[an error occurred while processing this directive] [an error occurred while processing this directive]

Minutes of the CSM Advisory Board Meeting

Thursday and Friday, 29 and 30 April 1999

NSF, Room 770, Arlington, VA


Members Attending: Ed Sarachik (Chair), Kevin Trenberth, Francis Bretherton, Bert Semtner, Bob Malone, Steve Running, Dick Berk, Michael Ghil, Steve Zebiak, and Isaac Held

Ex-Officio Members Attending: Rick Anthes, Bob Serafin, Jay Fein, Maurice Blackmon, and Anjuli Bamzai

Guests: Dick Greenfield, Joe Friday, Beth Robinson, Bob Borchers, Dave Bader, Mark Eakin, Steve Colman, Ted Elliott, Pam Stephens, Eric Itsweire, Steve Meacham, Ken Bergman, Cliff Jacobs, Jewel Prendeville, Scott Weidman, Ricky Rood, Eric Lindstrom, Dave Goodrich, Anne-Marie Schmoltner, Jerry Elwood, Jim Rosenberger, Ari Patrinos, Peter Schultz, Al Kellie, and Lydia Harper

1. Ed Sarachik, Chair of the Climate System Model (CSM) Advisory Board (CAB), welcomed the CAB members and guests. Dick Greenfield also welcomed everyone and said we should give special thanks to Rick Anthes and Francis Bretherton for their long-range views on climate modeling science and also to Michael Ghil as the former chair of CAB (then CSM Scientific Advisory Committee, SAC).

2. Sarachik asked that the topic of a U.S. national modeling plan be discussed later. A national modeling plan is needed, and it should include information on relationships to other sciences.

3. Maurice Blackmon thanked Jay Fein, Dick Greenfield, and Cliff Jacobs for hosting the CAB meeting at NSF.

Blackmon presented a history of CSM beginning in the early 90's. The Climate System Modeling Program (CSMP) was started at UCAR, funded by NSF's U.S. Global Change Research Program's (USGCRP) Climate Modeling, Analysis, and Prediction (CMAP) initiative, and also by DOE, NOAA, and NASA during its early period. A Science Advisory Council (SAC) was appointed by UCAR to provide advice to the UCAR President, the Director of NCAR, and the NSF Climate Dynamics Program Director on the CMAP program that included CSM. In 1994, several NCAR scientists prepared the first CSM Plan, which was reviewed extensively internally at NCAR, by Jay Fein at NSF, and by an anonymous NSF mail review. The first CSM Workshop was held in 1996, and the CSM Scientific Steering Committee (SSC) was appointed to provide scientific advice and oversight to the CSM Project. The second CSM Workshop was held in June, 1997. The third CSM Workshop was held in June, 1998, after which the SAC became the CSM Advisory Board (CAB), because CSM had evolved to be the major component of CMAP and the most complex. The next CSM Workshop will be in June, 1999.

The structure of the CSM project was reviewed. The CAB provides advice to the UCAR President, NCAR Director, and NSF Program Manager and to the SSC. There are presently nine working groups that make up the CSM project, and they report to the SSC. The CSM Program Office has provided information on the CAB, SSC, and Working Groups and CSM models and documentation via the World Wide Web, including the experiments completed and results obtained. The CSM website can be found at http://www.cesm.ucar.edu.

At present, the CSM scientists and support staff, along with many non-NCAR collaborators from universities and national laboratories, are working on the next version of the Community Climate Model (CCM4), an ocean model (NCOM-2), and improvements to the sea ice model. The CSM SSC is expecting the working groups to recommend testing and comparison of the improved models, so that the next version of CSM can be released in June, 2000 (CSM 2.0).

