In a recent paper, M. Pattantyús-Ábrahám and colleagues from the MiKlip project investigated the performance of the MiKlip decadal prediction system for the free atmosphere.
Homogenized radiosonde records offer more than 50 years of data and thus the longest data-set in the free atmosphere. To test the MiKlip climate model system the retrospective forecasts (hindcasts) of the MiKlip System were compared against radiosonde data in the European region.
Results show systematic differences between simulated and measured temperatures. The modelled temperatures are lower than observed throughout the atmosphere. This temperature difference increases with height up to the tropopause. In the stratosphere, the difference decreases again until the 100hPa pressure level. This behavior is seen above all of Europe. It suggests that the simulated atmosphere is too cold and less stable than observed. The modelled atmosphere is also more humid than observed, even when the lower temperatures are considered. Thus, raw results of the global model, e.g. regarding the probability for severe weather, should not be used directly. Instead, post-processing (downscaling to regional scale, calibrating atmospheric variables) is required before making detailed predictions about the future state of climate.
Pattantyús-Ábrahám, M., C. Kadow, S. Illing, W.A Müller, H. Pohlmann, W. Steinbrecht (2006): Bias and Drift of the Medium-Range Decadal Climate Prediction System (MiKlip) validated by European Radiosonde Data. Met. Zet. 25 p709–720, doi: 10.1127/metz/2016/0803