Beneath the Arctic landscape lies a thick layer of permafrost soil, containing organic matter accumulated over thousands of years, the vast majority of which has remained too cold to decompose. This layer, ranging from less than 1 meter to nearly 1,500 meters thick, stores a significant amount of carbon: The top 1–3 meters of permafrost alone hold about twice as much carbon as the atmosphere of Earth itself [Schuur et al., 2022].
The fate of the Arctic landscape has important implications for the health of Arctic communities and ecosystems, as well as for regional and global climate.
However, with global temperatures rising, this organic matter is vulnerable to decomposition that would release gases such as carbon dioxide and methane into the atmosphere. The warming effect of these greenhouse gases is expected to cause additional thawing through the permafrost carbon-climate feedback. Current estimates suggest permafrost carbon loss to the atmosphere over the next 100 years could be equivalent to the emissions from a large, developed country if 2019-level emissions are sustained over the next 100 years [Schuur et al., 2022].
The fate of the Arctic landscape has important implications for the health of Arctic communities and ecosystems, as well as for regional and global climate. Communities are already facing challenges from the physical changes occurring, such as collapsing roads and unstable infrastructure from thawing permafrost. Diminishing sea ice and shifting populations of key species are also affecting the livelihoods and lifestyles of communities across the Arctic.
Ecosystems are becoming greener as the Arctic warms and the range of plant species shifts northward [Berner and Goetz, 2022], a change that may enhance carbon uptake at high latitudes. However, warming temperatures and changing moisture conditions are also accelerating permafrost thaw, land subsidence and erosion, and an increased risk of wildfires.
All of these factors will affect whether the region acts as a net atmospheric carbon source or sink, and they point toward the crucial need to understand scientifically what’s happening to Arctic permafrost today and what may happen in the future. Modeling is a vital part of this effort, although existing Earth system models often disagree on how much land is underlain by permafrost and produce widely disparate estimates of permafrost degradation rates and carbon emissions.
To inform model development and improve our ability to simulate permafrost carbon dynamics, land surface modeling groups are working together to perform pan-Arctic permafrost simulations and compare how well models simulate carbon loss observed from a collection of air and soil warming experiments conducted across the region.
The Complex Problem of Permafrost Carbon Emissions
The release of permafrost carbon into the atmosphere is poorly accounted for in current climate projections from Earth system models [Schädel et al., 2024], primarily because of the challenge of realistically simulating permafrost thaw. This challenge arises in part because the Arctic is a heterogeneous region—its climate, topography, and landscape vary widely from place to place—and because permafrost carbon emissions depend on interactions among many factors, including permafrost history, soil moisture, the extent of disturbance, and vegetation types and amounts [Mekonnen et al., 2019; Schädel et al., 2024]. In addition, permafrost thaw can occur abruptly and at different spatial and temporal scales, and current models have difficulty capturing small-scale changes.
Future climate projections that do not account for permafrost thaw may underestimate the pace and magnitude of climate warming in coming centuries.
Yet robust projections of permafrost carbon dynamics are crucial for informing climate policy and action. Future climate projections that do not account for permafrost thaw may underestimate the pace and magnitude of climate warming in coming centuries.
Current models that do consider permafrost thaw offer very different projections of the fate of Arctic carbon. Some models predict the region will become a large carbon source to the atmosphere over the next 300 years, whereas others predict it will be a large sink [McGuire et al., 2018]. To characterize the future Arctic carbon balance more robustly, including average trends and uncertainties, scientists often run and evaluate multiple Earth system models together in model intercomparisons.
Here we discuss the Warming Permafrost Model Intercomparison Project (WrPMIP), which began to address these challenges in 2021. The main goals of WrPMIP are to evaluate model projections of permafrost carbon responses to warming, identify significant sources of discrepancies among models, and develop functional benchmarks that modelers can use to improve how physical and biogeochemical processes are represented in their models.
A Reality Check for Model Responses
WrPMIP is unique in that it is evaluating the skill of model simulations against data collected at experimental warming sites across the Arctic to see how well the models capture observed warming responses. Twelve modeling groups are participating in WrPMIP to perform a new set of Arctic-focused simulations using a suite of 14 land surface models. Simulation results are being compared to observational data from 56 open-top chamber and snow fence warming experiments that characterize respiration, soil moisture and temperature, and permafrost thaw depth across a variety of geographic settings [Maes et al., 2024] (Figure 1).
