Présentation orale au congrès conjoint SFE² GfÖ EEF 2022 (Metz, France) : Un appel à des stratégies de surveillance mixtes pour contenir l'invasion du scarabée japonais en Europe

International audienceThe Japanese beetle (Popillia japonica) is a polyphagous pest that spreads rapidly and is estimated to cost more than 460 M$/year in damages and control in the USA alone. Thus, it is listed as a priority pest according to the phytosanitary legislation of the European Union. This study’s goal was to build risk maps to inform surveillance strategies in continental Europe following the beetle’s introduction and successive spread in the last decade. We developed a species distribution model (SDM) using a machine-learning algorithm, considering factors relevant to the beetle’s biology, climate, land cover and human-related variables. This analysis was carried out at a fine spatial resolution (4x4 km); using presence-only data from native and invaded ranges (Japan, North America, Azores archipelago - Portugal). We gathered more than 30 000 presence data from citizen science platforms and standardized surveys. In order to balance presence-only data, we generated pseudo-absences using the target-group method (Coleoptera) to account for the sampling bias that is inherently associated with citizen-science data. Then we structured these data into environmental blocks and randomly sampled pseudo-absences within blocks using environmental distance to presence sites. Model training and performance evaluation have been performed with cross-validation using these environmental blocks. Our results show that climate, in particular seasonal trends, and human-related variables such as population and roads densities, are major drivers of the Japanese beetle distribution at the global scale. Risk maps show that Central Europe can be considered as suitable, whereas Southern and Northern European countries seem to be at lower risk. Furthermore, the currently occupied region in northern Italy and southern Switzerland is amongst the most suitable according to model predictions and represents less than 1% of the highly suitable area in Europe, which, overall, covers more than 182 thousand km2. A major cluster of high suitability areas is located near the currently infested zone, whereas others are scattered across the continent. This highlights the importance of designing hybrid surveillance strategies considering both active insect dispersal and the possibility of hitchhiking to reach distant areas