Local polynomial regression-based evaluation of general circulation models regarding teleconnections in boreal wintertime

General circulation models (GCMs) are effective tools to simulate the climate system and to assess its changings. Since large-scale atmospheric teleconnections are the imprints of the internal variability of the climate system, the evaluation of the GCMs concerning teleconnections became an important task in the field of climatology. Previously, a negative extrema-based method was applied on gridded 500 hPa geopotential height time series in the boreal wintertime. First, the Pearson correlations were calculated between all time series. Then, the strongest negative correlation was selected in each grid cell. Finally, the pairs of grid cells were selected which have the same associated negative correlation values. Based on the ECMWF’s ERA-20C and the NCEP/NCAR Reanalysis 1 datasets, it was found that, these pairs of grid cells are mostly located in a north‑south direction, surrounding the Northern Hemisphere in two belts. This means that the geopotential height in the two belts changes simultaneously in opposite directions. The two belts were represented by local polynomial regression (LOESS) curves. It was pointed out that the two belts can be detected in the first realizations of all available historical GCM outputs of the Coupled Model Intercomparison Project Phase 5 (CMIP5), but the courses of the LOESS curves were not captured well by some of the GCMs compared to the reanalyses, especially in case of the southerly located belt. However, different GCM realizations may produce significantly different LOESS curves. Therefore, the aim of this study is to quantify the differences of the LOESS curves obtained from multiple realizations of the GCMs disseminated by the Coupled Model Intercomparison Project Phase 6 (CMIP6). The future goal is to use LOESS curves to detect changes in the climate system, i.e. shifts of the oppositely varying belts.