Validation procedures for pesticide leaching models

Validation of pesticide leaching models is a complex procedure, involving t he balancing of the importance of the various components of the models. How ever, modelling is a problem-solving activity, and the criteria appropriate for the validation of any model depend on the ultimate use of the modellin g exercise. These points are illustrated by a consideration of the results From the Brimstone Farm experiment. For scientific validation, is considere d that models must be independently tested an evaluated for their hydrologi cal representation. Once models show an acceptable hydrological balance, th ere comes the problem of identifying the fit between the model predictions and the observed leaching of pesticides. Frequently there is a need to matc h a full set of model predictions with a small set of observational paramet ers. Under such conditions, there is a range of model test statistics, whos e behaviour and properties are still poorly understood, particularly when a pplied to time series data. For "what-if" scenario evaluation, the validati on procedures are more appropriately constructed around the sensitivity of the model to the parameters that are being varied. For regulation, model ca n be used either evaluate the fate of new chemicals, or range of possible o utcomes of its use. Here, the main concerns are not only accuracy, but the degree to which the scenario modelled is of general applicability.