I nik'd Sliilcs Kmironmenlal I'rolcclioii A»enc\ Office of Research ;iikI Dexclopmcnl National Kxposure Research Laboratory Research Abstract (iowrnmcni I'eiloniiiincc Kcsulis \ci (io;il hv\enlinu Polluliiin and Reducing Risk in Communilics. I Ionics. Workplaces and Ixosyslems Significant Research Findings: Advanced Pesticide Risk Assessment Technology Scientific Problem Predictive modeling is an important tool for assessing the environmental and Policy Issues safety of new pesticidal active ingredients and new uses for currently registered products, and for evaluating the implications of new findings in their environmental and product chemistries. Climate, soil properties, limnology, and agronomic practices influence exposure by controlling the movement of pesticides within the agricultural landscape and by governing the speed and products of transformation reactions. These factors vary with time and with location within the often continent-wide use patterns of agricultural chemicals. This variability, together with measurement uncertainties in the values of chemical properties, mandates a statistical and probabilistic approach to exposure assessment. Research Approach Pesticide dynamics and exposures depend upon highly variable properties of the atmosphere, agro-ecosystems, receiving waters, and resident biota. An effective pesticide modeling technology must include validated algorithms for transport and transformation of pesticides, extensive databases of agro-ecosystem scenarios (crop and soil properties, meteorology, limnology, fish community ecology) and graphical user interfaces to maximize the ease of production and interpretation of complex probabilistic analyses. Several agencies collect data of significance for environmental safety, but these data must be assembled in usable forms, organized by appropriate landscape units, and made accessible to simulation models if their potential is to be realized. Models that capture underlying mechanism and process are necessary for reliable extrapolation of laboratory chemical data into field conditions. For validation, these models require a major revision of the conventional model testing paradigm to recognize the primacy of model user's risk, as against developer's risk, in testing their predictive uncertainty. The predictive reliability of the models must be hypothesized and tested by methods that lead to conclusions of the form "the model predictions are within a factor-of two of reality at least 99% Results and Implications National Exposure Research Laboratory — October 2001 ------- of the time." Once predictive reliability is established, it can be treated as a "method error" within a probabilistic risk assessment framework. Under APM 131, entitled "Develop a Probability-Based Methodology for Conducting Regional Aquatic Ecosystem Exposure and Vulnerability Assessments for Pesticides" this project has developed a step-by-step process for establishing the predictive reliability of exposure models. Research Collaboration and Publications Monte Carlo simulation is the preferred method for capturing variability in environmental driving forces and uncertainty in chemical measurements. Latin Hypercube Sampling (LHS) software is under development to promote efficient computer simulation studies and production of tabular and graphical outputs. Desirable outputs include exposure metrics tailored to available toxicological data expressed as distribution functions (pdf, cdf) and, if needed, empirical distribution functions suitable for use in Monte Carlo risk assessments combining exposure and effects distributions. A combined climatological dataset for driving models of spray drift (AgDisp), cropland pesticide persistence (PRZM), and surface water exposure (EXAMS) is being assembled by combining two National Weather Service products: the Solar and Meteorological Surface Observation Network (SAMSON) data for 1961-1990, and the Hourly United States Weather Observations (HUSWO) data for 1990-1995. Together these provide coordinated access to sky cover, temperature, relative humidity, pressure, wind direction and speed, and precipitation. By using observational data for the models, "trace-matching" Monte Carlo simulation studies can transmit the effects of environmental variability directly to exposure metrics, by-passing issues of correlation (covariance) among external driving forces. Additional datasets in preparation include soils and land-use (planted crops) data summarized for the state parts of Major Land Resource Areas (MLRA), derived from National Resource Inventory (NRI) studies. APRAT was designed and is being conducted by a research team at the National Exposure Research Laboratory's Ecosystem Research Division in Athens, Georgia, Principal Investigator Lawrence A. Burns. Examples of recent publications from this study include: Barber, M.C. 2001. Bioaccumulation and Aquatic System Simulator (BASS). User's Manual Beta Test Version 2.1. EPA/600/R-01/035. Barber, M.C. 2001. "A comparison of models for predicting chemical bioconcnetration in fishes." MS in preparation for submission to the Canadian Journal of Fisheries and Aquatic Sciences. Bird, Sandra L., Steven G. Perry, Scott L. Ray, and Milton E. Teske. "Evaluation of the AgDRIFT® aerial spray drift model." Environmental Toxicology and Chemistry (Accepted 2001). Teske, Milton E., Sandra L. B ird, David M. Esterly, Thomas B. Curbishley, Scott L. Ray, and Steven G. P erry. "AgDRIFT®: A model for estimating near-field spray drift." Environmental Toxicology and Chemistry (Accepted 2001). National Exposure Research Laboratory — October 2001 ------- Burns, Lawrence A. "Probabilistic Aquatic Exposure Assessment for Pesticides." EPA Report. 2001. Burns, Lawrence A. "Performance standards and prediction uncertainty in exposure models." Environmental Toxicology and Chemistry (Submitted 2001). Burns, Lawrence A. "Exposure Analysis Modeling System (EXAMS): User Manual and System Documentation." EPA/600/R-00/081. 2000. Future Research The APRAT project began in 1999 and will conclude in 2008 with the release of studies serving as examples of the deployed technology and as validations of the underlying algorithms, datasets and probabilistic simulation methods. Modernization of individual computer codes to take advantage of Fortran 95's advanced memory management and modular sharing of common data structures and computational services is in progress. Assembly of the complete databases to serve the entire lower 48 states, development of specific scenarios for common classes of approved use patterns, and improvement of the models' internal algorithms are ongoing activities of this project. A number of major revisions of codes and algorithms are in the design phase, most notably improvements in linkage of shallow groundwater and tile drains between the PRZM and EXAMS models, and revisions of EXAMS' handling of sediment transport, benthic boundary layer exchange, and sorption kinetics. National Exposure Research Laboratory — October 2001 ------- Contacts for Questions and inquiries can be directed to: Additional Lawrence A. Burns Information (JS EPA, Office of Research and Development National Exposure Research Laboratory Athens, GA 30605-2700 Phone: 706/355-8119 E-mail: burns.lawrence@epa.gov M. Craig Barber US EPA, Office of Research and Development National Exposure Research Laboratory Athens, GA 30605-2700 Phone: 706/355-8110 E-mail: barber.craig@epa.gov Sandra L. Bird US EPA, Office of Research and Development National Exposure Research Laboratory Athens, GA 30605-2700 Phone: 706/355-8124 E-mail: bird.sandra@epa.gov National Exposure Research Laboratory — October 2001 ------- |