Burned area and fire emissions according to the fifth version of the Global Fire Emissions Database (GFED)

Quantifying burned area and associated fire emissions is paramount to understand how changing fire patterns affect radiative forcing and air quality. It is now well established that many fires are too small to be detected by coarse resolution satellite burned area products on which the Global Fire Emissions Database (GFED) relied. In the fifth version of GFED (GFED5) we therefore combine burned area derived from mapped coarse-resolution burned area from the MODIS sensor -which excels in detecting larger fires- with small-fire burned area. The latter is derived from MODIS active fire detections scaled to burned area using ratios constrained by higher-resolution burned area datasets from Landsat and Sentinel-2 for selected regions. Burned area in cropland regions was based on the Global Cropland Area Burned (GloCAB) dataset. Total global burned area is 61% higher than in GFED4s. We converted burned area to emissions using a simplified version of the CASA model used in previous GFED versions, but which now runs at a 500 m spatial resolution. This allows for better constrained modeled fuel loads based on field measurements. Although GFED5 emissions are aggregated to a 0.25 degree grid due to the statistical nature of deriving our burned area, we can now account for heterogeneity in fire processes within these large pixels. Emissions (3 Pg carbon per year) are roughly 50% higher than in GFED4 and we show how diverging trends in grassland versus forest ecosystems impact trends in total emissions. Finally, we show how converting fire carbon losses to trace gas and aerosol emissions is now better constrained due to the addition of several new emission factor measurement campaigns. In the savanna biome we now account for spatial and temporal variability in emission factors.