Examining Spatial Variation in the Effects of Japanese Red Pine (Pinus densiflora) on Burn Severity Using Geographically Weighted Regression

Abstract

Burn severity has profound impacts on the response of post-fire forest ecosystems to fire events. Numerous previous studies have reported that burn severity is determined by variables such as meteorological conditions, pre-fire forest structure, and fuel characteristics. An underlying assumption of these studies was the constant effects of environmental variables on burn severity over space, and these analyses therefore did not consider the spatial dimension. This study examined spatial variation in the effects of Japanese red pine (Pinus densiflora) on burn severity. Specifically, this study investigated the presence of spatially varying relationships between Japanese red pine and burn severity due to changes in slope and elevation. We estimated conventional ordinary least squares (OLS) and geographically weighted regression (GWR) models and compared them using three criteria; the coefficients of determination (R2), Akaike information criterion for small samples (AICc), and Moran’s I-value. The GWR model performed considerably better than the OLS model in explaining variation in burn severity. The results provided strong evidence that the effect of Japanese red pine on burn severity was not constant but varied spatially. Elevation was a significant factor in the variation in the effects of Japanese red pine on burn severity. The influence of red pine on burn severity was considerably higher in low-elevation areas but became less important than the other variables in high-elevation areas. The results of this study can be applied to location-specific strategies for forest managers and can be adopted to improve fire simulation models to more realistically mimic the nature of fire behavior.