Climate Modelling Group
School of Earth and Ocean Sciences

Steacie Fellowship Supplement

1 Preamble

The purpose of this proposal is to take advantage of the unique opportunity afforded by my Steacie Fellowship to develop new research into the understanding of past, present and future climate change and climate variability. The Steacie Fellowship Supplement will provide me with an opportunity to initiate new research as well as build upon my present research on the role of the ocean in climate change and climate variability. Specifically, I will be have the opportunity to: (i) address some challenging problems in my established area of research into the mechanisms of decadal-millennial timescale climate variability; (ii) branch out into a new and largely unexplored area of paleoclimatic modelling using fully coupled atmosphere-ocean models; (iii) attack the fundamental problem of boundary layer versus interior mixing in setting up the large-scale thermohaline circulation. Some of this research was proposed in my attached successful NSERC Strategic Grant program application.

During my three years as an NSERC URF at McGill University I discovered and documented a number of mechanisms behind the existence of internal modes of decadal-millennial variability in ocean models. This research has motivated many other researchers to undertake follow-up studies. While my earlier work has withstood the scrutiny of this subsequent research, it has become apparent that there is a pressing need for fully coupled atmosphere-ocean-ice model simulations to determine how the earlier mechanisms are modified when the ocean is allowed to fully interact with the atmosphere and cryosphere.

While at the University of Victoria my research group and I have invested a large amount of effort to develop an atmospheric model suitable for coupling to ocean models, for the purpose of undertaking the long-timescale integrations required to investigate climate variability on the decadal-millennial timescale. To this end we have coupled an ocean general circulation model (OGCM) to a newly-developed energy-moisture balance atmosphere model (EMBM), into which a thermodynamic ice model (TIM) has been incorporated. We have also recently incorporated a parameterization which allows for wind stress feedbacks. The virtue of the atmospheric component of the coupled model is that we do not need to employ flux adjustments to keep the simulation of the present climate stable. Thus, it is ideally suited for both climate and paleoclimate modelling. This fully-coupled model will be released to the climate modelling community along with a user manual by August 1997 via our web site ( We foresee releasing a parallelized version of this code towards the end of 1999.

With regard to the first area of proposed research mentioned above, the developmental phase of my research has been completed and I am now poised to make fundamental advances to our understanding of climate variability on the decadal-millennial timescale. I would like to further our understanding of the mechanisms of climate variability through the use of increasingly more sophisticated coupled models. The coupled energy-moisture balance atmosphere/ocean/ice model represents the simplest form of my proposed coupled modelling studies. I hope to use it to explore simple thermodynamic feedbacks and gain insight into what results we might expect and which experiments we should undertake with the more complicated Geophysical Fluid Dynamics Laboratory (GFDL) coupled model. The GFDL coupled model is more sophisticated than the aforementioned coupled model as it includes full atmospheric dynamics and physics. In addition, the GFDL coupled model will be used to investigate questions concerning the existence of variability in the coupled climate system and how it varies as the mean climatic state changes (i.e., does the decadal-interdecadal climate variability found in the coupled model change as CO2 is increased in the atmosphere?). I also intend to develop, a very simple analytic model explaining the essence of the decadal-interdecadal variability found in ocean models, thereby unifying the competing mechanisms which are beginning to appear in the literature. A detailed understanding of climate variability in the fully-coupled system would not only have a profound influence on the field of climate dynamics, but would also have great societal benefit.

While I have invested a lot of effort into understanding variability and processes within the present climate system, I have only recently become excited about the prospects of unraveling puzzles in the paleoclimatic record. Two important paradoxes exist in the paleoclimatic literature. The first of these concerns how it was possible for the Ordovician climate (~440 million years ago) to support glaciers when the atmosphere had CO2 levels 16 times higher than today. Similarly, during the Cretaceous, atmospheric CO2 levels were 8 times the present yet recent evidence has suggested that it was much cooler than previously thought, with tropical temperatures similar to those of today and polar temperatures hovering around freezing. I would like to invest the effort to try and unravel these profound questions through the use of our newly-developed coupled model. This new research will be undertaken in addition to that which was proposed in my NSERC Strategic grant application.

While the use of models to simulate past climatic events is an important avenue of investigation if one is to have confidence in their application to future climate change, it is also fundamental to understand the processes involved in the real ocean and how they are parameterized in these models. To this end, as part of my third area of proposed research, I would like to investigate how oceanic mixing in boundary layers, versus the interior of the ocean, affects the large-scale thermohaline circulation. More specifically, previous attempts at modelling the ocean's large-scale thermohaline circulation have ignored the observation that diapycnal mixing in the ocean is enhanced near lateral boundaries. I have already invested a great deal of intellectual effort in setting the stage for examining this problem and I am eager to carry it through to its completion. This problem is fundamental to ocean dynamics and the role of the ocean in climate since the oceans thermohaline circulation so critically depends on the parameterization of diapycnal mixing.

The single most important factor which limits my further progress in the above areas is the lack of sufficient computing time (e.g., it takes several months of continuous CPU time to integrate our models to equilibrium). I have recently acquired a four node IBM SP2 parallel processing machine as part of my NSERC Strategic grant (with additional support from IBM Canada and my Strategic Partners). All my machines are currently operating at 100% CPU usage. As I will now have 100% of my time dedicated to research I am hoping to add two more nodes to the SP2 in year 1 of the Supplement, as well as 54 GB of disk space for my SSA Subsystem in year 2 of the award. This will free up more computer time for my own personal research so that I do not need to compete with my students and research associates for resources. With the acquisition of this new hardware I am extremely confident that I will attain all the milestones outlined earlier.

