Abstract
Wildfires pose a significant threat to ecosystems, wildlife, and humancommunities, leading to habitat destruction, pollutant emissions, andbiodiversity loss. Accurate wildfire risk prediction is crucial for mitigatingthese impacts and safeguarding both environmental and human health. This paperprovides a comprehensive review of wildfire risk prediction methodologies, witha particular focus on deep learning approaches combined with remote sensing. Webegin by defining wildfire risk and summarizing the geographical distributionof related studies. In terms of data, we analyze key predictive features,including fuel characteristics, meteorological and climatic conditions,socioeconomic factors, topography, and hydrology, while also reviewing publiclyavailable wildfire prediction datasets derived from remote sensing.Additionally, we emphasize the importance of feature collinearity assessmentand model interpretability to improve the understanding of prediction outcomes.Regarding methodology, we classify deep learning models into three primarycategories: time-series forecasting, image segmentation, and spatiotemporalprediction, and further discuss methods for converting model outputs into riskclassifications or probability-adjusted predictions. Finally, we identify thekey challenges and limitations of current wildfire-risk prediction models andoutline several research opportunities. These include integrating diverseremote sensing data, developing multimodal models, designing morecomputationally efficient architectures, and incorporating cross-disciplinarymethods--such as coupling with numerical weather-prediction models--to enhancethe accuracy and robustness of wildfire-risk assessments.