International Workshop on Non-structural Measures for Water Management Problems
Flood Control Management System for Reservoirs as
Non-structural Measures
By
Cheng Chuntian
Department of Civil Engineering, Dalian University of Technology
Dalian,116024, P.R.China
ctcheng@dlut.edu.cn
K.W.Chau
Department of Civil and Structural Engineering, Hong Kong Polytechnic University
Hung Hom, Kowloon, Hong Kong
Ou Chunping
Intelligent Engineering Lab, Institute of Software, Chinese Academy of Sciences
Box 8718, 100080, Beijing, P.R.China
Abstract: Flood control operation of reservoirs can play a major role in alleviating flood losses.
Flood control management system for reservoirs, an important non-structural measure, will
benefit to make full use of the flood control capacity of the existing reservoir projects. The paper
is a summary of the national programming about the flood control management system for
reservoirs in China. The background, objectives, main challenges, contents and key technologies
of the programming are introduced. The issues of the integration method and technology of the
flood control management system are addressed. Emphasis is concentrated on object-oriented
design of system, integration of models and methods, as well as database development.
Keywords: flood control; integrated management; object-oriented; database; non-structural measure
Introduction
Flood disaster is one of the most damaging natural disasters in China, with annual average losses
more than 200 billion yuan(1 US$ equal to 8.3 CNY$) in recent years. As major structural
measures to defend against floods, more than total number of 86,000 reservoirs have been
established in the past fifty years. These reservoirs have great role on mitigating flood losses
together with other flood protection measures. A typical example is the flood control operation
of Gezhouba Reservoir, Geheyan reservoir and Danjiangkou Reservoir for 1998 floods of the
Yangtze River. The three large-scale reservoirs are at the main stem and tributaries of the upper
and middle reaches of the Yangtze River. Their joint operation with other flood control projects
have avoided the use of division flood regions and decreased the losses with up to more than
tens of billion yuan for the 1998 floods. The flood control operation of the four large-scale
reservoirs at the main stem of the Liaohe River for 1995 floods is another example. Dahuofang
reservoir, Qinhe Reservoir, Chaihe Reservoir and Guanyinge Reservoir had reduced the losses
by 14.3 billion yuan for 1995 floods with magnitude in more than 100 years return periods. After
1995 floods in the Liaohe River and 1998 floods in the Yangtze River, the governments from
national to local have realized that the flood control operation of reservoirs can play a major role
in alleviating flood losses but there are some problems in flood control management for
109
London, Canada, October 2001
reservoirs. Most of the existing flood control management systems for reservoir were established
for special purposes and are lack of data share and communication with governments, it is very
difficult to for decision- making departments to get real- time information in short time. In order
to make full use of the flood control capacity of existing reservoir projects and to improve the
national level of the flood control operation for reservoirs, the National Flood Control and
Drought Defying Chief Headquarters of China has commissioned Dalian University of
Technology (DUT), Hohai University, and Wuhan Hydroelectric University to develop an
integrated management system for flood control of reservoirs (IMSFCR) since 1998, with a
duration of five years. The objectives are to establish a standardized flood control software
system of multi- reservoir, integrated real-time data acquisition and processing, precipitation
analysis, flood forecasting, reservoir system analysis, information query, multi- media tutorial
and some of the recent methodologies of flood control based on large scale database
management system.
This paper addresses the issues of the software integration of flood control management system
for reservoirs. Emphasis is concentrated on object-oriented design of system, database
development as well as integration of models and methods. The main challenges, structure
framework and custom designed functions of IMSFCR are briefly described.
Main Challenges
The main challenges in developing IMSFCR lie in dealing with the complexity of typical
systems, the interface integration and standardization of software system. It is because China is a
country with a vast territory and there exist substantial differences in flood conditions
determined by variations such as physical geography, hydrological and meteorological
characteristics at different locations. Owing to these variations, methods employed in the
reservoir flood control system are determined to a large extent by the purposes and the scope of
the project. In addition, the data available has an effect on the choice of models. Hence,
extensive model libraries will be established, which may take a lot of time and effort. Advances
in computer technology have made it possible to simulate the flood control management
processes in an integrated and comprehensive framework. The flood control management system
for reservoirs involves directly with real-time data acquisition and processing, precipitation
analysis, flood forecasting, reservoir system analysis, information query, as well as multi- media
tutorial. Large-scale database management system (DBMS) is the basis of the integrated system.
