FIGURE 5

Three time series. Upper series: equatorial Pacific sea surface temperature anomalies for the region 160°E to 90°W, 5°N to 5°S, which corresponds to the regions included in Niño 3 and Niño 4 in the Climate Diagnostic Bulletin. Middle series: tropospheric temperature anomalies, as sensed by the Microwave Sounding Unit (MSU-2) carried aboard the TIROS N satellites, averaged over the tropical belt (20°N to 20°S). Lower series: surface air temperature over all land grid points within the tropical belt (20°N to 20°S). The temperature scale is the same for all three curves: one small tick mark is equivalent to 0.25K. MSU data courtesy of Roy Spencer, NASA Marshall Space Flight Center, Huntsville, Ala.. (After Yulaeva and Wallace, 1994; reprinted with permission of the American Meteorological Society.)

temperature anomalies is unmistakable. A correlation coefficient of about 0.8 can be obtained by performing a transformation on the SST time series, using it as the input for a stochastically forced climate model based on the formulation of Hasselmann (1976) (see also Frankignoul, 1985, Eq. 20). This formulation contains only two free parameters. One is analogous to a characteristic damping time of the forced climate system, which can be determined from the Stefan-Boltzmann Law, under the assumption that tropospheric temperature anomalies are radiatively damped. The other is analogous to a heat capacity (in this case, that of the troposphere and perhaps some part of the oceanic mixed layer in the passive regions of the tropics). This coefficient can be determined empirically to optimize the agreement between the output of the model and the tropospheric time series. The output of the model, shown as the middle curve in Figure 6, captures most of the features in the land-surface temperature time series immediately above it. For further details, see Yulaeva and Wallace (1994).

Vestiges of the same warm and cold episodes are evident in global mean MSU-2 time series shown in Figure 7, but the ENSO signal is diluted by about a factor of 3 by being combined with unrelated features in the extratropical time series, which are also shown in the figure. Time series of 850 to 300 hPa thickness based on zonal averages of longer records of rawinsonde observations for selected latitude belts also give the impression that the ENSO signal is largely confined to the tropical belt (Angell, 1988). It is evident from Figure 7 that the interannual variability of tropical tropospheric temperatures in association with the ENSO cycle is much larger than that of extratropical tropospheric temperatures. One obtains similar impressions from the results of EOF analysis of cosine-weighted, zonally averaged temperature anomalies derived from MSU-2 instruments. The leading mode is largely confined to the tropics and symmetric about the equator (Yulaeva and Wallace, 1994).

Figure 8 shows longer records of land surface air temperature for the same latitude belts. As in the previous figure, the temperature scales for all four curves are the same, but in this case a 3-month running mean filter has been applied to suppress the high-frequency variability in the extratropi-

FIGURE 6

Time series of (upper) equatorial Pacific SST, as in Figure 5, (lower) surface air temperature over all land gridpoints within the tropical belt (20°N to 20°S), and (middle) simulated tropical tropospheric response to the SST time series, based on a simple thermodynamic model as described in Yulaeva and Wallace (1994). One small tick mark on the vertical scale is equivalent to 0.5K.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement