Improved Model of the Carbon Cycle Can Help Verify Reported Emissions

Researchers at UC San Diego’s Scripps Institution of Oceanography have created a more accurate model of global carbon cycling.

The model better accounts for the contributions of Earth’s terrestrial ecosystems to atmospheric concentrations of the greenhouse gas carbon dioxide, a major source of uncertainty for scientists tallying global emissions.

The model’s improved accuracy could help humanity monitor and verify reported carbon emissions from the burning of fossil fuels at a time when tracking the sources of human-caused emissions is becoming a prominent component of the Paris Climate Agreement.

The new model can verify reported global emissions over a five-year timespan with half the uncertainty of another widely used model. That five-year timespan is especially relevant to the Paris Agreement because of a scheduled assessment called the global stocktake. Every five years, the stocktake evaluates the world's progress on lowering greenhouse gas emissions and other facets of addressing climate change. The first global stocktake is now in progress.

“Cutting the error in half helps us see if human-caused emissions are really going down or up,” said Ralph Keeling, study co-author and Scripps climate scientist.

The study, published today in Nature Climate Change and supported by the National Science Foundation, the Schmidt Futures program, and by Earth Networks (an AEM brand), improved a simple model’s performance by better accounting for the spatially and seasonally varying influences of temperature on land-based carbon cycling. The model accomplished this by incorporating the findings of previously published research that yielded a global map establishing how short-term temperature fluctuations impact the carbon balance of land ecosystems.

“The effect of temperature on carbon cycling on land depends on the season and it depends on where you are on Earth,” said Benjamin Birner, the study’s lead author and a former postdoctoral researcher at Scripps. “For example, a plant might not respond well to being super hot in the summer, but in spring it might grow better with more warmth.”

The model’s strong performance also suggests an unexpected stability in land ecosystems’ response to rising temperatures as climate change progresses. That runs counter to an influential 2014 paper which used similar atmospheric data to argue that tropical ecosystems had become more sensitive to temperature fluctuations in recent decades.

Refining terrestrial ecosystems’ piece of the carbon puzzle is particularly impactful because it fluctuates from year to year and has historically entailed the greatest level of uncertainty compared to the planet’s other three main carbon repositories (or “sinks”) and sources. Those three other main variables in Earth’s carbon budget are the oceans, changes in land use or land cover such as deforestation, and emissions from human activities such as burning fossil fuels and cement production.

Previously when researchers have tried to add up the contributions of these four sources and sinks (land use change, the oceans, land-based ecosystems, and fossil fuels and industry) to equal the measured changes in atmospheric carbon dioxide levels, there have been up to 760 million tons of carbon, equivalent to roughly 8% of current global fossil-fuel emissions, left unaccounted for. Researchers call this discrepancy the carbon budget imbalance. 

Until now, there have been two main types of models trying to explain year-to-year variations in exchanges of carbon between the land biosphere and the atmosphere.

The first kind is what Birner called “a relatively simple statistical approach” that aims to capture how the long-term build up of atmospheric carbon dioxide and important climate variables such as temperature, moisture, or the El Niño-Southern Oscillation (ENSO) impact the amount of carbon stored in the land biosphere. For the purposes of comparison, the study focused on models that factor in ENSO and tropical average land temperature anomalies.

The second kind of model is heavily featured by the Global Carbon Project (GCP), an international scientific effort that produces a new whole-Earth carbon budget annually. GCP combines results from highly complex computer programs which dynamically simulate global vegetation and all the associated biogeochemical and hydrological processes that eventually give rise to the concentrations of greenhouse gases scientists observe in the atmosphere.

The problem is that both of these approaches to balancing the carbon budget still leave large unexplained residuals. The ENSO-based and GCP models both leave roughly plus or minus 750 million tons of carbon unaccounted for at the year-to-year scale. At the decadal scale, the carbon budget imbalance is about plus or minus 2.4 gigatons of carbon for the ENSO model and 3.6 gigatons for the GCP model.

