Predicting Spatial Distributions of Vulnerability of Indiana State Aquifer Systems to Nitrate Leaching using a GIS

ABSTRACT

INTRODUCTION

OBJECTIVES

1. Evaluate Indiana's groundwater vulnerability to nitrate pollution potential using the DRASTIC and SEEPAGE models.
2. Determine the spatial auto-correlation of nitrate detections in groundwater across the state to determine whether point or non-point sources are likely the cause of such pollution.
3. Validate the accuracy of the approach by comparing the vulnerability maps with existing well water quality data sampled across the state.
   
   

   LITERATURE REVIEW

   DRASTIC
   
   DRASTIC is a groundwater quality index for evaluating the pollution potential of large areas using the hydrogeologic settings of the region (Aller et al., 1985, Aller et al., 1987, Deichert et al., 1992). This model was developed by EPA in the 1980's. DRASTIC includes various hydrogeologic settings which influence the pollution potential of a region. A hydrogeologic setting is defined as a mappable unit with common hydrogeologic characteristics. This model employs a numerical ranking system that assigns relative weights to various parameters that help in the evaluation of relative groundwater vulnerability to contamination. The hydrogeologic settings which make up the acronym DRASTIC are:

   [D] Depth to water table: Shallow water tables pose a greater chance for the contaminant to reach the groundwater surface as opposed to deep water tables.
   
   

   [R] Recharge (Net): Net recharge is the amount of water per unit area of the soil that percolates to the aquifer. This is the principal vehicle that transports the contaminant to the groundwater. The more the recharge, the greater the chances of the contaminant to be transported to the groundwater table.
   
   

   [A] Aquifer Media: The material of the aquifer determines the mobility of the contaminant through it. An increase in the time of travel of the pollutant through the aquifer results in more attenuation of the contaminant.
   
   

   [S] Soil Media: Soil media is the uppermost portion of the unsaturated / vadose zone characterized by significant biological activity. This along with the aquifer media decides the amount of percolating water to the groundwater surface. Soils with clays and silts have larger water holding capacity and thus increase the travel time of the contaminant through the root zone.
   
   

   [T] Topography (Slope): The higher the slope, the less is the pollution potential due to higher runoff and erosion rates which include the pollutants that infiltrate into the soil.
   
   

   [I] Impact of Vadose Zone: The unsaturated zone above the water table is referred to as the vadose zone. The texture of the vadose zone determines the time of travel of the contaminant through it. Authors of this model suggest that the layer that most restricts the flow of water be used.
   
   

   [C] Conductivity (Hydraulic): Hydraulic conductivity of the soil media determines amount of water percolating to the groundwater through the aquifer. For highly permeable soils, the travel time of pollutant is decreased within the aquifer.
   
   

   The major assumptions outlined in DRASTIC are:
   1. The contaminant is introduced at the surface
      
   2. The contaminant reaches groundwater by precipitation
      
   3. The contaminant has the mobility of water
      
   4. The area of the study site is greater than 100 acres
      
   DRASTIC evaluates pollution potential based on the above seven hydrogeologic settings. Each factor is assigned a weight based on its relative significance in affecting pollution potential. Each factor is further assigned a rating for different ranges of the values. The typical ratings range from 1-10 and the weights from 1-5. The DRASTIC Index, a measure of the pollution potential, is computed by summation of the products of rating and weights of each factor as follows:
   
   DRASTIC Index = DrDw + RrRw + ArAw + SrSw + TrTw + IrIw + CrCw
   
   Where
   
   Dr = Ratings to the depth to water table
   Dw = Weights assigned to the depth to water table
   Rr = Ratings for ranges of aquifer recharge
   Rw = Weights for the aquifer recharge
   Ar = Ratings assigned to aquifer media
   Aw = Weights assigned to aquifer media
   Sr = Ratings for the soil media
   Sw = Weights for soil media
   Tr = Ratings for topography (slope)
   Tw = Weights assigned to topography
   Ir = Ratings assigned to vadose zone
   Iw = Weights assigned to vadose zone
   Cr = Ratings for rates of hydraulic conductivity
   Cw = Weights given to hydraulic conductivity
   
   The higher the DRASTIC index, the greater the relative pollution potential. The DRASTIC index can be further divided into four categories: low, moderate, high, and very high. The sites with high and very high categories are more vulnerable to contamination and hence can be reviewed by a specialist. These weights are relative and a site with low pollution potential need not necessarily mean that it is free from groundwater contamination but it is relatively less susceptible to contamination compared to the sites with high or very high DRASTIC ratings.
   
