Changes in Temperature Records and Extremes: Are They Statistically Significant?

Abstract

Abstract The author investigates whether the increasing numbers of warm records and warm extremes in the extratropical Northern Hemisphere over the last decade are statistically significant. For the extremes, the focus is on summer mean temperatures; for warm records it is on daily and monthly means. Statistical significance is a highly nontrivial problem because the atmosphere is both spatially and temporally strongly autocorrelated. Therefore, a method is applied to produce an ensemble of surrogate fields that are statistically similar to the observed temperature field except that the surrogates are stationary. The significance is then estimated by comparing the number of records or extremes in the observations to similar numbers in the surrogates. The number of warm records and the number of extreme summers are found to have the same general temporal development, with a slow decrease from the late 1940s to a minimum in the 1970s followed by an increase to the present high values. However, there is a strong difference in the statistical significance of the different quantities. With very strong statistical significance, the recent large number of warm daily records and the number of extremely warm summers cannot be explained as chance occurrences. Both of these quantities show numbers of recent consecutive years with values above the 95% level that are much larger than any similar numbers found in the ensemble of 1000 surrogates. No significant change in the number of monthly warm records is found. The statistical significance weakens when considering the individual seasons or smaller regions like Europe.