Atmospheric thickness anomalies of the 850-200 hPa layer has been found to be highly correlated (R=0.85-0.95) with the Microwave Sounding Unit (MSU) channel 2 Tb (brightness temperature) anomalies. Monthly anomalies of the models' thickness and Tb were correlated by model grid point over the ten-year study period (N=120). All models demonstrate their highest degree of correlation in the near equatorial region. All of the models analyzed at this point have their highest correlation in the Eastern Tropical Pacific Ocean. All models tested share a common decreasing value of correlation in the poleward directions.
To show the variation of temperature sensitivity in the models, a global thickness index was derived. The global mean monthly anomalies of both 850-200 hPa thickness and Tb, were computed and the quotient of the respective standard deviations: s(Z')/s(Tb') was calculated. It may be thought of as the global mean thickness change which occurs for a corresponding global mean change of 1°C in the Tb. The authors have found a clearly wide range (40-64 gpm) in global mean layer thickness sensitivity to SST forcing among the models. The corresponding global mean index for 850-200 hPa thickness calculated by radiosonde versus co-located Tb radiances for 1979-88 is 55 gpm/degrees C. Most models display less sensitivity than is observed, indicating that the heat supplied from the SST anomalies is not effectively transported into the tropospheric layer globally.
Thickness variations at every gridpoint were converted to simulated MSU values by an empirically derived formula. From this a decadal trend could be calculated, all models demonstrate a strongly positive trend - values range from +0.08 to +0.23. The MSU2 demonstrates a significantly lower value of +0.025.
Principal component analysis was performed on the model thicknesses, and confirms the main atmospheric response pattern, that being associated with the ENSO. The NH has a relatively good agreement that diminishes in the extratropics and SH. The first empirical orthogonal function explains 9.8% of the total variance in the MSU, and a corresponding 16.4% of the variance for the ECMWF thicknesses. SST forcing alone appears to have little consequence in defining the mid-latitude circulation, as correlations between observed and model-derived Tb were universally near zero there.
Persistence of anomalies was investigated by calculating the variance for monthly, seasonal and annual gridded anomalies. A general characteristic of the models is that the magnitude of variance, relative to MSU variance, increases as averaging period increases indicating a tendency for anomalies in models to persist longer than observed.
In order to detect the impact of the El Chichón volcanic eruption of April, 1982, on surface temperature, we have examined the AMIP model output for the 10-year AMIP period and separated the ENSO and volcanic signals. We interpolated all model output to a 5 x 5 degree grid, the same grid as the Jones surface temperature data, and calculated the temperature anomalies with respect to the 10-year mean. We then selected 9 of the 21 models where the first EOF had a correlation greater than 0.71 with the SOI index. They were: CCC, CNRM, CSU, ECMWF, GLA, GSFC, JMA, MPI and UCLA.
We then compared the patterns produced by these models with the patterns
for the winters of 82-83 and 86-87. While all the models simulated the
86-87 winter well, for 82-83, the difference between the model output and
the observations shows a clear winter warming pattern from volcanic eruptions.