Multivariate analysis of solar spectral irradiance measurements

Principal component analysis is used to characterize approximately 7000 dow nwelling solar irradiance spectra retrieved at the Southern Great Plains si te during an Atmospheric Radiation Measurement (ARM) shortwave intensive op erating period. This analysis technique has proven to be very effective in reducing a large set of variables into a much smaller set of independent va riables while retaining the information content. It is used to determine th e minimum number of parameters necessary to characterize atmospheric spectr al irradiance or the dimensionality of atmospheric variability. It was foun d that well over 99% of the spectral information was contained in the first six mutually orthogonal linear combinations of the observed variables (flu x at various wavelengths). Rotation of the principal components was effecti ve in separating various components by their independent physical influence s. The majority of the variability in the downwelling solar irradiance (380 -1000 nm) was explained by the following fundamental atmospheric parameters (in order of their importance): cloud scattering, water vapor absorption, molecular scattering, and ozone absorption. In contrast to what has been pr oposed as a resolution to a clear-sky absorption anomaly, no unexpected gas eous absorption signature was found in any of the significant components.