[an error occurred while processing this directive] [an error occurred while processing this directive]
Minutes of the
CCSM Advisory Board
12 and 13 February 2001
NSF, Arlington, VA
CAB Attendees: Ed Sarachik (Chair), Tony Hollingsworth, Steve Running, Francis Bretherton, Steve Zebiak, Max Suarez, Eric Sundquist, Isaac Held, Bob Malone, Alan Thorpe, Rick Anthes, Maurice Blackmon, Tim Killeen, and Jay Fein
Other Attendees: David Verardo, NSF; Pam Stephens, NSF; J. Shukla, COLA; Anjuli Bamzai, NSF; Ferd Baer, University of Maryland; Jerry Elwood, DOE; Jack Kaye, NASA; Jim Hack, NCAR; Cliff Jacobs, NSF; Jerry Potter, DOE; Gordon Bonan, NCAR; Cecelia DeLuca, NCAR; Dick Moritz, University of Washington; Eric Lindstrom, NASA; Dave Bader, DOE; Chuck Hakkarinen, EPRI; Phil Arkin, NOAA; Yogesh Sud, NASA; Eric Itsweire, NSF; Steve Meacham, NSF; Louis Uccellini, NCEP; Jarvis Moyers, NSF; Vaughan Turekian, NRC Staffer; Patricia Esborg; and Ricky Rood, NASA
1. Status of CCSM-2. Blackmon briefly detailed the history of the project starting in 1994. At this time there is a community effort to improve all the model components that make up the Community Climate System Model (CCSM) and the coupler. The NCAR computer acquisitions that have been made over the last few years have slowed down the process due to different computer architectures.
Blackmon also reported that the Scientific Steering Committee (SSC) and the Working Group Co-Chairs met recently at NCAR, and the Atmosphere Model, Ocean Model, Land Model, and Polar Climate Working Group Co-chairs submitted their group's recommendations for the new component models that will become CCSM-2. Each working group also recommended that they be given more time to complete the new components before freezing the model. The SSC is considering the release of an interim CCSM model (iCCSM).
Ocean Component. Blackmon reported that the first ocean model component was based on MOM1 and NCAR physics that ran on a vector computer but did not run on parallel computers. The working group chose to abandon that model and use the POP model, which does run on parallel computers, and add NCAR physics to it. This new ocean model component is called POP1.4. The horizontal viscosity still needs further work; the Gent-McWilliams (G-M) and KPP implementations have been made consistent and G-M now allows different coefficients in advection and diffusion terms; and the river runoff and marginal sea balancing code needs to be tested and incorporated.
Future plans are to complete runs with the Visbeck et al. scheme that has been implemented and that has the diffusivity coefficients as functions of space; complete runs with the Beckmann and Doscher scheme that has been implemented to improve the bottom boundary layer; complete runs with the improved partial bottom cells scheme; and do more ocean-alone runs with new boundary conditions under sea-ice before running the x3 ocean model coupled to sea ice to equilibrium.
Sea Ice Component. Dick Moritz (co-chair of the Polar Climate Working Group) reported on this group's development of the sea ice model. The members of the core developer's group within the Polar Climate Working Group are Elizabeth Hunke (Los Alamos National Laboratory), Cecelia Bitz (University of Washington); Bruce Briegleb (NCAR); Bill Lipscomb (Los Alamos National Laboratory); Tony Craig (NCAR); and Julie Schramm (NCAR). This new sea ice model component includes enhanced sea ice thermodynamics that resolve vertical temperature profiles, energy conserving vertical boundary conditions, and explicit brine pocket parameterization; multi-category sea ice thickness distribution; plastic ice rheology with an elliptical yield curve that represents shear and normal stresses; uses a generalized orthogonal curvilinear coordinate grid; and runs efficiently on distributed shared-memory platforms. Hunke delivered the code to NCAR and it is running through the coupler.
Land Component. The Land Model Working Group (LMWG) has developed a new Common Land Model (CLM) that did not have all the conventions needed to be a part of CCSM. The group agreed to merge the new CLM with the NCAR Land Surface Model (LSM) that is used in CSM. Gordon Bonan (co-chair of the LMWG) gave a presentation on the merged version of the CLM and the LSM. The data structures issue was solved by adopting the CLM data structures and processing, and this task has been completed. The variable names and units issue was solved by using a standard set of variable names and units, and this task has been completed in CLM and is in progress in LSM. The constants issue of different names, different values, and different accessing was solved by using the LSM approach, and the CLM work is in progress. The vegetation and soil types issue was solved by using the LSM approach. The albedo issue was solved by using the LSM albedo and radiative transfer approach. Testing is in progress.
