7c44dad0-98aa-45ee-8011-4e3d2332c5c7 eng utf8 dataset dataset EnviroAtlas Coordinator U.S. Environmental Protection Agency, Office of Research and Development - Center for Public Health and Environmental Assessment (CPHEA), EnviroAtlas Geospatial Data Owner (919) 541-3832 109 T.W. Alexander Drive Research Triangle Park NC 27709 EnviroAtlas@epa.gov https://www.epa.gov/enviroatlas pointOfContact 2025-07-23 ISO 19139 Geographic Information - Metadata - Implementation Specification 2007 102039 Esri 8.1.2 EnviroAtlas - Predicted drinking water nitrate violations by 12-Digit HUC for the Conterminous United States 2020-12-29 publication 2022-01-31 revision U.S. Environmental Protection Agency (US EPA) originator U.S. EPA Office of Research and Development (ORD) - Center for Public Health and Environmental Assessment (CPHEA) Research Triangle Park, NC publisher tableDigital This map displays predicted probability of a 12-digit Hydrologic Unit (HUC12) being at risk of violating the drinking water nitrate standard. Risk is defined as there being greater than a 50% chance of having a drinking water violation. Observed violations of the nitrate standard in public drinking water was used to make predictions across the conterminous United States. The data on nitrate violations comes from the EPA's Safe Drinking Water Information System (SDWIS) and the data used as predictor variables in the prediction models comes from EPA's StreamCat database. The original predictions were made at the National Hydrography Dataset (NHD) Catchment scale and summarized at the HUC12 scale and/or 300 m pixel scale. The maps show that 19% of the conterminous United States is at risk of groundwater sourced drinking violations, while 16% of the U.S. is at risk of nitrate violations from surface water sources. Regions with the highest risk of nitrate drinking violations from groundwater sources are found in the southwest, south central plains, parts of north west, mid-Atlantic. Surface water sourced nitrate violations have the greatest risk primarily in the southwest and south-central areas of the U.S. This dataset was produced by the US EPA to support research and online mapping activities related to EnviroAtlas. EnviroAtlas (https://www.epa.gov/enviroatlas) allows the user to interact with a web-based, easy-to-use, mapping application to view and analyze multiple ecosystem services for the contiguous United States. The dataset is available as downloadable data (https://edg.epa.gov/data/Public/ORD/EnviroAtlas) or as an EnviroAtlas map service. Additional descriptive information about each attribute in this dataset can be found in its associated EnviroAtlas Fact Sheet (https://www.epa.gov/enviroatlas/enviroatlas-fact-sheets). This map can inform decisions of EPA's Office of Water and the Regions and States, NGOs and others throughout the U.S. involved in environmental management and protection activities. Specifically, the map predictions can help drinking water managers better target which systems or which states and regions need improved drinking water treatment or source water protection measures, to reduce the likelihood of drinking water violations, and thereby possibly reducing health impacts. The overall goal of EnviroAtlas is to employ and develop the best available science to map indicators of ecosystem services production, demand, and drivers for the nation. completed EnviroAtlas Coordinator U.S. Environmental Protection Agency, Office of Research and Development - Center for Public Health and Environmental Assessment (CPHEA), EnviroAtlas Geospatial Data Owner (919) 541-3832 109 T.W. Alexander Drive Research Triangle Park NC 27709 EnviroAtlas@epa.gov https://www.epa.gov/enviroatlas publisher asNeeded Alabama Arizona Arkansas California Colorado Connecticut Delaware Florida Georgia Idaho Illinois Indiana Iowa Kansas Kentucky Louisiana Maine Maryland Massachusetts Michigan Minnesota Mississippi Missouri Montana Nebraska Nevada New Hampshire New Jersey New Mexico New York North Carolina North Dakota Ohio Oklahoma Oregon Pennsylvania Rhode Island South Carolina South Dakota Tennessee Texas United States Utah Vermont Virginia Washington West Virginia Wisconsin Wyoming place EPA Place Names 2015-01-31 publication Contaminant Drinking Water Ecosystem Environment Ground Water Modeling Surface Water Water theme EPA GIS Keyword Thesaurus 2007-11-02 publication 020:072 theme Federal Program Inventory 2013-09-16 publication 12-digit HUCs Ecosystem Services EnviroAtlas