of the CCSM Scientific Steering Committee Meeting
2 and 3 April 2002
NCAR, Damon Room
Attendees: Jeff Kiehl (Chair), Jay Fein, Chris Bretherton, Cecilia Bitz, Ping Chang, Jim Hack, Maurice Blackmon, Danny McKenna, Peter Gent, Phil Merilees, and Lydia Shiver
Guests: Gordon Bonan, Marika Holland, Tony Craig
1. Welcome. Kiehl welcomed Phil Merilees as the new CCSM Coordinator. Merilees will start in that position 1 September 2002.
Kiehl also welcomed the new members of the SSC (Chris Bretherton and Cecilia Bitz) and thanked them for accepting to serve as a member of the SSC. He explained that the SSC oversees the CCSM project and decides on changes to each component model.
2. Overview of CCSM2 Simulations. Kiehl reported that the model control run is out to 430 years. The same physical constants for all component models were decided upon and those changes were made at 350 years. The physical constant changes had a significant effect on the atmosphere model, especially on the latent heat value. CCSM2 is running 5 years/day on ~128 PEs. The overall results are improved ocean and sea ice simulations and prognostic river runoff.
Kiehl reported that the 1% CO2 sensitivity simulation (without aerosols) was run and showed that doubling occurs at year 70, and quadrupling at year 140, and the simulation will stop at year 150.
The overall model results show that the climate sensitivity is slightly less than it was in CSM. It is believed that CCSM2 is less sensitive due to cloud feedbacks, e.g., longwave cloud forcing is getting weaker as the climate warms, clouds are brighter due to increasing shortwave cloud forcing.
Kiehl also reported that he and several NCAR scientists would visit GFDL in April to start collaborations to identify similarities and differences between the two atmospheric models and the key physical processes that contribute to the model differences.
The status of CCSM2 is that the physical algorithms were frozen in December 2001, the software engineering infrastructure changes are ongoing until May 2002, documentation is underway, and the model and control simulations will be released May 2002.
Kiehl reported that the biggest remaining issue is the simulations of the coupled model in the tropics.
3. Status of Atmospheric Simulations. Hack reported that the atmosphere model component is mostly frozen. He stated that the uncoupled configuration is undergoing final development. A new blended SST sea ice concentration data set has been completed that blends HadISST 1949-1980 record with 1979-2001 record and a control run is in progress. The improvements for CAM are in precipitable water, precipitation distribution, and eastern ocean solar energy budget. The degradations are in a colder troposphere, particularly the tropical tropopause, and a tendency for more zonal double-ITCZ structure. Many of the CAM characteristics carry over into the coupled model. The AMWG plans to begin an ensemble of ~50 year long control simulations conducted in coordination with the Climate Variability Working Group to serve as the basis for documenting simulation characteristics; to begin exploration of alternative dynamical frameworks (T31 paleo, T83 Eulerian and semi-Lagrangian, and finite-volume dynamical core); and to begin systematic investigation of specific simulation biases, e.g., eastern ocean upwelling deficiencies.
4. Status of the Ocean Model Simulations. Gent reported that improvements to the ocean model component are that the Arctic is much improved because of no Fourier filtering and adding river runoff from Russian rivers producing a realistic halocline; that the Pacific equatorial undercurrent is close to observations in CCSM2, unlike the very weak current in CSM1 (this is one factor in improving the ENSO variability in the CCSM2 compared to the CSM1); and that the transports in the Antarctic Circumpolar Current and between Florida and Cuba are now much closer to the observed values.
5. Status of Land Model Simulations. Bonan reported that there is a warm bias in the land model component but it is only +/-2 degrees, which is fairly good. He said the warm winter bias problem in the Northern Hemisphere is one of the top priorities for the Land Model Working Group members to solve. Another problem the group is working to solve is that the runoff comes in one month too early. Another priority this group is working on is the dryness of tropical America. Bonan stated that the annual cycle is well represented in the midlatitudes.
