[From the U.S. Government Printing Office, www.gpo.gov]









                                     Toward the Development of an
                                        Ocean Observing System
                                                    for
                                      Climate Study and Prediction


                             OOSDP Background Report Number I








                    The Role of Models in an Ocean Observing System




     67@                                        Prepared by

                                             Neville Smith



                                                  for the










                                                                                    EC








                                             Joint CCCO-JSC
                             Ocean Observing System Development Panel
                                             December 21, 1991

         GC
                                                                   UIC                   RON



         57
         . sc@ r;
         1992







                                         PZOPGrtY Of CSC Library



                             The Role of Models
                                           in an
                        Ocean Observing System






                                        Neville R. Smith
                             Bureau of Meteorology Research Centre
                                      Melbourne, Vic. 3001
                                            Australia




                                       DEPARTMENT OF COMMERCE NOAA
                                        AL HRVICES CENTER
                                     J   0 T H H 0 R S 0 @: A V E N (I E
                                  Ch@InLtSION 6(, @940b-2413







                              A background paper prepared on behalf of
                           the Ocean Observing System Development Panel




          7)-











                                                  FOREWORD

                     Much has been said and written, using sound scientific reasons, about
             instrumenting the ocean to monitor it systematically. The first major attempt to provide
             long-term, systematic, real- or near-real-time observations for predictions over a large-scale
             region was initiated by the Tropical Ocean Global Atmosphere Research Programme and its
             forerunners in the equatorial Pacific to research the El Nifio-Southern Oscillation. The
             amount of timely ocean observations is being increased by the on-going World Ocean
             Circulation Experiment and proposed Global Energy and Water Cycle Experiment,
             initiatives of the World Climate Research Programme, as well as the Joint Global Ocean
             Flux Study and other components of the International Geosphere-Biosphere Programme.

                     Although these research programs have limited 10-year lifetimes, they will leave a
             knowledge base on which to build a permanent ocean observing system. Building such a
             system is now an accepted international goal, and measurements associated with climate
             observations are given high priority. It generally is recognized that all marine activities
             stand to benefit from a comprehensive multi-purpose ocean observing system.

                     With the need for systematic global observations in mind, the Committee on
             Climate Changes and the Ocean (CCCO) of SCOR-IOC (Scientific Committee on Oceanic
             Research - Intergovernmental Oceanographic Commission) and the Joint Scientific
             Committee QSQ of ICSU-W`MO (International Council of Scientific Unions -World
             Meteorological Organization) jointly established the Ocean Observing System Development
             Panel (OOSDP). The OOSDP is charged with formulating the conceptual design of a long-
             term systematic observing system to monitor, describe, and understand the physical and
             biogeochemical properties that determine ocean circulation and the effects of the ocean on
             seasonal to decadal climate changes and to provide the observations needed for climate
             prediction.

                     Fundamentally, the ocean observing system must track in a timely manner: the
             fluxes of heat, fresh water, and carbon between ocean and atmosphere; such fluxes within
             the ocean; and the storage of these quantities within the ocean. Even with the application
             of new technology, it seems unlikely that any affordable system of observations alone will
             accomplish this task adequately. Models with appropriate constraints will be necessary to
             tie sparse observations together and produce products with the required time and space
             scales. The observing system must provide those measurements to develop, verify, and
             initiate the necessary models.

                     The OOSDP, a small group of scientists from five countries, spent its initial
             meetings dealing with the organization of the work, the approach to be followed, and the
             preparation of draft background papers to help identify system design considerations. The
             OOSDP plans that these informal background papers will identify: (1) the elements of the
             Ocean Observing System (OOS) as defined by scientific requirements (as per its charge); (2)
             existing capabilities available to meet these needs; and (3) new types of systems that should
             be encouraged in order to meet these needs more economically and efficiently in the future.
             These background reports are being distributed widely for information and comment.

                     The paper presented here is one of these background reports. It is being distributed
             for inforrnation and for review. Your comments, criticisms and suggestions may be sent
             either to Dr. Neville Smith (Bureau of Meteorology Research Centre, Box 1289K, Melbourne,
             Victoria 3001, Australia) or to the OOSDP chairman, Dr. Worth D. Nowlin, Jr. (Department
             of Oceanography, Texas A&M University, College Station, TX 77843-3146, USA).

                     Clearly the ultimate success of implementing an ocean observing system hinges on
             broad acceptance of its plan by the ocean science community and by the national agencies
             and intergovernmental organizations that will have to implement it. That, in turn, depends
             on broad inforrned participation. The Panel members accepted their appointments
             expecting to draw on the expertise of many other scientists, and from the concerned











            national and international groups, to produce a credible product by the December 1994
            target date. The effort will certainly fail if it is perceived to be the parochial product of only
            the small OOSDP core group. Accordingly, addressees are earnestly requested to give this
            document more than the ordinary attention they might otherwise give to unscheduled
            intrusions that place demands on their time. The task is difficult - the outcome is
            important - the feedback is essential.




                                                            Joint JSC-CCCO
                                                            Ocean Observing System Development Panel











                                                         CONTENTS


                                                                                                        Page
                        CONTENTS
                        PREFACE
                        SUMMARY                                                                           iv


                  1     INTRODUCTION                                                                       1


                  2     THE OCEAN SURFACE AND AIR-SEA EXCHANGES                                            4
                        2.1.1  Surface flux simulations                                                    4
                        2.1.2 Surface field simulations                                                    7
                        2.1.3  Ocean-ice interactions                                                      8
                  2.2   Assimilation and prediction at the surface                                        10
                  2.3   Network design and quality control for surface fields                             12
                  2.4   Developments                                                                      13


                  3     THE SEASONAL MIXED LAYER                                                          14
                        3.1.1  Models for Interpretation and Simulation of mixed layer physics            15
                        3.1.2 Physical-biological mixed layer models                                      17
                  3.2   Assimilation and prediction                                                       19
                  3.3   Network design and quality control for the mixed layer                            20
                  3.4   Developments for mixed layer models in OOS                                        20


                  4     THE TROPICAL OCEANS                                                               21
                  4.1   Simulations of tropical ocean circulation                                         22
                  4.2   Assimilation and prediction in tropical oceanography                              25
                  4.3   Tropical ocean observation networks and quality control                           27
                  4.4   Future developments in the tropical OOS                                           29


                  5     THE THERMOCLINE PROBLEM: EDDIES, GYRES AND VENTILATION                            30
                  5.1   Interpretation and Simulation                                                     32
                               (a)  Low vertical resolution models                                        32
                               (b)  Quasi-geostrophic models                                              33
                               (c)  A balanced model for ocean circulation                                33
                               (d)  Isopycnic models                                                      34
                  3.2 Assimilation and prediction of the thermocline circulation                          35
                  5.3 Observin   g network design and quality control                                     37
                  5.4 Toward operational thermocline prediction                                           39


                                                               i












                    6     THE DEEP WATER CIRCULATION                                                                       40
                    6.1   Ocean general circulation model simulations                                                      41
                          6.1.1    Early developments in ocean general circulation modelling                               42
                          6.1.2    Current OGCM activity                                                                   44
                          6.1.3    Ocean only simulations                                                                  44
                                   (a)   The Community Modelling Effort for WOCE                                           45
                                   (b)   The Fine Resolution Antarctic Model: FRAM                                         47
                                   (c)   A global eddy-resolving model                                                     48
                                   (d)   A model for global water mass formation                                           50
                                   (e)   Alternate OGCM configurations                                                     51
                          6.1.4    OGCMs, in a coupled model environment                                                   52
                          6.1.5    Modelling the transport and storage of tracers                                          55
                          6.1.6    Two- and three-dimensional physical-biological models                                   58
                    6.2   Assimilation and prediction: Inverse modelling                                                   59
                    6.3   OOS network design and quality control for the deep ocean                                        65
                    6.4   Developments required for the deep ocean component of OOS                                        67


                    7     REFERENCES                                                                                       69



                          APPENDIX - ACRONYMS                                                                              84










                                                          Preface

                          The OOSDP was established to formulate the conceptual design of an observing
                  system for the oceans and climate. The goals of the OOS include monitoring, under-
                  standing, simulating, and perhaps predicting, the physical, chemical and biological
                  changes in the ocean climate system. An important element of this will undoubtedly
                  be the effective implementation of a range of ocean models to help in the interpretation
                  and assimilation of the data. For the foreseeable future this will not be accomplished
                  by any single "super" global climate model, but by a suite of different models each
                  suited to a particular aspect of the ocean climate system.

                          The present background paper on the role of ocean modelling in the planned
                  OOS was suggested as a way of fostering discussion among the ocean (and general
                  climate) scientific community so that the Panel could be fully cognizant of the range in
                  opinions. The theme of my approach is the successful marriage of the observing and
                  modelling parts of the system for their mutual benefit, so that there is no effective
                  distinction or demarcation between modellers and observers. I have endeavoured,
                  within the limitations of my knowledge, to give an up-to-date account of modelling
                  activity which is likely to impact on the design and implementation of an operational
                  system. It is likely that too little attention has been paid in some areas, while in
                  others my personal interests may have led to undue em    phasis. Some aspects, such as
                  the oceanic carbon cycle and biogeochemical fluxes, will be discussed in more detail in
                  forthcoming OOSDP reports. I should stress that this paper is a review of modelling
                  and is not intended as an endorsement or recommendation for any specific model or
                  modelling strategy. I hope that any ensuing discussion between the Panel and
                  interested parties in the scientific community will help redress any imbalances or
                  misconceptions within the paper, and lead to a coherent treatment of modelling in the
                  final OOSDP report.

                          To ensure as wider coverage as possible, this paper will also be submitted to a
                  scientific journal and will be subject to the normal scientific review process. Any
                  comments received as a result of this initial limited distribution will be considered in
                  the preparation of the final manuscript.

                          This paper has benefited from the many vigorous discussions within the
                  OOSDP, and from the careful reading by members of the Panel. In particular, Worth
                  Nowlin, Peter Niiler, Liliane Merlivat, George Needler and Art Alexiou made many
                  helpful suggestions, while Alain Vdzina made a significant contribution to the bilogical
                  discussion. I thank all members of the Panel for their interest and assistance.
                  Encouragement and insightful comment was also forthcoming from several colleagues
                  and guests of the Panel. In particular I would like to thank Bill Budd, Peter McIntosh,
                  Scott Power, Carl Wunsch, Kirk Bryan, Dale Hess, Richard Iceeman, John Wilkin and
                  Andrew Moore for their help along the way. The comments of Ping Chang of TAMU
                  on an initial draft of this document were greatly appreciated. I am grateful. for the
                  support of the Bureau of Meteorology Research Centre in allowing me sufficient time
                  to complete this review.



                                              Neville Smith
                                              Omnet address: BMRC.AUSTRALIA (Attn: N.Smith)
                                              Internet address: [email protected]

                                                             iii








                                                         Summary

                  An overview of ocean modelling in the context of formulation of the conceptual design
                  of a long-term systematic ocean observing system is presented with the aim of
                  providing the necessary background for a discussion of the role of models within that
                  system. The modelling is not considered as separate and distinct from the observing
                  component but rather as an integral part of a complete system in which observing and
                  modelling elements operate in concert. For practical reasons the discussion is limited
                  to modelling of the surface fields, the seasonal mixed layer, the tropical oceans, the
                  thermocline waters, and the deep water. Neither the coastal zone nor the biological
                  and chemical components are discussed in any detail.

                       For each of these components the focus is on the degree of interfacing between the
                  models and observations. In the first instance models and observations can be treated
                  as independent elements, with the interfacing being a subsequent
                  validation/verification step. The models in this category are simulation or
                  interpretation tools. A more satisfactory arrangement is to configure the model so that
                  data are assimilated in the process of model solution, and the solution used as an
                  initial condition for a prediction of future conditions. In this category the models and
                  data operate jointly. The final level of sophistication allows the model to feed back
                  information into the observing system, either to assist in its design or to control the
                  operation of the observing system. In this case the emphasis turns to network design
                  and quality control of data.

                       Models of the ocean surface fields and of the exchanges of momentum, heat,
                  nutrients, gases, moisture and light across the interface are central to defining the role
                  of the oceans in climate change. The best estimates of surface fluxes are probably
                  from numerical weather prediction (NWP) centres where both surface and atmospheric
                  data are combined with a model to form a joint estimate. Nevertheless, traditional
                  bulk aerodynamic methods remain useful, particularly for estimates of the climatology.
                  Ocean models are playing an increasingly important role in flux estimation, first as a
                  validation tool for NWP estimates, and second as an independent source for estimating
                  net heat and moisture flux at climatological scales. Estimates of the surface fields are
                  usually obtained without any skilful input from' oceanic models, although this is
                  beginning to change as "operational" oceanographic centres come into being. Sea
                  surface temperature analyses are most developed, but lack of data essentially
                  precludes anything other. than climatological estimates of fields such as surface current
                  or surface salinity. Modelling and analysis of ice conditions is also discussed.

                       Developments in modelling the surface mixed layer are closely linked to
                  determinations of the surface fields. The various approachesto mixed layer modelling
                  are first reviewed, including research with biological-physical models, and their
                  relevance to the ocean observing system discussed. The role of mixed layer physics in
                  the process of combining observations and model estimates is unclear. Experience in
                  NWP is used as a guide to possible problems in ocean mixed layer data assimilation.
                  Some important areas for research and development are listed.

                       The tropical oceans offer unique opportunities for the development of operational
                  ocean prediction systems. The Tropical Ocean Global Atmosphere experiment has laid
                  the groundwork for much of this development and has been active in ocean modelling
                  and observing system design. The development of tropical ocean models is discussed,


                                                              iv









                principally with a view toward ocean-only assimilation and prediction systems, but
                also with regard to their role in coupled ocean-atmosphere systems. Several
                institutions have studied the feasibility of operating real-time ocean analysis and
                prediction systems and, in general, reported positive results. While further research
                and development is required in both the modelling and observing components, the
                indications are that the observation and model information can be combined usefully
                both for the purpose of analysis and forecasting, and for furthering our understanding
                of the tropical ocean system. Methods of network design and objective quality control
                have advanced considerably in this area.

                     Mesoscale eddies, gyres, ventilation, and subduction are all part of the
                thermocline problem. Knowledge of these processes has advanced considerably due to
                the skilful application of simplified models whereby compromises in the configuration
                or physics are made in order to make the solution of the problem tractable. The
                contribution of low vertical resolution models, quasi-geostrophic models, balanced
                models, and isopycnic models is discussed in relation to simulation of mesoscale eddies
                and subduction. The promise of "synoptic" global sea level coverage by satellites has
                prompted considerable research into the assimilation of data into such models,
                particularly eddy-resolving models. It is not yet clear whether there will be sufficient
                data to constrain the eddy field, or whether the smaller large-scale signal can be
                extracted from the mesoscale noise. Attempts to implement ocean "weather" analysis
                and forecast systems and the role of observing system simulation experiments are
                discussed.


                     Models of the deep ocean are central to the considerations of both the ocean
                observing system and the global climate observing system. Early developments in
                ocean general circulation modelling are reviewed and current developments,
                particularly within the World Ocean Circulation Experiment, are discussed. The
                discussion of the ocean-only models is centred around recent attempts at eddy
                resolving simulations for the North Atlantic and the Southern Ocean, and an attempt
                at a global eddy-resolving model. A model for water mass formation is also discussed
                in view of the important role such processes play in the global physical, chemical and
                biological cycles. Ocean models for global coupled ocean-atmosphere climate studies
                are also briefly reviewed, as are models for following the three-dimensional transport
                and storage of tracers and models for physical-biological interactions. Inverse methods
                are used to interface models and data from the deep ocean. The beta-spiral and inverse
                box models are discussed and their application in the determination of ocean
                circulation briefly reviewed. Both the inverse and prognostic approaches to the
                determination of the deep ocean circulation are extremely demanding of resources. In
                view of the limitations on both the observing and modelling components it is critical
                that the information derived from each be maximised. The results from the global
                research programs currently underway, and from planned experiments, will be critical
                in learning the best way to use the limited resources available for monitoring and
                understanding the deep ocean.










                                                            v









                   1. Introduction


                   It is clear that understanding of the ocean will not come from either observations or
                   modelling alone, but will depend on our ability to skilfully marry the two approaches.
                   From. the point of view of modelling within an ocean observing system (OOS), the
                   development and application of ideal "fluid" models independently of the guidance and
                   constraints provided by oceanic observations is likely to be of limited direct use.  Such
                   models, while theoretically interesting and satisfying, belong in the realm of
                   theoretical. fluid dynamics and are of little functional use in the design of an ocean
                   observing system. This somewhat narrow view of oceanic modelling immediately
                   restricts and focuses the scope of this review and draws the measurement and
                   modelling components closer together. In some cases the model development simply
                   cannot proceed without guidance from observations. For example, it is not clear at
                   this point whether the deep hydrographic measurements of the World Ocean
                   Circulation Experiment (WOCE) are representative of the equilibrium state of the
                   ocean, in which case equilibrium models of the ocean are appropriate, or whether they
                   are simply samples from a slowly evolving climate system thus requiring an altogether
                   different and more complex modelling strategy.


                       The case against an isolationist approach is equally clear from the observers point
                   of view. No matter how successful an OOS is at garnering resources it will never be
                   sufficiently dense in either space or time to provide an adequate description of the
                   three-dimensional, evolving ocean system. Some observing systems (e.g. satellites)
                   might provide a quasi-synoptic picture at one level of the ocean, perhaps even at the
                   requisite horizontal and temporal resolution, but we cannot expect to ever have a
                   comprehensive four-dimensional sampling system, nor be able to return to the past
                   and gather the long time series that are necessary for constructing a complete
                   understanding of the oceans. Models can at least provide some assistance in this
                   regard, enabling information to be interpolated and extrapolated in space and time, as
                   well as guiding scant observing resources toward locations and processes that are
                   critical to the observing system.


                       This paper aims to give an overview of ocean modelling in the context of the
                   design and development of a global OOS, concentrating on the temporal and spatial










               scales that are of most relevance to climate change. It has no pretensions to being all-
               encompassing even within the narrow scope outlined above, and through necessity
               must gloss over much of the interesting detail of individual models. The intention is to
               provide the reader who is interested in the design and development of an OOS with a

               background review as a basis for further discussion of the role of models within that
               system. In order to address the wide variety of models and model applications in an
               orderly manner the following discussion has been stratified first by system
               components.     This division is functional rather than unique, and the system
               components are far from being independent, yet it does provide a convenient
               demarcation for the present purposes. The components are:
                    ï¿½ Surface - models targeted specifically at the ocean surface or at the fluxes
                           through the surface. Also ocean-ice modelling.
                    ï¿½ Seasonal boundary1mixed layer - the physics of the upper ocean, principally
                           one-dimensional mixed layer models. Includes some discussion of physical-
                           biological models.
                    ï¿½ Tropical waters - a convenient component to distinguish models for the mixed
                           layer and thermocline from those developed specifically for seasonal and
                           interannual variability in the tropics.
                    ï¿½ The thermocline problem - models of the subtropical gyres, mesoscale eddies and
                           thermocline ventilation and subduction,
                    ï¿½ Deep (cold) water sphere - the realm of OGCMs and inverse models, but with
                           connections to models in previous categories and to global coupled models.
                           Includes deep convection, estimates of transport of energy, water,
                           nutrients, COD etc.


                    We will avoid a detailed discussion of modelling for the coastal zone which is
               covered elsewhere (e.g., the Coastal Ocean Prediction System, 1990).             Physical-
               biological modelling and modelling of the ocean carbon cycle will only be        discussed
               briefly here.   A separate discussion paper will deal with these issues         in detail
               (Merlivat and Vdzina, 1992). Each of the above components are developed in terms of
               the roles models will play in the development of an ocean observing system. These
               roles can be classified according to the interface established between the observing and
               modelling systems. Again this demarcation is used for organisational convenience.




                                                          -2-









                  (1) Interpretation and simulation.
                       Analysis methods (or data interpretation) and model simulations treat the
                  observing and modelling components as individual systems, with the interface being
                  provided by a validation/verification step (the data analysis is used to verify the model
                  or, on some occasions, the reverse).


                  (2) Assimilation and prediction.
                       The next level of sophistication is to allow the observations  and model to jointly
                  determine the solution. Such techniques have many names but we will use "inverse
                  modelling" to describe the class of models which seek equilibrium solutions, and
                  assimilation and prediction to describe the evolutionary system, whereby a solution is
                  mapped out in time by combining model and observed information. In the latter case
                  the concept of numerical ocean prediction (NOP) systems, along lines similar to
                  numerical weather prediction (NVVT), will be the theme.


                  (3) Network design- and quality control.
                       Since we are concerned with the conceptual design of an OOS a further important
                  role for models is their ability to feed information back to observing systems, aiding in
                  the design of the observing network, and helping to monitor the'quality of data. In the
                  former case observing system experiments (OSEs), where actual systems are varied,
                  and observing system simulation experiments (OSSEs), where synthetic systems are
                  tested, have important roles. However the technical complexit3f of such      experiments,
                  and the sometimes limited reward for effort, suggests scientific intuition and
                  experience could be effective alternate strategies.   Models play a crucial part in the
                  quality control of atmospheric data in NWP and it is to be expected that ocean models
                  will play a similar, perhaps even more important, part in the OOS.


                  (4) Developments.
                       Finally, for each of the components, we attempt to identify aspects whose
                  development will be important for OOS. At this stage, and without the benefit of
                  results from WOCE, the Joint Global Ocean Flux Study (JGOFS), the Global Energy
                  and Water Cycle Experiment (GEVVEX) and the last half of the Tropical Ocean Global
                  Atmosphere (TOGA) program, this part will involve some conjecture and is written in
                  the expectation that substantial modifications will be required as the planning for the
                  OOS ( by the Ocean Observing System Development Panel, OOSDP) proceeds.


                                                            -3-










               2. The ocean surface and air-sea exchanges


                       While information from all the ocean is needed to form a complete description
               of the ocean circulation, it is perhaps information from the interface with the
               atmosphere which will be most important in defining the ocean's role in climate
               change. Sea surface temperaturte (SST) has long been regarded as a crucial signature
               of the ocean's function in climate. For the OOS our list of variables is broader and
               includes sea surface salinity (SSS), sea surface state (waves), ocean colour, surface
               current, carbon dioxide, transparency, and so on. This section examines models which
               are specifically directed at the surface fields and, since we are at the surface, also
               examines models which provide information on the flux through the ocean surface
               (wind stress, heat, moisture, C02, light, dust, biogeochemical fluxes, etc.). For the
               most part we will be concentrating on systems where the role of the deeper ocean is
               trivialised, as in NWP and wave forecasting systems, so that surface information is
               garnered almost independently of the ocean sub-surface structure.




               2.1.1 Surface flux simulations



               Global air-sea flux estimates are vital for the development and understanding of ocean
               circulation, particularly in view of their key role in uncoupled ocean model
               applications. The first air-sea flux climatologies were developed from one-dimensional
               parameterizations of the planetary boundary layer (the bulk aerodynamic method) in
               which observations of SST, air temperature, humidity and near-surface wind speed
               were used to estimate the transports of momentum and heat through the ocean surface
               (e.g. Hellerman, 1967; Esbensen and Kushnir, 1981; Weare et al., 1981; Hellerman and
               Rosenstein, 1983; see also the discussions in Taylor, 1989). Similar techniques have
               been used in more recent times to provide improved estimates of the global ocean flux
               and surface drag coefficient (Oberhuber, 1988; Trenberth et al., 1989) and to provide
               regular analyses of surface stress over the tropical oceans (e.g., the Florida State
               University wind stress products, Legler and O'Brien, 1984). Oberhuber (1988) is a
               good example of climatological flux estimation using modern data bases such as the
               Comprehensive Ocean-Atmosphere Data Set (COADS). Such estimates are however



                                                        -4-









                  restricted by the limited data base (usually surface in situ data), poor data coverage,
                  and uncertainties in the planetary boundary layer parameterizations including model-
                  to-model variations (Weare, 1989).


                          In the near term it is reasonable to expect that the best estimates of air-sea
                  fluxes of momentum, heat and water will be derived from operational NVVTP schemes
                  that combine in situ and satellite data. While such schemes are also subject to
                  uncertainties with respect to the planetary boundary layer parameterizations (perhaps
                  even more so), and errors associated with the imperfect physics of the models, they are
                  able to incorporate a vast array of information in forming the air-sea flux estimate,
                  including data away from the surface, non-in situ data and information derived from
                  the interpolative and extrapolative skill of the prediction model. These products are
                  undergoing continual assessment (Arpe et al., 1988; Taylor, 1989; Simonot and Le
                  Treut, 1987; Burridge and Gilchrist, 1989; Barnier and Simonot, 1990; Arpe, 1991) and
                  are likely to benefit from specialist attention during WOCE (e.g. Jochens, 1990) and
                  TOGA (e.g. the TOGA Program on Seasonal-to-Interannual Prediction, UCAR, 1991),
                  and from efforts to expand the domain of influence of NWP to include ocean surface
                  processes explicitly within the data assimilation cycle (e.g. the Global Data
                  Assimilation Program (GDAP) for Air-Sea Fluxes, WCRP, 1989a). These programs aim
                  to add an important dimension to the determination of air-sea fluxes, namely
                  information at the time and space scales which are important for the ocean as distinct
                  from scales important for short-range forecasts. For example, sophisticated wave
                  models are now providing skilful evaluation and validation of NWP surface stress
                  products in near-real-time.


                          Other methods are available for. estimating the surface fluxes relevant to
                  climate studies. Atmospheric General Circulation Models (AGCMs) also diagnose
                  surface heat flux, but without the additional knowledge provided by ingestion of
                  atmospheric data (but with some knowledge of SST). Lambert and Boer (1988, 1989)
                  have undertaken a comparison of such products, the quality of which are, in general,
                  directly determined by the quality of the model. Randall et al. (1991) and Gutowski et
                  al. (1991) discussed the surface energy fluxes in AGCMs and the implications for
                  simulating global and regional climate change.          There are clearly significant
                  differences from model to model. The climate drift problem in coupled models is a
                  direct consequence of imperfections in the AG-CM and Ocean General Circulation


                                                            -5-










               Model (OGCM) components (Sausen et al., 1988; McCreary and Anderson, 1991). A
               further possibility for estimating fluxes is to derive surface flux as a diagnostic of a
               NOP or OGCM simulations. Initial tests (e.g. Gaspar et al., 1990; Leetmaa, 1990;
               Smith et al., 1992) have been promising and suggest that as such systems mature they
               will provide valuable independent estimates of net surface heat and, perhaps, moisture
               flux.



                        Estimation of the radiative sea surface budget is of interest to heat budget
               studies and biological models. Morcrette (1989) discusses the radiative balance in the
               ECMVVT NVVT scheme, which we can assume is typical of the highly complex
               operational NVVT products. While estimations at the top of the atmosphere are good
               (Morcrette and Fouquart, 1988), the surface fluxes are generally larger than
               climatological estimates (e.g. Esbensen and Kushnir, 1981). Simonot and Le Treut
               (1987) identified a systematic difference in the (old) ECMWF scheme of around 20
               W/m', with largest errors in the tropics. While the global annual mean and local

               distributions of surface heat flux are better in the new scheme there remains a
               systematic error (20 W/m') in the shortwave radiative flux (Barnier and Simonot,
               1990). They suggested this bias may be related to cloud cover and cloud optical
               properties. Gutowski et al. (1991) found that AGCMs (for coupled studies) often
               differed in their downward longwave and absorbed shortwave estimations, the
               discrepancies being larger than the climate change signals being sought (they
               concluded these discrepancies would not adversely affect the results). Yu et al. (1991)
               compared the spatial and temporal variability of observed and simulated radiance
               fields and concluded that, at the resolution of the GCM, there was reasonable accord,
               although the model tended to give slightly smaller results. For climate studies in
               general, and for coupled models in particular, such considerations are assuming
               increasing importance.


                        The above discussion is concerned principally with the estimation of fluxes of
               momentum, heat and moisture, but for the OOS fluxes of C02, dust, nutrients,
               chemical exchange, and a range of species are also important. However at this time
               models do not have a prominent role in these estimations.






                                                         -6-










                  2.1.2 Surface field simulations



                           For the surface fields themselves, analysis and interpretation of data has
                  usually been carried out with minimal interfacing to models, although this is now
                  beginning to change. The analysis of SST, for example, has traditionally been
                  performed without recourse to thermodynamic models of the ocean, instead relying on
                  simpler concepts of the temporal and spatial variability of SST to aid in merging,
                  interpolation and extrapolation of information (e.g. Reynolds, 1988; Levitus, 1982).
                  The mapping of other fields, such as SSS (Levitus, 1986; Delcroix and Henin, 1989) or
                  precipitation is primitive since what little is known suggests small scales of spatial
                  and temporal variability.


                           Most studies relevant to the simulation of surface fields, in particular SST,
                  have broader aims than just simply simulating the variability of the surface field itself,
                  and thus fit more neatly into the discussion of seasonal boundary layers, the tropical
                  ocean or deep water models. It is fair to say that, given good quality boundary
                  conditions (surface forcing), mixed layer models are capable of reproducing the diurnal
                  and seasonal variability of SST, SSS, phytoplankton distribution, etc. (Martin, 1985;
                  Miyakoda and Rosati, 1984; Woods, 1985; Seager et al., 1988; Seager, 1989; Woods,
                  1988; Clancy et al., 1990). These issues are developed further in the following section.


                           Finally mention should be made of models for sea surface state and, in
                  particular for ocean waves. There are a wide variety of models in use. The third-
                  generation wave prediction model (WAM) of the WAM Development and
                  Implementation Group (The WAMDI Group, 1988) is widely used and documented and,
                  for the purposes of the present discussion, can be taken as representative of the state-
                  of-the-art in this area. WAM computes the spectral density as a function of frequency
                  and direction over a global (or regional) grid, with the wind and pressure fields
                  providing the external forcing. WAM can be used at varying degrees of complexity
                  depending, among other things, on the resolution and complexity of the non-linear
                  interactions which are retained in the model. Ultimately the success of the model is
                  dependent on the verisimilitude of the forcing. At this time observations, for example
                  from wave-rider buoys or from remote-sensing devices, are mainly used as independent
                  verification points of the model simulation. However this situation is likely to change
                  in the near future as altimeter and SAR data become available from ERS-1 and other


                                                             -7-










               satellite missions, some of it in real-time. The WAM model has been actively used in
               the calibration and validation of both the altimeter and scatterometer of the ERS-1
               mission through real-time trials in parallel with the ECMVVT analysis and forecast
               system.




