An algorithm for simultaneous inversion of aerosol properties and surface reflectance from airborne GeoTASO hyperspectral data

Friday, 19 December 2014
Weizhen Hou1, Jun Wang1, Xiaoguang Xu1, Shouguo Ding1, Dong Han1, James W Leitch2, Thomas Delker2 and Gao Chen3, (1)University of Nebraska Lincoln, Lincoln, NE, United States, (2)Ball Aerospace, Boulder, CO, United States, (3)NASA Langley Research Ctr, Hampton, VA, United States
This paper presents an inversion method to retrieve aerosol properties from the hyperspectral data collected by airborne GeoTASO (Geostationary Trance gas and Aerosol Sensor Optimization). Mounted on the NASA HU-25C aircraft, GeoTASO measures radiation in 1000 spectral bands from 415 nm to 696 nm, and is a prototype for the TEMPO (Tropospheric Emissions: Monitoring of Pollution) instrument. It flew over Houston during September 2013 and gathered several days’ of airborne hyperspectral remote sensing data for our research.

Our inversion method, which is based on the optimization theory and different from the traditional lookup table (LUT) retrieval technique, can simultaneously retrieve parameters of atmospheric aerosols such as the aerosol optical depth and other aerosol parameters, as well as the surface reflectance albedo. To provide constraints of hyperspectral surface reflectance in the inversion, we first conduct principal component analysis (PCA) using 46 reflectance spectra of various plants and vegetation to identify the most influential components. With the first six principal components and the corresponding calculated weight vector, the spectra could be reconstructed with an accuracy of 1%.

UNL-VRTM (UNified Linearized Radiative Transfer Model) is employed for forward model calculation, and its outputs include not only the Stokes 4-vector elements, but also their sensitivities (Jacobians) with respect to the aerosol properties parameters and the principal components of surface spectral reflectance. The inversion is carried out with optimization algorithm L-BFGS-B (Large scale BFGS Bound constrained), and is conducted iteratively until the modeled spectral radiance fits with GeoTASO measurements. Finally, the retrieval results of aerosol optical depth and other aerosol parameters are compared against those retrieved by AEROENT and/or in situ measurements during the aircraft campaign.