Version 1 of the climate system model (CSM) is to be released to the research community next May. It is the first community-based model to link atmospheric, oceanic, biologic, cryogenic (ice-based), and chemical components. Investigators at UCAR institutions will be able to run the kind of sophisticated simulations of global change that heretofore have been limited to a handful of research centers.
In order to make the May debut, the CSM developers need to freeze the model components by the end of this year and then undertake several months of testing. That means the pressure is on for dozens of staff in CGD and the Atmospheric Chemistry Division. Meanwhile, the Scientific Computing Division is assembling a new hardware configuration--the Climate Systems Laboratory--that will be one of the world's most extensive for global climate modeling. (See sidebar for details on the CSL.)
"A lot of people here became convinced [in the early 1990s] that they needed coupled models to continue progressing in their work," says Byron. "This was happening all over the atmospheric science community." CGD's Warren Washington and Jerry Meehl developed a coupled ocean-atmosphere model in the mid-1980s that they've used with success in analyzing enhanced greenhouse effects and other complex phenomena. But Warren and Jerry lacked the base of documentation and computing support needed to offer their model to the world at large.
"NSF was beginning to see that a lot of institutions needed coupled models and weren't going to get them unless they built them themselves," says Byron. Going solo might have been feasible for modeling the atmosphere alone, but few institutions could handle the demands of creating and maintaining a fully coupled model.
As a central site for a new community model, NCAR made sense. The community climate model (CCM), developed here in the early 1980s and bolstered by SCD and CGD support, has been a standard for university users. "As a community tool," says Byron, "the CSM is really an outgrowth of the CCM, but is much broader in scope."
That breadth is partly due to the enormous progress in ocean-model development over the past decade. "NCAR has invested a lot in ocean modeling and its models have become important tools for the university community," says Bill. "This makes the issue of coupling these complex pieces a natural one for NCAR."
The new modeling effort ended up finding a home within the NSF's Climate Modeling, Analysis and Prediction (CMAP) program. "Jay Fein [at NSF] was extremely helpful in getting this thing going," says Byron. In late 1993, NSF approved the two-year proposal submitted by NCAR to develop a first cut at the CSM. The project has been overseen by a steering committee that includes the NCAR director, associate director, and directors of the four science divisions. Guidance and support has also come from the CMAP scientific advisory committee. Hands-on development is being led by a group of internal and external investigators headed by Byron and Bill.
"There were a lot of acrimonious debates," admits Byron in recalling the first stages of crafting the CSM proposal during 1993. "That's one of the reasons I ended up as a cochair--I was someone that everybody didn't hate for the moment. Also, the other logical people were heads of groups already."
One of the key challenges has been finding and nurturing the appropriate submodels that feed into the CSM. "In the past, coupled models had been undertaken by a small group of people using existing submodels that they may or may not have built themselves. Here, the scientists who built the component models are actively involved in the coupling process. Instead of one or two scientists providing input, you've got 20."
The four pieces of the CSM include
The fifth and final piece is unique: a coupler that will allow the four submodels to communicate with one other with a minimum of strings attached. For example, the ocean and atmosphere models can use different resolutions yet still be coupled without problems. "The idea is that you can replace any part easily and the coupler can deal with it," says Byron. "As far as we can tell, the coupler is an entirely original concept."
Over the past year or two, the CSM push has become a large-scale effort within NCAR. Almost all of CGD's scientists and software developers are in on the project to some degree, along with such collaborators as Guy Brasseur and David Erickson in ACD and Ray Roble in HAO.
The pace of internal preparation will continue to be furious up to the freeze date of 1 January 1996. In December, the CMAP advisory committee meets at the Mesa Lab. "These outside scientists will give us some advice," says CGD director Maurice Blackmon. "After we freeze the first version of the model, we'll run a lengthy simulation--perhaps 100 years. The results of that run will be presented next May at a workshop that will involve university collaborators." At that point, the model should be ready for release.
Once the pressure of getting version 1 out the door eases, the CSM developers will be able to work on refinements. "The really new pieces--the ability to couple in chemical and biogeochemical cycles--won't be in the first generation," says Byron. Some of those strands will come from the GENESIS model (Global Environmental and Ecological Simulation of Interactive Systems). This project, housed in CGD and supported mainly by the Environmental Protection Agency, has worked on biological and chemical modeling angles for several years under project leader Starley Thompson. "Some of the elements in GENESIS are ambitious and quite extensive," says Maurice, "so the experience with GENESIS will be quite valuable for the CSM."
As the model enters widespread use, the interchange between NCAR and university scientists will step up. That's where UCAR's Climate System Modeling Program comes in. Headed by Dave Schimel, this group was founded in 1990 to explore the potential of coupled models. It was folded into CMAP last year. "At that time," says Dave, "our role changed from that of a stand-alone program to one that's really closely coordinated with CSM."
The new CSMP has three charges:
It's gratifying for Dave to see the climate system model coming to fruition. Back in 1991, CSMP helped put together two workshops--one in Madison, Wisconsin, and the other in North Ryde, Australia--based on a new textbook (the first one specifically for climate system modelers) edited by CGD's Kevin Trenberth. Dave sees those workshops as a watershed for the new discipline.
"If you could point to the greatest thing UCAR's done in climate system modeling, I think it was the Madison workshop and the book. It led to a much broader understanding of climate systems science among a cohort of 50 or 60 atmospheric scientists. They really are the cream of the next generation."
CSMP has sponsored ten postdoctoral students over the past few years and brought dozens of others to various follow-on workshops. "The nice thing is that we're starting to see the results of our early efforts appearing as publications," says Dave. "We're seeing the first generation of students and postdocs trained by us now becoming influential researchers in their own right." Among them, Jay Famiglietti (University of Texas at Austin) is working on a river- hydrology submodel that conserves fresh water. Keith Hines (Byrd Polar Research Center, Ohio State University) is modeling physical and chemical aspects of the seasonal Antarctic vortex. And Xubin Zeng (University of Arizona) is using what Dave calls "a really revolutionary approach" that models turbulent energy by breaking it into mesoscale and larger-scale components.
Whatever their perspectives, the principals behind the CSM agree that it is a quintessential NCAR activity. "We are here to work with university scientists as well as perform our own studies," summarizes Maurice. "We are always unique in that sense." --BH