HR: 1330h
AN: B33D-04    [Abstracts]
TI: Estimating Shrub Abundance in Desert Grasslands Using Geometric-Optical Models
AU: * Chopping, M J
EM: chopping@pegasus.montclair.edu
AF: Montclair State University, 1 Normal Ave, Montclair, NJ 07043 United States
AU: Su, L
EM: su@pegasus.montclair.edu
AF: Montclair State University, 1 Normal Ave, Montclair, NJ 07043 United States
AB: This work examines the application of a geometric-optical model of canopy reflectance to provide measures of the physical structure of desert grasslands and shrublands which can provide the metrics required to expand the geographic extent over which ecological process models are able to operate. The work is aimed at estimating the structural attributes of desert grasslands and shrublands by providing a means of explaining the moderate resolution remotely-sensed signal in terms of empirically-derived background spectral directional reflectance and shrub number density. These structural attributes are important in surface hydrology and meteorology as well as for differentiating plant communities. The approach is through inversion of the non-linear simple geometric model (SGM) against red wavelength multiangle reflectance data from the Multiangle Imaging SpectroRadiometer (MISR) flown on the NASA Earth Observing System Terra satellite. These MISR data were corrected for atmospheric scattering and absorption, including estimates of aerosol optical depth and ozone. Observations from all nine cameras were used. The SGM has previously been tested against ground-based canopy maps and measurements but at the landscape scale it is necessary to account for the varying anisotropy of the soil-understory complex. While geometric-optical models often assume a Lambertian background for forest applications, the large fractional cover of exposed soil in arid regions means that this assumption cannot be made. In this study separation of background and upper canopy contributions was effected using a linear scaling of the parameters of the Walthall BRDF model with near-nadir brightness, using an optimization algorithm to adjust its parameters against MISR data for relatively dark and bright locations for which the number density and radius of the shrubs is known from high-resolution imagery. This approach makes the assumption that the understory contributes diproportionately to surface albedo than large shrubs in this environment. Shrub canopy statistics were compiled for each mapped 250 x 250 m observation using high resolution IKONOS panchromatic imagery using background subtraction and thresholding techniques. Shrubs were considered to have a minimum radius of 1 m. Inversion experiments were carried out for an area of 21 x 21 MISR multiangle observations mapped onto a 250 m grid using a direct search optimization algorithm to minimize the absolute root mean square error (RMSE) between observed and modeled bidirectional reflectance values. Constraints were imposed on the parameter space such that the retrieved canopy height is > zero and < 4 m and shrub number density is > zero. Model fits to observations generally provided a low RMSE with a mean of 0.005 (n=441). The results show that with some assumptions and constraints a reasonable relationship between retrieved and measured shrub density and fractional cover can be achieved. However the strongest relationship between modeled and observed density provided a coefficient of determination of only 0.2; more work is needed to determine whether the assumptions and approximations made are valid or too severe.
DE: 1640 Remote sensing
SC: Biogeosciences [B]
MN: 2005 Joint Assembly