050 2100 2150 2200 2250 1950 2000 2050 2100 2150 2200 2250
Comet Way End of Branch:

3.0 3.0
Cnode SLC to Comet Way
Cf & Cm
2.0 2.0
1.0 1.0
0.0 0.0
1950 2000 2050 2100 2150 2200 2250 1950 2000 2050 2100 2150 2200 2250
Figure 6.13: Simulated bromate concentrations for Scenario C at Harefield House, Hatfield Quarry, and
Comet Way using ‘best-case’, ‘typical-case’ and ‘worst-case’ parameters.
in a number of ways: the point of time in the contaminant evolution which is represented by the obser-
vations is not evident. A longer period of monitoring data would be necessary to identify trends which
could allow the curve to be calibrated with more certainty. Furthermore, observations of a particular
fissure concentration can represent one of two points on the simulated curve (Figure 6.14). Concurrent
observations of matrix porewater concentrations are required to identify which point in the evolution
such a fissure concentration represents.
Figure 6.14: Concurrent matrix and fissure concentrations are required to determine at which point along
the concentration-time graph a particular fissure concentration represents.
The network model has been less successful in representing concentrations at locations to the east
of the Palaeogene escarpment, where the karst flow system is believed to have a dominant effect on the
migration of bromate contamination across the catchment. This could be due to the uncertainty in the
relative fluxes of the karst system. More extensive tracer testing and flow gauging could help to reduce
this uncertainty.
The representation of the karst branches within the network model is a very simplistic approxima-
tion to the likely reality. The karst branches represented are only those that were identified by tracer
testing, but in reality there are likely to be additional connections. Also, very limited interaction between
karst branches and double-porosity branches is simulated by this model: the flow pathway between the
Hatfield Area and Arkley Hole and Lynchmill Spring is represented by either a single karst branch or
a single double-porosity branch. Further work could experiment with more extensive branching and
joining between karst branches and double-porosity branches.
The Network model represents a one-dimensional steady-state flow regime between nodes. There-
fore, unlike numerical models such as MODFLOW and MT3D, the network model model output does
not capture seasonal variations in concentrations, nor incorporate transient flow conditions such as those
introduced by changes in abstraction regimes.

Despite the uncertainties in the foregoing discussion, the network model has been successful in
highlighting the long-term effects of double-porosity nature of the chalk on the catchment-scale evo-
lution of bromate, which previous models of models of bromate in Hertfordshire have neglected. The
timescale of the catchment bromate contamination in Hertfordshire is similar to that of extensive chlo-
ride pollution of the Chalk of the Tilmanstone valley in Kent from coalfield brines. Watson (2004) used
a one-dimensional semi-analytical model (DP1D), incorporating Fickian diffusion between matrix wa-
ter and fracture water, was used to simulate chloride migration and was able to reproduce the observed
porewater and fracture water profiles. Forward modelling indicated that the double porosity diffusion ex-
tends the duration of contamination in the catchment by several decades. (Watson, 2004; Burgess et al.,
2005). The long-term effects of the double-porosity nature of the Chalk is relevant to catchment-scale
Nitrate (NO3– ) pollution in the Chalk (e.g. Williams et al. 2003; Silgram et al. 2005), and potentially
other aquifers which possess double-porosity characteristics, as it highlights the timescales over which
land management changes will take effect.
The double-porosity diffusion significantly extends the duration of elevated levels of contamination
by providing a secondary source of bromate. Concentrations are predicted to remain above regulatory
limits in the Vale of St Albans for around 200 years. The persistence of bromate within the Hertford-
shire Chalk aquifer for such long time scales has enormous financial implications for water resources
engineering: current treatment processes and scavenge pumping management measures may cease to be
sustainable, and finding alternative sources of water supply may become increasingly necessary. There
remains substantial uncertainty in the predicted time-series of bromate concentrations, and this could be
addressed by further sensitivity analyses and calibration as further monitoring data become available.
Nevertheless, the persistence of bromate for decades to centuries is the best available estimate at the
current time, and should be considered in management strategies in Hertfordshire.
268
Chapter 7
Conclusions
This chapter considers the contributions of the analysis and interpretation of data and the modelling work
presented in Chapters 3, 4, 5 and 6 in meeting the aims and objectives of the research outlined in Chap-
ter 1. For convenience, the research objectives are repeated and discussed in sections 7.1.1, 7.1.2 to 7.1.3.
7.1 Fulfillment of research aims and objectives

The overall aim for the EngD research presented in this thesis is:
To develop greater understanding of the processes controlling the spatial distribution

and temporal evolution of bromate contamination within the Hertfordshire Chalk aquifer,
including bromate release from the source zone, and to use this as the basis of predictive
models incorporating the effects of double-porosity diffusion on the long-term evolution of
bromate.
The research presented in this thesis has fulfilled this overall aim by developing a refined conceptual
model of bromate transport within the catchment and interpreting the spatial distribution of bromate in
light of this conceptual understanding, by conceptualising and quantifying a range of source history
scenarios for bromate input to the aquifer, and by using these advancements in understanding as a basis
for developing a network model that demonstrates the influences of double-porosity diffusion on the
long-term evolution of bromate at a catchment-scale.
7.1.1 Evolution of bromate contamination

Objectives:
• To develop a conceptual model for groundwater flow and contaminant transport in the Hertford-
shire Chalk aquifer system by review of existing data and interpretation of additional tracer testing
and geophysical testing;
• To use the available information and monitoring data to describe the spatial distribution and
temporal evolution of bromate across the catchment, and to interpret this in association with the
conceptual model of the flow and transport system.
The detailed analysis of bromate and bromide monitoring data presented in Chapter 4 has revealed
that bromate concentrations are affected by influences including recharge (soil moisture deficit, rainfall),
7.1. Fulfillment of research aims and objectives 269
water level, and abstractions. These relationships, integrated with the observations of the geology, hy-
drogeology and hydrology of the area affected by the bromate contamination, have been used to refine
the conceptual model of groundwater flow and transport of bromate within the catchment Chapter 3.
The conceptualisation supports double-porosity dominated transport of bromate within the Vale of St.
Albans area, which maintains a highly attenuated, relatively stable contaminant distribution west of
Hatfield. To the east of Hatfield, a significant karst network related to the position of the Palaeogene
overlap of the Chalk influences bromate trasport, dispersing bromate over large distances toward the
northern and middle Lea Valley. The revised conceptual understanding provides the basis for modelling
approaches applied to predict long-term, large-scale transport of bromate within the Hertfordshire Chalk
(section 7.1.3), and has allowed a new interpretation of the spatial distribution and evolution of bromate
and bromide within the catchment to be developed between 2000 and 2008.
However, the interpretation of the spatial and temporal evolution of bromate and bromide within the
catchment is hampered by a number of inadequacies in the available monitoring data: trends are difficult
to discern because monitoring data is available for a relatively short period of time, monitoring frequency
varies considerably between locations and varies over time at individual locations, and there are strong
seasonal influences. Data are generally for (non depth-specific) pumped groundwater samples so that
vertical distribution of bromate contamination cannot be investigated, nor can the matrix porewater con-
centrations be diagnosed. This is a severe limitation of the monitoring and investigation programme,
effectively preventing proper consideration of diffusive retardation of bromate which the thesis shows
may prolong the occurrence of bromate contamination by 200 years.
7.1.2 The source

Objectives:
In Chapter 5, the available site investigation data was assessed and interpreted to estimate the quan-
tity of bromate and bromide present on site in the unsaturated and saturated zone soils and groundwater
prior to the redevelopment of the site in the mid 1980s and subsequent to the discovery of bromate con-
tamination in early 2000s. These estimates have been used to constrain three source term scenarios for
bromate input to the aquifer beneath the site. There are many uncertainties associated with an incom-
plete knowledge of the history of the site, and the three source scenarios attempt to capture the range
of possible bromide and bromate source histories. This rigorous analysis of the source zone provides
a significant improvement in the characterisation of the bromate ‘source term’ compared to previous
representations, particularly the constant concentrations source terms used by Buckle (2003) and Atkins
7.1. Fulfillment of research aims and objectives 270
(2005). Nevertheless, it is recognised that due to the relative scarcity of data, these scenarios still repre-
sent crude estimates of the reality.
The one-dimensional double-porosity transport code, DP1D (Barker, 2005), has been used to simu-
late concentrations in groundwater down-gradient of the source site. Simulated concentrations using two
of the source term scenarios show relatively good agreement with observed groundwater concentrations
at locations 150 m, 500 m and 1000 m down-gradient of the source site. Either a ‘catastrophic release’
of bromide/bromate to the unsaturated zone followed by leaching to groundwater beneath the site, or a
‘direct release’ of bromate to the saturated zone, sometime between 1960 and 1970, result in the closest
fit to observed data. Porewater concentrations are not available for locations down-gradient of the source
site. The lack of porewater concentrations and the relatively short period of time for which groundwater
monitoring data (fissure concentrations) are available, combined with the large seasonal variations in
concentrations, means that the trends are difficult to discern, and robust conclusions cannot be made as
to whether or not the simulations are representative of the actual release scenarios.
7.1.3 Catchment-scale modelling of bromate transport

Objectives:
• To develop analytical network modelling of contaminant transport to allow representation of Fick-

ian double-porosity diffusion and to integrate karstic transport pathways within the network;
• To use this model to produce predictions for the likely bromate concentrations at key output loca-
tions over the long-term.
A novel analytical network model to represent the Hertfordshire Chalk catchment has been devel-
oped in Chapter 6, using code written by Prof. John Barker. The network model simulates Fickian
double-porosity diffusive exchange along interconnecting flow-lines, while allowing karstic branches to
be incorporated into the network. The model was parameterised by a combination of values found within
the literature, and the results of the single borehole dilution testing and catchment-scale natural gradient
tracer testing.
The network model, using the range of source terms developed in Chapter 5, has been successful in
simulating bromate and bromide concentrations of the order of magnitude of those observed at locations
within the Vale of St. Albans, west of the main karst system. The network model highlights the long-term
effects of double-porosity nature of the chalk on the catchment-scale evolution of bromate. The double-
porosity diffusion significantly extends the duration of elevated levels of contamination by providing a
secondary source of bromate: bromate concentrations within the Vale of St. Albans are predicted to
remain above regulatory limits for around 200 years.
The network model has been less successful in representing concentrations at locations to the east
of the Palaeogene escarpment, where the karst flow system is believed to have a dominant effect on the
migration of bromate contamination across the catchment. This is largely due to the uncertainty in the
relative fluxes of the karst system. The representation of the karst branches within the network model is
a very simplistic approximation to the likely reality. The karst branches represented are only those that
7.2. Recommendations for further work 271
were identified by tracer testing, and very limited interaction between karst branches and double-porosity
branches is simulated by this model. Further work could experiment with more extensive branching of
the karst network and closer interaction between karst branches and double-porosity branches.
The Network model represents a one-dimensional steady-state flow regime between nodes. There-
fore, unlike numerical models such as MODFLOW and MT3D, the network model model output does
not capture seasonal variations in concentrations, nor incorporate transient flow conditions such as those
introduced by changes in abstraction regimes. However, the analytical network model has a number of
advantages over the numerical models. In particular, the analytical model is computationally fast over
long time-scales, taking seconds to minutes to run compared to days to weeks for the catchment MODL-
FOW and MT3D-MS models (Atkins, 2005; Cook, 2010). Also, for the analytical network model, the
number of parameters which are varied during the calibration process is small compared to a distributed
parameter numerical model. No additional calibration was undertaken on the results within this thesis
due to insufficient monitoring data. Without measured bromate concentrations in porewater and longer
time-series of bromate concentrations in fissure water becoming available, no effective calibration will
be possible.
7.2 Recommendations for further work

The ability to validate the conceptualisations and modelling representations of bromate transport in the
Hertfordshire Chalk is dependent on the available of representative observation data. Although a bromate
monitoring programme has been underway since mid 2000, in comparison to the long time-scales pre-
dicted to be relevant to the evolution of bromate within the double-porosity Chalk aquifer, observations
are available over a very short period of time: generally for a maximum of eight years between 2000
and 2008, and for significantly less than this at a number of locations. The seasonal variations within the
observed monitoring data also makes trends difficult to discern. Furthermore, the vast majority of the
data relate to bromate concentrations withi the mobile fissure water; concurrent bromate concentrations
in the immobile matrix porewater are absent except in close proximity to the source site. Both concen-
trations in the fissure water and matrix porewater are necessary to fully characterise a system influenced
by double-porosity exchange and establish the extent of disequilibrium between fissures and matrix.
It is therefore considered essential that a number of cored investigatory boreholes are drilled across
the bromate affected area, and particularly in the the Vale of St. Albans area between the source site and
the Hatfield. These boreholes should be sampled for matrix porewater and fissure water. Samples should
be taken at specific depths to provide information about the vertical extent of contamination.
It would be informative to run the MODFLOW/MT3D-MS model of Cook (2010) as a steady-state
simulation, with the source terms used in this thesis, to directly compare simulated concentrations at the
locations also included in the Hertfordshire network model, with the results simulated by the network
model. This would allow the impacts of using a first order mass transfer coefficient to represent double-
porosity as opposed to representing it as a Fickian diffusion process to be assessed.
BIBLIOGRAPHY 272
Bibliography
Allen, D. J., Brewerton, L. J., Coleby, L. M., Gibbs, B. R., Lewis, M. A., MacDonald, A. M., Wagstaff,
S. J. and Williams, A. T. (1997), The physical properties of major aquifers in England and Wales.
British Geological Survey Technical Report WD/97/34, Technical Report, British Geological Survey.
Atkins (2002), St. Leonard’s Court, Sandridge, St. Albans. Environmental Site Investigation and Quan-
titative Pollutant Linkage Assessment., Technical Report, Atkins Ltd.
Atkins (2004), Bromate Contamination in the Lee Valley. Phase 1: Data Collation and Conceptualisa-
tion., Technical Report, Atkins Limited.
Atkins (2005), Bromate Contamination in the Lee Valley Phase 2: Modelling Report., Technical Report,
Atkins Limited.
Atkins (2006), Bromate Monitoring Data Review, Technical Report, Atkins Limited.
Atkinson, T. C. and Smith, D. I. (1974), Rapid groundwater flow in fissures in the chalk: an example
from south Hampshire, Quarterly Journal of Engineering Geology and Hydrogeology 7(2), 197–205.
Atkinson, T. and Smart, P. (1981), Artificial Tracers in Hydrology, in A Survey of British Hydrology,
Royal Society, London, pp.173–190.
Banks, D., Davies, C. and Davies, W. (1995), The Chalk as a karstic aquifer: evidence from a tracer test
at Stanford Dingley, Berkshire, UK, Quarterly Journal of Engineering Geology and Hydrogeology
28(Supplement1 ), S31 − −38.
Barker, J. A. (1982), Laplace transform solutions for solute transport in fissured aquifers, Advances in
Water Resources 5(2), 98–104.
Barker, J. A. (1985a), Block-Geometry Functions Characterizing Transport in Densely Fissured Media,

Journal of Hydrology 77(1-4), 263–279.
Barker, J. A. (1985b), Modelling the effects of matrix diffusion on transport in densely fissured media., in
Hydrogeology in the Service of Man. Memoirs of the 18th Congress of the International Association fo
Hydrogeologists., Vol. XVIII, International Association of Hydrogeologists, Cambridge, pp.250–269.
Barker, J. A. (1991), Transport in fractured rock, Applied groundwater hydrology pp.199–216.

BIBLIOGRAPHY 273
Barker, J. A. (1993), Modelling of flow and transport in the chalk, in R. A. Downing, M. Price and G. P.
Jones (eds.), The Hydrogeology of the Chalk of North-West Europe, Clarendon Press, Oxford, pp.35–38.
Barker, J. A. (2001), Mathemetical basis for the MAP model., Technical Report.
Barker, J. A. (2005), DP1D-Link Description of the code and issues. Version 6., Technical Report.
Barker, J. A. and Foster, S. S. D. (1981), A Diffusion Exchange Model for Solute Movement in Fissured
Porous Rock, Quarterly Journal of Engineering Geology 14(1), 17–24.
Barker, J. A., Wright, T. and Fretwell, B. (2000), A pulsed-velocity method of solute transport modelling.,
in A. Dassargues (ed.), Tracers and Modelling in Hydrogeology, Proceedings of the TraM’2000 Con-
ference, Vol. International Association of Hydrological Sciences Publication 262, IAHS Press, Liege,
Belgium, p.297302.
Barten, W. (1996), Linear response concept combining advection and limited rock matrix diffusion in a
fracture network transport model, Water Resources Research 32(11), 3285–3296.
Bear, J. (1972), Dynamics of Fluids in Porous Media., Elsevier, New York.
Bibby, R. (1981), Mass treansport studies in dual porosity media., Water Resources Research 17, 1075–
1081.
Bloomfield, J. (1996), Characterisation of hydrogeologically significant fracture distributions in the Chalk:

An example from the Upper Chalk of southern England, Journal of Hydrology 184(3-4), 355–379.
Bloomfield, J. P., Brewerton, L. J. and Allen, D. J. (1995), Regional trends in matrix porosity and dry
density of the chalk of England, Quarterly Journal of Engineering Geology 28, S131–S142.
Bloomfield, J. P., Farrant, A. R. and Cunningham, J. (2004), Structural Interpretation of the Chalk of the
Hertford District based on Slope Aspect Mapping (Draft). BGS Report Ref. CR/04/XXC, Technical
Report Ref. CR/04/XXC, British Geological Survey.
Board, W. R. (1972), The Hydrogeology of the London Basin, Technical Report, Water Resources Board.
Bristow, R., Mortimore, R. and Wood, C. (1997), Lithostratigraphy for mapping the Chalk of southern
England, Proceedings of the Geologists Association 108, 293–315.
Buckle, D. (2002), Bromate Groundwater Flow Study. Phase 1 (Conceptual Understanding) Report. Under-
taken on behalf of the Environment Agency and Three Valleys Water., Technical Report, Vivendi Water
Partnership.
Buckle, D. (2003), Bromate Groundwater Flow Study. Phase 2 Interim (Modelling) Report. Undertaken on
behalf of the Environment Agency and Three Valleys Water., Technical Report, Vivendi Water Partner-
ship.
BIBLIOGRAPHY 274
Burgess, W., Barker, J., Watson, S. and Fretwell, B. (2005), Contaminant retardation by double porosity
diffusive exchange in the Chalk: implications for monitoring., in Bringing Groundwater Quality Re-
search to the Watershed scale. Proceedings of the 4th International Groundwater Quality Conference..,
IAH, Waterloo, Canada.
Butler, R., Godley, A., Lytton, L. and Cartmell, E. (2005), Bromate environmental contamination: Review
of impact and possible treatment, Critical Reviews in Environmental Science and Technology 35(3), 193–
217.
Chatfield, C. (2004.), The Analysis of Time Series: An Introduction., sixth edition edition, Chapman and
Hall/CRC.
Chemfix (1984), Sandridge site bromide modelling exercise, Technical Report.
Chemfix (1985a), Evaluation of the results from the borehole situated 120m down dip from the Sandridge
site., Technical Report.
Chemfix (1985b), Field report for drilling of borehole C1 at House Lane, Sandridge on 21/22 January 1985,
Technical Report.
Chemfix (1985c), Report of the second phase of the field investigation at the Sandridge site, Technical
Report.
Chemfix (1985d), Report on the further modelling studies for the Sandridge site., Technical Report.
CL:AIRE (2002), Site characterisation in support of monitored natural attenuation of fuel hydrocarbons
and MTBE in a chalk aquifer in southern England., in Case study bulletin CSB1., Contaminated Land:
Applications in Real Environments (CL:AIRE).
Cook, S. J. (2010), The hydrogeology of bromate contamination in the Hertfordshire Chalk: Incorporating
Karst in Predictive Models., EngD, University College London.
Cushman, J. H. (1990), An Introduction to Hierarchical Porous Media, in J. H. Cushman (ed.), Dynamics

of Fluids in Hierarchical Porous Media, Academic Press Ltd, London, pp.1–6.
de Marsily, G. (1986), Quantitative Hydrogeology : Groundwater Hydrology for Engineers, Academic

Press, Orlando FL.
Domenico, P. and Schwartz, W. (1998), Physical and Chemical Hydrogeology., second edition edition,
Wiley.
EA (2005), Part IIA of the Rnvironmental Protection Act 1990. Decision Document for St Leonard’s Court.,
Technical Report, Environment Agency.
Edmunds, W. M. (1996), Bromine geochemistry of British groundwaters, Mineralogical Magazine

