Fire from policy, human interventions, or biophysical factors? Temporal-spatial patterns of forest fire in southwestern China

Abstract: Forest fire is a recurrent and serious problem which stirs major environmental, economic, and social concerns. Information on the temporal-spatial distribution of ignition probability and identifying the factors that drive the pattern of forest fire occurrences are the most challenging issues over the world. In this study, the spatial distribution and temporal variation in forest fire occurrence were examined by applying the Maximum Entropy (Maxent) algorithm. Redundancy analyses were used to determine the contributions of driving factors (forest fire policy, human interventions, and biophysical variables) to the forest fire occurrences. In addition, the strength of the relationship between the forest fire policy, human interventions, and biophysical variables and their direct and indirect effects on forest fire occurrences were determined using structural equation models (SEMs). Our results showed the Maxent model was suitable for predictive modeling of the distribution of forest fire occurrences in southwestern China. The predicted area of forest fire occurrences in southwestern China during 2000-2017 was 3.25 x 10(5) km(2) which mostly concentrated in the southern region. There was an increasing trend of forest fire crossing national and provincial borders. According to the temporal pattern of forest fire occurrences, additional ignition probability areas including the central Yunnan and the eastern Sichuan were found in spring when compared to those in winter. Moreover, forest fire policy is the main driver underlying the ignition, followed by biophysical factors and human interventions. These should be considered as key elements in the design of forest fire prevention programmes. Although the explanatory power of driving factors varied, national border distance (NBD) and conductive index (CI) that represented effects of forest fire policy were constantly the key driving variables. In addition, there is heterogeneity among the biophysical factors. The meteorological factors significantly affected on ignition probability in different temporal scale, while the effects of topographic factors were smaller. Human interventions had no significant direct effect on ignition probability, while they via fire policy had a significant positive indirect effect on the probability of fire occurrence. The results suggest that we should strengthen the administrative capacities to prevent forest fires, and improve the fire prevention publicity and education on the residents living near the national border and/or in low population density areas to get higher awareness and compliance of forest fire policies for local residents