Skip

Environmental Indicators of Pesticide Leaching and Runoff from Farm Fields

Robert L. Kellogg, Natural Resources Conservation Service, Washington D.C.

Richard Nehring, Economic Research Service, Washington D.C.

Arthur Grube, Office of Pesticide Programs, EPA, Washington D.C.

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

Steve Plotkin, Natural Resources Conservation Service, Amherst, Massachusetts

 

February 2000

Presented at a Conference on "Agricultural Productivity: Data, Methods, and Measures,"

March 9-10, 2000, Washington DC.

ABSTRACT

The purpose this paper is to present an indicator of how changes in pesticide use in agriculture have changed the potential for risk to human health and the environment from pesticide contamination of water leaving farm fields. Environmental indicators are designed to be relative estimates of potential risk that are based on pesticide use and the factors that are known to be important determinants of pesticide loss from farm fields, such as the intrinsic potential of soils to leach or runoff pesticides, the chemical properties of the pesticides, annual rainfall and its relationship to leaching and runoff, and changes in cropping patterns. The analytical framework consists of about 4,700 resource polygons representing the intersection of 48 states, 280 watersheds at the 6-digit Hydrologic Unit level, and 1,400 combinations of climate and soil groups. Twelve crops are included in the analysis: corn, soybeans, wheat, cotton, sorghum, barley, rice, potatoes, oats, sugarbeets, tobacco, and peanuts. Model estimates of pounds of pesticides applied, mass loss, and annual concentrations leaving the farm field (edge of field and bottom of root zone) were obtained for pesticides used on each of the 12 crops in each of the resource polygons for each year from 1960 through 1997. Indicators of potential risk are constructed from estimates of annual concentrations that exceed "safe" thresholds for chronic exposure to four target groups&emdash;humans, fish, crustaceans, and algae. It is expected that temporal and spatial trends of these indicators will closely track the change in potential risk to human health and the environment from agricultural use of pesticides.

Environmental Indicators of Pesticide Leaching and Runoff from Farm Fields

INTRODUCTION

Pesticides are a vital input in today's agriculture, protecting food and fiber from damage by insects, weeds, diseases, nematodes, and rodents. U.S. agriculture spends about 8 billion dollars annually on pesticides-about 70 percent of all pesticides sold in the country.1 It is estimated that each dollar invested in pesticide control returns approximately 4 dollars in crops saved.2 Nevertheless, pests still destroy nearly 13 percent of all potential food and fiber crops in the U.S. Farmers' expenditures on pesticides are about 4-5 percent of total farm production costs.

The dependence of agriculture on chemical pesticides developed over the last 60 years as the agricultural sector shifted from labor-intensive production methods to more capital and chemical intensive production methods. Sixty years ago most crops were produced largely without the use of chemicals. Insect pests and weeds were controlled by crop rotations, destruction of crop refuse, timing of planting dates to avoid high pest population periods, mechanical weed control, and other farming practices. While these practices are still in use, changes in technology, changes in prices, and government policies resulted in development of today's chemically intensive agriculture.

Usage of conventional pesticides on farms increased from about 400 million pounds (active ingredient) in the 1960s to over 800 million pounds in the late 1970s and early 1980s, primarily due to the widespread adoption of herbicides in corn production.3 Since that time, usage has been somewhat lower, ranging from about 700 to 780 million pounds per year. Pesticide usage in agriculture can vary considerably from year to year depending on weather, pest outbreaks, crop acreage, and economic factors such as pesticide prices and crop prices. Whereas the quantity of pesticides used by agriculture has fallen off slightly in recent years, total expenditures on pesticides by farmers are still increasing.

During the 1960's, agricultural pesticide use was dominated by insecticides, accounting for about half of all pesticides used. The quantity of insecticides applied fell as the organochlorines (DDT, aldrin, and toxaphene) were replaced by pyrethroids and other chemicals that required lower application rates. Today, 70 percent of the quantity of pesticides used in agriculture are herbicides. Corn leads all other crops-by a substantial margin-in total pesticide use. Rice, potatoes, vegetables, and fruits, however, actually use pesticides more intensively than corn and other crops. Minimum tillage practices are being adopted by many farmers, further reducing the need for machinery, labor, and energy inputs, but increasing agriculture's dependency on pesticides even more. Pesticide use trends can vary markedly from one part of the country to another as farmers respond to local pest problems and as crop production patterns vary.

Even as today's chemically intensive agriculture is partly responsible for providing abundant low-cost supplies of food and fiber, it has also created water quality problems. When the chemical revolution first started there was little concern about environmental consequences. Scientific testing indicated that DDT and other agricultural chemicals were generally not harmful to humans if used as directed. By the mid-1960's, however, there was a growing awareness that some agricultural chemicals were damaging the environment, and may have been affecting humans as well. Awareness that agricultural chemicals were not staying on the fields, but were being washed into streams and rivers and seeping into ground water, came about with the development of sensitive chemical testing procedures. These procedures did not become available for organochlorine pesticides (DDT, DDE, aldrin, dieldrin, heptachlor, and chlordane) until the late 1960s. The DDT problem was known before that time (Rachel Carson's book Silent Spring was released in 1962) largely because of bioaccumulation, resulting in detectable levels in animals high in the food chain.

