Daene C. McKinney Ximing Cai David R. Maidment  

A Prototype GIS-Based Decision Support System 
for River Basin Management 

Abstraction 

     This paper presents a prototype spatial decision support system (SDSS), which is a conjunctive application of GIS and DSS technologies, for river basin water resources planning and management. In this GIS-based DSS, a river basin is modeled as a collection of spatial objects which represent the river basin physical entities, and thematic objects. The thematic objects represent a river basin network, attributes for river basin entities, physical laws that govern the spatial objects in the real world, and the socio-economical policies that control the operation of the river basin (we define the laws and policies as relation objects), and inquiring schemes in which the network, attribute, and relation objects are connected under user interactions (we define the schemes as model objects). Therefore we use an object-oriented approach to integrate a river basin visual representation and logical representations of a river basin into an operational framework. Several object-oriented functions are developed, which include data input/update, network derivation, network modification, model generation, system connection, visual display, and the user interface. The system operation procedures are also discussed.  

Introduction 

     Geographic information system (GIS) based decision support systems (DSS), often known as spatial decision support systems (SDSS), are a class of computer systems in which the technologies of both GIS and DSS are applied to aid decision makers with problems that have a spatial dimension (Walsh, 1992). GIS is a general-purpose technology for handling geographic data in digital form, with the ability to preprocess data into a form suitable for analysis, to support analysis and modeling directly and to postprocess of results (Goodchild, 1993). GISs offer a spatial representation of water resource systems, but currently little predictive and related analytical capacities are available for solving complex water resource planning and management problems (Walsh, 1992; Parks 1993). DSSs are interactive programs, often with a graphical user interface (GUI), which embed traditional water resource simulation and optimization models, with adaptation of new approaches, to support users in semi-structural or ill-structural problem solving (Loucks and daCosta 1991). An extension of the DSS concept, a spatial decision support system (SDSS), which is the integration of DSS and GIS, was initiated by Densham and Goodchild (1989), and the research potential for SDSS in water resources was addressed by Walsh (1992). SDSSs integrate the spatial dimension and modeling capacity into an operational framework, so that DSS and GIS technology can both be more robust by their linkage and coevolution.  

      In this paper we consider a DSS for water management in river basins. That is to say, the DSS focuses on a specific rather than a generic problem, and provides aid for decision makers in river basin planning and management, rather than general users. The specific DSS is developed in the environment of a general-purpose GIS, and we call it the GIS-based DSS for river basin management. 

      River basin management, considered at the spatial scale of a river basin, is treated as a kind of decision analysis based on the physical conditions of the river basin and the socio-economic conditions in the basin. A river basin system includes three components (1) source components such as rivers, canals, reservoirs, and aquifers, and (2) demand components such as irrigation fields, industrial plants, and cities, and (3) intermediate components such as treatment plants, and water reuse and recycling facilities. A river basin system is made up of these components and the relations between them. To construct a model for river basin management, generally we often model it as a node-link network from the river basin map, with nodes representing lakes, reservoirs, and aquifers, whose length dimensions may be insignificant, and links representing rivers, canals, whose lengths are significant. (Loucks, 1996). In this paper we set up a network configuration, which is different from that of the traditional node-link network. In the network defined here, nodes represent all source, demand and intermediate components, no matter whether the length dimension of a component is significant or not. For example, a river is divided into a number of reaches, and each reach is represented as a node; while links in the network represent the spatial relations between two components. There are two kinds of links, one are the natural links, for example, a link between two consecutive river nodes; and the other are man-made links, which represent the water supply-demand relations, for example, a link between a reservoir node and a demand site node. Therefore, links in the network are abstract objects, which only represent one node linked to another node. 