Blackmon discussed the CSM goals for the future. At present we have a CSM model with variations of the component models having different resolutions and parameterizations. A convergence in various experimental models will take place, and it is hoped that CSM 2.0 will be ready in June, 2000. Further in the future as improvements are made in the model components, CSM 3.0 will be released. Some research areas have special needs, such as a different ocean model than the one now in CSM for season-to-interannual variability research. Paleoclimate research needs a low-resolution model and land modeling needs a high-resolution model. Another area for improvement is that the components of the atmosphere model run at different resolutions. Bob Serafin discussed how expensive it is to support several models, each needing resources for development, support, and computing. Jay Fein reported that in FY 1999 NSF awarded 7 small, one-time grants (approximately $25K) to young scientists working on CSM diagnostics and development. NSF plans to continue this practice, as needed. The research projects funded are reviewed by the CSM SSC, which recommends projects to be funded. (A list of NSF CSM-related current awards is available if requested.)

Blackmon reiterated that the CSM data and documentation is available via the web to anyone who wants it.

Ghil stated that a common computer language framework would be beneficial to all sciences. It was pointed out that this issue is on the agenda (see 5. below).

The CAB asked that the new CSM management plan describe how decisions are made.

4. Al Kellie, NCAR Scientific Computing Division Director, updated the CAB on the direction of computing at NCAR. He said that SCD would be a teraflop computing center with a mass storage system, data sets, and data visualization and analysis. SCD plans to have 1 teraflop computing power in 2004. The SCD short-term strategy is an immediate upgrade (June, 1999) using the best current-price performance machines with minimal disruption. There will be porting and optimization assistance available. The focus will be on a new microprocessor system. The C90 contract will not be renewed after October, 1999, and SCD will phase out the two old, classic J90s and HP SPP2000. SCD will also gain hands-on cluster experience. The long-term strategy is to continue to watch the market and evaluate performance curves, evaluate clustering multiple types of computers, and collaboration with PACI centers. (A copy of Kellie's viewgraphs is available if requested.)

5. Steve Zebiak reviewed the Common Infrastructure Initiative (CII) designed to avoid duplication and enhance cooperation among modeling groups. In August, 1998, a NCEP/NSF workshop was held on exploring ways to form more active partnerships between research scientists conducting climate modeling and operational global numerical weather prediction scientists. The outcome of the meeting was the endorsement of a common modeling infrastructure and common development efforts (seasonal-to-interannual, numerical weather prediction, and global change). In October, 1998, the first Infrastructure Workshop was held in Tucson, Arizona. Several subgroups were proposed to investigate column physics coding standards, common support tools, dynamical core compatibility, model/data repository, existing flux coupler designs, and a data format survey will be developed and conducted. In January, 1999, the second Infrastructure Workshop was held in Dallas, Texas, resulting in the column physics coding rules being revised, further discussion of a data repository and dynamical core experiment, and a convective parameterization repository experiment was initiated. In April, 1999, further discussion was held on standard support tools, data format standards (netCDF, HDF, GRIB), examining GDT and NCAR/CDM for metadata standards, considering software repository for common tools, and assessing BC and evaluation standards for models. By this time, successful use of the convective code repository had been achieved in the context of the Naval Research Laboratory operational weather forecast system. Zebiak also showed the website (authored by Max Suarez, Goddard Space Flight Center) to find information about this ongoing effort. The URL is http://janus.gsfc.nasa.gov/~suarez/infra. (A copy of Zebiak's viewgraphs is available if requested.) The CAB discussion concluded that there is a sense of optimism across the broad community for CII goals. It was suggested that the issues of common infrastructure, coding, and visualization standards are common to many disciplines and perhaps broad progress could be made through professional societies and the National Research Council (NRC).

6. Dick Berk reported on the National Academy of Sciences Committee on Applied and Theoretical Statistics Initiative to include more statisticians in climate modeling. It was suggested that Berk contact Doug Nychka, NCAR's Geophysical Statistics Project Leader.

7. Ed Sarachik discussed the CRC Report. The committee found that in climate modeling and research there is no integrated national strategy, no common infrastructure, supercomputing problems, and a real need to define needed resources specifically. The committee also found that a mechanism for coordinated, interagency cooperation is needed, a national model is needed, access to large computer resources is needed, access to model output is crucial, commercialization issues should be explored, interaction between scales (weather, seasonal-to-interannual, decadal, and long-term, e.g., Ice Age) is needed, and a facility is needed.