               2.1.3 Ocean-ice interactions



               For climate scales it is clearly important to take account of ocean-ice-atmosphere
               interactions at high latitudes. For the OOS we are particularly concerned with the
               effects on energy and water exchange. Sea ice effectively insulates the ocean from
               incoming radiation (increased albedo), and moderates the exchange of heat between
               the ocean and atmosphere. This exchange is felt most strongly in the oceanic mixed
               layer, but it is also important for the rates of formation of deep and bottom water
               masses (e.g. Antarctic Bottom Water). It is not appropriate for this document to detail
               the mechanics of ice growth, movement, and decay, nor to delve into the complex
               models that are now available for modelling the cryosphere. The WCRP Working
               Group on Sea Ice and Climate (e.g. WCRP, 1988a,1989b) and the associated Numerical
               Experimentation Group (e.g. WCRP, 1989c) discuss these issues, particularly within
               the context of coupled ice-ocean-atmosphere modelling. This section will instead simply
               summarise the state-of-the-art for ice modelling, and highlight issues relevant to OOS.


                       Washington and Parkinson (1986) gave a full account of current ice' model
               configurations. Sea ice models usually consist of four parts (Lemke, 1991):
               (a)     A surface energy balance which takes account of incoming and outgoing
                       radiation, atmospheric heating, and upward conduction of heat through the ice
                       to calculate the surface temperature.
               (b)     An ice thermodynamic model for determining the rate of heat conduction
                       through the ice.
               (c)     A momentum balance. Atmospheric and ocean stresses, Coriolis forcing, sea
                       surface tilt and internal ice strain, are used to predict the ice velocity.
               (d)     A water budget which takes account of surface accumulation, internal
                       movement, and melting to determine the spatial variability of ice thickness

                       and concentration.




                                                         -8-









                  Flibler (1979), Owens and Lemke (1990) and Parkinson and Washington (1979) are
                  typical of ice models being used in coupled studies.


                          Sophisticated sea ice models have so far only been applied to limited regions
                  (Arctic Ocean, Weddell Sea).         Global models usually adopt a much simpler
                  thermodynamic configuration (e.g. Washington and Meehl, 1986), and retain only very
                  simple advection of sea ice. Experiments with ice models coupled to mixed layer
                  models show that the vertical oceanic heat flux can be significant at the ice edge
                  (St6ssel et al., 1990). Lemke et al. (1990) have shown that the supply of heat from the
                  ocean to the ice is not constant, but is spatially and temporally variant, being largest
                  in winter and in ice-divergent areas (large freezing rates and enhanced convection).
                  Hibler and Bryan (1987) and Willmott and Mysak (1989) showed that advection 'of
                  heat by the ocean currents toward the ice pack is important for simulating the location
                  of the ice edge. These results clearly indicate ocean-ice coupling is important for
                  climate studies. The currently used purely thermodynamic configurations (for global
                  studies) are probably extra-sensitive to changes in the boundary conditions; dynamic
                  sea ice models have reduced sensitivity because of thermodynamic-dynamic
                  interactions which provide a stabilising effect (WCRP, 1988a). Both Lemke (1991.) and
                  WCRP (1988a) included tables of currently used ice model configurations in climate

                  studies.



                          Lemke (1991) discussed facets of (Arctic) ice modelling research which needed
                  further effort.   The role of ice rheology in sea ice - ocean interactions must be
                  investigated and validated against observations of ice drift and ice concentration
                  distribution. The thermodynamics of sea ice are known to be sensitive to the surface
                  energy balance, particularly the surface albedo specification (Meehl and Washington,
                  1990).   In turn, atmospheric models are known to be sensitive to the relative
                  distribution of sea-ice and open water (Budd et al., 1990). Within the ice the rate of
                  heat conduction, and its dependence on inhomogeneities in the ice coverage (e.g. brine
                  pockets and surface melt water), should be investigated. This problem is linked to the
                  ocean mixed layer processes which transport heat and water both vertically and
                  horizontally (e.g. Fichefet and Gaspar, 1989; Mellor and Kantha, 1989). For coupled
                  ice - ocean general circulation model studies the resolution of the ocean model near the
                  ice becomes an important factor. Uncertainties in the surface forcing (see Section
                  2.1.1) also hamper the development of ice models. The most consistent forcing fields


                                                             -9-










                are from NWP and these have been successfully used to drive Southern Ocean sea ice-
                mixed layer ocean models (St6ssel et al., 1990).


                        Recent developments in the understanding of ice dynamics and
                thermodynamics have been greatly enhanced by satellite data (passive microwave for
                ice concentration, Synthetic Aperture Radar (SAR) for concentration and velocity). The
                total data requirement for validation of sea-ice models (e.g. see Table II of WCRP,
                1988a) has many elements in common with OOS requirements. In particular the role
                of drifting buoys and upward-looking sonar on oceanographic moorings give
                information important for both. The use of NUT and AGCM products is also a

                common need.




                2.2 Assimilation and prediction at the surface


                The marriage of data and models can be seen in their most sophisticated form (for
                example, in NWP), or in their most primitive form (sea surface field analyses),
                depending upon which aspect of the surface is under attention. NVVT offers great
                promise for the accurate (in terms of climate) determination of sea surface fluxes of
                momentum and heat, and perhaps of water. In so far as the accuracy can be
                established from independent estimates, it would seem that NMT is good at producing
                surface stress estimates, but less successful in respect of radiative, sensible and latent
                heat flux estimates (Arpe et al., 1988; Lambert, 1988; Lambert and Boer, 1988;
                Burridge and Gilchrist, 1989). The confidence in such products is not uniform over the
                globe, mainly due to the disparate data coverage of the two hemispheres. Arpe (1991)
                suggested NVVT may now be able to deliver accurate estimates of the latent heat flux
                over the Northern Hemisphere extra-tropical oceans, but was less certain about the
                Southern Hemisphere and tropical estimates.         The accuracy of sea precipitation
                estimates is uncertain.      Arpe (1991) compared ECMVY'F estimates with both
                climatology (Jaeger, 1976) and satellite estimates (Janowiak and Arkin, 1991; Arkin
                and Janowiak, 1991) and, in the zonal mean, obtained respectable agreement., The
                precipitation estimates are affected by spin-up problems in both the tropics and
                Southern Hemisphere, so Arpe (1991) suggested using the 0.5-1.5 day forecast mean as
                a reasonable compromise.




                                                          _10-









                            Action at operational centres (e.g. Jochens, 1990; Janowiak et al., 1987) and
                   through programs like the Global Precipitation Climatology Project, the Tropical
                   Rainfall Measurement Mission (TRMM) and GEVVEX will no doubt improve this
                   situation.   It remains to be seen whether the necessary conditions for accurate
                   determination of surface flux include, first, a more sympathetic treatment and
                   assimilation of marine boundary layer measurements and, second, explicit
                   representation of ocean boundary layer physics rather than its proxy SST. The
                   problem of validating such products will be the subject of a separate discussion paper
                   (Taylor and Weller, in preparation).


                            For the sea state problem, research is already under way to incorporate wave
                   models such as WAM within the analysis-assimilation-prediction cycle of operational
                   NWP. With the possibility of remotely sensed sea state data becoming available, it is
                   feasible that such data can be used not only to correct the wave model prediction
                   through better initialisation, but also to assist in correcting the surface wind forcing.
                   The GDAP document (WCRP, 1989a) presented a vision which involved a
                   comprehensive data assimilation system capable of gathering information from many
                   different sources and in a variety of forms, with all parts working in concert to produce
                   atmospheric, sea surface flux and sea surface state analyses and predictions.


                           Of the sea surface variables, only SST is being assimilated and predicted in
                   any skilful way. The Optimal Thermal Interpolation System (OTIS) of the Fleet
                   Numerical Oceanography Centre (FNOC) produces nowcasts of ocean temperature by
                   combining data from various sources with both climatology and model forecasts
                   (Clancy et al., 1986; Clancy et al., 1990). Clancy (1989) and Cummings (1990) discuss
                   plans to interface "ocean feature models" with the thermal analysis scheme so that
                   ocean front and eddy structures can be mapped at both the surface and subsurface
                   levels, even in the absence of subsurface data. The Climate Analysis Centre uses a
                   variety of techniques to produce SST analyses, some of which combine stochastic
                   and/or deterministic model forecasts with in situ and AVHRR data (Reynolds, 1988),
                   and others which use a primitive equation equatorial ocean model combined with an
                   optimal interpolation scheme (Reynolds and Leetmaa, 1989). Folland et al. (1991) and
                   Folland et al. (1992) present an evaluation and intercomparison of several operational
                   SST analysis systems. Present indications are that deterministic models are not
                   quantifiably better than stochastic methods (e.g., persistence), but this will no doubt









               change as the sophistication and resolution of ocean data assimilation systems
               improve.


                        A related category of models are employed for predictability problems within
               the TOGA program. Systems such as Barnett's (1984) statistical prediction scheme for
               the El Nifto - Southern Oscillation (ENSO), or the Cane et al. (1986) dynamical
               prediction scheme for El Nifio are in effect predictions of anomalous SST in the Pacific
               Ocean. These schemes appear to provide predictive skill at least out to several
               seasons, and perhaps even to a few years (UCAR, 1991). However the schemes do not
               assimilate oceanographic data in the normal sense, but use prior analyses of wind and
               SST as predictors/forcing functions.


                        There is almost no activity in assimilating or predicting other surface physical,
               biological and chemical parameters, principally because of lack of data. One of the
               aims of GDAP was to incorporate remotely sensed surface data (e.g. altimeter,
               scatterometer, SAR, Coastal Zone Color Scanner (CZCS)) in a comprehensive data
               analysis and assimilation system.




               2.3 Network design and quality control for surface fields


               An essential component of NWP is the application of objective quality control
               techniques to ensure the integrity of ingested data. Major operational centres, such as
               the ECMWF (Palmer et al., 1990) and NMC (Kalnay et al., 1990; Gandin, 1988), have
               strong links between the analysis/prediction system and the data collectors. The
               ECMWF, for example, routinely check the outcome of quality control, feeding
               information back to measurement platforms where necessary. NMC have developed
               self-correcting techniques to improve the quality of ingested data. NOP centres, such
               as those being developed at NMC (Derber et al., 1990), LODYC/ORSTOM (Morliere,
               1990), FNOC (Clancy and Pollack, 1983; Phoebus, 1990), BMRC (Smith, 1991a,b), and
               elsewhere, are incorporating similar feedbacks but these have yet to be fully
               implemented operationally. None of the existing NWP quality control systems are
               particularly well attuned to the needs of ocean surface (flux) fields. Indeed in many
               systems much of the conventional marine-based data is not retained in the analysis




                                                         -12-









                  system due both to lack of confidence in the data and to problems in the model
                  planetary boundary layer.


                           There have been few OSE or OSSEs studies specifically directed at the oceanic
                  surface fields. As part of FGGE, many OSSEs and OSEs were performed to gauge the
                  influence of particular systems on NVVT (Bourke et al., 1985; Smith, 1989), but none of
                  these took account of the degradation/improvement of air-sea flux estimates as a result
                  of atmospheric observing system changes (one possible assumption might be that
                  improved model forecast skill does imply improved surface fluxes). Such experiments
                  should be done. Leetmaa (1990) was among the first to use oceanographic models
                  actively in the assessment of observing systems, in this case platforms for measuring
                  SST, and was able to quantify the relative accuracy and worth of different
                  measurement platforms.




                  2.4 Developments


                  In designing the surface component of OOS we should remember that NWP has played
                  a significant role in the establishment of the existing measurement system, and that
                  these needs may not be commensurate with the requirements of OOS. For example,
                  SST analyses are often smoothed at the scales of meteorology, thus omitting important
                  oceanographic features. There is a need for a more active role for ocean models in the
                  design. of measurement systems such as in the NMC example cited above. Such
                  investigations, together with carefully planned OSSE/OSEs, will be required if the
                  OOS is to provide good quality surface fields. For example, it may be possible to use
                  models in the design of a SSS measurement program.


                           It is likely that ocean modelling of surface fields and fluxes will be more
                  strongly linked to models for the subsurface circulation of physical, chemical, biological
                  and dynamical fields. Upper ocean mixing models play an important role in the
                  simulation and prediction of non-thermodynamical surface fields (Archer, 1990), and
                  conti nued experimentation and improvement will be important for the OOS.
                  Improvements will come from the merging of separate components which are
                  themselves well understood. In this respect regional analysis and assimilation, say for
                  the coastal regions where the skill of the model, observation density, and quality


                                                             -13-










              control can generally be better controlled than in a global system, could play a
              significant role. We might view the coastal regions as additional buffer zones for
              fluxes of nutrients, water, etc. into the upper and deep oceans, in much the same way
              as we view the surface boundary layer.





              3. The seasonal mixed layer


              The seasonal component of the ocean system is here defined as that portion of the
              water column which is continually changing on diurnal and seasonal time scales, be it
              through the action of winds and wintertime overturning as in mid-latitudes, or
              through the formation and decay of ice at high latitudes'.         From a modelling
              perspective we are mainly concerned with the so-called "mixed layer" class of
              oceanographic models although, as with other components, there is clearly an
              important function for other models in determining the evolution, structure, and
              physical, chemical and biological composition of the layer.


                      The mixed layer provides a buffer zone between the higher frequency scales of
              atmospheric forcing and the slow, large scale circulation of the thermocline and deep
              waters of the ocean. This buffering action is probably most critical for the chemical
              and biological components as, for example, in the regulation of the pCO, and pH of the
              oceans. Microscopic plants and animals (the plankton) consume some of the CO,
              dissolved in the mixed layer and sequester the carbon into the thermocline and deep
              waters, away from the immediate influence of high frequency processes at the
              atmosphere-ocean interface (Cubasch and Cess, 1990). The seasonal layer water
              gradually leaks into the permanent thermocline, thus determining the stratification
              and biogeochemical balance of the interior of the ocean (Pedlosky, 1990). These
              processes must be accurately observed and modelled if we are to understand the larger
              role of oceans in climate change.


                      The need to isolate the seasonal mixed layer as a distinct component in the
              context of the modelling discussion comes more from the recognition that mixed layer




                      Deep convection wiil be considered in section 6


                                                      -14-









                  models are distinctive entities in the global scheme of ocean modelling. The need for
                  parameterization of horizontal mixing by eddies can reasonably be expected to lessen
                  as eddy resolving models are implemented, but the representation of the dominant
                  processes in vertical mixing cannot be improved by simply adopting finer vertical
                  resolution. Judging from the experiences of NWP where the parameterizations for the
                  planetary boundary layer and convection have proved critical in the assimilation and
                  quality control of data, we might expect model developments for the oceanic boundary
                  layer to be similarly crucial for the implementation of simulation, assimilation and
                  prediction schemes for climate scales in oceanography.




                  3.1.1 Models for Interpretation and Simulation of mixed layer physics


                  By far the majority of studies of the seasonal boundary layer have been concerned with
                  the interpretation of data, leading to improvements in vertical parameterization, and
                  with the simulation of the vertical structure of the boundary layer (particularly mixed
                  layer depth). Niiler and Kraus (1977) gave a good account of the fundamentals behind
                  one-dimensional models of the upper ocean, while current research in the
                  parameterization of small-scale processes is discussed in M-aller and Henderson (1989).
                  The "Coastal Ocean Prediction System" workshop (COPS, 1990) also contained several
                  papers relevant to the issues discussed here. Archer (1990) reviewed the physical,
                  chemical, and biological processes which are relevant to the modelling of ocean
                  boundary layer variability, paying particular attention to the modelling of surface
                  PC02 and pH in the oceans. The discussion can be broken into three families of mixed
                  layer models according to the conceptual approach, but it should be remembered that
                  there is considerable overlap and commonality between these approaches.


                          Integral or bulk models (Kraus and Turner, 1967) are an expression of the
                  idea that the surface bound ary layer is well mixed, so that the variables within this
                  layer do not vary vertically. The mixed layer depth, and variable values within the
                  mixed layer, are constrained by the fluxes of turbulent kinetic energy (TYE) and
                  buoyancy through the surface, and by entrainment (detrainment) of water from (on to)
                  the seasonal thermocline. The credentials of the bulk approach have been established
                  by a range of studies (Stevenson, 1979; Niiler and Kraus, 1977; Garwood, 1977; Kim,
                  1976; Miyakoda and Rosati, 1984; Martin, 1985; Gaspar, 1988). Seager et al. (1988)


                                                          -15-









                and Seager (1989) described an SST prediction scheme based in essence on bulk
                turbulent closure theory. With a reasonably simple heat flux, and optimally tuned
                parameterizations (Blumenthal and Cane, 1989) they are successful in hindcasting
                SST, particularly in the tropical oceans.


                        A second class of models relies on the TKE generated by internal shear of the
                mean flow to drive vertical mixing. The mixing is parameterized in terms of the
                Richardson number, either in its bulk form for models based in the mixed layer
                concept (Pollard et al., 1973), or the gradient form for (local mixing) models with
                continuously varying shear (e.g. Pacanowski and Philander, 1981). Some more recent
                adaptions rely on a combination of both (e.g. Price et al-, 1986). The fact that these
                shear instability models develop the mixing parameterization in terms of the mean
                flow, and hence are cognisant of large-scale effects like the Coriolis effect, is the major
                conceptual distinction between the bulk and shear instability approaches. This can
                lead to improvements in the simulation of mixed layer variability since the inclusion of
                inertial effects provides a natural limit to the mixing and to mixed layer velocity.


                        The third family are the second moment closure (SMC) models whose
                foundations he in turbulence theory. The governing equations are formally derived by
                expanding the momentum and tracer equations (heat, salt, etc.) in terms of their mean
                and fluctuating (turbulent) components (Mellor and Yamada, 1982, Mellor, 1989a).
                Higher order expansions, and various closure hypotheses, are used to derive closed
                expressions for the Reynold's stress and eddy diffusion terms. The fact that the
                derivations are based on general turbulence theory, and empirical constants derived
                from laboratory results, means SMC models are in theory not site specific, but can be
                applied in a variety of different situations. Mellor and Yamada (1974). and Mellor
                (1985) present a hierarchy of models, ranging from the level 4 closure (anisotropic
                turbulence) down to the locally dissipative level 2 model (shear and buoyancy
                generated TKE is assumed to be dissipated locally). The SMC models have been
                applied in a variety of oceanic situations, finding particular favour in simulations of
                bottom boundary layer behaviour (Mellor and Yamada, 1982), coastal ocean prediction
                (the Princeton/Dynalysis Ocean Model; Blumberg and Mellor, 1987; Mellor, 1989b),
                flow in the marginal ice zone (Mellor et al., 1986, Mellor and Kantha, 1989) and
                tropical ocean simulations (Rosati and Miyakoda, 1988).




                                                          -16-









                          The comparative simulative skill of the various models has been discussed by
                  Niiler and Kraus (1977), Martin (1985), Archer (1990) and Mellor (1989a). The SMC
                  models are the most general, but are limited in their capacity to be tuned and tend to
                  be more computationally expensive. The SMC models do not take account of mixing
                  due to internal or surface gravity waves and are thus likely to perform less well when
                  compared with bulk or shear instability models which have this flexibility, albeit
                  through adjustable parameters (this is particularly so in the case of level 2 closure).
                  For the one-dimensional situations considered by Martin (1985) the SMC models
                  performed essentially like shear instability models. Mellor (1989a) points out that
                  several of the deficiencies noted by Martin (1985) could be corrected by "introducing" a
                  shear-term parameterization for the effects of internal waves. A further consideration
                  which will be discussed in more detail in a separate paper is the effect of biology on
                  the penetration of short-wave radiation (heating) in the upper ocean. In the tropics,
                  for example, where the vertical mixing of heat and momentum is critical in the
                  simulation of thermal and current structure (Pacanowski and Philander, 1981), we
                  might anticipate that differential heat absorption and consequent modification of the
                  vertical stability would impact on the circulation.




                  3.1.2 Physical-biological mixed layer models


                          Models that couple physical and biogeochemical components are needed to
                  interpret, and in the future assimilate, biological and chemical data collected from in
                  situ platforms and from remote sensing platforms. There is also a requirement for the
                  simulation and prediction of C02 fluxes between the atmosphere and the ocean and for
                  the transport of carbon in the ocean interior. The discussion here will be confined to
                  models in which the emphasis is on the physical aspects of the problem but with some
                  interface to biological processes. Modelling of the biological and chemical processes,
                  and of the oceanic carbon cycle, will be the subject of a separate discussion paper
                  (Merlivat and Vdzina, 1992, ). Two- and three-dimensional models will be briefly
                  reviewed in Section 6. Archer (1990) gave a general discussion of the role of vertical
                  mixing in biological -physical modelling.


                         Vertical one-dimensional models have a rich tradition in biological
                  oceanography due chiefly to the perception that biological production and fluxes can be


                                                          -17-









               related locally to vertical variations in the physical structure (currents, shear, stability,
               turbulent kinetic energy, etc.). These developments range from the study of large-scale
               and seasonal production patterns (e.g., Riley, 1942; Winter et al., 1975; Evans and
               Parslow, 1985; Fasham et al., 1983; Frost, 1987) and the impact of vertical motions on
               daily photosynthetic rates (Falkowski and Wirick, 1981; Woods and Onken, 1982; Wolf
               and Woods, 1988), through to investigations of the physical regulation of
               phytoplankton vertical structure, nutrient fluxes, oxygen and PC02 cycles (Jamart et
               al., 1977, 1979; Taylor et aL, 1986; Musgrave et al., 1988).         The models vary
               considerably in structure and complexity.


                       New production models based on one-dimensional wind-driven mixed-layer
               theory and specifications of the vertical nutrient concentration have been used to
               simulate the upward nutrient fiux (IGein and Coste, 1984; Lewis et al., 1986; Chen et
               al., 1988). Two- and three-dimensional new production models are being developed to
               simulate nutrient distributions in the ocean thermocline (e.g., Toggweiler, 1989). It
               may be possible to couple the mixed-layer models for vertical fluxes and the GCMs for
               transport to simulate upward nutrient fluxes into the photic zone thus yielding
               (indirect) estimates of the organic carbon inputs to the thermocline and deeper waters.


                       The majority of mixed layer models assume the effect of the biology on the
               physics can be ignored. However Sathyendranath et al. (1991), using ocean colour
               images of the Arabian Sea, have shown that phytoplankton distribution can influence
               the seasonal evolution of SST. Mazumber et al. (1990) demonstrated significant
               relationships between plankton biomass, light penetration and mixed layer depth for
               lake systems, strongly indicating a dependence of mixed layer evolution on the biology.
               There are presently very few models that incorporate biological feedbacks into the
               physics. Lewis et al. (1983) and Siegel and Dickey (1987) have produced optical-
               biological parameterisations which show that vertical heating rates by biological light
               absorption can be sufficient to generate upward convection in the water column.
               Simonot et al. (1988) attempted to couple physical and biological processes in a mixed-
               layer model using different vertical grids for the physical and biological sub-models.
               Together these and other studies pro-Vide strong evidence of a potentially important
               impact of marine biota on air-sea exchanges of heat and on vertical current structure
               (through modification of the vertical stability).




                                                        -18-









                  3.2 Assimilation and prediction


                  There have been relatively few studies which attempt to combine mixed layer models
                  with data in an assimilative/predictive mode. The OTIS model of FNOC (Clancy and
                  Pollack, 1983; Clancy et al., 1986; Clancy et al., 1989; Clancy et al., 1990), discussed
                  above in relation to surface fields, combines surface. and subsurface data into a mixed
                  layer model of the thermal state. Forecasts between one ingestion period and the next
                  are provided by a mixture of stochastic (e.g. climatological) and deterministic
                  predictions. Gaspar et al. (1990) took the view that remote sensing provides the best
                  opportunity for collecting the data necessary to constrain the evolution of models (wind
                  stress from scatterometers; radiation and SST from AVHRR). The ocean structure was
                  obtained by inverting the thermal evolution equations subject to. the surface
                  constraints, thereby providing predictions / corrections to the ocean surface fluxes.


                          The quasi-operational tropical forecast systems of NMC (Leetmaa and Ji,
                  1989) and BMRC (Smith, 1991b) both include shear-instability models for vertical
                  mixing. However the role and value of the mixed layer models in these systems is yet
                  to be documented. For example, are the characteristics of a particular mixed layer
                  model important for data assimilation or quality control, and can the model choice
                  impact on the predictive skill of the model? Experience in NVVT has indicated that the
                  planetary boundary layer formulation does have a substantial impact on forecast skill,
                  both through its local determination of the planetary boundary layer profile, and
                  through its influence on the large scale flow patterns.


                          For the tropical ocean prediction problem (see Section 4) there is at least some
                  guidance from the simplified coupled models used in tropical problems (e.g., Cane and
                  Zebiak, 1985). In developing models for interannual variability in the tropical oceans
                  it has been found that the inclusion of mixed layer physics, albeit in simplified form,
                  was essential for the successful coupling of the oceanic and atmospheric components.
                  The forecasts are, however, quite independent of oceanic data, relying instead on a
                  projection based on the accumulated information from past wind forcing. That these
                  models have been as successful as they have (UCAR, 1991) raises the possibility that
                  predictive skill (at least in the tropics at around seasonal time scales) may not be
                  dependent on observations of the mixed layer. However it should be borne in mind
                  that the mixed layer models used in the majority of tropical coupled ocean-atmosphere


                                                           _19-







                models employ only rudimentary physics and may thus be a         Ipoor guide to the utility of
                mixed layer models in the more general assimilation and prediction problem.




                3.3 Network design and quality control for the mixed layer


                There are few published accounts of upper ocean models, being used either explicitly or
                implicitly within larger models, for network design or controlling the quality of data
                from measurement systems. Phoebus (1990) described the quality control algorithms
                employed at FNOC as part of OTIS. The main techniques are climatological (first-
                guess) checks and "buddy checking". Again, we can get some insight into possible
                problems   -through the experience of NWP. There, both at the global scale and at
                regional scales, the experience has been that it is critical that the planetary boundary
                layer growth and convective activity be modelled well. If, for example, the top of the
                planetary boundary layer is anomalously high then data at that height may be
                incorrectly flagged and/or rejected. In the forecasting of fronts vertical mixing appears
                to be critical, so again the efficacy of the quality control is likely to be impacted if this
                aspect is inadequate. There is no reason to believe the situation in oceanograpby will
                be any different, and it may even be harder (the mapping and prediction of Gulf
                stream meanders is a good example; Cummings, 1990; Moore, 1991).




                3.4 Developments for mixed layer models in OOS


                The crucial role of seasonal boundary,layers as the "communication medium" between
                the interior of the ocean and both atmosphere-ocean interactions and land-ocean fluxes
                over the continental shelves underscores the importance of developing credible
                parameterizations for vertical mixing. Models for mixing of heat and momentum are
                at an advanced stage but there remain problems in capturing the evolution of the
                seasonal thermocline and in taking account of mixing generated remotely (say by
                storms) but effective locally through the breaking action of internal waves. A great
                deal of work remains to be done in order to understand the complex physical, biological
                and chemical interactions within the surface layer. For climate prediction, advances in
                this area, particularly with respect to the carbon cycle, will be critical.




                                                            -20-








                          The design and effectiveness of OSSE/OSE-type experiments should also be
                 addressed.    Experience in. NVVT suggests the ocean boundary layer will be an
                 important component of ocean prediction systems, first through its direct impact
                 locally on model-data conflicts, and second through its impact on the global-scale
                 simulation of currents, heat transport and biogeochemical fluxes. Much of the current
                 data will inevitably be collected in the mixed layer and, if this data is to be used
                 successfully as a constraint on the circulation, the ocean data assimilation systems will
                 need to have good representations of the upper layer physics. The problems associated
                 with sharp fronts, say between coastal and deep waters or in high-latitude regions
                 such as the Southern Ocean, or the patchiness of biological activity/production such as
                 in the North Atlantic spring bloom, pose considerable problems for interfacing models
                 with observations.



                         Perhaps the severest test of the integrity of ocean mixing models will come
                 with the advent of NOP as in UCAR (1991) and at the various quasi-operational
                 centres. The inadequacies of ocean boundary layer physics will be laid bare by the
                 continual requirement of matching real-time forecasts (rather than hindcasts) against
                 new, independent data.





                 4. The tropical oceans


                 The simulation and prediction of climate change in the tropical oceans offers unique
                 opportunities for the development of operational systems. It is the singular character
                 of tropical ocean circulation which largely warrants its consideration as a separate
                 model ling'component in the development of OOS.


                         The evolution of modelling and observation of the       tropical oceans is well
                 documented. The planning for TOGA (TOGA, 1985) and               the TOGA Numerical
                 Experimentation Group Reports (TOGA N-EG, 1989, 1990) gave details of the key
                 problems and advances made through the first phase of TOGA. Several review articles
                 have discussed developments in equatorial modelling (Moore and Philander, 1977,
                 Knox and Anderson, 1985; McCreary, 1985), while Philander (1990) presented an



                                                          -21-









                excellent overview of modelling and observational research related to El Niho and La
                Nifia. Charnock and Philander (1989), McCreary and Anderson (1991), Neelin et al.
                (1991), and the proceedings of the Li6ge Colloquia (Nihoul, 1985, 1990), gave a broad
                and up-to-date account of tropical ocean-atmosphere studies. For the specific problem
                of combining observations and model estimates for prediction, "A TOGA Program on
                Seasonal-to-Interannual Prediction" (UCAR, 1991) presented a good account of
                progress to date and highlighted key issues for future development. This document
                embraced much of which is relevant to the present discussion, namely a description of
                the simulative and predictive abilities of the present modelling and observing system,
                an assessment of developments which are required in the future, and a strategy for
                using models actively in the design of a (tropical) ocean observing network.




                4.1 Simulation of the tropical ocean circulation


                Knox and Anderson (1985) reviewed the history and advances in equatorial ocean
                theory. The genesis of the modern approach to equatorial ocean modelling is usually
                attributed to Lighthill (1969) who demonstrated why an equatorial ocean (in his case
                the Indian Ocean and Somali Current) could so quickly respond to changes in wind
                forcing, and that this response may not be locally driven. That is, changes in wind
                forcing at a particular site (but still within the vicinity of the equator) could generate
                significant responses in current and temperature at remote sites.           This concept
                generated much interest and activity in the problems of equatorial ocean circulation
                (see Moore and Philander, 1977) which, together with the growing interest in
                interannual variability and the ENSO phenomenon, was to lead to the inception of
                TOGA.



