FIRST: A Screening Model to Estimate Pesticide Concentrations in Drinking Water
May 1, 2001
This paper describes the development and use of the FIRST (FQPA Index Reservoir Screening Tool) model by the Environmental Fate and Effects Division (EFED) of the USEPA Office of Pesticide Programs (OPP). With the passage of the Food Quality Protection Act (FQPA) by Congress in August of 1996, EFED was asked to develop methods and tools to begin estimating human exposure to pesticides for use in the FQPA risk assessment process.
On this Page
- New Scenario for Lower Tiers
- Index Drinking Water Reservoir
- Percent Crop Area (PCA) Adjustments
- Differences between FIRST and GENEEC
- Spray Drift
- Maximum Percentage of Pesticide Dissolved in the Reservoir
- Flow Through the Reservoir
- Use of the Percent Cropped Area (PCA) Factor
- Reporting of Output Values
- Impact of Pesticide Partitioning
- Other Processes Simulated
- Limitations on Using FIRST for Drinking Water Exposure Assessments
EFED has been performing pesticide aquatic exposure assessments as a part of the ecological risk assessment process for a number of years. Ecological exposure assessments have been routinely carried out through computer modeling of a "standard" agricultural field-farm pond scenario and through use of field monitored pesticide concentration values. Although this "standard scenario" was designed to predict pesticide concentrations in the standard farm pond, it has been shown to be a good predictor of upper level pesticide concentrations in small but ecologically important upland streams (Effland et al., 1999). EFED has used the EPA-developed Pesticide Root Zone Model (PRZM) (Carsel et.al., 1997), the EXposure Analysis Modeling System (EXAMS) (BURNS et.al., 1997) and the GENeric Estimate Exposure Concentration (GENEEC) (Parker, et. al., 1995) models for these exposure assessments. GENEEC is a Tier 1, screening meta-model designed to mimic a PRZM/EXAMS simulation but requires fewer inputs and much less time and effort to use.
The linked PRZM and EXAMS models simulate the impact of daily weather on the treated agricultural field over a period of thirty-six years. During this time, pesticide is washed-off of the field into the water-body by twenty to forty rainfall/runoff events per year. Each new addition of pesticide to the water-body adds to the pesticide which has arrived earlier either through previous runoff events or through spray-drift and begins degradation on the day it reaches the water. FIRST, like GENEEC, on the other hand, are single event models. Both assume that one single large rainfall/runoff event occurs and removes a large quantity of pesticide from the field to the water all at one time.
GENEEC and the linked PRZM/EXAMS models with the "standard field-pond" scenario form the first two tiers of a tiered system of ecological exposure modeling. This tiered system is designed to minimize the amount of analysis which is required to evaluate any given chemical. Each of the tiers is designed to screen out pesticides by requiring higher, more complex levels of investigation only for those that have not passed the previous tier. Each tier screens out a percentage of pesticides from having to undergo a more rigorous review prior to registration or reregistration. "Passing" a given assessment tier indicates that there is a low possibility of risk to human health. "Failing" an assessment tier, however, does not mean the chemical is likely to cause health problems, but that the assessment should continue on to the next higher tier. As matter of policy, OPP does not take significant regulatory action based on the results of screening models.
New Scenario for Lower Tiers
For human health exposure assessments undertaken for FQPA, EFED has attempted to develop a consistent, parallel modeling system to replace that developed for ecological risk assessment. In recognition of the fact that very few people drink water derived from small farm ponds or from small, upland streams, EFED has developed an "index agricultural watershed-drinking water reservoir" or "index reservoir" scenario to replace the "standard field-pond" scenario. In addition, it is recognized that most watersheds large enough to support a community drinking water system (CWS) are not entirely planted in only one crop. The modeling system, therefore, has develop a method to consider a maximum percent cropped area (PCA) factor to account for this fact.
New Tier 2 modeling continues to use current versions of the linked PRZM and EXAMS models, but the "standard field-pond" scenario has been replaced by the "index reservoir-watershed" scenario. New Tier 1 modeling uses the FIRST program in place of the GENEEC program. The "FIRST" program is designed to mimic a more complex simulation using the linked PRZM3 and EXAMS 2.97.7 models but requires less time and effort to complete. The following describes these changes to the modeling system. Note: Neither of these tiers considers the potential impact of water treatment processes on removal of pesticide from the water that eventually reaches the consumer.