8. It was recommended that more CSM-data CD-ROMs should be produced, and Sarachik reported on how valuable the previously distributed CD-ROM had been. Blackmon stated that a survey could be done to decide what data is wanted on CD-ROM.

9. Beth Robinson, OMB Examiner for DOE, reported on the Office of Management and Budget's view of the Information Technology for the 21st Century (IT2) Initiative. She stated that "IF" the IT2 initiative is funded by Congress, it would add $366 million to the $1.5 billion High Performance Computing and Communications (HPCC) base program to augment key areas of HPCC research and development. The IT2 goals were set almost exclusively by PITAC (President's Information Technology Advisory Committee). The IT2 high-end computing goals are: 5 teraflop peak machine(s) available in 2001 to 2003, 40 teraflop peak machine(s) in 2004 to 2005, DOE and NSF will spearhead the effort to provide hardware advances, and the mechanisms for designing and siting the machines and allocating their cycles are yet to be determined.

Robinson stated that the USGCRP needs to partner with IT2, and it also needs to work with IT2 on the applications/high-end component to ensure IT2 becomes useful for science. She said that ACPI (Accelerated Climate Prediction Initiative) is part of the DOE Strategic Simulation Initiative (SSI) and ACPI could partner with IT2 also. (A copy of Robinson's viewgraphs is available if requested.)

10. Bob Borchers, NSF Division Director, Advanced Computational Infrastructure and Research, reported on high-end computing research priorities for IT2. The mission is to develop enabling technologies for computational science and to advance the state-of-the-art in high-end computing. The strategy to address the mission is to continue and expand scalable software support, to extend the scientific visualization program, to launch a major program for scientific applications (Terascale Opportunities for Promoting Science), and to acquire terascale hardware. The current status of the terascale acquisition is: NSF broadly recognizes the importance of terascale computing capability, NSF is committed to cooperation with DOE SSI, NSF ideas will be integrated in the broader multi-agency response to the PITAC interim report, $36 million included in NSF's FY 2000 Presidential budget, and NSF committed to OMB/OSTP "principles" for competitive acquisition. Current plans call for one facility. If funds increase, two may be possible. Robinson pointed out that currently the President's IT2 hardware budgets are flat beyond FY 2000.

The acquisition will be coordinated with DOE and the timeline is: in January, 1999, the Presidential FY 2000 budget included $36 million for high-end hardware; in April, 1999, NSF prepared a RFP to procure and operate a terascale facility; in June, 1999, the RFP will be released; in September, 1999, proposals will be received; in January, 2000, proposal review will be completed; in February 2000, a recommendation will be made to the National Science Board; in March, 2000, the awards will be announced; in June, 2000, the first deliveries of hardware will be made; and in August, 2000, "friendly users" will obtain access. (A copy of Borchers' viewgraphs is available if requested.)

11. Dave Bader, Pacific Northwest National Laboratories, a DOE contractor, reported on ACPI. ACPI (Accelerated Climate Prediction Initiative) is part of DOE's FY 2000 Strategic Simulation Initiative (SSI), which includes computational science and enabling technologies; computing facilities; and focussed applications (climate, combustion, and basic science). There is $70 million in the DOE SSI/IT2 2000 budget, $10 million for climate research plus part of a teraflop machine. Fein pointed out that ACPI represents great potential benefit to the climate research and applications community, and it is our only hope in FY 2000 for future progress. He suggested that everyone join together in support of the initiative; there is nothing else to take its place.

The meeting adjourned at 4:30 p.m. for a CAB Executive Session. Sarachik called the CAB meeting to order again at 8:30 a.m. on Friday, 30 April 1999.