A distributed model requires the analyst to acquire, maintain, and extensively utilize a
referenced database. It is greatly different from the traditional text files systems. Database
sources including input, output, calculation, query and temporary procedure should be used in all
operations. Another question, which arises here, is whether it is necessary to replace all the
existing software. A lot of mathematical models, including professional models and general
algorithms, are coded with a variety of programming languages such as Fortran, C++,
Powerbuilder, and so on. The user interface of the old software may be under traditional
database system or in previous versions. The choice in reusing or rewriting them under the new
environment, mainly depends on the quantity, complexity and quality of the existing software as
well as on the availability of time and resources. Furthermore, the programs must meet a
minimum standard of quality as far as reliability, efficiency and maintainability are concerned.
We chose to reengineer and redevelop all existing systems in order to satisfy the common rules
and to meet the requirements of reliability, efficiency and maintainability as far as possible. To
speed up the development procedure, the functions of old systems are refereed and made some
translation from one language to another.
110
International Workshop on Non-structural Measures for Water Management Problems
Selection of prototype system
In China, there are more than 5,000 rivers with area over 100km2 . There are about 86,000
reservoirs, and among them, more than 3,000 have storage capacity over 100 million cubic
meters. These reservoirs are classified as medium to large types that are the essential objects of
development for the national software project of reservoir flood control. 20 reservoirs were
chosen as prototype systems in the first trial project from the whole country in 1998. 40
reservoirs were chosen in the second batch in 1999 while 60 reservoirs were selected in the third
batch in 2000. Among them, the developing group s of DUT chose 10 reservoirs as prototype
systems in 1998, 14 ones in 1999, and 20 ones in 2000. From these prototype systems, we aim to
find the common features and to distinguish the differences among them. Furthermore, we are
planning to develop a general software system of reservoir flood control. On the other hand, the
calibration and validation of these common rules are also one of main objectives.
Structural framework of system
IMSFCR adopts client/server structure based on large-scale database. The databases are divided
into network databases and special databases. Network databases consist of real- time data
library, history records library and results library. Special databases include flood forecast
library and flood operation library. Network databases are shared resources involving original
records and public information. On the other hand, special databases are kept private to facilitate
simulation and analysis of flood forecasting and flood control operation by both technical and
non-technical personnel before formal or official results have been generated. Most data are
temporary in nature and are only valid in user’s machines. For non-technical users, it is
especially important that they can operate and learn from the system, without the threat that their
inadvertent action may erase some important and raw data. For technical users, they can simulate
any alternatives in an efficient way. This layout of databases enhances data security as well as
improves flexibility of the system application. The structural framework of the reservoir flood
control system is shown in Figure 1.
Custom designed functions
The reservoir flood control system includes six custom designed functions, i.e, “information
query”, “data processing”, “flood forecast”, “flood operation”, “result post-processing” and
“help”. Figure 2 shows the main function tree of the reservoir flood control system.
“Information query” provides several basic information inquiry for brief introduction of
reservoirs, dynamic real-time flood situation, flood forecasting results, flood operation
alternative results, historical records, operation rules, law and policy related to flood control
operation, and so on. Query can be activated through access to the databases via
internet/intranet. The query results may be shown in various formats such as tables, graphic,
maps, videos, and text.
111
London, Canada, October 2001
Reservoir flood operation
system
Forecast
Operation
Post-processing
Database
Operatio
Historica
Network
Real-
Special
Database
Help
Results
Preprocessing
Forecast
Information
Figure 1. Structural framework of the reservoir flood control operation system
Flood Control System of Reservoirs
Information
Preprocessing
Forecast
Operation
Introduction
Data input
Model choice
Real-time
Data processing
Initial condition
Historical
Real-time
monitoring
Real-time
forecast
Simulation
forecast
Revised forecast
Historical
analysis
Parameter
calibration
Initial
condition
Automatic
manner
Interactive
manner
Evaluation
Policy
Rules
Postprocessing
Results project
Graphic result
Table result
Report forms
Selection
Figure 2. Main function tree of the system
112
Help
About
system
Multimedia
International Workshop on Non-structural Measures for Water Management Problems
“Data preprocessing” provides the data input and modification requirements on historical
records, real- time flood data processing, and real- time flood monitoring. Large amounts of
historical flood records are very valuable to flood forecast and flood control operations. Since
data input is a time consuming task, various choices are allowed in the manner of input, such as
electronic formats developed for special tables, database translation connections from a database
into another, and special file formats similar to “excel”. General index and guidance are
established to facilitate the data input. All imported data, such as historical records, historical
files, and model parameters, can be reviewed and checked through “information query”.
