92nd American Meteorological Society Annual Meeting (January 22-26, 2012)

Monday, 23 January 2012: 4:15 PM
Usefulness of Benchmarking for Global Land Surface Model Development
Room 352 (New Orleans Convention Center )
Gianpaolo Balsamo, ECMWF, Reading, United Kingdom; and C. Albergel, M. Balzarolo, A. Beljaars, S. Boussetta, J. C. Calvet, E. Dutra, T. Kral, D. Papale, P. de Rosnay, and I. Sandu

Land surface model development for global Numerical Weather and Climate prediction applications is confronted to the lack of observability of some components (e.g. deep-soil conditions, 3-dimensional aspects of vegetation above and below the ground, etc.) for which remote sensing cannot be directly informative and in-situ observations are too sparse. Therefore, the level of sophistication needed for representing land surface boundary and ancillary conditions in models (e.g. in the root-zone or at the bottom of the resolved soil layers) is still subject to debate when it comes to global applications. The land surface benchmarking is an appealing context in which one could evaluate the need of increasing complexity of a given parametrization. For instance, during winter months it can be shown that momentum flux, snow sublimation/melting/accumulation and runoff are all interlinked. In summer months, rainfall, evaporation, soil moisture and runoff are also coupled. However in both seasons improvements in one of the aspects (e.g. related to a flux or to a land reservoir) does not necessarily propagate to the other physically-linked components. A land surface benchmarking in which fluxes and reservoirs (of both water and energy) can be compared with direct observations may therefore provide insights at process-level, allowing to evaluate the co-variant aspects of model changes, and helping to identify model deficiencies. Results will be shown from the recent land surface model developments at ECMWF, evaluated using a land surface benchmarking database gathered for this purpose. This is part of an effort to establish a process-oriented evaluation of different model versions against a range of diverse and independent observational datasets.

Supplementary URL: