Southeast Indian Subantarctic Mode Water in the CMIP6 Coupled Models

This study assesses the capability of 12 models in the Phase 6 of the Coupled Models Intercomparison Project (CMIP6) in simulating the Southeast Indian Subantarctic mode water (SEISAMW) by comparing to Argo observations. The results show that all of the analyzed CMIP6 coupled models can reproduce the SEISAMW and its seasonal cycle, albeit with discrepancies of formation region and properties among the models. The SEISAMW subduction rate shows significant interannual variability, which is primarily induced by lateral induction in these models, in agreement with observations. Furthermore, the longterm trends of the SEISAMW show volume loss in accordance with the descending trend of the subduction rate, which are co‐occurred with the ascending trend of the Southern Annular Mode (SAM) indices in most of the analyzed CMIP6 models, similar to those in the CMIP3 and CMIP5 coupled models. Meanwhile, the potential density of the SEISAMWs show decreasing trends in these volume loss models, which could be explained by the warming (SAM0‐UNICON, CESM2‐WACCM, CAS‐ESM2‐0 and CIESM), the freshening (IPSL‐CM6A‐LR, CAMS‐CSM1‐0, MRI‐ESM2‐0, and FGOALS‐f3‐L) or the both (FIO‐ESM‐2‐0, E3SM‐1‐0, and CanESM5) trends of the SEISAMWs in different CMIP6 coupled models. Decreases in the projected subduction rate and volume of SEISAMW imply a slowdown of Southern Indian Ocean circulation in the future, reducing the heat and carbon transport from atmosphere to ocean interior contributed by SEISAMW. Plain Language Summary Subantarctic mode water (SAMW) is formed between the subantarctic front and the subtropical front. The SAMW that formed in the South Indian Ocean is called Southeast Indian Subantarctic Mode Water (SEISAMW). When it forms, seawater at the base of mixed layer subducted into the ocean interior in late winter, providing subsurface oceanic reservoirs of heat, carbon, and other properties that can significantly affect the global climate. Using data from the historical simulation in 12 CMIP6 coupled models and Argo; we assess the capability of these models in simulating SEISAMW and reveal the interannual and longterm trend in SEISAMW properties (e.g., subduction rate, volume, temperature, salinity, and density). The results indicate that all of the analyzed CMIP6 models are able to reproduce the SEISAMW and its seasonal cycle. The interannual variability of the SEISAMW is attributed to lateral induction in all of the analyzed CMIP6 models, in agreement with observations. Bo