Sources of error in model predictions of pesticide leaching: a case study using the MACRO model

Uncalibrated predictions of soil water balance, water content, non-reactive solute transport (bromide) and pesticide leaching (bentazone) made by three users of a comprehensive mechanistic model (MACRO) are compared to measured data obtained for a sandy soil at Vredepeel in the Netherlands. The objective was to assess the significance of different sources of error for making predictions of pesticide leaching. Objective statistical indices were used to compare the simulations made by different users and to evaluate overall model performance. All three users predicted very similar water balances. Soil water contents were in good agreement with the measurements, with the simulation based on measured hydraulic functions giving somewhat better predictions than those based on automatic estimation procedures (pedo-transfer functions). Bromide movement was also satisfactorily predicted by all three users despite an inability to reproduce the strong retention near the soil surface caused by finger flow. Bentazone dissipation in the field was severely underpredicted by all three users based on laboratory measurements of degradation. This error overshadowed the effects of differences in parameterisation between users.