Abstract
Abstract
Forest fires pose significant environmental and socioeconomic risks, increasing the need for reliable methods for spatial mapping, susceptibility assessment, and risk prediction. This systematic review synthesizes research on the application of remote sensing, Geographic Information Systems (GIS), and machine learning for forest fire mapping. A total of 202 articles were initially identified from five academic sources, of which 51 studies met the predefined inclusion criteria. The reviewed literature shows increasing research activity, with publications spanning 2021 to 2026. Asia accounted for the largest proportion of studies (62.7%, n=32), while Africa was represented by only four studies, indicating a geographical research gap. Among remote sensing datasets, MODIS was most frequently used (27 studies), followed by Sentinel-2 (14), while VIIRS and Landsat 8 were each used in six studies. Machine learning approaches dominated the analytical methods, with Random Forest being the most frequently applied model (28 studies), followed by XGBoost (20), Support Vector Machine (13), and Logistic Regression (12). Common predictor variables included temperature, precipitation, slope, aspect, elevation, distance to roads and settlements, land use/land cover, and vegetation indices. NDVI was the most frequently used spectral index, appearing in 43 studies. Key limitations included spatial resolution, data quality, data availability, and model performance. The review demonstrates a clear shift toward integrated GIS, remote sensing, and machine learning frameworks, including cloud-based platforms such as Google Earth Engine. Future research should prioritize improved spatial and temporal data resolution, broader geographical coverage, particularly in Africa, integration of multiple datasets and seasonal risk indices, advanced deep learning methods, continuous monitoring, and robust validation to reduce bias and improve model transferability. These advances can strengthen forest fire management and support evidence-based decision-making.