Minutes of the March 20, 2012 
Exposure Modeling Public Meeting
Office of Pesticide Programs (OPP) 4[th] Floor Conference Center
2777 South Crystal Drive  -  Potomac Yard Two, North
Arlington, VA  22202	




Welcome and Introductions

The USEPA/OPP/EFED hosts biannual Exposure Modeling Public Meetings (EMPMs), which provide a forum for exchanging information between OPP/EFED and stakeholders with similar technical expertise on current issues related to pesticide exposure modeling in support of risk assessment.  Katrina White and Dirk Young chaired the March 20, 2012 meeting in the capacity of co-chairs of the Water Quality Technical Team.  The minutes and presentation slides for the March 20, 2012 EMPM are posted in the docket (EPA-HQ-OPP-2009-0879), which can be accessed at http://www.regulations.gov/.   The docket ID # is reported on the EMPM webpage: http://www.epa.gov/oppefed1/models/water/empm_top.htm

Brief Updates:
* EFED has updated the HUC8 percert cropped area (PCA) adjustment factors.  They can be found in the  guidance document titled, "Development and Use of Percent Cropped Area and Percent Turf Area Adjustment Factors in Drinking Water Exposure Assessments: 2012 Update" available at http://www.epa.gov/oppefed1/models/water/pca_adjustment_dwa.html
* Guidance for Evaluating and Calculating Degradation Kinetics in Environmental Media is now available at: http://www.epa.gov/oppfead1/international/naftatwg/guidance/degradation-kin.pdf

Presentations Abstracts and Notes:

1. NAFTA Guidance for Calculating Degradation Kinetics in Environmental Media 
   Dirk Young[1], Ian Kennedy[2]
   [1]Environmental Fate and Effects Division.  United States Environmental Protection Agency
   [2]Health Canada
   
The USEPA and Health Canada (NAFTA partners) have developed guidance for calculating degradation rates for transport model inputs (i.e., PRZM and EXAMS).  The NAFTA guidance specifies the conditions under which a first-order model can be used to represent pesticide degradation data in NAFTA assessments and provides a procedure to determine a representative rate when the data deviate from a first-order concept. The NAFTA guidance is based on the premise that there is no "correct" transformation of non-first-order data into first-order model parameters; therefore, any parameter transformation will require a subjective decision based on an agreed-upon level of protectiveness. The emphasis of the NAFTA procedure is therefore on standardization of a process rather than on statistics of determining "best fit" models.  In this regard the NAFTA group has arrived at a comparatively simple method that has preserved some standard assumptions used by FOCUS such as use of a DT90 as a reference point.

Notes:
In his presentation, Dirk Young explained the NAFTA guidance for calculating degradation kinetics that USEPA and Health Canada (NAFTA partners) have developed for transport model inputs. Dirk explained how EPA and Health Canada used data from aerobic soil metabolism studies to describe the challenges when the data did not follow first-order kinetics. Therefore, the main idea behind the NAFTA project was to find a standardization process to determine a representative single first-order rate for modeling when the data deviate from first-order degradation.  

The first step in the NAFTA guidance is to determine whether the data is significantly non-first order. If data are non-first order, then the representative half-life using the indeterminate order rate equation (IORE) and the double first order in parallel (DFOP) equation (this is the DFOP slow rate) is obtained. A representative single first order half-life using the IORE equation is called the TIORE.  The lower of TIORE or the DFOP slow rate is used as a representative single first-order model input value.  The NAFTA project is currently completed; however, the implementation plan and technical evaluation of the R program and package to calculate kinetics is yet to be completed. 

Questions:
   1. What is DFOP? Is it two-compartment decay? DFOP stands for Double First-Order in Parallel and it is a two-compartment decay.
   2. Are you combining studies to do this data analysis? Because it is conservative as it takes the 90[th] percentile confidence. No, we are not combining studies. You can have different soil studies and some of them will follow Single First-Order (SFO), while others will not follow SFO. Therefore, you will be using this approach for soils that do not follow SFO.

