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

Minutes of the CCSM Scientific Steering Committee Meeting
27 and 28 November 2001
NCAR, Damon Room

 
Attendees:  Jeff Kiehl (Chair), Dave Randall, Ping Chang, Byron Boville, Bob Dickinson, Peter Gent, Maurice Blackmon, Dick Moritz, Danny McKenna, Jay Fein, Anjuli Bamzai, and Lydia Shiver

1.  2002 Annual CCSM Workshop.  We will continue with parallel sessions of working group meetings, combine the poster session and the reception together, continue with time for a forum, and add one-half day on Friday at NCAR for a tutorial on model run scripts.

The working groups will be encouraged to have a mix of relevant presentations and time for discussion and not to hold AGU-like meetings with many presentations.  The Climate Variability Working Group (CVWG) will be encouraged to focus on ENSO, especially if one of the plenary speakers discusses this topic.  Kiehl will discuss this with the CVWG co-chairs, Hurrell and Alexander.

Shiver presented the results of the CCSM Tutorial survey.  The results showed that we should conduct a one-week CCSM Tutorial Workshop on the fully coupled model with professors, graduate students, and post-docs attending.  30% of the time should be focused on hands-on training and 70% should be on lectures on the structure of the model code, model scientific algorithms, methods of appropriately designing experiments with climate models, and data analysis and visualization tools (not on application of the model to research problems).

The vote was split evenly between every year and every two years on how often the tutorial workshop should occur.  A review committee should be established to evaluate applicants, the number of attendees should be limited to less than 30, and the material presented should be at the late graduate level.  Four people said the workshop should be held at NCAR, and the week before or week after the annual CCSM workshop was suggested.  The lecture materials should be available online.  No exam should be given to attendees, but they should evaluate the form and content of the workshop.  A periodic evaluation should be performed to determine how many workshop attendees use the coupled model.

The SSC recommended that the week long tutorial be conducted in winter or late fall, and the first workshop would be a pilot tutorial where attendees are recommended by their university as a representative.  Instead of an exam it was suggested that each attendee present results of an experiment based on what was learned during the tutorial.  Kiehl will contact the following universities for one representative:  University of Washington, Texas A&M, Duke, University of Massachusetts, Colorado State University, Georgia Tech, University of California at Berkeley, UCLA, Scripps, University of North Carolina, University of California at Santa Cruz, University of Illinois, University of California at Irvine, Penn State, University of Wisconsin at Madison, University of Colorado at Boulder, MIT, Princeton, and Lamont-Doherty.  Kiehl will prepare a draft syllabus for comment by the SSC members.

More criteria for workshop attendee funding will be implemented in the application form this year, such as a formal deadline after which funding decisions will be made.  An ad hoc selection committee will be appointed by Kiehl consisting of NCAR and non-NCAR SSC members.  The funding should be spread around to as many institutions as possible by providing a standard package of hotel, per diem, and shuttle costs.  The funding should be spent on researchers who have either worked with CCSM in the past or who have made a clear commitment to work with CCSM in the future.  Some funds should be reserved for graduate students and post-docs.  The working group co-chairs will be asked for their input before the selection committee meets.

Nominations for the 2002 Distinguished Achievement Award are due 15 December 2001.  Shiver will forward all nominations to the SSC for a vote after 15 December 2001.

2. Update on CCSM2.  Hack presented an update on the CCSM Atmosphere Model (CAM) and showed some results from the first coupled simulations of the CCSM2 using CAM.  The basic message was that most of the systematic biases that were identified in the uncoupled atmospheric model were also present in the coupled configuration.  Some of these biases are further amplified, such as the tendency to produce a double ITCZ.  One of the more puzzling features of the coupled simulation are the warm SST anomalies present in the eastern oceans, despite a sizable decrease in the net energy input to the system in these regions.  Preliminary sensitivity experiments point to inadequate mechanical forcing of upwelling in these regions, possibly related to anomalies in the representation of surface orography (e.g., the extension of the Andes orography too far to the west).

Final enhancements to the uncoupled configuration of the CAM are to apply fractional land and fractional sea-ice formulations, to add the CCSM sea-ice thermodynamics scheme, to complete a slab ocean model option, and to define the final set of boundary forcing datasets (e.g., SST and sea ice concentrations boundary datasets).  Further exploration of alternative dynamical frameworks, such as a T31 paleo configuration, T85 Eulerian and semi-Lagrangian configurations, and a finite-volume dynamical core, will continue.

Gent presented an update on the CCSM Ocean Model.  He reported that after 40 years the ocean component of the fully coupled CCSM2 run has the following improvements over the CSM1 coupled integrations.  The global heat content is much better, and the salinity distribution is improved. The latter is due to the quite realistic river runoff that is being given to the ocean. The full model conserves fresh water quite well. The Arctic Ocean is much improved because there is no numerical filtering, and the Bering Strait is open. The Arctic river input means that the halocline at the surface of the ocean is being realistically maintained; it was nonexistent in CSM1. The equatorial Pacific currents are now very realistic, and quite good in the Atlantic; they were both much too weak in the CSM1 integrations.

