Forest Fire Environmental Drivers and Predictive Modeling Approaches

Abstract

This article focuses on the risks of wildfires driven by heat waves, droughts, and human activities, and points out the research gap in the absence of transferable and interpretable early warning guidelines in data-scarce areas and urban-rural interface zones (WUI). Through a structured literature review and cross-regional comparison, we comprehensively evaluated the forest types and fuel structures in Southeast Asia, Europe, and the Americas, and examined the prediction methods ranging from process simulators (FARSITE, Prometheus) to machine learning pipelines. We proposed a normative framework that integrates spatial risk mapping with threshold-based early warnings. The results show that humidity and wind speed thresholds are common triggers for large fires; fuel continuity, peatlands, and WUI expansion significantly amplify spread and exposure; mixed modeling (using mechanism model outputs and multimodal data for training ML) can enhance prediction skills and transferability. Based on this, this article proposes an implementable “P4” path (perception - preparation - prediction - pre-processing) to support earlier and more operational risk management through integration of sensing, modeling, and operation.