   
2. Update: Implementation of PRZM-GW
   Reuben Baris, Michael Barrett, Rochelle F. H. Bohaty, Marietta Echeverria, Ian Kennedy, Greg Malis, Dirk Young, and James Wolf
   Environmental Fate and Effects Division.  United States Environmental Protection Agency, Arlington, VA
   
Estimation of pesticide concentrations in groundwater is one of the considerations as part of the exposure assessment in the pesticide registration process. For this reason, Canada and the United States combined efforts as part of the North American Free Trade Agreement (NAFTA) to develop a harmonized groundwater modeling protocol. This included the development of a common conceptual groundwater modeling scenario for regulatory purposes designed to be protective of even the most vulnerable drinking water supplies. The Pesticide Root Zone Model (PRZM) was selected as the NAFTA regulatory tool. Evaluation of PRZM was completed by comparing simulated pesticide concentrations to available targeted and non-targeted monitoring data. For the majority of chemicals tested, PRZM-predicted pesticide concentrations represent upper bound estimates of exposure in groundwater when conservative input parameters are used (e.g., maximum application rates, annually repeated applications, conservative degradation assumptions).  There are some pesticide detections in monitoring data that are not captured by PRZM model estimates. This may be a result of physical processes that are not accounted for in PRZM (e.g., preferential flow or macroparticle transport). With site specific adjustments to the PRZM input values estimated pesticide concentrations compare very well to monitoring data (the same order or within an order of magnitude). The evaluation demonstrates that PRZM is a versatile assessment tool that can be used both as a screening tool and a refined site-specific tool in risk assessment. 

Notes: 

The presentation began by outlining the guidance documents under which PRZM-GW and all EPA models are implemented: Guidance on the Development, Evaluation, and Application of Environmental Models produced by the Council for Regulatory Environmental Modeling (CREM) and Guidance for Quality Assurance Project Plans for Modeling produced by the Office of Environmental Information (OEI).  A new guidance, Guidance on Quality Assurance Project Plans, combines the two aforementioned guidance documents.  It was then explained how these guidances apply to the project.

The conceptual model is the same as the previous one presented to the EMPM, with the exception of a variable well screen length to replace a one-meter standard well screen length.  Of the 19 models that were considered,  3 models were evaluated  (PRZM, LEACHM, and PEARL), and  PRZM was  eventually selected.  The data quality objectives for the model specify  that non-targeted monitoring data should have lower values than the estimated value, and targeted monitoring results should have similar concentrations to simulations.  Some observed concentrations in non-targeted studies were above simulated concentrations; however, this may be attributed to  preferential flow and other unaccounted factors.  In the National Alachlor Well-Water Survey, PRZM provided conservative estimates for wells.  Even when the hydrolysis half-life was estimated to be 4 years, large differences in EDWCs could be seen between a long hydrolysis half-life and stability to hydrolysis since PRZM-GW simulations run for 30 to 100 years.  Label rates and soil degradation differences also contribute significantly to changing EDWCs.   The model is still in the evaluation phase within EFED, and guidance for consistent use among risk assessors is being developed concurrently.

Q: From Stuart Cohen: How do PRZM-GW concentrations compare to SCIGROW concentrations?
A: SCIGROW only models one year, whereas PRZM-GW models 30 to 100 years so results are not directly comparable.  Also, SCIGROW is regression based whereas PRZM-GW is mechanistic.  However, there are comparisons that will be released with other model documentation.

Q: For a hydrolytically stable compound, how does PRZM-GW compare to monitoring and SCIGROW?
A: The models are not directly comparable with respect to hydrolysis alone because SCIGROW does not include hydrolysis as an input.  Long term simulations like PRZM-GW need accurate hydrolysis estimates.  Chemicals with pH- dependent hydrolysis half-lives can have high monitoring detections in sites highly acidic or highly alkaline because extreme pH values are not modeled.  2,4-D is an example of this.

Q: PMRA: How is soil hydrolysis accounted for relative to aerobic soil metabolism?
A: Degradation from aerobic soil metabolism is restricted to the top one meter in the soil profile.  Degradation from hydrolysis is active in the full soil profile.
Note from Cal DPR: With respect to monitoring data, California Department of Pesticide Regulation has not found alachlor parent in samples, only its associated degradates.  Additionally, CalDPR is finding concentrations 3-4X higher in run-off areas as opposed to groundwater.  

Q: Cal DPR: The diuron example shows that the application rate is important.  How was the application rate determined? It should be noted that diuron is used as a desiccant at 2 lb/acre.
A: The application rate was calculated over the full 25 mi[2] application area.

Q: Syngenta: How is over-prediction accounted for in the model evaluation?
A: EFED is working with BEAD to better characterize use data and we have a number of other refinements that can be made.

