Predicting fire occurrence patterns with logistic regression in Heilongjiang Province, China

Abstract

Prediction of forest fire ignition may aid in forest fire vigilance and monitoring, and in prioritizing forest fuel treatments. In this paper, we chose easily obtained spatial variables pertaining to topography, vegetation types, meteorological conditions, climate, and human activity to predict forest fire ignition in Heilongjiang province, China, using logistic regression. Results showed fire ignition prediction through logistic regression had good accuracy. Climatic variables (e.g., average annual mean temperature and precipitation) and meteorological conditions (e.g., daily minimum temperature, daily minimum humidity, daily mean humidity, and mean wind speed) are the main determinants of natural forest fires. In the case of anthropogenic fires, vegetation types and human activity as indicated by distances to roads and settlements combined with suitable meteorological conditions (e.g., daily mean humidity) are the main driving factors. The fire ignition probability map can be easily used to prioritize areas for vigilance, to make decisions on allocating firefighting resources, and to select vulnerable spots for forest fuel treatments. It was found that forest fuel treatments should be focused on the Great Xing’an Mountains.