Gabriela Nunez-Mir

Aim Biological invasions are an escalating environmental challenge due to their substantial ecological and socio‐economic consequences. Accurate near‐term forecasts of future areas occupied by an invasive species could enhance the efficiency and efficacy of invasion monitoring and management but spread forecasting models have been developed and tested for few invasive species thus far. Location Northeastern USA. Methods We developed a quantitative model to forecast 1‐year‐ahead occupancy of Lymantria dispar dispar , an expanding invasive forest insect introduced into the eastern United States in 1869 that causes large‐scale defoliation of forests. We validated and tested the model using historical distribution and density data from a large‐scale network of pheromone‐baited traps. We first assessed how forecast accuracy depended on trap catch density thresholds for determining occupancy and on the spatial scale of local connectivity measures. Next, we tested how increasing the computational complexity and biological detail encoded in the models affected prediction accuracy. Results Models using lower occupancy thresholds and measuring connectivity over shorter distances tended to perform best. A simple model using only coarse and generic representations of habitat suitability and local diffusive spread potential performed at least as well as more complex models. Our best models achieved total accuracy and true‐positive rates exceeding 95%. Conclusions Our baseline model illustrates the utility of a mechanistic model to forecast year‐to‐year occupancy dynamics of L. dispar. Near‐term spread forecasting can be a valuable tool for invasion management, even for species without detailed a priori ecological knowledge.