     Both GISs and DSSs have been widely used in water resources, and SDSSs, as the conjunctive use of GISs and DSSs, have also been contributed to our understanding of water resources in recent years. Both GISs and DSSs have been widely used in water resources, and SDSSs, as the conjunctive use of GISs and DSSs, have also been contributed to our understanding of water resources in recent years.Compared to traditional DSSs, SDSSs have been improved through the incorporation of GIS in the aspects of data base, interface and model connection, etc. For data base, a GIS not only brings spatial dimensions into the traditional water resource data base, but also, more significantly, has the ability to integrate various social, economic and environmental factors related to water resources planning and management for use in a decision-making process. Therefore such a system helps to attain an integrated view of the world.(Lam and Swayne,1991; Cowan et al.,1996). For interface, the visual display capacity of GISs and the graphical user interface of DSSs complicates the user interface of a SDSS, which allows the user to take complete control of data input and manipulation. The sophisticated user interfaces can provide user-defined triggers, which allow the user to dictate how features will respond to environmental changes, and to construct rules to control the modeling process (Crosbie,1996). The ease and flexibility in which any water resource system can be defined, modified and visualized through the designed interface should bring ease and flexibility to the modeling and result analysis (Loucks et al.,1996).  

    The coupling of environmental models with a GIS is the major issue in a spatail decision support system (SDSS). For water resources problem solving, both the spatial representation of water resource systems and the insight into water resource problems are necessary (Walsh, 1992). GISs have been applied to provide the former, water resource models provide the later, and SDSSs provide the integration of both. There are several strategies and approaches for the coupling of environmental models with a GIS (Nyerges, 1993; Fedra, 1996), which can range from loose to tight coupling. A loose coupling is just the transfer of data between models and GIS, and it is based on two separate systems and generally separate data management. A tight coupling is one with integrated data management, in which GIS and models share the same database. The tightest of couplings is an embedded or integrated system, in which modeling and data are embedded in a single manipulation framework (Crosbie, 1996, Fedra, 1996; Fedra and Kubat, 1993, Djokic and Maidment, 1993, McKinney et. al. 1993, Burgin, 1995, etc.)  

     Object orientation as an integrated approach seems to be very promising for deep coupling of GIS and environmental models (Fedra, 1996; Crosbie 1996; Raper and Livingstone 1996; Densham and Goodchild 1990;) The idea beyond this approach is that the world is perceived as consisting of objects that interact in specific ways (Crosbie,1996). Object-oriented representation of the world includes spatial objects and thematic objects. Spatial objects represent real world entities, and thematic objects include attributes, methods, and topics. Beside the spatial attributes that can be directly derived from a GIS, there are external physical, environmental and socio-economic information related to the spatial entities. Methods are rule sets or functions defined for description or exploration of the relationships over the spatial objects. Topics represent tasks or objectives to be completed or reached, which is often identified through user interactions. Based on the given attributes, models and GIS functions are understood as methods for topics, and the integration of models and GIS functions becomes the pragmatic question of which method can perform the required task on the selected objects (Fedra, 1996). Several applications on object-oriented SDSS have appeared in water resources literature (Reitsma 1996; Fedra and Jamieson, 1996; Loucks et al. 1995 etc.).  

     In this paper we present a prototype spatial decision support system for water resource allocation in river basins. A river basin is represented as spatial objects and thematic objects in a GIS. Several object-oriented GIS functions are developed, which include (1) Data input/update, manipulating thematic objects for data preparation; (2) Network derivation, automatically deriving an abstract river basin network from a river basin map, based on the spatial relationships within the real world represented by the map; and (3) Network modification, for users to modify the network and build the representation they prefer, and users can also use this tool to draw a network based on the river basin image in raster format, and (4) Model generation, automatically generating a water resource allocation model through interactions with users, and (5) System connection, allowing the SDSS to call any external programs, and (6) Visual display. All these functions are built into a graphic user interface. The data, models and all operations are integrated in a GIS environment (ARCVIEW), and this GIS-based decision support system can support multiple objective analysis for river basin management issues, like water resources allocation in the river basin.  

     In the rest of this paper, a detail introduction of the modeling system is presented. A case study for the Kashkadarya River basin in Central Asia is given. Emphasis will be given to the aspects of using database, interface and GIS in modeling techniques. 

