Metapopulation modelling and area-wide pest management strategies evaluation. An application to the Pine processionary moth

•A method to rank area-wide pest management strategies is proposed.•A spatially explicit metapopulation model is used to predict pest dynamics.•The distance of predictions from a reference state is measured.•Strategies are ranked according to this distance.•Sensitivity analysis allows robust decision making.

Forecasting pest population abundance is a time and resource consuming task, and in particular for area-wide pest management is complicated by demographic and environmental stochasticity. These factors make difficult the development of quantitative tools to design and evaluate different management strategies performances by taking into account various form of variability and uncertainty. Pest management could benefit from methods supporting decision making based on models ease of development under scarce data and high uncertainty. Host plants for many agricultural and forest pests are often patchily distributed, therefore population dynamics can be suitably described in terms of metapopulations. Despite the fact that metapopulation models were originally proposed for pests, they remain a widely used tool in conservation biology but receive little attention in large scale pest management.The aim of this paper is to propose a framework allowing the ranking of the efficacy of area-wide pest control strategies, taking into account population spatial distribution in discrete patches. The Kullback–Leibler divergence, well known in Information Theory, Probability and Statistics, is used to measure how far the state of the metapopulation as predicted by a spatially explicit metapopulation model is from a suitable reference state.The method is applied to compare the efficacy of different types of predefined control strategies of the Pine processionary moth (Traumatocampa pityocampa (Den. and Schiff)). The analysis of a dataset on metapopulation dynamics of this moth from a fragmented Mediterranean pine forest allows to derive some rules of thumb for the rational allocation of control effort, in terms of spatial and temporal distribution of the interventions.