The behavior of vertical profiles of scattering coefficients is examined based on statistical analysis of the data obtained during whole day round lidar experiments conducted in a spring-summer period. The data were separated into subsets according to the type of air masses, season, and time of the day. The behavior of autocorrelation matrices was shown to vary with altitude whereas eigenvectors demonstrated their statistical stability. Four-layer altitude model is proposed taking into account peculiarities of behavior of the scattering coefficients' mean profiles and their statistical characteristics.