                         There were many thoughtful papers through this period which are relevant to
                this document. Busalacchi and O'Brien (1980, 1981) and Busalacchi et al. (1983)
                applied a reduced-gravity (that is, single vertical mode) model to the Pacific Ocean
                circulation, and were among the first to critically evaluate model results with observed
                fields. The hindcasts of pycnocline (or equivalently, sea-level) displacements were able
                to capture much of the observed variability in sea level and, if pyenocline displacement
                was used as a proxy for SST, at least sorae of the variability in SST. McCreary (1976),
                Cane and Sarachik (1977) and Cane (1979a,b) revealed the important role of equatorial


                                                          -22-








                 Kelvin and Rossby waves in the response of equatorial oceans. Gill (1983) used a
                 simple linear model with the long-wave approximation to simulate the Pacific Ocean
                 response during the 1972 El Nifio event, using eastern Pacific sea level as the forcing
                 function. Schopf and Cane (1983) redressed the lack of thermodynamics in the
                 reduced-gravity model by including mixed layer physics within the upper layer, adding
                 an extra surface level to accommodate the important influence of Ekman currents off
                 the equator. With only a modest increase in complexity they were able to take account
                 of the key thermodynamic processes which are critical to equatorial adjustment
                 (Schopf, 1983).


                         This period also saw the first general circulation models of the equatorial
                 circulation being constructed. Philander and Pacanow*ski (1980) used an idealised
                 equatorial- ocean driven by zonal winds to study thermodynamic and non-linear
                 interactions in the spin-up and spin-down (Philander, 1981) of equatorial circulations.
                 These scenarios were likened to the different phases of El Niflo. The first realistic
                 simulations of equatorial behaviour were due to Philander and Seigel (1985) (for the
                 Pacific Ocean; see also Philander et al., 1987a, and Philander and Hurlin, 1987) and
                 Philander and Pacanowski (1986) (for the Atlantic). The model included a shear
                 instability vertical mixing parameterization (Pacanowski and Philander, 1981) and
                 unprecedented horizontal and vertical resolution (for OGCM studies) to enable the
                 model to capture the shear and thermal structure characteristic of the equatorial
                 regions. The Philander and Seigel (1985) Pacific Ocean model was forced with NVVT
                 analyses of wind stress for the ENSO event of 1982-83 and a parameterized form of
                 surface heat flux in which latent heating (and its dependence on the wind speed)
                 dominated. This simulation captured the broad features of SST and current anomalies
                 through the event.


                         The proliferation of numerical ocean simulations for equatorial circulation can
                 be attributed in part to the success of these pioneering studies. Philander et al.
                 (1987a) demonstrated the effectiveness of the model in capturing the seasonal cycle of
                 the Pacific and improved our understanding of the complex dynamical and
                 thermodynamical mechanisms which help determine the Pacific Ocean energy budget.
                 Harrison et al. (1989, 1990), Harrison (1989a) and Harrison (1991) showed that Pacific
                 Ocean simulations are quite sensitive to differences in the wind forcing and to details
                 of the surface heat flux parameterizations. Latif (1987), using a slightly different


                                                         -23-









               model, also demonstrated a strong sensitivity to the wind forcing.           Rosati and
               Miyakoda (1988) studied the role of high frequency wind variations and of horizontal
               and vertical eddy parameterizations. With constant mixing and monthly mean winds
               their model developed systematic errors in several areas of the tropical Pacific; with
               grid-dependant horizontal eddy mixing, SMC vertical mixing (at level 21/2; see Section
               3.1.1), and high frequency variability in the wind forcing, the thermal and dynamical
               simulations were realistic.



                        Despite the successes of these and many like studies several problems remain.
               The extreme sensitivity of the models to uncertainties in the surface forcing, both for
               momentum and heating, severely reduces the confidence in the results. The success of
               tropical ocean simulation is inextricably linked to the successes and failures of the
               surface field component determinations (Section 2). The need for temporally consistent
               surface boundary conditions cannot be over-emphasised. The physics of the models
               should also be improved.


                        Most of the models discussed thus far have taken little account of the effect of
               salinity. Cooper (1988) demonstrated the importance of salinity in a simulation of the
               Indian Ocean, mainly through its direct contribution to steric height and the resultant
               geostrophic flow (order 10% of the total). However the more important role may be in
               its moderation of vertical mixing rates as suggested by the "barrier layer" effect in the
               western Pacific (Lukas and Lindstrom, 1987; Godfrey and Lindstrom, 1989; Smith et
               al, 1992). The vertical mixing parameterizations in general need further study (Rosati
               and Miyakoda, 1988, Smith and Hess, 1992). The role of deep temperatures (below the
               main thermocline) and interactions between equatorial circulations and mid-latitude
               circulations (such as might occur in the western boundary currents of the subtropical
               northwest Pacific) have yet to be fully understood.


                        It is now becoming apparent that both interannual and seasonal fluctuations
               in the Pacific are strongly influenced by ocean and atmosphere coupling (Neelin et al,
               1991). For this reason the OOS for the tropical regions cannot be developed without
               due regard for the coupled system. McCreary and Anderson (1991) identified three
               potentially important mechanisms for atmosphere-ocean coupling in the tropical
               oceans: the propagation of internal equatorial Kelvin and Rossby waves and the
               subsequent reflection at meridional barriers (the "delayed-action oscillator"); dual


                                                         -24-









                  equilibrium states, with a "trigger" switching system; and in situ growth of unstable
                  equatorial modes. The linear instability studies of Philander et al (1984), Hirst (1985,
                  1986, 1988), Battisti and Hirst (1989), and others, showed that instabihties of the
                  mean state could generate interannual variability in coupled models. Schopf and
                  Suarez (1988) attributed the reversal between warm and cold states to the propagation
                  of equatorial Rossby waves and their subsequent reflection at the western boundary.




                  4.2 Assimilation and prediction in tropical oceanography


                  The equatorial oceans provided one the first opportunities in oceanography to follow
                  the methodologies used so successfully in NVVT and perform numerical ocean
                  predictions (in this context "nume6cal" implies the use of objective analysis and
                  assimilation techniques). They also provided opportunities for oceanographers to apply
                  newer methods of data assimilation, such as variational techniques and adjoint
                  methods, and perhaps even provide a lead to meteorologists. Ghil (1989) made the
                  case for innovative techniques well, pointing out that oceanographers are handicapped
                  by vastly inferior observing networks and the comparative scales of oceanographic
                  weather. The discussion here will be restricted in the main to applications which have
                  been tested with real data, consistent with our commitment here to methods and
                  models which are relevant to the development of an OOS.


                           Research at NMC led to the first real-time application of equatorial ocean data
                  assimilation and prediction (Leetmaa and Ji, 1989). This application developed from
                  the work of Derber and Rosati (1989) who presented an application of variational
                  techniques to data assimilation in the global domain. In practice assessment of the
                  model was restricted to the equatorial Pacific and, in particular to the latter part of
                  1979 when the ship-of-opportunity program was beginning to provide partial coverage
                  of the region. It was clear from this limited study that data assimilation in the
                  equatorial regions was both feasible and effective. Leetmaa and Ji (1989) incorporated
                  this scheme to provide monthly hindcasts of oceanographic conditions in the tropical
                  Pacific, incorporating complex quality control measures to isolate poor data (Derber et
                  al., 1990; Leetmaa, 1990).       Hayes et al. (1989) showed the model capable of
                  reproducing many of the observed current variations near the equator, although in
                  certain regions (e.g. the southeast Pacific) lack of ingestable information and less-than-


                                                            -25-










                perfect model physics led to less satisfactory results.         Leetmaa and Ji (1989)
                emphasised the critical importance of surface forcing in the scheme whereby poor
                forcing could lead to systematic bias compared with observations. Leetmaa (1990)
                discussed how such a real-time system could be deployed to assess various aspects of
                the observing system and how it might be used to provide an effective alternate means
                for estimating heat and fresh-water fluxes into the ocean.


                        The TOGA program and the ensuing increased real-time             data flow have
                spawned several other analysis and assimilation programs. The ocean analysis facility
                at FNOC was discussed in the previous sections. White et al. (1988) discussed the
                Joint Environmental Data Analysis Centre which was established to provide quality
                control and analysis of real-time and delayed mode bathythermograph data. The
                analysis procedures are derived from the methods of optimum interpolation and were
                first applied in the interpretation of North Pacific (White and Bernstein, 1979) and
                equatorial Pacific (White et al., 1982, 1985) data. The analysis system at the BMRC is
                developed along similar lines (Smith et al., 1991), incorporating the results of Sprintall
                and Meyers (1991) and Meyers et al. (1991) for scales of spatial and temporal
                variability in the Pacific, and adapting the NWP methods described in Lorene (1981) to
                build complex objective quality controls (Smith, 1991a). A numerical model has since
                been added and a series of hindcasts from 1979 through to 1990 have been carried out
                to test the analysis -assimilation system (Smith, 1991b). An Atlantic Ocean analysis
                and assimilation system has also been established at the LODYC, Paris (Morliere et
                al., 1989) with monthly assimilations and analyses being published in the Bulletin
                Ocean Atlantique Tropical. Carrington et al. (1990) and Carrington (1991) described
                research efforts at the UKMO to establish an operational Indian Ocean forecast and
                assimilation system.


                        There are several other studies that, while not directly involved with
                operational centres, have provided insight into the appropriate methods of data
                assimilation (see also Ghil and Malanotte-Rizzoli, 1991). Philander et al. (1987b)
                showed that thermal data were extremely effective at initialising equatorial models,
                but that velocity data were less useful. Moore and Anderson (1989) presented results
                from an assimilation system for the Pacific Ocean based on a reduced gravity model
                and successive correction. This study demonstrated that ingested thermal data could
                quite rapidly correct the model circulation and that this information was retained for


                                                          -26-









                  extended periods. This problem was re-examined by Sheinbaum and Anderson
                  (1990a,b) who replaced the successive correction analysis scheme with a variational
                  approach. They showed that such a scheme could provide corrections to the model
                  forecast in areas remote from the site of the data. Thacker and Long (1988) and Long
                  and Thacker (1989a,b) used the adjoint method and synthetic equatorial ocean data to
                  investigate the conditions under which a control solution could be recovered. They
                  found surface elevation data needed to be supported by subsurface data if the
                  baroclinic state of the control model was to be regenerated. Bennett (1990) showed
                  that generalised inverse methods are tractable and feasible when combining XBT data
                  and an equatorial model. The key to his technique seems to be the efficient solution of
                  a two-point boundary value problem using representer functions.


                          The literature on data assimilation and inverse theory for tropical oceans has
                  grown enormously in recent years and, while many of these studies are not directly
                  relevant to OOS at present, it may transpire that some of them will find applications
                  in equatorial oceanography. Ghil and Malanotte-Rizzoli (1991) provided a valuable
                  overview of the application of data assimilation techniques in oceanography, much of
                  which is directly applicable to the present problem. Anderson and Willebrand (1989),
                  the special issue of Dynamics of Atmospheres and Oceans (Vol 13, Nos 3 and 4,
                  Haidvogel and Robinson, 1989), and the report of the workshop on "Inversion of Ocean
                  General Circulation Models" (WCRP, 1989d) together give a comprehensive account of
                  the healthy state of data assimilation in tropical oceanography.




                  4.3 Tropical ocean observation networks and quality control


                  Equatorial oceanography and TOGA have been pioneers in the study and design of
                  observation networks, and at least part of the success of TOGA must be attributed to
                  the careful attention paid to these details. White et al. (1982, 1985), Sprintall and
                  Meyers (1991), and Meyers et al. (1991) have used the over-sampled YCBT lines of the
                  ship-of-opportunity program to estimate the dominant temporal and spatial scales of
                  the equatorial Pacific Ocean, thereby optimising the information gathered by the
                  broad-scale sampling program. This information is crucial in a program where
                  resources are limited and excessive redundancy means wasted effort.




                                                          -27-









                        Dynamical models provide an important feedback 4 information to the
               observing system.      Existing equatorial ocean analysis and prediction schemes
               incorporate sophisticated quality control schemes (Leetmaa, and Ji, 1989; Smith,
               1991a,b), although the efficacy of such schemes would be enhanced by improved model
               skill and better measurement coverage. Hollingsworth and L6nnberg (1989) and
               Hollingsworth et al (1986) have emphasised the importance of these elements in
               quality control of atmospheric data. Leetmaa (1990) has discussed how information
               provided by the model and, in particular, the degree of compatibility between model
               forecast and data, can be used to delineate deficiencies in the ocean SST observing
               network. Leetmaa (1990) also discussed the relative worth of various observation
               platforms in terms of their impact in the data assimilation system and their value as
               independent validation of various fields. Derber et al (1990), Smith (1991a) and Smith
               et al. (1991) have discussed the ability of quality control systems to impact the efficacy
               of analysis and assimilation schemes.


                        The TOGA Program on Seasonal-and-Interannual Prediction (UCAR, 1991)
               emphasises the importance of OSSE/OSEs in observation network design and clearly
               future OOS enhancements should wherever possible be subject to evaluation in this
               manner. The only OSSE-style experiments carried out thus far have been of the
               "identical-twin" variety, but these have not usually been concerned with observation
               systems per se. Moore et al (1987) compared the relative influence of temperature and
               velocity data in initialising and updating an Indian Ocean model (the data were
               sampled from a separate Indian Ocean simulation). Temperature data were more
               effective in setting the model state, principally because of the greater proportion of
               potential to kinetic energy (Anderson and Moore, 1989). Philander et al. (1987b) also
               concluded that temperature data could be used effectively to initialise the tropical
               ocean circulation (again in identical-twin experiments). Miller (1990) investigated the
               impact the addition of ocean thermal data might have in hindcasts of sea level in the
               equatorial Pacific. To date this is the only true OSE that has been undertaken for the
               tropical oceans. His results suggested the TOGA Tropical Atmosphere Ocean (TAO)
               array would positively impact on hindcasts of monthly mean sea level ( a reasonable
               proxy for equatorial ocean heat content or dynamic height).







                                                          -28-









                  4.4 Future developments in the tropical OOS


                  The prior discussion and UCAR (1991) both underline the need for an increased,
                  positive approach to observing system design through active promotion of OSSEs and
                  OSEs. Although the TOGA program has been largely successful at meeting its half-
                  way goals, the move toward implementing an operational network for post-TOGA has
                  yet to begin. There are a variety of systems presently providing temperature, salinity
                  and velocity information in the tropical regions. These include the ship-of-opportunity
                  network, the TOGA TAO arrays, the Volunteer Observing Ships (VOS) conventional
                  surface measurements, current-meter moorings and drifters for velocity, the island tide
                  gauge network, and the promise of even more and better surface measurements from
                  satellites. The relative worth of each system must be quantified, and the only feasible
                  objective method available at present is impact evaluation in data assimilation
                  systems. This assessment is not likely to be simple, and the results are likely to be
                  somewhat ambiguous in any case. For this reason objective assessment may often be
                  subjugated by evaluations based on subjective scientific intuition. But it is important
                  that the oceanographic community learns to use such tools, and the tropical oceans
                  provide an ideal test bed. The TOGA Program on Interannual-to-Seasonal Prediction
                  (UCAR, 1991) and the TOGA Coupled Ocean Atmosphere Response Experiment (TOGA
                  COARE; TCIPO, 1991) will be important in this learning process.


                           There has, as yet, been no detailed assessment of the impact of salinity in
                  tropical ocean prediction systems, the principal reason being that little salinity data is
                  made available in real-time. The ORSTOM group have been systematically collecting
                  SSS on ship-of-opportunity lines (Delcroix and Henin, 1989) but to date there has been
                  no attempt to assess the impact of such data in a Pacific Ocean assimilation model.
                  Smith et al (1992) have shown that salinity effects are important for the surface
                  energy and water budget, which would suggest a role in prediction as well.














                                                            -29-









               5. The thermocline problem: eddies, gyres and ventilation


               In the processes and modelling chapter of the IPCC report on climate change Cubasch
               and Cess (1990) introduced discussion of the ocean by dividing it vertically into the
               seasonal boundary layer, the warm water sphere and deep water. According to their
               definition the warm water sphere, or equivalently the permanent thermocline water, is
               that part of the column which is ventilated by the seasonal boundary layer
               (exchanging heat, salt, nutrients, gases, etc.), and is pushed down to several hundred
               metres in gyres by the convergence of surface Ekman flow. To the modeller, the
               permanent thermocline has traditionally been the (sharp) demarcation between the
               wind-forced upper ocean flow, such as the mid-latitude gyres, and the less energetic
               deep waters. The problems in modelling and observing the thermodine circulation are
               compounded by the relatively weak signal of the permanent flow compared to the
               energetic ocean eddies. Woods (1991) pointed out that the eddies contain around 99%
               of the oceans kinetic energy. In addition most of the potential - kinetic energy
               interchange occurs at the smaller scales.


                       The majority of models are of an ideal-fluid type but, while providing powerful
               and interesting examples of thermocline circulation theory, they have usually not been
               tested in real ocean problems. This theory is discussed in Pedlosky (1987) and Gill
               (1982) and was the subject of excellent reviews by Pedlosky (1990) and Huang (1991).

               The class of models considered in this section tend to assume the existence of the
               thermocline as a matter of course, rather than considering its genesis and maintenance
               under the unified effect of surface wind and buoyancy forcing. The interested reader is
               referred to Huang (1991) for an account of these aspects.


                       The physical and chemical properties of the warm water sphere have been
               determined over many years by the gradual leakage from the seasonal layer. @ Western
               boundary   currents such as the Gulf Stream, large-scale mid-latitude gyres, and
               mesoscale (tens of kilometres) eddies are all part of the thermocline circulation. The
               circulation is no longer dominated by vertical mixing (as in the seasonal layer) but is
               in near-geostrophic balance, the perturbations being induced by, among other things,
               convergences and divergences due to surface wind and buoyancy forcing, and inherent
               instabilities in the mean flow. This near-geostrophic balance and the absence of deep



                                                        -30-









                   convection has enabled an hierarchy of model approaches. ranging from simple
                   Sverdrup balance (Sverdrup, 1947) and boundary layer models (Stommel, 1948; Munk,
                   1950) through the innovative ventilation models of Luyten, Pedlosky and Stornmel
                   (1983) and Woods (1985), to the complex and sometimes eddy-resolving ocean general
                   circulation models (e.g., Cox and Bryan, 1984; Cox, 1985). Our understanding of the
                   thermocline circulation has benefited from research at all planes of complexity but, in
                   order to keep the present discussion manageable, we limit our attention to just a few
                   of these studies.



                           The need for the models to be relevant to the development of a conceptual
                   design for an OOS immediately shifts the focus toward numerical models, though
                   usually not of the general circulation class (these will receive greater attention in the
                   following section). The dynamic models usually make explicit the notions of flow
                   following isopycnal surfaces (hence reduced-gravity models, isopycnic models) and of
                   the near-geostrophy of thermocline currents (hence the quasi-geostrophic assumption;
                   again refer to Pedlosky, 1987, or Gill, 1982, for details). For water mass formation (i.e.
                   the acquiring of characteristic physical and chemical properties) the layered ocean
                   concept is extended to allow for ventilation at locations where the thermocline layers
                   outcrop and intersect the seasonal boundary layer (Luyten et al., 1983; Huang, 1991)
                   and subsequent subduction beneath the lighter layers above. However it should be
                   recognised that the idealisation of the motion away from the surface as preserving
                   density is too simplistic. The results of Talley (1988) suggested potential vorticity is
                   not exactly conserved along flow lines and that buoyancy forcing of subducted layers
                   cannot be ignored (Pedlosky, 1990). Indeed in reality the fields of density and velocity
                   are non-linearly coupled and    to understand the distribution of either field requires
                   understanding of the full temporal and spatial structure of the ocean circulation.
                   Nevertheless these simplified models have much to contribute to the development and
                   implementation of an OOS and it is toward this contribution that the present
                   discussion is directed.      WOCE has outlined a major effort to enhance our
                   understanding of the mechanics of ventilation and subduction (the Subduction
                   Experiment; WCRP, 1988b,c; Jochens, 1990), the results of which will play a large role
                   in determining the future path of research.







                                                             -31-










               5.1 Interpretation and Simulation


               In this subsection we briefly report on   activities in interpretation of data and model
               simulation which have impacted our understanding of the thermocline circulation.
               While inverse models are the most general interpretive tool, and clearly have a major
               role in diagnosing the thermocline circulation, we postpone a discussion of their role
               until the deep water section (Section 6.2). The calculation of dynamic (or steric) height
               coupled with an assumption of geostrophic balance constitutes the simplest
               interpretive modelling tool for dynamical oceanography (Pickard and Emery, 1982).
               Coupled with comprehensive gridded oceanic data sets such as Levitus (1982) or
               Gordon and Molinelli (1982) the dynamic method can be a simple, but powerful
               interpretive tool (Gordon et al., 1978; Levitus, 1984; Godfrey, 1989). The weakness of
               the method lies mainly in the need to assume a level of no motion somewhere below

               the thermocline. The solution to this dilemma is to either measure the currents
               directly at some level, or include constraints in addition to that of geostrophy. This
               extra level of sophistication is pursued in more detail in Section 6.


                        We order our discussion of simulation tools according to the particular
               simplifying assumptions employed for each model. To model the oceanic gyres,
               mesoscale eddies and the ventilation/subduction process within a single model
               framework would require resources on a prohibitive scale, and might tend to obfuscate
               rather than illuminate the. fundamental processes. So it is usual to compromise and
               simplify the problem by tal;:ing explicit account of, the nature of the thermocline now.
               The approaches tend to divide into either mesoscale eddy genesis tools (Robinson,
               1983) or subduction tools (similar in concept to Stommel, 1979 and Luyten et al, 1983;
               see Huang, 1991), usually with the implicit consideration of gyre recirculation in the
               Sverdrup regime (Pedlosky, 1990).


               (a) Low vertical resolution models
                        The simplest solution to the resource dilemma is to limit the physics and
               vertical resolution. The reduced gravity model discussed in the context of equatorial
               modelling is an example. The numerical solution of such systems is highly efficient
               and enables truly eddy-resolving horizontal resolutions (Hurlburt, 1989). ffindle and
               Thompson (1989) utilised such a configuration in their study of 26- and 50-day waves
               in the Indian Ocean. Hurlburt et al. (1989) and Ydndle et al. (1989) discussed a North


                                                         -32-









                 Pacific Ocean simulation which included estimation of the Indonesian Archipelago
                 throughflow (the fine resolution enabled unprecedented resolution of the complex
                 geography of the throughflow region). Such models include sea surface elevation
                 explicitly and are thus well suited to simulations of sea level variability and are
                 potential users/assimilators of sea surface height data (Hurlburt, 1986; Kindle, 1986).


                 (b) Quasi-geostrophic models
                          These models are founded on the principle that thermocline motions are in
                 near-geostrophic balance, so that the quasi-geostrophic approximation to the full
                 equations is valid (Pedlosky, 1987).      A numerical model consistent with these
                 principals was first developed by Holland (1978) and, as a consequence of the greatly
                 reduced complexity, enabled eddy-resolving studies, mostly within idealised domains
                 (Holland et al., 1983; McWilliams et al., 1978). While such models are limited in their
                 application (e.g. no convection; not applicable at equator) they have provided
                 dynamical and kinematic insight that is not available from other models. For example
                 Schmitz and Holland (1982) were able to make comparisons between simulated
                 mesoscale variability and observed variability in the vicinity of the Gulf Stream. Such
                 models have enhanced our understanding of baroclinic instability in the oceans and of
                 the complex potential and kinetic energy exchanges which control the generation and
                 decay of mesoscale eddies. The quasi-geostrophic models have made a valuable
                 contribution to the understanding of mesoscale and gyre-scale processes, a knowledge
                 which is now being successfully applied to eddy-resolving OGCMs (e.g., Bryan        and
                 Holland, 1989; B6ning et al., 1991).


                 (c) A balanced model for ocean circulation
                         Gent and McWilliams (1984) have derived a set of balanced model equations
                 by combining an exact heat equation with truncated vorticity and divergence
                 equations, reducing the degrees of freedom of the primitive equations by 1/3. The
                 truncated divergence equation may be used in either linear or non-linear form and is
                 now diagnostic (no gravity waves). The model resolves mesoscale turbulence and
                 diabatic processes (c.f. quasi-geostrophic model), and appears to follow the primitive
                 equation slow manifold (e.g. Gulf Stream meanders and ring shedding) more closely
                 than its quasi-geostrophic counterpart. The reduced physics and associated saving in
                 resources make the balance equation models a viable alternative for climate studies,
                 for example in the study of sub-grid scale parameterizations.


                                                          -33-










                 (d) Isopycnic models
                         This class of model makes explicit use of the fact that the preferred plane of
                 flow is along the isopycnal surfaces; the fixed vertical grid ofAhe primitive equation
                 models is replaced by a set of isopycnic (material) surfaces, in essence a Lagrangian
                 approach to the vertical coordinate (see Huang, 1991, for an excellent account of the
                 development of such models). This approach can be traced to Parsons (1969) who
                 successfully used a two-layer model to describe the separation of the Gulf Stream.
                 Bogue et al. (1986) have compared the Parsons model solution and a one-layer model
                 for isopycnal outcropping with good agreement. Bleck and Boudra (1981) introduced a
                 hybrid isopycnic coordinate model to study mesoscale frontogenesis and the dynamics
                 of the Agulhas Current retroflection. The model was limited in its treatment of layer
                 outcropping and the intersection of layers with topography, but nevertheless has found
                 wide ranging applications (e.g. Bleck et al., 1988; see also Boudra, 1989). Huang and
                 Bryan (1987) used an isopycnic model to study wind-driven gyres in a non-eddy
                 resolving climate framework with substantial isopycnal outcropping. This model has
                 been extended to include thermocline ventilation, reminiscent of the Luyten et al.
                 (1983) model.    The model is able to produce the basic features of thermocline
                 ventilation and water mass formation.



                         These models are being implemented with realistic stratification and Ekman
                 pumping in order to test the ideas of Iselin (1939), who one of the first to note that the
                 acquired temperature and salinity properties of thermocline water are related to the
                 wintertime conditions at the outcropping point. The model developed by Woods (1985)
                 examines the connection between subducted waters and penetration of wintertime
                 mixing. Oberhuber (1991) has also developed an isopycnic model which has recently
                 been applied in a coupled atmosphere-ocean-cryosphere context (Oberhuber et al.,
                 1991). Bleck and Boudra (1986) introduced a pure-isopyenic version of the earlier
                 hybrid model, outcropping being accommodated through the Boris and Book (1973)
                 flux-corrected transport algorithm.     The model has been generalised to include
                 topographic and thermohaline driving, making it suitable for subduction studies and
                 coupled model work. Smith et al. (1990) carried out experiments with a seasonal
                 wind-driven isopycnic coordinate model and found the model capable of reproducing
                 the ma or features of the North Atlantic circulation, including western boundary
                 current separation and a Labrador current.



                                                          -34-









                           In addition to these conceptually simpler models there are models based more
                   on the general circulation model approach. These are particularly well suited to
                   thermocline c irculation and subduction experiments, some of which will be discussed in
                   the deep water section. Still other models (e.g. the model of Blumberg and Mellor,
                   1987) have been developed for confined basin and coastal studies but are now finding
                   applications more aligned with large-scale ocean studies.




                   5.2 Assimilation and prediction of the thermocline circulation


                   Assimilation and prediction research has focussed on the eddy field in the main,
                   although some of the inverse modelling studies are relevant to subduction (Section
                   6.2). The goals for the thermocline/warm water sphere are somewhat different from
                   those of equatorial prediction. Here the aim is more closely aligned with weather
                   prediction in meteorology, but now the weather is provided by mesoscale eddy activity,
                   and the predictability time scales are those of mesoscale eddy generation and decay
                   (i.e. weeks to months). So the model should be eddy resolving and efficient, and be
                   commensurate with the data being ingested. For these reasons the quasi-geostrophic
                   formulation has been favoured in the majority of applications with particular emphasis
                   or. methods for ingesting altimeter data.


                           The methods used for assimilating data vary from blending and nudging
                   techniques (e.g. Malanotte-Rizzoli and Holland, 1986, 1988, 1989), sequential optimal
                   interpolation (e.g., Robinson et al., 1986; White et al., 1990a,b,c), Kalman filters (e.g.,
                   Bennett and Budgell, 1987; Miller, 1989; Gaspar and Wunsch, 1989) and variational
                   methods (e.g., Long and Thacker, 1989a,b; Moore, 1991). The principal motivations for
                   data assimilation come from the prospect of having global, near-synoptic altimeter data
                   coverage, and from the pressing need for oceanographers to extract maximum
                   information from a sparsely distributed in situ observation network (Ghil and
                   Malanotte-Rizzoli, 1991).


                           Research in the assimilation of altimeter data has mostly been conducted with
                   identical-twin configurations. Marshall (1985) used a quasi-geostrophic model to show
                   that altimeter data assimilated into a model could improve the estimate of the geoid.
                   Webb and Moore (1986) , Hurlburt (1986) and De May and Robinson (1987) showed


                                                             --35-









                how altimeter data could constrain the thermocline and deep ocean flow through its
                projection onto the vertical modes. White et al. (1990a,b,c) used the Holland (1978)
                eddy-resolving quasi-geostrophic model to investigate the impact of continuous
                assimilation (by optimum interpolation) of (simulated) GEOSAT altimetric data. They
                found the data effectively constrained the linear part of the model domain but was
                unable to constrain the non-linear portion after initialisation. because of aliasing
                problems. Furthermore they showed that only wavelengths longer than the Nyquist
                sampling wavelength (twice the GEOSAT track separation) were positively impacted
                by assimilation. Kindle (1986) also addressed the problem of sampling frequency/
                spacing versus spatial scales of the mesoscale field, concluding that altimeter sampling
                must at least match the outer eddy radius if the mesoscale field is to be reproduced.
                Holland (1989) combined GEOSAT data with a quasi-geostrophic model, using a simple
                nudging technique, to predict the surface and deep eddy field and mean flow of the
                Agulhas retroflection region. The GEOSAT sampling in this region is seemingly
                sufficient to infer structure at the eddy scale. Ghil and Malanotte-Rizzoli (1991)
                provided an informative appraisal of the relative merits of direct ingestion of data
                versus blending/nudging techniques.


                        Robinson and Walstad (1987) discussed a model designed for data assimilation
                (Carter and Robinson, 1987) and forecasting. The model has been applied to both the
                Californian Current (Robinson et al., 1986) and to the Gulf Stream (GULFCAST;
                Robinson et al., 1989a,b). Their procedure utilised bathythermograph data, altimeter
                data and, in particular, AVHRR (SST) data to resolve the path and features of the Gulf
                Stream (hence "feature modelling"). They were able to produce weekly forecasts of the
                Gulf Stream path including meandering and ring shedding. Moore (1991) used a
                similar model and data but adopted adjoint methods in place of the Carter and
                Robinson (1987) optimal interpolation technique. The model was able to correct for
                large errors in the speed and position of the Gulf Stream and demonstrated the ability
                of the adjoint method to transfer information into data sparse regions. The advection
                of information in space and time (c.f. statistical extrapolation in conventional optimum
                interpolation) is an important advantage of this approach. However, in common with
                all regional models of the atmosphere and ocean, the uncertainty in the open boundary
                condition (i.e., the consideration of information from outside the domain which is
                advected and propagated through the boundaries) is important in determining error
                growth.


