NCAR

Executive Summary

The Challenge of Modeling Climate

A Step Towards Community Climate Models

Why NCAR?

The NCAR climate modeling groups propose an immediate focus on a climate system model (CSM).

A Start-Up Phase

CSM Plans

What Does the Research Community Get from a CSM at NCAR?

Climate researchers from universities, federal laboratories, the private sector, and NCAR get modeling tools and a framework and computing environment for using those tools. They get access to a range of model components to complement their own and a system in which to test their own components in a variety of configurations. Other coupled modeling efforts get access to and benefit from an open modeling environment and the opportunity to implement community CSM improvements in their own systems. The climate research community gets a much needed test bed, a way to incorporate and focus diverse achievements, and a set of recognized, documented, advanced coupled models to apply to national and international environmental policy issues. The climate impacts and policy research communities will have access to the CSM process and will help guide CSM development.

Changes in climate, whether anthropogenic or natural, involve a complex interplay of physical, chemical, and biological processes of the atmosphere, ocean, and land surface. As climate system research seeks to explain the behavior of climate over time scales of years to millennia, focus necessarily turns to behavior introduced by physical, chemical, and biogeochemical interactions among climate subsystems. The need to understand climate as a coupled system is demonstrated by the paleoclimate record, which reveals large, related changes in atmospheric and oceanic circulation and biogeochemistry (e.g., Broecker, 1987; Dansgaard et al., 1993). The challenges of modeling the role of anthropogenic emissions of carbon dioxide, of reactive trace gases, and of changing land use in the earth system likewise require a coupled climate system approach. While knowledge that land-ocean-atmosphere interactions influence climate is not new, the emergence of coupled climate system questions as central scientific concerns of geophysics constitutes a major change in the research agendas of atmospheric science, oceanography, ecology, and hydrology.

Two recent books, ``Climate System Modeling'' (Trenberth, 1992) and ``Modeling the Earth System'' (Ojima, 1992), document many aspects of climate system modeling and of the uses of such models for global change research and paleoclimate studies. Key questions include the coupling between processes occurring at different time scales (seasonal to interannual, interannual to decadal) that affect the longer term behavior of the system. The interaction between seasonal and El Nino-Southern Oscillation (ENSO) processes discussed by Tziperman et al. (1994) and Jin et al. (1994) provides one example. Evidence in the paleorecord and from models that changes in ENSO frequency are a component of long-term climate variability is a related but longer term example (Meehl et al., 1993; Thompson et al., 1989). The possible existence of thresholds or multiple equilibria in the climate system, arising from the thermohaline ocean circulation and perhaps related to the carbon cycle, is another issue of scientific importance and great relevance to society (Taylor et al., 1993). Adaptation to abrupt change would be more difficult than acclimatization to continuous change.

We believe that relatively simple models will continue to play an important role in climate system research, but the most credible characterizations of the important processes, as well as the best predictions for future climate change, will come from comprehensive models of the climate system. Many relevant examples of component subsystem models exist, and several groups around the world have produced interesting prototypes for a comprehensive climate system model. However, no mature model of this type yet exists. We believe it timely to establish a substantial, highly-focussed project at NCAR to develop, explore, test, and continuously improve a CSM. Extrapolating from previous general circulation model (GCM) experience, we strongly believe this project, and others like it, will require many years to come to maturity, although a sequence of useful models and scientific results will be produced over that period.

Development of a comprehensive CSM, which accurately represents the principal components of the climate system and the couplings between them, requires both wide intellectual participation and computing capabilities beyond those available to most U.S. institutions. A CSM, therefore, must involve an improved framework for coupling existing and future component models to permit rapid exploration of alternate formulations, an accessible coupled model within that framework that allows calculations at various complexities and resolutions according to scientific need and demand on computing resources, and an active series of simulations and evaluations using the improved model to address important scientific issues and problems of national and international policy interest.

There have been several coupled modeling studies over the last two decades and an ongoing debate over criteria for ``state-of-the-art'' climate models. It has been argued (by S. Manabe of the Geophysical Fluid Dynamics Laboratory (GFDL), S. Schneider of Stanford/NCAR, among others) that the choice of computational resolution involves scientific assessments balanced by resource (computing) availability, and that model skill in addressing specific scientific questions does not automatically result from detailed parameterizations or high resolution. We extract two principles from that debate pertinent to the development of a CSM. First, science issues must drive model development, although computing resources may force scientific compromises and may severely limit the number of simulations that can be performed. Second, the CSM development effort should not a priori eschew exploration of coupled behavior or solutions using simple, less computationally intensive, component models.

As a leading research institution in atmospheric science and associated disciplines, NCAR has a long history of developing, maintaining, and improving models of the atmosphere, ocean, and land surface. NCAR also pioneered basic studies of the earth system that led to current understanding of the strength and pervasive influence of land-ocean-atmosphere couplings on the climate system.

NCAR now plans to develop, maintain, and improve a comprehensive CSM. We commit to designing a scientifically sound, policy-relevant modeling system, accessible to a large user community. The CSM itself, a hierarchy of simpler subset models, and the results of a wide variety of experiments using the models will be available to NCAR, university, and federal laboratory scientists, and to any other interested parties. To ensure broad intellectual participation and efficient use of resources, the effort to develop a CSM will be coordinated and integrated across NCAR scientific divisions and the NCAR Scientific Computing Division, and will collaborate extensively with the university and federal laboratory research community.

As a focussed effort, the CSM project will address an important aspect of climate system research. In particular, the CSM project is aimed at developing and using comprehensive models to gain a predictive understanding of the climate system. Toward this end, the long-term goals of the CSM project are simple and ambitious. They are:

• to develop and maintain a comprehensive CSM that is at the forefront of international efforts in modeling the climate system;
• to maintain a hierarchy of less complete subset models;
• to continually improve the comprehensive model and its component models;
• to use the model to address important scientific questions about the climate system, including global change and interdecadal and interannual variability;
• to use appropriate versions of the model for calculations in support of national and international policy decisions; and
• to make the complete hierarchy of models and the results of major experiments available to the wider scientific community.

Complementary efforts using simplified models are also important and will be undertaken by many individuals, including CSM participants.

The CSM will have a simple conceptual structure: separate models for atmosphere, ocean, land surface, and sea ice that communicate through a flux coupler. The flux coupler will impose few (if any) rules about the resolutions and structures of component models. This structure will easily allow for the distribution of the component models among separate computers. This simple structure also allows the systematic maintenance of a hierarchy of climate system models. For example, the same structure can support alternative component models ranging from specified boundary conditions to interactive GCMs on either side of the atmosphere/ocean interface. The appropriate component models and their configuration for a particular problem can then be selected on the basis of the important processes expected to be involved.

The atmospheric component includes atmospheric dynamics, radiation, clouds, aerosols, and chemistry; it encompasses the troposphere and at least some (possibly large) portion of the middle atmosphere. The oceanic component includes ocean dynamics and biochemistry. For efficient modeling the ocean component supports different domains: regional or global and upper-ocean or full-depth. The land surface component includes physical, biological, and chemical processes; it extends from the top of the plant canopy through the upper soil layers. The sea ice component includes the dynamics and thermodynamics of sea ice formation and transport. The flux coupler is a separate computer code that interrogates components for state variables, calculates interfacial fluxes not produced by component models, and parcels out fluxes in an integrally-conservative manner, even when component models have different space-time discretizations. The flux coupler will also control the synchronization and pace of calculations.

The NCAR Community Climate Model (CCM) represents a successful example of an institutional modeling project. Many of its attributes should belong to the CSM:

• It is maintained as a community model, available to NCAR, university, and federal laboratory scientists who invest effort to learn about it and obtain computer time to use it;
• Its development and evaluation occurs through an open process that includes all interested scientists, but is accomplished in practice primarily by a relatively small team of scientists who make a substantial commitment of their time, cooperate closely, and take responsibility for the model's contents;
• Its development is guided and overseen by an internal advisory committee that meets frequently, an external panel of experts that meets approximately annually, and occasional workshops for users;
• Its code architecture and formats support modular subcomponents and portability to different computers and centers;
• Benchmark versions of the model are produced on explicitly declared (and approximately met) schedules; and
• The project commands resources, particularly in support staff and computer allocations, greater than those typically available to individual scientists.

The scientific scope of the CSM project is much greater than the scope of the CCM. In consequence, the CSM will be more complex than the CCM, and the CSM project will need to support much broader user communities than does the CCM. The CSM should also stand as a more nearly unique model. A few modeling groups have expertise that spans the climate system; however, in the U.S. none are institutionally committed to full community access to its models and computing environment. In its more complete configurations, CSM will require long integration times and computing resources substantially larger than those now available. To obtain the maximum benefit from the available scientific and computational resources, the CSM project will adhere to the following design criteria:

Establish a unified model framework. The CSM will establish a unified model framework, albeit one with flexibility and generality. When alternative component models exist or develop, the CSM project will endeavor to incorporate the best aspects of each in the unified model. However, the CSM will have to run on several different computer systems, and it is possible that separate implementations of some component models will have to be maintained for both massively parallel computers and moderately parallel vector computers.

Support variable domains, resolutions, and complexity. The CSM must support various domains and resolutions to allow locally high-resolution studies of specific problems or regions and to allow relatively low-resolution for long time scale (century and longer) studies. A hierarchy of components must be supported with, for example, different levels of physical or chemical complexity. The combination of the components for particular coupled calculations must be easily user selectable.

Support modular components. Component models should interact with the CSM in a modular way, so that scientists can experiment with alternative component models. Modularity offers the potential for fruitful cooperation with other groups pursuing climate system modeling.

Develop from existing components. In principle, CSM might wish to design entirely new component models, subject to all available advice. In practice, several CSM component models will grow from existing models. The CSM atmospheric physical model will start from CCM2. The first realization of the CSM oceanic physical model will derive from a GFDL-type ocean model. Other CSM component models have, a priori, less obvious initial design bases.

Begin with some components simplified. To efficiently explore new modes of solution behavior that occur with coupled component models, CSM calculations should begin with at least some components intentionally very simple.

Use computational science advances. To achieve the design criteria listed (modularity, variable domains, resolutions, and complexity, and selectable combinations of coupled components) and to support CSM on a wide range of computer systems, CSM must attend closely to code architectural aspects and other computational science issues. CSM teams should include computational science expertise.

Developing CSM and its components invokes a vital, ongoing need to judge how well the models perform, to diagnose erroneous performance, to improve those aspects or components deemed inadequate, and generally to document and validate the model. In carrying out these tasks, close collaboration with mathematicians and statisticians is essential. Even for individual model components, natural variability of the model may obscure the signals one seeks in the model solution. In the case of CSM, the scope of possible natural variability becomes much greater and less well documented, so validation presents a severe problem.

The lack of complete, global observations of many important variables and derived quantities presents major challenges to validation of component models of the atmosphere, ocean, and sea ice. In coupled models, concern with interactive system behavior exacerbates validation problems. However, a great deal of data is available that provides indirect information about aspects of the coupled system. For example, heat and fresh water budget constraints can be used to validate coupled atmosphere-ocean-sea ice components. Estimates of evaporation minus precipitation from atmospheric measurements or analyses can be used to examine fresh water transports, oceanic salinity, and thermohaline circulation. Similarly, the sum of convergence of heat transports by atmosphere and ocean must balance net radiation imbalances at the top-of-the-atmosphere. The combination of local budgets of these quantities in the atmosphere, the ocean, and at the interface produces over-specification requiring conciliation among different answers and, thus, provides constraints on system behavior. In the absence of sufficient quality observational data, these multiple internal checks will need to be extensively exploited.

Because requisite data sets may not exist or may exist in unusable forms, CSM development activities will connect closely with data base development and validation efforts across the community to make known the data needs of the CSM effort. A start in this direction is already underway within SCD and CGD, working collaboratively with universities and NOAA under the U.S. Global Change Research Program's Geosystem Database initiative.

The computational requirements of a comprehensive CSM are enormous and will require dedicated, up-to-date computer facilities. Inevitably, the development of CSM will require undesirable trade-offs between computational expense, resolution, and detailed treatments of physical, chemical, and biological processes. Incorporation of chemical and biogeochemical cycles will greatly increase the cost over that for the physical climate system alone. If the CSM project is to provide a truly state-of the-art model to the scientific community, it is imperative that the computational constraints not force excessive compromise.

There must be sufficient computer power available to simultaneously support: ongoing development of the coupled model and of the underlying component models; long (at least century) production integrations of the comprehensive coupled model; and decadal integrations of higher resolution, less comprehensive coupled models. Experience suggests that it will not be possible to maintain interest and participation in the CSM project if it is necessary to stop development to make long production simulations, or if the production simulations take many months each to complete.

The initial CSM computing requirements are being addressed through the Climate Simulation Laboratory (CSL) within the NCAR Scientific Computing Division. As of October, 1994, the CSL consists of 5 Cray Research Inc. computers, a YMP8-864, a 64 processor T3D, and three EL98's. These machines will meet the requirements of the CSM in the short term, and will allow reasonable resolution in the atmosphere and ocean components of the physical climate system. They will not be adequate to allow routine inclusion of chemical and biogeochemical cycles in the CSM for long integrations, and they are, in total, less powerful than the computers available at several climate research centers internationally. Substantially augmenting the initial CSL computers will be required if the CSM is to include chemistry and biogeochemistry and is to be widely used by the scientific community in the U.S.

The CSM project will be overseen by a steering committee consisting of the NCAR director, associate director, and the directors of the science divisions. Guidance and oversight will also come from the Scientific Advisory Committee for the National Science Foundation (NSF) Climate Modeling, Analysis and Prediction (CMAP) program (see appendix, section VIII). Scientific direction of CSM will come primarily from the CSM Investigator Group (CSMIG), composed of those NCAR and outside scientists actively participating in CSM. It is expected that the composition of this group will vary with time as some participants move on to other interests and new investigators become involved. The Chair of the CSMIG is appointed by the NCAR director in consultation with the CSMIG. During the start-up phase of CSM, Byron Boville and William Holland are co-chairing the CSMIG.

The CSMIG co-chairs will lead the collaborating team of scientists involved in the project. They will maintain a consensual, decision making approach and will attempt to resolve scientific debate within the project in a constructive fashion. They will describe and advocate the CSM project in a variety of venues, including at scientific meetings and at funding agency briefings. They will coordinate, and be responsible for, the development of plans and proposals. They will attempt to engage the participation of a broad spectrum of scientists in the CSM project. They will allocate the resources available to the project (including computer time and programming support) in consultation with the CSMIG.

CSM activities described in this plan will occur largely through the efforts of relatively small teams of scientists. These teams will work on component models and on coupling strategies, and all team members will be members of the CSMIG. Each team takes responsibility for developing and continually improving its component of the CSM consistent with the CSM goal of a fully-coupled model and with the CSM design criteria. Each team will declare its development priorities and work schedules. During the CSM start-up phase, CSM teams will consist largely, though not entirely, of NCAR scientists. The eventual success of the long-term CSM effort requires full and active participation by non-NCAR researchers as CSM members and an active effort to develop and maintain close collaborations over considerable distances.

The complexity of climate issues and the evident and as-yet undiscovered difficulties in designing, assembling, and testing a coupled climate system model require that the CSM project adopt and adhere to long-term goals and standards. Within and consistent with that long-term effort, we defined a start-up phase of approximately two years that began in January, 1994. We recognize the need to show sufficient progress and skill so that, by the end of the start-up phase, the CSM can serve as the centerpiece for a long-term community climate modeling research program and as a useful tool for addressing policy-driven questions about climate change. During this phase, CSM development will rely heavily on people and climate modeling research now underway at NCAR, coordinated toward the CSM goals and supplemented in many instances by direct solicitation of university and federal laboratory collaborators. Outside guidance and oversight will come from the Scientific Advisory Committee of CMAP.

