An Innovative Investigation on Predicting Forest Fire Using Machine Learning Approach

Abstract

Predicting forest fire occurrences can bolster early detection capabilities and improve early warning systems and responses. Currently, forest and grassland fire prevention and suppression efforts in China face significant hurdles due to the complex interplay of natural and societal factors. While existing models for predicting forest fire occurrences typically consider factors like vegetation, topography, weather conditions, and human activities, the moisture content of forest fuels is a critical aspect closely linked to fire occurrences. Additionally, it introduces forest fuel-related factors, including vegetation canopy water content and evapotranspiration from the top of the vegetation canopy, to construct a comprehensive database for predicting forest fire occurrences. Furthermore, the study develops a forest fire occurrence prediction model using machine learning techniques such as the random forest model (RF), gradient boosting decision tree model (GBDT), and adaptive augmentation model (AdaBoost).