Integrating a Distributional Approach to Using Percent Crop Area (PCA) and
Percent Crop Treated (PCT) into Drinking Water Assessments
U.S. Environmental Protection Agency
Office of Pesticide Programs
12/31/2019
1
-------
Table of Contents
1. EXECUTIVE SUMMARY 3
2. INTRODUCTION 4
2.1. Regulatory Context 4
2.2. Purpose Statement 4
2.3. Document Organization 5
3. OVERVIEW OF PCA AND PCT 6
3.1. Introduction 6
3.2. Percent Cropped Area (PCA) 6
3.3. Proposal to Incorporate Usage Data into Endangered Species Assessments 9
4. APPLICATION OF DISTRIBUTIONAL PCA REFINEMENTS TO DWAs 11
4.1. Purpose 11
4.2. Available Data 11
4.3. Method Description 12
4.3.1. Use Pattern Specific and Non-Standard Crop PCAs 12
4.3.2. Full Distribution of Watershed PCAs with Land Cover Class Overlap 13
4.3.3. Aggregated EDWC Based on Individual Crop PCA Values 15
4.4. Summary 17
5. APPLICATION OF STATE LEVEL PCT REFINEMENTS TO DWAS 18
5.1. Purpose 18
5.2. Available Data and Integration into DWA 18
5.3. Procedure 20
5.4. Conclusions 25
6. METHOD EVALUATION 25
7. PROPOSED USE/DRINKING WATER ASSESSMENT IMPLICATIONS 25
8. REFERENCES 27
Appendix A. Drinking Water Assessment Tiering Framework 29
Appendix B. Case Study 32
Appendix C. Determination of Overlap of Treated Area and Drinking Water Watersheds
Located in the 48 Conterminous States 33
Appendix D. Usage Data Sources 38
2
-------
1. EXECUTIVE SUMMARY
The Office of Pesticide Programs (OPP) estimates human exposure to pesticides as part of the
aggregate exposure assessment via all routes. Drinking Water Assessments (DWA) consider
whether use of a pesticide according to the approved label, whether the use be to agricultural
or non-agricultural sites, will lead to exposure via drinking water. Tier 1 and 2 DWAs assume
that the modeled watershed supplying a community water system (CWS) is 100% planted with
the crop of interest, and that 100% of the crop is treated with the pesticide of interest. These
assumptions are referred to as 100 % "Percent Cropped Area" (PCA) and 100 % "Percent Crop
Treated" (PCT).
One method OPP currently uses to refine its estimated drinking water concentrations (EDWC) is
to refine the assumption of 100% PCA. This is done by applying a PCA that represents the
multiple crops or uses for the pesticide. These PCAs are calculated from GIS land cover data and
USDA crop acreages, that estimate the percentage of land in particular agricultural or other
uses. The highest PCA from any single CWS watershed, either on a national or regional (HUC-2)
basis, is currently used to refine the EDWC at DWA Tier 2.
This White Paper proposes methods to apply the full range of PCA values from individual CWS
watersheds, not just the maximums, to the process of modeling EDWCs. Essentially, the
appropriate PCA for the modeled use(s) will be applied within the particular watershed, and the
resulting modified EDWCs presented as a distribution of EDWC across all CWS watersheds
within a HUC-2. This distribution is compared to the Drinking Water Level of Comparison
(DWLOC) supplied by the Health Effects Division, to determine what percentile of PCA-modified
EDWC falls below the DWLOC (passing) and what percentile falls above the DWLOC (failing).
The cut point between passing and failing is the "critical PCA," and this determines which CWS
watersheds require further refinement of their EDWC.
In this White Paper OPP also proposes the application of PCT data below the default
assumption of 100% to the DWA process. Data on PCT of various agricultural crops is supplied
by the Biological and Economic Analysis Division (BEAD) based on United States Department of
Agriculture (USDA) survey and Kynetec USA (i.e. private market survey data)1. This data is
summarized on a state level. PCT data for non-agricultural uses is typically available on a
national or regional basis. The method described here for applying PCT data to the DWA
process deals in large part with bridging the spatial scale mis-match between the PCT data
(state level or higher) and the PCA data, which is available at the watershed (sub-state, sub-
county or cross-state) level. Methods for dealing with the uncertainty of the geographic overlap
of cropped acres (PCA) and treated acres (PCT) are developed, in the context of combining
these two factors that modify the EDWC.
1 Kynetec USA, Inc. 2019. 'The AgroTrak* Study from Kynetec USA, Inc." Database Subset: 1998-2018
3
-------
2. INTRODUCTION
2.1. Regulatory Context
Pesticides are regulated in the United States under both the Federal Insecticide, Fungicide, and
Rodenticide Act (FIFRA) and the Federal Food, Drug and Cosmetics Act (FFDCA). Through these
statutes, the Environmental Protection Agency (EPA) Office of Pesticide Programs (OPP) must
ensure that aggregate exposure to the pesticide residues is safe, i.e., that "there is a reasonable
certainty of no harm" from exposure to the pesticide.
OPP's Environmental Fate and Effects Division (EFED) is responsible for conducting Drinking
Water Assessments (DWA), which include an analysis of the potential for and magnitude of
pesticide occurrence in drinking water from both surface water and groundwater sources.
USEPA OPP estimates drinking water concentrations (EDWCs) in surface water that supply
Community Water Systems (CWS) and compares them to benchmark values called drinking
water level of comparison (DWLOC) to determine if pesticide concentrations have the potential
to cause adverse effects to human health (USEPA OPP, 2019b, p. 13). EFED employs a robust,
tiered DWA process designed to efficiently screen out pesticides that do not pose a potential
risk to human health from those requiring more highly refined analyses to better understand
potential risks (e.g., in terms of when and where there may be concerns). Lower tier
assessments are intended to be conservative so that the assessor can confidently screen out
chemicals that represent a low risk (Appendix A). Higher tiers successively incorporate
refinements that draw on more focused chemical, spatial, temporal, and agronomic
information including consideration of available monitoring data to inform risk management
decisions. For additional information on USEPA OPP's tiered approach to DWAs, including a
detailed description of individual tiers, see the Draft Framework for Conducting Pesticide
Drinking Water Assessments for Surface Water (USEPA OPP, 2019b).
2.2. Purpose Statement
As part of OPP's tiered approach to DWAs, OPP estimates pesticide concentrations in surface
water using simulation models (e.g., Pesticides in Water Calculator (PWC), and Pesticide
Flooded Application Model (PFAM))2. At the lower tiers, OPP estimates the application of a
particular pesticide as defined by the label (use) and assumes the pesticide is applied to the
entire area that may be contributing to a surface water in a watershed. These conservative
assumptions lead to the generation of upper-end estimates of pesticide concentrations. In an
effort to advance methodologies used in DWAs, OPP developed approaches to refine these
estimates for higher tiered assessments described in this document. First, OPP developed an
approach to account for the variability in the PCA across watersheds nationally and within
specific regions that supply CWSs. Second, OPP developed a method to incorporate data on the
2 https://www.epa.gOv/pesticide-science-and-assessing-pesticide-risks/models-pesticide-risk-assessment#aquatic
4
-------
amount of a target crop or group of crops that are actually treated with a given pesticide
(referred to as percent crop treated or PCT). The goal of the PCA and PCT refinements are to
generate EDWCs that are protective of human health that reduce the magnitude of
overestimation due to variability in crop acreage and actual pesticide usage. While OPP
considers potential exposure to conventional pesticides in both surface and groundwater
sources of drinking water, this effort focuses on refining estimated pesticide concentrations in
surface water.
OPP currently uses a suite of PWC scenarios that have been developed over time for a variety of
reasons. In all cases, the scenario was developed with a goal of representing a high-end
exposure setting. The current scenarios span the United States for a range of crops/use sites
and while they were developed using consistent guidance, they do not all represent a similar
level of runoff potential. The current scenarios can be found at the following website.
https://www.epa.gov/pesticide-science-and-assessing-pesticide-risks/models-pesticide-risk-
assessment. Currently, OPP has a companion project to the PCA/PCT project to use spatial data
sets to develop a new suite of scenarios for use in surface water DWA. Once developed these
new scenarios will replace the existing scenarios and will provide a more quantitative level of
runoff potential.
The PCA/PCT project begins with EDWC that are developed using PWC scenarios available at
the time of the DWA. However, the approach is developed in such a way as to be independent
of how they were developed. The operating assumption is that in both cases the EDWC
resulting from a scenario at the national or regional scale represents an upper bound EDWC
representing a high-end exposure for the geography from which it was selected.
2.3. Document Organization
This document presents the background on and methods for applying PCA and PCT as
refinements to EDWCs in high-tier (i.e., tier 3 and tier 4) DWAs. Each section of the document is
dedicated to a separate component of the process. The sections in this document are organized
as follows:
Section 3 provides an overview of the previous PCA and PCT efforts to date.
Section 4 describes application of distributional PCA refinements to DWAs.
Section 5 describes the application of state-level PCT refinements to DWAs.
Section 6 describes the process for comparing the refined EDWC values to monitoring data.
Section 7 discusses the implications of the new PCA and PCT refinements to DWAs.
A case study of the application of these methods can be found in Appendix B.