IBM Canada recently expressed a strong interest in supporting my successful NSERC Strategic Grant project by offering to assist me in the acquisition of new parallel processing computer technology (IBM SP2) to undertake coupled atmosphere-ocean-ice modelling simulations. This NSERC Strategic Grant proposal had NOAA, the CICS and the Canadian Centre for Climate Modelling and Analysis as research partners. IBM Canada is a company which is committed to the environment and so is eager to assist in allowing me to meet the goals of the research proposed below. In addition, IBM will benefit from having my research group as one of the first to use their new parallel technology in climate modelling. They will provide personnel support to assist me in parallelizing the code of our coupled climate model. This we will in turn make available to the international climate modelling community via the world wide web. Thus, by investing in the parallelization of our climate modelling code, IBM Canada stands to take advantage of future business opportunities as other researchers who will use our code invest in parallel architecture.

As such, IBM Canada has agreed to offer me 50% discounts on the purchase of all hardware. This is 10% better than the preferred/best customer discount rate of 40% for the disk subsystem and 20% better than the preferred/best customer discount rate of 30% for the IBM SP2. Finally, IBM will provide direct personnel support to assist in the parallelization of our code to take advantage of the new SP2.

I was recently awarded a 5 year NSERC Strategic grant for which I am most grateful. This will be used to support all of the students and research associates working on this research program (with the exception of 1 additional postdoc whom I propose to take on as part of the Steacie Supplement). This proposal therefore seeks only modest support for new personnel and instead focuses on the acquisition of new, state-of-the-art parallel computer technology for my research group. The acquisition of this hardware will put my research group at the forefront in terms of available computer technology and will allow us to continue to make fundamental advances in our understanding of the ocean's role in climate and climate variability.

Finally, the University of Victoria Network and Technical Services Department has agreed to provide a dedicated Ethernet switch for the workstations in lab. This switch will allow communications to occur exclusively between my machines and will cut down on network traffic from other research groups on campus. In addition, the University will renovate a new room next to my present lab to install the new IBM SP2 and SSA Disk Subsystem. This renovation will cost about $20,000 to install appropriate cooling and power for the machines.

The Research Science Plan: NSERC Strategic Project
The Role of the Ocean in Climate Change and Climate Variability

1 Introduction

As I was recently awarded a 5 year NSERC Strategic Grant and as I am largely applying for the acquisition of hardware, I am attaching the Strategic proposal here. The research proposed under the Steacie Supplement is the same as discussed below with the addition of the two paleoclimatic projects and the boundary layer mixing project mentioned briefly above. The milestones given earlier therefore reflect the milestones of my Strategic Grant as well as the three additional milestones for the paleo and mixing projects.

2 Objectives

Over the next five years I hope to continue improving our understanding of the mechanisms of decadal-interdecadal climate variability through the development of increasingly more sophisticated coupled models. A coupled EMBM/TIM/OGCM will represent the simplest form of our coupled modelling studies. We shall use it to explore simple thermodynamic feedbacks and gain insight into what results we might expect and which experiments we should undertake with the more complicated GFDL and Canadian Climate Centre (CCC) coupled models. In addition, the GFDL coupled model will be used to investigate questions concerning the existence of variability in the coupled climate system and how it varies as the mean climatic state changes (i.e, does the decadal-interdecadal climate variability found in the coupled model change as CO2 is increased in the atmosphere?). The GFDL coupled model is less sophisticated and therefore less computationally costly than the CCC coupled model and it is hoped that the insight we gain from it will allow us to streamline the future experiments that will be performed with the more sophisticated CCC coupled model.

3 Introduction

This section has been partitioned into a number of subsections, each of which deals with a particular subcomponent of the project. Throughout the section scientific questions are posed and a brief description of how these questions will be addressed is also presented. More details of the methodology to be used can be found in section 4.

3.1 Decadal-interdecadal variability

There is substantial evidence for decadal climate variability in the air-sea-ice climate system. Recent hypotheses suggest that decadal variability in the Pacific may be either linked to changes in the El Niño/La Niña signal in the equatorial Pacific (Trenberth and Hurrell, 1994) or to midlatitude air-sea instabilities (Latif and Barnett, 1994). It is not clear whether similar mechanisms exist in the Atlantic or whether there is a relationship between the Pacific and Atlantic modes of variability. Below recent evidence is presented to show that decadal variability can exist in uncoupled ocean models in basins where deep water formation occurs.

Under mixed boundary conditions self-sustained internal variability on the decadal-interdecadal timescale can exist in ocean models (e.g., Weaver and Sarachik, 1991; Weaver et al., 1991). Often this variability is linked to the turning on and shutting off of high latitude convection and the subsequent generation and removal of east-west steric height gradients which cause the thermohaline circulation to intensify and weaken over a decadal timescale. Horizontal advection sets the oscillation timescale. Decadal internal oceanic variability still persists or may even be excited when a stochastic component is added to the freshwater forcing (Weaver et al., 1993; Weisse et al., 1994). It has also recently been shown that decadal internal oceanic variability can also exist in ocean models driven only by thermal forcing (Weaver et al., 1994; Greatbatch and Zhang, 1995; Winton, 1996).

The existence of such model results makes it difficult to interpret causes and effects of decadal variability in current coupled climate models which employ flux-adjustments (see Weaver and Hughes, 1996). This follows since if the flux adjustment is large in magnitude, one might expect that the oceanic variability is determined by this structure, with the higher frequency air-sea flux variability providing a stochastic forcing which simply excites it.

In this proposal advances to our understanding of decadal-interdecadal variability will be achieved through the use of coupled models of varying complexity which do not employ flux adjustments. This will expand upon the early uncoupled OGCM results. The use of the EMBM, developed by Fanning and Weaver (1996), will allow for simple thermodynamic feedbacks. In addition, the GFDL coupled model will be used to investigate the dependence of decadal-interdecadal climate variability on the mean climatic state. Furthermore, the use of OGCMs of varying resolution will allow for both an analysis of the effects of horizontal boundary layers (as discussed in Winton, 1996) and the role of eddies in decadal-interdecadal climate variability.

