A LUT approach for biophysical parameter retrieval by RT model inversion applied to data from wide field of view sensors.

Abstract

Since many years scientists have been working on techniques for retrieving biophysical and bio-chemical canopy parameters from remote sensing data. Recently the inversion of canopy reflec-tance models has been well established as a reliable alternative to the classical empirical ap-proaches. However, still a few obstacles have to be taken for making this technique fully opera-tional. In this context a lookup table approach was developed for the inversion of the combined leaf and canopy reflectance model PROSPECT+SAILH. The approach contains some elements that enable the use of scene specific information, such as the option of performing the inversion per user-defined land cover class. This may lead both to a more reliable parameter estimation as well as to a reduced calculation time since the algorithm does not have to take into account all possible reflectance cases within a scene. An important aspect of the inversion algorithm is the possibility to incorporate pixel based information on view geometry thus using the information con-tained in BRDF signatures instead of correcting for it. The model was tested on HyMap hyperspectral imagery of summer 2003 taken of the area sur-rounding Lake Waging-Taching in Southeast Germany. Scan angle and azimuth information con-tained in a file resulting from parametric geocoding was used to assess the spatial variation of view geometry to the separate pixels. Compared to the case in which only nadir view information is used, the model performed significantly better in the sense that it eliminated biased parameter estimations induced by BRDF effects. A comparison of parameter values estimated from data of the same area, but with different flight and sun configurations, shows that the results are much more consistent when view geometry is taken into account, even though measurements were not taken in the principal plane, and were supposed to be little affected by directional influences. A validation with measured field parameters is yet to be performed.