1. Sarachik led a discussion on the need for a national model. Held stated a national toolbox was needed, not a national model, to reduce duplication. Semtner stated that CSM is the de facto model but it lacks computer time for people to use. Bretherton stated that resource allocation was the problem and a common language and documentation were needed. Fein stated that researchers migrate to the best model and do not like to be told what model to use or how to develop a model. Fein also stated that NSF funds other models because they are well reviewed by the community. Ghil stated that scientists work with people not just a model and that what model is used depends on what problem you are researching. Running stated ecology models have evolved by deciding by consensus what modular components are absolutely required. Sarachik stated that for atmosphere and ocean general circulation models there is no agreement. Blackmon noted that the shift to parallel computing provides "self interest" forces for a common infrastructure. Blackmon and Sarachik also noted that sometimes a switch of models, e.g., MOM to POP, helps to create two communities of users (small computers (MOM), large computers (POP)). Bader and Zebiak stated that there are techniques for coding models that may be able to deal with the uncertain and constantly changing high-end computer architectures. Bader stated that we need model builders to partner with software engineers. Zebiak stated that a framework is needed, not just one model, and that the only incentive to consolidate efforts is a facility focussed on a few issues only. Zebiak stated that ACPI could build the facility. Jay Fein stated that a proposal on common infrastructure issues could be submitted to NSF. CSM could participate in the proposal. It was noted that a facility should have a small number of people devoted to both model development and applications (users). Anthes stated that the CSM is not only a model but also a facility for climate research that promotes healthy competition to improve models. Sarachik stated that we need to better articulate what we as a nation do in modeling, from high-end to simple. He said that CSM has evolved into a very large group of modelers. Maybe we need a "national voice" rather than a national model. Ghil stated that we need a hierarchy of models. Sarachik called for a follow-up NRC/CRC climate modeling report. Bretherton said the report should include how we evaluate the credibility of models, data assimilation, etc. Running suggested that the next report include issues from the user's perspective.

In response to the question, how will CSM interact with the CII, Blackmon said CSM has been involved from the beginning and will continue to be involved. A summary of the CAB consensus follows:

a. A national model? No

b. Common "toolbox"? Yes

c. Yes, CSM is part of a national framework and a community model but is not likely to be the only national framework or model.

d. What are our national modeling characteristics? We need a follow-up NRC study.

e. What are desirable characteristics of a facility?

1) one or more models/activities

2) computers/data/software engineering/etc.

3) standards

4) ease of use

5) connection to end users

6) expert advice

7) there should be more than one modeling center/facility.

2. A discussion followed on CSM and ACPI. Blackmon reported that CSM in partnership with LANL and LLNL will propose to ACPI to be a climate research consortium. Blackmon asked CAB for advice on how to manage the ACPI/CSM project and asked Bader and CAB for advice on how to include university participants. Sarachik asked if the Geophysical Fluid Dynamics Laboratory (GFDL) and if the International Research Institute for climate prediction (IRI) would be part of ACPI. Held said there were no plans at present. Zebiak said it would depend on whether seasonal-to-interannual prediction is included in ACPI. Bader stated that we need more than one ACPI consortium. Fein urged that ACPI be careful about funding limitations in going to more than one consortium in first round (FY 2000) competition.

3. Bader stated that DOE would get input from the climate modeling community for the SSI ACPI facility. He also stated two Announcements of Opportunity (AO) for regional climate centers and hardware would be forthcoming. An AO for research consortia will follow those. Further information on ACPI is available at http://www.epm.ornl.gov/ACPI.

4. The CAB suggested that more interactions with the Vegetation/Ecosystem Modeling and Analysis Project (VEMAP) scientists and Gordon Bonan could be beneficial in promoting interagency support of CSM.

5. The CAB suggested that the SSC include information in the CSM Plan on what could be done with additional funding and what cannot be done because there is no additional funding. Blackmon agreed to provide additional detail on the management structure and process at the June meeting.

6. The CAB will write an annual letter report, an assessment of CSM, after its meeting at the annual CSM Workshop.

7. The CAB will lend its support to ACPI.

8. Anyone interested should send suggestions about what CAB should be doing to Sarachik (sarachik@atmos.washington.edu).

9. The next meeting of the CAB will be on Friday, 25 June 1999, after the Fourth Annual CSM Workshop. The meeting will be in conjunction with the CSM SSC and CSM Working Group Co-chairs.