3.2 Centennial-timescale variability

A fundamental period for oceanic model variability also occurs on the overturning timescale (Mikolajewicz and Maier-Reimer, 1990; Winton and Sarachik, 1993; Weaver et al., 1993). The presence of a positive salinity anomaly in the low latitude surface regions tends to slow the meridional overturning slightly, since thermal effects tend to accelerate the thermohaline circulation and haline effects act to brake it. The weakened thermohaline circulation is then more affected by the specified flux on salinity which acts to intensify the positive anomaly at low latitudes and induce a negative salt anomaly at high latitudes. When the low latitude salinity anomaly reaches the high latitudes, convection and an intensified thermohaline circulation ensues. The whole process begins anew when the saline anomaly resurfaces at low latitudes. Thus the oscillation has a slow phase, which is associated with the saline anomaly being at low surface latitudes, and a rapid phase in which the salinity anomaly is at high latitudes or in the deep ocean.

Whether or not such centennial timescale variability exists in the coupled climate system, and the extent to which it is modified by allowing for feedbacks within the coupled system, is still an open question.

3.3 Millennial-timescale variability

While variability on this timescale is beyond the central theme of this proposal, we will be able to address an open and important question regarding the potential existence of the so-called flushes (discussed below). These catastrophic climate swings have previously only been observed in uncoupled OGCMs. If they are found to survive in coupled models then their mechanism serves as a potential explanation for variability found in past, and perhaps future, climates.

Ocean GCMs forced using mixed boundary conditions in which high latitude freshening is strong are often susceptible to polar halocline catastrophes. Associated with the polar halocline catastrophe is a collapsed thermohaline circulation state which is not stable since low latitude diffusion acts to make the deep waters warm and saline with horizontal diffusion acting to homogenize these waters laterally. Eventually, at high latitudes the deep waters become sufficiently warm so that the water column becomes statically unstable and rapid convection sets in. The result is a flush (Marotzke 1989; Weaver and Sarachik, 1991; Wright and Stocker, 1991) in which a violent overturning occurs whereby the ocean loses all the heat it had taken thousands of years to store in a matter of a few decades. At the end of the flush, high latitude freshening eventually suppresses convection and the thermohaline circulation once more collapses. The collapse/flush sequence repeats itself with the timescale between flushing events being set by diffusion.

The existence of flushes is linked to both the relative importance of freshwater flux over thermal forcing, and the strength of the wind forcing compared to the high latitude freshening. As the stratification is usually homogeneous in near-surface layers at high latitudes, the Ekman-driven overturning cells contribute very little to the meridional heat and salt transports. This situation changes once the polar halocline catastrophe has occurred as the surface layer becomes very fresh and the equatorward Ekman transport of fresh water is compensated for by a poleward return transport of more saline water, resulting in a net poleward salt transport. Moreover, the poleward salt transport due to the horizontal subtropical gyre increases substantially during the collapsed phase of the thermohaline overturning.

High-latitude surface freshening may be counteracted by the wind-driven salt transport to make the high-latitude surface waters sufficiently saline, so that deep convection resumes and the thermohaline circulation reestablishes itself. If the surface freshening cannot be compensated for by the wind-driven salt transport, the thermohaline circulation remains in the collapsed state until a flush sets in. When a stochastic term is added to the mean freshwater flux forcing field the frequency of flushing events increases while their intensity decreases, with increasing magnitude of the stochastic term. This follows since the probability of a sufficiently large evaporation anomaly increases so the ocean need not warm as much before a flush occurs.

As mentioned above, these flushes have never been found in coupled climate models. This may be due to the fact that feedbacks in the coupled system prevent their occurrence. On the other hand, no coupled model has ever been run long enough to quantitatively analyze this. Indeed, the coupled OGCM-AGCM of Manabe and Stouffer (1994) reveals that under 4xCO2 forcing, the thermohaline circulation collapsed and remained collapsed for 500 years. This situation with no ventilation of the deep waters cannot exist indefinitely as a stable equilibrium and so deep water must eventually form. Whether deep water will form through a flush or via a more gentle re-establishment is unknown. Through the use of our hierarchy of coupled models we will be able to address this question.

3.4 Surface boundary conditions

Much research has been conducted to develop improvements of the so-called restoring boundary conditions and mixed boundary conditions which have often been used by the ocean modelling community. In restoring boundary conditions, surface temperatures and salinities in an ocean model are relaxed, with a specified timescale, to some climatological values whereas under mixed boundary conditions, a specified flux on salinity is used in conjunction with the restoring condition on temperature.

Sea surface temperature anomalies, regardless of their scale, are strongly damped under restoring boundary conditions. In reality, the damping time should depend on the scale of the anomaly since latent and sensible heat loss are inefficient mechanisms for the removal of large-scale sea surface temperature anomalies. This follows since heat lost over one part of the ocean must be advected by the atmospheric winds away from the anomaly for it to be effectively removed. Therefore, as pointed out by Bretherton (1982), in the limit of a global scale anomaly, radiational damping (long timescale) is the only mechanism for its removal. These ideas have been utilized in the development of a number of improved parameterizations of the thermal surface boundary condition (e.g., Zhang et al., 1993; Mikolajewicz and Maier-Reimer, 1994; Rahmstorf and Willebrand, 1995).

Recent developments have also taken place in the improvement of the boundary condition on salinity in ocean models. Hughes and Weaver (1996) developed a new salinity boundary condition which incorporates the dependence of evaporation on sea surface temperature. The internal variability of the thermohaline circulation documented in sections 3.1-3.3 was found to still exist under this new boundary condition although it was modified slightly.

