The application of InSAR time series for landcover classification

Considering the wide coverage, the transparency from climate condition, Interferometric Synthetic Aperture Radar (InSAR) possesses a great potential for the landcover classification as shown in many precedent researches. In addition to the capability of InSAR signatures such as amplitude and phase coherence, the temporal variations of them provide highly useful information to discriminate the landcover classes of the target area. Therefore we devise a landcover classification scheme using the time series analysis of InSAR pairs. Combining the temporal signatures by the non linear dimension compression, it was expected to provide a highly reliable information to interpret landcover. The prototype implementation was tested in Mountain Baekdu located on the border between North Korea and China and Phased Array type L-band Synthetic Aperture Radar (PALSAR) of Advanced Land Observing Satellite (ALOS). The algorithms combining kernel Principle Component Analysis (kPCA) to analyze the phase coherence and backscattering coefficient was established. The results using InSAR signatures from two polarization modes of ALOS PALSAR were validated compared with Moderate resolution Imaging Spectrometer (MODIS) landcover classification and showed good reliability for mining landcover and vegetation information. In addition, the potential with C band ENVISAT ASAR packet composition were further examined. It will be upgraded for the application of next generation C and L band spaceborne SAR imagery.