5
-------
3. OVERVIEW OF PCA AND PCT
3.1. Introduction
Lower tier assessments are intended to be conservative in order to screen out chemicals that
represent a low risk (Appendix A). In estimating a pesticide's concentrations in surface water,
OPP routinely relies on a variety of pesticide use information representing how the pesticide
may be used as specified on the pesticide label and usage data which includes actual reported
information on how the pesticide is actually applied including where the pesticide is applied
and how much is applied. OPP incorporates more detailed pesticide use and usage data when
refining EDWCs (moving from low to high tier DWAs).
Pesticide use parameters are based on labels which define how a pesticide may be applied.
These label parameters generally include: use patterns (e.g., target pest(s), use sites, residential
uses), maximum application rates, application method (e.g., aerial or ground spray), minimum
retreatment intervals, and the maximum number of applications allowed per crop cycle or year.
Usage data describes actual pesticide applications. These data include information such as
actual application rates for a particular use site (e.g., corn), timing of application, methods of
application (e.g., aerial) and use intensity (PCT) at the state level. OPP obtains usage data from
a variety of public and proprietary surveys. The key difference between use and usage is that
the use represents potential applications of a pesticide (use sites and application rates as they
appear on the label) while usage represents what is typically applied for a given pesticide that
are observed to occur in the field (from growers and users). OPP calculates EDWCs employing
use data for low-tier assessment and integrates progressively more usage data in subsequent
tiers as describe in the DWA Framework (USEPA, OPP, 2019b).
PCA and PCT are key data that can be used to refine DWA analysis moving from low to high
tiers. PCA is defined as the areal fraction of a watershed with a land cover (e.g., corn, wheat,
vegetables, turf) that can be treated with a pesticide based on registered uses. PCT is defined as
the percent of base acres treated (acres treated with a given pesticide one or more times)
relative to the total number of acres grown. For low tier assessments, OPP conservatively
assumes 100% percent cropped area and 100% percent crop treated within a given watershed.
A better understanding of PCA and PCT allows for the calculation of EDWC values that more
accurately represent the potential for exposure to a given pesticide. The following sections
provide additional background on PCA and PCT and their use in DWAs.
3.2. Percent Cropped Area (PCA)
CWS watersheds large enough to support a drinking water facility rarely consist of a single crop
(e.g., apples) or land cover type (e.g., orchards). To account for the variability in use patterns, in
Tier 2-4 DWAs, OPP currently uses PCA adjustment factors to reflect the percentage of a
watershed that is covered by a particular use or land cover type. The application of PCAs to
DWAs has been extensively documented, reviewed, and utilized in OPP drinking water
6
-------
assessments (USEPA, 2014).3 OPP derived the existing PCA adjustment factors from National
Agricultural Statistics Service (NASS) Census of Agriculture (Ag Census) data, geospatial data
(National Land Cover Database, NLCD) and CWS DWI data that spatially overlapped the crop
acreage within each of the previously delineated CWS watersheds within the conterminous
United States (USEPA, 2014) for a subset of crops. OPP calculated PCA values for the land cover
class GIS layers for seven crops and groups of crop (corn, cotton, orchard, soybean, vegetables,
wheat and turf); and for selected combinations of crops (e.g., corn-wheat, soybean-wheat, turf-
corn, turf-vegetable, turf-wheat, vegetable-orchard); as well as for all agricultural land ("all-
agriculture") and all agricultural land plus residential turf ("all-agriculture plus turf"). The full list
of land cover classes with calculated PCAs can be found in Appendix C. It is important to note
that PCAs are not applicable to groundwater EDWCs or to areas outside the conterminous
United States (e.g., Alaska and Hawaii).
The current approach for using PCA refinements is to multiply the modeled base EDWCs by the
PCA of the crop(s) to give the refined EDWC values that are reflective of application to the
percentage of the watershed that could be treated with the pesticide. The PCA-adjusted
concentrations are then used as the EDWC in human health dietary risk assessment. For Tier 2
assessments, OPP uses the most conservative (maximum) PCA value of all watersheds within
the conterminous United States (for national assessments) or a specific hydrologic unit code
(HUC)-02 region (for regional assessments) to refine the EDWCs.4 OPP's current guidance for
the use of PCA refinements states that if the pesticide has registered use patterns on only one
crop/group of crops, then the PCA for that particular crop is used; if the pesticide is used on a
combination of crops with a calculated crop pair PCA, then the PCA for that particular pair of
crops is used; and, if the pesticide use include more or different crops than are captured in the
existing PCA combinations, the default "all-agriculture" PCA is applied. The maximum national
all-agriculture PCA is 100%, therefore PCA refinements will not alter the EDWCs or risk
3 In 2014 OPP identified 6,550 CWS DWI locations from EPA's Safe Drinking Water Information System (SDWIS). Of
the 6,550 locations, 74% (4,840) had unique, delineated watersheds. Of these, all but two (4,838) are located in
the continental United States. PCA values were calculated for these 4,838 DWI watersheds. The 1,710 CWS DWI
without validated watersheds (i.e., 6,550-4,840 = 1710) consists of the following: 195 were abandoned or outside
the continental US and are therefore not relevant to EPA risk assessments; 490 were water bodies that were not
appropriate to simulate with EPA standard watershed based runoff modeling (e.g., canals, aqueducts, off-stream
reservoirs, and large, incompletely-mixed water bodies such as the Great Lakes); 666 were not validated DWI
watershed; and 359 need more information on drinking water source to evaluate pesticide exposure for these
watersheds. The method described in this White Paper will ultimately be extended to all delineated CWS
watersheds and the additional HUC-12 surrogates described in the Development of Community Water System
Drinking Water Intake Percent Cropped Area Adjustment Factors for use in Drinking Water Exposure Assessments:
2014 Update (USEPA, 2014).
4 Hydrologic Units Codes are a hierarchical system developed by United States Geological Survey to catalogue
hydrological units within the United States. In this system, there are 18 individual HUC-02 regions in the
contiguous drainage areas in the United States with an average size of 177,560 mi2 (See Figure 1). The U.S. is
divided and sub-divided into smaller hydrologic units. These units are arranged within each other and identified by
a unique code consisting of two to eight digits based on the levels of classification in the hydrologic unit system.
Additional information can be found at https://water.usgs.gov/GIS/huc.html.
7
-------
conclusions in national assessments for any chemical with use patterns not covered by the
existing crop combination PCAs. The HUC-02 regional PCA for a crop may be lower than the
national PCA but still represents the most conservative PCA value for that HUC-02 region and is
used as a refinement in OPP's current DWA approach. Figure 1 shows the 18 HUC-02 in the
United States while Figure 2 shows the distribution of CWS watersheds with binned PCA values.
For additional information on the development of the CWS PCA values and use as a refinement
in DWAs see Development of Community Water System Drinking Water Intake Percent Cropped
Area Adjustment Factors for use in Drinking Water Exposure Assessments: 2014 Update (USEPA,
2014).
Figure 1. The 18 Hydrologic Unit Code (HUC) System Developed by United States Geological
Survey to Categorize Hydrologic Units fUSEPA 2014)
8
-------
DWI Watersheds
All Ag PCA
I I No delineated watershed
~ 0.00-0.10
~ 0.11 -0.25
~ 0.26-0.50
~ 0.51 -0.75
¦ 0.76-1.0
Figure 2. Geographic Distribution of the Community Water System Drinking Water Intake
Watershed and Percent Cropped Area (PCA)
3.3. Proposal to Incorporate Usage Data into Endangered Species Assessments
While OPP does incorporate some usage data into high-tier assessments {e.g., average use
rates, typical application dates), PCT has not historically been used in the standard tiered DWA
process.5 EPA is proposing to use it now with this peer review. Pesticide concentrations are
modeled on the watershed-scale, whereas PCTs are typically available at a larger scale that
reflect the available data. As such, PCTs are generally calculated at the national or state level.
The survey vendors that generate available usage data and OPP have determined that in most
cases the state level is the lowest level of aggregation that maintains the statistical power of
the studies underlying the available usage data across all years and crops. Appendix D lists the
various usage data sources used to calculate PCT values. The majority of drinking water
watersheds, however, are sub-state scale and many span portions of multiple states. The
difference in scale between state or national level PCT data and drinking water intake
watersheds has typically precluded the direct application of PCTs to DWA refinements. In
contrast, the Health Effects Division (HED) currently uses national-level PCT values as
5 PCT data has been used in certain high tier assessments, including the Organophosphorus Cumulative Risk
Assessment 2006 Update (USEPA, 2006).
9
-------
refinements for national dietary exposure analysis, but does not use state or regional level PCT
values in their assessments (USEPA, 2000).
OPP calculates PCT as the ratio of base acres6 of a crop or group of crops (e.g., dry beans)
treated with a particular pesticide in a given area, to the Crop Acres Grown (CAG) of the same
crop or group of crops in the same geographical area, based on available usage data using
Equation 1. Since OPP did not historically have a systematic method for refining EDWCs based
on PCT data, OPP generally assumed 100% PCT for a given watershed.
