Skip

Potential Priority Watersheds for Protection of Water Quality:

From Nonpoint Sources Related to Agriculture*

Poster Presentation at the 52nd Annual SWCS Conference

Toronto, Ontario, Canada, July 22-25, 1997
(Revised October 7, 1997)

Robert L. Kellogg, Susan Wallace, and Klaus Alt (retired)
Natural Resources Conservation Service

Don W. Goss, Texas Agricultural Experiment Station, Temple, Texas

Objective

National maps were developed to assist decision-makers in identifying priority watersheds for water quality protection from nonpoint sources related to agriculture. The purpose of these maps is to systematically identify where the greatest potential exists for water pollution based on factors known to be important influences on soil and chemical loss from farm fields, such as climate, soil characteristics, and pesticide and nitrogen loadings from agricultural sources. The basis for the analysis is 2,105 8-digit hydrologic units, or watersheds, in the 48 States (910,000 acres average size).

Summary

The potential for loss of pesticides, nutrients, and soil from farm fields in each watershed was assessed using national-level databases on land use, soils, climate, agricultural chemical use, and confined livestock populations. Maps of four pollution sources are presented for leaching and runoff. The top 400 watersheds (about 20%) were selected from each of the contributing maps and overlaid to identify which watersheds have the greatest potential for combinations of water pollution sources. Watersheds identified as having a high potential for more than one pollution source would have a high priority for implementation of conservation programs to reduce externalities associated with agricultural production.

NRI Modeling

Most of the maps shown here were produced using the National Resources Inventory (NRI) as a modeling framework. The NRI is a national survey of private land use conducted by the Natural Resources Conservation Service that is based on about 800,000 sample points throughout the US, including cropland, pastureland, rangeland, forest land, urban land, and other uses of private land. At each NRI sample point, information is collected on nearly 200 attributes, including land use and cover, cropping history, conservation practices, potential cropland, highly eroding land, water and wind erosion estimates, wetlands, wildlife habitat, vegetative cover conditions, and irrigation. The NRI is linked to a national soils database (SOILS5). Data from other sources, such as precipitation and agricultural chemical use, were imputed onto the NRI sample points using geographic and land use links.

A simulation of potential loss of agricultural chemicals from farm fields was conducted by treating each NRI sample point as a "representative field." Pesticide loss and nitrogen vulnerability indexes were estimated at individual NRI sample points using the site-specific information available from the NRI and other sources. Environmental indicators at the watershed level were constructed by aggregating measurements over the NRI sample points in each watershed. Expansion factors for each sample point were previously determined during the development of the NRI survey design. These expansion factors are the number of acres each sample point represents, and are used as weights to aggregate over the sample points.

NRI modeling was used here to create environmental indicators for cropland. The potential exists to create similar environmental indicators in this manner for carbon sequestration, chemical loss from forestland, and the contribution of soil loss and chemical loss to air pollution from nonpoint sources.

The NRI is based on Primary Sampling Units (PSUs) that are selected using a stratified random design. The standard PSU is 160 acres; smaller and larger PSUs were used in some areas of the country. Within each PSU, typically three points were selected for sampling. These points were used as the framework for simulation modeling.

About the Maps

Each of our maps is also available as a zipped postscript file and PDF file. These files can be used to produce high quality copies of our maps. For more infomation on postscript see our postscript help page. For PDF files Adobe Acrobat is required.

A variety of algorithms were used to create watershed values, generating a variety of units (pounds per watershed, index scores per watershed, etc.). To facilitate comparisons among the maps, the classes shown in each map were based on a consistent set of watershed rankings. The 200 watersheds with the highest scores are shown with the darkest color. The next highest 200 watersheds are shown with a slightly lighter color, and so on.

Caveats

  • Analyses do not show which watersheds will have water quality impairments related to agricultural production. The simulation models estimate soil loss, chemical loss, or vulnerability indexes for conditions at the edge of the field and the bottom of the root zone. Not included in the indicators are dynamics of fate and transport from the farm field to a water body. Dilution from runoff and recharge on noncropland areas in the watershed will also reduce the potential for actual water quality impairments.
  • Maps provide a relative ranking among watersheds. It is not known how the class breaks relate to observed water quality problems.
  • Farm management practices are not included in the determination of the indicators. Research has shown that soil, chemical, and water loss from farm fields can be substantially reduced by management practices. Where these practices are being used, the potential estimated by these indicators will be over-estimated.
  • The land base varies among the maps--7 crops for commercial nitrogen fertilizer, 13 crops for pesticides, cultivated cropland for climatic factors, all cropland for soil erosion, and all land for manure nitrogen. Where indicators are based only on the major crops, watersheds with large acreage of other crops may be mis-represented.

