A Height Dependent Evaluation of Wind and Temperature over Europe in the CMIP5 Earth System Models

Abstract:

Global Circulation Models (GCMs) are commonly evaluated on their performance at pressure levels of 1000, 850 and 500hPa. However, due to their improved interaction with the surface and increasing resolution, the GCMs are getting more and more realistic in representing variables at lower levels. For downscaling practices this implies that the near-surface variables might become suitable as predictors in statistical models and might increase the added value of dynamical models. This work performs a height dependent evaluation of six of the Earth System Models (ESMs) from the Coupled Model Intercomparison Project Phase 5. The ESMs are evaluated on the representation of the Probability Density Functions (PDFs) of wind and temperature in the lowest 1.5km of the atmosphere over Europe, using ERA-Interim reanalysis data as the reference. Results indicate that apart from small-scale biases, the surface wind speed PDFs north of 45°N are well represented by all ESMs. Therefore they can be considered skillful inputs for statistical downscaling practices. In such a way statistical downscaling models might profit from predictors which are more closely related to the predictands. South of 45°N, winds are affected by a large-scale bias originating from errors in the representation of the large-scale circulation, especially during winter. For temperature, near-surface levels as well as upper-atmospheric levels are affected by small-scale and large-scale biases. The latter are adopted by the downscaling models, underlining the importance of model evaluation before downscaling. In general, the height-dependent near-surface evaluation approach that is adopted in this work, gives more insight in the origin of large-scale biases, defines up to which altitude ESMs are influenced by small-scale phenomena and determines the lowest levels for which temperature and wind speed PDFs are suitable input variables for downscaling models.