應用統計分群及多期地表位移資料進行坡面分區活動性潛勢分析

Multitemporal interferometric synthetic radar (MTInSAR) is a remote sensing technique that can provide sequences of high-accuracy and wide-coverage interferograms, which can be used for analyzing transient surface deformations. MTInSAR has been used in hazard mitigation planning for potential deep-seated landslides and in long-term monitoring of slope activities in Taiwan. This study proposes to use a Gaussian mixture model for statistically clustering the surface deformation data points for further analyzing into the subzone activities of deep-seated landslides. Based on the model, associated clusters are defined to connect multitemporal deformation clusters; thus, time series of the deformation clusters can be derived. These time series deformation clusters enable investigating their relationships with other landslide activity influencing factors, such as precipitation. To demonstrate the proposed model, this study used two potential deep-seated landslide sites as examples. The results revealed that the deformation clusters and time series of the deformation clusters were indeed correlated with the subzone activities at the landslide sites. These encouraging results indicate that the proposed method can be extensively applied in the analysis of deep-seated landslides.