Discrimination of peatlands in tropical swamp forests using dual-polarimetric SAR and Landsat ETM data

This study investigates the potential of dual-polarimetric TerraSAR-X data to characterise different peatlands in tropical swamp forests of Central Indonesia. Discrimination of peatlands is important to assess the carbon content below the ground. The HH and VV polarisations of TerraSAR-X data are extracted from the single look slant range complex images. H/α/A (entropy/alpha/anisotropy) decomposition of the covariance matrix is used in this study. Our results show that scattering entropy/α-angle plane segmentation of sampled peatlands data was highly overlapping, except for very shallow peat in sparse forest. To estimate the importance of radar backscatter and its polarimetric properties for distinguishing between different peatlands, we used the canonical discriminant analysis method. We found that entropy and anisotropy were more important than the backscatter values and α-angle for discriminating the peatlands. Different inputs combinations of synthetic aperture radar (SAR) and enhanced thematic mapper (ETM) data were used for experiments to classify using the maximum likelihood classification method. We found that the classification using the backscatter values and entropy of SAR data combined with ETM bands 1, 2, 3, 4, 5 and 7 yielded 80% accuracy. Post-processing of the classified image increased the overall accuracy to 87% with κ = 0.85.