The Use of ARC/INFO GIS in the Building and Development of a Finite Element Ground/Surface Water Model (WestSIM) in California's Western San Joaquin Valley.

 This paper presents some of the methods used to develop a finite element ground/surface water model using a GIS. The GIS is used to create the mesh, analyze and populate much of the data, and develop a Graphical User Interface to edit and view various model layers. The purpose of the model is to create a tool that can attempt to better understand water movement in the San Joaquin Valley.

To build the model mesh, ArcInfo AML macro routines were developed to systematically build the vector based mesh as constrained by a set of rules. After the mesh is generated, it is viewed in conjunction with other ancillary data sets to determine which sets of elements are to be grouped into analysis units (subregions).

The GIS is used to estimate conjunctive use per element by performing weighted area calculations on those fields that fall fully or partially within a model element. It is also used to perform complex stream node allocations to the model nodes and elements.

ArcView applications and lightweight MapObjects (Esri) applications have been developed to facilitate input data development, review, editing, and refinement.

KEYWORDS: GIS, MODELING, FINITE ELEMENT, ArcInfo, GROUND WATER

For the past decade a tremendous amount of Hydrologic Modeling has been done with some degree of geo-spatial reference being made important and integral to the modeling effort. With the advent of GIS, adapting geo-spatial databases and methods has proven to be a valuable, and indeed at times, even a requirement. As a key player in California Water Resources, the U.S. Bureau of Reclamation (Reclamation) has had the need to model a myriad of hydrologic systems throughout the Central Valley of California, where most of the water resources are in demand for agricultural and municipal uses. This paper focuses on the integration of GIS to a model built to address issues related to ground water usage and land subsidence in the San Joaquin Valley region of California.

Historically, the San Joaquin Valley has experienced one of the largest volumes of land subsidence in the world. Land subsidence in the San Joaquin Valley is a consequence of declining ground water levels in aquifers that contain appreciable quantities of saturated, compressible clay. Pumping in excess of recharge, a situation that is projected to become more acute, causes declining ground water levels. Reclamation has begun a critical study to determine the location and quantity of future expected land subsidence in the San Joaquin Valley especially in areas where Reclamation has responsibilities for water deliveries. The study hopes to assess the water delivery disruptions, capacity reductions, and facility damage that would be caused by that land subsidence.

 

Part of this study involved the creation of the Western San Joaquin Integrated Ground water Surface Water Model (WESTSIM). This model is built using the Integrated Ground-Surface Water Model (IGSM) initially developed at the University of California Los Angeles (UCLA) by Dr. Young S. Yoon in 1976. It was used for studying small ground water basins in the 80's and subsequently converted to include surface water modeling as well. IGSM is a two-dimensional comprehensive numerical model which simulates various components of the hydrologic cycle and how they interact with each other. Essentially, IGSM performs a hydrologic water balance and has many important features built into it such as ground water flow simulation, surface flow simulation, soil moisture accounting, unsaturated flow simulation, stream/aquifer interactions, land use analysis, reservoir operation and water quality simulation. This finite element model structure is best suited for basin management, agricultural and urban use, site-specific analysis, ground water availability and conjunctive use operation. As with most models that simulate the real world, IGSM requires a significant amount of input data that inherently carries influence based on geo-spatial distribution.

 

Data items such as water demand based on land use, available surface water supply, types of land use, ground water pumping, urban water use and water diversions are all important input to IGSM. All these items inherently have a common need of having to have a geo-spatial relationship that is correct and meaningful for the proper interaction within the model geometry. A hydrologic model is based on mathematical equations attempting to describe the actual physical processes of the hydrologic cycle. However, a mathematical environmental model devoid of physical geo-spatial foundations has no place in modeling. "For a good mathematical model it is not enough to work well. It must work well for the right reasons. It must reflect, even if only in a simplified form, the essential features of the physical prototype." .V. Klemes; WRR(1986).

 

The remainder of this paper will be dedicated to how the physical geo-spatial characteristics of the model were addressed within the GIS. The GIS is comprised of ArcInfo (including GRID and TIN modules), and ArcView (including Spatial Analyst and 3-D Analyst). These software systems were used to create the initial model elements and nodes, provide input data, identify problems, and provide a variety of maps and reports.

