Wildfire Carbon Models Miss Underground Peat Fires by Up to a Factor of 14

From space, the most dangerous wildfires often look unremarkable. A camera pointed at a boreal landscape in northern Sweden may show a low-lying smolder with little visible flame - nothing like the dramatic orange fronts that dominate wildfire coverage. Yet beneath the visible surface, organic soils can burn for weeks or years, consuming material that has been accumulating since before written history. Each ton of that ancient carbon, once released, adds to the atmospheric burden that climate models try to track.

A study published in Science Advances by Johan Eckdahl, a postdoctoral scholar at UC Berkeley's Energy and Resources Group, and colleagues, reconstructed the carbon emissions from 324 wildfires that burned in Sweden in 2018 and compared those reconstructions against six of the most widely used global fire emissions models. The comparison revealed systematic errors - in both directions, but with a pattern that has clear implications for how much carbon boreal fires are actually releasing.

How the reconstruction was done

The team combined detailed national forest inventory data with field measurements at 50 of the burned sites - 19 from high-intensity fires and 31 from low-intensity fires. At each site, they measured the depth of the organic soil layer, which ranges from a few inches to several feet in boreal forests, and collected soil samples. Comparing carbon content between burned and unburned soils at matched sites allowed direct calculation of how much carbon the fire consumed at each location. This bottom-up approach is more labor-intensive than satellite-based methods but captures what satellites cannot see: what is happening underground.

Where the models go wrong - and by how much

The comparison between the field-based reconstruction and the global models showed large inaccuracies in both directions, depending on the type of fire.

In Gavleborg county, where high-intensity fires burned through drier forests with dramatic visible flames, the models tended to overestimate emissions - the satellite-visible fire looked bigger than it actually was in terms of total carbon consumed.

In the neighboring county of Dalarna, low-intensity fires that were barely visible from satellites burned into thick organic soil layers. There, the models underestimated carbon emissions by up to a factor of 14. The smoldering underground fire was largely invisible to the algorithms that global carbon budgets depend on.

"Many of the fires that matter most for the climate don't look dramatic from space," Eckdahl said. "Peatlands and organic soils can smolder for weeks to years, releasing enormous amounts of ancient carbon."

Why the boreal problem is larger than Sweden

Sweden is a useful study system because it has detailed national forest data and accessible fire sites. But it is small relative to the boreal and subarctic regions where peat and organic soils are most extensive. Siberia and Canada contain orders of magnitude more carbon stored in these soils, and both have experienced extreme fire seasons in recent years. The measurement infrastructure needed to do what Eckdahl's team did in Sweden barely exists in those regions.

"Sweden is a very large country, but it's quite small compared to Siberia and Canada," Eckdahl said. "We may be severely underestimating the impact of the recent extreme fire seasons in these regions."

If the factor-of-14 underestimate in Dalarna's low-intensity peat fires reflects something systematic rather than a local anomaly, global boreal wildfire carbon budgets could be substantially wrong - in a direction that makes the climate situation worse than current accounting suggests.

The soil biology question and next steps

Eckdahl is now part of the Western Fire and Forest Collaborative, working with UC Berkeley colleagues to apply similar approaches to fire-prone forests in the western United States. Those forests lack the extensive peat layers of boreal systems, but soil carbon variability driven by local climate, vegetation type, and microbial communities still influences fire emissions in ways that global models do not fully capture. His specific focus will be on soil bacteria and fungi and how they shape post-fire carbon recovery - a question with direct relevance to whether burned landscapes recover carbon over decades or remain net sources.

Co-authors on the Swedish study include Lars Nieradzik of Lund University and Louise Rutting of Brandenburg University of Technology.