This study tested the ability of quad-polarimetric L-band SAR data and polarimetric SAR features aiming at identifying forest degradation events on tropical peat swamp forests in SE Asia region. This study specifically considers the peatland forests in Kampar Peninsula, Riau Province, Sumatera, characterized with different forest disturbance, from forest plantation and oil palm concessions. Radar backscatter data (i.e. HH, HV, VH and VV), SAR polarimetric decomposition features (i.e. alpha angle, entropy and anisotropy), ratio of volume ground scattering amplitude and combined scattering matrix element values were used as ancillary data of the classification. Applying Maximum likelihood classification (MLC) method, the SAR classification yielded 77.8% of accuracy combining radar backscatter, polarimetric features, ratio of volume-ground scattering (RVOG_mv) and joint elements intensity (span_db). Multi-layer perceptron neural network (MLP-NN) classification outperformed the MLC method in terms of classification accuracy with 79.9% of overall accuracy using a combination of SAR backscatter and multi-spectral Landsat TM bands (Band 4,5,7) in the classification.
Topic: peatlands, swamps
Publication Year: 2013
Source: Proceedings of ACRS 2013 : 1-8 pages