Desert landscape scene simulation with simple geometric and radiosity models

Previous attempts to model landscape-level remotely-sensed scenes over Chihuahuan Desert grass-shrub transition zones using the simple geometric model (SGM) with measured large shrub densities did not result in good matches with multi-angle observations from the air. One of the findings of a previous study was that the understory of small shrubs and forbs such as broom snakeweed (Gutierezzia sarothrae) plays an important role in determining brightness. This has been further explored in modeling studies and shown to be important when using either simple geometric or radiosity modeling techniques. Good matches with observations were found using both techniques when models were driven by field-measured parameters (plant heights, radii, densities, and a background soil BRDF) and explicit plant maps constructed from air photograph and field survey data. The work described here extends modeling to the landscape level (over an area almost 1 km/sup 2/). This is achieved by estimating understory (small plant) density via a relation with mean image grayscale values masked for large plants in scanned aerial photographs and IKONOS panchromatic imagery. The results indicate that good representations of remotely-sensed scenes can be obtained by modeling as long as the soil anisotropy and the understory are taken into account. Future work must address improving estimates of understory density.