60(399), 275–284.
BIBLIOGRAPHY 275
Edmunds, W. M., Cook, J. M., Kinniburgh, D. G., Miles, D. L. and Trafford, J. M. (1989), Trace-element
occurrence in British groundwaters., Technical Report BGS Research Report SD/89/3, British Geologi-
cal Survey.
Edwards, A. J., Hobbs, S. L. and Smart, P. L. (1991), Effects of quarry dewatering on a karstified limestone
aquifer: A case study from the Mendip Hills, England., in Third Conference on Hydrogeology, Ecology,
Monitoring, and Management of Ground Water in Karst Terranes., National Water Well Association,
pp.77–92.
Entec (2000), River Mimram and Upper Lee Water Resources Sustainability Study Groundwater Model.
Phase 1 Report: Data Collation and Formulation of Conceptual Model. Undertaken on behalf of the
Environment Agency., Technical Report, Entec.
Entec (2002), River Mimram and Upper Lee Water Resources Sustainability Study Groundwater Model.
Phase 2 Model Development and Refinement. Undertaken on behalf of the Environment Agency., Tech-
nical Report, Entec.
Fitzpatrick, C. M. (2007), Hatfield Pumping Test. Interim Report for TVW and TWUL for the period July
2005 to December 2006., Technical Report, University College London.
Ford, D. C. and Williams, P. (2007), Karst hydrogeology and geomorphology, John Wiley, Chichester.
Foster, S. S. D. (1975), The Chalk groundwater tritium anomaly – A possible explanation, Journal of
Hydrology 25(1-2), 159–165.
Foster, S. S. D. and Milton, V. A. (1974), The permeability and storage of an unconfined chalk aquifer,
Hydrological Sciences Bulletin 19, 485–500.
Gale, A., Hancock, J., Bristow, R., Mortimore, R. and Wood, C. (1999), ’Lithostratigraphy for mapping
the Chalk of southern England’ by Bristow et al. (1997): discussion, Proceedings of the Geologists
Association 110, 65–72.
GoldsimTechnologyGroup (2007), GoldSim Contaminant Transport Module. User Guide, Technical Re-
port, GoldsimTechnologyGroup.
Goudie, A. (1990), The Landforms of England and Wales, Basil Blackwell, Oxford.
Grisak, G. E. and Pickens, J. F. (1980), Solute Transport through Fractured Media .1. The Effect of Matrix
Diffusion, Water Resources Research 16(4), 719–730.
Grisak, G. E. and Pickens, J. F. (1981), An Analytical Solution for Solute Transport through Fractured
Media with Matrix Diffusion, Journal of Hydrology 52(1-2), 47–57.
Hancock, J. M. (1975), The petrology of the Chalk, Proceedings of the Geologists’ Association 86(4), 499–
535.
BIBLIOGRAPHY 276
Harbaugh, A. W., Banta, E. R., Hill, M. and McDonald, M. (2000), MODFLOW-2000, The U.S. Geological
Survey modular ground-water modelUser Guide to modularization concepts and the ground-water flow
process., Technical Report, U.S. Geological Survey,.
Harold, C. H. H. (1937), Thirty-second Annual Report on the results of the chemcial and bacteriological
examination of the London waters for the twelve months ended 31st December 1937., Technical Report,
Metropolitan Water Board.
Headworth, H. (1972), The analysis of natural groundwater fluctuations in the chalk of Hampshire., Journal
of the Institute of Water Engineers 26, 107–124.
Headworth, H. G. (1978), Hydrogeological characteristics of artesian boreholes in the Chalk of Hampshire,

Quarterly Journal of Engineering Geology and Hydrogeology 11(2), 139–144.
Headworth, H. G., Keating, T. and Packman, M. J. (1982), Evidence for a shallow highly-permeable zone
in the Chalk of Hampshire, U.K, Journal of Hydrology 55(1-4), 93–112.
Helsel, D. and Hirsch, R. (1993), Statistical Methods in Water Resources, Elsevier.
Hijnen, W., Voogt, R., Veenendaal, H. and van der Jagt, H. (1995), Bromate reduction by denitrifying
bacteria., Applied Environmental Microbiology 61, 239.
Hill, D. (1984), Diffusion coefficients of nitrate, chloride, sulphate and water in cracked and uncracked
Chalk., Journal of Soil Science 35, 27–33.
Hydrotechnica (1988), Northern New River Wellfield Pump Testing. Final Report., Technical Report,
Hydrotechnica.
Ineson, J. (1962), A hydrogeological study of the permeability of Chalk, Journal of the Institution of Water
Engineers 16, 449–463.
Kachi, S. (1987), Tracer studies in the Chalk aquifer near Cambridge., PhD thesis, University of East
Anglia.
Klinck, B. A., Hopson, P. M., Lewis, M. A., MacDonald, D. M., Inglethorpe, S. D. J., Entwisle, D. C., Har-
rington, J. F. and Williams, L. (1998), The hydrogeological behaviour of the clay-with-flints in southern
England. British Geological Survey Technical Report, WE/97/5., Technical Report, British Geological
Survey.
Komex (2000), Site Investigation at St Leonards Court, Sandridge, St Albans., Technical Report 50598,
Komex.
Kruithof, J. and Meijers, R. (1995), Bromate formation by ozonation ans advanced oxidation and potential
options in drinking water treatment., Water Supply 13, 93.
BIBLIOGRAPHY 277
Lewin, K., Bradshaw, K., Blakey, N., Turrell, J., Hennings, S. and Fleming, R. (1994), Leaching Tests for
Assessment of Contaminated Land: Interim NRA Guidance. National Rivers Authority RD Note 301 .,
Technical Report, National Rivers Authority.
MacDonald, A., Brewerton, L. and Allen, D. (1998), Evidence for rapid groundwater flow and karst
type behaviour in the Chalk of Southern England., in N. Robins (ed.), Groundwater Pollution, Aquifer
Recharge and Vulnerability, Geological Society of London Special Publication No. 130, pp.95–106.
MacDonald, A. M. and Allen, D. J. (2001), Aquifer properties of the Chalk of England, Quarterly Journal
of Engineering Geology and Hydrogeology 34(4), 371–384.
MacDonald, M. G. and Harbaugh, A. W. (1984), A modular three-dimensional finite-difference groundwa-

ter flow model., Technical Report, U.S. Geological Survey.
Mathias, S. A., Butler, A. P., Jackson, B. M. and Wheater, H. S. (2006), Transient simulations of flow and
transport in the Chalk unsaturated zone, Journal of Hydrology 330(1-2), 10–28.
Mathias, S. A., Butler, A. P., McIntyre, N. and Wheater, H. S. (2005), The significance of flow in the matrix
of the Chalk unsaturated zone, Journal of Hydrology 310(1-4), 62–77.
Maurice, L. D., Atkinson, T. C., Barker, J. A., Bloomfield, J. P., Farrant, A. R. and Williams, A. T. (2006),
Karstic behaviour of groundwater in the English Chalk, Journal of Hydrology 330(1-2), 63–70.
Mortimore, R. N. (1993), Chalk water and engineering geology., in R. A. Downing, M. Price and G. P.
Jones (eds.), The Hydrogeology of the Chalk of North-West Europe, Clarendon Press, Oxford, pp.67–92.
Mortimore, R., Pomerol, B. and Foord, R. J. (1990), Engineering stratigraphy and paeleogeography for
the chalk of the Anglo-Paris Basin, in Proceedings of the International Chalk Symposium 1989, Thomas
Telford.
Oakes, D. (1977), The movement of water and solutes through the unsaturated zone of the Chalk of the
United Kingdom., in 3rd International Hydrology Symposium, Fort Collins, Colorado State University,
pp.447–459.
Ogata, A. and Banks, R. B. (1961), A solution of the differential equation of longitudinal dispersion in
porous media., United States Geological Survey Professional Paper 411-A.
Pollock, D. W. (1989), Documentation of computer programs to compute and display pathlines using
resultsd form the U.S. Geological survey modular three-dimensional finite-difference ground-water
model., Technical Report, U.S. Geological Survey.
Price, M. (1987), Fluid flow in the Chalk of England, in J. C. Goff and B. P. J. Williams (eds.), Fluid Flow
in Sedimentary Basins and Aquifers. Geological Society of London Special Publication 34, pp.141–156.
Price, M., Atkinson, T. C., Barker, J. A., Wheeler, D. and Monkhouse, R. A. (1992), A Tracer Study of the
Danger Posed to A Chalk Aquifer by Contaminated Highway Run-Off, Proceedings of the Institution of
Civil Engineers-Water Maritime and Energy 96(1), 9–18.
BIBLIOGRAPHY 278
Price, M., Atkinson, T. C., Wheerler, D., Barker, J. A. and Monkhouse, R. A. (1989), The Movement
of Groundwater in the Chalk Aquifer near Bricket Wood, Hertfordshire, and its Possible Pollution
by Drainage from the M25 Motorway. British Geological Survey Report WD/89/3., Technical Report,
British Geological Survey.
Price, M., Bird, M. J. and Foster, S. S. D. (1976), Chalk pore-size measurements and their significance,
Water Services 80(968), 596–600.
Price, M., Downing, R. A. and Edmunds, W. M. (1993), The Chalk as an aquifer, in R. A. Downing,
M. Price and G. P. Jones (eds.), The Hydrogeology of the Chalk of North-West Europe, pp.35–58.
Price, M., Low, R. G. and McCann, C. (2000), Mechanisms of water storage and flow in the unsaturated
zone of the Chalk aquifer, Journal of Hydrology 233(1-4), 54–71.
Price, M., Morris, B. and Robertson, A. (1982), A study of intergranular and fissure permeability in Chalk
and Permian aquifers, using double-packer injection testing, Journal of Hydrology 54(4), 401–423.
Rawson, P. F., Allen, P. M. and Gale, A. S. (2001), The Chalk Group - a revised lithostratigraphy., Geosci-
entist 11, 21.
Roberts, M. (2001), The Sandridge/Hatfield Bromate Pollution Investigation, Technical Report, Environ-
ment Agency.
Rouleau, A. and Gale, J. E. (1985), Statistical characterisation of the fracture system in the Stripa Granite,
Sweden, Int. J. Rock Mech. Min. Sci. Geomech. Abstr. 22, 353–367.
Shand, P., Tyler-Whittle, R., Besien, T., Peach, D. W., Lawrence, A. R. and Lewis, H. O. (2003), Baseline
Report Series: 6. The Chalk of the Colne and Lee river catchments, Technical Report British Geological
Survey Commissioned Report CR/03/069N. Environment Agency National Groundwater Contaminated
Land Centre Technical Report NC/99/74/6, British Geological Survey and Environment Agency.
Silgram, M., Williams, A., Waring, R., Neumann, I., Hughes, A., Mansour, M. and Besien, T. (2005),
Effectiveness of the nitrate sensitive areas scheme in reducing groundwater concentrations in England,
Quarterly Journal of Engineering Geology and Hydrogeology 38, 117–127.
Smith, D., Wearn, P., Richards, H. and Rowe, P. (1970), Water movement in the unsaturated zone of high
and low permeabilitystrata using natural tritium., in Isotope Hydrology, Atomic Energy Agency, Vienna,
pp.73–87.
Snow, D. T. (1968), Rock fracture spacing, opening and porosities, Journal of the Soil and Foundations
Division, Proceedings of the American Society of Civil Engineers 94, 19.
STATS (1983a), Interim report on site investigation, ref 83/3105., Technical Report, St Albans Testing
Services.
BIBLIOGRAPHY 279
STATS (1983b), Second report on site investigation, ref 83/3105A, Technical Report, St Albans Testing
Services.
STATS (1983c), Third report on site investigation, ref 83/3105C, Technical Report, St Albans Testing
Services.
STATS (1984), Report on further site investigations. ref AM/3554, Technical Report, St Albans Testing
Services.
Thomas, J. (2000), Note on the 1984 and 1985 modelling studies., Technical Report, Environment Agency.
Walsh, P. and Ockenden, A. (1982), Hydrogeological observations at the Water End swallow hole complex,
North Mimms, Hertfordshire., Transactions of the British Cave Research Association. 9, 184–194.
Ward, R. S. (1989), Artificial tracer and natural 222Rn studies of the East Anglian Chalk aquifer., PhD
thesis, University of East Anglia, Norwich.
Ward, R. S., Williams, A. T., Barker, J. A., Brewerton, L. J. and Gale, I. N. (1998), Groundwater tracer
tests: a review and guidelines for their use in British aquifers. RD Technical Report W160, Technical
Report, Environment Agency. NOT IN FILE.
Watson, S. J. (2004), Solute transport and hydrodynamic characteristics in the Chalk aquifer at Tilmanstone,
Kent, PhD, University College London.
White, W. B. (2003), Conceptual models for karstic aquifers., Speleogenesis and Evolution of Karst
Aquifers 1(1), 1–6.
Whittaker, H. (1921), Water Supply of Buckinghamshire and Hertfordshire. Geological Survey Memoirs.,
Technical Report, British Geological Survey.
Williams, A., Barker, J. A., Silgram, M., Mansour, M., Neumann, I. and Hughes, A. (2003), The use of
flow modelling to assess the impact of agricultural control measures on abstracted groundwater quality.,
in MODFLOW and More. Understanding through Modelling., pp.446–450.
Williams, A., Bloomfield, J., Griffiths, K. and Butler, A. (2006), Characterising the vertical variations in
hydraulic conductivity within the Chalk aquifer, Journal of Hydrology 330(1-2), 53–62.
Williams, R. (1987), Frost weathered mantles on the Chalk., in J. Boardman (ed.), Periglacial processes
and landforms in Britain and Ireland, Cambridge University Press, Cambridge, pp.127–133.
Woods, M. A. (2003), The Stratigraphy of the Chalk Group of the Hertford District: a preliminary study
for Veolia Water Partnership. BGS Report Ref.CR/03/216, Technical Report Ref.CR/03/216, British
Geological Survey.
Woods, M. A. (2006), UK Chalk Group stratigraphy (Cenomanian-Santonian) determined from borehole

geophysical logs, Quarterly Journal of Engineering Geology and Hydrogeology 39, 83–96.
BIBLIOGRAPHY 280
Woods, M. A. and Aldiss, D. T. (2004), The stratigraphy of the Chalk Group of the Berkshire Downs,
Proceedings of the Geologists Association 115, 249–265.
Worthington, S. . (2003), A comprehensive strategy for understanding flow in carbonate aquifers., Speleo-
genesis and Evolution of Karst Aquifers 1(1), 1–8.
Worthington, S., Ford, D. C., Beddows, P. A. and Dreybrodt, W. (2000), Porosity and permeability en-
hancement in unconfined aquifers as a result of solution., in A. Klimchouk, D. C. Ford and A. N. Palmer
(eds.), Speleogenesis: Evolution of karst aquifers., National Speleological Society, Huntsville, Alabama,
pp.463–472.
Younger, P. L. (1989), Devensian Periglacial Influences on the Development of Spatially-Variable Perme-

ability in the Chalk of Southeast England, Quarterly Journal of Engineering Geology 22(4), 343–354.
Younger, P. L. and Elliot, T. (1995), Chalk fracture system characteristics: implications for flow and solute
transport, Quarterly Journal of Engineering Geology and Hydrogeology 28(Supplement1 ), S39 − −50.
Zheng, C. (1990), MT3D: A modular three-dimensional transport model for simulation of advection, dis-
persion, and chemical reactions of contaminants in groundwater systems. Report to the U.S. Environ-
mental Porectection Agency., Technical Report.
Zheng, C. and Wang, P. P. (1999), MT3DMS: A modular three-dimensional multispecies transport model
for simulation of advection, dispersion, and chemical reactions of contaminants in groundwater systems;
documentation and users guide. Contract Report SERDP-99-1., Technical Report, U.S. Army Engineer
Research and Development Center.
281
Appendix A
Chronology of key events

Appendix A
Chronology of key events
This chronology of key events in the history of bromate contamination in Hertfordshire has been com-
piled based on information obtained from the Jon Newton at the Environment Agency, Rob Sage at
VWTV and Philip Bishop at TWUL.
1955 Planning permission granted for use of the site for the manufacture of specified chemicals.
1955–1980 The site occupied by Steetly Chemical Works and operated for the production of bromine-
based chemicals, including bromate.
1982 The works are decommissioned.
1983 Crest Nicholson Residential Ltd (Crest) purchase the factory site and adjoining land.
1984 Crest are granted planning permission for 30 houses and commence demolition and clearance of
buildings and hardstanding from the site.
1983–1985 Intrusive investigations on behalf of Crest. The site is found to be contaminated with or-
ganic bromide compounds. The potential for bromide pollution of groundwater is recognised and
assessed by consultants to Crest.
1985–1986 Crest granted increased planning permission for 66 houses on the site. Approval requires
the removal of contaminated soil to minimum of 1 m depth across the entire site. Excavations
completed by August 1986.
1986–1987 Construction of the residential development St Leonard’s Court (SLC) begins in November
1986 and is completed by October 1987.
1998–1999 VWTV detect bromate at Essendon PWS at concentrations of 10.8 µg l−1 in December
1999. The EA are informed.
2000 VWTV detect bromate concentrations of ∼100 µg l−1 , well in excess of the future drinking water
standard of 10 µg l−1 , at Hatfield PWS. The EA are informed and abstractions for public supply
are ceased. A sampling programme is undertaken that identifies St Leonard’s Court as the source
of the contamination. St Albans District council (SADC) commission intrusive investigations at
the SLC site. The bromate groundwater monitoring programme begins to establish the extent and
migration of the contamination.
2002 The monitoring network is expanded and special investigations of surface waters are commenced,
including the Ellenbrook and the River Colne. SLC is designated as ‘Contaminated Land’ based
upon the ‘significant pollutant linkage’ between bromate and bromide in the unsaturated zone
and groundwater in the underlying Chalk aquifer. The site is adopted as a ‘special site’ by the
Environment Agency (EA). The EA begin consultation and investigation as to the determination
of the ‘appropriate persons’.
2003–2004 Bromate concentrations continue to increase at Essendon PWS and the NNR wells, threat-
ening available water resources for pulic supply.
2005 Pumping trial is commenced at Hatfield PWS to assess the possibility of scavenge pumping to
protect down-gradient sources at Essendon PWS and the NNR wells. The EA issue a remediation
notice to two appropriate persons: Redland Minerals Ltd for the Bromate contamination, and Crest
Nicholson Residential Ltd for the Bromide contamination. Both parties appeal.
2007 Appeals to the remediation notice are heard by public inquiry. Scavenge pumping of the Hatfield
source is promoted as an interim remediation measure and is incorporated in the Inspectors Re-
mediation Notice. An abstraction license is granted by the EA to VWTV for up to 9 Ml/d for the
purposes of groundwater remediation only.
2009 Decision is reached by the Secretary of State to uphold a modified remediation notice against
Redland Minerals Ltd and against Crest Nicholson Residential Ltd. Both parties apply for judicial
review.
2010 The judicial review case is dismissed.

282
Appendix B
Business Case for the research

Appendix B
Business Case for the research
The bromate contamination plume represents a major threat to the long-term quality of a number of
strategic public water supply sources, and also many private supply sources. The main driver for the
research conducted so far, and the funding for the additional research, is the impact of the bromate
contamination on water quality and resource availability.
Bromate concentrations above the regulatory standard for drinking water of 10 µg l−1 bromate
(UK Water Supply (Water Quality) Regulations 2000) are exceeded, or close to exceedence in major
PWS sources operated by both VWP and TWUL. Elevated bromate concentrations in PWS sources are
resulting in significant cost (∼£2million since May 2000) to the Water Companies. These costs include:
• The loss of large 9ML/d source since May 2000 when Hatfield Bishops Rise was taken out of
supply;
• The cost of blending at Essendon. The future use of Essendon may be restricted depending on
future blending arrangements.
• The cost of extra treatment required from the NNR sources/River Lee at Hornsey WTW.
The impact on water resources has lead Veolia Water Partnership (VWP) to plan to install two al-
ternative sources outside of the plume area. These boreholes are likely to replace Hatfield and Essendon
by December 2008. The total cost of the relocation amounts to ∼£8 million. Additional bromate mitiga-
tion and treatment measures are being considered by TWUL to safeguard the continued use of the NNR
sources, which will entail further expenditure. The costs of these impacts of the bromate contamination
are being largely borne by the water companies (and ultimately the customers). Although the Contam-
inated Land regime being enforced by the Environment Agency may result in the identified ‘polluter’
being liable for costs, this process in likely to take many years and the outcome is uncertain.
Water companies are particularly concerned that the future movement of the plume may affect
additional large PWS sources. In order to evaluate the most appropriate strategies for the current and
future management of groundwater quality, it is essential that additional research focusing on bromate
behaviour and movement in the aquifer is undertaken. This will allow the water companies and the
Environment Agency to refine their understanding of the contamination within the aquifer and to make
predictions as to the future evolution of the plume.
For example, a better understanding of the nature of groundwater flow in the area, will help to
characterise and quantify the effect of pumping the Hatfield borehole has on concentrations at other key
locations (particularly Essendon and NNR sources). This will allow the feasibility of using scavenge
pumping as a means of controlling concentrations at these sources to be evaluated.
283
Appendix C
Hatfield Pumping Trial statistical analyses

Essendon bromate versus Hatfield abstraction
Analysis of Variance
Regression Analysis: Bromate (ug/L) versus T
The regression equation is Source DF SS MS F P
Bromate as BrO3 (ug/l) = 30.1 - 1.43 H(T) Regression 1 6207.5 6207.5 205.41 0.000
Residual Error 333 10063.4 30.2
Total 334 16270.9
335 cases used, 945 cases contain missing values
Predictor Coef SE Coef T P

Regression Analysis: Bromate as BrO3 (ug/l) versus H(T-3)
Constant 30.0819 0.6198 48.54 0.000
H(T) -1.4257 0.1253 -11.38 0.000 The regression equation is
Bromate as BrO3 (ug/l) = 30.0 - 1.41 H(T-3)
S = 5.93134 R-Sq = 28.0% R-Sq(adj) = 27.8%

