"The development of models for the prediction of climate change is the primary goal of WOCE. The SSG (Scientific Steering Group) emphasizes that models need good data sets to test them and that a close interaction between modellers and observations needs to be maintained. Noting the importance of coupled models for climate prediction, the SSG encourages national and international institutes to strengthen ocean and coupled ocean-atmosphere modelling and the comparison of models with observations. This needs to be in addition to the present climate modelling effort. The SSG emphasizes the need for significant investment in people and computer power well before the end of the WOCE intensive field period, and that data assimilation techniques need to be further developed before the preparation of WOCE data sets.My proposed research plan is aimed directly at this challenge. I propose to not only develop new models for the purpose of large scale ocean/climate prediction but also to use existing models. I further propose to undertake a comparison of these different models with the purpose of understanding their individual shortcomings/assets. One of the goals of this project is to obtain an analysis of the stability and variability properties of the global ocean's thermohaline circulation.
The SSG recognizes that resource problems vary from country to country (people, money, facilities), but emphasizes the importance of modelling as a central element of WOCE. Without substantial advances in modelling, WOCE will not meet its major objectives."
With regards to the coupling problem, I shall be investigating the role of simple atmospheric feedbacks on the stability and variability properties of the ocean's thermohaline circulation. Specifically, in collaboration with D. Harvey at the University of Toronto, I propose to couple an energy balance atmospheric climate model (see Harvey, 1992) to the Bryan-Cox Ocean General Circulation Model (OGCM). Below I shall also discuss other efforts which I shall undertake with my research team as a step towards eventually developing a a global ocean/ice model for the purpose of coupling to an atmospheric GCM. The fully coupled atmosphere-ocean-ice general circulation model will then be used for climate change/prediction simulations. This later project will involve collaboration with Dr. G. Boer and N. McFarlane at the Canadian Climate Centre towards the end of this round of WOCE.
For the sake of brevity I have kept the introduction short although I am also enclosing a copy of a recent review article which I wrote (Weaver and Hughes, 1992) regarding the current state of the art of large scale ocean modelling, the ocean's thermohaline circulation, and its link to climate.
1) - Conduct realistic geometry/topography experiments to examine the stability and variability properties of the thermohaline circulation in the North Atlantic. The goal of this project is to look at the timescales and nature of internal variability, if any, in the North Atlantic.
2) - Study the stability, variability and equilibria of the global thermohaline circulation in idealized geometry. The will allow for a thorough understanding of the role of competing forcings (wind stress, freshwater flux and heat flux) without being burdened by the complexity of a global ocean model.
3) - Examine the structure, stability and variability properties of the global thermohaline circulation by introducing realistic global geometry and topography. An understanding of the global thermohaline circulation and the importance of sub-surface topography will be obtained.
4) - To take the multiple equilibria found in project 2) and determine the relationship between the north-south pressure gradient and the east-west pressure gradient in different basins with and without deep water formation. A similar analysis will be done on the global OGCM fields associated with project 3). Such an analysis will be a useful diagnostic for users of the zonally-averaged model developed by Wright and Stocker (1991).
5) - Implement semi-Lagrangian advection schemes into the Bryan-Cox OGCM. This research will be supported from an AES contract for the first year of the next round of WOCE. Upon implementation of the new advective schemes, comparisons will be done in both idealized and global basins between the Bryan-Cox OGCM with and without the new schemes. In years 2 and 3 the adjoint of this code will be developed.
6) - Develop a finite element, semi-Lagrangian OGCM. Upon development of the code, comparisons will be done with the Bryan-Cox OGCM with and without the semi-Lagrangian advection schemes. The purpose of this project is to introduce more sophisticated numerical techniques (than the standard centred differences) into an ocean model.
7) - Couple an energy balance climate model to the Bryan-Cox OGCM. We shall begin by considering simple idealized basins and then move on to global geometry. The purpose of this project is to investigate the role of simple atmospheric feedbacks on the stability and variability properties of the thermohaline circulation.
8) - To force the Bryan-Cox OGCM with heat and freshwater fluxes and windstress obtained from the 2nd and 3rd generation Canadian Climate Centre (CCC) atmospheric GCM's present day and 2xCO2 equilibrium climates. The main question which we wish to address is whether or not the CCC atmospheric model fields will drive an ocean model to a reasonable climatology. This will provide important information on potential problems which may arise when the CCC atmospheric model is eventually coupled to an ocean model.
