Nolan, B.T. Hitt, K.J. Ruddy, B.C. 2002 1 Raster digital data Environmental Science and Technology Volume 36, Number 10, Pages 2138-2145 Reston, Virginia, USA U.S. Geological Survey https://water.usgs.gov/lookup/getspatial?gwrisk This data set is a national map of predicted probability of nitrate contamination of shallow ground waters based on a logistic regression (LR) model. The LR model was used to predict the probability of nitrate contamination exceeding 4 mg/L in predominantly shallow, recently recharged ground waters of the United States. The model contains variables representing (1) nitrogen (N) fertilizer loading, (2) percent cropland-pasture, (3) natural log of human population density, (4) percent well-drained soils, (5) depth to the seasonally high water table, and (6) presence or absence of unconsolidated sand and gravel aquifers. Observed and average predicted probabilities associated with deciles of risk are well correlated (r2 = 0.875), indicating that the LR model fits the data well. The likelihood of nitrate contamination is greater in areas with high N loading and well-drained surficial soils over unconsolidated sand and gravels. The LR model correctly predicted the status of nitrate contamination in 75 percent of wells in a validation data set from the National Water-Quality Assessment (NAWQA) program. Considering all wells used in both calibration and validation, observed median nitrate concentration increased from 0.24 to 8.30 mg/L as the mapped probability of nitrate exceeding 4 mg/L increased from less than or equal to 0.17 to greater than 0.83. Determining where shallow ground water is at risk of nitrate contamination can help managers allocate scarce resources for prevention, monitoring, and clean-up measures. Related_Spatial_and_Tabular_Data_Sets: The data set is provided as a GeoTIFF image (gwrisk.tif) as well as native ArcInfo Workstation GRID format (gwrisk.tgz). The following table describes the display of mapped probabilities in the GeoTIFF image: >Probability that nitrate Index RGB Values >concentration exceeds 4 mg/L Color Value R G B >----------------------------------------------------------------- > 0 to .17 Light yellow 1 255 255 235 > Greater than .17 to .33 Orange 2 255 214 153 > Greater than .33 to .50 Green 3 209 255 51 > Greater than .50 to .67 Blue 4 102 214 255 > Greater than .67 to .83 Purple 5 204 102 255 > Greater than .83 to 1 Red 6 255 0 0 > No data Gray 99 230 230 230 The gwrisk.clr (colormap) file specifies the colors to be used for displaying GRID cell values. This is an ascii file containing the index numbers followed by RGB values: >Index red green blue >---------------------------- > 0 0 0 0 > 1 255 255 235 > 2 255 214 153 > 3 209 255 51 > 4 102 214 255 > 5 204 102 255 > 6 255 0 0 > ... > 99 230 230 230 > ... > 255 0 0 0 The .aux (auxiliary) file stores coordinate system information for the .tif file and is recognized by ArcGIS. To extract the ArcInfo Workstation GRID from the .tgz archive file use TARARC or the following commands: >gunzip gwrisk.tgz >tar xvof gwrisk.tar The GRID "gwrisk" will be in a subdirectory called arctar00000. >arctar00000/ >arctar00000/gwrisk/ >arctar00000/gwrisk/dblbnd.adf >arctar00000/gwrisk/hdr.adf >arctar00000/gwrisk/log >arctar00000/gwrisk/metadata.xml >arctar00000/gwrisk/prj.adf >arctar00000/gwrisk/sta.adf >arctar00000/gwrisk/vat.adf >arctar00000/gwrisk/w001001.adf >arctar00000/gwrisk/w001001x.adf >arctar00000/info/ >arctar00000/info/arc.dir >arctar00000/info/arc0000.dat >arctar00000/info/arc0000.nit >arctar00000/info/arc0001.dat >arctar00000/info/arc0001.nit >arctar00000/info/arc0002.dat >arctar00000/info/arc0002.nit >arctar00000/info/arc0002r.001 >arctar00000/log 2002 The calibration data set was from 1230 wells sampled as part of the National Water-Quality Assessment (NAWQA) program from 1992-1995 and the validation data set was from 736 NAWQA wells sampled from 1996-1999. The N fertilizer input data represents 1992 and the population density is from 1990. Complete The map could be updated as necessary as better data sets become available and to reflect any improvements in the procedures to define the logistic regression model. -129.839567 -64.113097 51.784295 21.217696 USGS Thesaurus Ground water Ground water contamination Ground water pollution Ground water susceptibility Nutrients Nitrate National Water-Quality Assessment Program NAWQA inlandWaters ISO 19115 Topic Category geoscientificInformation inlandWaters environment Geographic Names Information System Conterminous United States None None None 2002 None. The national probability map is intended for regional use only. Areas of high probability on the map have high potential for nitrate contamination but are not necssarily contaminated. The map should not be used for local applications. Please acknowledge the U.S. Geological Survey in products derived from these data. Kerie J. Hitt U.S. Geological Survey mailing address