Constrained by the time interval stipulated in the flood forecast model, some of the original data
about rainfall, level and discharge must be processed in order to satisfy the requirements of the
real-time system. In addition, employing the real- time monitoring of flood event, the user can
dynamically perceive rainfall spatial- temporal distribution of the whole basin as well as level
process of some control sections. When alarming accumulated rainfalls or water levels arise, the
system will prompt the user to perform flood forecast and flood control operations.
“Flood forecast” includes the choice of flood forecasting models, initial condition set and
modification on antecedent soil moisture, real-time forecast, simulation forecast, revised
forecast, historical flood analysis, and model parameter calibration and validation. The system
has integrated most of the commonly used flood forecasting models in China, such as the famous
Xinanjiang model developed by Zhao(1992), typical rainfall-runoff models, unit hydrograph
model(Linsley et al.,1975; Hoggan, 1997).When a model is chosen, the system will set it as the
default in the next time. Flood forecast is classified into “real-time” and “simulation” depending
on the situation when real- time and simulation data are used respectively. Forecast results are
dynamically analyzed and shown with tables and graphics. Some characteristic values of the
flood , including total rainfall, total runoff, pure precipitation rate, peak inflow, peak time, the
largest flood inflow volume , occurrence time and corresponding frequency during the given
time interval, are displayed. When there is a large discrepancy between real data and prediction,
revised methods can be evoked. Historical flood analysis can be performed from the similar
historical floods based on pattern recognition. With the accumulation of more hydrological data,
the user can recalibrate and revalidate the model parameters, and choose another appropriate
model.
“Flood control operation” can deal with both single reservoir and multi-reservoirs system.
Reservoir flood control operation is operated in real time, which often differs very much from
other operations for planning purposes. The crucial difference between them is that decision
making of flood control is usually effective only for the current period or for the following
periods. Constrained by the updating forecasting inflow information at the current period,
decisions need to be made on a daily or even hourly basis during flood events. Multi-criteria
decision analysis has been shifting from optimization methods to more interactive decision
support tools (Bender and Simonovic, 2000). This system unitizes the recent fuzzy optimal
model for the flood control system developed by Cheng (1999) and Cheng and Chau( 2001). The
main features are the ability to quickly generate, select and evaluate alternatives and the
flexibility of these models to allow for the transient change of practical flood conditions and to
mimic the intuitions and experience of operators. “Automatically generated alternatives” uses
knowledge-based rules gleaned from flood control authorities (experts) and from analyzing a
mass of recorded historical data (Cheng and Chau, 2001). After these automatically generated
alternatives are evaluated, “interactively produced alternatives” may be activated to produce
more alternatives. Through fuzzy optimal model, one of the satisfied solutions is obtained.
113
London, Canada, October 2001
“Results post-processing” completes the task management for real- time information, flood
forecast, flood control operation, and report outputs. Tables and graphics are direct and distinct
demonstrations for problem and are easy to be understood by users and decision- makers. The
numerous tables and graphics are designed for the special tasks and can be printed as report
outputs. In addition, the formal results of flood forecast and flood control operation are only
accessible by the authorized users and institutions.
“Help” provides support and guide for users. The user can learn to use the system through
multimedia tutorials.
Database development
The database system is the basis of IMSFCR. Sybase and SQLServer are adopted as the DBMS
of reservoir flood control system. The two DBMS can ensure high data integrity, recovery, and
concurrency control. They support the high- level query language SQL and enable users to
perform sophisticated data retrievals. Most of the structural definitions of Sybase about data
properties is the same as those of SQLServer. One of the two DBMS can be chosen according to
the scale of the application system and the economic condition of the user. It is not necessary to
modify the programming source codes.
The design of the relational tables has a significant effect on the programming source codes and
the operation efficiency of the flood control system. The software system based on the database
is completely different from traditional files such as in HEC1-HEC5 packages(Hoggan, 1997).
