• Anklin, M., , R. Bales, , E. Mosley-Thompson, , and K. Steffen, 1998: Annual accumulation at two sites in northwest Greenland during recent centuries. J. Geophys. Res, 103 (D22) 2877528783.

    • Search Google Scholar
    • Export Citation
  • Banks, H. T., , R. A. Wood, , J. M. Gregory, , T. C. Johns, , and G. S. Jones, 2000: Are observed decadal changes in intermediate water masses a signature of anthropogenic climate change? Geophys. Res. Lett, 27 , 29612964.

    • Search Google Scholar
    • Export Citation
  • Broecker, W. S., 1991: The great ocean conveyor. Oceanography, 4 , 7989.

  • Collins, M., , and B. Sinha, 2003: Predictability of decadal variations in the thermohaline circulation and climate. Geophys. Res. Lett.,30, 1413, doi:10.1029/2002GL016776.

    • Search Google Scholar
    • Export Citation
  • Collins, M., , S. F. B. Tett, , and C. Cooper, 2001: The internal climate variability of HadCM3, a version of the Hadley Centre coupled model without flux adjustments. Climate Dyn, 17 , 6181.

    • Search Google Scholar
    • Export Citation
  • Cox, P. M., , R. A. Betts, , C. B. Bunton, , R. L. H. Essery, , P. R. Rowntree, , and J. Smith, 1999: The impact of new land surface physics on the GCM simulation of climate and climate sensitivity. Climate Dyn, 15 , 183203.

    • Search Google Scholar
    • Export Citation
  • Delworth, T. L., , and K. Dixon, 2000: Implications of the recent trend in the Arctic/North Atlantic oscillation for the North Atlantic thermohaline circulation. J. Climate, 13 , 37213727.

    • Search Google Scholar
    • Export Citation
  • Delworth, T. L., , and R. J. Greatbatch, 2000: Multidecadal thermohaline circulation variability driven by atmospheric flux forcing. J. Climate, 13 , 14811495.

    • 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.

    • Search Google Scholar
    • Export Citation
  • Delworth, T. L., , S. Manabe, , and R. J. Stouffer, 1993: Interdecadal variations of the thermohaline circulation in a coupled ocean–atmosphere model. J. Climate, 6 , 19932011.

    • Search Google Scholar
    • Export Citation
  • Delworth, T. L., , R. J. Stouffer, , K. W. Dixon, , M. J. Spelman, , T. R. Knutson, , A. J. Broccoli, , P. J. Kushner, , and R. T. Wetherald, 2002: Review of simulations of climate variability and change with the GFDL R30 coupled climate model. Climate Dyn.,19, doi:10.1007/ s00382-002-0249-5.

    • 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.

    • Search Google Scholar
    • Export Citation
  • Figueroa, H. A., 1994: Eddy resolution versus eddy diffusion in a double gyre GCM. Part II: Mixing of passive tracers. J. Phys. Oceanogr, 24 , 387402.

    • Search Google Scholar
    • Export Citation
  • Ganachaud, A., , and C. Wunsch, 2000: Improved estimates of global ocean circulation, heat transport and mixing from hydrographic data. Nature, 408 , 453457.

    • Search Google Scholar
    • Export Citation
  • Gill, A. E., 1980: Some simple solutions for heat-induced tropical circulation. Quart. J. Roy. Meteor. Soc, 106 , 447462.

  • Gordon, A. L., 1986: Inter-ocean exchange of thermocline water. J. Geophys. Res, 91 , 50375046.

  • Gordon, C., , C. Cooper, , C. A. Senior, , H. Banks, , J. M. Gregory, , T. C. Johns, , J. F. B. Mitchell, , and R. A. Wood, 2000: The simulation of SST, sea ice extents and ocean heat transports in a version of the Hadley Centre coupled model without flux adjustments. Climate Dyn, 16 , 147168.

    • Search Google Scholar
    • Export Citation
  • Hughes, T., , and A. Weaver, 1994: Multiple equilibria of an asymmetric two-basin model. J. Phys. Oceanogr, 24 , 619637.

  • Hurrell, J., 1995: Decadal trends in the North Atlantic Oscillation: Regional temperatures and precipitation. Science, 269 , 676679.

  • Kim, J-H., , and R. R. Schneider, 2003: Low-latitude control of interhemispheric sea-surface temperature contrast in the tropical Atlantic over the past 21 kyears: The possible role of SE trade winds. Climate Dyn.,21, doi:10.1007/s00382-003-0341-5.

    • Search Google Scholar
    • Export Citation
  • Klinger, B., , and J. Marotzke, 1999: Behavior of double-hemisphere thermohaline flows in a single basin. J. Phys. Oceanogr, 29 , 382399.

    • Search Google Scholar
    • Export Citation
  • Latif, M., , E. Roeckner, , U. Mikolajewicz, , and R. Voss, 2000: Tropical stabilization of the thermohaline circulation in a greenhouse warming simulation. J. Climate, 13 , 18091813.

    • Search Google Scholar
    • Export Citation
  • Mann, M. E., , and J. Park, 1994: Global-scale modes of surface temperature variability on interannual to century timescales. J. Geophys. Res, 99 , 2581925833.

    • Search Google Scholar
    • Export Citation
  • Mann, M. E., , R. S. Bradley, , and M. K. Hughes, 1999: Northern Hemisphere temperatures during the past millennium: Inferences, uncertainties, and limitations. Geophys. Res. Lett, 26 , 759762.

    • Search Google Scholar
    • Export Citation
  • Marotzke, J., 1997: Boundary mixing and the dynamics of three-dimensional thermohaline circulations. J. Phys. Oceanogr, 27 , 17131728.

    • Search Google Scholar
    • Export Citation
  • Marotzke, J., , and B. Klinger, 2000: The dynamics of equatorially asymmetric thermohaline circulation. J. Phys. Oceanogr, 30 , 955970.

  • 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.

    • Search Google Scholar
    • Export Citation
  • Nyberg, J., , B. A. Malmgren, , A. Kuijpers, , and A. Winter, 2002: A centennial-scale variability of tropical North Atlantic surface hydrography during the late Holocene. Palaeogeogr. Palaeoclimatol. Palaeoecol, 183 , 2541.

    • Search Google Scholar
    • Export Citation
  • Pardaens, A. K., , H. T. Banks, , J. M. Gregory, , and P. R. Rowntree, 2003: Freshwater transports in HadCM3. Climate Dyn, 21 , 177195.

  • Parker, D. E., , T. P. Legg, , and C. K. Folland, 1992: A new daily Central England temperature series. Int. J. Climatol, 12 , 317342.

  • Pope, V. D., , M. L. Gallani, , P. R. Rowntree, , and R. A. Stratton, 2000: The impact of new physical parametrizations in the Hadley Centre climate model—HadAM3. Climate Dyn, 16 , 123146.

    • Search Google Scholar
    • Export Citation
  • Proctor, C. J., , A. Baker, , and W. L. Barnes, 2002: A three thousand year record of North Atlantic climate. Climate Dyn.,19, doi:10.1007/s00382-002-0236-x.

    • Search Google Scholar
    • Export Citation
  • Rahmstorf, S., 1996: On the freshwater forcing and transport of the Atlantic thermohaline circulation. Climate Dyn, 12 , 799811.

  • Rahmstorf, S., 2003: The current climate. Nature, 421 , 699.

  • Ruehlemann, C., , S. Mulitza, , P. J. Mueller, , G. Wefer, , and R. Zahn, 1999: Warming of the tropical Atlantic Ocean and slowdown of thermohaline circulation during the last deglaciation. Nature, 402 , 511514.

    • Search Google Scholar
    • Export Citation
  • Sausen, R., , K. Barthel, , and K. Hasselmann, 1988: Coupled ocean– atmosphere models with flux corrections. Climate Dyn, 2 , 154163.

  • Schmidt, M. W., , H. J. Spero, , and D. W. Lea, 2004: Links between salinity variation in the Caribbean and North Atlantic circulation. Nature, 428 , 160163.

    • Search Google Scholar
    • Export Citation
  • Schmittner, A., , C. Appenzeller, , and T. Stocker, 2000: Enhanced Atlantic freshwater export during El Niño. Geophys. Res. Lett, 27 , 11631166.

    • Search Google Scholar
    • Export Citation
  • Spall, M. A., , and R. S. Pickart, 2001: Where does dense water sink? A subpolar gyre example. J. Phys. Oceanogr, 31 , 810826.

