Modeling species distributions using remote sensing data: the Eastern temperate forests of the US as a case study

Biodiversity is changing at rates only comparable to the major extinction events recorded in geologic history. Feasible approaches to monitor biodiversity at different spatial and temporal scales are urgently necessary. Development of these approaches are critical to advancing towards international biodiversity goals and for developing effective management strategies and conservation actions to face ongoing global changes. The use of remote sensing data coupled with in-situ biodiversity data has become the most comprehensive solution for monitoring components of biodiversity across the Earth surface. Predicting where species can occur under changing environments provides an important baseline for monitoring biodiversity. Most studies that aim to predict species distributions have focused on the use of interpolated climate surfaces as covariates, and these are widely used in macroecological studies. Much less work has been done exploring the use of remote sensing data as covariates to predict species distributions, even though variables derived from remote sensing frequently allow discrimination of local features and biotic components of the environment not captured by climatic variables. We modeled the distribution of forest species using satellite remote sensing products as covariates, drawing on empirical examples within ecologically and economically important tree lineages. Models were implemented in a fully Bayesian framework and evaluated using occurrences retrieved from the National Ecological Observatory Network (NEON) and the Forest Inventory and Analysis (FIA) Program of the U.S. Forest Service. Our empirical study provides insights for the prediction of species distributions and advance the construction of next-generation species distribution models.