Page 95 - Modelisation du devenir des pesticides...

3.3 Results and discussion

ted bromide leaching is underestimated. The bias ranges from -50 to -70% (Table
3.6), but it should be noticed that the observations can vary by up to 60%. As
for the water drainage, the correlation between observed and simulated accumula-
ted bromide leaching is correct (above 0.86) but due to the underestimation, the
efficiency is negative. Despite the underestimation of drainage flow predicted by
the model, the bias is within the range of model performances found in a similar
context with WAVE and LEACHP at Weiherbach. 33,34
At Thiverval-Grignon, the bromide content was monitored in the top 30cm
(while the water content was monitored in the top 5cm) by analysing two samples.
The discrepancies between the two samples could be linked to spatial variability at
the scale of the filed. There is better agreement between the simulated and measu-
red bromide content in the first year (absolute bias lower than 10% and efficiency
above 0.48) than in the second year (overestimation by more than 40%, negative
efficiency). These results might be linked to a fast leaching of the bromide in 1994,
wich is not represented. However, this cannot be confirmed by the experimental
data since the water content was only observed in the top soil before the bromide
application. Overall the bromide simulations are correct (bias <10%, Nash>0)
on two sites, Vredepeel and Thiverval-Grignon in 1993, but poor in Kerlavic for
both years, and poor in Thiverval-Grignon for 1994. However, as illustrated by
the comparison of models made by Francaviglia, 80 good prediction of the bromide
leaching does not necessary imply appropriate prediction of the pesticide loads in
the leachate.

3.3.4 Pesticide content

At Vredepeel, the bentazone was applied in late November (Table 3.1). As this
pesticide is weakly adsorbed to soil, the solute transfer is expected to dominate
and the simulated bentazon profiles are indeed, quite close to thepredicted bromide
residue profiles (Fig. 3.3.3). Nevertheless, the bentazon transfer is slower thanthat
of bromide, because of adsorption. Some sharp variations in the simulated profiles
are associated with initial conditions. Although the bias on bromide was low, the
bentazon residue is overestimated by 44%.
This suggests that the degradation of bentazon simulated by the model is too
weak. The degradation parameter used was set according to the measurements
made in the laboratory, which do not represent the degradation at the field scale
poorly. 62 Isoproturon is assumed to be more strongly adsorbed in soil than benta-
zon. Consequently, slower pesticide transfer can be expected at Kerlavic. However,
in the first year, isoproturon is detected only a few days after application (Fig.
3.3.3) which might be linked to a preferential flow. Isoproturon leaching varies in
a considerable range between replicates (400% compared to 60% for bromide in
second year). There are also a large differences between years, as the isoprotu-

83