nual variability wanes and waxes over the last 100 years. The corresponding curves for the standard deviation (not shown) reveal similar patterns, with increases in variance at the largest spatial scales with time but no consistent trends in the standard deviation of the smaller regions. As one would expect, there is a well-defined log-linear relationship between the magnitude of the interannual variability and the size of the area comprised by the various indices (see Figure 5). The inherently higher variability at smaller regional scales should be kept in mind when interpreting the paleoclimate record, which is based on limited regional samples.
The question of whether a warming climate will exhibit increased short-term variability is still open. Previous studies have been generally inconclusive on this subject (van Loon and Williams, 1978; Diaz and Quayle, 1980). Certainly, there is no consistent relationship between interdecadal changes in average temperatures and interdecadal changes in variability. Climatic variability may increase, decrease, or stay the same, in response to an arbitrary change in decadal mean temperatures. On the basis of the above results, we can say that, considering surface temperature variations during the last century, there has been a recent increase in variability at large spatial scales (continental to hemispheric-scale averages). This increase in variability is also concurrent with relatively large temperature increases over those areas. However, this is not the case for earlier warming episodes.
Prior to the late 1800s, reliable quantitative data on climate variations are sparse, and conclusions about the magnitude of these changes on large spatial scales necessarily contain significant degrees of uncertainty. There is ample evidence that multi-decadal temperature changes occurring over regions the size of Europe and China have been at least of order 1°C (Bradley and Jones, 1992). This suggests that it is quite plausible, if not likely, that temperatures as warm as those prevailing since the end of the Second World War may have been experienced at regional spatial scales of about 105 to 106 km2 for periods of one to several decades during some portions of the past thousand years under what amounts to natural (i.e., with negligible human influences) conditions.
Below we present some examples of long-term variability in a set of climate-sensitive paleotemperature records. The climate indices used here are discussed in several chapters of Bradley and Jones (1992). A listing of them, together with the source references and periods of record, appears in Table 1; each of them has been assigned a number for use in later tables comparing their data. Because of its very long record, the instrumental temperature series for central England (Manley, 1974) is considered in this section for comparison with other paleoclimate indices.
There is an extensive body of work relating climate variations to growth changes in particularly climate-sensitive trees (e.g., Fritts, 1976; Fritts et al., 1979; Hughes et al., 1982; Cook and Kairiukstis, 1990). We have selected a suite of climate-sensitive tree-ring records to study possible changes in climatic variability. Two main considerations were applied in the selection of these indices. First, we chose