_ __ Base Acres Treated with Pesticide X V^AA
PCTx— X lUU
Equation 1 Crop Acres Grown
In May 2019, OPP proposed a process to incorporate usage data into biological evaluations for
federally listed endangered and threatened species (listed species assessments) (USEPA,
2019a).7 The proposed method calculates the fraction of a listed species range that overlaps
with the area that a pesticide may be applied. This involves calculating state-level PCT
aggregated across multiple crops from available usage data. OPP then uses these aggregated
state-level PCT to calculate the treated acreage within a state based on the aggregated state-
level PCT and the cropped acres from the land cover GIS layer. OPP then compares the treated
acreage to species ranges (often at the sub-state level) within a state to determine the extent
of potential overlap between the treated area and the species range. The revised method for
listed species assessments can serve as a starting point for developing a method to apply PCT
data to refined DWA for human health assessments. From a conceptual and geospatial analysis
standpoint, a species range location and a drinking water watershed are similar; both areas
represent geographically defined locations where off-target pesticide exposure could result in
exposure to the populations dependent on that area. Although, the listed species assessment
and DWA application and data sources are different; they both use spatial layers of land cover
that were derived from satellite imagery. Therefore, it is possible to apply a similar
methodology used in the proposed endangered species approach for drinking water
watersheds as a refinement to higher tier drinking water assessments.
One area addressed in the revised endangered species assessment (ESA) method is uncertainty
in the actual location of pesticide applications associated with available usage data. Therefore,
the treated area could be located in any area within a given state where the crops are grown.
To account for the uncertainty in the degree of overlap between the treated area and the
species range, OPP proposed three methods for allocation of the base acres treated (BAT)
within a state to the species range by bounding the assumption of where treated acres may
occur using most conservative, least conservative, and mid-range assumptions of how treated
6 Base Acres Treated (BAT) = The number of unique acres of a crop that were treated at least once with a specified
active ingredient in a calendar year
7 https://www.epa.gov/endangered-species/draft-revised-method-national-level-endangered-species-risk-
assessment-process
10
-------
acres are distributed. These are bounding case assumptions designed to capture the
uncertainty in the distribution of the treated acres across the state and are described as the
"upper", "uniform", and "lower" methods. Specifically, the "upper" method is completed by
allocating all BAT to the species range until 100% of the species range is considered "treated",
with the excess acres allocated to areas outside the range but still in the state ("upper" overlap;
excess acres may then be allocated to areas outside the range but still in the state). The
"uniform" method allocates the BAT to potential use sites proportionally to the aggregated PCT
(the aggregation of state level usage data for all pesticide specific uses) of the land cover class,
while the "lower" method allocates the BAT to areas within the state but outside the species
range until 100% of the area outside the species range is considered "treated", with the excess
acres allocated to areas inside the species range.
As described in Section 5, OPP is proposing to apply a similar methodology to calculate the
treated area inside a drinking water watershed. Section 5 describes and graphically presents a
proposed method to apply state-level PCT and associated treated acreage data as a refinement
to higher tier drinking water assessments. OPP plans to incorporate the extent of the treated
area to derive watershed scale PCTs that can be used to refine EDWC.
4. APPLICATION OF DISTRIBUTIONAL PCA REFINEMENTS TO DWAs
4.1. Purpose
As previously stated, the standard method for applying PCA refinements to EDWCs is to
multiply the EDWC by the maximum national or HUC-02 regional PCA. For HUC-02 regions
where the standard PCA-refined EDWC still exceed the DWLOC, there are several different
options to further refine the EDWC values based on available PCA data. This section explains
three PCA options and how they can be applied to further refine exposure estimates. These
refinements can be applied individually or in sequence, based on the available data and use
patterns for a particular pesticide.
4.2. Available Data
OPP draws the maximum national and regional PCA values for each crop or group of crops from
the full data set of PCA values calculated for each of the previously delineated CWS watershed
within the conterminous United States. These data include the HUC-02 region for each CWS
watershed, the watershed area, and the watershed PCA values for each crop/group of crops
listed in Appendix C. OPP calculated crop pairing PCA values for nine crop combinations as the
sum of the individual crop/group of crops PCA values within the watershed.8
8 The nine crop pair combinations are corn-wheat, soybean-wheat, turf-corn, turf-orchards, turf-soybean, turf-
vegetables, turf-wheat, vegetables-orchards, and all-agriculture-turf.
11
-------
4.3. Method Description
The goal of the current PCA effort is to evaluate three different methods to refine EDWCs as
part of a Tier 3 DWA where the EDWCs exceed the DWLOC after the standard Tier 2 HUC-02
regional PCA refinement. Some methods will be more appropriate than others based on the
specific use pattern of a given pesticide. During the scoping process described in the DWA
framework (USEPA, OPP 2019b), it is essential to consider the use pattern of the pesticide to
determine which of these approaches will be most beneficial. These methods are described in
the following sections.
4.3.1. Use Pattern Specific and Non-Standard Crop PCAs
Use of the all-agriculture PCA when a pesticide is registered for use on crops not covered by
one of the existing crop pair combinations can lead to overestimation of the PCA value and
resulting EDWCs, particularly when the crops each have small acreage footprints relative to the
all-agriculture PCA (e.g., cotton plus orchards). The PCA value for any combination of crops
(e.g., corn-soybean-orchard) can be calculated from the full distribution of PCA values within
each watershed as the sum of their individual PCA values for each crop within that watershed.
Unlike in the crop pair PCA values, any number of individual PCA values can be combined to
give a use pattern specific PCA. These use pattern specific PCA values can be used to refine the
EDWC values in a similar fashion to the standard PCA values, i.e., multiplying the EDWC value by
the maximum national/HUC-02 regional use pattern specific PCA to give the refined EDWC
value.
As an alternative to the all-agriculture PCA for crops with no relevant standard PCA value (e.g.,
alfalfa, sugar beets, sorghum), a conservative estimate of the PCA for these non-standard crops
can be calculated as the difference between the all agriculture PCA (or all-agriculture + turf, if
the pesticide has residential turf or golf course applications) and the sum of the crop/group of
crops PCA values for each watershed (Figure 3). This is the maximum area that could be planted
with any crop that does not fall into standard PCA categories. Similarly, for non-agricultural
uses without a defined PCA, OPP can assume a PCA of 100% minus the all-agriculture + turf PCA
value, as that will provide a conservative estimate of all non-agricultural, non-turf uses in a
watershed. While these are likely an overestimation of the cropped area for any single one of
these use patterns (because it assigns all other acres to just one crop), the magnitude of the
overestimation is smaller than it would be to use the all-agriculture PCA.
12
-------
Corn
PCA=12.5%
Wheat
PCA= 12.5%
Non-Agricultural
Non-Standard
Crops
Turf
PCA=12.5%
All-Agriculture+Turf PCA = 50%
Other PCA = All-Agriculture+Turf PCA - Corn PCA - Wheat PCA-Turf PCA
= 50% -12.5% -12.5% -12.5% = 12.5%
Non-Agricultural PCA = 100% - All-Agriculture
=100% - 50% = 50%
Figure 3. Example calculation of non-standard and non-agricultural PCAs
The non-standard crops PCA can be combined with existing PCA values to calculate a use
pattern specific PCA value. For example, for a pesticide that can be applied to alfalfa, soybeans,
and cotton, the estimated alfalfa PCA could be combined with the existing soybean and cotton
PCA to calculate the alfalfa-soybean-cotton PCA for all watersheds and the maximum
national/HUC-02 regional PCAs can be used to refine the EDWCs in an identical fashion to a
standard PCA. It is important to note that, if there are multiple crops without standard PCA
values, the non-standard crops PCA should only be added once, as it represents the total area
that could be planted with all non-standard crops.
The use pattern specific and non-standard crop PCA approaches are suited for pesticides with a
broad range of use patterns spanning several crop/groups of crops with calculated PCAs and for
use patterns that do not have a calculated PCA. This method is not applicable to pesticides
whose use patterns are already covered by existing PCA values. If this method is not applicable
to the pesticide or, after applying the use pattern specific PCA value the national/regional
EDWC values still exceed the DWLOC, as a further refinement, the assessor can proceed to the
either the land cover class overlap analysis (Section 4.3.2) or the aggregate EDWC steps
(Section 4.3.3), as appropriate.
4.3.2. Full Distribution of Watershed PCAs with Land Cover Class Overlap
A more refined PCA analysis can use the full distribution of individual crop/groups of crops PCAs
(including use pattern specific and non-standard PCA) for all watersheds within a HUC-2 region.
In many cases the full distribution of PCA values (at either the regional or national scale) will
yield many CWS watersheds with low PCA values, including some with a PCA of 0 (i.e., regions
with no identified agricultural cropped or turf area). The assessor may use either the existing
13
-------
suite of PCA values (individually or grouped) or a new localized PCA as described in Section
4.3.1.
Next, OPP can back-calculate the PCA below which the EDWC would no longer exceed the
DWLOC, the "critical PCA". OPP will then overlay the critical PCA to the map. The critical PCA is
the PCA where the PCA-adjusted EDWC falls below the DWLOC (Equation 2).