   The USEPA (1992) analyzed the results of the National Survey of Pesticides in Drinking Water Wells data and the qualitative DRASTIC scores developed by Aller et al. (1987) at a 1:2,000,000 scale. County level DRASTIC scores (an aggregated score for each county) and subscores were computed and 90 counties were selected for analysis with nitrate data from wells sampled in the study. The results showed that DRASTIC performed very poorly for the selected counties. Hence for implementation of groundwater quality management plans at a regional scale more detailed vulnerability map is needed. The proposed study will use more detailed data at the 1:250,000 scale for estimating DRASTIC indices and additional data to improve vulnerability estimates.
   
   SEEPAGE
   
   The System for Early Evaluation of Pollution potential of Agricultural Groundwater Environments (SEEPAGE) model is a combination of three models that was adapted to meet SCS (Soil Conservation Service, recently renamed the Natural Resources Conservation Service) needs to assist field personnel (Moore et al., 1990, Richert et al. 1992, Engel et al. 1992). SEEPAGE considers various hydrogeologic settings and physical properties of the soil that affect groundwater vulnerability to pollution potential. This is also a numerical ranking model that considers contamination from both concentrated and dispersed sources.The SEEPAGE model considers the following parameters:
   
   
   1. Soil slope
   2. Depth to water table
   3. Vadose zone material
   4. Aquifer material
   5. Soil depth
   6. Attenuation potential
   The attenuation potential further considers the following factors:
   1. Texture of surface soil
   2. Texture of sub soil
   3. Surface layer pH
   4. Organic matter content of the surface
   5. Soil drainage class
   6. Soil permeability (Least permeable layer)
   Each factor is assigned a numerical weight ranging from 1-50 based on its relative significance, with the most significant parameter affecting the water quality assigned a weight of 50 and the least significant assigned a weight of 1. The weights are different for concentrated sources (site specific), and dispersed sources (non specific sources). Similar to DRASTIC, each of the factors can be divided into ranges and ratings assigned varying from 1-50. The ratings of the aquifer media and vadose zone are subjective and can be changed for a particular region. Once the scores of the six factors are obtained, these are summed to get the SEEPAGE Index Number (SIN). These values are representative of the pollution potential where a high SIN value implies relatively more vulnerability of the groundwater system to contamination. The SIN numbers are arranged into four categories of pollution potential: low, moderate, high, and very high. A high or very high SIN category indicates that the site has significant constraints for groundwater quality management (Richert et al., 1992).
   
   Engel (1992) used GRASS to carry out SEEPAGE analysis to evaluate the pollution potential of groundwater systems for the Kennedy Space Center, Florida. The data layers for carrying out the analyses were integrated in the GRASS GIS environment. In the research reported in this paper, SEEPAGE is used along with the DRASTIC model to carry out the regional scale studies. Both analyses were carried out in the ARC/Info GIS environment.
   
   

   METHODOLOGY

   Developing the Data Layers
   
   The STATSGO database from SCS comes at a scale of 1:250,000 and is distributed in different data formats. This database in the ARC/Info format was used in this study. The database is organized into map units which have as many as 21 components. These map components have information assigned to layers of soil horizons. Each of the layers are attributed various soil properties such as pH, organic matter content, etc. (SCS, 1992). Each of the properties is assigned a high and a low value for a mapunit. The STATSGO map for Indiana is available in the vector format and was used as the base map for the DRASTIC and SEEPAGE analyses. The hydrogeologic parameters required for the analysis were identified from the corresponding Info data tables and map layers for each of the layers were created using database management tools in ARC/Info. Codes were developed in Arc Macro Language (AML) to automate the process of extracting the information from the database files and assigning the corresponding ratings required for the analyses.
   
   The water table depth data layer was created by interpolation from a set of approximately 7200 data points of water table depth in well sites distributed across the state. Aquifer media and vadose media were extracted from the glacial geology map of Indiana. The aquifer recharge was computed using the percolation index (PI) in the NLEAP model (Deichert. et al., 1992, Follet et al., 1991, Follet et al., 1994). The hydrologic soil groups required for computing aquifer recharge were extracted from the STATSGO database and seasonal and annual precipitation distribution across Indiana was computed using forty year precipitation records from twenty weather stations. The land use data was obtained from the SCS and the fertilizer data extracted from agricultural statistics. These data layers, along with the hydrogeologic settings, were employed in this analysis.
   