Climatic Factors

Percolation and runoff factors were created for each NRI sample point for use in converting nitrogen inputs to vulnerability indexes.

Percolation Factor [GIF | Postscript | PDF ]. The percolation factor represents average annual percolation of water through the root zone in inches per year. It is based on precipitation and the hydrologic group of the soil. The calculation weighs precipitation during non-growing periods (fall and winter months) more than during growing periods to account for plant uptake. Hydrologic group is an attribute of the NRI. Precipitation data were obtained from a network of 1,473 climate stations throughout the US. A database of average monthly precipitation for each climate station was assembled using 25 years of daily precipitation data. Monthly precipitation data were imputed to NRI sample points on the basis of the proximity of the NRI sample points to the climate stations.

Runoff factors. Runoff factors represent runoff from a field in inches per year. Daily precipitation data for 25 years were used with the NRCS curve number method to calculate a monthly runoff factor for each of 1,473 climate stations. At each climate station, monthly runoff was estimated for twelve curve numbers ranging from 70 through 92. Monthly runoff values were imputed to NRI sample points according to the proximity of the sample point to one of the weather stations and according to curve number. Curve numbers were assigned to NRI sample points using information on the soil hydrologic group, tillage, conservation practice, and crop type. The annual runoff factor [GIF | Postscript | PDF ] was obtained by summing over the 12 monthly runoff factors. The two-month runoff factor [GIF | Postscript | PDF ] was obtained by summing over the two months following planting (determined from average planting dates for 55 regions), when the potential for runoff losses of soil and chemicals is often the greatest.

Soil Erosion

Tons of soil loss from sheet and rill erosion [GIF | Postscript | PDF ] per watershed were obtained directly from the NRI. The NRI estimated sheet and rill erosion (tons per acre) at each cropland sample point using the Universal Soil Loss Equation (USLE). These per-acre estimates were multiplied by the number of acres represented by the NRI sample point and aggregated over all the cropland sample points in each watershed. Erosion estimates are for 1992.

Manure Nitrogen Fertilizer

Manure Nitrogen Loadings from Confined Livestock [GIF | Postscript | PDF ]. Pounds of manure nitrogen loadings per watershed were estimated using data on livestock populations from the 1992 Agriculture Census. The average number of livestock present on each farm and the average length of time they lived on the farm were estimated for 16 types of livestock, adjusting the estimates to reflect the number of livestock held in confinement. Assuming all manure from confined operations was applied to the land, nitrogen loadings were estimated by multiplying the livestock population times the average amount of manure produced by each type of livestock, and then multiplying times an estimate of the average nitrogen content of the manure for each type of livestock. An additional adjustment was made for typical losses of nitrogen during storage and from volatilization during application. County totals for manure nitrogen were obtained by aggregating over the farms in the county. Watershed totals were obtained by multiplying county estimates by county-watershed conversion factors derived from a GIS calculation of the percentage of each county in each watershed. (Weights were adjusted using NRI information on the presence or absence of pastureland and cropland in each county-watershed polygon.)

Manure Nitrogen Leaching Vulnerability Index [GIF | Postscript | PDF ]. A watershed index score for the leaching potential for manure nitrogen was calculated by multiplying the total manure nitrogen estimate for the watershed times the average percolation factor for the watershed. The average percolation factor was based on values for all non-Federal rural land.

Manure Nitrogen Runoff Vulnerability Index [GIF | Postscript | PDF ]. A watershed index score for the runoff potential for manure nitrogen was calculated by multiplying the total manure nitrogen estimate for the watershed times the average 12-month runoff factor for the watershed. The average runoff factor was based on values for all non-Federal rural land.

Commercial Nitrogen Fertilizer

Nitrogen Loadings from Commercial Fertilizer Applications, Adjusted for Crop Uptake [GIF | Postscript | PDF ]. Pounds per watershed of nitrogen from commercial fertilizer applications was calculated based on production of 7 crops--corn, soybeans, wheat, cotton, barley, rice, and sorghum--comprising 162,343 NRI sample points. The nitrogen loading for each sample point was estimated as the difference between the amount of commercial nitrogen fertilizer applied and the amount taken up by the crop and removed at harvest. State data on nitrogen application rates and percent acres treated for 1992 were obtained from farmer survey data published by the Economic Research Service (Taylor). The amount of nutrients taken up by the crop was estimated by multiplying the percent of nutrients in the harvested portion times the average county per-acre yield, using county yield data published by the National Agricultural Statistics Service for 1988-1992. A nitrogen credit was estimated for legumes grown in the previous two years. The total pounds of "excess" nitrogen for each sample point was estimated by multiplying the per acre estimate by the percentage of acres treated in the state. Estimates of pounds per watershed were obtained by aggregating over the NRI sample points in each watershed.

