Atmospheric Predictability as Revealed by Naturally Occurring Analogues

Abstract

Two states of the atmosphere which are observed to resemble one another are termed analogues. Eitherstate of a pair of analogues may be regarded as equal to the other state plus a small superposed "error."From the behavior of the atmosphere following each state, the growth rate of the error may be determined. Five years of twice-daily height values of the 200-, 500-, and 850-mb surfaces at a grid of 1003 pointsover the Northern Hemisphere are procured. A weighted root-mean-square height difference is used as ameasure of the difference between two states, or the error. For each pair of states occurring within onemonth of the same time of year, but in different years, the error is computed. There are numerous mediocre analogues but no truly good ones. The smallest errors have an averagedoubling time of about 8 days. Larger errors grow less rapidly. Extrapolation with the aid of a quadratichypothesis indicates that truly small errors would double in about 2.5 days. These rates may be comparedwith a 5-day doubling time previously deduced from dynamical considerations. The possibility that the computed growth rate is spurious, and results only from having superposedthe smaller errors on those particular states where errors grow most rapidly, is considered and rejected. Thelikelihood of encountering any truly good analogues by processing all existing upper-level data appearsto be small.