Page 9 - Volume 4, Issue 1, January 2011

© Benaki Phytopathological Institute
Uncertainty in pest risk analysis 7
Figure 1A.
Uncertainty about parameter values
was described here by uniform distribu-
tions defined with lower and upper bounds
set equal to ± 20% of the estimated parame-
ter values reported by Magarey et al. (2005):
Tmin ~Unif(0.8, 1.2), Tmax~Unif(24, 36), Topt~Unif
(20, 30), Wmin~ Unif(38.4, 57.6), Wmax~ Unif (115.2,
172.8). The choice of ±20% was done here in
order to study the consequence of a mod-
erate uncertainty (the coefficient of varia-
tion of the uniform distribution was equal
to 28%) on the model output. Other choic-
es are of course possible.
Ten thousands parameter values were
randomly generated from the uniform dis-
tributions defined above by Monte Carlo
sampling. Due to an overlap of the distri-
butions of Topt and T max, a constraint on pa-
rameter values was considered at this step
in order to satisfy Tmax>Topt. The 10,000 corre-
sponding responses of W vs. T were comput-
ed and a sample of 20 out of the 10,000 re-
sponse curves was displayed in Figure 1B for
illustration. The distribution of the 10,000
response curves was summarized in Figure
1C by the percentiles 1%, 10%, 50% (medi-
an), 90%, 99%. The results showed that the
uncertainty was more important when the
temperature during the wetness period T
was close to 25°C i.e. the estimated optimal
temperature for the fungus (Figure 1C). The
distribution of W obtained for T=25°C was
skewed (Figure 1D); the median was equal to
Figure 1. Predicted wetness duration requirements for infection of bean foliage by Sclerotinia sclerotiorum. A: Predictions
obtained with the parameter values reported by Magarey et al. (2005). B: Sample of 20 response curves generated by Mon-
te Carlo simulation. C: Percentiles 1%, 10%, 50%, 90% and 99% of the 10,000 simulated wetness duration requirements in
function of the temperature. D: Distribution of the 10,000 simulated wetness duration requirements for T=25°C.