Today, pesticide levels in water are monitored routinely. Pesticide residues have been found in ground water, surface water, and rainfall. EPA began to emphasize ground water monitoring for pesticides in 1979 following discovery of DBCP and aldicarb in ground water in several states. In 1985, 38 States reported that agricultural activity was a known or suspected source of ground water contamination within their borders.4 Since then, several Federal and State agencies have developed programs to sample water resources and test for the presence of agricultural chemicals. Results published to date have shown that chemicals used in agricultural production have been found in ground water, sometimes at levels exceeding EPA's drinking water criteria.5 6 7 8 9 Monitoring for pesticides in surface water was frequent in the 1960s and 1970s as studies were conducted that led to the banning of chlorinated hydrocarbon insecticides. Sampling in the 1980s and 1990s found that the four leading herbicides in use during that time--atrazine, metolachlor, alachlor, and cyanazine--were frequently detected in surface waters in agricultural regions.10 11 12 13 14 Highest levels occurred after planting and during the early part of the growing season. Most of the pesticides commonly used presently and in the past have also been found in the atmosphere, including DDT, toxaphene, dieldrin, heptachlor, organophosphorous insecticides, triazine herbicides, alachlor and metolachlor.15 These airbourne pesticides return to the earth with rainfall to further contribute to water contamination. A recent report by USGS of a survey of pesticides in the Nation's waters concluded that pesticides were common in surface and shallow ground water in both urban and agricultural areas, but investigators were not able to determine if contamination is lessening or worsening.16

Chemical testing can detect the presence of a pesticide, and often can measure how much of the pesticide is in the water, but it cannot identify the source of the pesticide. It is not known what portion of observed residues originate from quasi-point sources within agriculture such as applicator loading and mixing sites or from nonagricultural sources. Since agriculture is the largest user of pesticides, it is likely that much of the pesticide residue found in the environment originated from agriculture. However, a significant--but unknown--portion of the pesticide residue originates from non-agricultural sources. Non-agricultural uses include: home, lawn, and garden use, industrial use, pest control in forestry, weed control along roadsides, ditches, railways, and rights-of-way, pest control by municipalities and local governments, golf courses, and the military.

Concerns about potential risks to human health and the environment resulted in the 1972 amendments to the Federal Insecticide, Fungicide, and Rodenticide Act (FIFRA), which increased the stringency of health and safety data required to support a pesticide registration. EPA first banned use of some organochlorine pesticides for agricultural purposes in the 1970's, and has since imposed use limitations on many other pesticides. The amendments also required that all existing pesticides be reregistered using current health and environmental standards. Chemical companies have responded to these regulatory pressures by marketing new chemicals that are thought to be less harmful to humans and the environment, or less likely to migrate from farm fields to contaminate ground water and surface water.

The purpose of the research reported here is to develop an indicator of how changes in pesticide use in agriculture has changed the potential for risk to human health and the environment from pesticide contamination of water. The ideal situation would be to have consistent data for all regions in the country on contamination of surface water and ground water that is exclusively due to use of pesticides in agricultural production. Such a database does not exist, nor could it ever be realistically produced. An alternative--and more tractable--approach is to model the potential for risk associated with pesticide loss from farm fields. The environmental indicators created here have been designed to be relative estimates of potential risk that are based on pesticide use and the factors that are known to be important determinants of pesticide loss from farm fields, such as the intrinsic potential of soils to leach or runoff pesticides, the chemical properties of the pesticides, annual rainfall and its relationship to leaching and runoff, and changes in cropping patterns. Indicators are constructed from model estimates of annual concentrations leaving the farm field (edge of field and bottom of root zone) that exceed "safe" thresholds for chronic exposure to four target groups-humans, fish, crustaceans, and algae.17

MODELING PESTICIDE LOSS FROM FARM FIELDS

Even when the label instructions are carefully adhered to, a small portion of pesticides applied reaches surface and ground water, as evidenced by the detection of pesticides in water quality monitoring studies. The greatest losses occur in the first few rainfall events after application. Losses in surface runoff are the greatest--up to 5 percent of the amount applied. Losses to subsurface drainage and percolation, however, are much less--typically less than 0.5 percent.18 Wind or water is required to transport pesticides from the field to surface or ground water. Whether and how much a pesticide migrates from the field where it was applied depends on a complex interaction of pesticide and soil properties with weather conditions and site characteristics.

Physical properties of pesticides (soil sorption propensity, vapor pressure, and solubility) and the persistence of the pesticide in the environment are important factors in the tendency of pesticides to move from the application site. Persistence is a pesticide's resistance to decompose through chemical, photochemical (sunlight), and microbial action. The half-life of a chemical is not a constant value, but can vary considerably depending on soil temperature and moisture, microbial populations present, amount and kind of organic material, pH, and soil type. Pesticides that persist longer in the environment are more likely to move off-site than less persistent pesticides because they are exposed to more leaching and runoff events. Sorption is the binding of the chemical to the soil. Pesticides that are strongly adsorbed tend not to leach, but rather are lost with the soil through soil erosion processes. Pesticides that are weakly adsorbed are lost mainly in surface runoff water and percolation. Vapor pressure is the measure of a pesticide's tendency to evaporate. Losses to the atmosphere of sprayed pesticides during application can be significant. Water solubility, which is related to soil sorption propensity, also influences the amount of pesticide that is likely to be removed by runoff or by leaching.