All preprocessing, calculation, query and post-processing are based on database and necessitates
the access to the data in the related tables. An optimized database design can render the system
easier to expand and minimize adjustment to the programming source codes, as well as improve
the efficiency.
The main works to achieve database optimization are to define the types of data queries and
requests, and to normalize the database relationships. The optimization products are table
structures where the table names, the name and data types of its fields, and the integrity
constraints are defined. More than 400 tables have been designed for the reservoir flood control
system. These tables form parts of the common rules for the national flood control system. The
database can be utilized to generate all the standard reports required by the user. Another major
objective is to provide the public with flood control information. The increasing use of the
Internet makes the World Wide Web an attractive vehicle for the dissemination of such
information. The flood control information about reservoirs can be accessed using
Internet/Intranet.
Design of object-oriented system
During recent years, object-oriented framework has become popular technology as effected by
object-oriented software development(Whitten et al., 2001). The technology has been applied to
water resources system development (Beclkhouche et al., 1999), as well as in other fields
(Cheng, 2000; Mattsson and Bosch, 2000). Object-oriented technology bears the promise of
reduced development effort through large-scale reuse and increased quality (Moser and
Nierstrasz, 1996). For the national flood control system, reliability, efficiency, maintainability
and reuse are the important objectives of the normalized and standardized flood control system
for reservoirs. The fourth generation languages (4GL), such as PowerBuilder 6.0, VC6.0 and
VB6.0, are object-oriented software development and are easier to be used. Most of components
in IMSFCR are developed by using PowerBuilder 6.0, some by VC6.0 and integrated by
ActiveX.
114
International Workshop on Non-structural Measures for Water Management Problems
Good quality object-oriented software system depends on the workflow analysis. The detailed
design of object involves layout of conceptual design, logical design, physical design and
implementation (Bekhouche et al, 2000). In other words, the whole system will be divided into a
mass of single objects. The attributes and behaviors of each object will be analyzed and defined.
A series of object-oriented models capturing attributes and methods of objects are established
based on the analysis of objects. The whole system is in fact the implementation of objectoriented models based on the logical and physical relationships of objects. Figure 2 is the main
function tree of the system, which also shows the logical and physical relationships of main
objects. The main interface of flood control system for reservoirs is an integrated object consists
of six integrated objects of second level function components. Each object of second level
function components is also an integrated object consists of lower level objects. The lowest level
component is a series of single object-oriented models. Figure 3 shows the framework of objectoriented models.
In Figure 3, rectangular boxes represent integrated objects; rectangular boxes with shadow
represent single objects. Adopting these object-oriented models, the interrelations that exist
among these components are established, specifying the behavior of the entire structure and each
component. The object-oriented framework technology has the following appealing traits: (1)
easy development by team groups thus speeding up the development, (2) large-scale reuse
because of relatively independent encapsulation of the object, (3) improved quality such as
reliability, efficiency and maintainability because of high portability.
Integration of mathematical models and methods
The two essential components are flood forecast and flood control operation. Numerous models
related to these two parts are integrated into this system. Most of the common methods and
models are coded with general functions through conversion from previously developed
software as far as possible. Only small proportions of them are developed recently under the new
environment. Interactive contents related to these functions such as initial conditions are
provided and executed through graphic user interface. Most of the mathematical models and
methods are embedded in methods of object through the call for mathematical model or method
functions.
The forecasting component consists of precipitation-runoff model libraries and flood routing
libraries. The precipitation-runoff model libraries have integrated five models up to now:
rainfall-runoff relationships (including P+Pa~R and P~Pa~R, flow forecast using unit graph
model), Xinanjiang model (Zhao,1992), Dahuofang Model( Liu 1985). The flood routing
libraries provide for Muskingum methods (Muskingum and Muskingum-Cunge), coefficients
methods and direct additional method after consideration of the time lag.
Real-time revised forecast results are useful in improving the forecasting quality. An effective
method for all cases is still under development because the flow routing time is very short for
most reservoirs. Two particular methods and technologies are used in this system. The first
method is dynamically adjusting the model parameter through calibration and validation of
model from those classified floods based on pattern recognition or from recent floods. Figure 4
represents a case application of the real-time revised forecast results through the above methods.
Another method is interactively modifying the forecasting results by hand with reference to
those similar historical floods. These two methods are effective for some reservoirs only.