3. Measuring and Estimating Concentrations in Drinking Water:  A Historical Perspective
   Russell L. Jones
   BayerCrop Science
   
After the adoption of the Food Quality Protection Act in 1996, risk assessments for registration had to include the contribution of residues in drinking water to dietary assessments.  This triggered work modeling and monitoring work by both EPA and industry to determine satisfactory methods for estimating and measuring concentrations in drinking water from both ground and surface water.  This presentation reviews the developments in modeling and monitoring of drinking water from surface water over the past 15 years, including identification of the most sensitive drinking water intakes, SWMI, WARP, and the Index Reservoir as well as the evolution in surface water monitoring studies.  The results of the past research will be used to suggest potential approaches to obtaining appropriate estimates of drinking water based on either modeling or monitoring.

Notes: 
The Food Quality Protection Act of 1996 required risk assessments for registration to include the contribution of residues in drinking water to dietary assessments.  This triggered modeling and monitoring activities by both EPA and industry to determine satisfactory methods for estimating and measuring concentrations in drinking water from both ground and surface water. The focus of this talk is on surface water assessments, which historically have been the driving exposure pathway for most drinking water risk assessments. Improvements made over the years include the addition of the index reservoir and percent crop area (PCA) factors in drinking water exposure assessments. The assessment methods have not changed significantly in the last 10 years. Often the predicted concentrations are much larger than monitored results, which "demonstrates the overly conservative nature of drinking water assessments." 

Examples were given from aldicarb, bromoxynil, carbaryl, ethoprophos, oxadiazon, and iprodione exposure assessments, using OPP models versus monitoring observations in raw and finished drinking water samples. Usually model estimates were orders of magnitude greater than measured concentrations*. (EPA Comment: However, it is true that it is often easy to miss peak concentrations at a site by a significant factor with all but the most intense monitoring programs.) Current OPP models for rice and cranberries may be especially conservative. One major factor increasing the likelihood of overestimation of drinking water concentrations is that the highest use intensities are usually in small watersheds where the small streams are not directly used for drinking water. Regression models such as the Surface-Water Mobility Index (SWMI) and Watershed Regressions for Pesticides (WARP) have advanced our ability to predict moderate- and long-term exposure levels by identifying input factors responsible for most of the variability in observed data  -  with emphasis on model inputs that can be generated from commonly available data sources in the US. EPA was involved in the early development of WARP but was unable to continue supporting its development for drinking water assessments "due to non-technical reasons."

The accuracy of drinking water assessments can be improved greatly by obtaining data to refine inputs to better reflect real world exposure scenarios. A major area for refinement is in the area of use intensity where improved representation of percent crop area, percent crop treated, and typical application rates is needed; all have been used by OPP in cumulative risk assessments. Pesticide sales data is often a good surrogate source for refining usage inputs when direct localized usage information is not available. Acute exposure estimation from monitoring data can also be improved by techniques like the use of "bias factors."  One should try to determine whether the underestimation bias from insufficiently frequent sampling to capture peak exposure is as important as the overestimation from conservative assumptions in modeling. An example of this was illustrated from the carbaryl data.  The sampling schedule used for carbaryl should have been more than adequate to accurately determine the 70[th] and 90[th] percentile concentrations.  Since the modeling was overestimating these concentrations by 1-3 orders of magnitude, then one could not attribute the difference between the modeled and measured concentrations at higher percentiles, but only to missing the peaks.

* Examples of acute and annual mean exposure modeling estimates and targeted monitoring result ratios:
Aldicarb: 1.5 x to 30x
Bromoxynil: 35x
Carbaryl: 100x to 2000x
Ethoprophos: 100x to 1000x
Oxadiazon: 300x
Iprodione: 600x
3,5-DCA: 5000x
[Above were the ratios of model predicted acute concentrations to observed concentrations, chronic (annual mean) concentrations were also compared.]