The SST errors in the equatorial regions and in coastal upwelling regions remain and are larger in magnitude than in CSM1. This is probably due to the higher resolution of the ocean component. There are also large errors in the central, far North Atlantic and in the Pacific sector of the Southern Ocean. In conjunction with these SST errors, the sea-ice is too extensive in the Labrador Sea region and is not extensive enough in the Pacific sector of the Southern Ocean.

Bonan presented an update on the CCSM Land Model.  He reported that previous simulations of the Community Land Model (CLM2) coupled to CCM3 showed that the new land model warms surface air temperature year-round, eliminating a pronounced Northern Hemisphere summer cold bias, improves the simulation of snow cover and depth, and improves the simulation of runoff. Simulations using CCSM also show these same basic features. The large winter Northern Hemisphere warm bias remains, as in other versions of the atmospheric model. However, the summer cold bias is eliminated. Prominent precipitation biases include a dry Amazon and wet central Africa. Regional analyses of monthly surface air temperature, precipitation, and runoff show good agreement with observations.

Holland presented an update on the CCSM Sea Ice Model.  She reported that the physics in the sea ice model component of the CCSM is frozen.  Documentation describing this physics is almost complete.  Additionally, a user's guide for the ice model is being written and should be available with the release of the CCSM2.

Analysis of the polar climate in the first forty years of the coupled model run was presented.  In the Northern Hemisphere, the time series of sea ice area and volume show considerable trends over the first 20 years.  By year 40, these trends appear to have diminished.  In general, the ice area annual cycle is enhanced compared to observations with relatively large wintertime areas and relatively small summertime areas.  However, compared to CSM1, the Northern Hemisphere ice areas are considerably improved and the maps of ice extent look good. The Northern Hemisphere ice thickness has some similar errors to those seen in the CSM1 integrations.  In particular, the ice thickness distribution across the Arctic is unrealistic, with the thickest ice present in the East Siberian Sea.  This is related to a bias in the simulated surface winds over the Arctic.  However, the wind forcing in the Fram Strait region looks reasonable and leads to a reasonable export of ice from the Arctic into the northern North Atlantic.  Additionally, the freshwater fluxes due to river runoff and precipitation over the Arctic basin look reasonable and allow for fresh shelf waters within the basin.

The Southern Hemisphere time series of ice volume and snow volume show large trends over the first five years of integration.  They then level out at unrealistically large values.  The ice area is reasonably simulated but is somewhat large compared to the observations.  This results from too much ice present in the Atlantic sector of the Southern Ocean and too little ice present in the Pacific sector.  The ice velocities in the Southern Hemisphere compare well to the observations.  However, excessive meridional ice volume transport is present due to the relatively thick ice cover that occurs.  This is consistent with a fresh bias in the sea surface salinity at the ice edge where melting occurs.

In general, many aspects of the polar climate simulation look promising.  However, there are still significant problems that need to be addressed as research issues.  Although not discussed at any length in the presentation, it also appears that there is some interesting variability present in the polar sea ice conditions that will require further analysis. 

Buja presented an update on the software engineering aspects of the CCSM coupled model.  He reported that the CCSM Software Engineering Group (CSEG) is working on porting and performance, documentation and training, and state of the art software engineering practices for the coupled CCSM.  The CCSM is currently running and being tested on the NCAR IBM SP, NCAR SGI Origin 2000, NCAR Compaq ES-40, Los Alamos National Lab's SGI Origin 2000, and NERSC IBM SP.  The performance at NERSC for the coupled model is 53 seconds per day for 1 year or approximately 130 years per month.  The CSEG is working on a User and Developer's Guide for each component, a Quick Start and User Guide for CCSM, internal interface documentation, and a growing library of CCSM software engineering documents that are online at http://www.cesm.ucar.edu/working_groups/Software and click on CCSM Software Engineering Developer's Guide.  The Software Engineering Coordination Plan is also available at the same website, and it contains information on the Change Review Board, repository, testing, status accounting, documents, planning and prioritization, management, and other issues.

The SSC discussed the need for all model components to have a web-based diagnostics page, and Kiehl will communicate this need to the working group co-chairs.  Mark Stevens has offered to help the working group co-chairs post their component's diagnostics page.

