Validation of the Lsa Saf Fvc and Lai Prototype Algorithm with Polder/adeos Data over Western Europe

The SAF on Land Surface Analysis (LSA SAF) is aimed to produce operationally land surface products taking full advantage of the new EUMETSAT meteorological satellites by the end of its Initial Operational phase (2005-2007). Our main concern in LSA SAF is to develop the algorithm for retrieving Leaf Area Index (LAI) and Fraction of Vegetation Cover (FVC), as well as to validate the derived products before being distributed to the Users Community. The prototype algorithm is based on the complementary use of two methods: (1) an optimised Spectral Mixture Analysis (VMESMA) method for retrieving FVC (the operational version is described in García-Haro et al., this issue) and, (2) a new canopy BRDF model for heterogeneous canopies, named DISMA, which allows to retrieve both FVC and LAI simultaneously. DISMA uses prior information derived from VMESMA, which is also used as backup solution for DISMA. The aim of this work is to demonstrate the feasibility of our prototype algorithm at coarse-resolution scales before implementing it in the LSA SAF system. In particular, we aim to know how the algorithm behaves as compared with other currently-used methods for retrieving vegetation FVC and LAI fields. The prototype VMESMA-DISMA has been applied on POLDER/ADEOS data acquired in April 1997. The outcomes of VMESMA (backup) and DISMA has been compared against two different methods for retrieving FVC and LAI from POLDER/ADEOS: (1) A method based on an empirical relationship with a vegetation index and (2) a method based on the inversion of a physically-based canopy model. Results show that the spatial distribution of VMESMA and DISMA retrievals is consistent. The overall uncertainty (RMS) between DISMA and reference POLDER products is on the order of 0.1 for FVC and 0.8 for LAI, very close to the RMS found between reference POLDER products (RMS of 0.1 for FVC and 0.7 for LAI). These results show the underlying physics sound and feasibility of the proposed algorithm for retrieving vegetation parameters in LSA SAF.