115
London, Canada, October 2001
First Level Object
Main interface of system
Second Level Object
Component 1
Second Level Object
Component 2
Second Level Object
……
Second Level Object
Component n2-1
Nth Level Object
Component 1
Nth Level Object
……
Nth Level Object
Component nn
Object 1
Object 2
Object …
Object m-1
Second Level Object
Component n2
Object m
Figure 3 Framework of object-oriented models
Calibration and validation of model parameters are also a time consuming process. A new
technology, i.e., genetic algorithms, is developed to automatically calibrate and validate model
parameters. In recent years, genetic algorithms have become one of the most widely used
techniques for solving a number of hydrology and water resources problems (Wang, 1991; Ritzel
et al, 1994; McKinney and Lin, 1994; Dandy et al, 1994; Cieniawski et al, 1995; Franchini and
Galeati, 1997;See and Openshaw, 1999; Wardlaw and Sharif, 1999) because of their advantages
over classical optimization methods. They have been used successfully as an optimization tool in
runoff model calibration (Wang,1991; Franchini and Galeati, 1997). A new genetic algorithm,
which combines a fuzzy optimal model with a genetic algorithm, is developed to solve the
multiple objective rainfall-runoff model parameters calibration problem (Cheng et al.,2001).
The parameter calibration includes optimization of multiple objectives: (1) peak discharge, (2)
peak time and (3) total runoff volume.. Furthermore, the graphic user interfaces related to the
genetic algorithm are designed and developed. Users can easily set the ranges of different
parameters, choose real and binary codes, and decide the approaches of crossover and mutation.
116
Observed
1500
Common method
Clustering
1000
Recent floods
0
500
discharge(m3/s)
2000
International Workshop on Non-structural Measures for Water Management Problems
2000-6-10 08
2000-6-12 05
2000-6-14 02
2000-6-15 23
2000-6-17 20
2000-6-19 17
time
Figure 4. A case study about the real-time revised forecast
After entering the initial conditions and evoking the “calibrate” button, the model parameters
will be automatically calibrated. The results will be shown in both graphics and tables. When a
group of satisfied parameters is available, choosing the “replace” button will update the old
parameters of the model. Figure 5 and Figure 6 show the part results of a case study about the
parameter calibration and validation based on our genetic algorithm.
500
1000
1500
2000
2500
Observed
Simulated
0
discharge (m3/s)
3000
3500
As mentioned above, flood control is operated in real-time. The decision making process is very
complex and related to all the parties interested in the problem under consideration. Traditional
techniques often simplify the problem by transforming the multi-objectives into a singleobjective, which are difficult to be related to intuitions and experience of operators.
1041
1081
1121
1161
routing period (3hr)
1201
1241
Figure 5 The simulated and observed hydrographs during calibration
117
1281
3000
Observed
1000
2000
Simulated
0
discharge (m3/s)
4000
London, Canada, October 2001
routing period (3hr)
Figure 6 The simulated and observed hydrographs during validation
Fuzzy optimal models are easy to deal with the complexity of a typical system with uncertainty
and to mimic intuitions and experience of operators (Cheng, 1999; Cheng and Chau, 2001).
Similar works can be found in Russell and Campbell (1996), Fontane et al.(1997), Despic and
Simonovic(2000), and Bender and Simonovic(2000). These models (Cheng, 1999; Cheng and
Chau, 2001) have been integrated into this system. When a new flooding event is imported, the
system will automatically generate a series of alternatives based on the rules in the knowledge
base and gives the evaluation of various alternatives (Cheng and Chau, 2001). Furthermore,
users can interactively propose new alternatives and evaluate all alternatives. Finally, one
alternative can be selected according to the evaluation.
Conclusion
The flood control management system for reservoirs has a significant role in alleviating flood
losses. The establishment of IMSFCR will speed up the development of the national flood
control system for reservoirs, make full use of the flood control capacity of existing reservoir
projects and improve the national standard of the flood control operation for reservoirs. Largescale database is the basis of IMFCR. IMFCR integrates descriptive knowledge (e.g., data and
information), the procedure knowledge (methods and algorithms) and reasoning knowledge
(rules), with robust and effective management capacity.
Acknowledgments
This research was supported by “The Teaching and Research Award Program for Outstanding
Young Teachers in Higher Education Institutions of Ministry of Education, P.R.C” (No.200026),
“Fok Ying Tung Education Foundation, P.R.C”(No.71072), the National Natural Science
Foundation of China (No. 60073037), and the Research Grants Council of Hong Kong
(PolyU5084/97E).