                                                         -36-









                           The assimilation of observations from the conventional observing network
                  poses a contrasting problem since we now have direct measurements of subsurface
                  structure, but only in a sparsely distributed network. Malanotte-Rizzoli and Holland
                  (1986, 1988) concentrated on the problem of assimilating hydro aphic data into both
                                                                                    : gr
                  steady and transient, eddy-resolving quasi-geostrophic models. They concluded that
                  such data might constrain the long-term bias of the model (i.e. the model and
                   observed" climatologies will be similar) but that it would be unable to resolve oceanic
                  weather. Malanotte-Rizzoli and Holland (1989) addressed the same problem, plus that
                  of initialisation, but in a primitive equation model. They found that baroclinic
                  information (i.e. hydrographic data) could quickly excite a realistic barotropic field, but
                  that depth-averaged data were unable to constrain the baroclinic component.
                  Unitialisation becomes an issue if, after data insertion, the model is out of balance;
                  this is not an issue with balanced equation or quasi-geostrophic models since their
                  formalism ensures balance by eliminating waves on the fast manifold.)


                           The growth of interest and knowledge in assimilation methods for the
                  thermocline circulation is attested by the large number of articles published in recent
                  years (Haidvogel and Robinson, 1989; WCRP, 1989d; Anderson and Willebrand, 1989;
                  the International Symposium on Assimilation of Observations in Meteorology and
                  Oceanography, WMO, 1990; Ghil and Malanotte-Rizzoli, 1991). A range of data and
                  models have been studied, and several different data-model merging techniques have
                  been used. This variety of approach is invaluable if the oceanographic community is to
                  learn the optimal strategy for combining information from observations with the
                  interpolative and interpretive skill of models.




                  5.3 Observing network design and quality control


                  GULFCAST (Robinson et al., 1989a,b) is a good example of the design and
                  implementation of a quasi-operational forecasting system with feedback to the
                  observing network.      The key data in the Gulf Stream observing network were
                  delineated by repeated trials with successively enhanced systems. By studying how
                  the forecast improved or degraded when a particular system or technique was added
                  (e.g. GEOSAT data in the Moore, 1991 study) they learned increasingly more about
                  both the modelling and observing systems. In essence they were performing OSEs.


                                                             -37-









                       The studies with simulated altimeter data (e.g. Webb and Moore, 1986; White
               et al., 1990b,c) are like OSSEs, but in this case it is a "with or without 0 data" style
               experiment.    To date no studies have looked at the case where simulated (e.g.
               altimeter) data is withheld from an existing non-trivial data assimilation system
               (rather than a data-less "control" model). Furthermore most of the studies have
               considered the altimeter data to be perfect (no noise or bias; White et al., 1990c do
               consider the addition of noise), and have thus not addressed the problem of quality
               controlling in a quasi-real-time situation (such problems have plagued NWP in the use
               of vertical temperature sounding data). This latter problem is not trivial. In all
               likelihood the number of altimeter data will be insufficient to define the eddy field and,
               as with any under-sampled system, this will make it make it extremely difficult to
               quality control and remove bias from an inevitably noisy system.


                       In the sense that all interpolation methods are sensitive to the temporal and
               spatial resolution of data, most of the studies cited above could be said to have
               relevance to array design.     Bennett (1985) specifically looked at the question of
               observation array design in the context of inverse methods. The success of models in
               resolving particular scales, as in the Holland (1989) and White et al. (1990a,b,c)
               studies, or in transporting information in time or space (Moore, 1991), all impacts on
               the design of observing networks. Ultimately, however, the answer to the network
               design problem for the thermocline circulation will depend on exactly what questions
               are being asked and the details of the accompanying specifications. If the requirement
               is to resolve eddies then the existing research is well on the way to providing a
               prescription, but perhaps not a solution to the logistical and implementation problems
               (i.e. a cost effective means for the requisite sampling density and rate).         If the
               requirement is to delimit frontal structures and, say the outer and inner boundaries of
               the Gulf Stream, then a great deal of work is yet to be done. There is tonfidenc;e,
               given sufficient resources, that we may be able to forecast the broad-scale weather
               patterns in the ocean (e.g. Gulf Stream meanders and ring movement) and, with
               further research, extract the weak    signal'of the large-scale  pattern froml the noisy
               mesoscale eddy field. But forecasts   of the genesis and passage of fronts (for example)
               may require a great deal more research.







                                                         -38-









                  5.4 Toward operational thermocline prediction


                  The promised growth in non-conventional oceanographic data, such as sea level
                  estimates from altimeters and wind-stress from scatterometers, is clearly an important
                  part of the prediction problem for the thermocline circulation.       It remains to be
                  determined what level of sampling is required to constrain the gyre circulation or the
                  eddy field. The temporal and spatial scales of ocean weather make resolution of the
                  circulation by conventional measurements unlikely, but they will almost certainly
                  remain important as ground truth/validation points for predictions based on non-
                  conventional data. The research to date suggests multiple altimeters may be required
                  to pin down the mesoscale field, but at this time it is not clear that such resources are
                  going to be available even for a short research period.


                          Resource limitations are an issue for both the subduction and dynamical
                  oriented streams. Trials with isopycnic models have been promising but much remains
                  to be learnt of the fundamental processes by which water masses acquire their
                  essential characteristics. Temporal effects and sub-surface buoyancy forcing are two
                  aspects which provide an immediate challenge. Theoretical modelling has already
                  provided excellent guidance in this problem (Huang, 1991; Pedlosky, 1990) and, it is to
                  be hoped, will continue to do so in future. Unrealistic rates of formation, and
                  unrealistic water mass characteristics are unfortunately a feature of most global ocean
                  climate models and, while increased resolution will partially alleviate the problem, a
                  great deal of experimentation with process models is still required. The data to
                  design, test and extend these models must be gathered. Eddy resolving studies are in
                  the main confined to truncated or simplified (e.g. quasi-geostrophic) configurations
                  where resource requirements are substantially reduced compared to OGCMs. Such
                  models have been thoroughly tested against primitive equation models in like domains
                  (usually idealised basins), but it remains to be seen whether they are the good
                  analogues they appear to be under more general conditions (thermohaline forcing,
                  finite/steep topography, deep & shallow convection). Uncertainty in the lateral (open)
                  boundary conditions, and in the initial conditions, remains a problem for regional
                  models, particularly in respect of their influence on the growth of errors. A similar
                  uncertainty exists for the upper boundary condition for global and basin-scale models
                  of the gyre recirculation and ventilation and subduction processes.




                                                           -39-










                       For the non-thermodynamic elements of the system research is under way to
               assimilate data in physical-biological models (two- and three-dimensional models are
               discussed briefly in Section 6 and in detail in Merlivat and Vezina, 1992). An example
               is the development of techniques to assimilate ocean colour measurements (a proxy for
               phytoplankton). A problem that must be faced in such systems is that of initialisation.
               Introduction of information on one component of the ecosystem immediately creates
               imbalances, and the remaining components must be adjusted to maintain equilibrium.
               Meteorologists and oceanographers have encountered similar problems with respect to
               physical and dynamical variables (Daley, 1981; Malanotte-Rizzoli et al., 1989). The
               use of variational and adjoint methods for assimilating a time-sequence of data
               appears to have removed some of the problems associated with initialisation. and it
               may be that if such methods (specifically, best fit solutions to all controlling variables)
               can be applied in physical-biological models then they too will suffer less from gross
               imbalances between the system components. However the very nature of biological
               and biogeochemical systems ("patchiness"; rapid growth and decay; sharp gradients)
               makes them an imposing challenge. These issues will be taken up again in a separate

               paper.






               6. The deep water circulation


               The motivation for considering the deep waters as a separate component is provided by
               their role in long-term climate change (Cubasch and Cess, 1990) where they are
               considered as the "memory" of the climate system. The global redistribution of heat,
               freshwater, and dissolved chemicals (e.g. carbon dioxide) through the deep ocean is a
               climate controlling process and one that will need to be simulated well if we are to
               describe the present state and future evolution. For the OOS and the Global Climate
               Observing System (GCOS) the deep water sphere is arguably the most critical
               component of all. The discussion here will concentrate on large temporal and spatial
               scales. However it is important to keep in mind that the large scale circulation is
               subject to forcing by the faster time-scale components considered previously (for
               example, Ekman pumping and subduction of heat and chemicals through the





                                                         -40-









                    thermocline, and high latitude deep thermohaline convection) and by surface and
                    bottom effects at much finer spatial scales.


                           Once again there is considerable overlap between issues considered previously
                    (e.g. eddies, water mass formation) and those that are relevant here; Likewise the
                    demarcation between models of the deep ocean and seasonal and thermocline models is
                    indistinct. In general the deep ocean is the domain of OGCMs and inverse models
                    since simulation and understanding of the "equilibrium" state of the oceans is the
                    primary consideration. Nevertheless it should be borne in mind that the oceans being
                    sampled now and in the OOS are in all likelihood not in equilibrium, but in a
                    continuous state of (slow) change. As was the case for tropical ocean models, it is
                    difficult to consider global oceans without including ocean-atmosphere interactions
                    (that is the global climate system; see WCRP, 1991a).


                           If equatorial ocean modelling for the OOS was in essence a discussion of
                    TOGA modelling activities, then deep ocean modelling is largely a discussion of
                    WOCE-related activities. The International WOCE Implementation Plan (WCRP,
                    1988b,c), and various national plans (e.g. Jochens, 1990) provided a good coverage of
                    the present status and priority issues. The problem can be conveniently divided into
                    prognostic models and inverse models. Consistent with the format of previous sections
                    prognostic models will be the subject of Section 6.1 (interpretation and simulation)
                    while inverse models will occupy the bulk of the discussion of Section 6.2 (assimilation
                    and prediction). While this division is convenient it is by no means straightforward.
                    For example, increased carbon dioxide experiments are usually described as
                    11 predictions" though they are forced by some (time varying) prescribed concentration of
                    greenhouse gases. Inverse models, on the other hand, could equally be thought of as
                    interpretation tools since they are aimed at revealing the equilibrium circulation given
                    a certain set of constraining observational and thermodynamic relationships.




                    6.1 Ocean general circulation model simulations


                    The development and application of OGCMs has been covered in several technical
                    papers (Bryan, 1969; Semtner, 1974; Hasselmann, 1982; Cox, 1984; Han, 1984), texts
                    (e.g. NAS, 1975; O'Brien, 1985, Washington and Parkinson, 1986, Anderson and


                                                            -41-










             Willebrand, 1989; Philander, 1990) and review articles (Holland, 1977, 1979; Bryan,
             1979; Meehl, 1990). This section will concentrate on milestones which marked new
             levels of model sophistication and understanding of the deep ocean circulation, and on
             some recent applications.




             6.1.1 Early developments in ocean general circulation modelling


             The "beginning" of numerical modelling for the large-scale ocean circulation is usually
             attributed to Bryan (1963) who followed the lead of atmospheric G-CM research and
             used the barotropic vorticity equation to model the oceanic circulation in a rectangular
             basin. (Sarkisyan, 1962, 1975 was another pioneer in this arena). The first global
             model with sufficiently general numerics to take account of real geometry and
             topography was presented by Bryan (1969), later updated to include better treatment
             of islands and improved efficiency on vector machines (Semtner, 1974).           Holland
             (1967) applied similar models in idealised basin studies.        At the time of "The
             Numerical Models of Ocean Circulation" meeting in 1972 (NAS, 1975) computer power
             was an extremely limiting factor in the design of global applications. Nevertheless
             that meeting introduced the first attempts at modelling the baroclinic response of
             ocean models to wind forcing in realistic domains (Takano, 1975; Cox, 1975). At that
             time the resolution was necessarily coarse (order 2'), the integration periods were
             short, and horizontal eddy parameterizations were over-dissipative.         Nevertheless
             many of the essential features of the circulation were captured, such as western
             boundary currents and the Antarctic Circumpolar Current.


                     Bryan et al. (1975), using essentially the same model as Cox (1975), published
             results from the first global coupled ocean-atmosphere model. While the scope of the
             study was limited the oceanic component did reproduce reasonable salinity
             distributions, such as the observed Atlantic-Pacific difference, and a credible dynamic
             circulation and thermohaline meridional overturning. The complexity of this study led
             Bryan and Lewis (1979) to examine and analyse the oceanic component in more detail,
             enabling somewhat finer resolution and seasonal forcing. The Bryan and Lewis (1979)
             study remained the benchmark for stand-alone global ocean simulations for many
             years (see also Bryan, 1979, Bryan, 1982a,b). The model was forced by Hellerman
             (1967) winds, and surface temperature and salinity were relaxed toward Levitus and


                                                      -42-









                  Oort's (1977) climatology. The integrations continued to near-equilibrium.            The
                  observed and computed potential temperature and salinity cross-sections were in
                  reasonable accord with observations although the thermocline was both too deep and
                  too thick, and the deep water too warm (Bryan, 1979). This work introduced improved
                  techniques for analysing oceanic simulations and assessing their sensitivity to
                  parameterizations and forcing. For example, Bryan and Lewis (1979) showed that the
                  thermocline depth is directly related to the strength of the wind forcing and to the
                  eddy mixing parameterizations.         The results demonstrated the importance of
                  meridional oceanic cells (akin to their atmospheric cell counterparts) in transporting
                  heat toward the poles and transferring (relatively warm) surface water from the
                  summer hemisphere to the winter hemisphere (Bryan, 1982b). The poleward heat flux
                  estimates were consistent with independent estimates from atmospheric and oceanic
                  data.



                           Several other groups were beginning to implement global ocean models at
                  ab out this time, mainly with a view toward coupled ocean-atmosphere-ice systems.
                  Washington et al. (1980) and Meehl et al. (1982) developed a coarse resolution ocean
                  model (5' by 4 levels) and tested its sensitivity to the magnitude of the vertical mixing
                  and horizontal diffusion. parameterizations, and to the wind forcing. Meehl et al.
                  (1982) pointed out the importance of seasonally varying surface wind-stress and heat
                  flux conditions, particularly for the oceanic heat storage and transport. Han (1984)
                  presented a numerical world ocean general circulation model, forced by fluxes derived
                  from the Esbensen and Kushnir (1981) climatology, and obtained realistic temperature
                  and salinity fields. An over-strong equatorial upwelling led to anomalous cooling in
                  the equatorial Pacific, and limitations on resolution resulted in too coarse western
                  boundary currents (Han-et al., 1985). Semtner (1984, 1986a,b) assessed developments
                  in ocean circulation modelling, including numerical methods and tuning of non-eddy
                  resolving models. Cox (1984) presented an updated version of the GFDL code for use
                  by the ocean modelling community, an important milestone in the development and
                  wider application of OG-CMs in climate research.











                                                            -43-









              6.1.2 Current OGCM activity


              In discussing the current state there are several distinct strands to the research
              activity that warrant attention. First, with the growing concern in climate and climate
              change, OGCMs have been deployed in a range of coupled atmosphere-ocean(-
              cryosphere) configurations in an attempt to understand the coupled climate system
              and diagnose its sensitivity to anthropogenic forcing. The range of models employed in
              these studies spans all constituents of the ocean system as defined in Section 1 but
              there is good reason to believe that the deep water component is especially significant.
              Though the heat capacity of the upper three metres of the ocean is equivalent to that
              of the entire atmosphere, it is the ability to sequester this heat in the deep ocean and
              transport it to higher latitudes which is the key to the coupled atmosphere-ocean
              system (Bryan et al., 1988). Another approach has been to use the increased computer
              power and improved numerical formulations to increase global model resolution and
              enable explicit representation of mesoscale eddies in ocean-only simulation
              experiments. This activity is more closely aligned with the aims of WOCE, furthering
              our understanding of the world ocean circulation and producing more realistic
              simulations of the velocity, temperature, salinity and other constituent fields of the
              circulation. Finally, there has been important progress in th e application of OGCMs to
              tracer and biogeochemical problems, once more encouraged by the growing interest in
              climate and climate change. For the OOS this is an area of special interest. These
              different aspects will be covered in more detail in the following sub-sections.




              6.1.3 Ocean only simulations


              Together with the application of OGCMs in coupled climate simulations there has been
              considerable recent activity in prescribed surface forcing (ocean-only) experiments,
              mostly associated with WOCE. A large part of this growth, particularly within the
              international ocean modelling community, can be directly attributed to the ready
              availability of the GFDL ocean model code (Cox, 1984) and the continuing free
              exchange of enhancements and experiences with that code (now represented by the
              Modular Ocean Model (MOM) version of the GFDL code; Pacanowski et al., 1991).
              Nevertheless it is important that a variety of models be used if only to ensure that
              model intercomparisons are meaningful and not simply a self-congratulatory exercise.


                                                       -44-










                  The Semtner and Chervin (1988), Hasselmann (1982) / Maier-Reimer et al. (1982) and
                  Haidvogel et al. (1990) models are good examples of such diversity.


                           The following discussion is not a comprehensive account of world ocean
                  modelling but rather an attempt to convey the scope and capabilities of the most
                  advanced models in use today, and to give an impression of the range of models being
                  used. Yet again, it is important to emphasise that issues discussed here transcend the
                  artificial boundaries imposed between the different oceanic components, and that
                  considerations in the surface, seasonal layer and warm ocean components, and in
                  particular research on improved parameterizations, will impact directly on the world
                  ocean models. The canonical world ocean model for OOS in the near-term will likely
                  be an amalgam of the following models combined with new initiatives in
                  parameterization, etc.


                  (a) The Community Modelling Effort for WOCE
                          This Community Modelling Effort (CME) aims to "design and execute a series
                  of baseline calculations of the wind- and thermohaline-driven, large-scale ocean
                  circulation, to make comparisons of these simulations with observations, and to
                  evaluate the performance of the models and identify needed improvements." While
                  CME is basically a response to Goal 1 of WOCE, the experience of CME could also be
                  viewed as a baseline modelling study for the OOS. The focus is on basin-scale, eddy
                  resolving models of the wind- and thermohaline-driven ocean circulation. The first
                  experiment is a simulation of the North Atlantic, including realistic geometry and
                  topography, salinity, and seasonally varying wind and thermohaline forcing (innovative
                  in the context of eddy-resolving models).


                          The Bryan-Semtner-Cox model is being incorporated in both medium (1
                  degree, 20 levels) and fine (1/3 degree, 30 levels) resolution configurations. The model
                  spans 15'S to 650N, including the Gulf of Mexico and Caribbean Sea (but excluding the
                  Mediterranean), uses rotated isopycnal coordinates (at low resolution) and the Camp
                  and Elsberry (1978) bulk mixed layer (F. Bryan and Holland, 1988, 1989; Jochens,
                  1990; Spall, 1990). This model will enable a thorough and detailed investigation of the
                  relative roles of the mean flow and eddy fields in the circulation, particularly in
                  respect of thermohaline balances, heat transport and thermocline ventilation.
                  Preliminary results show realistic spatial and temporal surface temperature patterns,


                                                           -45-










              vigorous eddy activity and sharp frontal features, a seasonally varying equatorial
              circulation with "21-day wave" instabilities, and a realistic flow field. Spall (1990)
              compared the model results with data from the Canary Basin and found the model
              introduced fine horizontal and vertical features which were not present in the initial
              condition but which were in accord with known characteristics of that region. In some
              cases the model impact was negative, particularly in regions influenced by the
              Mediterranean salty outflow and in respect of the eddy kinetic energy.


                       The sensitivity of the simulations to the resolution, surface forcing, various
              parameterizations and the open boundary conditions will be the subject of detailed
              analysis and further experimentation. Of the various controlling factors the surface
              forcing would appear to be among the most poorly defined. The Hellerman and
              Rosenstein (1983) winds, while perhaps adequate for low resolution models of the
              1980's, have neither the accuracy (Harrison, 1989b) nor the resolution required by such
              models. Furthermore the prescription for surface fluxes of heat and freshwater seem
              badly mismatched with the sophistication of the CME and like models.


                       The experiences of "community" modelling at this scale provide an ideal
              prototype for modelling within the context of the OOS. In addition to the experiments
              mentioned above several very fine resolution runs with improved northern boundary
              conditions, different friction coefficients, and different forcing are being run at the IfM
              Kiel (136ning et al., 1991). The design, running and analysis of such an experiment is
              not a trivial exercise and requires careful planning and coordination, as well as a will
              to communicate and combine resources over the international ocean modelling
              community. Even at basin scales, the resource requirements and data handling far
              exceed the capacity of any individual or group.          For a global model at similar
              resolution, with data assimilation and quality control facilities, the requirement may
              be an order of magnitude larger.


                      The CME has provided considerable experience with the management,
              archiving, analysing, and presentation of results from such a model.               As any
              operational meteorological centre will no doubt testify, the management and archiving
              of model integrations and output requires an enormous contribution of time and
              manpower. The shear volume of numbers produced in such an exercise precludes



                                                         -46-









                 traditional methods of analysis. There is a need for new techniques for visualisation
                 and synthesis of results for human and machine consumption.


                 (b) The Fine Resolution Antarctic Model: FRAM
                          FRAM is an initiative by United Kingdom scientists to address some of the
                 problems identified in Core Project 2 of WOCE (WCRP, 1988b) and, like CME, is built
                 aro und a "community" approach to the design, implementation and analysis of the
                 model. FRAM is a primitive equation model of the Southern Ocean from 24'S to the
                 Antarctic continent and is developed from the Cox (1984) version of the GFDL code.
                 The horizontal resolution is 1/2' longitude by 1/40 latitude (thus enabling explicit
                 resolution of eddies), and there are 32 levels in the vertical with high-quality
                 representation of bottom topography. The surface forcing strategy is similar to that
                 used in CME (FRAM intends to incorporate Esbensen and Kushnir, 1981, flux
                 estimates). Further details and background are given in Killworth and Rowe (1987),
                 WOCE (1988a), Killworth (1989), de Cuevas (1990), and The FRAM Group (1991).


                          Unlike the CME, the FRAM project has limited prior experience on which it
                 can draw and is concentrating on a region which is relatively poorly observed. The
                 exciting aspect is that the results from the model are potentially more valuable since
                 they could bridge a large gap in our understanding of the World Ocean circulation. A
                 substantial proportion of the initial effort was devoted to testing the model,
                 particularly in respect of topographic effects. The model is initially run in a robust
                 diagnostic mode for several years, rela5dng toward Levitus' (1982) annual mean fields.
                 Surface wind forcing is then (gradually) introduced, first as an annual mean then with
                 seasonal variations.    The northern boundary is open (Stevens, 1991).           Sharply
                 changing bathymetry and barotropic flow-bathymetry interactions can lead to
                 unacceptably large errors in the flow field (Killworth, 1989). With careful attention to
                 the smoothness of the bathymetry and to the spin-up strategy realistic flow and
                 thermohaline circulation patterns were achieved. However it is also clear from direct
                 observations that the fine detailed structure of the bathymetry (particularly trenches
                 and passages) is highly correlated with the deep current structure and without this
                 structure important currents, particularly those transporting Antarctic Bottom Water
                 away from the Antarctic continent, will be missing from the model. The degree to
                 which this omitted detail impacts the general circulation has not been determined.



                                                          -47-










              Sensitivity to various sub-grid scale parameterizations and to the open and surface
              boundary conditions is also yet to be determined.


                       The Antarctic Circumpolar Current (ACC) and its interaction with major
              topographic features such as the Kerguelen-Gaussberg Ridge, Campbell Plateau, the
              mid-ocean ridge of the South Pacific and Drake passage are all realistically
              represented. The main regions of eddy activity are in the Agulhas Current region and
              along the path of the ACC (there is also evidence of a filament-like structure to the
              ACC). With seasonal forcing there has been an increase in eddy intensity and the
              ACC transport through Drake Passage is oscillating around 160 Sv. The results from
              these integrations have recently been made available in atlas form (Webb et al., 1991).


                       Like CME, FRAM is extremely demanding of human and machine resources
              and could not be run outside a "community" ocean modelling environment. Key
              elements of the continuing development are the specification of surface wind and
              thermolialine forcing as FRAM is particularly handicapped by the poor state of
              knowledge of these fields, particularly during winter-early spring. The northern (open)
              boundary condition (Stevens, 1991) and interfacing witha snow and ice model are also
              high priorities. Considerable effort has already been expended on developing analysis
              and visualisation tools.



              (c) A global eddy-resolving model
                       The development and implementation of the Senitner and Chervin (1988)
              model (hereafter SC88) is significant for its design (specifically for modern
              supercomputers) and for being the first global (almost) eddy-resolving study (resolution
              of 112 x 112 globally. The eddy-rich structures in the western boundary currents and
              circumpolar regions, and the suggestion of a Pacific-Indian-Atlantic conduit in NADW
              formation are just two facets among many which warrant special consideration. The
              model derived from the Senitner (1974) and Cox (1984) versions of the GFDL code.
              Senitner (1986b) described a modified formulation of this code and provided a
              background to the logical development of the SC88 code. The code introduced several
              numerical enhancements not present in earlier versions (Chervin and Semtner, 1988).


                       At the present time the fastest computing machines available are still not
              capable of integrating a global eddy resolving model from an isothermal state of rest to


                                                        -48-









                   statistical equilibrium (under annual mean forcing) without considerable assistance in
                   the spin-up phase, even in the highly optimised form of SC88. A completely prognostic
                   primitive equation eddy-resolving ocean model, spun up with seasonal wind and
                   thermohaline forcing appears to be several years away. The strategy employed in
                   SC88 is similar to that of CME and FRAM, except that the grid is fixed through all
                   phases of the integration. The first 4 years of integration relax the global ocean
                   temperature and salinity toward Levitus' (1982) annual mean fields (restoring time 1
                   year), while at the same time gradually introducing Hellerman and Rosenstein (1983)
                   wind forcing. Laplacian horizontal mixing is used for both the momentum and
                   thermohaline components. The relaxation time scale is then reduced to three years for
                   the next 6 years of integration, and then removed altogether for the upper ocean
                   (except at the surface) for a further 10 years of integration. At this point the upper
                   ocean is supposed to have freely adjusted on decadal time scales (the thermocline
                   circulation in the parlance of this paper), but the deep ocean is still tied to climatology.
                   Finally, in consideration of the genesis of an eddy field, the Laplacian friction is
                   replaced by a scale-selective biharmonic form which allows the eddy field to reach
                   realistic energy levels (Semtner and Mintz, 1987). This configuration is integrated
                   from year 18 of the non-eddy field run until a quasi-statistical equilibrium has been
                   achieved. Like the CME and FRAM projects, SC88 required considerable human and
                   machine resources and broke new ground in the numerical methodology and in the
                   analysis and presentation of results.       The primary fields together with relevant.
                   statistics are archived every 3 days. The saving and retrieval of such large quantities
                   of data poses its own particular problems, a consideration which should not be over-
                   looked in the design of an OOS.


                           SC88 successfully provided a global map of the mean and eddy circulations
                   which is in broad agreement with existing observations (e.g. altimeter data, Cheney et
                   al., 1983). The model provided snapshots of the variability within a suite of western
                   boundary current regimes, providing unprecedented -scope for detailed study of the
                   interconnectivity of the western boundary current regimes. The most interesting
                   result was the apparent confirmation of the role played by the Indian Ocean as a
                   conduit for water travelling between the Pacific and Atlantic Oceans. Semtner and
                   Chervin (1990) also used the model to study the possible effects of mesoscale and
                   seasonal variability on acoustic travel times in the Munk and Forbes (1989) proposed
                   acoustic technique for measuring global ocean warming. Their results suggested the


                                                              -49-









             warming signal could be delineated from the noise of such variability. The scope for
             studying patterns of variability, from short through basin to global scales (e.g. inter-
             basin  gyre connectivity and cross equatorial flow), provides both challenge and
             encouragement in the design of an OOS. Semtner (1989) mentioned some of the
             aspects which are currently under investigation, and anticipates that a global 1/6
             degree 40-level ocean model might be feasible toward the end of WOCE.


             (d) A model for global water mass formation
                      Cox (1989) concentrated on a different aspect of the global ocean circulation -
             the processes by which the major Water masses of the World Ocean are formed. The
             Cox (1984) model is -employed once again, but with an idealised representation of the
             major ocean basins and topographic features. As such the experiments were not
             simulations of the World Ocean circulation, but served as a laboratory for testing the
             importance of different elements of the system, namely the presence of a circumpolar
             current in the Southern Ocean, the influence of wind-driving versus thermohaline
             driving, and the influence of a source of salty (Mediterranean) water in the North
             Atlantic. Cox also introduced a novel method for determining mixing ratios (of
             different water masses) in the model ocean via the introduction of multiple passive

             tracers.



                      The results showed that a thermohaline driven circulation without a

             circumpolar current or North Atlantic Deep Water source could only capture the
             coarsest aspects of the observed temperature and salinity patterns.          Subantarctic
             water dominated the water mass structure of the world ocean. The opening of Drake
             Passage and subsequent generation of a circumpolar current changed the meridional
             flow in a fundamental way, isolating the waters formed south of the current from the
             rest of the domain, and permitting waters of mid-latitude and northern origins to
             occupy a greater proportion of the domain. Wind-forcing increased the strength of the
             circumpolar flow and further isolated the subantarctic waters.           The simulated
             northward penetration of (intermediate) waters from north of the circumpolar region
             and southward from the North Atlantic and Pacific was also improved. Finally the
             addition of an upper ocean source of salty water in the eastern North Atlantic
             substantially improved the deep saline structure of all basins, and produced an overall
             improvement in the salinity structure.




                                                       -50-










                           This study demonstrated that current models, even in the absence of
                  mesoscale eddies, can capture several of the key water mass formation processes of the
                  world ocean and that the simulated distribution of these masses is consistent with
                  observations. The presence or not of sources for particular water types (and their
                  relative buoyancies) are crucial in determining the ultimate level to which these
                  waters circulate and the relative preponderance of the different masses. Bryan (1991)
                  has showed that for meridional transport of heat the model resolution is not of
                  paramount importance. The implication is that models with resolutions similar to Cox
                  (1989) do give meaningful results on water mass formation which may hold for finer
                  resolution and, hence, are relevant for designing observing systems which hope to
                  capture such aspects.


                  (e) Alternate OGCM configurations
                          The initiatives discussed above all derived from the GFDL model, although
                  each included enhancements and modifications to both the physics and numerical
                  design. However it is desirable to maintain a variety in the approach to any modelling
                  problem, no matter how well credentialled a particular model may be. Two of these
                  alternate approaches are mentioned here to demonstrate the possible benefits of
                  diversity for OOS.


