A multivariate analysis of Arctic climate is performed comparing the o
bserved climate with that simulated by four different global climate m
odels (GCMs). The focus is on the patterns of temporal and spatial var
iability in several climate parameters (sea level pressure, temperatur
e, specific humidity, and precipitation). There are broad similarities
between the observed data and all the GCM climates. There are, howeve
r, several major differences. The observed data show the Arctic climat
e to be dominated by the summertime pattern of temperature and humidit
y, which is decoupled from the atmospheric circulation. The winter pat
terns explain less of the observed variance but show a much closer ass
ociation between temperature and the large-scale circulation. The GCMs
, in contrast, overemphasize the winter season and show more of a larg
e-scale advective control on summertime temperature patterns. Possible
reasons for these differences are suggested, and their implications f
or GCM climate studies are discussed. The shortcomings in the GCMs poi
nt to the need for improvements in boundary layer rendition, in the tr
eatment of Arctic stratus, and in sea ice simulations through coupled
ocean models and the inclusion of ice dynamics.