Multispectral Analysis of Aerosols Over Oceans Using Principal Components

Abstract

Applying principal component analysis (PCA) to one month of Moderate Resolution Imaging Spectroradiometer (MODIS) narrow-band short-wave radiance data and comparing with the Goddard Global Ozone Chemistry Aerosol Radiation Transport (GOCART) model simulations, we show that aerosol size and speciation information can be inferred from multispectral radiance information without having to use other parameters, such as a fine mode fraction (FMF), that are difficult to validate. PCA was applied to seven highly correlated MODIS solar channels (0.47, 0.55, 0.66, 0.86, 1.24, 1.64, and 2.12 mum) to extract noncorrelated pseudochannels, each with a unique interpretation. The first pseudochannel (PCI) can be interpreted as the mean radiance across the seven channels, which is directly proportional to the aerosol concentration. The second pseudochannel (PC2) is sensitive to the aerosol size since different aerosol types scatter and absorb differently across the seven MODIS short-wave channels. PC3 is inversely related to the aerosol optical thickness (AOT) and the FMF and appears most sensitive to changes in sulfate and maritime sea-salt concentrations. Results indicate that high values of PCI are indicative of high dust aerosol concentrations comprising more than 40% of the total AOT, whereas high values of PC2 indicate anthropogenic aerosol concentrations (deduced from GOCART) in excess of 60%. Compared to simple 0.55-mum FMF thresholds, the PC channels are much more sensitive to dust aerosol concentrations and certain aspects of anthropogenic aerosols, with very low FMF values alone (< 0.2) being the best indicator of predominately sea-salt aerosol concentrations. Our results indicate that PCA could be used as an alternate method for inferring aerosol speciation information in future research over ocean and more complex land surfaces.