Equation 2
Where
Critical PCA = DWLOC/EDWCn
Critical PCA =
EDWCmax =
DWLOC =
PCA value below which the PCA-adjusted EDWC is less than the DWLOC
Maximum EDWC for all use patterns for a given pesticide
Drinking water level of concern
For watersheds with PCA below the critical PCA no further refinement is needed. For those that
remain above the critical PCA the map of the watersheds with a use pattern specific PCA value
equal or greater than the critical PCA is overlaid onto the map of the land cover dataset for all
of the crops to determine the number of watersheds with potential exposure concerns that
overlap with regions where the pesticide could be applied. This step entails overlaying
watersheds with information from the USDA NASS agriculture of census (e.g. Ag Census) data
which can show where individual crops are grown. To do this the map of watershed and
associated PCAs values would be overlaid with the use area of the pesticide based on Ag
Census. The overlap analysis defines watersheds with the generalized PCA landcover class (e.g.
orchards) relative to areas defined by Ag Census where the pesticide specific use site (e.g.
apples) are not present. Any watershed that exceeds the DWLOC based on PCA but overlaps
completely with counties without acreage for the specific use (e.g. apples) can be excluded
from further refinement because without the use site present the pesticide may not be applied.
For example, Figure 4 illustrates the overlap between watersheds with a vegetable critical PCA
greater than 0.6% and counties with reported vegetable acreage based on survey data. Any
watershed with a PCA>Critical PCA that overlaps with the use sites has potential exposure
concerns.
14
-------
Figure 4. Critical PCA Overlap Analysis for Vegetable Land Cover Class
Vegetables
No acreage reported
Acreage reported
Watersheds
Watershed PCA < Critical PCA
Watershed PCA > Critical PCA
Watersheds with PCACritical PCA
overlapping with potential use sites
For pesticides that are applied to a single land cover class, if there is no overlap between any of
the crops and a watershed with a PCA > Critical PCA, then the PCA refined EDWCs would be
expected to be below the DWLOC and no further refinements are necessary. If there is overlap
between the watersheds that are above the critical PCA and with the areas where the crop is
grown, then more refinements may be needed and an aggregate EDWC analysis would be
included in the refinements (See Section 4.3.3).
This method is most applicable to pesticides with a small set of use patterns, pesticides used
primarily on minor crops, and pesticides applied to crops grown in a small set of areas within the
United States {e.g., avocados, pecans, sugarcane). If there is overlap between the labeled crops
and watersheds with a PCA value greater than the critical PCA, then further refine the EDWCs by
considering the individual contributions of each crop to the total EWDC (Section 4.3.3).
4.3.3. Aggregated EDWC Based on Individual Crop PCA Values.
Typically at lower tiers of refinement assessors have applied the maximum PCA nationally and
regionally to the maximum EDWC for any use in that region. For example, a DWA with uses on
corn, soybean and orchards can have three unique maximum EDWC and three unique PCA, As
15
-------
a first step a DWA can multiply the maximum of the three EDWC by the aggregated PCA for all
three uses providing a conservative estimate for that region. However, If further refinements
are still indicated after conducting use pattern specific PCA refinements and/or the critical PCA
overlap analysis, a third option for PCA refinements is to calculate an aggregate EDWC (sum
total considering all uses and their relative contribution) for each CWS watershed. This is done
by using the watershed scale crop (e.g., apples) or group of crops (e.g., orchards) specific PCA
values to adjust the EDWC values for each individual crop/group of crops EDWCs (e.g., apples
and cherries would both be adjusted with an orchard PCA). As an example, for a pesticide with
two uses in a given CWS watershed with different EDWC and use specific PCA, each EDWC is
adjusted by its unique PCA and then the two PCA adjusted EDWC are added together to yield a
single aggregated EDWC. This mathematically defined using Equation 3.
Equation 3 Aggregate EDWC(PestXiWatershedY) = Z"=i(EDWCt * PCAt)
Where:
Aggregate EDWC (pestx, watershedy) = Aggregate EDWC for pesticide X within watershed Y
i = Crop/group of crops with registered use pattern for pesticide X
Number of crops/groups of crops with calculated PCA values
grown in watershed Y
EDWCi = Maximum modeled EDWC for crop or group of crops i
PCAi = Percent Cropped Area for crop or group of crops i
When calculating the aggregate EDWC using Equation 3, if multiple crops use the non-standard
crop PCA value calculated using the procedure in Section 4.3.1, then only include the use
pattern with the highest EDWC in the summation to prevent double counting the cropped area.
This method does not consider the differences in the timing of the maximum EDWC values for
the different modeled use patterns (maxima are added even if they are not coincident in time).
If the aggregated EDWC calculated with Equation 3 exceeds the DWLOC, it is appropriate to
account for the potential variation in the timing of exposure from each pesticide use in a given
region (national or HUC2) by combining 30-year daily chemographs (plots that show the
changes in pesticide concentrations with time) generated by PWC modeling to examine the
temporal variability in the EDWC values. This aggregation of chemographs has been done in
other highly refined DWA including the Organophosphate Cumulative Assessments9. Figure 5
provides an example of a single chemograph, or time series, of daily EDWCs from a modeled
use. Multiple time series which have been adjusted by their individual PCA would then be
added together at each daily time step to create a single aggregated time series. Similar to how
a time series is rank ordered by year to derive a l-in-10 year EDWC (USEPA OPP, 2019b) the
9 https://www.regulations.gov/document?D=EPA-HQ-OPP-2006-0618-000
16
-------
aggregated time series can be post processed to yield an aggregated 1 in 10 year EDWC which is
compared to the DWLOC,
Chemograph of Results
4.00E+02 |
2.50E+02
3 2.00E+02
1.50E+02
1.00E+02
S.00E+01
7/19/1969
8/23/1970
9/27/1971
5/19/1973
12/31/1968
2/4/1970
4/14/1972
10/31/1972
12/5/1973
Date
corn — orchard vegetable —Combined EE Cs
Figure 5. Example of Combined Time Series Chemograph (5 year duration)
The aggregated EDWC values are then ranked across watersheds nationally or within a HUC-2
region and compared to the DWLOC to determine what percentile of watersheds still exceed
the DWLOC, If none of the aggregated EDWCs exceed the DWLOC, then no further refinements
are necessary. If there are still watersheds where the EDWC are greater than the DWLOC, then
the next step is to consider PCT refinements.
4.4. Summary
These new applications of existing PCA data utilize the full distribution of PCA values to
calculate more granular, use pattern specific PCA values to better estimate potential pesticide
exposure. The three new refinement options discussed here will allow OPP to move away from
relying on the default all-agriculture PCA for chemicals with a wide range of use patterns. The
first method uses the full set of PCA values to calculate use pattern specific PCA values tailored
to the specific use patterns of a given pesticide. Second, the critical PCA overlap analysis helps
identify if the pesticide is expected to be used in regions where PCA value is high enough to
trigger risk concerns. Finally, the aggregate EDWC calculation using both peak and time series of
EDWC provides a method to utilize the granular PCA information to calculate watershed-scale
aggregate EDWCs to determine the number of watersheds within a region where the EDWCs
are expected to be above and below the DWLOC. If there are still DWLOC exceedances after
17
-------
these refinements, then consider the application of the PCT refinements described in the
subsequent section.
5. APPLICATION OF STATE LEVEL PCT REFINEMENTS TO DWA
5.1. Purpose
If any watersheds have EDWCs that are greater than the DWLOC after considering the full
distribution of PCAs, OPP can further refine the EDWCs by considering the treated area in the
watershed to better describe the exposure potential. This section describes a method for
applying PCT refinements to DWAs.
5.2. Available Data and Integration into DWA
OPP can apply PCT refinements to EDWCs by calculating the sum of the Base Acres Treated
(BAT) for all crops/group of crops within the state, distributing that sum of BAT to the individual
watershed(s) where the EDWC>DWLOC (using the maximum/minimum/uniform methods
described in Section 3.3), and then calculating the Treated Acreage Scaling Percentage (TASP),
the ratio between the treated acreage and the total area of the watershed (Equation 4).
_ Base Acres Treated within X
Equation 4 TASP(wate„hed x) Total area of X
The TASP is specific to each watershed and is analogous to the PCA, but it incorporates both the
percent cropped area (i.e. use pattern specific or non-standard) and the percent crop treated as
a single multiplicative scaling factor that can be applied as a refinement to the aggregated
modeled EDWC values derived as described in Section 4.3.3. The TASP-refined EDWC is equal to
the modeled EDWC multiplied by the TASP.
Calculating the watershed-level BAT requires applicable state-level BAT(s)forthe registered use
patterns and the overlap between the potential treated area and the CWS watershed. In many
cases a CWS watershed will reside wholly within a state and a single state level BAT (either for
single uses or summed across multiple uses) may be all that is needed. However, some CWS
watersheds will span multiple states and the watershed-level BAT across those states will be
allocated to the individual CWS watershed fraction within each state using an allocation
method similar to that employed for the ESA approach (ultimately the allocation can be an
automated process). The aggregated PCT for all of the relevant use patterns of a pesticide is
multiplied by the total cropped area for the registered use patterns. For instance, in an example
with three uses (e.g., apples, wheat and corn) with each having a unique number of acres
treated and unique land cover derived PCA (orchard, wheat and corn), the assessor would add
the acres treated across the three uses and divide by the summed acres across the three
landcover types. The total BAT for the state is the sum of the individual BAT for each land cover
class. The Biological and Economic Analysis Division (BEAD) derives these PCT values from state
level usage data and records them in Summary Use and Usage Matrix (SUUM) reports. The
usage data come from a variety of sources, including the U.S. Department of Agriculture's
18
-------
National Agricultural Statistics Service (USDA NASS), Kynetec USA (i.e. private market survey
data , and the California Department of Pesticide Regulation (DPR) Pesticide Use Reporting
(PUR) program. These data are based on 5 years of usage data as compiled by BEAD to generate
state- and national-level estimates of the maximum, minimum, and average PCT value for a
given crop or set of crops. A description of the method for the PCT calculation (including
surrogacy assumptions) and the underlying data sources can be found in Appendix C and
Appendix D, respectively.