  • Thorpe, R. B., , J. M. Gregory, , T. C. Johns, , R. A. Wood, , and J. F. B. Mitchell, 2001: Mechanisms determining the Atlantic thermohaline circulation response to greenhouse gas forcing in a non-flux-adjusted coupled climate model. J. Climate, 14 , 31023116.

    • Search Google Scholar
    • Export Citation
  • Timmermann, A., , M. Latif, , R. Voss, , and A. Grötzner, 1998: Northern Hemispheric interdecadal variability: A coupled air–sea mode. J. Climate, 11 , 19061931.

    • Search Google Scholar
    • Export Citation
  • Torrence, C., , and G. P. Compo, 1998: A practical guide to wavelet analysis. Bull. Amer. Meteor. Soc, 79 , 6178.

  • Vellinga, M., , and H. Banks, 2002: Climatological mean heat transport and its variability in the coupled climate model HadCM3. International WOCE Newsletter, No. 42, WOCE International Project Office, Southampton, United Kingdom, 6–8.

    • Search Google Scholar
    • Export Citation
  • Vellinga, M., , R. A. Wood, , and J. M. Gregory, 2002: Processes governing the recovery of a perturbed thermohaline circulation in HadCM3. J. Climate, 15 , 764780.

    • Search Google Scholar
    • Export Citation
  • von Storch, J. S., , P. Müller, , R. J. Stouffer, , R. Voss, , and S. F. B. Tett, 2000: Variability of deep-ocean mass transport: Spectral shapes and spatial scales. J. Climate, 13 , 19161935.

    • Search Google Scholar
    • Export Citation
  • Webster, P. J., 1981: Mechanisms determining the atmospheric response to sea surface temperature anomalies. J. Atmos. Sci, 38 , 554571.

    • Search Google Scholar
    • Export Citation
  • Wilks, D. S., 1997: Resampling hypothesis tests for autocorrelated fields. J. Climate, 10 , 6582.

  • Wu, P., , and C. Gordon, 2002: Oceanic influence on North Atlantic climate variability. J. Climate, 15 , 19111925.

  • Wunsch, C., 2002: What is the thermohaline circulation? Science, 298 , 11791180.

  • View in gallery

    Meridional overturning streamfunction in the Atlantic Ocean in HadCM3. Positive values (solid) indicate clockwise circulation, negative values (dashed) indicate counterclockwise circulation. (a) Average over 1600 yr. Units (Sv), interval of 2 Sv. (b) First EOF of annual-mean streamfunction. Linear correlation with annual-mean SST anomalies in the Niño-3.4 area is 0.40. (c) Second EOF of annual-mean streamfunction. Linear correlation with the NAO index (defined as the normalized, annual-mean sea level pressure difference between the Azores and Iceland) is −0.64. (d) First EOF of decadal-mean streamfunction. Fractional variance of each EOF, and linear correlation between the expansion coefficients of each EOF and the MOI is shown at the top. EOFs are scaled such that the standard deviation of their respective principal components is 0.1

  • View in gallery

    Composite analysis of upper-ocean currents (averaged over 0–800 m) of the decades with the 20 strongest minus decades with the 20 weakest values of the MOI. For clarity, arrows are only shown on every second grid point, and if larger than 0.2 cm s−1

  • View in gallery

    Detrended time series of maximum value of the annual-mean meridional streamfunction in the Atlantic at 45°N

  • View in gallery

    Wavelet power spectrum of time series of Fig. 3 using Morlet wavelet (gray shading, units Sv2). Model time (yr) runs along the horizontal axis, and period (yr) runs along vertical axis. The solid contour encloses areas where the wavelet power exceeds the 95% confidence level of an AR-(1) (or red-noise) process. One expects 5% of the points to do so by chance; here the contour encloses about 10% of the points. Dashed curve indicates where edge effects become important and wavelet power is underestimated. More details on wavelet analysis are given by Torrence and Compo (1998)

  • View in gallery

    Regression of decadally averaged SST anomalies onto MOI (units 0.01°C Sv−1, contours values at ±50, 25, 10, 5, 2.5, and 0). Positive values have solid contours, negative values have dash–dotted contours, and zero contours are heavy solid lines. Shading indicates where regression is significantly different from zero at 95% level

  • View in gallery

    Composite atmospheric anomaly patterns induced by THC fluctuations, obtained by averaging anomalies over 20 decades when the THC is strongest. Colors indicate where the null hypothesis of equal means is rejected at the 10% level (see text for details). (a) Surface air temperature anomalies in °C, with positive contours solid, negative contours dashed, and zero contour heavy. (b) As in (a) but for MSLP; units hPa. (c) As in (a) but for precipitation; units cm yr−1

  • View in gallery

    (a) Regression of vertically averaged density (0–800 m) onto MOI (units: 10−2 kg m−3 Sv−1). Shading indicates where regression is different from zero at the 95% confidence level. (b) Regression of vertical velocity at 700 m onto MOI (units: 10−5 cm s−1 Sv−1). Shading indicates where values are less than −2 × 10−5 cm s−1 Sv−1

  • View in gallery

    (a) Time series of decadal MOI (light) and zonal mean density anomalies at 60°N (heavy), averaged over the top 800 m of the water column. Data was low-pass filtered with a half-period of about 5 decades, and then normalized by standard deviation (shown in legend) to enable a direct comparison. Correlation between the two time series is shown in bottom-left corner. (b) As in (a) but for salinity-driven contributions to density Sσ. (c) As in (a) but for temperature-driven contributions to density Θσ

  • View in gallery

    Linear correlation between salinity contributions to density and zonal mean density at 60°N, both averaged over 0–800 m. Lead time of salinity (decades) over density at 60°N is shown at top. Contour interval is 0.1. Values larger than 0.4 are dotted, smaller than −0.4 hatched. To highlight the centennial fluctuation data were low-pass filtered with a half-period of about 5 decades

  • View in gallery

    (a) Salinity-driven density tendencies, ∂Sσ/∂t (solid), and temperature-driven density tendencies, ∂Θσ/∂t (dashed), regressed onto MOI and averaged between 0–800 m and over three regions of the Atlantic Ocean: (sub) Tropics (0°–35°N, black), North Atlantic Current (35°–48°N, red), and subpolar (48°–65°N, green). Horizontal axis shows lag in decades of density relative to MOI; at negative lags density leads MOI. (b) Area-integrated regression fields of all processes modifying salinity in subtropical domain. Colors and labeling identify individual processes explained in legend. The heavy black unlabeled curve is the net tendency (σ) as in (a). (c) As in (b), but for the North Atlantic Current region. (d) As in (b) but for the subpolar region

  • View in gallery

    Evolution (from 1 to 6 decades) of passive tracer concentration, averaged over 0–800 m. Contour interval is 0.1; concentrations larger than 0.4 are hatched. Tracer values may be interpreted as the fraction of water in the upper 800 m that has been at the surface between 0° and 15°N at any moment in the experiment

  • View in gallery

    Hovmöller plot of zonal mean salinity anomalies, averaged over 0–800 m, expressed as potential density anomalies; compare with Eq. (1) (colors, units kg m−3). Time runs down along vertical axis, latitude along horizontal axis. The solid curve centered at 50°N is the MOI time series, and the dotted curve centered at 10°N is the surface freshwater flux anomaly (evaporation–precipitation–runoff, labeled SFC), integrated over 0°–15°N. Positive anomalies are northward deflections, with scale indicated in (a). Positive anomaly of SFC is out of the ocean, making it saltier

  • View in gallery

    Regression of SST (shading, in 10−2 °C Sv−1), and surface wind stress (arrows, in N m−2 Sv−1) onto lagged MOI: SST and winds lead MOI by 6 decades

  • View in gallery

    Net surface freshwater flux into the tropical North Atlantic (between 0° and 15°N), vs cross-equatorial Atlantic SST gradient (0°– 15°N minus 0°–15°S). Solid triangles denote decades that are followed 60 yr later by one of the 20 decades with weakest THC. Solid circles denote the decades preceding the 20 decades with strongest THC by 60 yr

  • View in gallery

    Meridional cross section over the tropical Atlantic of the vertical motion ω in the troposphere (negative values indicate upward motion). The gray shading shows the time-mean field (in units of 10−3 Pa s−1; zero contour is highlighted by heavy solid line). The light contours show the regression of ω against MOI (units of 10−4 Pa s−1 Sv−1; negative contours are dashed)

