• Beismann, J.-O., , and B. Barnier, 2004: Variability of the meridional overturning circulation of the North Atlantic: Sensitivity to overflows of dense water masses. Ocean Dyn., 54, 92106, doi:10.1007/s10236-003-0088-x.

    • Search Google Scholar
    • Export Citation
  • Bentsen, M., , H. Drange, , T. Furevik, , and T. Zhou, 2004: Simulated variability of the Atlantic meridional overturning circulation. Climate Dyn., 22, 701720, doi:10.1007/s00382-004-0397-x.

    • Search Google Scholar
    • Export Citation
  • Biastoch, A., , C. Böning, , J. Getzlaff, , J.-M. Molines, , and G. Madec, 2008: Causes of interannual-decadal variability in the meridional overturning circulation of the midlatitude North Atlantic Ocean. J. Climate, 21, 65996615, doi:10.1175/2008JCLI2404.1.

    • Search Google Scholar
    • Export Citation
  • Böning, C., , M. Scheinert, , J. Dengg, , A. Biastoch, , and A. Funk, 2006: Decadal variability of subpolar gyre transport and its reverberation in the North Atlantic overturning. Geophys. Res. Lett., 33, L21S01, doi:10.1029/2006GL026906.

    • Search Google Scholar
    • Export Citation
  • Brauch, J. P., , and R. Gerdes, 2005: Response of the northern North Atlantic and Arctic Oceans to a sudden change of the North Atlantic Oscillation. J. Geophys. Res., 110, C11018, doi:10.1029/2004JC002436.

    • Search Google Scholar
    • Export Citation
  • Chang, E. K. M., 2007: Assessing the increasing trend in Northern Hemisphere winter storm track activity using surface ship observations and a statistical storm track model. J. Climate, 20, 56075628, doi:10.1175/2007JCLI1596.1.

    • Search Google Scholar
    • Export Citation
  • Chanut, J., , B. Barnier, , W. Large, , L. Debreu, , T. Penduff, , J. M. Molines, , and P. Mathiot, 2008: Mesoscale eddies in the Labrador Sea and their contribution to convection and restratification. J. Phys. Oceanogr., 38, 16171643, doi:10.1175/2008JPO3485.1.

    • Search Google Scholar
    • Export Citation
  • Comiso, J., cited 2012: Bootstrap sea ice concentrations from Nimbus-7 SMMR and DMSP SSM/I-SSMIS, version 2. National Snow and Ice Data Center, Boulder, CO, digital media. [Available online at http://nsidc.org/data/nsidc-0079.]

  • Cunningham, S. A., and Coauthors, 2007: Temporal variability of the Atlantic meridional overturning circulation at 26.5°N. Science, 317, 935938, doi:10.1126/science.1141304.

    • Search Google Scholar
    • Export Citation
  • Curry, R. G., , and M. S. McCartney, 2001: Ocean gyre circulation changes associated with the North Atlantic Oscillation. J. Phys. Oceanogr., 31, 33743400, doi:10.1175/1520-0485(2001)031<3374:OGCCAW>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Dai, A., , T. Qian, , K. Trenberth, , and J. Milliman, 2009: Changes in continental freshwater discharge from 1948 to 2004. J. Climate, 22, 27732791, doi:10.1175/2008JCLI2592.1.

    • Search Google Scholar
    • Export Citation
  • Danabasoglu, G., , W. G. Large, , and B. P. Briegleb, 2010: Climate impacts of parameterized Nordic Sea overflows. J. Geophys. Res., 115, C11005, doi:10.1029/2010JC006243.

    • Search Google Scholar
    • Export Citation
  • Danabasoglu, G., , S. C. Bates, , B. P. Briegleb, , S. R. Jayne, , M. Jochum, , W. G. Large, , S. Peacock, , and S. G. Yeager, 2012a: The CCSM4 ocean component. J. Climate, 25, 13611389, doi:10.1175/JCLI-D-11-00091.1.

    • Search Google Scholar
    • Export Citation
  • Danabasoglu, G., , S. G. Yeager, , Y.-O. Kwon, , J. Tribbia, , A. Phillips, , and J. Hurrell, 2012b: Variability of the Atlantic meridional overturning circulation in CCSM4. J. Climate, 25, 51535172, doi:10.1175/JCLI-D-11-00463.1.

    • Search Google Scholar
    • Export Citation
  • Danabasoglu, G., and Coauthors, 2014: North Atlantic simulations in Coordinated Ocean-Ice Reference Experiments phase II (CORE-II). Part I: Mean states. Ocean Modell., 73, 76107, doi:10.1016/j.ocemod.2013.10.005.

    • Search Google Scholar
    • Export Citation
  • de Boyer Montégut, C., , G. Madec, , A. S. Fischer, , A. Lazar, , and D. Iudicone, 2004: Mixed layer depth over the global ocean: An examination of profile data and a profile-based climatology. J. Geophys. Res., 109, C12003, doi:10.1029/2004JC002378.

    • Search Google Scholar
    • Export Citation
  • Delworth, T. L., , and M. E. Mann, 2000: Observed and simulated multidecadal variability in the Northern Hemisphere. Climate Dyn., 16, 661676, doi:10.1007/s003820000075.

    • Search Google Scholar
    • Export Citation
  • Delworth, T. L., , and F. Zeng, 2008: Simulated impact of altered Southern Hemisphere winds on the Atlantic overturning circulation. Geophys. Res. Lett., 35, L20708, doi:10.1029/2008GL035166.

    • Search Google Scholar
    • Export Citation
  • Dickson, B., , I. Yashayaev, , J. Meincke, , B. Turrell, , S. Dye, , and J. Holfort, 2002: Rapid freshening of the deep North Atlantic Ocean over the past four decades. Nature, 416, 832836, doi:10.1038/416832a.

    • Search Google Scholar
    • Export Citation
  • Doney, S. C., , S. G. Yeager, , W. G. Large, , and J. C. McWilliams, 2003: Modeling global oceanic interannual variability (1958-1997): Simulation design and model-data evaluation. NCAR Tech. Note NCAR/TN-452+STR, 48 pp.

  • Duchon, C. E., 1979: Lanczos filtering in one and two dimensions. J. Appl. Meteor., 18, 10161022, doi:10.1175/1520-0450(1979)018<1016:LFIOAT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Eden, C., , and J. Willebrand, 2001: Mechanism of interannual to decadal variability of the North Atlantic circulation. J. Climate, 14, 22662280, doi:10.1175/1520-0442(2001)014<2266:MOITDV>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Farneti, R., , and T. L. Delworth, 2010: The role of mesoscale eddies in the remote oceanic response to altered Southern Hemisphere winds. J. Phys. Oceanogr., 40, 23482354, doi:10.1175/2010JPO4480.1.

    • Search Google Scholar
    • Export Citation
  • Farneti, R., , and P. R. Gent, 2011: The effects of the eddy-induced advection coefficient in a coarse-resolution coupled climate model. Ocean Modell., 39, 135145, doi:10.1016/j.ocemod.2011.02.005.

    • Search Google Scholar
    • Export Citation
  • Farneti, R., , T. L. Delworth, , A. J. Rosati, , S. M. Griffies, , and F. Zeng, 2010: The role of mesoscale eddies in the rectification of the Southern Ocean response to climate change. J. Phys. Oceanogr., 40, 15391557, doi:10.1175/2010JPO4353.1.

    • Search Google Scholar
    • Export Citation
  • Fox-Kemper, B., and Coauthors, 2011: Parameterization of mixed layer eddies. III: Implementation and impact in global ocean climate simulations. Ocean Modell., 39, 6178, doi:10.1016/j.ocemod.2010.09.002.

    • Search Google Scholar
    • Export Citation
  • Gent, P. R., , and G. Danabasoglu, 2011: Response to increasing Southern Hemisphere winds in CCSM4. J. Climate, 24, 49924998, doi:10.1175/JCLI-D-10-05011.1.

    • Search Google Scholar
    • Export Citation
  • Gent, P. R., and Coauthors, 2011: The Community Climate System Model version 4. J. Climate, 24, 49734991, doi:10.1175/2011JCLI4083.1.

    • Search Google Scholar
    • Export Citation
  • Grist, J. P., and Coauthors, 2010: The roles of surface heat flux and ocean heat transport convergence in determining Atlantic Ocean temperature variability. Ocean Dyn., 60, 771790, doi:10.1007/s10236-010-0292-4.

    • Search Google Scholar
    • Export Citation
  • Grist, J. P., , S. A. Josey, , R. Marsh, , Y.-O. Kwon, , R. J. Bingham, , and A. T. Blaker, 2014: The surface-forced overturning of the North Atlantic: Estimates from modern era atmospheric reanalysis datasets. J. Climate, in press.

    • Search Google Scholar
    • Export Citation
  • Häkkinen, S., 1999: Variability of the simulated meridional heat transport in the North Atlantic for the period 1951-1993. J. Geophys. Res., 104, 10 99111 007, doi:10.1029/1999JC900034.

    • Search Google Scholar
    • Export Citation
  • Häkkinen, S., , P. B. Rhines, , and D. L. Worthen, 2011: Atmospheric blocking and Atlantic multidecadal ocean variability. Science, 334, 655659, doi:10.1126/science.1205683.

    • Search Google Scholar
    • Export Citation
  • Holland, M. M., , D. A. Bailey, , B. P. Briegleb, , B. Light, , and E. Hunke, 2012: Improved sea ice shortwave radiation physics in CCSM4: The impact of melt ponds and aerosols on Arctic sea ice. J. Climate, 25, 14131430, doi:10.1175/JCLI-D-11-00078.1.

    • Search Google Scholar
    • Export Citation
  • Hunke, E. C., , and W. H. Lipscomb, 2008: CICE: The Los Alamos sea ice model, documentation and software, version 4.0. Los Alamos National Laboratory Tech. Rep. LA-CC-06-012, 76 pp.

  • Hurrell, J. W., , J. J. Hack, , D. Shea, , J. M. Caron, , and J. Rosinski, 2008: A new sea surface temperature and sea ice boundary dataset for the Community Atmosphere Model. J. Climate, 21, 51455153, doi:10.1175/2008JCLI2292.1.

    • Search Google Scholar
    • Export Citation
  • Jochum, M., , G. Danabasoglu, , M. Holland, , Y.-O. Kwon, , and W. G. Large, 2008: Ocean viscosity and climate. J. Geophys. Res., 113, C06017, doi:10.1029/2007JC004515.

    • Search Google Scholar
    • Export Citation
  • Johnson, H. L., , and D. Marshall, 2002a: Localization of abrupt change in the North Atlantic thermohaline circulation. Geophys. Res. Lett., 29, doi:10.1029/2001GL014140.

    • Search Google Scholar
    • Export Citation
  • Johnson, H. L., , and D. Marshall, 2002b: A theory for the surface Atlantic response to thermohaline variability. J. Phys. Oceanogr., 32, 11211132, doi:10.1175/1520-0485(2002)032<1121:ATFTSA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Khatiwala, S., , P. Schlosser, , and M. Visbeck, 2002: Rates and mechanisms of water mass transformation in the Labrador Sea as inferred from tracer observations. J. Phys. Oceanogr., 32, 666686, doi:10.1175/1520-0485(2002)032<0666:RAMOWM>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Kistler, R., and Coauthors, 2001: The NCEP–NCAR 50-Year Reanalysis: Monthly means CD-ROM and documentation. Bull. Amer. Meteor. Soc., 82, 247267, doi:10.1175/1520-0477(2001)082<0247:TNNYRM>2.3.CO;2.

    • Search Google Scholar
    • Export Citation
  • Klinger, B. A., , and C. Cruz, 2009: Decadal response of global circulation to Southern Ocean zonal wind stress perturbation. J. Phys. Oceanogr., 39, 18881904, doi:10.1175/2009JPO4070.1.

    • Search Google Scholar
    • Export Citation
  • Knight, J. R., , R. J. Allan, , C. K. Folland, , M. Vellinga, , and M. E. Mann, 2005: A signature of persistent natural thermohaline circulation cycles in observed climate. Geophys. Res. Lett., 32, L20708, doi:10.1029/2005GL024233.

    • Search Google Scholar
    • Export Citation
  • Knight, J. R., , C. K. Folland, , and A. A. Scaife, 2006: Climate impacts of the Atlantic multidecadal oscillation. Geophys. Res. Lett., 33, L17706, doi:10.1029/2006GL026242.

    • Search Google Scholar
    • Export Citation
  • Kuhlbrodt, T., , A. Griesel, , M. Montoya, , A. Levermann, , M. Hofmann, , and S. Rahmstorf, 2007: On the driving processes of the Atlantic meridional overturning circulation. Rev. Geophys., 45, RG2001, doi:10.1029/2004RG000166.

    • Search Google Scholar
    • Export Citation
  • Large, W. G., , and G. Nurser, 2001: Ocean surface water mass transformation. Ocean Circulation and Climate—Observing and Modelling the Global Ocean, International Geophysics Series, Vol. 77, Academic Press, 317–336.

  • Large, W. G., , and S. G. Yeager, 2004: Diurnal to decadal global forcing for ocean and sea ice models: The data sets and climatologies. NCAR Tech. Note NCAR/TN-460+STR, 105 pp.

  • Large, W. G., , and S. G. Yeager, 2009: The global climatology of an interannually varying air–sea flux data set. Climate Dyn., 33, 341364, doi:10.1007/s00382-008-0441-3.

    • Search Google Scholar
    • Export Citation
  • Lavender, K. L., , R. E. Davis, , and W. B. Owens, 2002: Observations of open-ocean deep convection in the Labrador Sea from subsurface floats. J. Phys. Oceanogr., 32, 511526, doi:10.1175/1520-0485(2002)032<0511:OOOODC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Lee, S.-K., , W. Park, , E. van Sebille, , M. O. Baringer, , C. Wang, , D. B. Enfield, , S. G. Yeager, , and B. P. Kirtman, 2011: What caused the significant increase in Atlantic Ocean heat content since the mid-20th century? Geophys. Res. Lett., 38, L17607, doi:10.1029/2011GL048856.

    • Search Google Scholar
    • Export Citation
  • Levitus, S., and Coauthors, 1998: Introduction. Vol. 1, World Ocean Database, NOAA Atlas NESDIS 18, 346 pp.

  • Liu, Z., 2012: Dynamics of interdecadal climate variability: A historical perspective. J. Climate, 25, 19631995, doi:10.1175/2011JCLI3980.1.

    • Search Google Scholar
    • Export Citation
  • Locarnini, R. A., , A. V. Mishonov, , J. I. Antonov, , T. P. Boyer, , H. E. Garcia, , O. K. Baranova, , M. M. Zweng, , and D. R. Johnson, 2010: Temperature. Vol. 1, World Ocean Atlas 2009, NOAA Atlas NESDIS 68, 184 pp.

  • Lohmann, K., , H. Drange, , and M. Bentsen, 2009: Response of the North Atlantic subpolar gyre to persistent North Atlantic oscillation like forcing. Climate Dyn., 32, 273285, doi:10.1007/s00382-008-0467-6.

    • Search Google Scholar
    • Export Citation
  • Marsh, R., 2000: Recent variability of the North Atlantic thermohaline circulation inferred from surface heat and freshwater fluxes. J. Climate, 13, 32393260, doi:10.1175/1520-0442(2000)013<3239:RVOTNA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Marshall, J., and Coauthors, 1998: The Labrador Sea deep convection experiment. Bull. Amer. Meteor. Soc., 79, 20332058, doi:10.1175/1520-0477(1998)079<2033:TLSDCE>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Marshall, J., , H. Johnson, , and J. Goodman, 2001: A study of the interaction of the North Atlantic Oscillation with ocean circulation. J. Climate, 14, 13991421, doi:10.1175/1520-0442(2001)014<1399:ASOTIO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Medhaug, I., , and T. Furevik, 2011: North Atlantic 20th century multidecadal variability in coupled climate models: Sea surface temperature and ocean overturning circulation. Ocean Sci., 7, 389404, doi:10.5194/os-7-389-2011.

    • Search Google Scholar
    • Export Citation
  • Medhaug, I., , H. R. Langehaug, , T. Eldevik, , T. Furevik, , and M. Bentsen, 2012: Mechanisms for decadal scale variability in a simulated Atlantic meridional overturning circulation. Climate Dyn., 39, 7793, doi:10.1007/s00382-011-1124-z.