For each crop in the SUUM, OPP can calculate BAT by multiplying the reported PCT(s) by the
reported crop area grown (CAG). For crops that are not surveyed, OPP will make assumptions
about the PCT based on surrogate crop data (See Appendix C). OPP calculates the aggregated
land cover class PCT for a given land cover class (PCTtotj) for each state as the sum of the base
acres treated of all crops within the land cover class in the state (calculated from the individual
crop PCTs in the SUUM) divided by the total CAG for all crops within the land cover class grown
in the state (Equation 5).
_ Base acres treated _ PCTi*G{)
_ Cropped acres arown
Equation 5 11
Where:
PCTtotj = aggregated PCT (for land cover class j in state)
j = land cover class for registered use pattern (e.g., wheat, vegetables)
i = crop (within land cover class j) that is surveyed in state
n = number of crops (within land cover class j) with acres grown in state
PCTi = percent crop treated of crop i (from extended SUUM)
Gi = acres of crop i grown (in state) (from the Census of Agriculture)
The land cover classes represent potential use sites of a pesticide based on 18 classes that
represent aggregated agricultural/turf uses originally developed from the 2006 NLCD, the 2007
NASS Ag Census, and the 2008 USEPA Turf Layer as part of EPA's PCA work in support of refined
drinking water assessments (USEPA, 2014). These include 8 single crops/crop groups (e.g., corn,
orchards/vineyards, all-agriculture, residential turf), and 9 crop combinations (e.g., corn-wheat,
turf-soybeans, turf-all agriculture), in addition to a class for non-agricultural land. The complete
list can be found in Appendix C.
The approach described above for calculating aggregate BAT, PCTtotj, and TASP in a state
combines data that are at different spatial scales (land cover classes at 30-m pixel scale, Census
of Agriculture at county scale, and usage (i.e., PCTi, at state scale)). Due to the differences in
scale, the calculated PCTtotj values are only reliable at the state-scale level.
19
-------
5.3. Procedure
The following sections describe the general sequential process for applying watershed-scale
PCTtotj refinements:
1. Calculate the BAT for each land cover class within each state based on maximum,
minimum, and average PCT and the number of crop acres grown for crops in each land
cover class.
2. Allocate treated acreage to each CWS watershed.
3. Calculate the TASP (ratio of the treated area to the total watershed area) for each
watershed/PCT (min, max, avg)/distribution method (upper, lower, uniform)
combination (discussed below and in the documentation for the proposed ESA
Methods).
4. Calculate TASP-adjusted EDWCs
5. Compare TASP-adjusted EDWCs to DWLOC.
These steps are described in the subsequent sections. Ultimately, the process will be
automated.
Step 1. Calculate the Base Acres Treated for each land cover class within each state based on
maximum, minimum, and average PCT.
Due to possible differences in the CAG reported in the SUUM report used to calculate the PCTs,
and the CAG reported by the land cover class within each state, the usage data need to be
scaled proportionally to the CAG in the land cover class. OPP performs this in a three step
process: calculating the usage data BAT (Step la), converting that area to a state-level PCT
value (Step lb), and calculating the land cover class BAT from the usage data PCT value and
area of the land cover class GIS layer (Step lc). OPP calculates the BAT value for the maximum,
minimum, and average PCT to generate three BAT values for each land cover class within the
state. Each of these steps is described further below:
Step la. Calculate the aggregated usage data Base Acres Treated for a land cover class
within each state based on maximum, minimum, and average PCT.
OPP calculates the usage data BAT by multiplying the state-level PCT by the CAG from the
Census of Agriculture found in the SUUM report for each crop in the land cover class (Equation
4). For land cover classes with multiple crops, the land cover class BAT is then aggregated for
the individual crop BATs for the crops within the land cover class.
Step lb. Calculate the land cover class aggregate PCTs from the usage data Base Acres
Treated and the Crop Acres Grown.
OPP then uses the usage data BAT to calculate the aggregated usage data state-level PCT for
each individual land cover class, the aggregated BAT for the crops in the land cover class (e.g.,
20
-------
apples and wheat) is divided by the total CAG in the Census of Agriculture for those crops. OPP
uses this aggregated usage data state-level PCT to calculate the BAT for each land cover class
(e.g., apples to orchards landcover class and wheat to wheat landcover class).
Step lc. Calculate the BAT for each land cover class in the pesticide use pattern.
OPP calculates the BAT for each land cover class by multiplying the CAG for each relevant land
cover class by state-level aggregate land cover class PCT calculated in Step lb. This is the
maximum treated acreage for each land cover class that can be allocated to a single watershed
within the state.
For a given pesticide, there will be three treated acreage estimates of each land cover class,
based on the maximum, minimum, and average PCT values. A detailed description of the
methods to calculate the state-level treated acreage and the potential pesticide use sites in an
individual drinking water watershed can be found in Appendix C.
Step 2. Allocate treated acreage to each CWS watershed using upper, lower, and uniform
distribution methods.
To calculate the TASP for a watershed, it is necessary to calculate the number of treated acres
within the watershed use area. The use area where treated acres can be distributed is equal to
the overlap between the watershed and the relevant land cover classes for the pesticide. Since
there cannot be more treated acres within a watershed than there are cropped acres, the size
of the use area is capped at the PCA adjusted area of the watershed. For pesticides with
application to non-cropped areas, the total non-cropped area is calculated using the non-
agricultural PCA derived from the all-agriculture PCA (See Section 4.3.1)
While the total area of potential use sites in the watershed is known, the exact location of the
treated area within the state is not known. Although the exact sub-state location of usage is
unknown, because of the differences in scales, the usage of a pesticide can be limited to sub-
state areas by considering the finer scaled layer representing potential use sites within CWS
watersheds. To account for this unknown, OPP proposes several different methods for
distributing the treated acres to an individual watershed, described below: upper, lower, and
uniform distributions. These are bounding case assumptions designed to capture the
uncertainty in the distribution of the treated acres across the state.
Upper Distribution: This approach assumes that all the treated acres for a given land cover
class in a state can occur within a drinking water watershed boundary, up to the PCA adjusted
acreage of the watershed including non-agricultural uses (Figure 6). Treated acres are only
placed in counties within the drinking water watershed boundaries where there is at least 1
registered labeled use for the land cover class, as reported by the Census of Agriculture. If the
number of treated acres in a state is greater than number of acres of land cover class(es) that
overlap with the drinking water watershed boundaries, it is assumed that all acres within the
drinking water watershed boundaries are treated. If the number of treated acres in the state is
21
-------
less than the total area for the potential use sites within a drinking water watershed, then it is
assumed that all treated acres are located within the watershed.
Uniform Distribution: This approach assumes that the treated acres are distributed
proportionally to the aggregate land cover class PCT throughout the state(s) containing the
drinking water watersheds (Figure 7). The aggregated PCT is applied directly to the acres of the
land cover classes occurring within the drinking water watershed boundaries to calculate the
estimated treated acres. For uniform distribution, treated area = (acres within a drinking water
watershed boundary that overlaps with the land cover class) X (aggregate state-level PCT).
Lower Distribution: This approach assumes that all the treated acres for a given land cover
class occur outside the drinking water watershed boundary until there is no untreated area
outside the watershed, at which point any remaining treated acres are allocated to the drinking
water watershed (Figure 8). It is possible for the minimum distribution to allocate no treated
acreage to the watershed.
The treated acreage calculated for the maximum, minimum, and average PCT in Step 1 are
allocated to each watershed based on the three distribution methods to give nine separate
values that are utilized in the subsequent steps to calculate the TASP scaling factor.
Colorado
All treated
acres fall
within the
DWI
~
~
Agricultural
area (potential
watershed
~ ~
DWI
Watershed
Treated site
~
Example:
PCT = 10%
~
~
~
~ ~
~
~
~
~ ~
~
Figure 6. Conceptual Illustration of the "upper" distribution method
22
-------
Colorado
Treated
acres are
uniformly
distributed
throughout
state
Example:
PCT = 10%
Watershed
rP
d=h
Figure 7. Conceptual Illustration of the "uniform" distribution method
Colorado
Treated acres
concentrated
outside of
DWI
Watershed
Example:
PCT = 10%
%
4
~
~
~
~
m
~ ~
DWI
Watershed
~ n^_p ~
~
~
~
~ ~
~
~
~
Agricultural
area (potential
use site)
Treated site
Agricultural
area (potential
use site)
Treated site
Figure 8. Conceptual Illustration of the "lower" distribution method
Step 3. Calculate the Treated Area Scaling Percentage for each watershed for each
watershed/PCT/distribution method combination.
Once the treated area in a watershed has been allocated by the different distribution methods,
OPP calculates the TASP value by dividing the treated area within the watershed by the total
watershed area. OPP calculates a separate TASP for each combination of PCT (i.e., minimum,
maximum, and average) and distribution method (i.e., upper, lower, uniform), for a total of nine
23
-------
separate TASP values. For watersheds where the treated acreage is equal to the PCA-adjusted
watershed area, the TASP equals the PCA and the PCT refinement will not alter the EDWCs or
the potential risk picture.