Atmosphere Component. Jim Hack (co-chair of the Atmosphere Model Working Group) gave a presentation outlining the process used by this group to decide on a recommendation to the SSC for the next version of the atmosphere model for CCSM-2. First the group adopted a standard experimental strategy using 5-year climatological SST forcing to select the top candidates for additional evaluation. Then a 15-year Atmosphere Model Intercomparison Project (AMIP) II simulation was used to select the winning configuration, since a coupled integration to evaluate the coupled model performance was not yet possible.
The group also adopted a standard diagnostic strategy to facilitate a standard analysis of all the candidate models. A standard minimum set of diagnostics was exploited, starting with the WGNE standard climate diagnostics that was enabled with a "turn-key" diagnostic package by Mark Stevens at NCAR. A number of other diagnostic analyses were completed by members of the CCSM community, including an extensive diagnostic analysis from staff at PCMDI. All the candidate model simulations were posted on the Atmosphere Model Working Group (AMWG) web page and succeeded in entraining the broader community as participants in the analysis and discussions of the configuration options.
The AMWG recommendation consists of a computationally efficient, two-time level Semi-Lagrangian dynamical (SLD) core using a reduced linear grid, 26 vertical levels, prognostic cloud water, generalized cloud overlap, a modern treatment of water vapor absorption in the longwave radiative transfer parameterization, and a version of the Arakawa-Schubert cumulus parameterization developed at Colorado State University (CSU). The new cumulus convection parameterization includes the provision for multiple cloud bases and cloud tops, employs an equilibrium cumulus kinetic energy closure, allows for the handling of arbitrary tracer transport, and includes an extendable framework that can be directly coupled to stratiform cloud and boundary layer parameterizations. The simulation exhibits improvements in precipitable water, implied ocean heat transport, polar thermal bias, tropical variability, and the eastern-ocean surface solar energy budget.
There were also some weaknesses noted in the proposed model, including wintertime stationary wave structure; excessive precipitation maxima associated with major precipitation regimes in the eastern and western Pacific; no improvement in the Arctic surface wind field; a cold tropical tropopause, and non-conservative advection of tracers, a deficiency that is similar to the Community Climate Model version 3 (CCM3).
The AMWG feels that the proposed atmospheric model provides improvements to the physical and dynamical framework, improvements to the simulation, and optimal computational performance. Perhaps the greatest success of this activity has been the establishment of a "process" for continued model development work that involves the broader national scientific community.
Blackmon reported that the SSC requested that further work be done on the Semi-Lagrangian dynamical core and CSU physics atmosphere model component. He also reported that the Lin-Rood finite volume atmosphere model component was desirable as a candidate atmosphere model.
Software Engineering. Cecelia DeLuca (co-chair of the Software Engineering Working Group, SEWG) reported that a CCSM Programmers Group has been formed to establish version control, test and validation, and data file management procedures. They have already assembled a version of CCSM 2.0.x; ported it to the IBM; improved the coupler's performance; improved portability, flexibility, and usability of scripts and build; tested preliminary performance and identified bottlenecks; verified "suitable" performance for production; and updated the coupler interface for the new components.
She also reported that a CCSM Software Engineering Plan had been written and that day-to-day software management was established, a CCSM Software Developer's Guide is being written to collect conventions, and each time the Software Engineering Working Group meets they discuss and establish new procedures. In summary, she reported that the current projects and initiatives provide shallow coverage of software needs, there are gaps in computer science research and user support, and the SEWG is actively trying to coordinate and promote new activities.
2. Summary of Next 5 to 10 Year Plans of USGCRP. Dr. Jack Kaye, representing the Subcommittee for Global Change Research (SGCR), gave a presentation on the U.S. Global Change Research Program's (USGCRP) 10 Year Strategic Plan. This strategic plan calls for SGCR oversight, steering committee leadership, working group preparation of detailed plans for research elements, external research community participation, and an iterative process. Below are the details of the strategic plan.
The mission of the USGCRP is to develop and apply research-based capabilities required to improve forecasts of global change, assess vulnerability, and provide useful information.
Vulnerability is defined as the risk of adverse consequences, and the resilience is the ability to avoid, adapt to, or recover. Three components of vulnerability are exposure, sensitivity, and adaptive capacity. The USGCRP will continue support for basic research and the development of new capabilities that support decision making and adaptation capabilities. Research will be conducted at finer spatial scales and will include stronger interactions with users in acquiring and applying knowledge and broader involvement of management agencies.
The research elements are atmospheric composition, climate variability and change, carbon cycle, water cycle, land use and land cover change, and terrestrial and aquatic ecosystem resources. The attributes of the research elements are that they reflect strengths of research and federal communities, they were identified in the National Academy's "Pathways Report," and the goals reflect science and societal challenges
The integrating science questions are (1) what are the major natural and human forcings?; (2) what are the relationships between forcing and natural variability?; (3) what are natural and human induced changes?; and (4) what will evolve between human interactions with the environment?