A cautionary note should be added here regarding the variability found in OGCMs under mixed boundary conditions. It is evident from the new surface boundary condition parameterizations discussed in this subsection that the variability may be either damped or modified when more realistic surface boundary conditions are used. One should therefore view their results with some caution until they are verified with fully-coupled ocean-ice-atmosphere models.

Through the coupling of the OGCM to both an EMBM and a simplified dynamical AGCM we will be able to quantitatively examine whether or not feedbacks in the coupled system dampen, enhance or excite such variability.

3.5 Sub-grid-scale parameterizations

Recent advances have been made in the parameterization of sub-grid scale mixing associated with mesoscale eddies (Gent and McWilliams, 1990 -- hereafter GM; Danabasoglu et al. 1994) in coarse resolution OGCMs. Danabasoglu et al. (1994) have illustrated promising initial results from experiments using their parameterization for mesoscale eddy-induced mixing. When their parameterization was incorporated into a global ocean model the thermocline became sharper, the deep ocean became colder, the meridional transport of heat and salt became greater and the overturning in the North Atlantic expanded meridionally. All of these features are improvements on the results obtained in global models using traditional horizontal/vertical mixing schemes. Furthermore, in the Southern Ocean, the Deacon Cell vanished through an eddy-induced cancellation of the mean flow advection of tracers. The result was a reduction in the heat and salt transport by the ocean across the Antarctic Circumpolar Current.

Robitaille and Weaver (1995) examined three sub-grid scale mixing parameterizations (lateral/vertical; isopycnal/diapycnal; GM) using a global ocean model in an attempt to determine which yielded the best ocean climate. Observations and model CFC-11 distributions, in both the North and South Atlantic, were used in the model validation. While the isopycnal/ diapycnal mixing scheme did improve the deep ocean potential temperature and salinity distributions, when compared to results from the traditional lateral/vertical mixing scheme, the CFC-11 distribution was significantly worse due to too much mixing in the southern ocean. The GM parameterization, on the other hand, significantly improved the deep ocean potential temperature, salinity and CFC-11 distributions when compared to both of the other schemes. The main improvement came from a reduction of CFC-11 uptake in the southern ocean where the "bolus" transport canceled the mean advection of tracers and hence caused the Deacon Cell to disappear. These results suggest that the asymmetric response found in CO2 increase experiments, whereby the climate over the Southern Ocean does not warm as much as in the northern hemisphere, may be an artifact of the particular sub-grid scale mixing schemes used.

These initial results have inspired us to continue to use CFC-11 as a method for validating the climatology of our ocean models and in particular their sub-grid scale parameterizations. We will examine the effects of stability dependent diapycnal mixing which will be added to the GM parameterization. In addition, the role of boundary layer versus interior mixing will be analyzed through the inclusion of spatially varying diapycnal mixing. Furthermore, surface lateral mixing will be added as a parameterization of diabatic processes in the surface mixed layer arising from the presence of surface fluxes and small scale processes acting in series with the mesoscale eddies.

3.6 Eddy heat and salt transport

One of the most fundamental, yet unanswered, questions regarding the ocean's role in climate is whether or not eddies are important in transporting heat and salt poleward. Numerous studies (e.g., Bryan, 1991; Böning and Budich, 1992; Drijfhout, 1994) have suggested that eddies do not play a significant role in the transport of heat and salt poleward. They suggest that the heat and salt transport is dominated by the meridional overturning transport in the North Atlantic.

All of the above works suffer from a major shortcoming since all the experiments were conducted using ocean-only models. In the limit of specified fluxes of heat and freshwater, the oceanic heat and salt transports are necessarily predetermined at equilibrium since the divergence of the transport gives the zonally-averaged flux. Thus, whether or not eddies are resolved will not change the total heat or salt transport at equilibrium. All of the above studies also used restoring boundary conditions on temperature and salinity. While these boundary conditions do not impose an exact constraint on the oceanic poleward heat and salt transports at equilibrium, they do largely determine the thermocline structure and hence may clamp the ability of the ocean models to freely regulate their heat and salt transport. In addition, we will include salinity in our analysis. Previous works utilized buoyancy forcing alone so that eddies are aligned along isopycnals and hence no net heat transport occurs by their presence. In the present proposal, isotherms and isopycnals will no longer coincide and a net heat transport should be expected if eddies propagate across isopycnals.

To quantitatively address this problem we shall use an idealized OGCM of the North Atlantic with resolution ranging from 4o x 4o to 1/4o x 1/4o coupled to an EMBM. The only constraint on the coupled system will then be the incoming solar radiation at the top of the atmosphere. As the ocean model resolution increases, the coupled system will be allowed to adjust, subject to the specified constraint on the incoming solar radiation. By partitioning the heat transport into time mean and time dependent terms we will be able to see whether or not transient eddies are indeed important for the transport of heat and salt poleward. In addition, by examining the time-mean component we will be able to quantify the importance of the barotropic versus overturning transport.

3.7 Global teleconnections

The North Atlantic Ocean is fundamentally linked with the rest of the world's ocean via communication through the Antarctic Circumpolar Current and the Southern Ocean. The extent to which processes occurring in the North Atlantic affect the circulation in other oceans and vice versa is still unknown.