  • View in gallery

    Schematic of mechanism responsible for centennial THC fluctuation in HadCM3. When the THC is (left) strong ITCZ shifts northward, in response to enhanced SST gradient across equator. Fresh anomalies in the upper-ocean propagate northward and weaken the overturning. This results in the (right) weak phase

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 68 68 31
PDF Downloads 25 25 9

Low-Latitude Freshwater Influence on Centennial Variability of the Atlantic Thermohaline Circulation

View More View Less
  • 1 Hadley Centre for Climate Prediction and Research, Met Office, Exeter, Devon, United Kingdom
© Get Permissions
Full access

Abstract

Variability of the thermohaline circulation (THC) has been analyzed in a long control simulation by the Met Office's Third Hadley Centre Coupled Ocean–Atmosphere General Circulation Model (HadCM3). It is shown that internal THC variability in the coupled climate system is concentrated at interannual and centennial time scales, with the centennial mode being dominant. Centennial oscillations of the THC can impact surface climate via an interhemispheric SST contrast of 0.1°C in the Tropics and more than 0.5°C in mid- and high latitudes. A mechanism is proposed based on detailed process analysis involving large-scale air–sea interaction on multidecadal time scales. Anomalous northward ocean heat transport associated with a strong phase of the Atlantic THC generates a cross-equatorial SST gradient. This causes the ITCZ to move to a more northerly position with increased strength. The extra rainfall resulting from the anomalous ITCZ imposes a freshwater flux and produces a salinity anomaly in the tropical North Atlantic. Such sustained salinity anomalies slowly propagate toward the subpolar North Atlantic at a lag of 5–6 decades. The accumulated low-salinity water lowers upper-ocean density, which causes the THC to slow down. The oscillation then enters the opposite phase.

Corresponding author address: Michael Vellinga, Hadley Centre for Climate Prediction and Research, Met Office, FitzRoy Road, Exeter, Devon EX1 3PB, United Kingdom. Email: michael.vellinga@metoffice.gov.uk

Abstract

Variability of the thermohaline circulation (THC) has been analyzed in a long control simulation by the Met Office's Third Hadley Centre Coupled Ocean–Atmosphere General Circulation Model (HadCM3). It is shown that internal THC variability in the coupled climate system is concentrated at interannual and centennial time scales, with the centennial mode being dominant. Centennial oscillations of the THC can impact surface climate via an interhemispheric SST contrast of 0.1°C in the Tropics and more than 0.5°C in mid- and high latitudes. A mechanism is proposed based on detailed process analysis involving large-scale air–sea interaction on multidecadal time scales. Anomalous northward ocean heat transport associated with a strong phase of the Atlantic THC generates a cross-equatorial SST gradient. This causes the ITCZ to move to a more northerly position with increased strength. The extra rainfall resulting from the anomalous ITCZ imposes a freshwater flux and produces a salinity anomaly in the tropical North Atlantic. Such sustained salinity anomalies slowly propagate toward the subpolar North Atlantic at a lag of 5–6 decades. The accumulated low-salinity water lowers upper-ocean density, which causes the THC to slow down. The oscillation then enters the opposite phase.

Corresponding author address: Michael Vellinga, Hadley Centre for Climate Prediction and Research, Met Office, FitzRoy Road, Exeter, Devon EX1 3PB, United Kingdom. Email: michael.vellinga@metoffice.gov.uk

1. Introduction

Instrumental records show that climate varies at multidecadal to centennial periods at regional (Parker et al. 1992) as well as hemispherical scales (Mann and Park 1994). Understanding the processes that lead to climate fluctuations with a time scale of several decades or longer is obviously hampered by limited availability of sufficiently long instrumental records. The growing abundance of paleo data has led to improved estimates of past climate variability, and has shown that there are fluctuations in Northern Hemisphere temperature with periods of 70–80 yr (Mann et al. 1999). Natural climate fluctuations that occur on periods of several decades to a century are important, because of their potential to modulate anthropogenically forced climate change (Delworth and Dixon 2000).

Numerical models can be used to help understand the processes that give rise to the climate variability seen in paleo and instrumental records. There are indications that the current generation of climate models can indeed simulate aspects of multidecadal climate variability, as seen in the instrumental record (Knight et al. 2004, manuscript submitted to Nature, hereafter KNI) and proxy data (Delworth and Mann 2000). In several long simulations with climate GCMs, internal climate variability on multidecadal time scales has been linked to fluctuations in the ocean's thermohaline circulation (THC). [Although meridional overturning circulation is the correct term to refer to the large-scale overturning circulation of the Atlantic, we will use THC throughout this paper, because this term is used more widely. See Wunsch (2002) and Rahmstorf (2003) for a discussion]. In different versions of the Geophysical Fluid Dynamics Laboratory (GFDL) model there is internal THC variability with a dominant period of about 50 yr (Delworth et al. 1993, 2002; Delworth and Greatbatch 2000). In the European Centre Hamburg/Large-Scale Geostrophic model (ECHAM3/LSG), Timmermann et al. (1998) identified THC variability with a period of 35 yr. A common feature of the multidecadal variability in these models is that the THC responds to changes in upper-ocean density, caused by near-surface salinity changes over the North Atlantic sinking regions. The models differ in how these salinity anomalies arise: through modification of air–sea interaction over the midlatitude and subpolar North Atlantic in the ECHAM3/LSG model; or by changes in oceanic salt transport in response to density changes in the subpolar gyre for the GFDL model.

In this paper we will discuss THC variability simulated in the control integration of the Third Hadley Center Coupled Ocean–Atmosphere General Circulation Model (HadCM3). This is the most recent version of the Met Office's Hadley Centre climate model (Gordon et al. 2000). Our aims are 1) to identify the dominant mode of THC variability in the model, 2) to discuss its impact on surface climate, and 3) to analyze the mechanisms responsible. The layout of this paper is as follows: a brief description of the coupled model is given in section 2. In section 3 we describe the behavior of the THC. Relationships with surface climate are discussed in section 4. The analysis of the underlying physical processes is the subject of section 5. Conclusions and a discussion in section 6 end the paper.

2. Model description

HadCM3 is a coupled, global ocean–atmosphere–sea ice model. The model is described in detail by Gordon et al. (2000), Pope et al. (2000), and Cox et al. (1999). Here we only give a brief summary. The atmosphere component has a horizontal resolution of 2.5° × 3.75° and 19 vertical levels. The ocean component has a horizontal resolution of 1.25° × 1.25° with 20 vertical levels, 10 of which are in the upper 300 m. The sea ice model uses a simple thermodynamic scheme and contains parameterizations of ice drifts and leads. Finally, HadCM3 has a land surface scheme which includes freezing and melting of soil moisture. Contributions from 23 types of vegetation are also included. Distribution of the vegetation varies geographically, but is fixed in time.

An important improvement of HadCM3 over its predecessor HadCM2 is that the net radiative flux at the top of the atmosphere is close to zero. HadCM3 does not require flux adjustment to maintain a stable control climate (Sausen et al. 1988). One important implication of this is that surface fluxes and meridional transports (e.g., of heat) are physically consistent. A validation of the model simulation of some important climate quantities is given by Gordon et al. (2000). Simulation by HadCM3 of internal climate variability on an interannual time scale has amplitudes and patterns that are in overall agreement to the observed variability (Collins et al. 2001). It takes about 400 yr of integration before heat and freshwater budgets in the ocean attain a balance (Gordon et al. 2000; Pardaens et al. 2003). We will therefore use 1600 yr of model data, from year 400 onward. This model run uses fixed, preindustrial levels of greenhouse gas and aerosol concentrations. Unless stated otherwise, we will use decadally averaged model data.

3. Internal THC variability in HadCM3

The global ocean circulation that connects the major basins through large-scale warm and cold flows (Gordon 1986; Broecker 1991) is referred to as the THC. In ocean models, the meridional overturning streamfunction (Fig. 1a) is traditionally used as a proxy for the THC. In HadCM3 the mean meridional streamfunction in the Atlantic Ocean has indeed similar features as the mean THC inferred from estimates using observations (Ganachaud and Wunsch 2000): about 18 Sv (1 Sv ≡ 106 m3 s−1) of warm northward flow in the upper ocean, mainly in the Gulf Stream and North Atlantic Current; sinking in the Nordic and Labrador Seas and near the ridges between Greenland and Scotland; and a southward return flow at depth of North Atlantic Deep Water (NADW), that overlies a deep circulation cell of Antarctic Bottom Water of about 6 Sv. As a measure for the strength of the Atlantic THC, modellers often use the maximum value attained in the Atlantic by the meridional overturning streamfunction. Following Delworth et al. (1993) we will refer to this quantity as the meridional overturning index (MOI).

