A model of drought-related tree mortality based on in situ observations and remote sensing data

Climate projections suggest the world's forests will experience changes in the extent, duration, or intensity of drought conditions. A large number of drought-induced tree mortality events worldwide are reported in an expanding body of literature investigating these phenomena. Quaking aspen (Populus tremuloides) occupies one of the largest ranges of any tree in North America and has been the subject of multiple drought-induced mortality studies. As such, it is a suitable species to investigate how predisposing, inciting, and contributing factors vary over a large region and to then draw comparisons with other studies regarding the drivers of mortality. Using a multi-model inference framework with linear mixed effects models, we assessed quaking aspen mortality in Michigan, Minnesota, Wisconsin, Utah, and Colorado. Attributes for more than 70,000 trees were obtained from the Forest Inventory and Analysis (FIA) program and were combined with data from the U.S. General Soil map and a satellite-derived drought index. We compared the impact of a number of factors (e.g., stand age and relative density) to the impact of drought using log-odds probability. Results indicate variability in the drought index provides at least as much explanatory power as the mean or minimum of the same and confirm that drought is part of a complex of factors that lead to mortality. Future work will focus on predicting areas of high mortality based on ground observations routinely collected by the FIA program and satellite-based drought information.