Evaluation of Model-generated Cloudiness: Satellite-observed and Model-generated Diurnal Variability of Brightness Temperature

Abstract

In an attempt to validate the ECMWF model?s cloudiness, model output has been processed to reproduce satellite measurements as closely as possible. Brightness temperatures in the longwave window channel of Meteosat are simulated from cloudiness, temperature, and humidity fields produced by the forecast model, and compared to the equivalent ISCCP-B3/CX observations over a 5-day period in July 1983. On an instantaneous basis, clear-sky brightness temperatures from the model are generally in good agreement with observations, but errors arise from inaccuracies of the radiation transfer scheme, uncertainties in the input temperature and humidity profiles, and inadequate retrieval of the surface brightness temperature. The diurnal cycles of surface temperature and cloudiness, over a number of limited 100° ? 10° areas representative of different cloud regimes, are studied using evolution histograms of window brightness temperature. Comparison of histograms allows various deficiencies in the parameterization of the surface-cloud-radiation interactions to be pinpointed. Over tropical land areas, the model correctly reproduces the observed decrease in the amplitude of the diurnal cycle of surface temperature when the vegetation cover increases. The land surface temperature generally displays the correct phase but with too small an amplitude, linked to a combination of too little downward shortwave radiation, too large a surface albedo, or inadequate ground thermal resistance and surface-air coupling. Model brightness temperatures over high clouds are generally too high, showing deficiencies in the diagnosed cloud cover and cloud liquid-water content. The vertical extent of the convective cloudiness is rather well represented, but the deep clouds over land generally dissipate 6 h too early.