    • Search Google Scholar
    • Export Citation
  • Msadek, R., , and C. Frankignoul, 2009: Atlantic multidecadal oceanic variability and its influence on the atmosphere in a climate model. Climate Dyn., 33, 4562, doi:10.1007/s00382-008-0452-0.

    • Search Google Scholar
    • Export Citation
  • Nikurashin, M., , and G. Vallis, 2012: A theory of the interhemispheric meridional overturning circulation and associated stratification. J. Phys. Oceanogr., 42, 16521667, doi:10.1175/JPO-D-11-0189.1.

    • Search Google Scholar
    • Export Citation
  • Pickart, R. S., , D. J. Torres, , and R. A. Clarke, 2002: Hydrography of the Labrador Sea during active convection. J. Phys. Oceanogr., 32, 428457, doi:10.1175/1520-0485(2002)032<0428:HOTLSD>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Rahmstorf, S., , and M. H. England, 1997: Influence of Southern Hemisphere winds on North Atlantic deep water flow. J. Phys. Oceanogr., 27, 20402054, doi:10.1175/1520-0485(1997)027<2040:IOSHWO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Robson, J. I., , R. Sutton, , K. Lohmann, , D. Smith, , and M. D. Palmer, 2012a: Causes of the rapid warming of the North Atlantic Ocean in the mid-1990s. J. Climate, 25, 41164134, doi:10.1175/JCLI-D-11-00443.1.

    • Search Google Scholar
    • Export Citation
  • Robson, J. I., , R. Sutton, , and D. M. Smith, 2012b: Initialized decadal predictions of the rapid warming of the North Atlantic Ocean in the mid 1990s. Geophys. Res. Lett., 39, L19713, doi:10.1029/2012GL053370.

    • Search Google Scholar
    • Export Citation
  • Saenko, O. A., , and A. J. Weaver, 2004: What drives heat transport in the Atlantic: Sensitivity to mechanical energy supply and buoyancy forcing in the Southern Ocean. Geophys. Res. Lett., 31, L20305, doi:10.1029/2004GL020671.

    • Search Google Scholar
    • Export Citation
  • Seager, R., , Y. Kushnir, , M. Visbeck, , N. Naik, , J. Miller, , G. Krahmann, , and H. Cullen, 2000: Causes of Atlantic Ocean climate variability between 1958 and 1998. J. Climate, 13, 28452862, doi:10.1175/1520-0442(2000)013<2845:COAOCV>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Shaman, J., , R. M. Samelson, , and E. Skyllingstad, 2010: Air–sea fluxes over the Gulf Stream region: Atmospheric controls and trends. J. Climate, 23, 26512670, doi:10.1175/2010JCLI3269.1.

    • Search Google Scholar
    • Export Citation
  • Sijp, W. P., , and M. H. England, 2009: Southern Hemisphere westerly wind control over the ocean’s thermohaline circulation. J. Climate, 22, 12771286, doi:10.1175/2008JCLI2310.1.

    • Search Google Scholar
    • Export Citation
  • Smith, D. M., , R. Eade, , N. J. Dunstone, , D. Fereday, , J. M. Murphy, , H. Pohlmann, , and A. A. Scaife, 2010: Skilful multi-year predictions of Atlantic hurricane frequency. Nat. Geosci., 3, 846849, doi:10.1038/ngeo1004.

    • Search Google Scholar
    • Export Citation
  • Smith, R., and Coauthors, 2010: The Parallel Ocean Program (POP) reference manual: Ocean component of the Community Climate System Model (CCSM). LANL Tech. Rep. LAUR-10-01853, 141 pp.

  • Srokosz, M., , M. Baringer, , H. Bryden, , S. Cunningham, , T. Delworth, , S. Lozier, , J. Marotzke, , and R. Sutton, 2012: Past, present, and future change in the Atlantic meridional overturning circulation. Bull. Amer. Meteor. Soc., 93, 16631676, doi:10.1175/BAMS-D-11-00151.1.

    • Search Google Scholar
    • Export Citation
  • Steele, M., , R. Morley, , and W. Ermold, 2001: PHC: A global ocean hydrography with a high-quality Arctic Ocean. J. Climate, 14, 20792087, doi:10.1175/1520-0442(2001)014<2079:PAGOHW>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Sutton, R. T., , and D. L. R. Hodson, 2005: Atlantic Ocean forcing of North American and European summer climate. Science, 309, 115118, doi:10.1126/science.1109496.

    • Search Google Scholar
    • Export Citation
  • Thompson, D. W. J., , and S. Solomon, 2002: Interpretation of recent Southern Hemisphere climate change. Science, 296, 895899, doi:10.1126/science.1069270.

    • Search Google Scholar
    • Export Citation
  • Timmermann, A., , and H. Goosse, 2004: Is the wind stress forcing essential for the meridional overturning circulation? Geophys. Res. Lett., 31, L04303, doi:10.1029/2003GL018777.

    • Search Google Scholar
    • Export Citation
  • Toggweiler, J. R., , and B. Samuels, 1995: Effect of Drake Passage on the global thermohaline circulation. Deep-Sea Res. I, 42, 477500, doi:10.1016/0967-0637(95)00012-U.

    • Search Google Scholar
    • Export Citation
  • Toggweiler, J. R., , and B. Samuels, 1998: On the ocean’s large-scale circulation near the limit of no vertical mixing. J. Phys. Oceanogr., 28, 18321852, doi:10.1175/1520-0485(1998)028<1832:OTOSLS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Visbeck, M., , E. P. Chassignet, , R. Curry, , T. Delworth, , B. Dickson, , and G. Krahmann, 2003: The ocean’s response to North Atlantic Oscillation variability. The North Atlantic Oscillation: Climate Significance and Environmental Impact, Geophys. Monogr., Vol. 134, Amer. Geophys. Union, 113146.

    • Search Google Scholar
    • Export Citation
  • Wolfe, C. L., , and P. Cessi, 2010: What sets the strength of the middepth stratification and overturning circulation in eddying ocean models? J. Phys. Oceanogr., 40, 15201538, doi:10.1175/2010JPO4393.1.

    • Search Google Scholar
    • Export Citation
  • Wunsch, C., 2006: Abrupt climate change: An alternative view. Quat. Res., 65, 191203, doi:10.1016/j.yqres.2005.10.006.

  • Wunsch, C., , and R. Ferrari, 2004: Vertical mixing, energy, and the general circulation of the oceans. Annu. Rev. Fluid Mech., 36, 281314, doi:10.1146/annurev.fluid.36.050802.122121.

    • Search Google Scholar
    • Export Citation
  • Xu, X., , H. E. Hurlburt, , W. J. Schmitz Jr., , R. Zantopp, , J. Fischer, , and P. J. Hogan, 2013: On the currents and transports connected with the Atlantic meridional overturning circulation in the subpolar North Atlantic. J. Geophys. Res., 118, 502–516, doi:10.1002/jgrc.20065.

    • Search Google Scholar
    • Export Citation
  • Yashayaev, I., 2007: Hydrographic changes in the Labrador Sea, 1960-2005. Prog. Oceanogr., 73, 242276, doi:10.1016/j.pocean.2007.04.015.

    • Search Google Scholar
    • Export Citation
  • Yashayaev, I., , and J. W. Loder, 2009: Enhanced production of Labrador Sea water in 2008. Geophys. Res. Lett., 36, L01606, doi:10.1029/2008GL036162.

    • Search Google Scholar
    • Export Citation
  • Yeager, S. G., 2013: Understanding and predicting changes in North Atlantic sea surface temperature. Ph.D. thesis, University of Colorado, 176 pp.

  • Yeager, S. G., , and M. Jochum, 2009: The connection between Labrador Sea buoyancy loss, deep western boundary current strength, and Gulf Stream path in an ocean circulation model. Ocean Modell., 30, 207224, doi:10.1016/j.ocemod.2009.06.014.

    • Search Google Scholar
    • Export Citation
  • Yeager, S. G., , and G. Danabasoglu, 2012: Sensitivity of Atlantic meridional overturning circulation variability to parameterized Nordic Sea overflows in CCSM4. J. Climate, 25, 20772103, doi:10.1175/JCLI-D-11-00149.1.

    • Search Google Scholar
    • Export Citation
  • Yeager, S. G., , A. Karspeck, , G. Danabasoglu, , J. Tribbia, , and H. Teng, 2012: A decadal prediction case study: Late twentieth-century North Atlantic Ocean heat content. J. Climate, 25, 51735189, doi:10.1175/JCLI-D-11-00595.1.

    • Search Google Scholar
    • Export Citation
  • Zhang, L., , and C. Wang, 2013: Multidecadal North Atlantic sea surface temperature and Atlantic meridional overturning circulation variability in CMIP5 historical simulations. J. Geophys. Res., 118, 5772–5791, doi:10.1002/jgrc.20390.

    • Search Google Scholar
    • Export Citation
  • Zhang, R., 2010: Latitudinal dependence of Atlantic meridional overturning circulation (AMOC) variations. Geophys. Res. Lett., 37, L16703, doi:10.1029/2010GL044474.

    • Search Google Scholar
    • Export Citation
  • Zhang, R., , and T. Delworth, 2006: Impact of Atlantic multidecadal oscillations on India/Sahel rainfall and Atlantic hurricanes. Geophys. Res. Lett., 33, L17712, doi:10.1029/2006GL026267.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Time-mean fields from CONTROL: (a) AMOC; (b) AMOC at 26.5°N (gray) compared to RAPID observations (black); (c) barotropic streamfunction; and (d) SST difference from the merged Hadley–OI (see text) dataset. Averages are for 1988–2007 in (a),(c),(d) and from April 2004 to December 2007 in (b). Black and gray contour lines denote positive and negative, respectively.

  • View in gallery

    Time-mean (1988–2007) March MLD (color shade; contour level of 50 m) and sea ice edge (black contour, corresponding to an ice fraction of 15%) from CONTROL. Observed mean sea ice extent from SSM/I is also shown (red contour). Root-mean-square March MLD (computed over the 50 yr: 1958–2007) from CONTROL is overlaid in white contours (contour interval is 100 m, starting at 200 m). Thick black lines demarcate the Labrador Sea box region (60°–45°W, 53°–65°N) referred to in the text.

  • View in gallery

    Time series of March ice-covered area within the Labrador Sea box from CONTROL and SSM/I observations. Thin horizontal lines show the 1988–2007 mean values corresponding to the ice edge plotted in Fig. 2.

  • View in gallery

    Time series of anomalous potential temperature (shading) and potential density (σ2; contoured at 0.01 kg m−3; dashed lines show negative values) within the central Labrador Sea from (a) a compilation of hydrographic observations (Yashayaev 2007; Yashayaev and Loder 2009) and (b) CONTROL. (c),(d) As in (a),(b), but for anomalous salinity. The anomalies are computed relative to the 1960–2007 climatology at each depth level. CONTROL area averages were computed on depth levels within the box region (56°–49°W, 56°–61°N) in the vicinity of the Atlantic Repeat Hydrography Line 7 West (AR7W) section and include only grid cells where the bathymetry exceeds 3300 m. Model output from May of each year is used to reflect the spring timing of hydrographic measurements, although the difference from annual-mean output is small.

  • View in gallery

    First empirical orthogonal function (EOF1) of (a) merged Hadley–OI SST [contour interval (CI) = 0.1°C], (b) CONTROL SST (CI = 0.1°C), and (c) CONTROL AMOC (CI = 0.2 Sv). Gray shading is used for positive contours. (d) The associated normalized principal component time series. The domain used for computing the EOFs is the same as the region plotted [80°W–0°, 10°–70°N for (a),(b)]. All fields were first linearly detrended and smoothed with a 15-point Lanczos filter with cutoff period of 7 yr prior to EOF computation. Percentage of total variance explained (of the smoothed field) is given for each EOF.

  • View in gallery

    Hovmöller diagrams of annual AMOC strength anomaly (Sv) as a function of latitude and time from (a) the CONTROL simulation, (b) experiment M, (c) M+B (the sum of anomalies from these experiments), and (d) experiment B. No smoothing has been applied, either in the processing or in the plotting. Black (gray) circles in (d) indicate the approximate origins of positive (negative) AMOC anomalies referred to in the text.

  • View in gallery

    Comparisons with CONTROL of annual-mean AMOC strength as a function of latitude from experiments M and B as well as the sum of their anomalies (M+B), computed from (a),(c) raw annual-mean time series and (b),(d) low-pass-filtered time series. Temporal correlations are plotted in (a),(b) and root-mean-square differences from CONTROL are plotted in (c),(d). A 15-point Lanczos filter with cutoff period of 7 yr is used for (b),(d).

  • View in gallery

    As in Fig. 6, but for AMOC computed in density (σ2) space, such that the AMOC strength at each latitude is calculated as the maximum in density rather than depth space prior to the anomaly calculation.

  • View in gallery

    (top) Variance of annual-mean time series from CONTROL of (a) BSF, (b) SSH, and (c) depth-averaged upper-ocean (0–295 m) current speed. Also shown are the covariances of the same fields (d)–(f) between M and CONTROL and (g)–(i) between B and CONTROL. To a very good approximation, the sum of covariances plotted in (middle) and (bottom) equals the total variance from CONTROL plotted in (top). The contour levels are as follows: 0, 1, 2, 4, 6, 8, 10, 15, 20, 25, and 30 Sv2 for BSF; 0, 2, 4, 6, 8, 10, 15, 20, 30, 40, and 50 cm2 for SSH; and 0, 0.5, 1, 2, 3, 5, 7, 10, and 15 cm2 s−2 for velocity. Values are shaded above the first nonzero contour.

  • View in gallery

    Annual-mean time series of regionally averaged (a),(b) BSF and (c),(d) SSH anomalies from CONTROL, M, B, and the sum of anomalies M+B. The SPG region is the same as the Labrador Sea box (see Fig. 2); the STG region is defined as 80°–65°W, 26°–38°N. No smoothing has been applied. Note that positive BSF anomalies in the SPG in (a) correspond to a weaker (cyclonic) gyre, while positive BSF anomalies in the STG in (b) correspond to a stronger (anticyclonic) gyre.

  • View in gallery

    Lag correlations as a function of latitude of annual-mean Labrador Sea SSH (regionally averaged within the box 55°–50°W, 55°–60°N) with AMOC strength (computed as the maximum in depth) from (a) CONTROL and (b) experiment B. No time filtering has been used. Lead time is positive when Labrador Sea SSH precedes AMOC. The contour interval is 0.05 and values below 0.4 are not plotted.

  • View in gallery

    As in Fig. 6, but for (a) experiment B*, (b) experiment B.Q, (c) B.F+B.Q (the sum of anomalies from these experiments), and (d) experiment B.F. Experiment B* is identical to B, except that normal year winds are used for the computation of all turbulent fluxes (see text).

  • View in gallery

    As in Fig. 6, but for (a) experiment B.1, (b) experiment B.2, (c) experiment NYF, and (d) experiment M.SO.

  • View in gallery

    Anomalous annual-mean zonally averaged zonal wind stress τx (N m−2) in the Southern Hemisphere from CONTROL. The contour interval is 0.01 N m−2 with positive (negative) anomalies contoured in black (gray) with gray shading for positive values.

  • View in gallery

    Anomaly time series of winter fields from CONTROL regionally averaged over the Labrador Sea box region. (a) JFM-mean net air–sea freshwater flux and components converted into units of surface buoyancy loss (10−8 m2 s−3). (b) As in (a), but for JFM-mean net air–sea heat flux and components. (c) March-mean MLD, March-mean SSD, JFM-mean net surface buoyancy loss (−Bas; the sum of heat and freshwater buoyancy fluxes, multiplied by −1, so that positive values denote surface density gain), and December–March-mean NAO index. All fields are averaged over the ice-free ocean as determined by the March-mean ice fraction. Panel (c) is plotted in units of standard deviation of the respective time series. The anomalies are relative to the 1958–2007 time average.

