Effect of aggregation of soil parameters in spatially distributed pesticide leaching assessments

Models developed for the national scale cannot be used for local predication due to spatial aggregation

Spatially-distributed pesticide leaching models are now commonly used in pesticide registration and policy evaluation. In the Netherlands, the GeoPEARL model is used to calculate the regulatory endpoint, which is the 90th spatial percentile of the leaching concentration. To get an accurate estimate of this spatial percentile, the model must be run for a large number of unique combinations of soil type, weather district and groundwater depth group. In the current GeoPEARL schematisation, these unique combinations are aggregated to 6,405 larger spatial units. Within each unit the dominant soil profile is assumed to be representative for the whole unit.

We tested the effect of this spatial aggregation on nationwide predictions of pesticide leaching. Based on the soil map 1:50,000, each unit was subdivided into sub-units with the same soil type. Soil profile descriptions were obtained for all spatial sub-units. In this way, a more detailed spatial schematisation with 48,233 unique combinations was obtained. GeoPEARL was run for four substances using both schematisations. Comparison of results shows a minor effect on the nationwide frequency distribution of the predicted leaching concentration for all substances. Also, the 90th percentile leaching concentration was hardly affected by spatial aggregation of soil type within larger spatial units. However, when the detailed and aggregated maps are compared on a pixel basis, 30-40% of the pixels showed differences greater than 25%. These findings indicate that the current GeoPEARL schematisation works well for predicting pesticide leaching on the regional to national scale, but that local prediction is meaningfully affected by spatial aggregation of soil parameters.