Commercial Nitrogen Fertilizer Leaching Vulnerability Index [GIF | Postscript | PDF ]. The estimate of commercial nitrogen fertilizer loadings was multiplied by the percolation factor at each NRI sample point for each of the seven crops to obtain a nitrogen leaching vulnerability index score. For irrigated sample points, five inches was added to the percolation factor prior to the calculation. Watershed scores were obtained by summing the vulnerability index scores for the NRI sample points in the watershed for the seven crops.

Commercial Nitrogen Fertilizer Runoff Vulnerability Index [GIF | Postscript | PDF ]. The estimate of commercial nitrogen fertilizer loadings was multiplied by the runoff factor for two months following planting at each NRI sample point for each of the seven crops to obtain a nitrogen runoff vulnerability index score. Watershed scores were obtained by summing the vulnerability index scores for the NRI sample points in the watershed for the seven crops.

Pesticides

Pounds of Pesticides Applied to Crops by Watershed [GIF | Postscript | PDF ]. Pounds of pesticides per watershed were estimated using the NRI and the National Pesticide Use Database created by Gianessi and Anderson. Gianessi and Anderson organized data from publicly available reports and surveys from Federal and State agencies into a national database of pesticide use in US crop production for the period 1990-93. State average application rates and the percentage of acres treated for 141 pesticides were imputed to NRI sample points by crop and state for 13 crops: barley, corn, cotton, oats, peanuts, potatoes, rice, sorghum, soybeans, sugar beets, sunflowers, tobacco, and wheat. (These are the only crops specifically identified in the NRI.) A total of 170,219 sample points were included. For each chemical used on a specific crop in a specific state, the application rate (pounds per acre) was multiplied by the number of acres represented by the NRI sample point and by the percentage of acres treated in the state. Estimates of pounds per watershed were obtained by first aggregating over all chemicals at each sample point, and then aggregating over all sample points for the 13 crops in the watershed.

Potential Pesticide Leaching [GIF | Postscript | PDF ] and Runoff Loss [GIF | Postscript | PDF ] from Farm Fields. A National Pesticide Loss Database was created using the chemical fate and transport model GLEAMS. GLEAMS is a process model that estimates pesticide leaching and runoff loss using as inputs: soil properties, field characteristics (such as slope and slope length), management practices, pesticide properties, and climate. GLEAMS estimates were generated for 243 pesticides applied to 120 specific soils for 20 years of daily weather for each of 55 climate stations distributed throughout the United States. Separate GLEAMS estimates were made for irrigated and nonirrigated conditions. The maximum percent loss over the 20-year period was used to construct the indicator. Pesticide loss estimates were for movement beyond the edge of the field and beyond the bottom of the root zone, measured in percent loss per acre per year. Loss estimates were imputed to NRI sample points by soil group and proximity to climate stations. A total of 141 pesticides were included. Estimates of percent acres treated and average application rates from Gianessi and Anderson were imputed to the NRI sample points by crop and state. Each NRI sample point where corn was grown in Iowa, for example, included chemical use for 22 pesticides Gianessi and Anderson reported were used on corn in Iowa. The total loss of pesticides in pounds per watershed was estimated by 1) multiplying the estimate of percent loss per acre times the application rate to obtain the mass loss per acre for each pesticide at each NRI sample point, 2) calculating the number of acres treated for each pesticide by multiplying the estimate of percent acres treated by the number of acres associated with the sample point, 3) multiplying the number of acres treated by the mass loss per acre to obtain the mass loss for each pesticide at each NRI sample point, and 4) summing over the mass loss estimates for all the pesticides and sample points in the watershed.

Watersheds with High Potential for Chemical and Soil Loss from Farm Fields

The three maps of leaching indicators [GIF | Postscript | PDF ] and the four maps of runoff indicators [GIF | Postscript | PDF ] were overlaid to identify which watersheds have the greatest potential for combinations of water pollution sources. Watersheds identified as having a high potential for more than one pollution source (pesticides, nitrogen, or soil) would have a high priority for implementation of conservation programs. The top 400 watersheds (about 20%) were selected from each of the seven contributing maps as watersheds with a high potential for chemical or soil loss from farm fields.

Data Availability

Watershed estimates used in this analysis are available for downloading. The variable names are matched to the maps in a documentation table. If questions arise using the data, please contact Robert Kellogg.

Documentation Table

Watershed Database (Pipe delimited ASCII file)


* Some of the maps and numbers have been revised since the July presentation.