Boyle, James S.
The low frequency variation in the three dimensional air temperature fields of two reanalyses and two model simulations are described. The data sets used are the monthly mean temperature fields for the NCAR Climate Simulation Model (CSM, Boville and Gent, 1998) 300 year run, an NCAR Community Climate Model version 3 (CCM3, Kiehl et al., 1998) AMIP type simulation, and the NCEP/NCAR and ECMWF (ERA) reanalysis data sets. The variances and correlations are computed for the anomalies from the annual cycle for each data set.
In general, the reanalyses and models agree fairly well on the structure of the temperature variance. The models tend to have too much variance at the surface compared to the reanlyses. The CSM's poor simulation of the SST in the eastern Pacific leads to a much reduced variance in the Nino3 region. The enhanced variability over land appears to affect the midlatitude simulation of the CSM, in that the higher surface variability extends off the east coast of continents. This is not evident in CCM3 and reanalyses where the SSTs are prescribed.
At 200 hPa the CCM3 and reanalyses all evince the dumbbell pattern straddling the Equator in the eastern Pacific, attributed by Yulaeva and Wallace (1994) to ENSO variations. The CSM shows no such pattern. A CCM3 integration using climatological SSTs displays more variance than the CSM in this region. Apparently the coupling to an ocean in the CSM suppresses the atmospheric model's variability in this locale.
The correlations of the temperature fields with the surface air temperature show that the regions of subtropical subsidence are virtually uncorrelated to the surface at the 700 hPa level. The regions of the cold water off the west coast of continents evince decoupling with the surface at 850 hPa. In the region from 30S to 30N the zonal mean correlation falls to about 0.7 below 800 hPa, with this value extending up to about 600 hPa in mid and upper latitudes. These characteristics are consistent across all the data sets. Thus, the variations of vertically integrated measures such as MSU temperatures need not agree with observations of surface air temperatures at the time scales examined here.