Environment-Adaptive Preference Optimization for Wildfire Prediction

TL;DR

The paper introduces EAPO, a framework for adapting wildfire prediction models to changing environments by constructing distribution-aligned datasets and fine-tuning to improve detection of rare, high-impact events.

Contribution

EAPO is a novel method that combines distribution alignment and hybrid fine-tuning to enhance model robustness under environmental shifts in wildfire prediction.

Findings

EAPO achieves ROC-AUC of 0.7310 on wildfire prediction.

EAPO improves detection of extreme wildfire events.

EAPO maintains performance across environmental shifts.

Abstract

Predicting rare extreme events such as wildfires from meteorological data requires models that remain reliable under evolving environmental conditions. This problem is inherently long-tailed: wildfire events are rare but high-impact, while most observations correspond to non-fire conditions, causing standard learning objectives to underemphasize the minority class (fire) that matters most. In addition, models trained on historical distributions often fail under distribution shifts, exhibiting degraded performance in new environments. To this end, we propose Environment-Adaptive Preference Optimization (EAPO), a framework that adapts prediction to the target environment with long-tail distribution. Given a new input distribution, we first construct distribution-aligned datasets via $k$-nearest neighbor retrieval. We then perform a hybrid fine-tuning procedure on this local manifold, combining supervised learning with preference optimization, as well as emphasizing on rare extreme events. EAPO refines decision boundaries while avoiding conflicting signals from heterogeneous training data. We evaluate EAPO on a real-world wildfire prediction task with environmental shifts. EAPO achieves robust performance (ROC-AUC 0.7310) and improves detection in extreme regimes, demonstrating its effectiveness in dynamic wildfire prediction systems.