   Carrying out the Analyses
   
   The data layers of the hydrogeologic settings were assigned the corresponding ratings [tables 1 and 2] and were converted from vector to raster layers in ARC/Info (ESRI 1992a, ESRI 1992b, and ESRI 1992c). A Graphical User Interface was developed using the form menus in ARC/Info for carrying out the analyses. Using the GUI the DRASTIC and SEEPAGE analyses were carried out. The DRASTIC and SEEPAGE analyses were modified by considering the additional data layers land use and fertilizer usage. The agricultural statistics were used to extract the crop information (type of crop, yield, harvest date, etc.), and the fertilizer applied on a county scale. Based on the crop N uptake and N content of fertilizer, the excess fertilizer applied on farm land was computed for the state. The DRASTIC Index map and SEEPAGE Index Number (SIN) map were reclassified into four categories: [tables 1 and 2] low, moderate, high and very high.
   
   

   DRASTIC Index	Low	Moderate	High	Very High
   Values	1-140	141-180	181-230	> 230

   
   Table 1: Reclass Table for Modified DRASTIC Ratings
   
   
   

   SEEPAGE Index	Low	Moderate	High	Very High
   Values	1-89	90-144	145-209	> 210

   
   Table 2: Reclass Table for Modified SEEPAGE Ratings
   
   Spatial Correlation Studies to Eliminate Point Source Pollutants
   
   Spatial auto-correlation refers to the spatial ordering of a single variable and to the relationship between pairs of observations of this variable. The ordering of n observed values of some variable X is usually described with the aid of a connectivity matrix, C. Non-zero cij entries in the n x n matrix indicate that the corresponding polygons are juxtaposed. For data measured on a interval / ratio scale the statistics geary ratio and moran coefficient can be used (Griffith, 1987).
   
   Geary Ratio
   
   The Geary Ratio is an index for interval / ratio data that is based upon paired comparisons of juxtaposed map values. It may be calculated as :
   
   The meaning of this index is fairly straight forward: as similar values tend to clump together (cij = 1), then the Geary ratio approaches to zero. If dissimilar values tend to clump together, the geary ratio approaches 2.
   
   Moran Coefficient
   
   Another index for interval / ratio based data is Moran Coefficient and may be calculated as :
   
   The expected value of MC is (-1/n-1). For similar values in juxtaposition, MC-> 1 and for dissimilar values in juxtaposition, MC -> -1. As the Geary ratio deals with paired comparisons, the Moran coefficient deals with covariations.
   
   As the DRASTIC and SEEPAGE analysis predict vulnerability from non-point source pollution, it was desirable to eliminate detections due to point source pollutions. The water quality data containing the nitrate detections in well sites was imported as a point coverage into ARC/Info and Thiessen polygons were created for the point coverage. The spatial auto-correlation statistics, Moran Coefficient and Geary Ratio were computed for the datasets. The Moran Coefficient value was computed as 0.79 and Geary Ratio as -0.06 which indicated that a spatial correlation exists among the detects. It is assumed that computation of these spatial statistics for different combinations of nitrate detections will help in detecting values that significantly affect the Geary Ratio and Moran Coefficient, which might be an indication of point source pollution.
   
   Validation of the Accuracy of the Regional Scale Vulnerability Maps
   
   The modified DRASTIC and SEEPAGE results were compared with nitrate detections in 380 well sites sampled across Indiana. The nitrate detections were categorized into four categories: Low 0-5 ppm; Moderate 5-15 ppm; High 15-30 ppm; Very high > 30 ppm. As the well water quality samples for nitrate detections were not uniformly distributed across the state, the modified DRASTIC and SEEPAGE vulnerability ratings at the corresponding sites were extracted and compared with the nitrate detections in the well water quality database (Dou and Woldt, 1994). The results from the modified DRASTIC and SEEPAGE analysis were compared with the results from conventional DRASTIC and SEEPAGE analyses [tables 3, 4, and 5].

   RESULTS

   The DRASTIC and SEEPAGE analyses were carried out in ARC/Info, using the GUI, to create the GIS layers shown in figures 1 and 2. The modified techniques considering the additional data layers land use and fertilizer usage were carried out in the grid sub module. The results from the conventional DRASTIC model indicate that 58% of the groundwater systems in Indiana fall under the moderate vulnerability category and 23% under high and very high pollution potential. There was a 24% increase in the areas categorized as low vulnerability using the modified DRASTIC approach. The conventional SEEPAGE approach predicted around 75% of the state having moderate vulnerability and considering the additional datalayers land use and fertilizer usage did not change the predictions. Both models predicted 50% or more of Indiana aquifer systems to have moderate vulnerability ratings.
   
