Parametrization of the spectral dependence of the aerosol extinction coefficient in problems of atmospheric occultation sounding from space

An analysis of the published data and the results of interpretation of the data obtained by the Ozon-Mir spectrometer show that, during the occultatio n sounding of the atmosphere from space, the accuracy in determining the ch aracteristics of gas composition depends significantly on the method of par ametrizing the spectral dependence of the aerosol extinction coefficient (A EC). Errors in the spectral approximation of the aerosol extinction may res ult in 15-30% errors in determining the ozone content in the lower stratosp here, which significantly exceed the present-day requirements for the accur acy of remote measurements. The requirements are formulated for the paramet rization methods that are optimal in respect to the accuracy and computing power. Based on the numerical simulation of an ensemble of different atmosp heric aerosol conditions and the approximation of the AEC spectral dependen ces corresponding to these conditions, an idea of parametrizing the ABC spe ctral dependence through the expansion in terms of the empirical orthogonal basis is proposed. A model of variability of the microphysical parameters of the stratospheric aerosol, which takes into account the sulfuric acid ae rosols, volcanic dust, two meteoric-dust fractions, and two-layer particles , is developed. On the basis of a generated dataset of realizations of the microphysical parameters of stratospheric aerosol, the corresponding datase t of AEC values was calculated for the spectral range of the SAGE III measu ring instrumentation and the heights between 12 and 60 km. Two optimal para metrizations of the AEC spectral dependence, which are the expansion of the AEC spectral dependence (linear parametrization) and its logarithm (logari thmic parametrization) in terms of the covariance matrix eigenvectors, are obtained. It is shown that to have a minimum error in the parametrizations, one need use only five parameters in the linear parametrization and three parameters in the logarithmic parametrization. The behavior of relative err ors in the resulting parametrizations is studied. The advantage of the loga rithmic parametrization is shown. The obtained parametrizations are compare d with the Lumpe independent approach and others [30]. This comparison show s an evident advantage of the parametrizations proposed in this work: their relative rms error is 1%, and the error in the Lumpe parametrization for t he same set of AEC realizations is 4.4%.