Modeling wildfire drivers in Chinese tropical forest ecosystems using global logistic regression and geographically weighted logistic regression

Abstract

The tropics is an area with high incidence of wildfire all over the world in recent years, and the forest ecosystem in the tropics is extremely fragile. Thus, it is very important to identify drivers of wildfire in the tropics for developing effective fire management strategy. In this paper, global logistic regression (GLR) and geographically weighted regression (GWLR) models were employed to analyze the spatial distribution and drivers of tropical wildfires in Xishuangbanna and Leizhou Peninsula in tropical China from 2001 to 2018. The results show that the overall distribution of wildfire in Xishuangbanna and Leizhou Peninsula from 2001 to 2018 was spatially aggregated. In these tropical seasonal forest ecosystems, wildfire was mainly driven by meteorological factors, particularly by daily temperature range and precipitation. In Xishuangbanna (inland) peninsula, the impact of driving factors tended to be global, and the GLR model predicted the probability of wildfire occurrence better than the GWLR model. Drivers of wildfire in Leizhou Peninsula (coastal area) had clear spatial variation, and the GWLR model better explained the relationship. Furthermore, wildfire in Leizhou was more driven by human activities, especially management of agricultural lands. Our results demonstrate that effective forest management practice needs to adopt fire management practices with regional characteristics. The forest management strategy and traditional agriculture production system should pay more attention to changes in these driving factors and their relationship with wildfire.