Michael C. Brown, Project Engineer
Diane H. Davis, Project Hydrogeologist
Robert M. Greenwald, Principal Hydrogeologist
P. Srinivassan, Senior Engineer
David S. Ward, Vice President, Research and Development

ModelGIS was originally developed for the South Florida Water Management District (SFWMD) to facilitate the creation of a local groundwater model from a regional model. The code evolved from GWZOOM, which provided the technology to accommodate SFWMD's desire to create submodels from regional models. ModelGIS is an interface used to link ArcInfo to groundwater models. The U.S. Geological Survey (USGS) code MODFLOW is a modular finite-difference computer simulation code capable of simulating groundwater flow in three dimensions. It is an excellent tool used for the prediction of flow in complex groundwater systems. MODFLOW's modular structure allows for flexibility in its use, however the development of a complex model is limited by the lack of an efficient data pre-processor. The hydrogeologic complexity and detail of some sites facilitates the need for an interface to a batch oriented groundwater simulator. The use of a Geographic Information System (GIS) provides a powerful and efficient means of data preparation and visualization of simulation results. Arc Macro Language (AML) and Fortran 77 are used to create ModelGIS, which generates model grids, model layer elevations, aquifer properties, surface water data, and model output. ModelGIS may be used to compute regions of transmissivity and vertical leakance from hydrostratigraphic zones containing discrete properties of sand, silt, and clay. A case study is presented demonstrating the capabilities and flexibility of ModelGIS. This new approach allows for the development of a more accurate and complex numerical model that previously could not have been discretized to such a high level of detail.

Introduction

Groundwater models are powerful tools used to analyze many groundwater problems. For example, they may be used for reconnaissance studies prior to field investigations, data interpretation after field programs, predictive studies involving estimation of future groundwater system behavior, and, in combination with contaminant transport models, for evaluation of plume migration and remediation. An understanding of the hydrogeologic conditions of an aquifer system is necessary for the conceptualization and development of groundwater flow models. In order for groundwater flow models to be able to simulate actual aquifer behavior, they must have comprehensive data structures that allow for the utilization of various types of data describing the hydrogeologic conditions of the system. When, for example, parameter values describing the variability within the groundwater flow system show temporal trends in hydrologic stresses as well as past and future trends in water levels, there is a need to combine the spacial characteristics of the aquifer into a numerical representation. The complexities of the data make a Geographic Information System (GIS) a valuable tool for use in the development of models because of its ability to create, store, analyze, and present relational data.

ModelGIS, a GIS interface for the groundwater flow model MODFLOW, provides an efficient link between data stored in a GIS and the development of complex model data sets. Calculations of model parameters and construction of model layers are performed using simple menus. ModelGIS allows environmental specialists to focus on model conceptualization and analysis of groundwater flow by efficiently managing data within the GIS.

The most widely used three-dimensional groundwater flow model is MODFLOW. Developed by the U.S. Geological Survey (USGS), MODFLOW solves the three-dimensional groundwater equation using a block centered finite-difference technique. The program consists of a main module and a series of independent subroutines or modules. The modules are grouped into four primary packages: the basic package (BAS), detailing general grid information, the block center flow package (BCF), utilizing a block centered finite-difference formulation, the stress package (RIV, STR, WEL, DRN, EVR, and GHB), incorporating rivers, streams, wells, drains, evapotranspiration, and general head stress into the aquifer system, and the solver package (SIP, SOR, or PCG), an iterative solution to a system of simultaneous linear equations (McDonald, 1988). Individual features of the simulation are described by each package. In the well package for example, the effects of stresses on the system (i.e., pumping or injection) are simulated.

Case Study

In order to demonstrate the functionality of ModelGIS, a groundwater flow model was constructed utilizing hydrogeologic data for the Memphis Light, Gas, and Water Divisions (MLGW), Davis Well Field. The purpose of this model is to provide possible explanations for water quality degradation in two wells in the Davis Well Field. Located in the southwestern corner of Shelby County, Tennessee, the Davis Well Field is 1 of 10 producing fields operated by MLGW. Production began in 1971, with an estimated 13 million gallons per day (MGD) currently being withdrawn from 14 municipal wells. Production is from the Memphis aquifer. The hydrogeology in the vicinity of the site is quite complex, consisting of consolidated and semi-consolidated sediments of Cretaceous and Tertiary age dipping westward into the north-central portion of the Mississippi embayment, and southward along the axis of the trough of a broad syncline that approximates the Mississippi River. Loess covered bluffs serve both as a physiographic and hydrostratigraphic boundary between the fluvial deposits of the Gulf Coastal Plain to the west, and the alluvial deposits of the Mississippi Alluvial Plain to the east (Figure 1) (Parks, 1990).

