Modeling the probability of bark beetle-caused tree mortality as a function of watershed-scale host species presence and basal area

Abstract

In recent decades, bark beetle outbreaks have caused mass tree mortality in western US forests, which has led to altered wildfire characteristics, hydrological processes, and forest carbon dynamics. Understanding spatial variability in forest susceptibility to bark beetle outbreaks in the western US could inform strategic forest management to reduce wildfire risk, manage forest carbon, and plan for altered hydrology. The susceptibility of a forest stand to mortality by bark beetles depends on the availability and characteristics of trees of the host tree species. For multiple bark beetle species that affect coniferous forests in the western US, field studies have demonstrated that outbreak probability and severity increase with the basal area of host trees. However, field plot data with information on species composition and basal area are not available in a spatially complete manner across the western US. In this study, we estimated susceptibility to bark beetle-caused tree mortality based on stand characteristics in coniferous forest watersheds in the western US. We used a dataset of forest inventory plot data imputed to all forested area in the western US at 30 m resolution to calculate host basal area and model its relationship to the probability and amount of bark beetle-caused mortality for eight of the most damaging bark beetle species in the western US in recent decades. To estimate tree mortality from individual bark beetle species, we used raster datasets quantifying the proportion of area with mortality by each bark beetle species from USDA Forest Service Aerial Detection Surveys. We evaluated model performance on a temporally-separated validation dataset and calculated the area under the receiver operator curve (AUC) and correlation between predicted and observed probability (AUC) and amount (correlation) of mortality. We found that models using bark beetle species-specific host basal area and average climate conditions were able to predict the probability of mortality, with AUC values ranging from 0.721 to 0.952, but had a limited ability to predict the amount of mortality. Our results identify forests in the western US with characteristics that lead to a higher susceptibility to bark beetle outbreak. Our analysis included less well-studied species that have caused significant mortality in drought-influenced forests in recent years, specifically fir engraver and western pine beetle, and revealed that large forested areas in the western US are susceptible to mortality by these two species.