Inversion of vegetation canopy reflectance based on variationnel multiscale approach

Radiative transfer modeling (RTM) is largely used in the remote sensing community to simulate the vegetation canopy reflectance as a function of the vegetation biophysical properties. Such models allow particularly for the retrieve of the latter by inversing remotely sensed data in visible and infrared domains. This estimation is not simple to develop because of measurement and model uncertainties, and also due to model non-linearity. In this paper, we adopt an iterative variational scheme to estimate leaf area index (LAI) and chlorophyll amount (Cab) by inversion of the well-known PROSAIL RTM. In addition to consider a prior information about the biophysical properties to reduce the uncertainty, we propose a multiscale approach that start by solving easier problems and increases the complexity until converging to the original problem solution. Such a technique allows to deals with non-linearity: both overcome local minima and prevent the algorithm divergence. Performance evaluation is done using PROSAIL simulated data. Particularly, model robustness against noise level is studied and promising results are shown.