In enabling and integrating the activities, human dimensions of global environmental change, regional research and interactions, observations, modeling, data and information, etc., are involved. The characteristics of enabling and integrating these activities are: cross-cutting intellectual or enabling elements; high priority for GCRP but difficult to support, organize, and manage; each activity must be pursued along two tracks and closely coordinated with the six research elements; and the need for further development and consensus building.
The human dimensions show that the framework is better developed than the mechanism for supporting research, and further interagency discussion is needed. Regional research and interaction is important for applications, in particular, in regions under multiple stresses. As a first step, regional pilots should be designed with an integrated approach to observations, process studies, modeling, and information. To obtain the observations we must cooperate with other observation programs, and we need to protect and improve surface and balloon-based measurements and connect to new (e.g., satellite) measurements. Financial considerations alone suggest that adding new components will not be simple. We need to improve connections between research and operational systems, and we need to continue enlargement in international activities (e.g., GOOS).
In the data and information management area, we need volume and diversity of data and there is some progress, we need to be linked to model and assessments of worldwide data, we need integrated data and information, and we need to consider possible alternative implementation strategies.
In the modeling area, we need to continue to improve quality of earth system models, increase capabilities to provide large numbers of runs for assessment, develop closer relationships with operations modeling, focus on intercomparison and evaluation of models, and integrate modeling with other research and assessment activities.
In the program development, prioritization, and review area, the basic elements have been described as the need for unprecedented cooperation and coordination to attain stated objectives and greater focus on deliverables; interagency working groups should be responsible for periodic implementation plans, coordination, and deliverables; prioritization should be according to criteria of scientific merit, importance, tractability, and linkages; and there should be program review and evaluation involving program management and stakeholders on a periodic basis.
Further development is needed with regard to connections between elements and activities; relationship between assessments and decision support; terrestrial and marine ecosystems; relationship between carbon cycle, ecosystem, land use and coverage, and human dimensions; integrated assessment; and operating and management relationships.
The next steps are review by the research community, working groups meetings to revise the research elements, expansion of the interagency steering committee, and interaction with the Bush administration.
3. Agency Climate Modeling Plans
NASA. Eric Lindstrom reported that NASA's climate modeling program consists of Jim Hanson's research at Goddard Institute for Space Studies (GISS); seasonal-to-interannual research work being done at Goddard Space Flight Center; research being performed with NSF and Office of Naval Research (ONR) on estimating the circulation and climate in the ocean; research being done at the Center for Ocean-Land-Atmosphere Studies (COLA); research being done at the International Pacific Research Center in Hawaii on the Pacific and Indian Oceans; and the CLIVAR program. Another area of interest at NASA is decadal-to-centennial climate variability.
DOE. Dave Bader reported that the DOE Climate Change Prediction Program (CCPP) core activities prior to 2001 consisted of ocean and sea ice model development at Los Alamos National Laboratory (LANL), model intercomparison and diagnostics work at the Program for Climate Model Diagnosis and Intercomparison (PCMDI), long-term climate change prediction research through the NCAR Cooperative Agreement, and computational support at LANL and Oak Ridge National Laboratory. In addition, university grants and single PI DOE lab projects were funded.
He then discussed the changes to the CCPP in 2001, which are a refocus at PCMDI with increased emphasis on diagnostics; closer interactions with modeling development such as participation in the analysis of CCSM atmosphere model candidates; linking with the DOE Atmospheric Radiation Measurement (ARM) project by using a Lin-Rood based AGCM as a parameterization testbed; Climate Model Intercomparison Program (CMIP) runs; diagnostic analysis tools and software; and community outreach and support. Further changes are the incorporation of ACPI (Accelerated Climate Prediction Initiative) and SciDAC (Scientific Discovery through Advanced Computing) with a new funding announcement, elimination of small, single PI lab projects, and the differentiation of lab versus academic roles in that labs will contribute to U.S. modeling infrastructure, lab research will be for intermediate timeframes (2 to 5 years), and academic research will be in the longer term (5 to 10 years).
The CCPP/SciDac RFP includes grant applications for predictability and climate dynamics, parameterization development, numerical methods and algorithms, and diagnostics; cooperative agreement applications for the climate model of the future that includes integrated, multi-disciplinary research of climate, applied math, and computational science; and emphasizes multi-disciplinary graduate training. The lab part of the RFP includes model development on a 2 to 5 year timeframe that consists of multi-institutional, comprehensive integrated proposals for model development, requires self organization, and requires science and computational science and software engineering.
