.

OTC 7963
@

Using 3D seismic and attribute analysis to track fluid movement in the

subsurface
M.N. Luheshi, S.G. Ahsvihare and J. Evans, BP Exploration

COPWWt

l-.

OFFSHORE

TECHNOLOGY

presented
mt lb
Otf$hore

This paper we$

Texas, 6-9 May ISW
Ttu papw
nfwmat!cm

considerable development for routine use. In certain

situations the rock properties of an interval change
significantly enough to create a measurably different
seismic response. [n order to apply this technology it is
necessary to establish an understanding of

CONFERENCE
Tochno@f

Wsabdad
by h
OTC Pr~
ccmtansd m an abslrad 8uhiHti

Crmferancu

tmtd h Houslcm

Cmnntia
fOlkYUWW mvnw c1
by thn autt?ar(s) CaWnM
cd h

-W
- w-d,
have rmt bsal revw
by h
06Bt!cf0 T&SmOlOgy ad am
subIe@ to uxrOctIC+-I by ttw suhcqs).
The matwial,
a9 pmwntsd
does not
nac.essaiy
refkl
any powtmn d uw Oifstwfe
Tm5wmhxsy Ccmferencn or m
Mkw9

Panm$skm

Hlustmtmiw

may

acknow!dgemaot

to mpy
no!

is malnded
coped

10 an abslrW

The

abstracl

C4 ml mom man SW Wfds

shouki

mntain

cxmvp!wous

of wimre and by whom the paper was presented

Abstract

In 1993, there was a well control incident while

drilling a prospect offshore Vietnam. This incident
necessitated the implementation of an extensive
seismic monitoring programme. It became rapidly
apparent that conventional high resolution seismic
techniques for identifying and monitoring gas in the
subsurface were inadequate due to loss of signal power
through the system.

The distribution of rock properties in the interval

of interest
The scale of the change in seismic response
caused by different fluid fills (water, oil, gas)
Whether the quality of the available seismic data
has the amplitude, time and depth resolutions
required to detect these changes
And, for monitoring cases, whether the quality of
the seismic is controlled sufficiently to enable
comparisons between successive surveys

In the case presented here, the interest was in

monitoring possible subsurface gas movements
following a well control incident. The technical
problems and the approach, however, are quite
generic; in that the same principles apply for fluid
monitoring experiments in conventional production
situations.

This problem drove the need for an innovative

approach to acquiring a Medium Resolution 3D
Suwey. This had two important characteristics,
namely
significantly
higher
bandwidth
than
conventional 3D data and also having the penetration
necessary in order to image the target reservoir,
through a disturbed overburden.

For the particular problem presented here, it was

necessary to establish the history of gas movement in
the subsurface using a series of seismic data sets which
had different acquisition and processing histories. The
need was to understand to what extent it was possible
to measure the extent of gas charged sands using these
seismic data sets.

This paper describes the technology used in

acquiring and processing the data, as this has relevance
to use of this type of acquisition for conventional
reservoir description purposes. It also describes the
attribute analysis developed to compare multiple
vintages of seismic in order to be able to track fluid
movement in the subsurface.

The general analytical process was to build an

understanding of the seismic signature by
constructing a detailed analysis of the rock
properties distribution in the intervals of interest
t%om this, use seismic modelling to predict the
likely signature of reservoir quality sands under
various conditions of fluid charge and depth
analyse the quality of the various surveys acquired
before and after the incident to establish their
repeatability
perform a detailed geophysical mapping and
attribute analysis exercise to aHow comparison of
specific sequences using the various seismic data
sets
condition the resulting maps to allow comparison
of results from the different seismic vintages

The enhanced resolution achieved has relevance to

enabling better reservoir characterisation in general.
The attribute analysis is an example of how these type
of data can then be conditioned in order to allow
tracking of fluid in the subsurface, in appropriate
situations.

