The Common Land Model (CLM) is a project that is multi-disciplinary and multi-institutional and whose intent is to develop a state of the art land surface model for use in climate studies. It requires broader expertise than found in any one research group or institution. This report documents the initial prototype version of this model.
The CLM derives from the premise that if a land model could be constructed from components used in most land models and in a way acceptable to the community, then individual groups that are interested in land modeling but lacking the necessary resources or expertise, could focus on the new aspects without excessive repetition of past efforts. Furthermore, all the users of such a CLM, in the spirit of ‘open source’ code development, could then share the improvements and refinements provided by individual groups.
The CLM was originally proposed to provide a framework for a truly community-developed land component of the NCAR CSM at a workshop of the NCAR CSM Land Working Group in February 1996. An opportunity to advance this concept was provided by a visit to the US of Yongjiu Dai from the Beijing / China (Institute of Atmospheric Physics, Chinese Academy of Sciences), hosted by Xubin Zeng (University of Arizona) and supported with funding from a NASA /EOS Interdisciplinary Science Project. Initial specifications of the software to be developed were given by Robert Dickinson, who suggested that the goal of a truly common land model was unlikely to be achieved within existing institutions but that it would be useful to provide a prototype that could demonstrate at least some groups were motivated and capable of joining forces to produce a shared model.
One important starting point was Gordon Bonan’s (NCAR) well-documented LSM land code, currently in use as part of the NCAR CSM climate model. Dai had already demonstrated extensive familiarity with the IAP LSM (version 1994, IAP94) code (his PhD), as well as the BATS code that had been the starting point of his IAP94 code. Thus, it was decided that Dai, with sufficient help, could effectively implement a state of the art model that combined best features of the LSM, BATS, and IAP94 codes. Hence, early in 1998, Dai, Dickinson, and Zeng reviewed the content of these codes and developed initial software specifications. It was also seen as highly desirable that the resulting model not only be targeted as a next generation model for the land component of the NCAR CSM but also as a candidate for next generation used by other US groups requiring extensive revisions of their current land surface models.
Various conversations indicated interest in potential application for the needs of the GSFC DAO modeling structure, which was initially implementing Koster’s MOSAIC model, and COLA scientists tasked with revision of their SSiB model. Ties to groups concerned with carbon cycle and ecological modeling were also desired. Initial specifications of the design were reviewed and commented on by an ad hoc CLM scientific steering committee, chaired by CSM LWG chairs (Dickinson and Bonan), and with initial membership besides Dai and Zeng, was Paul Dirmeyer (Center for Ocean-Land-Atmosphere Studies), Jay Famiglietti (University of Texas-Austin), Jon Foley (University of Wisconsin), and Paul Houser (GSFC). Dirmeyer and Houser provided especially substantive guidance as to how the code could be improved in ways that would better meet their institutional requirements. Zeng took on the role of coordinating this community input and in working with Dai to oversee its implementation. The improved software implementation plan consequently developed was discussed at the June 1998 annual CSM Workshop and finalized by Dai and Zeng. Dai subsequently constructed an initial code that was scrutinized for 3 days in March 1999 by an outside review committee consisting of Bosilovich, (GSFC), Dirmeyer, and Houser.
At this point an extensive period of “beta-testing” of this code was initiated and lasted for more than a year. At the University of Arizona, Zong-Liang Yang joined Dai, Dickinson, and Zeng in this effort, whereas the participation of Bonan at NCAR was strengthened by Keith Oleson. COLA provided Schlosser to join Dirmeyer’s participation, and GSFC/DAO’s participation continued through Houser and Bosilovich. Denning and Baker of Colorado State University also joined this model testing as part of their effort to establish a next generation SiB model. Initial results from this testing were presented during the CLM workshop at COLA in November 1999. Site data used for validation included all the PILPS sites and others, in particular that of Cabauw, Valdai, FIFE, ARME, ABRACOS, HAPEX, BOREAS, and WLEF, the regional data of the Red-Arkansas basin, and the Global Soil Wetness Project (GSWP) data. Validation of applications requires actually testing in those applications. Initial such applications have included use in the multi-agency Land Data Assimilation project (LDA), and coupling to CCM3.
The code design specification has emphasized various aspects of modularity. Three logically separate elements were defined: a) the core single-point model process code, b) the boundary condition data, and c) a driver that interfaces the land model to the atmosphere (including the scaling procedures within a climate model required to interface atmospheric model grid-square inputs to land single-point processes). Such separation allows the best science to be used for each of these elements, and in particular, to insure that a) the core model can be tested with single-point field data, b) the latest boundary condition data sets as derived from NASA satellite data can be incorporated when available, and c) the latest concepts as to scaling procedures can be adopted. This report primarily documents the point model treatment. Some advances in data sets and scaling are also described here, but these questions require much further effort. The code has been developed in FORTRAN, currently FORTRAN 90, to be most compatible with existing codes and scientist’s programming expertise. Houser developed an interface (coupler) routine to isolate the land code from the requirements of needing model data structures.
Some aspects of the CLM are relatively complex as required to be general enough to satisfy a wide variety of anticipated applications. Managing such complexity is not easy. However, we anticipate that good documentation and the open scrutiny of many scientists will eliminate any serious errors. More limited applications may require more simplified treatments. For example, the multi-layer soil and snow structure is needed to provide accurate simulation over a wide variety of time scales, and hence to be useful for such disparate applications as model data assimilation of surface properties, or correctly determining soil temperatures beneath snow for matching measurements of soil respiration. The suggestion has been made that a variety of simpler models also be provided. In particular, a much simpler slab treatment could be used for the limited requirement of providing sensible and latent fluxes to the atmosphere. However, our limited resources have not yet permitted taking on such an effort. Furthermore, there is considerable interest in further improvements, and likely increased complexity, for many aspects of the current code. These include the parameterization of runoff, and better integration into models of vegetation dynamics and soil biogeochemistry.