What is the resolution of the data?
Spatial resolution is 0.25d (1997-2001 1.0d to reflect larger uncertainty). Temporal resolution of the burned area and emissions is monthly, but we distribute those emissions over the days using active fire detections.
Differences with other datasets
Is GFED5 data compatible with data from the Global Fire Assimilation System (GFAS)?
No, GFAS is tuned to match GFED3, and GFAS emission estimates are therefore conservative as GFED5 estimates are on a global scale roughly 60% higher than GFED3
Does GFED use the same approach as the Fire INventory from NCAR (FINN) ?
Yes, to some degree. Both FINN and GFED estimate emissions using the Seiler and Crutzen (1980) approach, multiplying burned area with fuel consumption per unit burned area. The burned area in FINN is derived from active fires and scalars to go to burned area while the basis of GFED is mapped burned area. Fuel consumption in FINN is based on look-up tables and varies between fire types while in GFED these are based on a biogeochemical model and are calculated for each month and grid cell. FINN has a much higher spatial resolution (1km) compared to GFED (0.25d).
Uncertainties
How large is the uncertainty in GFED emissions?
Substantial but quantifying its actual uncertainty is difficult. Many top-down studies based on atmospheric carbon monoxide (CO) show that predicted atmospheric CO concentrations reasonably match observed ones when GFED is used as source of CO, but they also show that mismatches still exist. Over larger regions we expect uncertainties (2 sigma) to be 50% but they are larger for smaller regions and especially over regions where peatlands or deforestation are important sources of fire emissions. We do not expect a low bias in emissions anymore, something that was the case in earlier GFED versions.
How about uncertainties in aerosol emissions? Many models need to boost the fire emissions to match aerosol optical depth (AOD).
While many studies using CO as a constraint show GFED is providing reasonable estimates, those using AOD often indicate emissions are far too low. It is unlikely that the difference between the reasonable results using CO and underestiation when using AOD is related to emission factors. Simply boosting fire emissions to match AOD may obscure errors in other processes that are relevant in the emissions to AOD conversion (see for example
Zhong et al. (2023))
Climate change and fires
Are fires increasing due to climate change?
There are two trend related to fire that are well established. First, on a global scale, burned area is decreasing. This is mostly in grass-dominated vegetation types such as savannas which dominate global burned area numbers, and it is because of landscape fragmentation and an increase in the amount of land used for agriculture. Second, weather conditions are becoming more favorable for fires in many forested ecosystems due to climate change. Longer fire seasons, more lightning, and longer and more intense drought periods result in an increase in forest fires in many regions, most prominent in high latitudes. While globally burned area is declining, fire carbon emissions are relatively stable, and emissions of reduced gases are rising. This is because per unit burned area, forests have higher fuel loads and higher rates of emissions of reduced gases.
How important are CO2 emissions from fires?
All fires emit CO2 and with about 3 Gt of carbon per year fires emit the equivalent of 30% of fossil fuel emissions. In general, however, fire is thought to be carbon-neutral because regrowth compensates for those emissions. If there is no regrowth after a fire, for example because fires are used in the deforestation process, their CO2 emissions do add to the build-up of greenhouse gases. The same holds for fires in tropical peatlands which are often drained prohibiting regrowth. Finally, if fire frequencies increase, part of the fire CO2 emissions are not balanced anymore.