A soft computing approach to the elaboration of satellite data

In this work a soft computing approach, based on a neural network methodology, is presented, aimed to retrieve atmospheric parameters of meteorological interest such as temperature, water vapour and ozone profiles from high resolution infrared sensor spectra. Specific neural network has been developed on basis of the specifications of the Infrared Atmospheric Sounding Interferometer (IASI). The performance of the neural network based inversion methodology has been evaluated by considering an inversion case in which test cases are retrieved. Given the nature of available data, a preliminary elaboration of spectra by means of principal component analysis has made, and an online adaptive pruning of the network during training is performed. The adaptive pruning is based on a bi-local technique and allows to optimize the network size and architecture.