An Information Theory–Based Evaluation of General Circulation Models Regarding Atmospheric Oscillations and Their Effects on the Carpathian Basin

The winter climate of the Carpathian Basin located in Central/Eastern Europe is mainly governed by large-scale atmospheric circulation systems, i.e. oscillation phenomena (hereinafter oscillations). Because of this, the scale of their effects and their projected changes in the 21st century is a crucial topic amid climate-related studies. In this study, we aim to assess the strength of the relationship between oscillations and atmospheric variables in the region of the Carpathian Basin in the 21st century. To reach this goal, multivariate statistical analysis is performed on reanalysis datasets and on the simulation outputs of general circulation models (GCMs) that are available in the framework of the Coupled Model Intercomparison Project Phase 5 (CMIP5). First, oscillations are identified and their effects are quantified in reanalysis datasets in the winter months (December, January and February) of the second half of the 20th century. For this purpose, geopotential height field at the 500 hPa isobaric surface is used for the Northern Hemisphere. Then, historical GCM outputs are compared to the results obtained from reanalysis datasets. This step is required to evaluate GCMs with respect to the oscillations. Subsequently, GCMs in which oscillations distribute similarly to the reanalysis datasets, are determined as the subject of further study. The selection of these GCMs is a quite challenging task. The application of commonly used measures (e.g. standard deviation, root-mean-square error) seems to be not the most advantageous way to rank GCMs because they failed to extract important information from the data. Consequently, an information theory-based approach is applied to facilitate the selection of the best-performing GCMs. The computation of diversity indices (e.g., Shannon index) are more advantageous because they enable us to distinguish grid points containing a substantial amount of information within the analyzed field. Furthermore, this method facilitates the complex analysis that takes into account both linear and nonlinear effects simultaneously. The final aim of our study is to examine Representative Concentration Pathway (RCP) simulations of the best-performing GCMs to provide improved climate predictions for the Carpathian Basin for the 21st century that can assist impact studies in the region and act later as a key input in developing appropriate adaptation and mitigation strategies on national and regional level.