 

IGSM works by using the finite element numerical technique to solve the governing ground water flow equations. It can simulate any combination of multiple layered confined, unconfined and leaky aquifer systems. The model contains a soil moisture accounting technique, which can simulate direct runoff, evapotranspiration, and percolation resulting from irrigation or rainfall. It also allows for configuring various water diversions and pumping schemes. Stream flow simulation is achieved by dividing the stream into segments bounded by stream nodes that are coincident with ground water nodes. The structure of the model consists of a series of elements connected by nodes. The elements carry a variety of characteristics that affect how the nodes interact with them in tracking the movement of water throughout the system. The model typically uses a monthly time step for basin planning. Important data input to the model include rainfall data, land use data, surface water diversion, ground water pumping, and urban water use. IGSM has multiple levels of interaction, which consist of surface water application, the soil zone, the unsaturated zone, and the aquifer. Mass balance, water budget accounting, governing ground water flow simulation equations are computed for the layers and interactions are conducted through the nodes that connect the elements. Data are provided to the model via a series of formatted files. The most widely used data sets from the GIS have been stream and reservoir locations, land use data by specific crop type categories and selected habitat types.

 

View of an IGSM Structure

 

View of Hydrologic Cycle in IGSM

 

At the onset of this effort, decisive action was taken to create the model element and node geometry with the GIS to ensure spatial alignment to important hydrologic features and water district boundaries. It was also important to spatially correlate with an existing MODFLOW ground water model study. The MODFLOW model of interest was the Belitz model. The Belitz model was referenced for recharge calibration and data exchange purposes. The first order of business was to deliver a working map showing all the features the model development team would need. Using such a base map, the modelers were able to make a decision on the geographic location of the finite element mesh. Upon realizing the geographic scope and extent, the GIS lab generated an even-sized square mile element-node upright grid mesh. The mesh was then rotated in space to conform to the physical layout of the valley as well as line up with the mesh for the Belitz model. This was done using some geometric calculations with parameters obtained from the original base map. After the rotational positioning of the mesh, the base map was updated by the modelers to reshape some elements as triangles to better describe water districts and other hydrologic features. The mesh coverage was subsequently updated. The element and node configurations were built for the model using Arc Macro Language (AML) routines. The routines were designed to build unique numbered elements and nodes and know when elements were square or triangular. The final model mesh had 2716 elements with 2602 nodes. Once this procedure was completed and verified, the water budgeting subregions were developed using spatial dissolving methods. For this model, the subregions were designed to follow the water district boundaries as closely as possible.

 

View of model mesh development and rotation

 

View of hydrologic elements and model position

 

View of model mesh with Belitz model: Note the red mesh representing the Belitz MODFLOW model

 

View of final refined model with Water Districts

 

View of model with subregions The subregions boundaries were formed by using water district boundaries. The subregions are where the model performs water budgeting using monthly time steps. Water demand and conjunctive use calculations are performed primarily from information gleaned from land use crop summary information derived from the GIS.

 

 

Within our GIS there exists an ample amount of detailed crop land use data for many of the agricultural counties in the Central Valley of California. The California Department of Water Resources (CDWR) collects this data every five years for each county. The fact that the crop data may not reflect the most recent condition in the counties was found acceptable, as most crop use does not change rapidly. For the purposes of this model, the several county crop land use coverages that intersected the model were map-joined and prepared for overlay with the model subregions. Prior to the union overlay process, the crop codes within the CDWR domain were selected and grouped as to the 17 crop type domains IGSM needed. These crop type domains are developed from categorizing crop type into Agricultural, Urban, Riparian and Native Vegetation areas. The Agricultural categories are further broken down into groups of crops based on the type of water consumption they have. For example, rice and tomatoes are classified separately instead of being grouped into large classes such as truck crops or field crops. This filtering task was automated using AML macros. Once the IGSM crop type field was populated, the land use coverage was overlaid with the model subregions. After the overlay process was verified and checked, using frequency summarizing of intersected crop type per subregion, a datafile was created suitable for IGSM model input.

 

View of Crop Type within WestSIM This area is zoomed into the Westlands and Coalinga water districts in the southern portion of the San Joaquin Valley. The subregions are highlighted in a thick black line, and the canals are depicted in purple.

