Amazon wildfire emissions up to three times higher than estimated

Fires are a recurring phenomenon in central South America, often intensified by drought and deforestation. In 2024, wildfire activity reached its highest levels in 20 years, affecting vast areas of the Amazon rainforest and the Cerrado – the world’s most biodiverse tropical savanna, stretching across one-fifth of Brazil and extending into Bolivia and Paraguay.

A paper, published in Geophysical Research Letters, used artificial intelligence (AI) to analyse satellite observations of carbon monoxide during the August–September 2024 fire season. Scientists used this gas as an indicator of carbon dioxide emissions, combining the satellite data with wildfire models.

The results suggest that current scientific methods significantly underestimate carbon emissions, with actual carbon output potentially between 1.5 and three times higher.

This has important implications for climate models and global carbon budgets, which rely on accurate wildfire emission estimates.

Why current estimates may be too low

The research, funded by ESA, was led by the Technical University of Dresden in cooperation with the Royal Netherlands Meteorological Institute (KNMI) and BeZero Carbon, the London-based carbon ratings agency. It also found that prolonged smouldering was a significant contributing factor to carbon emissions during 2024.

Jos de Laat, senior scientist at KNMI and lead-author, said, “We studied an area of around 4 million sq km, where the most intense fires and pollution were concentrated near the Brazil-Bolivia border. This had severe air quality impacts across the region.

“We found significant gaps between modelled and observed levels of air pollution. Current methods cannot reproduce what satellites are actually seeing, suggesting that important sources of emissions are being missed.”

To address this, researchers trained an AI system to accelerate their advanced emission calculations. This made it possible to analyse multiple years and regions despite the high computational demands.

The research also combined data from several Sentinel missions (Sentinel-2, Sentinel-3 as well as Sentinel-5P) to improve both the estimation and evaluation of wildfire emissions, noting that the synergy of these instruments is crucial for progress.

Carbon monoxide as a proxy for carbon dioxide

Carbon monoxide is a colourless, odourless and toxic gas that is released when organic matter such as vegetation burns incompletely. Carbon dioxide, on the other hand, is the biggest contributor to anthropogenic emissions. Smoke plumes from wildfires contain both gases.

However, carbon monoxide is easier for satellites to detect than carbon dioxide, making it a useful proxy for estimating wildfire emissions.

While carbon dioxide is the main greenhouse gas, it is already naturally present in the atmosphere at high and almost constant concentrations (around 430 parts per million, or ppm) making small changes difficult to detect from space – a bit like trying to see a piece of white paper on snow. Carbon monoxide, in contrast, naturally exists at much lower levels (less than 0.2 ppm) and is much more variable, so increases are easier to spot, more like looking for a piece of white paper against a dark background.

ESA research improves wildfire emission estimates

The research into wildfires in the Cerrado and the Amazon rainforest was part of the Sense4Fire international research project, which is funded by ESA. The project investigates conditions that make wildfire ignition more likely and aims to improve estimates of carbon emissions generated by flames and smouldering embers.

The research uses a wide range of satellite data, including Sentinel satellites and additional sources. It works on increasing the scientific understanding of fire dynamics and their role in the carbon cycle by integrating observations from the Sentinels into new Earth observation datasets and models. The Sense4Fire project applies advanced techniques, harnessing more complex computational analysis as well as richer data about vegetation, made possible by advanced remote-sensing satellite instruments.

Stephen Plummer, ESA Earth Observation Applications Scientist, noted “The findings from this paper raise questions on how we are calculating carbon emissions from fire, and in particular CO2, a primary greenhouse gas and driver of climate warming. Earth observing satellites such as the Sentinels are contributing to increasingly accurate datasets, giving us a far clearer understanding of how our Earth system is reacting and evolving. The contribution from space provides an important benchmark to assess global carbon and climate models which are vital for underpinning the climate policies that decision-makers need to put in place.”

Copernicus Sentinel-5P tracks wildfire pollution from space

Sentinel-5P, which was launched in October 2017, was the first Copernicus mission dedicated to monitoring the atmosphere. Its state-of-the-art spectrometer, Tropomi, measures trace gases such as nitrogen dioxide, ozone, formaldehyde, sulphur dioxide, methane, carbon monoxide and aerosols. It provides daily global coverage with unprecedented spatial resolution. This makes Sentinel-5P exceptionally well-suited to measure carbon monoxide. Thanks to its much finer spatial resolution combined with much better detectors, Tropomi has advanced the measuring and monitoring of air pollution.

Tropomi’s observations include detailed information about vegetation, fuel moisture and surface conditions, enabling more accurate emission estimates compared to traditional methods, which relied mainly on burned area and fire radiative power.

KNMI’s Jos de Laat noted, “The methodologies and data we created during this project will be integrated into upcoming European Horizon projects and the Copernicus Atmospheric Monitoring Service (CAMS), ensuring broader application and continued development.”