changes varies for the smaller regional averages, with most areas of the contiguous United States experiencing their greatest decadal changes during the first half of this century.
Recent studies have documented some rather pronounced and relatively rapid shifts in the climate system that occurred on decadal time scales. In particular, the atmospheric circulation in the extratropical North Pacific underwent a major shift in the mid-1970s that lasted over 10 years (Trenberth, 1990; Ebbesmeyer et al., 1991). This change was accompanied by a tendency for climate patterns in the tropical Pacific to exhibit conditions reminiscent of El Niño/Southern Oscillation (ENSO) warm-event conditions—weaker trades, warmer equatorial sea surface temperature (SST), anomalous rainfall patterns, etc.—relatively more frequently than during previous decades. Gordon et al. (1992) have also reviewed some of the relatively abrupt changes in water properties and surface climate in the Atlantic Ocean that have occurred in the past several decades, including sudden changes in SST anomalies in the South Atlantic compared to those of the North Atlantic, and the so-called "great salinity anomaly" in the North Atlantic from about the late 1960s to the early 1980s (Dickson et al., 1988). Certainly the sudden decrease of rainfall in the Sahel of Africa, a feature that has lasted for two decades, is a good example of how the climate can undergo significant, sudden, and prolonged change over relatively large spatial scales.
Clearly, increased knowledge of the behavior of climatic variability, from the interannual through decadal and century time scales, is needed to improve our assessments of any future changes in climate from regional to global scales. Indeed, as was noted by Ghil and Vautard (1991), the ability to distinguish a warming trend from natural variability is critical for separating out the greenhouse-gas-induced signal.
As we noted at the outset, an important element of climate that is too often overlooked is the variance or the characteristic variability of climatic means. In determining climate impacts, the importance to society of short- to medium-term climatic instability (i.e., climate fluctuations occurring on interannual to interdecadal time scales) is perhaps equal to or greater than that of slow changes in the background mean. Below we have analyzed some aspects of low-frequency changes in the temperature variance of different regional means.
We have examined contemporaneous variation of two measures of temperature variability, the standard deviation in running 15-year segments, and decadal means of the root-mean-square differences between successive yearly values of seasonal and annual regional temperature anomalies. The latter index is defined as
Since the xi are deviations from a reference mean and have approximately zero mean,
where s2 is the series variance, and r1 is the autocorrelation of the time series with a lag of I year. We will use this measure, rather than the standard deviation, as the key index of interannual variability. Note that this index, Ind V, amplifies changes in the high-frequency part of the variance spectrum.
Figures 3 and 4 illustrate the changes in the interannual variability of the regional groupings discussed earlier. The curves differ from one season to another and from one region to another. For the largest continental-sized regions, there appears to be an increase in interannual variability in the last couple of decades. At smaller spatial scales, there does not seem to be a consistent trend; instead, the interan-