Among forest disturbance agents, insect pests are known to be some of the most destructive, causing outbreaks that result in widespread tree damage and mortality with cascading ecological impacts to forest structure and biogeochemical cycles. Due to their critical role in disturbance regimes and their increasing threat to forests under novel climate conditions, there is considerable interest in both quantifying the impacts of forest pests and forecasting their future dynamics. A crucial step in developing generalized frameworks for understanding and forecasting plant-pest interactions is teasing apart the factors that influence pest disturbance magnitudes following outbreaks. While studies of pest outbreaks often take place on the ground, field-based approaches rarely capture the full extent and severity of pest disturbance events due to their ephemeral nature and heterogeneous distributions across the landscape. To address this challenge, this project leverages data from a recently developed Landsat-based forest canopy condition monitoring product to assess the spatial heterogeneity in disturbance magnitudes following the recent outbreak of a non-indigenous moth species, Lymantria dispar (formerly European gypsy moth), across southern New England. We aimed to identify the environmental and anthropogenic drivers of spatial heterogeneity in L. dispar outbreak severity at a regional scale.
We tested the relationships between a suite of environmental variables and L. dispar disturbance magnitudes at 5000 sites across our study area. We also assessed the impacts of human-related drivers, such as population densities, on L. dispar defoliation severities by testing relationships between disturbance magnitudes and anthropogenic variables at those sites Our covariates were incorporated into a pixel-level Bayesian state-space time series model to generate forecasts of L. dispar disturbance magnitude at individual points in southern New England. The outcome of our model is a forecast of summer 2022 forest canopy condition for 5000 individual sites across our study area. These results include a discrete outbreak probability for each site followed by a forecast of recovery rate for the following growing seasons which charts the return to the pre-disturbance greeness value in each pixel. Additionally, our results showed a positive relationship between disturbance magnitude and recovery rates post-outbreak. Therefore, this modeling framework also lays the foundation for our next step, which focuses on quantifying and forecasting the recovery phase of L. dispar outbreaks.