4. Improved Characterization of the Temporal and Spatial Variability of Potential Surface Water Drinking Water Exposure by Using Environmental and Historic Monitoring Databases
   Paul Hendley[1], Wenlin Chen[1], Robert Joseph[1],  Ron Williams[1], Chris Harbourt[2] & Paul Miller[2]
   [1]Syngenta Crop Protection LLC, Greensboro, NC
   [2]Waterborne Environmental Inc., Champaign, IL
   
The extensive atrazine surface water monitoring database totals over 300,000 samples including ~120,000 drinking water samples from high frequency seasonal monitoring programs covering more than 15 years and >210 Community Water Systems (CWS). This wealth of frequent monitoring data sampled across a wide range of years and sites allows for robust conclusions to be drawn with very high statistical certainty regarding the highest centiles of potential atrazine exposures. To complement the monitoring data, site-specific details have also been accumulated on CWS watersheds, source waters and water handling procedures; this information has been used to better understand the expected temporal concentration patterns at individual CWS. We will outline how these and related data (e.g. daily flow and chemical monitoring data from Heidelberg University) have been used to examine approaches for optimizing sampling strategies for assessing potential drinking water concentrations in surface water. Scenarios with different sampling frequencies have been simulated and the results have been compared to measured data for each year and site. This talk will demonstrate how lower frequency monitoring data can be used to estimate potential maximum shorter duration concentrations.

Notes:
For larger sampling sizes (e.g.,> 7000), there is a greater confidence in capturing the highest centile concentration (99[th]%-tile) compared to smaller sampling sizes.

Two considerations/enhancements to the database for capturing peaks in monitoring (i.e., did you miss peaks?)
1) biased factors and error ratios
2) supplement monitoring data with model estimates (PRZM-hybrid) 

Because the atrazine dataset is so large, it may be possible to apply conclusions from these studies to the general monitoring study design (e.g., baseline for using year-to-year variation).  

Questions:
1) What is the effect and applicability of chemical specific biased factors? Can they be applied across chemicals?
- Potentially, but this would need some additional research.
2) How much data are available across water systems?
- 150 or greater water systems. 


5. Estimating Upper Centile Pesticide Concentrations and Sample Size Requirement
Wenlin Chen[1], Paul Mosquin[2]
[1]Syngenta Crop Protection, LLC
 [2]RTI International

A survey sampling approach is presented for estimating upper centiles of aggregate distributions of surface water pesticide concentrations obtained from monitoring with large sample sizes but variable sampling frequency. It is applied to three atrazine monitoring programs of Community Water Systems (CWS) that used surface water as their drinking water source: 1. the nationwide Safe Drinking Water Act (SDWA) data, 2.the Syngenta Voluntary Monitoring Program (VMP), and 3. the Atrazine Monitoring Program (AMP). Requisite sample sizes are determined using statistical tolerance limits, relative standard error, and the Woodruff interval sample size criterion. These analyses provide 99.9% confidence that the existing data include the 99.9th centile atrazine concentration for CWS raw and finished water in the Midwest atrazine high-use areas and in the nationwide SDWA dataset, assuming no bias due to nonrandom selection of CWS and monitored days during these years. The general validity of these approaches is established by a simulation that shows estimates to be close to target quantities for weights based on sampling probabilities or time intervals between samples. 

Notes:
Additional detail on the analysis is included in Mosquin et al., 2012 (JEQ publication)

There are two approaches to focus monitoring design (precision and sample size) on confidence in capturing upper centile concentration and relative standard error (Slides 12 and 13, respectively)

Question:
1) How is right-handed bias handled in Woodruff; how is it used? What is the Woodruff method?
-Woodruff is the skewness of a binomial approximation
-Enough samples as defined by Woodruff allow you to have good coverage at X-centiles, and at x-sample size.

   
6. Estimation of Upper Percentiles of Chlorpyrifos Surface Water Concentration from Yearly Monitoring Program Data
   Paul L. Mosquin[1], G. Gordon Brown[1], Roy W. Whitmore[1], Nick Poletika[2]
   [1]RTI International, Department of Statistics & Epidemiology, Research Triangle Park, NC, 27709, United States 
   [2]Dow AgroSciences LLC, Department of Field Exposure and Effects, Indianapolis, IN, 46268, United States 

This talk presents a finite population method to estimate annual 95[th], 99[th] and 99.9[th] centiles of surface water chlorpyrifos concentrations and associated confidence intervals using historical surface water pesticide monitoring programs with large number of measurements. The method provides reliable estimates, accounting for infrequent sample collection, measurements below reporting limits, and the strong seasonality present in these datasets. Data from three monitoring programs are considered: the USGS National Water-Quality Assessment Program (NAWQA), the Heidelberg University National Center for Water Quality Research (NCWQR) monitoring data, and the Washington State Department of Agriculture/ Department of Ecology surface water monitoring program. Results for temporal trends will be provided, including a comparison of estimates before and after chlorpyrifos was withdrawn from residential use in 2001.

