Initial development of an iterative near-term forecast of Lymantria dispar moth defoliation events and post-disturbance canopy recovery.
Monday, August 2, 2021
Link To Share This Presentation: https://cdmcd.co/v9Djxx
Charlotte Malmborg and Michael Dietze, Boston University, Jaclyn Matthes and Valerie Pasquarella, Biological Sciences, Wellesley College, Wellesley, MA, Audrey A. Barker Plotkin, Department of Environmental Conservation, University of Massachusetts, Amherst, MA, Richard MacLean, Massachusetts Dept. of Conservation & Recreation
Background/Question/Methods Lymantria dispar is a generalist, foliage-feeding insect pest species that has become one of the most damaging invasive pests in northeastern forests since its introduction in the mid-nineteenth century. L.dispar currently only occupies about one quarter of its potential invasive range. However, its rate of spread is projected to increase over the coming years as a result of changing climate conditions and increasing rates of human-mediated dispersal through trade and transportation pathways. Consequently, understanding the future impacts of this invasive forest pest is a high priority for researchers and managers of forest resources. Of particular interest in recent years is the development of reliable near-term forecasts of pest outbreaks in space and time, as well as rates of ecosystem recovery, to gain a deeper understanding of uncertainties that underlie invasive pest dynamics. In this study, we used remotely-sensed data of change in forest condition over time as a proxy for observing L.dispar population dynamics during eruptive events. From these data, we conducted time series analyses to construct informative priors for a Bayesian modeling framework for L.dispar forecasts. Results/Conclusions We used growing-season Landsat observations of changing forest conditions from April through September between 1995 and 2020 to conduct time series analyses of L.dispar outbreaks at randomly sampled sites across Massachusetts and Connecticut. These observations notably include the particularly severe defoliation events that took place across southern New England between 2015 and 2017. Bayesian state space models developed from these forest condition metrics included a discrete outbreak probability prediction followed by a forecast of recovery rate during subsequent years. Hindcasts of the remote sensing data detected the 2015-2017 outbreaks with a calibrated canopy rate of recovery on the order of two to three years following the major change in forest condition. Our model was validated using L.dispar population data from field samples taken at the Quabbin Reservoir and Harvard Forest. Moving forward we aim to improve models by including a broader suite of environmental covariates, enabling us to better capture spatial heterogeneity in outbreak risk and recovery rates. We will also report preliminary results from an automated outbreak and recovery forecast for summer 2021.