Permafrost is a critical component of Earth's climate system and yet many Earth System models do not adequately represent permafrost processes. The complexity of Earth System Models and the urgent need to reduce model uncertainty requires innovative approaches to evaluate and analyze model results. In the Warming in Permafrost Model Intercomparison Project (WrPMIP), we use manipulative field‐based experiments to inform long‐term projections about the magnitude and underlying mechanisms of permafrost carbon dynamics in a changing Arctic. Specifically, we perform and analyze multi-model simulations across the pan-Arctic and at specific sites, aligning them with open-top chamber and snow-fence experiments in tundra ecosystems. To simulate experimental warming in models, we modified aerodynamic resistance to mimic open-top chamber air warming and reduced the snow thermal conductivity to mimic snow-fence induced soil warming.

At the regional scale, we harmonized output from 14 ecosystem and Earth system models and compared environmental variables (e.g., soil temperature, active layer depth, water table depth) and variables associated with ecosystem carbon exchange (ecosystem respiration and gross primary productivity) with synthesis results of 56 open-top chamber in-situ warming experiments (Maes et al. 2024). Model responses strongly varied, with some models showing little increase in ecosystem respiration or gross primary productivity with warming, while others showed up to a 30% increase in ecosystem respiration, aligning with in-site responses of open-top chamber experiments. Generally, model responses were linear over the 20-year perturbation period, whereas field experimental responses show non-linear patterns across a 20-year time window, highlighting the complexity of carbon flux responses to warming that models do not capture.