Page 87 - Modelisation du devenir des pesticides...

3.2 Materials and methods

and used cultivar values recommended for STICS 46 while the agricultural practises
were prescribed according to the experimental data available. Pesticide properties
were set according to measured data when possible (Table 3.3).

Pesticide Solubility Koc Kne ads Kne des DT50
3
-3
-1
-1
-1
-1
mg/dm dm kg d d KJmol d
Atrazine 35 70 0.9 0.6 45.4 26
references 42 57 20 20 60 57
Isoproturon 70.2 40 0.85 0.085 46.8 25
references 42 42 59 59 60 56
bentazone 570 5.5 0.0 0.0 73.8 62
references 42 55 - - 60 62

Table 3.3 – Pesticide characteristics used : Koc soil organic carbon sorption co-
efficient, non-equilibrium sorption (Kneads) and disorption (knedes)
coeefficient, the molar enthalpy for transformation, DT50 degrada-
tion time for 50% opf compound

The Koc value for Bentazone was graphically determined from the laboratory
measurements provided by Boesten and Van der Pas. 55 The non-equilibrium sorp-
tion was not taken into account for bentazon because of the weak sorption of the
bentazon onto the soil particles. Non–equilibrium sorption rate are difficult to mea-
sure and data available in the literature are very scarces. The values selected for
atrazine were provided by Rat. 20 For isoproturon, as suggested by Guimont, 56
we used a ratio of 10 between desorption and sorption at non-equilibium, and the
sorption rate was set to 0.85 according to the adsorption rate recommended for
chlortoluron by Altfelder. 59 The DT50 for the three pesticides selected were esta-
blished from batch experiments in the laboratory. The molar enthalpy values were
taken from EFSA. 60 A 3-year spin-up was used to reach realistic initial values for
soil moisture.

3.2.5 Model evaluation methodology

The methodology for evaluatingthe model is divided into two approaches, the
firstbeing a graphical display comparison and the second is a statistical analysis
with the criteria detailed in this subsection. The statistical indicators selected for
the model assessment are the Root Mean Square Error (RMSE), The Modelling
Efficiency (EF), the Bias (ERR) and the Pearson Correlation Coefficient (CR).

100 s P (P i − O i )
RMSE = ¯ × (3.2.10)
O n

75