Low-frequency variations of the large-scale ocean circulation and heat transport in the North Atlantic from 1955–2008 in situ temperature and salinity data T. Huck, A. Colin de Verdière, P. Estrade, F. Gaillard, R. Schopp, P. Bellec, R. Dussin Laboratoire de Physique des Océans (UMR 6523 CNRS IFREMER IRD UBO), Brest, France Abstract On interdecadal timescales, the Atlantic meridional overturning circulation (AMOC) is thought to be in phase with the North Atlantic Sea Surface Temperatures (as measured by the Atlantic Multidecadal Oscillation – AMO – index). However, it appears that we have entered a positive phase of the AMO since 1995-2000 although we fear the Atlantic meridional overturning may be on a declining trend, as suggested by several observational and modelling studies. Here we constrain ocean models with temperature and salinity fields built on observations, and compare the results with various simple methods (namely diagnostic, robust diagnostic and prognostic), models (North Atlantic and global configurations at various resolutions), and forcings. Mean transports of heat and mass are sensitive to the method and model configuration, but their decadal variability is much more coherent and does not depend explicitly on the variations of the surface forcing, its influence being imprinted in the thermohaline structure. Multidecadal variations are of the order of 20% (0.15 PW in heat transport and 4 Sv in overturning), with large transports in the subpolar gyre in the early 1960’s and mid 1990’s, and low values in the mid 1970’s. Declining transports of heat and mass are coherent in several models and methods since 1995, especially in the subpolar gyre, and opposite to the long term tendency from 1958 to 2008. 1. Introduction Variations in the oceanic thermohaline structure have been documented over the last decades: surface intensified warming and changes in salinity, as well as deep water properties and formation rates [Dickson et al. 1996, 2002]. However the associated changes in the large-scale ocean circulation are poorly known, and deserve much interest in the context of the ongoing global warming and possible decay of the thermohaline circulation [Bryden et al. 2005; Gregory et al. 2005], or recent decline observed in the North Atlantic subpolar gyre [Häkkinen and Rhines 2004]. Several ocean models have been forced by atmospheric reanalysis forcings, but these forcings have significant uncertainties and well-known heterogeneities over the last 50 years. The main model deficiencies lie in formulation of subgrid-scale mixing with consequences on deep-water formation, usually impacting the overturning circulation on the long term. In situ data assimilation in such models on long time scales requires complex tools and delicate choices on the method, that largely influence the results. On the other hand, to avoid the need for accurate surface fluxes of heat and freshwater, one can use the observed temperature and salinity (TS) fields. Density providing the baroclinic velocities through the thermal wind relation, the barotropic part is obtained from the vorticity equation forced by the wind and a bottom pressure torque [Sarkisyan and Keonjiyan 1975]. Mellor et al. [1982] integrated this equation along f /H contours, whereas Holland and Hirschman [1972] used the dynamical part of numerical ocean models, although some adjustment of the bottom density field may be necessary [Ezer and Mellor 1994]. These methods have been applied to compare the pentads 1955–59 and 1970–74 [Greatbatch et al. 1991; Ezer et al. 1995], and more recently for 7 pentads from 1950 to 1994 using a finite element formulation [Myers et al. 2005]. NODC has made available global fields of TS pentadal anomalies from 1955– 59 to 1994–98 based on hydrographic data. We will diagnose mean ocean currents from these fields to investigate the low-frequency variations of mass and heat transports in the North Atlantic. We first use three simple, well-documented methods: constant tracers, robust diagnostic, and short prognostic. Although the methods provide different results on the mean state, the low-frequency variations are rather coherent. Then, we implement only the robust diagnostic method in a global model continuous simulation with the seasonal cycle and TS anomaly fields updated to 2008. 1