9) - To develop simple ice models for coupling to the aforementioned GCMs. The development of an ice model is fundamental if one wishes to eventually couple an ocean model to a fully prognostic atmospheric GCM.
10) - Develop simple analytical box models to understand the stability and variability properties of the thermohaline circulation obtained in GCM experiments.
In the section below I provide a detailed discussion regarding the methodology, rationale and importance of these individual projects. Their order is not meant to be illustrative of their individual priority/importance.
In all our runs, when there was no freshwater input directly into the Labrador Sea internal, self-sustained 22 year period variability existed throughout the entire integration. The poleward heat transport at 39 degrees N associated with the oscillation varied from a maximum of 0.8 petawatts, when the thermohaline circulation was most intense, to a minimum of 0.5 petawatts. The variability was most pronounced in the western North Atlantic, especially in the region of the Labrador Sea. The eastern North Atlantic also underwent variability which was out of phase with the west. The oscillations could, however, be suppressed by adding a slight transport of freshwater directly into the northern regions of the Labrador Sea which caused to the system to evolve into a stable equilibrium. The variability was relatively insensitive to the amount of imposed freshwater flux into the East Greenland Current. These results are, however, very recent so that a complete analysis of them has yet to be done. Whether or not this sort of variability is observed in the real ocean is not known, although, the Greenland ice core data of Hibler and Johnsen (1979) does exhibit 20 year timescale variability, adding an element of credibility to the model results.
The role of bottom topography on the stability and variability of the thermohaline circulation has also not been treated adequately. The above uncoupled ocean GCM simulations used a flat bottom topography. Moore and Reason (1992) suggest that some of the internal variability of the thermohaline circulation might be sensitive to the inclusion of bottom topography although it is not clear how well this topography is resolved in their 12 level model. We are currently including realistic bottom topography in the 3 degrees x 3 degrees North Atlantic experiments discussed above (with 20, 40, 80 and 160 vertical levels) to investigate its role quantitatively. Our initial results suggest that while topography acts to reduce the variability slightly, it still exists especially at high vertical resolution or when a stochastic component is added to the freshwater flux forcing field. The main role of topography appears to be linked to how deep convection can penetrate. With shallow topography, convection occurs to shallower depths so that the subsequently set up east-west pressure gradients are reduced compared to integrations with no topography. I expect that this project will be completed early in the first year of my WOCE research program and hence only request partial support to aid in its completion.
All of the long integrations which were discussed in this section were done using non-eddy resolving ocean models without an ice component. To extend this research project I shall examine whether or not the conclusions which have been drawn from the coarse resolution simulations carry over to high-resolution, eddy-resolving models. This analysis will begin using idealized geometry GCM experiments of the North Atlantic (as in Weaver and Sarachik, 1991a,b) and then move on to the realistic geometry/topography model discussed above. I expect to take on a new MSc or PhD student to look at this project. Similarly, I hope to take on a Postdoctoral Fellow/Research Associate to develop a simple ice model to couple to the ocean model in order to investigate possible feedbacks in the ice-ocean system (see project 9). Furthermore, the mechanism for the generation of the 22 year variability discussed above no longer appears to be the same as the "loop oscillation" described by Weaver and Sarachik (1991), Weaver et al. (1991), and Winton and Sarachik (1992). I will be taking on a PhD student (S. Zhang) in September who will try and develop simple box models to explain this new type of thermohaline variability (see project 10).
Hughes and Weaver (1992) have recently extended the two-basin analysis of Marotzke and Willebrand (1991) to allow for asymmetric basin widths/extents, asymmetric forcing (as in Wells and Mead, 1990), and by adding a deep sill across the now fully prognostic Antarctic Circumpolar Current. Their model was spun up using restoring boundary conditions which were derived from zonal averages of the Levitus (1982) data set for each basin separately. The present day Conveyor Belt equilibrium under mixed boundary conditions, with order 9 Sv of deep water sinking to intermediate depths in the North Atlantic and overriding AABW from the south, was obtained under a diagnosed hydrological cycle with net evaporation over the North Atlantic and net precipitation over the North Pacific.
Under these diagnosed mixed boundary conditions, steady states with no North Atlantic Deep Water (NADW) sinking and deep NADW sinking (with overturning in excess of 20 Sv) were also found; however North Pacific Deep Water formation was never found. The northward extension of the North Atlantic rendered the Northern Sinking and Inverse Conveyor Belt equilibria unattainable even under symmetric temperature forcing in each basin and a hydrological cycle with 1.5 times as much precipitation over the North Atlantic as evaporation in the previous experiment (and almost zero net precipitation in the North Pacific). This preference of the model for Conveyor Belt type circulations was found to be partly set up by the wind stress forcing. In addition, Hughes and Weaver (1992) found multiple versions of the Southern Sinking equilibrium in which deep water formed either only on the inside of the ACC, or also at the polar boundary.
