Abstract. Registrant use of computer simulation modeling is generally accepted by regulators in lieu of some experimental data for risk assessment, provided worst-case assumptions and extremely conservative criteria for negligible risk are used. These requirements taken together mean that only extreme cases escape the requirement for actual environmental data. This conservatism reflects the uncertainty in both exposure estimates by the models and in the hazard estimates from toxicologists. Until the credibility of estimated environmental concentrations given by models is improved by more experimental data and experience on the part of both registrants and regulators, the models will continue to be considered not-validated by many. There are four current developments that provide a compromise between the two impossible extremes of total model validation and total field measurement: (1) as models develop more process detail the parameters become more fundamental in nature and these are generally more amenable to independent estimation or measurement; (2) the reporting of probabilistic analyses and probable errors of results by modelers will enhance credibility; (3) the development of data sets for specific crop/site/pesticide/weather scenarios will allow users to modify those parts of the input they are concerned with while retaining confidence that parameters with which they are unfamiliar will have reasonable values; and (4) meso-scale and micro-scale physical simulation experiments can fill the gap between laboratory and field results. The most important need for enhancing model credibility is full access by the pesticide science community to all the information available, so that we may increase our understanding. Legal mechanisms should be considered for longer ownership of exclusive marketing rights of active ingredients, in exchange for complete public disclosure of the environmental fate and toxicology data submitted to regulatory agencies.

Additional index words: Registration, regulation, pesticide runoff, pesticide leaching, water quality, nonpoint pollution.

* This paper was originally published in Weed Technology. 1992. Volumbe 6:753-759.
** Res. Chem., U.S. Dep. Agric.., Agric. Res. Serv.., Nematodes Weeds and Crops Research, Univ. Georgia Coastal Plain Exp. Stn. Tifton, GA 31794.

 

INTRODUCTION

 

The practical need for models. It is not possible, ever, to know enough about a pesticide's behavior in the environment to be absolutely certain of its safety. We can try to learn enough, however, by experiment and experience, so that if the pesticide is released into the environment under the proper conditions, it is reasonable to hope that no unpleasant surprises will occur. In pursuit of this understanding, the amount of information demanded of a manufacturer to register a pesticide for use has become large and extremely expensive to acquire.

The object of this Symposium is to assess the current progress and future potential of computer simulation models to predict pesticide risks. We want these models to replace some very expensive field experiments in generating data required for registration. Field experiments to measure environmental concentrations or aquatic impacts require several seasons of data (even more if the weather doesn't cooperate) and several locations and can cost millions of dollars. If we have a credible computer simulation model for any of the processes by which pesticides impact on the environment, this could provide a relatively inexpensive way of evaluating that impact. The models replace experiments with an extrapolation from experience. Even the most sophisticated runoff simulation model can be set up and run for a fraction of the cost of a field experiment-but this is worth doing only if the results will be believed.

Most of the presentations in this Symposium are concerned with pesticide transport into water by leaching and runoff. This is because most of the agricultural environmental risk assessment simulation modeling research has been done in that area. Other pesticide environmental processes have been studied and simulated in various degrees of detail. For example some data and insights are available for plant uptake and herbicide efficacy, foliar degradation and washoff, crop pesticide residues, and volatilization. Although these will not be covered here, the time is ripe for them to be incorporated into a comprehensive, "mainstream" environmental simulation models. Some of these processes are being developed as subsystems of the USDA Root Zone Water Quality Model (RZWQM) (12). The purpose of this paper is to attempt to assess the sources of computer simulation model non-credibility, and to suggest how and why this situation may improve in the future.

 

THE PROBLEM WITH WATER QUALITY MODELS

 

The agricultural system that produces pesticide runoff and leaching-an interacting combination of soil, crop, pesticide, and weather-is extremely complicated (48, 49). No matter how detailed our understanding of each of its subsystems and/or our mathematical description of it, new complexity and sub-subsystems are always revealed by further research. There is never a complete data set to test the newest model. No matter how sophisticated, all computer simulation models are simplifications of reality. Whether we are modeling a well-characterized system or trying to extend our models to predict behavior in a less well known system, at some point we surpass our knowledge and must estimate effects. This often means lumping spatial and/ or temporal variability, or using an empirical function that fits one case in the hope it fits another.