                          Hasselmann (1982) proposed that the conventional primitive equation model
                  could be made more efficient by filtering out faster moving gravity waves using the
                  quasi-geostrophic approximation.      The circulation is divided into a barotropic
                  component with associated surface displacement, and a baroclinic component which
                  includes prognostic equations for temperature, salinity and other tracer fields of
                  interest. On the time scales of ocean climate the flow is in approximate geostrophic
                  balance and the depth-averaged flow responds near-instantaneously to surface stress
                  and thermodynamic forcing. The model is closed through the addition of frictional
                  layers at solid boundaries and at the equator (additional terms are required in the
                  vicinity of the equator). Additional detail is given in Maier-Reimer et al. (1982). The
                  model has been employed in a variety of climate-related situations (e.g. Maier-Reimer
                  and Hasselmann, 1987; Cubasch, 1989).


                          Haidvogel et al. (1990) (see also Haidvogel, 1990) proposed a semi-spectral
                  primitive equation ocean circulation model.       The equations of motion are the


                                                           _51-










              hydrostatic primitive equations, much as in the GFDL model, with an explicit Price et
              al. (1986) style mixed layer at the surface and options for periodic or closed
              boundaries, as well as various forms of frictional dissipation. Of particular interest are
              the application of a bathymetry-following (sigma) coordinate system in the vertical,
              and a horizontal orthogonal curvilinear coordinate system for improved handling of
              lateral boundaries. The vertical dependence of the variables are expressed in terms of
              a continuously varying set of orthogonal structure functions (Chebyshev polynomials)
              which offer better resolution near the top and bottom surfaces and improved accuracy.
              This resolution may not always be placed to best advantage, for example, when dealing
              with subsurface jets. Haidvogel (1990) discussed a variety of applications. To date,
              the model has not been applied in global or basin scale ocean climate studies but this
              should be feasible. The improved accuracy and efficiency of the Haidvogel et al. (1990)
              semi-spectral model offers a viable alternative for situations where the resolution of
              vertical mixing and advection is of paramount importance.




              6.1A OGCMs in a coupled model environment


              The state of coupled climate modelling has been the subject of several reviews (e.g.
              Schlesinger and Mitchell, 1987, Meehl, 1990) and was a principal concern of the recent
              IPCC scientific assessment of climate change (e.g. Cubasch and Cess, 1990 and
              Bretherton et al., 1990). The principal concern here is an assessment of the current
              state-of-the-art systems with respect to simulations of the deep ocean circulation. The
              first coupled experiments for doubled carbon dioxide scenarios employed slab or mixed
              layer oceans (see Cubasch and Cess, 1990). Bryan et al. (1988) were the first to
              suggest that such models may not be capturing key interactions between the upper
              and deep ocean, principally the formation of deep and intermediate waters and the
              consequent transport of heat and salt. The obvious solution was to incorporate an
              OGCM, but resources were to provide a tight constraint on such remedies.


                      In contrast to the pioneering study of Bryan et al. (1975), where the aim was a
              realistic simulation of present climatic conditions, most modem coupled model studies
              have been concerned with issues of climate change; the lack of realism in the control
              simulation was only important in so far as it contributed to the uncertainty of the
              predictions. It soon became apparent that deficiencies in the oceanic and atmospheric


                                                        -52-










                   components had an immediate impact on the way such integrations reached
                   equilibrium. The separate components differed in their simulations of the surface heat
                   and freshwater fluxes, and in turn differed. from the "best" observed estimates, so that
                   the coupled system tended toward an unrealistic climate.          The usual solution (if
                   indeed a solution is needed; Meehl, 1990) was to adjust the surface exchanges to
                   account for this mismatch, usually referred to as "flux correction" (Sausen et al., 1988)
                   or "flux adjustment" (Manabe and Stouffer, 1988). The root cause of the ocean model -
                   atmosphere model mismatch remains unclear, but the likely candidates on the oceanic
                   side are poor resolution, uncertainty in eddy parameterizations, poor upper ocean
                   physics, and poor representation of the deep thermolialine field (i.e. the deep
                   recirculation of heat, salt, etc.).


                           The OGCMs employed in coupled climate studies are generally similar to
                   those which were being used in the late 1970's and early 80's for ocean-alone
                   simulations (e.g. Han, 1984; Meehl et al., 1982; Bryan et al., 1975; Bryan and Lewis,
                   1979). Running in coupled mode (particularly in synchronous mode) is extremely
                   demanding in terms of resources, the major impositions being restricted horizontal and
                   vertical resolution and limited range of integrations (equilibrium may not always be
                   reached, e.g. Schlesinger and Jiang, 1988, and important seasonal effects must be
                   excluded, e.g. Bryan et al., 1988). Horizontal resolution is typically around 4' and the
                   number of vertical levels ranges from the bare minimum of two (Gates and Potter,
                   1990), through coarse resolution (four as in Washington and Meehl, 1989) to
                   intermediate resolution (twelve as in Manabe et al., 1990). We will assume the latter
                   two studies are representative of OGCMs currently incorporated in coupled
                   simulations, and use the zonally averaged oceanic temperature and salinity structures
                   from these studies as a guide to the performance.


                           SSTs are typically too cool in low latitudes and too warm in high latitudes.
                   The former deficiency is partly attributable, either directly or indirectly, to the coarse
                   horizontal and vertical resolution whereby the unique meridional structure of the
                   equatorial zone (e.g. the equatorial wave guide and eastern basin upwelhng) is only
                   partially represented. The crude representation of vertical eddy mixing is also a
                   factor. This is consistent with the relative performance of the Washington and Meehl
                   (1989) (their Fig. 10) and Manabe et al. (1990) (their Fig. 3) models. The relatively
                   warm high latitude SSTs can be traced to exaggerated transport of surface heat,


                                                            -53-










              particularly in high southern latitudes, and the crude representation of the seasonally
              varying thermohaline circulation. The vertical distribution of temperature is usually
              characterised by a thermocline which is both too deep and broad (as in ocean-only
              models), and by a deep ocean that is far too warm. Manabe et al. (1990) reported deep
              equilibrium temperatures of 40C which they attributed to lack of wintertime salinity-
              forced deep convection in their model (see also England, 1992). Washington and Meehl
              (1989) reported much better correspondence with observed fields (their Fig. 12), but
              this may be due in part to their initialisation. and integration strategy (e.g., the <2'C
              water in their computed fields does not appear to be ventilated at high latitudes thus
              suggesting it may be left over from the initial state). The adequacy of the salinity
              simulations is difficult to gauge due to the influence of surface water flux adjustment
              (Manabe et al., 1990) and initial conditions (Washington and Meehl, 1989). Both
              simulations capture some of the observed structure, e.g. the Arctic halocline and the
              subtropical surface maxima, but fail to capture more subtle features such as the
              salinity minimum signature of Antarctic Intermediate Water and the deep penetration
              of salty water near 40'N. The root cause of these deficiencies is not immediately clear
              but poor physics and poor resolution, along with the lack of a seasonal cycle, appear
              prime candidates. The report of the first session of the WCRP Steering Group on
              Global Climate Modelling (WCRP, 1991b) discussed the present state of coupled
              modelling and issues which should be addressed.


                      The degradation of surface fields in coupled models is significant and, not
              surprisingly in view of the deterministic nature of the oceanic component, does lead to
              abnormal excursions in the equilibrium climate. For the time being "flux correction"
              techniques are the only recourse for climate sensitivity experiments where the
              equilibrium climatic state is required to be in reasonable accord with observations.
              For global climate modelling it is clearly important to correctly simulate the
              ventilation of the sub-tropical gyres and the deep circulation of the sub-polar gyres.
              These processes, along with heat transport by the boundary currents and unresolved
              deep topographic transports are the key to realistic coupled climate model simulations.
              One encouraging aspect of the model results published thus far is the apparent natural
              formation of variability at interannual and decadal time scales. On the debit side, the
              treatment of cryospheric interactions in coupled models appears far from satisfactory
              (WCRP, 1991b), and the oceanic seasonal cycle is yet to be captured within any skill.



                                                       -54-









                  6.1.5 Modelling the tran port and storage of tracers


                  The aim here is to present a brief overview of ocean modelling in the context of deep
                  circulation of geochemical tracers. The present discussion focuses on the global
                  transport of ocean constituents through the combined effect of upper ocean (mixing)
                  and deep ocean (advective, convective) processes. Sarmiento and Toggweiler (1984a,b)
                  pointed out that 75% of the ocean volume, principally those waters of the cold water
                  sphere, are ventilated through about 4% of the ocean surface, mainly at high latitudes.
                  It is this renewal of the deep water sphere at times scales of the order 1000 years that
                  is the principal focus of large-scale tracer models.


                          There are basically two approaches to the prognostic problem (note that
                  inverse theory is also widely used in this context). One is to treat the ocean as a
                  collection of boxes, as in Sarmiento and Toggweiler (1984a,b) and Broecker and Peng
                  (1986, 1987), with the exchange between boxes being determined by prescribed
                  diffusion and advection rates. The other is to treat carbon dioxide, tritium and other
                  non-reactive constituents as passive tracers and adopt a prognostic dynamic ocean
                  model (e.g. Holland, 1971; Bryan and Lewis, 1979), coupled with an air-sea exchange
                  model, to determine the advection, diffusion and convection rates.


                          Sarmiento and Bryan (1982) and Sarmiento (1983) performed a bomb-tritium
                  simulation in the North Atlantic using the GFDL OGCM. Seasonal convection was
                  found to be as important as downward advection and diffiision in determining tracer
                  penetration. Indeed the effective "diffusion" of the model is dominated by advective
                  and convective contributions rather than the prescribed vertical tracer diffusion. Such
                  a complex behaviour cannot easily be captured in simpler box models with prescribed
                  vertical mixing rates and no thermodynamic circulation (e.g. Oescheger et al., 1975).
                  Sarmiento and Toggweiler (1984b) pointed out that if the box-model approach is to
                  work for the deep ocean then it is essential that the deep water be ventilated
                  appropriately through a high latitude box, thus making it possible to form deep water
                  with the appropriate characteristics. The implications for the carbon cycle are that an
                  increase in high latitude carbon productivity coupled with a reduced thermohaline
                  overturning could lead to significant changes in atmospheric C02 concentrations, such
                  as occurred between the last glacial and interglacial periods.




                                                           _55-










                      Bryan and Sarmiento (1985) made a similar point based on perturbations to
              the surface temperature and passive tracer forcing of OGCMs. They note that the
              effective vertical penetration of surface tracer or heat perturbations is over five times
              that by prescribed vertical diffusion. This is due to the effect of deep convection at
              high latitudes, the implication being that vertical mixing rates can be quite high over
              localised regions, and that any change in the high latitude overturning rate will have a
              substantial effect on the net downward diffusion of tracers.


                      A related issue is the possibility of positive and negative feedbacks in deep
              water recirculation. Bryan and Sarmiento (1985) argue that warm anomalies at the
              surface would suppress high-latitude convection and reduce ocean ventilation (cooling
              in the case of the ocean), and thus reduce the net heating of the atmosphere. This was
              a key element of the Manabe et al. (1990) and Manabe et al. (1991) coupled studies. A
              similar sensitivity to climate change is at the heart of the suggestions by Broecker et
              al. (1985), Broecker and Denton (1990) and Manabe and Stouffer (1988), among others.
              Increased fluxes of freshwater into the surface water of the North Atlantic (from
              melting glaciers or reduced evaporation/increased precipitation) would reduce the
              surface salinity and suppress deep mixing, while at the same time reducing the upper
              ocean northward flux of warm water; The continents adjacent to the far Northern
              Atlantic would be chilled while mid-latitudes would get warmer.


                      Maier-Reimer and Hasselmann (1987) modelled the levels of inorganic carbon
              in the ocean as a passive tracer advected by the ocean circulation. The model carbon
              cycle is closed at the surface by a one-layer diffusive (motionless) atmosphere which is
              interfaced to the ocean using a standard C02 flux relation and a chemically active
              mixed layer. This model excludes biological sources and sinks. The model reproduces
              the broad pattern of C02 distribution in the ocean but, due to the absence of a
              biological pump, the surface PC02 may be underestimated in upwelling regions by
              about a factor of 1.5 (e.g., compared with Bacastow and Maier-Reimer, 1991).
              Comparisons with other models suggested the response function for C02 impulses
              could be captured with relatively simple box models. They also suggested that
              determination of future C02 concentrations would depend on estimations of the
              detailed time history of the oceanic C02 uptake since this effects the ability of the
              ocean to sequesterC02 in the deep.




                                                        -56-










                            Toggweiler et al. (1989a,b) discussed simulations of radiocarbon distribution in
                    a coarse-resolution world ocean model (similar to the Manabe and Stouffer, 1988
                    model). Toggweiler et al. (1989a,b) recognised the deficiencies of such models and the
                    adverse effects they might have on inferences drawn from the model. The five
                    experiments examined robust diagnostic methods (e.g. Sarmiento and Bryan, 1982)
                    versus purely prognostic simulations, and the effect of various surface gas exchange
                    coefficients. Biological production was ignored. Toggweiler et al. (1989a) examined the
                    steady state (order 2000 years of integration) pre-bomb distributions. The model
                    reproduced the mid-depth `C minimum observed in the North Pacific and the strong
                    front at 45'S. In the Atlantic penetration of relatively old water from the Antarctic
                    produces an unrealistic layering of the distribution. Spatial variation in the surface
                    gas exchange rates did not influence the deepwater radiocarbon values. Ventilation in
                    the circumpolar region is controlled by the "Deacon cell" of the model (the lack of such
                    a cell in the North Pacific and Indian basins restricted the ventilation there). They
                    found a fully prognostic model gave better simulations than semi-diagnostic models in
                    spite of the clearly poorer interior temperature and salinity simulation, principally
                    because the observations prevented significant overturning and deep ventilation.


                            In Toggweiler et al. (1989b) a pulse of `C is added to the atmosphere, using
                    Toggweiler et ar. (1989a) as the initial condition, and used to follow the post-bomb     14C
                    distributions. The model successfully reproduced the GEOSECS inventories (Broecker
                    et al., 1985), but predicted a significantly different pattern of 14C uptake in the decade
                    after GEOSECS. The post-GEOSECS buildup is confined to the sub-thermocline
                    layers of the North Atlantic and the lower thermocline of the South Atlantic, and to
                    2000m  and above in the circumpolar region. Subantarctic Mode Water formation is a
                    key process in carrying `C into the thermoclines of the southern hemisphere, but the
                    model has a restricted domain for mode water f6rmation and so fails to ventilate in the
                    appropriate regions. The movement of bomb `C into the deep circumpolar waters and
                    the deep North Atlantic is too slow. Sarmiento et al. (1991) have argued that tracers
                    provide an important independent validation of the water mass and transport
                    realisations, from OGCMs and are useful for isolating weaknesses in the model
                    formulation. For example, in the Sarmiento et al. (1991) model. the bomb-produced
                    radiocarbon inventory is too low compared to observations, so they inferred their
                    estimate of the oceanic uptake of carbon was probably a lower limit to the true value.



                                                               57









             6.1.6 Two- and three-dimensional physical-biological models


             The chemical models discussed above do include some biological processes but usually
             in a simplistic form. Multidisciplinary programs such as JGOFS have spurred the
             development and testing of spatial models with both physical and complex biological
             systems. Fasham et al. (1990) coupled the OGCM used by Sarmiento et al. (1988) with
             a 7-compartment biological model. This general plankton food web model has been
             used to investigate a variety of biogeochemical processes in the upper ocean
             (Toggweiler, 1990). These models are extremely demanding of resources particularly if
             the biological component is to be kept realistic.       Realistic simulation of general
             production and vertical nutrient flux would also require eddy-resolving physical scales,
             an unrealistic demand given the present resources.


                      In the absence of the resources required to run global-scale physical-biological
             models, regional models have appeared, spurred by availability of biological
             information from remote sensing platforms. For example, Walsh et al. (1988, 1989)
             developed models for the Middle Atlantic Bight and Gulf of Mexico, respectively, while
             the southeastern US continental shelf has been modelled by Hoffman (1988) and
             Ishizaka (1990a,b,c). Specialised basin-scale models have also been developed for
             several regions including the North Atlantic (Wroblewski et al., 19'88; Wroblewski
             1989). The physical models are generally eddy resolving but the interface between the
             physics and biology varies (e.g., no horizontal advection, diffusion or no vertical
             advect"ion). The biological models tend to be somewhat simpler than that employed by
             Fasham et al. (1990) (e.g., 1-3 compartments), focusing on the primary producers and
             the nitrogen pools, with simple closures for losses to respiration, grazing and sinking.


                      Ocean-margin coupling is important since the continental shelves act as a
             buffer zone for the biogeochemistry of the ocean interior. Regional models for these
             zones are in an advanced stage of development but require improved knowledge of the
             forcing and better validation data sets. In the context of using physical-biological
             models within an OOS it is important to remember that parameterizations developed
             for a particular location or scale may not be transferable to other regions. The
             problem of scaling-up knowledge from the small scales to the large scales of an OOS is
             a particularly difficult problem. It is important that the in situ observations be
             gathered to enable such model developments to take place. The 10-year JGOFS time


                                                        58-









                   series is unlikely to be capable of assessing the absolute accuracy of biological models
                   given that biological communities can vary over decadal time scales and that the
                   signal-to-noise ratio of such fields is likely to be small. The processes being considered
                   are extremely complex and not well understood, suggesting that for the near-term, the
                   OOS will need to concentrate on simulation and validation aspects rather than

                   assimilation.




                   6.2 Assimilation and prediction: Inverse modelling


                   The prognostic models discussed above are not well suited for systematic testing of
                   model parameters, model forcing, or for assessing compatibility with observations. In
                   a sense the models have used only part of the information available, namely
                   deterministic equations for the circulation and some estimate of the external forcing of
                   the system. No allowance has been made for errors in these components, and the vast
                   information available from observations has not been used explicitly. A better strategy
                   might be to combine this prognostic skill with data to produce optimal estimates of the
                   oceanic state, together with estimates of the uncertainty in the product and in the
                   constituent parameterizations. For the deep (or global-scale) ocean these techniques
                   are usually referred to as inverse methods, but they are closely related to the
                   variational and control techniques discussed previously under the heading of "data
                   assimilation" (Wunsch, 1989a,b; Ghil and Malanotte-Rizzoli, 1991). In either case the
                   focus is on combining data with models and, for the global deep water sphere, this is a
                   special and subtle problem. The various approaches have been discussed in Anderson
                   and Willebrand (1989), and have been the subject of several recent workshops (e.g.
                   "Inversion of Ocean General Circulation Models", WCRP, 1989d; "Assimilation of
                   Observations in Meteorology and Oceanography", WMO, 1990). Wunsch (1989b) gave
                   a particularly clear account of the inverse approach and the reader is referred there for
                   details and an extensive reference list. The advent of WOCE and advances in

                   computer resources have given added impetus to the application of such techniques to
                   global hydrographic and geochemical data sets. It is perhaps in this latter aspect, the
                   use of non-hydrographic data, that sets the applications discussed here apart from
                   those previously discussed in tropical ocean and mesoscale ocean "weather" prediction.
                   Tracer data provide additional, potentially powerful constraints on the inferred

                   circulations.


                                                            -59-










                      One approach to the problem might be to adapt the existing and new
              techniques of seasonal and ocean weather prediction. Sequential data assimilation, as
              practiced in NAT and lately in NOP, would not seem to be viable for the large space
              and time scales of the deep ocean - even if WOCE were successful on all accounts it
              would hardly provide sufficient data to initialise a model, let alone provide the time-
              space history needed for a synoptic analysis and sequential assimilation. However, if
              the questions being asked were modified, say to concentrate on the annual cycle, then
              some of the methods might become feasible. Thacker et al. (1989) developed an adjoint
              model for the GFDL OGCM which closely followed the computational strategy of the
              forward model. While some aspects of the physics, particularly those which include
              non-linear time-dependent logic (e.g., to convectively adjust or not), do pose significant
              problems, this approach appears to offer considerable scope for studying the sensitivity
              of ocean climate models. The practical application of the adjoint method to real global-
              scale problems has yet to be tested.


                      Inverse methods grew out of the inadequacies of the dynamic method for
              determining ocean circulation (Worthington, 1976; Wunsch, 1989b). The velocity shear
              could be determined from hydrographic data and an equation of state but the method
              did not ensure consistency in space or with the distribution of non-hydrographic data,
              and relied on a poorly known level of no motion for determining the absolute velocity.
              Inverse methods sought to remove these inconsistencies by requiring the circulation to
              simultaneously satisfy a variety of constraints and observed distributions. The vast
              majori ty of inverse models have been applied to the problem of determining the mean
              (steady-state, equilibrium) circulation of the oceans consistent with, but not necessarily
              exactly fitting, various dynamical and conserving principles and observed data. Just
              as in most prognostic OGCMs, some important temporal and spatial variability must
              be represented by appropriate parameterizations until resources are sufficient for
              explicit resolution. The dependency on horizontal parameterisations will presumably
              diminish as resolution is improved but the same may not be true of vertical eddy
              parameterisations. The power of inverse models is in their ability to systematically
              incorporate information from disparate sources, either in the form of observations or
              relational equations, and provide an estimate not only of the oceanic state (as
              represented by the constituent fields) and associated errors, but also of the various
              model parameters (e.g. mixing coefficients).




                                                        -60-









                           The beta spiral method (Schott and Stommel, 1978) required the motion to be
                   geostrophic and for the potential vorticity to be locally balanced. When this relation is
                   combined with the thermal wind equations and conservation of potential density w         *e
                   get a relationship between the spiralling of the lateral velocity vector with depth and
                   the vertical velocity. By applying this equation at many levels we get a formally over-
                   determined inverse system of equations for the unknowns, lateral and vertical velocity
                   at the reference level. A variety of methods are available to solve the system including
                   singular value decomposition. Olbers (1989a,b) and Wunsch (1989b) discussed some of
                   the issues involved.



                           Schott and Stommel (1978), Olbers and Wenzel (1989) and Killworth and Bigg
                   (1988) demonstrated the usefulness of the beta spiral method. Olbers et al. (1985)
                   applied the method to North Atlantic data and, in addition to the determination of the
                   absolute flow field, showed that the method could be used to infer diffusion rates. The
                   method appeared to work well in areas where diffusion was a dominant process in the
                   tracer balance, but the results were less compelling where this was not the case (e.g.
                   mid-gyre in the North Atlantic). Noise in the data may be a problem. It is also not
                   clear how much the pre-processing of data (e.g. to achieve gridded values) affected the
                   determinations of diffusion rates - excessive smoothing may indirectly lead to spurious
                   diffusion rates. The problem of noise may also be dependent on the details of the
                   actual implementation, for example the use of tight and loose constraints within the
                   implementation. Olbers (1989b) and Olbers, and Wenzel (1989) presented similar

                   calculations for the Southern Ocean and these estimates seemed consistent with our

                   knowledge of mixing in this region (large values near the strong circumpolar current).
                   The results were also consistent with the few observations available (from buoys and
                   altimeter data) but it was difficult to quantify the accuracy of the estimates without
                   prior knowledge of the variance in the ingested hydrographic data.


                           As Olbers (1989a,b), Olbers and Wenzel (1989) and McDougall (1988) point
                   out, there are several aspects which require careful attention. The fact that the
                   method is applied locally means that conservation of mass is not assured. This can be
                   rectified by seeking a velocity field which is close to the beta-spiral inverse field but
                   non-divergent, and that takes account of Ekman pumping at the surface and solid-
                   boundary conditions. For the North Atlantic this generally results in a smoother field
                   (Olbers and Wenzel, 1989).      The successful implementation of the method also


                                                            -61-










              (naturally) requires that the lateral velocity vector veer with depth. If it does not,
              then the inverse problem may approach ill-posedness, as manifested by extremely low
              values of the matrix condition value (Olbers et al., 1985). A related problem occurs in
              the Antarctic Circumpolar Current where the zonal velocity varies little with depth.
              The eigenvalue associated with the zonal flow is relatively small thus making the
              determination of the reference zonal velocity sensitive to error. This can be overcome
              by including further conditions on the flow, such as steering by bottom topography
              (Olbers, 1989a,b). McDougall (1988) showed that the beta-spiral method is best
              written for neutral surfaces rather than isopycnal surfaces. Conservation of tracers
              such as isopycnal potential vorticity is achieved naturally on neutral surfaces. Olbers
              (1989a) examined the determination of the diffusion tensor for cartesian, isopycnal and
              neutral surfaces and could not clearly identify a superior orientation of the tensor.
              The fact that these issues have a large bearing on the information derived from the
              method suggests they might well be important for the development of a OOS.


                      Inverse box models have now been applied in a wide variety of problems
              (Wunsch, 1977, 1978; Roemmich, 1980; Pollard, 1983; Fu, 1981, 1986; Wunsch, Hu and
              Grant, 1983; Rintoul, 1988; Schlitzer, 1987, 1988; Bolin et al., 1987). The use of such
              techniques is largely superseding and supplanting the use of traditional dynamic
              methods. It is sensible that diagnosis of the flow field should not simply be a function
              of the hydrographic data, but should use as much model and other dynamic and tracer
              data as is available. The problem then becomes one of fixing the (relative) accuracy of
              these measurements, choosing the appropriate dynamical and tracer constraints (e.g.
              geostrophy, Ekman dynamics, conservation, diffusion), and determining internal and
              external sources and sinks (e.g. wind forcing, biological productivity/consumption). The
              aims include determination of the absolute velocity field, the various property budgets,
              and determination of mixing coefficients. The inverse method provides estimates of
              solution variance and resolution for testing against a priori assumptions, and an
              ability to incorporate acquired knowledge of the fields (mainly statistical; Wunsch,
              1989b). The temperature and salinity data are the dominant determinants for the
              velocity field (through geostrophy) but some recent experiments with non-linear
              inversion methods (Mercier et al., 1989; Mercier, 1989) have shown that the density
              field can also be corrected using direct measurements.





                                                       -62-










                           Nutrients, oxygen, tritium and radiocarbon have all been used in inversion
                  studies, the relative usefulness of the data base components usually being determined
                  by the adequacy of knowledge of the measurement errors (e.g. instrument and
                  sampling problems) and of the sources and sinks for the various components. Nutrient
                  data have proven useful in constraining (lowering the estimated error of) the
                  circulation and the diapycnal mixing rates (e.g. Rintoul, 1988, 1989). The fact that
                  chemical data contain information independent of the physical measurements suggests
                  the inverse determinations of the circulation from properly synthesised chemical and
                  physical data should be superior. The use of transient tracers for determining ocean
                  transport is a more difficult problem. Wunsch (1984) showed that the equatorial
                  upwelling in the Atlantic could be determined (constrained) by radiocarbon data.
                  Wunsch (1987) discussed the regularisation problem for transient tracers. Jenkins
                  (1989) showed that bomb-tritium data alone could only weakly constrain the
                  circulation, but suggested that its use in combination with helium-3 data may prove
                  more powerful. The cause of the problem here is the uncertainty in the surface
                  forcing. M6mery and Wunsch (1990) tested the feasibility of constraining ocean models
                  with transient tracer data. They found that existing tritium observations would only
                  weakly constrain the interior ocean circulation even if it were assumed that
                  atmospheric transfer rates and open boundary conditions were known with reasonable
                  accuracy. Nevertheless they conclude that, with the appropriate tracer and good
                  knowledge of the tracer input function (use a closed domain with good surface
                  information), such techniques should yield useful information on the circulatiori.


                          WCRP (1989d) discussed at some length the issues facing inverse theory
                  practitioners, and these issues are also relevant here. The great advantage of inverse
                  theory, as against traditional methods of analysis (e.g. Gordon et al., 1982; Levitus,
                  1982), is that an estimate of the error in the diagnosed field arises naturally out of the
                  methodology. True, this estimate of the error is dependent on reliable estimation of
                  the variance in the field and knowledge of the spatial and temporal variability. The
                  estimates of the state must be representative of the true equilibrium state of the field
                  (errors of representativeness have been discussed at length in NVVP, e.g. Lorenc, 1986).
                  Wunsch (1989b) made a strong case for pursuing inverse and other estimation methods
                  beyond estimation of the field, to look at the soundness of a priori assumptions
                  through calculation of estimated field variances and covariances.




                                                            -63-










                       The identification of systematic errors, due to forcing error or model
              deficiencies, is also an important issue. WCRP (1989d) discussed the format of data
              being used for inversion, weighing the use of raw data against already gridded
              products (the analogous issue in NWP is the use of super-observations to represent
              several individual pieces of information). For practical reasons gridded data were
              preferred but the  problems with pre-processing, especially smoothing (Olbers, 1989b;
              Rintoul, 1989) were important. A related problem is that of estimating solution
              variance which, for the under-determined problem (no redundancy in the data), usually
              requires a reduction in the effective degrees of freedom through smoothing. Such
              estimates need to evaluated on many realisations of the same situation.


                      For both the practitioners and end users of inverse theory applications the
              sensitivity of the solutions to the minimisation criterion is an important issue. The
              essence of the under-determined inverse problem is that additional information is
              needed to close the problem and this information usually takes the form of a
              minimisation (say diffusion) or smoothness criterion. Ideally this criterion should have
              a sound physical basis.       The inverse model circulation and estimated error
              characteristics, and comparisons with independent data, ultimately attest to the
              appropriateness of the a priori assumptions.


                      Wunsch (1989a) suggested that the inverse modelling community might
              (should) adopt a similar strategy to the Community Modelling Effort and develop a
              compr ehensive model (the North Atlantic was the obvious domain), using state-of-the-
              art computing and numerical techniques to reduce the box size and enable
              consideration of many different fields and constraints. Wunsch suggested that a global
              one degree square, ten level model is feasible, so long as a community-based modelling
              approach was adopted. Even for the steady model this would require considerable
              effort to prepare the data (with error estimates), configure the domain and model, and
              decide on the best solution approach. The WCRP (1989d) workshop recognised the
              value of a variety of approaches to the inversion of oceanic models, and that work
              should now begin on developing suitable time-varying inverse models.









                                                       -64-









                  6.3 OOS network design and quality control for the deep ocean


                          Quality control of deep ocean data is a difficult problem, and is usually
                  accomplished through a mix of subjective and simple objective methods. The cost of
                  gathering deep ocean data makes it crucial that its integrity be assured and that little
                  data is wasted. On the other hand the data base for the deep ocean is sparse in time
                  and space and, since quality control inevitably depends upon previous information, will
                  always operate under extremely difficult circumstances. The accuracy of global-scale
                  ocean models is such that they cannot yet be used to guide quality control of data;
                  However inverse models inevitably incorporate a form of quality control within their
                  formulation since they are in essence a data-fitting model - bad data will fit poorly.
                  We are not aware of any published accounts of the use of inverse models for detecting
                  errors in deep ocean data.