The latitude at which this maximum value is reached may change with time. This can make a physical interpretation of MOI variability difficult. To analyze the spatial patterns associated with variability of the THC, we have calculated EOFs of 1600 yr of annual-mean meridional overturning streamfunction. We look at interannual variability first, and applied a high-pass filter to the data that removes variability at periods longer than 10 yr. The first EOF (Fig. 1b) of annual-mean overturning consists of a circulation cell that is most pronounced in the upper 4000 m of the subtropical and tropical Atlantic (south of about 30°N). This EOF can be associated with ENSO variability: the expansion coefficient of EOF1 has a correlation with the Niño-3.4 index (5°N–5°S, 120°–170°W) of 0.40. In El Niño years there is anomalous downwelling in the Atlantic between 0°–30°N, and upwelling near 30°S. The second EOF of the annual-mean meridional overturning (Fig. 1c) is most pronounced in the subtropical and midlatitude North Atlantic. Prominent features of this EOF are two cells of opposite sign that extend from the surface to the bottom of the ocean. The expansion coefficient of EOF2 has a correlation with the model's North Atlantic Oscillation index of −0.64. In (NAO; Hurrell 1995) years of a positive NAO index, there is anomalous downwelling in the Atlantic near 40°N, and upwelling to the north and south. A direct regression of Niño-3.4 SST anomalies or the NAO index onto the overturning data gives patterns (not shown) that are similar to those of Figs. 1b,c, but are more confined to the South Atlantic for Niño-3.4 index, or midlatitude North Atlantic for the NAO index. The same patterns also emerge for regressions of overturning data onto wind stress anomalies at 20°S and 30°N (not shown). This suggests that EOFs 1 and 2 capture the projection in the vertical–meridional plane of the wind-driven ocean response, associated with ENSO and NAO variability, respectively. The latter is consistent with the barotropic ocean response to NAO wind forcing described by Eden and Willebrand (2001). The leading EOF of the meridional streamfunction in four older-generation climate models resembles either our EOF 1 or EOF 2 of Figs. 1b,c (von Storch et al. 2000). The differences in the leading EOF in those four models may thus be caused by models having largest variance in wind-driven fluctuations at either low latitudes (as in HadCM3) or at midlatitudes.

Time evolution of these spatial patterns of variability in the annual meridional overturning is only partly captured by the MOI: linear correlation between the MOI and the expansion coefficients of EOFs 1 and 2 is −0.02 and −0.29. However, when we use 160 decades of decadally averaged data to calculate EOFs of the meridional overturning, we find that the MOI does correlate well (0.82) with the expansion coefficient of the leading EOF of the decadal-mean circulation. This EOF has a basinwide structure (Fig. 1d) which suggests that on decadal and longer time scales the MOI does capture a spinup/spin-down of the entire NADW cell. A composite analysis of upper-ocean currents during decades with high and low values of the MOI confirms this (Fig. 2). Fluctuations in the MOI correspond to anomalies that span large parts of the upper limb of the NADW cell: the Southern Hemisphere Brasil Current, and in the Northern Hemisphere the Antilles Current and Gulf Stream and North Atlantic Current.

We conclude that the MOI is not a useful measure to quantify variability of the THC on an interannual time scale. On time scales of decades or longer the MOI does provide a useful description of THC variability. To quantify the strength of the THC at various time scales it is preferable to use the maximum of the meridional streamfunction at a fixed latitude. We choose 45°N, that is, near the center of the NADW cell. The time series using annual-mean data is shown in Fig. 3. Apart from the interannual variability, which at this latitude is predominantly the ocean response to the NAO, the most striking features are the large excursions of the THC on time scales of several decades to centuries. The variability on a centennial time scale varies with time, with periods of strong and weak activity. Wavelet analysis provides information on how variance of periodic signals within a time series vary with period, and time. Figure 4 shows a wavelet analysis of the annual-mean maximum overturning streamfunction at 45°N (Torrence and Compo 1998). Variance exceeding levels that may be explained by a red-noise process shows broad local maxima on interannual time scales and time scales of 70–200 yr. At intermediate time scales of 1–3 decades wavelet power is intermittent. The wavelet analysis shows that variability at a centennial time scale is the dominant mode of internal THC variability in the control simulation of HadCM3.

4. Climate impact of the centennial mode

a. Ocean

In HadCM3, the mean meridional flow carries most of the mean heat transport in the Atlantic. It also contributes most to heat transport variability (Vellinga and Banks 2002). Figure 5 presents a regression of SST onto the MOI, showing how centennial THC fluctuations and its associated heat transport could affect the upper ocean. SST anomalies associated with centennial THC fluctuations have an interhemispheric pattern: when the THC is stronger, SST over most of the Northern Hemisphere becomes warmer, while over a large part of the Southern Hemisphere SST cools slightly. The SST signal is strongest in the Atlantic Ocean, and clearer in the Northern Hemisphere than in the Southern Hemisphere. A similar pattern of global, interhemispheric SST anomalies but with larger amplitude was identified in HadCM3, after the THC had been forced to collapse almost completely (Vellinga et al. 2002). The typical amplitude of SST anomalies in the Atlantic varies with latitude, from about 0.1°C Sv−1 in the (sub)tropics, to over 1°C Sv−1 in the Greenland Sea. As this pattern persists for several decades, the atmosphere could be expected to respond, which is discussed later.

b. Atmosphere

Low-frequency THC variability could provide a source of decadal predictability of climate (Collins and Sinha 2003). Studies have also tried to quantify the links between multidecadal changes in the instrumental surface temperature record and the THC (KNI). Here we describe anomalies in the atmosphere associated with the centennial THC fluctuation, with an emphasis on significance. An important point, not explicitly addressed by these previous studies, is whether the signal-to-noise ratio of THC-induced atmospheric anomalies is large enough to be noted in individual seasons, when high-frequency (background) variability is present. This high-frequency variability is typically eliminated, by using smoothed or decadal data. For practical purposes we will only consider variability on time scales of seasons and longer.

Composite atmospheric anomaly patterns induced by THC fluctuations were obtained by averaging anomalies over 20 decades when the THC is strongest. This is done separately for each of the four seasons, but only results for boreal winter [December–January–February (DJF)] are shown here. Significance of anomalies in each grid point is assessed by testing the null hypothesis that they were taken from a distribution that has the same mean as that of the anomalies in the remaining decades. To account for serial correlation in the fields we have used the moving blocks bootstrap method of Wilks (1997) to estimate sample variance.

Shown in Fig. 6 are DJF anomalies of surface air temperature, precipitation, and mean sea level pressure (MSLP). Only significant values (at the 10% level) are colored. Surface air temperature anomalies resemble the SST regression pattern of Fig. 5, and vary from about 0.1°C in the Tropics and subtropics, to over 1°C over the Greenland Sea. At high latitudes sea ice cover is reduced when a strong THC brings in warmer water, and amplifies atmospheric warming. Surface air temperature anomalies are significant (in the sense as defined earlier) mainly over the Atlantic and North Pacific Oceans, as well as parts of Europe and large parts of Asia. Lower MSLP appears near the Icelandic low and extends over large parts of Europe, Fig. 6b. Such a pattern would have a projection on the NAO, possibly reflecting some oceanic (THC) influence on the NAO at decadal and longer time scales (Marshall et al. 2001; Wu and Gordon 2002). The most striking features in precipitation (Fig. 6c) appear in the tropical Atlantic, where rainfall anomalies of opposite sign exceeding ±10 cm yr−1 occur on each side of the equator. The pattern clearly indicates a shift of the ITCZ associated with the THC anomolies. A strong THC is accompanied by a northward shift of the ITCZ. As further discussed later in section 5e, the freshwater flux resulted from a such a shift of the ITCZ plays a crucial role in the centennial oscillation of the THC. Anomalous rainfall in the Tropics is not limited to the Atlantic sector: anomalies can be readily communicated elsewhere via the equatorial waveguide (Gill 1980). Precipitation anomalies over NH midlatitude land areas weak and less significant.