  
System Design - General Introduction 

      The idea for the system design is to put the data, the model and the decision analysis process all together into the environment of the GIS. The GIS we use is Arcview, a GIS software package with a high-level object-oriented programming language, namely, Avenue. Figure 1 shows the structure of the GIS-based DSS. A river basin is represented by spatial objects, which represents the real world entities, thematic objects which include the network, attributes, logical and policy relations, and models. A mathematical programming model is generated based on the network, attributes, physical laws and control policies, and users' interaction. Currently the SDSS can generate two models. One is a linear optimization model for water resources allocation in a river basin without considering water quality, the other is non-linear optimization model with both water flow balance and salt balance. Other models, like economic models for river basin management are under development. For some models, like the optimization models for water allocation in a river basin, an external solver is needed. In our case, we use GAMS (General Algebraic Modeling System), to solve the optimization models. The system writes the model in the GAMS language. When the model is ready, The optimization solver GAMS is called to run in a window within GIS Arcview, and the results from GAMS are read into Arcview data base. 

     The GIS-based DSS is designed as an adaptive system, which can evolve as new spatial objects, and thematic objects including attributes, logical and policy relations, and models, and can also adapt user-defined operating rules or user-preferred changes are modified, added to, or deleted from the system. This adaptive ability makes the system general to water management problems in any river basin.  

      Since all operations are done within the GIS, all the primary functions of the GIS are available to users, and all the new functions, which are not in the current GIS, are developed in the object-oriented programming language of the GIS, namely Avenue, and they are embedded in the same interface as the primary GIS functions.                                  Figure 1 Structure of the SDSS  


Representation of A River Basin 

      The river basin used for a case study in this paper is the Kashkadarya River basin, which is a sub-basin of the Amudarya River basin in the Aral Sea region of Central Asia. Figure 2 shows a digitized map of the basin. A major canal, the Mosco canal, diverts water from the Kashkadarya River to a region outside of the basin, namely Mosk. Since the local water in the basin is not sufficient for water supply, another major canal (Karshi Canal) diverts water from the Amudarya River to the Kashkadarya River basin. There are several reservoirs in this basin, and the major ones include the Chmik Reservoir on the Kashkadarya River, and the Talim reservoir on the Kashi Canal. The reservoirs have enough capacity to control the normal natural inflow and diversion from the Amudarya River (McKinney and Karimov, 1997). Although the water demand is mainly supplied by surface water, groundwater in several aquifers in this basin plays an important role in water supply. There is intensive irrigated agiculture in this basin, and irrigation water occupies the major part of the total water demand in this basin. Figure 2 shows 8 demand sites, which are delineated based on the administrative districts and irrigation field distribution. For these demand sites, beside the water supply canals, there are collector canals that collect drainage from the irrigation fields and send it back to reservoirs, rivers or aquifers. 

     The natural semi-arid climate and the intensive irrigation create two typical problems in this basin: water quantity shortage and salinization. Our primary research purpose is to develop an efficient water allocation tool to aid in the resolution of the problems in this basin.  

Spatial Objects 

For our study purpose, the required spatial items in the river basin include (Figure 2):  
   1.the main river, divided into reaches, 

2.the tributaries and canals which divert water to the river basin, 

3.the canals, which divert water from the main river to demand sites, 

4.the collector canals, which collect agricultural drainage or other return flows from demand sites, 

5.the reservoirs and lakes,  

6.the groundwater sources (aquifers), 

7.the demand sites, which include cities and irrigation districts or fields,  

8.water treatment plants, 

9.hydropower stations, and 

10. administrative districts or regions.       To input these entities into the GIS-based DSS, one can digitize river basin maps using the GIS's digitizing capacity . An alternative way is to read the river basin image files in raster format into the GIS, but some of the system functions, like network derivation, does not work in this way.                                  Figure 2   A river basin map (the Karshikardaya River basin in Central Asia) 

      We define each of these entities as a spatial object, and we can use them to make any required spatial analysis, but generally, a mathematical model, which is an abstraction of the real world can not be built directly on real world maps. To overcome this obstacle, we will introduce thematic objects in the following.  

 Thematic Objects  

      We can use the GIS to extract a river basin network from the real world representation, and then the network becomes a bridge between the real world and mathematical models. In the river basin network, the spatial objects are treated as nodes, and the spatial relations between the objects are treated as directed arcs or links. In the river basin map, we see the spatial relations between two spatial objects, for example, a tributary intersects with a demand site; while in the network view, we see a link (line) between them, which means the tributary may supply water to the demand site. Figure 3 shows a network for the case study river basin. 

