Pre-Ignition Wildfire Burn Severity Modeling with Polarimetric UAVSAR Data and Machine Learning

Wildfire occurrence and the increase in their destruction and severity due to climate change has made the prediction, mapping, and techniques of mitigation a very timely area of research. Remote sensing data is used post-fire to guide vegetation planning, map burn severity, and create dynamic models for predicting fire spread. Synthetic Aperture Radar data shows potential to contribute to these estimates compared to the limitations of traditional optical-based models. This research presents a model of burn severity prediction based on pre-fire static variables with correlations to burn severity using data available pre-ignition for insight of wildfire propagation within a known ignition perimeter.