3. Participation in CCSM Development.  Moritz discussed his views on improvements that could be made in the CCSM project.  He stated that it is amazing how far the CCSM project has come and that the website needs to continue and be improved.  He said that hallmarks, such as community involvement in model development, no flux adjustments, etc., should be identified and broadcast to the community to gain credibility, visibility, and base of users for CCSM.  He stated that it is important that the SSC not dictate science specifics, but that the SSC must set priorities and make decisions to lead the project.  A major goal pointed out was that communications need to be designed to allow participation and not just inform and update, especially in the coupled model results arena.  Video conferencing of the weekly scientific meetings was suggested, and Shiver will find out the details of NCAR's capabilities.  Another suggestion was to continue the weekly report from the scientific meeting but post it on the web.  The SSC members should find a way to assess resource needs and address them, such as support for non-NCAR scientists.  It was suggested we be realistic about projections of model release dates so the community will have a good idea of when to expect the next version of the model.  Another area that needs to be addressed is reward and recognition of participants.  Kiehl will ask the working group co-chairs if they know of any ways to reward significant contributions.

Moritz proposed that the SSC request proposals from interested scientists for workshops and working group meetings focused on a scientific problem.  He suggested that the first such topical workshop focus on Tropical Pacific Seasonal-to-Interannual Variability.  Ping Chang said he would follow up by coordinating a proposal for such a workshop.  Moritz also suggested that CCSM should identify resources and establish procedures for providing support to select, non-NCAR investigators that are addressing high priority CCSM problems.  A current example is the work by Chris Bretherton on a new CCSM atmospheric boundary layer formulation, which could be greatly accelerated with additional person-power at NCAR to work with Bretherton.  It was suggested to help the SSC and non-NCAR scientists understand what the NSF-funded model component liaisons do is to list their duties on a webpage, along with the software engineers who support the project, and have them give presentations to the SSC about their work.

4. NCEP Collaborations.  Fein and Tribbia updated the SSC on discussions with NCEP on collaborations to analyze the growth of the systematic errors as forecasts proceed and to explore the utility of an isentropic coordinate in the CCSM atmospheric model.  NCEP staff will visit Fein at NSF on 10 December 2001, and Fein will ask for clarification of NCEP's plans for collaborating with NCAR.

5.  DOE Collaborations.  Kiehl reported that DOE is eager to coordinate activities with CCSM and does not want to be in conflict or competition.  Dave Bader will be stepping down from the program manager position, so if anyone knows someone who would be interested in the position, they should encourage him or her to apply.

6.  NASA Collaborations.  DeLuca gave a presentation on the status of the Earth System Modeling Framework (ESMF) project that SCD and CGD (Boville) are collaborating with NASA on.  The broad collaboration is developing a model coupling superstructure and utility infrastructure over a three year period.  CCSM is one of 15 testbed applications that will utilize ESMF tools.  ESMF is developing time management, timing/profiling and logging, error handling, input/output tools, and parameter specification utilities.  It will have an object-based design, a standard software development cycle, and multiple language interfaces, such as C and Fortran 90.  The time management and date and time utilities have already been integrated in the CAM and the ESMF library is linked in to CAM.  The current focus of ESMF is to prepare the code, documentation, and website for distribution of the CCSM.  The benefits of the ESMF to CCSM are that more extensive and better supported infrastructure will enable easier model development and give a buffer against complexity and changes in computing platforms, so introduction of new codes will be easier.

Boville reported that collaborations with the Data Assimilation Office (DAO) are occurring at an irregular frequency and that the finite volume dynamical core is running as well as other core schemes tested.  Boville pointed out that we need to run the finite volume model at NCAR routinely, since NCAR physics and DAO physics packages are different.  When the CCSM is released, the atmosphere component will have three dynamical cores to choose from, including the finite volume dynamical core.

7. CCSM Computing.  The CCSM is running on the NCAR IBM SP, NCAR SGI Origin 2000, NCAR Compaq ES-40, LANL SGI Origin 2000, and NERSC IBM SP.  Fein and Jacobs will work with the CSL Allocation Review Panel regarding stabilizing the allocation for CCSM.  Kiehl will send Fein information on what percentage of the CSL is needed to give adequate capability to move forward in the future.  Kiehl will also send Fein information about the potential needs for CCSM to complete simulations for the IPCC.

8.  New SSC Members and New AMWG Co-Chair.  Kiehl announced that Jim Hack, Chris Bretherton, and Cecilia Bitz will be new members of the SSC as of 1 January 2002.  He also announced that William Collins will be one of the co-chairs of the Atmosphere Model Working Group.

9.  Response to CAB Letter.  A few minor suggestions were made for the response to the CAB letter.  Shiver will prepare the letter for signature and distribution.

10. Eos Article Status.  Dickinson will prepare a draft article for the SSC to comment upon in January.

11.  CCSM Approved and Freeze/Release Dates.  The 4 new component models (atmosphere, ocean, land, and sea ice) were formally approved by the SSC as the components to be used in the next version of the CCSM to be released in May 2002.  The model will be frozen as soon as the control run shows no huge problems, probably the first week of December 2001.  All model documentation must be ready when the model is released, so Shiver will ask the working group co-chairs what documentation they will have for their component.  Also, Shiver will gather information from each working group on peer-reviewed papers that are in process and who the authors will be.