The ARC/INFO* geographic information system (GIS) was used along with the groundwater model MODFLOW, to model a groundwater aquifer system. GIS was used to create coverages and generate input to and display output from MODFLOW, a U.S. Geological Survey (USGS) finite-difference groundwater model. Point and polygon coverages of well locations, impervious area, active/inactive model areas, and boundary conditions were generated in ARC from a well-location data file or by digitizing. Surfaces of water-level elevation, bedrock surface, and saturated thickness were generated from elevation items in the well point coverage using the SPLINE function in GRID. The SPLINE function was chosen over TIN because it produced a more realistic surface and it extended interpolation past the boundaries of a triangular irregular network (TIN) and beyond the boundaries of the modeled area. A suite of Arc Macro Language (AML) programs was used to facilitate data input to MODFLOW and to visualize data output from the model. Two AMLs, REALGRIDARRAY and ZONEARRAY, automated the conversion of spatial data into ASCII input files for MODFLOW. Another pair of AMLs, SHADEMODFLOWCELLS and ROWCOLOFFSET, assisted in the visualization of data, and reformatted MODFLOW output for input into MODPATH3, a particle-tracking program. Because of the short timeframe, it is unlikely that model results could have been used to assist in interim corrective measure implementation without the use of GIS.
* The use of trade names is for descriptive purposes only and does not imply endorsement by the U.S. Government.
In 1996, the U.S. Geological Survey began site characterization and scenario testing for the installation of an interim corrective measure for a groundwater contaminant plume. The origin of the plume was two leaking underground tanks that were used from 1978 to 1988 to store waste solvent. Recent investigations have found that the alluvial groundwater plume extended beyond the eastern boundary of the facility.
An compliance order issued by the state required implementing an interim corrective measure in the summer of 1996 to contain the groundwater plume onsite. The selected containment method consisted of a subsurface treatment wall interrupted by four treatment gates; the wall funneled gGroundwater flow into the gates for in-place treatment of the groundwater as it passed through (Starr and Cherry, 1994). A series of site-characterization tasks were begun to determine specifications for the containment. Because of time constraints associated with the project, these characterizations were underway during mobilization of construction activities. Model findings were rapidly incorporated into the final implementation of the interim corrective measure.
A two-dimensional, finite-difference model of groundwater flow in the alluvium was developed by the USGS using MODFLOW (McDonald and Harbaugh, 1988). The model was based on limited data from previous investigations as well as new data from the current site-characterization studies. Several possible funnel-and-gate configurations were simulated using the flow model. The site-specific modeling allowed consideration of different environmental factors, such as lateral variations in aquifer thickness and hydraulic conductivity, as well as physical constraints on the location of the system. The final configuration was a wall approximately 1,200 feet long including four separate gates, each 40 feet wide, which is the largest multiple-gate and funnel system implemented to date. This paper discusses how ARC/INFO was used to assist the site-characterization and modeling effort and how the ease of editing model-input parameters and visualization in the GIS made modeling possible within a short timeframe.
A polygon coverage of a two-dimensional finite-difference mesh for MODFLOW was created, after a unit cell size was defined, using the AML MODELGRID (Kernodle and Winkless, 1996). The cell size chosen was 10 feet by 10 feet and resulted in a mesh of 390 rows by 400 columns having a total of 156,000 cells. MODELGRID created an arc coverage of the mesh lines, a polygon coverage of the mesh, and a point coverage of the polygon label points.
Grid surfaces of bedrock elevation (bottom of aquifer) and water-table elevation were generated using the GRID function SPLINE from an attributed point coverage of well locations, and a saturated thickness surface was calculated from the bedrock- and water-elevation grids. Vector coverages, also used as input to the model, included polygon coverages of impervious areas, polygons defining active and inactive portions of the model area, boundary conditions, polygons defining the spatial extent of remediation features, and polygons defining areas of different hydraulic conductivity.
Because the distribution of data points (well locations) around the study area did not extend to the edge of the modeled area, the GRID function SPLINE was used instead of the TIN module.
Map of modeled area showing locations of wells, the interim corrective measure, and extent of modeled area.
It was a requirement of this model that data extend beyond the irregular area interpolated by TIN all the way to the stream boundary to the south and to the model area boundaries to the west, north, and east. Because the spatial extent of an ARC/INFO grid is a box enclosing all the data points, the SPLINE function required little or no use of artificial points to force extrapolation to the boundaries of the active model area.
Map showing the differences in boundaries calculated by TIN and SPLINE interpolation methods.
The ASCII files required by MODFLOW were created using two AMLs; REALGRIDARRAY and ZONEARRAY (Jennifer Sieverling and Ned Banta, U.S. Geological Survey, written commun., 1996). Instead of selecting polygons and coding the 156,000-node mesh manually, the AMLs ZONEARRAY and REALGRIDARRAY were used to perform the coding automatically.