Hydrostratigraphic units at the site include, from youngest to oldest, a shallow water-table aquifer, the Jackson-upper Claiborne confining unit, and the Memphis aquifer. The water table aquifer consists mainly of alluvial and fluvial deposits. Fluvial depositional features include point bars, abandoned channels and natural levees, characterized by sand, gravel, and

Figure 1. Location of the Davis Well Field and major physiographic features in the vicinity of the site.

A hydrogeologic connection, or absence of the confining unit between the water table aquifer and the confining unit, exists in a monitor well near the Davis Well Field (Brown, 1993). A possible explanation for increased leakage across the confining layer in this area could be explained by extensive pumping within the Davis Well Field, localized faulting, or erosional features formed as a result of the Mississippi River (Parks, 1990). These and many other hydrostratigraphic heterogeneities describing the groundwater system have been incorporated into model data sets via the techniques described in GWArc, the first component of ModelGIS.

ModelGIS Interface

Figure 2. Main ModelGIS menu.

Converting data from ArcInfo into MODFLOW data sets is accomplished using a series of ArcInfo commands, AML functions, and Fortran 77 programming. The ModelGIS interface is written using the standards that Esri developed for ArcTools (Esri's menu interface to ArcInfo) (Esri, 1994). Figure 2 shows the main ModelGIS menu. GWArc converts coverages to model based data while GWPre assembles the data into MODFLOW data packages. An option to execute the MODFLOW program is also available (GeoTrans, 1996). The Davis Well Field contains site features defined in the GIS as points (wells), lines (rivers and streams), polygons (lakes and hydraulic zones), and three-dimensional surfaces (groundwater elevations and hydrostratigraphic surfaces).

Once the GWArc module is invoked, a menu appears containing several options, shown in Figure 3. The construction of the finite-difference model grid is usually the first activity associated with the development of a model. The user is presented with the choice of creating a grid interactively or from a MODFLOW row and column file (found in the BCF package), Figure 3. ModelGIS has the flexibility to create a grid using uniform or variably spaced cells as well as a standard or rotated grid (GeoTrans, 1996). For this paper, MODFLOW row and column file was used to construct the model grid.

After creation of the grid, the user may construct model based coverages by selecting the Model Data option in the GWArc menu (Figure 3). The creation of matrix oriented data is accomplished by intersecting a polygon coverage or a three-dimensional

Figure 3. GWARC main menu and model grid operator's menu.

surface with the model grid. The Polygon Intersection menu, in Figure 4, allows the user to create matrix-oriented data from polygon coverages. An option is also presented to assign each grid block a value based on the grid center rule or a maximum area rule. Other types of data stored as polygons, such as recharge zones, wetlands, water bodies, or hydraulic conductivity zones, may also be intersected with the grid using the same procedure.

Three-dimensional surfaces most commonly created are potentiometric surfaces and hydrostratigraphic surfaces. The creation of matrix based coverages is accomplished by the intersection of polygon coverages or three-dimensional surfaces with the model grid (GeoTrans, 1996).

List oriented data is created by intersecting a point, line, or polygon coverage with the model grid. Wells, rivers, water

Figure 4. Polygon intersection menu allowing for the creation of matrix-oriented data from polygon coverages.

bodies, or hydrogeologic zones may be conceptually modeled by using MODFLOW's list oriented well, river, or drain packages. For MODFLOW's river package, layer, row, column, stage, and conductance are required for each cell defined as a river. In reality, river stage gage measurements may only exist at a few locations along a river. ModelGIS has the ability to linearly interpolate along the river between data points in order to provide data points for all intersected grid cells (Figure 5).

Another unique feature of ModelGIS allows for the generation of three-dimensional surfaces from existing hydrogeologic data. A five layer model was developed for the Davis Well Field from the interpretation of geophysical logs and two-dimensional profiles. The water table aquifer a sand and gravel unit and the Jackson-upper Claiborne confining unit, predominantly comprised of clay, were modeled as layers one and two, respectively. The Memphis Sand aquifer, modeled as three layers, is approximately 800-feet thick. Layers three and four each have a thickness of 100 feet, while layer five is approximately 600-feet thick. Within the Memphis Sand aquifer, well screens are variably placed due to the heterogeneity of this aquifer unit.

Figure 6 illustrates the process used to generate three-dimensional surfaces from two-dimensional profiles using ModelGIS. To determine layer elevations, data is extracted from each profile using a sampling interval, usually the same as the smallest grid spacing. Then, the elevation data is related to the same interval along the profile line (plan view) and stored as a point attribute. The resulting point coverage is used to generate a three-dimensional surface.

