Challenges in improving Arctic freshwater simulations: An evaluation of CMIP6 models in the Beaufort Gyre region

The performance of CMIP6 models in simulating freshwater content (FWC) in the Beaufort Gyre remains unclear. This study evaluated 17 CMIP6 models using both observational and reanalysis datasets. Additionally, a global ice-ocean coupled model based on Finite Volume Community Ocean Model (Global-FVCOM) was incorporated for reference. The results revealed a significant inter-model spread among the CMIP6 models in spatiotemporal variations of FWC, with discrepancies relative to the evaluation data that were larger than those exhibited in Global-FVCOM. These discrepancies were primarily attributed to simulation errors of the salinity structure within the CMIP6 models. Over half of the models indicated that the primary source of FWC error originated from the layers above the base of halocline, where most models underestimated FWC, while others suggested the error originated from the layers between the base of the halocline and the 34.8 psu isohaline, where models tended to overestimate FWC. Based on an overall evaluation using observational and reanalysis datasets, EC-Earth3, MRI-ESM2-0, and FIO-ESM-2-0 showed better performance relative to other CMIP6 models. However, these three models, along with the multi-model mean, exhibited larger errors than Global-FVCOM, suggesting that current CMIP6 models still face challenges in FWC simulation relative to some ice-ocean coupled models. The main aspects contributing to the errors, including discrepancies in uncertainties induced by internal variability, numerical configurations, vertical mixing schemes, model resolutions, freshwater inputs, and atmospheric forcings were further discussed in this study. This study enhances understandings of CMIP6 models’ capabilities to simulate FWC in the Beaufort Gyre region, providing valuable insights for future model improvements.