Developing 30 m resolution monthly burned area products in Africa (2017–2025)

Accurate burned area (BA) mapping is essential for quantifying wildfire extent, estimating carbon emissions, and assessing ecosystems, particularly in sub-Saharan Africa (SSA), which contributes the largest portion of BA and fire-related carbon emissions in the globe. However, there is still a lack of multi-year high spatial resolution (≥30 m) monthly BA products for SSA. To address this issue, this study presents a novel optical and SAR data coupling framework for producing multi-year 30 m resolution monthly BA in SSA from integrated Landsat 8/9, Sentinel-2, and Sentinel-1 data. We evaluated pre-/post-fire and change metrics including thermal infrared, red-edge, and SAR-derived indicators to select optimal features, input them into a random forest model to compute burned probabilities. Following the refinement of the initial probability map via SAR-based BA detection, soft suppression of cropland misclassifications, active fire confirmation, and vegetation filtering, region growing was performed to delineate the final BA map. Leveraging this framework, the novel 30 m resolution Monthly Burned Area products for SSA from 2017 to 2025 (MBA-SSA1725) were generated and can be freely downloaded. Validation against two existing reference datasets and a specifically designed reference dataset (RBA-SSA1725) showed Dice Coefficient (DC) of 0.8462, 0.8614 and 0.8751, respectively, with lower omissions and higher precision compared to the existing BA products. We further analyzed the spatiotemporal patterns of BA in SSA, including seasonal fluctuations, and interannual variations. This work fills critical gaps in multi-year, high-resolution monthly BA monitoring for Africa, which is important for enhancing fire-induced carbon emissions estimation, trend evaluation, and ecosystem assessment in the most fire-prone region on the earth.