GFED Home

Introduction

Fires are an important source of atmospheric trace gases and aerosols and represent the most important disturbance agent at the global scale. In addition, deforestation, tropical peatland fires, and areas experiencing increasing fire frequency contribute to the accumulation of atmospheric CO₂.

The Global Fire Emissions Database (GFED) project quantifies global fire extent and emissions. GFED combines satellite observations of fire activity with information on vegetation productivity to estimate gridded monthly burned area and fire emissions, as well as scaling factors that can be used to derive emissions at higher temporal resolution. The resulting datasets are available for download from this website and are intended for use in large-scale atmospheric and biogeochemical studies. The data are updated daily and can also be used to assess the current fire situation across larger regions and compare current fire activity with previous years.

Fraction of each grid cell that burned annually, averaged over the 2002–2025 period. Red areas represent regions that burn frequently, such as savannas and agricultural areas, where fires primarily consume grasses and crop residues. In total, approximately 750 million hectares burn each year, an area roughly equivalent to the size of Australia.

Average annual fire carbon emissions for each grid cell, averaged over the 2002–2025 period. Although the spatial pattern is similar to that of the burned area shown above, regions with high fuel loads, such as forests and peatlands, stand out more clearly because emissions per unit of burned area are substantially higher.

Methodology and key datasets

Fire emissions in GFED are calculated as the product of burned area, fuel consumption, and emission factors. Fuel consumption is determined by the total fuel load or biomass and the fraction of that biomass that is actually consumed by fire, commonly referred to as combustion completeness. Emission factors are used to convert the amount of combusted biomass into emissions of CO₂ and other greenhouse gases, aerosols, and a range of atmospheric constituents.

Burned area

Burned area for all vegetation types except croplands is based on Chen et al. (2023). This burned area dataset builds on the MODIS MCD64A1 Collection 6 burned area product developed by Giglio et al. (2018), while accounting for errors of omission and commission. Burned area for croplands is derived from the GloCAB dataset of Hall et al. (2023).

Fuel consumption

Fuel consumption, expressed as carbon emissions per square meter of burned area, is based on Van Wees et al. (2022).

Emission factors

The resulting carbon losses are used to estimate emissions of trace gases and aerosols using emission factors. For savannas, and for CO₂, CO, CH₄, and N₂O only, emission factors are based on Vernooij et al. (2023). For other biomes, as well as for other species in savannas, the NEIVA (Next-generation Emissions InVentory Expansion of Akagi) emission factors are used; see Binte Shahid et al. (2024). For the boreal region, and for CO and CH₄, these data have been further expanded using the tower-based study of Wiggins et al. (2021).

Emissions

Emissions of carbon, trace gases, and aerosols are calculated by combining these three components, as described in Van der Werf et al. (2025). This study covers the core GFED period from 1997 to 2022. For the period from 2023 onwards, active fire detections are directly scaled to burned area and emissions, following the approach described by Chen et al. (2026).

Time period covered and spatial detail

The current version is GFED5, which provides data at a spatial resolution of 0.25 degrees for the period from 2001 onwards. Data for 1997–2000 are provided at a spatial resolution of 1.0 degree, reflecting the greater uncertainties associated with this earlier period.

For the period from 2023 onwards, emissions are based on VIIRS active fire data, using relationships between VIIRS active fire detections, burned area, and emissions established during the overlapping period. Data for 2023 onwards are currently provided as a beta version and remain subject to change until the accompanying paper has been accepted for publication.