Once the structural aspects of the aquifer are included in the model, parameters describing the flow through the model layers must be assigned. The creation of a hydraulic conductivity surface based on a harmonic mean of the percentage of stratigraphic types is performed using a ModelGIS utility. A model profile containing the different hydrostratigraphic zones is intersected with a data sample

interval, which is then linked to the location along the profile. This procedure is similar to the previous method of converting profile data into a surface. A three-dimensional surface containing the harmonic mean data is then intersected with the model grid and multiplied by the hydraulic conductivity coverage. The use of the profile utility allows for the creation of data sets using data from all available sources, thus resulting in a model most accurately depicting the actual conditions.

After all model data sets are created, the user can invoke the GWPre module. This module allows for the preparation of matrix and list oriented data into MODFLOW format, Figure 7. The construction of the basic package, flow package, and other packages are performed in this module. The basic module allows the selection of other packages, determination of the grid size (based on grid dimensions in the GIS workspace), linking of boundary, and starting head arrays, as well as input of other non-matrix and non-list based model data. The BCF module has features similar to that of the basic module as well as built-in logic to assist in the creation of the BCF package. Depending on the layer type, ModelGIS automatically determines the files required for the package. Once all packages have been assembled, the user returns to the main module and executes MODFLOW (GeoTrans, 1996).

Once the model is executed, ModelGIS can also be used to evaluate model results and subsequently modify existing data. These changes may be made to individual model grid cells or an entire matrix based on model results. For example, a model run may generate dry nodes, or cells where the water level is below the model layer, in discrete locations due to the oscillation of the solution package. The resultant water levels can be read by ModelGIS. Based on the stratigraphic elevation of the layer, the water level at the locations of the dry nodes may be adjusted to ensure grid cell saturation. This new matrix of water level data can then be used for the next model run resulting in fewer dry nodes and a more rapid convergence.

ModelGIS can be used to develop data sets for most any finite-difference model such as FTWORK, MT3D, MOC, Prickett-Lonnquist, and Trescott-Pinder-Larson models. A unique feature of ModelGIS for example, is the ability to sum pumping rates for multiple wells in a grid block and average multiple water levels for those same wells within the same grid block. Depending on the type of data used, values within a grid cell may represent minimum, maximum, average, or a summation of multiple data entries within a grid cell.

Other Utilities and Future Modifications

Several modules currently exist in ModelGIS that were not used in this case study, but can be used in other model applications. Two specific modules include a coupled surface water and groundwater flow model (MODBRANCH), and a parameter estimation module (PEST). MODBRANCH consists of two programs, BRANCH a one-dimensional, numerical model used to simulate unsteady flow in open channel networks developed by the USGS and using the groundwater flow model MODFLOW. PEST, a nonlinear parameter estimation module, adjusts model parameters in order to minimize the discrepancies between observed and measured data. Modules currently under development include interfaces for the MODPATH particle pathline and endpoint processor and a new version of MODFLOW that simulates three-dimensional contaminant transport (MODFLOWT).

Conclusions

ModelGIS is a powerful tool that can be used to develop data sets for most finite difference models. It provides a dynamic connection between data stored in a GIS and the generation of complex data sets. Using simple menus, calculations of model parameters, construction of model layers, and development of other model features may be performed. From the interpretation of model results, ModelGIS can be used to modify individual grid cells or an entire matrix. Based on this case study, the development of a sophisticated numerical groundwater model can be easily developed using ModelGIS. The use of ModelGIS expedites the conversion of the hydrogeologic framework from GIS point, line, and polygon coverages into model based data. The modular construction of ModelGIS allows for the evolution of additional utilities, such as the contaminant transport code MODFLOWT and the parameter estimation code PEST, that assist in environmental studies.

Acknowledgments

The authors would like to thank Dr. John W. Smith and James E. Outlaw of the Ground Water Institute (GWI), University of Memphis for supplying the data used in this paper. Appreciation is also expressed to Alex H. Vincent of GeoTrans, Inc., for his technical assistance and support.

References

Brown, M.C., 1993. A Study of the Aquifer System at the Davis Well Field. Master's Thesis, Memphis State University.

Esri, 1994. Understanding GIS the ArcInfo Method. Version 7 for UNIX and Open VMS, Environmental Systems Research Institute, Redlands, California.

GeoTrans, Inc. 1996. ModelGIS User Documentation, Version 1.4 March 1996.

McDonald, M.G., and A.W. Harbaugh, 1988. A Modular Three-Dimensional Finite-Difference Ground-Water Flow Model. Techniques of Water Resources Investigations of the U.S. Geological Survey, Book 6, Chapter A.1, Denver, Colorado.

Parks, W.S., 1990. Hydrogeology and Preliminary Assessment of the Potential for Contamination of the Memphis aquifer in the Memphis area, Tennessee. U.S. Geological Survey Water Resources Investigations Report 90-4092.