Interannual Variability and Predictability of 500 mb Geopotential Heights over the Northern Hemisphere

Abstract

The interannual variability of monthly mean 500 mb heights in a 15-year sample of observed data is compared to the variance expected from sampling errors associated with high-frequency fluctuations using the analysis of variance approach. The monthly mean “signal” stands out significantly from the “noise” over a substantial fraction of the Northern Hemisphere during the winter. The expected spectrum of variance at very low frequencies is assumed to be white at frequencies lower than (30 days)⁻¹ rather than the (96 days)⁻¹ cutoff used by Madden. This difference is justified by observing that the effective time between independent samples T₀ is relatively insensitive to changes in the maximum lag over which the local autocorrelation is integrated to calculate T₀. Further non-randomness in the variance of 500 mb heights is evidenced by the correlation between monthly mean height and contemporaneous daily variability.