Introduction/background

Using seismic data to track fluid movement is a fairly

well known technique, but one that still needs
99

USING 3D SEISMIC AND AITRIBUTE

ANALYSIS TO TRACK FLUID MOVEMENT

avoid adverse weather conditions. However, this

experience shows that in spite of the poor weather
during acquisition, the 3D process is very robust at
noise reduction, so that it is feasible to obtain higher
bandwidths. The ultimate limitation on this will be the
natural earth absorption effects which will eventually
limit the bandwidth achieved.

Interpret the resulting attribute maps in terms of

gas distribution and possible changes of these over
time.

Sei8mic Quality Issues

In order to appreciate the scale of the technical
challenge, it is worth reviewing the characteristics of
the data available for this study.

The lesson tlom this example is that it is worth

checking under local condhions whether for the given
target depth this Mid Res acquisition configuration
could provide greater resolutions, Certainly the
experience reported here, would indicate that this
method is very much worth consideration given the
great added value that can be obtained from such data.
The added value in improved reservoir description can
be quite considerable, depending on target depth.

There were several vintages of seismic acquired in

the area of interest. Five surveys were conventional
shallow high resolution seismic (one pre-drill and four
post drill). in order to understand the pre-drill
subsurface condition, data from conventional 2D and
3D seismic were also analysed.
The high resolution seismic had very high
bandwidth but relatively shallow signal penetration
(the prospect is at about 1.6 km depth). The
conventional 2D and 3D seismic haa good depth
penetration but low resolution. There was a need to
construct an image that had resolution approaching
that of shallow surveys and penetration equivalent to
that of conventional deep 2D or 3D seismic.

Rock Properties Analyde

Understanding the geophysical acoustic properties of

the rocks through which the seismic signal propagates
is key to using attribute techniques for fluid or
Iithology prediction. Here the intent was to use an
seismic amplitude based technique for detecting gas
presence. This immediately raises the question of the
range of amplitudes one would expect in the section of
interest and how different these reflections would be if
brine tilled sands were replaced with gas,

Medium resolution 3D seismic. It was thus decided to

acquire a special 3D seismic data set using
conventional 3D technology, with very large sources
and long cables, but with parameters selected so as to
get greatly enhanced vertical resolution (shallow gun
and cable depths were used). This is refereed to as the
Medium Resolution 3D (or Mid Res 3D).

Seismic sections will generaily have a wide range

of reflection amplitudes; some of which will be due to
purely Iithological changes and others will be affected
by the presence of hydrocarbons. The goal of the rock
properties analysis is to understand the range of
possible reflection coefficients using well data and
where necessary fluid substitution techniques.

The Mid Res 3D succeeded in achieving the

design goals of enhanced resolution and depth of
penetration. This was achieved in spite of very poor
weather conditions. An example of the outcome is
shown in Figs. 1a - d. These figures show a
comparison of the different vintages of data. 1ss
particular a comparison of Fig. ld with the remaining
panels in Fig. 1 shows that the Mid Res 3D has
good depth of penetration
excellent lateral resolution
resolution
and
vertical
conventional 3D

about

The analysis is based mostly on fairly standard

petrophysical techniques. From the weil data it is
possible to generate distributions of, for example,
velocity, density and acoustic impedance of various
Iithologies and plot a trend curve with depth (see Fig.
3), From such trend curves it is then possible to
construct generic property curves for each Iithology
which can be used as input to seismic modelling (see
below),

twice

The impact of this has been a greatly enhanced

understanding of the stratigraphy and of the
depositional systems in the interval of interest. This is
best illustrated by an example of a time slice (see Fig.
2) which shows very detailed depositional features.
i%istype of information is invaluable in producing the
geological description needed for understanding
potential fluid migration pathways.
The enhanced resolution achieved is due to the
shallow gun and cable depths used (4m & Sm
respectively). Generally these depths are set lower to