Notes: 
The authors discussed a finite population method to estimate annual 95[th], 99[th] and 99.9[th] centiles of chlorpyrifos concentrations in surface water and associated confidence intervals using historical surface water monitoring data from three different monitoring programs.  The three monitoring programs included: the USGS National Water-Quality Assessment Program (NAWQA), the Heidelberg, University National Center for Water Quality Research (NCWQR) monitoring data, and the Washington State Department of Agriculture/ Department of Ecology surface water monitoring program. In their talk, the authors presented a comparison of chlorpyrifos concentration estimates before and after the removal of residential chlorpyrifos uses as well as a comparison of chlorpyrifos concentrations in agricultural and urban watersheds. 

The following issues were highlighted by the presenters:  1) the monitoring sampling is generally not from a probability design; 2) there are potentially only a few observations per year at a given site; 3) and non-detections are common. 

The correlation of the monitoring datasets and actual chlorpyrifos use -- application timing, crops, etc. -- was not discussed.  Therefore, the representation of the annual 95[th], 99[th] and 99.9[th] centiles of chlorpyrifos concentrations in surface water was not clear. 


7. Sampling Plans for Water Quality Assessment
   John W. Green, Ph.D. 
   DuPont Applied Statistics Group

Sampling plans for monitoring chemical concentrations in rivers were investigated for two California streams of differing order to assess the relative merits of daily sampling versus less frequent sampling. The assessments were based on the ability to estimate time-weighted total and peak concentrations on a monthly or annual basis and to detect the presence of spikes in these concentrations. A secondary purpose was to determine the feasibility of predicting spikes in chemical concentrations from the river discharge rate. In these streams, some less frequent sampling plans were shown to provide almost the same utility as daily sampling for the first purpose and discharge rate was shown to have little correlation with chemical concentration.

Notes: 
Sampling plans for monitoring chemical concentrations in rivers were investigated for two California Rivers (Orestimba Creek at River Road near Crows Landing and Sacramento River at I Street Bridge) to assess the merit of daily sampling versus less frequent sampling. The assessments were based on the ability to estimate time-weighted total and peak pesticide concentrations on a monthly or annual basis and to detect the presence of spikes in these concentrations. The feasibility of predicting spikes in chemical concentrations from the river discharge rate was also investigated. The presenter suggested that for these datasets, less frequent sampling plans provide almost the same utility as daily sampling and that pesticide concentrations did not correlate with the discharge rate.

   
8. Update on Development of Drinking Water Intake Watershed PCAs
   Jim Carleton, Rochelle Bohaty, Marietta Echeverria, Nelson Thurman, Michelle Thawley, Katrina White, Mohammed Ruhman 
   Environmental Fate and Effects Division.  United States Environmental Protection Agency, Arlington, VA

EFED uses percent crop area (PCA) adjustment factors in its drinking water assessments to account for the fact that watersheds supplying surface drinking water sources may not be devoted entirely to agriculture.  The first PCAs were developed in the late 1990s and mid 2000s, based on cropped acreage within 8-digit HUCs.  More recently, EFED has developed delineations for watersheds associated specifically with drinking water intakes that supply community water systems.  EFED is planning to use this new dataset to develop drinking water intake watershed-specific PCAs.  This talk will discuss the intake watershed dataset and PCA development methodologies in general terms, along with unique issues and challenges associated with the use of these new data for PCA development.

Notes: 
Drinking water intake (DWI) percent cropped area (PCA) adjustment factors are being developed to represent delineated watersheds with surface water intakes serving community water systems (CWS).  These PCAs use the same crop distribution data as other PCA projects (Ag Census and NLCD), but are based on the 4,840 DWI catchments judged "reasonable" by the USGS.  Some DWIs were not included because they are not associated with NHD+; however, surrogate PCAs can be used in these instances.  The catchments vary greatly in size (1/3 are less than 10 mi[2] and (1/4) are greater than 1,000 mi[2]).  These PCAs are limited by the resolutions of the datasets involved (AgCensus only provides data at the county level), and there is temporal variation in cropland data.  Also, there are concerns with releasing drinking water intake data as it relates to homeland security.  Next steps in this project are to develop draft PCAs, validate the GIS process, evaluate draft PCAs, finalize PCAs, and finalize the report and guidance document.

