Hyperspectral and multispectral remote sensing of aerosols based on surface spectral reconstruction by PCA

Abstract

A series of studies of hyperspectral remote sensing had been carried out to develop a hyperspectral remote sensing technique for aerosol retrieval in the previous works, including the theoretical framework, information content analysis and application to the real data, in which a hyperspectral inversion algorithm was developed to simultaneously retrieved the aerosol and surface properties, and the surface reflectance spectra were decomposed into different principal components, thus only several weighting coefficients of principal components (PCs) were needed to be retrieved. In this study, based on the optimal estimation (OE) framework, we extend the OE-based hyperspectral inversion algorithm to multispectral remote sensing, and the synthetic multispectral intensities of Polarized Scanning Atmospheric Corrector (PSAC) centered in 410, 443, 555, 670, 865, 1610 and 2250 nm are used to test the inversion framework. Principal component analysis (PCA) has been conducted for the spectral dataset of 4 typical surface types with 7 channels of PSAC, in which the PC’s contribution and spectra, the spectral reconstruction results and constraints of PC’s weighting coeffects are discussed. Unified Linearized Vector Radiative Transfer Model (UNL-VRTM) is used as the forward model, and 1% Gaussian distribution errors has been added to the simulated radiance at the top of the atmosphere for multispectral inversion test. The iterative process of multispectral normalized intensities and the reconstructed surface reflectance during the OE iteration are investigated, and the normalized cost function values are well convergent. This study can provide key support to the development of OE-based inversion algorithms for multispectral remote sensing