   The results from the conventional and modified DRASTIC and SEEPAGE approaches were compared with the database containing observed nitrate detections in wells. The conventional DRASTIC and SEEPAGE models predicted 80% of the high and very high vulnerable areas correctly. There was a 20% increase in accuracy in predicting low vulnerability areas using the modified DRASTIC technique [table 4]. The consideration of additional data layers in SEEPAGE did not improve the accuracy of predictions [table 5].
   
   

   	DRASTIC Ratings
   Observed	Low	Moderate	High	Very High
   Low	25	146	138	0
   Moderate	1	17	63	0
   High	0	2	18	2
   Very High	0	2	2	1

   
   Table 3: Comparison of Conventional DRASTIC Ratings With Observed Nitrate Detections
   
   

   	Modified DRASTIC Ratings
   Observed	Low	Moderate	High	Very High
   Low	86	85	138	0
   Moderate	7	11	63	0
   High	0	1	20	1
   Very High	0	2	2	1

   
   Table 4: Comparison of Modified DRASTIC Ratings With Observed Nitrate Detections
   
   

   	SEEPAGE Ratings
   Observed	Low	Moderate	High	Very High
   Low	9	183	117	0
   Moderate	2	22	57	0
   High	0	2	20	0
   Very High	0	2	3	0

   
   Table 5: Comparison of Modified SEEPAGE Ratings With Observed Nitrate Detections
   
   The Pearson Correlation Coefficient (r) was computed as 0.61 between conventional DRASTIC predictions to the observed nitrate detections. The addition of data layers land use and fertilizer application improved the r value to 0.67. DRASTIC and SEEPAGE ratings were generally conservative in prediction. The areas predicted to have low vulnerability but which had observations of high or very high nitrate concentrations are of concern and are being investigated further. Such high observations of nitrates may be the result of point sources. Using SAS (Statistical Analysis Software) was used to carry out regression analysis on all the possible combinations of DRASTIC factors including the additional data layers to determine the best possible combination that predicts the nitrate detections accurately. The aquifer recharge data layer was found to have the least significant effect on nitrate detections and could be dropped from the model. This is largely due to the relatively small difference of this parameter within the study area. Studies are also underway to perform sensitivity analysis on the weights assigned to each of the factors.
   
   
   Figure 1: Groundwater Vulnerability Map to Nitrate Pollution Using Modified DRASTIC Technique
   
   
   Figure 2: Groundwater Vulnerability Map to Nitrate Pollution Using Modified SEEPAGE Technique
   

   SUMMARY

   Regional scale groundwater vulnerability studies to nitrate pollution from non-point sources using DRASTIC and SEEPAGE analyses were conducted at the 1:250,000 scale for Indiana. These models were integrated with the ARC/Info GIS environment. DRASTIC and SEEPAGE analyses categorized most of the Indiana aquifer systems as moderately vulnerable to pollution. Additional data layers, land use and fertilizer usage, were considered in DRASTIC and SEEPAGE in an attempt to improve the results. The results from conventional and modified DRASTIC and SEEPAGE analyses were compared with the nitrate detections in well water quality samples across the states. The results indicated that the models performed well in predicting the sites with high and very high nitrate detections.
   
   Statistics, Geary ratio and Moran coefficient were computed to assess the spatial auto-correlation of nitrate detections. The results indicated that spatial correlation exits between the occurrences of nitrate pollution. This suggests that regional-scale factors and processes are of importance and indicating that non-point sources of nitrates are important to the quality of groundwater.
   
   The modified DRASTIC and SEEPAGE analysis show a great deal of potential as screening tools for policy decision making in groundwater management. These analyses should not be used to replace detailed studies but should be applied to screen areas of high and very high vulnerability that a site-specialist should concentrate on. These analyses can be further modified by including additional data showing the excess amount of nitrate applied, crop rotations, and livestock nutrient value applied to farms. It is anticipated that this will improve the accuracy of the vulnerability estimates.
   