During the start-up phase, the prototype CSM will be developed, tested, and probably revised extensively. Development of the prototype model is already proceeding, with initial coupled atmosphere-ocean calculations having begun in April, 1994. The initial release of the CSM for wider use by the scientific community will take place at the end of the start-up phase. The highest priority short-term goals of the CSM project are: to develop a global coupled atmosphere, ocean, land surface, sea ice model from existing component models; to determine the ability of the model to reproduce the current climate, including interannual and interdecadal variability; to attempt to significantly reduce the largest errors of the initial coupled model; to use the resulting model for global change studies, including the Intergovernmental Panel on Climate Change (IPCC) global warming scenarios; to develop a hierarchy of subset models for addressing specific problems on shorter time scales, e.g., a global atmosphere coupled to a tropical Pacific upper ocean model to address the ENSO-related variability; and to couple carbon and sulfur cycle models into the CSM.

There are also several lower priority goals that the CSM will actively address, largely taking advantage of existing efforts and interests in specific areas. These lower priority goals address problems that will be important to the CSM project in the longer term and will entrain wider participation in the early stages of CSM development without significantly impacting the highest priority projects. These goals include:

• development of a middle atmosphere GCM including interactive ozone chemistry; and
• coupling of an upper atmosphere model to the lower atmosphere of the CSM for studies of the interaction of the lower atmosphere with the mesosphere and thermosphere.

At the end of the CSM start-up phase, we intend to have publishable solutions for a variety of coupled phenomena (see below); to have a set of models that can be used for climate system research by a broad segment of the scientific community; and to be in a position to contribute to the scientific basis of national and international policy discussions relating to global change by running scenarios for changes in greenhouse gases. The models available at the end of the start-up phase will serve as the basis for continuing development based on the research interests of the expanding user base.

According to the design criteria of the previous section, CSM will develop from existing components and attempt, during initial calculations, to couple those components in somewhat simplified form. In choosing initial calculations, we attend to the need to demonstrate that the CSM project has: engaged the active collaboration of a wide selection of scientists; developed a viable approach to building a CSM; and demonstrated substantial progress in learning about coupled behavior during the CSM start-up phase. Our work on these particular initial calculations will allow us to meet those needs and lead to more complex and comprehensive coupling efforts as skill develops.

We propose the following strategy for selecting and conducting initial calculations:

• Work on coupled problems in parallel with work on component models, testing the latter at least partly in uncoupled mode;
• Attempt to remove as many biases as possible in the surface exchanges of energy and momentum of component models in uncoupled mode;
• Start with equilibrium climate as our principal target problem to address climate drift of the coupled system from the beginning;
• Start with primarily physical coupling and rapidly work toward including chemical, biogeochemical, and ecological aspects;
• Define our target resolution as the maximum computationally-feasible resolution over the next two years for coupled simulations of at least 100 years.
• Evaluate the benefits of this target resolution compared to lower resolutions in the face of vastly increased costs for high resolution; and
• Sustain effort on several specific problems motivated by particular scientific issues rather than emphasizing repeated calculations of just one comprehensive problem.

The initial CSM calculations have been selected according to the above guidelines and computational constraints imposed by the phased upgrades planned for the CSL. Extensive coupled calculations could not begin prior to the availability of a YMP8 in October, 1994. However, a great deal has been accomplished in the interim period.

The first nine months of the CSM project (January--September, 1994) has been a preliminary development phase. During this phase several activities have taken place:

• Configure and test an updated atmosphere model based on CCM2;
• Configure and spin up the ocean GCM;
• Compare the ocean heat transport simulated by the ocean model to the implied ocean heat transport in the atmosphere model with observed sea-surface temperatures. The implied ocean heat transport is the transport that would be required to maintain the fixed sea-surface temperatures against the simulated surface fluxes;
• Complete the implementation and testing of the prototype flux coupler;
• Run the atmosphere model driven by the flux coupler with observed sea-surface temperatures instead of an ocean model;
• Run the ocean model driven by the flux coupler and state variables from the previous simulation with the atmosphere model;
• Run the atmosphere model driven by the flux coupler and sea-surface temperatures of the previous simulation with the ocean model; and
• Begin low-resolution coupled simulation (R15 atmosphere, ocean and sea ice).

Initial Calculations

The CSM coupled calculations during the start-up phase will focus on two problems, equilibrium climate and seasonal to interannual variability. The first problem forms the basis for all climate change experiments, including greenhouse gas induced global warming, and consequently for contributions to the IPCC process. It also supports the modeling component of focus 2 of the World Climate Research Programme (WCRP) CLIVAR program. The second problem contributes to the modeling component of CLIVAR focus 1 and the GOALS program in the U.S.

Equilibrium Climate

This is the essential problem to be addressed during the start-up phase of the CSM project. This calculation addresses the question of whether the improvements that have been made in the atmosphere and ocean model components over the last several years are sufficient to allow coupled simulations that do not drift greatly from the presently observed climate. Extensive analysis of the simulated annual cycle and interannual variability will be required. We intend to run the coupled model without ``flux adjustment'' at the atmosphere-ocean interface. Based on previous experience at NCAR and elsewhere, it is expected that the initial coupled calculation will drift away from the present climate and that several revisions of the atmosphere and ocean models will be required. We will not resort to ``flux adjustment'' unless it is absolutely required to produce a usable model within the two year start-up phase.

This problem presents an excellent first test to CSM because it is the prerequisite to all global change experiments and to interdecadal variability studies. Although later revisions of the CSM will be integrated for a least 100 years, integrations of a decade or so will probably be sufficient to establish the initial drift of the early experiments.

Given this physical model as a base, we can investigate additional phenomena with developmental, partially-coupled models on the same space and time scales: (a) stratospheric ozone cycle; (b) tropospheric sulfur cycle; (c) carbon cycle with terrestrial ecosystems and an abiotic ocean; (d) oceanic primary productivity and phytoplankton distributions with a nitrogen based marine ecosystem model; and (e) lower atmospheric driving (one-way coupling) of the upper atmosphere. Terrestrial and ocean biogeochemical models will be added as diagnostic components to the physical models at an early stage. The research plans for the component models provide descriptions of the approaches to these additional calculations.

Component models:

• Atmosphere model based on CCM2 with many improvements to its numerical formulation and physical parameterizations;
• Land surface (biophysical exchange) model (Bonan, 1994);
• Full-depth, ocean with isopycnal tracer mixing;
• Flato-Hibler sea ice with extended thermodynamics (snow, etc.); and
• Prototype flux coupler for ocean, ice, and marine atmosphere. The variables exchanged and the exchange protocols have been established and the prototype coupler is complete.

Initial conditions:

Atmosphere and land initial conditions will come from an uncoupled CCM2 solution. Ocean initial conditions will derive from a near-equilibrium solution calculated with a full-depth ocean model using mean seasonal forcing of climatological wind stress and temperature and salinity boundary conditions that restore towards climatology. The ocean will then be run for several decades using boundary conditions from the atmosphere model to yield initial conditions for the coupled model.

Issues:

The target resolutions for this calculation are T42 for the atmosphere (18 levels) and for the ocean (45 levels) and ice components. Alternative resolutions are also being explored and the benefits of the target resolution will have to be established. The duration of the initial calculations and the criteria for validation have not yet been established. We expect to resolve these issues quickly. However, the availability of computer time will inevitably play an important role in the choices made. We will improve the quality of the solution iteratively through successive integrations and explore several alternative choices.

Seasonal to Interannual Variability

We will design seasonal and interannual calculations to address the mechanisms responsible for the coherent, coupled variations of the ocean and atmosphere on seasonal and interannual time scales. Such variations will include ENSO, midlatitude ocean-atmosphere fluctuations, and variability in the thermohaline circulation. We will use the ocean model in different configurations to address the different phenomena mentioned above. For example, to address ENSO the ocean model will require resolution sufficient to resolve important frontal structures and wave dynamics in the equatorial region. To address midlatitude ocean-atmosphere fluctuations, for example in the North Atlantic region, the ocean model will require eddy-resolving resolution in the North Atlantic Ocean and coarser (non-eddy-resolving) resolution elsewhere in the global ocean. The sea ice component will also be configured appropriately for different phenomena; i.e., it is not important to ENSO variability but it is absolutely vital to studies of variability in the thermohaline circulation. Eventually it will be possible to have a truly eddy-resolving global ocean model as part of the CSM, but this will require much greater computational resources.

The CCM2 and land models will be run with sufficiently high resolution (T42 or higher for these studies), and the prototype flux coupler for ocean, ice, and marine atmosphere will also be used. The flux coupler approach allows regional ocean models to be combined with observations from the rest of the ocean in an easy and consistent way. The initial conditions for these various studies will also be chosen according to the phenomena being addressed.

At the end of the startup phase, a CSM version 1 will be frozen and documented extensively. The source code and accompanying technical descriptions and users guides and the results of control and greenhouse gas scenario experiments will be made available to the scientific community. Plans are still being developed for the distribution and support of the CSM and for fostering the continuing involvement of an active user community in the ongoing development and use of the CSM. These plans will be completed in consultation with the Scientific Advisory Committee for CMAP and will depend on the availability of resources to support the CSM. This work will require the efforts of a dedicated core group of programmers as well as the input from NCAR scientists, patterned after the successful development and application of the CCM atmospheric model and its easy availability to the outside community. The personnel requirements will be discussed in Section VI. Here we will discuss some aspects of how the NCAR CSM activities relate to the wider climate modeling community.

The model will be used by CSM investigators (from within and outside NCAR) to address a wide variety of scientific problems. The CSM will also be freely available to any interested parties, whether or not they desire any further involvement with the onging CSM project. Experience with the CCM suggests that to foster maximal use and benefit from this new tool for climate studies, NCAR will need to not only make the basic CSM model easily available, but to supply documentation and other support to outside users. Thus the CSM will be made available at the appropriate time by anonymous FTP, extensive documentation of the code will be provided, workshops will be undertaken by NCAR to actively aid a users group, and a visitor/postdoc program will be supported at a level depending upon resources. Following the CCM experience, a first workshop in the summer of 1996 (and one every two years thereafter) would allow a rapid dissemination of knowledge and expertise on how to use the model to the outside community of users.

Development and support of analysis tools will also be a vital part of the CSM project and will be required by outside users in addition to the CSM itself. Experience with the CCM indicates that support of large analysis package across the widely distributed and diverse computing environments of the CSM users would be a major undertaking. Such a project would probably be beyond the resources likely to be available to the CSM project and would probably not lead to a product satisfactory to a large number of the CSM users in any case. Plans for CSM analysis tools are still being developed, and will evolve with user needs, but will probably focus on standardizing data formats to those supported by many analysis and visualization packages (e.g. netCDF). Widely available commercial and public domain software packages could then be used to analyze CSM results. A set of tools to manipulate the data files and visualize the results will be developed, in large part as modules within existing software packages.

There are already strong links between NCAR modeling activities and the outside community. Many scientists at NCAR have very active collaborations with the university community and close involvement with activities in other agencies and programs supporting climate research [DOE (CHAMMP, LLNL, LANL); EPA (GENESIS); NASA (EOS, GSFC, JPL, WSRP); NOAA (Climate data reanalysis)]. Discussions on cooperation between CSM and some of these programs have already started. Thus it is expected that involving the outside community in CSM will occur naturally and rapidly in the early stages of CSM development and use. Almost half of the planned CSM budget will be used for short and long term visitors, postdoctoral fellows, and workshops.

Scientific Problems to be Addressed by CSM

Several important scientific problems require a CSM beyond the scope of existing NCAR models and include models of physical, chemical, and biological processes that transcend disciplinary boundaries. The CSM project will not only develop the model and make it available, but will also use the CSM and its subset models to address many of these problems. Examples, roughly in order of priority to the CSM project, include:

Equilibrium physical climate. The first priority of the CSM project is to develop a model that more accurately represents the current state of the physical climate system, including its mean state and annual and interannual variability. This is required for increasing confidence in global change studies. Simulating the present climate requires coupled climate system models, including the atmosphere, ocean, land surface, and sea ice. It is the starting point for studies of chemical and biogeochemical cycles.

Global warming. The possibility of significant global warming motivates much research on the climate system. Determining the extent of greenhouse gas warming, the impacts of warming on other aspects of climate (precipitation, solar radiation at the land surface), and the spatial variability of warming effects require coupled climate system models. Determining the time evolution of greenhouse effects also requires coupled models: it depends upon the complex behavior of the coupled systems (e.g., land, carbon cycle, ocean). NCAR has a long history of global warming research and participates in IPCC simulations with coupled models.

Comprehensive annual cycle. Simulation of temporal and spatial components of the annual cycle presents a demanding and attractive test to coupled models. Simulation of the annual cycle involves all aspects of the climate system (atmospheric physics and chemistry, the hydrologic cycle, ocean dynamics, and biology/biophysics) but without important and difficult longer term processes of deep ocean circulation, vegetation change, and carbon cycle dynamics. Accurate annual cycle simulations should include physical climatology and annual cycles of important trace gases (O, CH, CO) and of terrestrial characteristics, such as soil moisture and leaf area.

Interannual variability. Interannual variability on time scales of a few years involves interactions of the atmosphere and the land surface through changes in surface wetness and snow cover. Sea ice variations are also important. However, interactions of the atmosphere and the upper ocean, particularly the tropical oceans, may be of dominant importance on this time scale. Studies on this time scale will require more detailed upper ocean models than will be used for equilibrium climate interdecadal variability. The CSM project begins with a solid base of experience in coupling atmospheric and tropical upper ocean models, in coupling atmosphere and land surface models, and in coupling atmospheric and global ocean models.

Interdecadal variability. Detecting long-term climate change requires an understanding of climate variability on decadal time scales. Interdecadal variability can be forced by processes external to the atmosphere-ocean-land-ice system, such as volcanic emissions, solar variability, and anthropogenic effects. However, interdecadal variability is also an intrinsic aspect of the atmosphere-ocean-land-ice system, resulting from internal dynamics in the atmosphere and from atmospheric interactions with oceans and land. For example, interdecadal variabilities in land, ocean, and atmosphere systems probably play roles in Sahelian drought cycles, monsoon variations, and other low-frequency phenomena.

Sulfur cycle. Climate researchers have long recognized the importance of sulfate aerosols in climate, and this problem has recently returned to the forefront of scientific interest. Studies of sulfur effects require that coupled models include elements of tropospheric chemistry and represent anthropogenic and biological controls over tropospheric SO and NH concentrations and vertical distributions within the coupled physical climate system.

Carbon cycle. Understanding the role of the carbon cycle in the climate system presents a major scientific and practical challenge to earth and life sciences. Only coupled models can consider the full range of factors influencing the carbon cycle: human land use, CO-fertilized growth in regrowing and natural areas, climate feedbacks to ecosystems, and climate effects on carbon storage through physical and biological processes. The CSM effort begins an essentially new effort in this area, but one that starts from existing strengths in modeling terrestrial carbon dynamics and in tracer transport in the atmosphere and ocean (e.g., Schimel et al., 1994; Erickson et al., 1994).

Middle atmosphere circulation and ozone depletion. Catalytic ozone destruction caused by anthropogenic gases is one of the key global change problems. Until recently, most modeling of ozone chemistry and transport has taken place using two-dimensional models (in the latitude-height plane). It is now generally recognized that several key transport processes require a three-dimensional model to treat them credibly (e.g., convection, stratosphere-troposphere exchange). NCAR has existing expertise in modeling dynamics and transport in the middle atmosphere (Boville, 1991; Rasch et al., 1993) and in modeling ozone chemistry (Brasseur et al., 1990). Work on a coupled dynamical-photochemical model for ozone has already started. Eventually, global warming problems may be treated with an interactive ozone chemistry component, although these problems will be treated separately in the early stages of the CSM project, and the ozone chemistry will be done in models without an interactive ocean.

Glacial-interglacial transitions. There is conflicting observational evidence suggesting that transitions between extreme climate conditions (e.g., between ice ages and warmer interglacial periods) can occur on short, perhaps even decadal, time scales with several transitions within the span of a few centuries. Investigating such phenomena requires a reasonably complete and skillful model of the coupled climate system. This is a long-term goal of CSM and will be an active area of research in the later stages of the project.

Further Developments

The particular model components that comprise the CSM at the start of phase II are being selected by NCAR scientists and a small number of external collaborators. These choices depend on a number of factors, including, but not limited to, scientific merit.

NCAR management and CSM scientists are committed to the principle of continuous evaluation of the CSM and its components, and the replacement of existing components with others, from wherever, that are of higher scientific quality. This commitment includes the task of defining a process for such evaluations.