Soil characteristics and their interaction with physical properties of pesticides also affect the potential for pesticides to migrate from the field. Three major soil characteristics that affect chemical movement are soil texture, soil permeability, and soil organic matter. Soil texture is an indication of the relative proportions of sand, silt, and clay in the soil. Pesticides tend to be adsorbed mostly on clay and organic matter. The higher the clay content, the greater the number of binding sites for pesticide retention. Coarse, sandy soils generally allow water to move rapidly downward and offer few opportunities for adsorption. Soil permeability is a general measure of how fast water can move downward in a particular soil. Permeability is controlled by soil structure, which is the way individual soil particles clump together, creating pore space between the soil aggregates. This pore space is key to water retention and movement through the soil. Organic matter content is the most important variable affecting sorption of pesticides onto soil particles. Organic matter provides binding sites, is very reactive chemically, and promotes microbial populations that contribute to pesticide degradation. Soil organic matter also influences how much water the soil can hold before movement occurs.

Pesticide loss from farm fields was estimated using a process model called GLEAMS, which incorporates these complex interactions of weather, pesticide properties, and soil characteristics.19 GLEAMS was initially developed to simulate edge-of-field and bottom-of-root zone loading of sediment and chemicals to evaluate alternative management practices. It operates on a daily time step with daily climatic data. Gleams consists of three major components: hydrology, erosion/sediment yield, and pesticides. Soil-water accounting procedures represent the principal hydrologic processes of infiltration, runoff, water application by irrigation, soil evaporation, plant transpiration, and soil water movement within and through the root zone. A modification of the Universal Soil Loss Equation (USLE) is used to simulate storm-by-storm rill and interrill soil erosion in overland flow areas. The pesticide component of GLEAMS is designed to allow simulation of interactions among pesticide properties, climate, soils, and management. Adsorption characteristics are coupled with the hydrologic component to route pesticides within and through the root zone. Upward movement of pesticides and plant uptake are simulated with soil evaporation and plant transpiration, respectively. Pesticide degradation into metabolites is tracked for compounds that have potentially toxic daughter products; in the output, metabolites are included in the concentration calculation for the parent compound.

A National Pesticide Loss Database was constructed using GLEAMS estimates for 243 pesticides applied to 120 generic soils for 20 years of daily weather from each of 55 climate stations. This resulted in 1,603,800 runs of 20 years each. Separate GLEAMS estimates were made for irrigated and nonirrigated conditions. The crop in these simulations was a generic row crop behaving similar to corn, soybeans, cotton, or sorghum, planted in straight rows. The model estimates irrigation timing and amounts depending on soil moisture.

Soil Parameters

The 120 generic soils were selected on the basis of soil texture and organic matter content. Twelve textures were combined with four organic matter contents for surface horizons. Four textures were used for subsurface horizons, which were not coarser in texture than the surface horizon. Organic matter content of a subsurface horizon was 20 percent of the surface horizon. The thirty horizon combinations times four organic matter contents made up 120 soils, summarized in table 1. The hydrologic group (HG) was assigned according to subsurface texture. Using tables found in the GLEAMS manual and relations in the USDA Soil Survey Handbook, the other soil parameters required by GLEAMS were estimated. The Curve Number (CN) is also required, and was assigned as shown in table 1.

Climate Parameters

GLEAMS includes a climate generator that simulates daily rainfall and temperature. Fifty-five climatic stations across the United States were chosen. The program generates daily weather using the mean, standard deviation, and skew for monthly precipitation, maximum temperature, minimum temperature, and the mean and standard deviation of monthly solar radiation. Two other monthly values required are: 1) the probability of having a wet day after a wet day, and 2) the probability of having a dry day after a wet day. A twenty year record was simulated with the climate generator from a 40-year frequency distribution to produce a distribution of pesticide loss estimates that would reasonably represent most weather conditions. Planting and harvest dates were estimated for each of the 55 climate stations based on mean and standard deviation of monthly low temperatures from the climatic record.

Pesticide Application Parameters

GLEAMS results were simulated for the 243 pesticides in the SCS/ARS/CES pesticide properties database.20 The pesticide database is included in the GLEAMS model pesticide parameter editor, including foliar characteristics constructed using the procedure by Willis and McDowell. 21 The Insect Control Guide22 and the Weed Control Guide23 were used to define the action of each compound, when applied, how frequently applied, and recommended rates and methods of application.

Pesticide application timing was based on the planting date, harvest date, and purpose. Pesticide application method was based on planting date and purpose. Some herbicides were designated only for pre-plant application and some only for post-plant application. Those herbicides with applications designated as "all methods" were included with the pre-emergent herbicides. Preplant pesticides were simulated with application seven days before planting. Pre-emergent pesticides were simulated with application on the planting date. Post-plant pesticides were simulated with application fourteen days after the planting date. Over the top insecticides, fungicides, and miticides were simulated in 3 repeat applications commencing after one third of the growing season was completed. For example, an insecticide with a recommended repeat application every five days was first applied one-third of the way through the growing season, and then repeated every five days for a total of three applications. Growth regulators were applied after one fourth of the growing season was completed. Defoliants were applied 5 days before harvest date.

Some soil insecticides and nematicides were incorporated in the soil; some surface applied, and some applied over-the-top of foliage. The SCS/ARS/CES pesticide properties database also includes growth regulators and defoliants, both of which are applied on foliage. The insecticides used as foliar applications were applied at label recommended frequency. The Insect Control Guide included recommendations on the frequency of application, i.e. 3-5 days, 5 days, or 7 days. In GLEAMS, insecticides were applied every 3 days for the 3-5 day recommendation, every 5 days for the 5 day recommendation, and every 7 days for the 7 day recommendation.

