Uncertainty and stochastic sensitivity analysis of the GeoPEARL pesticide leaching model

GeoPEARL is a spatially distributed model describing the fate of pesticides in the soil–plant system. It calculates the drainage of pesticides to local surface waters and the leaching into groundwater. GeoPEARL plays an important role in the evaluation of Dutch pesticide policy plans. This study analysed how uncertainties in soil and pesticide properties propagate through GeoPEARL for three representative pesticides. The GeoPEARL output considered is the 90th percentile of the spatial distribution of the temporal median of the leaching concentration (P90). The uncertain pesticide properties are the coefficient of sorption on organic matter and the half-life of transformation in soil. Both were assumed uncorrelated in space and were represented by lognormal probability distributions. Uncertain soil properties considered were horizon thickness, texture, organic matter content, hydraulic conductivity and the water retention characteristic. Probability distributions were derived from meta-data stored in the Dutch soil information system. A regular grid sample of 258 points covering the agricultural area in the Netherlands was randomly selected. At the grid nodes, realizations from the probability distributions of uncertain inputs were generated and used as input to a Monte Carlo uncertainty propagation analysis. The results show large uncertainties in P90, with interquartile ranges larger than the median for all three pesticides. Taking input uncertainty into account also leads to a systematic shift of the P90 towards greater values. Stochastic sensitivity analysis showed that the pesticide half-life is the main source of uncertainty and that the coefficient of sorption to organic matter and uncertainty in soil organic matter contribute to a lesser extent. Uncertainty contributions from other soil properties were negligible. These results suggest that improved assessment of soil properties will only marginally improve the accuracy of the predicted pesticide leaching. Instead, more accurate assessment of the pesticide properties, in particular the pesticide half-life, is required. This is, however, difficult, because the pesticide half-life depends on highly variable soil microbial properties in a way that is as yet poorly understood.