Detectability of changes to the Atlantic meridional overturning circulation in the Hadley Centre Climate Models

• Published in: Climate Dynamics
• Main Authors: Roberts, Christopher D., Palmer, Matthew D.
• Format: Article in Journal/Newspaper
• Language: English
• Published: 2012
• Subjects: AMOC; Atlantic meridional overturning circulation; Climate model; Detection; MOC; North Atlantic
• Online Access: https://hdl.handle.net/1983/23878c8a-b696-4c0c-a198-e6c331b43489; https://research-information.bris.ac.uk/en/publications/23878c8a-b696-4c0c-a198-e6c331b43489; https://doi.org/10.1007/s00382-012-1306-3; http://www.scopus.com/inward/record.url?scp=84868094123&partnerID=8YFLogxK

The Atlantic meridional overturning circulation (MOC) is responsible for a climatically significant northward heat transport that is expected to decrease in response to anthropogenic global warming. Here, simulations from an ensemble of UK Met Office Hadley Centre Climate Models (HadGEM1, HadGEM2 and a 22 member perturbed physics ensemble of HadCM3-like models) are used to evaluate detection times for different MOC observing strategies. Six different detection statistics are compared, including direct observations of the MOC at two latitudes (26°N and 50°N) and several multivariate detection variables based on an optimal fingerprint of MOC change previously identified using HadCM3 (Vellinga and Wood in Geophys Res Lett 31(14):L14203, 2004). Using these models, and assuming perfectly observed conditions, we find no evidence to suggest that detection times would be significantly reduced by measuring the MOC at 50°N instead of (or in addition to) measurements at 26°N. Our results suggest that complementary observations of hydrographic properties in the North Atlantic may help reduce MOC detection times, but the benefits are not universal across models, nor as large as previously suggested. In addition, detection times calculated using optimal fingerprint methods are sensitive to the model-dependent estimates of covariances describing internal climate variability. This last result presents a strong case for deriving fingerprints of MOC change using dynamical/physical arguments, rather than statistical methods, in order to promote more robust results across a range of models.