Source DF SS MS F P Constant 30.0122 0.5830 51.48 0.000
Regression 1 4555.7 4555.7 129.49 0.000 H(T-3) -1.4105 0.1160 -12.16 0.000
Total 334 16270.9
S = 5.81719 R-Sq = 30.7% R-Sq(adj) = 30.5%
Regression Analysis: Bromate as BrO3 (ug/l) versus H(T-1) Analysis of Variance

Bromate as BrO3 (ug/l) = 29.9 - 1.44 H(T-1) Regression 1 5002.3 5002.3 147.82 0.000
Total 334 16270.9
Constant 29.8580 0.5608 53.24 0.000 The regression equation is
H(T-1) -1.4417 0.1155 -12.48 0.000 Bromate as BrO3 (ug/l) = 29.8 - 1.31 H(T-4)
S = 5.76994 R-Sq = 31.9% R-Sq(adj) = 31.7% 335 cases used, 945 cases contain missing values
Analysis of Variance Predictor Coef SE Coef T P

Constant 29.8401 0.6273 47.57 0.000
Source DF SS MS F P H(T-4) -1.3084 0.1211 -10.80 0.000
Regression 1 5184.6 5184.6 155.73 0.000
Total 334 16270.9 S = 6.01551 R-Sq = 25.9% R-Sq(adj) = 25.7%
Source DF SS MS F P
The regression equation is Regression 1 4220.9 4220.9 116.64 0.000
Bromate as BrO3 (ug/l) = 30.5 - 1.58 H(T-2) Residual Error 333 12050.1 36.2
Total 334 16270.9
Predictor Coef SE Coef T P Essendon Bromate V Hatfield Abstraction (T-2)
Constant 29.7661 0.6775 43.94 0.000
H(T-5) -1.2757 0.1308 -9.76 0.000 Normal Probability Plot Versus Fits
99.9
99 4
S = 6.11821 R-Sq = 22.3% R-Sq(adj) = 22.0%
90
2
Analysis of Variance 50
0
Percent
Source DF SS MS F P 10
-2
Regression 1 3562.6 3562.6 95.17 0.000 1
Standardized Residual
0.1 -4
Total 333 15990.2 -4 -2 0 2 4 15 20 25 30
Standardized Residual Fitted Value
Regression Analysis: Bromate as BrO3 (ug/l) versus H(T-6) Histogram Versus Order
80
The regression equation is 4
60
2
40
333 cases used, 947 cases contain missing values 0
Frequency
20 -2
Predictor Coef SE Coef T P 0 -4
Constant 28.6746 0.6490 44.19 0.000 -3.0 -1.5 0.0 1.5 3.0 4.5 1 00 00 00 00 00 00 00 00 00 00 00 00
H(T-6) -1.0912 0.1281 -8.52 0.000 Standardized Residual
1 2 3 4 5 6 7 8 9 10 11 12
Observation Order
S = 6.26975 R-Sq = 18.0% R-Sq(adj) = 17.7%
Residuals Versus Date

Analysis of Variance (response is Bromate as BrO3 (ug/l))
Regression 1 2851.0 2851.0 72.53 0.000
Total 332 15862.6 4
3
2
The regression equation is
Bromate as BrO3 (ug/l) = 28.4 - 1.08 H(T-7) 1
0
-1
Constant 28.4450 0.6316 45.04 0.000 -2

H(T-7) -1.0846 0.1288 -8.42 0.000
-3
S = 6.28231 R-Sq = 17.6% R-Sq(adj) = 17.4%
-4
01/01/2006 01/01/2007 01/01/2008 01/01/2009
Date
Source DF SS MS F P
Regression 1 2798.9 2798.9 70.92 0.000
Total 332 15862.6
Residuals Versus SMD Essendon Bromate V Hatfield Abstraction
(response is Bromate as BrO3 (ug/l)) Bromate as BrO3 (ug/l) = 30.47 - 1.582 H(T-2)
5 60 Regression
95% CI
4 95% PI
50
3 S 5.49730
R-Sq 38.2%
2 40 R-Sq(adj) 38.0%
1
30
0
-1 20
Bromate as BrO3 (ug/l)
-2
10
-3
-4 0
0 20 40 60 80 100 120 140 0 1 2 3 4 5 6 7 8 9
SMD H(T-2)
SMD vs Date
140
Fitted Line Plot
120 Bromate as BrO3 (ug/l) = 30.47 - 1.582 H(T-2)
60 Regression
100 95% CI
95% PI
50
80 S 5.49730
R-Sq 38.2%
40 R-Sq(adj) 38.0%
60
40 30
20
6140 - Chilterns - East - Colne

20
0
10
01/01/2005 01/01/2006 01/01/2007 01/01/2008 01/01/2009
Date
0
0 1 2 3 4 5 6 7 8 9
H(T-2)
Regression Analysis: Bromate as BrO3 (ug/l) versus H(T-2), SMD - 6140

Bromate as BrO3 (ug/l) = 27.8 - 1.66 H(T-2) + 0.0543 SMD - 6140 (response is Bromate as BrO3 (ug/l))
5
4
Constant 27.7737 0.5847 47.50 0.000 3
H(T-2) -1.6599 0.1008 -16.47 0.000
SMD - 6140 0.054310 0.006448 8.42 0.000
2
S = 4.99745 R-Sq = 49.0% R-Sq(adj) = 48.7% 1
Source DF SS MS F P -1
Regression 2 7979.4 3989.7 159.75 0.000
Total 334 16270.9 -2
-3
Source DF Seq SS
H(T-2) 1 6207.5 01/01/2006 01/01/2007 01/01/2008 01/01/2009
SMD - 6140 1 1771.9
Date
Essendon Bromate V Hatfield Abstraction & SMD-Colne

Normal Probability Plot Versus Fits
99.9
99 4
90
2
50
0
Percent
10
-2
1
0.1 -4
-4 -2 0 2 4 15 20 25 30 35
Histogram Versus Order
4
60
45 2
30 0
Frequency
15 -2
-4
0
-3 -2 -1 0 1 2 3 4 1 00 00 00 00 00 00 00 00 00 00 00 00
1 2 3 4 5 6 7 8 9 10 11 12
Observation Order
Amwell Hill bromate versus Hatfield abstraction
S = 5.04850 R-Sq = 1.3% R-Sq(adj) = 0.6%
Regression Analysis: Bromate as BrO3 (ug/l) versus H(T)
The regression equation is Analysis of Variance

Bromate as BrO3 (ug/l) = 9.81 - 0.246 H(T)
Source DF SS MS F P
Regression 1 45.04 45.04 1.77 0.186
140 cases used, 1140 cases contain missing values Residual Error 136 3466.27 25.49
Total 137 3511.32

Constant 9.8098 0.7629 12.86 0.000
H(T) -0.2463 0.1530 -1.61 0.110 Regression Analysis: Bromate as BrO3 (ug/l) versus H(T-3)

S = 5.04930 R-Sq = 1.8% R-Sq(adj) = 1.1% Bromate as BrO3 (ug/l) = 9.70 - 0.192 H(T-3)
Analysis of Variance 137 cases used, 1143 cases contain missing values
Source DF SS MS F P
Regression 1 66.10 66.10 2.59 0.110 Predictor Coef SE Coef T P
Residual Error 138 3518.36 25.50 Constant 9.6970 0.7308 13.27 0.000
Total 139 3584.46 H(T-3) -0.1923 0.1478 -1.30 0.196
S = 5.01945 R-Sq = 1.2% R-Sq(adj) = 0.5%


Source DF SS MS F P
Regression 1 42.64 42.64 1.69 0.196
Total 136 3443.94

Constant 9.6549 0.7078 13.64 0.000 Regression Analysis: Bromate as BrO3 (ug/l) versus H(T-4)
H(T-1) -0.2262 0.1479 -1.53 0.128
S = 5.05383 R-Sq = 1.7% R-Sq(adj) = 1.0%

Source DF SS MS F P Predictor Coef SE Coef T P

Regression 1 59.78 59.78 2.34 0.128 Constant 10.1753 0.7859 12.95 0.000
Residual Error 138 3524.69 25.54 H(T-4) -0.2632 0.1485 -1.77 0.079
Total 139 3584.46
S = 4.97687 R-Sq = 2.3% R-Sq(adj) = 1.6%

Bromate as BrO3 (ug/l) = 9.58 - 0.194 H(T-2) Source DF SS MS F P
Regression 1 77.86 77.86 3.14 0.079
138 cases used, 1142 cases contain missing values Total 134 3372.17
Predictor Coef SE Coef T P Regression Analysis: Bromate as BrO3 (ug/l) versus H(T-5)
Constant 9.5796 0.6867 13.95 0.000
H(T-2) -0.1943 0.1462 -1.33 0.186 The regression equation is
Total 127 3015.83
Predictor Coef SE Coef T P Residual Plots for Bromate as BrO3 (ug/l)

Constant 10.5789 0.7819 13.53 0.000
H(T-5) -0.3325 0.1514 -2.20 0.030
Residuals from Bromate as BrO3 (ug/l) vs Date
S = 4.91246 R-Sq = 3.6% R-Sq(adj) = 2.8%
Residuals from Bromate as BrO3 (ug/l) vs 6600 - Lee Chalk

Amwell Hill Bromate V Hatfield Abstraction (T-5)
Source DF SS MS F P
Regression 1 116.44 116.44 4.82 0.030 Normal Probability Plot Versus Fits
99.9 2
Total 131 3253.63
99
90 1
50 0
Percent
10 -1
1
0.1 -2
130 cases used, 1150 cases contain missing values -4 -2 0 2 4 8 9 10 11
Predictor Coef SE Coef T P Histogram Versus Order

Constant 10.5225 0.7461 14.10 0.000
H(T-6) -0.3107 0.1502 -2.07 0.041 16 2
12 1
S = 4.86986 R-Sq = 3.2% R-Sq(adj) = 2.5%
8 0
Frequency
4 -1
0 -2
Source DF SS MS F P
-1.50 -0.75 0.00 0.75 1.50 1 00 00 00 00 00 00 00 00 00 00 00 00
Regression 1 101.44 101.44 4.28 0.041 1 2 3 4 5 6 7 8 9 10 11 12
Observation Order
Total 129 3137.03


Constant 10.0003 0.7198 13.89 0.000
H(T-7) -0.1658 0.1537 -1.08 0.283
S = 4.86991 R-Sq = 0.9% R-Sq(adj) = 0.1%
Source DF SS MS F P
Residuals Versus Date Amwell Hill Bromate V Hatfield Abstraction (T-5)
2 Regression
20 95% CI
95% PI
S 4.91246
1 15 R-Sq 3.6%
R-Sq(adj) 2.8%
10
0
5
-1
0
-2
01/01/2006 01/01/2007 01/01/2008 01/01/2009 0 1 2 3 4 5 6 7 8 9
Date H(T-5)
Residuals Versus SMD Regression Analysis: Log(BrO3) versus H(T-5)

(response is Bromate as BrO3 (ug/l)) The regression equation is
Log(BrO3) = 0.971 - 0.0255 H(T-5)
2
1 Predictor Coef SE Coef T P

Constant 0.97134 0.05276 18.41 0.000
H(T-5) -0.02553 0.01021 -2.50 0.014
0 S = 0.331481 R-Sq = 4.6% R-Sq(adj) = 3.9%
-1 Source DF SS MS F P
Regression 1 0.6864 0.6864 6.25 0.014
Residual Error 130 14.2844 0.1099
Total 131 14.9708
-2
0 20 40 60 80 100 120 140
6600 - Lee Chalk
log10(Amwell Hill Bromate) V Hatfield Abstraction (T-5)
Residuals Versus SMD
Normal Probability Plot Versus Fits (response is Log(BrO3))
99.9
99 1
90 1
0
50
-1
Percent
10
-2
1
0
0.1 -3
-4 -2 0 2 4 0.75 0.80 0.85 0.90 0.95
-1
20 1
15 0 -2
10 -1
Frequency
5 -2
-3
-3
0
-2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1 00 00 00 00 00 00 00 00 00 00 00 00 0 20 40 60 80 100 120 140
1 2 3 4 5 6 7 8 9 10 11 12
Standardized Residual 6600 - Lee Chalk
Observation Order
Residuals Versus Date Log10(Amwell Hill Bromate) V Hatfield Abstraction (T-5)

(response is Log(BrO3)) Bromate as BrO3 (ug/l) = 10.58 - 0.3325 H(T-5)
Regression
20 95% CI
1
95% PI
S 4.91246
15 R-Sq 3.6%
0 R-Sq(adj) 2.8%
10
-1
-2
0
-3
01/01/2006 01/01/2007 01/01/2008 01/01/2009 0 1 2 3 4 5 6 7 8 9
Date H(T-5)
Fitted Line: Bromate as BrO3 (ug/l) versus H(T-5)

Regression Analysis: Log(BrO3) versus H(T-5), 6600 - Lee Chalk

Log(BrO3) = 1.17 + 0.00429 H(T-5) - 0.00617 6600 - Lee Chalk (response is Log(BrO3))
3
Predictor Coef SE Coef T P 2

Constant 1.16628 0.03976 29.34 0.000
H(T-5) 0.004287 0.007452 0.58 0.566
6600 - Lee Chalk -0.0061678 0.0005116 -12.06 0.000
1
S = 0.228179 R-Sq = 55.1% R-Sq(adj) = 54.4%

0
Source DF SS MS F P -1
Regression 2 8.2543 4.1272 79.27 0.000
Total 131 14.9708
-2
Source DF Seq SS
H(T-5) 1 0.6864 01/01/2006 01/01/2007 01/01/2008 01/01/2009
6600 - Lee Chalk 1 7.5679
Date
Log10(Amwell Hill Bromate) V Hatfield Abstraction (T-5) & SMD-Lee

99.9
99 2
90 1
50 0
Percent
10 -1
1 -2
0.1
-4 -2 0 2 4 0.4 0.6 0.8 1.0 1.2

24 2
18 1
12 0
Frequency
-1
6
-2
0
-2.4 -1.6 -0.8 0.0 0.8 1.6 2.4 1 00 00 00 00 00 00 00 00 00 00 00 00
1 2 3 4 5 6 7 8 9 10 11 12
Observation Order
Hoddesdon bromate versus Hatfield abstraction S = 9.96880 R-Sq = 12.1% R-Sq(adj) = 11.6%
Regression Analysis: Bromate as BrO3 (ug/l) versus H(T) Analysis of Variance

Total 160 17978.0

Constant 30.092 1.430 21.05 0.000
H(T) -1.4977 0.2935 -5.10 0.000
S = 9.96038 R-Sq = 13.9% R-Sq(adj) = 13.4%

Constant 30.595 1.352 22.63 0.000
Source DF SS MS F P
H(T-3) -1.5435 0.2744 -5.63 0.000
Regression 1 2584.1 2584.1 26.05 0.000
Total 162 18556.7
S = 9.63218 R-Sq = 16.8% R-Sq(adj) = 16.2%

Total 158 17502.6

Constant 28.858 1.340 21.53 0.000
H(T-1) -1.3023 0.2885 -4.51 0.000
S = 10.1149 R-Sq = 11.2% R-Sq(adj) = 10.7%

Constant 32.913 1.401 23.49 0.000
Source DF SS MS F P
H(T-4) -1.8999 0.2673 -7.11 0.000
Regression 1 2084.5 2084.5 20.37 0.000
Total 162 18556.7
S = 9.16736 R-Sq = 24.6% R-Sq(adj) = 24.1%

Total 156 17271.8

Constant 28.900 1.281 22.57 0.000
H(T-2) -1.2927 0.2762 -4.68 0.000
Hoddesdon Bromate V Hatfield Abstraction (T-4)

Constant 31.891 1.358 23.48 0.000 Normal Probability Plot Versus Fits
H(T-5) -1.7775 0.2702 -6.58 0.000 99.9 4
99
S = 9.23033 R-Sq = 22.1% R-Sq(adj) = 21.5% 90 2
50
Percent
10
Source DF SS MS F P 1 -2
Regression 1 3688.3 3688.3 43.29 0.000
0.1
Residual Error 153 13035.5 85.2 -4 -2 0 2 4 15 20 25 30 35
Total 154 16723.7 Standardized Residual Fitted Value

4
30
20
0
Frequency
-2
0
Predictor Coef SE Coef T P -2 -1 0 1 2 3 1 00 00 00 00 00 00 00 00 00 00 00 00

Constant 31.456 1.314 23.94 0.000 Standardized Residual
1 2 3 4 5 6 7 8 9 10 11 12
H(T-6) -1.7190 0.2686 -6.40 0.000 Observation Order
S = 9.25704 R-Sq = 21.3% R-Sq(adj) = 20.8%

(response is Bromate as BrO3 (ug/l))
4
Source DF SS MS F P
Regression 1 3508.8 3508.8 40.95 0.000
Residual Error 151 12939.6 85.7 3
Total 152 16448.5
2

0
-1

Constant 31.699 1.351 23.46 0.000
H(T-7) -1.7299 0.2768 -6.25 0.000 -2
S = 9.27060 R-Sq = 20.8% R-Sq(adj) = 20.2% -3

01/01/2006 01/01/2007 01/01/2008 01/01/2009
Analysis of Variance Date
Source DF SS MS F P
Regression 1 3355.4 3355.4 39.04 0.000
Total 150 16161.1
Residuals Versus SMD Log(BrO3) = 1.48 - 0.0288 H(T-4)
4 157 cases used, 1123 cases contain missing values
3 Predictor Coef SE Coef T P

Constant 1.48051 0.02381 62.19 0.000
H(T-4) -0.028756 0.004541 -6.33 0.000
2
S = 0.155743 R-Sq = 20.6% R-Sq(adj) = 20.0%
1
0
Source DF SS MS F P
Regression 1 0.97253 0.97253 40.09 0.000
-1 Residual Error 155 3.75967 0.02426
Total 156 4.73220
-2
Residual Plots for Log(BrO3)
-3
0 20 40 60 80 100 120 140 Normal Probability Plot Versus Fits
99.9 3.0
6600 - Lee Chalk 99
90 1.5
Hoddesdon Bromate V Hatfield Abstraction (T-4) 50 0.0
Percent
-1.5
Regression 1
60 0.1 -3.0
95% CI
-4 -2 0 2 4 1.25 1.30 1.35 1.40 1.45
95% PI Standardized Residual Fitted Value
50 S 9.16736
R-Sq 24.6% Histogram Versus Order
40 R-Sq(adj) 24.1% 30 3.0
1.5
20
30 0.0
10
Frequency
-1.5
20
0 -3.0
-3 -2 -1 0 1 2 1 00 00 00 00 00 00 00 00 00 00 00 00

1 2 3 4 5 6 7 8 9 10 11 12
10 Standardized Residual
Observation Order
0
Regression Analysis: Bromate as BrO3 versus H(T-4), 6600 - Lee Chalk
0 1 2 3 4 5 6 7 8 9 The regression equation is
H(T-4) Bromate as BrO3 (ug/l) = 31.1 - 2.19 H(T-4) + 0.0602 6600 - Lee Chalk

Regression Analysis: Log(BrO3) versus H(T-4) Constant 31.082 1.450 21.43 0.000
H(T-4) -2.1915 0.2712 -8.08 0.000
6600 - Lee Chalk 0.06018 0.01717 3.51 0.001
S = 8.85068 R-Sq = 30.2% R-Sq(adj) = 29.2%
3
Regression 2 5208.3 2604.1 33.24 0.000
Total 156 17271.8
1
Source DF Seq SS
H(T-4) 1 4245.5 0
6600 - Lee Chalk 1 962.7
-1
Hoddesdon Bromate V Hatfield Abstraction (T-4), SMD-Lee
Normal Probability Plot Versus Fits -2
99.9
3.0
99
-3
90 1.5
01/01/2006 01/01/2007 01/01/2008 01/01/2009
50 0.0 Date
Percent
10
-1.5
1
0.1 -3.0 Residuals Versus SMD
-4 -2 0 2 4 15 20 25 30 35
Histogram Versus Order 3

40 3.0
30 1.5 2
20 0.0
1
Frequency
10 -1.5
0 -3.0
-2 -1 0 1 2 3 1 00 00 00 00 00 00 00 00 00 00 00 00 0
1 2 3 4 5 6 7 8 9 10 11 12
Observation Order
-1
-2
-3
0 20 40 60 80 100 120 140
6600 - Lee Chalk
Middlefield Road bromate versus Hatfield abstraction S = 7.86399 R-Sq = 4.1% R-Sq(adj) = 3.5%
Bromate as BrO3 (ug/l) = 20.1 - 0.599 H(T) Source DF SS MS F P
Regression 1 430.95 430.95 6.97 0.009
Residual Error 164 10142.15 61.84

Constant 20.107 1.115 18.03 0.000
H(T) -0.5987 0.2260 -2.65 0.009

Source DF SS MS F P
Total 166 10624.89 H(T-3) -0.4167 0.2211 -1.88 0.061
S = 7.94870 R-Sq = 2.1% R-Sq(adj) = 1.5%

Regression 1 224.34 224.34 3.55 0.061
Residual Error 163 10298.64 63.18

Constant 20.002 1.042 19.19 0.000
H(T-1) -0.6138 0.2193 -2.80 0.006

Source DF SS MS F P
Total 166 10624.89 H(T-4) -0.6717 0.2178 -3.08 0.002
S = 7.76930 R-Sq = 5.6% R-Sq(adj) = 5.0%


Source DF SS MS F P
Regression 1 573.85 573.85 9.51 0.002
Total 162 10292.13
Constant 19.7126 0.9846 20.02 0.000
H(T-2) -0.5575 0.2112 -2.64 0.009

