Capabilities and limitations of methods for BRDF characterization in imaging spectrometry data

Anisotropic reflectance behavior is typical for all natural surfaces. The target- and wavelength-specific characteristics of this physical phenomenon may be expressed by the conceptual quantity of the Bidirectional Reflectance Distribution Function (BRDF). On the one hand, characterization of the BRDF may enable to better estimate biophysical and biochemical as well as structural parameters of the observed surface. On the other hand, reflectance anisotropy often is considered an interfering effect in airborne or spaceborne Imaging Spectrometry data. Changes in within-scene across-track radiometry, which are caused by the sensor view angle variation, can lead to misclassification or improper estimation of the surface properties of interest. Anisotropy is especially pronounced for sensor systems that feature wide field- of-view optics, as it is the case for a number of current and future instruments, both airborne and spaceborne. Accurate quantitative data analysis most often requires a normalization of the existing reflectance anisotropy, especially for the derivation of albedo products. State- of-the-art sensor systems have implemented dedicated, operational BRDF analysis steps into their data processing chain (e.g. MODIS). Full BRDF characterization requires a number of multi-angular observations; proper correction for reflectance anisotropy therefore is more error-prone and less amenable to validation in airborne single-pass imagery. This paper reports on present achievements in the analysis of both empirical, scene-based and semi- empirical methods suited for the characterization and quantification of anisotropy and its correction in Imaging Spectrometry data. It summarizes capabilities and limitations of the methods, with respect to sensor properties and acquisition geometry that determine the range of available angular observations and limit the accuracy of BRDF characterization. Furthermore, it focuses on spectral pre-classification, which has authoritative influence on the success of some of the methods. RSL's spectral database SPECCHIO contains a large number of field-measured spectra that can be both used for spectral pre-classification of data and validation of the anisotropy normalization results in the future.