      Each of the nodes and links in the river basin network is defined as a thematic object. The node objects inherit spatial characteristics from the corresponding spatial objects. This inheritance will be addressed later in the discussion of network derivation functions. However, the link objects do not have any digital spatial characteristics, and they only specify "which node is linked to which other node". The links defined in a river basin network are listed in Table 1. 

Table 1 Links in a river basin network       Corresponding to each node object in the network, there are two sets of attributes, one is the spatial characteristics, like coordinates, length for lines, and area and perimeter for polygons, etc., and this set of attributes can be derived by the GIS directly; the other is any external information which has spatial characteristics, including additional physical information and social, economic and environmental information, and these attributes are available from related external sources and they should be input into the GIS. For example, for a reservoir object, the spatial characteristics include the reservoir's area, perimeter, and the coordinates of the reservoir area's center. The external information for a reservoir includes the reservoir topographic relations (e.g., elevation vs. area, elevation vs. volume), reservoir inflow and pollution load, average water temperature in different seasons, and some parameters related to reservoir operation for water supply, electric power generation, water release for ecological and environmental purpose, and flood control. Link objects, which specify "which is linked to which", do not have any spatial characteristics, but there is some external information associated with them. For example, for a link from an aquifer to a demand site, pumping capacity and maximum allowed pumping are necessary information for the link.       In the GIS, the attributes of a node or link object are stored in tables. For each node or link object, there is such an attribute table. In the GIS-based DSS, the attribute tables are also defined as thematic objects, subordinate to the node and link objects in the network.  

     We define another class of thematic objects to describe the physical laws that govern the spatial objects in the real world, and the socio-economic policies that control the operation of the river basin system. The network objects (node and link objects) should also obey these laws and policies. For example, water balance and salt mass balance for a reservoir node, or operation rules for a reservoir. This class of thematic objects represent advanced relations between each network object and any other related network objects. We name these objects relation objects. 

     Finally, models for river basin analysis are defined as a class of thematic objects, which represent a set of inquiring schemes in which the network objects (node and link objects), attribute objects, and relation objects are specified by the user interactions. We will illustrate how to construct the model objects in the following discussion on model generation function.  

     Through this system, a river basin is visually represented by spatial objects, and is logically represented by thematic objects. Visual representation allows users to see what the river basin looks like, while logical representation, which is logically and mathematically operational for users, provides a window for users to see what happens or what will happen in the river basin under various scenarios. This is what allows us to perform optimization or simulation, a logical program on the system to evaluate and control the system performance.  

  

Models for Water Allocation in a River Basin  

     Before we go further to discuss the functions of the GIS-based DSS, we discuss the models applied for water allocation decision analysis in a river basin. Currently we just consider optimization models. A multiple objective analysis (MOA) approach is used in the optimization models to deal with the complexity of water allocation involving multiple purposes.  

   Objectives 

     The optimization models may include multiple objectives, such as     

Users can select one or more, or all, of the above objectives. By specifying different objective weights, multiple objective analysis can be performed and tradeoffs between those objectives can be evaluated (McKinney and Cai, 1996).  

     Constraints 

     There are three kinds of constraints: physical constraints (e.g., mass balances), policy constraints (e.g., upper and lower bounds on variables), and system control constraints (e.g., to maintain feasibility). The physical constraints result from by the model generation function directly, based on the river basin network. The generation of policy and system control constraints requires user interactions. 

      The physical constraints are the major part of the model constraints. The concept of this kind of constraint is a mass balance, including water mass balance and salt mass balance in rivers and tributaries, reservoirs and lakes, aquifers and demand sites. Water flow and salt transport in the system are described by the mass balance equations. The physical constraints also include some physical limits, such as river and canal diversion capacity, groundwater pumping capacity, hydropower power generation capacity, and waste water treatment plant capacity. 