Environment Environmental Atlas theme User 2022-01-11 publication None. Please check sources, scale, accuracy, currency and other available information. Please confirm that you are using the most recent copy of both data and metadata. Acknowledgement of the EPA would be appreciated. EPA Public Domain License licenseUnrestricted https://edg.epa.gov/EPA_Data_License.html unclassified https://project-open-data.github.io/schema/#accessLevel textTable eng utf8 Microsoft Windows 10 Version 10.0 (Build 19042) ; Esri ArcGIS 12.9.0.32739 true -128.028392 -65.203364 22.734889 51.67736 2011-01-01T00:00:00 2011-12-31T00:00:00 2013-01-01T00:00:00 2017-12-31T00:00:00 This dataset can be joined to the EnviroAtlas feature class NHDPlusV2_WBDSnapshot_EnviroAtlas_CONUS using the field 'HUC_12.' EnviroAtlas uses the best data available, but there are still limitations associated with these data. These data are based on models and large national geospatial databases. Calculations based on these data are estimations of the truth founded on the best available science. The percentage of catchments correctly classified by the random forest model was 79% for ground water (GW) public water systems and 90% for surface water (SW) public water systems, for the conterminous U.S. (CONUS), but this varied geographically. Typically, performance was better in areas with lower violation risk, such as the south east, north east, and parts of lower Mississippi watershed, but had greater false positive predictions in regions with higher violation risk. The predictions of nitrate violations and concentrations is not based on ambient surface or groundwater data, but "finished" water in drinking water systems that is ready to be distributed to the public. This means that there is not necessarily a direct relationship between environmental variables and nitrate concentration or violation data, due to the possibility that the source water has been treated. Additionally, the models used to create the map were only able to incorporate environmental or geologic factors in its predictions, and was not able to include "engineered" factors within the drinking water treatment plant (e.g., ability of a public water system (PWS) operator to switch to or mix with other water sources, use improved treatment technologies, or use other management strategies). This may explain why some regions had false positives (i.e. low observed nitrate violations but high predicted violation risk), due to the presence of nitrate treatment technologies helping to reduce the violation rate in the presence of environmental factors that create high risk. Lastly, it should be noted that the model predictions of drinking water nitrate concentrations were only made for NHD catchments that were predicted to have &gt;0.5 probability of having a nitrate violation, in the random forest classification model. File Geodatabase Table https://enviroatlas.epa.gov/download/CONUS_metrics_FGDB.zip download download https://enviroatlas.epa.gov/download/CONUS_metrics_CSV.zip download download https://www.epa.gov/enviroatlas/enviroatlas-data information information https://www.epa.gov/enviroatlas information information https://enviroatlas-epa.hub.arcgis.com/search?tags=EA-DrinkingWaterNitrate information information dataset Tests for integrity have not been performed. EPA National Geospatial Data Policy Features have been reviewed by U.S. EPA EnviroAtlas Project Managers for public release. true Features represented have not been tested for completeness EPA National Geospatial Data Policy Features have been reviewed by U.S. EPA EnviroAtlas Project Managers for public release. true Obtained drinking water facility point location coordinates from EPA's Office of Water (Renee Morris at Office of Water). 2017-01-01T00:00:00 Obtained the model response variable: drinking water nitrate violations dataset from SDWIS. 2018-01-01T00:00:00 SDWIS Federal Reports Advanced Search Downloaded predictor variables from StreamCat Website (which includes land use, geology, soil, weather, etc.) 2018-01-05T00:00:00 StreamCat Dataset Created the predictor variables for aquifer type by downloading the "us_aquifers" shapefile from the U.S. Geological Survey (USGS) and summarizing percent aquifer type (e.g. % sandstone, % semiconsolidated sand aquifers) per NHD catchment. 