A proper determination of the BRDF is of interest for land surface studies in different topics such as albedo estimation, correction of anisotropy effects, and retrieval of vegetation parameters by defining optimal geometries. In this paper, we evaluate a set of parametric models widely-used for BRDF characterisation (Roujean model, Ambrals combinations, non-linear RPV and the empirical Walthall's model). These models are inverted and tested against atmospherically-corrected BRF measurements acquired with the CHRIS (Compact High Resolution Imaging Spectrometer) instrument on board the PROBA (Project for On-Board Autonomy) satellite over an agricultural test site located in Barrax (Spain) during the SPARC (SPectra bARrax Campaign) 2003 campaign. The study area presents different land crops with high variability in LAI values from 0 to 6.
The objectives of the present study are to determine how well the different parametric BRDF models are able to fit CHRIS/PROBA's observed multiangular reflectances in order to determine the nadir-zenith reflectance, which is the optimal geometry to retrieve the fractional vegetation coverage (FVC), and to describe the anisotropy of vegetation canopies, which can be useful to estimate accurately the leaf area index (LAI). To do so, performance indicators are obtained for the different models. The results of this study show that all the tested models are fairly accurate in the entire spectral range (RMS<0.016 at 674 nm and RMS<0.025 at 803 nm) and thus are suitable for normalisation purposes. However, most of them are not able to describe BRDF features such as the hot spot, which hampers the use of these models for exploiting the directional information. There are no significant differences, for the experimental conditions, among those evaluated although the best models appear to be the linear Ross-Li model (low RMS) and the non-linear RPV model (more realistic BRDF).