  • View in gallery

    Anomaly time series of JFM-mean fields from CONTROL regionally averaged over the Labrador Sea box region. (a) The 10-m atmospheric potential temperature θ and specific humidity q together with the net surface buoyancy flux (this differs from the curve in Fig. 15c by a factor of −1). (b) As in (a), but after smoothing with a 15-point Lanczos filter with cutoff period of 7 yr. All curves are normalized by the standard deviations of the respective unfiltered time series. The regional averages of θ and q are computed over the entire Labrador Sea box region, while Bas is averaged over the ice-free subdomain.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 23 23 18
PDF Downloads 16 16 11

The Origins of Late-Twentieth-Century Variations in the Large-Scale North Atlantic Circulation

View More View Less
  • 1 National Center for Atmospheric Research,* Boulder, Colorado
© Get Permissions
Full access

Abstract

Surface forcing perturbation experiments are examined to identify the key forcing elements associated with late-twentieth-century interannual-to-decadal Atlantic circulation variability as simulated in an ocean–sea ice hindcast configuration of the Community Earth System Model, version 1 (CESM1). Buoyancy forcing accounts for most of the decadal variability in both the Atlantic meridional overturning circulation (AMOC) and the subpolar gyre circulation, and the key drivers of these basin-scale circulation changes are found to be the turbulent buoyancy fluxes: evaporation as well as the latent and sensible heat fluxes. These three fluxes account for almost all of the decadal AMOC variability in the North Atlantic, even when applied only over the Labrador Sea region. Year-to-year changes in surface momentum forcing explain most of the interannual AMOC variability at all latitudes as well as most of the decadal variability south of the equator. The observed strengthening of Southern Ocean westerly winds accounts for much of the simulated AMOC variability between 30°S and the equator but very little of the recent AMOC change in the North Atlantic. Ultimately, the strengthening of the North Atlantic overturning circulation between the 1970s and 1990s, which contributed to a pronounced SST increase at subpolar latitudes, is explained almost entirely by trends in the atmospheric surface state over the Labrador Sea.

The National Center for Atmospheric Research is sponsored by the National Science Foundation.

Corresponding author address: Stephen Yeager, National Center for Atmospheric Research, P.O. Box 3000, Boulder, CO 80307-3000. E-mail: yeager@ucar.edu

Abstract

Surface forcing perturbation experiments are examined to identify the key forcing elements associated with late-twentieth-century interannual-to-decadal Atlantic circulation variability as simulated in an ocean–sea ice hindcast configuration of the Community Earth System Model, version 1 (CESM1). Buoyancy forcing accounts for most of the decadal variability in both the Atlantic meridional overturning circulation (AMOC) and the subpolar gyre circulation, and the key drivers of these basin-scale circulation changes are found to be the turbulent buoyancy fluxes: evaporation as well as the latent and sensible heat fluxes. These three fluxes account for almost all of the decadal AMOC variability in the North Atlantic, even when applied only over the Labrador Sea region. Year-to-year changes in surface momentum forcing explain most of the interannual AMOC variability at all latitudes as well as most of the decadal variability south of the equator. The observed strengthening of Southern Ocean westerly winds accounts for much of the simulated AMOC variability between 30°S and the equator but very little of the recent AMOC change in the North Atlantic. Ultimately, the strengthening of the North Atlantic overturning circulation between the 1970s and 1990s, which contributed to a pronounced SST increase at subpolar latitudes, is explained almost entirely by trends in the atmospheric surface state over the Labrador Sea.

The National Center for Atmospheric Research is sponsored by the National Science Foundation.

Corresponding author address: Stephen Yeager, National Center for Atmospheric Research, P.O. Box 3000, Boulder, CO 80307-3000. E-mail: yeager@ucar.edu

1. Introduction

Changes in the strength of the Atlantic meridional overturning circulation (AMOC) lead the decadal variations in North Atlantic sea surface temperature (SST) in long coupled control simulations (Delworth and Mann 2000; Knight et al. 2005; Danabasoglu et al. 2012b) as well as in forced coupled simulations of the twentieth century (Medhaug and Furevik 2011; Zhang and Wang 2013). The Atlantic multidecadal variability (AMV; a climate index based on the detrended, basin-scale average of North Atlantic SSTs; see, e.g., Sutton and Hodson 2005), which has been observed over the past century, is therefore assumed to be related, at least in part, to overturning variations that at various times have opposed or added to the secular warming trend associated with anthropogenic forcings. There are adequate surface observations to establish a further link between these slow changes in the surface temperatures of the North Atlantic and far-reaching climate impacts from the Americas to India, including modulation of Atlantic hurricane activity (e.g., Sutton and Hodson 2005; Knight et al. 2006; Zhang and Delworth 2006; D. M. Smith et al. 2010). Given the prominent role that AMOC is hypothesized to play in decadal climate variations, there is a recognized need for advances in our understanding of the processes that modulate the strength of the AMOC and, more generally, the basin-scale circulation of the North Atlantic and our ability to accurately represent those processes in climate models (Liu 2012; Srokosz et al. 2012).

Coupled general circulation models (CGCMs) will necessarily be key tools for investigating AMOC-related decadal climate variability for the foreseeable future, because observational sampling of the ocean is not nearly sufficient to permit in-depth study of basin-scale variations on such long time scales. However, there are many reasons to question the fidelity of AMOC variability and associated mechanisms diagnosed from the current generation of CGCMs used for the Intergovernmental Panel on Climate Change (IPCC) assessment reports [e.g., Fifth Assessment Report (AR5)]. The magnitude, preferred time scale, and dominant mechanism of AMOC variability can vary substantially from model to model (e.g., Liu 2012), and there is sensitivity to the subgrid-scale parameterizations used in any particular model (e.g., Fox-Kemper et al. 2011; Danabasoglu et al. 2012b; Yeager and Danabasoglu 2012). The long (multicentury) control simulations used to study AMOC-related intrinsic climate variability generally lack resolved eddies and tend to be characterized by large mean biases in the North Atlantic that are exacerbated by coupling. The AMOC variability obtained in such CGCM simulations cannot be rigorously evaluated for realism, because the observed records, even of surface climate fields, are too short. Long coupled experiments often lack the effects of historical radiative forcings but, even if those are included, the expectation of phase differences between the internal variability of the model and that of nature precludes a close comparison with observation. The presumption that CGCM experiments accurately simulate the key variability mechanisms at work in the climate system can really only be tested in initialized (e.g., decadal prediction) experiments but, even then, the inexact initialization forces a comparison of the single realization of Earth’s climate history with an ensemble of coupled model trajectories.

Insight into the connections between Atlantic variability of the past few decades and AMOC can be gained from forced ocean general circulation model (OGCM) simulations whose fidelity can be tested by direct comparison with available observations. The confidence in mechanisms diagnosed from such simulations depends upon the model’s ability to reproduce known features of the mean and time-varying ocean state. Reliable historical forcing data are only available from about 1950 onward, and so this technique cannot address the topic of intrinsic multidecadal variability in the absence of strong external radiative forcings, but it can help to identify important AMOC driving mechanisms. A fairly robust result that emerges from such OGCM hindcast studies is that there was a strengthening of the AMOC and northward heat transport (NHT) over the last few decades of the twentieth century, and this decadal-scale spinup of the North Atlantic circulation was associated with changes in the North Atlantic Oscillation (NAO) (e.g., Häkkinen 1999; Eden and Willebrand 2001; Bentsen et al. 2004; Beismann and Barnier 2004; Böning et al. 2006; Biastoch et al. 2008; Robson et al. 2012a). In these and other studies (e.g., Marshall et al. 2001; Visbeck et al. 2003; Brauch and Gerdes 2005; Lohmann et al. 2009), the slow AMOC variations of the recent past are interpreted as delayed baroclinic adjustments to pronounced NAO-related forcing changes. The buoyancy forcing of the subpolar gyre (SPG) is strongly tied to NAO conditions, and thus variations in the formation rates of the water masses that comprise the North Atlantic Deep Water (NADW) are found to covary with the NAO (Marsh 2000; Khatiwala et al. 2002). The enhanced production of Labrador Sea water (LSW), in particular, during the high-NAO winters of the 1970s, 1980s, and early 1990s would appear to explain the increase in AMOC and SPG circulation strength over the last three decades of the twentieth century. The associated increase in NHT has been identified as a significant contributor to the abrupt SPG warming observed in the mid-1990s, and it accounts for the significant skill at predicting high-latitude North Atlantic SST in several recent decadal prediction studies (Grist et al. 2010; Robson et al. 2012a,b; Yeager et al. 2012). The NAO would appear to be of primary importance in driving recent Atlantic circulation changes, but we note that recent studies have also highlighted the potential significance of the east Atlantic pattern of sea level pressure variation and associated atmospheric blocking events (Msadek and Frankignoul 2009; Häkkinen et al. 2011).

While it seems evident that North Atlantic thermohaline forcing is an important driver of changes in NADW properties and hence of AMOC variations, the relative impacts of various remote and local wind and buoyancy forcings on the AMOC remains unclear. In the limit of weak interior diapycnal mixing, energetic considerations suggest a fundamental role for winds, particularly Southern Ocean winds, in sustaining the global overturning circulation (e.g., Toggweiler and Samuels 1995, 1998; Wunsch and Ferrari 2004; Kuhlbrodt et al. 2007). Some models indeed show weakened overturning when the momentum flux into the ocean is completely switched off (Timmermann and Goosse 2004; Saenko and Weaver 2004), but the magnitude of this sensitivity would appear to depend on the model representation of air–sea flux feedbacks (Rahmstorf and England 1997). Based on its importance in the steady-state energy budget of the ocean, it has been hypothesized that mechanical forcing variations may also play a significant role in driving transient AMOC changes with important climate implications on long (glacial–interglacial) time scales (Wunsch 2006), and this is supported by idealized model studies that indicate that AMOC scales linearly with the magnitude of Southern Ocean zonal wind stress (Nikurashin and Vallis 2012). On the decadal time scales of interest here (far shorter than the equilibration time scale of the global overturning circulation), the role of remote mechanical forcing in driving AMOC variability is presumably less important. However, one recent OGCM study (Lee et al. 2011) finds that most of the AMOC-driven increase in North Atlantic upper-ocean heat content since the mid-twentieth century was caused by the strengthening of Southern Ocean zonal winds associated with the recent trend in the southern annular mode (SAM; Thompson and Solomon 2002).

The aim of this paper is to contribute to our understanding of the mechanisms of AMOC variability by identifying the key forcing elements that explain the simulated historical variability of the large-scale North Atlantic circulation between 1948 and 2007. We analyze a global, non-eddy-resolving ocean–sea ice configuration of the Community Earth System Model, version 1 (CESM1), which is forced at the surface with historical, interannually varying atmospheric state fields. Our control experiment is one of several such integrations performed by modeling groups around the world, as phase II of the Coordinated Ocean-Ice Reference Experiments (CORE-II) organized by the Climate Variability and Predictability (CLIVAR) Working Group on Ocean Model Development (WGOMD; http://www.clivar.org/organization/wgomd/core). A multimodel comparison of these CORE-II experiments (Danabasoglu et al. 2014) highlights some significant differences in Atlantic mean state that have been attributed to differences in subgrid-scale parameterizations and parameter choices as well as to differences in grid resolution. Nevertheless, in ongoing work that has not yet been published, many common features of simulated interannual to decadal variability have been identified across the suite of participating models, such as the aforementioned AMOC increase over the last three decades of the twentieth century and an associated slow spinup of the cyclonic subpolar gyre circulation, which the present study seeks to elucidate.

A powerful technique for probing mechanisms in OGCM simulations is to perform sensitivity experiments in which the variability of certain fluxes (e.g., wind, buoyancy) is selectively suppressed (e.g., Eden and Willebrand 2001; Böning et al. 2006; Biastoch et al. 2008; Robson et al. 2012a). We adopt this approach and extend it to systematically assess the relative impacts on AMOC of year-to-year changes in various atmospheric forcing components (fluxes of heat, freshwater, and momentum) and subcomponents (e.g., the latent and sensible heat fluxes), and in particular examine the impacts of atmospheric state variability in two key regions: the Labrador Sea and the Southern Ocean. The NAO-related air–sea fluxes that have the greatest impact on AMOC are thus identified and their effects contrasted with those associated with the observed SAM trend. The present study is similar in many respects to Biastoch et al. (2008), who used an earlier version of CORE forcing, and this permits some direct comparisons with their findings but it is worth enumerating some salient differences: 1) we use entirely different ocean and sea ice models, which both have somewhat coarser resolution than theirs; 2) we apply a much weaker global sea surface salinity restoring (relaxation time scale of 4 yr as opposed to their 180 days over 50 m); 3) we do no restoring of model temperature or salinity anywhere below the surface (in contrast to their use of a “robust diagnostic” method to prevent polar water mass drift); 4) our experiments should include any effects associated with Nordic Seas overflows because the model has an overflow parameterization (Danabasoglu et al. 2010) and we enforce only a weak constraint on Nordic Sea temperature and salinity variability (see difference 2 above); and 5) the scope of our sensitivity analysis is considerably broader, since we aim to quantify the relative influence of various forcing subcomponents and geographic regions on AMOC variability.

After describing the model and experimental setup in section 2, we begin by establishing the fidelity of the hindcast control simulation by comparing the simulated North Atlantic mean and variability to a variety of observations (section 3). The relative contributions of buoyancy and wind forcing to Atlantic Ocean variability on interannual and longer time scales is examined in section 4. The impacts of regional forcing variations are then assessed for the Labrador Sea and Southern Ocean in sections 5 and 6, respectively. Having established the dominance of Labrador Sea forcing in driving AMOC variations of the late twentieth century, we turn in section 7 to an examination of the origins of Labrador Sea flux variability. Section 8 contains a final discussion and conclusions.

2. Experimental setup

a. The model

All of the simulations to be analyzed are coupled ocean–sea ice configurations of the CESM1, whose general description is given in Gent et al. (2011). The ocean model is the Parallel Ocean Program, version 2 (POP2; R. Smith et al. 2010b), at nominal 1° horizontal resolution with 60 vertical levels. This is coupled to the Los Alamos sea ice model, version 4 (CICE4; Hunke and Lipscomb 2008), which runs on the same horizontal grid as the ocean. Further details regarding the POP2 and CICE4 models as implemented in CESM1 can be found in Danabasoglu et al. (2012a) and Holland et al. (2012), respectively.

b. The forcing

The coupled ocean–sea ice model is forced at the surface with the CORE-II historical atmospheric datasets, which are freely distributed together with release notes and support code by the WGOMD. This dataset, which spans 1948–2007,1 is based primarily on the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis (Kistler et al. 2001) but includes flux parameters and prescribes adjustments based on other sources in an attempt to assemble the best forcing suite for ocean and sea ice modeling (Large and Yeager 2004, 2009). No temperature restoring is used, but there is a very weak global restoring of model surface salinity to observed climatology with a piston velocity of 50 m per 4 yr to prevent salinity drift. The model uses a virtual salinity flux at the surface as opposed to true freshwater flux. Further details of the model configuration and how the solution from our present control simulation differs from other comparably forced models can be found in Danabasoglu et al. (2014).

We will be referring to the effects of specific fluxes, and so it is useful to review here the ocean model forcing methodology outlined in detail in Large and Yeager (2004) and Large and Yeager (2009). The fluxes of momentum , freshwater F, and heat Q that drive the ocean model are partitioned into air–sea (as) and ice–ocean (io) fluxes depending on the fraction of ice-free ocean fo within a particular ocean grid cell,
e1
e2
e3
In the CORE-II dataset, a monthly, interannually varying continental discharge dataset (Dai et al. 2009) is used to prescribe a river runoff flux R. This continental runoff is distributed as a surface freshwater flux over ocean cells in the vicinity of river mouths. We will account for it in this study by including it in the Fas term even though it is not an air–sea flux (in particular, it is not masked in sea ice regions). It is important to note that the dataset used for river runoff in CORE-II uses model-derived regression relationships to specify runoff from Greenland, rather than actual measured discharge rates, and so any freshwater flux variation associated with the melting of glaciers on Greenland is absent in these experiments.
Our experiments are performed in a coupled ocean–sea ice configuration, and so fo, , Fio, and Qio will in general be unconstrained, prognostic fields. Our focus will be on exploring controlled perturbations to , Fas, and Qas, which have the following dependence on the atmospheric and oceanic states:
e4
e5
e6
e7
e8
where P is precipitation; E is evaporation; QS is shortwave radiation; QL is longwave radiation; and QE and QH are the latent and sensible heat fluxes, respectively. The longwave radiation splits into downward and upward components, with the latter solely a function of the SST. Bulk formulas parameterize the turbulent fluxes (, E, QE, and QH) as functions of differences between the prescribed near-surface atmospheric state (wind , potential temperature θ, specific humidity q, and density ρ) and the evolving oceanic state (horizontal surface velocity and SST). All of the turbulent fluxes depend on the wind speed through the term , and they require specification of transfer coefficients for drag CD, sensible heat CH, and evaporation CE. The computation of evaporation relies on the assumption that the specific humidity of air at the ocean surface is saturated at qsat(SST), and the latent heat flux is related to the evaporation through the latent heat of vaporization Λυ. Finally, cp is the specific heat of air.

