Vol. 13, issue 10, article # 9

Abstract:

Using an array of sensing data obtained with a micro-pulse aerosol lidar data as an example, two spectral statistical methods of data processing are compared. The methods compared are the non-parametric method that uses a fast Fourier transform (FFT) and the parametric one based on a model of "autoregression moving average" (ARMA). The calculations have been carried out following the two-channel spectral estimation scheme by the Nuttall-Strand method for the ARMA model. Spatiotemporal distributions of the coherency and phase spectra were calculated from spatiotemporal distribution of atmospheric aerosol scattering coefficient. The advantages of parametric approach ensuring more high frequency resolution and higher accuracy of obtained spectral estimates are shown. At interpretation of aerosol lidar data, the coherency spectrum was indicative of the regions in space, where temperature inversion could happen. The phase spectrum makes it possible to detect zones in the troposphere of slowly ascending and descending aerosol inhomogeneities. It is proposed to start spectral processing of lidar data with the ARMA model of the second order to obtain smoothed spectral estimates.