Model Output

The daily mass of pesticide that was removed by leaching, runoff in solution, and runoff with sediment were recorded for each model run. Annual totals for each were obtained by summing the daily estimates. The daily volume of water that leached below the root-zone or ran off as surface water was recorded and summed over the year, and the daily mass of sediment loss was recorded and summed over the year. Output from the GLEAMS simulations included annual estimates of:

  1. Mass loss in leachate as a percent of the amount applied.
  2. Mass loss dissolved in runoff as a percent of the amount applied.
  3. Mass loss adsorbed to eroded soil as a percent of the amount applied.
  4. Volume of water percolation in centimeters.
  5. Volume of water runoff in centimeters.
  6. Sediment loss in kilograms.

 

Annual concentrations were calculated as mass loss per volume of water. Mass loss was expressed as a percentage of the amount applied for the model runs so that loss estimates could be derived for any application rate. This is necessary because, in practice, application rates often vary from the recommended rates used to make the model runs. This provides the flexibility to apply different application rates across the country to reflect differences in use, or to simulate the effects of proposed policies on pesticide loss. The annual concentration was also expressed as a ratio of concentration to the mass of pesticide applied so that, when multiplied by the application rate, the concentration adjusted to reflect actual pesticide use rates would be obtained.

The 20-year distribution of mass loss and concentration estimates were used to derive prediction equations corresponding to any percentile of the distribution. The 95th percentile, which is the mass loss or concentration that would be expected to be exceeded only five percent of the time, was used to develop the environmental risk indicators.

DEVELOPMENT OF ENVIRONMENTAL RISK INDICATORS

Overview

A model was constructed to estimate environmental risk indicators for 1960 through 1997 using the following data and information:

  • The National Pesticide Loss Database.
  • Annual estimates of pesticide use by crop and state from farmer surveys, primarily the Doane farm panel surveys and the USDA pesticide use surveys.
  • Annual county estimates of acres planted.
  • Soil distribution from the National Resources Inventory.
  • Irrigated acreage from the National Resources Inventory.
  • Water quality thresholds corresponding to drinking water standards (or equivalent derived from mammalian chronic toxicity data) and the maximum safe levels for chronic exposure of fish, algae, and crustaceans to pesticides.

 

The analytical framework consists of about 4,700 resource polygons representing the intersection of 48 states (sufficient data do not exist to include Hawaii or Alaska), 280 watersheds at the 6-digit Hydrologic Unit level, and 1,400 combinations of climate and soil groups identified in the National Pesticide Loss Database. Twelve crops are currently included in the analysis: corn, soybeans, wheat, cotton, sorghum, barley, rice, potatoes, oats, sugarbeets, tobacco, and peanuts. A total of 94 pesticides were included in estimates for 1960-86, and up to 192 pesticides were included in estimates for 1987-97. Estimates of pounds of pesticides applied, mass loss, and annual concentrations were obtained for pesticides used on each of the 12 crops in each of the resource polygons for each of the 38 years.

Environmental risk was assessed using Threshold Exceedence Units (TEUs). TEUs account for both the extent that the annual concentration exceeds a water quality threshold, and the number of acres on which pesticides are applied that produces the threshold exceedence. A concentration-threshold ratio was calculated for each pesticide used in each resource polygon, representing a per-acre estimate of potential risk. Thresholds differ considerably among pesticides, but with this approach, all pesticides have the same per-acre estimate of potential risk (equal to 1) at the point where the threshold and the concentration coincide. As application rates increase, the concentration increases, and the per-acre estimate of risk increases; the per-acre risk score for pesticides with relatively high toxicity increases faster than pesticides with lower toxicity. Where the threshold is exceeded, the per-acre estimate of potential risk was multiplied by the acres treated to obtain TEU estimates that, when summed over a state or region, represent a regional estimate of potential risk. Where the concentration is below the threshold, the potential risk score is zero because the procedures used to derive the thresholds emulate those used by EPA to define "safe" thresholds for chronic exposure.

TEUs were derived in this manner specifically for the purpose of comparing risk from multiple pesticides over space and time. They are a relative measure of potential risk (the higher the TEU score, the higher the potential risk), but they do not measure actual risk.24 It is expected that temporal and spatial trends of these indicators will closely track the change in potential risk to human health and the environment from agricultural use of pesticides.

Management factors (such as tillage practices or grass waterways and stream buffers) that may mitigate pesticide loss from farm fields are not included in the indicators.

The general algorithm for the potential risk index aggregated to the national level is:

TEUt = Sigma Sigma Sigma Exceedence per acre treatedt,r,k,p * Acres treatedt,r,k,p
  r k p  
Exceedence per acre treated = [(RELCONC*APPRATE)/THRESH)-1], negative values discarded.
Acres treated = ACRES*PCTTREAT*PCTSOIL

where:

TEUt = Threshold Exceedence Units for a given year.
RELCONC = Relative concentration associated with the 95th percentile pesticide loss per acre for a specific pesticide on a specific soil group and a specific climate group.
APPRATE = Application rate for a specific pesticide.
THRESH = Threshold concentration below which the pesticide loss concentration is defined to be "safe" for chronic exposure.
ACRES = Acres of crop planted in the state-watershed combination.
PCTTREAT = Percentage of acres treated with a specific pesticide.
PCTSOIL = Percentage of state-watershed combination in a specific soil group.
k = 12 irrigated and 12 nonirrigated crops.
p = Pesticides
r = Approximately 4,700 resource polygons
t = Years from 1960 through 1997.

