ESTIMATION OF VEGETATION PARAMETERS FROM POLARIMETRIC SAR DATA

摘要

This work presents the analysis of the capability to use the radar backscatter coefficient in semi-arid zones to estimate the vegetation crown in terms of Leaf Area Index (LAI). The research area is characterized by the presence of a pine forest with shrubs as an underlying vegetation layer (understory), olive trees, natural grove areas and eucalyptus trees. The research area was imaged by an airborne RADAR system in L-band during February 2009. The imagery included multi- parameter radar images. All the images were fully polarized i.e., HH, VV, HV polarizations. We measured LAI using the (DELTA)T Sun Scan Canopy Analysis System. Verification was done by analytic calculations and digital methods for the leaf's and needle's surface area. In addition, we estimated the radar extinction coefficient of the vegetation volume by comparing point calibration targets (trihedral corner reflectors with 150cm side length) within and without the canopy. The radar extinction in co- polarized images was approx27dB and approx24dB for pines and olives respectively, compared to the same calibration target outside the vegetation. The variability among corner reflectors in the open was +/- 3dB. We used smaller trihedral corner reflectors (41cm side length) and covered them with vegetation to measure the correlation between vegetation density, LAI and radar backscatter coefficient for pines and olives under known conditions. Reversed correlation between the radar backscatter coefficient of the trihedral corner reflectors covered by olive branches and the LAI of those branches was observed. The correlation between LAI and the optical transmittance was derived using the Beer-Lambert law. In addition, comparing this law's principle to the principle of the radar backscatter coefficient production, we developed a model associating the radar backscatter coefficient with the LAI.