Burned Area Mapping Using ΔBAI-Otsu from Landsat 8 Imagery in Bukit Anak Dara East Lombok

Abstrak

Forest and land fires are recurrent in Indonesian tropical mountain savannas and threaten biodiversity, carbon stocks, and local livelihoods, yet spatially explicit burned-area information is still limited. This study aimed to evaluate the performance of the Burn Area Index (BAI) from Landsat 8 OLI–TIRS imagery for mapping the 2024 fire in Bukit Anak Dara, East Lombok. Burned and unburned pixels were classified by applying a two-class Otsu threshold to the ΔBAI histogram for the full scene extent. The resulting burned-area map was validated against high-resolution polygons obtained from visual interpretation of Sentinel-2A imagery and against fire hotspots from the SiPongi+ system. Compared with Sentinel-2A polygons, the ΔBAI–Otsu method produced a burned-area estimate of 275.49 ha versus 318.87 ha from the reference and achieved an overall accuracy of 0.97, precision of 0.94, recall of 0.81, and an F1-score of 0.87. Validation against hotspot data yielded lower performance (overall accuracy 0.87, precision 0.40, recall 0.41, F1-score 0.41), reflecting conceptual and spatial-scale differences between point-based active-fire detections and patch-based burned-area mapping. Burned pixels were concentrated on west–northwest facing slopes dominated by dry savanna, highlighting the role of topography and fuel characteristics in fire spread. Overall, the results therefore indicate that the ΔBAI–Otsu approach is a rapid, transparent, and reproducible tool for post-fire burned-area mapping in tropical mountain ecosystems and has strong potential for routine operational monitoring.