NORMAL-MODES OF THE ATMOSPHERE AS ESTIMATED BY PRINCIPAL OSCILLATION PATTERNS AND DERIVED FROM QUASI-GEOSTROPHIC THEORY

The principal oscillation pattern (POP) analysis is a technique to emp irically identify time-dependent spatial patterns in a multivariate ti me series of geophysical or other data. In order to investigate medium -scale and synoptic waves in the atmosphere it has been applied to tro pospheric geopotential height fields of ECMWF analyses from 1984 to 19 87. The data have been subjected to zonal Fourier decomposition and to time filtering so that variations with periods between 3 and 25 days were retained. Analyses have been performed separately for each zonal wavenumber 5-9 on the Northern Hemisphere in winter and on the Souther n Hemisphere in summer (DJF). POPs can be seen as normal modes of a li near approximation to a more complex dynamical system. The system matr ix is estimated from observations of nature. This concept is compared with conventional stability analysis where the system matrix of the li near system is derived from theoretical, in this case quasigeostrophic , reasoning. Only the mean basic flow depends on time- and space-avera ged fields of observed wind and temperature from the ECMWF data. It tu rns out that the most significant POPs are very similar in time and sp atial structure to the most unstable waves in the stability analysis. They describe the linear growth phase of baroclinic, unstable waves th at propagate eastward with periods of 3-7 days. Since the POPs are pur ely derived from observations, the results indicate the appropriatenes s of the assumptions usually made in linear stability analysis of zona lly symmetric flows to explain high-frequency atmospheric fluctuations . Moreover, the POP analysis reveals patterns that are not found in th e linear stability analysis. These can possibly be attributed to the n onlinear decay phase of baroclinic waves. Eliassen-Palm cross sections help clarify the interpretation of the POPs in terms of the life cycl e of nonlinear baroclinic waves.