12201 Sunrise Valley Drive, MS-413

Reston VA 20192 USA 1-888-275-8747 703-648-6693 kjhitt@usgs.gov 9:00am-5:30pm EST Please contact through email https://water.usgs.gov/nawqa/nutrients/pubs/est_v36_no10/ See figure 3 of the published report (PDF format) for a map of the probability that nitrate exceeds 4 mg/L in shallow ground waters of the United States, based on the LR model. PDF https://water.usgs.gov/GIS/browse/gwrisk.gif Map of the probability that nitrate exceeds 4 mg/L in shallow ground waters of the United States, based on the LR model. GIF Microsoft Windows 2000 Version 5.0 (Build 2195 Service Pack 3); ArcInfo Workstation version 8.3 Nolan, B.T., and Ruddy, B.C. 1996 1 USGS Fact Sheet 092-96 http://water.usgs.gov/nawqa/FS-092-96.html Nolan, B.T., Ruddy, B.C., Hitt, K.J., and Helsel, D.R. 1997 1 Environmental Science and Technology Volume 31, Number 8, Pages 2229-2236 This data set is the one that previously was distributed as "GWRISK" and now is being replaced. Nolan, B.T., Ruddy, B.C., Hitt, K.J., and Helsel, D.R. 1998 1 Water conditioning and purification Volume 39, Number 12, Pages 76-79 http://water.usgs.gov/nawqa/wcp/ Nolan, B.T. 2001 1 Ground Water Volume 39, Number 2, Pages 290-299 http://water.usgs.gov/nawqa/nutrients/pubs/gw_v39_no2/ Nolan, B.T., Hitt, K.J., and Ruddy, B.C. 2002 1 Raster digital data Environmental Science & Technology Volume 36, Number 10, Pages 2138-2145 http://water.usgs.gov/nawqa/nutrients/pubs/est_v36_no10/ The LR model results were checked using standard USGS review procedures. Not applicable for raster data. The data spans the conterminous United States. An output cell is assigned a value of "no data" if any of the corresponding input data cells at that location is "no data." U.S. Geological Survey 19990601 Version 1.0 raster digital data Sioux Falls, SD U.S. Geological Survey http://edc.usgs.gov/products/landcover/nlcd.html On-line 1986 1993 early to mid 1990s NLCD The source data consists of land cover data at 30-m resolution. The 30-m data was aggregated at 1-km resolution to indicate the percent of each land cover type in the cell. Curtis Price Rick Clawges 1999 1.0 digital raster data Open-File Report 99-78 Rapid City, SD U.S. Geological Survey https://water.usgs.gov/lookup/getspatial?ofr99-78_popdeng On-line 1990 1990 POPDENG 1-km grid of 1990 population of the conterminous United States, converted from an ASCII file obtained from the Consortium for International Earth Science Information Network (CIESIN). U.S. Department of Agriculture, Natural Resources Conservation Service 1994 map Miscellaneous Pub. 1492 Fort Worth, TX U.S. Department of Agriculture http://soils.usda.gov/survey/geography/statsgo/ CDROM 1994 Publication date of STATSGO data STATSGO Weighted averages for percent well-drained soils and depth to seasonally high water table. James A. Miller (compiler) 19980701 Version 1.0 map Madison, Wisconsin, USA U.S. Geological Survey http://www.nationalatlas.gov/mld/aquifrp.html On-line 1998 Publication date of principal aquifers data USAQV1 Surface outcrop or near-surface locations of unconsolidated sand and gravel aquifers (excluding glaciated sediments and alluvial aquifers along major rivers). David R. Soller and P.H. Packard 1998 1 map USGS Digital Data Series 38 Reston, Virginia U.S. Geological Survey https://pubs.usgs.gov/dds/dds38/ On-line 1998 Data compiled prior to 1987 QSURF Location of coarse-grained surficial Quarternary sediments. Preliminary version of the ground water risk map published in USGS fact sheet FS-092-96. 1996 Revised version of the ground water risk map published in Environmental Science and Technology, volume 31, number 8, pages 2229-2236. 