Separate pesticide risk indicators were constructed for pesticides in leachate and pesticides dissolved in runoff. In addition, the mass loss of pesticides (in leachate, dissolved runoff, and adsorbed runoff) was estimated using a similar algorithm. Databases were constructed by aggregating to three levels: 1) the national level, 2) state level, and 3) watersheds at the 6-digit Hydrologic Unit level. Twelve pesticide-related time series were produced for each level of aggregation:

  • Pounds of pesticides applied
  • Leaching mass loss, pounds
  • Dissolved runoff mass loss, pounds
  • Adsorbed runoff mass loss, pounds
  • Pesticide leaching risk index for protection of drinking water, TEUs
  • Pesticide runoff risk index for protection of drinking water, TEUs
  • Pesticide leaching risk index for protection of fish, TEUs
  • Pesticide runoff risk index for protection of fish, TEUs
  • Pesticide leaching risk index for protection of algae, TEUs
  • Pesticide runoff risk index for protection of algae, TEUs
  • Pesticide leaching risk index for protection of crustaceans, TEUs
  • Pesticide runoff risk index for protection of crustaceans, TEUs

The leaching risk index for protection of drinking water is a proxy for risk associated with drinking water taken from shallow wells. The runoff risk index for protection of drinking water is a proxy for risk associated with drinking water taken from surface water supplies without land use controls in the upstream watershed. This indicator would not be appropriate for the New York water supply, for example, because much of the source area for that watershed is protected. Some of the other surface water supplies in the country have similar protections. The runoff risk indexes for protection of fish, algae, and crustaceans are proxies for risk associated with exposure in surface water during times of the year when runoff is the predominant source of flow. The leaching risk indexes for protection of fish, algae, and crustaceans are proxies for risk associated with exposure during times of the year when ground water is the predominant source of flow (such as mid-winter low flow conditions) or in waterbodies that are predominately fed by ground water sources (spring-fed ponds, lakes, and streams). Both estimates of risk for fish, algae, and crustaceans are necessary because most waterbodies have both a ground water and a surface water component. For example, USGS reports that in a sample of 54 streams from throughout the U.S., ground water contributions ranged from 14 percent to 90 percent, with a median of 55 percent.25

The Analytical Framework

The basis for the analytical framework underlying these estimates is the National Resources Inventory (NRI). The NRI is a national survey of private land use that is based on about 800,000 sample points, 300,000 of which are on cropland.26 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 that includes information on soil texture, hydrologic group, and organic matter content, which were the soil characteristics used to define the 120 soil groups for which pesticide loss estimates were simulated using GLEAMS. Using this soil information, each NRI cropland sample point was assigned a soil ID corresponding to one of the 120 soil groups. Similarly, each NRI cropland sample point was assigned one of the 55 climate stations based on geographic proximity. The pesticide use databases provided information on the suite of pesticides used for each crop and state. Linking these data sources together, a list of pesticides was imputed onto each NRI cropland sample point, as well as estimates of pesticide loss (mass loss and concentration) relative to application rate.

The framework was constructed by identifying resource polygons as the intersection of state, watershed (6-digit level), soil ID, and climate station. To reduce the number of resource polygons to a manageable number, very small polygons were excluded. This procedure was executed using NRI data for both 1982 and 1992. It resulted in a total of 4,891 resource polygons for 1982 and 4,459 for 1992. State-watershed combinations numbered 544 for 1982 and 517 for 1992. The analytical framework based on 1982 was used for calculating environmental indicators for 1960 through 1985, whereas the analytical framework based on 1992 was used for 1986 through 1997.

The percentage of a state-watershed combination that corresponded to each soil-climate combination within a state-watershed combination was estimated using NRI acreage of all 12 crops combined (PCTSOIL in the above algorithm). This percentage was used in the algorithm to distribute acres by crop for each state-watershed combination to the soil-climate groups.

Land Use Time Series

Data on acres planted by county were used to capture changes in cropping patterns over space and time. County data were obtained from electronic files published by the National Agriculture Statistics Service (NASS). These data generally cover 1972 to the present, but not all years are included for each of the 12 crops. Gaps in the record were filled one of two ways: 1) state totals were distributed to counties using county shares from a nearby year, or 2) state totals were distributed to counties using county shares derived from the Census of Agriculture. State totals were obtained from NASS publications. The latter approach was used for all states and crops for 1960-71, and for potatoes through 1982 and sugarbeets through 1976. For these estimates, the Census of Agriculture county data were used as follows: 1959 county shares used to estimate acres planted for 1960-61, 1964 county shares used for 1962-66, 1969 county shares used for 1967-71, 1974 county shares used for 1972-76, 1978 county shares used for 1977-80, and 1982 county shares used for 1981-84. Table 2 summarizes the number of states with county estimates of acres planted for each crop and the source of the data.

County data on acres planted were converted to state-watershed combinations using conversion weights derived from the NRI, which provides acreage estimates by watershed and county for each of the 12 crops. Separate weights were developed for 1982 and 1992. The 1982 weights were used for years prior to 1986, and the 1992 weights were used for 1986-97. Weights were crop specific; that is, county corn acres planted were converted to state-watershed acres planted using county-watershed proportions derived from NRI corn acreage estimates. In some counties, however, the NRI did not find any acres, even though acres (usually a small number) were reported in the Census of Agriculture or in the NASS county acreage database. This occurred more frequently in the early years of the series, where the 1982 NRI did not represent spatial distribution of crops as well as in the later years of the series. In these cases, a conversion weight based on the sum of all 12 crops was used.