Regression 1 857.65 857.65 14.38 0.000

H(T-5) -0.8018 0.2238 -3.58 0.000
S = 7.73725 R-Sq = 7.5% R-Sq(adj) = 6.9%

Source DF SS MS F P
Regression 1 768.65 768.65 12.84 0.000
Constant 21.110 1.034 20.41 0.000
H(T-8) -0.8979 0.2174 -4.13 0.000
Total 160 10287.19
S = 7.65848 R-Sq = 9.4% R-Sq(adj) = 8.9%

Regression 1 1000.2 1000.2 17.05 0.000

H(T-6) -0.7553 0.2291 -3.30 0.001
S = 7.79163 R-Sq = 6.5% R-Sq(adj) = 5.9%

Source DF SS MS F P
Regression 1 659.74 659.74 10.87 0.001
Constant 20.4131 0.9986 20.44 0.000
H(T-9) -0.7342 0.2090 -3.51 0.001
Total 158 10191.13
S = 7.76026 R-Sq = 7.0% R-Sq(adj) = 6.4%


Source DF SS MS F P
Regression 1 742.84 742.84 12.34 0.001
Total 165 10619.18

Constant 21.456 1.111 19.32 0.000
H(T-7) -0.8609 0.2270 -3.79 0.000
S = 7.72186 R-Sq = 8.5% R-Sq(adj) = 7.9%
Middlefield Rd Bromate V Hatfield Abstraction (T-8)
Normal Probability Plot Versus Fits (response is Bromate as BrO3 (ug/l))
99.9 4
99 4
90 2
50
3
0
Percent
10
1
2
-2
0.1
-4 -2 0 2 4 14 16 18 20 22
Standardized Residual Fitted Value 1

4
0
40
2
30 -1
20 0
Frequency
10 -2
-2
0
-2 -1 0 1 2 3 1 00 00 00 00 00 00 00 00 00 00 00 00 0 20 40 60 80 100 120 140
1 2 3 4 5 6 7 8 9 10 11 12
Observation Order
Residuals Versus Date Middlefield Rd Bromate V Hatfield Abstraction (T-8)

4 50 Regression
95% CI
95% PI
3
40 S 7.65848
R-Sq 9.4%
2 R-Sq(adj) 8.9%
30
1
20
0
10
-1
-2 0
01/01/2006 01/01/2007 01/01/2008 01/01/2009 0 1 2 3 4 5 6 7 8 9

Date H(T-8)
Regression Analysis: Log(BrO3) versus H(T)

Regression Analysis: Log(BrO3) versus H(T-8) Residuals Versus Date
The regression equation is (response is Log(BrO3))
Log(BrO3) = 1.28 - 0.0200 H(T-8)
3
1
Constant 1.28012 0.02699 47.43 0.000 0
H(T-8) -0.020048 0.005675 -3.53 0.001
-1
S = 0.199886 R-Sq = 7.1% R-Sq(adj) = 6.5%
-2
Analysis of Variance -3
Source DF SS MS F P
Regression 1 0.49866 0.49866 12.48 0.001 -4
Residual Error 164 6.55252 0.03995
Total 165 7.05119 -5
-6
Log10(Middlefield Rd Bromate) V Hatfield Abstraction (T-8)
01/01/2006 01/01/2007 01/01/2008 01/01/2009
Normal Probability Plot Versus Fits Date
99.9
2
99
90 0
50 -2 (response is Log(BrO3))
Percent
10 -4 3
1
0.1
-6 2
-6 -4 -2 0 2 1.10 1.15 1.20 1.25 1.30
Standardized Residual Fitted Value 1
Histogram Versus Order 0
40 2
-1
30 0
20 -2
-2
Frequency
10 -4 -3
-6
0
-4
-6.0 -4.5 -3.0 -1.5 0.0 1.5 1 00 00 00 00 00 00 00 00 00 00 00 00
1 2 3 4 5 6 7 8 9 10 11 12
Observation Order
-5
-6
0 20 40 60 80 100 120 140
6600 - Lee Chalk
412 4.39 0.0273 1.1694 0.0300 -1.1421 -5.78R
432 0.01 1.6854 1.2564 0.0377 0.4289 2.18R
Log10(Middlefield Rd Bromate) V Hatfield Abstraction (T-8) 860 7.80 0.4624 1.1209 0.0273 -0.6585 -3.32R
Log(BrO3) = 1.280 - 0.02005 H(T-8)
R denotes an observation with a large standardized residual.
1.8 Regression
95% CI
1.6 Residual Plots for Log(BrO3)
95% PI
1.4 S 0.199886
99.9
R-Sq 7.1% 2
99
1.2 R-Sq(adj) 6.5%
90 0
1.0 50 -2
Percent
0.8 10
-4
Log(BrO3)
1
0.6
0.1 -6
-6 -4 -2 0 2 1.10 1.15 1.20 1.25 1.30
0.4 Standardized Residual Fitted Value
0.2 Histogram Versus Order

40 2
0.0
30 0
0 1 2 3 4 5 6 7 8 9
H(T-8) 20 -2
Frequency
10 -4
Regression Analysis: Log(BrO3) versus H(T-8), 6600 - Lee Chalk -6

0
-6.0 -4.5 -3.0 -1.5 0.0 1.5 1 00 00 00 00 00 00 00 00 00 00 00 00
1 2 3 4 5 6 7 8 9 10 11 12
The regression equation is Standardized Residual
Log(BrO3) = 1.29 - 0.0191 H(T-8) - 0.000323 6600 - Lee Chalk Observation Order

Constant 1.29369 0.03101 41.72 0.000
H(T-8) -0.019106 0.005776 -3.31 0.001
6600 - Lee Chalk -0.0003230 0.0003625 -0.89 0.374
S = 0.200012 R-Sq = 7.5% R-Sq(adj) = 6.4%
Source DF SS MS F P
Regression 2 0.53042 0.26521 6.63 0.002
Residual Error 163 6.52077 0.04000
Total 165 7.05119
Source DF Seq SS
H(T-8) 1 0.49866
6600 - Lee Chalk 1 0.03176
Unusual Observations
Obs H(T-8) Log(BrO3) Fit SE Fit Residual St Resid

119 5.82 0.7005 1.1505 0.0240 -0.4500 -2.27R
(response is Log(BrO3))
3
-1
-2
-3
-4
-5
-6
01/01/2006 01/01/2007 01/01/2008 01/01/2009
Date
Rye Common versus Hatfield abstraction S = 6.64265 R-Sq = 0.3% R-Sq(adj) = 0.0%
Regression Analysis: Bromate as BrO3 (ug/l) versus H(T) Analysis of Variance

Total 161 7083.87

Constant 9.4817 0.9441 10.04 0.000
H(T) -0.0619 0.1917 -0.32 0.747
S = 6.65172 R-Sq = 0.1% R-Sq(adj) = 0.0%

Constant 9.5952 0.9208 10.42 0.000
Source DF SS MS F P
H(T-3) -0.0919 0.1898 -0.48 0.629
Regression 1 4.60 4.60 0.10 0.747
Total 161 7083.87
S = 6.64902 R-Sq = 0.1% R-Sq(adj) = 0.0%

Total 161 7083.87

Constant 9.9169 0.8838 11.22 0.000
H(T-1) -0.1817 0.1883 -0.96 0.336
S = 6.63461 R-Sq = 0.6% R-Sq(adj) = 0.0%

Constant 10.1653 0.9752 10.42 0.000
Source DF SS MS F P
H(T-4) -0.2124 0.1869 -1.14 0.257
Regression 1 40.99 40.99 0.93 0.336
Total 161 7083.87
S = 6.62719 R-Sq = 0.8% R-Sq(adj) = 0.2%

Total 161 7083.87

Constant 9.7070 0.8341 11.64 0.000
H(T-2) -0.1332 0.1810 -0.74 0.463
Total 160 7068.74

H(T-5) -0.2887 0.1899 -1.52 0.130
S = 6.61969 R-Sq = 1.4% R-Sq(adj) = 0.8%

Source DF SS MS F P
Regression 1 101.31 101.31 2.31 0.130
Constant 10.7462 0.9074 11.84 0.000
H(T-8) -0.3884 0.1935 -2.01 0.046
Total 160 7068.74
S = 6.58474 R-Sq = 2.5% R-Sq(adj) = 1.9%


Source DF SS MS F P
Regression 1 174.70 174.70 4.03 0.046
Total 160 7068.74

H(T-6) -0.2688 0.1893 -1.42 0.157
S = 6.62574 R-Sq = 1.3% R-Sq(adj) = 0.6%

Source DF SS MS F P
Regression 1 88.57 88.57 2.02 0.157
Constant 10.1108 0.8765 11.54 0.000
H(T-9) -0.2256 0.1848 -1.22 0.224
Total 160 7068.74
S = 6.63661 R-Sq = 0.9% R-Sq(adj) = 0.3%

Total 160 7068.74

Constant 11.0406 0.9371 11.78 0.000
H(T-7) -0.4355 0.1903 -2.29 0.023
S = 6.56048 R-Sq = 3.2% R-Sq(adj) = 2.6%
Source DF SS MS F P
Regression 1 225.39 225.39 5.24 0.023
Rye Common Bromate V Hatfield Abstraction
Residuals Versus SMD-6600
Normal Probability Plot Versus Fits (response is Bromate as BrO3 (ug/l))
99.9
99 3 4
90 2
50 1 3
Percent
10 0
1 -1 2
0.1
-4 -2 0 2 4 7 8 9 10 11
1
16 0
3
12 2
8 1 -1
0
Frequency
4
-1
-2
0
-1.50 -0.75 0.00 0.75 1.50 2.25 3.00 1 0 0 0 0 0 0 0 0 0 0 0 0 0 20 40 60 80 100 120 140
10 20 30 40 50 60 70 80 90 100 110 120
Observation Order
Regression Analysis: Log(BrO3) versus H(T-7)

Log(BrO3) = 0.867 - 0.0122 H(T-7)
4

3
2 Constant 0.86739 0.05813 14.92 0.000
H(T-7) -0.01220 0.01180 -1.03 0.303
1 S = 0.406946 R-Sq = 0.7% R-Sq(adj) = 0.0%
0 Analysis of Variance
Source DF SS MS F P
-1 Regression 1 0.1769 0.1769 1.07 0.303
Residual Error 159 26.3312 0.1656
Total 160 26.5081
-2
01/01/2006 01/01/2007 01/01/2008 01/01/2009
Date
Log10(Rye Common Bromate) V Hatfield Abstraction
99.9 2 Regression Analysis: Bromate as BrO3 versus H(T-7), 6600 - Lee Chalk
99
1
90 The regression equation is
0
50 Bromate as BrO3 (ug/l) = 11.8 - 0.350 H(T-7) - 0.0206 6600 - Lee Chalk
-1
Percent
10
-2
1
0.1
-4 -2 0 2 4 0.750 0.775 0.800 0.825 0.850
Constant 11.786 1.034 11.40 0.000
2
20
H(T-7) -0.3499 0.1961 -1.78 0.076
1 6600 - Lee Chalk -0.02063 0.01239 -1.66 0.098
15
0
10 -1
Frequency
5 -2 S = 6.52423 R-Sq = 4.9% R-Sq(adj) = 3.7%
0
-2.25 -1.50 -0.75 0.00 0.75 1.50 1 0 0 0 0 0 0 0 0 0 0 0 0
10 20 30 40 50 60 70 80 90 100 110 120
Observation Order
Source DF SS MS F P
Residuals Versus Date Regression 2 343.38 171.69 4.03 0.020
(response is Log(BrO3)) Residual Error 158 6725.37 42.57
2 Total 160 7068.74
1
Source DF Seq SS
H(T-7) 1 225.39
6600 - Lee Chalk 1 117.98
0
-1
Bromate
as BrO3
Obs H(T-7) (ug/l) Fit SE Fit Residual St Resid
-2
278 0.00 27.300 11.699 1.012 15.601 2.42R
286 0.41 33.200 11.554 0.963 21.646 3.35R
307 0.01 25.406 11.210 0.936 14.196 2.20R
-3
01/01/2006 01/01/2007 01/01/2008 01/01/2009 R denotes an observation with a large standardized residual.
Date
Rye Common Bromate V Hatfield Abstraction & SMD-Lee

Residuals Versus 6600 - Lee Chalk Normal Probability Plot Versus Fits
(response is Log(BrO3)) 99.9
99 3
2
90 2
50 1
Percent
1 0
10
1
-1
0.1
0 -4 -2 0 2 4 5.0 7.5 10.0 12.5

-1 40
3
30 2
-2 20 1
0
Frequency
10
-1
0
-3
-1.6 -0.8 0.0 0.8 1.6 2.4 3.2 1 00 00 00 00 00 00 00 00 00 00 00 00
0 20 40 60 80 100 120 140 Standardized Residual
1 2 3 4 5 6 7 8 9 10 11 12
6600 - Lee Chalk Observation Order
Turnford bromate versus Hatfield abstraction
Residuals Versus Date Regression Analysis: Bromate as BrO3 (ug/l) versus H(T)
4 The regression equation is
3

2

1
Constant 21.041 1.279 16.45 0.000
H(T) -1.0437 0.2567 -4.07 0.000
0
S = 8.92022 R-Sq = 9.3% R-Sq(adj) = 8.7%
-1
-2 Analysis of Variance
01/01/2006 01/01/2007 01/01/2008 01/01/2009
Date Source DF SS MS F P
Regression 1 1315.0 1315.0 16.53 0.000
Residuals Versus SMD-6600 Total 163 14205.4
4
3 Bromate
as BrO3
Obs H(T) (ug/l) Fit SE Fit Residual St Resid
2 5 0.03 40.813 21.010 1.272 19.803 2.24R
12 0.23 39.123 20.801 1.230 18.322 2.07R
1 286 0.00 43.500 21.041 1.279 22.459 2.54R
300 0.01 40.816 21.030 1.277 19.786 2.24R
307 1.84 45.790 19.121 0.919 26.669 3.01R
0 319 2.99 37.711 17.920 0.760 19.791 2.23R
495 0.00 0.600 21.038 1.278 -20.438 -2.32R
-1 1146 4.80 34.400 16.035 0.714 18.365 2.07R

-2
0 20 40 60 80 100 120 140
6600 - Lee Chalk Regression Analysis: Bromate as BrO3 (ug/l) versus H(T-1)


Constant 20.435 1.190 17.18 0.000
H(T-1) -0.9752 0.2501 -3.90 0.000
S = 8.95332 R-Sq = 8.6% R-Sq(adj) = 8.0%
Source DF SS MS F P
Regression 1 1219.1 1219.1 15.21 0.000
Total 163 14205.4
Unusual Observations 164 cases used, 1116 cases contain missing values
Bromate
as BrO3 Predictor Coef SE Coef T P
Obs H(T-1) (ug/l) Fit SE Fit Residual St Resid Constant 20.172 1.184 17.04 0.000
5 0.00 40.813 20.435 1.190 20.378 2.30R H(T-3) -0.8819 0.2407 -3.66 0.000
12 0.44 39.123 20.005 1.103 19.118 2.15R
271 0.65 38.100 19.801 1.062 18.299 2.06R
286 0.00 43.500 20.435 1.190 23.065 2.60R S = 8.99859 R-Sq = 7.7% R-Sq(adj) = 7.1%
300 0.37 40.816 20.074 1.116 20.742 2.33R
307 1.98 45.790 18.504 0.841 27.286 3.06R
319 3.00 37.711 17.509 0.731 20.202 2.26R Analysis of Variance
495 0.00 0.600 20.431 1.189 -19.831 -2.23R
1146 4.80 34.400 15.758 0.738 18.642 2.09R Source DF SS MS F P
Regression 1 1087.5 1087.5 13.43 0.000
R denotes an observation with a large standardized residual. Residual Error 162 13117.9 81.0
Total 163 14205.4

Bromate
as BrO3
5 0.01 40.813 20.163 1.182 20.650 2.31R
12 0.00 39.123 20.172 1.184 18.951 2.12R
271 0.01 38.100 20.163 1.182 17.937 2.01R
286 0.00 43.500 20.172 1.184 23.328 2.62R
300 0.01 40.816 20.163 1.182 20.653 2.32R
Constant 19.849 1.138 17.44 0.000
307 0.01 45.790 20.163 1.182 25.627 2.87R
H(T-2) -0.8673 0.2447 -3.54 0.001
319 3.00 37.711 17.526 0.740 20.185 2.25R
495 0.00 0.600 20.169 1.183 -19.569 -2.19R
1146 4.79 34.400 15.946 0.731 18.454 2.06R
S = 9.02099 R-Sq = 7.2% R-Sq(adj) = 6.6%
Source DF SS MS F P Regression Analysis: Bromate as BrO3 (ug/l) versus H(T-4)

Regression 1 1022.1 1022.1 12.56 0.001
Residual Error 162 13183.3 81.4 The regression equation is
Total 163 14205.4 Bromate as BrO3 (ug/l) = 22.1 - 1.24 H(T-4)
Unusual Observations 164 cases used, 1116 cases contain missing values
Bromate
as BrO3 Predictor Coef SE Coef T P
Obs H(T-2) (ug/l) Fit SE Fit Residual St Resid Constant 22.115 1.265 17.49 0.000
5 0.00 40.813 19.849 1.138 20.964 2.34R H(T-4) -1.2359 0.2426 -5.09 0.000
12 0.01 39.123 19.840 1.136 19.283 2.15R
271 0.00 38.100 19.849 1.138 18.251 2.04R
286 0.00 43.500 19.849 1.138 23.651 2.64R S = 8.69378 R-Sq = 13.8% R-Sq(adj) = 13.3%
300 0.01 40.816 19.840 1.136 20.976 2.34R
307 0.01 45.790 19.840 1.136 25.950 2.90R
495 0.00 0.600 19.846 1.138 -19.246 -2.15R
Regression 1 1961.1 1961.1 25.95 0.000
R denotes an observation with a large standardized residual. Residual Error 162 12244.2 75.6
Total 163 14205.4

Bromate
as BrO3
5 0.12 40.813 21.967 1.240 18.846 2.19R
286 0.00 43.500 22.115 1.265 21.385 2.49R S = 8.75548 R-Sq = 9.4% R-Sq(adj) = 8.8%
300 0.01 40.816 22.103 1.263 18.713 2.18R
307 0.01 45.790 22.103 1.263 23.687 2.75R
495 0.00 0.600 22.111 1.264 -21.511 -2.50R
530 0.00 4.026 22.111 1.264 -18.085 -2.10R Source DF SS MS F P
1133 5.04 33.300 15.892 0.696 17.408 2.01R Regression 1 1277.4 1277.4 16.66 0.000
1146 4.79 34.400 16.192 0.686 18.208 2.10R Residual Error 161 12342.0 76.7
Total 162 13619.4
Bromate
as BrO3
12 0.02 39.123 20.827 1.256 18.296 2.11R
286 0.00 43.500 20.847 1.260 22.653 2.61R
300 3.00 40.816 17.784 0.751 23.032 2.64R
307 0.01 45.790 20.837 1.258 24.953 2.88R
319 2.99 37.711 17.794 0.752 19.917 2.28R
495 0.01 0.600 20.841 1.259 -20.241 -2.34R
Constant 21.684 1.245 17.42 0.000
H(T-5) -1.1898 0.2420 -4.92 0.000
S = 8.57627 R-Sq = 13.1% R-Sq(adj) = 12.5% Regression Analysis: Bromate as BrO3 (ug/l) versus H(T-7)

Analysis of Variance Bromate as BrO3 (ug/l) = 21.3 - 1.15 H(T-7)
Source DF SS MS F P
Regression 1 1777.5 1777.5 24.17 0.000 163 cases used, 1117 cases contain missing values
Total 162 13619.4
Constant 21.315 1.261 16.90 0.000
Unusual Observations H(T-7) -1.1460 0.2547 -4.50 0.000
Bromate
as BrO3 S = 8.66839 R-Sq = 11.2% R-Sq(adj) = 10.6%
12 0.02 39.123 21.660 1.241 17.463 2.06R
300 1.09 40.816 20.387 1.032 20.429 2.40R
319 2.99 37.711 18.127 0.746 19.584 2.29R Regression 1 1521.7 1521.7 20.25 0.000
495 0.00 0.600 21.679 1.244 -21.079 -2.48R Residual Error 161 12097.7 75.1
1133 4.92 33.300 15.826 0.687 17.474 2.04R Total 162 13619.4

Regression Analysis: Bromate as BrO3 (ug/l) versus H(T-6) Bromate

as BrO3
12 0.03 39.123 21.280 1.255 17.843 2.08R
286 0.41 43.500 20.845 1.174 22.655 2.64R
300 3.00 40.816 17.877 0.742 22.939 2.66R
307 0.01 45.790 21.303 1.259 24.487 2.86R
319 3.00 37.711 17.877 0.742 19.834 2.30R

