Page 90 - Modelisation du devenir des pesticides...

Chapitre 3 Evaluation du modèle STICS version pesticide sur différents sites
expérimentaux
are within the range of the crops statistics at the European scale 72 for the wheat
sites of Vredepeel and Kerlavic, and a bit high for the maize at the Thiverval-
Grignon site. The maximal LAI are in the range of those observed for winter
wheat and maize by Serrano et. al. 73 It can be noted that, although the wheat
yields are comparable at Vredepeel and Kerlavic, the maximum Leaf Area Index
and maximaum root depth are about 30% higher at Vredepeel.

3.3.2 Soil water profiles and drainage

Fig. 3.3.1 compare the observed and simulated soil water profiles several days
after application (DAA) or the time evolution of the drainage flows. At Vredepeel,
the values are plotted for 4 different dates (from December to August). The model
represents the temporal evolution of the soil water content poorly (especially on
August 278DAA).
Table 3.6 presents the mean statistical scores calculated at 22, 42, 103, 214 and
474 DAA : the full amount of water is underestimated by 5% and the efficiency is
negative.
Similarly to Klein et al.’s findings, 26 the simulated soil moisture stays close to
the field capacity at each depth except during a short period in summer when the
evaporation creates a soil moisture depletion, while observations show a larger and
longer depletion in summer. This is probably associated with the inability of the
model to take the impact of the seasonal fluctuations of the shallow aquifer into
account. The soil parameters used in the model (see Table 3.2) were set according
to the study of Boesten and Van der Pas 55 and lead to quite different values for
the wilting point and field capacity compared to classical pedological functions
associated with soil texture. 74,75 This allows the model to store more water in
the soil, indirectly accounting for the dampening effect due to the presence of the
watertable. However, the watertable dynamic is not simulated, and thus the tem-
poral evolution of the soil moisture is not correct. At Kerlavic, the accumulated
drainage flow is underestimated by the model from 21% to 25% in the model com-
pared to the average value of the lysimeters (Table 3.6). As previously mentioned
the 3 lysimeters were filled with disturbed soil one year prior to the monitoring
and cracks at the surface of the lysimeter boxes have been reported by Guimont. 56
Consequently, the drainage flow observation is probably influence by preferencial
flow. Moreover, the 3 lysimeters give contrasted results in terms of bromide and
isoproturon leaching (see below) and thus, it is not clear how the simulation can
be compared to the average observation. Although the STICS model can consider
cracks, which can lead to preferential flow, 46 we did not have enough data to
constrain this parameterization. Nevertheless the flow is well correlated with the
observations (Table 3.6). Soil water contents at Grignon were only observed at
5cm, and thus, are not representative of the first meter of soil. The observations

78