Boundary-Aware Uncertainty Quantification for Wildfire Spread Prediction
For practitioners of wildfire spread prediction, this work provides a more operationally relevant evaluation protocol for uncertainty quantification, but the results are incremental as the student model matches rather than exceeds ensemble performance.
The paper introduces the Fire-Centered Evaluation Region (FCER) framework for evaluating uncertainty quantification in wildfire spread prediction, and shows that a distilled single-pass student model achieves comparable calibration and complementary uncertainty ranking to an ensemble model on the WildfireSpreadTS dataset.
Reliable wildfire spread prediction is vital for risk-aware emergency planning, yet most deep learning models lack principled uncertainty quantification (UQ). Further, for boundary-sensitive cases like wildfire spread, evaluating models with global metrics alone is often insufficient. To shift the focus of UQ evaluation toward a more operationally relevant approach, the Fire-Centered Evaluation Region (FCER) framework is introduced as a spatially conditioned protocol to characterize UQ within critical fire zones. Using FCER, an Ensemble is compared against an distilled single-pass student model on the WildfireSpreadTS dataset. The student model demonstrates comparable calibration and complementary uncertainty ranking in boundary-relevant regimes. Code is available at https://github. com/jonasvilhofunk/WildfireUQ-FCER