A fundamental question regarding the North Atlantic and its relationship with the rest of the world's oceans concerns the reason why the North Atlantic forms deep water yet the Pacific does not. Warren (1983) singles out the more stable stratification of the North Pacific (where surface waters are on average 32.8 psu and deep waters 34.6-34.7 psu compared to 34.9 and 34.9-35.0 psu in the North Atlantic) as the explanation and identifies a number of causes for this. For example, there is nearly twice as much evaporation over the North Atlantic as over the North Pacific; the water introduced into the North Atlantic from lower latitudes is more saline than its counterpart in the North Pacific, and the residence time of this water in the region of net precipitation at high latitudes is shorter. However, as Warren (1983) concedes, none of these factors is independent of the already existing thermohaline circulation in the North Atlantic. The salinities of the surface and bottom water masses are similar because one is being actively converted into the other; the higher evaporation is related to higher sea surface temperatures, which are due in part to the greater northward advection of warm subtropical water by the Gulf Stream -- but the Gulf Stream itself is partly thermohaline-driven. Finally, the thermohaline contribution to the western boundary current and the active conversion of surface to deep waters accounts for the shorter residence time in the North Atlantic.

Many geographical clues also exist regarding the asymmetry of the thermohaline circulation in the two oceans. The first and most obvious one is that the North Atlantic extends farther north than does the Pacific, and has a deeper connection with the Arctic. Furthermore, the Pacific is twice as wide as the Atlantic. In connection with the Broecker et al. (1990) argument that the interbasin atmospheric water vapour transport drives the conveyor, the narrow Isthmus of Panama at low latitudes allows westward freshwater export via the trade winds (Weyl, 1968), while the Rocky Mountains block an opposite flow at higher latitudes. Schmitt et al. (1989) have proposed that the narrower width of the Atlantic compared to the Pacific would cause a greater fraction of its area to be susceptible to the incursions of cold dry continental air that favour evaporation. Within the ocean, the salty Mediterranean outflow assists in preconditioning intermediate water flowing into the Norwegian Sea to undergo deep convection, as does the exchange with South Indian waters off the Cape of Good Hope. Finally, Reid (1961) has hypothesized that the poleward extension of South America compared to South Africa might impede the transport of freshwater out of the Pacific by the ACC.

The global coupled EMBM-TIM-OGCM will be used to try and unravel which of the different mechanisms is the most crucial in determining the observed preference for deep water formation in the Atlantic instead of the Pacific.

As discussed in section 3.1, numerical models have revealed decadal-interdecadal variability of the North Atlantic thermohaline circulation. It is not clear, however, to what extent this variability affects the rest of the world oceans. In addition, it is not clear how changes in other basins would affect the existence of North Atlantic decadal-interdecadal variability. For example, Toggweiler and Samuels (1993, 1995) suggest that the southern ocean winds may regulate the rate of North Atlantic Deep Water (NADW) formation. Hughes and Weaver (1994), however, suggest that the winds over the Southern Ocean are only one of a number of ways of regulating NADW formation. In order to examine the interrelationships between the North Atlantic and global oceans, the global coupled EMBM-TIM-OGCM will once more be used in various configurations as outlined in section 4.2.

Furthermore, we wish to address the debate as to where the return flow of NADW occurs. Coarse resolution models (e.g. Hirst and Godfrey, 1993) suggest that most of the return flow happens in the cold water route through Drake Passage. 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 Agulhas Current (Semtner and Chervin, 1992). Observations suggest that some combination of the two routes is appropriate (Broecker 1991). In this project we shall apply perturbations (both passive and active) to the North Atlantic thermohaline circulation to examine, the global equilibrium response, the transient adjustment phase as well as the paths through which the perturbations travel.

4 Methodology

In this section more details are provided as to the methodology which will be used to address the questions posed in section 3.

4.1 The EMBM-TIM

Recently Augustus Fanning, a PhD student working under my supervision, developed a diffusive heat transport energy balance model (EBM), that has been tested in both simplified and global domains. The EBM is loosely based upon the models of Budyko (1969), Sellers (1969), and North (1975). We have extended these models to allow coupling with the GFDL OGCM (Pacanowski et al., 1993) by allowing latent, sensible and radiative heat transfers between the ocean and atmosphere. In an effort to completely couple the ocean-atmosphere system, a moisture balance equation has also been added to the EBM so that freshwater fluxes can be predicted for the ocean model.

The resultant EMBM has been run in a global 2o x 2o domain with fixed sea surface temperatures (Fanning and Weaver, 1996). Under climatological oceanic conditions, the surface air temperatures, specific humidities and surface fluxes are comparable to direct estimates. As an extension to the climatological forcing case, a simple perturbation experiment was considered in which the 1955-59 pentad was compared to the 1970-74 pentad by driving the model under the respective sea surface temperatures. The model exhibits global, as well as basin-mean temperature changes in the latter pentad comparable to direct estimates (Jones, 1988).

A version of the fully coupled ocean-atmosphere model (EMBM coupled to the GFDL-MOM) has been run in a single-hemisphere (60o x 60o) basin, driven by zonally uniform wind stress and solar insolation forcing. A series of several experiments of varying horizontal resolution (ranging from 4o x 4o to 1/4o x 1/4o) and viscosity will be conducted to assess the effect on the components of the net poleward heat transport. We will analyze the relative contributions of the mean and time variant components of the heat transport. These include the effects of the barotropic gyre transport (in the horizontal plane), the meridional overturning transport (in the zonal plane), the baroclinic gyre transport, as well as the eddy and diffusive heat transport components.

It is not clear whether or not the coupled atmosphere-ocean system can actually allow multiple equilibrium states when full coupling of both the heat and freshwater are allowed. This follows since a strong constraint is placed on the heat transport of the coupled system through the incoming solar radiation at the top of the atmosphere.

Ocean-only studies under mixed boundary conditions (Bryan, 1986; Marotzke and Willebrand, 1991; Hughes and Weaver, 1994) do reveal the existence of multiple equilibria. However, in these studies the freshwater flux forcing is specified and the atmosphere is allowed to have infinite heat capacity. The coupled two-dimensional zonally-averaged ocean model-EBM of Stocker et al. (1992) also revealed multiple equilibria but once more the freshwater flux forcing was essentially specified. While the fully coupled AGCM-OGCM of Manabe and Stouffer (1988) displayed multiple equilibria of the North Atlantic overturning, flux adjustments an order of magnitude larger than climatological mean forcing fields had to be used.