It is important to note that not all forcing fields in the CORE-II dataset vary interannually over the full period 1948–2007. While the atmospheric state fields (, θ, q, and ρ) are available at 6-hourly resolution over the full time period, monthly precipitation is available only after 1979 and daily downward shortwave and longwave radiation are available only after 1984. Furthermore, the lack of adequate observations of high-latitude precipitation obliges us to use a monthly climatology for each year north of 68°N. Climatological values are used to fill in years for which historical values are lacking for a particular field [see Large and Yeager (2009) and CORE-II release notes].

Finally, our sensitivity experiments make use of the “normal year” forcing (NYF) dataset described in Large and Yeager (2004), which is a single repeatable annual cycle of forcing fields at the same temporal resolution as in the full multiyear dataset. This single cycle is climatological, but it is not a simple average of forcing fields. In particular, it is designed to preserve the high-frequency variance (frequencies of annual period and higher) present in the full dataset as well as preserve the coherent propagation of weather signals.

c. The experiments

The initial condition of the control experiment (CONTROL) was a state of rest and the January-mean potential temperature and salinity climatology from the Polar Science Center Hydrographic Climatology (PHC2), a blending of the Levitus et al. (1998) dataset with Arctic Ocean modifications based on Steele et al. (2001). It was then spun up through five consecutive 60-yr cycles of 1948–2007 forcing such that, after every 60 yr of model integration, the atmospheric state transitioned from December 2007 directly into January 1948 and then the cycle was repeated. Our analysis focuses on the final 50 yr of the final cycle of integration (simulation years 251–300; forcing years 1958–2007), in order to avoid some of the transient behavior associated with the forcing transition (Doney et al. 2003).

Flux perturbation experiments that isolate the momentum-forced and buoyancy-forced interannual variability—referred to as experiments M and B, respectively (Table 1)—were branched from CONTROL at the start of simulation year 242 (near the start of the fifth forcing cycle) and integrated through simulation year 300. In these sensitivity experiments, NYF was used to greatly suppress interannual variability in the fluxes of either buoyancy or momentum, although the dependence on the time-evolving ocean state in the turbulent fluxes as well as the upward longwave means that the use of NYF for these flux components does not completely eliminate interannual variability (see Table 1 caption). This effect is small compared to the signals of interest here.

Table 1.

Guide to the forcings used in each experiment. An asterisk indicates that the variables in question are interannually varying between 1948 and 2007 to the extent permitted by the CORE-II dataset [e.g., implies that atmospheric ρ, q, and are all fully varying as in CONTROL], while the absence of an asterisk indicates that normal year fluxes and/or state fields were used. Note that there will be some interannual variation in normal year fluxes that are functions of the ocean state, since , Δθ = θ − SST*, Δq = qqSAT(SST*), and . The Labrador Sea box is defined as the region in 60°–45°W and 53°–65°N.

Table 1.

Variations on the M and B experiments were run to assess the impacts of particular fluxes and to probe regional effects. The relative impacts of surface-forced temperature T and salinity S variability are isolated in experiments B.Q and B.F, which make use of interannually varying forcing only for the buoyancy fluxes of heat and freshwater, respectively. Experiment B.1 examines the ocean response to variations in the turbulent buoyancy fluxes, and B.2 probes the effects of time-varying turbulent buoyancy forcing within a Labrador Sea box region (60°–45°W, 53°–65°N; see Fig. 2). We examine one variant of the M experiment in which interannual wind variability is restricted to Southern Ocean latitudes south of 35°S (experiment M.SO).

A pure NYF case was also branched from CONTROL at year 242 in order to quantify the drift associated with forward integration with NYF buoyancy forcing, because the initial condition obtained after four cycles of CORE-II forcing is quite different from the state that would correspond to a NYF forcing equilibrium. It is found that there is a drift toward weaker AMOC at all latitudes in the NYF experiment (Fig. 13c) that results from the adjustment of the ocean density structure to NYF buoyancy forcing. We have looked at the 60-yr linear trends in AMOC in all experiments to ascertain which experiments are strongly impacted by the long-term drift associated with NYF forcing (not shown). A comparison of experiments NYF and B.1 revealed that changing from NYF to interannually varying turbulent buoyancy fluxes changes the sign of the AMOC trend from weakly negative to strongly positive. Thus, the downward drift signal is presumably present in experiments that use NYF forcing for turbulent buoyancy fluxes (M and M.SO), and so we remove the drift diagnosed from the NYF experiment (the anomalies plotted in Fig. 13c) before plotting anomaly time series from these experiments. In the experiments that only partially use NYF turbulent buoyancy fluxes (B.2, B.F, and B.Q), the argument for such a drift correction is less clear. We have opted to fully drift correct experiment B.2, and we remove half of the NYF drift (Fig. 13c) from each of experiments B.F and B.Q. There is some sensitivity to the drift-correction choice in the B.F and B.Q metrics listed in Table 2, but otherwise our results are not strongly impacted by these choices. Unless otherwise indicated, the final 20 yr of the simulations (corresponding to 1988–2007) are used for computing time-mean quantities.

Table 2.

Select correlations and root-mean-square differences (in parentheses; in Sverdrups) between annual-mean AMOC strength anomalies as a function of latitude and year [AMOC(ϕ, t)] shown in the Hovmöller diagrams of Figs. 6, 8, 12, and 13. These statistics are based on years 1958–2007 and on the full latitude range of 30°S ≤ ϕ ≤ 70°N. The first row in each entry gives the raw annual values, and the second row gives the low-pass-filtered statistics after smoothing with a 15-point Lanczos filter with cutoff period of 7 yr. Rows with “(σ2)” give values based on the AMOC maximum in density (Fig. 8); all others are based on the AMOC maximum in depth.

Table 2.

3. Overview of North Atlantic mean and variability from the CORE-II control simulation

The CONTROL experiment appears to have a reasonable mean overturning circulation, with a maximum strength of roughly 26 Sv (1 Sv ≡ 106 m3 s−1) in a small recirculation cell centered at about 37°N (Fig. 1). Direct comparison with the observed AMOC profile at 26.5°N [Rapid Climate Change (RAPID); Cunningham et al. 2007] shows excellent agreement in terms of the mean magnitude and shape of the meridional transport distribution in the upper 2500 m, but the deep, southward return flow is shallower than in RAPID and there is no discernible Antarctic Bottom Water (AABW) circulation in the zonal integral of abyssal velocity in the model. The strength of the barotropic circulation in the subtropical gyre (STG) is about 40 Sv and exceeds 40 Sv in the SPG (Fig. 1c). STG transport is low compared to available observation-based estimates due to the coarse model resolution and the correspondingly high horizontal viscosity needed for numerical reasons (Jochum et al. 2008). However, the SPG circulation strength is actually comparable to observational estimates of depth-integrated equatorward transport near 55°N (Pickart et al. 2002), and it is quite consistent with Xu et al. (2013), who report 37–42 Sv near 53°N based on available observations as well as simulations with an eddy-resolving model. The barotropic streamfunction (BSF) and SST fields exhibit unrealistic spatial structures that are ubiquitous in this class of (non-eddy-resolving) model: the too-broad Gulf Stream (GS) separates too far north of Cape Hatteras and then remains too zonal, resulting in 4°–5°C SST biases of opposite sign along the North American coast and off the Grand Banks of Newfoundland (Figs. 1c,d). The merged SST dataset put together by Hurrell et al. (2008) (freely available at https://climatedataguide.ucar.edu/climate-data/merged-hadley-noaaoi-sea-surface-temperature-sea-ice-concentration-hurrell-et-al-2008), referred to as the merged Hadley–optimum interpolation (OI) SST, is used here as the observational benchmark. The subpolar seas are characterized by positive SST [and sea surface salinity (SSS); not shown] biases, which are most pronounced in the Labrador Sea.

Fig. 1.
Fig. 1.

Time-mean fields from CONTROL: (a) AMOC; (b) AMOC at 26.5°N (gray) compared to RAPID observations (black); (c) barotropic streamfunction; and (d) SST difference from the merged Hadley–OI (see text) dataset. Averages are for 1988–2007 in (a),(c),(d) and from April 2004 to December 2007 in (b). Black and gray contour lines denote positive and negative, respectively.

Citation: Journal of Climate 27, 9; 10.1175/JCLI-D-13-00125.1

Notwithstanding these familiar gyre circulation biases, the model maintains a fairly realistic winter sea ice edge compared to satellite observations [Special Sensor Microwave Imager (SSM/I); Comiso 2012] and the region of most active deep convection is along the ice edge in the western Labrador Sea, where mixed layer depths (MLDs) in late winter reach to about 1500 m on average (Fig. 2). The fact that the deepest and most variable Atlantic MLDs in the model are found in the Labrador Sea and that MLDs there regularly exceed 1000 m is broadly consistent with observational estimates (de Boyer Montégut et al. 2004). However, the region where winter MLDs exceed 1200 m in Fig. 2 would appear to be much more extensive than seen in observations from this region (Marshall et al. 1998; Pickart et al. 2002; Lavender et al. 2002), which show only very localized patches of such deep winter mixing mainly in the southwestern quadrant of the Labrador Sea. Danabasoglu et al. (2014) compare the March-mean MLD from this NCAR hindcast to the World Ocean Atlas climatology (Locarnini et al. 2010) using a standard density-based MLD criterion. This comparison likewise suggests that the winter convection in our simulation is probably too robust, although one must bear in mind that the observational sampling of this region in wintertime and in the vicinity of the ice edge remains poor. The interannual variability of the winter MLD in the Labrador Sea is concentrated along the sea ice edge but with a lobe of high variability concentrated in the southeast corner of the Labrador Sea box region. The region of greatest variability is thus somewhat displaced from the region of deepest mean mixing, a phenomenon that has been noted in fully coupled simulations of CESM1 (Yeager and Danabasoglu 2012; Danabasoglu et al. 2012b). A second, weaker and less extensive center of convective activity is apparent in the Norwegian Sea, off the coast of Svalbard (Fig. 2).

Fig. 2.
Fig. 2.

Time-mean (1988–2007) March MLD (color shade; contour level of 50 m) and sea ice edge (black contour, corresponding to an ice fraction of 15%) from CONTROL. Observed mean sea ice extent from SSM/I is also shown (red contour). Root-mean-square March MLD (computed over the 50 yr: 1958–2007) from CONTROL is overlaid in white contours (contour interval is 100 m, starting at 200 m). Thick black lines demarcate the Labrador Sea box region (60°–45°W, 53°–65°N) referred to in the text.

Citation: Journal of Climate 27, 9; 10.1175/JCLI-D-13-00125.1

It is difficult to assess the fidelity of the simulated variations in Labrador Sea MLD, but other fields can readily be evaluated against observed time series to get a sense of the model hindcast skill in this key deep-water mass formation region. A straightforward comparison of Labrador Sea March sea ice coverage with satellite observations reveals a too-weak response of the sea ice component of the model to the intense positive NAO winters of the early 1980s and 1990s (Fig. 3). This is probably related to ocean model biases (in particular, the too warm Labrador Sea surface waters) that prevent ice growth in this region, but forcing and sea ice model deficiencies are likely also to blame. The correlation with the observed time series is nevertheless quite high (r = 0.86).

Fig. 3.
Fig. 3.

Time series of March ice-covered area within the Labrador Sea box from CONTROL and SSM/I observations. Thin horizontal lines show the 1988–2007 mean values corresponding to the ice edge plotted in Fig. 2.

Citation: Journal of Climate 27, 9; 10.1175/JCLI-D-13-00125.1

Long-term monitoring of the region via repeat hydrographic sections and floats has revealed decadal variations in the T and S properties of central Labrador Sea water (Dickson et al. 2002; Yashayaev 2007; Yashayaev and Loder 2009). A direct comparison with the observation-based T and S profile time series analyzed in these studies (not shown) mostly highlights the model mean biases in this region (warm and salty), which are less pronounced in density because of the compensating effects of T and S contributions. A more encouraging correspondence is seen when deviations from the long-term (1960–2007) time mean are compared (Fig. 4), although the comparison is rough. As in Yashayaev (2007), we generate the time series by averaging over the “central Labrador Sea” defined using the bathymetric contour of 3300 m, but we perform the averaging in depth coordinates rather than density coordinates. As in observations, the variability of area-averaged central Labrador Sea T and S in CONTROL is dominated by a decadal-scale evolution from warm/salty conditions in the 1960s and early 1970s to cold/fresh conditions in the 1990s and then back to warm/salty anomalies in the 2000s. The density anomalies are largely set by temperature in both model and observation. There are many differences in the details: the model generally has weaker T and S anomalies than observed except in the most recent years when it shows much stronger anomalies; the strong anomalies of the mid-1980s are of the wrong sign in CONTROL; and the model shows more negative salinity anomalies than observed in the upper 2000 m prior to about 1996. The pronounced freshening of LSW between the 1960s and 1990s, particularly at around 2000-m depth, is therefore less dramatic in CONTROL than in observations (Dickson et al. 2002). Nevertheless, the low-frequency density variations are quite well represented, particularly later in the record when the quality of the observed record is highest (here and throughout, the phrase “low frequency” refers to variations on decadal and longer time scales). The fact that Labrador Sea density anomalies in CONTROL are comparable to or somewhat weaker than observed helps to allay our concerns about the impacts of anemic sea ice variability (Fig. 3) on deep-water mass formation.

Fig. 4.
Fig. 4.

Time series of anomalous potential temperature (shading) and potential density (σ2; contoured at 0.01 kg m−3; dashed lines show negative values) within the central Labrador Sea from (a) a compilation of hydrographic observations (Yashayaev 2007; Yashayaev and Loder 2009) and (b) CONTROL. (c),(d) As in (a),(b), but for anomalous salinity. The anomalies are computed relative to the 1960–2007 climatology at each depth level. CONTROL area averages were computed on depth levels within the box region (56°–49°W, 56°–61°N) in the vicinity of the Atlantic Repeat Hydrography Line 7 West (AR7W) section and include only grid cells where the bathymetry exceeds 3300 m. Model output from May of each year is used to reflect the spring timing of hydrographic measurements, although the difference from annual-mean output is small.

Citation: Journal of Climate 27, 9; 10.1175/JCLI-D-13-00125.1

A crucial test of any hindcast simulation is its ability to reproduce observed spatiotemporal patterns of SST variability. This is not guaranteed in simulations driven by bulk flux formulas without temperature restoring, particularly in regions where ocean heat transport is a significant component of the upper-ocean heat budget. The CONTROL hindcast does quite well at reproducing the observed AMV pattern over the latter half of the twentieth century (Fig. 5), which we define here as the first empirical orthogonal function (EOF) of detrended and low-pass-filtered2 North Atlantic annual-mean SST. The merged Hadley–OI product shows the familiar pattern (see, e.g., Sutton and Hodson 2005) of basinwide anomalies of a single sign with variability concentrated in four high-latitude regions: the Labrador Sea, the Irminger Sea, north of Iceland, and east of Newfoundland. The simulated AMV is also dominated by a broad, single-signed pattern with largest amplitude in the subpolar gyre, but differences are apparent. The model shows excessive variability in the central SPG and the region of opposite-signed anomalies just south of Nova Scotia is much stronger than observed (Fig. 5b), presumably because of unrealistic GS separation. Despite these disparities, the pattern correlation of the two EOFs is about 0.9 and the temporal correlation of the principal component (PC) time series is 0.85. In fact, the CONTROL simulation has one of the best representations of the AMV out of all the models participating in a recent CORE-II intercomparison. The second EOF of low-pass-filtered North Atlantic SST explains another 20% of decadal SST variance in both observations and CONTROL, with correlations of 0.7 and 0.8 for the EOF patterns and PC time series, respectively (not shown).

Fig. 5.
Fig. 5.

First empirical orthogonal function (EOF1) of (a) merged Hadley–OI SST [contour interval (CI) = 0.1°C], (b) CONTROL SST (CI = 0.1°C), and (c) CONTROL AMOC (CI = 0.2 Sv). Gray shading is used for positive contours. (d) The associated normalized principal component time series. The domain used for computing the EOFs is the same as the region plotted [80°W–0°, 10°–70°N for (a),(b)]. All fields were first linearly detrended and smoothed with a 15-point Lanczos filter with cutoff period of 7 yr prior to EOF computation. Percentage of total variance explained (of the smoothed field) is given for each EOF.

