Use of field data to diagnose land-surface interaction.
The land-surface interaction is as important to the climate of a global model as the sea surface boundary condition. However over land, there is no measured input field analogous to the sea surface temperature, which controls the surface sensible and latent heat fluxes (together with surface wind, air temperature and humidity). The fluxes over land in contrast are driven on diurnal time scales by the net radiation, and the partition into latent and sensible heat, which depends on the availability of water (either in the soil or in surface reservoirs) for evaporation. The accuracy of the down-welling fluxes depends on the model radiation physics, and the specification of clouds and aerosols. The outgoing fluxes depend on the calculated surface radiation or skin temperature, and the albedo, which in turn depends on snow cover, vegetation type and season (for example, leaf-out). Thus even the calculation of net radiation at the surface involves many physical parameterizations. The availability of water for evaporation depends on both a realistic hydrological model (balancing model precipitation, surface and soil water reservoirs and runoff) and a realistic vegetation model (to extract soil water for transpiration as a function of photosynthetic processes). At higher latitudes, soil water only becomes available after the ground melts, so the soil thermal balance processes, and the timing of snow melt (snow insulates the ground) also control the seasonal cycle of transpiration. However global fields for soil water and soil temperature are not available for analysis, so a model must derive them from its own physical parameterizations, using near surface atmospheric measurements as constraints (e.g. Douville et al., 1999).
However the coupling between surface fluxes and the convective boundary layer is so tight, that errors in any of the physical parameterizations, whether sub-surface hydrology or thermal transfers, vegetation parameterizations, or stable/unstable boundary layer parameterizations can all interact to give an erroneous diurnal cycle of the mixed layer, and in turn an erroneous diurnal cycle of convective precipitation in a model. This paper will address efforts to use data from both field programs and on the scale of river basins to assess errors in the model formulation of the land surface interaction.
Betts, A.K., P. Viterbo, A.C.M. Beljaars and B. Van den Hurk, 1999: Use of field data to diagnose land-surface interaction. Proc. ECMWF Semina on Diagnosis of Models and Data Assimilation systems. Sept. 6-10, 1999, Reading, United Kingdom, 347-364. http://www.ecmwf.int/publications/library/ecpublications/pdf/seminar/1999/sem1999betts.pdf