Acres planted data included both irrigated and dryland acreage. The proportion of each crop in each state-watershed combination that was irrigated was estimated using the NRI. These proportions were multiplied by the crop acreage to obtain separate estimates of irrigated and nonirrigated acres for each of the 12 crops in each state-watershed combination.

Pesticide Use Time Series

An annual time series of application rates and percent acres treated for each crop and each state that had acres planted data was constructed for 1960 through 1997 using all sources of information available, summarized in Table 3. The primary sources were the Doane farm panel surveys and the USDA National Agricultural Statistics Service (NASS) pesticide use surveys. For 1990 through 1997, two separate sets of estimates of application rates and percent acres treated were constructed, one with Doane data and another with NASS data, so that results from the two sources could be compared. For 1960 through 1989, a single set of estimates was constructed from published reports of a series of annual surveys by Doane (1970-89), intermittent surveys conducted by Doane and NASS, and a small number of surveys from other sources.

The Doane Pesticide Profile Study is an annual survey of farmers conducted by Doane Marketing Research, Incorporated, on behalf of the pesticide industry.27 The Doane Pesticide Profile Study includes all crops and all states for 1987 to the present. Sample sizes ranged from 12,252 farms in 1987 to 18,208 farms in 1991. The database is electronic, and estimates of application rates and percent acres treated were obtained by querying the database.

USDA NASS conducted annual pesticide use surveys for selected crops in selected states from 1990 to the present. Sample sizes were comparable to those in the Doane Pesticide Profile Study when compared on a crop-by-crop basis. These data were used to construct a second set of estimates of application rates and percent acres treated for 1990 through 1997 for the selected states and crops.

The Doane Countrywide Farm Panel Surveys for insecticides and herbicides, precursors to the Doane Pesticide Profile Study, were available (although not in electronic form) annually for most years from 1970 through 1986 for corn, soybeans, cotton, wheat, and sorghum.28 Results were reported for the 7 Doane regions, shown below, and for most states in regions 3 and 5 for corn and soybeans. State estimates from the survey were used for corn and soybeans in regions 3 and 5. State estimates for cotton, wheat, sorghum, and corn and soybeans in regions other than 3 and 5 were obtained by assigning the regional estimates to each of the states within each region, implicitly assuming homogeneity within the regions. These estimates, particularly application rate estimates, were supplemented with additional data from NASS surveys and surveys by Batelle.

Doane Region States Crops
1 Maine, Connecticut, Massachusetts, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, Vermont Corn
2 Delaware, Florida, Georgia, Maryland, North Carolina, South Carolina, Virginia, West Virginia Corn, soybeans, cotton
3 Illinois, Indiana, Ohio, Michigan, Wisconsin Corn, soybeans
4 Alabama, Arkansas, Kentucky, Louisiana, Mississippi, Tennessee Corn, soybeans, cotton, sorghum
5 Iowa, Minnesota, Missouri, Nebraska, Kansas, North Dakota, South Dakota Corn, soybeans, wheat, sorghum
6 Arizona, California, Colorado, Idaho, Montana, Nevada, New Mexico, Oregon, Utah, Washington, Wyoming Wheat, cotton
7 Texas, Oklahoma Wheat, cotton, sorghum

Pesticide use estimates for these 5 crops for the early years not covered in the Doane Countrywide Panel Surveys were largely constructed using information in three USDA pesticide use surveys for 1964, 1966, and 1971 for each of the 10 U.S. production regions and national use estimates from Doane for most years. The national Doane data together with regional shares from 1970-73 were used to generate time series for acres treated by region. Application rates were taken from all sources.

These procedures and databases provided estimates for most states, years, and pesticides for these five crops, but a few additional estimates were needed to complete the series. Missing data on application rates were filled using application rates for nearby years or nearby states. Occasionally it was possible to derive percent acres treated data for a state (or groups of states) or pesticide using the acres treated residuals derived from regional or national data. Percent acres treated data for some missing years were constructed using regional shares from nearby years to distribute national totals of acres treated. In cases where acres planted data indicated that crops were grown in states not covered by the pesticide use surveys, pesticide use data from a nearby state or region was used.

Pesticide use data for the remaining 7 crops for 1960-86 was comparatively sparse, and considerable extrapolation was required to construct complete series. There was very little state specific data available, so the focus was on constructing regional time series. State estimates were made by assigning the regional estimates to each of the states within each region. Missing data for application rates were filled by extending data from survey years to years without data. In cases where only national data on acres treated were available, acres treated were distributed to regions using regional shares from survey years and percent acres treated estimates calculated using acres planted data. Pesticide use data for small grains and wheat were used for oats and sometimes for barley. Most other missing data were filled using procedures already presented for the first 5 crops. In all cases, the series were constructed using all the information available and using the most precise extrapolation procedures possible.

Not all pesticides reported in the surveys could be used to estimate the environmental risk indicators because information on pesticide loss or water quality thresholds were not available. The 194 pesticides included in the estimates are summarized by 6-7 year time intervals in table 4. Most of the pesticides not included were very minor. The few pesticides with significant or frequent use that were not included were: acetochlor, clethodim, dimethenamid, DSMA, flumetsulam, halosulfuron, heptachlor, imazethapyr, nicosulfuron, primisulfuron, prosulfuron, tefluthrin, and zetamethain.