OTC 7963

The need here is to understand the impact of gas

on the different sands penetrated in this section. Fluid
substitution is used to estimate the effect of gas
saturation on the acoustic impedance of potential
reservoir sands, Fig. 4 shows an example for one sand,
of the impact of gas saturation on acoustic impedance
(Al). This shows the typical behaviour of a rapid
reduction of AI as gas saturation is increased from
zero to about 104 and then the curve becomes
relatively insensitive to saturation.
100

OTC 7963

M.N, LUHESHI, S.G. ALUVIHARE, J. EVANS

Thus the general conclusions from the rock

properties and seismic calibration analyses is that

Seismic Calibrstionlmodelling
The interesting information that is needed from the

rock properties analysis, is the impact of Iithology and

fluid saturation changes on the likely reflectivity that
will be observed on a seismic section. An
understanding of this will help identifi the range of
amplitudes that can be used to uniquely identify gas
presence and those where the interpretation is likely to
be more ambiguous.

gas presence can be detected using interval

RMS attribute
gas brightens RMS by 300% to 700%
seismic is insensitive to changes in gas
saturation above IOO/O.

Clearly the amplitude effect will reduce with

increasing depth. Eventually it will be very difticrrlt to
unambiguously identi~ gas charged sand. The key in
this type of analysis is to understand the scale of the
effect and how it varies with depth. This is a generic
requirement for alI fluid detection techniques based on
seismic and requires very careful methodical analysis
of the geophysical rock properties using well data.

The technique used here was to build simple

wedge models consisting of three layers (se Fig. 5),
with the middle layer being sand of a given set of
properties and the encasing layers being shales. The
wedge geometry allows the investigation to include
thickness variations, since these also have a significant
effect on the seismic response.
The next step is to calculate the seismic response
of each model. Two examples are shown in Figs. 6a &
6b. Fig. 6a shows the case where the sand is brine
filled and 6b is gas filled. There are quite clear
differences between these two cases, both in terms of
amplitude and also in terms of the phase of the Top
Sarrd reflector.

Seismic Data Considerations

Having established the scale of the signal that needs to
be detected, then for time lapse analysis, the next
question is whether the various vintages of seismic
survey are comparable. To detect fluid movement
using seismic it is necessary to compare a survey
acquired after the movement has occurred with one
that predates it.

Clearly it would be possible to use such models as

a tool to hunt for anomalous seismic amplitudes (both
in terms of brightness and phase). However, in this
case, there is a large volume of data to be scanned in
order to try and identifi significant anomalies and then
to interpret these. Hence it was necessary to construct a
simpler less Iabour intensive method to identi~
anomalies. A method based on average Root Mean
Square (RMS) energy in a defined interval was used.

A direct numerical comparison requires that the

only difference between the two surveys must be due
to changes in fluid distribution. This condition is
generally extremely difticult to achieve, even in cases
where the data are acquired in precisely the same
fashion. Seismic products are sensitive to both
acquisition and processing methodologies and to
achieve rigorous repeatability requires extremely close
quality control, down to the need for controlling
versions of processing sothvare as well as the obvious
requirements of repeatable acquisition and processing
parameters.

This RMS attribute process was based on analysis

of several examples of the types of models shown in
Figs. 5 & 6. fire technique consisted of defining an
interval between two geological horizons, as picked
horn seismic, Then the RMS energy for this interval at
every trace location was calculated. If this is done for a
range of wedge models using the properties defined in
the petrophysical analysis, then the results can be
shown in RMS plots such as Fig. 7. This figure shows
the RMS energy calculated from a variety of wedge
models using several different combinations of sand
and shale properties, with and without gas charge.

In the case described here, the operational

considerations were such that there was an explicit
need to be able to compare radically different seismic
data sets, This then required a method to render these
data sets comparable to the extent that anomalous
events due to gas charge could be identified.