5. Mechanism of centennial THC fluctuations in the Atlantic

a. Relation of THC to density anomalies

From the geostrophic balance one expects a relation between zonal mean meridional flow and zonal pressure difference. Theoretical considerations (Marotzke 1997) and GCM experiments for idealized geometry (Klinger and Marotzke 1999) have shown that the maximum east–west density difference scales as the north–south surface density difference. This may explain the strong empirical relation between the equilibrium strength of the THC and the large-scale meridional density difference found in several GCMs with more realistic geometry (Hughes and Weaver 1994; Rahmstorf 1996; Thorpe et al. 2001). We find that in HadCM3 low-frequency THC fluctuations are also associated with a pattern of basin-scale density anomalies. Figure 7a is a linear regression of the top 800-m vertically averaged density field on the MOI. Strongest anomalies occur along the western edge of the North Atlantic, that is, in the Labrador Sea and along the east coast of Greenland. This anomaly pattern sets up an enhanced west–east density difference which drives a thermal wind shear, that results in the vertically averaged flow of Fig. 2. The convergence of this anomalous flow feeds into the downwelling branch of the THC, (Fig. 7b). This effect is strongest adajcent to steep topography, anomalous downwelling in the open ocean is much weaker. Topographic boundaries can support the density gradient necessary for downwelling, as described by Spall and Pickart (2001). In the northern subtropical gyre are density anomalies of opposite sign, that set up a north– south density contrast across the North Atlantic Current. There is only a weak density signal in the South Atlantic, in contrast with the equilibrium case (Thorpe et al. 2001). This relation between centennial fluctuations in the MOI and upper-ocean density also holds for zonal mean density anomalies. The relation is particularly strong at 60°N (Fig. 8a), consistent with Thorpe et al. (2001).

b. Relation of density to salinity anomalies

To understand the cause of density anomalies (σθ) in Figs. 7a and 8a, they are written as the sum of contributions from potential temperature (θ′) and salinity (S′) anomalies:
i1520-0442-17-23-4498-e1
where angled brackets denote vertical average over depth range of 0–800 m, and prime denotes decadal mean anomaly from long-term mean. The derivatives with respect to density are for local θ and S and take into account the nonlinearity of the equation of state. Time series of decomposition (1) at 60°N show that a strong THC occurs almost always when there are anomalously saline conditions at 60°N, Fig. 8b. Temperature anomalies at 60°N, on the other hand, have no consistent relation to THC anomalies, Fig. 8c. Sometimes they occur in phase with the THC, sometimes in antiphase, and overall correlation is weak. In what follows we will therefore concentrate on the origin of salinity anomalies.

First we map changes in salinity of the upper ocean in the decades before dense conditions at 60°N occur. For a clearer salinity signal we want to eliminate at this stage the effects of local feedbacks that amplify or damp the anomalies. We therefore use a lagged correlation analysis between 〈(∂σθ/∂S)S′〉 and zonal mean density at 60°N. Earliest signs of positive salinity anomalies appear in the subtropical west Atlantic about 6 decades before a positive density anomaly at 60°N, Fig. 9a. This lead time corresponds to about half the period of the centennial mode. At this stage most of the northern North Atlantic is fresher than normal, consistent with the THC being in a weak phase. These fresh conditions at high latitudes gradually diminish and give way to more saline conditions over the following decades (Figs. 9b–d). It takes until lead time of 2 decades for positive anomalies to traverse the path of the North Atlantic Current, (Fig. 9e). After that most of the Atlantic north of 45°N is anomalously saline, while there is some indication of negative anomalies building up in the tropical North Atlantic, Fig. 9f. Similar salinity correlation patterns emerge if the MOI is used instead of the density time series.

c. Salinity budget for typical THC fluctuation

Since salinity anomalies play a key role in centennial THC fluctuations, we will concentrate on the process through which salinity anomalies are created and removed over one THC cycle. A budget analysis is carried out using model diagnostics of the rate of change due to various terms in the tracer equations (advection, mixing, surface fluxes, etc.). Figure 10 shows the regionally integrated contributions of different terms to density changes as a function of lag time in decades, which is based on regression fields of individual terms onto the MOI. For positive values of the lag, density lags MOI, for negative lag density anomalies lead the MOI. The North Atlantic is divided into three regions: the northern Tropics, the North Atlantic Current (NAC) region, and the subpolar area.

Density increase due to salinity peaks in the subtropics about 7 decades before an MOI maximum (Fig. 10a). Farther north in the NAC region salinity increase peaks at 4–5 decades, and in the subpolar region, which spans the area where most anomalous downwelling occurs (cf. Figs. 1d, 7b), it peaks at leads of 2–3 decades. Confirming what we found previously (cf. Fig. 8c), the temperature effect on density in the sinking area is rather weak compared to salinity, and, if anything, tends to follow MOI rather than to lead it: it makes the subpolar North Atlantic warmer (i.e., more buoyant) before the MOI reaches a maximum, and denser when the THC is weakening (Fig. 1a, dashed green curve), acting as a negative feedback.

The dominant balance in the tropical salinity budget is between surface flux and advection (green and red curves, Fig. 10b). Surface flux dominates at a lead of 6 decades, and tends to increase salinity. Advection partially removes this anomaly with a tendency to freshen the Tropics. In the two more northerly subregions advection dominates the net rate of change, but here it makes the water more saline (Figs. 10c,d). So advection is seen to remove salt from low to high latitudes in the decades that precede a strong THC. Other feedbacks, for example, associated with sea ice formation or diffusion of salt, are clearly of less importance. Once the THC is growing stronger, the effect of surface freshwater flux in the Tropics starts to reverse, and causes freshening at a lead of 2 decades. Again it is advection that, at positive lags, communicates the freshening signal to NAC and sinking regions. After the MOI has peaked at lag zero, the reverse cycle starts, with strongest freshening tendency in the subpolar sinking region at lag = 2 decades. The freshwater budget of the Arctic Ocean (not shown) resembles that of the subpolar regions as being “passive”: salinity change is dominated by advection, and varies in phase with the THC anomalies rather than lead it.

d. Tracer propagation from low to high latitudes

In the previous section it was found that advection plays a leading role in transferring salinity anomalies from low to high latitudes. This reinforces the notion of northward propagation of salinity anomalies obtained from Fig. 9. Temporal and spatial scales are, however, inconsistent with those of a salinity anomaly simply being advected by surface boundary currents, which would take a few years to travel the distance from the Tropics to the subpolar gyre. Recirculation and diapycnal transfer will dilute a tracer signal carried by low-latitude surface waters, and will lengthen the time before salinity can build up at high latitudes.

To investigate how signals that originate in the tropical Atlantic propagate northward, a passive tracer-release experiment was carried out in the same model. It was run for 100 yr, with tracer value in the tropical Atlantic (between 0°–15°N) set uniformly to one at the surface only. Decadal mean tracer concentrations, averaged over the top 800 m are shown in Fig. 11. It takes 5 to 6 decades before concentrations in the Greenland and Norwegian Seas exceed 0.4. Largest concentrations occur off Bermuda, where the model forms its subtropical-mode water (Banks et al. 2000). It is in this area that most of the tracer is transferred downward, and fills the upper 800 m of the water column. In contrast, there are low tracer concentrations between 0°–15°N, that is, in the region where the tracer actually enters the ocean via the surface. The low ventilation rate here is consistent with strong stratification and shallow mixed layer depth. There is no clear crossing by concentration maxima of the NAC. Instead, high isopleths run parallel to the southern boundary of the NAC and bypass it at the far eastern boundary of the basin (Figs. 11d–f). This may be because southward advection by the subpolar gyre of tracer minima from the north will subdue tracer concentrations in the northwest Atlantic. Also, the absence of mixing by eddies in a model of this resolution may underestimate the degree of meridional mixing across the NAC (Figueroa 1994).

e. Tropical air–sea interaction

From the previous two sections, it is found that surface freshwater flux anomalies in the tropical North Atlantic are key in initiating a typical centennial THC fluctuation (Fig. 10b). This relationship can be further summarized in a time–latitude Hovmöller plot, Fig. 12. Salinity anomalies (colors) in the Tropics and subtropics tend to covary with tropical surface freshwater flux (left curve). This suggests a possible large-scale air–sea interaction over the tropical Atlantic Ocean and an important role of the ITCZ.

