Article révisé par les pairs
Résumé : Insect pest monitoring can be seen as a resource allocation problem under sensing capacity constraints, as it is impossible to inspect the entire field at each sampling time. Accordingly, the main question in entomological monitoring is where should data be collected to maximise the information on the infestation level. This study aimed to analyse and propose a solution to this problem by combining physiologically-based models and state observers. These algorithms compare the simulated and the real output of the system (described by the model and measurements, respectively), and provide a more reliable estimation of its state. Notably, we explored the combination of a metapopulation physiologically-based model described by Ordinary Differential Equations (ODEs) with an Extended Kalman Filter (EKF). We set up an ad hoc optimisation problem which provide, as solution, the parcels where data should be collected. The framework hereby introduced minimises the overall uncertainty associated with the state of the outbreak by providing the optimal allocation of the resources available (maximum number of parcels that can be inspected at each sampling time). The performance of the proposed technique, denoted as “smart sampling”, has been compared in silico, through simulations, with two static and one dynamic sampling strategies commonly applied in entomology, with particular reference to visual inspections. Results showed that the dynamic sampling strategy based on the EKF (the smart strategy) had a higher performance with respect the traditional ones. This work provided relevant theoretical insights that can improve data collection protocols in entomology, and that can be further extended to other case studies beyond insects.

Documents en relation

DI-fusion