   GIS are an effective tool for carrying out groundwater vulnerability studies. The integration of the groundwater quality models within the GIS environment results in easy implementation of the analyses for different agricultural management practices. Automated data extraction to develop data layers required for the DRASTIC and SEEPAGE analysis from the STATSGO database will enable analysis for the whole Unites States at a 1:250,000 scale. When more data becomes available, the field scale simulations of more detailed models like NLEAP can be used for regional scale studies. It is anticipated that the studies on the spatial and scale variability of inputs on field scale models will lead to specific criteria with regards to level of detail of inputs required for conducting regional scale analyses.

   REFERENCES

   Aller, L., Bennett, T.,Lehr, J. H., and Petty, R.J., 1985, "DRASTIC : A standardized system for evaluating groundwater pollution potential using hydrogeologic settings," U.S. EPA, Robert S. Kerr Environmental Research Laboratory, Ada, OK, EPA/600/2-85/0108, 163pp.
   
   Aller, L., Bennet, T., Lehr, J. H., Petty, R. J., and Hackett, G., 1987, "DRASTIC : A standardized system for evaluating groundwater pollution potential using hydrogeologic settings," EPA-600/2-87-035.
   
   Deichert, L. A. and Hamlet, J. M., 1992, "Non-point groundwater pollution potential in Pennsylvania," ASAE International Winter Meeting, Nashville, Tennessee, 15-18 December, 1992, Paper No. 922531.
   
   Dou, C and Woldt., D. E., 1994, "Application of geostatistics for mapping nitrate contaminated groundwater," ASAE International Summer Meeting, Kansas City, Missouri, 20-23 June, 1994, Paper No. 942095.
   
   DRASTIC User Manual, 1987, Robert S. Kerr Environmental Laboratory, U.S. Environmental Protection Agency, ADA, OK.
   
   Engel, B. A., 1992, "Water quality modeling using Geographic Information System (GIS) data," 1992 NASA/ASEE summer faculty fellowship program, John F. Kennedy Space Center, University of Central Florida, Aug. 10, 1992.
   
   ESRI, 1992a, "ARC/Info user's guide," Cell based modeling with grid. March, 1992.
   
   ESRI, 1992b, "ARC/Info user's guide," AML user's guide. May, 1992.
   
   ESRI, 1992c, "ARC/Info command references," Arc command references. May, 1992.
   
   Evans, Barry M., 1990, "GIS in groundwater systems management," Journal of Soil and Water Conservation, Vol. 45:246-8.
   
   Follet, R.F., Keeny, D. R., and Cruse, R. M., 1991, "Managing nitrogen for groundwater quality and farm profitability," SSSA, Madison, WI.,pp. 285-322.
   
   Follet, R. F., Shaffer, M. J., Brodahl, M. K., and Reichman, G. A., 1994, "NLEAP simulation of residual soil nitrate for irrigated and non- irrigated corn," Journal of Soil and Water Conservation, Vol. 49(4):375-82.
   
   Griffith, D. A., 1987, "Spatial auto-correlation - A primer", Association of American Geographers.
   
   Indiana Department of Environmental Management, 1989, "Groundwater section," Indiana groundwater protection, A guide book. June 1989.
   
   Moore, and John S., 1990, "SEEPAGE: A system for early evaluation of the pollution potential of agricultural groundwater environments," USDA. SCS, Northeast Technical Center, Geology Technical Note 5.
   
   Richert, S. E., Young, S. E., and Johnson, C, 1992, "SEEPAGE: A GIS model for groundwater pollution potential," ASAE 1992 International Winter Meeting, Nashville, Tennessee, 15-18 December, 1992, Paper No. 922592.
   
   SCS, 1992, "State Soils Geographic Database (STATSGO) User's Guide, 1992," SCS, Publication No. 1492, August 1992.
   
   USEPA, 1990, "National Pesticide Survey: Phase I Report," USEPA Washington DC.
   
   USEPA, 1992, "Another Look: National Survey of Pesticides in Drinking Water Wells Phase II Report," USEPA 579/09-91-020. EPA Washington, DC.
   
   

   ---

   Kumar C.S Navulur, Graduate Research Assistant, Department of Agricultural and Biological Engineering, Purdue University, W. Lafayette, IN 47907-1146, ph 317-494-1196, email: navulur@ecn.purdue.edu
   
   Bernard A. Engel, Associate Professor, Department of Agricultural and Biological Engineering, Purdue University, W. Lafayette, IN 47907-1146, ph 317-494-1198, email: engelb@ecn.purdue.edu.