Predicting pesticide concentrations in river water with a hydrologically calibrated basin-scale runoff model

Abstract

Hydrological diffuse pollution models require calibration before they can be used to make accurate long-term predictions for a range of hydrological and meteorological conditions. As such, the applicability of the models to the dispersion of new pesticides is limited due to the lack of calibration data. In this study, the performance of a GIS-based basin-scale runoff model for predicting the concentrations of paddy-farming pesticides in river water was examined when calibrated using hydrological data alone, without optimization based on empirical pesticide concentration data. The prediction accuracy on a daily or hourly scale was somewhat unsatisfactory due to inevitable compromises concerning rice farming schedules. However, the month-averaged pesticide concentrations were satisfactorily accurate; more than 50% of predicted values were between half and twice the observed values, considering the deficiencies of the input data, particularly for pesticide usage, which may include up to 50% error.