"From the viewpoint of natural science, and indeed from any viewpoint concerned with truth, a disquieting aspect of computer based modeling is the gap between the model and real-world events. There is reason to fear that the gap will not grow smaller and that worry about it may ultimately just fade away" (38).

This complaint by Philip has been made by every experimental scientist who has tried a model and has enough experience with the difficulties of studying real systems to recognize how complex reality is. Nevertheless, the regulator, and the regulatee, will both make two replies:

 

1. Complex systems require complex models***. A process such as runoff cannot be analyzed without a complex model. There are too many dimensions to the problem. Even to make sense of measured results on such a system requires a model.

 

2. Models are required because for risk assessment we must extrapolate from measured data to unmeasurable (or at least highly unlikely to be observed) scenarios, especially the "worst case" scenarios of the regulators.

Generally we assume that our models will evolve into more universal descriptions of reality. We hope they will be more accurate as we learn more about fundamental processes and incorporate them into the software (11) thereby decreasing our dependence on empirical functions. For instance, for the process of pesticide sorption by soils we might replace a simple linear, equilibrium adsorption isotherm with (in increasing order of refinement) a nonlinear isotherm, a nonequilibrium (kinetic) single-step process equation, or a multi-step kinetic process. Each such increase in explicitly accounted-for complexity should increase the accuracy of simulation of reality and the ability of the equation to describe different situations (i.e., the universality of the equation increases).

Unfortunately, each increase in complexity of a model also increases the amount of input information that must be known or estimated for the model and also increases the available output. More input information must be known; more output information must be sorted through; and the model becomes more difficult to use and interpret. An ironical paradox arises: the more realistic model has less credibility. The less the user knows of the subtleties involved, the more suspicious he/she becomes of the models as their difficulty increases.

The reverse is also true. The more one knows about the system the more one recognizes the vast simplification of reality and the extremity of the assumptions used in the simplest models such as the SCS pesticide screen (17, 18) or other indexing procedures (19, 23, 24, 27). Of course, if you only want a rule of thumb you may be able to use a screening model. But no one is ever satisfied with screening-there is always an immediate demand for the next level of discrimination. Does more complexity indeed give better results? DeCoursey (11) argues that as you replace calibrated empirical algorithms with physical-based process algorithms you increase the precision of your predictions at the expense of accuracy. In other words "research" or "physically based" models give better calculations of relative differences. But because there are more input parameters, which will themselves have uncertain values, the absolute values of predictions will probably not be more accurate. Indeed, they will probably be less accurate if the values assigned to the input parameters are less accurate.

How can the more realistic but more complex models be made more credible? There are four current developments that will help.FOUR CURRENT DEVELOPMENTS THAT WILL IMPROVE CREDIBILITY FOR ADVANCED NONPOINT POLLUTION COMPUTER SIMULATION MODELS

 

*** Roger Smith. U.S. Dep. Agric., Agric. Res. Serv., Ft. Collings. CO. Personal communication, the Arlington, Texas CREAMS workshop. 1978.

1. More detailed physical process understanding will ultimately give more accurate predictions. In spite of the increase in input requirements, the more physically based a model, the more fundamental (in a physical sense) are its input parameters. Also, the more fundamental the parameters the better the procedures for independently estimating or measuring them are likely to be. For example, consider the replacement of the pesticide soil sorption coefficient K_d (21), a quantity which must be measured, with the more fundamental "organic carbon" sorption coefficient K_oc. For nonpolar pesticides, the latter is a more fundamental property related to the hydrophobicity of the pesticide molecule and not the behavior of the pesticide with a specific soil. Of course, to use K., the fractional organic carbon content of the soil F_oc must be known or measured since K_d = K_oc x F_oc. Thus, you still need a measurement of F_oc for the specific soil: you have increased the number of parameters needed to describe the soil sorption of the pesticide from one to two. But the two parameters K_oc and F_oc are easier to estimate separately than their product K_d; K_oc estimates can be made independently of the soil (21), and F_oc is usually available.