Constant 20.847 1.260 16.55 0.000
H(T-6) -1.0213 0.2502 -4.08 0.000
Regression Analysis: Bromate as BrO3 (ug/l) versus H(T-8) Unusual Observations
Bromate
as BrO3
12 0.00 39.123 20.243 1.142 18.880 2.17R
229 1.99 36.686 18.330 0.816 18.356 2.10R
271 0.00 38.100 20.243 1.142 17.857 2.06R
286 0.00 43.500 20.243 1.142 23.257 2.68R
300 0.01 40.816 20.233 1.140 20.583 2.37R
307 0.01 45.790 20.233 1.140 25.557 2.94R
Constant 20.296 1.176 17.26 0.000
319 3.00 37.711 17.360 0.716 20.351 2.33R
H(T-8) -0.9783 0.2479 -3.95 0.000
495 0.01 0.600 20.235 1.140 -19.635 -2.26R
1139 5.04 33.800 15.402 0.740 18.398 2.11R
S = 8.78231 R-Sq = 8.8% R-Sq(adj) = 8.3%
Turnford Bromate V Hatfield Abstraction
Source DF SS MS F P
Regression 1 1201.6 1201.6 15.58 0.000
99.9 3.0
Residual Error 161 12417.8 77.1 99
Total 162 13619.4 90 1.5
50 0.0
Percent
Unusual Observations 10
-1.5
1
Bromate 0.1 -3.0

-4 -2 0 2 4 10 15 20
as BrO3 Standardized Residual Fitted Value
12 0.00 39.123 20.296 1.176 18.827 2.16R Histogram Versus Order
229 3.00 36.686 17.361 0.719 19.325 2.21R 40 3.0
271 0.00 38.100 20.296 1.176 17.804 2.05R
30 1.5
278 0.65 37.600 19.660 1.049 17.940 2.06R
286 0.00 43.500 20.296 1.176 23.204 2.67R 20 0.0
300 1.97 40.816 18.368 0.830 22.448 2.57R
Frequency
10 -1.5
307 0.37 45.790 19.934 1.103 25.856 2.97R
0 -3.0
319 2.99 37.711 17.370 0.720 20.341 2.32R -2 -1 0 1 2 3 1 0 0 0 0 0 0 0 0 0 0 0 0

10 20 30 40 50 60 70 80 90 100 110 120
R denotes an observation with a large standardized residual. Observation Order
Regression Analysis: Bromate as BrO3 (ug/l) versus H(T-9) Residuals Versus Date
2
1

Constant 20.243 1.142 17.73 0.000 0
H(T-9) -0.9610 0.2364 -4.07 0.000
-1
S = 8.75886 R-Sq = 9.3% R-Sq(adj) = 8.7%

-2
-3
Source DF SS MS F P 01/01/2006 01/01/2007 01/01/2008 01/01/2009
Regression 1 1267.9 1267.9 16.53 0.000 Date
Total 162 13619.4
Constant 22.413 1.374 16.31 0.000
Residuals Versus SMD-6600 H(T-4) -1.1862 0.2588 -4.58 0.000
(response is Bromate as BrO3 (ug/l)) 6600 - Lee Chalk -0.00937 0.01673 -0.56 0.576
3
S = 8.71226 R-Sq = 14.0% R-Sq(adj) = 12.9%
2
1
Source DF SS MS F P
0 Regression 2 1984.89 992.45 13.08 0.000
Residual Error 161 12220.46 75.90
Total 163 14205.35
-1
-2 Source DF Seq SS
H(T-4) 1 1961.10
6600 - Lee Chalk 1 23.79
-3
0 20 40 60 80 100 120 140
6600 - Lee Chalk
Turnford Bromate V Hatfield Abstraction & SMD-Lee
Regression Analysis: Log(BrO3) versus H(T-4) 99.9 3.0
99
90 1.5
Log(BrO3) = 1.26 - 0.0244 H(T-4) 50 0.0
Percent
10
-1.5
1
164 cases used, 1116 cases contain missing values 0.1 -3.0
-4 -2 0 2 4 10 15 20
Predictor Coef SE Coef T P Histogram Versus Order

Constant 1.26135 0.03632 34.73 0.000 40 3.0
H(T-4) -0.024447 0.006967 -3.51 0.001
30 1.5
20 0.0
S = 0.249658 R-Sq = 7.1% R-Sq(adj) = 6.5%
Frequency
10 -1.5
0 -3.0
-2 -1 0 1 2 1 0 0 0 0 0 0 0 0 0 0 0 0
Analysis of Variance Standardized Residual
10 20 30 40 50 60 70 80 90 100 110 120
Observation Order
Source DF SS MS F P
Regression 1 0.76735 0.76735 12.31 0.001
Residual Error 162 10.09733 0.06233 Residuals Versus Date
Total 163 10.86468 (response is Bromate as BrO3 (ug/l))
3
2
Obs H(T-4) Log(BrO3) Fit SE Fit Residual St Resid
495 0.00 -0.2218 1.2613 0.0363 -1.4831 -6.00R 1
530 0.00 0.6049 1.2613 0.0363 -0.6564 -2.66R
R denotes an observation with a large standardized residual. 0
Regression Analysis: Bromate as BrO3 versus H(T-4), 6600 - Lee Chalk -1

The regression equation is -2

Bromate as BrO3 (ug/l) = 22.4 - 1.19 H(T-4) - 0.0094 6600 - Lee Chalk
-3
164 cases used, 1116 cases contain missing values 01/01/2006 01/01/2007 01/01/2008 01/01/2009
Date

Broxbourne bromate versus Hatfield abstraction
Residuals Versus SMD 6600 Fitted Line: Bromate as BrO3 (ug/l) versus H(T)
3
2
1
-1
Constant 25.024 1.186 21.10 0.000
-2 H(T) -1.4599 0.2415 -6.04 0.000
-3
S = 7.59608 R-Sq = 20.8% R-Sq(adj) = 20.2%
0 20 40 60 80 100 120 140
6600 - Lee Chalk
Source DF SS MS F P
Turnford Bromate V Hatfield Abstraction Regression 1 2107.6 2107.6 36.53 0.000
Bromate as BrO3 (ug/l) = 22.12 - 1.236 H(T-4) Residual Error 139 8020.4 57.7
50 Total 140 10128.0
Regression
95% C I
95% PI
40 Unusual Observations
S 8.69378
R-Sq 13.8%
R-Sq(adj) 13.3% Bromate
30
as BrO3
20 12 0.23 40.651 24.689 1.139 15.962 2.13R
19 0.20 41.973 24.732 1.145 17.241 2.30R
286 0.00 42.700 25.024 1.186 17.676 2.36R
10 307 1.84 47.352 22.338 0.846 25.014 3.31R
313 2.99 36.112 20.659 0.697 15.453 2.04R

0
403 5.81 0.665 16.542 0.757 -15.877 -2.10R
530 0.00 5.911 25.021 1.185 -19.110 -2.55R
571 2.01 6.900 22.096 0.821 -15.196 -2.01R
0 1 2 3 4 5 6 7 8 9 R denotes an observation with a large standardized residual.
H(T-4)


Constant 23.456 1.144 20.50 0.000
H(T-1) -1.1542 0.2405 -4.80 0.000
S = 7.90584 R-Sq = 14.2% R-Sq(adj) = 13.6%
Source DF SS MS F P
Regression 1 1440.2 1440.2 23.04 0.000
Total 140 10128.0 Regression Analysis: Bromate as BrO3 (ug/l) versus H(T-3)

Bromate
as BrO3
12 0.44 40.651 22.949 1.060 17.702 2.26R
19 0.17 41.973 23.260 1.111 18.713 2.39R
40 2.99 36.335 20.005 0.699 16.330 2.07R
Constant 23.593 1.107 21.31 0.000
271 0.65 39.300 22.706 1.021 16.594 2.12R
H(T-3) -1.1563 0.2236 -5.17 0.000
286 0.00 42.700 23.456 1.144 19.244 2.46R
300 0.37 39.185 23.029 1.073 16.156 2.06R
307 1.98 47.352 21.171 0.806 26.181 3.33R
S = 7.81728 R-Sq = 16.1% R-Sq(adj) = 15.5%
313 3.00 36.112 19.994 0.698 16.118 2.05R
400 5.15 0.897 17.512 0.733 -16.615 -2.11R
530 0.00 5.911 23.454 1.144 -17.543 -2.24R
571 0.00 6.900 23.454 1.144 -16.554 -2.12R
Source DF SS MS F P
Regression 1 1633.7 1633.7 26.73 0.000
Total 140 10128.0
The regression equation is Unusual Observations

Bromate
as BrO3
141 cases used, 1139 cases contain missing values Obs H(T-3) (ug/l) Fit SE Fit Residual St Resid
12 0.00 40.651 23.593 1.107 17.058 2.20R
19 0.01 41.973 23.582 1.106 18.391 2.38R
Predictor Coef SE Coef T P 40 2.99 36.335 20.136 0.695 16.199 2.08R
Constant 23.420 1.059 22.11 0.000 271 0.01 39.300 23.582 1.106 15.718 2.03R
H(T-2) -1.1981 0.2251 -5.32 0.000 286 0.00 42.700 23.593 1.107 19.107 2.47R
300 0.01 39.185 23.582 1.106 15.603 2.02R
307 0.01 47.352 23.582 1.106 23.770 3.07R
S = 7.77993 R-Sq = 16.9% R-Sq(adj) = 16.3% 313 3.00 36.112 20.124 0.694 15.988 2.05R
417 4.63 34.924 18.240 0.674 16.684 2.14R
530 0.00 5.911 23.591 1.107 -17.680 -2.28R
Analysis of Variance 571 0.00 6.900 23.590 1.107 -16.690 -2.16R
Source DF SS MS F P R denotes an observation with a large standardized residual.

Regression 1 1714.7 1714.7 28.33 0.000
Total 140 10128.0 Regression Analysis: Bromate as BrO3 (ug/l) versus H(T-4)

Bromate
as BrO3
12 0.01 40.651 23.408 1.058 17.243 2.24R
19 0.01 41.973 23.408 1.058 18.565 2.41R
40 3.00 36.335 19.826 0.674 16.509 2.13R
Constant 25.036 1.170 21.40 0.000
271 0.00 39.300 23.420 1.059 15.880 2.06R
H(T-4) -1.3929 0.2261 -6.16 0.000
286 0.00 42.700 23.420 1.059 19.280 2.50R
300 0.01 39.185 23.408 1.058 15.777 2.05R
307 0.01 47.352 23.408 1.058 23.944 3.11R
S = 7.56504 R-Sq = 21.5% R-Sq(adj) = 20.9%
313 2.99 36.112 19.837 0.674 16.275 2.10R
530 0.00 5.911 23.418 1.059 -17.507 -2.27R
571 0.00 6.900 23.416 1.059 -16.516 -2.14R
Source DF SS MS F P
Regression 1 2173.0 2173.0 37.97 0.000 The regression equation is
Residual Error 139 7954.9 57.2 Bromate as BrO3 (ug/l) = 24.7 - 1.40 H(T-6)
Total 140 10128.0

Bromate Predictor Coef SE Coef T P

as BrO3 Constant 24.723 1.157 21.36 0.000
Obs H(T-4) (ug/l) Fit SE Fit Residual St Resid H(T-6) -1.3998 0.2312 -6.05 0.000
12 0.02 40.651 25.008 1.166 15.643 2.09R
19 0.08 41.973 24.924 1.155 17.049 2.28R
286 0.00 42.700 25.036 1.170 17.664 2.36R S = 7.47451 R-Sq = 21.0% R-Sq(adj) = 20.4%
307 0.01 47.352 25.022 1.168 22.330 2.99R
313 2.99 36.112 20.871 0.706 15.241 2.02R
403 5.15 0.665 17.862 0.663 -17.197 -2.28R Analysis of Variance
417 6.96 34.924 15.341 0.870 19.583 2.61R
571 1.86 6.900 22.449 0.849 -15.549 -2.07R Regression 1 2047.9 2047.9 36.66 0.000
R denotes an observation with a large standardized residual. Total 139 9757.8
Bromate
as BrO3
12 0.02 40.651 24.695 1.153 15.956 2.16R
19 0.06 41.973 24.639 1.146 17.334 2.35R
286 0.00 42.700 24.723 1.157 17.977 2.43R
300 3.00 39.185 20.523 0.689 18.662 2.51R
307 0.01 47.352 24.709 1.155 22.643 3.07R
400 3.60 0.897 19.683 0.646 -18.786 -2.52R
Constant 25.570 1.139 22.45 0.000
522 0.06 9.200 24.636 1.145 -15.436 -2.09R
H(T-5) -1.5311 0.2191 -6.99 0.000
S = 7.22703 R-Sq = 26.1% R-Sq(adj) = 25.6%

Source DF SS MS F P Bromate as BrO3 (ug/l) = 24.1 - 1.30 H(T-7)
Regression 1 2550.1 2550.1 48.82 0.000
Total 139 9757.8 140 cases used, 1140 cases contain missing values
Unusual Observations Predictor Coef SE Coef T P

Constant 24.132 1.183 20.39 0.000
Bromate H(T-7) -1.3009 0.2441 -5.33 0.000
as BrO3
12 0.02 40.651 25.540 1.135 15.111 2.12R S = 7.65760 R-Sq = 17.1% R-Sq(adj) = 16.5%
19 0.17 41.973 25.310 1.108 16.663 2.33R
124 0.00 11.098 25.567 1.139 -14.469 -2.03R
300 1.09 39.185 23.902 0.946 15.283 2.13R
417 4.43 34.924 18.788 0.611 16.136 2.24R Regression 1 1665.6 1665.6 28.40 0.000
530 1.33 5.911 23.528 0.906 -17.617 -2.46R Residual Error 138 8092.2 58.6
Total 139 9757.8
Bromate
as BrO3 H(T-9) -1.2799 0.2238 -5.72 0.000
12 0.03 40.651 24.093 1.177 16.558 2.19R
19 0.23 41.973 23.833 1.137 18.140 2.40R S = 7.56065 R-Sq = 19.2% R-Sq(adj) = 18.6%
271 0.20 39.300 23.872 1.143 15.428 2.04R
286 0.41 42.700 23.599 1.101 19.101 2.52R
307 0.01 47.352 24.119 1.181 23.233 3.07R
558 0.00 7.800 24.130 1.183 -16.330 -2.16R Regression 1 1869.2 1869.2 32.70 0.000
R denotes an observation with a large standardized residual. Total 139 9757.8
Bromate
as BrO3
12 0.00 40.651 23.780 1.073 16.871 2.25R
19 0.01 41.973 23.768 1.071 18.205 2.43R
271 0.00 39.300 23.780 1.073 15.520 2.07R
286 0.00 42.700 23.780 1.073 18.920 2.53R
300 0.01 39.185 23.768 1.071 15.417 2.06R
307 0.01 47.352 23.768 1.071 23.584 3.15R
Constant 23.634 1.119 21.12 0.000
400 3.88 0.897 18.815 0.639 -17.918 -2.38R
H(T-8) -1.2457 0.2375 -5.24 0.000
403 3.60 0.665 19.173 0.641 -18.508 -2.46R
445 0.01 8.547 23.768 1.071 -15.221 -2.03R
538 0.00 8.600 23.777 1.072 -15.177 -2.03R
S = 7.67845 R-Sq = 16.6% R-Sq(adj) = 16.0%
558 0.00 7.800 23.777 1.072 -15.977 -2.13R

Source DF SS MS F P
Regression 1 1621.5 1621.5 27.50 0.000 Turnford Bromate V Hatfield Abstraction
Total 139 9757.8
99.9
3.0
99
90 1.5
50 0.0
Percent
Bromate 10
-1.5
as BrO3 1
Obs H(T-8) (ug/l) Fit SE Fit Residual St Resid 0.1 -3.0

-4 -2 0 2 4 10 15 20 25
12 0.00 40.651 23.634 1.119 17.017 2.24R Standardized Residual Fitted Value
19 0.44 41.973 23.086 1.036 18.887 2.48R
229 3.00 35.822 19.897 0.679 15.925 2.08R Histogram Versus Order
271 0.00 39.300 23.634 1.119 15.666 2.06R 40 3.0
286 0.00 42.700 23.634 1.119 19.066 2.51R
30 1.5
300 1.97 39.185 21.180 0.786 18.005 2.36R
307 0.37 47.352 23.173 1.049 24.179 3.18R 20 0.0
558 0.00 7.800 23.631 1.119 -15.831 -2.08R
Frequency
10 -1.5
0 -3.0
R denotes an observation with a large standardized residual. -2 -1 0 1 2 3 1 0 0 0 0 0 0 0 0 0 0 0 0

10 20 30 40 50 60 70 80 90 100 110 120
Observation Order


Constant 23.780 1.073 22.17 0.000
Bromate as BrO3 (ug/l) = 25.0 - 1.65 H(T-5) + 0.0203 6600 - Lee Chalk
3
1
Constant 24.997 1.209 20.67 0.000
0 H(T-5) -1.6493 0.2347 -7.03 0.000
6600 - Lee Chalk 0.02030 0.01473 1.38 0.170
-1
S = 7.20359 R-Sq = 27.1% R-Sq(adj) = 26.1%
-2
-3 Analysis of Variance
01/01/2006 01/01/2007 01/01/2008 01/01/2009
Date
Source DF SS MS F P
Regression 2 2648.6 1324.3 25.52 0.000
Total 139 9757.8
3
Source DF Seq SS
H(T-5) 1 2550.1
6600 - Lee Chalk 1 98.6
2
Bromate
0 as BrO3
-1 19 0.17 41.973 26.982 1.641 14.991 2.14R
98 0.00 * 27.670 1.902 * * X
99 0.00 * 27.695 1.916 * * X
-2 108 0.00 * 27.616 1.871 * * X
124 0.00 11.098 26.741 1.419 -15.643 -2.21R
-3 286 0.00 42.700 25.084 1.189 17.616 2.48R
0 20 40 60 80 100 120 140 300 1.09 39.185 23.694 0.955 15.491 2.17R
6600 - Lee Chalk 307 0.01 47.352 25.544 1.134 21.808 3.07R
400 6.96 0.897 16.024 1.155 -15.127 -2.13R
403 6.96 0.665 16.064 1.176 -15.399 -2.17R
Turnford Bromate V Hatfield Abstraction 417 4.43 34.924 19.858 0.987 15.066 2.11R
Bromate as BrO3 (ug/l) = 25.57 - 1.531 H(T-5) 530 1.33 5.911 22.797 1.047 -16.886 -2.37R
50 Regression R denotes an observation with a large standardized residual.
95% C I
X denotes an observation whose X value gives it large leverage.
95% PI
40
S 7.22703
R-Sq 26.1%
R-Sq(adj) 25.6%
30
20
10

0
0 1 2 3 4 5 6 7 8 9
H(T-5)
Turnford Bromate V Hatfield Abstraction & SMD-lee
Amwell Marsh bromate versus Hatfield abstraction
99.9
3.0
99
90 1.5 The regression equation is

50 0.0
Percent
10
-1.5
0.1 -3.0
-4 -2 0 2 4 10 15 20 25
Histogram Versus Order Constant 13.9156 0.5481 25.39 0.000
3.0 H(T) -0.5823 0.1147 -5.08 0.000
30
1.5
20
0.0
S = 4.12680 R-Sq = 12.3% R-Sq(adj) = 11.9%
Frequency
10 -1.5
-3.0
0
-2 -1 0 1 2 3 1 0 0 0 0 0 0 0 0 0 0 0 0 Analysis of Variance
10 20 30 40 50 60 70 80 90 100 110 120
Observation Order
Source DF SS MS F P
Regression 1 438.71 438.71 25.76 0.000
Residuals Versus Date Residual Error 183 3116.57 17.03
(response is Bromate as BrO3 (ug/l)) Total 184 3555.28
2 Regression Analysis: Bromate as BrO3 (ug/l) versus H(T-1)

1
0
-1
-2 Constant 13.5558 0.5222 25.96 0.000
H(T-1) -0.5233 0.1131 -4.63 0.000
-3
01/01/2006 01/01/2007 01/01/2008 01/01/2009
Date
S = 4.17033 R-Sq = 10.5% R-Sq(adj) = 10.0%
Source DF SS MS F P
Regression 1 372.61 372.61 21.42 0.000
3
Total 184 3555.28
2
1
0
-1
-2 185 cases used, 1095 cases contain missing values
-3 Predictor Coef SE Coef T P

0 20 40 60 80 100 120 140 Constant 13.2494 0.4969 26.66 0.000
6600 - Lee Chalk H(T-2) -0.4672 0.1101 -4.24 0.000
Source DF SS MS F P
Total 184 3555.28 H(T-5) -0.7611 0.1073 -7.09 0.000
S = 3.91184 R-Sq = 21.7% R-Sq(adj) = 21.2%


Source DF SS MS F P
Regression 1 770.07 770.07 50.32 0.000
Total 183 3555.13

Constant 13.8630 0.5152 26.91 0.000
H(T-3) -0.5836 0.1078 -5.41 0.000 Regression Analysis: Bromate as BrO3 (ug/l) versus H(T-6)

Source DF SS MS F P
Total 184 3555.28 H(T-6) -0.6804 0.1095 -6.22 0.000
S = 4.01414 R-Sq = 17.5% R-Sq(adj) = 17.1%


Source DF SS MS F P
Regression 1 622.50 622.50 38.63 0.000
Total 183 3555.13

Constant 14.7466 0.5310 27.77 0.000
H(T-4) -0.7362 0.1044 -7.05 0.000 Regression Analysis: Bromate as BrO3 (ug/l) versus H(T-7)

Source DF SS MS F P
Total 184 3555.28 H(T-7) -0.6859 0.1112 -6.17 0.000
S = 4.01934 R-Sq = 17.3% R-Sq(adj) = 16.8%