While component models for the atmosphere and ocean have been available for some years, and at least partial land models for nearly as long, relatively few coupled simulations have been conducted. The rate of progress has been retarded by the difficulties of coupling models, by persistent problems with coupled simulations (e.g., ``climate drift''), and by the lack of adequate computing resources at every major institution involved in such calculations. In view of the level of effort involved in developing coupled models and of the resources required to run them, it is important to learn as much as possible from existing and planned calculations.

The first coupled ocean-atmosphere model calculations (see review by Meehl, 1990) were conducted at GFDL in the late 1960's with a simple sector model (Manabe and Bryan, 1969). That group continued coupled model development through the 1970's leading to a global ocean-atmosphere coarse grid model with asynchronous coupling (Manabe et al., 1975; Bryan et al., 1975).

The first global coupled ocean-atmosphere-sea ice model calculation at NCAR also used asynchronous coupling in light of limited computing resources (Washington et al., 1980). Several other groups began global coarse grid coupled modeling activities in the 1980's (e.g., Max Planck Institute in Germany, the Hadley Centre in England). A transient CO increase calculation performed by researchers at NCAR with a global coupled coarse grid model (Washington and Meehl, 1989) was one of several model simulations that contributed to the IPCC reports (IPCC, 1990, 1992). In that model earlier results performed with simple sector coupled models were confirmed in that the oceanic thermohaline circulation weakened with increased CO in the atmosphere. Similar results with a global coupled coarse grid model were documented about the same time by researchers at GFDL (Stouffer et al., 1989). A streamlined version of the GFDL global coarse grid coupled model has recently been run for several 500-year integrations to document century time scale variability of the thermohaline circulation with various CO increase scenarios (Manabe and Stouffer, 1993). The results show a suppression of the thermohaline circulation, which appears to be irreversible in the highest CO scenario.

Currently, Washington and Meehl are running a global coupled model with increased ocean resolution (, 20 levels). Control and transient CO increase experiments are being performed with and without flux correction. This coupled model will be used to simulate the effects of IPCC CO scenarios on climate processes and to study further inherent variability in the coupled system on various time scales. While this experiment will be conducted outside the CSM flux coupler framework, it will include CCM2, a CSM component. The study will be a useful reference point for the subsequent development and testing of the flux coupler concept described below. The NCAR contribution to the current IPCC 1995 assessment will be based on the above experiments, and contributions to subsequent assessments will be based on the new coupled model described in this document.

The focussed research within the CSM project is built upon a general base of NSF-funded climate research at NCAR, as well as several closely affiliated climate modeling projects funded by other agencies. The scope and quality of this research are greatly enhanced by extensive collaborations with university and federal laboratory scientists.

A large part of the climate research at NCAR takes place within the Climate and Global Dynamics Division (CGD). The broad research objectives of CGD are to promote further understanding of the physical causes of present and past climates and of large-scale atmospheric and oceanic dynamics, thus contributing to the basis for the scientific prediction of weather and climate. During the last few years, CGD has engaged in the following research activities geared toward achieving these objectives: (1) release of CCM2 to the scientific community; (2) investigation, with the CCM and other models, of the mechanisms of global climate variability and change and exploration of the basis for long-range prediction of weather and climate; (3) study of the dynamics of global oceans and interactions with the atmosphere; (4) evaluation and analysis of global observational data sets for the study of climate variability and validation of climate models; (5) study and modeling of clouds, their radiative properties, and their effects on global energy balance from satellite observations; (6) description and modeling of the interface between the atmosphere and the oceans, cryosphere, chemosphere, and land--biosphere within the concept of coupled climate systems; (7) examination of the mesoscale implications of large-scale climate changes; and (8) study of the relationship between ecosystem processes and the atmosphere. The CSM project will draw upon this ongoing research for many of its modeling components, as well as analysis and validation of simulations with observations.

NCAR scientists have continued development of, and research with, the Global Environmental and Ecological Simulations of Interactive Systems (GENESIS) earth system model. The GENESIS model is designed to facilitate research in terrestrial surface processes and paleoclimate. Part of this research supports the Paleoclimate of Arctic Lakes and Estuaries (PALE) program. The current GENESIS model (version 1.02) has been in use since January, 1992, by a number of groups within and outside NCAR. This work has been supported by the Environmental Protection Agency (EPA).

The climate sensitivity to increased CO work of Washington and Meehl, described above, is supported by the Department of Energy (DOE) Global Change Research Program. The development of versions of the CCM and other climate modules for a variety of massively parallel processor computers is being pursued with funding from the Computer Hardware, Advanced Mathematics and Model Physics (CHAMMP) program (DOE). Also under CHAMMP is the development of a semi-Lagrangian version of the CCM2 and a simplified, workstation-based, numerical framework for parameterization development, as well as benchmark models for testing numerical methods. Important studies in radiation and clouds using a two-dimensional cumulus ensemble model continue with support from the Atmospheric Radiation Measurements (ARM) program, also funded by DOE. In addition, scientists continue to participate in the Atmospheric Model Intercomparison Project (AMIP) sponsored by the Environmental Sciences Division of DOE.

A modeling component, especially with regard to land surface processes, terrestrial ecosystems, and the use of satellite data for validation, is included in an interdisciplinary project of the National Aeronautics and Space Administration's (NASA) Earth Observing System (EOS). The latter also supports satellite and other data set development for use with models and development of an adjoint model for examining sensitivity to input data and parameterizations. Further, EOS funding supports the development of ecosystem dynamics modeling for use in global applications and coupled earth system modeling of the carbon cycle.

Several NCAR projects are supported by the National Oceanic and Atmospheric Administration's (NOAA) Office of Global Programs under the Climate and Global Change Program that help develop the ocean component of climate models, explore coupling atmospheric and ocean models especially for ENSO studies, and explore processes important in interannual and low-frequency variability. In addition, NCAR scientists are working collaboratively with NOAA in the production of reanalyzed long-term climate data sets.

The Japanese Central Research Institute of Electric Power Industry (CRIEPI) helps support development of a high-resolution regional climate model RegCM2 that can be driven by global analyses as boundary conditions. Also contributing is the Italian Agency for New Technologies, Energy and Environment (ENEA) by supporting regional climate modeling, studies of coupled ocean-atmospheric forcing, and the effects of resolving boundary features in global warming scenarios. Other support in providing computer time for climate modeling has come from the Model Evaluation Consortium for Climate Assessment (MECCA).

NCAR scientists participate in many national and international programs that are also relevant to CSM. Our participation in the CMAP program, Global Tropospheric Chemistry Program (GTCP), Tropical Ocean and Global Atmosphere (TOGA) and the TOGA Coupled Ocean--Atmosphere Response Experiment (TOGA COARE), World Ocean Circulation Experiment (WOCE) especially through the WOCE Community Modeling Effort (CME), Earth Radiation Budget Experiment (ERBE), Role of Clouds, Energy and Water (ROCEW)/Global Energy and Water cycle Experiment (GEWEX), Atlantic Climate Change Program (ACCP), and International Satellite Cloud Climatology Project (ISCCP) is strong evidence of our contribution to the U.S. Global Change Research Program. Efforts are ongoing in Geosystems Databases and data reanalysis with the development, documentation, analyses, verification, and distribution of data sets to a wide user community.

The core NCAR NSF support, together with other agency funding for specific projects, provides a strong base for the NCAR CSM. University and federal laboratory scientists will contribute to the effort by collaborating with NCAR scientists or as independent investigators. The results of applications of the models through these efforts are expected to feed into the CSM, while the CSM develops new modules and models that can be employed and exploited by these efforts. The CSM will also draw upon the diverse activities ongoing within CGD, the Atmospheric Chemistry Division (ACD), the Mesoscale and Microscale Meteorology Division (MMM), and the High Altitude Observatory (HAO), thus providing opportunities for alternate model components to the CSM. Through its various versions, the CSM will eventually be tailored to such specific problems as droughts, global change, ozone depletion, long-range prediction, ENSO phenomena, solar-weather relationships, land surface changes, paleoclimate studies, and ecosystem dynamics.

The component model plans appearing below describe the research and development plans of a large number of individuals gathered into several groups that roughly correspond to the subheadings under the major components. Each group or individual has plans to build better component models, and the priority given to various parts of the plans will be guided by the needs of the CSM project and by the interests of the individual scientists. These plans will certainly change as experience is gained with the components in the CSM context.

Chemistry appears in some form within each of the component models rather than as a separate component because the important processes and compounds are different within the atmosphere, ocean, and land. For example, CO in the atmosphere is an inert, nearly well mixed gas whose concentration is controlled by sources and sinks at the surface, modified by atmospheric transport (Erickson et al., 1994; Tans et al., 1990). However, CO emission and uptake by the land and ocean biota are controlled by complex biogeochemical processes that are crucial components of the climate system (Schimel et al., 1990; Sarmiento et al., 1993).

The most mature component models are the atmospheric and oceanic GCMs and, to a lesser extent, the land biophysical model. Therefore, their plans describe specific, current research issues that are expected to result in model improvements in the immediate future. The chemical and biogeochemical models of the atmosphere, ocean, and land are less mature than the physical models. In consequence, the plans given below are more general for these models, although specific projects are given for the short term.

NCAR CSM: Atmosphere Component

Overview

Atmospheric modeling, with roots in early atmospheric climate simulations begun in the mid 1950's, represents the most mature aspect of climate system modeling. Indeed, much of what is currently understood about the role of the atmosphere in the climate system has been discovered through the judicious use of the experimental testbed afforded by atmospheric general circulation models (AGCMs). Despite this long history, climate system modeling presents a demanding scientific challenge to those modeling the atmosphere. The atmospheric model will be required to simulate atmospheric behavior in a coupled interactive context, a more demanding requisite than merely simulating the atmosphere with the strong constraints of specified ocean surface temperatures and sea ice distributions. The atmospheric model will also be required to accurately transport chemical constituents, and in some configurations, drive an upper-atmospheric model with momentum, heat, and constituent fluxes.

NCAR has a long history of developing and using AGCMs, starting with the grid-point models of the late 1960's and early 1970's designed by Kasahara and Washington and progressing to the spectral CCMs of the present day. The latest model in this series, the NCAR CCM2, forms the basis for the atmosphere component of CSM. CCM2 is usually configured as a tropospheric model with a top in the upper stratosphere but can also be configured as a middle atmosphere model, extending from the surface to the upper mesosphere. A land biophysics model (Biosphere-Atmosphere Transfer Scheme (BATS)) is presently included in CCM2, but will soon become part of a separate land component model.

Radiative forcing is directly influenced by the abundance in the atmosphere of trace gases such as water vapor, carbon dioxide, methane, nitrous oxide, ozone and the chlorofluorocarbons. Sulfate aerosols also affect the radiative budget of the Earth/ atmosphere system. Industrial, as well as agricultural, activities have perturbed significantly the chemical composition of the atmosphere at the regional and global scales, and hence the climate forcing. It, therefore, is important that the interactions between biogeochemical cycles and the physical climate system be accurately represented in the CSM. We will develop chemical modules to be included in the atmospheric component of the CSM and to account for the exchanges of chemical compounds with the continental biosphere and the oceans. The chemical modules will be integrated into the full CSM after careful testing and validation. The emphasis will be on the cycles of carbon and sulfur, and their impact on the climate system. In addition, the role of other radiative gases such as ozone and methane, which contribute to the ``greenhouse effect,'' will be investigated.

We also present subsections on a nested regional modeling capability that will be included in the CSM, and an upper-atmosphere model that may eventually lead to a merging of CCM2 with upper-atmosphere components to give a full atmosphere component for the CSM.

Community Climate Model

Over the last three years, the Climate Modeling Section integrated a large number of simulation capabilities into a single cohesive modeling framework, the CCM2. The most recent version (CCM2.1) includes BATS for computing surface exchanges over land. The individual and collective behavior of various model components were validated against available observational data. These components were modified accordingly, followed by major refinements to the overall implementation for optimal computational performance.

The result of this effort, the NCAR CCM2, is a new AGCM for which most aspects of the formulation represent improvements over the CCM1. The principal algorithmic approaches carried forward from CCM1 are the use of a semi-implicit, leap-frog time integration scheme, the use of the spectral transform method for treating dry dynamics, the use of a bi-harmonic horizontal diffusion operator, and the large-scale condensation process. The CCM2 makes use of new algorithms for both resolved dynamics and parameterized physics, which are documented in Hack et al. (1993). The standard model configuration uses a horizontal spectral resolution of T42 ( transform grid), 18 vertical levels, and a top at 2.917 mb. It employs a 20 minute time step by dynamically adjusting the spectral resolution of the top layer to maintain a Courant number of less than one. A middle-atmosphere version of CCM2 is also maintained. This version uses 44 levels between the surface and 0.017 mb ( km).

Two major improvements are included in the CCM2 dynamical formalism. The first is the incorporation of a hybrid vertical coordinate that follows terrain near the surface (traditional sigma) and transitions to a pure pressure coordinate above about 100 mb. A second major change to the resolved dynamics is the incorporation of a shape-preserving semi-Lagrangian transport (SLT) scheme (Williamson and Rasch, 1994) for advecting water vapor. This scheme can also be used to transport an arbitrary number of other scalar fields (e.g., cloud water variables, chemical constituents, etc.) as required by a CSM. The use of the SLT method largely addresses the many numerical problems exhibited by the spectral advection approach used in earlier versions of the CCM.

Aspects of the CCM2 dynamic, thermodynamic, and radiative climate statistics are documented in Hack et al. (1994) and Kiehl et al. (1994). The physical representation of a wide range of key climate processes in CCM2 is much more physically realistic than in previous versions of the CCM. Hurrell et al. (1993) present a comparison of the climatology of CCM2 with CCM1, GENESIS, and the atmospheric component of the coupled ocean-atmosphere model of Washington and Meehl (1993). Overall, the model climatology is substantially better than climates produced by previous versions of the model, which generally suffered from a variety of large biases in the mean climate state. The most serious of these earlier biases, a systematically cold and dry model troposphere, has to a large extent been eliminated. Although the simulated climate continues to be cold, it is now generally within to K of what is analyzed, exhibiting excellent agreement with analyses and observations in the tropics. The simulated three-dimensional water vapor distribution is also in very good agreement with both analyses and selected observational station data. The most serious coupled problems in CCM2 are likely to be its excessive latent heat fluxes, precipitation, and surface stresses in the tropics.

Actions during Start-up Phase

During the CSM start-up phase, the developers of CCM2 and other atmospheric modelers will work together with the ocean modelers to carry out and analyze simulations with the coupled land-atmosphere-ocean model. The atmospheric scientists will work on problems that arise during coupled calculations in a collaborative mode with other members of the CSMIG. Many improvements will be included in the model, based on the long-term research projects listed below.

The standard T42, 18-level CCM2 with BATS has been restructured to interface with the flux coupler and will provide the land and atmosphere components for the initial development of the CSM. An independent land surface model, with its own history carrying and restart capabilities, is being developed based on existing land components of the CCM2 code. The independent land model will couple to CCM2 through the flux coupler to facilitate introduction of alternative land surface models.

Long-Term Research

NCAR and other scientists will continue to concentrate on improving atmosphere components of the CSM. In the following we list areas of immediate concern and scientists who are committed to work on them. Although this section is referred to as long-term research, most efforts are currently underway. When improvements are documented and acceptable to the CSMIG, they will be incorporated in the CSM.