Index Drinking Water Reservoir.
In July, 1998, OPP presented to the FIFRA SAP a proposed "index" reservoir scenario to replace the "field pond" scenario used at that time in its screening models to estimate pesticide concentrations in drinking water derived from surface water. The concept behind use of a model of an "index" reservoir to screen pesticides is that the chosen reservoir - and its associated characteristics - would become the standard set of conditions by which EPA would judge the potential of a pesticide to contaminate drinking water derived from surface water. The "index" reservoir would be selected from a group of reservoirs that provide drinking water to communities throughout the country. EPA would pick a particular reservoir that has characteristics associated with a higher potential for pesticide contamination of surface water and use those real world characteristics in its mathematical screening model. Because the "index" reservoir models real world characteristics, it is likely to produce more realistic estimates of pesticide concentrations in surface water. Because the "index" reservoir has characteristics that are associated with a higher potential for pesticide contamination of surface water, the model is likely to be protective of other drinking water sources which are less vulnerable to contamination.
The actual reservoir simulated is a small drinking water reservoir located in Shipman, Illinios. Shipman City Lake is 13 acres in area, 9 feet deep, and has a watershed area of 427 acres. Its ratio of drainage area to capacity (volume of water in the lake) is approximately 12. As a comparison, the "field pond" currently used has a ratio of 5. Shipman City Lake is one of a number of mid-western reservoirs that tend to be small and shallow with small watersheds, and frequently have Safe Drinking Water Act (SDWA) compliance problems with atrazine, a herbicide widely used on corn grown in these watersheds. The Index Drinking Water Reservoir characteristics have been incorporated into the linked PRZM \ EXAMS and FIRST models and are implemented in conjunction with percent cropped area adjustment (FIFRA SAP, 1999). While estimates of pesticide concentrations in drinking water based on a Midwestern index drinking water reservoir may not be representative of residue levels in drinking water sources in other parts of the country, the scenario provides an effective screening tool to determine the need for more extensive refinements. The modeling scenarios currently account for region-specific rainfall, soil, and hydrologic/runoff factors. Steps to develop scenarios for regional reservoirs for advanced tiers of modeling have been hampered by the lack of monitoring data outside of the Midwest that is of sufficient quality and extent to develop scenarios for additional reservoirs.
Percent Crop Area (PCA) Adjustments
The PCA is a generic adjustment which represents the maximum percent of any watershed that is planted to the crop or crops being modeled and, thus, may potentially be treated with the pesticide in question. PCA factors are generated from Geographic Information System (GIS) overlays of cropping area and watershed delineations and are applied to estimates of index reservoir surface water pesticide concentration values from the PRZM/EXAMS and FIRST models. The output generated by these models is multiplied by the maximum decimal fraction of cropped area in any watershed generated for the crop or crops of interest. To be effective as an adjustment to screening model estimates, the PCA should result in estimated concentrations that are closer to, but not less than, actual pesticide concentrations in vulnerable (prone to pesticide-laden runoff) surface water sources. While it moves away from assuming that the entire watershed would be treated at the same time, the PCA is still expected to be a screen because it represents the highest percentage of crop cover of any large watershed in the U.S. and it assumes that the entire crop is being treated. Current PCA values and a description of how the values are derived and applied can be found in US EPA OPP (2000a). The PCA adjustment is only applicable to pesticides applied to agricultural crops. The PCA is not to be applied to groundwater modeling assessments.
As the "Index Reservoir" replaced the "standard field-pond", the FIRST model replaced GENEEC as the surface water screen for screening level model for FQPA pesticide exposure assessments. Like the more complex and resource-intensive PRZM and EXAMS models which it mimics, FIRST estimates pesticide concentrations in a vulnerable index reservoir. As with GENEEC, FIRST considers adsorption of the pesticide to soil or sediment, incorporation of the pesticide at application, direct deposition of spray drift into the water body, and degradation of the pesticide in soil before runoff and within the water body. Simulation results are adjusted to account for the percentage of the watershed in the crop being assessed. FIRST is expected to be exceeded by measured pesticide concentrations in drinking water only very rarely due to the conservative nature of the model assumptions. It represents a small drinking water reservoir surrounded by a runoff-prone watershed, uses maximum pesticide application rates and assumes that no buffer exists between the reservoir and the treated fields. The simulation assumes runoff from a single, large rainfall event.