As discussed by Weaver and Hughes (1996), the use of flux adjustments places a strong constraint on the implied oceanic heat and salt transports. Therefore, multiple equilibria studies using flux adjustments do not allow the ocean to freely adjust. Our fully coupled OGCM-TIM-EMBM will avoid the need for flux adjustments and will allow free coupling of both heat and freshwater between the ocean and the atmosphere. Once we have analyzed the results from the two-hemisphere studies we will extend our OGCM to two-basin geometry (as in Hughes and Weaver, 1994).

Initial results from the single-hemisphere coupled OGCM-TIM-EMBM have revealed spontaneous decadal variability in the coupled system. These results will be analyzed during this study and the effects of increasing horizontal resolution (i.e., the role of eddies) will also be studied. Initial benchmarks with our IBM workstation cluster have shown that these experiments are feasible with existing machines.

In addition, by integrating the coarse resolution version of the coupled model for many thousands of years, and by applying freshwater perturbations (as in Bryan, 1986) to the high latitude salinity budget or by applying stochastic forcing to the system (as in Weaver et al., 1993), we will be able to investigate the processes of centennial and millennial timescale variability discussed in sections 3.2 and 3.3, respectively.

4.2 The global OGCM

A global ocean model was recently developed in collaboration with Warren Lee and others for coupling of the CCC AGCM. Two versions of this model now exist: The first version is a high resolution model (1.8o x 1.8o x 29 levels) and has now been coupled to the CCC AGCM to investigate the climatic response to increasing atmospheric greenhouse gases and aerosols. The second version of the model is at slightly coarser resolution (3.6o x 1.8o x 19 levels) and is currently being used to understand the structure, stability and variability of the global ocean thermohaline circulation. The description of this coarse resolution model may be found in Weaver and Hughes (1996).

In order to address the scientific questions posed in section 3, we will use the coarse resolution version of this model in both coupled (EMBM-TIM) and uncoupled forms. We will investigate the response of the global oceans to perturbations in the North Atlantic by introducing passive tracers in the North Atlantic and examining the paths of the perturbations. A subset of these experiments (i.e., those that yield interesting scientific results) will be repeated using the higher resolution version of the GCM. Furthermore, the knowledge gained from the more idealized-geometry experiments discussed in the last section will be used to investigate mechanisms and processes involved in North Atlantic decadal-interdecadal variability and their global teleconnections.

As discussed above, the EMBM has been extended to a global domain (Fanning and Weaver, 1996). This model will be coupled to the global OGCM to address the question as to why the Atlantic forms deep water instead of the Pacific. We will use the global OGCM in various configurations to investigate the importance of the different mechanisms outlined in section 3.7. For example, basin geometries will be modified; the hydrological cycle will be perturbed in various regions; the winds will be varied. Hughes and Weaver (1994) suggest that the North Atlantic overturning is proportional to the depth-integrated steric height gradient from the tip of South America to the latitude of deep water formation. They further suggest that changes in the aforementioned external parameters simply act to change this depth-integrated steric height gradient and hence the rate of NADW formation. This hypothesis will be reexamined in the coupled system through the analysis of the steric fields.

4.3 Ocean model validation using CFC-11

As discussed in section 3.5 we will continue to use CFC-11 to validate the climatology and sub-grid scale parameterizations of both our global and North Atlantic models. The theoretical analysis leading to the improvements of the GM scheme will be developed in collaboration with Drs. Amit Tandon and Chris Garrett. In particular, we shall advance this sub-grid scale parameterization through the inclusion of the processes discussed in the last paragraph of section 3.5.

At the surface of our ocean models, the flux of CFC-11 across the air-sea interface is expressed as
[Equation 1],
where lambda is the gas transfer velocity (taken from Liss and Merlivat, 1986), modified for use with CFC-11 (as in Wanninkhof, 1992); Cesw is the concentration of CFC-11 at equilibrium for a given water temperature and salinity (taken from Warner and Weiss, 1985); C is the concentration of CFC-11 at the top level of the model.

In ice-covered regions, the gas transfer velocity for the CFC-11 is reduced using the formula:
[Equation 2],
where F is the fraction of sea-ice cover in tenths (0 = no ice; 10 = complete ice cover). So far we have only considered annual mean forcing so that an annual mean ice cover climatology was obtained from the monthly ice concentrations of Gloersen et al (1992) and used to determine the appropriate gas transfer velocity.

The initial validation process of our global OGCM was conducted by Robitaille and Weaver (1995). Inspired by the success of this analysis we will analyze the climatologies of the OGCMs using each of the different sub-grid scale parameterizations outlined in section 3.5. By quantitatively assessing the ability of the OGCMs to simulate present CFC-11 distributions the strengths and weakness of each parameterization will be realized.

Due to the reduction of vertical mixing when the GM scheme was incorporated, numerical problems associated with vertical grid Peclet violations were found to occur. A flux-corrected transport (FCT) scheme (Gerdes et. al 1991) was therefore implemented into the GFDL OGCM and the consequences of using this advection scheme to eliminate these numerical problems will be investigated. Several integrations comparing mixing and advection schemes, in a simple model, demonstrate that it may be necessary to use a more sophisticated advection scheme (like FCT) when using isopycnal mixing parameterizations.

