Abstract

C.S. Frederiksen(1), A.P. Kariko(1) and X. Zheng(2)

(1) Bureau of Meteorology Research Centre, Melbourne, Australia
(2) National Institute of Water and Atmospheric Research, Wellington, New Zealand

In this paper, a modified Empirical Orthogonal Function (EOF) analysis is proposed for extracting long-range potentially predictable patterns of meteorological seasonal mean fields, that is, the patterns arising from slowly varying external forcings and slowly varying internal dynamics. EOF analysis, or Principal Component Analysis (PCA), is very commonly used to derive spatial patterns of inter-annual variability in climate data. The relationship between these spatial patterns and external forcings, such as sea surface temperatures (SSTs), form the theoretical basis of many seasonal forecasting schemes. However, there is no guarantee that the dominant patterns derived by conventional EOF analysis are closely related to slowly varying external forcings and slowly varying internal dynamics. This is particularly the case for regional meteorological fields in the extra-tropics, where the chaotic, that is, day-to-day weather, components of atmospheric variability and co-variability are reasonably high.

Our proposed method, based on an analysis of variance and covariance and using daily data, provides a means of decomposing the covariance/correlation matrix of a seasonal mean field into covariance/correlation matrices for the potentially predictable and the chaotic, or weather-noise, components, separately. The dominant patterns, arising from the potentially predictable covariance/correlation matrix, are shown to be more closely related to slowly varying external forcings and slowly varying internal dynamics than those from a conventional EOF analysis. We also briefly discuss the weather-noise patterns of variability.

Full Paper (Size: 2.9 MB)