Differences Between FIRST and GENEEC
Differences between FIRST and GENEEC include
the physical scenario,
the treatment of spray drift,
the maximum percentage of applied pesticide which can reach the water-body,
the flow through the reservoir,
the use of a PCA and
the output values which are reported by the model.
The physical size of the watershed-reservoir system is much larger that the standard field-pond system. The index reservoir scenario assumes a 172.8 hectare watershed channels runoff to a 144,000 cubic meter water-body. The standard field-pond scenario assumes a 10 hectare field provides runoff to a 20,000 cubic meter water-body. The reservoir is 2.74 meters deep and the standard pond is 2.00 meters deep.
Reservoir water concentrations may be increased due to deposition of spray drift into the feeding stream or directly into the reservoir itself. The user of the FIRST model may simulate application by aerial spray, air blast spray, ground spray or broadcast application of granular material. For aerial application the program assumes that there is sixteen percent spray drift directly into the reservoir with 95 percent landing on the field (application efficiency). For air blast application to orchards, groves and vineyards, FIRST assumes 6.3 percent spray drift directly into the reservoir with 99 percent landing on the field. For ground spray, 6.4 percent goes directly to the reservoir with 99 percent being deposited on the field. For granular broadcast application, 100 percent application efficiency is assumed with no pesticide drifting directly to the stream or reservoir. Biological and abiotic degradation of the spray drift in the reservoir begins immediately. Degradation of the pesticide reaching the reservoir via runoff begins two days later (on the day it reaches the reservoir). NOTE: The application efficiency values and the spray drift values do not add to 100% because the application efficiency is a percentage deposited on each hectare of the watershed and the spray drift is a percentage of the application rate. GENEEC Version 2.0 uses a subroutine developed by the Spray Drift Task Force (SDTF) which calculates spray drift based on the upper 90th percentile of expected drift percentages. The SDTF subroutine allow the user to calculate a drift reduction based on the type of application and the spray droplet size distribution. It also can simulate the impact of a buffer between the treated field and the modeled water body.
Maximum Percentage of Pesticide Dissolved in the Reservoir.
The FIRST model assumes that up to eight percent of the pesticide applied to this 427 acre (172.8 hectare) watershed is washed into the reservoir by one large storm. Eight percent was chosen in order to yield output values which exceed both PRZM/EXAMS modeled values and upper level monitored values in all but a very few rare cases. The actual amount which appears as the dissolved (bioavailable) concentration estimate is a function of the equilibrium partition coefficient (Kd) or the organic carbon normalized equilibrium partition coefficient (Koc). Either of these parameters may be chosen by the model user to partition the pesticide in this watershed/reservoir system into two separate phases: a dissolved (in water) phase and an adsorbed (to field soil or reservoir bottom sediments) phase. The EXAMS parameter PRBEN remained at its default value of 0.5 to equally divide influent pesticide between the water column and the bottom sediments. It is pesticide in the dissolved phase in the water column which is considered to be toxicologically available when consumed. It is therefore only the mass of dissolved pesticide which enters into the calculation of the concentration in the reservoir. The GENEEC model assumes that between ten percent and 0.1 percent of the pesticide applied to the field ends up in the water body in a dissolved (bioavailable) form. The actual fraction is also a function of the Kd or Koc.
Flow Through the Reservoir.
FIRST also assumes that flow from the watershed is sufficient for two full reservoir turnovers each year. In other words, the annual flow through the reservoir is equal to twice the reservoir volume of approximately 144,000 cubic meters. This is equivalent to a flow of approximately 33 cubic meters per hour (EXAMS parameter STFLO) through the reservoir. The reduction in pesticide concentration as a result of this flow is related to the partition coefficient (Kd) of the chemical. At a Kd of 1.0 or less, PRZM program simulation shows an approximate 2.5% reduction of the peak concentration and a 30% reduction of the annual average concentration. The reduction is greatest at a Kd of 10 with a 7% reduction of the peak concentration and a 35% reduction of the annual average concentration. Above a Kd of 10,000 there is no reduction as a result of flow. This reduction is programmed into FIRST. GENEEC does not account for flow or pond turnover in its simulations.
Use of the Percent Cropped Area (PCA) Factor.
