COSMOS, a spatially explicit model for the epidemiology of banana weevil (Cosmopolites sordidus, Germar.)
Introduction
Fabrice Vinatier
1, Philippe Tixier
1, Christophe Le Page
2, Claude Bruchou
3, Pierre-François Duyck
1& Françoise Lescourret
41CIRAD, UPR 26, Martinique, France fabrice.vinatier@cirad.fr, 2CIRAD, UPR GREEN, Bangkok, Thailand, 3 INRA, UR 0546 BioSP, Avignon, France,4 INRA, UR 1115 PSH, Avignon, France
Distribution of attacks of C. sordidus
Understanding the main rules that explain the epidemiology of a pest is of primary importance to control it. Spatial dynamics of banana weevil (Cosmopolites sordidus) is greatly influenced by local movements and individual behavior. We present a stochastic Individual- Based Model (IBM) based on biological parameters of C. sordidus. COSMOS includes:
z egg-laying depending on density of adults and phenological stage of host-plant
zthermal development of immature stages of C. sordidus and its host-plant
z local dispersion by walking
Spatial dynamics of C. sordidus
Density of plants
0 10 20 30 40 50 60
0.00 0.01 0.02 0.03 0.04 0.05
0.06 Initialisationwith attack at end of first cropping cycle
Comparison between observation and simulation at the end of the second cropping cycle
---(Obs) __ (Sim)
Circumference of banana plant attacked (in cm)
Mean of attacks PE: Eggs predation MRL: mortality rate of larvae DE: density effect on egg-laying FH: fecundity of females G: diameter of burrow of larvae DC1,DC2: dispersion coefficients of adults
LM: longevity of adults Larva in its burrow
Measure of attacked circumference
Dispersal and demographic parameters contribute greatly to the explanation of attacks.
Model validation
for 40 banana plotsConclusions
Sensitivity analysis
(Morris method) on model parameters.Creating a mechanisticmodel necessitates to choose the best biological parameters that influence the epidemiology of a tropical pest.
Comparing model outputs to real data is a key step of the modeling process.
Sensitivity analysisis useful for categorizing the most influent parameters.
Scaling upthe model at field size and for long periods of time permits emergence of population properties (aggregation) resulting from individualbehaviour(local dispersion by walking).
Banana lines (in m) Simulation of spatial dynamics of C. sordidus starting
from an initial random distribution on a large plot Banana rows (in m)
Adults number per banana plant
Initialization
T=50 days
T=200 days
Aggregation of individuals, not always linked to the heterogeneity of resources, is well simulated by the model at
the field scale
0 cm
20 cm
40 cm
60 cm
The COSMOS model reproduces with a good accuracy the intensity and variability of attacks for
small plots.
PE
MRL
FH
G DE LM
µ*
σ
0.5 1.0 1.5
0.51.01.5
Delay between the two cropping cycles
1.
2. 3.
1.
2.
3.
Non linear effects and/or with interactions Insignificants effects
Linear effects
0 20
40 60
80 0
20 40
60 80 0
20 40
60 80
0
20
40
60
80 0
20
40
60
80 0
20
40
60
80 051015202530051015202530
051015202530
(in m)
(in m) (in m)
(in m)
µ* and σare respectively mean and standard deviation of absolute value of elementary effect of
each factor