1997 Characteristics of nitrogen loading and aquifer susceptibility to contamination were evaluated to determine their influence on contamination of shallow ground water by nitrate. A set of 13 explanatory variables was derived from these characteristics, and variables that have a significant influence were identified using logistic regression (LR) (Nolan, 2001). The model was developed using data collected from more than 1200 wells by the first 20 NAWQA study units that began in 1991 and contained these variables: 1) N fertilizer loading to the land surface; 2) percent cropland-pasture; 3) natural log of human population density; 4) percent well-drained soils; 5) depth to seasonally high water table; and 6) presence or absence of a fracture zone within a surficial aquifer. The model was used to identify variables that significantly influence nitrate contamination of shallow ground water, but was not validated with an independent data set and was not used in prediction. 200103 The previous LR model was refined (Nolan and others, 2002). The new model was recalibrated with three updated variables that represent improved sources of data and was validated using data collected from over 700 wells by a different set of NAWQA study units that began in 1994. Three variables were updated: N loading, percent cropland-pasture, and presence or absence of rock fractures. The current model uses separate estimates of farm and nonfarm fertilizer loading (USGS, unpublished data, 2001). The county-level fertilizer estimates were allocated to NLCD land cover classes instead of being allocated equally to farm and nonfarm uses. Farm fertilizer N was allocated to NLCD classes orchards/vineyards, row crops and small grains. Nonfarm fertilizer N was allocated to NLCD classes low-intensity residential and urban/recreational grasses. Cropland-pasture is represented by the aggregation of NLCD classes pasture/hay, row crops, small grains, and fallow land. In the previous model, cropland-pasture was derived from Anderson land use/land cover data updated with 1990 population data. A binary indicator variable representing the presence or absence of unconsolidated sand and gravel aquifers was substituted for the fracture zone variable. The other variables remain the same: natural log of human population density, percent well-drained soils, and depth to seasonally high water table. Each of the variables in the LR model (farm-nonfarm N fertilizer loading, percent cropland-pasture, natural log of population density, percent well-drained soils, depth to the seasonally high water table, and presence or absence of unconsolidated sand and gravel aquifers) was compiled within 1-km grid cells for prediction of nitrate contamination at the national scale. All of the grid layers line up with a template representing the conterminous U.S. consisting of 3070 rows x 4855 columns, about 14.9 million total cells. About 7.6 million cells contain data; the rest are "no data." Each cell measures 1-km x 1-km. The original 30-m NLCD were aggregated to give the percent of each category of land use within 1-km cells. The area (square km) of each type of land use in each cell is the percent divided by 100. The aggregated 1-km NLCD data were used to compute farm-nonfarm N fertilizer loading and percent cropland-pasture. Farm-nonfarm N was determined by combining NLCD areas of farm and nonfarm land use with fertilizer application data by county. The farm N application rate was calculated by dividing the total farm N applied (kg) in each county by the area of farm land within the county. Farm N was applied to appropriate farm land classification areas within each 1-km cell by multiplying the application rate by the farm land area within the cell. The procedure was duplicated for nonfarm land use and nonfarm fertilizer data by county. Total farm-nonfarm N is the sum of farm N plus nonfarm N. [Grid fertn92g.] The NLCD percent cropland-pasture is the sum of the percent of all categories of cropland-pasture land use in each 1-km cell. [Grid tcroppas.] The population data was derived from 1990 Census block group data representing people per square km. The log is calculated as ln(pop90g + .00001) for each 1-km cell. [Grid logpop90.] The STATSGO soil mapping units were gridded at 1-km