As reported above, FIRST considers reduction of the area within the reservoir which is planted to the modeled crop but considers that pesticide is applied to all of the crop which is planted. GENEEC does not consider any non-cropped area within the ten hectare field simulated and therefore does not use a PCA.
Reporting of Output Values.
The FIRST model generates two concentration values: the peak value which occurs on the day of the single large rainstorm and the annual average value which is the mean of the peak and the values over the next 364 days. The peak value is used for acute exposure assessments and the annual average value is used for chronic and cancer exposure assessments. GENEEC Version 1.0 estimated the peak value which occurs on the day of the single large rainstorm as well as multiple day averages over periods of 4, 21, and 56 days. During the period while the OPP Health Effects Division (HED) has used GENEEC values for FQPA exposure assessments, the peak value was used directly for acute assessments and the 56-day average concentration value was used but divided by 3 when comparing it to the chronic DWLOC. Division by 3 is not done with the annual average (chronic exposure value) from FIRST. GENEEC Version 2.0 estimates the peak value which occurs on the day of the single large rainstorm as well as multiple day averages over periods of 4, 21, 60 and 90 days.
Impact of Pesticide Partitioning
The impact on the system of changing the partition coefficient was determined through repetitively increasing the Kd value within a PRZM/EXAMS simulation, running each new simulation and then recording the resulting instantaneous concentration for each value. This gave a series of dissolved pesticide concentration values as a function of Kd. A dissolved versus adsorbed relationship to Kd was established and programed into FIRST. The value of the Kd parameter in the EXAMS program controls not only the final equilibrium partitioning of the chemical between the dissolved and adsorbed phases, but also determines the time it takes to reach this equilibrium. For very large Kd values, the binding takes place largely within the first day while for smaller values of Kd, the process may not be complete for almost a year.
In order to accurately mimic this EXAMS process in the FIRST model, an empirical procedure was carried out to simulate a pseudo-binding rate as a function of Kd. This rate was determined by "turning off" all degradation processes within the PRZM and EXAMS models and calculating an apparent rate constant that would account for the continuous decline in concentration values. A four parameter Morgan-Mercer-Flodin (1975 ) type function was then fit to calculate these apparent binding constants as a function of Kd. This ongoing adsorption within the reservoir is then programmed to occur simultaneously with chemical and biological degradation processes.
OTHER PROCESSES SIMULATED
Overall Degradation Rate
Calculating degradation in the reservoir is for the purpose of estimating annual average concentration values for chronic exposure assessment. The EXAMS program, upon which the aquatic degradation is based, considers the amount of light penetration in the simulated water body to calculate an effective photolysis rate from the direct photolysis rate in clear water which is used as input. Based on EXAMS simulation, the photolysis rate constant in the relatively murky index reservoir is 124 times slower than that in clear water. The overall degradation rate in the reservoir is calculated by summing the aerobic aquatic metabolism rate constant, the abiotic hydrolysis rate constant and 1/124 th of the aquatic direct photolysis rate constant. This sum is the combined rate given in the output table.
The program also allow the user to account for those pesticides which are incorporated at the time of application. Incorporation reduces the mass of pesticide available to runoff by a factor equal to the depth of incorporation in inches up to a maximum of six inches.
Degradation in the Field
Some pesticides are designed to degrade very quickly in the field so that in a short time, little residue remains. In order to give "credit" to such pesticides, FIRST was designed to allow a two day degradation period for the pesticide in the treated agricultural field prior to the single rainfall event which washes the pesticide into the stream feeding the reservoir and eventually to the reservoir itself. The aerobic soil metabolism rate is used to simulate the rate of decline during this two day period as well as during the period between multiple applications prior to the rainfall/runoff event. (For pesticides whose label requires that the pesticide be "wetted-in" at the time of application either through irrigation or rainfall, the two day period is not used).
Use of the Binding Parameter, Kd, in Preference to Koc to Represent Pesticide Attachment to Soil, to Organic Matter or to Water-body Bottom Sediments
Adsorption (binding) tests are performed on soils of several textural classes prior to pesticide registration. The soil/water equilibrium partition coefficient (Kd) is defined as the ratio between the concentration in soil and the concentration in water. It can therefore be used to estimate the dissolved or the adsorbed fraction in a soil-water system for any chemical. For this reason it is in general the preferred parameter for this purpose and is recommended for use in GENEEC version 2.
