Atmospheric CO 2 inversion reveals the Amazon as a minor carbon source caused by fire emissions, with forest uptake offsetting about half of these emissions

Tropical forests such as the Amazonian rainforests play an important role for climate, are large carbon stores and are a treasure of biodiversity. Amazonian forests are being exposed to large scale deforestation and degradation for many decades which declined between 2005 and 2012 but more recently has again increased with similar rates as in the 2007/2008. The resulting forest fragments are exposed to substantially elevated temperatures in an already warming world. These changes are expected to affect the forests and an important diagnostic of their health and sensitivity to climate variation is their carbon balance. In a recent study based on CO2 atmospheric vertical profile observations between 2010 and 2018, and an air column budgeting technique to estimate fluxes, we reported the Amazon region as a carbon source to the atmosphere, mainly due to fire emissions. Instead of an air column budgeting technique, we use here an inverse of the global atmospheric transport model, TOMCAT, to assimilate CO2 observations from Amazon vertical profiles and global flask measurements. We thus estimate inter- and intra-annual variability in the carbon fluxes, trends over time and controls for the period 2010-2018. This represents the longest Bayesian inversion of these atmospheric CO2 profile observations to date. Our analyses indicate that the Amazon is a small net source of carbon to the atmosphere (mean 2010-2018 = 0.13 +/- 0.17 PgC y(-1), where 0.17 is the 1-sigma uncertainty), with the majority of the emissions coming from the eastern region (77 % of total Amazon emission). Fire is the primary driver of the Amazonian source (0.26 +/- 0.13 PgC y(-1)), however the forest uptake likely removes around half of the fire emissions to the atmosphere (-0.13 +/- 0.20 PgC y(-1)). The largest net carbon sink was observed in the western-central Amazon region (72 % of the fire emissions). We find larger carbon emissions during the extreme drought years (such as 2010, 2015 and 2016), correlated with increases in temperature, cumulative water deficit and burned area. Despite the increase in total carbon emissions during drought years, we do not observe a significant trend over time in our carbon total, fire and net biome exchange estimates between 2010 and 2018. Our analysis thus cannot provide clear evidence for a weakening of the carbon uptake by Amazonian tropical forests.