Monitoring open landfill fires using integrated UAV and Sentinel-2 satellite imagery

Landfill fires pose severe environmental and public health risks, particularly in developing regions where waste accumulation is poorly managed. This study develops a multi-sensor machine-learning model for monitoring open landfill burning, using the 2023 Sarimukti Landfill fire in West Bandung, Indonesia, as a case study. UAV imagery and Sentinel-2 data were integrated into map fire-affected zones using Random Forest (RF) and Gradient Tree Boosting (GTB) algorithms. The model achieved high classification performance, with overall accuracies of 80.55% and 81.31% for two UAV datasets and high cross-validated AUC values that were comparable across classifiers and across the UAV-only and integrated inputs. A fire-mitigation priority model was then developed by combining the Burned Severity Index (BSI) and a Digital Surface Model (DSM) to evaluate the influence of elevation on fire persistence. The current study is limited to a single landfill site; future work should extend the model to multiple landfill locations and incorporate additional variables, such as thermal and meteorological data, to improve generalizability. These findings highlight the potential of UAV-satellite integration and machine learning for rapid fire monitoring and risk assessment in waste-management systems.