One problem area associated with the growth and evolution of a successful program, such as CCSM, is multi-agency support. The different agencies (NSF, DOE, NASA, NOAA) have common problems but different priorities; therefore, the CCSM structure must recognize the conflicts and make sure all the multi-agency sponsors are satisfied.
Bader said that the DOE approach to CCSM will be (1) to focus on IPCC-type climate change simulations that Warren Washington is leading, (2) to contribute to the general development of CCSM where possible within the DOE mission, (3) to support extra effort needed to meet DOE missions, and (4) to try to accomplish a delicate balance between (2) and (3) that is required when conflict and controversy arise.
NOAA. Phil Arkin, NOAA Office of Global Programs, reported that NOAA's mission is climate services. NOAA is supporting the Applied Research Centers (ARCS) such as the Climate Prediction Center at the International Research Institute for climate prediction (IRI) that is involved in assessment and evaluation programs; Geophysical Fluid Dynamics Laboratory (GFDL), where a new director may change the focus of that program; COLA; Scripps; Florida State University, etc.
NSF. Jay Fein detailed the present NSF budgets relating to climate modeling. NSF funds USGCRP research at NCAR at $1 million dollars annually that provides funding for community liaisons, meetings and workshops, and research. NSF also funds university grants (PIs) at $1 million dollars annually, including about $0.25 million per year in small grants with highly focussed near-term objectives relevant to CCSM priorities. In addition, there is about $3 million per year provided to CCSM-related research from NCAR's base funding. NSF future plans are to support fundamental research applied to model development, testing, and application; to engage a broad community of scientists working on and with CCSM; and to develop explicit links to national and international assessments.
The meeting adjourned so participants could attend the seminar given by Tony Hollingsworth on Research on Medium-Range and Seasonal Forecasting at ECMWF.
4. NAS Report. On Tuesday, 13 February 2001, the meeting began with a report by Ed Sarachik on the National Academy of Sciences (NAS) Panel on Modeling. He reported that a survey of modeling programs was done with questions relating to machines and people. Every program reported via the survey that more computing capabilities were needed and human resources was an issue because everyone needed additional and replacement staff. Some of the concerns noted in the report include: duplication of effort and lack of federal agency coordination and common infrastructure. The NAS report is in committee review and has not been published yet.
Killeen stated that NCAR is writing a computing strategic plan. Right now NCAR has 1/2 teraflop computing capability and plans to have a 10-teraflop peak in 2004.
5. High-End Climate Science: Development of Modeling and Related Computing Capabilities. A Report to the USGCRP from the ad hoc working group on Climate Modeling. Ricky Rood gave a presentation (copies available upon request) outlining the findings of this committee regarding what needs to be done to address significant shortcomings in the U.S.'s ability to provide needed climate services. A recommendation was that if the nation is going to provide "climate products" for prediction, assessment, etc., then a policy decision needs to be made to that effect. This will require the development of a "product-driven" research capability; the development of a research organization that the current agency culture is not well suited to support; a timely response is reliant upon existing resources that currently reside in the agencies; and there is conflict of how to focus agency-held resources "outside" of existing agency culture. The formation of a product-driven climate service would be built from existing expertise, new funding would be supplemented, there would be a clear separation from current agency activities, and the service would be "outside" of the present agency culture. For this change to be successful, leadership, strategic thinking, and culture change will be required. Otherwise, the current components (and funding) will just be rearranged.
The report also stated that the shortage of human resources with scientific and information technology expertise is an issue.
The report has been published by the USGCRP and has been posted on the web at http://www.usgcrp.gov/usgcrp/Library/models2001/default.htm.
6. Miscellaneous. Alan Thorpe reported that the Hadley Centre was inherently part of the U.K. Met Office and they share in model research and operations. He said there was tremendous synergy in being part of the national meteorological service. The U.K. universities use the model, but they do not participate in developing it. Also, the IPCC is not a customer of the Hadley Centre since they do not give money to them. The U.K. has one agency that funds the Hadley Centre and it uses the research to make policy. The Hadley Centre uses a Cray T3E. Thorpe stated that in the U.S. research area that institutional protection seemed to be a problem and that it was clear that multi-agency cooperation is critical but a big problem.
Some CAB members see the CCSM as a product, but Blackmon stated that the CCSM was research and not a product. Blackmon also stated that CCSM is a volunteer based project. Fein stated that if CCSM goes towards production, the project would lose people who want to do research and model development. Suarez suggested that the SSC freeze the CCSM and also have another version as a research-based model for community use. Blackmon stated that the CCSM project needs to release a new version of the model to keep up interest. It was suggested that instead of using the term "freezing the model" that we use "release of new version with improvements." Another issue that was mentioned was how long NCAR would support older versions of the model.