The thermohaline circulation is driven by the flux of buoyancy through the ocean surface. This buoyancy flux can be broken down into two competing components - heat and freshwater (P-E) fluxes. High latitude cooling and low latitude heating tend to drive a poleward flowing surface flow, high latitude sinking and a deep equatorward return flow, whereas high latitude excess precipitation over evaporation and low latitude excess evaporation over precipitation (except in a relatively narrow belt at the Intertropical Convergence Zone) tend to brake this thermally driven overturning. The existence of multiple equilibria and the stability and variability properties of the thermohaline circulation depend fundamentally on the competing properties of temperature and salinity in the net surface buoyancy forcing of the ocean and in particular, on the fundamental difference in the coupling of temperature and salinity between the ocean and the atmosphere. Variations in the stability or variability properties of the oceans' thermohaline circulation and hence its associated poleward transport of heat would have significant impact on both local and global climate.
The atmosphere and ocean exchange momentum, heat, water vapor, carbon dioxide and other trace substances on all space and time scales. The exchanges depend on, and in turn, alter, the climates of the atmosphere and ocean. Projections of future climates must be carried out with a coupled atmosphere-ocean general circulation model that has sufficient intricacies to be of challenge to scientists and sufficient geographic resolution to be of interest to governments and decision-makers in both the government and private sectors.
In the past forty years, there have been significant advances in the development, validation and application of atmospheric and oceanic general circulation models (AGCM and OGCM). However, atmospheric and oceanic models have evolved separately, so that atmospheric models have specified or highly parameterized oceans, and vice versa. By its nature, a coupled model will spotlight and, in general, reveal gaps or inadequacies in physics previously held fixed. For example, climate drift is seen to occur when an AGCM and an OGCM are coupled into a single interactive system (Manabe and Stouffer, 1988; Washington and Meehl, 1988). In principle, if an AOGCM faithfully reproduces the physics and dynamics of the atmospheric and oceanic subsystems and their interactions, the drift may be a manifestation of nonlinear interactions in the system (Lorenz, 1968, 1982). In practice, however, the drift is most likely caused by errors or inaccuracies in the calculated air-sea fluxes of momentum, heat and P-E. Indeed, when AGCMs and OGCMs are run separately for present-day conditions, air-sea fluxes of heat and P-E calculated directly from AGCMs are different from those implied by ocean models (e.g. Zaucker et al., 1994), and this incompatibility in a coupled model may cause a climate drift. To prevent climate drift, a flux correction (or adjustment) method is used in which artificial, nonphysical terms are added to the conservation equations of heat, water and momentum (Sausen et al., 1988; Manabe and Stouffer, 1988; Manabe et al., 1991). Climate variability is a manifestation of the internal workings of the climate system. Its study therefore provides not only insights into the interactions of the coupled system, but also a dynamic test of the processes represented in the coupled model. El Niño-Southern Oscillation (ENSO) is a familiar manifestation of natural year-to-year variability of the coupled atmosphere-ocean system. To date, success in simulating, and predicting, different aspects of the ENSO phenomena using an atmospheric model with prescribed ENSO sea surface temperatures (SST), ocean models with specified ENSO surface forcing, and AOGCMs (e.g. Lau et al., 1992; Philander et al., 1992; Schneider and Kinter, 1994) are encouraging the notion that physics of interannual variation may be understood, at least to first order, and may be realistically captured in general circulation models. Little is known about atmospheric variability on timescales longer than a season, largely because of the limited observational record. It is this longer-term variability which must be quantified before one can interpret decade-long trends in any climate record or simulation.
Recent ocean modelling studies (eg. Weaver et al., 1991; Weaver et al., 1993) have indicated that internal variability of the thermohaline circulation may exist on the decadal-century timescale. The most important component of the system which ultimately determines the model's response is the strength of the P-E forcing versus the thermal forcing. This has important implications in the interpretation and modelling of both palaeo and present climate and climate change. If the results of the idealized ocean models hold for comprehensive climate models, a relatively small error in the P-E exchange between ocean and atmosphere may produce dramatic changes in the thermohaline circulation, with impact on the global climate.
