By constructing the simultaneous and lag correlation and covariance fields of the ECMWF variables, properly time filtered by using the wavelet analysis technique, which were regressed with a key time series of the 1000 hPa southerly winds at a grid point (0, 30 ¢ªW), we were able to identify the distinguished structure and evolution of the decadal oscillations globally. At first we examined the distribution of the standard deviation of various parameters. The values are large for the tropics and subtropics for variables such as ozone, water vapor. For geopotential and temperature, there is very strong variation close to the poles throughout the troposphere and the lower stratosphere.

Horizontally the decadal oscillation is clearly identified in the ozone and relative humidity in the tropical regions with a reliable signature in the extra-tropics and also far to the polar region. For all the meteorological parameters the upper levels show strong signals of 11-years oscillation. Near the surface at the equatorial Atlantic, there exists clear decadal variability in the meridional winds, which was used for composites for other variables such as relative humidity, specific humidity, temperature, winds in their time mean fields and variances and correlation and covariance statistics. While vertically the decadal signatures of variability clearly manifest themselves to the 10 hPa level for the ozone density, zonal winds, and relative humidity near the tropopause level. The maximum variability of ozone mass mixing ratio appeared near 20 hPa level approximately. The evolution of the decadal components of atmospheric motions is very consistent with between the fields of geopotential, temperature, winds, ozone, and relative humidity.

Strong and consistent signatures in the structure and their evolution of the decadal oscillations clearly supersede the restrictions in the data record length and some defects of too strong convective motions over the tropics, which are the inherent propensity of the ECMWF reanalysis for studying oscillations with periods of about 10 years or more. The reality of 11-year oscillation was checked by using the conventional FFT, wavelet coherency, and simple statistics combined with wavelet spectrum analysis using zonal mean zonal winds.