Estimates of the Autocorrelations and Spectra of Seasonal Mean Temperatures over North America

Abstract

Lag one-year autocorrelations and spectra for summer and winter seasons are estimated from 64-Year time series at 87 North American stations. Linear trends are first removed from the data to eliminate the effects of very long-period (longer than 64 years) variations. Summer season temperatures appear to have more year-to-year correlation and ?redder? spectra than those of winter seasons. High correlations and red spectra are not necessarily found more frequently at near-ocean stations than at interior stations. Excepting variance near the quasi-biennial period, which is not judged because of its proximity to the folding frequency, no spectral ?peaks? are found. Because of their differing autocorrelations and spectra it is concluded that evidence for potential long-range (more than two but less than 32 years) predictability is greater in summer than in winter, and that this is due, at least in part, to less natural variability or climatic noise in summer than in winter.