2018-01-08T00:00:00 Principal Aquifers of the 48 Conterminous United States, Hawaii, Puerto Rico, and the U.S. Virgin Islands Calculated the predictor variable water input (kg2/cm): ratio of agricultural land that is irrigated divided by mean annual precipitation. Irrigation Data was obtained from: https://www.agcensus.usda.gov/Publications/2012/Full_Report/Volume_1,_Chapter_2_County_Level/. Percent irrigation on agricultural lands per catchment was calculated using the U.S. Department of Agriculture (USDA) agricultural census data allocated to National Land Cover Dataset (NLCD) agricultural lands (https://www.mrlc.gov/data). 2018-01-12T00:00:00 2012 Census of Agriculture National Land Cover Database Calculated the predictor variable mean percent hillslope per catchment based on NHDPlusV2 elevation rasters. 2018-01-12T00:00:00 NHDPlus (National Hydrography Dataset Plus) Calculated the predictor variable Hortonian Overland Flow per catchment, using zonal statistics, based on a raster from: https://water.usgs.gov/GIS/metadata/usgswrd/XML/gwava-dw_hor.xml 2018-01-16T00:00:00 Vulnerability of shallow ground water and drinking-water wells to nitrate in the United States: Model of predicted nitrate concentration in U.S. ground water used for drinking (simulation depth 50 meters) -- Input data set for Hortonian overland flow (gwava-dw_hor) Calculated the predictor variable precipitation surplus (precipitation minus evapotranspiration) for each NHD catchment and watershed using PRISM (Parameter-elevation Regressions on Independent Slopes Model) mean monthly precipitation and temperature data (1994-2016) 2018-01-17T00:00:00 PRISM climate data - 30-year normal mean precipitation (mm) PRISM climate data - 30-year normal mean temperature (°C) Calculated the predictor variable density of septic systems per NHDPlusV2 catchment, based on 1991 Census dataset with number of septic systems per county. 2018-01-18T00:00:00 US Census Calculated the predictor variable total fresh surface-water withdrawals in agricultural land per catchment area (L/day). Downloaded Excel File of USGS Water Use data for 2010. Calculate the number of 30m-by-30m pixels that are agriculture per county. Merge withdrawal data with number of pixels per county and calculate withdrawal as mgd per pixel - by dividing the withdrawal per county by the number of pixels in each county. Use Raster Calculator (spatial analyst) to multiply the raster by agricultural NLCD raster to allocate withdrawal data to just agricultural land to produce final raster. 2018-01-29T00:00:00 Estimated Use of Water in the United States County-Level Data for 2010 Calculated the predictor variable NHD slope (longitudinally) along each stream within a catchment using data within NHDplusV2. Slope is calculated as the max elevation minus the min elevation divided by the stream length (with the following NHDPlusV2 variables: SlopeLenKm, MinElevSmo, MaxElevSmo. 2018-02-18T00:00:00 NHDPlusV2 Calculated the predictor variable surface water nitrogen flux (kg N per HUC12 watershed) and summarized these values at the NHD catchment scale. 2018-06-12T00:00:00 Calculated the predictor variable Nitrogen Surplus (kg N / yr) per catchment (excluding biological N Fixation), based on methods of Sabo et al. 2019 Journal of Geophysical Research (JGR) Biogeosciences. N surplus is the sum of agricultural fertilizer, urban fertilizer, total nitrogen deposition, combined biological nitrogen fixation (CBNF), N inputs from manure, and human waste N, minus crop N removal. All variables were obtained or calculated from StreamCat variables, except urban fertilizer and crop N removal. Based on Sabo et al. 2019 JGR Biogeosciences, human waste = 4.7 * population density. Crop N removal per catchment is calculated using the N crop removal per county dataset from NuGIS (Nutrient Use Geographic Information System), and allocating the crop removal values to agricultural lands only, using NLCD 2011 (values 81 and 82) data. Urban fertilizer per catchment was based off county level data (https://www.sciencebase.gov/catalog/item/5851b2d1e4b0f99207c4f238) that was allocated to developed lands based on NLCD 2011 data values (21, 22, 23) and then converted to a raster summarize the values per catchment and watershed. 