Predicting Wildfire Susceptibility in Napa County, California using Machine Learning

Abstract

Wildfires have long been a part of the natural environment, however through climate change and increased human activity, they have become a significant problem to both humans and wildland. Stopping the expansion of wildfires would be critical in mitigating the dangerous outcomes of them. Firefighters stopping the spread of wildfires must know which parts of the environment are most vulnerable to the spread of wildfires, and vegetation is one of the key determining factors in the wildfire susceptibility of a given area. Previous works have used several different machine learning algorithms for the purpose of determining wildfire susceptibility. The algorithm used in this study for wildfire susceptibility prediction is a random forest applied to a vegetation dataset of Napa County, California provided by the California Department of Fish and Wildlife (CDFW). The random forest works by creating a set of decision trees to get an overall probability for each vegetation area. The model has a 91.7% accuracy in predicting wildfire burn probability in a vegetation area.