Supervised Machine Learning Approaches on Multispectral Remote Sensing Data for a Combined Detection of Fire and Burned Area

Bushfires pose a severe risk, among others, to humans, wildlife, and infrastructures. Rapid detection of fires is crucial for fire-extinguishing activities rescue missions. Besides, mapping burned areas also supports evacuation accessibility emergency facilities. In this study, we propose generic approach detecting based on machine learning (ML) approaches remote sensing data. While most studies investigated either the or areas, addressed evaluated, in particular, combined three selected case study regions. Multispectral Sentinel-2 images represent input data supervised ML models. First, generated reference target classes, burned, unburned, fire, since no were available. Second, regional fire datasets preprocessed divided into training, validation, test subsets according defined schema. Furthermore, an undersampling ensured balancing datasets. Third, seven classification used including tree-based models, self-organizing map, artificial neural network, one-dimensional convolutional network (1D-CNN). All achieved satisfying results. Moreover, they performed highly accurate detection, while separating unburned was slightly more challenging. The 1D-CNN extremely randomized tree best-performing models with overall accuracy score 98% subsets. Even unknown dataset, high accuracies. This generalization even valuable any use-case scenario, organization fire-fighting civil protection. proposed could be extended enhanced crowdsourced further studies.