The organic carbon normalized soil/water equilibrium partition coefficient (Koc) may be preferred for pesticides for which there is a strong positive correlation between the Kd value and the organic carbon content of the soils on which the adsorption tests were performed. If there is correlation, the multiple Koc values will be less variable than the multiple Kd values and the Koc is likely to be a more accurate estimator. If there is no correlation, use of the Kd is preferable. The Kd / Koc conversion is based on an organic matter content of 2 percent and an organic carbon content of 1.16 percent. If neither the Kd nor the Koc is available, it is recommended to use 0.35 times the Kow value.
There may be cases in which the amount of pesticide which is washed off of the treated field as a result of a large rainstorm exceeds the amount which can be dissolved in the water body. In this case a precipitate is formed and remains in that form until the concentration in the overlying water column decreases to a point below the solubility of the chemical and additional chemical can dissolve. In the FIRST program, the solubility serves as an upper limit on the amount which is dissolved in the water. It does not enter into program calculations in any other way.
Limitations on Using FIRST for Drinking Water Exposure Assessments
The FIRST program does not consider the impact of water treatment processes. Removal and transformation of most pesticides by water treatment processes is highly variable from CWS to CWS and from day to day in the same CWS and is therefore difficult to predict on a consistent basis. If consistently high removal across most CWS systems has been documented for a specific pesticide, FIRST will overestimate concentrations in drinking water to that extent.
FIRST is designed to yield concentration values which exceed those predicted by the linked EPA PRZM and EXAMS models for all but the most extreme sites, application patterns and environmental fate properties. PRZM/EXAMS predictions may exceed FIRST predictions under the following circumstances:
Applications to crops in managed environments known to produce excessive runoff (eg. crops grown over plastic mulch).
Applications at sites with hydrologic group D soils which also receive excessively high rainfall (eg. EFED sweet potato scenario in southern Louisiana).
Multiple applications over a window of 30 days or longer in exceptionally high rainfall areas (eg. far southeastern US). In each of these cases, FIRST will exceed PRZM/ EXAMS estimated peak concentrations values, but not always the annual average concentration values. Even then PRZM/EXAMS would not be expected to exceed the FIRST values by more than a factor of 2.
For applications of chemicals with half-life values of 5 days or less at exceptionally high runoff sites the PRZM/EXAMS concentrations values may exceed both the FIRST peak and annual average values by a factor of 2. Allowing these few exceedences for extreme conditions makes FIRST a more reasonable predictive tool for the rest of the country.
Burns, L.A. 1991. Exposure Analysis Modeling System: Users Guide for EXAMS II version 2.94, Environmental Research Laboratory, Office of Research and Development, U.S. Environmental Protection Agency, Athens, GA.
Burns, L.A. March 1997. Exposure Analysis Modeling System (EXAMSII) Users Guide for Version 2.97.5, Environmental Research Laboratory, Office of Research and Development, U.S. Environmental Protection Agency, Athens, GA.
Carsel, R.F., C.N. Smith, L.A. Mulkey, J.D. Dean and P. Jowise. 1984. Users manual for pesticide root zone model (PRZM): Release 1, Rep.EPA-600/3-84-109, 219 pp. Environmental Research Laboratory, Office of Research and Development, U.S. Environmental Protection Agency, Athens, GA.
Effland, W.R., Thurman, N.C., Kennedy, I. Proposed Methods For Determining Watershed- Derived Percent Cropped Areas and Considerations for Applying Crop Area Adjustments To Surface Water Screening Models; USEPA Office of Pesticide Programs; Presentation To FIFRA Science Advisory Panel, May 27, 1999.
R.F. Carsel, J.C.Imhoff, P.R.Hummel, J.M.Cheplick and J.S.Donigian, Jr. 1997. PRZM-3, A Model for Predicting Pesticide and Nitrogen Fate in Crop Root and Unsaturated Soil Zones: Users Manual for Release 3.0; Environmental Research Laboratory, Office of Research and Development, U.S. Environmental Protection Agency, Athens, GA.
Parker, R.D., R.D. Jones and H.P. Nelson., 1995. GENEEC: A Screening Model for Pesticide Environmental Exposure Assessment.,in Proceedings of the International Exposure Symposium on Water Quality Modeling; American Society of Agricultural Engineers, pp. 485-490; Orlando, Florida.