CLASSIFICATION OF RADAR IMAGERY OVER BOREAL REGIONS FOR METHANE EXCHANGE STUDIES

Airborne synthetic aperture radar (SAR) data acquired over Alaska are used to investigate the ability of SAR to distinguish between land cov er classes of differing methane exchange rates. Land cover within the study area is divided into four classes: forest, bog, water, and fen, with fen having the highest methane emission. Accurate classification is achieved using both statistical and neural network techniques appli ed to fully polarimetric L- and C-band data. Similar classification ac curacies are also obtained using non-polarimetric subsets of the data, analogous to data that would be available by combining SAR observatio ns from ERS-1/2, JERS-1 (Fuyo-l), and RADARSAT. Accurate classificatio n of fens, however, is possible only when the non-polarimetric subset includes L-band data.