Open-top chambers are transparent chambers, about 0.5–1 meter in diameter, constructed on the ground around measurement plots. These chambers raise air and soil temperatures by suppressing heat exchange with the overlying atmosphere [Henry and Molau, 1997]. Snow fences increase snow accumulation on their leeward sides, thereby insulating and warming the soil [Natali et al., 2011]. Data from the warming experiments used in WrPMIP indicate that there is, on average, 0.5°C to 1.5°C of air warming and 0.5°C to 1°C of topsoil warming in the summer in the open-top chambers and 1°C to 3°C of winter soil warming as a result of snow fences.
Results from the first set of regional simulations from WrPMIP indicate that the models tested disagree substantially on the amount of ecosystem respiration (carbon loss) occurring at lower Arctic latitudes, where respiration fluxes are higher (Figure 2). In addition, whereas the experimental site observations generally show an immediate increased ecosystem respiration response to warming followed by a decline in respiration after about 5 years [Maes et al., 2024], the regional simulations so far have not captured these observed temporal trends. For example, some models show slower respiration increases that level out with warming, whereas other models show little change in respiration over time. However, future WrPMIP site-level simulations may more accurately capture responses to warming, as the models will be more tuned to individual sites characteristics.
An added challenge to identifying causes of model disagreement is that Earth system models are becoming increasingly complex. Even when a model reproduces observations well, it is difficult to know whether it’s getting the correct answer for the correct reasons (i.e., by simulating the relevant physical properties accurately) [Huntzinger et al., 2020]. As models become more complex, it is increasingly important to understand the reasons behind a model arriving at its results.
One way to dig deeper, beyond simply comparing modeled carbon stocks and fluxes to observations, is to use observed functional responses as model performance benchmarks. Functional responses characterize how a physical process, say, ecosystem respiration, responds to a single factor, such as temperature. These responses provide information not only about a model’s estimate of a given stock or flux but also about the sensitivity of that estimate to changes in environmental conditions. Studying functional responses can help shed light on whether models are capturing the necessary processes to simulate permafrost thaw and the resulting carbon release accurately.
With the experimental warming data and models WrPMIP is evaluating, we can test the sensitivity of environmental and biogeochemical responses to warming and determine whether those responses are similar in models and experimental observations [Schädel et al., 2018]. In particular, we are investigating whether the first set of model simulations can reproduce the magnitudes and rates at which ecosystem respiration is increasing with temperature. After a second set of model simulations is completed, we will also be evaluating model results using the performance metrics within the International Land Model Benchmarking project [Collier et al., 2018]. These standardized performance metrics, such as the calculation of a model’s bias and the seasonal cycle of respiration compared to observations, can be applied across models and model versions for consistent evaluation.
Working Together for Model Improvement
Another benefit of the Warming Permafrost Model Intercomparison Project has been the formation of teams of modelers and experimentalists working together to solve grand challenges of permafrost carbon modeling.
Beyond the scientific work and results of WrPMIP, another benefit of this project has been the formation of teams of modelers and experimentalists working together to solve grand challenges of permafrost carbon modeling. The resulting increases in communication and collaboration are helping accelerate progress toward accurately incorporating permafrost carbon processes into Earth system models.
In September 2023, 29 scientists from seven countries gathered at a 3-day workshop at the Woodwell Climate Research Center in Massachusetts to discuss the state of permafrost modeling, individual model capabilities, and priorities for improving modeling of the future carbon balance and the effects that permafrost will have on the Arctic. Priorities identified included the need to incorporate significant landscape disturbance processes in the Arctic, such as thermokarst and wildfires, into models.
Other major needs discussed at the workshop were consistent benchmarking throughout the model development process, which is one of the main objectives for WrPMIP; more collaboration among modelers, experimentalists, and policymakers; and codevelopment of models by different modeling groups. As a result of this workshop, WrPMIP participants discussed plans to share model code and make model documentation more publicly available to facilitate codevelopment.