Thus, in principle, more detailed, more physically-based models should ultimately be more accurate as well as more precise.

 

2. The use of probabilistic and sensitivity analyses in the presentation of results will enhance credibility. A suspicion of any result presented without a probable error estimate is normal. Most of us have had the experience of reviewing a paper in which the author appeared to have no idea of the precision of his/her results. The response is a healthy skepticism about the results themselves.

 

Models do not have to be guilty of this. In the simplest deterministic models or equations, a propagation-of-errors analysis can be used to determine uncertainty in the predicted value. As models become more complex, this procedure becomes impossibly unwieldy and other methods must be used. Fontaine et al. (15) demonstrate powerful software techniques for analyzing quite complex models for input error sensitivity. Mills and Leonard (34) demonstrated how one might incorporate the stochastic nature of pesticide runoff probabilities. Bevin and Jakeman (6) discuss both stochastic (input probability estimates) and input error sensitivity analysis. They show how these reduce the dimensionality (number of parameters of concern of a model and resolve the impasse between the desire for as much physical understanding as possible to be incorporated into a model, and the overparameterization that results.

 

3. Scenario development: doing complex calculations only once. This is a straightforward way to both enhance credibility and make a model more friendly, as suggested by Oliver (37). A realistic environmental model is a re-creation of a piece of the world and must include a wide range of technical disciplines. Every user will have a specific interest and range of expertise, and may be uninformed or uninterested in the subtleties of the algorithms outside his/her field. A reasonable set of values for all parameters for specific scenarios will allow that user to explore the effects of changes in the parts of the input he/she is interested in, while being reasonably certain that the rest of the data (and especially that part of the input for which the user has no idea what the values should be) are at least reasonable. Unfortunately, the user may not always recognize that changes in his/her parameter may affect others, and a common error is the use of incompatible parameters. Once a scenario has been constructed and the calculations made, the results can be applied by the nonexpert to similar situations. An example is the set of scenarios in the STREAM Manual (13) in which detailed HSPF model calculations of runoff of pesticides with different combinations of persistence and K_oc have been run on crops on five large watersheds. A person with knowledge of soil science and pesticide environmental behavior can determine pesticide load and concentration exceedance frequencies for a specific pesticide from the curves plotted as a result of the calculations.

 

The Agricultural Research Service Root Zone Agricultural Model (RZWQM) currently in the development stage (12) will have scenarios for homogeneous single-crop watersheds consisting of complete data for a specific combination of crop, soil, cultural practice, and weather year. Cotton (Gossypium hirsutum L.) in the southeastern United States coastal plain and conventional and conservation-tillage corn (Zea mays L.) in the northern and southern corn belt are examples. All data necessary to run the model including crop growth, soil microbiology, macropore flow, pesticide ionization potential and other processes will be supplied. These data sets will not be real-no such complete data set exists. But the data will be reasonable values supplied by the model builders, and will be validated against those parts of the data which are available.

A hybrid system: combining physical and mathematical simulation. Acquiring "Real" pesticide nonpoint pollution data-water quality data taken in fullscale experiments on real agricultural fields under real weather and measuring all parameters-is excruciatingly difficult and expensive. One is at the mercy of the weather and developing a data base covering a representative range of weather conditions takes many years. Measuring pesticide concentrations in runoff and leachate are especially difficult because the amounts transported are very sensitive to rainfall timing (7, 30, 48, 51). Add to this the need for registrants to demonstrate "worst-case" results for the regulators, and field experiments which produce acceptable results can run very long indeed.

 

In addition to weather variability, large spatial variability in soil properties and crop means that any individual measurement obtained is the mean of a very diverse population. This means that making a deterministic connection of that value to any specific processes or variables can be difficult.