                          Models have been used at various stages in the design of ocean observing
                  networks for the deep ocean. Bretherton et al. (1976) were among the first to look at
                  the oceanographic experiment design problem. Their task was to design a current
                  meter array that would give the best possible map of the mesoscale eddy structure, as
                  part of the Mid-Ocean Dynamics Experiment. Their approach drew on experience in
                  NWP where, with information on the noise and signal levels of the observations (and
                  their spatial coherence), an estimate of the expected error in the mapped field could be
                  obtained, subject to the specification of an objective function (e.g. least rms error in the
                  mapped field). While this approach has proved extremely useful in many different
                  contexts (e.g. Meyers et al., 1991), it remains sensitive to the particular statistical
                  assumptions being used, and is not easy to use to "predict" the best array. Bretherton
                  and McWilliams (1980) did approach the problem from this perspective, using a
                  slightly more general statistical and information theory.


                          The key to such applications is often in the choice of the objective function.
                  Bennett (1985) chose an objective function to characterise acoustic tomography arrays
                  (Barth and Wunsch, 1990 used a similar function). Bennett (1985) assessed the ability
                  of a set of arrays to map a continuous (synthetic) field using inverse techniques. Barth
                  and Wunsch (1990) discussed a similar problem, but carried out the optimisation of the
                  objective function by simulated annealing. They concluded that such techniques, when




                                                            -65-









              carefully applied, would give results superior to those of traditional methods of array
              design.


                      OSSE/OSE-type experiments are often carried out within the context of
              inverse modelling. For example, comparing the circulation obtained with hydrographic
              data alone compared with that obtained by including chemical tracers. The problem is,
              however, in the validation of the experiments. In NVYT, and in prognostic modelling in
              general, validation of the improved/decreased skill of the model is ascertained by
              comparison of forecasts and analyses (i.e. comparison against independent data). The
              very nature of the inverse methodology ensures that additional information, so long as
              it is not biased or systematically in error, will lower the expected error of the ocean
              state simulation. It is often more difficult to judge whether a particular system makes
              a significant impact on the estimation of, say, the lateral velocity field. As Olbers,
              (1989b) pointed out in his inversion study of the Southern Ocean, it is almost
              impossible to quantify the accuracy of estimates for, say, diffusion since little
              independent evidence exits. One approach, as in Killworth and Bigg (1988) is to use
              synthetic data generated from OGCMs. However, as borne out by the experiences of
              FGGE, results from simulated observations do not always translate over to real data
              systems. Nevertheless the data generated by high-resolution modelling projects such
              as CME, FRAM and Semtner and Chervin (1988) do offer opportunities to test a priori
              assumptions and variance estimates on "known" fields; The reduced dependence of
              these models on eddy mixing parameterisations should engender greater confidence in
              the results of such tests.



                      There have been no attempts to use OGCMs directly in the design of
              observational networks. Certainly programs such as FRAM and the CME provide
              many results which may stimulate observational programs, or help define the relative
              roles of particular components within an observation program. But as yet resource
              limitations prevent the repeated experiments which would be necessary to implement
              true OSSE/OSEs.












                                                       -66-










                  6.4 Developments required for the deep ocean component of an OOS


                           The development of models for the deep ocean is inextricably linked to
                  progress in each of the other components - surface forcing, mixed layer physics (for
                  tracers and heating) and ventilation of the thermocline. We should also not overlook
                  the tropical oceans. Research there is concentrating on seasonal and interannual
                  variability, but it is in these regions (e.g. the western Pacific warm pool) where many
                  of the important air-sea interactions are taking place and it may well be that these
                  processes are critical for understanding the global circulation. The failure of coupled
                  models with poor resolution of the equatorial regions is perhaps a forewarning of this
                  key role. On the same theme, the equilibrium climate of the oceans may also be
                  sensitive to the interannual reorganisation of the tropical oceans. For coupled models
                  a severe test of their integrity will be their ability to realistically simulate interannual
                  variability in the tropics. However the first priority must be to successfully produce
                  the observed seasonal cycle, an ability that is presently absent from most coupled

                  models.



                           The primary focus of modelling at this time is on simulations of the current
                  climatic conditions. Computer power has now reached a stage where truly eddy-
                  resolving models of the global ocean may be possible. Both prognostic and inverse
                  models are being used to build a better understanding of the ocean circulation as it
                  now. This is the primary objective of WOCE, but the design and construction           of an
                  OOS must be mindful that this building of knowledge will almost certainly continue
                  well beyond WOCE. Models can play a critical role in determining how best to gather
                  this information, and the modelling community should be encouraged to develop and
                  define this strategy. In future, tried and tested strategies should be in place for
                  evaluating the relative worth of different observing systems in the determination of
                  the global ocean circulation.


                           The ultimate goal is to implement global ocean prediction systems. The
                  confide nce in such predictions will be based partly on the ability of models to
                  reproduce features of the present circulation. Some of the most severe tests will come
                  in biogeochemistry where models will be required to reproduce current patterns of
                  trace gas and carbon exchange, as well as the observed biological and chemical cycles
                  in the ocean (e.g. primary productivity). This will of course require substantial


                                                             -67-










             progress in our understanding of the important non-conservative processes in the
             ocean. While there are many who would argue that credible climate perturbation
             experiments can be performed on less-than-perfect equilibrium climates, there will
             always be just as many doubters who will be suspicious of any prediction based on
             poor initial conditions.    This will be particularly so if we cannot quantify the
             uncertainties in the deterministic and observed components of the prediction system.


                      Finally, as an example of how we might systematically build the components
             of OOS, we examine an initiative within GEV*TEX to investigate the North Atlantic
             water budget (Schmitt and Bryan, 1991, unpublished manuscript).             This is not
             intended as an endorsement of this plan over any other, but rather to illustrate a way
             of developing observing and modelling systems in accord for their mutual benefit.
             Schmitt and Bryan point out the large uncertainties in our present knowledge of the
             net E-P over the North Atlantic - climatological estimates differ by as much as 30
             mm/month. This uncertainty means we are not even sure of the sign of the mean
             annual buoyancy flux. We also know, from historical and contemporary data, that the
             North Atlantic deep circulation can be interrupted by surface salinity anomalies
             (Dickson et al., 1988). We know how to run models with a variety of surface
             freshwater fluxes to test their sensitivity; however without suitable data we cannot be
             sure of the appropriateness of the model parameterizations, of the validity of the
             forcing functions, or whether the predictions are sensible. Schmitt and Bryan     suggest,
             among other things, that simple water mass formation models (e.g. Huang et al., 1991)
             and GCMs under plausible forcings and different mixing parameterizations should be
             used to explore the water budget. Inversion, and other classical methods, should be
             applied to existing data sets to improve current estimates of the net E-P flux inter-
             basin water exchange, and ocean-cryosphere interactions. By developing a better
             understanding of one basin we can better formulate OOS design for the global ocean.
             The testing and development of models in such an experiment is clearly of considerable

             benefit.














                                                       -68-











                      7. References


                      Anderson, D.L.T., and A.M. Moore, 1989: Initialisation of equatorial waves in ocean models. J. Phys.
                           Oceanogr., 19,116-121.
                      Anderson, D.L.T., and J. Willebrand (Eds), 1989: Oceanic Circulation Models: Combining Data and
                           Dynamics. D. Anderson and J. Willebrand, eds., Kluwer Academic, Hingham, Mass., 604 pp.
                      Archer, D., 1990: Modelling pCO, in the upper ocean: a review of relevant physical, chemical, and
                           biological processes. U.S. Department of Energy, Carbon Dioxide Research Program, Report TRO50,
                           61 pp.
                      Arkin, P.A., and J.E. Janowiak, 1991: Analyses of the global distribution of precipitation. Dyn. Atmos.
                           Oceans, 16, 5-16.
                      Arpe, K., 1991: The hydrological cycle in the ECMWF short range forecasts. Dyn. Atmos. Oceans, 16, 33-
                           59.
                      Arpe, K., D. Burridge, and A. Gilchrist, 1988: Comparison of surface stresses and latent heat fluxes over
                           oceans in short range forecasts by ECMVvT, UKMO and NMC. Proceedings, 13th Annual Climate
                           Diagnostics Workshop, Cambridge, Massachusetts.
                      Bacastow, R., and E. Maier-Reimer, 1991: Dissolved organic carbon cycles in modelling oceanic new
                           productions. Global Biogeochem. Cycles, 5, 71-85.
                      Barnett, T.P., 1984: Prediction of El Nifio of 1982-83. Mon. Wea. Rev., 112, 1403-1407.
                      Barnier, B., and J.-Y. Simonot, 1990: Surface heat flux over the North and South Atlantic in 1985-1986
                           from day 1 predictions of the European Center for Medium Range Weather Forecasts. J. Geophys.
                           Res., 95, 13301-13311.
                      Barth,N., and C. Wunsch, 1990: Oceanographic experiment design by simulated annealing. J. Phys.
                           Oceanogr., 20, 1249-1263.
                      Battisti, D.S., and A.C. Hirst, 1989: Interannual variability in a tropical atmosphere-ocean model:
                           Influence of the basic state, ocean geometry and nonlinearity. J. Atmos. Sci., 46, 1687-1712.
                      Bennett, A.F., 1985: Array design by inverse methods. Prog. Oceanogr., 15, 129-156.
                      Bennett, A.F., 1990: Inverse methods for assessing ship-of-opportunity networks and estimating
                           circulation and winds from tropical expendable bathythermograph data. J. Geophys. Res., 95, 16111-
                           16148.
                      Bennett, A.F., and W.P. Budgell, 1987: The Kalman smoother for a linear quasi-geostrophic model of
                           ocean circulation. Dyn. Atmos. Oceans, 13, 219-267.
                      Bleck, R., and D.B. Boudra, 1981: Initial testing of a numerical ocean circulation model using a hybrid
                           (quasi -isopyenic) vertical coordinate. J. Phys. Oceanogr., 11, 755-770.
                      Bleck, R., and D.B. Boudra, 1986: Wind-driven spinup in eddy-resolving ocean models formulated in
                           isopycnic and isobaric coordinates. J. Geophys. Res., 91, 7611-7621.
                      Bleck, R., R. Onken, and J.D. Woods, 1988: A two-dimensional model of mesoscale frontogenesis in the
                           ocean. Q. JI Roy. Met. Soc., 114, 347-371.
                      Blumberg, A.F., and G.L. Mellor, 1987: A description of a three-dimensional coastal ocean circulation
                           model. In: Three-dimensional Coastal Ocean Models, Vol. 4, N. Heaps (Ed.), Amer. Geophys. U., 208
                           PP.
                      Blumenthal, M.B., and M.A. Cane, 1989: Accounting for parameter uncertainties in model verification: An
                           illustration with tropical sea surface temperature. J. Phys. Oceanogr., 19, 815-830.
                      Bogue, N.M., RX Huang, and K. Bryan, 1986: Verification experiments with an isopycnal coordinate
                           ocean model. J. Phys. Oceanogr., 16, 985-990.
                      Bolin, B.A., A. Bjorkstrom, K. Holm6n, and B. Moore, 1987: On inverse methods for combining chemical
                           and physical oceanographic data: A steady state analysis of the Atlantic Ocean. Tech. Rep.,
                           Meteorologiska Institutionen (MISU), Stockholm, Universitet, Arrheniuslaboratoriet, 134 pp.
                      B6ning, C.W., R. Dbscher, and R.G. Budich, 1991: seasonal transport variations in the western subtropical
                           North Atlantic: Experiments with an eddy-resolving model. J. Phys. Oceanogr., 21, 1271-1289.
                      Boris, J.P., and D.L. Book, 1973: Flux-corrected transport. 1: SHASTA, a fluid transport algorithm that
                           works. J. Comput. Phys., 11, 38-69.
                      Boudra, D.B., 1989: Use of isentropic coordinate numerical models in studies of ocean circulation. In:
                           Global Ocean Prediction Systems, INO symposium, Princeton, NJ, July 1989, 45-49.
                      Bourke, W., R. Seaman and K. Puri 1985. Data assimilation. Adv. Geophys., 28B, 123-155.
                      Bretherton, F.P., and J.C. McWilliams, 1980: Estimations from irregular arrays. Rev. Geophys. Space
                           Phys., 18, 789-812.



                                                                        -69-











                 Bretberton, F.P., K. Bryan, and J.D. Woods, 1990: Time-dependent greenhouse-gas-induced climate
                      change. In: Climate Change. The IPCC Scientific Assessment. J.T. Houghton, G.J. Jenkins and J.J.
                      Ephraums, eds, WMO/UNEP, University Press, 173-194.
                 Bretherton, F.P., R.E Davis and C.B. Fandry 1976. A technique for objective analysis and design of
                      oceanographic experiments applied to MODE-73. Deep-Sea Res., 23, 559-582.
                 Broecker, W.S., and G.H. Denton, 1990: What drives glacial cycles? Scientific American, 262, 49-56.
                 Broecker, W.S., and T.-H. Peng, 1986: Carbon cycle 1985: Glacial to interglacial changes in the operation
                      of the global carbon cycle. Radiocarbon, 28, 309-327.
                 Broecker, W.S., and T.-H. Peng, 1987: The role of CaCO, compensation in the glacial to interglacial
                      atmospheric CO, change. Global Biogeochem. Cycles, 1, 15-29.
                 Broecker, W.S., D. Peteet, and D. Rind, 1985: Does the ocean-atmosphere system have more than one
                      stable mode of operation. Nature, 315, 21-26.
                 Bryan, F.O., and W.R. Holland, 1988: Progress in the US-WOCE Ocean Community Modelling Effort.
                      WOCE Newsletter, No. 6, 3-7.
                 Bryan, F.O., and W.R. Holland, 1989: A high-resolution simulation of the wind- and thermolialine-driven
                      circulation in the North Atlantic Ocean. In: Proceedings of the Aha Hulikola Hawaiian Winter
                      Workshop on Parameterizarion of Small-scale Processes, Honolulu, Hawaii, P. Muller and D.
                      Henderson, eds., Hawaii Institute of Geophysics Special Publ., 99-116.
                 Bryan, K., 1963: A numerical investigation of a nonlinear model of a wind-driven ocean. J. Atmos. Sci., 20,
                      594-606.
                 Bryan, K., 1969: A numerical method for the study of the world ocean. J. Comput. Phys., 4, 347-376.
                 Bryan, K., 1979: Models of the World Ocean. Dyn. Atmos. Oceans, 3, 327-338.
                 Bryan, K., 1982a: Poleward heat transport in the oceans: Observations and models. Ann. Rev. Earth
                      Planet. Sci., 10, 15-38.
                 Bryan, K., 1982b: Seasonal variation in meridional overturning and poleward heat transport in the
                      Atlantic and Pacific Oceans: a model study. J. Mar. Res., 40 (supplement), 39-53.
                 Bryan, K., 1991: Poleward heat transport in the ocean. A review of a heirarchy of models of increasing
                      resolution. Tellus, 43 AE, 104-115.
                 Bryan, K., S. Manabe, and R.C. Pacanowski, 1975: A global ocean-atmosphere climate model. Part 11. The
                      oceanic circulation. J. Phys. Oceanogr., 5, 30-46.
                 Bryan, K., and L.J. Lewis, 1979: A water mass model of the World Ocean. J. Geophys. Res., 34, 2503-2517.
                 Bryan, K., and J.L. Sarmiento, 1985: Modeling ocean circulation. In: Issues in Atmospheric and Oceanic
                      Modelling. Part A. Climate Dynamics. S. Manabe, ed., Academic Press, 433-460.
                 Bryan, K., S. Manabe, and M.J. Spelman, 1988: Interhemispheric asymmetry in the transient response of
                      a coupled ocean-atmosphere model to a CO, forcing. J. Phys. Oceanogr., 18, 851-867.
                 Budd, W.F., I.H. Simmonds, and X. Wu, 1990: The physical basis for a dynamic Antarctic sea-ice model for
                      use with an atmospheric GCM. IGS Symposium on Ice-Ocean Dynamics and Mechanics. Hanover,
                      New Hampshire, August 1990.
                 Burridge, D., and A. Gilchrist, 1989: Estimates of surface fluxes from global operational numerical
                      weather prediction schemes. Phil. Trans. R. Soc. Lond. A, 329, 303-315.
                 Busalacchi, A.J., and J.J. O'Brien, 1980: The seasonal variability in the tropical Pacific. J. Phys.
                      Oceanogr., 10, 1929-1952.
                 Busalacchi, A.J., and J.J. O'Brien, 1981: Interannual variability in the equatorial tropical Pacific in the
                      1960's. J. Geophys. Res., 86, 10901-10907.
                 Busalacchi, A.J., K. Takeuchi, and J.J. O'Brien, 1983: Interannual variability of the equatorial Pacific,
                      revisited. J. Geophys. Res., 88, 7551-7562.
                 Camp, N.T., and R.L. Elsberry, 1978: Oceanic thermal response to strong atmospheric forcing, II. The role
                      of one-dimensional processes. J. Phys. Oceanogr., 8, 215-224.
                 Cane, M.A., 1979a: The response of an equatorial ocean to simple wind stress patterns: I. Model
                      formulation and analytical results. J. Mar. Res., 37, 233-252.
                 Cane, M.A., 1979b: The response of an equatorial ocean to simple wind stress patterns: II. Numerical
                      results. J. Mar. Res., 37, 253-299.
                 Cane, M.A., and E.S. Sarachik, 1977: Forced baroclinic ocean motions: 11. The linear equatorial bounded
                      case. J. Mar. Res., 35, 395-432.
                 Cane, M.A., and S.E. Zebiak, 1985: A theory for El Niho and the Southern Oscillation. Science, 228, 1085-
                      1087.
                 Cane, M.A., S.C. Dolan, and S.E. Zebiak, 1986: Experimental forecasts of the 1982/83 El Nifio. Nature,
                      321, 827-832.
                 Carrington, D.J., 1991: Use of wind stresses from operational NVVT models to force an OGCM of the
                      Indian Ocean. Climate Research Technical Note, CRTN 10, Hadley Centre, UKMO, 46 pp.



                                                                    -70-











                       Carrington, D.J., D.L.T. Anderson, and C. Gordon, 1990: A serni -operational OGCM of the Indian Ocean.
                            International Symposium on 'Assimilation of Observations in Meteorology and Oceanography",
                            Clermont-Ferrand, France, 9-13 July, 1990, 427-428.
                       Carter, E.F., and A.R. Robinson 1987. Analysis models for the estimation of oceanic fields. J. Atmos.
                            Oceanic Technol., 4, 49-74.
                       Charnock, H., and S.G.H. Philander (Eds), 1989: The dynamics of the coupled atmosphere and ocean.
                            Proceedings of a Royal Society Discussion Meeting, H. Charnock and S.G.H. Philander, Eds, The
                            Royal Society, London, 315 pp.
                       Chen, D., S.G. Horrigan, and D.-P. Wang, 1988: The late summer vertical nutrient mixing in Long Island
                            Sound. J. Mar. Res., 46, 753-770.
                       Cheney, R.E., J.G. Marsh, and B.D. Beckley, 1983: Global mesoscale variability from Seasat collinear
                            altimeter data. J. Geophys. Res., 88, 4343-4354.
                       Chervin, R.M., and A.J. Semtner, Jr., 1988: An ocean modeling system for the supercomputer
                            architectures of the 1990's. In: Proceedings of the NATO Advanced Research Workshop on Climate-
                            Ocean Interaction, M. Schlesinger, Ed., YJuwer, Dordrecht, Netherlands.
                       Clancey, R.M., 1989: Validation of the Optimum Thermal Interpolation System (OTIS) at FNOC. In:
                            Global Ocean Prediction Systems, INO symposium, Princeton, NJ, July 1989, 71-74.
                       Clancy, R.M., and K.D. Pollack, 1983: A real-time synoptic ocean thermal analysis/forecast system. Prog.
                            Oceanogr., 12, 383-424.
                       Clancey, R.M., K.D. Pollack, and J.A. Cummings, 1986: Technical description of the Optimal Thermal
                            Interpolation System (OTIS) Version 1.0: A model for oceanographic data assimilation. FNOC Tech.
                            Note 422-86-02, Fleet Numerical Oceanography Center, Monterey, CA, 65 pp.
                       Clancey, R.M., P.A. Phoebus, and K.D. Pollack, 1989: Technical description of the Optimum Interpolation
                            System, Version 1: A model for oceanographic data assimilation. Naval Ocean Research and
                            Development Activity, Rep. 240, Stennis Space Center, Mississippi, 28 pp.
                       Clancey, R.M., P.A. Phoebus, and K.D. Pollack, 1990: An operational global-scale ocean thermal analysis
                            system. J. Atmos. Oceanic Technol., 2, 233-254.
                       Coastal Ocean Prediction Systems, 1990: The Coastal Ocean Prediction Systems Program: Understanding
                            and Managing Our Coastal Ocean, Vol. II. Rept. of planning workshop, Univ. New Orleans,
                            November, 1989, 354 pp.
                       Cooper, N.S., 1988: The effect of salinity on tropical ocean models. J. Phys. Oceanogr., 18, 697-707.
                       Cox, M.D., 1975: A baroclinic numerical model of the World Ocean. In: Numerical Models of Ocean
                            Circulation. National Academy of Sciences, Washington, D.C., 107-118.
                       Cox, M.D., 1984: A primitive equation three-dimensional model of the ocean. Tech. Rep. 1, Geophysical
                            Fluid Dynamics Laboratory, NOAA, Princeton University, Princeton, N.J., 250 pp.
                       Cox, M.D., 1985: An eddy resolving numerical model of the ventilated thermocline. J. Phys. Oceanogr., 15,
                            1312-1324.
                       Cox, M.D., 1989: An idealised model of the World Ocean. Part 1. The global-scale water masses. J. Phys.
                            Oceanogr., 19, 1730-1752.
                       Cox, M.D., and K. Bryan, 1984: A numerical model of the ventilated thermocline. J. Phys. Oceanogr., 14,
                            674-687.
                       Cubasch, U., 1989: Coupling a global atmospheric model with a global ocean model using the flux
                            correction method. In: Aspects of Coupling Atmosphere and Ocean Models, R. Sausen, Ed, Rep. No.
                            6., Met. Inst. Univ. Hamburg, Germany.
                       Cubasch, U., and R.D. Cess, 1990: Processes and Modelling. In: Climate Change. The IPCC Scientific
                            Assessment. J.T. Houghton, G.J. Jenkins and J.J. Ephraums, eds, WM0/UNEP, University Press, 69-
                            93.
                       Cummings, J.A., 1990: Water mass empirical orthogonal functions in the NW Atlantic with application to
                            optimal field estimation. Unpubl. manuscript.
                       Daley, R., 1981: Normal mode initialisation. In: Dynamic Meteorology. Data Assimilation Methods. L.
                            Bengtsson, M. Ghil, and E. Kall6n, Eds, Springer-verlag, New York, 77-110.
                       De Cuevas, B.A., 1990: FRAM - the Fine Resolution Antarctic Model. WOCE Newsletter, No. 10, p 14.
                       De May, P., and A.R. Robinson, 1987: Assimilation of altimeter eddy fields into a limited-area quasi-
                            geostrophic model. J. Phys. Oceanogr., 17, 2280-2293.
                       Delcroix, T., and C. Henin, 1989: Mechanisms of subsurface thermal structure and sea surface
                            thermohaline variabilities in the south western tropical Pacific during 1979-85 - a preliminary report.
                            Proceedings of the Western Pacific International Meeting and Workshop on TOGA COARE, Noum6a,
                            New Caledonia, May 1989, 581-590.
                       Derber, J., A. Leetmaa, M. Ji, and R. Shinn, 1990: Oceanic data assimilation at the National
                            Meteorological Center. International Symposium on "Assimilation of Observations in Meteorology and
                            Oceanography", Clermont-Ferrand, France, 9-13 July, 1990, 59-61.


                                                                         -71-











                Derber, J., and A. Rosati, 1989: A global oceanic data assimilation system. J. Phys. Oceanogr., 19, 1333-
                     1347.
                Dickson, R.R., J. Meincke, S. Malmberg, and A Lee, 1988: The "Great Salinity Anomaly" in the northern
                     North Atlantic 1968-82. Prog. Oceanogr., 20, 103-151.
                England, M.E., 1992: On the formation of Antarctic Intermediate and Bottom Water in ocean general
                     circulation models. J. Phys. Oceanogr., submitted.
                Esbenson, S.K., and Y. Kushnir, 1981: The heat budget of the global ocean: An atlas based on estimates
                     from surface marine observations. Rep. 29, Clim. Res. Inst., Oreg. State Univ., Corvallis.
                Evans, G.T., and J.S. Parslow, 1985: A model of annual plankton cycles. Biol. Oceanogr., 3, 327-347.
                Falkowski, P.G., and C.D. Wirick, 1981: A simulation model of the effects of vertical mixing on primary
                     productivity. Mar. Biol., 65, 69-75.
                Fasham, M.J.R., P.M. Holligan, and P.R. Pugh, 1983: The spatial and temporal development of the spring
                     phytoplankton bloom in the Celtic Sea, April 1979. Prog. Oceanogr., 12,87-145.
                Fasham, M.J.R., H.W. Ducklow, and S.M. McKelvie, 1990: A nitrogen-based model of plankton dynamics
                     in the oceanic mixed layer. J. Mar. Res., 48, 591-639.
                Fichefet, Th.., and P. Gaspar, 1988: A model study of upper ocean - sea ice interactions. J. Phys.
                     Oceanogr., 18, 181-195.
                Folland, C.K., D.W. Reynolds, M. Gordon, and D.E. Parker, 1991: The WMO sponsored intercomparison of
                     sea surface temperature data sets for long range forecasting - conclusions and recommendations.
                     Long-Range Forecasting Research Report, No. 14, ICTP/WMO International Technical Conference on
                     Long-Range Weather Forecasting Research, Trieste, Italy, 8-12 April, 1991, 89-92.
                Folland, C.K., D.W. Reynolds, M. Gordon, and D.E. Parker, 1992: A study of six operational sea surface
                     temperature analyses and suggestions for improvement. J. Climate, submitted.
                Frost, BW., 1987: Grazing control of phytoplankton stock on the open subarctic Pacific Ocean: a model
                     assessing the role of mesozooplankton, particularly the large calanoid copepods Neocalanus sp.. Mar.
                     Ecol. Prog. Ser., 39, 49-68.
                Fu, L., 1981: The general circulation and meridional heat transport of the subtropical South Atlantic
                     determined by inverse methods. J. Phys. Oceanogr., 11, 1171-1193.
                Fu, L., 1986: Mass, heat and freshwater fluxes in the South Indian Ocean. J. Phys. Oceanogr., 16, 1683-
                     1693.
                Gandin, L.S., 1988: Complex quality control of meteorological observations. Mon. Wea. Rev., 116, 1137-
                     1156.
                Garwood, 1977: An ocean mixed layer capable of simulating cyclic states. J. Phys. Oceanogr., 7, 455-468.
                Gaspar, P., 1988: Modelling of the seasonal cycle of the upper ocean. J. Phys. Oceanogr., 18, 161-180.
                Gaspar, P., and C.I. Wunsch, 1989: Estimates from altimeter data of baTotropic Rossby waves in the
                     northwestern Atlantic Ocean. J. Phys. Oceanogr., 19, 1821-1844.
                Gaspar, P., J.-C. Andr6, and J.-M. Lefevre, 1990: The determination of the latent and sensible heat fluxes
                     at the sea surface viewed as an inverse problem. J. Geophys. Res,, 95, 16169-16178.
                Gates, W.L., and G.L. Potter, 1990: Simulation of the climate effects of increased CO, with a coupled
                     atmosphere/ocean GCM. Clim. Dyn., to appear.
                Gent, P.R., and J.C. McWilliams, 1984: Balanced models in isentropic coordinates and the shallow water
                     equations. Tellus, 36A, 166-171.
                Ghil, M. 1989: Meteorological data assimilation for oceanographers. Part I: Description and theoretical
                     framework. Dyn. Atmos. Oceans, 13,171-218.
                Ghil, M., and P. Malanotte-Rizzoli, 1991: Data assimilation in meteorology and oceanography. Adv.
                     Geophysics, to appear.
                Gill, A.E., 1982: Atmosphere-Ocean Dynamics. Academic Press, New York, 661 pp.
                Gill, A.E., 1983: An estimation of sea-level and surface-current anomalies during the 1972 El Niho and
                     consequent thermal effects. J. Phys. Oceanogr., 13, 586-606.
                Godfrey, J.S., 1989: A Sverdrup model of the depth-integrated flow for the World Ocean allowing for
                     island circulations. Geophys. Astrophys. Fluid Dynamics, 45, 89-112.
                Godfrey, J.S., and E.J. Lindstrom, 1989: The heat budget of the equatorial western Pacific surface mixed
                     layer. J. Geophys. Res., 49, 8007-8017.
                Gordon, A.L., E. Molinelli, and T. Baker, 1978: Large-scale relative dynamic topography of the Southern
                     Ocean. J. Geophys. Res., 83, 3023-3032.
                Gordon, A.L., and E. Molinelli, 1982: Southern Ocean Atlas: Thermohaline-Chemical Distributions and the
                     Atlas Data Set, Columbia University Press, New York.
                Gutowski, W.J., Jr., D.S. Gutzler, and W.-C. Wang, 1991: Surface energy balances of three general
                     circulation models: Implications for simulating regional climate change. J. Climate, 4, 121-134.




