Open AccessNeural network for aerosol retrieval from hyperspectral imagery
Author(s) -
Steffen Mauceri,
B. C. Kindel,
Steven T. Massie,
P. Pilewskie
Publication year - 2019
Publication title -
atmospheric measurement techniques
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.679
H-Index - 88
eISSN - 1867-8548
pISSN - 1867-1381
DOI - 10.5194/amt-12-6017-2019
Subject(s) - modtran , aerosol , single scattering albedo , radiative transfer , albedo (alchemy) , hyperspectral imaging , remote sensing , environmental science , atmospheric radiative transfer codes , atmospheric correction , a priori and a posteriori , artificial neural network , atmospheric sciences , meteorology , computer science , radiance , reflectivity , geology , artificial intelligence , physics , optics , art , philosophy , performance art , art history , epistemology
. We retrieve aerosol optical thickness (AOT) independentlyfor brown carbon, dust and sulfate from hyperspectral image data. Themodel, a neural network, is trained on atmospheric radiative transfercalculations from MODTRAN 6.0 with varying aerosol concentration and type,surface albedo, water vapor, and viewing geometries. From a set of testradiative transfer calculations, we are able to retrieve AOT with a standarderror of better than ±0.05. No a priori information on the surfacealbedo or atmospheric state is necessary for our model. We apply the modelto AVIRIS-NG imagery from a recent campaign over India and demonstrate itsperformance under high and low aerosol loadings and different aerosol types.