This document presents a shadow detection method for remote sensing images that combines very high resolution hyperspectral and LIDAR data. The method uses line-of-sight analysis of a digital surface model from LIDAR to create a rough shadow image. Large shadow and non-shadow regions are identified and used to train a support vector machine classifier on the hyperspectral data. The classifier produces a final shadow detection result, which is then post-processed. Results show the method can surprisingly accurately identify shadows despite registration errors between the LIDAR and hyperspectral data.