Geophysical Research Abstracts, Vol. 8, 00333, 2006
SRef-ID:
© European Geosciences Union 2006
Inclusion of spatial variability in Monte Carlo
simulations of pesticide leaching
B. Leterme (1), M. Vanclooster (2), A. Tiktak (3), A.M.A. van der Linden (4),
M.D.A. Rounsevell (1)
(1) Dept. of Geography, Université catholique de Louvain, Louvain-la-Neuve, Belgium, (2)
Dept. of environmental sciences and land use planning - Génie Rural, Université catholique de
Louvain, Louvain-la-Neuve, Belgium, (3) Milieu- en Natuurplanbureau (MNP), Bilthoven, the
Netherlands (4) National Institute for Public Health and the Environment, Bilthoven, the
Netherlands(leterme@geog.ucl.ac.be / Phone: +32-10-472842)
Pesticide leaching is one of the high priority issues in both the EU Thematic Strategy
on the Sustainable Use of Pesticides and the registration of plant protection prod-
ucts. So far deterministic models have been used in the evaluation process and more
recently spatially distributed modelling has also been introduced. It is however well
known that, at the point scale, variability in input parameters might have a large effect
on the predicted leaching concentrations. It is unknown to what extend spatially vari-
ability affects uncertainty in predicted leaching concentrations. This paper deals with
an uncertainty analysis of the GeoPEARL model, with special emphasis on the in-
clusion of spatial distribution knowledge. GeoPEARL is a process-based, distributed
pesticide leaching model (Tiktak et al., 2003), used for calculating spatial percentiles.
The most important endpoint in pesticide registration is the 80
th
percentile in space.
The main objective of this study is to determine to what extent this percentile is af-
fected by the inclusion of the spatial variability of model inputs. The Monte Carlo
(MC) uncertainty analysis applied to the arable part of the Dyle river catchment (cen-
tral Belgium) involves n model runs for each of the k unique plots. The analysis is
performed for atrazine and bentazone.
The most sensitive input parameters are selected through literature review and dis-
cussion with model developers. The present research defines two categories of input
parameters that depend on the type of uncertainty associated with them. The first cat-
egory contains parameters denoted as ‘plot’ parameters because their spatial distribu-