                                                                 -72-










                     Haidvogel, D.B., 1990- Semi-spectral primitive equation regional ocean circulation model. In: The Coastal
                           Ocean Predictive Systems Program: Understanding and Managing Our Coastal Ocean, Vol H. Report
                           of planning workshop, University of New Orleans, Nov. 1989, 61-76.
                     Haidvogel, D.B., and A.R. Robinson, 1989: Dynamics of Atmospheres and Oceans. Special Issue on Data
                           Assimilation, 13, Nos 3 and 4.
                     Haidvogel, D.B., J.L. Wilkin, and R. Young, 1990: A semi-spectral primitive equation ocean circulation
                           model using vertical sigma and orthogonal curvilinear horizontal coordinates. J. Comput. Phys., 94,
                           151-185.
                     Han, Y.-J., 1984: A numerical World Ocean general circulation model, Part II. A baroclinic experiment.
                           Dyn. Atmos. Oceans, 8,141-172.
                     Han, Y.-J., M.E. Schlesinger, and W.L. Gates, 1985: An analysis of the air-sea interaction simulated by
                           the OSU coupled atmosphere-ocean general circulation model. In: Coupled Ocean-Atmosphere Models,
                           J. Nihoul ed., Elsevier, Amsterdam, 167-182.
                     Harrison, D.E., 1989a: Local and remote forcing of ENSO ocean waveguide response. J. Phys. Oceanogr.,
                           19, 691-695.
                     Harrison, D.E., 1989b: On climatological monthly mean wind stress and wind stress curl fields over the
                           world ocean. J. Climate, 2, 57-70.
                     Harrison, D.E., 1991: Equatorial sea surface temperature sensitivity to net surface heat flux: some ocean
                           circulation model results. J. Climate, 4, 539-549.
                     Harrison, D.E., W.S. Kessler, and B. Giese, 1989: Hindcasts of the 1982-83 El Nifio: thermal and dynamic
                           variability along the ship of opportunity XBT tracks. J. Phys. Oceanogr., 19, 397-418.
                     Harrison, D.E., B. Giese, and E.S. Sarachik, 1990: Mechanisms of SST change in the equatorial waveguide
                           during the 1982-83 ENSO. J. Climate, 3, 173-188.
                     Hasselmann, K., 1982: An ocean model for climate variability studies. Prog. Oceanogr., 11, 69-92.
                     Hayes, S.P., M.J. McPhaden, and A. Leetmaa, 1989: Observation verification of a quasi-real-time
                           simulation of the tropical Pacific Ocean. J. Geophys. Res., 94(C2), 2147-2158.
                     Hellerman, S., 1967: An updated estimate of the wind stress of the World Ocean. Mon. Wea. Rev., 95, 593-
                           606 (See also Corrigendurn Mon. Wea. Rev., 96, 63-74.)
                     Hellerman, S., and M. Rosenstein, 1983: Normal monthly wind stress over the World Ocean with error
                           estimates. J. Phys. Oceanogr., 13, 1093-1104.
                     Hibler, W.D., 111, 1979: A dynamic thermodynamic sea ice model. J. Phys. Oceanogr., 9, 815-846.
                     Hibler, W.D., III, and K. Bryan, 1987: A diagnostic ice-ocean model. J. Phys. Oceanogr., 17, 987-1015.
                     Hirst, A.C., 1985: Free equatorial instabilities in simple coupled atmosphere-ocean models. In: Coupled
                           Ocean-Atmosphere Models, J. Nihoul ed., Elsevier, Amsterdam, 153-165.
                     Hirst, A.C., 1986: Unstable and damped equatorial modes in simple coupled atmosphere-ocean models. J.
                           Atmos. Sci., 43, 606-630.
                     Hirst, A.C., 1988: Slow instabilities in tropical ocean basin-global atmosphere models. J. Atmos. Sci., 45,
                           830-852.
                     Hofmann, E.E., 1988: Plankton communities in the outer southeastern U.S. continental shelf. Part III: A
                           coupled physical -biological model. J. Mar. Res., 46, 919-946.
                     Holland, W.R., 1967: On the wind-driven circulation in an ocean with bottom topography. Tellus, 19, 582-
                           599.
                     Holland, W.R., 1971: Ocean tracer distributions. Part 1. A preliminary numerical experiment. Tellus, 23,
                           371-392.
                     Holland, W.R., 1977: Oceanic general circulation models. In: The Sea, vol. 6, John Wiley, New York, 3-45.
                     Holland, W.R., 1978: The role of mesoscle eddies in the general circulation of the ocean: Numerical
                           experiments using an wind-driven quasi-geostrophic model. J. Phys. Oceanogr., 8, 363-392.
                     Holland, W.R., 1979: The general circulation of the ocean and its modelling. Dyn. Atmos. Oceans, 3, 111-
                           142.
                     Holland, W.R., 1989- Altimeter data assimilation into ocean circulation models - some preliminary results
                           in oceanic circulation models. In: Combining Data and Dynamics, D.L.T. Anderson and J.
                           Willebrand, Eds, Kluwer Academic Publ., 203-230.
                     Holland, W.R., D.E. Harrison, and A.J.,Semtner, Jr., 1983: Eddy-resolving models of large-scale ocean
                           circulation. In: Eddies in Marine Science, A.R. Robinson, ed., Springer-Verlag, Berlin and New York,
                           379-403.
                     Hollingsworth, A. and P. Lbrinberg 1989. The verification of objective analyses: diagnostics of analysis
                           system performance. Meteorol. Atmos. Phys., 40, 3-27.
                     Hollingsworth, A., D.B. Shaw, P. L6nnberg, L. Illari, K. Arpe, and A.J. Simmons, 1986: Monitoring of
                           observation system quality by a data assimilation system, Mon. Weather Rev., 114, 861-879.
                     Huang, RX, 1991: The three-dimensional structure of wind-driven ventilation and subduction. Rev.
                           Geophys., suppl., 29, 590-609.


                                                                        -73-











                 Huang, FL, J. Luyten, and H. Stommel, 1991: Multiple equilibrium states in combined thermal and saline
                      circulation. J. Phys. Oceanogr., submitted.
                 Huang, RX, and F, Bryan, 1987: A multilayer model of the thermobaline and wind-driven ocean
                      circulation. J. Phys. Oceanogr., 17, 1909-1924.
                 Hurlburt, H.E., 1986: Dynamic transfer of simulated altimeter data into subsurface information by a
                      numerical model. J. Geophys. Res., 91, 2372-2400.
                 Hurlburt, H.E., 1989: A cost-effective global ocean model for Navy applications. In: Global Ocean
                      Prediction Systems, INO symposium, Princeton, NJ, July 1989, 37-40.
                 Hurlburt, H.E., J. Kindle, E.J. Metzger, and A. Wallcraft, 1989: Results from a global ocean model in the
                      western tropical Pacific. Proceedings of the Western Pacific International Meeting and Workshop on
                      TOGA COARE, Noum6a, New Caledonia, May 1989, 343-354.
                 Iselin, C.O'D., 1939: The influence of vertical and lateral turbulence on the characteristics of the waters at
                      mid-depths. Eos, 7@-ans. Amer. Geophys. Union, 20, 414-417.
                 Ishizaka, J., 1990a: Coupling of coastal zone color scanner data to a physical-biological model of the
                      southeastern U.S. continental shelf ecosystem, 1. CZCS data description and Lagrangian particle
                      tracing. J. Geophys. Res., 95, 20167-20181.
                 Ishizaka, J., 1990b: Coupling of coastal zone color scanner data to a physical-biological model of the
                      southeastern U.S. continental shelf ecosystem, 2. An Eulerian model. J. Geophys. Res., 95, 20183-
                      20199.
                 Ishizaka, J., 1990c: Coupling of coastal zone color scanner data to a physical-biological model of the
                      southeastern U.S. continental shelf ecosystem, 3. Nutrient and phytoplankton fluxes and CZCS data
                      assimilation. J. Geophys. Res., 95, 20201-20212.
                 Jaeger, L., 1976: Monatskarten des Niederschlags fur die Ganze erde. Berichte des Deutschen
                  , Wetterdienstes, Nr. 139, Band 18, Offenbach A.M., 38 pp.
                 Jamart, B.M., D.F. Winter, K. Banse, G.C. Anderson, and R.K. Lam, 1977: A theoretical study of
                      phytoplankton growth and nutrient distribution in the Pacific Ocean off the northwestern U.S. coast.
                      Deep-Sea Res., 24, 753-773.
                 Jamart, B.M., D.F. Winter, and K. Banse,1979: Sensitivity analysis of a mathematical model of
                      phytoplankton growth and nutrient distribution in the Pacific ocean off the northwestern U.S. coast.
                      J.Plankt. Res., 1, 267-290.
                 Janowiak, J.E., V.E. Kousky, and P.A. Arkin, 1987: The effect of changes in observation and assimilation
                      systems on CAC climate data sets. Proceedings of the 12th Annual Climate Diagnostics Workshop,
                      Salt Lake City, Utah, NTIS PB88-191747, Natl. Tech. Inf. Serv., Springfield, Va.
                 Janowiak, J.E., and P.A. Arkin, 1991: Rainfall variations in the tropics during 1986-89. J. Geophys. Res.,
                      in press.
                 Jenkins, W.J., 1989: Tritium and helium-3 in the North Atlantic in 1981 - distribution, source functions
                      and time history. CCCO/WOCE Workshop on Inversion of Ocean General Circulation Models, Royal
                      Society, London, WCRP 28, WMO/TD No. 331, A53-59.
                 Jochens, A.E., 1990: The U.S. contribution to WOCE air-sea flux estimates. U.S. WOCE Planning Report
                      Number 15, U.S. WOCE Office, College Station, TY, 46 pp.
                 Kalnay,E., M. Kanamitsu, and W.E. Baker, 1990: Global numerical weather prediction at the National
                      Meteorological Center. Bull. Amer. Meteor. Soc., 71, 1410-1428.
                 Killworth, P.D., 1989: The Flne Resolution Antarctic Model: FRAM. In: Global Ocean Prediction Systems,
                      INO Symposium, Princeton, NJ, July 1989, 111-114.
                 Killworth, P.D., and G.R. Bigg, 1988: An intercomparison of inverse methods using an eddy resolving
                      general circulation model. J. Phys. Oceanogr., 18, 987-1008.
                 Killworth, P.D., and MA. Rowe, 1987: FRAM - the Fine Resolution Antarctic Model. Ocean Modelling, 74,
                      11.
                 Kim, J-W., 1976: A generalized bulk model of the oceanic mixed layer. J. Phys. Oceanogr., 6, 686-695.
                 Kindle, J., and J.D. Thompson, 1989: The 26- and 50-day oscillations in the western Indian Ocean: model
                      results. J. Geophys. Res., 94, 4721-4736.
                 Kindle, J., H.E. Hurlburt, and E.J. Metzger, 1989: On the seasonal and interannual variability of the
                      Pacific to Indian Ocean througliflow. Proceedings of the Western Pacific International Meeting and
                      Workshop on TOGA COARE, Noum6a, New Caledonia, May 1989, 355-366.
                 Kindle, J.G., 1986: Sampling strategies and model assimilation of altimetric data for ocean monitoring and
                      prediction. J. Geophys. Res., 91, 2418-2432.
                 Mein, P., and B. Coste, 1984: Effects of wind stress variability on nutrient transport into the mixed layer.
                      Deep-Sea Res., 31, 21-37.
                 Knox, R.A. and D.L.T. Anderson 1985: Recent advances in the study of the low-latitude circulation. Prog.
                      Oceanogr., 14, 259-317.



                                                                     -74-











                    Kraus, E.B., and J.S. Turner, 1967: A one-dimensional model of the seasonal thermocline, II. The general
                          theory and its consequences. Tellus, 19, 98-106.
                    Lambert, SJ., 1988: A comparison of operational global analyses from the European Centre for Medium
                          Range Weather Forecasts (ECMWF) and the National Meteorological Center (NMC). Tellus, 40A,
                          272-284.
                    Lambert, SJ., and G.J. Boer, 1988: Surface fluxes and stresses in general circulation models. Canadian
                          Climate Centre Report, No. 88-1, CCRN 18, 183 pp.
                    Lambert, S.J., and G.J. Boer, 1989: Atmosphere-ocean heat fluxes and stresses in general circulation
                          models. Atmos. Ocean, 27, 692-715.
                    Latif, M., 1987: Tropical ocean circulation experiments. J. Phys. Oceanogr., 17, 246-263.
                    Leetmaa, A., 1990: A basic ocean observing system for climate forecasting. In: Global Ocean Observing
                          Systems Workshop Report, National Ocean Service, Alexandria, Virginia, September 1-3, 1990.
                    Leetmaa, A., and M. Ji, 1989: Operational hindcasting of the tropical Pacific. Dyn. Atmos. Oceans, 13, 465-
                          490.
                    Legler, D.M., and J.J. O'Brien, 1984: Atlas of Tropical Pacific wind stress climatology 1971 - 1980. Florida
                          State University, Department of Meteorology, Tallahassee, U.S.A, 182 pp.
                    Lemke, P., 1991: Ocean - sea coupling and deep nihing. Arctic Climate System Study Workshop, June
                          1991, Bremerhaven, Germany.
                    Lemke, P., W.B. Owens, and W.D. Hibler, 1990: A coupled. sea ice-mixed layer-pyenocline model for the
                          Weddell Sea. J. Geophys. Res., 95,9513-9525.
                    Levitus, S., 1982: Climatological Atlas of the World Ocean. NOAA Prof. Paper 13, 173 pp., 17 microfiche,
                          U.S. Govt. Printing Office, Washington, DC.
                    Levitus, S., 1984: Annual cycle of temperature and heat storage in the World Ocean. J. Phys. Oceanogr.,
                          14, 727-746.
                    Levitus, S., 1986: Annual cycle of salinity and salt storage in the World Ocean. J. Phys. Oceanogr., 16,
                          322-343.
                    Levitus, S., and A.H. Oort, 1977: Global analysis of oceanographic data. Bull. Amer. Meteorol. Soc., 58,
                          1270-1284.
                    Lewis, M.R., J.J. Cullen, and T.Platt, 1983: Phytoplankton and thermal structure in the upper ocean:
                          consequences of non-uniformity in chlorophyll profile. J. Geophys. Res., 88, 2565-2570.
                    Lewis, M.R., W.G. Harrison, N.S. Oakey, D. Hebert, and T. Platt, 1986: Vertical nitrate fluxes in the
                          oligotrophic ocean. Science, 234, 870-873.
                    Lighthill, M.J., 1969: Dynamic response of the Indian Ocean to the onset of the Southwest Monsoon. Phil.
                          Trans. R. Soc. London, Ser A 265, 45-93.
                    Long, R.B., and W.C. Thacker, 1989a: Data assimilation into a numerical equatorial ocean model, 1. The
                          model and the assimilation algorithm. Dyn. Atmos. Oceans, 13, 379-412.
                    Long, R.B., and W.C. Thacker, 1989b: Data assimilation into a numerical equatorial ocean model, II.
                          Assimilation experiments. Dyn. Atmos. Oceans, 13, 413-440.
                    Lorene, A., 1981: A global three-dimensional multivariate statistical interpolation scheme. Mon. Wea.
                          Rev., 109, 701-721.
                    Lorene, A., 1986: Analysis methods for numerical weather prediction. Q. Jl Roy. Met. Soc., 112, 1177-1194.
                    Lukas, R., and E. Lindstrom, 1987: The mixed layer of the western equatorial Pacific Ocean. In:
                          Proceedings of the Aha Hulikoa Hawaiian Winter Workshop on the Dynamics of the Oceanic Surface
                          Mixed Layer, Honolulu, Hawaii, P. Muller and D. Henderson, eds., Hawaii Institute of Geophysics
                          Special Pub., 67-94.
                    Luyten, J.R., J. Pedlosky, and H. Stommel, 1981 The ventilated thermocline. J. Phys. Oceanogr., 13, 292-
                          309.
                    Maier-Reimer, E., K. Hasselman, D. Olbers, and J. Willebrand, 1982: An ocean circulation model for
                          climate studies. Tech. Rep., Max-Planck-Institut ffir Meteorologie.
                    Maier-Reimer, E., and K. Hasselmann, 1987: Transport and storage of CO, in the ocean - An inorganic
                          ocean circulation carbon cycle model. Clim. Dyn., 2, 63-90.
                    Malanotte-Rizzoli, P., and W.R. Holland, 1986: Data constraints applied to models of the ocean general
                          circulation. Part I: the steady case. J. Phys. Oceanogr., 16, 1666-1682.
                    Malanotte-Rizzoli, P., and W.R. Holland, 1988: Data constraints applied to models of the ocean general
                          circulation. Part 11: the transient eddy-resolving case. J. Phys. Oceanogr., 18, 1093-1107.
                    Malanotte-Rizzoli, P., and W.R. Holland, 1989: Assimilation of altimeter data into an ocean circulation
                          model: space versus time resolution studies. J. Phys. Oceanogr., 19, 1507-1534.
                    Malanotte-Rizzoli, P., R.E. Young, and D.B. Haidvogel, 1989: Initialization and data assimilation
                          experiments with a primitive equation model. Dyn. Atmos. Oceans, 13, 349-378.
                    Manabe, S., and R.J. Stouffer, 1988: Two stable equilibria of a coupled ocean-atmosphere model. J.
                          Climate, 1, 841-866.


                                                                     -75-











                Manabe, S., K. Bryan, and M.J. Spelman, 1990: Transient response of a global ocean-atmosphere model to
                     a doubling of atmospheric carbon dioxide. J. Phys. Oceanogr., 20, 722-749.
                Manabe, S. R.J. Stouffer, M.J. Spelman, and K. Bryan, 1991: Transient responses of a coupled ocean-
                     atmosphere model to gradual changes of atmospheric CO, Part I: Annual mean response. J. Climate,
                     4,785-818.
                Marshall, J.C., 1985: Determining the ocean circulation and improving the geoid from satellite altimetry.
                     J. Phys. Oceanogr., 15, 330-349.
                Martin, P.J., 1985: Simulation of the mixed layer at OWS November and Papa with several models. J.
                     Geophys. Res., 90, 903-916.
                Mazumber, A., W.D. Taylor, D.J. McQueen, and D.R.S. Lean, 1990: Effects of fish and plankton on lake
                     temperature and mixing depth. Science, 247, 312-315.
                McCreary, J.P.,Jr., 1976: Eastern tropical Pacific response to changing wind systems with application to
                     El Nifio. J. Phys. Oceanogr., 6, 632-645.
                McCreary, J.P., Jr., 1985: Modeling equatorial ocean circulation. Ann. Rev. Fluid Mech., 17, 359-409.
                McCreary, J.P., Jr., and D.L.T. Anderson, 1991: An overview of coupled ocean-atmosphere models of El
                     Nifto and the Southern Oscillation. J. Geophys. Res., 96 (suppl.), 3125-3150.
                McDougall, T.J., 1988: Neutral-surface potential vorticity. Prog. Oceanogr., 20, 185-221.
                McWilliams, J.C., W.R. Holland, and J. Chow, 1978: A description of numerical Antarctic circumpolar
                     currents. Dyn. Atmos. Oceans, 2, 213-291.
                Meehl, G.A., 1990: Development of global coupled ocean-atmosphere general circulation models. Clim.
                     Dyn., 5,19-33.
                Meehl, G.A., W.M. Washington, and A.J. Semtner, Jr., 1982: Experiments with a global ocean model
                     driven by observed atmospheric forcing. J. Phys. Oceanogr., 12, 301-312.
                Meehl, G.A., and W.M. Washington, 1990: C02 climate sensitivity and snow-sea ice albedo
                     parameterization in an atmospheric GCM coupled to a mixed layer ocean model. Climate Change, 16,
                     283-306.
                Mellor, G.L., 1985: Ensemble average, turbulence closure. Advances in Geophysics, Vol. 28, Part B,
                     Saltsman and S. Manabe, Academic Press, New York, 432 pp.
                Mellor, G.L., 1989a: Retrospective on oceanic boundary layer modeling and second moment closure. In:
                     Proceedings of the Aha Hulikoa HawaiianWinter Workshop on Parameterizarion of Small-scale
                     Processes, Honolulu, Hawaii, P. Muller and D. Henderson, eds., Hawaii Institute of Geophysics
                     Special Pub., 251-272.
                Mellor, G.L., 1989b: The Princeton / Dynalysis ocean model. In Global Ocean Prediction Systems, INO
                     Symposium, Princeton, NJ, July 1989, 45-49.
                Mellor, G.L., and T. Yamada, 1974: A heirarchy of turbulence closure models for planetary boundary
                     layers. J. Atmos. Sci., 31, 1791-1806. (Corrigenda, 1977, J. Atmos. Sci., 34, 1482.)
                Mellor, G.L., and T. Yamada, 1982: Development of a turbulence closure model for geophysical fluid
                     problems. Rev. Geophys. Space Phys., 20, 851-875.
                Mellor, G.L., M.G. McPhee, and M. Steele, 1986: Ice-seawater turbulent boundary layer interaction with
                     melting or freezing. J. Phys. Oceanogr., 16, 1829-1846.
                Mellor, G.L., and L. Kantha, 1989: An ice-ocean coupled model. J. Geophys. Res., 94 (CS), 10937-10955.
                M6mery, L., and C. Wunsch, 1990: Constraining the North Atlantic circulation with tritium data. J.
                     Geophys. Res., 95(C4), 5239-5256.
                Mercier, H., 1989: A study of the time-averaged circulation in the western North Atlantic by simultaneous
                     inversion of hydrographic and current meter data. Deep-Sea Res., 36, 297-313.
                Mercier, H., M. Ollitrault, and P.V. Letraon, 1989: A non-linear inverse model of the mid-latitude North
                     Atlantic Ocean - combining float and bydrographic data. CCCOAVOCE Workshop on Inversion of
                     Ocean General Circulation Models, Royal Society, London, WCRP 23, WMO/`I`D No. 331, A24-26.
                Merlivat, 1. and A. Vezina, 1992: The role of the ocean in the carbon dioxide balance. A OOSDP discussion
                     paper, in preparation.
                Meyers G., H. Phillips, N.R. Smith and J. Sprintall 1991. Space and time scales for optimum
                     interpolation of temperature - Pacific Ocean. Prog. Oceanogr., 28, ???-218.
                Miller, R.N., 1989: Direct assimilation of altimetric differences using the Kalman filter. Dyn. Atmos.
                     Oceans, 13, 317-334.
                Miller, R.N., 1990: Tropical data assimilation experiments with simulated data: The impact of the TOGA-
                     TAO array. J. Geophys. Res., 95, 11461-11482.                                       -
                Miyakoda, K, and A. Rosati, 1984: The variation of sea surface temperature in 1976 and 1977, 2. The
                     simulation with mixed layer models. J. Geophys. Res., 89, 6533-6542.
                Moore, A.M., 1991: Data assimilation in a quasi-geostrophic open ocean model of the Gulf Stream region
                     using the adjoint method. J. Phys. Oceanogr., 21, 398-427.



                                                                 -76-










                     Moore, A.M., and D.L.T. Anderson, 1989: The assimilation of XBT data into a layer model of the tropical
                          Pacific Ocean. Dyn. Atmos. Oceans., 13, 441-464.
                     Moore, A.M., N.S. Cooper, and D.L.T. Anderson, 1987: Initialisation and data assimilation in models of the
                          Indian Ocean. J. Phys. Oceanogr., 17, 1965-1977.
                     Moore, D.W., and S.G.H. Philander, 1977: Modeling of the tropical oceanic circulation. In: The Sea, M.N.
                          Hill, ed., vol 6, Wiley (Interscience), New York, 319-361.
                     Morcrette, J.-J., 1989: Impact of changes to the radiation scheme in the ECMWF model. Research Dept
                          Tech. Report No. 64, European Centre for Medium-Range Weather Forecasts, Reading, U.K., 52 pp.
                     Morcrette, J.-J., and Y. Forquart, 1988: Comparison of ERBE measurements with model-generated
                          radiation fields. IRS88: Current Problems in Atmospheric Radiation, J. Lenoble and J.-F. Geleyn,
                          Eds., A. Deepak-Publishing, Hampton, Va., 331-334.
                     Morliere, A., J. Merle, J. Servain, and J.-P. Rebert, 1990: An operational 3 dimensional simulation of the
                          tropical Atlantic Ocean with assimilation of the in situ observations. International Symposium on
                          "Assimilation of Observations in Meteorology and Oceanography", Clermont-Ferrand, France, 9-13
                          July, 1990, 414.
                     MUller, P., and D. Henderson, (eds) 1989: Parameterization of Small-Scale Processes, Hawaiian Winter
                          Workshop, University of Hawaii at Manoa, January 17-20, 1989, 354 pp.
                     Munk, W.H., 1950: On the wind-driven ocean circulation. J. Meteor., 7, 79-93.
                     Munk, W.H., and A.M.G. Forbes, 1989: Global ocean warming: An acoustic measure. J. Phys. Oceanogr.,
                          19, 1765-1778.
                     Musgrave, D.L., J. Chou, and W.J. Jenkins, 1988: Application of a model of upper ocean physics for
                          studying seasonal cycles of oxygen. J. Geophys. Res., 93, 15679-15700.
                     NAS, 1975: Numerical Models of Ocean Circulation. National Academy of Sciences, Washington, D.C.,
                     Neelin, J.D., M. Latif, MA-F. Allaart, M.A. Cane, U. Cubasch, W.L.Gates, P.R. Gent, M. Ghil, C. Gordon,
                          N.C. Lau, G.A. Meehl, C.R. Mechoso, J.M. Oberhuber, S.G.H. Philander, P.S. Schopf, K.R. Sperber,
                          A. Sterl, T. Tokioka, J. Tribbia, and S.E. Zebiak, 1991: Tropical air-sea interaction in general
                          circulation models. Clim. Dyn., in press.
                     Nihoul, J.C.J. (Ed), 1985: Coupled Ocean-Atmosphere Models, Elsevier Oceanography Series, 40, 767 pp.
                     Nihoul, J.C.J. (Ed), 1990: Coupled Ocean-Atmosphere Models, Elsevier Oceanography Series, in press.
                     Niiler, P.P., and E.B. Kraus, 1977: One dimensional models of the upper ocean. In: Modelling and
                          Prediction of the Upper Layers of the Ocean, E.B. Kraus, Ed., Pergamon Press, 143-172.
                     O'Brien, J.J. (Ed.), 1985: Advanced Physical Oceanographic Numerical Modelling. Proe. NATO ASI,
                          Banyuls-sur-Mer, France, June 1985, 608 pp.
                     Oberhuber, J.M., 1988: An atlas based on the COADS data set: The budget of heat, buoyancy and
                          turbulent kinetic energy at the surface of the global ocean. Max-Planck-Institute for Meteorology
                          Report No. 15, Hamburg 13, Germany.
                     Oberhuber, J.M., 1991: Simulations of the Atlantic circulation with a coupled sea ice-mixed layer-isopycnal
                          general circulation model. J. Phys. Oceanogr., to appear.
                     Oberhuber, J.M., F. Lunkeit, and R. Sausen, 1991: CO, doubling experiments with a coupled global
                          isopyenic ocean-atmosphere circulation model. MPI Rep., Max-Planck-Institut fur Meteorologie,
                          Hamburg.
                     Oescheger, H., U. Siegenthaler, U. Schatterer, and A. Gugelmann, 1975: A box-diffusion model to study
                          the carbon dioxide exchange in nature. Tellus, 27, 168-192.
                     Olbers, D.J., 1989a: Diffusion parameterisations for the climatological circulation of the North Atlantic
                          and Southern Ocean. In: Proceedings of the Aha Hulikoa Hawaiian Winter Workshop on
                          Parameterizarion of Small-scale Processes, Honolulu, Hawaii, P. Maller and D. Henderson, eds.,
                          Hawaii Institute of Geophysics Special Publ., 181-204.
                     Olbers, D.J., 1989b: Inverse modelling activities at the AWL CCCO/WOCE Workshop on Inversion of
                          Ocean General Circulation Models, Royal Society, London, WCRP 23, WMO/TD No. 331, A39-A43.
                     Olbers, D.J., and M. Wenzel, 1989: Determining diffusivities from hydrodynamic data by inverse methods
                          with applications to the Circumpolar Current. In: Oceanic Circulation Models: Combining Dynamics
                          and Data. D.L.T Anderson and J. Willebrand, eds, Kluwer Academic, Hingham, Mass., 95-140
                     Olbers, D.J., M. Wenzel, and J. Willebrand, 1985: The inference of North Atlantic circulation patterns
                          from climatological hydrographic data. Rev. Geophys., 23, 313-356.
                     Owens, W.D., and P. Lemke, 1990: Sensitivity studies with a sea ice - mixed layer - pyenocline model in
                          the Weddell Sea. J. Geophys. Res., 95, 9527-9538.
                     Pacanowski, R.C., and S.G.H. Philander, 1981: Parameterization of vertical mixing in numerical models of
                          tropical oceans. J. Phys. Oceanogr., 11, 1443-1451.
                     Pacanowski, R.C., K. Dixon and A. Rosati, 1991: The G.F.D.L Modular Ocean Model Users Guide version
                          1.0", GFDL Ocean Group Technical Report #2.