OPP calculates both an aggregate TASP and a land cover class specific TASP from the usage
data. The aggregate TASP equals the total treated acreage for all land cover classes allocated to
the watershed divided by the watershed area; the land cover specific TASP equals the treated
acreage for a single land cover class divided by the watershed area. OPP uses the aggregate
TASP for the initial PCT refinement for Tier 3 EDWC calculations, while the land cover specific
TASP values further refines the EDWC by considering the contributions of the individual use
patterns to the overall EDWC, as described in Step 4.
Allocating treated acreage to watersheds that span multiple states is handled differently since
OPP calculates PCTs (and the resulting BAT values) on a state-by-state basis. The total treated
acreage that is allocated to the multistate watershed is the total of the treated acreage for each
separate state that overlaps with the watershed. The maximum acreage that can be allocated
from each state is equal to the area of the watershed that overlaps with that state, and the
maximum total treated acreage from all states is equal to the PCA adjusted area of the
watershed. Since the extent of land cover class overlap is a property of the watershed and not
specific to a pesticide, it can be calculated for all multi-state watersheds and incorporated into
an automated PCT process.
Step 4. Calculate TASP-adjusted EDWC
Once OPP calculates the TASP values for a given watershed, it can directly apply the values to
the individual use EDWCs. Like the application of watershed-scale PCA refinements, the TASP-
adjusted EDWC = (watershed EDWC) x TASP. For the initial evaluation, OPP uses the maximum
EDWC for all use patterns and the aggregate TASP to generate a conservative TASP-adjusted
EDWC value. If that value exceeds the DWLOC, OPP can further refine the EDWC by calculating
the aggregate EDWC as the sum of the individual TASP-adjusted EDWC values calculated from
the specific crop scenario EDWCs and the associated land cover specific TASP value using a
similar approach to calculating the aggregate PCA adjusted EDWC using Equation 3.
It is important to note that the TASP incorporates both PCA and PCT, so it should not be applied
to the PCA-adjusted EDWC value. As previously stated, if the treated area within the watershed
is equal to or greater than the PCA-adjusted area, then the TASP-adjusted EDWC will equal the
PCA-adjusted EDWC.
Step 5. Compare TASP-adjusted EDWCs to DWLOC
The TASP-adjusted EDWCs can be directly compared to DWLOC to determine the percent of
watersheds where there is still a potential risk. The TASP for a given watershed, and by
extension the refined EDWC, will vary depending on the method used to calculate the BAT in
Step 1 and to distribute the treated acres in Step 2 of the PCT method. With three different PCT
24
-------
values (max, min, average), and three different treated acreage distribution models (upper,
lower, uniform), there are nine potential EDWC values for each watershed, each representing
different degrees of conservativeness and assumptions regarding the use pattern and behavior
of the chemical. This will be discussed in Section 7.
5.4. Conclusions
This new method utilizes state-level PCT data to refine watershed-scale EDWC values in high
tier risk assessments (tier 3 and tier 4). It utilizes available state-level usage data to calculate
the total treated acreage within a state and then allocates the treated acreage to the potential
use sites within the individual watersheds using three different distributional models. The
watershed-scale treated acreage is used to calculate the TASP, a multiplicative scaling factor
that incorporates both the PCA and PCT, that is used to refine the EDWCs. A Case Study
demonstrating the application of the PCA and PCT methods described above can be found in
Appendix B.
6. METHOD EVALUATION
When fully employed and consistent with the DWA Framework (USEPA OPP, 2019b) modeled
EDWCs refined using the approaches outlined above will be evaluated by considering available
surface water monitoring data. USEPA OPP currently uses surface water monitoring data as a
"ground truth" against modeled estimates. The goal of this comparison is not to consider the
monitored values as the "true" exposure but to make sure that refinements do not provide
estimates of exposure below values observed in relevant monitoring data. This comparison
needs to account for the relevance, duration of exposure concern and uncertainty in both
modeled and monitored values but is intended to ensure the final values used for the dietary
assessment remain appropriate without underestimating expected exposures.
7. PROPOSED USE/DRINKING WATER ASSESSMENT IMPLICATIONS
OPP plans to incorporate the PCA and PCT refinements discussed in this paper to pesticides that
present potential human health risks after incorporating refinements such as crop- and region-
specific modeling scenarios, average pesticide use rates, and national- and regional-scale PCA
adjustments.
As applicable, OPP plans to incorporate the distributional PCA method presented in this paper
into Tier 3 analysis. The proposed approach to use the full suite of CWS PCA values builds on
OPP's current surface water modeling approach. Unlike the existing approach which uses a
maximum PCA either nationally or by HUC-2 watershed, the proposed approach will use the
unique PCA values for all 4800+ CWS watersheds. The proposed approach will instead identify
unique CWS watershed specific EDWCs which can be compared with the pesticide specific
DWLOC. Using the step wise refinement process laid out above, successive refinements should
be able to identify the percentage of CWS watersheds whose EDWC remain above the DWLOC.
25
-------
Using the full set of PCA data to calculate use pattern specific PCA values for a given pesticide
allows for more granular application of PCA data and can easily be applied to Tier 3 risk
assessments for chemicals with multiple use patterns.
As stated in Section 5, there are 9 potential combinations of PCT and treated acreage
distribution, which leads to up to 9 different sets of EDWC value. Selecting the most
appropriate value will depend on the specific properties and use patterns of the pesticide. For
example, the maximum PCT and "upper" distribution method are the most conservative
options and generate the upper bound concentration estimates of the 9 options, and therefore
can serve as a high-end screening tool. If a pesticide has a long-term (>1 year) exposure
concern, then the average PCT may more accurately represent the pesticide use level over the
relevant exposure window. If the usage data for the chemical indicates that it is applied to the
majority of the crops within a given region, then the "uniform" treated acreage distribution
may be more appropriate than the "upper" distribution. The OPP chemical team must examine
the available data and determine which set of parameters most accurately represents the use
pattern for a given chemical.
When interpreting application of PCT using the methods describe above consideration must be
given to the strengths, weaknesses and uncertainty associated with the available usage data
and methods to distribute state and national level usage data to watersheds within a state.
Similarly, for watersheds that span multiple states PCT data will be allocated from multiple
states to the watershed.
When considering the results of the options for PCT (state level maximum, average, and
minimum of 5 years of data) and distribution methods (upper, uniform, and lower) the strength
and weakness of each option within the matrix of options must be weighed. This WoE concept
can point to the most likely, or reasonable outcome from the analysis.
For example, in cases where the 5 years of PCT data yield consistent estimates across all usage
statistics (i.e., average, maximum, and minimum) adds confidence to the outcome when
considering all potential outcomes in the WoE approach. Conversely, in cases where there are
wide differences between the minimum, average and maximum PCT consideration of other
lines of evidence (e.g., longer term trends in usage) can provide further confidence in which
PCT value provides the strongest line of evidence. With distribution methods, the upper
distribution method provides the most conservative approach but the assumption that all
treated acres will be present in every watershed in a state is not likely. Similarly, the lower
distribution method where all treated acres are outside the watershed with overlap limited to
treated acres in excess of the area outside of the watershed is equally unlikely. The uniform
distribution method may provide a more reasonable assumption compared with the upper and
lower methods but leaves open the possibility of underestimation when pest pressure drives
usage into site specific regions within a state or watershed.
Ultimately, the decision whether a specific set of CWS watersheds remain above the DWLOC
after consideration of the full distribution of PCA and the matrix of options for considering PCT
26
-------
will have to weigh the strengths, weaknesses, and uncertainty associated with the underlying
data sets.
Including PCT data onto the PCA approach allows for more realistic application of usage data
which defines where the pesticide is actually being applied as opposed to where it could be
applied based on landcover data. Utilizing the approach developed for ESA assessments for PCT
represents a new high-end Tier 3 analysis for refining drinking water assessments and moving
from the theoretical to the actual. While significantly more resource-intensive than the full
distribution PCA analysis, it represents a new tool for refining high tier risk assessments.
Ultimately, application of the proposed PCA/PCT should allow for more realistic DWA, is more
fully consistent with the approaches described in the DWA Framework, and provides
opportunity to focus options for Tier 4 evaluations on specific areas.
8. REFERENCES
USEPA, 1999. https://archive.epa.gov/scipoly/sap/meetings/web/html/052599 mtg.html
Liess, M., Schulz, R., Liess, M. H. D., Rother, B., & Kreuzig, R. 1999. Determination of insecticide
contamination in agricultural headwater streams. Water Research, 33(1), 239-247.
N. Moriasi, D., W. Gitau, M., Pai, N., & Daggupati, P. 2015. Hydrologic and Water Quality
Models: Performance Measures and Evaluation Criteria. Transactions of the ASABE,
58(6), 1763-1785.
Rabiet, M., Margoum, C., Gouy, V., Carluer, N., & Coquery, M. 2010. Assessing pesticide
concentrations and fluxes in the stream of a small vineyard catchment - Effect of
sampling frequency. Environmental Pollution, 158(3), 737-748.