 

 

One of the needs of the IGSM stream flow/ground water  interaction simulator is the need to know which stream node each of the nodes within the model drains into. This would then allow the grouping of elements into stream node drainage units. Using the hydrologic features available within the GIS, the modelers came up with nodes that are designated as stream nodes. These nodes were isolated out into a new stream node coverage. The idea was for a given model node to traverse the terrain using flow direction rules until it reached a stream node and then assign itself the stream node unique identifier. This had to be done for 2602 nodes. Automation was critical once again. A flow-direction grid was computed for the model area from the digital elevation model (DEM). Several issues had to be addressed to avoid runaway traversing and missing closest stream node assignment. As a selected node traversed the terrain using the flow-direction grid as a guide, it will eventually approach a stream node, thereby implying the presence of a stream channel. Once it arrives at the channel, it will continue to travel following the elevation/flow direction profile, which does not actually reflect the stream. Therefore, there has to be a way to assign this node the closest stream node available, and also be aware that it has landed into a stream channel. In order to achieve this, the known streams where stream nodes existed were buffered to 1000 feet and converted to an ArcInfo grid. A Euclidean point-distance grid was also computed for the extent of the model area. Each cell in this grid contained the identification number of the closest stream node relative to itself. An AML routine was developed to cycle through the nodes in the model and find its corresponding stream node. The routine traversed the selected node over the terrain using the flow-direction grid, checking cell position for stream channel presence. The stream channel grid is a simple grid that has the value of 100 in valid cells representing stream channel and 0 otherwise. Once a query for stream channel verifies presence, the node is assigned the stream node value stored in the point-distance-allocation grid. Once this process was completed, the elements of the model were aggregated by the stream nodes they were assigned, and dissolved out to create polygons that had common stream nodes.

 

View of Nodes and Stream Node Buffer

 

View of Node Allocation system :Although it is built for the model boundary area, this node allocation system will be used within stream channels only.

 

View of Stream Node Assignment :Note node number 1456 on the westside traverses to Stream node number 46 on the eastside.

 

Animation of Stream Node Assignment

 

 

 

The subsurface layers of the model area are an important component in determining aquifer characteristics. The information describing these subsurface layers comes from well data and other published stratigraphy data. For the WestSIM model, seven layers were considered. The modeling team had a need to review the layer data and provided the GIS lab with the elevation profile files required to generate the TIN structures for each layer. Each TIN structure was loaded into ArcView 3-D Analyst for visual inspection. All the layers were also loaded simultaneously referencing the top surface and working downwards to better represent the relationship of the layers.

 

View of sub-surface layers with nodes

 

View of sub-surface layers with elements

 

 
 
 
 
 

The modeling process with WestSIM continues at the writing of this paper. The modelers have become more involved in using GIS tools to aid in visualization and calibration efforts. They are primarily using ArcView with Spatial Analyst. There are current needs to rerun the node allocation sequences since subsequent expert knowledge is available, and the nodes designated as stream nodes may change. There are efforts in place to create small watershed boundaries to simulate stream events outside the model boundary. This effort will use GIS techniques within ArcInfo and ArcView using detailed elevation profiles. The advantages gained in using GIS to prepare and visualize this model have been apparent to all parties involved. There are current works in progress to utilize such tools as Microsoft Access and Esri MapObjects to facilitate a graphical user interface (GUI) to IGSM. This will make data management and model rerun very efficient, as IGSM currently uses data files in ASCII format, which have been known to be cumbersome to use or update, never mind introducing errors.

Using GIS in environmental modeling has become a de-facto standard as more modeling professionals realize the power of GIS databases and software engines. The ability to link object oriented mapping technologies such as the new ArcInfo 8.0 Desktop suite and MapObjects with database objects opens a new realm into how modeling and GIS are going to become even more synergistic than ever before. It is not very difficult to imagine hydrologic features operating in a geo-spatial realm with object class methods and behaviors driven by modeling scenarios. We have already begun to visualize other IGSM models within a MapObjects environment where time series data are linked to model subregions. In the near future, it is anticipated that WestSIM will also inherit powerful object technology and become even more robust.
 
 

References or Acknowledgments:

Montgomery Watson Consulting Engineers Inc., Documentation and User Manual for Integrated Ground Water and Surface Water Model (IGSM), 1990

Belitz, Kenneth R., Character and evolution of the ground-water flow system in the central part of the western San Joaquin Valley, California, OF 87-0573, U.S.G.S.,1988

Grey, Larry, Hydrologist, U.S. Bureau of Reclamation, Sacramento, Ca.

Author Information:

Michael Sebhat
Project Manager, U.S. Bureau of Reclamation Mid-Pacific Region GIS Service Center
U.S. Bureau of Reclamation
2800 Cottage Way, Sacrmento, CA. 95825
(916)-978-5272
msebhat@mp.usbr.gov

Thomas Heinzer
Asst. Project Manager, U.S. Bureau of Reclamation Mid-Pacific Region GIS Service Center
U.S. Bureau of Reclamation
2800 Cottage Way, Sacrmento, CA. 95825
(916)-978-5273
theinzer@mp.usbr.gov