With growing population, expansion of settlements and overexploitation in mountainous areas, catastrophic slope failures have caused extensive buildings damage and threatened human lives globally in the past decades (Dai et al., 2002; Nadim et al., 2006; Keefer and Larsen, 2007; Highland and Bobrowsky, 2008). There were more than fifty thousand people affected by catastrophic landslides in the period of 2000-2018 around the world based on the statistic of EM-DAT (Emergency Events Database) in International Disaster Database (Guha-Sapir et al., 2019). While in Taiwan, according to the database from National Science and Technology Center for Disaster Reduction (NCDR), there were over 8,000 slope failures and landslide events happened during 2000-2017, which resulted in about 970 deaths.
Many studies have investigated the basic physics of landslide triggering or reactivation, and the main factors of landslide occurrence are generally considered to be heavy rainfall and severe seismic shaking (e.g., Iverson, 2000; Keefer, 2000; Guzzetti et al., 2007; Jibson, 2007; Crosta and Frattini, 2008; Wasowski et al., 2011). Slow-moving or creeping slopes may further evolve to rapid and destructive landslides (Chigira, 1992; Dramis and Sorriso-Valvo, 1994; Kilburna and Petley, 2003; Petley et al., 2005).
doi:10.6342/NTU201901977
2
Therefore, detecting and monitoring slope deformation can provide better clues to landslide forecast and identify potential landslide sites, and are crucial in hazard risk management and assessment. The location and scale of creeping landslides could be preliminary interpreted by the features of topography from remote sensing images, with supporting information such as geological maps, slope aspects, in situ monitoring data and field observations (Kääb, 2002; Glenn et al., 2006; Kasai et al., 2009; Lillesand et al., 2014). However, there still remain difficulties for detailed characterization of slope deformation over greater spatial and temporal scales.
For the purpose of reducing casualties and property losses, not only mapping the areas susceptible to potential landslides is important but also landslide-affected area has to be considered the, as well as disaster prevention education and evacuation of protected targets for the disaster mitigation.
For landslide hazard assessments, the monitoring of potential deep-seated landslide is a significant issue. The application of Synthetic Aperture Radar Differential Interferometry (D-InSAR) to detect an unstable slope indicated the potential capability of this technique (Fruneau et al., 1996), and the development of Multi Temporal Interferometry (MTI) technique overcame the main factors including coherence loss in vegetated areas and
atmospheric effects limiting the performance of D-InSAR in landslide investigation (Hilley et al., 2004; Bovenga et al., 2006; Colesanti and Wasowski, 2006). In Taiwan, MTI techniques are also applied for the monitoring of potential landslides (Chen et al., 2017; Dong, 2017). However, because the deformation is too large to be derived from phase change in radar wave, D-InSAR and MTI techniques could hardly detect movement of landslide body with several meters.
By comparing the ground objects in remote sensing images between pre-event and post-pre-event, the horizontal displacement over 10 m could be derived.
(Lin et al.,2004; Lin et al., 2014). The limitations of comparing the ground objects are that the distribution of ground objects is uneven, and the displacement with wide distribution cannot be derived. Therefore, an alternative which can remedy the limitation of MTI techniques and ground objects has to be considered.
Particle Image Velocimery (PIV) technique could be another attempt to analyze displacement of landslides. Originally, the PIV technique was applied in the hydromechanics field (Landreth rt al., 1988; Adrian, 1991; Lecordier et al., 1994). With the advantage of analyzing displacement of whole image by cross-correlation method, the PIV technique can provide displacement of
doi:10.6342/NTU201901977
4
landslide with large movement in wide distribution. Dominquez et al. (2003) used SPOT satellite image with the cross-correlation method to measure horizontal coseismic displacement along the Chelungpu fault in the 1999 Chi-Chi earthquake. In other research work (Kääb et al.,2000; Kääb et al., 2002), the cross-correlation method was applied for monitoring glacial movement from repeated air- and spaceborne optical data. Chan et al. (2004) and Chen and Lee (2005) analyzed the surface rupture resulted from the 1999 Chi-Chi earthquake and the near-fault surface displacement for the 2003 Chengkung earthquake in eastern Taiwan with the PIV technique. Tseng et al. (2009) analyzed non-catastrophic landslide triggered by the 1999 Chi-Chi earthquake in central Taiwan with the PIV technique applied to digital aerial photos.
Slope stability and geometry of landslide subsurface are also crucial for landslide hazard assessments. For stability evaluation of a slope, accurate geotechnical site characteristics are essential. With limited site investigation data constrained by exploration techniques, subsurface information used in subsequent slope stability analysis involves uncertainty. Varnes (1978) and Carter and Bentley (1985) introduced graphical methods for inferring landslide subsurface from ground displacement. Bishop (1999) estimated the
depth of landslide sliding surface in translational slide by geometrically balancing along downslope cross-section, and this approach is similar to balanced cross-section method in structural geology (Woodward, 1989).
