Free and Forced Changes of the Surface Atmospheric Circulation and Storminess Extremes over the last 140 years

To gain a clearer understanding of natural versus forced changes of the near-surface climate and storminess over the last 140 years, we have investigated large ensembles of three distinct types of observational datasets and model simulations of the 1874 to 2010 period. The first type is a 56-member ensemble of global atmospheric observational reanalyses at 6-hourly resolution, the 20CRv2 dataset. The second type of dataset, that we refer to as the AMIP20C dataset, is a 56-member ensemble of atmospheric GCM simulations of the 1874 to 2010 period generated by us using the same NCEP AGCM used to produce the 20CRv2 dataset, and with identical specifications of time-varying SST, sea ice, and radiative forcings. The third dataset type is a multi-model ensemble of 62 CMIP5 coupled climate model simulations of the same 1874 to 2010 period with observed radiative forcings, downloaded from the PCMDI archive at daily resolution.

These three types of datasets enable cleaner separations of the radiatively forced versus internal natural climate variations than previously possible. We do this by interpreting the long-term variability in the 20CRv2 observational dataset as a combination of internal chaotic, SST-forced, and radiatively forced variations; the variability of the ensemble-mean responses in the AMIP20C simulations as a combination of responses to natural and radiatively forced SST variations; and the variability of the multi-model ensemble (MME) mean responses in the CMIP5 simulations as responses to radiative forcing variations.

The most important result from this study is that the observed trends in many circulation variables evident over the second half (1943-2010) of the record are much weaker or non-existent when considered over the full record (1874-2010). The weak longterm circulation trends are associated with even weaker longterm storminess trends (where storminess is defined at each gridpoint as the r.m.s. of 24-hr SLP differences). These results have important implications for the atmospheric circulation response to global warming, and cast doubt on inferences about this response drawn in many studies from considering only the second half (or an even shorter subset) of the record. Consistent with the weak observed longterm circulation trends, the longterm trends in the ensemble-mean AMIP20C and MME-mean CMIP5 simulations are also weak over most of the globe, except in the extratropical southern hemisphere. In the extratropical northern hemisphere, not just the long-term MME-mean trend but even the MME-mean multidecadal variations are weak.