Large warm bias in simulated land surface temperature over the Northern Hemisphere during July. Other aspects of surface climate, such as precipitation and surface energy exchanges, are strongly affected by this bias. The principal source of this problem is the diagnosis of cloud optical properties. Techniques that more realistically diagnose cloud droplet effective radius as a function of space and time and cloud liquid water path as a function of the space- and time-dependent atmospheric state have been independently explored by Kiehl (1994) and Hack (1994). The combination of these two approaches substantially reduces many temperature and precipitation biases observed in CCM2 Northern Hemisphere summer circulation. (Kiehl, Hack)

Anomalous positioning of deep convection in the Western Pacific. This anomalous position affects wave propagation and long-wave positioning in January simulations, contributing to excessive ridging over the North Pacific and a reduction in the height field over Western North America. The incorporation of improved cloud optical properties results in a northward displacement of the diabatic heating, a correspondingly improved Australian monsoon, and more realistic ridging throughout the troposphere along western North America. The magnitude of the precipitation maxima remains unreasonably large, and its position remains too far west, apparently associated with excessive boundary layer transport of water vapor in deep convective regimes. (Hack, Tribbia)

Wind stress over the tropical Pacific remains too strong compared to estimates from atmospheric analyses. In this region model stresses actually converge with increasing resolution but to values different from the atmospheric estimates (Williamson et al., 1994). These model stresses are expected to have serious consequences when coupled to an ocean model. Research will be directed toward identifying and removing this deficiency. (Tribbia, Gent, Boville)

Simulated polar tropopause temperatures are significantly colder than observed, particularly in summer. This problem is common to almost all AGCMs and its source is currently being investigated. Possible deficiencies in both the thermodynamic (cloud/ radiation) and dynamic (gravity wave) parameterizations are being examined. (Boville, Saravanan)

Orographic locking of precipitation. Spurious precipitation over high terrain can have a serious impact on land surface models. Two possible causes of the spurious precipitation are being investigated, elevated heat sources and numerical errors associated with terrain following vertical coordinates. (Williamson)

Numerical methods for transport of chemical constituents. The semi-Lagrangian method of CCM2 is a good starting point but not ideal. It addresses the most serious problems associated with the spectral transform method. Most chemistry researchers would be more comfortable with schemes that locally enforce conservation. Research will be direct toward adapting such schemes for global applications. (Rasch, Williamson)

The need to pre-specify the order of parameterizations within a time step. Physical parameterizations in CCM2 are approximated in a time split manner to allow efficient solutions of the implicit components incorporated in them. Since these parameterizations depend on each other, at least one of the connections is lagged one time step. The current ordering, arrived at through evolution of the CCM series, may not be appropriate for the coupled system. We will consider alternative orderings in atmospheric and coupled simulations. (Williamson)

The Antarctic winter vortex in the stratosphere is much too strong and the polar temperatures are too cold by up to 40 K in the middle atmosphere version of CCM2. This deficiency has been traced to inadequate mesospheric gravity wave drag by Garcia and Boville (1994). The middle atmosphere version of CCM2 currently uses Rayleigh friction to parameterize the effects of breaking gravity waves in the mesosphere. A single vertical profile of coefficients is employed that is horizontally uniform. This scheme does not provide a reasonable, physically-based context for improvements. More physically-based parameterizations are being considered and will be included in future versions of the model. (Boville)

Atmospheric Chemical Transport Models

ACD, jointly with CGD, has developed a hierarchy of three-dimensional chemical transport models (CTMs) that will contribute to development of an integrated CSM. One of them (Rasch et al., 1993) is run ``on-line'' with the middle atmosphere version of CCM2 to simulate the three-dimensional distribution of approximately 25 species that govern production and destruction rates of ozone in the stratosphere and lower mesosphere. The calculated distribution of radiatively-active gases such as ozone and water vapor can be used to calculate radiative heating rates so that coupling between stratospheric chemistry and dynamics is adequately represented. A tropospheric version of the CTM will be run with an appropriate chemical scheme and detailed surface emission rates and deposition velocities both within CCM2 and ``off-line'', driven by dynamical variables provided by CCM2. An ``off-line'' stratospheric version of the CTM is also available. Finally, preliminary global model studies have been performed by prototype models, such as the three-dimensional Intermediate Model for Annual and Global Evolution of Species (IMAGES).

Start-up phase

During the start-up phase of CSM, we will focus on validation of the above models based on available observations. At the same time, to address the needs of CSM, simplified approaches to represent chemical processes and their roles in the climate system will be developed. Preliminary coupling studies will also be conducted.

Model validation will be performed by comparing calculated distributions of chemical constituents with observed fields. Comparison is relatively straightforward for several stratospheric species for which satellite data are often available (LIMS, TOMS, UARS, etc.). It is more difficult for reactive species in the troposphere that have only episodic measurements from the ground, balloons, and aircraft.

Carbon Cycle Modeling

CO plays a central role in all proposed scenarios related to the climate of the 21st century. The mixing ratio of CO in the atmosphere is largely controlled by exchanges with the ocean and the terrestrial biosphere. To study the complex cycles that influence carbon on the planetary scale and to address the relative contribution of various surface CO emission scenarios in Earth's climate, several atmospheric CO tracers can be included in the CCM2 (Erickson et al., 1994). Each tracer is related to different surface source and sink representations in the model, such as air-sea exchange, fossil fuel combustion, biomass burning/deforestation, and exchange with the terrestrial biosphere.

The air-sea exchange of CO is directly controlled by the distribution of CO in the surface ocean. An observational data base for surface ocean CO is being compiled by several researchers around the world. The exchange of CO between the terrestrial biosphere and the atmosphere is a complex process that involves a variety of biological and geophysical variables. A satellite-based estimate of the magnitude and phasing of the CO uptake and release by the terrestrial biosphere is being used presently. An interactive biosphere is being developed whereby the CO exchange process is influenced at each time step by the dynamical variables in the atmospheric models. All model runs are compared to a comprehensive observational data base on atmospheric CO, at 35 sites worldwide, supplied via collaboration with the NOAA-CMDL group. A smaller observational data base dealing with the isotopic composition of atmospheric CO will also be used to constrain the global carbon cycle in an independent fashion.

The CSM effort will support the further developments of the global three-dimensional carbon modeling effort in conjunction with development of the aforementioned chemistry transport models. The global three-dimensional GCM modeling activity will be complemented by several other sub-models of varying spatial and temporal resolution. These models include soil carbon dynamics models, terrestrial biosphere CO exchange models, and surface ocean biogeochemical models, which are discussed in other sections of this document.

Sulfur Cycle, Aerosols, and Atmospheric Radiation Modeling

Several observational and theoretical studies have suggested that atmospheric aerosols influence the scattering and absorption of radiation in the atmosphere with a direct effect on surface temperatures. Atmospheric sulfur, in the form of sulfate aerosols, is thus an important component of the CSM. Preliminary modeling efforts have already computed the global distribution of anthropogenic sulfate aerosols and the possible influence on Earth's climate using CCM2 and other AGCMs. A sulphate aerosol model will be developed and coupled to the radiation calculation in the atmospheric model to assess their possible climatic influence.

Other aerosols, such as carbonaceous aerosols that are related to biomass burning and other combustion processes and mineral dust aerosols originating in desert regions, may also influence the radiative properties of the atmosphere. We will attempt to model the distributions of these aerosols based on estimates of their surface sources and incorporate their radiative effects in the atmospheric model.

Tropospheric Ozone Modeling

Tropospheric ozone is important as a greenhouse gas and as an oxidant, and its abundance has increased dramatically over the last century. The abundance of tropospheric ozone is determined by a number of complex chemical processes involving anthropogenic and biogenic precursors. Although much attention has been given to the mechanisms affecting ozone in the boundary layer, the chemical budget of ozone in the upper troposphere, where its effect on the radiative forcing is largest, remains poorly quantified (Hauglustaine et al., 1994).

A chemical transport model of the troposphere will be developed that simulates the processes that lead to chemical production and destruction of ozone and other oxidants and takes into account the surface emission and deposition of the chemical constituents, as well as their three-dimensional advective and convective transport. This model is also intended to derive the spatial and global distribution of the hydroxyl radical, which determines the global lifetime of many chemical species including methane and contributes to the conversion of reduced sulfur species into sulfuric acid (and hence sulfate aerosols). The chemical scheme used in this particular model will be based on the classic Ox, NOx, CH, CO scheme with a simplified formulation of the effect of non-methane hydrocarbons.

The prototype model for this project is IMAGES, which includes a rather detailed chemical scheme, but is limited by its simplified transport formulation. The model is also constrained by observational data, limiting its ability to treat interactions with the physical climate system. A more advanced framework will be provided by an ``off-line'' version of CCM2. In this case, the transport will be driven by the dynamical variables (winds, temperature, convection, clouds, precipitation, etc.) provided at selected time intervals. The CTM will be coupled to the land surface model of CSM, which will provide ecosystem distributions, as well as biogenic fluxes of several chemical compounds. The precipitation of other gases including nitrates provided by the CTM will be used by the land surface model as input for its vegetation scheme. Finally, an ``on-line'' version of the CTM will be developed. This will probably require that the chemical scheme be simplified to reduce the computational burden. Rather than calculating explicitly the concentration of all chemical species, a possible approach is to derive at each location of the atmosphere variations in the concentrations from given climatological values, making use of appropriate assumptions and simplifications.

Stratospheric Ozone Modeling

Ozone depletion in the stratosphere is one of the most dramatic changes observed in the Earth system. It is expected to affect the thermal structure and possibly the circulation of the atmosphere, as well as the level of biologically damaging UV-B radiation at the surface. We expect to further develop the stratospheric version of the CTM. The model will account for the effects of industrially manufactured chlorofluorocarbons and include explicitly the impact on the ozone layer of compounds belonging to the oxygen, hydrogen, nitrogen, chlorine, and bromine families. Special attention will be given to the importance of heterogeneous processes occurring on the surface of solid particles inside polar stratospheric clouds and of liquid/solid sulfate aerosol particles at all latitudes. The radiatively active species derived by the CTM (such as ozone and water vapor) will be coupled to the GCM and hence affect directly the calculated temperature and circulation in the stratosphere.

Regional Atmospheric Model

In the last several years, a one-way nested regional climate modeling capability has been developed at NCAR in which a limited area model is driven either by analyses of observations or by GCM output (Giorgi, 1990; Giorgi et al., 1993; Giorgi et al., 1994). This nested modeling technique allows us to produce long-term regional climate simulations at resolutions of a few to several tens of km. This capability can be useful to CSM in at least three respects: 1) in developing and testing the performance of physical parameterizations at high resolution; 2) in carrying out pilot coupling studies, whereby regional ocean, sea ice, hydrology, and ecosystem models can be interactively coupled with the regional atmospheric model over selected areas of interest; and 3) in bridging the scale gap between climate models and impact models, since climate scenarios at a few tens of km resolution, or less, can be produced.

Start-up phase

Recently, a second generation version of the NCAR regional climate model (RegCM2) has been completed. This version of the model includes BATS, the CCM2 planetary boundary layer and radiative transfer calculations, and three cumulus parameterization options (including that of the CCM2). In the start-up phase we plan to 1) complete a 10-year run with RegCM2 driven by the European Center for Medium-Range Weather Forecasts (ECMWF) analyses of observations over a domain encompassing the whole continental U.S. at 50 km resolution. This run will allow careful validation of the model climatology over the U.S. and identify and correct main model biases; and 2) produce multi-year runs of present day climate conditions with RegCM2 (50 km resolution) driven by CCM2 output over the U.S. and Europe. These runs will serve as validation of the coupled RegCM/CCM modeling system.

Long-term research

In the long term we plan to: 1) produce and make available to the community increasingly accurate and computationally efficient versions of the RegCM suitable for regional climate simulations at resolutions of up to 10--30 km. This will likely require development of parameterizations, e.g., for precipitation and surface processes, suitable for such resolutions; 2) carry out pilot studies with the RegCM coupled with regional ocean, sea ice, hydrology, and biosphere models. Examples of regions of interest for such coupling studies are the continental U.S. (in particular the Great Lakes basin), the northern Atlantic, the Mediterranean and Baltic Sea Basins, and eastern Asia; 3) produce high-resolution regional climate change scenarios with the RegCM driven by climate change simulations performed with the CSM. It is anticipated that, depending on computing resources, among the regions for which such scenarios may be generated will be the continental U.S., Europe, eastern Asia, and the Sahelian region.

Upper Atmosphere Model

The middle and upper atmosphere must be considered to obtain accurate models for long-term climate studies. The middle and upper atmosphere are coupled both physically and chemically to the lower atmosphere with mutual interactions important for understanding both interannual variability and long-term climate change. Dynamical interactions between planetary waves, gravity waves, and the mean circulation introduce considerable variability to the basic structure and dynamics of the middle and upper atmosphere that is not well understood. It seems certain that the dynamical state of the middle atmosphere (at least through the middle stratosphere) has a significant impact on the circulation of the troposphere through mechanisms that are also not well understood. Transport of chemically and radiatively active gases from the troposphere can significantly alter chemical and radiative balances of the middle and upper atmosphere, leading to changes in its basic structure and seasonal evolution. Effects of solar variability occur on time scales ranging from that of the Maunder Minimum through the 80-year Gleisberg cycle, the 11-year sunspot cycle, the 27-day solar rotation cycle, to shorter term solar flare, solar proton event, and geomagnetic storm variations. Finally, electrodynamic couplings between the upper atmosphere wind dynamo, solar wind/magnetosphere dynamo, and thunderstorms in the global electric circuit could have important long-term variations that influence the climate system (e.g., more lightning activity with global change).

Start-up phase

The CSM flux coupler will be used to couple the NCAR Thermosphere-Ionosphere-Mesosphere-Electrodynamics General Circulation Model (TIME-GCM) to CCM2. The TIME-GCM solves for the circulation, temperature, compositional, and electrodynamic structure of the atmosphere between 30 (10mb) and 500 km. The flux coupler will be applied at the 10 mb constant pressure surface and initially will be a one-way coupling, using fluxes of heat, momentum, and composition from the lower atmosphere as inputs to the TIME-GCM to evaluate their impact on the dynamics and structure of the upper atmosphere. Once upward coupling is examined, mutual interactions between the two models may be attempted.

This coupling represents the most expedient way of evaluating the importance of upper atmosphere structure and dynamics on CSM. Eventually, some subset of TIME-GCM physics and chemistry will be extracted and incorporated into a CSM atmosphere model that extends from ground to thermosphere, and includes couplings with the solar wind through auroral processes to achieve a consistent model structure instead of two dissimilar models coupled at the 10 mb surface. Nevertheless, much science and information can be obtained with the initial effort.

Long-term research

Problems involved in modeling the middle and upper atmosphere include:
Non-LTE radiation. Radiative cooling from CO and O is non-LTE (local thermodynamic equilibrium) above about 60 km, introducing complexity into calculation of cooling rates in the upper atmosphere. Furthermore, collisions with O strongly enhance CO radiation near 100 km, albeit with poorly-known rate coefficients, an interaction important for determining the structure of the mesopause region.

Chemical effects of particle fluxes. Reactive chemical species (e.g., NO and HO) are produced when energetic electrons and protons bombard the upper and middle atmosphere during geomagnetic storms, solar flares, and auroral sub-storms. The particle fluxes are highly variable both spatially and temporally, and during strong solar proton events, can produce as much NO in the middle atmosphere as a full year photodissociation of NO. We need to understand this variability for possible natural influences on global change.

NSF Supported Staff for CSM

Non-NSF Supported Staff Contributing to CSM

Several scientists and programmers at NCAR are supported by non-NSF funds to work on specific projects that are closely related to CSM and contribute directly to its goals. For atmospheric modeling these include:

Collaborations

Collaborators on atmospheric modeling and validation are: V. Ramanathan on the surface energy budget of the Pacific basin; B. Albrecht and W. Frank (Pennsylvania State University) on planetary boundary layer and deep convection parameterization; R. Turco (University of California at Los Angeles) on biogeochemical tracer modeling; P. Tans (NOAA Climate Modeling Diagnostics Laboratory) on the carbon cycle; and R. Oglesby (Purdue University) on climate effects of sulphur. CSM activities include links to DOE CHAMMP Program initiatives on the development of efficient component model codes and computational techniques for coupling component models on a variety of hardware platforms. Atmospheric scientists from NCAR and computer scientists from the NCAR Scientific Computing Division, Oak Ridge National Laboratory, Argonne National Laboratory, Los Alamos National Laboratory (LANL), and Lawrence Livermore National Laboratory are participating in this effort, coordinated by R. Malone of LANL.

