Uncertainty and sensitivity analysis of GeoPEARL

Computations using GeoPEARL with the standard schematisation were compared with those obtained with a schematisation at higher spatial resolution. For all three pesticides considered the nationwide spatial frequency distribution of the median annual leaching concentration (PEC50) and the spatial 90th percentile of the PEC50 (SP90) were hardly affected by spatial aggregation of soil type within larger spatial units. For the assessment of the propagation of uncertainties in the input, only soil properties and the most important pesticide properties, i.e. half-life in soil (DT50) and the coefficient for sorption on organic matter (Kom) were considered.

First, the uncertainties in the soil data and the pestcide properties were quantified. Next, a regular grid sample of points covering the whole of the agricultural area in the Netherlands were randomly selected.At the grid nodes, realisations from the probability distributions of uncertain inputs were generated and used as input to a Monte Carlo uncertainty propagation analysis. Uncertainties in DT50 and to a lesser extent Kom contributed most to the uncertainty in PEC50 and SP90. The uncertainty about the PEC50 at point locations is greater than that about the SP90. When taking the uncertainties into account, the SP90 of the leaching concentration shifted towards greater values. Recommendations are made for further improvement of the model predictions, in particular by reducing the uncertainty in DT50.