Future plans for this working group are to put carbon into the CCSM and to do carbon cycle simulations and define how to represent subgrid heteorgenity and plant functional types. Peter Thornton is presently working on changing the data structure of the model, and the Land Model Working Group will bring this recommended change to the SSC in June.
6. Status of Sea-Ice Model Simulations. Holland reported that as compared to observations, these simulations show low ice area in the summer and high ice area in the winter for the Northern Hemisphere. The ice is generally too thin in the Arctic. In the Southern Ocean, the ice is too extensive due to excessive area in the Atlantic sector and is generally too thick.
7. Status of CCSM2 Model Code and Development Process. Craig reported there would be a CCSM2 Release Home Page and all documentation (user's guides, scientific documentation, reference guides) would be located here. He stated that production runs are ongoing and the release date is scheduled for 17 May 2002. On 100 processors of the IBM, the coupled model runs about 4 simulated years per day and requires about a month to run one century.
Craig reported on the DOE SciDac project and the NASA ESMF project that are providing 3 new software engineers to leverage work on CCSM.
The software engineering "issues" are model development, infrastructure, modern software engineering practices, performance, portability and hardware issues, modern coding practices (F90+, C, C++), and RISC versus vector. The overall goals of the CCSM Software Engineering Group (CSEG) are to improve quality; to improve coordination, communication, and control; to improve productivity; to continue to allow "individual" development; to recognize that CCSM is a community project; and to minimize overhead and burden of "process."
Craig proposes that Change Review Boards (CRB) be created to improve communication and coordination of both scientific and software tasks, to better control production versions, to better prioritize development tasks, to improve quality by implementing reviews, and to formally distinguish tasks that are under development and tasks that are ready for production. Each model component and the coupled model will have a CRB that will keep track of tasks under development; submit change requests; review model changes; ask community for input on change requests, working closely with working groups and the SSC; review for "improvement," performance, testing status, coding quality, and documentation; and accept, reject, defer, or return request at the end of the process. To help keep track of tasks and changes, a status accounting tool will be chosen or developed, and the repository policy has been changed to create consistency between the repository and the CRBs.
Craig also stated that the goals for repository access are to open the CCSM repository to more developers; formalize repository use policy and procedures; improve coordination and control of development efforts; clearly separate development tasks in the repository from production versions of the models; have all development occur on branches; and place some limitations on individual access and have an ability to enforce limitations. The process for repository access is to continue to use CVS, developers formally request access, all development will take place on CVS branches, developers will provide regular progress updates, and CCSM will have tools to monitor usage and retract access if necessary.
A CCSM2 script tutorial will be convened at NCAR the Friday after the CCSM workshop to help the community use the new model.
8. Discussion on Climate Variability Working Group. Hurrell and Alexander showed a draft agenda for this year's CVWG meeting and reported that they have invited speakers to give presentations on aspects of the CCSM2 control simulations.
The SSC, along with Hurrell and Alexander, discussed plans for a topical session on coupled model simulations of the tropics at this year's workshop. Gent will give a 15 minute talk on fully coupled sensitivity studies that might shed light on sources of biases in the CCSM2, someone will give a 15 minute talk on atmosphere alone sensitivity studies, and Large will give a 15 minute presentation on ocean alone sensitivity studies. Bretherton said that CLIVAR might be interested in hosting a workshop on this subject to focus on ways to solve the problem. The planning committee (Chang, Bretherton, Hurrell, and Alexander) will update the SSC on plans for a workshop at the June SSC meeting.
9. Seventh Annual CCSM Workshop. The open forum will be the same as last year with SSC and Working Group Co-chairs answering questions from the workshop participants. A formal poster session will take place on Tuesday evening during the reception. Shiver outlined a schedule for workshop funding requests to be decided, and an ad hoc committee will rank CCSM workshop funding requests after the working group co-chairs give their recommendations. Craig discussed plans for the script tutorial after the workshop that will be held at NCAR, and it was requested that feedback be obtained in 3 to 6 months from participants about the tutorial. Bader will be invited to the SSC meeting, and Sarachik and Held will be the CAB members joining the SSC meeting at the workshop. Fein stated that he would be glad to announce that travel funding to the CCSM workshop would be an appropriate request on NSF proposals/grants.