 

For these reasons several experimental techniques using small-scale physical simulations of pesticide dissipation processes have been developed, in which weather, especially rainfall, is simulated and therefore under control. These physical models are intermediate in scale between full field experiments and laboratory process studies. Each of the smaller-than-full-scale systems is an attempt to control variation by minimizing spatial and climatic variability, i.e., to look at a small volume within a huge multidimensional variable space. Some characteristics of each scale and some example references are given in Table 1, for five major dissipative processes for pesticides. From left to right in Table I the cost-and realism-of the studies described go up by about an order of magnitude per column.

 

The smallest-scale, laboratory, tests may be thought of as single-parameter measurements. At the next level, small-scale simulations are process experiments in which spatial variability is minimized. They allow cause-and-effect relationships to be ascertained and have in the case of runoff generated much new detailed process information. Small-scale experiments in Table I are used for "early-tier" risk assessment (the lysimeter tests are required in Germany).

 

The intermediate-scale runoff simulations ("mesoplots"-so named because they are analogous to the mesocosms in scale) are designed to be of a size realistic enough to allow crop- and management-scale variability, e.g., canopy effects, tillage effects, banded pesticide applications, etc.

 

For water pollution processes-runoff, leaching, and aquatic ecosystem impacts the intermediate-scale and small-scale experiments listed are currently under much active investigation in Europe and the United States. These experiments have the potential to fill some of the gap between model simulations and field results d to replace full-scale field experiments in many cases. They can, because of weather "control" provide better investigations of worst-case scenarios than field experiments. They are good enough reproductions of reality to provide useful data for environmental decision-making. When combined with a simulation model like CREAMS (28) GLEAMS (31), HSPF (26), RUSTIC (10), or others (2, 13, 14, 29, 30, 53, 55) and extrapolated to other conditions, these experiments should provide better risk assessments for a new pesticide than field experiments, for much less time and money. For additional security, an internal standard pesticide whose behavior has been extensively investigated should be included in the physical simulation.

 

 

For regulation purposes, modeling is best viewed as a tool we may use to organize and analyze the vast array of information that must be assembled on a pesticide's fate and behavior in the environment. Obviously, our models are only as good as our understanding of processes and our data base. We will approach our ultimate goal-credible risk analysis-only as this information base grows.

Sadly, however, much of the data-information generated by the industry to fulfill registration requirements is under lock and key at EPA. Although the research conducted to obtain those data has often been mandated by specific protocols with limited goals, the data should be quite useful for other purposes: the companies usually have recognized that it is in their interest to obtain the best data by the best science possible.

But the companies can't share their results. The patent protection of the chemical is too short-lived. If the information becomes public it may be used by others to support registrations of the same active ingredient. So EPA is bound by law to keep much of the companies' data confidential.

Everyone loses in this arrangement. The public and the pesticide science community cannot evaluate these data, and worse, no one can learn from them for the future. EPA does not have the resources to develop principles from the results; they are overwhelmed with the mechanics of the registration process. Thus, each company has to "reinvent the wheel": the companies cannot even learn the most basic information from each other. For example, currently there are several companies making huge investments in mesocosm experiments on quite similar compounds; each must operate as if the other's work does not exist. It is not necessary to explain who ultimately pays for all this redundant research.

The final irony is that the "privileged data" process has not even been able to accomplish its objective of fostering free trade; the amount of data required for registration has become so large and expensive that only the largest companies and the largest markets get new compounds.

One solution for this dilemma is an international agreement allowing pesticide manufacturers exclusive marketing, for a much longer time after registration, of the active ingredients for which they develop toxicology and environmental safety data. This would be in return for complete public disclosure of all environmental fate and toxicological information submitted to any regulatory agency as part of registration. The result would allow the entire scientific community to scrutinize the data. Pesticide science would take a significant leap forward both as a science and as a protector of the environment.