Fig. 7 shows the very clear impact of gas charge

on the RMS attribute. There is a very large change in
brightness almost regardless of thickness. For the
range of depths of interest it was possible to quantify
the brightening effect of the gas. This brightening
was found to be in the range of 300A to 700%. This
gives us the essential quantification of the impact of
gas charge on the seismic attribute of interest (here,
interval RMS energy for each sequence).

101

A method for conditioning the data was developed

to allow this comparative analysis. This method relied
on the simple observation that the reflectivity of a
section of the subsurface wilI be invariant with time
unless there has been some movement of fluids in the
time between the acquisition of the various surveys.
Hence if one could identifi an area of subsurface and a
depth range within which no change was expected, it
would be possible to condition the RMS attribute for
this zone so that it was repeatable between the
different surveys.

USING

3D SEISMIC

AND ATTRIBUTE ANALYSIS TO TRACK FLUID MOVEMENT

OTC 7963

significant anomalies in each of these maps are closely

repeatable, which implies little gas movement in this
sequence. The significance of this figure is that in spite
of the observed non-repeatability of the surveys, the
conditioning process described above, has managed to
substantially cure these differences. The observable
differences are well below the threshold needed for
indicating gas presence at this level (namely an RMS
brightening of about 300Yo).

This is a variant on the technique of balancing

seismic amplitudes using a reference horizon.
As an example of the extent of the variability
observed on the data sets used in this study, Fig. 8
shows the RMS energy for a 500ms thick interval for
an area of seismic from four data sets. These data
segments were selected from an area where there was
convincing evidence of an undisturbed subsurface.
Hence each bar in Fig. 8 should be identical in theory
if the seismic had been acquired and processed in
exactly the same way. Clearly there are very
significant amplitude differences between these
surveys, which could mask the effects that need to be
detected.

Key Lessons for Dstectlon of Fluid Movement

The ideal situation for detecting the movement of fluid
fronts in the subsurface from seismic data are

1. Ensure exactly repeatable time lapse seismic

2. The hydrocarbon effect on a seismic attribute is
significantly higher than noise and/or purely
lithoIogicaI events

The RMS attribute analysis based on the synthetic

models discussed above, suggests a possible practical
conditioning technique. If the RMS energy for an
interval is estimated from a survey and then mapped,
the mapped
then the frequency distribution of
amplitudes will be approximately normal (see Fig. 9a).
The points at the high end of the distribution (those
with RMS values greater than, say, 3 times the mean
value) are very likely to be due to gas charged sand
(based on the rock properties analysis). If the same
calculation is performed using another seismic data set
then a similar distribution will be found (see Fig. 9b).
The main part of both these distributions (those points
around the mean) should be due to reflections from
mud/mud interfaces or mud!brine sand interfaces.

The work described here, shows a way of

understanding at least to what extent (2) is ttue, for a
given situation. This is a systematic rigorous
methodology for assessing the scale of the
hydrocarbon signal and the ability of a seismic
experiment to detect this size of change.
As to the issue of repeatability of the seismic, this
work has shown one pragmatic way of overcoming the
inevitable variability that appears in practice. [t is a
simple robust method that can be applied efficiently to
large volumes of data.
The key lessons can be summarised as follows

Geophysical

attribute analysis.

the
properties/seismic
1. Understand
rock
relationship
2. Understand the quality of your seismic

Hence as a pragmatic tool for conditioning the

different data seta, a process which causes the
frequency distributions of RMS maps estimated from
different data sets to overlay, should be robust enough
for our needs, The process that was used in this case is
as follows

HOWrepeatable was the experiment

3.

Interpret the same horizons from the different

seismic data sets
Estimate interval RMS energy for each sequence
from every seismic data set
Produce an RMS map for each sequence of
interest fkom each seismic survey
Calculate the frequency distribution of the RMS
values from the maps
Scale the RMS maps so that the RMS range tlom
mean vaiue to mean value plus 3 times standard
deviation, was identical for each map ( see Fig.
1o)
Plot the resulting maps

Fig. 11 shows an example for a particular

sequence. This figure shows a comparison of RMS
maps produced from 2 surveys (one was a high
resolution survey and one was the Mid Res 3D). The

Find a conditioning function to allow time

lapse comparisons
For both fluid detection and reservoir description
work, go for maximum lateral and vertical
resolution using 3D seismic
Consider High Resolution 3D
3D technology is a very
powerful noise reduction tool

robust

and

The enhanced resolution from surveys of the 3D Mid

Resolution type, have a very high value in geological
description.

