Amazonian forest wildfires are likely to become more frequent under a hotter and drier climate and could contribute significantly to Brazilian CO2 emissions. Currently, the annual estimates of CO2 emissions and removals from LULUCF in Brazil are estimated following IPCC guidelines by the environment ministry system (SIRENE). However, the SIRENE platform covering national and subnational emissions, does not include emissions from wildfires. This is a potentially important omission as previous studies have pointed to the non-recovery of carbon stocks of burned forests within 30 years in the Brazilian Amazon. Quantifying these emissions is not simple - computing changes in carbon stock from live and dead pools require a dynamic system modelling and repeated measurements from forest plots. Here we assessed the long-term carbon balance of burned forests and developed an approach considering that part of the carbon lost from fire is immediately emitted to the atmosphere, and another part is transferred to dead organic matter, being emitted slowly through decomposition. We developed a dynamic spatial-temporal model in Google Earth Engine that integrates biome-specific parameters, such as maps of carbon stocks, dead wood and litter combustion factors, tree mortality, decomposition and turnover rates to quantify net CO2 at a year-to-year basis with high spatial resolution burned area maps. Our model brings novelty for integrating two unique datasets — before-and-after fire censuses of 40 permanent forest plots in the Brazilian Amazon; and a high resolution burned area map (Mapbiomas Fire) ranging from 1990 to 2020. We estimated that in 30 years, burned forests in the Brazilian Amazon have emitted 1,274 Tg of CO2. These emissions stem only from fires in forests that remained up to 2020, e.g., it does not include burned forests that were deforested. Since 2009, forest wildfires emissions are equivalent to 41% of deforestation emissions. Our analysis indicates more burned forests are likely to remain in future years under a scenario of controlled deforestation — during the most recent period in our analysis (2012-2020), 47% of burned forests were not deforested but acted as a carbon source. Until now, there has been no spatially explicit method of incorporating forest fire emissions into Brazil’s carbon accounting from LULUCF and assigning those emissions to specific years. While our proposed approach is preliminary and requires refinement, we were able to apply it to the whole Amazon basin, revealing for the first time how fire emissions progress over time across across forest stands and fire regimes.
forest wildfires, emissions, carbon, dynamic modelling, google earth engine