1. Ahuja, L. R., A. N. Sharpley, M. Yamamoto. and R. G. Menzel. 1981. The depth of rainfall-runoff-soil interaction as determined by 32p. J. Environ. Qual. 12:34-40.
2. Arnold, J. G., J. R. William, A. D. Nicks, and N. B. Sammons. 1990. SWRRB-a basin scale simulation model for soil and water resources management. Texas A. & M. Press, 210 p.
3. Baker, D. B. and R. P. Richards. 1991. Herbicide concentrations in Ohio's drinking water supplies: a quantitative exposure assessment. p. 9-29 in D. L. Weigmann, ed. Pesticides in the Next decade: the Challenges Ahead. Proc. 3rd Nat. Res. Conf. on Pesticides, Nov. 8-9, 1990. Virginia Water Res. Cent., Blacksburg. 832 p.
4. Baker, J. L., J. M. Laflen, and H. P. Johnson. 1978. Effect of tillage systems on runoff losses of pesticides, a rainfall simulation study. Trans. Am. Soc. Agric. Eng. 1978:886-892.
5. Bergstrom, L. 1990. Use of lysimeters to estimate leaching of pesticides in agricultural soils. Environ. Pollut. 67:325-347.
6. Bevin, K. J. and A. J. Jakeman. 1990. Complexity and uncertainty in models. p. 555-576 in D. G. DeCoursey, ed. Proc. Int. Symp. Water Quality Modeling on Agricultural Non-Point Sources, June 19-23. Utah State Univ., Logan, UT, U.S. Dep. Agric., Agric. Res. Serv. Rep. ARS81, June, 1990. 881 P.
7. Burgoa, B. and R. D. Wauchope. 1991. Pesticides in runoff and surface waters in T. R. Roberts and P. C. Kearney, eds. Environmental Chemistry of Pesticides John Wiley and Sons, Sussex, England (in preparation).
8. Burgoa, B. and R. D. Wauchope. 1991. The effect of formulations on runoff and leaching losses of atrazine and alachlor from tilted-beds. Abstr. Weed Sci. Soc. Am. 31:63.
9. Coody, P. N., J. W. White, and R. L. Ganey. 1990. A "small plot" approach to predicting pesticide runoff and aquatic exposure. Proc. Soc. Environ. Toxicol. Chem. 11^th Annu. Mtg., Nov. 11-15, Washington, D.C.
10. Dean, J. D., P. S. Huyakom, A. S. Donigian, K. A. Voos, R. W. Schanz Y. J. Meeks. and R. F. Carsel. 1989. Risk of unsaturated transport and transformation of chemical concentrations (RUSTIC) Volume 1: theory and code verification. U.S. Environ. Prot. Agency Rep. EPA/600/3-89/048a.
11. DeCoursey, D. G. 1991. Developing models with more process detail: do more algorithms give more truth? Weed Technol. (this issue).
12. DeCoursey, D. G. and K. W. Rojas. 1990. RZWQM-A model for simulating the movement of water and soluted in the root zone. p. 813-821 in D. G. DeCoursey, ed. Proc. Int. Symp. Water Quality Modeling on Agricultural Non-Point Sources, June 19-23. Utah State Univ.. Logan, UT, U.S. Dep. Agric., Agric. Res. Serv. Rep. ARS-81, June, 1990. 881 p.
13. Donigian, A. S.. D. W. Meier, and P. P. Jowise. 1986. Stream transport and agricultural runoff of pesticides for exposure assessment: a methodology. Environ. Prot. Agency Rep. EPA/600/3-86/01 Ia. March 1996, 760 p. Nat. Tech. Inf. Serv., Springfield, VA.
14. Ferreria. V. A. and R. A. Smith. 1990. OPUS: an advanced simulation model for nonpoint-source pollutant transport at the field scale overview. in DeCoursey. D. G., ed. Proc. Int. Symp. Water Quality Modeling of Agricultural Non-Point Sources, Part 2 U.S. Dep. Agric., Agric. Serv. Rep. ARS-1. June, 1980. p. 823434.
15. Fontaine, D. D., P. L. Havens, G. F. Blau, and P. M. Tillotson. 1991. The role of sensitivity analysis in groundwater risk modeling for pesticides. Weed Technol. (this issue).
16. Giddings, J. M. 1980. Types of aquatic microcosms and their research applications. p. 248-266 in J. P. Giesy, ed. Microcosms in ecological research. CONF-781 101. Tech. Inf. Cent.. Oak Ridge, TN.
17. Goss, D. W. 1991. Screening procedure for soils and pesticides remove to potential water quality impacts. Weed Technol. (this issue).
