Abstract: Regional Climate Models are widely used tools to add detail to the coarse resolution of global simulations. However, these are known to be affected by biases. Usually, published model evaluations use a reduced number of variables, frequently precipitation and temperature. Due to the complexity of the models, this may not be enough to assess their physical realism (e.g. to enable a fair comparison when weighting ensemble members). Furthermore, looking at only a few variables makes difficult to trace model errors. Thus, in many previous studies, these biases are described but their underlying causes and mechanisms are often left unknown. In this work the ability of a multi-physics ensemble in reproducing the observed climatologies of many variables over Europe is analysed. These are temperature, precipitation, cloud cover, radiative fluxes and total soil moisture content. It is found that, during winter, the model suffers a significant cold bias over snow covered regions. This is shown to be related with a poor representation of the snow-atmosphere interaction, and is amplified by an albedo feedback. It is shown how two members of the ensemble are able to alleviate this bias, but by generating a too large cloud cover. During summer, a large sensitivity to the cumulus parameterization is found, related to large differences in the cloud cover and short wave radiation flux. Results also show that small errors in one variable are sometimes a result of error compensation, so the high dimensionality of the model evaluation problem cannot be disregarded.

Authorship

Download reference