Better Predictions of Permafrost Thaw
Vast stores of Arctic permafrost carbon long frozen below Earth’s surface are proving increasingly vulnerable to release to the atmosphere, although the magnitude and rate of the release over the next century and beyond will depend on how quickly permafrost thaws. This thawing, in turn, will depend on the extent of continuing greenhouse gas emissions—and attendant warming—resulting from fossil fuel burning, land use change, and other contributors. For example, a 25% loss of permafrost is expected under a low-emissions scenario (Representative Concentration Pathway (RCP) 2.5), whereas the loss under a high-emissions scenario (RCP 8.5) could be up to 70% [Schuur et al., 2022].
WrPMIP’s efforts to evaluate modeled permafrost responses to experimental warming should inform needed improvements in Earth system models by addressing a grand challenge in future climate projections: illuminating when, how much, and what forms of permafrost carbon will be released into the atmosphere in a warming world.
Acknowledgments
We thank additional members of the WrPMIP team, including Ted Schuur and Anna Gagne-Landmann, for their contributions to this Science Update. We also thank the many researchers who have provided data to the WrPMIP project and thank workshop participants for their contributions of ideas. The writing of this article was supported by the National Science Foundation Graduate Research Fellowship Program under grant 1938054. We acknowledge support from U.S. Department of Energy grant DE-SC0022116 and partial funding from Permafrost Pathways through the TED Audacious Project. W.J.R. was funded by the Reducing Uncertainties in Biogeochemical Interactions through Synthesis and Computation (RUBISCO) Scientific Focus Area, Office of Biological and Environmental Research, U.S. Department of Energy Office of Science under contract DE-AC02-05CH11231.
References
Berner, L. T., and S. J. Goetz (2022), Satellite observations document trends consistent with a boreal forest biome shift, Global Change Biol., 28(10), 3,275–3,292, https://doi.org/10.1111/GCB.16121.
Collier, N., et al. (2018), The International Land Model Benchmarking (ILAMB) system: Design, theory, and implementation, J. Adv. Model. Earth Syst., 10(11), 2,731–2,754, https://doi.org/10.1029/2018MS001354.
Henry, G. H. R., and U. Molau (1997), Tundra plants and climate change: The International Tundra Experiment (ITEX), Global Change Biol., 3, 1–9, https://doi.org/10.1111/j.1365-2486.1997.gcb132.x.
Huntzinger, D. N., et al. (2020), Evaluation of simulated soil carbon dynamics in Arctic-boreal ecosystems, Environ. Res. Lett., 15(2), 025005, https://doi.org/10.1088/1748-9326/AB6784.
Maes, S. L., et al. (2024), Environmental drivers of increased ecosystem respiration in a warming tundra, Nature, 629, 105–113, https://doi.org/10.1038/s41586-024-07274-7.
McGuire, A. D., et al. (2018), Dependence of the evolution of carbon dynamics in the northern permafrost region on the trajectory of climate change, Proc. Natl. Acad. Sci. U. S. A., 115(15), 3,882–3,887, https://doi.org/10.1073/PNAS.1719903115.
Mekonnen, Z. A., et al. (2019), Expansion of high-latitude deciduous forests driven by interactions between climate warming and fire, Nat. Plants, 5, 952–958, https://doi.org/10.1038/s41477-019-0495-8.
Natali, S. M., et al. (2011), Effects of experimental warming of air, soil and permafrost on carbon balance in Alaskan tundra, Global Change Biol., 17(3), 1,394–1,407, https://doi.org/10.1111/J.1365-2486.2010.02303.X.
Schädel, C., et al. (2018), Divergent patterns of experimental and model-derived permafrost ecosystem carbon dynamics in response to Arctic warming, Environ. Res. Lett., 13(10), 105002, https://doi.org/10.1088/1748-9326/AAE0FF.
Schädel, C., et al. (2024), Earth system models must include permafrost carbon processes, Nat. Clim. Change, 14, 114–116, https://doi.org/10.1038/s41558-023-01909-9.
Schuur, E. A. G., et al. (2022), Permafrost and climate change: Carbon cycle feedbacks from the warming Arctic, Annu. Rev. Environ. Resour., 47, 343–371, https://doi.org/10.1146/annurev-environ-012220-011847.
Author Information
Jeralyn Poe (jmp838@nau.edu) and Jon Wells, Northern Arizona University, Flagstaff; Christina Schädel, Woodwell Climate Research Center, Falmouth, Mass.; Deborah N. Huntzinger, Northern Arizona University, Flagstaff; and William J. Riley, Lawrence Berkeley National Laboratory, Berkeley, Calif.