Set Prediction for Next-Day Active Fire Forecasting

2026-05-11

AI Summary

This study addresses the challenge of accurately forecasting localized wildfire ignitions at sub-kilometer scales, a task poorly served by existing methods that output fire probabilities on coarse kilometer-scale grids. To this end, we reformulate next-day wildfire prediction as a high-resolution (375 m) point-set prediction problem and propose WISP—the first end-to-end, query-based architecture to apply sparse set prediction to wildfire forecasting. WISP integrates multi-source covariates, including 48-hour meteorological, vegetation, land surface, and historical fire data, and introduces an asymmetric classification–localization matching strategy with dynamic loss weighting to effectively coordinate query assignment, ranking, and activation. Evaluated on a newly constructed global hourly benchmark, WISP achieves an average precision of 38.2% in cross-regional testing, captures 53.4% of fire radiative power–weighted fire clusters, and successfully localizes 54.1% of observed fire clusters within 5 km.

Abstract

Accurate next-day active fire forecasts can support early warning, disaster response, forest risk assessment, and downstream estimation of fire-related carbon emissions. Existing machine learning approaches to wildfire forecasting typically predict wildfire danger or fire probability on kilometre-scale daily grids, which is useful for regional warning but does not directly represent localized fire events. We propose Wildfire Ignition Set Predictor (WISP), a query-based model that reformulates next-day active fire forecasting as point-set prediction. From 48 hours of covariates including meteorology, satellite vegetation products, static land, and fire history, WISP predicts a fixed-size ranked set of future active fire cluster centres on a 375 m grid across globally distributed regions. The model is trained end-to-end with Hungarian matching; to address the conflicting roles of the classification score in assignment, ranking, and query activation, we use asymmetric classification-localization weighting in matching and loss. We further construct a globally distributed, hourly, multi-source benchmark for this task. On a held-out test set spanning fire regions worldwide, the best WISP variant achieves 38.2% average precision (AP) for ranked fire-centre detections, covers 53.4% of fire cluster mass weighted by fire radiative power (FRP), and localizes 54.1% of observed clusters within 5 km. These results establish sparse set prediction as a viable formulation for high-resolution wildfire forecasting and provide a benchmark for future work in this regime.

Problem

• active fire forecasting
• point-set prediction
• wildfire ignition
• high-resolution prediction
• fire cluster localization

Innovation

• set prediction
• wildfire forecasting
• query-based model
• Hungarian matching
• asymmetric weighting