The aforementioned ocean modelling studies have offered the tempting speculation that a major source of decadal to century climatic variability resides in the intrinsic dynamics of the ocean's thermohaline circulation. Such a hypothesis must be carefully tested and quantified using more complicated and realistic coupled atmosphere-ocean-ice models (as in Delworth et al. 1993). Quantification of climate variability also defines the "noise" level above which anthropogenic climate change may be detected. Another question is to what extent the variability, or the magnitude of the "noise" level, is dependent on the equilibrium state or evolves as climate changes (eg. as in Weaver and Hughes, 1994; Tang and Weaver, 1994).
It is evident that the development of a quantitative understanding of decadal-century climate variability is in its very early stages. The modelling and prediction of decadal-century scale climate variability needs to be refined and expanded through a hierarchy of both coupled and uncoupled ice/ocean/atmosphere models. Model simulations must be carefully analysed and compared to both existing and future observations. The central goal of this proposal is to tackle this challenging problem.
Together with my students and postdocs, I have submitted or have had a number of manuscripts appear over the last year. Briefly, Weaver et al. (1993) was a detailed sensitivity analysis of the stability and variability properties of the thermohaline circulation. We showed how thermohaline variability on fundamental timescales (diffusive, meridional overturning, horizontal advective) could be excited depending on the relative importance of thermal vs P-E vs wind forcing. T. Hughes and I also wrote a manuscript concerning the existence of multiple equilibria in an idealized global ocean model (Hughes & Weaver, 1994). Under present day forcing we showed that there was a clear preference for the "conveyor belt" equilibrium. Furthermore, our model results suggest that there are three possible modes for the North Atlantic conveyor: 1) the present day "normal" state; 2) a state with no North Atlantic overturning (colder); 3) a state with enhanced thermohaline overturning (warmer).
In another paper (Weaver & Hughes 1994) I further showed how transitions between these modes could be excited through stochastic atmospheric forcing. This theory provides a possible explanation for the recent, and highly publicized, Greenland ice core data for the Eemian (GRIP 1993). That is, I specifically addressed the question as to why the stability of our present Holocene is different from that of the Eemian interglacial, and quantitatively showed that the difference was most likely linked to an enhanced hydrological cycle associated with the warmer mean climate of the Eemian. Recent coupled atmosphere-ocean simulations (Manabe & Stouffer 1993) of the climatic response to increasing atmospheric CO2, for which I was asked to write a Nature News & Views (Weaver 1993), have also noted that the hydrological cycle intensifies as the climate warms. If the variability found in the GRIP ice core data is corroborated, then the Eemian may offer us a glimpse as to the type of rapid climate variability which we might expect in a future climate warmed through anthropogenic greenhouse gas emissions.
An interesting phenomenon which I observed in writing the News & Views piece was that for different equilibria obtained under normal, 2xCO2, 4xCO2, 8xCO2 forcing in coupled GCMs (and indeed in the uncoupled Canadian Climate Centre atmospheric GCM - Boer et al. 1992; McFarlane et al. 1992), the total planetary heat transport was fairly constant (in a global warming or cooling scenario there was net heat loss or gain by the planetary system but at equilibrium, the radiation balance at the top of the atmosphere was similar). This phenomenon was exploited in the coupled atmosphere-ocean box model developed by Tang & Weaver (1994). The results of this simple coupled model suggest, as did the uncoupled ocean experiments of Weaver & Hughes (1994), that if the earth were to warm by a few degrees then we should expect rapid climate variability as perhaps seen in the last interglacial period.
Two research projects were also recently completed. In the first of these (Reynaud et al. 1994) we analysed archived data from the Labrador Sea region of the North Atlantic. Several diagnostic models were used to study the climatological mean summer circulation in the area. We have recently extended this to examine the circulation during different decades over the past century. In the second project (Weaver et. al 1994) the "observed" North Atlantic P-E field, Levitus SST and Hellerman and Rosenstein wind stress data were used to drive a "realistic" geometry North Atlantic OGCM. In these experiments we found 20 year self-sustained variability of the thermohaline circulation and hence the poleward heat transport which was driven by changes in Labrador Sea convection. This is an important result as there have been many observations of decadal variability in and around the North Atlantic.