Elastic dislocation model is an alternative for estimating subsurface slip and commonly performed for modeling static elastic deformation. (Steketee, 1958; Okada, 1985, 1992). Nikolaeva et al. (2014) applied elastic model to invert surface displacement of a landslide estimated by InSAR and inferred the subsurface.
Slope stability analysis by finite element method (FEM) have been developed in the past years (Zienkiewicz et al., 1975; Griffiths and Lanem 1999; Griffiths and Marquez, 2007; Zheng et al., 2006, 2009; Hammouri et al., 2008; Le, 2014; Liu et al. 2015; Javankhoshdel et al. 2017; Zhang et al., 2018). With the FEM based shear strength reduction analysis, stress, strain, displacements and location of the critical slip surface could be obtained (Ugai and Leshchinsky, 1995; Duncan et al., 1996; Griffiths, 1999; Matsui and San, 1992; Xu et al., 2005; Xu and Low, 2006; Zhang and Dai, 2010; Chen et al., 2018; Liu et al., 2018; Moallemi et al.,2018). In contrast with the balanced-cross section and elastic dislocation model, strength reduction analysis based on the FEM is independent on ground displacement. This method depends on
doi:10.6342/NTU201901977
6
the properties of material and geometry of the slope. Moreover, the failure process and development of strain distribution could be shown by the FEM based shear strength reduction analysis (Duncan et al., 1996; Griffiths, 1999).
In this study, the aim is to examine the characteristics of landslide induced by the Typhoon Morakot in central Taiwan. The Typhoon Morakot influenced Taiwan during August 6, 2009 to August 10, 2009 and triggered more than 1,500 landslides. For example, the landslide event in the Shiaolin Village, Kaohsiung caused 400 deaths and the landslide body were over 25 million m3 in volume. Another example is the Yucheliao Landslide, which is in Chiayi County. In contrast with the case in the Shiaolin Village, the Yucheliao Landslide is also triggered by the rainfall brought by the Typhoon Morakot, but the landslide body almost remained and stayed on the slope (Fig.
1-1).
The purpose of this research is to estimate and characterize the Yucheliao Landslide. The work includes: (1) Cross-correlate the orthoretified aerial photographs with the PIV technique. (2) Discuss the PIV analysis to evaluate the calculated results of ground displacement. (3) The FEM based shear strength reduction analysis via numerical simulation. (4) Inverse the geometry of the landslide subsurface with ground displacement by the FEM.
In this study, there are three orthoretified aerial photos acquired from Aerial Survey Office, Forestry Bureau, and they are taken in 2001, 2007 and 2009 respectively. These photos keep the information about displacement of ground objects due to the Typhoon Morakot (Fig. 1-2). With orthoretified images, the Yucheliao Landslide can be analyzed and measured by the cross-correlation method of the PIV technique. The results of the horizontal displacement near the Yucheliao Landslide are presented as a displacement vector map. In addition, the PIV technique can show distinguishable regions by the magnitude of displacement on the study area. Based on the results from the PIV analysis, some inferences of the Yucheliao Landslide can be made.
For example, the characteristics of slope failure can be inferred, and the mechanism of the Yucheliao Landslide would be discussed by combining with other geological data, such as rock mechanics, well drilling, or geological maps.
In numerical modeling, the shear strength reduction method is applied for slope stability analysis. The Shear strength reduction method was introduced by Dawson et al. (1999), and widely applied for slope stability analysis and estimating potential landslide subsurface (Fig. 1-3). A simplified slope model for the Yucheliao Landslide is introduced in this study, and the
doi:10.6342/NTU201901977
8
result from the shear strength reduction will be examined by limit equilibrium method, which is another method applied for slope stability. With the ground displacement from the PIV analysis and the FEM based shear strength reduction method, the interpretation of the landslide subsurface will be on trial with the FEM analysis. The result from the numerical modeling can provide the stability of the Yucheliao Landslide, and the deformation mechanism of the landslide block can be examined with the ground deformation data.
Figure 1-1. The main scarp and lateral scarp are obvious in the photo taken near Yuchelaio landslide (Lin et al., 2014). This photo is taken in the direction toward north.
Figure 1-2. Aerial photo taken after Typhoon Morakot overlain on the one taken in the pre-event period. Green lines and green polygons represent roads and construction before failure, and red lines and red polygons represent moved roads and constructions after failure.
doi:10.6342/NTU201901977
10
Figure 1-3. Slope model which Dawson et al. applied strength reduction method.
(a) Numerical mesh of the slope model composed by homogenous soil. (b) Velocity field at collapse, along with critical log-spiral surface. (Dawson et al., 1999)