Overlapping multi-angular spectral reflectance images were collected over desert grassland and grass-shrub transition communities (USDA, ARS, Jornada Experimental Range, Las Cruces, NM) using a DuncanTech multispectral digital camera mounted on a tilting bracket installed in an ARS research aircraft. Flightlines were oriented in the principal plane with three flights providing observations at three solar zenith angles (37 deg, 47 de and 60 deg). The 0.65um band images were calibrated to spectral radiance (W m2 um-1 sr-1) using laboratory measurements, corrected for atmospheric attenuation using the 6Sv4.2 code and resampled to a 2m UTM grid. The surface reflectance estimates were checked with air- and ground-based observations of 8m2 reference tarps. The images were convolved with a 50m2 pseudo-Gaussian point spread function and sampled at 25m to simulate samples from a sensor with a larger instantaneous field-of-view to properly ycapture the bidirectional reflectance distribution function (BRDF; intrinsic surface length scales are greater than 10m). A BRDF model based on geometric optics and volume scattering functions was developed to account for variation in directional reflectance signal with viewing and illumination angles. It was adjusted to the set of multi-angular observations for each location using the Hooke and Jeeves algorithm (1961), inverting the model for 2 (facet area index and plant protrusion density) and 3 parameters (facet area index, plant protrusion density and width) simultaneously and for these parameters plus a crown shape factor with stepwise inversion of each parameter sequentially. The retrieved parameter maps were assessed for value distributions and compared with 1m panchromatic and 4m vegetation index images from the IKONOS satellite.