ZONEARRAY created input files that were based on vector coverages. Polygons defined discrete zones such as hydraulic conductivity, impervious areas, and active versus inactive areas of the model. In this model, a zone was usually a polygon attributed with an integer number to represent a feature or value. The AML executed ARC commands to overlay the various polygon themes with the model mesh to determine which zone each cell in the mesh belonged to for each theme. The appropriate zone value was added as an attribute to the polygon attribute table (PAT) of the model mesh.
Map showing sample overlay of the model mesh with a polygon coverage of hydraulic conductivity zones.
The AML then rearranges the values from the PAT and stores them in a sorted temporary INFO file. To create the ASCII file that would be given to MODFLOW, the AML passed the zone values in each cell from the temporary INFO file to a Fortran program. The Fortran program then output the values in a predefined format to a MODFLOW-readable ASCII file.
The second AML, REALGRIDARRAY, was used to create input files based on continuous surfaces such as water-table elevation, bedrock elevation, and saturated thickness. The AML executed GRID commands that rasterized the polygon coverage of the model mesh and used the resulting grid to compute zonal statistics (geometric mean or arithmetic mean) for each cell of the model mesh and added that value as an attribute to the PAT of the model mesh. As with ZONEARRAY, this AML passed values from an INFO file to a Fortran program that formatted the values for input into MODFLOW .
Example of an ASCII file created by running the AML RELAGRIDARRAY which required a file defining output format, a surface grid, and the model mesh coverage.
These AMLs made updating the input ASCII files for MODFLOW much easier than manually editing the thousands of numbers by simply running them again using updated coverage and grid themes. For example, editing the spatially referenced hydraulic conductivity coverage while it was overlaid on other coverage information and rerunning the parameterizing AMLs was much easier than trying to determine and modify corresponding numbers in the ASCII MODFLOW input files as cells values changed in the GIS. The AMLs were re-run as needed, as changes were made to surfaces and polygon coverages for model calibration and sensitivity analysis.
The MODFLOW model outputs both graphic and unformatted binary files. However, the graphics are primitive and are not spatially referenced. ARC/INFO was used to import these graphics and enhance and display them as graphics overlaid on a base map of the study area.
Map of MODFLOW-calculated water-table elevation contours displayed in ARC with other spatially referenced information.
The MODFLOW graphics files were CGM files that imported directly into ARC using the ARCPLOT command PLOT. Scaling the graphics was established by matching the boundary box of the MODFLOW graphic with that of the geo-referenced model mesh boundary.
MODFLOW also produces unformatted binary files that contain either model-calculated head or drawdown values. The AML, SHADEMODFLOWCELLS (Jennifer Sieverling and Ned Banta, U.S. Geological Survey, written commun., 1996), helped with visualization of these data. The AML generated a postscript file for a graphic that showed a gradational shading of the head or drawdown values. The postscript graphic also contained an annotated legend about the source data, data type, model layer number, model stress period, time step, and shading legend.
The final AML used from the model-interface suite was ROWCOLOFFSET (Jennifer Sieverling and Ned Banta, U.S. Geological Survey, written commun., 1996). This AML calculated the row and column identification of the mesh cell that each well point fell within and added it to the well coverage PAT. The x and y offset of the wells from the centroid of its corresponding mesh cell also was recorded in the well coverage PAT. These values are required by MODPATH3, a particle-tracking package that computes particle-path lines based on output from simulations created by MODFLOW (Pollock, 1989).
The display, editing, parameterization, and spatial referencing capabilities of ARC/INFO greatly minimized the amount of effort involved in preparing data for input into MODFLOW and visualizing data output from the model. Edits to coverages could be made quickly and the AMLs re-run to automatically change the input files for MODFLOW. Parameter changes, such as boundary conditions, hydrologic parameters, and configuration of the interim corrective measure, for scenario testing and model calibration, was facilitated by ARC/INFO; it was easier editing input data as coverages, and also the representation of the input data as spatially referenced data, instead of ASCII files, ensured a more easily visualized representation of the study area with overlaid related spatial data. The GIS also assisted in the visual exploration of model output so that any changes required in input parameters could be easily assessed.
Kernodle, J.M., and Winkless, D.O., 1996, MODELGRID.AML: Geographic Information System software to generate geo-referenced finite-difference-flow-model grids, electronically released program.
McDonald, M.G., and Harbaugh, A.W., 1988, A modular three-dimensional finite-difference groundwater flow model: Techniques of Water Resources Investigations of the U.S. Geological Survey, Book 6, Chapter A1.
Pollock, David W., 1989, Documentation of computer programs to compute and display pathlines using results from the U.S. Geological Survey modular three-dimensional finite-difference groundwater flow model: U.S. Geological Survey Open-File Report 89-381, 188 p.
Starr, R.C., and Cherry, J.A., 1994, In-situ remediation of contaminated groundwater-The funnel-and-gate system; Ground Water, v. 32, p. 465-476.
1 U.S. Geological Survey, Box 25046, MS 415, Denver Federal Center, Lakewood, CO 80225