These imbalances make it more challenging to accurately verify reported carbon emissions from the world’s countries and harder to understand how Earth’s systems are responding to rising temperatures.

To improve on these models, Keeling and Birner had the idea of leveraging the work of Christian Rödenbeck, a climate researcher at the Max Planck Institute for Biogeochemistry and co-author of the paper.

Rödenbeck developed a way to derive how temperature at different places and times of year affects the amount of carbon land ecosystems store or release into the atmosphere. He used detailed records of atmospheric concentrations of carbon dioxide and a database of temperatures from 196 measurement stations around the world. His research produced a global map of the seasonal temperature sensitivities of land carbon cycling for different locations.

Birner and Keeling used Rödenbeck’s map to inform a statistical model that is very similar to the ENSO model of carbon cycling except that it substitutes Rödenbeck’s seasonal temperature sensitivities into the equation. The team then used the record of atmospheric carbon dioxide concentrations measured at Hawaii’s Mauna Loa and the South Pole between 1958 and 2021 to assess the model’s performance.

The new model outperformed both the simple ENSO model and GCP’s complex dynamic vegetation model. At the interannual scale, the new model left 500 million tons of carbon unaccounted for at the year-to-year scale, an improvement of around 250 million tons of carbon compared to the ENSO and GCP models. At the decadal scale the new model did even better, with a carbon budget imbalance of plus or minus 1.6 gigatons of carbon, which is less than half of the imbalance from the GCP model.

The model’s smaller carbon budget imbalance reduces uncertainty and makes it a better tool than the ENSO or GCP models for verifying reported carbon emissions. At a five-year timespan the new model can verify emissions with just 4.4% uncertainty, half of the 8.8% uncertainty associated with the more complex GCP model.

“The performance of this model says to me that it’s not just temperature or anomalous warmth that matters for carbon cycling, it’s also the timing and location of that anomalous warmth,” said Birner.

Birner also noted the surprising stability of the land biosphere’s temperature sensitivities.

He explained that the model’s temperature sensitivities were identical for every year they ran it. If, in the real world, the temperature sensitivities of terrestrial ecosystems had been changing significantly, then the model would not have successfully accounted for the measured growth of carbon dioxide in the atmosphere.

Birner said this is surprising because of how intimately plants interact with temperature, and the capacity of living things to evolve and adapt. It would be easy to imagine plants altering their responses to changes in temperature as a kind of adaptation to a shifting climate. Such adaptations could change the relationship that Birner and Keeling’s model was using to derive terrestrial carbon cycling. And yet, that’s not what the researchers observed. Instead, the temperature sensitivities of land ecosystems around the world remained very consistent for the 60 years the team used to check their model’s performance.

The authors, however, also note that performance breaks down beyond about a 20-year time horizon. Additionally, they cautioned against extrapolating the stability implied by the model’s success too far into the future.

“We are showing a high degree of stability, but that doesn’t mean that the carbon cycle isn’t sensitive to temperature,” said Keeling. “That stability may not hold forever. It’s likely there are thresholds in the system beyond which that stability may disintegrate and we could indeed see higher temperatures leading to higher carbon emissions from the land biosphere.”

The next step for this model would be to operationalize it to “see if the last five years of carbon dioxide growth in the atmosphere are in alignment with reported emissions,” said Keeling.

But he said the model can also help us better understand the fundamentals of carbon cycling on Earth – a key to forecasting and combating climate change.

“If we have a year of unusual carbon dioxide growth we can also use this model to check our understanding of why that took place,” said Keeling. “If this current El Niño is like past ones we might see extra carbon dioxide growth. The fundamental assumptions of this model provide us with a new hypothesis about what’s driving the carbon cycle that we can test against what we see each year.”

In addition to Birner, Keeling, and Rödenbeck, the study’s other co-authors are Julia Dohner of Scripps Oceanography and Armin Schwartzman of UC San Diego’s Division of Biostatistics and Halıcıoğlu Data Science Institute.