Citation: Journal of Climate 27, 9; 10.1175/JCLI-D-13-00125.1

Yeager et al. (2012) showed that time series of SPG-averaged heat content and SST anomalies from CONTROL are in excellent agreement with observations, better than might be surmised from the basin-scale comparison of Fig. 5. In that study, a heat budget of the SPG region revealed that changes in advective heat transport convergence associated with a multidecadal spinup of the large-scale overturning and high-latitude gyre circulations (we will refer to these jointly as the “thermohaline” or “buoyancy driven” circulation) was responsible for the abrupt warming of the SPG in the mid-1990s. The link between basin-scale thermohaline circulation variations and high-latitude decadal SST variability is suggested by comparing the PC time series of the dominant low-frequency AMOC and SST EOFs (Fig. 5), although the time series are clearly too short to establish a statistically significant relationship between AMOC and AMV. The AMOC variance is dominated by a large-scale decadal modulation that roughly matches the AMV transition from high (warm) in the 1950s and 1960s, to low (cold) in the 1970s and 1980s, back to high in the 1990s and 2000s. The overturning exhibits a large-scale weakening between the early 1950s and late 1970s followed by a strengthening that lasted into the mid-1990s. This decadal AMOC variation is most pronounced in midlatitudes, between about 30° and 50°N, but is coherent at all latitudes in the Atlantic (Fig. 5c). The AMOC strengthening leads the recent increase in subpolar SST (Fig. 5d), consistent with the explanation for the mid-1990s warming of the SPG offered by Yeager et al. (2012).

4. Buoyancy- and momentum-forced variability

Consistent with several other recent ocean hindcast studies (e.g., Böning et al. 2006; Biastoch et al. 2008; Robson et al. 2012a), we find that the late-twentieth-century AMOC variability in CONTROL can, to an excellent degree of approximation, be understood and analyzed as a linear superposition of anomalies associated with time-varying momentum and buoyancy forcing (isolated in experiments M and B, respectively; see Table 1). In addition, the low-frequency component of North Atlantic AMOC variability is primarily a response to high-latitude buoyancy forcing anomalies. Hovmöller diagrams of annual-mean AMOC strength (the maximum in depth of the AMOC streamfunction) as a function of latitude and time show that, consistent with the first AMOC EOF, the AMOC variability of CONTROL is dominated at all latitudes by the transition to relatively strong overturning beginning in the mid-1980s following relatively weak overturning in the preceding decades (Fig. 6a). Computing AMOC strength at a fixed depth level (e.g., 1000 m) gives qualitatively similar results. North of the equator, the low-frequency variability in CONTROL appears most associated with buoyancy forcing (Fig. 6d) while much of the higher-frequency (interannual) variability at all latitudes is most associated with momentum forcing (Fig. 6b). The linear superposition of M and B anomalies explains almost all the AMOC variability of CONTROL (Fig. 6c). Note that the M anomalies have had the NYF drift signal (Fig. 13c) removed as explained in section 2c.

Fig. 6.
Fig. 6.

Hovmöller diagrams of annual AMOC strength anomaly (Sv) as a function of latitude and time from (a) the CONTROL simulation, (b) experiment M, (c) M+B (the sum of anomalies from these experiments), and (d) experiment B. No smoothing has been applied, either in the processing or in the plotting. Black (gray) circles in (d) indicate the approximate origins of positive (negative) AMOC anomalies referred to in the text.

Citation: Journal of Climate 27, 9; 10.1175/JCLI-D-13-00125.1

The relative contributions of momentum and buoyancy forcing to late-twentieth-century variations in AMOC strength are quantified as functions of latitude in Fig. 7, as correlations with and root-mean-square (rms) differences from the AMOC strength in CONTROL. When interannual fluctuations are included in the analysis, both metrics show a dominant influence of B north of about 35°N, except for the 50°–60°N latitude band, while M explains most of the total variance south of 35°N (Figs. 7a,c). The variance maximum at 35°N (Fig. 6a) is a feature of this model associated with DWBC interaction with topography off Cape Hatteras whose realism is unclear. The introduction of the Nordic Seas overflow parameterization greatly reduces the prominence of this variance maximum but does not eliminate it (Yeager and Danabasoglu 2012). The model circulation response does not cleanly split into momentum- and buoyancy-forced perturbations at this location, resulting in a relative minimum and maximum of the M+B correlation and rms distributions, respectively (Fig. 7). Temporal smoothing with a decadal filter (Figs. 7b,d) confirms the visual impression obtained from the Hovmöller diagrams regarding the low-frequency AMOC behavior: B largely explains the decadal variability in AMOC north of the equator, while M accounts for most of the decadal variability south of the equator.

Fig. 7.
Fig. 7.

Comparisons with CONTROL of annual-mean AMOC strength as a function of latitude from experiments M and B as well as the sum of their anomalies (M+B), computed from (a),(c) raw annual-mean time series and (b),(d) low-pass-filtered time series. Temporal correlations are plotted in (a),(b) and root-mean-square differences from CONTROL are plotted in (c),(d). A 15-point Lanczos filter with cutoff period of 7 yr is used for (b),(d).

Citation: Journal of Climate 27, 9; 10.1175/JCLI-D-13-00125.1

The B experiment suggests a southward propagation of AMOC anomalies originating in the high northern latitudes with latitudinally dependent propagation speeds (Fig. 6d). The positive anomalies that emanate from close to 60°N in 1973, 1984, and 1990 correspond to the appearance of anomalously dense water in the central Labrador Sea in those years (Fig. 4), while the opposite is true for the negative anomalies identified in 1970 and 1979. The relatively slow propagation between about 45° and 35°N presumably reflects the existence of interior advective pathways of NADW between Newfoundland and Cape Hatteras, with fast coastal wave processes dominant elsewhere, as discussed in Zhang (2010), who analyzed AMOC in density space from a coupled climate simulation. The conclusions drawn from Fig. 6 are not much changed when AMOC strength anomalies from CONTROL, M, and B are computed in density coordinates, rather than depth coordinates (Fig. 8). As Zhang (2010) explains, AMOC in density space has a maximum north of 45°N because the strong (horizontal) subpolar gyre circulation, which largely cancels in the zonal integral in depth coordinates, is now tallied as part of the “overturning.” It follows that the AMOC variance maximum shifts from subtropical to subpolar latitudes, but we still find that buoyancy forcing accounts for most of the decadal AMOC variability north of the equator (Fig. 8). To the extent that AMOC in density space corresponds to horizontal gyre circulation north of about 45°N, it follows from Fig. 8 that low-frequency variations in the strength of the subpolar gyre circulation are largely buoyancy driven, rather than wind driven, with bottom pressure torque playing a significant role in the barotropic vorticity balance. A vorticity budget of the CONTROL simulation shows that this is indeed the case (Yeager 2013). The correspondences between the Hovmöller plots of Figs. 6 and 8 is quantified in Table 2, which lists the correlation coefficients and rms differences of the AMOC strength anomaly patterns from M, B, and M+B with that of CONTROL over the full Atlantic domain. The metrics in Table 2 succinctly convey many of the points highlighted above: there is a high degree of linearity of the model AMOC response to momentum and buoyancy forcing perturbations; low-pass filtering reduces (enhances) the amount of AMOC variance explained by momentum (buoyancy) forcing, such that buoyancy forcing explains most of the decadal AMOC signal; and analyzing AMOC in sigma coordinates highlights the decadal, buoyancy-driven variability (experiment B is more highly correlated with CONTROL than is M, even without any time filtering).

Fig. 8.
Fig. 8.

As in Fig. 6, but for AMOC computed in density (σ2) space, such that the AMOC strength at each latitude is calculated as the maximum in density rather than depth space prior to the anomaly calculation.

Citation: Journal of Climate 27, 9; 10.1175/JCLI-D-13-00125.1

Changes in the large-scale horizontal gyre circulation of CONTROL can likewise be reconstructed quite accurately as the simple linear superposition of momentum- and buoyancy-forced anomalies. The interannual variances of CONTROL BSF, sea surface height (SSH), and upper-ocean flow strength (as represented by 0–295-m depth-averaged current speed) are shown in Fig. 9, together with the covariances of those CONTROL fields with corresponding anomalies from the M and B simulations. The sum of the covariances is very nearly equal to the total variance in CONTROL (not shown), which supports the linearity of the gyre response to momentum and buoyancy forcing perturbations. Momentum forcing accounts for most of the variance in these horizontal flow metrics at subtropical latitudes (south of about 35°N), but the influence of buoyancy forcing on BSF, SSH, and near-surface western boundary current flow is apparent even at such low latitudes. Farther north, buoyancy forcing clearly becomes the dominant contributor to CONTROL variance in SSH and near-surface flow; in particular, it accounts for almost all of the variability in the North Atlantic Current (NAC) of CONTROL. The variance in the barotropic subpolar gyre circulation is more complex (Fig. 9, left); B is mostly dominant north of about 40°N but there is a dominant M influence seen along the Labrador shelf and across the gyre center at roughly 50°N. The 35°–45°N latitude band appears to be a transition region where both forcing components contribute significantly to the high horizontal flow variability. This latitude range encompasses the GS extension region after its separation from the Atlantic coast. M and B contribute to variance along the southern and northern flanks of the GS, respectively, with B showing a greater influence than M in the central part of the basin (Figs. 9a,d,g). The time-scale dependence of the momentum and buoyancy contributions to horizontal circulation variability can be readily seen in regionally averaged time series from the western edge of the basin where the gyre flow is strongest (Fig. 1c). Figure 10 shows that fluctuations in the BSF and SSH are largely decadal and buoyancy-driven in the SPG and largely interannual and momentum-driven in the STG. Again, the sum of M and B anomalies nicely reproduces the CONTROL anomalies. Thus, the principal forcing components that drive anomalous gyre circulations are found to vary with latitude and time scale in much the same way as for the overturning circulation (Figs. 6, 7).

Fig. 9.
Fig. 9.

(top) Variance of annual-mean time series from CONTROL of (a) BSF, (b) SSH, and (c) depth-averaged upper-ocean (0–295 m) current speed. Also shown are the covariances of the same fields (d)–(f) between M and CONTROL and (g)–(i) between B and CONTROL. To a very good approximation, the sum of covariances plotted in (middle) and (bottom) equals the total variance from CONTROL plotted in (top). The contour levels are as follows: 0, 1, 2, 4, 6, 8, 10, 15, 20, 25, and 30 Sv2 for BSF; 0, 2, 4, 6, 8, 10, 15, 20, 30, 40, and 50 cm2 for SSH; and 0, 0.5, 1, 2, 3, 5, 7, 10, and 15 cm2 s−2 for velocity. Values are shaded above the first nonzero contour.

Citation: Journal of Climate 27, 9; 10.1175/JCLI-D-13-00125.1

Fig. 10.
Fig. 10.

Annual-mean time series of regionally averaged (a),(b) BSF and (c),(d) SSH anomalies from CONTROL, M, B, and the sum of anomalies M+B. The SPG region is the same as the Labrador Sea box (see Fig. 2); the STG region is defined as 80°–65°W, 26°–38°N. No smoothing has been applied. Note that positive BSF anomalies in the SPG in (a) correspond to a weaker (cyclonic) gyre, while positive BSF anomalies in the STG in (b) correspond to a stronger (anticyclonic) gyre.

Citation: Journal of Climate 27, 9; 10.1175/JCLI-D-13-00125.1

To the extent that low-frequency overturning and subpolar gyre circulation variations are both primarily driven by changes in surface buoyancy forcing and associated water mass formation, there may be potential for monitoring and predicting AMOC variations by tracking high-latitude gyre circulation indices. This idea was explored by Böning et al. (2006) using forced model hindcast simulations, as in the present study. In contrast to those authors, we find that interannual wind stress variability does not factor significantly in Labrador Sea SSH variability, and therefore Labrador Sea SSH may well serve as an easily observable proxy for Labrador Sea deep convection and thermohaline circulation change. Böning et al. (2006) compared a heat flux–forced (rather than buoyancy-forced) sensitivity experiment with their control hindcast to ascertain the nonnegligible effects of wind variations on Labrador Sea SSH. In our experiments, almost the entire interannual SSH variance in the Labrador Sea is accounted for by B without a significant wind-forced residual (Figs. 9, 10). Experiment B explains much of the BSF variance in the Labrador Sea (Fig. 10a) and almost all of the Labrador Sea SSH signal (Fig. 10c), and there is a clear correspondence between Labrador Sea SSH variations and buoyancy-driven AMOC variations, viewed either in depth space (Fig. 6d) or in density space (Fig. 8d). Positive SSH anomalies in the Labrador Sea circa 1970 and 1980 (Fig. 10c) were associated with weak barotropic cyclonic circulation, especially in the western SPG (Fig. 10a; note that positive BSF anomalies here correspond to weaker cyclonic circulation), and the opposite was true in the early 1970s, mid-1980s, and early 1990s. Positive (negative) SSH and BSF anomalies in the Labrador Sea correspond to the development of negative (positive) AMOC anomalies in depth space, which subsequently propagated equatorward [Fig. 6d; note the years indicated by black (gray) circles]. Furthermore, these buoyancy-forced anomalies are all clearly associated with the (largely) temperature-driven density anomalies in the central Labrador Sea (Fig. 4), which are linked to winter NAO variations in both observations (Curry and McCartney 2001; Yashayaev 2007) and in CONTROL (see Fig. 15).

This result suggests that it may indeed be possible to monitor slow, buoyancy-driven AMOC variations by observing Labrador Sea SSH changes, with clear potential for advance prediction of slow AMOC change at lower latitudes. In CONTROL, variations in SSH in the central Labrador Sea correlate reasonably well (r > 0.6) with AMOC strength when the former leads the latter, with the lead time increasing with lower latitude, consistent with southward propagation of density anomalies (Fig. 11). South of about 35°N, the correlation structure becomes more complex, with distinct correlation maxima at lead times of about 5 and 9 yr, which may reflect different propagation mechanisms. As expected, the correlation of Labrador Sea SSH with AMOC is much stronger in experiment B, which does not have the “noise” associated with time-varying momentum forcing. In that experiment, Labrador Sea SSH is a strong predictor of AMOC variations at all Northern Hemisphere latitudes (r > 0.8), with the correlation at subtropical latitudes maximized at a lead time of about 4 yr. We do not have an explanation at this time for the different correlation structures evident in Fig. 11, which imply different propagation speeds of density anomalies with and without wind variations. Nevertheless, these results suggest that variations in Labrador Sea SSH can explain and predict roughly 36% (r2) of total AMOC variance and over 70% of the decadal, buoyancy-driven AMOC variance, at subtropical latitudes.

Fig. 11.
Fig. 11.

Lag correlations as a function of latitude of annual-mean Labrador Sea SSH (regionally averaged within the box 55°–50°W, 55°–60°N) with AMOC strength (computed as the maximum in depth) from (a) CONTROL and (b) experiment B. No time filtering has been used. Lead time is positive when Labrador Sea SSH precedes AMOC. The contour interval is 0.05 and values below 0.4 are not plotted.

Citation: Journal of Climate 27, 9; 10.1175/JCLI-D-13-00125.1

5. Identifying the key components of NAO-related forcing

The decadal variability in AMOC north of the equator in CONTROL is primarily due to decadal variation in Labrador Sea water properties that are set by large-scale, high-latitude atmospheric variations: that is, the atmospheric variability represented by the NAO. This finding is in line with numerous other studies already mentioned that have analyzed the ocean response to NAO. In this section, we seek to further elucidate the most salient aspects of NAO-related forcing by examining the relative roles of various flux components and the roles of local versus nonlocal forcing of Labrador Sea deep convection.

First, we note that the buoyancy forcing used in experiment B is dependent on the surface wind speed because this term appears in the turbulent fluxes of evaporation and latent and sensible heat flux [Eqs. (6) and (8)]. To what extent are NAO-related wind speed variations driving the buoyancy flux variations implicated in the recent decadal circulation changes of the North Atlantic? To answer this question, we have run an experiment B* (not listed in Table 1), which is equivalent to B except that NYF wind speed is used in the computation of all turbulent fluxes [, etc. (see Table 1)]. The resulting AMOC signal (in depth space; Fig. 12a) shows only minor differences from the experiment with full variability in buoyancy fluxes (cf. Fig. 6d). We conclude that interannual wind speed variations are relatively unimportant as drivers of the decadal AMOC signals of interest. This is in line with the findings of Seager et al. (2000), who conclude that, north of 40°N, the impact of wind speed variations on the turbulent heat fluxes is considerably less than the impact of wind direction changes (the latter being most associated with changes in the advection of temperature and moisture).