USEPA. 2000. Available Information on Assessing Pesticide Exposure in Food: A User's Guide.
EPA-HQ-OPP-2007-0780-0001. June 21, 2000. Health Effects Division, Office of Pesticide
Programs, Environmental Protection Agency.
USEPA. 2006. USEPA. Organophosphorus Cumulative Risk Assessment 2006 Update. July 31,
2006. Office of Pesticide Programs, USEPA.
USEPA. 2009. USEPA. Guidance on the Development, Evaluation, and Application of
Environmental Models. EPA/100/K-09/003. March 2009. Office of the Science Advisor,
USEPA.
USEPA. 2014. Development of Community Water System Drinking Water Intake Percent
Cropped Area Adjustment Factors for use in Drinking Water Exposure Assessments: 2014
Update. 9/9/14. Environmental Fate and Effects Division. Office of Chemical Safety and
Pollution Prevention. U.S. Environmental Protection Agency. Available at
https://www.epa.gov/pesticide-science-and-assessing-pesticide-risks/development-
community-water-system-drinking-water (Accessed December 8, 2016).
USEPA. 2019a. USEPA. DRAFT EPA Proposed Revised Method for National Level Endangered
Species Risk Assessment Process for Biological Evaluations of Pesticides. EPA-HQ-OPP-
2019-0185-0002. May 16, 2019. Environmental Fate and Effects Division, Office of
Pesticide Programs, Environmental Protection Agency.
27
-------
USEPA, 2019b. Framework for Conducting Pesticide Drinking Water Assessments for Surface
Water. Office of Pesticides Programs, U.S. Environmental Protection Agency. August 22,
2019. Available at https://www.regulations.gov/document?D=EPA-HQ-QPP-2019-0417-
0006
28
-------
Appendix A. Drinking Water Assessment Tiering Framework
The Framework for Conducting Pesticide Drinking Water Assessments for Surface Water
documents the standard tiered assessment approach for estimating pesticide concentrations in
surface water sources of drinking water. In summary, EPA utilizes a four-tiered DWA approach
to conserve resources by only using time-intensive, refined risk assessment methods requiring
more complex data input for pesticides failing [i.e., the estimated concentration of pesticide in
the source water exceeds the DWLOC] at lower tiers (Figure A.l). Most DWAs completed by
EPA are categorized as Tier 1 or Tier 2 assessments; a few DWAs have required a Tier 3 or Tier 4
level assessment.
As part of the tiered approach for conducting pesticide DWAs, EPA utilizes aquatic model
estimates and measured pesticide concentrations from surface water monitoring programs
when data are available. The temporal and spatial variability of pesticide concentrations in
surface water is typically not well characterized by periodic discrete samples, which represent
snapshots of pesticide occurrence in specific locations. Short-term peaks in pesticide
concentrations in surface water can occur because of seasonal or event-driven pesticide
applications (e.g., infestation or public-health hazard) and streamflow conditions (Liess et a!.,
1999; Rabiet et a!., 2010). Most monitoring data are collected on a non-daily basis and the
number of sites is often limited. As a result, EPA often uses pesticide concentrations estimated
from aquatic models as quantitative inputs to the human health dietary exposure model rather
than monitoring data because of the uncertainty that available monitoring programs capture
potentially high concentrations of pesticides that may have occurred on non-sampling days, or
in areas that were not sampled.
29
-------
LOW EFFORT, MINIMAL DATA
• Provides bounds on
potential p*st»c^e
concentration! an
drinking water
• Not intended to
present actual
drinking water
exposure
concentration*
• UHdt»fimnout
pest»c*des having How
risk potential, or
identify pesticides
with potential high
rak
INTENSE EFFORT, ALL AVAILABLE DATA
driving risk
Scoping
¦ Point estimate
provided as upper
bound estimated
drinking water
concentrations
(fOWCs}
£DWC covers
ma«nsum labeled
rates for afl uses and
representative use
sites
EDWC der^ based
on bghlv vulnerable
pestfcid* application
locations from models
¦ Identify pesticides
with a low potential
risk tor CWSs on a
national or regional
basis
¦ Distributions) of
£DWCrt& at Tier 2 on a
regional basrt
* CeftskJentypaiuw
rates and practices
* Refine EPWG for
pesticides h*nng a
high paternal risk on
a national or regional
basis
* Utilizes bath model
estimated
concentrations as
well as iwonftortng
data
* Further etntnates
pesticides where
exposure in drinking
water is not a
concern for CWSs
Provides time sertes
of lOWCs for typical
use on a regional or
svib-reg*onal CWSs
basis
KsgJiert temporal and
spatial resolution
with watershed scafie-
based EOWCs
Uuiitet both model
estimated
concentrations as
well as monitoring
data
1 Oefkrves regions or
areas of the country
with a hgh degree of
confidence where the
use of a pestkide
may pose risk to
human health
Tier 1 Tier 2 Tier 3 Tier 4
Methods and Resources:
• Soeeflftlgl
calculate*
owioqsji
Label use reports
Wei hods *nd Additional
Mwwret*:
* Peslscide m Water
CakUtator |PWO
• Pe*fcyson
« Dietary frpos-ure
Evaluate* Model - food
Commodity Intake
Database C*l use information
(rates, application dates,
methods)
Methods and Additional
Resources;
* Spatial Aquatic Model
ISAM} (under
rfewlopmentl
* 5&AWAVE-QEK
* Warmhed regression!)
laddrtiona' work
* pniicide specifkSftfs
¦ DCEM'FOO/CaAende-K
* Typical we information
[rates, applications
daces, methods}
SCREENING LEVEL
HIGHIV REFINED & COMPLEX, MOST REALISTIC
Figure A.l Tiered Drinking Water Assessment Framework [from the Framework for
Conducting Pesticide Drinking Water Assessments for Surface Water
In a DWA, EPA estimates pesticide concentrations in surface water sources resulting from the
labeled uses, which may include a single use site {e.g., corn) or multiple use sites {e.g., corn,
cotton, and turf). While the aquatic model inputs such as soil and weather may be localized
{e.g., limited to field, county, or statewide analyses), the resulting estimated concentrations a
typically used to cover, or be protective of, a larger geographical scale (regional or national),
30
-------
encompassing a wider range of environmental conditions, use patterns, and agricultural
practices and thus a wider range of potential pesticide concentrations.
To complete the work for DWAs, EPA reviews pesticide specific environmental fate and
transport data to derive environmental persistence (i.e., half-lives) and mobility inputs used in
aquatic models. These data are considered along with the physical-chemical properties (e.g.,
vapor pressure and solubility) of the pesticide to complete aquatic modeling. EPA also evaluates
existing water monitoring data for the pesticide if available. However, rarely are monitoring
data quantitatively used to estimate pesticide concentrations in drinking water. Stakeholders
have expressed concern with over-reliance on models and underutilization of monitoring data.
31
-------
Appendix B. Case Study
See Attached file.
32
-------
Appendix C. Determination of Overlap of Treated Area and Drinking Water Watersheds
Located in the 48 Conterminous States
The extent of overlap of any pesticide's potentially treated area and a given drinking water
watershed integrates information on potential use sites and usage data. To address
uncertainties associated with how treated acres may be distributed within a state (relative to a
drinking water watershed), and the magnitude of usage on any given year, approaches are
employed to represent upper, uniform, and lower estimates of potential overlap. These
different estimates are considered in a weight of evidence when deciding whether use of a
pesticide is likely or not likely to exceed a DWLOC in any given group of drinking water
watershed. This summary describes the approach for determining the extent of overlap
between drinking water watershed and treated acres.
Potential Use Sites
The land cover classes representing potential use sites in the 48 conterminous states and
District of Columbia (CONUS) are represented by selected crops (e.g., corn), groups of crops
(e.g., vegetable, orchards), turf, and general non-agricultural land. Potential use sites of a
pesticide are represented by 18 land cover classes that represent aggregated uses originally
represented as part of EPA's PCA work in support of refined drinking water assessments
(USEPA, 2014). The CWS PCAs were primarily developed to better understand pesticide
exposure in drinking water resulting from agricultural uses. In terms of non-agricultural uses,
PCAs were developed for residential and golf course turf.
Although PCAs for other non-ag uses are not currently available in USEPA 2014, it is possible to
estimate the maximum extent of these use sites using the agricultural and turf PCAs. The upper
limit of the potential non-agricultural/non-turf PCA consists of all areas not used for agriculture
or turf, i.e., Non-Agricultural PCA = 100% - (all ag-turf PCA). This would capture all potential use
sites not covered by other PCA values and provide a conservative estimate of the non-
agricultural use area. This method would likely overestimate the non-agricultural PCA since it
does not distinguish between different non-agricultural use sites (e.g., rights-of-way,
residential, commercial) and therefore would lead to upper end EDWC estimates that are
appropriate for human health risk assessment.
The current DWA land cover classes are presented below.
