Internal atmospheric noise characteristics in twentieth century coupled atmosphere–ocean model simulations

Abstract

The statistical characteristics of the atmospheric internal variability (hereafter internal atmospheric noise) for surface pressure (PS) in twentieth century simulations of a coupled general circulation model are documented. The atmospheric noise is determined from daily post-industrial (1871–1998) Community Climate System Model 3 simulations by removing the SST and externally forced responses from the total fields. The forced responses are found from atmosphere-only simulations forced by the SST and external forcing of the coupled runs. However, we do not address the influence of the SST variability on the synoptic scale high frequency weather noise.The spatial patterns of the main seasonal modes of atmospheric noise variability are found for boreal winter and summer from empirical orthogonal function analyses performed globally and for various regions, including the North Atlantic, the North Pacific, and the equatorial Pacific. The temporal characteristics of the modes are illustrated by power spectra and probability density functions (PDF) of the principal components (PC). Our findings show that, for two different realizations of noise, the variability is dominated by large scale spatial structures of the atmospheric noise that resemble observed patterns, and that their relative amplitudes in the CGCM and AGCM simulations are very similar. The regional expression of the dominant global mode, a seasonally dependent AO-like or AAO-like pattern is also found in the regional analyses, with similar time dependence. The PCs in the CGCM and the corresponding SST forced AGCM simulations are uncorrelated, but the spectra and PDFs of the CGCM and AGCM PCs are similar.The temporal structures of the noise PCs are white at timescales larger than few months, so that these modes can be thought of as stochastic forcings (in time) for the climate system. The PDFs of the noise PCs are not statistically distinguishable from Gaussian distributions with the same standard deviation. The PDFs do not change substantially between the first and second half of the twentieth century.