The validation of pesticide leaching models

The validation of pesticide leaching models presents particular problems where the number of model predictions is far in excess of the observed data. Normally, however, there are more frequent held observations for other parameters (notably the site hydrology) than for pesticide concentrations in either water or soil. A five-stage validation procedure which takes advantage of the most frequently available observations and which tests each of the components of the model in a cumulative way, is thus advocated: Stage 1: Parameterisation of the model using only independently measured parameters. Stage 2: Hydrological validation: the validation of the predictions of water movement and water content of the soil. Stage 3: Solute movement validation: where field data are available for solutes other than pesticide, the model should first be validated for them, especially if they are more abundant than the pesticide observations. Conserved solutes such as chloride or bromide are preferred, although nitrate may be used for short periods. Stage 4: Pesticide fate in the soil: models should use parameters of pesticide fate derived from independent studies. Stage 5: Pesticide leaching: only in the last stage are the relatively small number of pesticide observations compared with the model predictions with respect to patterns and orders of magnitude of occurrence. With this scheme, the results of each stage are carried forward to the next, and confidence in the model is built with each stage. This is illustrated using the CRACK-P model and hydrological, nitrate and pesticide data from the Brimstone Farm Experiment Oxfordshire, UK.