It is well known that the tropical atmosphere is sensitive to SST anomalies (Webster 1981). Figure 13 shows the regressions of SST and surface wind stress onto MOI, when these fields are leading the MOI by 6 decades, that is, about half the period of the centennial fluctuation. At this phase, north of the equator are cold SST anomalies, while in the South Atlantic there are warm SST anomalies. This pattern of SST anomalies is consistent with the THC being in a weak phase. Associated with this SST anomaly pattern is a strengthening of the Northeast Trades, and a weakening of the Southeast Trades. The anomalous surface wind stress pattern has an anomalous southward cross-equatorial component. This suggests a change in the Hadley circulation, in particular a southward shift of the ITCZ. In the model there is a strong correspondence between cross-equatorial SST contrast, and meridional excursions of the ITCZ over the tropical Atlantic, as inferred from various proxies, for example, net surface freshwater flux north of the equator, Fig. 14: if the SST contrast is large, so is the net freshwater flux into the tropical North Atlantic. This is the result of anomalies in precipitation and runoff, but not of evaporation.

Moreover, there is a direct connection between centennial fluctuations of the THC and changes to the Hadley circulation over the tropical Atlantic (Fig. 15). When the THC is strong there is stronger than average rising north of the time-mean position of the ITCZ, and weaker than usual at its southern flank, and vice versa when the THC is weak. The freshwater transport that the ocean carries northward across 15°N is larger when the ITCZ is shifted into a more northerly position (i.e., when the THC is strong), and weaker when it is in a more southerly position. When the THC is weak, this change in ITCZ position and anomalously low ocean freshwater transport create a positive salinity anomaly. When it reaches the North Atlantic, it increases the strength of the THC, as seen in Fig. 12. A strong overturning enhances the cross-equatorial SST contrast (Fig. 5), which moves the ITCZ southward, and closes the cycle.

It is interesting to see if there is a link between the magnitude of the cross-equatorial SST gradient and flux anomalies on the one hand, and the response by the THC 6 decades later on the other hand. In Fig. 14 we have marked the data points that occur 6 decades before the 20 decades with the strongest MOI (filled circles) and the weakest MOI (filled triangles). The THC tends to be weak 6 decades after the freshwater flux and SST contrast were large, and strong 6 decades after the freshwater flux and SST contrast were low. However, about 25% of the 20 strongest and weakest decades end up in the “wrong” quadrant (e.g., a strong THC follows 6 decades after a strong SST gradient/strong NH precipitation). It is important to realize that the scatterplot shows unfiltered data that contain decadal “noise,” unrelated to THC variability on a centennial time scale. It does suggest, though, only limited scope for THC predictability, at least with these predictors.

Anomalous freshwater export from the tropical Atlantic to the tropical Pacific via the atmosphere has been shown as a possible process by which the THC may stabilize under increased greenhouse gas forcing (Latif et al. 2000; Thorpe et al. 2001). Schmittner et al. (2000) have shown that, under constant CO2 levels, a similar effect can be caused by fluctuations in frequency or amplitude of El Niño events. Comparison between time series of Niño-3.4 SSTs, MOI, and tropical freshwater flux in the Atlantic (not shown) indicates some dependency between ENSO and THC variability at periods of 10–20 yr, but not at the centennial time scale discussed in this paper.

Whereas the importance of temperature for the centennial THC variability is rather small in the subpolar sinking regions (as discussed in sections 5a,b), it is through the large influence of SST anomalies on the tropical atmosphere that temperature changes have a crucial role in the model's centennial THC variability.

6. Conclusions and discussion

We have analyzed THC variability in the HadCM3 control simulation using 1600 yr of model data after 400 yr of spinup. A wavelet analysis has revealed strong power peaks at interannual and centennial time scales, with the centennial mode dominant. In the Atlantic, interannual variability of the THC reflects the influence of ENSO and the NAO. The centennial mode has a basin-scale structure and interhemispheric impacts on surface climate.

Process analysis and the investigation of mechanism for the centennial THC oscillation has been a major part of this paper. It has revealed a large-scale air–sea interaction in the Atlantic sector with a centennial time scale. The mechanism can be summarized in the schematic picture of Fig. 16. Anomalous northward ocean heat transport associated with a strong phase of the Atlantic THC generates a cross-equatorial SST gradient. This causes the ITCZ to move to a more northerly position with increased strength. The extra rainfall resulting from the anomalous ITCZ imposes a freshwater flux and produces a salinity anomaly in the tropical North Atlantic. Such sustained salinity anomalies slowly propagate towards the subpolar North Atlantic at a lag of 5–6 decades. The accumulated low-salinity water lowers upper-ocean density, which causes the THC to slow down. The oscillation then enters the opposite phase.

The time scale of the centennial oscillation depends on the process of salinity anomalies moving from the Tropics to high latitudes where it affects the THC. This time scale was found to be consistent with propagation of an idealized tracer from low to high latitudes in the model. It is, however, different from the advective time scales of the surface boundary currents. A similarly slow meridional tracer propagation in HadCM3 was described by Thorpe et al. (2001). They noticed that it took salinity anomalies, created in the tropical Atlantic under anthropogenic climate forcing, several decades to reach subpolar regions. Marotzke and Klinger (2000) also found that adjustment of the THC can occur at a slow advective time scale of several decades.

The dominant time scale of low-frequency internal THC variability in HadCM3 is longer than that in other climate models (Delworth et al. 1993; Timmermann et al. 1998). First of all, propagation of tracers (such as salinity) depends on resolution, numerical schemes, and will differ between models. In HadCM3 salinity anomalies need to cover a long distance from the regions where they are formed (low latitudes) to where they can affect the THC (high-latitude sinking region). In the other climate models, salinity anomalies are created at subpolar and midlatitudes, that is, closer to the sinking regions. The precise response of the atmosphere to a given SST anomaly probably varies between climate models. This will lead to differences between the models in where the atmosphere responds most strongly to THC-driven SST anomalies, and how it feeds back onto the ocean.

Indirect validation of the model's centennial THC variability using global surface air temperature response shows that amplitude and pattern appear broadly consistent with what is seen in the instrumental record (KNI). Various proxy data from paleo records support the notion of centennial variability in the North Atlantic climate (Anklin et al. 1998; Proctor et al. 2002), and surface salinity in the Caribbean (Nyberg et al. 2002) during the last 1000–2000 yr. By synchronizing high-resolution paleo records of the tropical Atlantic during glacial climate (e.g., Ruehlemann et al. 1999; Kim and Schneider 2003) and those of the North Atlantic circulation (as done by Schmidt et al. 2004 for changes at the millenial time scale), it may be possible to verify if the mechanisms described in this paper have a resemblance in the real climate system. If they do, then our results suggest that the source of significant change to the THC may not reside solely at high latitudes, but that changes in the tropical ocean–atmosphere system also have the potential of affecting the THC on this centennial time scale.

Acknowledgments

We thank Richard Wood, Jonathan Gregory, and two reviewers for suggestions and discussions, and Helene Banks, Steven Spall, and Rachel Stratton for their technical advice. This work was funded by the Department for the Environment, Food and Rural Affairs, under the Climate Prediction Program PECD/ 7/12/37.

REFERENCES

  • Anklin, M., , R. Bales, , E. Mosley-Thompson, , and K. Steffen, 1998: Annual accumulation at two sites in northwest Greenland during recent centuries. J. Geophys. Res, 103 (D22) 2877528783.

    • Search Google Scholar
    • Export Citation
  • Banks, H. T., , R. A. Wood, , J. M. Gregory, , T. C. Johns, , and G. S. Jones, 2000: Are observed decadal changes in intermediate water masses a signature of anthropogenic climate change? Geophys. Res. Lett, 27 , 29612964.

    • Search Google Scholar
    • Export Citation
  • Broecker, W. S., 1991: The great ocean conveyor. Oceanography, 4 , 7989.

  • Collins, M., , and B. Sinha, 2003: Predictability of decadal variations in the thermohaline circulation and climate. Geophys. Res. Lett.,30, 1413, doi:10.1029/2002GL016776.

    • Search Google Scholar
    • Export Citation
  • Collins, M., , S. F. B. Tett, , and C. Cooper, 2001: The internal climate variability of HadCM3, a version of the Hadley Centre coupled model without flux adjustments. Climate Dyn, 17 , 6181.

    • Search Google Scholar
    • Export Citation
  • Cox, P. M., , R. A. Betts, , C. B. Bunton, , R. L. H. Essery, , P. R. Rowntree, , and J. Smith, 1999: The impact of new land surface physics on the GCM simulation of climate and climate sensitivity. Climate Dyn, 15 , 183203.

    • Search Google Scholar
    • Export Citation
  • Delworth, T. L., , and K. Dixon, 2000: Implications of the recent trend in the Arctic/North Atlantic oscillation for the North Atlantic thermohaline circulation. J. Climate, 13 , 37213727.