Analysis of Variance Amwell Marsh Bromate V Hatfield Abstraction
Source DF SS MS F P 99.9
2
99
Regression 1 614.90 614.90 38.06 0.000
Residual Error 182 2940.22 16.16 90
0
Total 183 3555.13 50
Percent
10
-2
1
0.1
Regression Analysis: Bromate as BrO3 (ug/l) versus H(T-8) -4 -2 0 2 4 8 10 12 14
The regression equation is Histogram Versus Order

20
15 0
Frequency
5 -2
Predictor Coef SE Coef T P 0
-3.00 -2.25 -1.50 -0.75 0.00 0.75 1.50 1 0 0 0 0 0 0 0 0 0 0 0 0
Constant 13.8845 0.5215 26.62 0.000 Standardized Residual
10 20 30 40 50 60 70 80 90 100 110 120
H(T-8) -0.6197 0.1151 -5.39 0.000 Observation Order

S = 4.10470 R-Sq = 13.7% R-Sq(adj) = 13.3% (response is Bromate as BrO3 (ug/l))
2
Regression 1 488.69 488.69 29.00 0.000
Total 183 3555.13 0
-1
-2
-3
184 cases used, 1096 cases contain missing values 01/01/2006 01/01/2007 01/01/2008 01/01/2009
Date

Constant 13.5639 0.5094 26.63 0.000 Residuals Versus SMD
H(T-9) -0.5313 0.1098 -4.84 0.000 (response is Bromate as BrO3 (ug/l))
2
S = 4.16036 R-Sq = 11.4% R-Sq(adj) = 10.9%
1
0
Source DF SS MS F P
Regression 1 404.96 404.96 23.40 0.000
Residual Error 182 3150.17 17.31 -1
Total 183 3555.13
-2
-3
0 20 40 60 80 100 120 140

6600 - Lee Chalk
Residual Plots for Log(BrO3)
Amwell Marsh Bromate V Hatfield Abstraction
Bromate as BrO3 (ug/l) = 14.79 - 0.7611 H(T-5) Normal Probability Plot Versus Fits
99.9
25 Regression 99 1
95% C I 90 0
95% PI
50 -1
20 S 3.91184
Percent
R-Sq 21.7% 10 -2
R-Sq(adj) 21.2% 1 -3
15 0.1
-4 -2 0 2 4 0.9 1.0 1.1 1.2
10 Histogram Versus Order

40
1
30 0

5
20 -1
-2
Frequency
10
0 -3
0
0 1 2 3 4 5 6 7 8 9 -3.00 -2.25 -1.50 -0.75 0.00 0.75 1 0 0 0 0 0 0 0 0 0 0 0 0
10 20 30 40 50 60 70 80 90 100 110 120
H(T-5) Standardized Residual
Observation Order
Regression Analysis: Log(BrO3) versus H(T-5)

The regression equation is (response is Log(BrO3))
Log(BrO3) = 1.16 - 0.0327 H(T-5)
1

0

Constant 1.15814 0.02736 42.33 0.000 -1
H(T-5) -0.032743 0.005485 -5.97 0.000
-2
S = 0.200004 R-Sq = 16.4% R-Sq(adj) = 15.9%
-3
-4
Source DF SS MS F P 01/01/2006 01/01/2007 01/01/2008 01/01/2009
Regression 1 1.4252 1.4252 35.63 0.000 Date
Total 183 8.7055
Residuals Versus 6600 - Lee Chalk
(response is Log(BrO3))
-1
-2
-3
-4
0 20 40 60 80 100 120 140
6600 - Lee Chalk
Amwell Marsh Bromate V Hatfield Abstraction (T-5) & SMD-Lee
Bromate as BrO3 (ug/l) = 16.5 - 0.567 H(T-5) - 0.0463 6600 - Lee Chalk Normal Probability Plot Versus Fits
99.9 3.0
99
184 cases used, 1096 cases contain missing values 1.5
90
50 0.0
Percent
Constant 16.4972 0.5259 31.37 0.000 10 -1.5
H(T-5) -0.56654 0.09821 -5.77 0.000
6600 - Lee Chalk -0.046287 0.006332 -7.31 0.000 1
0.1 -3.0
-4 -2 0 2 4 5.0 7.5 10.0 12.5 15.0
S = 3.44667 R-Sq = 39.5% R-Sq(adj) = 38.9% Standardized Residual Fitted Value

3.0
24
Source DF SS MS F P
1.5
Regression 2 1404.93 702.46 59.13 0.000 18
Residual Error 181 2150.20 11.88 0.0
Total 183 3555.13 12
Frequency
-1.5
6
Source DF Seq SS 0 -3.0
H(T-5) 1 770.07 -2.25 -1.50 -0.75 0.00 0.75 1.50 2.25 1 00 00 00 00 00 00 00 00 00 00 00 00

6600 - Lee Chalk 1 634.85 Standardized Residual
1 2 3 4 5 6 7 8 9 10 11 12
Observation Order
Residual Plots for Bromate as BrO3 (ug/l) Residuals Versus Date

3
Residuals from Bromate as BrO3 (ug/l) vs Date
-1
-2
-3
01/01/2006 01/01/2007 01/01/2008 01/01/2009
Date
Hatfield bromate versus Hatfield abstraction S = 42.3414 R-Sq = 22.7% R-Sq(adj) = 22.5%
Bromate as BrO3 (ug/l) = 347 - 11.4 H(T) Source DF SS MS F P
Regression 1 264752 264752 147.68 0.000
Residual Error 504 903568 1793
509 cases used, 771 cases contain missing values Total 505 1168321

Constant 346.696 4.538 76.39 0.000
H(T) -11.4384 0.7956 -14.38 0.000
Bromate as BrO3 (ug/l) = 322 - 7.69 H(T-3)
S = 40.6581 R-Sq = 29.0% R-Sq(adj) = 28.8%
Source DF SS MS F P Predictor Coef SE Coef T P

Regression 1 341709 341709 206.71 0.000 Constant 321.597 3.808 84.46 0.000
Residual Error 507 838112 1653 H(T-3) -7.6940 0.7199 -10.69 0.000
Total 508 1179822
S = 43.5241 R-Sq = 18.5% R-Sq(adj) = 18.4%

Bromate as BrO3 (ug/l) = 336 - 10.2 H(T-1) Regression 1 216388 216388 114.23 0.000
Total 503 1167353

Constant 336.382 4.039 83.28 0.000
H(T-1) -10.1684 0.7372 -13.79 0.000 The regression equation is
Bromate as BrO3 (ug/l) = 317 - 6.67 H(T-4)
S = 41.0422 R-Sq = 27.3% R-Sq(adj) = 27.2%

Source DF SS MS F P Constant 317.162 3.893 81.47 0.000
Regression 1 320518 320518 190.28 0.000 H(T-4) -6.6668 0.7293 -9.14 0.000
Total 507 1172855
S = 44.6489 R-Sq = 14.3% R-Sq(adj) = 14.1%

Bromate as BrO3 (ug/l) = 326 - 8.60 H(T-2) Regression 1 166603 166603 83.57 0.000
Total 503 1167353

Constant 326.356 3.783 86.28 0.000
H(T-2) -8.5973 0.7075 -12.15 0.000
Hatfield Bromate V Hatfield Abstraction Residuals Versus SMD
Bromate as BrO3 (ug/l) = 346.7 - 11.44 H(T) (response is Bromate as BrO3 (ug/l))
700 Regression
95% C I 300
95% PI
600
S 40.6581
R-Sq 29.0% 200
500 R-Sq(adj) 28.8%
100
400
Residual
300 0

200 -100
100 -200
0 1 2 3 4 5 6 7 8 9 0 20 40 60 80 100 120 140
H(T) 6140 - Chilterns - East - Colne
Hatfield Bromate V Hatfield Abstraction Regression Analysis: Bromate as BrO3 versus H(T), 6140 - Chilterns
99.9 400
99 Bromate as BrO3 (ug/l) = 338 - 12.5 H(T) + 0.245 6140 - Chilterns - East - Colne
90
200
50
Percent
Residual
10
0
1
0.1 -200 Predictor Coef SE Coef T P
-200 0 200 400 250 275 300 325 350
Residual Fitted Value
Constant 338.259 4.644 72.85 0.000
H(T) -12.5002 0.7939 -15.75 0.000
Histogram Versus Order 6140 - Chilterns - East - Colne 0.24523 0.04290 5.72 0.000
150 400
100 200 S = 39.4447 R-Sq = 33.3% R-Sq(adj) = 33.0%
Residual
50
Frequency
0 -200
-160 -80 0 80 160 240 320 1 00 00 00 00 00 00 00 00 00 00 00 00
Residual
1 2 3 4 5 6 7 8 9 10 11 12 Source DF SS MS F P
Observation Order Regression 2 392546 196273 126.15 0.000
Total 508 1179822
Residual Plots for Bromate as BrO3 (ug/l)

300
200 Residuals from Bromate as BrO3 (ug/l) vs Date
100
Residuals from Bromate as BrO3 (ug/l) vs 6140 - Chilterns - East - Colne
Residual
0
-100
-200
01/01/2005 01/01/2006 01/01/2007 01/01/2008 01/01/2009
Date
Hatfield Bromate V Hatfield Abstraction & SMD-Colne
Chadwell Spring versus Hatfield abstraction
99.9
99 300 Regression Analysis: Bromate as BrO3 (ug/l) versus H(T)
90 200
50 100 The regression equation is
Percent
0 Bromate as BrO3 (ug/l) = 1.58 - 0.0996 H(T)
Residual
10
-100
1
0.1
-100 0 100 200 300 250 275 300 325 350 141 cases used, 1139 cases contain missing values
Residual Fitted Value

160
300 Constant 1.5843 0.1394 11.37 0.000
120 200 H(T) -0.09963 0.02872 -3.47 0.001
80 100
0
Residual
Frequency
40 S = 0.871765 R-Sq = 8.0% R-Sq(adj) = 7.3%
-100
0
-160 -80 0 80 160 240 320 1 00 00 00 00 00 00 00 00 00 00 00 00
1 2 3 4 5 6 7 8 9 10 11 12
Residual
Observation Order
Source DF SS MS F P
Residuals Versus Date Regression 1 9.1432 9.1432 12.03 0.001
(response is Bromate as BrO3 (ug/l)) Residual Error 139 105.6365 0.7600
400 Total 140 114.7797
200 Bromate
as BrO3
100 292 1.97 3.3000 1.3881 0.0960 1.9119 2.21R
607 4.35 3.2000 1.1505 0.0737 2.0495 2.36R
Residual
0
614 3.04 3.7000 1.2817 0.0798 2.4183 2.79R
621 3.96 3.2000 1.1901 0.0736 2.0099 2.31R
627 3.96 3.0000 1.1901 0.0736 1.8099 2.08R
-100 943 6.00 3.2000 0.9870 0.0910 2.2130 2.55R
950 5.99 3.5000 0.9872 0.0910 2.5128 2.90R
965 5.99 3.3000 0.9871 0.0910 2.3129 2.67R
-200
1008 5.99 3.5000 0.9875 0.0909 2.5125 2.90R
01/01/2005 01/01/2006 01/01/2007 01/01/2008 01/01/2009
1014 3.99 3.4000 1.1868 0.0735 2.2132 2.55R
Date
Residuals Versus 6140 - Chilterns - East - Colne
400
300 Bromate as BrO3 (ug/l) = 1.51 - 0.0861 H(T-1)
200
100
Residual
Constant 1.5104 0.1328 11.38 0.000
0 H(T-1) -0.08607 0.02814 -3.06 0.003
-100
S = 0.879593 R-Sq = 6.3% R-Sq(adj) = 5.6%
-200
0 20 40 60 80 100 120 140 Analysis of Variance
6140 - Chilterns - East - Colne
Regression 1 7.2375 7.2375 9.35 0.003
Residual Error 139 107.5421 0.7737
Source DF SS MS F P
Regression 1 10.933 10.933 14.63 0.000
Total 140 114.780


Constant 1.6991 0.1438 11.82 0.000
Regression 1 5.7692 5.7692 7.36 0.008
Residual Error 139 109.0105 0.7842
Source DF SS MS F P
Regression 1 13.063 13.063 17.75 0.000
Total 139 114.650


Constant 1.6839 0.1383 12.18 0.000
Regression 1 7.7802 7.7802 10.11 0.002
Residual Error 139 106.9995 0.7698
Source DF SS MS F P
Regression 1 13.604 13.604 18.58 0.000
Total 139 114.650

Chadwell Spring Bromate V Hatfield Abstraction (T-6)
140 cases used, 1140 cases contain missing values Normal Probability Plot Versus Fits
99.9
3
99
2
Constant 1.5990 0.1382 11.57 0.000 90
H(T-7) -0.10402 0.02876 -3.62 0.000 1
50
0
Percent
S = 0.871130 R-Sq = 8.7% R-Sq(adj) = 8.0% 10
-1
1
0.1
Analysis of Variance -4 -2 0 2 4 0.6 0.9 1.2 1.5 1.8
Source DF SS MS F P
Regression 1 9.9265 9.9265 13.08 0.000
Residual Error 138 104.7238 0.7589 Histogram Versus Order
Total 139 114.6503 24 3
2
18
Regression Analysis: Bromate as BrO3 (ug/l) versus H(T-8) 1
12
0
Frequency
6
Bromate as BrO3 (ug/l) = 1.65 - 0.119 H(T-8) -1
0
-1.50 -0.75 0.00 0.75 1.50 2.25 3.00 1 00 00 00 00 00 00 00 00 00 00 00 00
140 cases used, 1140 cases contain missing values Standardized Residual
1 2 3 4 5 6 7 8 9 10 11 12
Observation Order
Constant 1.6474 0.1373 12.00 0.000
H(T-8) -0.11877 0.02932 -4.05 0.000
S = 0.861689 R-Sq = 10.6% R-Sq(adj) = 10.0% (response is Bromate as BrO3 (ug/l))
3
Source DF SS MS F P
Regression 1 12.184 12.184 16.41 0.000 2
Total 139 114.650
0
-1
-2
01/01/2006 01/01/2007 01/01/2008 01/01/2009
Date
Residuals Versus 6600 - Lee Chalk log10(Bromate) = 0.135 - 0.0438 H(T-6)
3

Constant 0.13503 0.05058 2.67 0.009
2 H(T-6) -0.04383 0.01013 -4.33 0.000
S = 0.313056 R-Sq = 11.9% R-Sq(adj) = 11.3%

1
0 Source DF SS MS F P
Regression 1 1.8348 1.8348 18.72 0.000
Residual Error 138 13.5246 0.0980
Total 139 15.3593
-1
log10(Chadwell Spring Bromate) V Hatfield Abstraction (T-6)

-2 Normal Probability Plot Versus Fits
0 20 40 60 80 100 120 140 99.9
99 2
6600 - Lee Chalk
90 1
50
0
Percent
-1
10
Chadwell Spring Bromate V Hatfield Abstraction (T-6) -2
1
0.1
-4 -2 0 2 4 -0.3 -0.2 -0.1 0.0 0.1
4 Regression Standardized Residual Fitted Value
95% CI
95% PI Histogram Versus Order
3 S 0.855697 24 2
R-Sq 11.9%
18 1
R-Sq(adj) 11.2%
2 0
12
-1
Frequency
6
-2
1
0
-2.25 -1.50 -0.75 0.00 0.75 1.50 2.25 1 00 00 00 00 00 00 00 00 00 00 00 00
1 2 3 4 5 6 7 8 9 10 11 12
Observation Order
0

-1
0 1 2 3 4 5 6 7 8 9
H(T-6)
Regression Analysis: log10(Bromate) versus H(T-6)

(response is log10(Bromate))
Constant 0.16476 0.05279 3.12 0.002
H(T-6) -0.03742 0.01065 -3.51 0.001
2 6600 - Lee Chalk -0.0012969 0.0007160 -1.81 0.072
1 S = 0.310501 R-Sq = 14.0% R-Sq(adj) = 12.7%
0
Source DF SS MS F P
Regression 2 2.1510 1.0755 11.16 0.000
Residual Error 137 13.2083 0.0964
-1 Total 139 15.3593
-2 log10(Chadwell Spring Bromate) V Hatfield Abstraction (T-6) & SMD-Lee
99.9
-3 2
99
01/01/2006 01/01/2007 01/01/2008 01/01/2009
90
Date 0
50
Percent
10
Residuals Versus 6600 - Lee Chalk -2

1
(response is log10(Bromate)) 0.1

-4 -2 0 2 4 -0.3 -0.2 -0.1 0.0 0.1
2
20 2
1 15
0
10
0
Frequency
5
-2
0
-2.25 -1.50 -0.75 0.00 0.75 1.50 1 00 00 00 00 00 00 00 00 00 00 00 00

-1 Standardized Residual
1 2 3 4 5 6 7 8 9 10 11 12
Observation Order
-2
-3
0 20 40 60 80 100 120 140
6600 - Lee Chalk
Regression Analysis: log10(Bromate) versus H(T-6), 6600 - Lee Chalk

log10(Bromate) = 0.165 - 0.0374 H(T-6) - 0.00130 6600 - Lee Chalk
(response is log10(Bromate))
-1
-2
-3
01/01/2006 01/01/2007 01/01/2008 01/01/2009
Date
284
Appendix D
Single Borehole Dilution Testing

Appendix D
Single Borehole Dilution Testing
D.1 Introduction
A series of single borehole dilution tests (SBDT) within existing boreholes in the Hertfordshire Chalk
were undertaken during 2008. These single borehole dilution tests were undertaken in conjunction with
a programme of point-to-point Natural Gradient Tracer Testing using Bacteriophage which are described
by Cook (2010).
D.2 Objectives
The objectives were:
• To determine the hydraulically active horizons within the selected boreholes in order to guide
injection strategies for the natural gradient point-to-point tracer testing;
• To use uniform injection SBDTs to obtain a direct measurement of horizontal specific discharge
(Darcy velocity);
• To use point injection SBDTs to determine vertical flow rates within the boreholes;
The scope of works comprised:
• Single borehole dilution testing
• Geophysical logging (Temperature & Conductivity, Gamma & Caliper, Formation Resistivity and
Impellor Flowmeter)
• Down-hole CCTV
Four boreholes were identified as suitable based on criteria outlined by Maurice (2009):
1. Nashe’s Farm Borehole (location 019)
2. Comet Way Borehole (location 402)
3. Harefield House Borehole (location 226)
4. Hatfield Business Park Borehole (location 002)
The location at Hatfield Business Park was not tested due to difficulties in obtaining permission.
D.3 Methodology
The field methodology followed Ward et al. (1998) and Maurice (2009). To summarise, a concentrated
solution of table salt (sodium chloride) was injected as a tracer into the water column via a weighted
hosepipe. For uniform injections, tracer solution was injected throughout the full length of water column;
for point injections, tracer solution was added at a specific interval. A Solinst Levelogger LTC probe was
used to measure the electrical conductance profile before and after injection to monitor the dilution of
the salt within the water column to background concentrations over time. The probe measured depth
below the water table using a pressure transducer. Corrections were made for changes in atmospheric
pressure, and temperature adjustment periods.
A proforma and groundwater risk assessment was submitted to the Environment Agency for ap-
proval in November 2007.
D.4 Results and Interpretation

Maurice (2009) used a simple model to simulate the EC profiles produced by different types of flowing
features, and the results from this are used in the interpreation of the EC profiles below.
Figure D.1: Borehole construction details for the boreholes used for the single borehole dilution testing.
D.4.1 Nashe’s Farm BH
The borehole construction and water level details are shown schematically in Figure D.1.
D.4.1.1 Geophysical testing

The results from the Geophysical testing are detailed in the memo from Jessica Randle (2008). The
results pertinent to the dilution testing are summarised below:
• The calliper trace indicates that the open hole section of the borehole has a variable diameter, with
some fractures/fissures widening to 300 mm. In particular there are increases in borehole diameter
between 20.5 and 22.5 m below datum (bD) and between 24.5 and 25.5 m bD.
• The impellor flow logs detected some upward flow within the borehole between 23.3 m bD and
24.8 m bD.
D.4.1.2 Tracer efflux times

The electrical conductance (EC) profile for the uniform injection (Figure D.2) shows that the tracer is
diluted to background concentrations by 21 hours after injection, and has declined to ∼10 % of the initial
concentration by 6 hours 13 minutes after injection. Extrapolation of the graph of summed tracer con-
centration indicates that the tracer concentrations are likely to have diluted to background concentrations
by ∼10 hours.
D.4.1.3 EC profile interpretation - Uniform Injection

Assuming a uniform column of tracer immediately after injection, the initial profile indicates that rapid
dilution occurred in the top section of the profile (20.0 to 22.5 m bD), mid water column (24.5 to
26.0 m bD) and bottom part of the profile (30.0 to 31.5 m bD) before the initial log was run (Fig-
ure D.2). Subsequent profiles show gradual dilution of tracer throughout the borehole column, with
profiles becoming near vertical after 2 hours 50 minutes and dilution apparently continuing at a more
uniform, slower rate.
The horizons at 20.0-22.5 m bD and 24.5-26.0 m bD correspond to depths of increased borehole
diameter (Section D.4.1.1). The reduced tracer concentrations (’troughs’) at these horizons compared
to tracer concentrations in the rest of the borehole column can therefore be explained (at least in part)
by increased dilution due to increased borehole diameter. For an increase in diameter from 200 mm
to 240 mm between 24.5 and 26.0 m bD, the dilution factor is 1.44, which would reduce EC from
2.2 mS cm−1 to 1.5 mS cm−1 . For an increase in diameter from 200 mm to 300 mm between 20.0
and 22.5 m bD, the dilution factor is 2.25, which would reduce EC from 2.2 mS cm−1 to 1.0 mS cm−1 .
These dilutions are larger than those measured, but this could be due to incomplete mixing. The rapid
dilution at these concentration ‘trough’ locations could also be a result of inflow or crossflow from
fractures/fissures at these horizons (Maurice, 2009). The concentration ‘trough’ at 24.5-26.0 m bD
appears to become less prominent with time, which would be consistent with dilution as a result of
increased diameter. In contrast, the ‘trough’ at 30.0-31.5 m bD persists, which could indicate a flowing
feature at this location.
Figure D.2: Electrical Conductance profile for uniform injection (column dilution) SBDT and point injection SBDT at Nashe’s Farm BH. Tracer was injected between 24.5
and 25.5 m bD for the point injection. Vertical flow rate is based on the vertical position of the point injection concentration peak.
The sharp boundary between 24.5 and 22.0 m bD in the initial (t=0:00) profile appears to move
up the borehole to ∼21.0 m bD by the profile at 11 minutes, where it then remains. This is consistent
with upward flow recorded by the impellor flow meter between 23.3 m bD and 24.8 m bD (section ??).
Apart from this, there do not appear to be any definite features on the uniform injection EC profile which
indicate vertical flow within the borehole (Maurice, 2009).
D.4.1.4 EC profile interpretation - Point Injection

In order to investigate the flow behaviour in the mid section of the profile, a pulse of tracer was injected to
between 5.86 and 4.86 below the water table (24.55 and 25.55 m bD). The EC profile immediately after
injection shows a relatively broad concentration peak (EC above background between approximately
22.5 and 27.5 m bD), with a sharper concentration spike between 24.5 and 25.5 m bD (Figure D.2).
Subsequent profiles show that the tracer peak moves down the borehole (vertical flow) until it reaches
between 28.0 and 29.5 m bD, after which the peak flattens out and concentrations decrease towards
background levels.
Summed tracer concentrations for the point injection indicate a low rate of mass loss over the
first 100 minutes as the peak moves between 25.0 and 26.8 m bD. There is then significant mass loss
between 100 and 300 minutes as the peak moves between 26.8 and 28.8 m bD. Vertical flow rates show
a reduction in flow rate below ∼27.5 m bD, which reduces to zero at ∼28.5 m bD. There also appears to
be a decrease in flow rate at ∼26.5 m bD. These observations could indicate outflow (or crossflow with
net outflow) at these horizons (Maurice, 2009).
D.4.2 Comet Way BH


The results from the Geophysical testing are detailed in Jessica Randle (2008). The tests were limited
due to the small amount of water in the borehole and poor verticality of the boreole. The results pertinent
to the dilution testing are summarised below:
• The impeller flowmeter logs detected a slight downward flow within the borehole from 17.0mbD,
in the slotted section of casing.

diluted to background concentrations by 1 hour after injection in the section with slotted casing, and by
2 hours in the section of plain casing.

Due to a misinterpretation of borehole construction information, tracer was added to the full 3.8 m
of water column (between 15.9 and 19.8mbD), including the section of water column within the plain
casing.
Figure D.3: Electrical Conductance profile for uniform injection (column dilution) SBDT at Comet Way
BH.
Assuming a uniform column of tracer immediately after injection, the initial profile indicates that
rapid dilution occurred in the slotted casing (below 16.9 m bD), particularly between 17.0 and 19.0 m bD,
before the initial log was run. The subsequent profiles (17 minutes to 2 hours) are near vertical in the
cased section and indicate dilution continuing at a more uniform, slower rate.
The tracer in the section of plain casing (between 15.9 and 16.9 m bD) also gradually dilutes
to background concentrations, which implies that the tracer moves downwards into the slotted casing
section. This concurs with the downward flow detected by the impellor flow meter from 17.0 m bD.
However, since the profiles do not show a tracer front moving down the borehole, this vertical flow
appears to happen rapidly in relation to horizontal flow out of the borehole.
D.4.3 Harefield House BH


Geophysical test results are not available for Harefield House Borehole as the borehole diameter was too
small to accommodate the equipment.

diluted to background concentrations over the majority of the profile by 4 hours 4 minutes after injection.
Elevated concentrations remain in the bottom section of the borehole below 25.5 m bD. Extrapolation
of the graph of summed tracer concentration indicates that the tracer concentrations are likely to have
diluted to background concentrations by ∼1800 minutes (∼30 hours).

Tracer solution was only placed in the section of borehole with slotted casing, with fresh water in the
cased section. Assuming a uniform column of tracer immediately after injection, the initial profile indi-
cates that rapid dilution occurred in the upper section of the profile (16-21.5 m bD), before the initial log
was run. Subsequent profiles show gradual dilution along the length of the borehole. Summed concen-
trations indicate rapid exponential mass loss until 41 minutes after injection. Total mass remains almost
constant until 1490 minutes (24 hours 50 minutes), after which it decreases approximately exponentially
to background concentrations.
D.4.3.4 EC profile interpretation - Point Injection

In order to investigate the flow behaviour in the mid section of the profile, a pulse of tracer was injected
at 22.9 m bD (Point injection 1) and a separate pulse was injected at 19.5 m bD (Point injection 2).
Point Injection 1 (22.9 m bD)
The EC profile immediately after injection shows elevated concentrations between ∼21.5 and
∼25.0 m bD in addition to the high concentrations seen at the bottom of the borehole in the back-
ground profile and subsequent profiles. Subsequent profiles show that the tracer front moves up the
borehole (vertical flow) until it reaches ∼17.0 m bD at 26 minutes. The summed tracer concentrations
Figure D.4: Electrical Conductance profile for uniform injection (column dilution) SBDT and point injection SBDT at Harefield House BH. Two point injections were carried
out: one with tracer injected at 22.9 m bD and a separate one with tracer injected at 19.5 m bD. Vertical flow rate is based on the vertical position of the point injection
concentration peak.
indicate that the mass appears to increase from the initial profile until the profile at 26 minutes. There
do not appear to be any consistent ‘nick points’ which would indicate outflow horizons over the section
of vertical flow.
Point injection 2 (19.5 m bD)
The EC profile immediately after injection shows elevated concentrations between ∼81.5 and
∼22.5 m bD, with a relatively sharp peak at ∼20.5 m bD. Subsequent profiles show that the tracer peak
moves up the borehole (vertical flow) as it dilutes, until it returns to background concentrations by 37
minutes. There do not appear to be any consistent ‘nick points’ which would indicate outflow horizons
over the section of vertical flow. Summed tracer concentrations for point injection 2indicate mass loss
over the first 16 minutes as the peak moves between ∼20.5 and ∼17.0 m bD. Summed concentrations
hover around background concentrations for the remainder of the dilution until 74 minutes.
D.5 Calculation of horizontal specific discharge (Darcy Velocity)

The results for the
D.5.1 Methodology
Methodology is outlined in Figure D.5
D.5.2 Results
Ct −Cb
Plots of ln C0 −Cb
for Nashe’s Farm, Harefield House, and Comet Way are included.
Figure D.5: Methodology for determination of specific discharge (darcy velocity) from the results of the
Single Borehole Dilution Tests.
Harefield House 16.0 mbD Harefield House 16.50 mbD Harefield House 17.00 mbD
time (minutes) time (minutes) time (minutes)
0 50 100 150 200 250 300 350 0 50 100 150 200 250 300 350 0.0 50.0 100.0 150.0 200.0 250.0 300.0 350.0
0.0 0.0 0.0
y = -0.0122x - 1.0401 y = -0.0135x - 0.4139 y = -0.005x - 0.3944
-0.5 R2 = 0.6068 -0.5 2 -0.5 2
R = 0.7228 R = 0.3279
-1.0 -1.0 -1.0
-1.5 -1.5 -1.5
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)

-2.0 -2.0 -2.0
-2.5 -2.5 -2.5
-3.0 -3.0 -3.0
-3.5 -3.5 -3.5
-4.0 -4.0 -4.0
-4.5 -4.5 -4.5
-5.0 -5.0 -5.0
0.0 50.0 100.0 150.0 200.0 250.0 300.0 350.0 0.0 50.0 100.0 150.0 200.0 250.0 300.0 350.0 0.0 50.0 100.0 150.0 200.0 250.0 300.0 350.0
0.0 0.0 0.0
y = -0.0098x - 0.2461 y = -0.0129x - 0.1637 y = -0.0117x - 0.2124
-0.5 2 -0.5 -0.5 2
R = 0.667 2
R = 0.7083 R = 0.7091
-1.0 -1.0 -1.0
-1.5 -1.5 -1.5
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)

-2.0 -2.0 -2.0
-2.5 -2.5 -2.5
-3.0 -3.0 -3.0
-3.5 -3.5 -3.5
-4.0 -4.0 -4.0
-4.5 -4.5 -4.5
-5.0 -5.0 -5.0
0.0 50.0 100.0 150.0 200.0 250.0 300.0 350.0 0.0 50.0 100.0 150.0 200.0 250.0 300.0 350.0 0.0 50.0 100.0 150.0 200.0 250.0 300.0 350.0
0.0 0.0 0.0
y = -0.0075x - 0.7238 y = -0.0085x - 0.3981 y = -0.0106x - 0.1718
-0.5 -0.5 2 -0.5 2
R2 = 0.6619 R = 0.8361 R = 0.886
-1.0 -1.0 -1.0
-1.5 -1.5 -1.5
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)

-2.0 -2.0 -2.0
-2.5 -2.5 -2.5
-3.0 -3.0 -3.0
-3.5 -3.5 -3.5
-4.0 -4.0 -4.0
-4.5 -4.5 -4.5
-5.0 -5.0 -5.0
0.0 50.0 100.0 150.0 200.0 250.0 300.0 350.0 0.0 50.0 100.0 150.0 200.0 250.0 300.0 350.0 0 50 100 150 200 250 300 350
0.0 0.0 0.0
y = -0.0116x - 0.1636 y = -0.0141x + 0.0045 y = -0.0168x - 0.9786
-0.5 2 -0.5 2
R = 0.8885 R = 0.9568 -1.0
2
R = 0.9189
-1.0 -1.0
-1.5 -1.5 -2.0
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)

-2.0 -2.0
-3.0
-2.5 -2.5
-4.0
-3.0 -3.0
-3.5 -3.5 -5.0
-4.0 -4.0
-6.0
-4.5 -4.5
-5.0 -5.0 -7.0
0 50 100 150 200 250 300 350 0 50 100 150 200 250 300 350 0 50 100 150 200 250 300 350
0.0 0.0 0.0
y = -0.0228x - 0.8718 y = -0.0186x - 1.0068 y = -0.0172x - 1.008
-1.0 2 2
R = 0.8668
2
R = 0.8403
R = 0.9405 -1.0 -1.0
-2.0
-2.0 -2.0
-3.0
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)

-4.0 -3.0 -3.0
-5.0 -4.0 -4.0
-6.0
-5.0 -5.0
-7.0
-6.0 -6.0
-8.0
-9.0 -7.0 -7.0
0 50 100 150 200 250 300 350 0 50 100 150 200 250 300 350 0 50 100 150 200 250 300 350
0.0 0.0 0.0
y = -0.016x - 0.8292 y = -0.0176x - 0.5657 y = -0.0234x - 0.8242
2
R = 0.8902 2 -1.0 2
R = 0.8471
-1.0 R = 0.9405
-1.0
-2.0
-2.0
-2.0 -3.0
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)

-3.0 -4.0
-3.0
-4.0 -5.0
-4.0 -6.0
-5.0
-7.0
-5.0
-6.0
-8.0
-6.0 -7.0 -9.0
0 50 100 150 200 250 300 350 0 50 100 150 200 250 300 350 0 50 100 150 200 250 300 350
0.0 0.0 0.0
y = -0.0222x - 0.7961 y = -0.0152x - 1.0156 y = -0.0027x + 0.1185
2 2 2
-1.0 R = 0.9022 -1.0 R = 0.5299 -0.2 R = 0.5694
-2.0
-2.0 -0.4
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)

-3.0
-3.0 -0.6
-4.0
-4.0 -0.8
-5.0
-5.0 -1.0
-6.0
-7.0 -6.0 -1.2
-8.0 -7.0 -1.4
Comet Way 16.00 mbD Comet Way 16.20 mbD Comet Way 16.40 mbD
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140
0.0 0.0 0.0
y = -0.0500x - 0.1471 y = -0.0466x - 0.1038 y = -0.0387x - 0.2325
2 2 2
R = 0.9908 R = 0.9909 R = 0.9913
-1.0
-1.0 -1.0
-2.0
-2.0 -2.0
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)

-3.0
-3.0 -3.0
-4.0
-4.0 -4.0
-5.0
-5.0 -5.0
-6.0
-7.0 -6.0 -6.0
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140
0.0 0.0 0.0
y = -0.0445x - 0.1675 y = -0.0409x - 0.3681 y = -0.062x - 0.0094
2 2 2
R = 0.9973 R = 0.9682 -1.0 R = 0.9962
-1.0 -1.0
-2.0
-2.0 -2.0
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)

-3.0
-3.0 -3.0 -4.0
-5.0
-4.0 -4.0
-6.0
-5.0 -5.0
-7.0
-6.0 -6.0 -8.0
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140
0.0 0.0 0.0
y = -0.1031x + 0.0239 y = -0.0941x - 0.0515 y = -0.0834x - 0.0924
2 2
-2.0 R = 0.985 2
R = 0.9976 R = 0.9901
-2.0 -2.0
-4.0
-4.0 -4.0
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)

-6.0
-6.0 -6.0
-8.0
-8.0 -8.0
-10.0
-10.0 -10.0
-12.0
-14.0 -12.0 -12.0
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140
0.0 0.0 0.0
y = -0.0856x - 0.1049 y = -0.0859x - 0.132 y = -0.1302x + 0.3779
2 2
2
R = 0.9898 R = 0.9846 -2.0 R = 0.962
-2.0 -2.0
-4.0
-4.0 -4.0
ln(Ct - Cb/Co - Cb)

ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)
-6.0
-6.0 -6.0 -8.0
-10.0
-8.0 -8.0
-12.0
-10.0 -10.0
-14.0
-12.0 -12.0 -16.0
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140
0.0 0.0 0.0
y = -0.1071x - 0.0921 y = -0.1106x - 0.0669 y = -0.1014x - 0.184
2 2
-2.0 R = 0.9913 -2.0 R = 0.9937 -2.0 R2 = 0.9772
-4.0 -4.0 -4.0
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)

-6.0 -6.0 -6.0
-8.0 -8.0 -8.0
-10.0 -10.0 -10.0
-12.0 -12.0 -12.0
-14.0 -14.0 -14.0
Comet Way 19.40 mbD
Comet Way 19.00 mbD Comet Way 19.20 mbD
time (minutes)
time (minutes) time (minutes) 0 20 40 60 80 100 120 140
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140
0.0
0.0 0.0
y = -0.1126x - 0.2793
y = -0.046x - 1.3119 y = -0.1018x - 0.3029
2
-2.0 R2 = 0.9624
-1.0 R = 0.8032 -2.0
2
R = 0.9839
-4.0
-2.0
-4.0
ln(Ct - Cb/Co - Cb)

ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)

-6.0
-3.0
-6.0
-8.0
-4.0
-8.0 -10.0
-5.0
-10.0 -12.0
-6.0
-12.0 -14.0
-7.0
-16.0
-8.0 -14.0
Nashe's Farm 20.50 mbD Nashe's Farm 21.00 mbD
Nashe's Farm 21.50 mbD
time (minutes) time (minutes)
0 20 40 60 80 100 120 140 160 180 200 0 50 100 150 200 250 300 time (minutes)
0 200 400 600 800 1000 1200 1400
0.0 0.0
y = -0.0203x - 0.6072 y = -0.0196x + 0.1566 0.0
-0.5 2 y = -0.0132x + 0.278
R = 0.5796 -0.5 R2 = 0.9648
-2.0 R2 = 0.968
-1.0 -1.0
-4.0
-1.5 -1.5
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)
-6.0
ln(Ct - Cb/Co - Cb)

-2.0 -2.0
-2.5 -2.5 -8.0
-3.0 -3.0 -10.0
-3.5 -3.5 -12.0
-4.0 -4.0 -14.0
-4.5 -4.5
-16.0
-5.0 -5.0
-18.0
time (minutes)
0 200 400 600 800 1000 1200 1400 time (minutes) Nashe's Farm 23.00 mbD
0 200 400 600 800 1000 1200 1400
0.0 time (minutes)
y = -0.0077x + 0.2125 0.0 0 200 400 600 800 1000 1200 1400
-1.0 2 y = -0.0051x + 0.0333 0.0
R = 0.9766
R2 = 0.974 y = -0.0046x - 0.2068
-1.0
-2.0 2
R = 0.9718
-1.0
-3.0 -2.0
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)
-4.0 -2.0
ln(Ct - Cb/Co - Cb)

-3.0
-5.0
-3.0
-6.0 -4.0
-4.0
-7.0
-5.0
-8.0 -5.0
-9.0 -6.0
-6.0
-10.0
-7.0
-7.0
Nashe's Farm 24.00 mbD time (minutes)
time (minutes) 0 200 400 600 800 1000 1200 1400
0 200 400 600 800 1000 1200 1400 time (minutes)
0 200 400 600 800 1000 1200 1400 0.0
0.0
0.0 y = -0.0054x - 0.2556
y = -0.0056x - 0.3108 2
y = -0.0052x - 0.3953 -1.0 R = 0.9568
-1.0 R2 = 0.9895 2
-1.0 R = 0.9082
-2.0
-2.0
-2.0
ln(Ct - Cb/Co - Cb)

ln(Ct - Cb/Co - Cb)
-3.0
ln(Ct - Cb/Co - Cb)

-3.0
-3.0
-4.0
-4.0
-4.0
-5.0
-5.0
-5.0
-6.0
-6.0
-6.0
-7.0
-7.0
-7.0
-8.0
-8.0
-8.0
time (minutes)
time (minutes) time (minutes) 0 200 400 600 800 1000 1200 1400
0 200 400 600 800 1000 1200 1400 0 200 400 600 800 1000 1200 1400
0.0
0.0 0.0 y = -0.0046x - 0.1851
y = -0.0052x - 0.2366 y = -0.0052x - 0.1916 2
2 2
R = 0.9931
-1.0 R = 0.9492 -1.0 R = 0.9574 -1.0
-2.0 -2.0 -2.0
ln(Ct - Cb/Co - Cb)

ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)
-3.0 -3.0 -3.0

-4.0 -4.0
-4.0
-5.0 -5.0
-5.0
-6.0 -6.0
-6.0
-7.0 -7.0
-7.0
-8.0 -8.0
Nashe's Farm 26.50 mbD Nashe's Farm 27.00 mbD Nashe's Farm 27.50 mbD
0 200 400 600 800 1000 1200 1400 0 200 400 600 800 1000 1200 1400 0 200 400 600 800 1000 1200 1400
0.0 0.0 0.0
y = -0.0042x - 0.2768 y = -0.0043x - 0.3235 y = -0.0041x - 0.3321
2 2 2
R = 0.9888 R = 0.9881 R = 0.9857
-1.0 -1.0 -1.0
-2.0 -2.0 -2.0
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)

-3.0 -3.0 -3.0
-4.0 -4.0 -4.0
-5.0 -5.0 -5.0
-6.0 -6.0 -6.0
time (minutes)
0 200 400 600 800 1000 1200 1400 time (minutes) time (minutes)
0 200 400 600 800 1000 1200 1400 0 200 400 600 800 1000 1200 1400
0.0
0.0 0.0
y = -0.0041x - 0.335
2 y = -0.004x - 0.3494 y = -0.0041x - 0.3558
R = 0.9813 2 2
-1.0 R = 0.976 R = 0.9745
-1.0 -1.0
-2.0
ln(Ct - Cb/Co - Cb)
-2.0 -2.0
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)

-3.0
-3.0 -3.0
-4.0
-4.0 -4.0
-5.0
-5.0 -5.0
-6.0
-6.0 -6.0
time (minutes)
0 200 400 600 800 1000 1200 1400 time (minutes) time (minutes)
0 200 400 600 800 1000 1200 1400 0 200 400 600 800 1000 1200 1400
0.0
0.0 0.0
y = -0.0038x - 0.3887
2 y = -0.0047x - 0.2252 y = -0.0049x - 0.0665
R = 0.9548 2 2
-1.0 R = 0.9825 R = 0.9868
-1.0 -1.0
-2.0 -2.0 -2.0
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)
ln(Ct - Cb/Co - Cb)

-3.0 -3.0
-3.0
-4.0 -4.0
-4.0
-5.0 -5.0
-5.0
-6.0 -6.0
-6.0
-7.0 -7.0
Memorandum Date: 12/02/08
To: Ciara Fitzpatrick UCL