Water Quality Thresholds

The EPA Office of Water sets drinking water standards for some pesticides. A Health Advisory (HA) is the maximum concentration of a chemical in drinking water that is not expected to cause any adverse noncarcinogen effect over a lifetime exposure with a margin of safety. A Maximum Contaminant Level (MCL) is the maximum permissible pesticide concentration allowed in a public water source. HAs and MCLs were used as water quality thresholds to calculate the environmental risk indicators for drinking water. Many pesticides, however, do not have HAs or MCLs. For pesticides categorized by EPA Office of Water as noncarcinogens or only "possible" carcinogens, "safe" thresholds were estimated from EPA published Reference Dose (RFD) values using procedures similar to those used to derive HAs. For pesticides categorized by EPA Office of Water as probable or known carcinogens, EPA published data on cancer slopes were used to estimate the CHCL (Chronic Human Carcinogen Level), which produces concentrations comparable to the MCL.

Thresholds for fish, algae, and crustaceans were estimated from published toxicity data. Since so little data on long-term exposures are available, Maximum Acceptable Toxicant Concentrations (MATCs) were calculated from 96-hour LC50's using the method of Barnthouse, Suter and Rosen.29 Most 96-hour LC50's used in this study were taken from EPA's Office of Pesticide Toxicity Database (EPA, 1997). MATCs are equivalent to the maximum concentration permitting survival for chronic exposures. When toxicity data for more than one species was available, the "safe" threshold for the most sensitive species was used.

The estimated water quality thresholds are shown in table 6 for the pesticides included in this study. Thresholds are constant over space and time, but the suite of pesticides included in the indicator differs depending on the target group. For example, more toxicity results were available for humans than for the other groups. The potential risk indicators are thus constrained to represent temporal and spatial changes only for the suite of pesticides for which toxicity data were available for each target group. Additional information on toxicity data and derivation methods can be found in Plotkin et. al.30

Combining Risk Indicators Based on Alternative Pesticide Use Sources for 1991-97

The separate pesticide use databases for NASS 1990-97 and Doane 1990-97 allowed separate risk indicators to be estimated for pesticides that were common to both surveys. In the majority of cases, risk indicators did not differ by more than about 20 percent when aggregated to the state level and compared by pesticide. When aggregated over pesticides, there was even less difference. For the risk indicators presented in this paper, the two estimates were combined. For each pesticide in each resource polygon, the average of the two risk scores was calculated wherever pesticide use estimates from both surveys existed (see table 3 for NASS coverage).

 

TRENDS IN THE POTENTIAL FOR ENVIRONMENTAL RISK

Pounds of pesticides applied to the 12 crops included in the study steadily increased throughout the 1960s and 1970s to peak in the early 1980s at about 560 million pounds, and then fell to about 450-500 million pounds throughout the 1990s (figure 1). Pesticide loss estimates generally followed a similar trend (figure 2). As shown in the table below, the highest average percent losses are for dissolved runoff and the average lowest are for leaching for most time periods; however, percent leaching losses were much higher and percent runoff losses lower in the 1960s than in later years. The sum of loss estimates for both leaching and runoff ranged from about 4 to 5.5 percent of the amount applied. However, this represents an upper bound on total loss because loss results correspond to the 95th percentile, which is a nearly "worst" case.

95th percentile loss as a percent of amount applied 1960s 1970s 1980s 1990s
Leaching 1.5% 1.0% 0.5% 0.5%
Dissolved Runoff 1.9% 3.4% 3.5% 3.1%
Adsorbed Runoff 1.0% 1.1% 1.2% 1.5%
Sum 4.4% 5.5% 5.2% 5.1%

Corn was generally, but not always, the dominant crop associated with high potential risk scores at the national level. Corn was dominant for all time periods for runoff and leaching risk for protection of drinking water, runoff and leaching risk for protection of algae, and all but the early 1960s for runoff and leaching risk for protection of crustaceans (table 6). Cotton, potatoes, soybeans, tobacco, and sorghum were frequently the second or third most important crop in determining the national potential risk scores for these 6 indicators. For fish risk indicators, however, several crops were dominant at one time or another: corn, peanuts, tobacco, cotton, and potatoes. Relationships between potential risk scores and crops at the state and regional level will often be quite different than that shown for the Nation, however, because no single region or state has the same crop mix as the entire Nation.

The 8 pesticide risk indicators varied markedly over both space and time. The spatial distributions for the two indicators associated with drinking water are shown in map 1 and map 2. The runoff risk indicator is greater in the Midwest states and the Mississippi embayment region than in other parts of the country. The leaching risk indicator is high in some of these same areas, but is also high in the Southeast and Mid-Atlantic states.

National aggregates demonstrate the general temporal trends. For all four target groups, pesticide risk in runoff was much greater than pesticide risk in leachate. The pesticide runoff indicator for protection of drinking water showed a dramatic reduction in risk from the 1960s and early 1970s (figure 3). Leaching risk for drinking water, however, had highest scores in 1996 and 1997, and the series appears to be trending upward (figure 4). For fish, there was little overall trend in the runoff indicator (figure 5), but leaching risk in 1996-97 was less than half the level in 1976-78 figure 6). Potential risk levels for algae were, overall, over 25 times greater than risk levels for the other three groups, reflecting algae's sensitivity to herbicides. Pesticide risk for algae showed modest reductions in recent years from the high levels in the late 1970s for both runoff and leaching (figure 7). Crustacean risk had trends generally similar to fish (figure 8).

Overall, it appears that pesticide risk for all groups initially increased in the early 1960s to high levels in the 1970s, and then either decreased or remained about constant through the 1980s and 1990s.

