Spatiotemporal Pattern Of Wenchuan Earthquake Landslides In Longmenshan Area From Time Series Radar Interferometry

The devastating Ms 8.0 Wenchuan earthquake on May 12,2008 generated a serious surface rapture extending ～300km along the Yingxiu-Beichuan fault and ～70km along the Guanxian-Jiangyou fault in the direction of northeast. The event triggered about 50,000 coseismic landslides and covered a large range of 750km2. Due to the strong impact of the event on mountainous geological environment, the slope body shatter and relaxation occurred in the disaster area. So the frequency and scale of landslide hazards in the mountainous area would tend to increase markedly in the following 3-5 years after the Wenchuan earthquake. Therefore, the monitoring and analysis on the post-seismic landslide hazards are of great significance, which could help to gain the scientific knowledge regarding to the distribution pattern and development characteristics of landslides. The study is also helpful for the reduction of hazards and the post-seismic homeland reconstruction.Spaceborne Interferometric Synthetic Aperture Radar (InSAR) is a newly-developed powerful space geodetic technique with the advantages of the all-weather and day-and-night imaging capacity, fine resolution, high precision and wide spatial coverage. However, it’s challenging to use the conventional InSAR technique for surface displacements measurements and landslide monitoring due to the complex geological and geomorphologic setting in Longmenshan mountainous area. This easily leads to both spatiotemporal decorrelation and atmospheric artifacts. On one aspect, the rugged terrain features and the rapid elevation change could cause the interferometric phase jump, making it difficult to obtain the valuable deformation information. On the other aspect, the large terrain elevation leads to the significant variation on atmospheric humidity near the surface. And the phase delay due to SAR signal propagation in different atmosphere could cause the errors of deformation measurements.To overcome above problems, the long wavelength component of atmospheric phase correction model with three dimensional geo-spatial factors is presented. The landslide distribution pattern along the main fault triggered by the rainstorm is obtained by the combination of the terrain slope factor and the LOS deformation threshold. And the mass wasting volume associated with giant landslide induced by Wenchuan earthquake and the spatial characteristics of the post-earthquake landslide also can be determined quantitatively by InSAR. In the thesis, the landslide hazards monitoring and analysis in Longmenshan mountainous area are discussed from the following aspects including the SAR image co-registration, three dimensional reconstruction from the two-sided radargrammetry, the atmospheric phase correction and the PS-InSAR technique applied to detect the feature of landslides in the mountainous area.A new co-registration method based on moving coherence surface fitting is proposed in the thesis. The position of the homonymous point is obtained by computing the maximum value of coherence fitted surface function, instead of step searching for the discrete maximum coherence value. This approach solves the co-registration technique difficulties on SAR images covering Longmenshan area, enhancing the calculation efficiency and reducing the coherence loss from the matching error. And the precision and reliability of the co-registration results are further improved.In order to improve mountainous DEM precision, the two-sided radargrammetry approach of combining the ascending and descending SAR images to form a stereo pair with a ultra-long baseline (> 1000km) is discussed in the thesis. The slant-range Doppler equations are employed to establish the stereoscopic intersection model. The quantitative functional relationship between radargrammetric elevation uncertainty and baseline-height ratio, image co-registration error is obtained. The PALSAR images stereo-pairs with the baseline up to 1080km long covering Longmenshan area are used in the study to generate DEM after the Wenchuan earthquake. The DEM accuracy assessment is performed with reference to GPS observations. The in-situ measurement results validate the feasibility and reliability of the proposed method.To investigate the spatial pattern of landslides along the major fault in Longmenshan area, the PALSAR images with short baseline obtained before and after the rainfall period respectively are utilized in the study. The atmospheric phase correction model with three dimensional geo-spatial factors is proposed to eliminate the long wavelength components. The LOS deformation threshold is set to reduce the short wavelength phase error. The terrain slope factor is introduced to analyze the spatial pattern of post-earthquake landslides triggered by the rainstorm. The results show that the active landslides present a linearly concentration distribution along both sides of Yingxiu-Beichuan fault. The statistical data suggest that almost 90% of the post-earthquake landslides are located in the elevation range of 1000-3000m and the terrain slope factor of 15～35°and 70% are on the hanging wall side of the faults. There are significant differences between the hanging and foot wall of the faults.In order to analyze quantitatively the mass wasting volume associated with giant landslide (Daguangbao landslide) induced by the Wenchuan earthquake, the available twenty interferometric pairs with good coherence from post-seismic PALSAR images covering Daguangbao area are used in the study, which were obtained during 2008～2011. The neighborhood differential PS network is established on the basis of the persistent scatterer objects. Both the ground deformation and the elevation error are solved by the least squares adjustment, thus the quantitative spatial pattern of the giant landslide is obtained. The results show that the Daguangbao landslide is with 4350m long along the NE-SW direction and an extension width of 3400m, and a peak height change of 535m in the vertical direction. The affected area of landslide mass movement reaches 7.25km2 with the volume up to 1.28 billion m3. It validates the the feasibility and the reliability of detecting quantitatively giant landslide volume by PS-InSAR.