Climate in and around the Atlantic Ocean has been observed to vary in broad spatial patterns and on a variety of timescales during the twentieth century. In the northern regions the dominant mode of variability is the North Atlantic Oscillation (NAO, e.g. Hurrell and Van Loon 1997). The NAO involves an oscillation in mass between the subtropics and the high latitudes. In the high index phase the Icelandic Low is anomalously low, the Azores high is anomalously high, the mid-latitude surface westerlies are strong and there is a strong storm track that trends from the United States coast towards the British Isles and Scandinavia. In the low index phase both the Icelandic low and the Azores high are weaker, the westerlies are weaker, and storms tend to move from the United States into the Labrador Sea region while those that do make it across the Atlantic move into Southern Europe and the Mediterranean. The NAO has a coherent signal in sea surface temperature (SST) involving a tripole pattern of almost zonally oriented anomalies with subtropical and high-latitude SSTs varying in phase and mid-latitude SSTs varying out of phase (e.g. Kushnir 1994, Cullen and DeMenocal 1999). Spectral analyses of NAO time series reveal enhanced variance at periodicities of around two years and at some decadal periods (Hurrell and Van Loon 1997). Further, the NAO has revealed some long term trends, most recently in the form of the tendency towards a deeper Icelandic low and stronger Azores high from 1960 to the 1990s (Hurrell 1995).
In the tropical regions the dominant mode of variability involves variations in the cross-equatorial SST gradient and SST anomalies that are off equatorial and encompass the entire subtropical oceans (Nobre and Shukla 1996). When one hemisphere warms anomalous winds tend to blow across the equator into the warmer hemisphere. There has been some debate about whether the SSTs of the subtropical oceans vary out of phase (Houghton and Tourre 1992) with, most recently, Rajagopolan et al. (1998) concluding that the SSTs of the two hemispheres are not related to each other. The remote effects of the El Niño-Southern Oscillation creates off-equatorial SST anomalies in the Atlantic Ocean (e.g. Saravanan and Chang 1999, Giannini et al. 1999). In addition, the subtropical SST anomalies vary strongly on decadal timescales in a way that seems independent of ENSO. The equatorial Atlantic also exhibits a weak equatorial pattern of variability that is akin to the El Niño-Southern Oscillation phenomena in the Pacific Ocean but which is not self-sustained (Zebiak 1993).
Dividing Atlantic Ocean climate variability into tropical and mid-latitude modes may be useful but is not necessarily valid. For example, the NAO is associated with variations of winds and SST in the northern subtropical Atlantic Ocean. Further, Ragajopolan et al. (1999) present statistical evidence that SSTs in the subtropical South Atlantic are associated with variations in the NAO. This connection might work via the impact of South Atlantic SSTs on Amazon rainfall; the latter influencing the NAO via atmospheric teleconnections or changes in the Hadley Cell (Robertson et al. 1999).
Recently, several investigations have concluded that interannual variations of Atlantic SSTs are primarily driven by the atmosphere via changes in surface fluxes. However, the work described here will show that anomalous Ekman flows can also make an important contribution. The concept of flux driven SST variability was first suggested on the basis of analyses of SSTs and marine meteorological data by Cayan (1992a,1992b) and has been supported by modeling studies (Battisti et al. 1995, Delworth and Mehta 1998). In contrast, it has been suggested that the longer timescale variations might involve a more active role for the ocean including changes in ocean heat transport (e.g. Kushnir 1994, Grötzner et al. 1998). Appealing to an active role for the ocean is attractive in that the long timescales associated with ocean dynamics make it easy to explain decadal fluctuations and long periods of persistent oceanic anomalies. It may also be the case that the ocean's role is restricted to the ability of near surface mixing to sequester heat content anomalies from one winter to another below the summer mixed layer (e.g. Battisti et al. 1995). These explanations for low frequency variations invoke atmosphere-ocean coupling and require that the mid-latitude atmosphere be responsive to underlying SST anomalies. The latter has proven elusive to demonstrate. Some models do show a coherent response to North Atlantic SST anomalies (e.g. Grötzner et al. 1998, Ferranti et al. 1994, Rodwell et al. 1999) while others do not (e.g. Pitcher et al. 1988, Lau and Nath 1994; see also Peng et al. 1995 and the review by Kushnir and Held 1996). To date it has not been resolved whether different models respond differently to the same SST anomalies or whether the apparently inconsistent results are explained by differences in the imposed SST anomalies, experimental design, length of integration and so on.
In this paper we will report on efforts to understand the variability of Atlantic Ocean climate from 1958 to 1998, which is the period for which reliable atmospheric data are available from the National Centers for Environmental Prediction (NCEP) Reanalysis. We attempt to model the SST over this period using ocean mixed layer models, in which the ocean heat transport is held at its climatological value, and also with a fully dynamical ocean general circulation model (OGCM). All models are coupled to a simple thermodynamic model of the well mixed atmospheric layer that forms the lower component of the marine convecting boundary layer (Seager et al. 1995). In this manner, the models are forced only by the time-varying wind speed and direction while the SST and the boundary layer temperature and humidity are computed according to balances between the surface fluxes, ocean heat transport (if allowed to vary), advection and eddy transports of heat and moisture in the atmospheric mixed layer, entrainment across the top of the atmospheric mixed layer and radiative cooling. Since, even in the case where the atmosphere is forcing the ocean, the atmospheric temperature and humidity come into balance with the SST, imposing them in the heat flux boundary conditions of an ocean model informs the model what the SST was. While this is commonly done in ocean modeling studies it ensures that the simulated SST will track that observed while making interpretation of that result confusing. In the ocean modeling work reported here we instead attempt to properly model the coupling between the ocean and the atmospheric boundary layers. This is very clearly an improved experimental set up that allows the SST full freedom to evolve (Seager et al. 1988). We will examine the extent to which surface fluxes and ocean heat transport determine the SST variability and, in turn, why these components of the ocean surface heat budget vary. To what extent can variations in Atlantic Ocean climate be understood as the atmosphere forcing the ocean or vice versa?
We begin by using NCEP reanalyses to examine the terms in the thermodynamic energy budget of the lowest level of the atmosphere. This allows us to assess the different roles that changes in wind speed, advection, subsidence and atmospheric eddy transports have in generating the flux anomalies that influence the SST. In Section 3 we present some preliminary calculations in which we use the atmospheric mixed layer model to simulate the observed changes in surface latent and sensible heat flux given the observed SST. The modeled fluxes are in good agreement with those observed so, in Section 4, we couple the AML model to a uniform depth ocean mixed layer in which the ocean heat transport is assumed to remain at its climatological, seasonally varying, values. The coupled AML-OML model is used to simulate the SST from 1958 to 1998 forced by the time-varying NCEP wind speed and direction. This experiment is analyzed and demonstrates that much of the observed SST variability can be explained in terms of surface fluxes without the need to invoke changes in ocean heat transport. We then repeat this calculation using a variable depth ocean mixed layer in order to assess the role of ocean mixing. In Section 5 we model the SST from 1958 to 1998 using the full ocean GCM which allows the mixed layer depths and ocean heat transport to vary. This run is analyzed, in comparison to the AML-OML experiments, to isolate the role of ocean heat transport. Conclusions are offered in section 6.