NCAR CSM: Ocean Component

Background

Modeling of large-scale ocean circulation has a long history at NCAR. Model development and application has been a central theme of research in CGD's Oceanography Section for 20 years. Early work focussed upon process studies in limited areas to examine fundamental oceanic processes at work in isolation. At the same time, eddy-resolving oceanic GCMs were developed to examine the role played by mesoscale eddy processes in large-scale gyre dynamics. This array of work evolved in the mid-1980s into the WOCE-driven CME that was begun at NCAR to examine, for a single ocean basin (the North Atlantic), the oceanic general circulation at a variety of resolutions (, and ) in anticipation of eventually moving to the global domain (Bryan and Holland, 1989; Boning et al., 1994; Holland and Bryan, 1993a,b). This work has involved an assessment of the wind and thermohaline--driven circulation and particularly its climate characteristics (heat transports, temperature/salinity property distributions, overturning mass transports, etc.) with the view of building a better large-scale ocean circulation model.

Models of the global ocean based upon the GFDL model have also been implemented at NCAR and a variety of calculations carried out. A series of equilibrium calculations, each run for approximately 10000 years, has been carried out with a low-resolution ( latitude/longitude) version of the Modular Ocean Model (MOM), making use of many of the improvements in physical parameterization discussed below (Danabasoglu et al., 1994). Other global calculations have been carried out for shorter times at higher resolutions (Semtner and Chervin, 1992) with the original GFDL physics as a demonstration of the capability to carry out global, eddy--resolving calculations. These, together with the North Atlantic experiments mentioned above and the ENSO modeling discussed below, will form the physical basis for our future CSM ocean component model.

Coupled calculations are already underway using a variety of ocean model components. For example, Gent and collaborators have been exploring climate variability connected with ENSO. It is one of the strongest coupled modes, and, by a widely accepted hypothesis, may only involve the ocean importantly in the upper tropical Pacific. ENSO phenomena have been studied at NCAR both in a stand alone model of the upper, tropical Pacific Ocean (Gent and Cane, 1989; Gent, 1991) and in coupled mode with the global CCM2 atmospheric model (Gent and Tribbia, 1993). For the present, the tropical Pacific model will continue to be used for coupled experiments of ENSO behavior until the CSM model provides a satisfactory alternative. Similarly, low-resolution coupled experiments (e.g., Washington and Meehl, 1989) addressing IPCC issues are in progress, giving us additional experience with the oceanic role in addressing coupling questions.

Finally, there have been important developments in the implementation of alternative software architectures for running the GFDL model efficiently on other computer systems in a massively parallel fashion (Smith et al., 1992). The intention of the ocean team effort in CSM is to combine the best choices for ocean physics (i.e., parameterizations of mixing in the ocean interior; a better mixed layer parameterization; more suitable numerical algorithms for advection; etc.) with a computational structure for the ocean model component that best fits the NCAR computing environment. A new level of sophisticated physical parameterizations is in hand; now we must move toward implementing this in an optimal computing environment.

Strategy

We are currently in a period of very active development of the traditional GFDL model in the areas of parameterized physics (isopycnal mixing, surface boundary layer), numerical formulation (coordinate systems, discretization, transport schemes), and software architecture (modularity, parallel implementation). We can expect rapid evolution of the ``standard" CSM ocean component with a range of alternative configurations to be available in the next year.

While some options with potentially improved solutions will be available in the short term, we will begin by providing a suite of global modules based on GFDL MOM Code 1.1, with varying resolution and our current ``best choice" set of reasonably well-tested physical parameterizations. The ocean team will be responsible for assembling and regularly improving this global ocean GCM for use in a wide variety of climate problems. In particular, this model will be judged by its performance both in an uncoupled mode with specified surface boundary conditions, as well as when coupled to other component models of the climate system. While the scope of the ocean model includes physical transport processes, biogeochemical processes, and sea ice, these will come along at different rates and be included in different complexities as they are ready.

The general approach is to create some early versions of the model; establish benchmark solutions in an uncoupled mode with surface boundary conditions from either empirical data or atmospheric model output; use these solutions as initial conditions for prototype coupled calculations focusing initially on relatively short time scales (annual and interannual); and then establish a long-term cycle of investigating possible model improvements, computing new solutions in both uncoupled and coupled modes for different phenomenological regimes, and comparing these solutions to observations. The coupled solutions, in particular, will be designed and analyzed in collaboration with other participants in the CSM effort.

Physical Ocean Models

The physical circulation and transport model includes all processes, either in its resolved dynamics or through subgrid scale parameterizations, that are necessary for geographically realistic calculations of the global ocean circulation and its variability on all climatic time scales from the daily cycle to paleoclimatic evolution. Virtually all aspects of the ocean general circulation are potentially of interest, but the first focus will be upon the thermohaline convective circulations, the wind-driven equatorial and Antarctic zonal currents and extra-tropical gyres, the sea-surface temperature and surface boundary layer profiles, and the water-mass distributions of temperature and salinity. The model's design criteria include variable resolution, alternative domain specifications (so that regional or upper-ocean configurations can be used where apt), and a code architecture that supports many variants in its elements --- all within a common model framework so that systematic solution intercomparisons can be made.

In future implementations, other alternative versions of the GFDL code will be considered. This is important because the CSM in its whole must be tailored architecturally to the NCAR computing environment. The ocean code structure for the future will thus evolve, making use of the best choices that derive from efforts at GFDL, DOE, and NCAR, which have alternative structures for the basic GFDL code. Melding the numerics with the improved physical parameterizations will give us a powerful new tool for CSM and for ocean-only studies as well.

Issues to be addressed to improve ocean models in coupled and uncoupled mode can be divided into three categories: (a) improved parameterization of physical processes in the model formulation, (b) improved numerical algorithms and efficiency in the ocean model, and (c) improved data sets used to force the ocean model in uncoupled runs and to critically test its ability to reconstruct the temperature--salinity properties of the global ocean. The issues currently being explored in these categories and the scientists committed to working on them are:

Physical Issues

The effects of mesoscale eddies must be parameterized in ocean models that do not resolve them. A recent proposal based upon isopycnal form drag and tracer mixing (Gent and McWilliams, 1990; McWilliams and Gent, 1994; Gent et al., 1994) has been implemented in a global model and has shown (Danabasoglu et al., 1994) very promising improvements in the simulation of several quantities important in climate models such as northward heat transport. (Danabasoglu; Gent; McWilliams)

Interaction between the atmosphere and ocean occurs in boundary layers. Having a realistic representation of this complex of processes is important in an ocean model. A new parameterization (Large et al., 1994) for vertical mixing that uses a K-profile form in the planetary boundary layer (PBL) and a local diffusivity in the interior has been developed. (Large; Doney)

Numerical Issues

Different vertical coordinate systems have different effects on the results and need exploration for the climate problem. A stretched-sigma alternative to the standard vertical height coordinate has been formulated and is being tested, including the reformulation of a simplified equation of state that can be evaluated at arbitrary height, that is formally more accurate and provides a different numerical treatment of the topographic lower boundary condition. (Danabasoglu; Gent)

An upper-ocean model configuration with a stretched-sigma coordinate and a climatological lower boundary condition beneath the thermocline (taken from data or from a full-depth solution) will permit better vertical resolution, greater computational efficiency, shorter time to approach equilibrium, and less climate drift for calculations of short-term climate variability (up to decadal). (Large; Gent)

Modified grid structure and adequate filtering is needed in the Arctic basin to avoid issues concerning the polar singularity that occurs in a regular latitude-longitude grid. (Gent; Bryan)

Alternative algorithms for tracer advection (e.g., Smolarkiewicz, 1984) may be needed at some resolutions to prevent unphysical concentrations due to numerical errors. Various choices are being examined. (Hecht; Bryan; Holland)

Alternative formulations of the external-mode, reference-pressure, or free-surface solution method for computational efficiency and accuracy (e.g., Pinardi et al., 1994) are under development. These will be implemented as necessary. (Chervin; Holland)

Data Issues

Adequate surface boundary conditions for uncoupled experiments are needed. Data sets need to be improved for the specification of mean annual cycle of surface buoyancy forcing (restoring to climatological temperature and salinity or prescription of the actual fluxes of heat and fresh water) and wind stress forcing fields for uncoupled solutions. (Large; Holland)

Improved climatologies of temperature and salinity and other chemical properties of the interior of the ocean are needed as models improve. Critical tests of the success of uncoupled and coupled model experiments will depend upon having a high-quality observed climatic state of the ocean. (Owens; Holland; Chow)

These issues will be the primary foci during the next two years of the CSM project. On a longer time scale, other developments will be pursued, including a reconsideration of the underlying grid structure and numerical algorithms in the model. Also, different spatial resolutions to establish the cost/benefit curve for global solutions will be explored. The coarsest grid planned to work with is 20 levels. Refinements to 45 levels and 45 levels are being tested. Given computational costs to run the coupled system to equilibrium and given the need to explore alternative model specifications, reaching mesoscale-eddy resolution for climate studies in the immediate future except on an exploratory and regional basis is not anticipated. In this latter respect coupled models with very high resolution in the equatorial region or in the North Atlantic region will allow exploration of interactions in the coupled system in a regime where the oceanic and atmospheric components are each capable of generating substantial unforced variability.

For coupled calculations, development options will be selected, guided substantially by progress in testing them. For a particular selection, a full-depth, near-equilibrium solution for specified surface forcing will be obtained. That solution will then serve as initial conditions (and lower boundary conditions for an upper-ocean model configuration) for the coupled calculation. A variety of different forcing fields will be examined to minimize drifts in the coupled solution.

Sea Ice Models

The sea ice model has two primary components: local ice formation or melting due to thermodynamic forcing from above and below and horizontal advection of sea ice. The first connects crucially to the oceanic PBL model and with the flux coupler that handles atmospheric surface fluxes. The latter must include a dynamic model of the sea ice with a proper parameterization of horizontal, internal ice stress.

The formation or melting of sea ice due to vertical heat fluxes has been considered from the outset in the formulation of the K-profile planetary boundary layer model (Large et al., 1994). In addition to predicting the mass of ice at a grid point, the model must also include the fraction of ice cover (compactness), since the large difference in ice temperature at its upper surface and the ocean surface temperature (near the freezing point when ice is nearby) produces significantly different heat fluxes to the atmosphere, so that the two situations must be considered separately. The formulation will follow the standard conservation statements for ice mass and compactness (Hibler, 1979; Parkinson and Washington, 1979). In contrast to earlier models that included the interaction of sea ice with ocean mixed layers (Lemke et al., 1990), the formation of ice occurring within the upper grid points of the ocean boundary layer model and the melting of ice due to the difference between the atmospheric and oceanic heat fluxes will be considered.

A model for the advection of sea ice involves solving horizontal momentum equations for sea ice, which crucially depend on proper parameterization of internal ice stress. From previous modeling studies (e.g., Owens and Lemke, 1990), it appears that one may need to include a viscous-plastic rheology as formulated by Hibler (1979). However, for some purposes an alternative, simplified treatment of the sea ice rheology, a version of the formulation by Flato and Hibler (1990) that includes only shear stresses, may be useful. A version of this latter model has recently been recoded at NCAR and has been adapted to communicate with the flux coupler. In the near future, a form of the Hibler model with the complete rheology will also be implemented to work through the flux coupler and the user may use either version.

Biogeochemical Ocean Models

The long-term goal is to develop a global model of the major marine biogeochemical cycles (e.g., carbon, sulfur, nitrogen, oxygen), including a reasonable level of biological detail in the upper ocean and coupled fluxes of CO and other radiatively important gases with the atmosphere. As preliminary steps towards that goal, three intermediate problems have been identified for the next one to two years: the inorganic CO solubility pump, the seasonal cycle of productivity and phytoplankton biomass, and the ocean uptake and transport of inert and radioactive tracers.

The full-depth ocean model configuration will be used to examine the ocean transport of inorganic CO forced by solubility effects in the abiotic ocean, focusing on the large-scale patterns driven by the thermohaline circulation. The model simulations of the inorganic CO distribution and the geographic and temporal patterns of the air-sea CO flux will be compared with observations and previous model work (Maier-Reimer and Hasselmann, 1987). The sensitivity of the CO solutions to the new vertical and isopycnal mixing parameterizations will be explored. Early coupled calculations involving an abiotic, inorganic CO model and an interactive terrestrial carbon storage model are also envisioned (see Biogeochemical and Ecosystem Processes: NCAR CSM Land Component).

A nitrogen-based marine ecosystem model, adapted from the Fasham model (Sarmiento et al., 1993), will be incorporated into a global, upper-ocean configuration of the CSM ocean component, including the new parameterization of the surface boundary layer (Large et al., 1994). The bottom boundary conditions for the coupled ecosystem model will be specified from full-depth physical solutions together with available nutrient climatologies (Levitus et al., 1993). The behavior of the ecosystem model has been examined in detail in the vertical (Doney and Najjar, 1993), and alternative parameterizations of key biological processes (e.g., grazing, particle remineralization, micronutrient fertilization, and trophic and size class structure) will be explored in both the one-dimensional and global models. The coupled biological-physical model will be used to study a variety of questions on the seasonal to interannual time scale: the evolution of inorganic nutrients, phytoplankton, and zooplankton biomass; the rates of new and primary production; and the biological role of horizontal and vertical transport. Emphasis will be placed on understanding the seasonal cycle of spring blooms in the mid- and high latitudes and on the biological response to large-scale interannual variability such as ENSO. Validation of the marine ecosystem model will utilize data from individual time-series sites, historical nutrient and chlorophyll data sets, and the satellite ocean color data from the upcoming Sea-viewing Wide-field-of-view Sensor (SeaWiFS) project.

Both of these biogeochemistry problems will be carried out in conjunction with physical transport studies using transient and natural chemical tracers, such as radiocarbon, tritium, and chlorofluorocarbons that are essentially non-reactive or have well known radioactive decay rates (e.g., Toggweiler et al., 1989a,b). For this problem the surface atmospheric fluxes will be prescribed from observations.

In the second phase of the CSM project, the abiotic CO and upper-ocean ecosystem models will be combined into a full-depth, biotic carbon model. The transition from the nitrogen-based ecosystem model to a full representation of the carbon system is relatively straightforward, involving the addition of variable C/N ratios for biological pools, CO air-sea exchange, and the calcium carbonate cycle.

NSF Supported Staff for CSM

Non-NSF Supported Staff Contributing to CSM

Several scientists and programmers at NCAR are supported by non-NSF funds to work on specific projects that are closely related to CSM and contribute directly to its goals. For ocean modeling these include:

Collaborations

Dr. W.B. Owens, Woods Hole Oceanographic Institution (WHOI), is directly involved in the development of the sea ice model and in the development of an improved temperature/salinity climatology. Other collaborators are Dr. G. Flato of the Institute of Ocean Sciences (IOS, Canada) on sea ice modeling, Dr. David Glover of WHOI on the validation of the marine ecosystem model with satellite ocean color data, Dr. F.F. Jin of the University of Hawaii on empirical studies to determine the feedback boundary conditions, Dr. N. Pinardi of the University of Modena, Italy, on development of a surface-pressure solver and on empirical studies of the surface-flux climatology, Dr. W. Schmitz of WHOI on global equilibrium circulation, Dr. A.J. Semtner on the architecture of ocean model codes for moderately parallel computers, and Dr. R. Toggweiler of GFDL on equilibrium circulation and biogeochemical tracer distributions.

NCAR CSM: Land Component

The land is an integral part of the Earth system, influencing climate through a variety of processes. Key processes include surface energy exchange, the exchange of trace gases, and the generation and transport of runoff, which in turn affects ocean circulation and biogeochemical cycles via river input. The degree of coupling between terrestrial and climate processes varies, as does the time scale over which that coupling occurs. For example, the coupling between stomatal conductance, influencing water and CO exchange, and boundary layer climates is so close that in effect plant canopy biophysics must be considered integral for the atmosphere. At the other extreme, soil carbon, a significant component in the global carbon budget, responds to climate and influences atmospheric CO, but only on time scales of decades to millennia. Coupling on intermediate time scales occurs from the emission of biogenic compounds (e.g., methane, non-methane hydrocarbons) that have a profound influence on the ``chemical climate'' of the troposphere.

