Modeling macroscale patterns of forest productivity in United States forests using geographically weighted regression

Keywords: net primary productivity, forest management, climate change, non-stationarity spatially explicit model
Abstract Type: Paper Abstract

Abstract

Little is known about the relationship between management approaches and forest productivity at the macroscale. Using annual modeled estimates from ~400 random selected 10 km x 10 km grid cells in the U.S. Southeast (SE) and Pacific Northwest (PNW) for the period 2000-2016, we used geographically weighted regression (GWR) to estimate NPP and NEE as a function of soil, topography, climate, and management practices. These regions together contain 32% of the U.S. forested lands, are managed by a suite of different forest practices across a variety of environmental drivers, and are experiencing climate change in different ways. GWR was selected, as it is a spatially explicit method that accounts for spatial non-stationarity, allowing for relationships between variables to change over space. Results indicated that the responses of NPP and NEE to environmental variables in PNW were more complex than those of SE. Soil, topography, and management were not as strongly correlated with NPP and NEE in the SE. In contrast, they had a key role in the PNW forests. Furthermore, the magnitude and direction of the impacts of seasonal climate variables varied spatially, such that summer and spring precipitation were more important in predicting NEE and NPP in SE versus PNW. The findings of this study may provide a useful framework for forest management climate change strategies that could be extended across the continental U.S.