     The policy constraints are specified by setting upper and lower bounds of variables. For example, the water diversion from a reservoir to a demand site may not be allowed to exceed a given amount and the system control constraints are used to maintain feasibility of the model solution, and they are also used to control the model performance, e.g., making the solution more stable or more realistic.  

     Currently, the optimization models developed here include a linear model and a non-linear model. The linear model considers the water quantity balance, and the non-linear model can conjunctively consider water and salt mass balance, and it can also consider the non-linear power generation equations.  

Object-Oriented Functions 

  Data Input/Update  

     On a graphical screen showing the river basin network, the data input/update functions allow users to click any node or link object, and then a tabular window is opened for users to input/update external information for the node or link object that they selected. The experienced GIS users can use the attribute input methods to directly edit the attribute tables in the GIS. 

 Network Derivation 

     To write a mathematical model for the simulation or optimization of a river basin , one starts by extracting the system network from river basin maps, and traditionally this work is done by hand. The network deriving tool can automatically derive the abstracted river basin network from a river basin map, based on spatial relationships within the real world;  

     In the GIS , several spatial relations can be used to resolve problems related to location, such as proximity, adjacency, and containment. Those spatial relations are defined for a selected spatial object, the one doing the selecting, and a target spatial object, the one from which items are being selected. First, the network derivation tool searches all spatial objects, identifies them as nodes in the network, and transfers the spatial attributes from the spatial objects (parent objects) to the nodes; second, the tool searches all spatial relations between the spatial objects, and identifies them as links in the network, and finally, all nodes and links are defined as thematic objects and are displayed in a view. Figure 4 shows a simplified example for deriving a network.    
     The spatial relation functions that Arcview provides include:                 Intersect - selects features in the target themes that intersect the features in the target. Intersection implies that at least one point is common to both the selector and the target or one of them is  
                               completely within the other. If the selector and target are the same, Intersect will select adjacent features; and 

            Are Within Distance Of - selects features in the target themes that are within a specified distance of the selector theme's features. We can specify the type of distance units in the View Properties dialog 
                              box. 

      We incorporate all these spatial relation functions into an Avenue program, which can be used to search and identify every kind of spatial relations in a coverage, and represent those relations as corresponding links in a network. For some cases, the system can specify the links quantitatively. For example, considering the links from a demand site to aquifers, because one demand site may spatially intersect with more that one aquifer. To measure the potions that the demand site overlays different aquifers, a grid coverage is designed, and the system identifies the common grids that the demand site and each of the aquifers, and then calculate the intersected areas by the number of the grids and the grid size.  

      If the river basin map is input as a raster image file, then the network deriving tool will not work, and users need to draw the network according to entities shown in the image by using the network modifying tool, which is introduced below. 


Network Modification  

      The network deriving function identifies links only by spatial relations, such as proximity, adjacency, and containment. For some cases, the links may not exist in the real world. For example, if a tributary intersects a demand site on the map, the network deriving tool will define a link from the tributary to the demand site, which means the tributary supplies water to the demand site. However in the real world, this supply may not exist because of some physical, social and political limits. For another case, Since the river basin map may only represent the current conditions, and user may want to plan some changes to the system, the network deriving tool will not define network nodes and links for the variants since they may not exist in the original map. In these cases, it is necessary for the user to use a tool to modify or update the network for this purpose. The network modify tool allows users add (delete) new (existing) nodes or links to build a variant network. The tool is designed as a user-defined trigger (Crosbie, 1996), by which users can build the model construction to suit their interests. When an existing node or link is deleted, the attributes associated with it will be made inactive; while when a new node or link is added, a window prompt will allow the user to input any required external information. 

Model Generation 

      As mentioned in the river basin representation section, models for river basin analysis are also defined as a class of thematic objects. The generation of this class of thematic objects is shown in Figure 5. The model generation function generates the GAMS codes for the optimization model. The generating process involves user interactions, such as selecting the model type (linear or nonlinear), specifying initial conditions and targets for objectives, setting preferences (weights) for the objectives, and setting policy control constraints.  

      The model objects inherit characteristics from network objects, attribute objects and relation objects, and they are also affected by user interactions. Therefore, the model objects represent variants for any change in the network configuration, input data, control policies, or user interactions.  

