Author
 |
CHOPPING, MARK |
|
Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal Publication Acceptance Date: 3/22/2000 Publication Date: 6/15/2000 Citation: N/A Interpretive Summary: The aim of this study was to test the Roujean linear semi-empirical kernel-driven (LiSK) bidirectional reflectance distribution function (BRDF) model in terms of ability to fit observed distributions and to explain observed directional phenomena. The tests were achieved with two independent ground-based azimuthal bidirectional reflectance factor (BRF) datasets acquired with field radiometers over semi-arid grasslands in Xilingol, Inner Mongolia (P.R. China), and Walnut Gulch, Arizona (USA). The results showed that the model is able to describe the observed distributions with high accuracy and with low sensitivity to number of angular inputs. Bidirectional reflectance observations in the forward-scattering direction are shown to be important in constraining model inversions. The values of retrieved parameters (kernel weights) indicated that one or more of the simplifying assumptions made in the model lderivation is likely to be too severe for modeling BRDF in the near-infrared region. The model was shown to provide a means of correcting data from wide-swath or multi-angular sensors for non-Lambertian surface scattering phenomena, allowing greater precision and consistency than hitherto possible in the retrieval of surface spectral reflectance measures for grassland monitoring over large areas with optical remote sensing methods. Technical Abstract: This study provides an assessment of the Roujean linear semi-mpirical kernel-driven (LiSK) bidirectional reflectance distribution function (BRDF) model with respect to the ability to describe and explain the directional reflectance phenomena observed over contrasting semi-rid grassland canopy-soil complexes on two continents in the northern hemisphere. Reference data are provided two independently-collected ground-based azimuthal bidirectional reflectance factor (BRF) data sets in the visible and near-infrared regions, acquired with field radiometers over semiarid grasslands in Inner Mongolia (P.R. China) in 1996 and at Walnut Gulch, Arizona in 1990. The behavior of the retrieved model parameters was investigated, and values were compared to effective biophysical parameters. The results demonstrate that the model is able to replicate observed phenomena over both the Inner Mongolia and Arizona sites with a high degree eof accuracy (root-mean-square deviations less than 0.01 reflectance and coefficients of determination greater than 0.86 in all cases). Anisotropy factors based on retrieved model parameters are shown to reduce directional dependence importantly in both datasets, an important practical application of the model. Interpretation of model parameters in terms of physical phenomena or biophysical parameters in terms of cover fraction, leaf area index, and fresh weight was problematic. Retrieval of negative parameter values indicated that the simplifying assumption of an optically thick media made in the derivation of the volume scattering kernel is likely to be too severe for modeling BRDF in the near-infrared region. A sensitivity analysis showed that model inversion is robust with respect to number of observations but requires an adequate range of acquisition angles. |
