To address solar variability, ground based whole-sky imaging (WSI) techniques have been developed for solar forecasting. In WSI the red-blue-ratio (RBR) from a few clear days is used as a reference to detect clouds in the imagery throughout the year. However, due to changes in atmospheric aerosol concentration (parameterized by Aerosol Optical Thickness, AOT) and size distribution (parameterized by Angstrom Parameter, AP), the solar spectrum significantly varies between clear days. As a result, cloud detection methods are unable to distinguish optically thin clouds from clear skies with high aerosol content. In this work we examine the correlation between RBR and AOT and AP for 18 clear days in San Diego, California. The AP was only weakly correlated to RBR, while a strong correlation to between RBR and AOT was observed. Consequently, AOT can be modeled based on sky imagery. Alternatively, RBR measurements can be auto-calibrated to yield better dynamic thresholds for cloud detection in WSI.