Research Background and Motivation

Chapter 1 Introduction

1.1 Research background and motivation

Motivated by the tremendous loss and damages of Taiwan due to natural disasters caused by the vulnerable physiographic environment (NFA, 2012) and the newly availability of airborne LiDAR technology due to the launch of a national airborne LiDAR mapping program (Liu and Fei, 2011) and advances in the researches on the use of LiDAR in landslide investigations (Jaboyedoff et al., 2010&2012), this study is devoted to explore the applicability of airborne LiDAR data for investigation of landslides in Taiwan.

Nearly three-quarters of the territory of Taiwan, and 95% of its population, are exposed to frequent natural hazards (Dilley et al., 2005). In the aftermath of Typhoon Morakot, which dramatically affected southern Taiwan on August 8, 2009, and August 9, 2009, and caused the worst flooding in a century, authorities realized that the country is lacking detailed, accurate, and current elevation data and aerial imagery covering the entire territory of 36 000 km². To address this problem, a national mapping program, spanning 2010 to 2015, was launched to capture an entire territory of the country with airborne LiDAR (Light Detecting And Ranging) and digital imagery (Liu and Fei 2011). A LiDAR DEM (Digital Elevation Model) and DSM (Digital Surface Model) and color orthophotos represent a core part of this national spatial data infrastructure.

2

Taiwan is located on the active collision zone between the Eurasian plate and the Philippine Sea plate. Mountains have a high slope and high relief, and rock formations are highly fractured and fragile. These physiographic settings are unfavorable to slope stabilities. Taiwan is also located on the path of typhoons in northwest Pacific area. Torrential rainfall during the typhoon season often triggers geological hazards. Landslides are one of the most important primary disasters.

In Taiwan, a typhoon can trigger hundreds, even thousands, of shallow landslides in mountainous areas (Lin and Jeng, 2000; Cheng et al., 2005; Lin et al., 2006). These landslides can deliver large amounts of sediment into local reservoirs, reducing their water storage capacity (Dadson et al., 2004; Mikos et al., 2006). In addition, the turbidity of the water in the reservoirs has a negative effect on the sustainable operation of water supply reservoirs. The assessment and inventory of landslides is essential for effective watershed management and sustainable development. However, because of the steep terrain in Taiwan’s mountainous watersheds, most landslides are unreachable. The detailed topographic mapping required for emergency mitigation measures cannot be completed within a short period using conventional on-site surveying. Therefore, improving the efficiency and accuracy of landslide monitoring and mapping using remote sensing techniques has become an important research issue (Liu, 1987; Raju and Saibaba, 1999; Rau et al., 2007; Borghuis et al., 2007; Herva et al., 2003).

3

In planning optimal measures of disaster mitigation, researchers often use remote sensing images and digital elevation models to map disaster features and to predict disaster susceptibility. During or immediately after a disaster event, ground survey or photogrammetry, in addition to remote sensing images, can be used to obtain detailed topography data of the subjected area. Because of its ability to obtain high-density point clouds and direct geo-referencing, LiDAR can be used to obtain a more accurate and detailed topographic survey. LiDAR generates accurate 3D coordinates of discrete measurements. Subsequently, DEM and DSM can be produced with high efficiency. In tropical and sub-tropical zones of Taiwan, most of the terrains are covered by dense forestry.

Ground surface would be normally predicted by the surface of canopy in photogrammetry if the ground points cannot be seen from two different perspectives of a stereo-pair. One of the most important advantages of airborne LiDAR compared with conventional photogrammetry is that photogrammetry requires two different lines of sight to both see the same points on the ground from two different perspectives, but LiDAR only needs a single laser pulse to penetrate through the trees to measure the ground beneath. This means that LiDAR will have far fewer areas where the terrain is obscured by trees that block the lines of sight. The images of bare ground before and after the event are thus derived from LiDAR surveys to understand changes in the landscape and their possible consequences. The geomorphometric features become good tools for landslide detection, and are adopted in this study.

4

The general feature of a rainfall-induced landslide on aerial photograph is a fresh landslide scar with an elongated shape located on a relatively steep slope.

Landslides can occur in any kind of geology, as there are some weathered overburdens on steep slopes. In aerial photographs, landslide features include a bright tone, bare surface, and the other features shown in Table 1. Manual interpretation uses both 2D and 3D features of the landslides for recognition: 2D features include tone, location, and shape, and 3D features include location, direction, slope, and shadow effects. A sound consideration of the automation of landslide recognition should consider all these aspects.

Geomorphometry is a major concern in manual interpretation.

Geomorphometry, also known as geomorphological analysis, terrain morphometry, terrain analysis, and land surface analysis (Hengl and Reuter, 2009), is the science of quantitative land surface analysis. The purpose of geomorphometry is to extract surface parameters and objects using input from digital terrain models. Pike (1988) used a dozen groups of parameters as terrain descriptors by manually digitized digital terrain models. Pike used the resulting

"geometric signature or topographic signature" to categorize terrain characteristics, and suggested the degree of landslide danger. Topographic signature of life and their processes are deemed to be strongly influenced by biota (Dietrich and Perron, 2006). Guth (2001&2003) used terrain fabric as measures of a point property of the digital terrain models and the underlying topographic surface. This technique is also called topographic fingerprinting

5

(Densmore and Hovius, 2000), and determines the location of a landslide on the slope. State-of-the-art technology such as high resolution satellite images, digital aerial photography, and airborne LiDAR has opened a new era in the automation of landslide recognition, especially the possibility of applying geomorphometrics. The extraction of land surface parameters is becoming increasingly attractive for both stochastic and process-based modeling, as it makes use of all the levels of detailed digital terrain models. Topographic-based analyses can be used to objectively delineate landslide features, generate mechanical inferences about landslide behavior, and evaluate recent landslide activity (Glenn et al., 2006; Mckean and Roering, 2004). Surface roughness derived from LiDAR DTM allows the objective measurement of landslide topography. Eigenvalues of surface normals are an effective parameter for differentiating shallow landslides and debris flows (Woodcock, 1977). Expert knowledge of the geomorphometric properties of landslides may be required to establish an automatic interpretation method. High resolution and high accuracy LiDAR DEM and DSM and orthophotos are now basic constituents of NSDI in Taiwan (Liu and Fei, 2011). Therefore, it is high time to further apply geomorphometry in active landslide study (Liu et al., 2009).

Airborne LiDAR make it possible to map and evaluate landslides in a survey type of regional level (typically at scales ranging from 1:10,000 down to 1:4,000,000 or even smaller), whereas the accuracy can be as good as or community level or site-specific level (typically vary from 1:1,000 to 1:10,000).

6