Simon Cook
From: Jessica Randle TVW Water Resources
CC: Rob Sage TVW Water Resources
Re: Comet Way BH5 Downhole Inspection
Dear Ciara/Simon
Please find below a summary of the downhole inspection undertaken in Borehole No. 5 at Comet Way,
Hatfield, conducted on Tuesday 29th January.
Figure 1: Borehole Headworks
K:\CMF Thesis\Appendices\Jess_Memos\CometWay_Memo.doc
1
Figure 2: Borehole Chamber
Geophysical Logging
Figure 3 displays the fully processed composite log of the geophysical tools ran during the survey.
The fluid temperature and conductivity sonde was not used during the inspection as there was only
3.8m of water in the borehole. The tool itself is over two and a half metres long. The borehole is
cased to total depth, therefore the formation resistivity tool could not be run. During logging, the
legs on the caliper sonde did not open. This is likely to be due to the poor verticality seen in the
borehole.
The filtered gamma log displays a count rate of 80 counts per second (CPS) until 2.8mbD,
suggesting a clay rich soil or made ground overlying the Chalk. Beneath here, the rate decreases to
19CPS, suggesting a lithological boundary with the Chalk. At 12.7mbD, the count rate is seen to
increase again to a maximum of 50CPS before falling when the water level is encountered, which is
to be expected as the gamma reading is attenuated when measured in water. The final two metres of
the profile displays a constant value of 11CPS.
The impeller flowmeter logs detected a slight downward flow within the borehole from 17.0mbD,
in the slotted section of casing. Flow horizons, if present, are indicated by changes in the impeller
rotation rate and time per revolution that are not associated with changes in the cable speed.
2
Well Name: BH 5
File Name: S:\WATERR~1\TVW\CENTRAL\BROMAT~1\DATA\DOWNHO~1\COMETW~1\BH5COMP.HDR
Location: Comet Way Datum: Ground Level
Metres Filtered Gamma Down Cable speed Down rotation rate Down time per rev Up Cable speed Up rotation rate Up time per rev
0 (CPS) 200 0 (m/min) 1000 0 (RPM) 80 0 (ms) 4000 0 (m/min) 1000 0 (RPM) 100 0 (ms) 3000
0
-5
-10
-15
-20
Figure 3: Comet Way Borehole No. 5 Composite Geophysical Log
3
CCTV Inspection
A summary of the features seen during the CCTV inspection is shown in Table 1.
Depth Features
(mbD)
0 Top of Casing
1.2 Plain steel casing, in good condition
2.4 Casing joint, poor contact between sections, overhang present
CCTV camera knocks against the sides of the borehole. Beneath this
depth, borehole is off vertical
15.8 Rest Water Level, surface is dirty, water underneath is clear
16.9 Casing is slotted. Slots are clean, horizontal and arranged in three
columns. Off white Chalk is visible through the slots
19.4 Beer can is present in the borehole.
Estimated base of borehole is at 19.6mbD. Base is even and covered with
sediment
Table 1. Summary of CCTV Inspection Findings for Borehole No. 5
Regards
Jessica Randle
Water Resources
(01923) 814249/ #4249
4
Memorandum Date: 12/02/08
To: Ciara Fitzpatrick UCL

Simon Cook
From: Jessica Randle TVW Water Resources
CC: Rob Sage TVW Water Resources

Maria Teneke
Re: Nashes Farm Downhole Inspection
Dear Ciara/Simon
Please find below a summary of the downhole inspection undertaken at Nashes Farm, conducted on
Monday 4th February.
Figure 1: Borehole Headworks
K:\CMF Thesis\Appendices\Jess_Memos\Nashes_Memo.doc
1
Figure 2: Borehole Chamber
Geophysical Logging
Figure 3 shows the geophysical logs produced by the fluid temperature and conductivity sondes run
in the borehole. During logging, the depth scale did not record properly, although all other
parameters are ok. Note the top of the log is at 19.6mbD and the base of the log is at 32mbD, giving
a 12.4m section. Figure 4 displays the fully processed composite log of the remaining geophysical
tools ran during the survey.
The fluid temperature profile displays 11.2C at the rest water level (RWL), which is encountered at
19.6mbD and remains constant until the base of the log. Sudden changes in the geothermal gradient
may be due to possible flow horizons. The differential temperature log indicates that no discernable
fluctuations are evident within the borehole.
The fluid conductivity profile gradually decreases from 704S/cm at the RWL to 644S/cm at the
base of the log. Again, changes in the profile may infer flow horizons. The differential conductivity
log indicates no significant fluctuations relating to flow.
After initially high values reflecting the pumping house and surrounding made ground, the filtered
gamma log displays a count rate of 2.7 counts per second (CPS) until 8.7mbD. Beneath here, the
rate increases slightly to 3.6CPS. A decrease to 2.4CPS is observed from the rest water level
(RWL) at 19.6mbD, which is to be expected as the gamma reading is attenuated when measured in
water. This value is maintained until the base of the borehole. No obvious lithological boundaries
are identifiable from the gamma profile.
2
The caliper trace confirms the internal diameter of the plain casing to be 160mm. The open hole
section of the borehole appears well fractured and displays a variable diameter, with some fissures
widening to 300mm.
The formation resistivity varies between a minimum of 38.5m at 21.6mbD and 29.3m at
23.5mbD. With only 10m of saturated open hole section, it is difficult to stratigraphically place
where the borehole lies in terms of Chalk formations. There are no obvious marker horizons present
and the resistivity trace is strongly linked to fluctuations with the caliper log. The shallow depth of
the borehole, the well fissured nature of the open hole section and the presence of large sections of
flint would suggest that the borehole comprises Upper Chalk. The sudden increase in resistivity at
the bottom of the log is due to a test measurement carried out during logging to ensure the sonde is
calibrated correctly.
The impeller flowmeter logs detected some upwards flow within the borehole between 23.3mbD
and 24.8mbD. Flow horizons, if present, are indicated by changes in the impeller rotation rate and
time per revolution that are not associated with changes in the cable speed.
Well Name:
File Name: S:\WATERR~1\TVW\CENTRAL\BROMAT~1\DA
Location: Nashes Farm Datum: Pump house Floor
Fluid Temp Diff Fluid Temp Fld Cond @ 25°C Diff Fld Cond
8 (°C) 16 -1 (°C/m) 1 400 (µS/cm) 800 -60 (µS/cm/m) 60
Figure 3: Nashes Farm Borehole TCDS Geophysical Log
3
Well Name:
File Name: S:\WATERR~1\TVW\CENTRAL\BROMAT~1\DATA\DOWNHO~1\NASHES~1\NASHCOMP.HDR
Location: Nashes Farm Datum: Pump House Floor
Metres Filtered Gamma BH Diameter Form Resistivity Down Cable speed Down rotation rate Up Cable speed Up rotation rate
0 (CPS) 40 0 (mm) 400 0 (Ohm-m) 100 0 (CPM) 1000 0 (RPM) 60 0 (CPM) 1000 0 (RPM) 60
0
-5
-10
-15
-20
-25
-30
Figure 4: Nashes Farm Borehole Composite Geophysical Log
CCTV Inspection
A summary of the features seen during the CCTV inspection is shown in Table 1.
Depth Features
(mbD)
0 Pump House floor
1.2 Plain steel casing, slightly rusted, in reasonable condition
5.9 Base of casing
Chalk is stained brown, smooth and stable
4
Depth Features
(mbD)
7.4 Borehole circumference appears irregular
8.0 Chalk opens out to one side, flints present
11.6-12.3 Chalk surface becomes more rough
13.8 Chalk opens out to one side. Beneath here the Chalk is whiter
16.5 Chalk opens out to one side
17.3 CCTV camera gets stuck on a large piece of flint protruding from the side
of the borehole
19.5 Rest water level, water is reasonably clear
19.6 Chalk appears smooth
20.4 Borehole opens out and the circumference appears irregular
22.5 CCTV camera knocks against a ledge on the side of the borehole.
Chalk beneath here is smooth
23.1-24.4 Large vertical groove appears to one side of the borehole
24.4 Borehole opens out, circumference becomes irregular
30.7 CCTV Camera gets stuck on the sides of the borehole, survey terminated
*Please note the borehole was plumbed to a depth of 32.70m
Table 1. Summary of CCTV Inspection Findings for borehole
Regards
Jessica Randle
Water Resources
(01923) 814249/ #4249
5
285
Appendix E
MAP, GoldSim and DP1D comparison

Validation of GSIM-MAP-DP1D
1 Replication of Old Chalford Network
1.1 Comparison of complete streamtube set in MAP & GSIM

Based on the parameters given in the MAP input files received from Ann Williams, the complete
streamtube set (a total of 103 streamtubes, with 5 output locations) was replicated in GSIM.
Initially, the model was run without matrix diffusion zones. The parameters used for the MAP
simulation including matrix diffusion were unavailable (I’m waiting to see if Ann can find these input
files). Using best estimates of the parameters relating to matrix diffusion, the GSIM model was re‐
run including matrix diffusion zones.
1.1.1 Advection Only (no matrix diffusion)

Simulated mobile zone concentrations are identical (Figure 1), except the GSIM output is smoothed
by the effect of mandatory dispersion in GSIM (minimum 10% of the path length).

Figure 1: Comparison of simulated mobile zone concentrations (no matrix diffusion) using GSIM and MAP.
1.1.2 Advection + matrix diffusion

Simulated mobile zone concentrations from the MAP and GSIM models are very similar (Figure 2),
with differences presumably a result of differences in parameter values for diffusion and matrix
geometry.
Concentrations in the immobile zone were simulated using a ‘diffusion cells’ approach (see note
below). Since MAP is not currently coded to give immobile zone concentrations, the simulated
concentrations could not be validated against MAP.

Figure 2: Comparison of simulated mobile zone concentrations with matrix diffusion using GSIM and MAP.
2 Immobile zone concentrations
2.1 Comparison of single pipe GSIM-DP1D

In order to compare average immobile zone concentrations, a single streamtube was modelled with
both GSIM and DP1D.
The single streamtube modelled with DP1D is an average of the 9 streamtubes leading to the output
‘Sink 2’. This was chosen so that an approximate comparison could be made between this
GSIM/DP1D single streamtube and MAP output for Sink 2.
Despite using the same parameters in both simulations, simulated mobile and immobile zone
concentrations are significantly different (Figure 3; Figure 4). . (Mobile zone concentrations remain
very similar in GSIM and MAP outputs). In particular, the relative difference between mobile and
immobile concentrations is larger in DP1D than in GSIM, indicating that more diffusion into the
matrix is occurring in GSIM than in DP1D (Figure 4).

Figure 3: Comparison of mobile zone concentrations with matrix diffusion using GSIM and MAP & DP1D.

Figure 4: Comparison of mobile and immobile zone concentrations using GSIM, MAP & DP1D. Solid red line is
DP1D mobile concentration, dashed red line is DP1D immobile concentration. Other lines represent GSIM immobile zone
concentration at distances into the matrix block, until half block thickess.
Possible explanations that were considered are summarised below:
2.1.1 Differences in Block Geometry Functions (BGFs)

I have done my best to work through the maths of GSIM pipe pathways and I think that the BGFs for
slab and sphere geometry are those in Barker (1985), which are presumable used in MAP and DP1D.
2.1.2 Differences in input boundary conditions

GSIM uses a mass balance boundary condition (mass flux) rather than a concentration boundary
condition. To account for this, the parameter δ was set to δ=1 in DP1D. This did not result in a
significant improvement to the match between DP1D and GSIM outputs.
2.1.3 Fickian diffusion Vs Mass transfer

It is recognised that the ‘diffusion cell’ approach does not represent Fickian diffusion, but rather
used a constant mass transfer coefficient. However, although this will lead to differences in
predicted immobile concentrations by this method compared to a Fickian representation,
particularly at early and late times, in GSIM this will not have an effect on mobile concentrations as
along the streamtubes Fickian diffusion is simulated analytically (as in MAP). The ‘diffusion cells’
approach merely allows an estimate of immobile concentrations at a point in the absence of coding
to give the analytically calculated immobile concentrations as an output.
2.1.4 Conclusions
Simulations with GSIM and MAP, using identical/best‐estimate parameters, produce identical/similar
output concentrations for mobile concentrations with and without inclusion of matrix diffusion.
However, it is not possible to compare simulated immobile zone concentrations as neither model is
coded to produce this as an output.
An approximation for immobile zone concentrations can be made by setting up a series of coupled
‘diffusion cells’ which are linked to the output mobile zone concentration. However, there is no
MAP output to validate/verify this approach. When a single pathway is simulated with DP1D and
GSIM, mobile and immobile zone concentrations are substantially different.
NOTE: ‘Diffusion cells’ approach to estimate matrix concentrations in GSIM
This approach is represented diagrammatically in Figure 5.
A series of GSIM ‘cell pathways’ (mixing cells) are generated to represent half a matrix block. The
first cell contains water only, and its concentration is fixed to the output concentration from the pipe
pathway (streamtube) mobile zone at that point. The rest of the cells contain saturated porous
matrix material and represent thin slices of matrix moving from the fracture into the matrix block.
Each matrix cell is coupled to the adjacent cells by a ‘diffusive mass transfer link’. This link
represents diffusive exchange of mass between immobile porewater in adjacent matrix cells, and is
described using a constant mass transfer rate. The linked cells therefore form a finite difference
network. Concentrations are calculated at each time step.

Figure 5: The ‘diffusion cells’ approach to estimate immobile concentrations in GSIM.
Diffusion Cell Set‐up
Describe set‐up
X1
Fracture Y1 0
Matrix Y2 0.001
Matrix Y3 0.003
Matrix Y4 0.005
Matrix Y5 0.01
Matrix Y6 0.02
Matrix Y7 0.03
Matrix Y8 0.05
Matrix Y9 0.1
Matrix Y10 0.2
Matrix Y11 0.4
Matrix Y12 0.6
The numbers in the cells specify the Y‐coordinate for the start of the corresponding cell element. In the last row
the number specifies the Y‐coordinate for the end of the last Cell.
Fracture cells contain water only
Matrix cells contain chalk with porosity and water within the pores.
The fracture cell has a ‘specified concentration’ boundary condition which is specified as the
concentration output of the pipe element.
Cells are linked by a diffusive mass transfer links.
The diffusive flux f from pathway i to pathway j is computed as follows:
f ij = D(ci – cj)
where D = diffusive conductance for the species in the mass flux link [L3/T], ci and cj are dissolved
concentration of the species in the medium (water) within cell i and cell j respectively.
Diffusive conductance terms are computed as follows:
D = A
Li + Lj
d*tpi*npi d*tpj*npj
where Li and Lj are the diffusive lengths for the diffusive mass flux link in cell i and cell j (the default
is distance from the centre of the cell to the edge or interface of the cell)
d is the diffusivity for species in water
tpi is the tortuosity for the porous medium in cell i
npi is the porosity for the porous medium in cell i
GoldSim calculates the effective diffusion coefficient for the pipe pathway as Dim = d*t*n.
I have set d, t and n to give a Dim of 1.74 x 10‐10 m2/s, which is the value in DP1D, therefore n = 0.5,
t = 0.7 and d = 0.50 x 10‐9 m2/s.

286
Appendix F
Parameters for Hertfordshire Network Model

Appendix F
Parameters for Hertfordshire Network Model
Table F.1: Parameters used in the Hertfordshire Network Model
Q ta tcb sigma tcf alpha/x

3
Double-porosity Branches (m /d) (hours) (hours) (hours)
typical 6.08E+00 9.44E+00 2.89E+04 3.80E+02 2.00E-01 1.00E-01

Sandridge to Harefield House worst case 6.08E+00 1.89E-02 6.51E+04 4.05E+03 3.97E-03 1.00E-01
best case 6.08E+00 5.56E+01 7.23E+03 2.50E+01 1.16E+01 1.00E-01
typical 6.08E+00 2.50E+01 2.89E+04 3.80E+02 2.00E-01 1.00E-01

Harefield House to Hatfield Quarry worst case 6.08E+00 5.00E-02 6.51E+04 4.05E+03 3.97E-03 1.00E-01
best case 6.08E+00 1.47E+02 7.23E+03 2.50E+01 1.16E+01 1.00E-01
typical 6.08E+00 2.60E+00 2.89E+04 3.80E+02 2.00E-01 1.00E-01

Hatfield Quarry to Comet Way worst case 6.08E+00 9.56E-03 6.51E+04 4.05E+03 3.97E-03 1.00E-01
best case 6.08E+00 9.06E+00 7.23E+03 2.50E+01 1.16E+01 1.00E-01
typical 6.08E+00 1.99E+00 2.89E+04 3.80E+02 2.00E-01 1.00E-01

Comet Way to Join worst case 6.08E+00 7.31E-03 6.51E+04 4.05E+03 3.97E-03 1.00E-01
best case 6.08E+00 6.93E+00 7.23E+03 2.50E+01 1.16E+01 1.00E-01
typical 6.08E+00 6.34E+01 2.89E+04 3.80E+02 2.00E-01 1.00E-01

Join to Arkley Hole (DP) worst case 6.08E+00 1.27E-01 6.51E+04 4.05E+03 3.97E-03 1.00E-01
best case 6.08E+00 3.73E+02 7.23E+03 2.50E+01 1.16E+01 1.00E-01
typical 6.08E+00 1.74E+02 2.89E+04 3.80E+02 2.00E-01 1.00E-01

Arkley Hole to Lynchmill (DP) worst case 6.08E+00 3.49E-01 6.51E+04 4.05E+03 3.97E-03 1.00E-01
best case 6.08E+00 1.03E+03 7.23E+03 2.50E+01 1.16E+01 1.00E-01

Q ta tcb sigma tcf alpha/x
3
Karst Branches (m /d) (hours) (hours) (hours)
Join to Arkley Hole (Karst) 2.60E+00 1.65E+00 4.38E+05 1.90E+01 1209.125 0.001
Join to Lynchmill (Karst) 7.22E+00 3.12E+00 9.04E+04 1.90E+01 249.9371 0.001
Harefield House to Arkley Hole (Karst) 3.04E-03 2.30E+01 4.38E+05 1.90E+01 1209.125 0.01
Harefield House to Lynchmill (Karst) 1.31E-02 5.29E+01 9.04E+04 1.90E+01 249.9371 0.01
Comet Way to Arkley Hole (Karst) 8.23E-01 9.02E+00 4.38E+05 1.90E+01 1209.125 0.01
Comet Way to Lynchmill (Karst) 1.61E+00 9.02E+00 9.04E+04 1.90E+01 249.9371 0.01

Hertfordshire Chalk PDF

Uploaded by

Copyright:

Available Formats

Hertfordshire Chalk PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Hertfordshire Chalk PDF

Uploaded by

Copyright:

Available Formats

The hydrogeology of bromate contamination

in the Hertfordshire Chalk: double-porosity

Ciara Marie Fitzpatrick

A dissertation submitted in partial fulfilment

Department of Earth Sciences

The hydrogeology of Bromate Contamination in the Hertfordshire Chalk Aquifer: In-

1 formerly Three Valleys Water Ltd (TVW)

BGS British Geological Survey

MDL Method Detection Limit

NNR Northern New River

OBH Observation Borehole

P.S. Pumping Station

PWS Public Water Supply

STW Sewage Treatment Works

SMD Soil Moisture Deficit

SLC St. Leonard’s Court

TWUL Thames Water Utilities Limited

2 Groundwater flow and transport in the Chalk 30

2.9 Solute transport in the saturated zone of the Chalk . . . . . . . . . . . . . . . . . . . . . 50

3.5.7 Statistical relationships and bromate transport in Hertfordshire . . . . . . . . . . 107

4 The evolution of bromate contamination in the Hertfordshire Chalk 117

5 The Bromate Source 167

5.6.3 Groundwater monitoring in the vicinity of the source site . . . . . . . . . . . . . 196

6 Multiple Analytical Pathways Approach 238

A Chronology of key events 281

B Business Case for the research 282

C Hatfield Pumping Trial statistical analyses 283

D Single Borehole Dilution Testing 284

E MAP, GoldSim and DP1D comparison 285

F Parameters for Hertfordshire Network Model 286

3.1 Lithostratigraphy of the Hertford district. After Bloomfield et al. (2004). . . . . . . . . . 67

3.1 Location of study area, including topography and hydrology . . . . . . . . . . . . . . . 63

3.22 Regression coefficients: means and 95% confidence intervals . . . . . . . . . . . . . . . 108

4.1 All bromate monitoring locations 2000-2008. . . . . . . . . . . . . . . . . . . . . . . . 119

5.7 Spatial distribution of the bromide contamination based on investigations undertaken

5.25 Relationship between bromide and bromate concentration in groundwater samples

5.41 Bromate source history for Scenario C. . . . . . . . . . . . . . . . . . . . . . . . . . . . 227

1.2 Bromate transport in the Hertfordshire Chalk aquifer

1.3 The bromate source

1.4 Research aims and objectives

Evolution of bromate contamination

Catchment-scale modelling of bromate transport

1.6 Structure of thesis

1.7 Environmental Hydrochemistry of Bromate and Bromide

Aquifer Bromide (mg l−1 ) Chloride (mg l−1 ) Br/Cl r2

1.7.2 Environmental Behaviour

Groundwater flow and transport in the Chalk

2.2 The Chalk

2.2.3 Tectonic History

2.2.4 Influence of periglaciation

2.3 The Chalk as an aquifer

2.4 Karstic behaviour of the Chalk

2 basedon tracer tests

• Surface drainage patterns;

• The presence of explorable caves and cave systems;

• Incidence of chalk boreholes pumping sand;

• Construction details and descriptions from adits;

• Water level response to pumping and recharge;

• Occurrence of indicator bacteria in abstraction boreholes.