Land Model Components

The objective of including a terrestrial component in the CSM is to be able to examine coupling between the climate system and land processes on multiple time scales. To address such interactions, a land model needs a series of biophysical and biogeochemical components. The structure proposed separates these components into (a) biophysics and hydrology and (b) biogeochemistry and ecosystem processes. We propose to bring these components into the program in a phased fashion that reflects the readiness of the components within the land program and the overall plan of experiments within CSM.

Biophysics and Hydrology

The land surface influences atmospheric physical processes through radiative fluxes, turbulent transfer, and the partitioning of energy between latent and sensible heat. Key modeling issues in coupling the land surface to the atmosphere include:

Subgrid scale heterogeneity. Most land surface models have the same spatial resolution as the atmosphere, neglecting variation in vegetation and soil at smaller spatial scales. Three approaches have been adopted to account for this subgrid scale heterogeneity: (a) a finer land surface grid than the atmosphere, e.g., a 2 by 2 land grid and 4.5 by 7.5 atmospheric grid (Bonan et al., 1992; Pollard and Thompson, 1994); (b) explicit representation of several surface types for each grid with the necessary fractional areas to average surface fluxes to the grid (Avissar, 1992; Koster and Suarez, 1992); and (c) closed-form analytical expressions that integrate over distributions of important heterogeneity (Entekhabi and Eagleson, 1989; Johnson et al., 1993).

Runoff. The current generation of land surface models have focussed on the coupling between the land and the atmosphere. Future models must examine the coupling between the land and oceans. Ocean circulations are affected by the magnitude of runoff through its impact on temperature and salinity distributions. River runoff delivers a significant amount of nutrients and dissolved organic carbon to the coastal ocean, and thus affects oceanic biogeochemical cycling. The riverine input of dissolved organic carbon in particular may be a significant component of the global carbon cycle (Siegenthaler and Sarmiento, 1993). Improved runoff prediction will require better schemes for runoff generation, including subgrid scale heterogeneity in vegetation and soil types. It will also require a runoff routing model that considers lateral flow of runoff between adjacent grid cells, which current land surface models ignore. This affects not only the timing and magnitude of runoff to oceans but also soil water and latent heat exchange. Once the runoff reaches a channel some form of river routing must be used.

Lakes and other wetlands. Lakes are a substantial area of the terrestrial surface but are neglected in current land surface models. However, in regions with large areas of standing water (e.g., high latitudes in the Northern Hemisphere), this can have a significant effect on surface fluxes. The ability to include lakes and other wetlands in the land surface model depends on the parameterization of subgrid scale heterogeneity because even on a 0.5 by 0.5 grid most lakes are too small to be resolved. In addition, lakes and wetlands are important agents of biogeochemical cycling, serving as sources, sinks, and transformers of biological materials. Wetlands are significant sources of groundwater recharge and act as natural buffers against floods. Consequently, the representation of lakes and wetlands must be closely tied with the subgrid scale parameterization, hydrology, and biogeochemistry.

Model complexity. The degree of complexity with which surface energy exchange should be represented is poorly understood and is the subject of a land surface model intercomparison project (Pitman et al., 1991). This issue is important because of model sensitivity to land surface characteristics that are poorly known, e.g., soil water holding capacity, plant physiology. The complexity of a land surface model also affects the parameterization of subgrid scale heterogeneity and the links to biogeochemistry and ecosystem processes. In addition, there are considerable discrepancies in how micrometeorologists parameterize surface exchange models for comparison with tower measurements and the surface exchange models used in global climate models. The micrometeorological models are much more detailed, but this complexity is prohibitive in terms of computer time. The large discrepancy between models needs to be further investigated.

Plant physiology. Recent land surface model development has recognized the close linkage of biophysical and plant physiological processes and the benefits of coupling these into a comprehensive land surface model (Sellers et al., 1992). For example, latent heat exchange and CO exchange are linked through stomatal resistance. This coupling of stomatal resistance and photosynthesis provides for additional land-atmosphere feedbacks, e.g., diurnal variation in atmospheric CO concentration in the boundary layer and long-term changes in atmospheric CO concentration. Adding vegetation CO fluxes also provides new means to test land surface models.

Plant growth and community structure. The fluxes of energy, water vapor, and momentum from the atmosphere to the land depend in a critical way on the structure of the plant community. Properties such as albedo and surface roughness that are required by biophysical models are closely related to the state of the biota (e.g., plant biomass and height, leaf area index, root-stem allocation), which changes on annual time scales due to the seasonal growth and die back at temperate and high latitudes and on longer time scales from shifts in vegetation composition (e.g., broadleaf deciduous tree to grassland). These interactions can be studied either in isolation or in a coupled mode whereby the feedbacks between surface fluxes and plant behavior are allowed to occur (see Biogeochemical and Ecosystem Processes).

The initial implementation of the physical component of the land surface model will use an existing global land surface model while leaving the above issues as research projects. Emphasis will be on the physical land-atmosphere coupling at the seasonal time scale. Improved parameterization of subgrid scale heterogeneity, hydrology (lakes, wetlands, runoff generation), and links to biogeochemistry and ecosystem processes will be brought on-line in due time as appropriate. Subgrid scale heterogeneity, lakes, and wetlands can be added quickly (i.e., within the first year). The linkage of photosynthesis with energy exchange also has sufficient development to be brought on-line within the first year (Bonan, 1994). Coupling to ecosystem processes, such as growth, nutrient cycling, and vegetation dynamics, is possible but will require more time (see below). These additional components must be brought on-line in a systems approach that addresses the coupling, linkages, and common processes and state variables rather than treating biophysics, hydrology, biogeochemistry, and ecosystem processes as independent components.

Biogeochemical and Ecosystem Processes

Ecosystem and biogeochemical processes influence land-atmosphere interactions, generally on longer time scales than effects of surface biophysical processes. Key processes include storage of carbon in living and non-living components of the terrestrial biosphere, influence of the nitrogen cycle on carbon fixation (photosynthesis) and storage, and long-term effects of vegetation structural change (e.g., forestgrasslands, coniferdeciduous) on biogeochemistry and on biophysical exchanges through changes in physiology, rooting depth, height, and albedo.

For early (though not initial) experiments in CSM, a model of terrestrial biospheric storage and exchange of carbon is the primary target. Such a model will simulate carbon fixation as a function of climatic variables (providing a forcing from climate), atmospheric carbon dioxide concentration, and the nitrogen cycle (providing internal, nonlinear dynamics). The model will simulate decomposition as a function of soil physical properties, climate, the N cycle, and plant organic chemistry. The model will simulate the gross effects of land management on carbon storage, as this is a crucial driver for ``realistic'' simulations of the carbon cycle.

This model will be based on the Century model developed by Parton, Schimel, and coworkers at Colorado State University, but with significant simplification of carbon-nitrogen interactions (Parton et al., 1994; Schimel et al., 1994). The Century component plays two important roles in the coupled model. It provides nitrogen to the land surface parameterization, which constrains canopy development and conductance. Second, it computes terrestrial carbon storage as a function of climate forcing and CO, modified by internal C/N dynamics, and thus, modifies atmospheric CO concentrations over time.

In coupling to the physical land surface model, carbon fixation will be derived from calculations of photosynthesis carried out in the land surface parameterization to provide consistency between the carbon and water cycle calculations. Soil temperature and water, needed to simulate microbial activity, will also be derived from the land surface parameterization. However, the biogeochemical model can simulate carbon fixation and soil water itself at longer temporal resolutions, so that the model can also be run uncoupled from the short-term (i.e., less than 30 minute time step) biophysical model for long calculations (e.g., 100s of years).

Three-dimensional calculations of CO effects on climate generally prescribe a rate of CO increase based on extrapolation of historical trends. In fact, non-linear processes control the relationship between CO forcing and atmospheric CO concentrations: a) the non-linear chemistry of CO oceanic uptake; b) the highly non-linear effects of CO in biological uptake and storage of C in the terrestrial biosphere; c) the potential effects of altered ocean surface and thermohaline circulations on CO; and d) the effects of physical climate changes (temperature, precipitation, solar radiation) on the biosphere. Simple models suggest that growth rate and lifetime of CO are sensitive to the inclusion of a function that increases terrestrial uptake as atmospheric CO increases. These effects presumably affect the rate of change of climate for any rate of change of anthropogenic CO, possibly with significant interactions. In more realistic ecosystem models, the ``'' function relates to the N cycle and to soil moisture, so direct CO effects need to be simulated with a full ecosystem model. A proposed first experiment with a coupled climate system model including the terrestrial biogeochemical component would be to simulate the evolution of climate, atmospheric CO, and terrestrial C storage during a 100-year transient calculation with CO forcing according to a prescribed scenario. This calculation would require at a minimum a geochemical CO uptake mechanism in the ocean model but could be done with an upper ocean model. Inclusion of a full-depth ocean model would add additional possibilities to behavior of the physical climate system (e.g., Manabe and Stouffer, 1993) but would not be required for the initial carbon cycle calculation. The terrestrial model required for this calculation could be ready in approximately one year with additional programming support.

Two other areas of ecosystem interactions are of interest: exchange of trace gases other than carbon dioxide and effects of changing vegetation types on biogeochemistry (CO and other species) and biophysics. For trace gas exchanges, no comprehensive model exists but a number of components exist, including models for NO, non-methane hydrocarbons (Guenther et al., 1994), and methane (Holland et al., 1993). Modeling of trace gases can be thought of in two categories. First are species such as NO (and CO), greenhouse gases, but unreactive with weak influence in tropospheric reactions. These gases may be modeled like CO with no atmospheric reactions simulated. Inclusion of such gases allows the exploration of their effects on climate. In a second category are reactive chemical species, such as the NMHCs, methane, and NO. The emissions and uptake of these species is sensitive to climate and land surface conditions, may influence climate through the lifetimes of greenhouse species, and have profound effects on the OH-O chemistry of the troposphere. Coupling of simulations of the exchange fluxes of reactive species can only be of interest when the terrestrial model is coupled to an atmospheric model containing chemistry. Since the simulations of chemistry require the fluxes of a fairly large number of species, the ``jump'' from a CO model to a CO-trace gas model is large. Intermediate calculations may be possible, such as calculations containing a methane source model and a simplified atmospheric destruction model. The timetable on coupled atmospheric-terrestrial biogeochemistry models is somewhat indeterminate but will require two to three years.

Tropical deforestation and ``boreal" deforestation experiments with coupled land- atmosphere models raise the possibility of coupled climate-vegetation dynamics in which the biogeographic distribution of vegetation affects and is affected by climate (Bonan et al., 1992; Henderson-Sellers, 1993). In addition, recent steady state calculations of terrestrial C storage under doubled CO climates with a biogeochemical model (TEM: Melillo et al., 1993) showed that changing vegetation distributions significantly affect C storage beyond the direct effects of climate on biogeochemical processes. A number of equilibrium global vegetation models exist, including a model developed as part of the GENESIS project at NCAR, a model developed by the Henderson-Sellers group, the MAPSS model of Neilson, the DOLY model of Woodward. Considerable effort is underway to make these equilibrium models time-dependent, e.g., linking N and water availability to changes in leaf area, and changes in leaf area to changes in growth form. A dynamic vegetation distribution model, which links with the short-term biophysics and biogeochemistry, should eventually be a component of the land team activity. CSM scientists are currently involved in an EPRE-NASA-USFS-CSMP supported intercomparison of vegetation models (Vegetation/Ecosystem Modeling and Analysis Project VEMAP) which should provide insight into the behavior of these models. Because the time table for developing credible time-dependent vegetation type models is unclear, no timetable for this activity can be given.

Validation

Validation of land surface models will utilize global and in situ data sets. A number of important data sets exist. The new Advanced Very High Resolution Radiometer (AVHRR) Pathfinder activity produces a satellite product for the land fundamentally superior to the GAC and GVI products previously used, and improved algorithms for the extraction of leaf area, interception of photosynthetically active radiation, and net primary productivity now exist. Records of seasonal and interannual variability of CO, C of CO, O, and oxygen isotopes provide fundamental constraints on simulations of terrestrial carbon and water cycling. A significant number of globally-distributed, but point or regional measures, also exist. These include results of field campaigns in France, West Africa, the central U.S., Northern Canada, Scandinavia, and Brazil, in which comprehensive measurements exist. Also available are several new compendia of plant biomass and soil carbon and nitrogen dynamics from the U.S. Long Term Ecological Research network, assembled globally by a SCOPE project and recently published. Finally, data on river runoff in major river basins and dissolved and particulate carbon loading were recently published and made available for testing runoff components of surface hydrology. Critical gaps in land surface data sets are soil moisture, important in biophysical and biogeochemical calculations, for which no satisfactory remote techniques exist, and snow cover, for which satellite products have low spatial resolution and poor estimation of depth and water content.

NSF Supported Staff for CSM

Collaborations

J. Famiglietti (University of Texas, Austin) is working on issues of land surface hydrology/runoff and W. Parton (Colorado State University) is involved in the Century Model.

NCAR CSM: Flux Coupler

Description

The flux coupler (FC) provides the framework of the CSM by providing overall computing control and interfacial physical coupling among the component system models. Control involves software engineering issues, such as model synchronization and data communication between components. The coupling function addresses issues such as the computation of physical quantities exchanged among CSM components, as well as numerical questions such as those associated with interpolation between disparate grids of the component models.

The control and coupling functionality are designed to allow each component to run with its own time stepping scheme and to have its own grid structure. The FC will make as few demands upon component models as possible, and conversely, the models will make minimal demands upon the coupler. The FC and model components will exchange minimal amounts of information with each bit of information preferably transmitted only once. Thus, each component model must maintain some control structure. If, for example, a model requires lagged time information, that component must store data received from the FC and use it appropriately. The FC, for its part, will only require that each component model accept fluxes and flux-related quantities, such as albedo, as boundary conditions.

The FC diverges from the most common practice of atmosphere-ocean model coupling in two important ways. First, coupled models have traditionally been implemented as single monolithic programs by ``subroutinizing" the component models. In the FC framework, each component model remains as a stand-alone program. Second, interaction between the atmosphere and the surface has generally been handled completely within the atmospheric component model. This severely limits the flexibility of the system and makes it difficult to accommodate component models with disparate spatial resolutions. In the FC quantities computed on the surfaces between models are treated independently of the models themselves. This set of choices is based on recent experience at NCAR in modeling ocean-atmosphere interaction and draws on recent developments at other institutions in the areas of ocean-atmosphere and atmosphere-land surface coupling. Mechoso et al. (1993) and Barth and Smith (1993), for example, have demonstrated the viability of the message-passing approach. The approach adopted for computing interfacial fluxes is very similar to that used to represent sub-grid scale heterogeneity in land-surface process modeling (e.g., Koster and Suarez, 1992; Avissar and Pielke, 1989). These issues are discussed further below. Two aspects of coupling not specifically being addressed by the CSM FC project at the present time are (a) asynchronous integration strategies (Dickinson, 1981) and (b) flux adjustment procedures (Sausen et al., 1988).

Software Architecture and Computing Environments

Given current computing trends, the FC is designed with the capability to operate in a distributed, heterogeneous processor environment, as well as with a more traditional single-system, shared-memory environment. We have constructed a system in which the component models, as well as the FC itself, operate as a set of separate, cooperating executables. Communication between the components is facilitated through an explicit message-passing interface. Each model sends just those data required by the FC, which then computes updated fluxes and returns them to the components, allowing them to advance.

A message-passing FC allows orderly program development. Models should require far less modification to operate within the message-passing system than if it were demanded that each model be ``subroutinized'' for coupling. This greatly increases the flexibility of the CSM system, permitting different component models to be easily substituted. Existing components should still run in their usual stand-alone mode with only the invocation of the message exchanges, and some coordination functions, required for use in the coupled mode.

In developing the FC, attention will be given to software-engineering principles and adoption of programming standards similar to those already in use among atmospheric modelers (Kalnay et al., 1989). These standards require ``plug compatibility'' for internal FC subroutines; this modular approach should facilitate modifications, such as alternative flux formulations, interpolation routines, or extensions to accommodate additional processes, e.g., chemistry.