18. Goss, D. W. and R. D. Wauchope. 1990. The SCS/ARS/CES pesticide properties database: 11. Using it with soils data in a screening procedure. p. 471-493 in D. L. Weigmann, ed. Pesticides in the Next Decade: the Challenges Ahead. Proc. 3rd Nat. Res. Conf. on Pesticidcs, Nov. 8-9. 1990. Virginia Water Res. Cent.. Blacksburg. 831 p.
19. Gustafson, D. 1. 1989. Groundwater ubiquity score: a simple method for assessing pesticide leachability. Environ. Toxicol. Chem. 9: 339-357.
20. Hamacer. J. W. 1972. Decomposition: quantitative aspects. in C.A.I. Goring and J. W. Hamaker, eds. Organic Chemicals in the Soil Environment. Marcel Dekker, Inc., New York. 440 p.
21. Hamaker. J. W. and J. M. Thompson. Adsorption- p. 49-144 in C.A.I. Goring and J. W. Hamaker. eds. Organic Chemicals in the Soil Environment. Marcel Dekker. Inc., New York. 440 p.
22. Harper, L. A., A. W White, Jr., R. R. Bruce, A. W. Thomas and R. A. Leonard. 1976. Soil and microclimate effects on trifluralin volatilization. J. Environ. Qual. 5:236-242.
23. Hollis. J. M. 1990. Assessments of the vulnerability of aquifers and surface waters to contamination by pesticides- Research Rep. for U.K. Ministry for Forestry and Fisheries (MAFF), Soil Survey and Land Research Centre, Silsoe Campus, Silsoe. Bedford MK45 4DT. England. 27 p.
24. Hornsby, A. G. 1991. Site-specific pesticide recommendations: the final step in environmental impact prevention. Weed Technol. (this issue).
25. Hubbard, R. K., R. G. Williams, M. D. Erdman, and L. R. Marti. 1989. Chemical Transport from coastal plain soils under simulated rainfall: II. Movement of cyanazine, sulfometuron-methyl and bromide. Trans. Am. Soc. Agric. Eng. 32:1250-1257.
26. Johanson, R. C.. 1. C. Imhoff, J. L. Kittle, and A. S. Donigian. Jr. 1984. Hydrological simulation program-fortran (HSPF): user's manual for release B.O. U.S. Environ. Prot. Agency Rep. No. EPA June 1984. 767 p.
27. Jury, W. A.. D. C. Focht and W. J. Famier. 1987. Evaluation of pesticide groundwater pollution potential from standard indices of soil chemical adsorption and degradation. J. Environ. Qual. 16:422-428.
28. Knisel. W. G. ed. 1990. CREAMS: a field scale model for Chemicals, Runoff, and Erosion from Agricultural Management Systems. USDASEA Conserv. Rcs. Rep. No. 26, 643 p.
29. Kördel, W., M. Klein, J. Baust, and B. von Oepen. 1990. Simulation model using SESOIL for scenarios Germany, avail. Fraunhofer-Institut für Umweltchemie und Ökotoxicologie, Schmallenberg/Grafschaft. Germany. 29 p.
30. Leonard, R. A. 1990. Movement of pesticides into surface waters. p. 303-349 in H. H. Cheng ed. Pesticides in the Soil Environment – Processes, Impacts, and Modelling. Soil Sci. Soc. Am- Book Ser., No. 2, Soil Sci. Soc. Am., Madison, WI 53711. 530 p.
31. Leonard, R. A., W. G. Knisel, and D. A. Still. 1987. GLEAMS: groundwater loading effects of agricultural management systems. Trans. Am. Assoc. Agric. Eng. 30:1403-1418.
32. Leonard, R. A., G. W. Langdale, and W. G. Fleming. 1979. Herbicide losses from four upland Piedmont watersheds – data and implications for modeling pesticide transport. J. Environ. Qual. 8:223-229.
33. McDowell, L. L., G. H. Willis, L. M. Southwick. and S. Smith. 1994. Methyl parathion and EPN washoff from cotton plants by simulating rainfall. Environ. Sci. Technol. 18:423-427.
34. Mills, W. C. and R. A. Leonard. 1994. Pesticide pollution probabilities. Trans. Am. Soc. Eng. 27:1704-1710.
35. Nash, R. G. 1983. Determining environmental fate of pesticides with microagroecosystems. Residue Rev. 85:199-215.
36. Obenhuber, D. C. and R. Lowrance. 1991. Reduction of nitrate in groundwater microcosms by carbon additions. J. Environ. Qual. 20: 255-258.
37. Oliver, G., J. Burt, and R. Solomon. 1990. The use of surface runoff models for water quality decisions-a user's perspective. p. 197-204 in D. G. DeCoursey, ed. Proc. Int. Symp. Water Quality Modeling on Agricultural Non-Point Sources, June 19-23, Utah State Univ., Logan UT, U.S. Dep. Agric., Agric. Res. Serv. Rep. ARS-81, June, 1990. 881 p.