4.4 The dynamical coupled AGCM-TIM-OGCM

Late last year we acquired the GFDL coupled climate model for use on my local work station cluster. Drs. S. Valcke, S. Zhang and I, now have this model up and running and are using it (in close collaboration with Ron Stouffer and Suki Manabe) to investigate questions concerning the existence of climate variability in the coupled climate system and how it varies as the mean climatic state changes (i.e, does the decadal-interdecadal climate variability found in the coupled model change as CO2 is increased in the atmosphere?). We are also investigating the role of flux adjustments on interdecadal climate variability. The numerical simulations of Delworth et al. (1993), using the GFDL coupled model, revealed interdecadal variability of the thermohaline circulation in the North Atlantic. It is not clear to what extent the variability in that study is preconditioned by the heat and salt flux adjustment fields required to prevent climate drift in the coupled model. It is also unclear whether or not this variability is linked to coupled ocean-atmosphere dynamics or to ocean dynamics alone. In order to do elucidate this, the oceanic part of this model will be run under fixed-flux boundary conditions, made up of atmospheric fluxes (diagnosed from the atmospheric model at equilibrium) and the flux adjustment terms. If similar variability as in the fully coupled experiments is found, we can conclude that the variability is due to internal ocean dynamics alone. In addition, we propose to test the ideas of Weaver and Hughes (1996) to see whether or not we can reduce the necessary flux adjustments by simply adjusting the zonal mean fluxes of heat and salt (and hence the implied oceanic heat and salt transports).

The experiments proposed above are analogous and complementary to those which will be performed with the CCC coupled model. Since the GFDL coupled model is simpler and hence more computationally efficient than the CCC coupled model, it is hoped that the insight we gain from it will allow us to better streamline the future experiments performed with the more sophisticated CCC coupled model

4.5 Decadal variability in OGCMs with various subgrid-scale parameterizations

During the last decade, the tuning of large-scale ocean models towards observations has been achieved by adjusting tracer mixing processes and surface boundary conditions. However, few alternatives to the traditional Laplacian closure of Reynolds stress have been implemented in OGCMs. The influence of the momentum dissipation parameterization and dynamical boundary conditions is not likely to be negligible at coarse-resolution and is worth being precisely evaluated. Therefore, an ocean model has been developed for a coarse-resolution, box-geometry, mid-latitude beta-plane, based on the planetary geostrophic equations and allowing for different choices of momentum dissipation (linear, harmonic, biharmonic or none) and associated boundary conditions (no-slip, free-slip or vorticity closure). These models were first compared to the GFDL OGCM with the same geometry and forcing to validate the planetary geostrophic dynamics. Results from this analysis will be written up shortly.

We shall investigate the effect that different momentum dissipation parameterizations have on the internal decadal-interdecadal variability found in ocean models. Through the use of these efficient ocean models we have found that atmospheric forcing plays the leading role in generating decadal variability in ocean models: flux boundary conditions are the most likely to allow variability as no damping applies to surface anomalies, although the spatial distribution is important. A parameter sensitivity study of the oscillatory behaviour has also been carried out. Initial results suggest that the horizontal tracer diffusivity has a critical damping effect, while increasing the vertical diffusivity strongly enhances the oscillations. The parameterization or even inclusion of convection is found not to be necessary in sustaining the decadal variability, although it is necessary to remove static instabilities. As pointed out by Winton (1996), the variation of the Coriolis parameter with latitude is also not necessary implying that Rossby wave propagation is not important for the oscillations. Greatbatch and Peterson (1995) proposed an explanation in terms of Kelvin waves propagating around the basin. This mechanism was investigated by moving the boundaries or by forcing an f-plane model with a symmetric (about the meridional centre of the basin) heat flux. None of these major changes remove the oscillatory behavior; therefore we conclude that Kelvin wave propagation is not important for the oscillation. As the variability is mainly observed in the region of separation of the western boundary current, we are now looking at 2-layer and 2-dimensional models to investigate an advective mechanism, as originally proposed by Weaver and Sarachik (1991). We believe that we will also be able to develop a very simple one spatial dimension (meridional), non-linear partial differential equation to explain the essential characteristics of the decadal variability found in coarse-resolution OGCMs.

4.6 Simulation of the Younger Dryas event

The use of models to simulate past climatic events is an important avenue of investigation if one is to have confidence in their application to future climatic changes. To this end we shall undertake a simulation of the Younger Dryas event (hereafter YD). The realistic geometry, global, coupled OGCM-EMBM will be used to investigate the transition between the last glaciation and the present Holocene. During the transition, an abrupt return to glacial climatic conditions, known as the YD occurred. The YD cold episode was particularly pronounced in regions bordering the North Atlantic, and is evidenced in northern European and maritime Canadian lakes and bogs; North Atlantic marine sediments; and northwestern European and central Canadian glacial moraines. Although strongest in the northern Atlantic region, further evidence indicates the impacts of the YD were felt throughout the globe.

While the exact cause of the YD is still unknown, a general consensus has emerged that it was linked to an oceanographic phenomena. The amplification of the YD signal in the northeastern North Atlantic suggests a primary role for the THC, particularly NADW production. The question naturally arises as to what source could supply the necessary excess freshwater needed to reduce NADW formation. The obvious sources are the polar ice caps and the Laurentide and Fennoscandian ice sheets (LIS and FIS, respectively). The traditional viewpoint is that the YD was triggered by the diversion of meltwater (due to the retreating LIS) from the Gulf of Mexico to the St. Lawrence (Broecker et al., 1988). However, the fact that deep water is usually formed in local high-latitude regions of small extent suggests that not only the amount, but also the location of meltwater introduced is crucial for interrupting the North Atlantic Conveyor.