2018-07-10T00:00:00 Nutrient Use Geographic Information System National Land Cover Database (NLCD) 2011 County-Level Estimates of Nitrogen and Phosphorus from Commercial Fertilizer for the Conterminous United States, 1987-2012 Calculated the predictor variable Net Anthropogenic Nitrogen (NANI) within catchments, based on the sum of farm fertilizer, urban fertilizer, NOx deposition, combined biological nitrogen fixation (CBNF), Human and Livestock N demand, minus Livestock N content, and minus crop N removal. All variables were from existing StreamCat variables except urban fertilizer, human and livestock food demand, and crop N removal. Based on Sabo et al. 2019 JGR Biogeosciences, Human N demand is calculated as population density by a constant 6.21, livestock N demand is calculated as manure N inputs times a constant 1.37, and livestock N content is calculated as one fourth of livestock N demand. Crop N removal per catchment is calculated using the N crop removal per county dataset from NuGIS, and allocating the crop removal values to agricultural lands only, using NLCD 2011 (values 81 and 82) data. 2018-07-11T00:00:00 Nutrient Use Geographic Information System National Land Cover Database (NLCD) 2011 Calculated predictor variables based on 2010 Census block group data (e.g. median household income, percent with high school education, etc.) by summarizing the block group data at the NHDplusV2 catchment scale. 2018-07-11T00:00:00 Calculated the predictor variable wastewater treatment plants density per catchment (number/ km2). The EPA Facility Registry Service (FRS) data was filtered down to just active WWTPs and then saved as either all, major, and minor WWTPs. 2018-07-18T00:00:00 EPA Facility Registry Service (FRS): Wastewater Treatment Plants Obtained rock N (kg per square km) (N from rock weathering) from Ben Houlton (Houlton et al. 2018 in Science) as raster dataset and summarize rock N per NHD catchment. 2018-07-20T00:00:00 Nitrogen from rock weathering Convergent evidence for widespread rock nitrogen sources in Earth’s surface environment 2018-04-06 publication Science 6384 58-62 Downloaded treatment information from the Safe Drinking Water Information System and summarized the number of public water systems with nitrate treatment technologies (e.g. reverse osmosis, ion exchange) per catchment. Note that SDWIS does not provide a complete list of sites with nitrate treatment technology, thus this predictor variable was not a complete dataset and was not used in the full model. 2018-07-21T00:00:00 Safe Drinking Water Information System - Federal Reports Advanced Search All predictor variables were prepossessed to create raster layers in order to perform a type of zonal statistics (through the StreamCat algorithm) to average the variables per NHDPlusV2 catchment and upstream watershed. 2018-08-01T00:00:00 Used drinking water facility information to calculate nitrate violations and mean concentrations per NHDPlusV2 catchment, which were used as response variables in the models. 2018-08-01T00:00:00 Compiled and prepared predictor and response variable datasets for modeling. 2018-08-11T00:00:00 Completed random forest classification modeling to make predictions of nitrate drinking water violation risk for all 2.6 million NHDPlusV2 catchments. 2018-08-13T00:00:00 Completed random forest regression modeling to make predictions of violation concentration for the NHDPlusV2catchments predicted to have >50% probability of a nitrate violation, based on the classification modeling. 2018-08-13T00:00:00 Summarized model predictions at HUC12 scale by using a NHDPlusV2 catchment to HUC12 conversion table created using a spatial join 2018-08-14T00:00:00 Predicted probabilities (Pred_Viol_Prob_GW and Pred_Viol_Prob_SW) were calculated as the average of all catchment values within a HUC12. For predicted violation risks (Pred_Viol_Risk_GW and Pred_Viol_Risk_SW), the most frequent category within a HUC12 was used. 2020-03-05T00:00:00 Creating Catchment to HUC12 Converter: 1) Using R, a spatial join was done, with a methodology that allowed for the calculation of the area of intersection. a. Used the HUC12s used by EnviroAtlas (“NHDPlusV2_WBDSnapshot_EnviroAtlas_CONUS”) b. NHD Catchments were processed by hydroregion 2) The results for each hydroregion were then combined together (using rbind()). R Code for Hydroregion 1: Cat01_HUC12_i <- st_intersection(st_make_valid(huc12[,c('HUC_12','HUC12','geometry')]), st_make_valid(Cat01[,c('COMID','geometry')])) %>% mutate(intersect_area = st_area(.)) %>% # create new column with shape area dplyr::select(HUC_12, COMID, intersect_area) %>% # select columns needed to merge st_drop_geometry() Summarizing the catchment data by the EnviroAtlas version of the NHDPlus V2 HUCs. 2022-01-01T00:00:00 NHD Catchments by hydroregion EnviroAtlas - NHDPlus V2 WBD Snapshot, EnviroAtlas version - Conterminous United States