                                                                       -77-











                 Palmer, T.N., C. Brankovic, F. Molteni, S.Tibaldi, L. Ferranti, A- Hollingsworth, U. Cubasch, and E.
                       Minker, 1990: The European Centre for Medium-Range Weather Forecasts (ECMWF) program on
                       extended-range prediction. Bull. Amer. Meteor. Soc., 71, 1317-1330.
                 Parkinson, C.L., and W.M. Washington, 1979: A large-scale numerical model of sea ice. J. Geophys. Res.,
                       84, 311-337.
                 Parsons, A.T., 1969: A two-layer model of Gulf Stream separation. J. Fluid Mech., 39, 511-528.
                 Pedlosky, J., 1987: Geophysical Fluid Dynamics. Second edition, Springer-Verlag, New York, 710 pp.
                 Pedlosky, J., 1990: The dynamics of the subtropical gyres. Science, 248, 316-322.
                 Philander, S.G.H., 1981: The response of equatorial oceans to a relaxation of the trade winds. J. Phys.
                       Oceanogr., 11, 176-189.
                 Philander, S.G.H., 1990: El Niho, La Niha, and the Southern Oscillation. Academic Press, San Diego,
                       California, 293 pp.
                 Philander, S.G.H., and R.C. Pacanowski, 1980: The generation of equatorial currents. J. Geophys. Res., 85,
                       1123-1136.
                 Philander, S.G.H., T. Yamagata, and R.C. Pacanowski, 1984: Unstable air-sea interactions in the tropics.
                       J. Atmos. Sci., 41, 604-613.
                 Philander, S.G.H., and A.D. Seigel, 1985: Simulation of the El Nifto of 1982-83. In: Coupled
                       Ocean-Atmosphere Models, J.C.J. Nihoul, Ed., Elsevier Oceanogr. Ser., Vol 40, Elsevier, New York,
                       517-542.
                 Philander, S.G.H., and R.C. Pacanowski, 1986: A model of the seasonal cycle in the tropical Atlantic
                       Ocean. J. Geophys. Res., 91, 14212-14220.
                 Philander, S.G.H., and W.J. Hurlin, 1987: The heat budget of the tropical Pacific Ocean in a simulation of
                       the 1982-83 El Nifio. J. Phys. Oceanogr., 17, 926-931.
                 Philander, S.G.H., W.J. Hurlin, and A.D. Seigel, 1987a: Simulation of the seasonal cycle of the tropical
                       Pacific Ocean. J. Phys. Oceanogr., 17, 1986-2002.
                 Philander, S.G.H., W.J. Hurlin, and R.C. Pacanowski, 1987b: Initial conditions for a general circulation
                       model of the tropical ocean. J. Phys. Oceanogr., 17, 147-157.
                 Phoebus, P.A., 1990: Quality control algorithms for ocean temperature data. Naval Ocean Research and
                       Development Activity, Rep. 243, Stennis Space Center, Mississippi, 21 pp.
                 Pickard, G.L., and W.J. Emery, 1982: Descriptive Physical Oceanography. Pergamon Press, New York, 249
                       PP.
                 Pollard, R.T., 1983: Mesoscle (50-100 km) circulations revealed by inverse and classical analysis of the
                       JASIN hydrographic data. J. Phys. Oceanogr., 13, 377-394.
                 Pollard, R.T., P.B. Rhines, and R.O.R.Y. Thompson, 1973: The deepening of the wind-mixed layer.
                       Geophys. Fluid Dyn., 4, 381-404.
                 Price, J.F., R.A. Weller, and R. Pinkel, 1986: Diurnal cycling: observations and models of the upper ocean
                       response to diurnal heating, cooling, and wind-mixing. J. Geophys. Res., 91, 8411-8427.
                 Randall, D.A., R.D. Cess, J.P. Blanchet, G.J. Boer, D.A. Dazlich, A.D. Del Genio, M. Deque, V. Dymnikov,
                       V. Galin, S.J. Ghan, A.A. Lacis, H. Le Treut, Z.-X. Li, X.-Z. Liang, B.J. McAvaney, V.P. Meleshko,
                       J.P.B. Mitchell, J.J. Morcrette, G.L. Potter, L. Rikus, E. Roeckner, J.F. Royer, U. Schlese, D.A.
                       Shenin, J. Slingo, A.P. Sokolov, K.E. Taylor, W.M. Washington, R.T. Wetherald, 1. Yagai, and M.-H.
                       Zhang, 1991: Intercomparison and interpretation of surface energy fluxes in atmospheric general
                       circulation models. J. Geophys. Res., submitted.
                 Reynolds, R.W., 1988: A real-time global sea surface temperature analysis. J. Climate, 1, 75-86.
                 Reynolds, R.W., and A. Leetmaa, 1989: Evaluation of NMC's operational surface fluxes in the tropical
                       Pacific. Proceedings of the Western Pacific International Meeting and Workshop on TOGA COARE,
                       Noum6a, New Caledonia, May 1989, 535-542.
                 Riley, G.A., 1942: The relationship of vertical turbulence and spring diatom flowerings. J. Mar. Res., 5, 67-
                       87.
                 Rintoul, S., 1988: Mass, heat and nutrient fluxes in the Atlantic Ocean determined by inverse methods.
                       PhD Thesis, Massachusetts Institute of Technology/Woods Hole Oceanographic Institution, 275 pp.
                 Rintoul, S., 1989: Mass, heat, oxygen and nutrient fluxes and budgets in the North Atlantic Ocean.
                       CCCO/WOCE Workshop on Inversion of Ocean General Circulation Models, Royal Society, London,
                       WOCE IPO, A51-52.
                 Robinson, A.R., 1983: Eddies in Marine Science, Springer-Verlag, 609 pp.
                 Robinson, A.R., J.A. Carton, N. Pinardi, and C.N.K. Mooers, 1986: Dynamical forecasting and dynamical
                       interpolation: An experiment in the California Current. J. Phys. Oceanogr., 16, 1561-1579.
                 Robinson, A.R., and L.J. Walstad, 1987: The Harvard open ocean model: calibration and application to
                       dynamical process, forecasting and data assimilation studies. Appl. Numer. Meth., 3, 89-131.
                 Robinson, A.R., M.A. Spall, L.J. Walstead, and W.G. Leslie, 1989a: Data assimilation and dynamical
                       interpolation in GULFCAST experiments. Dyn. Atmos. Oceans, 13, 301-316.


                                                                    -78-










                    Robinson, A.R., S.M. Glenn, MA. Spall, L.J. Walstad, G.M. Gardiner, and W.G. Leslie, 1989b: Forecasting
                          Gulf Stream meanders and rings. Eos, Trans. Amer. Geophys. Union, 70, 1464-1473.
                    Roemich, D., 1980: Estimation of meridional heat flux in the North Atlantic by inverse methods. J. Phys.
                          Oceanogr., 10, 1972-1983.
                    Rosati, A., and K. Miyakoda, 1988: A general circulation model for upper ocean simulation. J. Phys.
                          Oceanog?%, 18, 1602-1626.
                    Sarkisyan, A.S., 1962: On the dynamics of wind current appearance in the barochnic ocean. Oceanology,
                          2(3), 395-409.
                    Sarkisyan, A.S., 1975: Review of numerical ocean circulation models. In: Numerical Models of Ocean
                          Circulation. National Academy of Sciences, Washington, D.C., 76-93.
                    Sanniento, J.L., 1983: A simulation of bomb tritium entry into the Atlantic ocean. J. Phys. Oceanogr-,
                          1924-1939.
                    Sarmiento, J.L., and J.R. Toggweiler, 1984a: A preliminary model of the role of upper ocean chemical
                          dynamics in determining oceanic 0, and atmospheric CO, levels. Proc. NATO ARI on Dynamic
                          Processes in the Chemistry of the Upper Ocean, NATO Conference Series rV: Marine Sciences,
                          Plenum Press.
                    Sarmiento, J.L., and J.R. Toggweiler, 1984b: A now model for the role of the oceans in determining
                          atmospheric pCO,. Nature, 308, 621-624.
                    Sarmiento, J.L., and K Bryan, 1982: An ocean transport model for the North Atlantic. J. Geophys. Res.,
                          87, 394-408.
                    Sarmiento, J.L., J.R. Toggweiler, and R. Najjar, 1988: Oceanic carbon cycle dynamics and atmospheric
                          pCO,. Phil. Trans. R. Soc., Ser. A 325, 3-21.
                    Sarmiento, J.L., J.C. Orr, and U. Siegenthaler, 1991: A perturbation model of C02 uptake in an Ocean
                          General Circulation Model. J. Geophys. Res., in press.
                    Sathyendranath, S., A.D. Gouvela, S.R. Shetya, P. Ravindran, and T. Platt, 1991: Biological control of
                          surface temperature in the Arabian Sea. Nature, 349, 54-56.
                    Sausen, R., R.K. Barthels, and K. Hasselman, 1988: Coupled ocean-atmosphere models with flux
                          correction. Clim. Dyn., 2, 154-163.
                    Schlesinger, M.E., and J.P.B. Mitchell, 1987: Climate model simulations of the equilibrium climate
                          response to increased carbon dioxide. Rev. Geophys., 25, 760-798.
                    Schlesinger, M.E., and X. Jiang, 1988: The transport of CO,7induced warming into the ocean: An analysis
                          of simulations by the OSU coupled atmosphere-ocean general circulation model. Clim. Dyn., 3, 1-17.
                    Schlitzer, R., 1987: Renewal rates of East Atlantic deep water estimated by inversion of "C data. J.
                          Geophys. Res., 92, 2953-2961.
                    Schlitzer, R., 1988: Modeling the nutrient and carbon fluxes of the North Atlantic. Part 1: Circulation,
                          mixing coefficients, and heat fluxes. J. Geophys. Res., 93, 10699-10723.
                    Schmitz, W.J., Jr., and W.R. Holland, 1982: A preliminary comparison of selected numerical eddy-
                          resolving general circulation experiments with observations. J. Mar. Res., 40, 75-117.
                    Schopf, P.S., 1983: On equatorial waves and El Nifio. 11. Effects of air-sea coupling. J. Phys. Oceanogr., 13,
                          1878-1893.
                    Schopf, P.S., and MA. Cane, 1983: On equatorial dynamics, mixed layer physics and sea surface
                          temperature. J. Phys. Oceanogr., 13, 917-935.
                    Schopf, P.S., and M.J. Suarez, 1988: Vacillations in a coupled ocean-atmosphere model. J. Atmos. Sci., 45,
                          549-566.
                    Schott, F., and H. Stomell, 1978: Beta spirals and absolute velocities from different oceans. Deep-Sea Res.,
                          25,961-1010.
                    Seager, R., 1989: Modeling tropical Pacific sea surface temperature. J. Phys. Oceanogr., 19, 419-434.
                    Seager, R., S.E. Zebiak, and M.A. Cane, 1988: A model of the tropical Pacific sea surface temperature
                          climatology. J. Geophys. Res., 93, 1265-1280.
                    Senitner, A.J., Jr., 1974: An oceanic general circulation model with bottom topography. Numerical
                          Simulation of Weather and Climate, Tech. Rept No. 9, Department of Meteorology, University of
                          California, Los Angeles, 99 pp.
                    Senitner, A.J., Jr., 1984: Modeling the ocean in climate studies. Ann. Glaciol., 5, 133-140.
                    Semtner, A.J., Jr., 1986a: History and methodology of modelling the circulation of the World Ocean. In:
                          Advanced Physical Oceanographic Numerical Modelling, Proc. NATO ASI, Banyuls-sur-Mer, June
                          1985, Reidel Publ., Holland, 23-32.
                    Semtner, A.J., Jr., 1986b: Finite-difference formulation of a World Ocean model. In: Advanced Physical
                          Oceanographic Numerical Modelling, Proc. NATO ASI, Banyuls-sur-Mer, June 1985, Reidel Publ.,
                          Holland, 187-202.
                    Senitner, A.J., Jr., 1989: An ocean prediction model for the supercomputing architecture of the 1990's. In:
                          Global Ocean Prediction Systems, INO symposium, Princeton, NJ, July 1989, 59-62.


                                                                       -79-










                   Semtner, A.J., Jr., and Y. Mintz, 1987: Numerical simulation of the Gulf Stream and mid-ocean eddies. J.
                        Phys. Oceanogr.,  7,208-230.
                   Semtner, A.J., Jr., and R.M. Chervin, 1988: A simulation of the global ocean circulation with resolved
                        eddies. J. Geophys. Res., 93, 15502-15522.
                   Semtner, A.J., Jr., and R.M. Chervin, 1990: Environmental effects on acoustic measures of global ocean
                        warming. J. Geophys. Res., 95 (C8), 12973-12982.
                   Sheinbaum, J., and D.L.T. Anderson, 1990a: Variational assimilation of XBT data. Part I. J. Phys.
                        Oceanogr., 20, 672-688.
                   Sheinbaum, J., and D.L.T. Anderson, 1990b: Variational assimilation of XBT data. Part II. Sensitivity
                        studies and use of smoothing constraints. J. Phys. Oceanogr., 20, 689-704.
                   Siegel, D.A., and T.D. Dickey, 1987: On the parameterisation of irradiance for open ocean photoprocesses.
                        J. Geophys. Res., 92,14648-14662.
                   Simonot, J.-Y., and H. Le Treut, 1987: Surfkce heat fluxes from a numerical weather prediction system.
                        Climate Dyn., 2, 11-28.
                   Simonot, J.-Y., E. Dollinger, and H. Le Treut, 1988: Thermodynamic-biological-opticaI coupling in the
                        oceanic mixed layer. J. Geophys. Res., 93, 8193-8202.
                   Smetacek, V.S., 1985: Role of sinking in diatom life-history cycles: ecological, evolutionary and geological
                        significance. Mar. Biol., 84, 239-251.
                   Smith, L.T., D.B. Boudra, and R. Bleck, 1990: A wind-driven isopyenic coordinate model of the North and
                        Equatorial Atlantic Ocean. J. Geophys. Res., 95, 13105-13128.
                   Smith, N.R., 1989: Improved in situ measurements for initialising atmospheric forecast models. In: Report
                        of the Working Group on In Situ Measurements for Fluxes, WCRP-23, WMO/TD No. 304, La Jolla,
                        California, March, 1989, II.2.1-IL2.24.
                   Smith, N.R., 1991a: Objective quality controls in an oceanic subsurface thermal analysis. scheme. J.
                        Geophys. Res., 96, (suppl.), 3279-3287.
                   Smith, N.R., 1991b: Assimilation experiments using a Pacific Ocean general circulation model: preliminary
                        results. In: Second Annual BMRC Modelling Workshop - Data Assimilation Systems, September
                        1990, BMRC Research Report No. 27, Bureau of Meteorology, Melbourne, 174-181.
                   Smith, N.R., and G.D. Hess, 1992a: On vertical eddy mixing and equatorial ocean models. Part 1, The
                        parameterisations. J. Phys. Oceanogr., submitted.
                   Smith, N.R., and G.D. Hess, 1992b: On vertical eddy mixing and equatorial ocean models. Part 2. The
                        thermodynamic circulation. J. Phys. Oceanogr., submitted.
                   Smith, N.R., J.E. Blomley, and G. Meyers, 1991: A univariate statistical interpolation scheme for
                        subsurface thermal analyses in the tropical oceans. Prog. Oceanogr., 28, 219-256.
                   Smith, N.R., R. Kleeman, and G.D. Hess, 1992: The surface air-sea heat and water budget as diagnosed by
                        equatorial Pacific Ocean models. J. Climate, submitted.
                   Spall, M.A., 1990: Circulation in the Canary Basin- A model/data analysis. J. Geophys. Res., 95, 9611-
                        9628.
                   Sprintall, J., and G. Meyers, 1991: An optimal Y[BT sampling network for the eastern Pacific Ocean. J.
                        Geophys. Res., 96, 10539-10552.
                   St6ssel, A., P. Lemke, and W.B. Owens, 1990: Coupled sea ice - mixed layer simulations for the Southern
                        Ocean. J. Geophys. Res., 95, 9539-9555.
                   Stevens, D.P., 1991: The open boundary conditions in the United Kingdom Fine-Resolution Antarctic
                        Model. J. Phys. Oceanogr., 21, 1494-1499.
                   Stevenson, J.W., 1979: On the effect of dissipation of seasonal thermocline models. J. Phys. Oceanogn, 9,
                        57-64.
                   Stommel, H., 1948: The westward intensification of wind-driven ocean currents. Trans. Amer. Geophys.
                        Union, 29, 202-206.
                   Stommel, H., 1979: Determination of water mass properties of water pumped down from the Ekman layer
                        to the geostrophic flow beneath. proc. Nat. Acad. Sciences U.S.A., 76, 3051-3055.
                   Sverdrup, H.U., 1947: Wind-driven currents in a baroclinine ocean: with application to the equatorial
                        currents of the eastern Pacific. Proc. Nat. Acad. Sci., 33, 318-326.
                   Takano, K. 1975: A numerical simulation of the World Ocean circulation. In: Numerical Models of Ocean
                        Circulation. National Academy of Sciences, Washington, D.C., 121-132,                                                   Z
                   Talley, L.D., 1988: Potential vorticity distribution in the North Pacific. J. Phys. Oceanogr., 18, 89-106.
                   Taylor, A.H., and I. Joint, 1990: A steady-state analysis of the 'microbial loop'in stratified systems. Mar.
                        Ecol. Prog. Ser., 59,1-17.                                                                                              01
                   Taylor, A.H., A.J. Watson, M. Ainsworth, J.E. Robertson, and D.R. Turner, 1991: A modelling
                        investigation of the role of phytoplankton in the balance of carbon at the surface of the North
                        Atlantic. Global Biogeochem. Cycles, 5, 151-171.



                                                                       -80-










                     Taylor, A.H., J.R.W. Harris, and J. Aiken, 1986: The interaction of physical and biological processes in a
                          model of the vertical distribution of phytoplankton under stratification. In: Marine Interfaces
                          Ecohydrodynamics, J.C.J, Nihoul, Ed., Elsevier Oceanography Series, 42, Elsevier, New York, 313-
                          330.
                     Taylor, P.K., 1989: WOCE surface flux determinations - a strategy for in situ measurements. Report of the
                          Working Group on In Situ Measurements for Fluxes, WCRP - 23, La Jolla, California, March, 1989.
                     TCIPO (TOGA COARE Internation Project Office), 1991: TOGA COARE experiment design. TOGA
                          COARE International Project Office, UCAR, Boulder, USA, October, 1991.
                     Thacker, W.C., and R.B. Long, 1988: Fitting dynamics to data. J. Geophys. Res., 93, 1227-1240.
                     Thacker, W.C., R.B. Long, S.-M. Hwang, R. Sabina, and Y.-H. Daneshzadeh, 1989: Adjoint of an oceanic
                          general circulation model. In: CCCO / WOCE Workshop on Inversion of Ocean General Circulation
                          Models, Royal Society, London, 10-12 July 1989, A4-7.
                     The WAMDI-Group: Hasselmann, S., K. Hasselmann, E. Bauer, P.A.E.M. Janssen, G.J. Komen, L.
                          Bertotti, P. Lionello, A. Gillaume, V.C. Cardone, J.A. Greenwood, M. Reistad, L. Zambresky, and J.A.
                          Ewing, 1988: The WAM model - a third generation ocean wave prediction model. J. Phys. Oceanogr.,
                          18,1775-1810.
                     The FRAM Group, 1991: Initial results from a fine resolution model of the Southern Ocean. Eos, Trans.
                          Amer. Geophys. Union, 72, 169,174-5.
                     TOGA, 1985: TOGA Implementation Plan, International TOGA Project Office, Boulder Colorado.
                     TOGA Numerical Experimentation Group, 1989: Report of the 3rd session of TOGA NEG, Hamburg, FRG,
                          18-20 September, 1989, WMO/TD-No. 339.
                     TOGA Numerical Experimentation Group, 1990: Report of the 4th session of TOGA NEG, Palisades, New
                          York, U.S.A., 21-23 May, 1990, WMO/TD-No. 398.
                     Toggweiler, J.R., 1989: Is the downward dissolved organic matter (DOM) flux important in carbon
                          transport? In: Productivity of the Ocean: Past and Present. W.H. Berger, V.S. Smetacek and G.
                          Wefer, Eds, John Wiley & Sons, New York, 65-83.
                     Toggweiler, J.R., 1990: Modelling workshop offers first look at new simulations of equatorial Pacific. U.S.
                          JGOFS News, 2, 1.
                     Toggweiler, J.R., K. Dixon, and K. Bryan, 1989a: Simulations of radiocarbon in a coarse-resolution World
                          Ocean model, 1. Steady state pre-bomb distributions. J. Geophys. Res., 94, 8217-8242.
                     Toggweiler, J.R., K. Dixon, and K. Bryan, 1989b: Simulations of radiocarbon in a coarse-resolution World
                          Ocean model, 2. Distributions of bomb-produced carbon 14. J. Geophys. Res., 94, 8243-8264.
                     Trenberth, K.E., W.G. Large and J.G. Olson, 1989: The effective drag coefficient for evaluating wind stress
                          over the oceans. J. Climate, 2, 1507-1516.
                     UCAR, 1991: Prospectus for a Tropical-Oceans-Global-Atmosphere (TOGA) Program on Seasonal-to-
                          Interannual Prediction. M. Cane and E. Sarachik, eds, NOAA Climate and Global Change Program
                          special Report No. 4, UCAR, 1991.
                     Walsh, J.J., D.A. Dieterle, and M.A. Meyers, 1988: A simulation analysis of the fate of phytoplankton
                          within the Mid-Atlantic Bight. Cont. Shelf Res., 8, 757-787.
                     Walsh, J.J., DA_ Dieterle, M.A. Meyers, F.E. Muller-Karger, 1989: Nitrogen exchange at the continental
                          margin: a numerical study of the Gulf of Mexico. Prog. Oceanogr., 23, 245-301.
                     Washington, W.M., A.J. Semtner, Jr., G.A. Meehl, D.J. Knight, and T.A. Mayer, 1980: A general
                          circulation experiment with a coupled atmosphere, ocean and sea ice model. J. Phys. Oceanogr., 10,
                          1887-1908.
                     Washington, W.M., and C.L. Parkinson, 1986: An Introduction to Thr@e-dimensional Climate Modelling.
                          University Science Books, Mill Valley, U.S.A, 422 pp.
                     Washington, W.M, and G.A. Meehl, 1986: General circulation model CO, sensitivity experiments: Snow-
                          sea ice albedo parameterization and globally averaged surface air temperature. Climate Change, 8,
                          231-241.
                     Washing-ton, W.M., and G.A. Meehl, 1989: Climate sensitivity due to increased C02: experiments with a
                          coupled atmosphere and ocean general circulation model. Clim. Dyn., 4, 1-38.
                     WCRP, 1988a: Working Group on Sea Ice and Climate. Report of the 3rd session, Oslo, Norway, June
                          1988, WCRP-18, WMO/TD No. 272.
                     WCRP, 1988b. World Ocean Circulation Experiment Implementation Plan, Vol. L Detailed Requirements.
                          WCRP-11, WMO/TD No. 242.
                     WCRP, 1988c. World Ocean Circulation Experiment Implementation Plan, Vol. H. Scientific Background.
                          WCRP-12, WMO/TD No. 243.
                     WCRP, 1989a: Global Data Assimilation Program (GDAP) for Air-Sea Fluxes. Report of the JSCICCCO
                          Working Group on Air-Sea Fluxes, October 1988, WCRP-16, WMO/TD-No. 259.
                     WCRP, 1989b: Working Group on Sea Ice and Climate. Report o@ the 4th session, Rome, Italy, November
                          1989, WCRP-41, WMO/TD No. 377.


                                                                       -81-











                WCRP, 1989c: Sea Ice Numerical Experimentation Group. Report of the ist session, Washington, D.C.,
                     May 1989, WCRP-45, VtrMO/TD No. 384.
                WCRP, 1989d: Inversion of Ocean General Circulation Models. Report of the CCCO/WOCE Workshop,
                     London, 10-12 July 1989, WMO/TD-No. 331.
                WCRP, 1991a: The Global Climate Observing System. Report of a meeting convened by the Chairman of
                     the JSC, Winchester, UK, 14-15 January, 1991, WCRP-56, WMO/TD No. 412.
                WCRP, 1991b: Global Climate Modelling. Report of first session of WCRP Steering group on Global
                     Climate Modelling, March 1991, WCRP-55, WMO/TD No. 411.
                Weare, B.C., 1989: Uncertainties in estimates of surface heat fluxes derived from marine reports over the
                     tropical and subtropical oceans. Tellus, Ser. A, 41, 357-370.
                Weare, B.C., P.T. Strub, and M.D. Sarnual, 1981: Annual mean surface heat fluxes in the tropical Pacific
                     Ocean. J. Phys. Oceanogr., 11, 705-717.
                Webb, D.J., and A.M. Moore, 1986: Assimilation of altimeter data into ocean models. J. Phys. Oceanogr.,
                     16,1901-1913.
                Webb, D.J., P.D. Killwortb, A.C. Coward, and S.R. Thompson, 1991: The FRAM Atlas of the Southern
                     Ocean. Natural Environmental Council, Swindon, UK.
                White, W.B., and R.L. Bernstein, 1979: Design of an oceanographic network in the midlatitude North
                     Pacific. J. Phys. Oceanogr., 13, 531-544.
                White, W.B., G. Meyers, and K. Hasunuma, 1982: Space/time statistics of short-term climatic variability
                     in the western North Pacific. J. Geophys. Res., 87, 1979-1989.
                White, W.B., G. Meyers, J.R. Donguy, and S.E. Pazan, 1985: Short-term climatic variability in the
                     thermal structure of the Pacific Ocean during 1979-82. J. Phys. Oceanogr., 15, 917-935.
                White, W.B., S.E. Pazan, G.W. Withee, and C. Noe, 1988: Joint Environmental Data Analysis (JEDA)
                     Center for the scientific quality control of upper ocean thermal data in support of TOGA and WOCE.
                     Eos, Trans. Amer. Geophys. Union, 69,122-123,131. .
                White, W.B., C.-K. Tai, and W.R. Holland, 1990a: Continuous assimilation of GEOSAT altimetric sea level
                     observations into a numerical synoptic ocean model of the California Current. J. Geophys. Res., 95
                     (C3), 3127-3148.
                White, W.B., C.-K. Tai, and W.R. Holland, 1990b: Continuous assimilation of simulated GEOSAT
                     altimetric sea level into an eddy-resolving numerical model, 1. Sea level differences. J. Geophys. Res.,
                     95 (C3), 3219-3234.
                White, W.B., C.-K. Tai, and W.R. Holland, 1990c: Continuous assimilation of simulated GEOSAT
                     altimetric sea level into an eddy-resolving numerical model, 2. Referenced sea level differences. J.
                     Geophys. Res., 95 (C3), 3235-3252.
                Willynott, A.J., and L.A. Mysak, 1989: A simple steady-state coupled ice-ocean model, with application to
                     the Greenland-Norwegian Sea. J. Phys. Oceanogr., 19, 501-518.
                Winter, D.F., K. Banse, and G.C. Anderson, 1975: The dynamics of phytoplankton blooms in Puget Sound,
                     a fjord in the Northwestern United States. Mar. Biol., 29, 139-176.
                WMO, 1990: International symposium on 'Assimilation of Observations in Meteorology and
                     Oceanography", Clermont-Ferrand, France, 9-13 July, 1990, extended abstracts.
                Wolf, K.U., and J.D. Woods, 1988: Langrangian simulation of primary production in the physical
                     environment - the deep chlorophyll maximum and nutrieline. In: Toward a Theory on Biological-
                     Physical Interactions in the World Ocean. B.J. Rothschild, Ed., Kluwer Academic, 51-70.
                Woods, J.D., 1985: The physics of thermocline ventilation. In: Coupled Ocean-Atmosphere Models, J.C.J.
                     Nihoul, Ed., Elsevier Oceanogr. Ser., Vol 40, Elsevier, New York, 543-590.
                Woods, J.D., 1988: Scale upwelling and primary production. In: Toward a Theory on Biological-Physical
                     Interactions in the World Ocean. B.J. Rothschild, Ed., Kluwer Academic, 7-38.
                Woods, J.D., 1991: A global ocean observing system. In The Global Climate Observing System. Report of a
                     meeting convened by the Chairman of the JSC, Winchester, UK, 14-15 January, 1991, Appendix 6.
                Woods, J.D., and R. Onken, 1982: Diurnal variation and primary production in the ocean - preliminary
                     results of a Lagrangian ensemble model. J. Plankt. Res., 4, 735-756.
                Worthington, L.V., 1976: On the North Atlantic Circulation, John Hopkins U. Press, Baltimore, 110 pp.
                Wroblewski, J.S., 1989: A model of the spying bloom in the North Atlantic and its impact on oceanic
                     optics. Limnol. Oceanogr., 34, 1563-1571.
                Wroblewski, J.S., J.L. Sarmiento, and G.R. Flierl, 1988: An oceanic basin scale model of plankton
                     dynamics in the North Atlantic. 1. Solutions for the climatological oceanographic conditions in May.
                     Global Biogeochem. Cycles, 2, 199-218.
                Wunsch, C., 1977: Determining the general circulation of the oceans: A preliminary discussion. Science,
                     196, 871-875.
                Wunsch, C., 1978: The North Atlantic general circulation west of 50*W determined by inverse methods.
                     Rev. Geophys., 16, 583-619.


                                                                -82-












                     Wunsch, C., 1984: An eclectic Atlantic Ocean circulation model. Part 1: The meridional flux of heat. J.
                         Phys. Oceanogr., 14, 1712-1733.
                     Wunsch, C., 1987: Using transient tracers: The regularisation problem. Tellus, 39B, 477-492.
                     Wunsch, C., 1989a: The next generation of inverse models. CCCO/WOCE Workshop on Inversion of Ocean
                         General Circulation Models, Royal Society, London, WOCE IPO, A36-38.
                     Wunsch,C. 1989b: Tracer inverse problems. In: Oceanic Circulation models: Combining Dynamics and
                         Data. D.L.T Anderson and J. Willebrand, eds, Kluwer Academic, Hingham, Mass., 1-77.
                     Wunsch, C., D.-X. Hu, and B. Grant, 1983: Mass, heat, salt and nutrient fluxes in the South Pacific
                         Ocean. J. Phys. Oceanogr., 13, 725-753.
                     Yu, W. G. Sobze, H. Le Treut, and M. Desbois, 1991: Comparison of radiance fields observed by satellite
                         and simulated by the LMD general circulation model. Dyn. Atmos. Oceans, 16, 147-165.























































                                                                     -83-










                                   APPENDIX



                                LIST OF ACRONYMS


          ACC         Antarctic Circumpolar Current
          AGCM        Atmospheric General Circulation Model
          ATLAS       Autonomous Temperature Line Acquisition System
          AVHRR       Advanced Very High Resolution Radiometer
          BMRC        Bureau of Meteorology Research Centre
          CAC/NMC     Climate Analysis Center/National Meteorological Center
          CCCO        Committee on Climate Changes in the Ocean
          CME         Community Modelling Effort
          COADS       Comprehensive ocean Atmosphere Data Set
          COPS        Coastal Ocean Prediction System

          CzCS        Coastal Zone Color Scanner

          ECMWF       European Centre for Medium Range Weather Forecasts
          ENSO        El Niho/Southern Oscillation

          ERS-1       Earth Remote Sensing Satellite
          FNOC        Fleet Numerical Oceanography Center
          FRAM        Fine Resolution Antaictic Model

          FSU         Florida State University

          GCM         General Circulation Model

          GCOS        Global Climate Observing System
          GDAP        Global Data Assimilation Program

          GEOSAT      Geodetic Satellite Mission

          GEOSECS     Geochemical Oceans Sections Study
          GEWEX       Global Energy and Water Cycle Experiment
          GFDL        Geophysical Fluid Dynamics Laboratory
          GOOS        Global Ocean Observing System
          IPCC        Intergovernmental Panel on Climate Change
          JGOFS       Joint Global Ocean Flux Study

          JSC         Joint Scientific Committee
          LODYC       Laboratoire, d@Oc6anographie Dynamique, et de Climatologie
          mom         Modular Ocean Model




                                       -84-









               NAS             National Academy of Sciences
               NCAR            National Center for Atmospheric Research
               NEG             Numerical Experimentation Group
               NMC             National Meteorological Center
               NOP             Numerical Ocean Prediction

               NWP             Numerical Weather Prediction

               OGCM            Ocean General Circulation Model
               OOS             Ocean Observing System
               OOSDP           Ocean Observing System Development Panel
               ORSTOM          Office de la Recherche Scientifique et Technique Outre Mer
               OSE             Observing System Experiment
               OSSE            Observing System Simulation Experiment
               OSU             Oregon State University
               OTIS            Optimal Thermal Interpolation System
               SAR             Synthetic Aperture Radar
               SINEG           Sea Ice Numerical Experimentation Group
               SMC             Second Moment Closure
               SSS             Sea Surface Salinity
               SST             Sea Surface Temperature
               TAO             Tropical Atmosphere Ocean (array)
               TKE             Turbulent Kinetic Energy
               TOGA            Tropical Ocean and Global Atmosphere
               TOGA COARE      TOGA Coupled Ocean Atmosphere Response Experiment
               TRMM            Tropical Rainfall Measurement Mission
               UKMO            United Kingdom Meteorological Office
               VOS             Volunteer Observing Ship
               XBT             eXpendable BathyThermograph

               WAM             WAve Model

               WCRP            World Climate Research Program
               WGSIC           Working Group on Sea Ice and Climate
               WOCE            World ocean Circulation Experiment









                                                     -85-






                                                                                                                     NOAA COASTAL SERVICES M LIBRARY



                                                                                                                     3 6668 14111030 6