Coupling Principles

CSM model components will couple primarily through fluxes across component boundaries, which requires formulating each component with flux boundary conditions. For example, the atmosphere component need not know sea-surface temperature (SST) but only fluxes that the FC computes using SST from the ocean component. Conservation of fluxes becomes the most important principle. Even though fluxes may be modified in some manner by the receiving model due to such functions as smoothing or filtering, each CSM component must see the same globally-averaged flux as another at interfaces shared by those models.

Spatial Aspects of Interaction. Two component models interacting through one of the interfaces will generally not have the same horizontal resolution or identical grid layout. This necessitates spatial averaging or interpolation of quantities to be communicated between the models. In this system a distinction is made between two types of fluxes passing across an interface. The first type is a flux computed by a component model. These generally involve computationally intensive calculations over the three-dimensional domain of the component model. Examples include the precipitation rate or downward radiative fluxes computed by the atmosphere or the potential freezing rate computed by the ocean. The second type is a flux computed at the interface itself based on the (two-dimensional) surface states of the component models. Examples include turbulent heat fluxes across the air-sea interface or the stress between the ocean and sea ice. The general philosophy followed here is that to the extent possible, the second type of flux should be computed on the finer of the two grids defining an interface. The outputs (both state variables and output fluxes) of the model at the coarser resolution are interpolated to the finer resolution grid. Fluxes requiring state information from both systems are computed on the finer grid by the coupler. These fluxes are then spatially averaged back to the coarser model grid in a conservative manner. In the case where a component model (say atmospheric) grid cell contains more than one surface type (say land and ocean), the fluxes from each interface are averaged together in an area weighted manner into an aggregate flux. This particular scheme was selected to allow for subgrid-scale heterogeneity of surface types, to be conservative, and to take maximum advantage of any high-resolution information available in the system.

Temporal Aspects of Interaction. Just as the component models may operate with different spatial discretizations, they will generally have different time steps or require updated fluxes at different time intervals. We follow a similar procedure as used for handling disparities in spatial resolution in handling temporal interactions. Fluxes on an interface are computed at the frequency required by the model with the faster time scales and temporally averaged over an interval called the synchronization period before being passed to the model with the slower time scales. The state variables on the interface from the ``slow" model are held fixed during this interval, while the output fluxes and state variables of the ``fast" model are updated an integer number of times. At the end of the synchronization period, all component models will have seen the same flux passing through the interface.

Work Plan

The FC will be developed in two phases. The first, already well along, involves the implementation of a basic system for atmosphere--ocean--sea ice interactions, with the land model kept initially within the present (CCM2) atmospheric component. Modifications to the component models to allow communication of two-dimensional fields of state variables and fluxes are underway. In addition, tuning/optimization of the communication structures are being examined. Our goal is to begin coupled experiments using the flux coupler immediately so as to both test the FC aspects of the coupling and to begin to carry out useful coupled experiments from the physical point of view.

The second phase of FC development will involve extensions to include a separate land model, to include chemical processes that require FC passing of information between components, and to include the thermosphere as a separate component. In the long term, the FC will provide an effective framework to switch in and out various sub-components of the complete climate system model according to the desires of the investigator.

NSF Supported Staff for CSM

Non-NSF Supported Staff Contributing to CSM

Several scientists and programmers at NCAR are supported by non-NSF funds to work on specific projects that are closely related to CSM and contribute directly to its goals. For the flux coupler these include:

VI. Resource Requirements for CSM Personnel Requirements Summary

NCAR scientists have a very strong interest and commitment to the CSM. There is however a need for a long term commitment of dedicated resources to support the activities of the CSM project. This involves establishing a core group of programmers for:

• developing, maintaining, documenting and
• disseminating the CSM code;
• managing the computations and data sets created and making them available to the outside user community;
• aiding in the development of new components or revisions of existing components;
• providing general support to CSM activities, including developing and maintaining analysis tools.

Based upon past experience with the CCM core group, we anticipate needing four programmers for these purposes and propose bringing onboard such people over the course of the start-up period, i.e. the first two years. At steady state, about 360K per year would be needed for the duration of the CSM project. These programmers may eventually be combined with the existing CCM core group to support all aspects of the CSM.

In order to promote active participation by university faculty and training of young scientists we propose that a minimum of two long term visitor and two postdoctoral positions should be funded (300K per year). These positions could be coordinated through UCAR-CSMP in order to broaden the the outreach to the scientific community and to reduce overhead.

The CSM project also requires resources for short-term visitors collaborating with NCAR scientists on CSM development and analyses and for those wanting hands-on experience with the model (50K per year). Regular workshops on the model and its use are needed to stimulate the involvement of a broader spectrum of users and to increase their understanding of the capabilities and limitations of the CSM (24K, every other year). Finally, the CSM core group and visitors will need adequate workstations to work effectively (25K per year).

A comprehensive CSM will be computationally expensive. The development of any complex model, such as the CSM, inevitably involves compromises either to control computational costs, or because of insufficient scientific understanding. Below we show an estimate of the computational cost of the components in the first implementation of a modest-resolution CSM version 1. We also discuss, component by component, the rationale for these choices as well as give an indication of the relative costs of other (future) choices. It should be remarked that it is relatively straight-forward to enhance the resolution of any of the components and, for some climate issues, this will be required.

Several strategies are being followed to limit the expense of the CSM, such as intermediate horizontal resolution and limited complexity. The target resolutions for the atmospheric components have been chosen near the minimum resolutions at which credible representations of the most important large scale dynamical processes are possible. The only chemical or biogeochemical option which will probably be incorporated in coupled simulations during the startup phase is an extremely simplified treatment of the atmospheric concentrations of the principal greenhouse gases and sulphate aerosol. More detailed chemistry and biogeochemistry will be tested primarily in uncoupled simulations until both significant skill and a need for coupling is demonstrated. Nevertheless, the CSM will tax the capabilities of currently available supercomputers. As increasing computer power is available, pioneering calculations with higher resolution and increased complexity will be performed. However, assuming that the important dynamical and thermodynamical processes are adequately represented, more will be learned from frequent simulations at modest resolution than from one or two simulations at the highest possible resolution. It is extremely important that a large proportion of the available computing time be used for multiple simulations in order to understand the model behavior and its sensitivity to a variety of parameter choices.

The computational cost of the physical components of the climate system are reasonably well known because relatively mature models exist. The initial configuration of the CSM has been designed so that the physical climate system from the ocean floor to the middle stratosphere can be simulated on the existing computers in the CSL. However, transient greenhouse gas experiments will stretch the available capacity to the limit. For example, a 150 year integration from preindustrial greenhouse gas concentrations would take about 2 months on a dedicated YMP8.

COMPUTATIONAL REQUIREMENTS
YMP Processor hours / simulated year

The computational costs of the chemical and biogeochemical components of the climate system are not well known because only prototype models currently exist. However, it is certain that any reasonably comprehensive treatment of these processes will be expensive. The cost of treating chemical compounds in three-dimensional models can be easily seen from the increase in the required number of variables. An atmospheric or oceanic GCM needs three dimensional equations to predict only 4 variables: zonal and meridional momentum, temperature, and water vapor or salt concentration. In addition, the vertical velocity must be diagnosed. Each chemical compound whose concentration must be advected adds an additional three dimensional variable and could be expected to add about 20 to the computational cost. The actual increase is generally less than 20 because the parameterized physics in the GCM significantly increase its basic cost and the numerics can often eliminate redundant calculations for additional variables. However, it is clear that simulating the concentrations of the principal greenhouse gases will be time consuming, even with highly simplified chemistry and biogeochemistry and limited numbers of constituents.

How Increased Computer Resources Can Be Used

1. Atmosphere

Dynamics

The advent of semi-Lagrange dynamics allows an increase of the spectral truncation to T63 with no increase in cost. Developing of a reduced grid (at the poles) model will allow a further modest increase in resolution (or decrease in cost). At least a factor of five increase in computing power would be required to make a significant increase in the horizontal resolution.

The vertical domain can be increased to include the full stratosphere and lower mesosphere at about 2.5 times the cost of the standard model. This increase will be required for a credible treatment of ozone chemistry and its effects on climate.

Physics

The radiative transfer parameterization is fairly detailed for both solar and long wave radiation. The long wave effects of the principal greenhouse gases are being included and the short wave effects of aerosols can be included easily in the short wave without significantly altering the cost.

The boundary layer employs one of the best available parameterizations for models with modest vertical resolution. A modest increase in vertical resolution ( 10) would improve the ability to represent the boundary layer depth and turbulent fluxes, but little more computational cost is required unless significantly better parameterizations are developed.

The treatment of convection in CCM2 is particularly simple and more elaborate (and expensive) schemes are being tested. A significant increase in the model cost (10--20) will probably be required to incorporate a better convection, if the benefits can be demonstrated.

Chemistry

Atmospheric chemistry can use enormous amounts of computing power. Each chemical species added to the model must be advected by the dynamics and mixed vertically by the boundary layer and convection. Computing chemical and photolytic reaction rates can also be very expensive, especially when expensive algorithms are required to deal with the numerical problems posed by reaction time scales which vary from nanoseconds to hours or longer.

A reasonably comprehensive treatment of stratospheric ozone chemistry approximately doubles the cost of the model even though the chemistry is still somewhat simplified and highly optimized. Half of the cost increase results from including the photochemical reactions and half results from the cost of transporting 14 chemical constituents in the GCM.

Detailed tropospheric chemistry will be much more expensive both in terms of the number of species which must be advected, and in terms of the computational cost of chemistry itself.

The initial CSM will include only a highly simplified chemistry for the principal greenhouse gas and sulphate aerosol. Even this limited set will probably represent 50 of the cost of the atmospheric model.

More detailed chemical models will be pursued outside the coupled model framework, but may require greatly expanded ( 5--10 times) computer resources for significant advances.

2. Ocean

Physics

The mesoscale eddy parameterization included in the ocean model results in considerable net savings in computation time. The increase in the accuracy of the simulation of many large scale ocean properties, including the poleward heat fluxes, allows the model to be run at coarser horizontal resolution than would be possible in a model employing only a standard horizontal diffusion in height coordinates. The reduction in horizontal resolution easily offsets the 40 increase in computation time resulting from the calculation of the density slopes required by the eddy parameterization.

The next major change in the model physics will be the inclusion of an ocean boundary layer, similar in nature to the atmospheric boundary layer included in the atmosphere component. The boundary layer parameterization will increase the cost of the ocean model by a further 25, based on initial testing, but should greatly improve the accuracy of the simulation of vertical mixing in the upper layers of the ocean.

Numerics

An acceleration technique has been used to run the ocean model to equilibrium with fixed atmospheric forcing. The ocean adjustment time increases as a function of depth to about 1000--3000 years for the deep ocean. The acceleration technique allows deep ocean temperature and salinity timesteps 100 times greater than the time step for the upper layer ocean velocities. This technique has been shown to work well in obtaining equilibrium solutions under constant forcing conditions but cannot be used in coupled model runs because of the variations in atmospheric forcing. Indeed, several coupled modeling groups have abandoned acceleration techniques entirely because of concerns that it leads to the wrong equilibrium solution when even seasonal variations in forcing are included. This means that ocean component of the coupled simulations will take much longer (10--100 times) than ocean spinup runs of comparable length.

Most of the early CSM simulations will be run at a horizontal resolution in the ocean model of 2longitude and averaging 2latitude (higher near the equator and at high latitudes, lower in midlatitudes). Some simulations will be performed at half of this grid spacing ( 1), but clearly neither configuration resolves ocean eddies, which is the reason that the mesoscale eddy parameterization has been included. If an eddy resolving global ocean model is considered, then the computation cost would increase enormously. Although considerable success at simulating ocean eddies has been achieved at 1/6resolution, truly eddy resolving models will require 1/10resolution and cost at least 400 times more than the 2model being used in the initial CSM simulations.

The ocean model is a considerable compromise in horizontal resolution. The ability to represent some of the important processes involved in ENSO has been sacrificed to achieve an affordable cost. Since the model has a minimum latitudinal spacing of 1.2near the equator, it cannot adequately resolve the equatorial wave guide and is not expected to be able to represent some of the essential dynamics of tropical Pacific variability. Experience suggests that the 1 model, with it's 0.6equatorial resolution, will be required to properly represent ENSO.

The 45 level vertical resolution of the ocean model is considerably greater than that being used by most other coupled modeling groups. This resolution has been chosen to represent the crucial effects of variations in the ocean depth on the circulation and, particularly, on water mass formation. Many aspects of the ocean circulation are determined by the restrictions on the flow produced by shallow sills and the isolation of basins by intervening ridges.

Other issues, such as alternative advection algorithms, filtering schemes and vertical coordinates would not greatly affect the computational cost of the ocean model.

Biogeochemistry

The simplest ocean biogeochemistry calculations using an abiotic, inorganic CO2 model increase the cost of the ocean model by 40.

Early tests using a marine ecosystem model adapted from the ``Fasham'' model have shown that it increases the ocean model cost by a factor of 3.

Inclusion of a full-depth biotic carbon and nitrogen-based ecosystem model would probably increase the computational cost of the ocean model by a factor of 5.

3. Land Surface

Physics

The land surface grid is currently the same as the horizontal resolution of the atmosphere (approximately 2.8 x 2.8). Moving the land surface to its own grid with finer resolution would allow for better characterization of vegetation and soil inhomogeneities that are thought to significantly affect land-atmosphere interactions. However, this gain in spatial resolution is expense. A 2 x 2 grid increases the overhead of the land surface model by a factor of two; a 1 x 1 grid increases the overhead by a factor of eight.

Most land surface models simulate the average soil water within a grid box. Independent of the horizontal resolution of the land, the distribution of soil water within a grid box and the non-linearities associated with surface fluxes and soil hydrology need to be considered. Simplying adding two subgrid fractions (wet, dry) with explicit surface flux calculations and hydrology for each fraction would increase the overhead of the land surface model by a factor of two.

Empirical and theoretical studies suggest the terrestrial fresh water flux into oceans is an important determinant of past climatic changes. Models that transport runoff from individual land surface points to appropriate oceans are being developed and must be included in future CSM activities. The cost of these models is not known, but can be high if one fully represents the two-dimensional lateral transport of water between grid boxes.

Biogeochemistry

The land surface model simulates CO fluxes associated with plant photosynthesis, plant respiration, and microbial respiration, and the feedbacks from the nitrogen cyclke that govern transient responses. This model will also serve as a platfrom for the eventual simulation of biogenic CH, NO, NO, and hydrocarbon fluxes. One of the future tasks for the CSM is to use these fluxes to simulate atmospheric CO concentrations. The cost of adding an inert chemical constituent such as CO to the atmosphere is approximately 1.2 CPU hours per simulated year.

The cost of running a complete terrestrial ecosystem model at T42 resolution is approximately 1 CPU hour per simulated year.

4. Sea Ice

The initial sea ice component of CSM will use a cavitating fluid rheology, partly to reduce the computation time. Developments such as full rheology physics and more advanced thermodynamics are likely in the near future. These developments may increase the computational cost of the sea ice model several times, but it will remain small compared to other components of the CSM.

VII. References

97, 2729--2742.

Avissar, R., and R.A. Pielke, 1989: A parameterization of heterogeneous land surfaces for atmospheric numerical models and its impact on regional meteorology. Mon. Wea. Rev. 117, 2113--2136.

Barth, N., and S. Smith, 1993: Coupling numerical models of the atmosphere and ocean using the Parallel Virtual Machine (PVM) Package. In: Proceedings of the Sixth SIAM Conference on Parallel Processing for Scientific Computation. R. Sincovec, D. Keyes, M. Leuze, L. Petzold, and D. Reed (Eds.), SIAM, pp. 71--75.

Bonan, G.B., 1994: Comparison of the land surface climatology of the NCAR CCM2 at R15 and T42 resolutions with implications for sub--grid land surface heterogeneity. J. Geophys. Res., submitted.

Bonan, G.B, D. Pollard, and S.L. Thompson, 1992: Effects of boreal forest vegetation on global climate. Nature 359, 716--718.

Boning, C.W., F.O. Bryan, and W.R. Holland, 1994: Thermohaline circulation and poleward heat transport in a high resolution model of the North Atlantic. Submitted.

Boville, B.A., 1994: Middle atmosphere version of CCM2: Annual cycle and interannual variability. J. Geophys. Res., submitted.