Summarizing Predicted Violation Probability by HUC_12: 1) Merge on Model predictions with Cat_HUC converter created from spatial join 2) Calculate the percent of total area for intersecting catchments per HUC12 a. Merge on the total HUC12 area b. Calculate % of total catchment intersection within the whole HUC 3) Since the shapefiles were not perfectly aligned, all the small intersections need to be removed: a. If sum of intersection area less than 50% of total area of HUC12, then don't include b. If individual catchment's intersection area is less than 1% of HUC area, don't include 4) Summarize nitrate violation probability by HUC12 using the intersection area as weight (Get weighted mean value). R code: # Merge on Model predictions with Cat_HUC converter created from spatial join tempa = merge(RFC_GW_pred_prob,Cat_HUC12_i,by='COMID') # Get the area % of intersecting cats per HUC12 area_sum = tempa %>% dplyr::group_by(HUC_12) %>% summarize(sum_intrsec_area_m2 = sum(intersect_area_m2)) area_prc = merge(tempa,area_sum,by='HUC_12') # Merge onto original area_prc = subset(area_prc, seletc = -c(intersect_area)) # remove unnecessary column # Merge on HUC12 area area_prc2 = merge(area_prc,huc12_df[c('HUC_12','huc_area_m2')],by='HUC_12') # calc % of total catchment intersection within the whole HUC area_prc2$total_area_perc = 100*area_prc2$sum_intrsec_area_m2/area_prc2$huc_area_m2 # If sum of intersection area < 50% of total area of HUC12, then don't include # Subset out HUC12s that have >50% coverage by catchments with data area_prc3 = area_prc2[area_prc2$total_area_perc >= 50, ] # If individual catchment's intersection area is < 1% of HUC area, don't include area_prc3$cat_area_perc = 100*area_prc3$intersect_area_m2/area_prc3$huc_area_m2 area_prc4 = area_prc3[area_prc3$cat_area_perc >= 1, ] # summarize by HUC12 (using intersection area as weight) Pred_Viol_Prob_GW_huc_w = area_prc4 %>% dplyr::group_by(HUC_12) %>% summarize(Viol_Prob = weighted.mean(x=Pred_Viol_Prob_GW, w=intersect_area_m2, na.rm=T)) 2022-01-02T00:00:00 Summarizing Predicted Violation Risk by HUC_12: 1) Merge catchment model violation risk predictions with catchment to HUC12 converter created with the spatial join 2) Calculate the percent of total area for intersecting catchments per HUC12 a. Merge on the total HUC12 area b. Calculate % of total catchment intersection within the whole HUC 3) Since the shapefiles were not perfectly aligned, all the small intersections need to be removed: a. If sum of intersection area less than 50% of total area of HUC12, then don't include b. If individual catchment's intersection area is less than 1% of HUC area, don't include 4) Calculate the amount of area taken up by each risk category per HUC12 5) Assign the risk category with the most area to the HUC12 R Code: # Merge catchment model violation risk predictions with catchment to HUC12 converter created with the spatial join tempa = merge(RFC_GW_pred_risk,Cat_HUC12_i,by='COMID') tempa$risk_cat = as.factor(tempa$Pred_Viol_Risk_GW) # add on risk category tempa = subset(tempa, select= -c(Pred_Viol_Risk_GW,intersect_area,HUC_12)) # remove unneeded columns # Get the area % of intersecting cats per HUC12 area_sum = tempa %>% dplyr::group_by(HUC12) %>% summarize(sum_intrsec_area_m2 = sum(intersect_area_m2)) %>% data.frame() area_prc = merge(tempa,area_sum,by='HUC12') # Merge onto original # Merge on HUC12 area area_prc2 = merge(area_prc, huc12_df[c('HUC_12','huc_area_m2')], by.x='HUC12', by.y='HUC_12') # calc % of total catchment intersection within the whole HUC area_prc2$total_area_perc = 100*area_prc2$sum_intrsec_area_m2/area_prc2$huc_area_m2 # If sum of intersection area < 50% of total area of HUC12, then don't include # Subset out HUC12s that have >50% coverage by catchments with data area_prc3 = area_prc2[area_prc2$total_area_perc >= 50, ] # If individual catchment's intersection area is < 1% of HUC area, don't include area_prc3$cat_area_perc = 100*area_prc3$intersect_area_m2/area_prc3$huc_area_m2 area_prc4 = area_prc3[area_prc3$cat_area_perc >= 1, ] # Calculate the amount of area taken up by each risk category per HUC12 tempa2 = area_prc4 %>% group_by(HUC12,risk_cat) %>% summarize(risk_sum = sum(intersect_area_m2)) %>% data.frame() # Assign the risk category with the most area to the HUC12 Pred_Viol_Risk_GW_huc_w = tempa2 %>% group_by(HUC12) %>% summarize(maxarea_m2 = max(risk_sum), risk_cat = risk_cat[which(risk_sum == maxarea_m2)]) %>% data.frame() 2022-01-03T00:00:00 asNeeded 2025-07-30