Corn
Cotton
Orchard/Vineyards
Soybean
Vegetable
Wheat
All-agriculture
Residential Turf
33
-------
Corn-Wheat
Soybean-Wheat
Turf-Corn
Turf-Orchard
Turf-Soybean
Turf-Vegetable
Turf-Wheat
Vegetable-Orchard
Turf-All Agriculture
Non-agricultural lands
Applying Usage Data to the PCA Land Cover Classes
The goal of this approach is to determine the area within each state that is treated one or more
times with a given pesticide (referred to as "Base Acres Treated" or BAT). This is accomplished
by combining data representing the potential use sites, including all land cover classes and
acres grown from the 2006 National Land Cover Dataset, the 2007 National Agricultural
Statistics Service Agricultural Census, and the 2008 USEPA Turf Layer and available usage
data.10
For agricultural uses (and some non-agricultural uses) of a pesticide, usage data are available to
quantify the PCT. This can then be used to adjust the extent of the potential use area
overlapping with a drinking water watershed's boundaries to represent the potential extent of
overlap that is directly treated with a pesticide. PCT data are available for specific crops and
states. A pesticide's usage data are summarized in the Science Information and Analysis Branch
(SIAB) Use and Usage Matrix (SUUM) which is provided to the chemical team by BEAD. The
pesticide's SUUM reports PCT data based on usage that occurred for a given 5-year range.
Three statistics for PCT are reported for each state and crop combination (where states and
crops are surveyed): average, minimum and maximum annual. The method discussed below is
applied separately to the average, minimum and maximum annual PCT data in order to quantify
the overlap of drinking water watershed boundaries and exposure areas, while accounting for
variability in usage over time.
Usage data are applied to the 18 land cover classes discussed above. For categories
represented by a single crop in both the SUUMs and land cover classes (i.e., corn, cotton,
soybean and wheat), the available PCT data for a given state are applied directly to the acres of
the land cover class in that state to calculate the acres treated (acres treated = acres grown x
PCT). If the PCT is not available for a specific state/crop combination, a surrogate PCT is applied
using the process described in the next section.
10 Full details available at https://www.epa.gov/sites/production/files/2015-
07/documents/development and use of community water svstem.pdf
34
-------
For those categories representing multiple crops in either the SUUMs or land cover classes
(e.g., corn and wheat), an aggregated PCT is calculated. In order to calculate the aggregated
PCT, the acres grown and PCT for each crop in the category is needed by state. Both pieces of
information are found in the SUUM for each state/crop combination with reported usage.
Acreage in the SUUM can come from a variety of sources, including, but not limited to,
agricultural market research data, USDA's National Agricultural Statistics Service (NASS),
California's Pesticide Use Reporting (PUR). If the state/crop combination does not have
reported usage, the information in the SUUM is supplemented with data from the 201211
Census of Agriculture (USDA-NASS, 2012). The Census of Agriculture is also used to account for
crop group categories in the land cover classes that are not registered (e.g., vegetables,
orchards). In this situation, information is not provided for these crops in the SUUM, but the
crops would be included in the land cover class. This process results in three scenarios:
The land cover class crop/state combination is found in the SUUM, and the acres grown and
crop specific PCT from the SUUM are used directly. The land cover class crop is registered and
the crop/state combination is surveyed by one or more usage sources, but no state specific
usage information reported in the SUUM. In this scenario usage is assumed to be de minimus.
Thus, the acres grown for the state are extracted from the SUUM and the acres grown for the
state are extracted from Census of Agriculture and a PCT of 0 is used in the calculation of the
aggregated PCT. The land cover class crop is registered and the crop/state combination is not
surveyed by any applicable usage sources, thus there is no state specific usage information
reported in the SUUM. In this scenario the acres grown for the state are extracted from the
Census of Agriculture and a surrogate for the crop specific PCT is assigned using the method
described in the next section. The crop is not a registered use. In this scenario, the acres grown
for the state are extracted from Census of Agriculture and a PCT of 0 is used in the calculation
of the aggregated PCT. This is done to account for crops found in the land cover class that are
not registered.
At the end of this process all state/crop combinations found in the Census of Agriculture are
accounted for, with acres grown and a crop specific PCT as described in Section 5.2. For
state/crops combinations with usage data, the acres grown are extracted from the SUUM; for
state/crops combination without usage the acres grown are extracted from the Census of
Agriculture.
Acres treated for land cover classes with multiple crops are calculated by multiplying this
aggregated PCT by the area of the land cover class for the state. The total area of the land cover
class for the state only includes those counties with at least 1 registered use as reported in the
Census of Agriculture. The land cover classes often include multiple crops, and some of the
crops included may not be registered. The Census of Agriculture is used to identify counties
where all crops for a land cover class are reported as not grown. These counties are excluded
from the totals prior to calculating the treated acres.
35
-------
Applying Surrogate Usage Data
Some uses are not surveyed for usage at all and some uses are only surveyed for usage in some
states. For crops that are included in aggregated land cover classes, but for which no survey
usage data exists, usage data from the same state for other crops in the land cover class will be
used as surrogates. If no data for crops within a PCA are available in a state, surrogate PCT will
be applied using data available for the same crop or land cover class but a different state. If a
land cover class has no usage data for any state, the highest available PCT from all state-crop
combinations will be used. The decision tree below (Figure 1) outlines the approach for
determining which data will be used as surrogates.
The surrogacy approach is designed to use the best available data to identify the likely extent of
treated area when usage data are not available for a given crop. Surrogate data are ideally
assigned using crops within the same land cover class and then using data from the same crop
but different spatial location. If the first two options are not possible, then a conservative
approach is employed where the greatest extent of usage on any crop-state combination is
used as the surrogate. Use of surrogate data represents an uncertainty. In cases where a
drinking water watershed has potential risk concerns, a weight of evidence analysis will be
conducted prior to deciding that EDWC in a watershed is likely below the DWLOC. In this weight
of evidence analysis, the impact of the surrogacy assumptions on the overlap analysis will be
considered. This will be done by quantifying the extent of overlap when all crops with no PCT
data are assumed to have no treated acres (i.e., PCT is assumed to be 0). For state/crop
combination where a surrogate PCT was applied, the surrogate will also be assumed to be 0, an
aggregate PCT will be re-calculated for this assumption, and the overlap adjusted for
comparison. The weight of evidence will consider the difference in the extent of overlap of the
two approaches in order to consider whether the surrogacy assumptions impact the likelihood
that an individual of a listed drinking water watershed will be exposed.
36
-------
Use crop specific
PCT for assessed
state
Yes
No
Use highest PCT
of all uses within
PCL and state
Yes
No
Use highest PCT
for crop from any
of the 48 states
Yes
No
Use highest PCT
from all crops
and states
Is use surveyed in other states?
Is crop-specific PCT available for assessed state?
Is a PCT available for a use within the same PCL
and state?
Figure 1. Decision framework for applying surrogate usage data.
Calculation of Extent of Overlap of Drinking Water Watershed and Treated Acres
Once the total treated area in the state has been calculated, the treated acreage can be
distributed to the individual watershed based on the upper, lower, and uniform distribution
approaches described in Section 5 to determine the treated acreage within a given watershed.
37
-------
Appendix D. Usage Data Sources
The usage data utilized to calculate the state-level PCT values come from the following public
and proprietary sources.
Kynetec USA, Inc. 2019. "The AgroTrak® Study from Kynetec USA, Inc." Database Subset: 1998-
2018 - proprietary pesticide usage data from 1998 to 2018 for historical use trend and 2013 to
2017 for current usage estimates. These data are collected and sold by a private market
research firm. The data are collected by annual surveys of agricultural users in the continental
United States and provides pesticide usage data for about 60 crops, including both specialty
and row crops. The survey design targets at least 80 percent of US acreage/production of the
surveyed commodities. Survey methodology provides statistically valid results, typically at the
state-level.
United States Department of Agriculture's National Agricultural Statistics Service (NASS) -
publicly available pesticide usage data from 2013 to 2017. NASS data are based on surveys that
focus on the top-producing states that together account for the majority of U.S. acres or
production of the surveyed commodity. NASS survey design targets a minimum of 80 percent of
the acreage/production for every fruit, vegetable, and field crop surveyed. Operation level data
are combined during summary and, pending compliance with disclosure rules, published at the
state and national levels. NASS does not collect data annually for each crop, but surveys for
various commodities on a rotating schedule.
California Department of Pesticide Regulation (CADPR) Pesticide Use Reporting (PUR) - publicly
available pesticide usage data for 2012 to 2016. The PUR database contains detailed records
and summaries of agricultural applications of pesticides on crops based on application permits.
All agricultural growers must submit their production agricultural pesticide use reports monthly
and pest control businesses must submit pesticide use reports within 7 days after application.
As such, CADPR data is a census of all usage rather than a survey. The Pesticide Use Summary
reports are published annually.
California Agricultural Statistics Review (CASR) - publicly available California crop production
data for 2012-2016. CASR data are used as the primary source for CAG data when calculating
PCT estimates for California crops and based on acres planted.
California County Agricultural Commissioners' Report (CCACR) - publicly available California
crop production data for 2012-2016. CCACR data are used as a secondary source to calculate
California crop PCT estimates in instances where CASR data are not available. PCT estimates
using CCACR data are based on acres harvested.
Non-Agricultural Market Research Data (NMRD) - proprietary data source that provides market
research data for agrochemicals/specialty pesticides for various market sectors, including
professional turf and ornamental plants, professional pest control, consumer pesticides, and
vegetation management. Market reports reflect usage by class/market segment and chemical
38
-------
and are based on sales information (manufacturer and retail) and end-user surveys. Study dates
vary by market sector.
39
------- |