Study On Landslide Recognition In Minjiang Basin Based On MTI Technology

In the vast western part of China, landslide survey is very difficult because of cloud、 rain、bad traffic, and so on. In order to overcome these limitations, this paper uses ENVISAT ASAR data to explore the Stanford Method for Persistent Scatterers/Multi-Temporal Interferometry(StaMPS/MTI)-based landslide recognition theory and method. Minjiang Basin is chosen as the study area to analyze the spatial pattern of Coherent Target(CT). Besides, landslide extraction model is established, and the precision test of landslide recognition is conducted. Here are the results and conclusions:(1) The distribution pattern of CT calculated from ascending and descending pass data is distinctively different, and CTs extracted are more favorable in monitoring slope facing east and west respectively. A "CT visibility map" is created by incorporating SAR imaging geometry and terrain feature, which shows that the density of CT is well in accordance with CT visibility zone. By using only ascending or descending geometry, CTs cannot cover the whole study area, because about 40% of the study area has few or none CT, which is the radar blind space.(2) Based on SAR image and land use data, the interferometric characteristics of typical landslide are analysed. Compared with grassland、forestland、waters、habitation and farmland, landslide has relative low value in back scattering map, and high value in differential map of coherence coefficient(cc35-cc70). In combine with 35 days’ coherence coefficient map, a landslide extraction model can be established by setting proper threshold values.(3) Overlaying the MTI-based deformation rate map and the slope-unit enables the zonation of "CT anomalous area", and the "potential landslide" can be identified by wiping out those areas where landslide cannot happen. On this basis, landslide recognition and boundary determination can be realized with the landslide extraction model and mean back scattering map. In the study area with an extension of 437.92 square kilometers,18 potential landslides are detected.(4) It can be inferred from the accuracy test that:①Cross validation based on different pass SAR data reveals that the "CT anomalous area landslide recognition" method proposed is somewhat universal, and the extraction model can be applied to SAR data of different pass.②Verification of morphology based on optical image shows that the landslide accuracy rate is 13/18, and the false positives may be attributed to the motion of dangerous rock body.③Verification of location based on historical landslide points proves that the proposed method has recognized one third of known landslides, three new landslides are detected. Others are located in the blind space of radar or in dormant state. In other cases, the density of CT is relatively low.