Evaluation of midlatitude cloud properties in a weather and a climate model: Dependence on dynamic regime and spatial resolution

Publisher Website
Google Scholar
Abstract
In this study, the midlatitude cloud fields produced by a climate (GISS) and a weather (ECMWF) model are evaluated against satellite observations. Monthly ensembles of model cloud property distributions for the four seasons are compared with similar ensembles from satellite retrievals. The weather model is run in both forecast and “climate” mode in order to evaluate the importance of the exact representation of the atmospheric conditions in these ensemble comparisons. The weather and climate models are evaluated at different resolutions that cover the range used in today's climate and weather prediction simulations. Cloud property evaluations are separated into broadly defined dynamic regimes that cover the range of large‐scale midlatitude motions. Quantitative evaluation tables are produced that rank the performance of the different model versions used in the study. The evaluation analysis reveals several common features between the two models. Those are the overestimation of cloud optical depth in all dynamic regimes, the underestimation of cloud cover in the large‐scale descent regime and the underestimation of cloud top height in the large‐scale descent regime. It is also shown that, in the radiative balance calculations, the models compensate for the overestimation of cloud optical depth through the underprediction of cloud cover. The comparison of the forecast and “climate” runs of the ECMWF model shows remarkably similar statistical properties of the clouds in the two runs. The analysis of runs with different resolutions reveals large improvement when going from a 4° × 5° 9‐layer to a 2° × 2.5° 32‐layer run with the GISS GCM, much of which is caused by the increase in vertical resolution. A comparison of a T42 and a T106 run of the same vertical resolution with the ECMWF GCM does not show considerable differences between the two model versions.