The analytical methods for solution of some ill-posed inverse problems in application to the problem of multifrequency laser sensing of aerosol atmosphere has been reviewed briefly. It has been shown that under conditions of significant multiplescattering noise, which takes place in the case of sensing of cloud aerosol, it is preferable to use flexible, so-called intellectual, algorithms for reconstruction of the microphysical parameters of aerosol. The hybrid scheme, in which the learning procedure in the method of artificial neural networks is optimized based on the genetic search, appears to be the most efficient. The ways for further modification of the genetic algorithm with the use of specific features of the multi-wavelength lidar sensing technology have been described. The results of closed numerical simulation shown in Figures 5.15, 5.16, and 5.19 demonstrate the rather high quality of reconstruction of the a priori preset spectra of cloud droplet size distribution up to optical depths 1.5—2 times exceeding the estimates known until now.

< Prev   CONTENTS   Source   Next >