Abstract
Integrating factors affecting the success of invasive insect pests into dynamical models can help assessing their invasion risks and control. Here, we model the spread of a gall-forming hymenopteran parasite of chestnut trees, Dryocosmus kuriphilus, and its control agent, Torymus sinensis, across 23 natural forest sites located in the French Eastern Pyrenees. The integration of field estimates of the levels of bottom-up (frequency, density and genetic susceptibility of chestnut trees) and top-down (hyperparasitism by native insects and fungi) control of the pest in a Nicholson-Bailey model allowed to identify source and sink sites for the invasive species and its control agent. Comparisons with the observed levels of hyperparasitism by T. sinensis showed that it was found in 7/23 sink sites. The extension of our modelling into a two-site model showed that dispersal rates as low as 1‰ can be responsible for the persistence of T. sinensis in sinks, regardless of the precise dynamical regime of D. kuriphilus-T. sinensis coexistence in the source. Although dispersal promotes the persistence of the control agent and tends to homogenize its effectiveness in both sites, it was also shown to reduce the global biological control effectiveness at high rates of coupling.