Page 102 - Modelisation du devenir des pesticides...

Chapitre 3 Evaluation du modèle STICS version pesticide sur différents sites
expérimentaux
the Agriflux model was calibrated to provide a better estimation of the drainage
flow. This was not done in STICS resulting in a significant underestimation of
water drainage and isoproturon flow. At Thiverval-Grignon, the simulation of the
atrazine is slightly better by STICS than by LEACHP, PRZM or VARLEACH.
As mentioned by Baer and Calvet only linear sorption was activated for the
58 58
three model applications. The discrepancies between STICS and the other models
are probably due to the non-equilibrium sorption implemented in STICS. Only
STICS is able to simulate the presence of atrazine 6-month after application in an
amount comparable to the observations (Fig. 3.3.3).

3.4 Conclusion

The aim for the pesticide module included in STICS is to apply it at the ba-
sin scale in order to estimate the diffuse pollution associated with pesticides in
addition to the simulation of the nitrogen transferred to the hydrosystem. The as-
sessment presented above with 3 different sites and 3 different pesticides provides
an idea of the model strengths (genericity) and weaknesses (mostly concerning soil
water transfer in complex sites with shallow aquifer or preferential flow). However,
these results were obtained mostly using parameters measured in situ or parame-
ters extracted from a literature review of the experimental site. Such in-situ data
will not be available at catchment-area scale. In order to gain an idea of the error
that could be associated with a poor determination of the parameters, a simple
one-at-time sensitivity test was performed (Appendix A). As widely reported in
the litterature DT50 and Koc are the most sensitive parameter for pesticide pro-
perties (Boesten, 81 Dubus et al. , 82 and Heuvelink et al.), 83 whereas the field
capacity has most influence on solute transfer. These findings clearly reinforce the
necessity to obtain quality estimates of Koc and DT50 in various soil types and
climates for a spatially distributed application of the model. 84,85 To sum up, the
model provided results as reasonable as those of other pesticides fate models. As a
capacity model STICS encountered the same difficulties in representing the water
transfer at Vredepeel owing to the presence of a fluctuating shallow water table,
and at Kerlavic where preferential flows seem to occur on the reconstructed soil
contained in the lysimeters. However, even at those sites, STICS was able to cor-
rectly represent the difference in behavior between the conservative tracer and the
pesticides, which is assumed to be associated to a realistic representation of the
pesticide sorption and degradation processes. This study is an additional illustra-
tion of the difficulty of assessing the behavior of pesticides in soil by means either
soil profile samples or lysimeters, due to the uncertainty associated with the expe-
rimental approach, variation among replicates, and the partial vision of the whole
processes, as only soil pesticide content profiles or leaching fluxes (but not both)

90