Résumé
Fires across sub-Saharan Africa (SSA) are a dominant source of global carbon emissions, yet their true magnitude remains uncertain due to the limitations of coarse-resolution satellite products. In this study, we developed a high-resolution fire emission inventory prototype for SSA using active fire detections from the VIIRS sensor (375 m) and a top-down approach based on fire radiative power (FRP). Emissions were estimated through the integration of FRP to fire radiative energy (FRE), conversion to dry matter burned using biome-specific combustion coefficients, and application of emission factors for carbon dioxide. A parallel MODIS-based dataset was also produced using the same methodology to isolate sensor-specific effects. To evaluate detection and modelling differences, the VIIRS-based product (VIIRS-EM) was compared against six widely used global fire emission inventories. In addition, a subset of emissions from small fires (defined as FRP < 10 MW) was derived and assessed separately. Over the period 2013-2022, VIIRS-EM estimated average annual carbon emissions of 3.0 Pg C, which is 50-75 % higher than most MODIS-based inventories. Emission hotspots were identified in agricultural and savanna regions, particularly in West and Central Africa. Small fires contributed significantly to early and late fire-season emissions and revealed widespread underestimation in existing products. Our findings underscore the importance of high-resolution detection and FRP-based modelling for capturing the full extent of African fire activity. The VIIRS-EM inventory provides improved spatial and temporal resolution, with implications for atmospheric composition modelling, greenhouse gas accounting, and regional fire policy development.