Boville, B.A., 1991: Sensitivity of simulated climate to model resolution. J. Climate 4, 469--485.

Brasseur, G.P., M.H. Hitchman, S. Walters, M. Dymek, E. Falise, and M. Pirre, 1990: An interactive chemical dynamical radiative two-dimensional model of the middle atmosphere. J. Geophys. Res. 95, 5639--5655.

Broecker, W.S., 1987: The biggest chill. Natural History Magazine, October, 74--82.

Bryan, F.O., and W.R. Holland, 1989: A high resolution simulation of the wind- and thermohaline--driven circulation in the North Atlantic Ocean. In: Parameterization of Small Scale Processes. aha Huilko'a Proceedings Winter Hawaii Workshop, P. Muller and D. Henderson (Eds.), Spec. Publ. 99--116, Hawaii Inst. Geophys.

Bryan, K., S. Manabe, and C. Pacanowski, 1975: A global ocean-atmosphere climate model. Part II. The oceanic circulation. J. Phys. Oceanogr. 5, 30--46.

Danabasoglu, G., J.C. McWilliams, and P.R. Gent, 1994: The role of mesoscale tracer transports in the global ocean circulation. Science, in press.

Dansgaard, W., S.J. Johnsen, H.B. Clausen, D. Dahl-Jensen, N.S. Gundestrup, C.U. Hammer, C.S. Hvidberg, J.P. Steffensen, A.E. Sveinbjornsdottir, J. Jouzel, and G. Bond, 1993: Evidence for general instability of past climate from a 250-kyr ice-core record. Nature 364, 218--220.

Dickinson, R.E., 1981: Convergence rate and stability of ocean-atmosphere coupling schemes with a zero-dimensional climate model. J. Atmos. Sci. 38, 2112--2120.

Doney, S.C., and R.G. Najjar, 1993: The effect of vertical transport in a one--dimensional coupled biological--physical model of the upper ocean. Preprint.

Entekhabi, D., and P.S. Eagleson, 1989: Land surface hydrology parameterization for atmospheric general circulation models including subgrid scale spatial variability. J. Climate 2, 816--831.

Erickson, D.J., III, P.J. Rasch, P.P. Tans, and P. Friedlingstein, 1994: The NCAR CCM2-based global atmospheric CO cycle model: Initial validation experiments. J. Geophys. Res., submitted.

Flato, G.M., and W.D. Hibler, III, 1990: On a simple sea ice dynamics model for climate studies. Adv. Geophys. 14, 72--77.

Garcia, R.R., and B.A. Boville, 1994: Downward control of the mean meridional circulation and temperature distribution of the polar winter stratosphere. J. Atmos. Sci., in press.

Gent, P.R., 1991: The heat budget of the TOGA--COARE domain in an ocean model. J. Geophys. Res. 96, 3323--3330.

Gent, P.R., and M.A. Cane, 1989: A reduced gravity, primitive equation model of the upper equatorial ocean. J. Comp. Phys. 81, 444--480.

Gent, P.R., and J.C. McWilliams, 1990: Isopycnal mixing in ocean circulation models. J. Phys. Ocean. 20, 150--155.

Gent, P.R., and J.J. Tribbia, 1993: Simulation and predictability in a coupled TOGA model. J. Climate 6, 1843--1858.

Gent, P.R., J. Willebrand, T.J. McDougall, and J.C. McWilliams, 1994: Parameterizing eddy--induced tracer transports in ocean circulation models. J. Phys. Ocean., submitted.

Giorgi, F., 1990: Simulation of regional climate using a limited area model nested in a general circulation model. J. Climate 3, 941--963.

Giorgi, F., G.T. Bates, and S.J. Nieman, 1993: The multiyear surface climatology of a regional atmospheric model over the western United States. J. Climate 6, 75--95.

Giorgi, F., C. Shields Brodeur, and G.T. Bates, 1994: Regional climate change scenarios over the United States produced with a nested regional climate model: Spatial and seasonal characteristics. J. Climate 7, 375--399.

Guenther, A., C. Hewitt, D. Erickson, R. Fall, C. Geron, T. Graedel, P. Harley, L. Klinger, M. Lerdau, W. McKay, T. Pierce, R. Scholes, R. Steinbrecher, R. Tallamraju, J. Taylor, and P. Zimmerman, 1994: A global model of natural volatile organic compound emissions. J. Geophys. Res., submitted.

Hack, J.J., B.A. Boville, J.T. Kiehl, P.J. Rasch, and D.L. Williamson, 1994: Climate statistics from the NCAR Community Climate Model (CCM2). J. Geophys. Res., in press.

Hack, J.J., 1994: Sensitivity of the simulated climate to liquid water path length specification. J. Geophys. Res., in preparation.

Hack, J.J., B.A. Boville, B.P. Briegleb, J.T. Kiehl, P.J. Rasch, and D.L. Williamson, 1993: Description of the NCAR Community Climate Model (CCM2). NCAR Technical Note/TN--382+STR, Boulder, Colorado, 108 pp.

Hauglustaine, D.A., C. Granier, G.P. Brasseur, and G. Megie, 1994: The importance of atmospheric chemistry in the calculation of radiative forcing on the climate system, J. Geophys. Res. 99, 1173--1186.

Henderson-Sellers, A., 1993: Continental vegetation as a dynamic component of a global climate model: a preliminary assessment. Climate Change 23, 337--377.

Hibler, W.D., III, 1979: A dynamic thermodynamic sea ice model. J. Phys. Ocean. 9, 815--846.

Holland, E.A., C. Coxwell, D.S. Schimel, and D. Valentine, 1993: A model of methane production in soils. Bulletin of the Ecological Society of America, 78th Annual ESA Meeting, Ecology and Global Sustainability, University of Wisconsin, Madison, Wisconsin, July 31--August 4, 1993 74(2).

Holland, W.R., and F.O. Bryan, 1993a: Sensitivity studies on the role of the ocean in climate change. Ocean Processes in Climate Dynamics: Global and Mediterranean Examples. P. Malanotte-Rizzoli and A. R. Robinson (Eds.), NATO ASI Series, Kluwer Academic Publishers, Dordrecht, 111--134.

Holland, W.R., and F.O. Bryan, 1993b: Modeling the wind and thermohaline circulation in the North Atlantic Ocean. Ocean Processes in Climate Dynamics: Global and Mediterranean Examples. P. Malanotte-Rizzoli and A. R. Robinson (Eds.), NATO ASI Series, Kluwer Academic Publishers, Dordrecht, 135--156.

Hurrell, J.W., J.J. Hack, and D.P. Baumhefner, 1993: Comparison of NCAR Community Climate Model (CCM) climates. NCAR Technical Note NCAR/TN--395+STR, Boulder, Colorado, 335 pp.

IPCC, 1990: Climate Change: The IPCC Scientific Assessment, J.T. Houghton, G.J. Jenkins, and J.J. Ephraums (Eds.), Cambridge University Press, 365 pp.

IPCC, 1992: Climate Change 1992: The Supplementary Report to The IPCC Scientific Assessment, J.T. Houghton, B.A. Callander, and S.K. Varney (Eds.), Cambridge University Press, 200 pp.

Jin, F.-F., J.D. Neelin, and M. Ghil, 1994: El Nino on the Devil's Staircase: Annual Subharmonic Steps to Chaos. Nature 264 70--72.

Johnson, K.D., D. Entekhabi, and P.S. Eagleson, 1993: The implementation and validation of improved land-surface hydrology in an atmospheric general circulation model. J. Climate 6, 1009--1026.

Kalnay, E., M. Kanamitsu, J. Pfaendtner, J. Sela, M. Suarez, J. Stackpole, J. Tuccillo, L. Umscheid, and D. Williamson, 1989: Rules for interchange of physical parameterizations. Bull. Amer. Meteor. Soc. 70, 620--622.

Kiehl, J.T., 1994: Sensitivity of the simulated climate to effective drop size specification. J. Geophys. Res., in preparation.

Kiehl, J.T., J.J. Hack, and B.P. Briegleb, 1994: The simulated earth radiation budget of the NCAR CCM2 and comparisons with the Earth Radiation Budget Experiment (ERBE). J. Geophys. Res., in press.

Koster, R.D., and M.J. Suarez, 1992: A comparative analysis of two land surface heterogeneity representations. J. Climate 5, 1379--1390.

Large, W.G., J.C. McWilliams, and S.C. Doney, 1994: Oceanic vertical mixing: A review and a model with a nonlocal K-profile boundary layer parameterization. Rev. Geophys., submitted.

Lemke, P., W.B. Owens, and W.D. Hibler, III, 1990: A coupled sea ice--mixed layer--pycnocline model for the Weddell Sea. J. Geophys. Res. 95, 9513--9525.

Levitus, S., M.E. Conkright, J.L. Reid, R.G. Najjar, and A. Mantyla, 1993: Distribution of nitrate, phosphate and silicate in the world oceans. Prog. Oceanog. 31, 245--273.

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.

Manabe, S., K. Bryan, and M.J. Spelman, 1975: A global ocean-atmosphere climate model. Part I. The atmospheric circulation. J. Phys. Oceanogr. 5, 3--29.

Manabe, S., and K. Bryan, 1969: Climate calculations with a combined ocean-atmosphere model. J. Atmos. Sci. 26, 786--789.

Manabe, S., and R.J. Stouffer, 1993: Century-scale effects of increased atmospheric CO on the ocean-atmosphere system. Nature 364, 215--218.

Mechoso, C.R., C.-C. Ma, J.D. Farrara, and J.A. Spahr, 1993: Parallelization and distribution of a coupled atmosphere-ocean general circulation model. Mon. Wea. Rev. 121, 2062--2076.

Meehl, G.A., 1990: Development of global coupled ocean-atmosphere general circulation models. Clim. Dyn. 5, 19--33.

Meehl, G.A., G.W. Branstator, and W.M. Washington, 1993: Tropical Pacific interannual variability and CO climate change. J. Climate 6, 42--63.

Melillo, J.M., A.D. McGuire, D.W. Kicklighter, B. Moore, III, C.J. Vorosmarty, and A.L. Schloss, 1993: Global climate change and terrestrial net primary production. Nature 363, 234--240.

McWilliams, J.C., and P.R. Gent, 1994: The wind--driven ocean circulation with an isopycnal--thickness mixing parameterization. J. Phys. Ocean. 24, 46--65.

Ojima, D., Editor, 1992: Modeling the Earth System. UCAR/Office for Interdisciplinary Earth Studies, Boulder, Colorado, 488 pp.

Owens, W.B., and P. Lemke, 1990: Sensitivity studies with a sea ice-mixed layer-pycnocline model in the Weddell Sea. J. Geophys. Res. 95, 9527--9538.

Parkinson, C.L., and W.M. Washington, 1979: A large--scale numerical model of sea ice. J. Geophys. Res. 84, 311--337.

Parton, W.J., D.S. Schimel, D.S. Ojima, and C.V. Cole, 1994: A general model for soil organic matter dynamics: sensitivity to litter chemistry, texture and management. Soil Science Society of America Journal, in press.

Pinardi, N., A. Rosati, and R.C. Pacanowski, 1994: The sea surface pressure formulation of rigid lid models. Implications for altimetric data assimilation studies. Preprint.

Pitman, A.J., Z.-L. Yang, J.G. Cogloy, and A. Henderson-Sellers, 1991: Description of Bare Essentials of Surface Transfer for the Bureau of Meteorology Research Centre AGCM. BMRC Research Report, Bureau of Meteorology Research Center, Melbourne, Victoria, Australia, 32, 117 pp.

Pollard, D., and S.L. Thompson, 1994: Use of a Land-Surface-Transfer Scheme (LSX) in a global climate model: The response to doubling stomatal resistance. Global and Planetary Change, submitted.

Rasch, P.J., X.X. Tie, B.A. Boville, and D.L. Williamson, 1993: A three-dimensional transport model for the middle atmosphere. J. Geophys. Res. 99, 999--1018.

Sarmiento, J.L., R.D. Slater, and M.J.R. Fasham, 1993: A seasonal three--dimensional ecosystem model of nitrogen cycling in the North Atlantic euphotic zone. Global Biogeochem. Cycles. 7, 417--450.

Sausen, R., K. Barthel, and K. Hasselmann, 1988: Coupled ocean-atmosphere models with flux correction. Clim. Dyn. 2, 145-163.

Schimel, D.S., B.H. Braswell, E.A. Holland, R. McKeown, D.S. Ojima, T.H. Painter, W.J. Parton, and A.R. Townsend, 1994: Climatic, edaphic and biotic controls over storage and turnover of carbon in soils. Global Biogeochemical Cycles, in press.

Schimel, D.S., W.J. Parton, T.G.F. Kittel, D.S. Ojima, and C.V. Cole, 1990: Grassland biogeochemistry: links to atmospheric processes. Climatic Change 17, 13--25.

Sellers, P.J., J.A. Berry, G.J. Collatz, C.B. Field, and F.G. Hall, 1992: Canopy reflectance, photosynthesis, and transpiration. III. A reanalysis using improved leaf models and a new canopy integration scheme. Remote Sensing of Environment 42, 187--216.

Semtner, A.J., and R.M. Chervin, 1992: Ocean general circulation from a global eddy-resolving model. J. Geophys. Res. 97, 5493--5550.

Siegenthaler, U., and J.L. Sarmiento, 1993: Atmospheric carbon dioxide and the ocean. Nature 365, 119--125.

Smith, R.D., J.K. Dukowicz, and R.C. Malone, 1992: Parallel ocean general circulation modeling. Physica D 60, 38--61.

Smolarkiewicz, P.K., 1984: A fully multidimensional positive definite advection transport algorithm with small implicit diffusion. J. Comp. Phys. 54, 325--362.

Stossel, A., P. Lemke, and W.B. Owens, 1990: Coupled sea ice--mixed layer simulations for the Southern Ocean. J. Geophys. Res. 95, 9539--9555.

Stouffer, R.J., S. Manabe, and K. Bryan, 1989: Interhemispheric asymmetry in climate response to a gradual increase of CO. Nature 342, 660--662.

Tans, P.P., I.Y. Fung, and T. Takahashi, 1990: Observational constraints on the global atmospheric CO budget. Science 247, 1431--1438.

Taylor, K.C., G.W. Lamorey, G.A. Doyle, R.B. Alley, D.M. Grootes, P.A. Mayewski, J.W.C. White, and L.K. Barlow, 1993: The ``flickering switch" of late Pleistocene climate change. Nature 361, 432--435.

Thompson, L.G., E. Mosely-Thompson, M.E. Davis, J.F. Bolzani, J. Dai, T. Yao, N. Gunderstrup, X. Wu, L. Klein, and Z. Xie, 1989: Holocene-late Pleistocene climatic ice core records from Quinhai-Tibetan Plateau. Science 246, 474--477.

Toggweiler, J.R., K. Dixon, and K. Bryan, 1989a: Simulations of radiocarbon in a coarse--resolution world ocean model, 1, Steady state prebomb 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, 1, Distributions of bomb--produced carbon 14. J. Geophys. Res. 94, 8243--8264.

Trenberth, K.E., Editor, 1992: Climate System Modeling. Cambridge University Press, 788 pp.

Tziperman, E., L. Stone, M.A. Cane, and H. Jarosh, 1994: El Nino Chaos: Overlapping of Resonances Between the Seasonal Cycle and the Pacific Ocean-Atmosphere Oscillator. Science 264 72--74.

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 G.A. Meehl, 1989: Climate sensitivity due to increased CO: Experiments with a coupled atmosphere and ocean general circulation model. Clim. Dyn. 4, 1--38.

Washington, W.M., and G.A. Meehl, 1993: Greenhouse sensitivity experiments with penetrative cumulus convection and tropical cirrus albedo effects. Clim. Dyn. 8, 211--223.

Williamson, D.L., and P.J. Rasch, 1994: Water vapor transport in the NCAR CCM2. Tellus 46A, 34--51.

Williamson, D.L., J.T. Kiehl, and J.J. Hack, 1994: Climate sensitivity of the NCAR Community Climate Model (CCM2) to horizontal resolution. Clim. Dyn., submitted.

VIII. Appendix -- CMAP SAC Members

IX. Acronyms