This will be a four year project involving a systematic procedure for model development. We have started with the barotropic vorticity equation which allows for full topography and specified stratification (Myers and Weaver, 1994). More recently we have added time-dependence to the nonlinear model and it has been extended to spherical coordinates and global geometry. Upon development of the three-dimensional code, comparisons will be done with the Bryan-Cox OGCM, initially using idealized geometry and proceeding to the global domain (eg. project 2). The model will eventually be used for climate simulations (P. Myers, a PhD student, is working on this project).
The model has already proved to be extremely useful in diagnosing the transport of the North Atlantic (Myers et al. 1994). In this manuscript we suggest that the reason why 3-D ocean models do not get the Gulf Stream to separate at the correct latitude is due to a poor representation of the density field in the upper ocean. Research is also progressing well into the development of semi-Lagrangian advection algorithms appropriate for ocean models (eg. Das and Weaver 1994).
We shall also use the finite element global model to investigate the effects of Bering and Indonesian throughflow, as well as Mediterranean outflow, on the global thermohaline circulation. Furthermore, we wish to address the debate as to where the return flow of North Atlantic Deep Water occurs. Coarse resolution models (e.g. Hirst & Godfrey 1993) suggest that most of the return flow happens in the cold water route via the Aghulas Retroflection. Eddy resolving models, which essentially prescribe the deep temperature and salinity structure of the ocean, suggest that most of the return flow happens in the warm water route via eddy generation in the Aghulas Current (Semtner & Chervin 1988, 1992). Observations suggest that some combination of the two routes is appropriate (Broecker 1991).
Before undertaking the sensitivity experiments it is important that the global model successfully reproduces the present day climatology. A major test that we will use is the simulation of ocean Freon distributions. I recently obtained surface boundary condition information, from Dr. R. Weiss at Scripps, which will be injected into the ocean model and the predicted and observed Freon distributions will be compared (D. Robitaille, a PhD student, will work on this project).
This global ocean model will be the main model with which we shall undertake the coupling with the global energy balance model as well as the AGCM. The coupling with the AGCM will be done in collaboration with the Canadian Climate Centre which has recently relocated to the campus of the University of Victoria. The AGCM which will be used was developed by this group (McFarlane et al. 1992) - see Appendix below.
Before doing this a research associate, Dr. B. Tang, and I will couple the Semtner (1976) ice model to a single-hemisphere OGCM to investigate the first-order role of ocean-ice feedbacks on the stability and variability properties of the thermohaline circulation. This analysis will eventually be extended to the global ocean models.
We envision analysing possible mechanisms for interdecadal variability in the ocean/atmosphere system such as: Stronger thermohaline circulation; more evaporation; more high latitude precipitation; slower thermohaline circulation; and again. Furthermore, in the ice/ocean system we should see interdecadal variability as in Yang and Neelin (1993): Stronger thermohaline circulation; more ice melt; weaker thermohaline circulation; more ice growth; and again. It is not clear how these two independent oscillations would interact in the fully coupled system (L. Zhang, a PhD student, and B. Tang, a research associate, will work on this project).
In an ocean model, the time scales are determined by dynamics internal to the model. As mentioned above, what has not been investigated is how the modes of thermohaline circulation are manifested in a coupled model, and, in particular, how the time scales of thermohaline circulation variability are related to the time scales of atmospheric variability. In a coupled system, changes in one part of the ocean can be communicated very rapidly to another part of the ocean via the atmosphere. A central research question is whether there are thermohaline circulation modes on shorter or longer timescales that depend upon the interaction between the ocean and atmosphere, and thus have been heretofore missing in separate ocean GCM experiments.
Similarly, despite the apparent lack of long physical timescales in the atmosphere, do such timescales appear when the atmosphere is coupled to oceanic variability? We will examine whether there are other modes on longer timescales that depend upon the interaction between the ocean and atmosphere, and thus have been heretofore missing in separate ocean GCM and atmosphere GCM experiments. The goal is therefore to identify the modes, and their time scales, in the coupled system and the critical processes for each. The analysis will be important, not just for the interpretation of the recent climate record, but also for elimination of climate drift in coupled models.