A Multi-Angle Approach to Mapping Forest and Shrub Canopy Structure in the Southwestern United States

TitleA Multi-Angle Approach to Mapping Forest and Shrub Canopy Structure in the Southwestern United States
Publication TypeConference Paper
Year of Publication2006
AuthorsChopping M., Su L., Laliberte, Andrea S., Rango A., Moisen G., Martonchik J.V.
Conference NameNASA Biodiversity and Terrestrail Ecosystem Joint Workshop
Date PublishedAugust 21-25, 20
ARIS Log Number199443
KeywordsMISR, multi-angle, red wavelength, shrub canopy, southwestern United States
AbstractRed wavelength data from NASA's Multi-angle Imaging SpectroRadiometer (MISR) acquired at a nominal 275 m and in nine viewing directions were used to invert a simple geometric-optical (GO) model to retrieve canopy structure parameters over parts of S.E. Arizona and S. New Mexico (>159,556 km2). This area encompasses desert grassland, often with woody shrub encroachment; riparian woodland; and upland forest. The combined soil and understory signal -- represented by the Walthall model -- was estimated a priori using regression relationships with MISR nadir data and the red band isotropic, geometric, and volume scattering kernel weights of the LiSparse-RossThin semi-empirical, kernel-driven bidirectional reflectance distribution function model, using measurements extracted from Ikonos panchromatic imagery over 19 locations in a grass-shrub transition zone with contrasting upper/lower canopy configurations. The GO model was adjusted using the Praxis minimization algorithm and the merit function min(|RSME|), with no constraints. Distributions of crown cover and mean canopy height for forested areas show good matches with maps from the USDA Forest Service developed from field survey as part of the Forest Inventory Analysis (FIA). Within upland forest, the mean canopy height map shows a better match with the corresponding FIA map than the cover map. Some areas with known shrub cover are predicted to have low or no woody plant cover, indicating a need to adjust the background calibration. Retrievals are very rapid almost 3 million inversions were completed in < 15 minutes allowing application of this method over very large areas.