Permafrost, Peatland, and Cropland Regions Are Key to Reconciling North American Carbon Sink Estimates

Persistent discrepancies between bottom‐up, terrestrial biosphere models (TBMs), and top‐down, atmospheric inversions, have made it difficult to quantify the magnitude of the North American terrestrial carbon sink. Previous studies have compared aggregated continent‐scale estimates of carbon fluxes from TBMs and inversions for all of North America, but this provides limited insights into finer‐scale mismatches that contribute to the overall discrepancies. Here we evaluate agreement between TBM and inversion carbon flux estimates at 1° × 1° resolution to provide more direct insights into where models disagree and what underlying factors drive discrepancies. We find that the additional carbon uptake estimated by inversions, in just 16% of the area of North America, is large enough to account for the discrepancy between TBMs and inversions across the whole continent. The majority of these differences occur in permafrost, peatland, and cropland regions. In these regions, we find a higher likelihood of potential biases in the weaker sink estimates from TBMs, suggesting that the stronger sink implied by inversions is more likely to be realistic. However, the current observational coverage is insufficient for fully assessing the causes of discrepancies or the magnitude of biases in either approach. Encouragingly, improved representation of agricultural processes in a TBM led to better agreement with inversions in croplands. Efforts to accurately model cropland dynamics will help improve agreement between TBMs and inversions. Overall, this work presents a clear path for reconciling the discrepancies between inversion and TBM estimates of the North American carbon sink that have persisted for two decades.

K.T.F., W.S., A.M., B.B.-L., and A.M.M. acknowledge support from the NASA Terrestrial Ecology Program (Grant 80NSSC22K1253). K.T.F., W.S., and A.M.M. acknowledge support from the Carnegie Institution for Scienceendowment. We thank all modelers of the Trends in Net Land-Atmosphere Exchange project (TRENDY; https://blogs.exeter.ac.uk/trendy/). We thank the CarbonTracker2022 Team for the CarbonTracker CT2022 results provided by NOAA GML, Boulder, Colorado, USA from the website at http://carbontracker.noaa.gov. We thank the CarbonTracker-Lagrange team for terrestrial CO₂ flux data. We thank the Atmospheric and Environmental Research, Inc. (AER)—particularly, Thomas Nehrkorn, John Henderson, and Janusz Eluszkiewicz—for conducting WRF-STILT simulations and providing transport footprints. We thank the CarbonTracker-Lagrange project team for proving the WRF-STILT transport footprints. The ESA CCI LC data used here contain modified Copernicus Climate Change Service information [2024]. Neither the European Commission nor ECMWF is responsible for any use that may be made of the Copernicus information or data it contains. Computations presented in this study were conducted through the Carnegie Institution for Science's partnership in the Resnick High Performance Computing Center (https://www.hpc.caltech.edu), a facility supported by Resnick Sustainability Institute at the California Institute of Technology.