Custom Loss Functions in Fuel Moisture Modeling
Jonathon Hirschi
TL;DR
The paper investigates whether custom loss functions that emphasize accuracy on drier fuels can improve wildfire ROS forecasts derived from FMC predictions. It compares 12 loss schemes (including exponentially weighted and ROS-based weights) across three static ML models (Linear Regression, Random Forest, XGBoost) using spatiotemporal cross-validation on RAWS data, then converts FMC forecasts to ROS via a 3 m/s wind curve. Results show only modest overall gains in ROS RMSE, with stronger improvements during the driest period (ROS-weighted loss up to ~4.7% reduction; exponential weight ~3.4%). The work demonstrates that simple, physically motivated loss weighting can yield practical gains for ROS forecasting, though further validation with recursive models and broader datasets is needed to confirm real-time wildfire simulation benefits.
Abstract
Fuel moisture content (FMC) is a key predictor for wildfire rate of spread (ROS). Machine learning models of FMC are being used more in recent years, augmenting or replacing traditional physics-based approaches. Wildfire rate of spread (ROS) has a highly nonlinear relationship with FMC, where small differences in dry fuels lead to large differences in ROS. In this study, custom loss functions that place more weight on dry fuels were examined with a variety of machine learning models of FMC. The models were evaluated with a spatiotemporal cross-validation procedure to examine whether the custom loss functions led to more accurate forecasts of ROS. Results show that the custom loss functions improved accuracy for ROS forecasts by a small amount. Further research would be needed to establish whether the improvement in ROS forecasts leads to more accurate real-time wildfire simulations.