10. Eighth Annual CCSM Workshop. Shiver and the SSC discussed dates for the next two workshops, since Fein and others had requested that the CCSM Workshop be moved in 2004 due to an international CLIVAR meeting being scheduled the 4th week of June. The dates for the workshop in 2003 are 24 to 26 June and the dates of the workshop in 2004 are 29 June to 1 July 2004.
Shiver gave a presentation on site visits to several hotels and conference centers where future CCSM workshops could be held. The SSC voted to continue to hold the workshop in Breckenridge, Colorado.
11. Rotation of Working Group Co-Chairs. The SSC discussed working group co-chairs.
12. Membership of SSC, CAB, and Working Groups. The SSC discussed these memberships.
13. CCSM Computational Resource Issues. Kiehl announced that the CSL proposal for CCSM will be in 3 documents: control/production simulations; development of component and coupled models; and report on usage of CSL allocation over the last two years. Kiehl stated that a list of potential experiments using CCSM was collected from the community, and if all the experiments were done, it would take 75% of the total CSL. It was decided that the IPCC-type runs would be factored into the CCAWG portion of the proposal.
The SSC discussed whether CCSM should include a vector architecture option and maybe have both vector and massively parallel codes for CCSM. It was stated that it would be very costly to have software engineers and scientists working to redo the code for vector architecture and to keep both vector and massively parallel codes up to date simultaneously.
14. CCSM Code Copyright Issue. Catherine Shea, UCAR lawyer, attended the SSC meeting to gather information and present ideas about using a copyright or a general purpose license for the release of CCSM2. Hack stated that CCSM2 is a community effort and not UCAR property. It was decided that a notice and disclaimer statement drafted by Shea could be used. Shiver will check on the University of California copyright issue with LANL and forward all information to Shea. (Note: The notice and disclaimer statement was completed on 13 May 2002 by Shea and distributed to the SSC and Working Group Co-chairs for comment.)
15. CCSM/DOE Data Policy Issues. The SSC agreed that all CCSM downloadable data and code should be accessible after each user identifies him/herself, their affiliation, and their purpose for using the code. DOE has this structure in place already. Kiehl will estimate the data storage need for CCSM and communicate the need to SCD. The SSC will discuss how non-NCAR researchers will be able to obtain access to the CCSM data on NCAR computers at their June meeting.
16. CCSM and U.S. CLIVAR Issues. Bretherton reported that the U.S. CLIVAR SSC wanted to make CLIVAR more connected to climate modeling, so they have begun to organize Climate Process Teams (CPTs). There is some funding from NSF that will be available in FY03, so CLIVAR is now trying to choose topics and put teams together. The idea is for an organized and focused research effort designed to improve serious problems in current coupled climate model simulations. The effort must involve both observational/process researchers and climate modelers from, e.g., CCSM, GFDL. Some suggested topics to form teams around are small-scale ocean mixing processes, deep convection, and atmospheric boundary layer. Other players are GFDL and NCEP. Bretherton asked that comments and suggestions be sent to him or David Legler for one atmosphere and one ocean problem that could be agreed upon, so that GFDL, NCAR, NCEP, and CLIVAR collaborators can form a CPT and work collaboratively on the problem(s). Kiehl will take the lead to generate a list of topics, NCAR participants, and estimate of resources needed. Fein stated that NSF has approximately $1 million to fund CPTs in FY03. NSF OCE and NOAA OGP may also have funds available for CPTs. An announcement of opportunity would probably come out late summer and would be informal.
17. Teleconferencing Weekly NCAR Scientists Meetings. Shiver reported there was not much interest from SSC and Working Group Co-chairs to be involved in this weekly meeting. Conference calls can be set up and are an option. Another option is sending the URLs and data information to the SSC and Working Group Co-chairs weekly or biweekly.