Validation of the boundary layer representation in the ECMWF model.
In this paper we concentrate on the boundary layer budgets of heat and moisture over land. The boundary layer (BL) is the region between the surface and the free atmosphere where vertical diffusion due to turbulent motion takes place. The BL parametrization determines together with the land surface scheme the fluxes of heat moisture and momentum and redistributes the surface fluxes over the boundary layer depth. A realistic representation of the boundary layer in large scale models is important because (i) the large scale atmospheric budgets are affected on a time scale of a few days through the surface fluxes, (ii) the boundary layer interacts with other processes e.g. clouds, radiation, convection and (iii) boundary layer variables are important forecast products. The latter is obvious for wind at the 10 m level and the temperature at screen level, but also more and more use is made of advanced boundary layer parameters as surface fluxes, boundary layer height and boundary layer wind fields. These parameters are for instance needed to compute diffusion and advection of air pollution. In this paper we will try to assess the quality of BL related parameters in the ECMWF model like surface fluxes of heat and moisture, boundary layer depth and thermodynamic structure. The main comparison is with FIFE data (Kansas, USA, see Sellers et al. 1988) for August and October 1987, but we will also show some results from a comparison with data from Cabauw in the Netherlands and operational radio sondes over Europe. On the basis of these comparisons a number of model deficiencies were identified, which have led to the development of a revised boundary layer scheme. It will be shown how the new parametrization improves the results and which problems remain to be solved.
Comparison with FIFE data
During 1987 an extensive series of surface data was collected (Sellers et al. 1988) over the Konza prairie in Kansas during the First ISLSCP (International Satellite Land Surface Climatology Project) Field Experiment (FIFE). This data was gathered over an area of 15x15 km, reduced to a single mean time series (Betts and Ball 1992) and compared with time series of the closest grid-point from ECMWF model forecasts. Short range forecasts (48 hours) were selected for this comparison based on the idea that large scale meteorological features are fairly accurate in this time-range and that at the same time the boundary layer has had enough time to adjust to its forcing (boundary layer processes are relatively fast). To resolve the diurnal cycle it was necessary to rerun the model with diagnostics that archives model variables and fluxes every time step for selected grid points. A recent model version was used (cy38 with physics updates for cy39) which was started from the operational analysis of October 1987. For August 1987 the re-analysis, performed with the same model cycle, was used. Unfortunately, the re-analysis for August had the wrong deep soil climate field (the June climate, which is much more moist). The surface fluxes of heat and moisture should therefore not be interpreted in the absolute sense, but since they are so close to observations, the boundary layer budget of heat and moisture can still be compared with data to identify other model problems. For more details on FIFE data and the comparison with short range model forecasts we refer to Betts et al. (1993), details on the BL parametrization are given by Louis (1979) and Louis et al. (1982); the land surface scheme is described by Blondin (1991).
Beljaars, A. C. M. and A. K. Betts, 1992: Validation of the boundary layer representation in the ECMWF model. ECMWF Semina proceedings, 7-11 Sept. 1992, Validation of models over Europe, Vol II, 159-195. http://www.ecmwf.int/publications/library/ecpublications/pdf/seminar/1992/validation2beljaars.pdf