It should be noted that if, as an additional forcing, anomalous amounts of freshwater were taken out of the North Pacific and placed in the North Atlantic (a strong reversal of the present day hydrological cycle) an Inverse Conveyor Belt equilibrium could arise. However, once the external reversal of the hydrological cycle was removed, so that the model was now only forced by the original mixed boundary conditions, the Inverse Conveyor Belt equilibrium became unstable and a transition to either the Conveyor Belt or Southern Sinking equilibria ensued.
The analysis of this work is by no means complete and we are currently investigating the role of different forcing mechanisms (freshwater flux vs heat flux vs wind stress) in determining the stability and variability properties of the global thermohaline circulation. By keeping our geometry simple we are therefore able to focus directly on the important processes involved in driving the global ocean circulation. We expect this project to be completed during the first year of the next round of WOCE.
To assist me with this research I propose to hire a Research Associate. I have already had communications with M. England, M. Winton, D. Holland, and F. Yin regarding this project. They have all shown interest in coming to the University of Victoria. M. England would be the ideal candidate as he has already developed a global Bryan-Cox OGCM in which he took a great deal of care to reproduce the present day water masses of the global ocean (see England, 1992a,b). His initial experiments were aimed at tuning the model to represent the present day climatology and he has not done any of the experiments which I propose above to focus specifically on the stability and variability properties of the global thermohaline circulation.
As the Bryan-Cox OGCM is currently the most widely used and tested ocean model. We will use it as a control to which we will compare the results of the semi-Lagrangian GCM (project 5) and finite element GCM (project 6). Dr. R. Greatbatch and his Research Associate S. Zhang at Memorial University will also be developing a global version of their planetary geostrophic model (Zhang et al. 1992a,b). Furthermore, Josef Cherniawsky at the Institute of Ocean Science is starting to develop a global version of the Oberhuber (1990) isopycnal coordinate model. We will collaboratively undertake a comparison of the climatologies obtained from all these models to identify their individual assets/shortcomings. As an important first stage, comparisons will be done in idealized flat-bottomed geometries as in the work of Weaver and Sarachik, (1991a,b) and Zhang et al. (1992a).
The parameterization used by Wright and Stocker (1991) is:
Wright and Stocker (1991) took the aforementioned ocean GCM output and plotted the left hand side of (4) as a function of the right hand side, for 27 different latitudes and for two depths. They concluded that for the particular GCM simulation which they analysed, a constant value of epsilon = 0.3 represented a good fit except above 500 m where epsilon decayed towards 0 at z=0, consistent with the use of zonally uniform restoring boundary conditions in the OGCM, which essentially demanded rhoE - rhoW about equal to 0 at z=0.
The model of Wright and Stocker (1991) was extended by Stocker and Wright (1991) and Wright and Stocker (1992) to allow transport between various zonally-averaged oceanic basins. Stocker and Wright (1991) connected a zonally-averaged Atlantic Ocean basin to a zonally-averaged Pacific Ocean basin via a zonally-averaged Southern Ocean, in which they allowed for the lack of meridional barriers by choosing epsilon = 0.0001 in some experiments.
Wright and Stocker (1992) extended the model still further by incorporating an Indian Ocean basin, parameterizing the effects of meridional Ekman transport, and allowing for transport through the Indonesian Archipelago and the Bering Strait. They showed that when fluxes from marginal seas such as the Mediterranean and the Red Seas were incorporated into the model, it gave a fairly realistic representation of zonally-averaged water mass characteristics in each of the three ocean basins. They further undertook a detailed scaling analysis and parameter sensitivity study of their model (see Wright and Stocker, 1992, for more details).
Although Stocker and Wright (1991a) point out that the closure parameter epsilon used in their model should vary inversely with the basin width, there is no apparent reason why it should have the same value in basins both with deep water formation and without. Furthermore, it is not clear to what extent the parameterization is valid if wind-stress forcing is included or if one is near a region of deep convection. I propose to quantitatively investigate this by analysing the fields obtained in projects 2 and 3 above. Such an analysis will provide useful information to those users of the zonally-averaged model. T. Hughes, a PhD student, will be undertaking this research.
