Geostationary Himawari-9 Reveals Missing Agricultural Burning Outside LEO Satellite Detection Windows and Improves Burned-Area Estimates in Mainland Southeast Asia

Biomass burning emissions in Southeast Asia (SE Asia) remain highly uncertain due to limited ground data and persistent model-observation discrepancies. A major source of uncertainty arises from differences in how emission inventories estimate burned area. The Fire Inventory from the National Center for Atmospheric Research version 2.5 (FINNv2.5) relies on low Earth orbit (LEO) Terra/Aqua MODIS (1km) and SNPP VIIRS (375m) active-fire detections, assuming full-pixel burns. Because these sensors pass over the region only twice daily, they miss late-afternoon and evening agricultural fires that often dominate burning activity in this region. Himawari-9, a geostationary (GEO) satellite, provides hourly observations and captures diurnal burning patterns but at a coarser 2 km resolution. To leverage this temporal coverage, we compared burned areas from Himawari-9 and SNPP VIIRS during their overlapping daytime overpass (~12-2 pm local time) and derived a corrected Himawari-9 pixel area of 0.14 km², equivalent to a VIIRS pixel. This correction enables a combined GEO-LEO approach that retains VIIRS-level spatial detail. We integrated Himawari-9 detections with corrected pixel sizes into FINNv2.5, removed tropical duplicate-pixel routine which can inflate burned-area estimates, particularly given MODIS’s coarse 1-km pixel size, and developed additional configurations using NOAA-20 and NOAA-21 VIIRS. Across February-March 2024, the configuration using Himawari-9 and all VIIRS sensors performed best when evaluated against independent burned-area datasets (GISTDA LANDSAT8/9 for Thailand, Copernicus Sentinel-3, and NASA VNP64A1). Relative to default FINNv2.5, this configuration detected substantially more small distinct fires: +73-319% in croplands, +59-113% in grass/savanna, and +59-93% in woody shrubs across Vietnam, Laos, Myanmar, Thailand, and Cambodia. Against March 2024 GISTDA burned areas in Thailand, correlations improved across crop types, with the largest gains in rice (r = 0.29 to 0.41) and sugarcane (r = 0.19 to 0.31). These results demonstrate how combining high-temporal GEO fire detections with fine-spatial VIIRS observations can enhance fire detection and strengthen emission inventories in SE Asia and beyond.