38. Philip, J. R. 1991. Soils, natural science, and models. Soil Sci. 151: 91-98.
39. Pionke, H. B., D. E. Glotfelty, A. D. Lucas, and J. B. Urban. 1988. Pesticide contamination of ground waters in the Mahantango Creek Watershed. J. Environ. Qual. 17:76-84.
40. Skipper, H. D., J. G. Mueller, V. L. Ward. and S. C. Wagner. 1986. Microbial degradation of herbicides. p. 457-475 in N. D. Camper, ed. Research Methods in Weed Science. 3^rd Edition, Southern Weed Science Society, Champaign, IL 61820. 496 p.
41. Smith, C. N., D. S. Brown, J. D. Dean, R. S. Parrish, R. F., A. S. Donigian. 1985. Field Agricultural runoff monitoring (FARM) Manual. Environ. Prot. Agency Rep. No. EPA/600/3-85/043, June, 1985. 230 p. Nat. Tech. Inf. Serv., Springfield, VA.
42. Smith, C. N., R. A. Leonard, G. W. Langdale, and G. W. Bailey. 1978. Transport of agricultural chemicals from small upland Piedmont watersheds. Environ. Prot. Agency Rep. No. EPA/600/3-78-056. Nat. Tech. Inf. Serv., Springfield, VA.
43. Spencer, W. F., M. M. Cliath, J. W. Blair, and R. A. LeMert. 1985. Transport of pesticides from irrigated fields in surface runoff and tile drain waters. U.S. Dep. Agric., Agric. Res. Serv. Conserv. Res. Rep. No. 31, Nat. Tech. Inf. Serv., Washington, D.C. 71 p.
44. Suntio, L. R., W. Y. Shiu, D. Mackay, J. N. Seiber, and D. Glotfelty. 1988. Critical review of Henry's law constants for pesticides. Residue. Rev. 103:1-59.
45. Taylor, A. W., D. E. Glotfelty, B. L. Glass, H. P. Freeman, and W. M. Edwards. 1976. Volatilization of dieldrin and heptachlor from a maize field. Agric. Food Chem. 24:625-631.
46. Taylor, A. W. and W. F. Spencer. 1990. Volatilization and vapor transport processes. p. 213-269 in H. N. Cheng. ed. Pesticides in the Soil Environment-Processes Impacts, and Modelling. Soil Sci. Soc. Am. Book Ser., No. 2, Soil Sci. Soc. Am., Madison, WI 53711. 530 p.
47. Touart, L. W. 1988. Aquatic mesocosm tests to support pesticide registrations. Hazard Evaluation Division Technical Guidance Document, Office of Pesticide Programs, U.S. Environ. Prot. Agency Rep. EPA/540/09-88-035. Nat. Tech. Inf. Serv., Springfield, VA.
48. Wagenet, R. J. and P.S.C. Rao. 1990. Modeling pesticide fate in soils. p. 351-399 in H. H. Cheng, ed. Pesticides in the Soil Environment, Processes, Impacts, and Modelling. Soil Sci. Soc. Am. Book Ser., No.
49. Wauchope, R. D. 1978. The pesticide content of surface water draining from agricultural fields-a review. J. Environ. Qual. 7:459-472.
50. Wauchope, R. D. 1981. Runoff studies and pesticide registration. p 200-205 in G. Zweig and M. Berom eds. Test Protocols for Environmental Fate and Movement of Toxicants. Assoc. Off. Anal. Chem.. Arlington, VA. 330 p.
51. Wauchope, R. D., T. M. Buttler. A. G. Hornsby. P.W.M. Augustijn-Beckers, and J. P. Burt. 1991. The SCS/ARS/CES Pesticide Properties Database for Environmental Decision-Making- Rev. Environ. Contam. Toxicol. (in press).
52. Wauchope, R. D. and D. G. DeCoursey. 1986. Measuring and predicting losses of herbicides in runoff water from agricultural areas. p. 135-154 in N. D. Camper, ed. Research Methods in Weed Science, 3^rd Edition. Southern Weed Science Society, Champaign, IL 61820, 486 p.
53. Wauchope, R. D. and R. A. Leonard. 1980. Maximum pesticide concentrations in agricultural runoff. a semiempirical prediction formula. J. Environ. Qual. 9:665-672.
54. Weber, J. B. 1985. Soils. herbicide sorption and model plant-soil systems. p. 155-188 in N. D. Camper, ed. Research Methods in Weed Science, 3^rd Edition. Southern Weed Science Society, Champaign, IL 61820. 486 p.
55. Williams, J. R., C. A. Jones, and P. T. Dyke. 1984. A modeling approach to determine the relationship between erosion and soil productivity. Trans. Am. Soc. Agric. Eng. 27-:129-144.
56. World Wildlife Fund. 1991. Improving Aquatic Risk assessment under FIFRA: a report of the Aquatic Effects Dialog Group. In Press.