      The spatial objects, i.e. the node in the network, are the central objects in the model generation function. They are classified into classes corresponding to the physical entities, including reservoirs, river reaches, tributaries, aquifers, canals, treatment plants, and demand sites. The model generation function takes action on each member in each of these classes. The procedures for the model generation function are illustrated in the following table. 

Table 2 Procedures for the model generation function                Figure 5   Model generation process in the SDSS 


System Connection  

     In the SDSS prototype developed here, a commercial optimization package, GAMS, is used to solve the models. The system connection function can call GAMS from within Arcview. When the GAMS codes for the optimization model is prepared by the model generation function, the system connection function calls GAMS, and within Arcview a window is opened to show GAMS solving information. When GAMS completes execution, the system connection function writes the solution into the attribute objects.  

Visual Display  

    Besides the primary display capacity of the GIS , an enhanced display function can display error, help or progress information during the execution of all the functions described above; and it can display primary data or modeling result in tables or charts, for each network (node and link) object that the user selects.  

User Interface 

   Since the prototype SDSS is designed within GIS Arcview, and the interface of Arcview is available to users. New interface is customized by using Avenue for the specific purposes of the prototype SDSS. Through the new interface, a user can easily conduct the functions described above, including data input or update, network derivation and modification, model construction and operation, and result display. The interface provides help information such as step by step procedures, options for some alternatives such as model types and initial condition for modeling, and visual view of modeling processes. Various graphics, message boxes, menus, and windows provide friendly interface to users. 

   The interface is especially significant for model construction process. As described in model generation, the model construction process involves the interactions between users and the SDSS. River basin management models are made up not only by physical and socio-economic issues, but also by some individual decision preference. During the model construction process, the SDSS asks a user some questions, or prompts to him some options. The user's answers to the questions or choices to the options are incorporated into the model. For example, the SDSS may ask the user if he wants to set a goal for the salt concentration in the downstream flow, or set a limit for pumping from a aquifer; if the user chooses multiple objectives for the river basin management, then he will be asked to specify a weight for every objective. The interface designed in the SDSS provides a dialogue environment between the system and users. 

  
SDSS Operation Procedures 

1. Activate ARCVIEW, and open the project for the SDSS. Figure 6 shows the menus, tools, and buttons designed for the SDSS.               Figure 6  Menus, tools and buttons  
  

      2. load a river basin geographic themes into a map view. The geographic themes respectively represent rivers, canals, reservoirs and lakes, aquifers and recharge zones, treatment plants, irrigation fields, and towns and cities, etc.  

      3. Create a network from the geographic themes in two steps: adding nodes and adding links. First the SDSS searches each kind of nodes corresponding the geographic themes and add these nodes into a network view. Secondly, the SDSS examines the spatial relationships between the geographic themes, identify links based on the spatial relationships, and add these links into a network view. Figure 3 shows an example for a map view and a network view. 
  
     4. Modify the network There are several tools for users to modify a network, which include "add nodes tool", "add links tool", and "delete nodes/links tool". To add a node, a user first click a point in the network view to specify where to add the node, and then specify the type (a reservoir, an aquifer, a river reach, etc.) and the name for node. When a new node is added, the data records for this node are automatically added in the data base, and the system will mention the user to input the necessary information for this new node. To add a link, the SDSS asks the user to specify the type of the link to be added ( like a link from a reservoir to a demand site), and then asks the users to click the from-node (like a reservoir node), and the to-node (like a demand site). If the node that the user clicks is not compatible to the link type, then the SDSS will give error message, and inform the user to click a right node. To delete a node or a link, the user just activates the "delete nodes/links tool", and then click the node or link to be deleted. If an existing node or a link is deleted, the information with this node or link in the data base will be set to be inactive.  

     5. Input/Update Data. To input or update the attribute data, activate the "data input & update" tool, and then select the node or link, for which the information will be input/updated. The system will show the default data in a table and in a chart. The user can modify the data in the table or just take the default value A simple way for experienced ARCVIEW users to input/update data is to edit the corresponding attribute tables directly.  

