Forest Fire Forecasting Leveraging Modis Satellite Fire Data Using Machine Learning For Jharkhand State, India

Forest fires pose significant environmental and economic threats, necessitating robust predictive models for better management and prevention. This study focuses on the temporal and spatial forecasting of forest fires using machine learning (ML) techniques, leveraging historical fire incident data from the Jharkhand state of India, spanning the period from 2001 to 2022, and fire radiative power (FRP) as key indicators. Two distinct approaches are explored for forecasting Forest fires: one based on the frequency of forest fire occurrences (FRO) and the other using Fire Radiative Power FRP as a parameter of fire intensity over a 72-month period using machine learning techniques. Both approaches involve data preprocessing, including monthly aggregation and dividing the dataset into training and testing, followed by the application of various ML regression models. The models are evaluated using metrics such as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), Mean Absolute Percentage Error (MAPE), and R-squared ($\mathrm{R}^{2}$), and it is observed that in both approaches, ensemble methods like Gradient Boosting outperformed simpler models, demonstrating lower error rates. The findings of this work underscore the importance of advanced machine learning techniques in accurately predicting forest fire occurrences and intensities, offering valuable insights for future fire management and mitigation strategies.