I have recently received a contract from the Atmospheric Environment Service to hire Dr. S. Das as Research Associate in the School of Earth and Ocean Sciences. He will start working on replacing the numerical advection algorithms in the Bryan-Cox GCM by semi-Lagrangian algorithms (following Robert, 1981; Bates and McDonald, 1982; Ritchie, 1991; Coté and Staniforth, 1991,). This will entail using the GFDL MOM (Modular Ocean Model) version of the code (Pacanowski et al., 1991) which has been written in a much more directly accessible way than the Cox (1984) version. Dr. R. Bermejo, an expert on semi-Lagrangian numerical techniques will visit us at the University of Victoria to discuss progress in both this project and project 6.
The contract for Dr. Das expires as of June 30, 1994. During the first year of the contract he will develop the schemes and test them against the old version of the Bryan-Cox model. I am herein requesting funding for Dr. S. Das for only the last two years of the WOCE proposal. Dr. Das did his PhD work on data assimilation using the adjoint method. The adjoint of the Bryan-Cox model was developed by Thacker, Long and Hwang and further enhanced by Marotzke (1992). I am proposing that Dr. Das develop the adjoint of his semi-Lagrangian advection algorithms for implementation into the Marotzke (1992) version of the Thacker, Long and Hwang code. This will involve collaborative research with J. Marotzke at MIT. Furthermore, once the semi-Lagrangian advection algorithms have been developed and tested we will develop a global ocean model using them for direct testing with the model developed in project 3.
In this project I am seeking full support for a PhD student (P. Myers) to develop a finite element, semi Lagrangian OGCM. This will be a four year project involving a systematic procedure for model development. He will start with the linearized barotropic vorticity equation. Nonlinear terms will then be included, followed by topography and finally stratification. Upon development of the code, comparisons will be done with the Bryan-Cox OGCM with and without the semi-Lagrangian advection schemes in idealized basin geometry. Eventually the model will be extended to a global domain for comparison with the aforementioned global models but this latter project is not within the scope of P. Myers' PhD thesis.
Dr. Salil Das, who will be implementing the semi-Lagrangian advection algorithms into the MOM version of the GFDL model, will also assist in this project. Furthermore, R. Bermejo who has recently returned to Spain has agreed to come out to Victoria a couple of times to provide some additional guidance.
As a first step towards coupling an OGCM to an AGCM, Danny Harvey and I will couple an energy balance climate model (EBCM) to the Bryan-Cox OGCM. D. Harvey will spend part of his upcoming sabbatical (summer, 1992) at the School of Earth and Ocean Sciences, University of Victoria. We will couple a quasi-2 dimensional EBCM based on Harvey (1992) in which an idealized hydrological cycle is to be parameterized. D. Harvey is presently refining his EBCM to include a hydrological cycle which will be ready for coupling in August 1992. This project will involve coupling a hierarchy of models, starting from an idealized one-hemisphere OGCM, then moving to the two-basin model of project 2 and finally to the fully global model of project 3.
In April 1992, T. Hughes and I visited AES to discuss this research with N. McFarlane and B. Gough and we picked up the AGCM fields. T. Hughes, a PhD student under my supervision, is starting to develop a global ocean model in order to look at the stability and variability properties of the global thermohaline circulation. The last part of her thesis will be to undertake these CCC AGCM flux/wind forced global ocean experiments.
Before undertaking this task I propose to couple the Bryan (1969) ice model to a single-hemisphere OGCM to investigate the first-order role of ocean-ice feedbacks to the stability and variability properties of the thermohaline circulation. This analysis will be extended to two-hemisphere/two basin and global ocean models as the knowledge of the coupled system is increased.
Richard Greatbatch and Sheng Zhang at Memorial University will be undertaking similar experiments using their planetary geostrophic model and a close collaboration between our two groups will ensue. Furthermore, Greg Flato and Greg Holloway at the Institute of Ocean Sciences will be using the Hibler ice model to study the Arctic Ocean. We will also be in close communication as our research develops.
I am applying for full support for a Research Associate who will work on this project as well as project 3.
It appears that the variability found in the OGCMs may be linked to either low-latitude Welander (1982) or Welander (1989) oscillations (project 1) or "loop-oscillations" (Winton and Sarachik, 1992; Weaver et al., 1991) based upon the "Howard-Malkus loop" presented by Welander (1986). A detailed analysis of model output fields and the development of simple analytic box models will shed light on the physical mechanisms causing them.