HR: 1340h
AN: A23C-0808    [Abstracts]
TI: Monitoring Aerosol Optical Properties Over the Mediterranean From SeaWiFS Images Using a Neural Networks Inversion
AU: * Jamet, C
EM: cjamet@eos.ubc.ca
AF: Department of Earth and Ocean Sciences,, University of British Columbia, 6339 Stores Road,, Vancouver, BC V6T 1Z4 Canada
AU: Moulin, C
EM: Cyril.Moulin@cea.fr
AF: LSCE/IPSL, bat, 709, L'Orme des Merisiers, Gif-sur-Yvette Cedex, 91191 France
AU: Thiria, S
EM: thiria@lodyc.jussieu.fr
AF: LODYC/IPSL, tour 45-55, case courrier 100, Universite Pierre et Marie Curie, 4 place Jussieu, Paris Cedex 05, 75252 France
AB: The SeaWiFS archive provides a unique opportunity to study aerosol optical properties over oceans since October 1997. Standard SeaWiFS aerosol products are however not suitable because optical thicknesses are limited to 0.35 and Angstr\"om coefficient to 1.5. We developed an inversion based on neural networks to retrieve both optical thickness and Angstr\"om exponent from SeaWiFS read and near infrared channels. Neural networks are capable of approximating non-linear inverse functions and of processing efficiently large amounts of data. Neural networks were trained with radiative transfer computations for wide ranges of optical thickness and Angstr\"om exponent. All SeaWiFS images of the Mediterranean for years 1998, 1999 and 2000 were processed and monthly mean maps of aerosol optical thickness and Ansgtr\"om exponent were derived. A comparison with ground-based measurements at three AERONET stations in the Mediterranean for the year 2000 shows the good accuracy of the method, as well as the improvement compared to operational SeaWiFS aerosol products. The neural networks are able to retrieve high values of the optical tickness and of the Angstr\"om exponent.
DE: 0305 Aerosols and particles (0345, 4801)
SC: Atmospheric Sciences [A]
MN: 2004 AGU Fall Meeting