Modeller subjectivity in estimating pesticide parameters for leaching models using the same laboratory data set

User-dependent subjectivity in the process of testing pesticide leaching mo dels is relevant because it may result in wrong interpretation of model tes ts. About 20 modellers used the same data set to test pesticide leaching mo dels (one or two models per modeller). The data set included laboratory stu dies on transformation and sorption of ethoprophos and bentazone in soil fr om the top 25 cm. at two or three temperatures. All modellers received the raw data from these studies without guidance for deriving the model input p arameters. The modellers were asked to provide the values of the half-lives and sorption coefficients which the model considered would use for this so il layer at 10 degrees C (and at field capacity for the half-lives). The ha lf-life of ethoprophos ranged from 92 to 346 days with an average of 191 da ys and a coefficient of variation of 29%. The half-life of bentazone ranged from 33 to 204 days with an average of 83 days and a coefficient of variat ion of 46%. The linear and Freundlich sorption coefficients of ethoprophos ranged from 1.7 to 4.3 dm(3) kg(-1) with an average of 3.3 dm(3) kg(-1) and a coefficient of variation of 21%. The linear and Freundlich sorption coef ficients of bentazone ranged from 0.08 to 0.14 dm(3) kg(-1) with an average of 0.11 dm(3) kg(-1) and a coefficient of variation of 13%. This variabili ty caused by the interpretation of the modeller is so large that it overrul es conceptual differences between models in many cases. The most important cause of the variability in the half-lives was the expert judgement involve d in establishing the relationship between transformation rate and soil tem perature. Differences in fitting procedures played only a minor role for th e half-lives but they were an important cause of the variability in the lin ear sorption coefficient. Some recommendations are proposed to reduce the e ffect of user subjectivity on modelling results in future model tests. (C) 2000 Elsevier Science B.V. All rights reserved.