Fig. 12.
Fig. 12.

As in Fig. 6, but for (a) experiment B*, (b) experiment B.Q, (c) B.F+B.Q (the sum of anomalies from these experiments), and (d) experiment B.F. Experiment B* is identical to B, except that normal year winds are used for the computation of all turbulent fluxes (see text).

Citation: Journal of Climate 27, 9; 10.1175/JCLI-D-13-00125.1

Both heat and freshwater fluxes would appear to be important contributors to the buoyancy-forced variability in CONTROL. As expected given the nonlinearity of the equation of state, splitting the buoyancy forcing into separate thermal and haline components in experiments B.Q and B.F, respectively, results in AMOC anomaly signals whose sum is weaker than in the total response from experiment B (cf. Figs. 12c, 6d; also note that the correlation of B.F+B.Q with B is only 0.9 in Table 2). Nevertheless, the correspondence is sufficient to draw some conclusions from this decomposition. Changes in heat and freshwater forcing in the high-latitude North Atlantic are about equally important in driving slow changes in ocean dynamics (Table 2) and there tends to be constructive interference of thermal- and haline-forced signals such that the strength of AMOC anomalies in B is quite a bit larger than those in either B.Q or B.F individually. This suggests a particularly important role for evaporation, which drives same-signed contributions to density in both the temperature and salinity equations. The episodic generation of high-latitude AMOC anomalies in certain years (i.e., those identified in Fig. 6d) seems to be associated with the time-varying heat fluxes of B.Q (Fig. 12b), while surface freshwater forcing contributes to a multidecadal modulation of AMOC that emanates from high northern latitudes (Fig. 12d).

Experiments B.1 and B.2 are designed to hone in further on the most important components of high-latitude buoyancy forcing. All of the essential characteristics of low-frequency AMOC variability from experiment B are captured in B.1, which uses interannually varying forcing only for the turbulent buoyancy fluxes (Fig. 13a). The B.1 anomalies are somewhat larger than in B, implying that variability in the other heat and freshwater fluxes (P, R, QS, and QL) tend to damp the Labrador Sea density variations induced by evaporation and the latent and sensible heat fluxes. The close match between B.1 and B (cf. Figs. 13a, 6d; see Table 2) also demonstrates that the use of climatological precipitation and downward radiation fluxes prior to 1979 and 1984 does not significantly affect the variability simulated in B; these are relatively inconsequential fluxes compared to E, QE, and QH for driving large-scale circulation variations. Experiment B.2 is identical to B.1 except that flux variability is confined to the Labrador Sea box region, with NYF applied elsewhere. Most of the aforementioned variability features are still evident in the AMOC anomaly time series (Fig. 13b), but with somewhat reduced amplitude. The correlation of B.2 with B is only slightly lower than that of B.1 (Table 2). Comparing B.1 and B.2 (or B.1 and B) gives a sense of the impact of local, turbulent flux forcing of deep convection with and without the “preconditioning” of Labrador Sea waters by surface buoyancy forcing variations over the larger Atlantic basin. The preconditioning is presumably mainly associated with large-scale air–sea flux anomalies over the larger SPG region and tends to modulate the amplitude of Labrador Sea density anomalies, but this is a second-order effect compared to local buoyancy forcing variations over the deep convection region. We conclude that most of the decadal variability in AMOC over the last half of the twentieth century can be traced to variations in the turbulent heat and freshwater forcing over the Labrador Sea alone.

Fig. 13.
Fig. 13.

As in Fig. 6, but for (a) experiment B.1, (b) experiment B.2, (c) experiment NYF, and (d) experiment M.SO.

Citation: Journal of Climate 27, 9; 10.1175/JCLI-D-13-00125.1

6. The role of Southern Ocean winds

We now turn to the question of the relative role of Southern Ocean (SO) wind variability on low-frequency AMOC changes in the recent past. The trend in the SAM over the last few decades of the twentieth century (Thompson and Solomon 2002) is present in the NCEP–NCAR reanalysis surface winds used in CORE-II, and it can clearly be seen in a Hovmöller plot of the zonally averaged zonal wind stress that drives the CONTROL simulation (Fig. 14). The positive wind stress trend is most pronounced poleward of 40°S in the region of the Antarctic Circumpolar Current (ACC). The hypothesis that SO wind variations may have a controlling influence on rates of overturning in the Atlantic has been explored in numerous recent studies, with mixed results (e.g., Delworth and Zeng 2008; Klinger and Cruz 2009; Sijp and England 2009; Farneti and Delworth 2010; Wolfe and Cessi 2010; Lee et al. 2011). The impact of SO wind variations appears to depend critically on the fidelity of the model representation of mesoscale eddies (Farneti and Delworth 2010; Farneti and Gent 2011; Gent and Danabasoglu 2011), which contributes to the diversity of sensitivities found in the literature. In particular, the use of a constant coefficient in the ocean eddy parameterization induces a rather strong response in the Northern Hemisphere AMOC to changes in SO winds. Another potential source of confusion, however, is that studies focused on SO effects often employ idealized models and rarely place the results in context by comparing to AMOC variability resulting from realistic high-latitude Northern Hemisphere buoyancy flux variations.

Fig. 14.
Fig. 14.

Anomalous annual-mean zonally averaged zonal wind stress τx (N m−2) in the Southern Hemisphere from CONTROL. The contour interval is 0.01 N m−2 with positive (negative) anomalies contoured in black (gray) with gray shading for positive values.

Citation: Journal of Climate 27, 9; 10.1175/JCLI-D-13-00125.1

As noted in the discussion of Figs. 6 and 7, momentum forcing accounts for most of the AMOC variance south of about 30°N, and most of the decadal variance south of the equator. Experiment M.SO looks specifically at the nonlocal impacts of the trend in Southern Ocean wind stress, with interannual variations in atmospheric surface winds applied only south of 35°S. The effect of this forcing is clearly discernible on AMOC north of 30°S, with northward propagating anomalies reflecting the sign of the SO zonal wind stress anomalies (Fig. 13d). This signal explains a large fraction of the low-frequency variability in M south of the equator, which in turn dominates the decadal variability in CONTROL at those latitudes. However, the correlation of M.SO with M is only about 0.6 on decadal time scales when considering the whole Atlantic domain (Table 2). The M.SO signal is very weak north of the equator and is far weaker than buoyancy-driven AMOC signals north of about 20°N. The cross-equatorial attenuation of and delayed response to momentum-forced signals emanating from the Southern Ocean (experiment M.SO; Fig. 13d) or of buoyancy-forced signals emanating from the Labrador Sea region (experiment B.2; Fig. 13b) is likely attributable to the “equatorial buffer” effect proposed by Johnson and Marshall (2002b). The equatorial buffer implies a much longer adjustment time scale for the branch of the overturning circulation in the opposite hemisphere from where a perturbation originates. Simple theoretical arguments suggest that the equator acts as a low-pass filter of overturning circulation anomalies driven by high-latitude forcing anomalies in either the Labrador Sea or the SO (Johnson and Marshall 2002a). It is interesting to note that the positive AMOC trend induced by SO wind forcing is more or less coherent with the positive trend induced by Labrador Sea buoyancy forcing over all latitudes south of about 20°N; this explains the relatively high correlation of M.SO with CONTROL (Table 2), although the rms difference with CONTROL is quite high. We return to this point in the discussion. In line with Johnson and Marshall (2002a), our experiments suggest that the recent decadal variations in SO wind forcing were much less important than NAO-related buoyancy forcing in driving recent changes in the North Atlantic AMOC, but that south of the equator (and certainly at 30°S), SO wind variations were at least as important as SPG buoyancy forcing in driving decadal AMOC variability.

7. The origins of Labrador Sea flux variability

We have shown with experiment B.2 that most of the decadal AMOC variability in the North Atlantic between 1958 and 2007 can be traced to turbulent fluxes of heat and freshwater in the Labrador Sea. An examination of the fluxes in this region offers further clues about the origin of the decadal time scale of AMOC in CONTROL. We are interested in the relative impacts on surface buoyancy of the various flux components, and so we have converted monthly Qas and Fas terms to surface buoyancy fluxes following Large and Nurser (2001). Year-to-year variations in wintertime [January through March (JFM) mean] air–sea buoyancy flux in the Labrador Sea box region are clearly dominated by changes in sensible heat loss, with changes in the latent heat loss contributing significantly as well (Fig. 15b). Changes in evaporation, which impact sea surface density (SSD) by altering SSS, are the third most important contributor to the interannual changes in the net surface buoyancy flux (Fig. 15a), but it is important to bear in mind that precipitation variability is lacking prior to 1979 and there is no representation of the potentially significant Greenland glacier melt in the CORE-II forcings. Nevertheless, the buoyancy forcing variations due to freshwater forcing are much smaller than those due to heat forcing in this region (note the scale change between Figs. 15a,b). In the vicinity of the sea ice edge, however, ice–ocean fluxes (in particular, buoyancy fluxes related to ice melt) dominate the buoyancy flux (e.g., Yeager and Jochum 2009), but our focus here is on the factors that influence deep convection in the open ocean of the Labrador Sea. We find that, to first order, variations in deep convection can be understood as resulting from changes in the local air–sea fluxes of both heat and freshwater, which together determine the net surface buoyancy flux (Bas; Fig. 15c). The sign convention for fluxes is positive into the ocean, so episodes of intense Labrador Sea convection were contemporaneous with anomalously strong buoyancy loss (positive −Bas) from the surface.

Fig. 15.
Fig. 15.

Anomaly time series of winter fields from CONTROL regionally averaged over the Labrador Sea box region. (a) JFM-mean net air–sea freshwater flux and components converted into units of surface buoyancy loss (10−8 m2 s−3). (b) As in (a), but for JFM-mean net air–sea heat flux and components. (c) March-mean MLD, March-mean SSD, JFM-mean net surface buoyancy loss (−Bas; the sum of heat and freshwater buoyancy fluxes, multiplied by −1, so that positive values denote surface density gain), and December–March-mean NAO index. All fields are averaged over the ice-free ocean as determined by the March-mean ice fraction. Panel (c) is plotted in units of standard deviation of the respective time series. The anomalies are relative to the 1958–2007 time average.

Citation: Journal of Climate 27, 9; 10.1175/JCLI-D-13-00125.1

Because anomalous evaporation is always accompanied by anomalous latent cooling, the E and QE fluxes are perfectly correlated in terms of their contributions to the net surface buoyancy flux; furthermore, anomalies of QH over the Labrador Sea are highly correlated with the evaporative buoyancy fluxes (Fig. 15b). The high correlation between QE and QH follows from the Clausius–Clapeyron relation: anomalously cold air is anomalously dry and vice versa. The three turbulent buoyancy fluxes thus work in tandem to generate large surface density tendencies, which explain much of the simulated variability in SSD and MLD in the Labrador Sea (Fig. 15c). Of course, to fully account for variations in Labrador Sea SSD and MLD, one must take into account lateral physics, including processes that set the deep density structure, but we find that the essential features of MLD variability in our CONTROL hindcast are largely dictated by Bas variations. The large, negative AMOC anomalies that originated in the early and late 1970s (Fig. 6d) can be traced to negative MLD and SSD anomalies in the Labrador Sea following winters of particularly weak surface buoyancy loss due to weaker than normal surface heat loss. The three positive AMOC signals that originated in the mid-1970s, mid-1980s, and early 1990s (Fig. 6d) can be traced to positive MLD and SSD anomalies that apparently resulted from strong fluxes of buoyancy out of the ocean during cold, dry air outbreaks in winter months of those years. While we cannot rule out the possibility that model error may contribute to the dominance of air–sea forcing of convection in this region, especially given the role that unresolved eddies might be expected to play in mixing buoyant shelf waters into the Labrador Sea interior, the correspondence of the flux time series of Fig. 15 with both the simulated and observed temperature and density profiles in the central Labrador Sea should be noted (Fig. 4). Changes in central Labrador Sea density in the model are clearly related to the history of net wintertime surface buoyancy forcing there, with the heat flux forcing playing a particularly important role in the water mass transformation. As already noted, the simulated temperature/density anomalies compare reasonably well with hydrographic observations, except in the early 1970s and mid-1980s, and we cannot explain the rather large deviations from observed salinity in the Labrador Sea.

The turbulent buoyancy fluxes are functions of both the atmospheric and oceanic states [Eqs. (6) and (8)], but the interannual variability of E, QE, and QH over the Labrador Sea and thus the variability in Bas are almost entirely driven by changes in atmospheric surface temperature θ and humidity q. The correlations of seasonally and regionally averaged θ and q with Bas over the Labrador Sea (Fig. 16) are 0.92 and 0.93, respectively. Such high correlations are not that surprising given that this is an uncoupled ocean–sea ice run that precludes oceanic feedbacks onto the atmospheric state, but the result nevertheless sheds light on the mechanisms at work in such forced hindcast simulations. Low-pass filtering of these time series reveals that a pronounced downward trend in buoyancy flux (i.e., upward trend in density flux) into the surface ocean between the early 1960s and mid-1990s was driven by corresponding trends in atmospheric temperature and humidity over this region (Fig. 16b). The slowly changing atmospheric state–induced trends in net heat and freshwater fluxes into the Labrador Sea region, which tended to increase SSD by decreasing SST and increasing SSS over multiple decades in the late twentieth century. We conclude that the ultimate source of the enhanced AMOC in the late 1980s and 1990s in CONTROL is the multidecadal shift toward colder and drier atmospheric conditions over the Labrador Sea in winter.

Fig. 16.
Fig. 16.

Anomaly time series of JFM-mean fields from CONTROL regionally averaged over the Labrador Sea box region. (a) The 10-m atmospheric potential temperature θ and specific humidity q together with the net surface buoyancy flux (this differs from the curve in Fig. 15c by a factor of −1). (b) As in (a), but after smoothing with a 15-point Lanczos filter with cutoff period of 7 yr. All curves are normalized by the standard deviations of the respective unfiltered time series. The regional averages of θ and q are computed over the entire Labrador Sea box region, while Bas is averaged over the ice-free subdomain.

Citation: Journal of Climate 27, 9; 10.1175/JCLI-D-13-00125.1

8. Discussion and conclusions

We have explored the forcing contributions to decadal variations in the large-scale overturning and gyre circulations in a CORE-II coupled ocean–sea ice hindcast simulation run with the latest version of the CESM1. As shown in section 3 and in Yeager et al. (2012), there are many quite realistic features of the mean and variability of this CONTROL solution that support its use as a tool to study mechanisms of ocean variability in the recent past. There are also many known (and no doubt unknown) inadequacies of the model that will necessarily qualify any conclusions drawn from it.

First, mesoscale eddies are parameterized in the model, and while the parameterization used is state of the art (Danabasoglu et al. 2012a), the Labrador Sea and Southern Ocean are two regions that are particularly sensitive to the representation of eddies (Chanut et al. 2008; Farneti and Gent 2011; Danabasoglu et al. 2012b). A CORE-II simulation using the eddy-resolving version of CESM1 is planned, but not available at this time for comparison with CONTROL. The studies by Farneti et al. (2010), Farneti and Delworth (2010), and Gent and Danabasoglu (2011) suggest that, if anything, our model underestimates the eddy-induced overturning response to SO wind increase, and thus overestimates the SO wind impact on AMOC. Another caveat related to model resolution is the perennial issue of a poor NAC representation, which could prolong the time scale of AMOC variability by eliminating the relatively quick advective feedback of warm/salty/buoyant NAC water into the central Labrador Sea following intense convection and gyre spinup. Work is underway to assess the impacts of this bias on model variability. We speculate that this may explain some of the discrepancy with the observed temperature and density anomalies in the central Labrador Sea, particularly in the mid-1980s (Fig. 4), but the general agreement between model and observations in this region gives us confidence that model shortcomings are not catastrophic.

The CORE-II forcings used here to drive the ocean and sea ice models, while considered to be among the best available surface boundary conditions for historical ocean/sea ice reconstructions, almost certainly contain biases that impact our results. As already mentioned, the forcing suite lacks the freshwater input associated with land ice melt, and this could explain some of the noted differences between observed and simulated salinity in the central Labrador Sea. The dominance of turbulent heat flux terms in our experiments could be related to the fact that other flux terms are only available at coarser temporal resolution (daily for downward radiative fluxes and monthly for precipitation and runoff) or that the near-surface atmospheric state fields in CORE-II are biased. The extent to which these uncertainties qualify our conclusions is not known, and this should be explored in future work. However, one recent study suggests that the uncertainties in reanalysis products does not significantly impact the overturning circulation variability inferred from historical atmospheric state fields (Grist et al. 2014).

