Comparative analysis of burn area between google earth engine and manual digitization using the NBR algorithm

Abstract

Background: Indonesia, as the third-largest tropical forest country in the world, is experiencing significant forest degradation driven by illegal logging, land-use conversion, and recurrent wildfires. Peatland ecosystems, particularly in Kubu Raya District, West Kalimantan, are highly susceptible to fire due to their organic-rich composition and seasonal desiccation. This study aims to assess the spatial distribution and severity of forest and land fires in Kubu Raya from 2019 to 2023 using remote sensing and geographic information system (GIS) techniques. Methods: Hotspot data from the Fire Information for Resource Management System (FIRMS) MODIS were analyzed to determine fire occurrences, while Sentinel-2 imagery was utilized to calculate the Normalized Burn Ratio (NBR) index for burn severity estimation. Image analysis was conducted using both manual digitization and the Google Earth Engine (GEE) platform to compare accuracy, efficiency, and spatial representation of burned-area detection. Findings: The findings indicated that 2023 recorded the largest burned area, covering 832,188.98 ha, predominantly within peatland zones. Accuracy assessment demonstrated that the GEE-based method achieved higher reliability, with overall accuracy and kappa statistic values of 86% and 74%, respectively, outperforming the manual approach. The spatial distribution of fire hotspots revealed that peat-dominated areas were more vulnerable to large-scale fires due to their hydrological characteristics. Conclusion: The results highlight that GEE provides a rapid, consistent, and accurate technique for burn area detection and fire severity analysis. Integrating cloud-based remote sensing with conventional GIS enhances monitoring capabilities for sustainable peatland management. Novelty/Originality of this article: The novelty of this research lies in its comparative accuracy evaluation between automated and manual burn area mapping. This study provides new methodological insights for fire monitoring across Indonesia’s tropical peatlands, demonstrating the advantages of cloud-based platforms for large-scale environmental assessments.