    • Search Google Scholar
    • Export Citation
  • Delworth, T. L., , and R. J. Greatbatch, 2000: Multidecadal thermohaline circulation variability driven by atmospheric flux forcing. J. Climate, 13 , 14811495.

    • 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.

    • Search Google Scholar
    • Export Citation
  • Delworth, T. L., , S. Manabe, , and R. J. Stouffer, 1993: Interdecadal variations of the thermohaline circulation in a coupled ocean–atmosphere model. J. Climate, 6 , 19932011.

    • Search Google Scholar
    • Export Citation
  • Delworth, T. L., , R. J. Stouffer, , K. W. Dixon, , M. J. Spelman, , T. R. Knutson, , A. J. Broccoli, , P. J. Kushner, , and R. T. Wetherald, 2002: Review of simulations of climate variability and change with the GFDL R30 coupled climate model. Climate Dyn.,19, doi:10.1007/ s00382-002-0249-5.

    • 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.

    • Search Google Scholar
    • Export Citation
  • Figueroa, H. A., 1994: Eddy resolution versus eddy diffusion in a double gyre GCM. Part II: Mixing of passive tracers. J. Phys. Oceanogr, 24 , 387402.

    • Search Google Scholar
    • Export Citation
  • Ganachaud, A., , and C. Wunsch, 2000: Improved estimates of global ocean circulation, heat transport and mixing from hydrographic data. Nature, 408 , 453457.

    • Search Google Scholar
    • Export Citation
  • Gill, A. E., 1980: Some simple solutions for heat-induced tropical circulation. Quart. J. Roy. Meteor. Soc, 106 , 447462.

  • Gordon, A. L., 1986: Inter-ocean exchange of thermocline water. J. Geophys. Res, 91 , 50375046.

  • Gordon, C., , C. Cooper, , C. A. Senior, , H. Banks, , J. M. Gregory, , T. C. Johns, , J. F. B. Mitchell, , and R. A. Wood, 2000: The simulation of SST, sea ice extents and ocean heat transports in a version of the Hadley Centre coupled model without flux adjustments. Climate Dyn, 16 , 147168.

    • Search Google Scholar
    • Export Citation
  • Hughes, T., , and A. Weaver, 1994: Multiple equilibria of an asymmetric two-basin model. J. Phys. Oceanogr, 24 , 619637.

  • Hurrell, J., 1995: Decadal trends in the North Atlantic Oscillation: Regional temperatures and precipitation. Science, 269 , 676679.

  • Kim, J-H., , and R. R. Schneider, 2003: Low-latitude control of interhemispheric sea-surface temperature contrast in the tropical Atlantic over the past 21 kyears: The possible role of SE trade winds. Climate Dyn.,21, doi:10.1007/s00382-003-0341-5.

    • Search Google Scholar
    • Export Citation
  • Klinger, B., , and J. Marotzke, 1999: Behavior of double-hemisphere thermohaline flows in a single basin. J. Phys. Oceanogr, 29 , 382399.

    • Search Google Scholar
    • Export Citation
  • Latif, M., , E. Roeckner, , U. Mikolajewicz, , and R. Voss, 2000: Tropical stabilization of the thermohaline circulation in a greenhouse warming simulation. J. Climate, 13 , 18091813.

    • Search Google Scholar
    • Export Citation
  • Mann, M. E., , and J. Park, 1994: Global-scale modes of surface temperature variability on interannual to century timescales. J. Geophys. Res, 99 , 2581925833.

    • Search Google Scholar
    • Export Citation
  • Mann, M. E., , R. S. Bradley, , and M. K. Hughes, 1999: Northern Hemisphere temperatures during the past millennium: Inferences, uncertainties, and limitations. Geophys. Res. Lett, 26 , 759762.

    • Search Google Scholar
    • Export Citation
  • Marotzke, J., 1997: Boundary mixing and the dynamics of three-dimensional thermohaline circulations. J. Phys. Oceanogr, 27 , 17131728.

    • Search Google Scholar
    • Export Citation
  • Marotzke, J., , and B. Klinger, 2000: The dynamics of equatorially asymmetric thermohaline circulation. J. Phys. Oceanogr, 30 , 955970.

  • 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.

    • Search Google Scholar
    • Export Citation
  • Nyberg, J., , B. A. Malmgren, , A. Kuijpers, , and A. Winter, 2002: A centennial-scale variability of tropical North Atlantic surface hydrography during the late Holocene. Palaeogeogr. Palaeoclimatol. Palaeoecol, 183 , 2541.

    • Search Google Scholar
    • Export Citation
  • Pardaens, A. K., , H. T. Banks, , J. M. Gregory, , and P. R. Rowntree, 2003: Freshwater transports in HadCM3. Climate Dyn, 21 , 177195.

  • Parker, D. E., , T. P. Legg, , and C. K. Folland, 1992: A new daily Central England temperature series. Int. J. Climatol, 12 , 317342.

  • Pope, V. D., , M. L. Gallani, , P. R. Rowntree, , and R. A. Stratton, 2000: The impact of new physical parametrizations in the Hadley Centre climate model—HadAM3. Climate Dyn, 16 , 123146.

    • Search Google Scholar
    • Export Citation
  • Proctor, C. J., , A. Baker, , and W. L. Barnes, 2002: A three thousand year record of North Atlantic climate. Climate Dyn.,19, doi:10.1007/s00382-002-0236-x.

    • Search Google Scholar
    • Export Citation
  • Rahmstorf, S., 1996: On the freshwater forcing and transport of the Atlantic thermohaline circulation. Climate Dyn, 12 , 799811.

  • Rahmstorf, S., 2003: The current climate. Nature, 421 , 699.

  • Ruehlemann, C., , S. Mulitza, , P. J. Mueller, , G. Wefer, , and R. Zahn, 1999: Warming of the tropical Atlantic Ocean and slowdown of thermohaline circulation during the last deglaciation. Nature, 402 , 511514.

    • Search Google Scholar
    • Export Citation
  • Sausen, R., , K. Barthel, , and K. Hasselmann, 1988: Coupled ocean– atmosphere models with flux corrections. Climate Dyn, 2 , 154163.

  • Schmidt, M. W., , H. J. Spero, , and D. W. Lea, 2004: Links between salinity variation in the Caribbean and North Atlantic circulation. Nature, 428 , 160163.

    • Search Google Scholar
    • Export Citation
  • Schmittner, A., , C. Appenzeller, , and T. Stocker, 2000: Enhanced Atlantic freshwater export during El Niño. Geophys. Res. Lett, 27 , 11631166.

    • Search Google Scholar
    • Export Citation
  • Spall, M. A., , and R. S. Pickart, 2001: Where does dense water sink? A subpolar gyre example. J. Phys. Oceanogr, 31 , 810826.

  • Thorpe, R. B., , J. M. Gregory, , T. C. Johns, , R. A. Wood, , and J. F. B. Mitchell, 2001: Mechanisms determining the Atlantic thermohaline circulation response to greenhouse gas forcing in a non-flux-adjusted coupled climate model. J. Climate, 14 , 31023116.

    • Search Google Scholar
    • Export Citation
  • Timmermann, A., , M. Latif, , R. Voss, , and A. Grötzner, 1998: Northern Hemispheric interdecadal variability: A coupled air–sea mode. J. Climate, 11 , 19061931.

    • Search Google Scholar
    • Export Citation
  • Torrence, C., , and G. P. Compo, 1998: A practical guide to wavelet analysis. Bull. Amer. Meteor. Soc, 79 , 6178.

  • Vellinga, M., , and H. Banks, 2002: Climatological mean heat transport and its variability in the coupled climate model HadCM3. International WOCE Newsletter, No. 42, WOCE International Project Office, Southampton, United Kingdom, 6–8.

    • Search Google Scholar
    • Export Citation
  • Vellinga, M., , R. A. Wood, , and J. M. Gregory, 2002: Processes governing the recovery of a perturbed thermohaline circulation in HadCM3. J. Climate, 15 , 764780.

    • Search Google Scholar
    • Export Citation
  • von Storch, J. S., , P. Müller, , R. J. Stouffer, , R. Voss, , and S. F. B. Tett, 2000: Variability of deep-ocean mass transport: Spectral shapes and spatial scales. J. Climate, 13 , 19161935.

    • Search Google Scholar
    • Export Citation
  • Webster, P. J., 1981: Mechanisms determining the atmospheric response to sea surface temperature anomalies. J. Atmos. Sci, 38 , 554571.

