Top-down carbon monoxide fire emissions over South America correlated with global climate indices

South America (SA) has suffered a multitude of extreme, drought-induced fires in recent years, including 2024 which saw fire emissions across the continent 263 Tg C (84%) above average^[1]. Burned area and fire carbon emissions in SA are projected to increase over the coming decades due to higher temperatures and drier conditions associated with climate change^[2]. These effects are already being seen in the Amazon, where fire is driving the rainforest towards being a net carbon source^[3] and threatening existential climate tipping points. Meanwhile to the South, the ecologically diverse Pantanal wetlands have undergone a step-change in wildfire activity, with 2019–2021 experiencing a 408% increase in annual carbon monoxide (CO) emissions relative to the 2013–2018 average.

CO is a major trace gas released from fires. Its emissions can be used to quantify wildfire carbon impacts and investigate correlations between fire activity and global climate indices. Despite this, there remains considerable disagreement between fire inventory products, with mean annual CO emissions ranging from 284–625 Tg yr^-1 globally, and predictions diverging further at smaller spatial scales. These large uncertainties originate from the underlying assumptions of the inventory methodologies and the imperfect sensitivity of their satellite data inputs. Satellite observations of atmospheric total column CO, combined with inverse modelling techniques, provide a direct, top-down method to constrain these estimates, allowing more accurate CO emissions to be determined.

Here, we derive fire emission estimates between 2019–2024 for SA using the INVICAT 4D-Var inverse chemical transport model, assimilating TROPOspheric Monitoring Instrument (TROPOMI) total column CO satellite observations into the model for the first time. Six fire inventories (GFEDv4.1s, GFEDv5.1, GFASv1.2, QFEDv2.6r1, FINNv1.5, FINNv2.5) are used as priors in separate CO inversions, from which posterior result sensitivity is quantified and prior biases are assessed. We use emission ratios to determine, spatially and temporally, the total carbon flux into the atmosphere from fires in SA. We find that the 2024 extreme fire season in SA is poorly captured by the fire inventory products currently available, with peak atmospheric CO over SA observed to be 12.8 Tg (46%) larger than forward-modelled inventory emissions predict. Additionally, we create a multilinear regression model to predict the spatial distribution of CO anomalies across tropical SA by correlating the inversion posterior emissions to key global climate indices at various lag times. This novel method can provide spatial forecasts of the wildfire vulnerability arising from the global state of the climate months in advance.

 

[1] Kelley et al., 2025, State of Wildfires 2024–2025, Earth System Science Data

[2] Burton et al., 2021, South American fires and their impacts on ecosystems increase with continued emissions, Climate Resilience and Sustainability

[3] Basso et al., 2022, Atmospheric CO₂ inversion reveals the Amazon as a minor carbon source caused by fire emissions, with forest uptake offsetting about half of these emissions, Atmospheric Chemistry and Physics