Forecasting Burned Areas of Wildfires: A Case Study of Mae Hong Son Province in Thailand

Wildfire prediction plays a critical role in preventing and managing forest fires, which pose a significant threat to ecosystems, economies, and human lives. This study focused on developing accurate prediction models to estimate the total burned area, especially during wildfire events. The research explored various approaches, including feature engineering, machine learning algorithms such as regression tree and multiple linear regression, and deep learning models like Long Short-Term Memory (LSTM) networks. The dataset included information on forest fire areas and relevant climatic conditions collected from Mae Hong Son province in Thailand. The models were trained, fine-tuned, and evaluated using performance metrics such as mean square error (MSE). The results demonstrated the effectiveness of the LSTM model with external features in predicting the burned areas, particularly on fire occurrences. Overall, the prediction of burned area in wildfires holds significant potential for supporting decision-making processes, resource allocation, and proactive measures to minimize the impact of wildfires.