    • Search Google Scholar
    • Export Citation
  • Wilks, D. S., 1997: Resampling hypothesis tests for autocorrelated fields. J. Climate, 10 , 6582.

  • Wu, P., , and C. Gordon, 2002: Oceanic influence on North Atlantic climate variability. J. Climate, 15 , 19111925.

  • Wunsch, C., 2002: What is the thermohaline circulation? Science, 298 , 11791180.

Fig. 1.
Fig. 1.

Meridional overturning streamfunction in the Atlantic Ocean in HadCM3. Positive values (solid) indicate clockwise circulation, negative values (dashed) indicate counterclockwise circulation. (a) Average over 1600 yr. Units (Sv), interval of 2 Sv. (b) First EOF of annual-mean streamfunction. Linear correlation with annual-mean SST anomalies in the Niño-3.4 area is 0.40. (c) Second EOF of annual-mean streamfunction. Linear correlation with the NAO index (defined as the normalized, annual-mean sea level pressure difference between the Azores and Iceland) is −0.64. (d) First EOF of decadal-mean streamfunction. Fractional variance of each EOF, and linear correlation between the expansion coefficients of each EOF and the MOI is shown at the top. EOFs are scaled such that the standard deviation of their respective principal components is 0.1

Citation: Journal of Climate 17, 23; 10.1175/3219.1

Fig. 2.
Fig. 2.

Composite analysis of upper-ocean currents (averaged over 0–800 m) of the decades with the 20 strongest minus decades with the 20 weakest values of the MOI. For clarity, arrows are only shown on every second grid point, and if larger than 0.2 cm s−1

Citation: Journal of Climate 17, 23; 10.1175/3219.1

Fig. 3.
Fig. 3.

Detrended time series of maximum value of the annual-mean meridional streamfunction in the Atlantic at 45°N

Citation: Journal of Climate 17, 23; 10.1175/3219.1

Fig. 4.
Fig. 4.

Wavelet power spectrum of time series of Fig. 3 using Morlet wavelet (gray shading, units Sv2). Model time (yr) runs along the horizontal axis, and period (yr) runs along vertical axis. The solid contour encloses areas where the wavelet power exceeds the 95% confidence level of an AR-(1) (or red-noise) process. One expects 5% of the points to do so by chance; here the contour encloses about 10% of the points. Dashed curve indicates where edge effects become important and wavelet power is underestimated. More details on wavelet analysis are given by Torrence and Compo (1998)

Citation: Journal of Climate 17, 23; 10.1175/3219.1

Fig. 5.
Fig. 5.

Regression of decadally averaged SST anomalies onto MOI (units 0.01°C Sv−1, contours values at ±50, 25, 10, 5, 2.5, and 0). Positive values have solid contours, negative values have dash–dotted contours, and zero contours are heavy solid lines. Shading indicates where regression is significantly different from zero at 95% level

Citation: Journal of Climate 17, 23; 10.1175/3219.1

Fig. 6.
Fig. 6.

Composite atmospheric anomaly patterns induced by THC fluctuations, obtained by averaging anomalies over 20 decades when the THC is strongest. Colors indicate where the null hypothesis of equal means is rejected at the 10% level (see text for details). (a) Surface air temperature anomalies in °C, with positive contours solid, negative contours dashed, and zero contour heavy. (b) As in (a) but for MSLP; units hPa. (c) As in (a) but for precipitation; units cm yr−1

Citation: Journal of Climate 17, 23; 10.1175/3219.1

Fig. 7.
Fig. 7.

(a) Regression of vertically averaged density (0–800 m) onto MOI (units: 10−2 kg m−3 Sv−1). Shading indicates where regression is different from zero at the 95% confidence level. (b) Regression of vertical velocity at 700 m onto MOI (units: 10−5 cm s−1 Sv−1). Shading indicates where values are less than −2 × 10−5 cm s−1 Sv−1

Citation: Journal of Climate 17, 23; 10.1175/3219.1

Fig. 8.
Fig. 8.

(a) Time series of decadal MOI (light) and zonal mean density anomalies at 60°N (heavy), averaged over the top 800 m of the water column. Data was low-pass filtered with a half-period of about 5 decades, and then normalized by standard deviation (shown in legend) to enable a direct comparison. Correlation between the two time series is shown in bottom-left corner. (b) As in (a) but for salinity-driven contributions to density Sσ. (c) As in (a) but for temperature-driven contributions to density Θσ

Citation: Journal of Climate 17, 23; 10.1175/3219.1

Fig. 9.
Fig. 9.

Linear correlation between salinity contributions to density and zonal mean density at 60°N, both averaged over 0–800 m. Lead time of salinity (decades) over density at 60°N is shown at top. Contour interval is 0.1. Values larger than 0.4 are dotted, smaller than −0.4 hatched. To highlight the centennial fluctuation data were low-pass filtered with a half-period of about 5 decades

Citation: Journal of Climate 17, 23; 10.1175/3219.1

Fig. 10.
Fig. 10.

(a) Salinity-driven density tendencies, ∂Sσ/∂t (solid), and temperature-driven density tendencies, ∂Θσ/∂t (dashed), regressed onto MOI and averaged between 0–800 m and over three regions of the Atlantic Ocean: (sub) Tropics (0°–35°N, black), North Atlantic Current (35°–48°N, red), and subpolar (48°–65°N, green). Horizontal axis shows lag in decades of density relative to MOI; at negative lags density leads MOI. (b) Area-integrated regression fields of all processes modifying salinity in subtropical domain. Colors and labeling identify individual processes explained in legend. The heavy black unlabeled curve is the net tendency (σ) as in (a). (c) As in (b), but for the North Atlantic Current region. (d) As in (b) but for the subpolar region

Citation: Journal of Climate 17, 23; 10.1175/3219.1

Fig. 11.
Fig. 11.

Evolution (from 1 to 6 decades) of passive tracer concentration, averaged over 0–800 m. Contour interval is 0.1; concentrations larger than 0.4 are hatched. Tracer values may be interpreted as the fraction of water in the upper 800 m that has been at the surface between 0° and 15°N at any moment in the experiment

Citation: Journal of Climate 17, 23; 10.1175/3219.1

Fig. 12.
Fig. 12.

Hovmöller plot of zonal mean salinity anomalies, averaged over 0–800 m, expressed as potential density anomalies; compare with Eq. (1) (colors, units kg m−3). Time runs down along vertical axis, latitude along horizontal axis. The solid curve centered at 50°N is the MOI time series, and the dotted curve centered at 10°N is the surface freshwater flux anomaly (evaporation–precipitation–runoff, labeled SFC), integrated over 0°–15°N. Positive anomalies are northward deflections, with scale indicated in (a). Positive anomaly of SFC is out of the ocean, making it saltier

Citation: Journal of Climate 17, 23; 10.1175/3219.1

Fig. 13.
Fig. 13.

Regression of SST (shading, in 10−2 °C Sv−1), and surface wind stress (arrows, in N m−2 Sv−1) onto lagged MOI: SST and winds lead MOI by 6 decades

Citation: Journal of Climate 17, 23; 10.1175/3219.1

Fig. 14.
Fig. 14.

Net surface freshwater flux into the tropical North Atlantic (between 0° and 15°N), vs cross-equatorial Atlantic SST gradient (0°– 15°N minus 0°–15°S). Solid triangles denote decades that are followed 60 yr later by one of the 20 decades with weakest THC. Solid circles denote the decades preceding the 20 decades with strongest THC by 60 yr

Citation: Journal of Climate 17, 23; 10.1175/3219.1

Fig. 15.
Fig. 15.

Meridional cross section over the tropical Atlantic of the vertical motion ω in the troposphere (negative values indicate upward motion). The gray shading shows the time-mean field (in units of 10−3 Pa s−1; zero contour is highlighted by heavy solid line). The light contours show the regression of ω against MOI (units of 10−4 Pa s−1 Sv−1; negative contours are dashed)

Citation: Journal of Climate 17, 23; 10.1175/3219.1

Fig. 16.
Fig. 16.

Schematic of mechanism responsible for centennial THC fluctuation in HadCM3. When the THC is (left) strong ITCZ shifts northward, in response to enhanced SST gradient across equator. Fresh anomalies in the upper-ocean propagate northward and weaken the overturning. This results in the (right) weak phase

Citation: Journal of Climate 17, 23; 10.1175/3219.1

Save