resolution. Each mapping unit has multiple components. For well-drained soils, each cell contains the total percent of hydrologic groups A and B, computed for all soil components within the mapping unit. [Grid welldrg.] For depth to seasonally high water table, each 1-km cell contains the weighted average value of the depth to water table, based on the relative percents of all soil components within the mapping unit. [Grid wtdepavg.] The binary unconsolidated sand and gravel aquifer indicator variable is a combination of unconsolidated sand and gravel aquifers (Miller, 1998) and coarse-grained stratified sediments (Soller and Packard, 1998). The principal aquifer boundaries were converted to 1-km grid format with all aquifers having a rock name of "Unconsolidated sand and gravel aquifers" coded 1. All other aquifers were coded 0. [Grid bsgaqg2.] Soller's ALLQSURF map was converted to a 1-km grid having "coarse-grained sediments" (code 101) coded as 1 and everything else coded as 0. [Grid qcgsed2.] The two input grids are combined so that any cell having either sand and gravel = 1 or coarse-grained = 1 is coded as 1. [Grid newbaqg.] Explanatory variables in logistic the regression model are: > Estimated Wald >Variable coefficient p value >----------------------------------------- ----------- ------- >constant -5.541 <0.001 >1992 fertilizer N (kg/ha) 0.004 <0.001 >NLCD cropland-pasture (%) 0.016 <0.001 >ln(1990 population density) 0.229 <0.001 > (ln(people/km2) >well-drained soils (%) 0.025 <0.001 >depth to seasonally high water table (m) 1.088 <0.001 >presence or absence of unconsolidated 0.424 0.002 > sand and gravel aquifers To make the national probability map, the values from the 1-km grid cells are put in to the LR equation to calculate for each cell the probability that nitrate in shallow ground water exceeds 4 mg/L. An output cell is assigned a value of "no data" if any of the corresponding input data cells at that location is "no data." The following ArcInfo Workstation GRID commands produced the probability grid: >/*Probability grid with revised binary aquifer variable (using Soller's >/* coarse grained stratified sediments) >/* >/*snap to window specified by xmin,ymin, xmax,ymax (geolg grid) >setwindow -2505700.250 122020.555 2349299.750 3192020.555 d:/ancill/wolock/geol/geolg >setcell 1000 >&if [exists pgridrev -grid] &then kill pgridrev >/*This plugs the values for the input variables into the LR equation: >pgridrev = (exp (-5.5408 + (.00373 * fertn92g) + (.0162 * tcroppas) ~ > + (.2286 * logpop90) + ~ >(.0249 * welldrg) + (1.0879 * wtdepavg) + (.4245 * newbaqg) )) / ~ >(1 + (exp (-5.5408 + (.00373 * fertn92g) + (.0162 * tcroppas) + (.2286 * logpop90) + ~ >(.0249 * welldrg) + (1.0879 * wtdepavg) + (.4245 * newbaqg) ))) GRID cells were randomly selected from this dataset and checked by hand to ensure the correct probability values were calculated using the logistic regression equation and the input data values. In most NAWQA study units, the predicted exceedence probability reasonably approximates the observed proportion of wells with nitrate exceeding 4 mg/L. This indicates that the model adequately simulates regional N loading and aquifer susceptibility in these areas. The model inaccurately predicts the probability in some areas, such as the Rio Grande Valley. 200111 The LR values were grouped by category. The resulting grid has 6 categorical values corresponding to the range of probablities (1-6) and 1 value for "no data" (99). 200111 The GRID was converted to a GeoTIFF image for distribution. >gridimage gwrisk gwrisk.clr gwrisk tiff 20030904 Received the following comments from Curtis Price and responded to them. 1. The TIFF actually has your index values stored with an embedded colormap. So I'd add the index values to your table of concentrations in the supp info section, as you did in the attribute overview. [done] 2. I think you can and should skip the world file, as the georeferencing should be stored in the TIFF header, accessible to AV3, ArcGIS and any other software that recognizes GeoTIFF. But you may want to *add* the AUX file generated by ArcCatalog for the TIFF as that would make the coordinate system ("projection" in workstation parlance) of the TIFF available to ArcGIS. [done] 3. Suggest using > instead of ! to format tables [either would work] 4. ARC/INFO = ArcInfo or ArcInfo Workstation. I use the terminology [ArcInfo Workstation] GRID as the software; "grid" is a data set. You seem to be using this convention some of the time in your narratives, but not all the time. The downside of this is that reviewers may want you to define GRID because it looks like an acronym. It isn't. [done] 5. Suggest specifying processing window in your AML as x1 y1 x2 y2 rather than referring to the grid "geolg" and its snapping which was projected from Soller's map?? [done] 6. Suggest instead of "using this aml" just say "using the following ArcInfo Workstation GRID commands" [done] 20030908 Raster Grid Cell 3070 4855 1 Albers Conical Equal Area 29.5 45.5 -96 23 0.00000 0.00000 coordinate pair 1000.0 1000.0 METERS North American Datum of 1983 GRS80 6378137.000000 294.257222 GWRISK.VAT Value attribute table for the grid ArcInfo Workstation GRID software VALUE Categorical value of probability that nitrate nitrate exceeds 4 mg/L or no data USGS 1 0 to .17 probability that nitrate exceeds 4 mg/L based on LR model USGS 2 more than .17 to .33 probability that nitrate exceeds 4 mg/L based on LR model USGS 3 more than .33 to .50 probability that nitrate exceeds 4 mg/L based on LR model USGS 4 more than .50 to .67 probability that nitrate exceeds 4 mg/L based on LR model USGS 5 more than .67 to .83 probability that nitrate exceeds 4 mg/L based on LR model USGS 6 more than .83 to 1 probability that nitrate exceeds 4 mg/L based on LR model USGS 99 no data USGS COUNT Count of cells in each VALUE category Calculated by ArcInfo Workstation GRID software 18820 7276548 GWRISK.STA Statistics table that accompanies the grid ArcInfo Workstation GRID software MIN Minimum GRID cell VALUE Computed 1 Computed This is the minimum value of the categorical values (1,2,3,4,5,6,99) in the grid. MAX Maximum GRID cell VALUE Computed 99 Computed This is the maximum value of the categorical values (1,2,3,4,5,6,99) in the grid. MEAN Mean GRID cell VALUE Computed 49.152 Computed automatically by GRID software The MEAN is meaningless for the categorical values. STDV Standard Deviation of GRID cell VALUEs Computed 48.691 Computed automatically by ArcInfo Workstation GRID software The STDV is meaningless for the categorical values. Each 1-km grid cell stores the value of the category of predicted probability that nitrate concentration in shallow ground water exceeds 4 mg/L, as calculated by the multi-variate LR model. No data is indicated by a value of 99. The values of the categories are: > Grid Probability that nitrate exceeds 4 mg/L > cell in shallow ground waters of the > value United States, based on the LR model >-------------------------------------------------- > 1 0 to .17 > 2 More than .17 to .33 > 3 More than .33 to .50 > 4 More than .50 to .67 > 5 More than .67 to .83 > 6 More than .83 to 1 > 99 No data To obtain an ArcInfo grid containing the actual probabilities calculated in the LR model, contact the NAWQA Nutrients Synthesis team by emailing kjhitt@usgs.gov. None. U.S. Geological Survey Michael Ierardi IT Specialist mailing and physical

445 National Center

Reston Virginia 20192 USA 1-888-275-8747 (1-888-ASK-USGS) mierardi@usgs.gov Although this data set has been used by the U.S. Geological Survey, U.S. Department of the Interior, no warranty expressed or implied is made by the U.S. Geological Survey as to the accuracy of the data and related materials. The act of distribution shall not constitute any such warranty, and no responsibility is assumed by the U.S. Geological Survey in the use of this data, software, or related materials. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government. Other Arc Info grid zipped 2.709 https://water.usgs.gov/GIS/dsdl/gwrisk.zip Other Arc Info grid zipped 1 https://water.usgs.gov/GIS/dsdl/gwrisk.tgz Other Color map file for grid 1 https://water.usgs.gov/GIS/dsdl/gwrisk.clr Other Geo TIFF image zipped 1 https://water.usgs.gov/GIS/dsdl/gwrisk.tar.gz None. This dataset is provided by USGS as a public service. 20041108 U.S. Geological Survey Ask USGS -- Water Webserver Team mailing

445 National Center

Reston VA 20192 1-888-275-8747 (1-888-ASK-USGS) mierardi@usgs.gov FGDC Content Standards for Digital Geospatial Metadata FGDC-STD-001-1998