Q: Will drinking water intake PCAs replace HUC-derived PCAs?
A: This is a policy decision that will be made in the future.  However, a HUC PCA may be more appropriate in some cases as DWIs are not available for all localities.

Q: The analysis being done for DWI PCAs seems very similar to an approach currently underway for assessing endangered species.  Are these efforts duplicative?
A: These efforts are not the same as there are pixel value differences between the data sets being used and the land cover data is being used for very different purposes (e.g., proximity analysis versus percent of a watershed with a specified land cover).

Q: Is any work being done beyond GIS work to determine the appropriateness of including water storage reservoirs among the DWI PCA catchments?
A: No, only GIS analysis is being used to identify catchments.  EFED does not have the resources to individually verify all water storage reservoirs.

Q: Will a specific representative DWI catchment be identified as being protective and representative for specific crop(s)?
A: For the HUC PCAs, specific crops and regions are identified that address this issue; however, identifying crops and regions for DWI PCAs cannot be done easily as many of the catchments are nested within large catchments complicating which catchment is representative.

Q: What are the minor crops in consideration for crop specific PCAs?
A: The same groups as for HUC PCAs: all agriculture, corn, soybean, cotton, wheat, turf, orchard/vineyard, vegetable, and rice

Q: What is the measure of success for this project?
A: When verification, comparison to monitoring, comparison to old PCAs, and implementation for use in EFED is completed, the project will be considered a success.

Q: What is included in the turf layer?
A: The turf layer used to calculate the HUC8, HUC10, and HUC12 PCAs recently completed covers residential turf and sod farms.  There was uncertainty for that turf layer for sod farms as many acres of sod farms are not reported in Nationall Agricultural Statistis (NASS) data.  A new turf layer has been created that is specific to residential turf (e.g., did not account for sod farm turf) and that layer was used with the DWI-PCA calculation.  The all agricultural PCA is expected to be used for sod farms for PCAs completed in the future.

9. Exposure Assessment for Pronamide Drinking Water Residues in California Central Coast Lettuce Production Areas 
   Nick Poletika[1], Brian Bret[2], Mike Winchell[3]
   [1]Dow AgroSciences LLC
   [2]Dow AgroSciences LLC, Regulatory and Government Affairs NAFTA
   [3]Stone Environmental, Inc.  Applied Information Management

Regulatory pesticide human exposure models are intentionally designed to be protective of individuals in populations.  Consequently, they tend to be quite conservative in early assessment tiers, and predictions are generally refined with data in subsequent assessments.  For drinking water obtained from surface water, pesticide exposure model concentration estimates for agricultural use generally are in the ug L[-1] to tens of ug L[-1] range, which can lead to concern if human health endpoints for a specific pesticide are low values.  Currently US EPA OPP accepts only targeted monitoring studies for refinement of model estimated concentrations.  We present an alternative assessment method to targeted monitoring for a pesticide with a limited range of label crops and geographic regions of use that relies on proximity analysis of use sites and drinking water intakes.  We also address national security issues related to sensitive intake location data and consider the general utility of the approach.

Notes:
Using the case study of pronamide on leaf lettuce in the central California coastal region, the presentation focused on an approach for evaluating drinking water exposure by examining public records and data sets to assess potential contamination of surface water sources.  By using historical surface water drinking water intake locations, SDWIS data, Internet resources (e.g., Google Earth), and discussions with community water system (CWS) employees, a dataset was compiled of active drinking water intakes.  The exposure potential for the active drinking water intake was assessed by using the ArcGIS tool NHDFlowline and use data from California's Pesticide Use Reports.  Sites were eliminated from future consideration if the pesticide was not used in the county, if there was no hydrologic connection (as determined by NHDFlowline), or if the intake was located upstream of the use sites.  One revelation of the research was that in the central California coast region a number of CWS receive raw water from remote sources, which tend to be in the elevated regions of the State, as opposed to the valley regions where lettuce is grown.  Based on the pronamide case study, of the 233 locations evaluated, none had the potential for pronamide contamination and exposure.  The recommendation was made that this approach could be extended to other regions of California, but that routine updates would need to be conducted to demonstrate that no exposure changes had occurred.  As drinking water intake locations are considered sensitive information, it was also recommended that EPA work with the USGS to more easily conduct this type of analysis in the future.  