     6. Construct a model Activate the "model generating tool" and select the model type. Figure 7 gives a demonstration for the interactions between the system and users during the model constructing process, which include specifying the initial modeling conditions, setting policy targets, selecting management objectives and specifying objective weights etc. The system gives step-by-step "what to do" message to users, and users just need to follow the steps, and answer the questions, or make the choices. For some cases, the system provides default values or conditions, and the user either take or modify them.  
Figure 7 (1) - (7) Interactions between the system and a user during model constructing process


     7. Call a model solver. When the model is ready, GAMS is called to run in a window within Arcview, and the results from GAMS are read into Arcview data base.  

     8. View Results click the "result view" button, and then follow the message boxes to make some choices and see the results that are of interest. The results are displayed as charts, tables, and message boxes. Figure 11 shows some examples for the result view. 
 

Figure 8 (1) - (3)  Examples for Result view

Summary 

     In this paper we give a prototype GIS-based decision support system for river basin management, and we use this prototype system to show how the spatial decision support system (SDSS), which is a conjunctive application of GIS and DSS technologies, can improve water resources management in a river basin. An object-oriented approach is used to integrate a river basin visual representation and logical representation into an operational framework, in which management analysis can be made based on the river basin spatial dimension, physical laws embedded in the river basin system, and management policies and decision preference.  

     The prototype system provides several GIS-based functions, among which the network deriving function and model generating function developed in this prototype allows some work in river basin management to be done by computers automatically, rather than by hand in the traditional way. Of course the most significant thing is that the models generated in the prototype are tightly connected to the real world spatial entities and their attribute data, so that the system allows users to do water management analysis in a more efficient and convenient way. The further development of these functions should be promising for water resources planning and management. 

     This prototype system is still under development. More work is needed to complete the current prototype. For example, the model generation function can only generate optimization models, and this function is to be updated so that it can generate appropriate hydrologic simulation models (Yeh, 1996) water quality simulation models, and economic models for river basin management.  

References 

Bennett, D. A., Armstrong, M. P., and Weirich, F. (1996), "An object-oriented model base management  

system for environmental simulation", In GIS and Environmental Modeling: Progress and Research  

Issues, edited by M. F. Goodchild et al., GIS World Book, Fort Collins, 1996, 439-443. 

Brooke, A., Kendrick, D., and Meeraus, A.(1988). GAMS: a User's Guide, Scientific Press, san Fransisco. 

Burgin, J., (1995), Automating the allocation of water supplies in texas using an expert system to control  

the interaction of a geographic information system and LP solution algorithm, Ph.D. diss. , the University of Texas at Austin.  

Csillag, F (1996), "Variation on hierarchies: toward linking and integrating structures", In GIS and 

Environmental Modeling: Progress and Research Issues, edited by M. F. Goodchild et al., GIS World 

Book, Fort Collins, 1996, 433-437. 

Cowan, D. D., Grove, T. R. Mayfield, C. I. et al. (1996), "An integrative information framework for 

environmental management and research", In GIS and Environmental Modeling: Progress and Research 

Issues, edited by M. F. Goodchild et al., GIS World Book, Fort Collins, 1996, 423-427. 

Crosbie, P. (1996), "Object-oriented design of GIS: a new approach to environmental modeling", In GIS 

and Environmental Modeling: Progress and Research Issues, edited by M. F. Goodchild et al., GIS  

World Book, Fort Collins, 1996, 383-386. 

Densham, P. J., and Goodchild, M. F. (1989), "Spatial decision support systems: a research agency." 

GIS/LIS'89 Proc. Vol2, American Congress on Surveying and Mapping, Bethesda, Md. 

Djokic, D., and Maidment, D. R. (1993), "Creating an expert geographical information system: the 

ArcInfo - Nexpert object interface", In ASCE Monograph on Integration Issues in Expert Systems 

Technology. New York: ASCE. 

Esri (Environmental System Research Institute, Inc., Redlands, Carlifornia, USA), Arcview 2.1 (3.0) 

User's Manual, 1995 (1996). 