These national-level trends, however, obfuscate the underlying spatial-temporal diversity. In spite of the sharp fall in the national indicator for runoff risk for drinking water (decrease of 60 percent from 1973-74 to 1996-97, ending up at a level far below levels in the 1960s), for example, there are areas of the country where that risk is increasing, as well as areas where the potential risk has fallen more dramatically. Runoff risk for protection of drinking water in Louisiana generally increased throughout the time period (figure 9). Runoff risk in Alabama and Mississippi increased to a peak in 1976-77, followed by a decline through 1988, and then increased again through most of the 1990s (figure 10). In Wisconsin, runoff risk increased steadily through the 1960s and 1970s and then remained relatively high except for the last 3 years in the series (figure 11). New York and Pennsylvania had trends similar to Wisconsin, but the downward trend began earlier (figure 12). Runoff risk in Michigan and Minnesota also increased steadily through the 1960s and 1970s, but then mostly decreased through the 1980s and 1990s to levels similar to those in the 1960s (figure 13). Texas showed a jump in runoff risk in the 1970s, but otherwise showed little trend (figure 14). Runoff risk in four Northern Plains states fluctuated throughout the period, but showed little or no overall trends (figure 15). Runoff risk in Iowa, which had the highest risk score of any state throughout most of the time period, doubled between 1960 and 1973-74, and then gradually fell (with some fluctuation) to levels below those at the beginning of the series (figure 16). Midwest states (Ohio, Illinois, Indiana, and Missouri) had trends similar to the national trends, except that the percent reduction was greater (75-90 percent decrease from 1973-74 to 1996-97 for Ohio, Indiana, and Missouri) (figure 17). The Southeast showed trends similar to the Midwest, but the decreases in risk occurred slightly earlier and generally were sharper (figure 18). Runoff risk scores for these 25 states comprised 96-97 percent of the total risk score for all 48 states.

The other 7 pesticide risk indicators similarly show this kind of diversity in temporal trends at the state and regional levels. Thus, the national level trends are useful as overall trends for the agriculture sector, but do not necessarily represent specific production areas.

The indexes of potential risk change over space and time because of: 1) changes in acres planted, 2) changes in percent acres treated, 3) changes in the suite of pesticides used, and 4) changes in application rates. In addition, it is the changes in these factors for the more mobile and/or persistent pesticides and on the more vulnerable acres (high intrinsic potential for pesticide loss) that influence the indexes the most. Sometimes these factors work together to increase/decrease pesticide risk, and sometimes they offset. Changes in acres planted and changes in application rates would both generally be expected to be gradual; however, these changes can introduce sharp fluctuations when relative prices shift quickly or when government programs are introduced. The Payment in Kind (PIK) program in 1983, for example, sharply reduced acreage in field crops in most areas of the country in 1983, but levels returned in 1984 when the program was discontinued. Sharp changes in the indicators would also occur during times when the relative price of corn and soybeans changes, since corn generally has had a higher risk score than soybeans. The introduction and rapid adoption of new pesticides or the eradication of a pest problem (such as the boll weevil in cotton production) can also result in sharp changes in the risk indicators in specific areas. Fluctuations in the state-level time series for the pesticide risk indicators can be tracked back to one or more of these factors.

Table 1. 120 Basic Soils Represented By Surface And Subsurface Texture

Table 2. Source of county estimates of acres planted for 12 crops, 1960-97.

Table 3. Pesticide use surveys used to construct the pesticide use database for 1960-97.

Table 4. Pesticides used to estimate environmental risk indicators.

Table 5. Water quality thresholds (concentrations in parts per billion) for pesticides included in the potential risk indicators.

Table 6. Percentage of Pesticide Risk Indicator (TEUs) by Crop.

Map 1. Pesticide Runoff Risk for Protection of Drinking Water, 1996 - 1997 Average.

Map 2. Pesticide Leaching Risk for Protection of Drinking Water, 1996 - 1997 Average.

Figure 1. Trend in Quantity of Pesticides Applied to 12 Crops.

Figure 2. Trends in Quantities of Pesticides Lost from Farm Fields for 12 Crops.

Figure 3. Pesticide Risk Indicators for Protection of Drinking Water.

Figure 4. Pesticide Leaching Risk Indicator for Protection of Drinking Water.

Figure 5. Pesticide Risk Indicators for Protection of Fish.

Figure 6. Pesticide Leaching Risk Indicator for Protection of Fish.

Figure 7. Pesticide Risk Indicators for Protection of Algae.

Figure 8. Pesticide Risk Indicators for Protection of Crustaceans.

Figure 9. Pesticide Runoff Risk Indicator for Protection of Drinking Water, Louisiana.

Figure 10. Pesticide Runoff Risk Indicator for Protection of Drinking Water, Mississippi and Arkansas.

Figure 11. Pesticide Runoff Risk Indicator for Protection of Drinking Water, Wisconsin.

Figure 12. Pesticide Runoff Risk Indicators for Protection of Drinking Water, New York and Pennsylvania.

Figure 13. Pesticide Runoff Risk Indicators for Protection of Drinking Water, Michigan and Minnesota.

Figure 14. Pesticide Runoff Risk Indicator for Protection of Drinking Water, Texas.

Figure 15. Pesticide Runoff Risk Indicators for Protection of Drinking Water, 4 Northern Plains States.

Figure 16. Pesticide Runoff Risk Indicator for Protection of Drinking Water, Iowa.

Figure 17. Pesticide Runoff Risk Indicators for Protection of Drinking Water, 4 Midwest States.

Figure 18. Pesticide Runoff Risk Indicators for Protection of Drinking Water, Southeast States.