Assessment of deep learning models integrated with weather and environmental variables for wildfire spread prediction and a case study of the 2023 Maui fires

2025-09-05

arXiv.org

AI Summary

Wildfire spread prediction remains challenging due to complex spatiotemporal dynamics and limitations of both physics-based and data-driven models. Method: This study systematically evaluates five deep learning models—including ConvLSTM and its attention-enhanced variants—for wildfire spread forecasting using a multimodal spatiotemporal dataset integrating fire progression, wind speed, humidity, topography, and other environmental variables; it further applies explainable AI (XAI) to identify dominant drivers and conducts the first quantitative benchmark against the physics-based FARSITE model on the 2023 Maui wildfire. Contribution/Results: ConvLSTM-based models demonstrate superior input flexibility and environmental interpretability over FARSITE, whereas FARSITE achieves marginally higher overall accuracy and F1 score. The work establishes a novel cross-paradigm evaluation framework bridging AI and physical modeling, clarifying the applicability boundaries and enhancement pathways for data-driven wildfire forecasting, thereby advancing trustworthy, interpretable, and hybrid wildfire prediction paradigms.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision
• Humans and AI: Explainable AI (XAI) for Human Understanding

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web search models and ranking

Abstract

Predicting the spread of wildfires is essential for effective fire management and risk assessment. With the fast advancements of artificial intelligence (AI), various deep learning models have been developed and utilized for wildfire spread prediction. However, there is limited understanding of the advantages and limitations of these models, and it is also unclear how deep learning-based fire spread models can be compared with existing non-AI fire models. In this work, we assess the ability of five typical deep learning models integrated with weather and environmental variables for wildfire spread prediction based on over ten years of wildfire data in the state of Hawaii. We further use the 2023 Maui fires as a case study to compare the best deep learning models with a widely-used fire spread model, FARSITE. The results show that two deep learning models, i.e., ConvLSTM and ConvLSTM with attention, perform the best among the five tested AI models. FARSITE shows higher precision, lower recall, and higher F1-score than the best AI models, while the AI models offer higher flexibility for the input data. By integrating AI models with an explainable AI method, we further identify important weather and environmental factors associated with the 2023 Maui wildfires.

Problem

• Evaluating deep learning models for wildfire spread prediction accuracy
• Comparing AI models with traditional fire spread models like FARSITE
• Identifying key weather and environmental factors in wildfire prediction

Innovation

• Integrating weather and environmental variables into deep learning models
• Comparing ConvLSTM and attention-based models with traditional FARSITE
• Using explainable AI to identify key wildfire factors