n . 1 A framework for estimating unresolved spectral shade

Spectral Mixture Analysis (SMA) is a standard way of analyzing spectral images in terms of fundamental components of the scene. For images in reflected sunlight, much of the image variance is caused by lighting variations shadowing and photometric shading that is accounted for by using a shade endmember located close to the origin in a spectral DN space. Under control of the lighting and viewing geometry, shade mixes with the tangible spectral endmembers such as soil and green vegetation to produce the observed spectral radiances. In many scenes, the landscape is vegetated and shade comprises topographic shading and shadowing ("hillshade"), which results from unresolved shadows cast by the canopy ("treeshade") and shadows cast by elements of the canopy ("leafshade"). Hillshade is commonly estimated using digital elevation models (DEMs) and assuming unvegetated surfaces are Lambertian. Deviations from hillshade include treeshade and leafshade. In general, we use LiDAR DEMs with 1-m resolution to model hillshade (“bare earth” or “last arrival”) and treeshade ("first arrival” minus bare earth). In this study of a low-relief forested area in Maryland, USA, we use LiDAR to estimate treeshade and SMA to calculate the shade endmember fractions for an ASTER image of the same area taken near the same time of year (leaf-on). The differences between the LiDAR-based model and the shade image are used to parse shade into its basic constituents and give the first remote-sensing estimates of the relative magnitude of leafshade and treeshade in a forest dominated by deciduous trees.