Fedra, K, "Distributed models and embedded GIS: integration strategies and case studies", In GIS and 

Environmental Modeling: Progress and Research Issues, edited by M. F. Goodchild et al., GIS World 

Book, Fort Collins, 1996, 414-417. 

Fedra, K. and Jamieson, D. G., (1996), "An object-oriented approach to model integration: a river basin 

information system example", HydroGIS'96, IAHS publ. no 235, 1996, 669-676. 

Goodchild, M. F. (1993), "Data models and data quality: problems and prospects", In Environmental 

Modeling with GIS, edited by M. F. Goodchild, B.O. Parks, and L. T. Steyaert, 8-15. New York: Oxford 

University Press. 

Lam, D. C., and Swayne, D. A., (1991), "Integrated database, spreadsheet, graphics, GIS, statistics, 

simulation models, and expert systems: experience with the RAISON system on microcomputers" In 

NATO, ASI Series, Vol. 26, edited by D. P. Loucks and J. R. de Costa, 429-59, Germany.  

Leipnik, M. R., Kemp, K. K. and Loaiciga H. A. (1992), "Implementation of GIS for water resources 

planning and management", J. Wat. Resour. plng. Mgmt. : 119 (2), 184-205. 

Loucks, D. P., French, P. N., and Taylor, M. R., (1996), "Development and use of map-based simulation 

shells for creating shared-vision models", HydroGIS'96, IAHS publ. no 235, 1996, 695-702. 

Loucks, D. P. , and da Costa, J. R. (1991). Decision support systems water resources planning, NATO  

ASI Series, Springer-Verlag, Berlin, Germany. 

Mckinney, D. C., and Cai, X. "Multiobjective optimization model for water allocation in the Aral Sea 

Basin", the 2nd American Institute of Hydrology (AIH) and Tashkent Institute of Engineers for 

Irrigation (IHE) conjunct conf. on the Aral Sea basin water resources problems, Tashkent, Uzbekistan, 

July, 1996.  

Mckinney, D.C., D. R. Maidment, and M.Tanriverdi,"Expert geographic information sytem for Texas water  

planning", J. of Wat. Resour. Plng. And Mgmt.,119 (2): 170-183. 

Nyerges, T. L. (1993), "Understanding the scope of GIS: its relationship to environmental modeling", In 

Environmental Modeling with GIS, edited by M. F. Goodchild, B.O. Parks, and L. T. Steyaert, 75-84. 

New York: Oxford University Press. 

Pundt, H. , Hitchcock, A., Bluhm, M., and Streit, U. A., (1996), "A GIS-supported freshwater information 

system including a pen-computer component for field data recording", HydroGIS'96, IAHS publ. no.  

235, 1996, 703-711. 

Raper, J., and Livingstone, D. (1996), "High-level coupling of GIS and environmental process modeling", 

In GIS and Environmental Modeling: Progress and Research Issues, edited by M. F. Goodchild et al., 

GIS World Book, Fort Collins, 1996, 387-390.  

Reitsma, R. (1996), "Bootstrapping river basin models with object orientation and GIS topology", In GIS 

and Environmental Modeling: Progress and Research Issues, edited by M. F. Goodchild et al., GIS 

World Book, Fort Collins, 1996, 457-461. 

Walsh, M. R. (1992), "Toward spatial decision support systems in water resources", J. of Wat. Resour. 

Plng. and Mgmt. 109 (2): 158-169. 

Yeh, Zichung (1996), "Object-oriented hydrologic modeling ", Ph.D. diss., the University of Texas at Austin.  

  

Authors: 

Daene C. McKinney, Associate Professor Dept. of Civil Engineering, the University of Texas at Austin, Austin, Tx 78712; telephone: (512) 471-1807, email address: daene_mckinney@mail.utexas.edu. 

Ximing Cai, Graduate Research Assistant, Dept. of Civil Engineering, the University of Texas at Austin, Austin, Tx 78712; telphone: (512) 471-0073, email address: xcai@crwr.utexas.edu. 

David R. Maidment, Professor, Dept. of Civil Engineering, the University of Texas at Austin, Austin, Tx 78712; telephone: (512) 471-0065, email address: maidment@crwr.utexas.edu.