Another important caveat pertains to the underestimation of mean and variability of simulated sea ice coverage in the Labrador Sea (Figs. 2, 3). This results in reduced insulation of ocean surface waters from the extremely cold, dry Arctic air, and therefore greatly increases the buoyancy flux out of the ocean. In previous CESM hindcasts that had much less sea ice in the Labrador Sea, this resulted in excessive surface water mass transformation and an overly strong AMOC mean and variance (Yeager and Jochum 2009). The fact that CONTROL exhibits a reasonable match to observed hydrographic variations in the central Labrador Sea (Fig. 4), particularly in terms of the timing and magnitude of density anomalies, suggests that the sea ice bias in the Labrador Sea may be tolerable. However, there are indications that the magnitude and extent of winter convection in the Labrador Sea in CONTROL is excessive.

The finding that historical AMOC variability in the North Atlantic can be quite cleanly split into momentum- and buoyancy-forced components that are characterized predominately by interannual and decadal time scales, respectively, with the latter associated with NAO-driven deep convection in the Labrador Sea is in line with previous work done with a variety of models (Eden and Willebrand 2001; Böning et al. 2006; Biastoch et al. 2008; Robson et al. 2012a). In this study, we have further investigated the spatial dependence of AMOC and gyre circulation variability on surface forcing constituents and find that buoyancy forcing is the dominant driver of decadal AMOC variability north of the equator and of horizontal gyre variability north of about 40°N. Given that most of the variance in the high-latitude gyre (SPG) and AMOC circulations derive from this common forcing, we have explored the potential for monitoring AMOC using Labrador Sea SSH variations as a proxy for the strength of the thermohaline circulation. Going beyond the buoyancy/momentum decomposition, the forcing perturbation technique has been used here to systematically assess the relative impacts on AMOC of heat and freshwater forcing, wind speed variations, the trend in SO zonal winds, turbulent buoyancy fluxes, and Labrador Sea atmospheric conditions. The fidelity of our results is supported by the good correspondence of our CONTROL with available observations and by the fact that, in contrast to the previous studies cited, our model includes a parameterization for Nordic Seas overflows that is known to impact AMOC variability, primarily by enhancing the mean deep stratification of the Labrador Sea (Danabasoglu et al. 2012b; Yeager and Danabasoglu 2012). Furthermore, our use of NYF allows us to filter the power spectrum of forcing fields more effectively than doing simple time averaging to construct climatological forcing; with NYF, variance at annual and higher frequencies is retained.

With the aforementioned caveats in mind, this analysis has led us to the following conclusions:

  • High-northern-latitude buoyancy forcing accounts for almost all of the decadal variability in AMOC and SPG strength over the period 1958–2007, including the positive trend in North Atlantic overturning and gyre strength in the 1980s and 1990s that contributed to the large SST increase north of 45°N in the mid-1990s.
  • Both heat and freshwater forcing play important roles in driving recent decadal AMOC changes and in particular the turbulent buoyancy fluxes (evaporation and sensible and latent heat flux) account for almost all of the buoyancy-driven variability.
  • Variations of atmospheric surface temperature and humidity over the Labrador Sea region drive the variations in turbulent winter buoyancy loss; in turn, these local surface buoyancy fluxes largely determine the variations in SSD, deep convection, and water mass characteristics that ultimately drive the decadal component of AMOC variability. The preconditioning of Labrador Sea water by surface forcing outside of the Labrador Sea and the influence of wind speed on the turbulent buoyancy fluxes both appear to be second-order effects.
  • While NAO-related buoyancy forcing is the dominant driver of decadal AMOC variability north of the equator, momentum forcing is implicated in the slow variability farther south in the Atlantic. Much of the increasing trend in AMOC in the Southern Hemisphere is related to observed trends in SO westerly winds.
  • Labrador Sea SSH changes are largely buoyancy driven and thus may be an excellent proxy for monitoring slow AMOC variations.

The fact that the ultimate source of most of the low-frequency AMOC variability in this and other CORE-II simulations appears to be low-frequency atmospheric variability over the Labrador Sea associated with the NAO raises many questions that will be the focus of future work. The downward trend in winter surface air temperature and humidity in this region between 1960 and 1990 (Fig. 16b) clearly has huge ramifications for ocean dynamics. What is the origin of this low-frequency power? Does it derive from coupled exchanges with the surface ocean (which are absent here but occurred in nature) or from external forcing, as is hypothesized for the trend in SAM? Observed trends in the Northern Hemisphere storm track are clearly implicated (Chang 2007; Shaman et al. 2010) in the slow variations in Labrador Sea θ and q. Are externally forced changes in the midlatitude storm tracks implicated in the coherent upward trends in AMOC in both hemispheres driven by Labrador Sea buoyancy forcing and SO momentum forcing, respectively (Fig. 13)? If these are externally forced, it raises the question of why phase 5 of the Coupled Model Intercomparison Project (CMIP5) simulations of the twentieth century fail to exhibit synchronized decadal AMOC variations, together with in-phase NAO variations, in the latter decades of that century (Zhang and Wang 2013). Alternatively, the slow variations in atmospheric state over the main deep-water formation region could have arisen from purely internal coupled air–sea interactions of the sort that govern decadal AMOC variability in some CGCMs (e.g., Medhaug et al. 2012), in which case we would not expect synchronicity in CMIP5 simulations. The deconstruction of the Atlantic circulation response to historical surface forcing perturbations presented here provides an observation-based benchmark for evaluating AMOC variability mechanisms in coupled model simulations. In the coupled framework, the prominence of Labrador Sea buoyancy forcing as a driver of AMOC variations will vary depending upon the nature of model biases. For instance, biases in surface salinity or sea ice extent can strongly inhibit the surface-forced deep convection, which features so prominently in our CONTROL experiment, and model error may preclude the generation of large amplitude, low-frequency variations in atmospheric state over the Labrador Sea of the sort that are seen in the observational record. Future work will also focus on several aspects of this work that bear on the decadal prediction of AMOC. First, it may be possible to identify strongly buoyancy-forced oceanic observables (e.g., Labrador Sea SSH) that will enable prediction of low-latitude decadal AMOC variability. Second, our analysis suggests that progress in decadal climate prediction in the Atlantic may require a dedicated focus on improving our understanding of and model representation of the processes that govern surface air temperature and humidity over the Labrador Sea region.

Acknowledgments

We are very grateful to Igor Yashayaev for sharing his Labrador Sea hydrographic time series. Peter Gent provided helpful criticism on an early version of this manuscript. This work was supported by the NOAA/Climate Program Office under Climate Variability and Predictability Program Grant NA09OAR4310163. This research used computing resources provided by NCAR’s Computational and Information Systems Laboratory. We thank Anand Gnanadesikan and two anonymous reviewers for their insightful comments.

REFERENCES

  • Beismann, J.-O., , and B. Barnier, 2004: Variability of the meridional overturning circulation of the North Atlantic: Sensitivity to overflows of dense water masses. Ocean Dyn., 54, 92106, doi:10.1007/s10236-003-0088-x.

    • Search Google Scholar
    • Export Citation
  • Bentsen, M., , H. Drange, , T. Furevik, , and T. Zhou, 2004: Simulated variability of the Atlantic meridional overturning circulation. Climate Dyn., 22, 701720, doi:10.1007/s00382-004-0397-x.

    • Search Google Scholar
    • Export Citation
  • Biastoch, A., , C. Böning, , J. Getzlaff, , J.-M. Molines, , and G. Madec, 2008: Causes of interannual-decadal variability in the meridional overturning circulation of the midlatitude North Atlantic Ocean. J. Climate, 21, 65996615, doi:10.1175/2008JCLI2404.1.

    • Search Google Scholar
    • Export Citation
  • Böning, C., , M. Scheinert, , J. Dengg, , A. Biastoch, , and A. Funk, 2006: Decadal variability of subpolar gyre transport and its reverberation in the North Atlantic overturning. Geophys. Res. Lett., 33, L21S01, doi:10.1029/2006GL026906.

    • Search Google Scholar
    • Export Citation
  • Brauch, J. P., , and R. Gerdes, 2005: Response of the northern North Atlantic and Arctic Oceans to a sudden change of the North Atlantic Oscillation. J. Geophys. Res., 110, C11018, doi:10.1029/2004JC002436.

    • Search Google Scholar
    • Export Citation
  • Chang, E. K. M., 2007: Assessing the increasing trend in Northern Hemisphere winter storm track activity using surface ship observations and a statistical storm track model. J. Climate, 20, 56075628, doi:10.1175/2007JCLI1596.1.

    • Search Google Scholar
    • Export Citation
  • Chanut, J., , B. Barnier, , W. Large, , L. Debreu, , T. Penduff, , J. M. Molines, , and P. Mathiot, 2008: Mesoscale eddies in the Labrador Sea and their contribution to convection and restratification. J. Phys. Oceanogr., 38, 16171643, doi:10.1175/2008JPO3485.1.

    • Search Google Scholar
    • Export Citation
  • Comiso, J., cited 2012: Bootstrap sea ice concentrations from Nimbus-7 SMMR and DMSP SSM/I-SSMIS, version 2. National Snow and Ice Data Center, Boulder, CO, digital media. [Available online at http://nsidc.org/data/nsidc-0079.]

  • Cunningham, S. A., and Coauthors, 2007: Temporal variability of the Atlantic meridional overturning circulation at 26.5°N. Science, 317, 935938, doi:10.1126/science.1141304.

    • Search Google Scholar
    • Export Citation
  • Curry, R. G., , and M. S. McCartney, 2001: Ocean gyre circulation changes associated with the North Atlantic Oscillation. J. Phys. Oceanogr., 31, 33743400, doi:10.1175/1520-0485(2001)031<3374:OGCCAW>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Dai, A., , T. Qian, , K. Trenberth, , and J. Milliman, 2009: Changes in continental freshwater discharge from 1948 to 2004. J. Climate, 22, 27732791, doi:10.1175/2008JCLI2592.1.

    • Search Google Scholar
    • Export Citation
  • Danabasoglu, G., , W. G. Large, , and B. P. Briegleb, 2010: Climate impacts of parameterized Nordic Sea overflows. J. Geophys. Res., 115, C11005, doi:10.1029/2010JC006243.

    • Search Google Scholar
    • Export Citation
  • Danabasoglu, G., , S. C. Bates, , B. P. Briegleb, , S. R. Jayne, , M. Jochum, , W. G. Large, , S. Peacock, , and S. G. Yeager, 2012a: The CCSM4 ocean component. J. Climate, 25, 13611389, doi:10.1175/JCLI-D-11-00091.1.

    • Search Google Scholar
    • Export Citation
  • Danabasoglu, G., , S. G. Yeager, , Y.-O. Kwon, , J. Tribbia, , A. Phillips, , and J. Hurrell, 2012b: Variability of the Atlantic meridional overturning circulation in CCSM4. J. Climate, 25, 51535172, doi:10.1175/JCLI-D-11-00463.1.

    • Search Google Scholar
    • Export Citation
  • Danabasoglu, G., and Coauthors, 2014: North Atlantic simulations in Coordinated Ocean-Ice Reference Experiments phase II (CORE-II). Part I: Mean states. Ocean Modell., 73, 76107, doi:10.1016/j.ocemod.2013.10.005.

    • Search Google Scholar
    • Export Citation
  • de Boyer Montégut, C., , G. Madec, , A. S. Fischer, , A. Lazar, , and D. Iudicone, 2004: Mixed layer depth over the global ocean: An examination of profile data and a profile-based climatology. J. Geophys. Res., 109, C12003, doi:10.1029/2004JC002378.

    • Search Google Scholar
    • Export Citation
  • Delworth, T. L., , and M. E. Mann, 2000: Observed and simulated multidecadal variability in the Northern Hemisphere. Climate Dyn., 16, 661676, doi:10.1007/s003820000075.

    • Search Google Scholar
    • Export Citation
  • Delworth, T. L., , and F. Zeng, 2008: Simulated impact of altered Southern Hemisphere winds on the Atlantic overturning circulation. Geophys. Res. Lett., 35, L20708, doi:10.1029/2008GL035166.

    • Search Google Scholar
    • Export Citation
  • Dickson, B., , I. Yashayaev, , J. Meincke, , B. Turrell, , S. Dye, , and J. Holfort, 2002: Rapid freshening of the deep North Atlantic Ocean over the past four decades. Nature, 416, 832836, doi:10.1038/416832a.

    • Search Google Scholar
    • Export Citation
  • Doney, S. C., , S. G. Yeager, , W. G. Large, , and J. C. McWilliams, 2003: Modeling global oceanic interannual variability (1958-1997): Simulation design and model-data evaluation. NCAR Tech. Note NCAR/TN-452+STR, 48 pp.

  • Duchon, C. E., 1979: Lanczos filtering in one and two dimensions. J. Appl. Meteor., 18, 10161022, doi:10.1175/1520-0450(1979)018<1016:LFIOAT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Eden, C., , and J. Willebrand, 2001: Mechanism of interannual to decadal variability of the North Atlantic circulation. J. Climate, 14, 22662280, doi:10.1175/1520-0442(2001)014<2266:MOITDV>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Farneti, R., , and T. L. Delworth, 2010: The role of mesoscale eddies in the remote oceanic response to altered Southern Hemisphere winds. J. Phys. Oceanogr., 40, 23482354, doi:10.1175/2010JPO4480.1.

    • Search Google Scholar
    • Export Citation
  • Farneti, R., , and P. R. Gent, 2011: The effects of the eddy-induced advection coefficient in a coarse-resolution coupled climate model. Ocean Modell., 39, 135145, doi:10.1016/j.ocemod.2011.02.005.

    • Search Google Scholar
    • Export Citation
  • Farneti, R., , T. L. Delworth, , A. J. Rosati, , S. M. Griffies, , and F. Zeng, 2010: The role of mesoscale eddies in the rectification of the Southern Ocean response to climate change. J. Phys. Oceanogr., 40, 15391557, doi:10.1175/2010JPO4353.1.

    • Search Google Scholar
    • Export Citation
  • Fox-Kemper, B., and Coauthors, 2011: Parameterization of mixed layer eddies. III: Implementation and impact in global ocean climate simulations. Ocean Modell., 39, 6178, doi:10.1016/j.ocemod.2010.09.002.

    • Search Google Scholar
    • Export Citation
  • Gent, P. R., , and G. Danabasoglu, 2011: Response to increasing Southern Hemisphere winds in CCSM4. J. Climate, 24, 49924998, doi:10.1175/JCLI-D-10-05011.1.

    • Search Google Scholar
    • Export Citation
  • Gent, P. R., and Coauthors, 2011: The Community Climate System Model version 4. J. Climate, 24, 49734991, doi:10.1175/2011JCLI4083.1.

    • Search Google Scholar
    • Export Citation
  • Grist, J. P., and Coauthors, 2010: The roles of surface heat flux and ocean heat transport convergence in determining Atlantic Ocean temperature variability. Ocean Dyn., 60, 771790, doi:10.1007/s10236-010-0292-4.

    • Search Google Scholar
    • Export Citation
  • Grist, J. P., , S. A. Josey, , R. Marsh, , Y.-O. Kwon, , R. J. Bingham, , and A. T. Blaker, 2014: The surface-forced overturning of the North Atlantic: Estimates from modern era atmospheric reanalysis datasets. J. Climate, in press.

    • Search Google Scholar
    • Export Citation
  • Häkkinen, S., 1999: Variability of the simulated meridional heat transport in the North Atlantic for the period 1951-1993. J. Geophys. Res., 104, 10 99111 007, doi:10.1029/1999JC900034.

    • Search Google Scholar
    • Export Citation
  • Häkkinen, S., , P. B. Rhines, , and D. L. Worthen, 2011: Atmospheric blocking and Atlantic multidecadal ocean variability. Science, 334, 655659, doi:10.1126/science.1205683.

    • Search Google Scholar
    • Export Citation
  • Holland, M. M., , D. A. Bailey, , B. P. Briegleb, , B. Light, , and E. Hunke, 2012: Improved sea ice shortwave radiation physics in CCSM4: The impact of melt ponds and aerosols on Arctic sea ice. J. Climate, 25, 14131430, doi:10.1175/JCLI-D-11-00078.1.

