Large diversity in AMOC internal variability across NEMO-based climate models

• Published in: Climate Dynamics
• Main Authors: Zhao, Alcide, Robson, Jon, Sutton, Rowan, Lai, Michael W. K., Mecking, Jennifer V., Yeager, Stephen, Petit, Tillys
• Format: Article in Journal/Newspaper
• Language: English
• Published: Springer 2024
• Subjects: Arctic; Greenland; Denmark Strait; Iceland
• Online Access: https://centaur.reading.ac.uk/114411/; https://centaur.reading.ac.uk/114411/9/114411%20VoR.pdf; https://centaur.reading.ac.uk/114411/1/Zhao_CLDY-D-23-00391R2_accepted.pdf

We characterise, and explore the drivers of, differences in the internal variability of the Atlantic Meridional Overturning Circulation (AMOC) across five NEMO-based CMIP6 class climate models. While the variability of AMOC variability is dominated by its lower dense limb in all models, there is large diversity in the timescale, multidecadal variability, and latitudinal coherence of AMOC across models. In particular, the UK models have much weaker AMOC multidecadal variability and latitudinal coherence. The model diversity is associated with differences in salinity-governed surface density variations which drive high-density water mass transformation (WMT) in the Greenland-Iceland-Norwegian Seas (GIN) and the Arctic. Specifically, GIN Seas WMT shows large multidecadal variability which has a major impact on AMOC variability in non-UK models. In contrast, the smaller variability in GIN Seas WMT in the UK models has limited impact on the lower latitude AMOC via the Denmark strait overflow mass transportation. This leads to a latitudinally less coherent and weaker multidecadal variability of the AMOC lower limb. Such differences between UK and non-UK models are related to differences in model mean states and densification processes in the Arctic and GIN Seas. Consequently, we recommend further in-depth studies to better understand and constrain processes driving salinity changes in the Arctic and GIN Seas for more reliable representation of the AMOC in climate models.