Neural networks for the dimensionality reduction of GOME measurement vector in the estimation of ozone profiles

• Published in: Journal of quantitative spectroscopy & radiative transfer Vol. 92; no. 3; pp. 275 - 291
• Main Authors: Del Frate, F., Iapaolo, M., Casadio, S., Godin-Beekmann, S., Petitdidier, M.
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01-05-2005; Elsevier
• Subjects: Atmospheric and Oceanic Physics; GOME; Neural networks; Ozone profiles retrieval; Physics; GOME; Neural networks; Ozone profiles retrieval; Ozone atmosphérique
• ISSN: 0022-4073; 1879-1352
• DOI: 10.1016/j.jqsrt.2004.07.028

Dimensionality reduction can be of crucial importance in the application of inversion schemes to atmospheric remote sensing data. In this study the problem of dimensionality reduction in the retrieval of ozone concentration profiles from the radiance measurements provided by the instrument Global Ozone Monitoring Experiment (GOME) on board of ESA satellite ERS-2 is considered. By means of radiative transfer modelling, neural networks and pruning algorithms, a complete procedure has been designed to extract the GOME spectral ranges most crucial for the inversion. The quality of the resulting retrieval algorithm has been evaluated by comparing its performance to that yielded by other schemes and co-located profiles obtained with lidar measurements.