    • Search Google Scholar
    • Export Citation
  • Hunke, E. C., , and W. H. Lipscomb, 2008: CICE: The Los Alamos sea ice model, documentation and software, version 4.0. Los Alamos National Laboratory Tech. Rep. LA-CC-06-012, 76 pp.

  • Hurrell, J. W., , J. J. Hack, , D. Shea, , J. M. Caron, , and J. Rosinski, 2008: A new sea surface temperature and sea ice boundary dataset for the Community Atmosphere Model. J. Climate, 21, 51455153, doi:10.1175/2008JCLI2292.1.

    • Search Google Scholar
    • Export Citation
  • Jochum, M., , G. Danabasoglu, , M. Holland, , Y.-O. Kwon, , and W. G. Large, 2008: Ocean viscosity and climate. J. Geophys. Res., 113, C06017, doi:10.1029/2007JC004515.

    • Search Google Scholar
    • Export Citation
  • Johnson, H. L., , and D. Marshall, 2002a: Localization of abrupt change in the North Atlantic thermohaline circulation. Geophys. Res. Lett., 29, doi:10.1029/2001GL014140.

    • Search Google Scholar
    • Export Citation
  • Johnson, H. L., , and D. Marshall, 2002b: A theory for the surface Atlantic response to thermohaline variability. J. Phys. Oceanogr., 32, 11211132, doi:10.1175/1520-0485(2002)032<1121:ATFTSA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Khatiwala, S., , P. Schlosser, , and M. Visbeck, 2002: Rates and mechanisms of water mass transformation in the Labrador Sea as inferred from tracer observations. J. Phys. Oceanogr., 32, 666686, doi:10.1175/1520-0485(2002)032<0666:RAMOWM>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Kistler, R., and Coauthors, 2001: The NCEP–NCAR 50-Year Reanalysis: Monthly means CD-ROM and documentation. Bull. Amer. Meteor. Soc., 82, 247267, doi:10.1175/1520-0477(2001)082<0247:TNNYRM>2.3.CO;2.

    • Search Google Scholar
    • Export Citation
  • Klinger, B. A., , and C. Cruz, 2009: Decadal response of global circulation to Southern Ocean zonal wind stress perturbation. J. Phys. Oceanogr., 39, 18881904, doi:10.1175/2009JPO4070.1.

    • Search Google Scholar
    • Export Citation
  • Knight, J. R., , R. J. Allan, , C. K. Folland, , M. Vellinga, , and M. E. Mann, 2005: A signature of persistent natural thermohaline circulation cycles in observed climate. Geophys. Res. Lett., 32, L20708, doi:10.1029/2005GL024233.

    • Search Google Scholar
    • Export Citation
  • Knight, J. R., , C. K. Folland, , and A. A. Scaife, 2006: Climate impacts of the Atlantic multidecadal oscillation. Geophys. Res. Lett., 33, L17706, doi:10.1029/2006GL026242.

    • Search Google Scholar
    • Export Citation
  • Kuhlbrodt, T., , A. Griesel, , M. Montoya, , A. Levermann, , M. Hofmann, , and S. Rahmstorf, 2007: On the driving processes of the Atlantic meridional overturning circulation. Rev. Geophys., 45, RG2001, doi:10.1029/2004RG000166.

    • Search Google Scholar
    • Export Citation
  • Large, W. G., , and G. Nurser, 2001: Ocean surface water mass transformation. Ocean Circulation and Climate—Observing and Modelling the Global Ocean, International Geophysics Series, Vol. 77, Academic Press, 317–336.

  • Large, W. G., , and S. G. Yeager, 2004: Diurnal to decadal global forcing for ocean and sea ice models: The data sets and climatologies. NCAR Tech. Note NCAR/TN-460+STR, 105 pp.

  • Large, W. G., , and S. G. Yeager, 2009: The global climatology of an interannually varying air–sea flux data set. Climate Dyn., 33, 341364, doi:10.1007/s00382-008-0441-3.

    • Search Google Scholar
    • Export Citation
  • Lavender, K. L., , R. E. Davis, , and W. B. Owens, 2002: Observations of open-ocean deep convection in the Labrador Sea from subsurface floats. J. Phys. Oceanogr., 32, 511526, doi:10.1175/1520-0485(2002)032<0511:OOOODC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Lee, S.-K., , W. Park, , E. van Sebille, , M. O. Baringer, , C. Wang, , D. B. Enfield, , S. G. Yeager, , and B. P. Kirtman, 2011: What caused the significant increase in Atlantic Ocean heat content since the mid-20th century? Geophys. Res. Lett., 38, L17607, doi:10.1029/2011GL048856.

    • Search Google Scholar
    • Export Citation
  • Levitus, S., and Coauthors, 1998: Introduction. Vol. 1, World Ocean Database, NOAA Atlas NESDIS 18, 346 pp.

  • Liu, Z., 2012: Dynamics of interdecadal climate variability: A historical perspective. J. Climate, 25, 19631995, doi:10.1175/2011JCLI3980.1.

    • Search Google Scholar
    • Export Citation
  • Locarnini, R. A., , A. V. Mishonov, , J. I. Antonov, , T. P. Boyer, , H. E. Garcia, , O. K. Baranova, , M. M. Zweng, , and D. R. Johnson, 2010: Temperature. Vol. 1, World Ocean Atlas 2009, NOAA Atlas NESDIS 68, 184 pp.

  • Lohmann, K., , H. Drange, , and M. Bentsen, 2009: Response of the North Atlantic subpolar gyre to persistent North Atlantic oscillation like forcing. Climate Dyn., 32, 273285, doi:10.1007/s00382-008-0467-6.

    • Search Google Scholar
    • Export Citation
  • Marsh, R., 2000: Recent variability of the North Atlantic thermohaline circulation inferred from surface heat and freshwater fluxes. J. Climate, 13, 32393260, doi:10.1175/1520-0442(2000)013<3239:RVOTNA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Marshall, J., and Coauthors, 1998: The Labrador Sea deep convection experiment. Bull. Amer. Meteor. Soc., 79, 20332058, doi:10.1175/1520-0477(1998)079<2033:TLSDCE>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Marshall, J., , H. Johnson, , and J. Goodman, 2001: A study of the interaction of the North Atlantic Oscillation with ocean circulation. J. Climate, 14, 13991421, doi:10.1175/1520-0442(2001)014<1399:ASOTIO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Medhaug, I., , and T. Furevik, 2011: North Atlantic 20th century multidecadal variability in coupled climate models: Sea surface temperature and ocean overturning circulation. Ocean Sci., 7, 389404, doi:10.5194/os-7-389-2011.

    • Search Google Scholar
    • Export Citation
  • Medhaug, I., , H. R. Langehaug, , T. Eldevik, , T. Furevik, , and M. Bentsen, 2012: Mechanisms for decadal scale variability in a simulated Atlantic meridional overturning circulation. Climate Dyn., 39, 7793, doi:10.1007/s00382-011-1124-z.

    • Search Google Scholar
    • Export Citation
  • Msadek, R., , and C. Frankignoul, 2009: Atlantic multidecadal oceanic variability and its influence on the atmosphere in a climate model. Climate Dyn., 33, 4562, doi:10.1007/s00382-008-0452-0.

    • Search Google Scholar
    • Export Citation
  • Nikurashin, M., , and G. Vallis, 2012: A theory of the interhemispheric meridional overturning circulation and associated stratification. J. Phys. Oceanogr., 42, 16521667, doi:10.1175/JPO-D-11-0189.1.

    • Search Google Scholar
    • Export Citation
  • Pickart, R. S., , D. J. Torres, , and R. A. Clarke, 2002: Hydrography of the Labrador Sea during active convection. J. Phys. Oceanogr., 32, 428457, doi:10.1175/1520-0485(2002)032<0428:HOTLSD>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Rahmstorf, S., , and M. H. England, 1997: Influence of Southern Hemisphere winds on North Atlantic deep water flow. J. Phys. Oceanogr., 27, 20402054, doi:10.1175/1520-0485(1997)027<2040:IOSHWO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Robson, J. I., , R. Sutton, , K. Lohmann, , D. Smith, , and M. D. Palmer, 2012a: Causes of the rapid warming of the North Atlantic Ocean in the mid-1990s. J. Climate, 25, 41164134, doi:10.1175/JCLI-D-11-00443.1.

    • Search Google Scholar
    • Export Citation
  • Robson, J. I., , R. Sutton, , and D. M. Smith, 2012b: Initialized decadal predictions of the rapid warming of the North Atlantic Ocean in the mid 1990s. Geophys. Res. Lett., 39, L19713, doi:10.1029/2012GL053370.

    • Search Google Scholar
    • Export Citation
  • Saenko, O. A., , and A. J. Weaver, 2004: What drives heat transport in the Atlantic: Sensitivity to mechanical energy supply and buoyancy forcing in the Southern Ocean. Geophys. Res. Lett., 31, L20305, doi:10.1029/2004GL020671.

    • Search Google Scholar
    • Export Citation
  • Seager, R., , Y. Kushnir, , M. Visbeck, , N. Naik, , J. Miller, , G. Krahmann, , and H. Cullen, 2000: Causes of Atlantic Ocean climate variability between 1958 and 1998. J. Climate, 13, 28452862, doi:10.1175/1520-0442(2000)013<2845:COAOCV>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Shaman, J., , R. M. Samelson, , and E. Skyllingstad, 2010: Air–sea fluxes over the Gulf Stream region: Atmospheric controls and trends. J. Climate, 23, 26512670, doi:10.1175/2010JCLI3269.1.

    • Search Google Scholar
    • Export Citation
  • Sijp, W. P., , and M. H. England, 2009: Southern Hemisphere westerly wind control over the ocean’s thermohaline circulation. J. Climate, 22, 12771286, doi:10.1175/2008JCLI2310.1.

    • Search Google Scholar
    • Export Citation
  • Smith, D. M., , R. Eade, , N. J. Dunstone, , D. Fereday, , J. M. Murphy, , H. Pohlmann, , and A. A. Scaife, 2010: Skilful multi-year predictions of Atlantic hurricane frequency. Nat. Geosci., 3, 846849, doi:10.1038/ngeo1004.

    • Search Google Scholar
    • Export Citation
  • Smith, R., and Coauthors, 2010: The Parallel Ocean Program (POP) reference manual: Ocean component of the Community Climate System Model (CCSM). LANL Tech. Rep. LAUR-10-01853, 141 pp.

  • Srokosz, M., , M. Baringer, , H. Bryden, , S. Cunningham, , T. Delworth, , S. Lozier, , J. Marotzke, , and R. Sutton, 2012: Past, present, and future change in the Atlantic meridional overturning circulation. Bull. Amer. Meteor. Soc., 93, 16631676, doi:10.1175/BAMS-D-11-00151.1.

    • Search Google Scholar
    • Export Citation
  • Steele, M., , R. Morley, , and W. Ermold, 2001: PHC: A global ocean hydrography with a high-quality Arctic Ocean. J. Climate, 14, 20792087, doi:10.1175/1520-0442(2001)014<2079:PAGOHW>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Sutton, R. T., , and D. L. R. Hodson, 2005: Atlantic Ocean forcing of North American and European summer climate. Science, 309, 115118, doi:10.1126/science.1109496.

    • Search Google Scholar
    • Export Citation
  • Thompson, D. W. J., , and S. Solomon, 2002: Interpretation of recent Southern Hemisphere climate change. Science, 296, 895899, doi:10.1126/science.1069270.

    • Search Google Scholar
    • Export Citation
  • Timmermann, A., , and H. Goosse, 2004: Is the wind stress forcing essential for the meridional overturning circulation? Geophys. Res. Lett., 31, L04303, doi:10.1029/2003GL018777.

    • Search Google Scholar
    • Export Citation
  • Toggweiler, J. R., , and B. Samuels, 1995: Effect of Drake Passage on the global thermohaline circulation. Deep-Sea Res. I, 42, 477500, doi:10.1016/0967-0637(95)00012-U.

    • Search Google Scholar
    • Export Citation
  • Toggweiler, J. R., , and B. Samuels, 1998: On the ocean’s large-scale circulation near the limit of no vertical mixing. J. Phys. Oceanogr., 28, 18321852, doi:10.1175/1520-0485(1998)028<1832:OTOSLS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Visbeck, M., , E. P. Chassignet, , R. Curry, , T. Delworth, , B. Dickson, , and G. Krahmann, 2003: The ocean’s response to North Atlantic Oscillation variability. The North Atlantic Oscillation: Climate Significance and Environmental Impact, Geophys. Monogr., Vol. 134, Amer. Geophys. Union, 113146.

    • Search Google Scholar
    • Export Citation
  • Wolfe, C. L., , and P. Cessi, 2010: What sets the strength of the middepth stratification and overturning circulation in eddying ocean models? J. Phys. Oceanogr., 40, 15201538, doi:10.1175/2010JPO4393.1.

    • Search Google Scholar
    • Export Citation
  • Wunsch, C., 2006: Abrupt climate change: An alternative view. Quat. Res., 65, 191203, doi:10.1016/j.yqres.2005.10.006.

  • Wunsch, C., , and R. Ferrari, 2004: Vertical mixing, energy, and the general circulation of the oceans. Annu. Rev. Fluid Mech., 36, 281314, doi:10.1146/annurev.fluid.36.050802.122121.

    • Search Google Scholar
    • Export Citation
  • Xu, X., , H. E. Hurlburt, , W. J. Schmitz Jr., , R. Zantopp, , J. Fischer, , and P. J. Hogan, 2013: On the currents and transports connected with the Atlantic meridional overturning circulation in the subpolar North Atlantic. J. Geophys. Res., 118, 502–516, doi:10.1002/jgrc.20065.

    • Search Google Scholar
    • Export Citation
  • Yashayaev, I., 2007: Hydrographic changes in the Labrador Sea, 1960-2005. Prog. Oceanogr., 73, 242276, doi:10.1016/j.pocean.2007.04.015.

    • Search Google Scholar
    • Export Citation
  • Yashayaev, I., , and J. W. Loder, 2009: Enhanced production of Labrador Sea water in 2008. Geophys. Res. Lett., 36, L01606, doi:10.1029/2008GL036162.

    • Search Google Scholar
    • Export Citation
  • Yeager, S. G., 2013: Understanding and predicting changes in North Atlantic sea surface temperature. Ph.D. thesis, University of Colorado, 176 pp.

  • Yeager, S. G., , and M. Jochum, 2009: The connection between Labrador Sea buoyancy loss, deep western boundary current strength, and Gulf Stream path in an ocean circulation model. Ocean Modell., 30, 207224, doi:10.1016/j.ocemod.2009.06.014.

    • Search Google Scholar
    • Export Citation
  • Yeager, S. G., , and G. Danabasoglu, 2012: Sensitivity of Atlantic meridional overturning circulation variability to parameterized Nordic Sea overflows in CCSM4. J. Climate, 25, 20772103, doi:10.1175/JCLI-D-11-00149.1.

    • Search Google Scholar
    • Export Citation
  • Yeager, S. G., , A. Karspeck, , G. Danabasoglu, , J. Tribbia, , and H. Teng, 2012: A decadal prediction case study: Late twentieth-century North Atlantic Ocean heat content. J. Climate, 25, 51735189, doi:10.1175/JCLI-D-11-00595.1.

    • Search Google Scholar
    • Export Citation
  • Zhang, L., , and C. Wang, 2013: Multidecadal North Atlantic sea surface temperature and Atlantic meridional overturning circulation variability in CMIP5 historical simulations. J. Geophys. Res., 118, 5772–5791, doi:10.1002/jgrc.20390.

    • Search Google Scholar
    • Export Citation
  • Zhang, R., 2010: Latitudinal dependence of Atlantic meridional overturning circulation (AMOC) variations. Geophys. Res. Lett., 37, L16703, doi:10.1029/2010GL044474.

    • Search Google Scholar
    • Export Citation
  • Zhang, R., , and T. Delworth, 2006: Impact of Atlantic multidecadal oscillations on India/Sahel rainfall and Atlantic hurricanes. Geophys. Res. Lett., 33, L17712, doi:10.1029/2006GL026267.

    • Search Google Scholar
    • Export Citation
1

Periodically updated, the dataset has since been extended through 2009.

2

Here and in what follows, we use a 15-point Lanczos filter (Duchon 1979) with a cutoff period of 7 yr to isolate decadal and longer time scales. This filter passes almost all of the variance at periods longer than 10 yr and half of the variance at a period of 7 yr.

Save