Characterizing and comparing control-run variability of eight coupled AOGCMs and of observations. Part 1: temperature

PDF: http://data.dhkim.info/monograph/CD/RCD4XQVUTVWHCF2J.pdf

Article  

  Climate Dynamics
Publisher: Springer-Verlag Heidelberg
ISSN: 0930-7575 (Paper) 1432-0894 (Online)
DOI: 10.1007/s00382-003-0357-x
Issue: Volume 21, Numbers 7-8

Date:  December 2003
Pages: 619 - 646  
Characterizing and comparing control-run variability of eight coupled AOGCMs and of observations. Part 1: temperature
L. D. D. Harvey1  and T. M. L. Wigley2

(1)  Department of Geography, University of Toronto, 100 St. George Street, Toronto M5S 3G3 Canada,  
(2)  National Center for Atmospheric Research, Boulder, CO 80307, USA,  

Received: 17 December 2002  Accepted: 20 August 2003  Published online: 15 November 2003

Abstract   We examine the spatial patterns of variability of annual-mean temperature in the control runs of eight coupled atmosphere–ocean general circulation models (AOGCMs) and of observations. We characterize the patterns of variability using empirical orthogonal functions (EOFs) and using a new technique based on what we call quasi-EOFs. The quasi-EOFs are computed based on the spatial pattern of the correlation between the temperature variation at a given grid point and the temperature defined over a pre-determined reference region, with a different region used for each quasi-EOF. For the first four quasi-EOFs, the reference regions are: the entire globe, the Niño3 region, Western Europe, and Siberia. Since the latter three regions are the centers of strong anomalies associated with the El Niño, North Atlantic, and Siberian oscillations, respectively, the spatial pattern of the covariance with temperature in these regions gives the structure of the model or observed El Niño, North Atlantic, and Siberian components of variability. When EOF analysis is applied to the model control runs, the patterns produced generally have no similarity to the EOF patterns produced from observational data. This is due in some cases to large NAO-like variability appearing as part of EOF1 along with ENSO-like variability, rather than as separate EOF modes. This is a disadvantage of EOF analysis. The fraction of the model time-space variation explained by these unrealistic modes of variability is generally greater than the fraction explained by the principal observed modes of variability. When qEOF analysis is applied to the model data, all three natural modes of variability are seen to a much greater extent. However, the fraction of global time-space variability that is accounted for by the model ENSO variability is, in our analysis, less than observed for all models except the HadCM2 model, but within 20% for another three models. The space-time variation accounted for by the other modes is comparable to or somewhat larger than that observed in all models. As another teleconnection indicator, we examined both Southern Oscillation Index (SOI) and its relation to tropical Pacific Ocean temperature variations (the qEOF2 amplitude), and the North Atlantic Oscillation Index (NAOI) and its relation to North Atlantic region temperatures (the qEOF3 amplitude). All models exhibit a relationship between these indices, and the qEOF amplitudes are comparable to those observed. Furthermore, the models show realistic spatial patterns in the correlation between local temperature variations and these indices.