We shall conduct experiments to reinvestigate the climatic implications of the geographical and temporal change in the runoff from the LIS and FIS, utilizing estimated meltwater and precipitation runoff from drainage basins in and around the North Atlantic, before, during, and after the YD (Teller, 1990). While Maier-Reimer and Mikolajewicz (1989), using an ocean-only model, found they were capable of shutting down NADW production within 200 years from the time they deflected 347 km3/yr from the Gulf of Mexico into the St. Lawrence valley (roughly half that estimated to have occurred), our initial results suggest the traditional meltwater diversion theory is incapable of inducing a shutdown of NADW. If, however, we apply the runoff estimates previous to the YD, the conveyor is pushed to the brink, allowing the diversion of LIS waters to completely halt NADW production. In an additional experiment we will consider the role of the FIS meltwater on the YD climate.

As an extension of this paleoclimatic experiment we will also conduct a sequence of experiments with the coupled model to investigate the consequences of opening and closing oceanic gateways such as the Isthmus of Panama and Drake Passage (about 3.5 and 30 Million years ago, respectively). In particular, attention will be focussed on the changes in meridional oceanic heat transport and the relation to glacial events in the paleoclimatic record.

5 References not in NSERC Personal Data Form 100

Böning, C.W. and R.C. Budich, 1992: J. Phys. Oceanogr., 22, 361-381.

Bretherton, F.P., 1982: Prog. Oceanogr., 11, 103-129.

Broecker, W.S. 1991. Oceanogr., 4, 79-89.

Broecker, W.S. et al., 1988: Paleoceanogr., 3, 1-19.

Broecker, W.S. et al., 1990: Clim. Dyn., 4, 73-79.

Bryan, F., 1986: Nature, 323, 301-304.

Bryan, K., 1991: Tellus, 43, 104-115.

Budyko, M.I., 1969: Tellus, 21, 611-619.

Danabasoglu, G. et al., 1994: Science, 264, 1123-1126.

Delworth, T. et al., 1993: J. Climate, 6, 1993-2011.

Drijfhout, S.S., 1994: J. Phys. Oceanogr., 24, 353-369.

Gent, P.R. and J.C. McWilliams, 1990: J. Phys. Oceanogr., 20, 150-155.

Gerdes, R. et al., 1991: Clim Dynamics 5, 211-226.

Gloersen, P. et al., 1992: NASA Report SP-511, 290pp.

Greatbatch, R.J., and S. Zhang, 1995: J. Climate, 8, 81-91.

Greatbatch, R.J. and K.A. Peterson, 1996: J. Phys. Oceanogr., in press.

Hirst, A.C. and Godfrey, J.S. 1993: J. Phys. Oceanogr., 23, 1057-1086.

Jones, P.D., 1988: J. Climate, 1, 654-660.

Latif M. and T.P. Barnett, 1994: Science, 266, 634-637.

Maier-Reimer, E. and U. Mikolajewicz, 1989: MPI, Rep. #39, 13 pp.

Manabe, S. and R.J. Stouffer, 1988: J. Climate, 1, 841-866.

Manabe, S. and R.J. Stouffer, 1994: J. Climate, 7, 5-23.

Marotzke, J., 1989: In: Oceanic Circulation Models: Combining Data and Dynamics, D.L.T. Anderson and J. Willebrand, eds., NATO ASI series, Kluwer, 501-511.

Marotzke, J., and J. Willebrand, 1991: J. Phys. Oceanogr., 21, 1372-1385.

Mikolajewicz, U., and E. Maier-Reimer, 1990: Clim. Dyn., 4, 145-156.

Mikolajewicz, U., and E. Maier-Reimer, 1994: J. Geophys. Res., 99, 22633-22644.

North, G.R., 1975: J. Atmos. Sci., 32, 2033-2043.

Pacanowski, R. et al., 1993: GFDL Ocean Group Tech. Rep. #2, 46pp.

Rahmstorf, S. and J. Willebrand, 1995: J. Phys. Oceanogr., 25, 787-805.

Reid, J.L., 1961: Deep-Sea Res., 26, 1199-1223.

Schmitt, R.W. et al., 1989: J. Phys. Oceanogr., 19, 1208-1221.

Sellers, W.D., 1969: J. Appl. Meteorol., 8, 392-400.

Semtner, A.J., 1976: J. Phys. Oceanogr., 6, 379-389.

Semtner A.J. and R.M. Chervin, 1992: J. Geophys. Res., 97, 5493-5550.

Stocker, T.F. et al., 1992: J. Climate, 5, 773-797.

Teller, J.T., 1990: Paleoceanogr., 5, 897-905.

Toggweiler, J.R. and B. Samuels, 1993: In: The Global Carbon Cycle, M. Heimann, ed., NATO ASI series, Vol. 115, Springer-Verlag, 303-331.

Toggweiler, J.R. and B. Samuels, 1995: Deep-Sea Res., 42, 477-500.

Trenberth, K.E., and J.W. Hurrell, 1994: Can. J. Fish. Aquat. Sci., 51, in press.

Wanninkhof, R., 1992: J. Geophys. Res., 97, 7373-7382.

Warner, M.J., and R.F. Weiss, 1985: Deep-Sea Res., 32, 1485-1497.

Warren, B.A., 1983: J. Mar. Res., 41, 327-347.

Weisse, R. et al., 1994: J. Geophys. Res., 99, 12411-12421.

Weyl, P.K., 1968: Causes of Climatic Change, Meteor. Monogr., No. 8, Amer. Meteor. Soc., 37-62.

Winton, M., 1996: J. Phys. Oceanogr., 26, 289-304.

Winton, M., and E.S. Sarachik, 1993: J. Phys. Oceanogr., 23, 1389-1410.

Wright, D.G., and T.F. Stocker, 1991: J. Phys. Oceanogr., 21, 1713-1724.

Zhang, S. et al., 1993: J. Phys. Oceanogr., 23, 287-299.

Return to the Climate Modelling Group WWW page
This page is maintained by
Last updated: