| description abstract | The information available on different scales in the atmosphere for a number of different variables is explored using the global ECMWF analyses by examining the spatial spectra at T106 resolution. In most atmospheric spectra, a low wavenumber regime can be identified that does not follow a power law and is dominated by the stationary forced part of the flow. A higher wavenumber regime, where an approximate power law does appear to hold, can also usually be found. For the rotational part of the flow in the upper troposphere, the observed spectra follow quite closely that expected for quasi-two-dimensional geostrophic turbulence between about wavenumbers 12 and 70, with a kinetic energy spectrum falling off as n?3, where n is the total spherical harmonic wavenumber. In the lower troposphere, there is more power at high wavenumbers than would be expected from geostrophic turbulence, most likely due to the influence and close proximity of the lower boundary. Changes in the global analyses since 1979 have mainly influenced the spectra in the lower troposphere and the more recent analyses for 1988 have more power at higher wavenumbers. In the stratosphere, the spectra at high wavenumbers do not follow a power law behavior very well. The widespread practice of using a coarse grid without the appropriate truncation or smoothing first can result in unresolved scales being aliased and excessively noisy fields; an example is the 2.5° gridded fields made available by ECMWF. Appropriate procedures are described for truncating the T106 ECMWF spectral archive for scalar and vector fields. T42 resolution is an adequate representation for diagnostic calculations depicting most quantities within a few percent accuracy, although some spatial structure, which may partly be noise, is lost for the w, divergence, and moisture fields. In contrast, errors greater than 10% can occur at T21 or R15 resolution, although these truncations can be useful for displaying results. | |
