Landsat-based fire maps reveal higher fire emissions from larger area of low-severity burnings than coarse resolution data in Southeast Asia

Abstract

Accurately describing the distribution of burned area and fire emissions is essential for fire management and sustainable development. Medium-resolution datasets, such as MCD64A1 Version 6.1 and the Global Fire Emissions Database (GFED), indicate a decrease in tropical and subtropical fire activity over the past 20 years. However, these datasets often miss small burned patches, leading to underestimations in heterogeneous landscapes. Southeast Asia (SEA) is a major hotspot for fire activity and biodiversity. This study provides a high-resolution (30 m) dataset of burned area (1990–2023) and fire-related carbon emissions (2001–2023) across SEA, derived from Landsat data. We curated training data from MODIS, VIIRS, and Landsat imagery, used a Random Forest algorithm to predict burned areas, and created monthly burned area maps with 30-meter resolution. Additionally, we simulated fire-related carbon emissions using the GFED framework and high-resolution maps. Our findings reveal an average annual burned area of 125,212 km² and emissions of 270 Tg C, both significantly exceeding previous estimates from coarse-resolution datasets by 11,771 km² and 83.67 Tg C, respectively. Both burned area and carbon emissions have remained stable over time, with agricultural and forest fires contributing significantly to the total burned areas and emissions. Forest burning accounts for 49 % of the total burned area, leading to substantial forest loss (12 % of total forest loss) and carbon emissions (76 % of total fire emissions), with an increase since 2001. Agricultural burning contributes 28 % of the burned area and 12 % of carbon emissions in SEA. These results emphasize the importance of high-resolution monitoring for improving fire prediction and management in fragmented landscapes.