Q: Can this analysis be done for other crops, or is it  only applicable for lettuce? 
A: Using the PUR data in general, this analysis  could be conducted  for other crops.

   
10. Hybrid PRZM:"Filling in the Gaps" in Field Sampling Data using Realistic Simulation Modeling
   Paul Miller, PhD[1], Chris Harbourt, PhD[1], William J. Northcott, PhD[1], Paul Hendley, PhD[2], and Jessie J. Prenger[1]
   [1]Waterborne Environmental, Inc.
[2]Syngenta

The Hybrid-PRZM approach can be used to combine observations from field monitoring with results from simulation modeling  based on local rainfall for the same period to "fill in" gaps between sampling occasions. The approach uses Pesticide Root Zone Model (PRZM) simulations based on local rainfall data together with site-specific watershed soil and agronomic characteristics. The model is used to estimate edge-of-field atrazine concentrations occurring on days with rainfall; the simulation results are scaled to watershed conditions without calibration. The PRZM model results supplement field monitoring data with estimates of potential runoff events on days when runoff inducing rain falls in a watershed between sampling occasions.  This technique captures the strengths of field and simulation approaches to develop conservative daily time series using quality sampling results and appropriate simulation assumptions.  Significant developments are presented here that result in improved exposure simulation results; these include the use of NEXRAD local precipitation data combined with a land workability algorithm to develop appropriate distributions of application dates across a watershed.

Notes:
The hybrid PRZM techniques for estimating peak concentrations / acute exposure levels in surface waters are being developed at Waterborne Environmental, Inc. in cooperation with Syngenta Crop Protection. 
Why is this needed? There's a concern that monitoring programs may be missing peak concentrations, in some cases by quite a large margin. 
The use of bias factors is one promising method for extrapolating from moderate-frequency monitoring data to conservative estimation of what the peak daily exposures might look like if higher frequency monitoring data were available for the pesticide of interest. 
The author proposed a more "realistic" way of estimating high frequency data (i.e., acute exposure levels). The method utilizes regionally specific data for inputs into the PRZM modeling system to set up simulations that are calibrated (but with no "model specific" calibration required) with the available monitoring data to fill in the exposure distribution. This method has been applied to predict site and event specific concentrations of atrazine at the edge of field. The local information includes: 
   * Nexrad rain temp solar data 
   * Planting date, crop maturity 
   * Soil characteristics 
   * Cropland data layer, 
   * Historical and current weather (Growing degree days, rainfall etc.)
   * Crop yield and soil moisture models
   * Etc.
      
Locally measured weather and pesticide application data are used. The watershed PCT (percent crop treated) calculations are based on Cropland Data Layer (CDL) land-use data from the NASS and grower survey data (local but not specific  -  see presentation authors for details). The atrazine example presented crop reporting district - based application rates.
SSURGO soils data is used for site-specific inputs on corn and sorghum in the example presented.
Algorithms have been developed to simulate farmer behavior and agronomic timing based on local conditions to give more realistic inputs on a daily basis. The method accounts for daily changes in land workability and distribution of practices over the watershed  -  there is a LOCAL response regarding whether land can be worked.  The reluctance of farmers to plant too early is accounted for.
The first planting date is determined from 14 day moving average of air temperature when 55 degrees and soil moisture is less than 0.85 and date is April 1 or later then planting is possible (for Midwest?)
Several adjustments to the inputs were made to improve the precision and accuracy of modeling of runoff events. The modeled window for potential applications was started 7 days prior to the accumulation of 630 GGD (growing degree days). If local variation is not considered, predictions are still acceptable. 
Conclusions:
PRZM hybrid shows considerable promise for supplementing 7-day interval monitoring data.
The method predicts the relative magnitude of peaks but over predicts (the overall peak?) by >20 ppb- The method reflects reality in that peaks are predicted only during real runoff periods. The new land workability approach provided improved estimates of atrazine application timing.
The authors will be evaluating larger community water supply watersheds as well - the examples in this presentation were for watersheds of 10 to 1000 square miles that were part of the atrazine ecological monitoring program. Instead of a mechanistic approach, a WARP-type (USGS model) regression approach for larger watersheds may be used. Probabilistic approaches will also be investigated.