Address:
USDA - Agricultural Research Service
Nematodes, Weeds and Crop Research
P.O. Box 748
Tifton, GA 31793

Phone: (912) 386-3892
Fax: (912) 386-7225
E-Mail: don@tifton.cpes.peachnet.edu

Dr. Wauchope holds a B.S. in Chemistry from the University of North Carolina at Chapel Hill, 1965. He received his MS in Inorganic Chemistry from North Carolina State University at Raleigh, 1969. He thesis was titled: A Light-Scattering Study of Polymeric Complex Formation in Strongly Acidic Solutions of Titanium. He holds a Ph.D. in Inorganic and Physical Chemistry, also from North Carolina State University at Raleigh. His Dissertation was about The Temperature Dependence of Solubilities of Solid Aromatic Hydrocarbons in Water.

Dr. Wauchope is a Research Chemist for the USDA-Agricultural Research Service, Nematodes, Weeds and Crops Research Unit at the Coastal Plain Experiment Station in Tifton, GA. Since 1988, he is the Lead Scientist in pesticide environmental chemistry conducting groundwater and surface water nonpoint pollution studies. He is also involved in pesticide residue analysis, pesticide environmental risk assessment and nonpoint pollution modeling. He has conducted field studies of surface and groundwater pollution by pesticides. Dr. Wauchope was the director of pesticide residue and food safety laboratory from 1988 to 1993 in which he conducted laboratory research on pesticide analysis, pesticide mobility in soils, and pesticide physical chemistry.