118
International Workshop on Non-structural Measures for Water Management Problems
References
B.Beclkhouche, I.Demtchouk, and L.J.Steinberg(1999). “Design of object-oriented water quality
software system”. Journal of Water Resources Planning and Management, 125(5), 289-296
Brian J. Ritzel, J.Wayland Eheart, and S. Ranjithan(1994). “Using genetic algorithms to solve a
multiple objective groundwater pollution containment problem”. Water Resources Research,
30(5), 1589-1603.
Cheng Chuntian(1999). “Fuzzy optimal model for the flood control system of upper and middle
reaches of the Yangtze River”. Hydrological Sciences Journal, 44(4),573~582
Cheng Chuntian(1999). “Fuzzy optimal decisions for multi-objective flood control system of
cascade reservoirs”. Journal of Hydrodynamics ,Series B, 11(3), 6~13
Cheng Chuntian, K.W.Chau(2001). “Fuzzy iteration methodology for reservoir flood control
operation”, Journal of the American Water Resources Association, 37(5), 1381~1388
Cheng Chuntian, Ou Chunping, K.W.Chau(2001). "Combining a fuzzy optimal model with a
genetic algorithm to solve multiobjective rainfall-runoff model calibration", Jour nal of
Hydrology (in review)
Daene C. McKinney, and Min-Der Lin(1994). “Genetic algorithm solution of groundwater
management models”. Water Resources Research. 30(6), 1897-1906
Daniel H.Hoggan(1997). “Floodplain hydrology and hydraulics”. The McGraw-Hill Companies,
Inc.
E.C.Cheng(2000). “A object-oriented organization model to support dynamic role-based access
control in electric commerce”. Decision Support Systems, 29, 357~369
Graeme C. Dandy, Angus R. Simpson, and Laurence J. Murphy(1996). “An improved genetic
algorithm for pipe network optimization”. Water Resources Research, 32(2), 449-458.
Linda See, and Stan Openshaw(1999). “Applying soft computing approaches to river level
forecasting”. Hydrological Science Journal. 44(5), 763-778.
Liu(1985). “Dahuofeng rainfall-runoff model ”. The selected papers on hydrological forecasting
conference. The hydroelectric publisher of China, Beijing, China, 35~48
Jeffrey L.Whitten, Lonnie D.Bentley and Kevin C.Dittman(2001). System analysis and design
methods. McGraw-Hill Companies, Inc.
Ozren Despic, S.P.Simonovic(2000). “Aggregation operators for soft decision making in water
resources”. Fuzzy Sets and Systems, 115, 11~33
Q. J. Wang(1991). “The genetic algorithm and its application to calibrating conceptua l rainfallrunoff models ”. Water Resources Research. 27(9), 2467-2471.
119
London, Canada, October 2001
Ray K.Linsley,JR., M.A.Kohler and J.L.H.Paulhus(1975). “Applied Hydrology”. Tata McGrawHill Publishing Company Limited, New Delphi-110015
Robin Wardlaw, and Mohd Sharif(1999). “Evaluation of genetic algorithms for optimal reservoir
system operation”. Journal of Water Resources Planning and Management, 125(1), 25-33
Marco Franchini, and Giorgio Galeati(1997). “Comparing several genetic algorithm schemes for
the calibration of conceptual rainfall-runoff models ”. Hydrological Science Journal. 42(3), 357378.
Michael J.Bender, S.P.Simonovic(2000). “A fuzzy compromise approach to water resource
systems planning under uncertainty”. Fuzzy Sets and Systems, 115, 35~44
M.Mattsson, and Jan Bosch(2000). “Stability assessment of evolving industrial object-oriented
frameworks”. Journal of Software Maintenance: Research and Practice, 12, 79~102
Moser S, Nierstrasz O(1996). “The effect of object-oriented frameworks on developer
productivity”. IEEE computer, 29(9), 45~51
Scott E. Cieniawski, J. Wayland Eheart, and S. Ranjithan(1995). “Using genetic algorithms to
solve a multiobjective groundwater monitoring problem”. Water Resources Research. 31(2),
399-409.
Zhao R.J(1992). “The Xinanjiang model applied in China”. Journal of Hydrology, 135, 371~381
120