Polarimetric Decomposition Analysis of ALOS PALSAR Observation Data before and after a Landslide Event

Abstract

Radar scattering mechanisms over landslide areas were studied using representative full polarimetric parameters: Freeman–Durden decomposition, and eigenvalue–eigenvector decomposition. Full polarimetric ALOS (Advanced Land Observation Satellite) PALSAR (Phased Array type L-band Synthetic Aperture Radar) datasets were used to examine landslides caused by the 2008 Iwate-Miyagi Nairiku Earthquake in northern Japan. The Freeman–Durden decomposition indicates that areas affected by large-scale landslides show dominance of the surface scattering component in both ascending and descending orbit data. The polarimetric parameters of eigenvalue–eigenvector decomposition, such as entropy, anisotropy, and alpha angle, were also computed over the landslide areas. Unsupervised classification based on the H- plane explicitly distinguishes landslide areas from others such as forest, water, and snow-covered areas, but does not perform well for farmland. A landslide area is difficult to recognize from a single-polarization image, whereas it is clearly extracted on the full polarimetric data obtained after the earthquake. From these results, we conclude that 30-m resolution full polarimetric data are more useful than 10-m resolution single-polarization PALSAR data in classifying land coverage, and are better suited to detect landslide areas. Additional information, such as pre-landslide imagery, is needed to distinguish landslide areas from farmland or bare soil.