102

a25

1.25

1.50
1.75
2.W

0.00

0.25
0.50
0.75
1.00
1.25
i .50
1.75
2.00

Figure 1 b.

1992 Conventional

103

2D

Figure Ic.

1993 High Resolution 2D (Site Survey)

0.00
0.25

f
0

0-50
0.75

# 1.00
c
=
1.25

1.50
1.75

2.00

104

Time-Slice

1400ms

Figure 2. Time Slice from Medium Resolution 3D

Acoustic Impadanca(g/cc.ft/s)

Danaity (g/cc)

Compraaslonal Valocity (W*)

xl

Kc

*:

.
lax

100C

..- *

a14wEau40a

laoo~ OASWND8
J

lYX

tw

\ l\

Figure 3. Velocity Density and Al Trends

l-

Hard alwla

E?

----

Im

WOW -----

-----

-----

---.

--

14000-

12000.

l.~
0%

lmb20n30%40%

so%

60%70%00%90%1-

Water Saturation

Figure 4.

Predicted variation of Acoustic Impedance

with Water Saturation

106

t
1 OOm

5m

Figure S.
20

Example Input to Wedge Models

40

so

100

120

140

1s0

1s0

SP

100
120
140

1s0
1s0
200
o

$i
E

220
240
m
2s0
200
220
240

Flguro6s. Wedge Model for Brine-Fiiied Sand with

Soft Shale above and Hard Shale below
20

40

so

100

120

140

1s0

1s0

Ioa
lza
140
160
180
200
m
240
2M
260
w
320
240

Figure 6b.

Wedge Model for Gas-Filled Sand with

Soft Shale above and Hard Shale below
107

SP

16000

RMS

amplituda
RJ;
%

RMS
amplituda
ran
r
tor;p
?

o
0

20

40

80

60

100

160

140

120

160

200

Thickness

Figure 7.

Range of RMS Amplitudes for Gas and

Brine Wedge Models

4s00

RMS AJl@tUCb

Amplltuda Ranga

4000
3s00

3000

II

25LW
2aoo
lW
lm
Soo
o
1

Seismic Survey

Fkmre 8.

Example RMS Amplitude and Range of

Amplhudes for Grey Sequence

108

40

10%

30

Effect ol gas influx

-... $.

s-i

20
5%

o%
0.0

1.5

0.5

2.0

Amplitude

Figure 9a.

Scaled RMS Energy for Orange Sequence,

Reshoot

10% -

40

30

s%

20

>

10

0
0.0

0,5

1.0

1.5

2.0

Amplitude

Figure 9b.

Scaled RMS Energy for Orange Sequence,

Reshoot 1

109

-5.0

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

-4.0

-s o

5.0

-4.0

-3.()

-2.0

-1.(.) 0.0

1.0

2.0

3.0

-4.0

1046012696
t4915

Reshoot 4

-?5M

-786

23

-5.0
In

-4.0

-3.0

632

1642

-2.0

-1.0

2451

3261

4070

4660

5669

64g6

0.0

1.0

2.0

3.0

-4.0

-72k6

-5041

-2623

-606

1611

3626

6046

6263

4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

-4.0

5 )

Mid Res

-29a

*163

-x&

&

2i67

&

tizI

10&7

13i74

I&l

UL

..

,,.

Figure 10. Distribution of RMS Interval Energy for 4 Surveys

-/

,-.,

..

--.

--

. -.,

-.

J*.,.

49.*.

< .

.:

,,,

-,

./i;

111