Landslide hazard analysis for Hong Kong using landslide inventory and GIS

Abstract

This paper presents a landslide-inventory-based and GIS-based framework for systematic landslide hazard analysis by employing historical landslide data in Hong Kong, coupling with geological, geomorphological, population, climatic, and rainfall data. Based on 1448 landslide data from 1984 to 1998, the diurnal and seasonal distributions of landslides are established and compared with the seasonal rainfall variation. The cumulative fatalities and injuries caused by landslides increase with the cumulative rainfall in Hong Kong, indicating a strong correlation between rainfall and landslide consequences. The average annual fatality and injury rates in Hong Kong caused by landslide are 11.35 and 11.63, respectively. In terms of being hit by a landslide, squatter areas and roads on Hong Kong Island are at the highest risk. A frequency–volume relation for Hong Kong Island was established, and, using this relation, it was estimated that the return period of a 26,000 m³ landslide (the size of 1995 Shum Wan Road Landslide) is about 3.12 years. A hazard zonation map for Hong Kong Island is established by using historical data. The potential use of GIS technology to incorporate various layers of information is illustrated using Hong Kong Island as an example. Both landslide hazard and risk maps are proposed using raster calculation.

Introduction

Landslides can manifest themselves in many different forms, including rockfalls, rockslides, debris flows, soil slips, rock avalanches, and mud-flows. Some infrequent landslides may lead to catastrophes. For example, it was estimated that 200,000 people died in 1920 from an earthquake-induced loess flow in Kansu Province China (Hansen, 1984). More recently, 4000 people died after an earthquake-induced debris avalanche occurred at north peak of Nevados Huascaran, in Peru in 1962 (Hansen, 1984). Considering the scale of these events, they are basically unpreventable. The most reliable way to prevent landslide-induced casualties and economic losses is to avoid building towns or cities in the vicinity of steep terrains. But, this is considered impracticable or impossible in many countries due to the rapid growth of human population or due to the expensive cost in relocating of ancient or historical cities. Thus, regional landslide hazard analysis and management is becoming an important task for city planners and officials.

Landslide hazard and risk analysis for a town or city normally involves a sequence of mountain ranges surrounding the area, and thus the use of hazard map is found necessary. Landslide hazard map is very useful in estimating, managing and mitigating landslide hazard for a region (Anbalagan et al., 1993; Kienholz, 1978; PIARC, 1997). Many review articles have addressed the issue of landslide hazard and risk analysis and management. They include Leroi (1997), Hansen (1984), Fell and Hartford (1997), Einstein (1988), Einstein (1997), Varnes (1984), and Hungr (1997). The scale of landslide hazard can be expressed in either regional basis, community basis or a site basis. When landslide hazard is estimated, we have to consider the historical records, the local geology (e.g. shallow or deep-seated landslides), lithology (e.g. physical and chemical behaviors of rocks and soils), structure (e.g. stratigraphic sequences, joint sets etc), geomorphology (e.g. steepness of slopes), hydrologic conditions (e.g. groundwater level), vegetation (e.g. form and type of vegetative cover), and climate (e.g. precipitation and temperature). The local stress condition is also an important factor that may relate to uplift, erosion, deposition, and groundwater fluctuation. We should also borne in mind that many of these parameters and conditions actually change with time.

Ideally, a reliable landslide hazard map should carry appropriate weights from historical landslide events, from geomorphological analysis, and from mechanics analysis of slides, falls, and flows of the earth mass. Since all three aspects of hazard analysis involve the handling and interpreting a large amount of factual, geological and simulated data, the use of computer or information technology is crucial to the success of such analysis. Since the mid 1980s, geographical information system (GIS) becomes a very popular technology used in calculating and managing natural hazards, including landslides (Coppock, 1995). Some of these GIS-based hazard analyses focus on earthquake-induced landslides (e.g. Luzi et al., 2000; Miles and Ho, 1999; Miles and Keefer, 1999; Refice and Capolongo, 2002), and some on hydrological-condition-induced landslides (e.g. Miller and Sias, 1998). GIS analysis has also been proposed to produce rockfall hazard map (e.g. Cancelli and Crosta, 1994). However, the reliability of the hazard analysis does not depend on which GIS software or platform we used but on what analysis method we employ (Carrara et al., 1999). Therefore, various methods of analysis have been proposed by many different authors (e.g. Dikau et al., 1996; Leroi, 1997; Guzzetti et al., 1999; Dai and Lee (2002a), Dai and Lee (2002b); Carrara et al., 1999).

GIS technology has been used virtually everywhere in the world, including both developed and developing countries. For example, GIS technology has been applied to examine landslide hazard in North America: Los Angeles area, USA by Jibson et al. (2000), Glenwood-Springs, Colorado, USA by Mejianavarro et al. (1994), Moresby, Canada by Davis and Keller (1997), and Idaho, USA Gritzner et al. (2001); in South America: Manizales in Central Colombia by van Westen and Terlien (1996); in Asia: Jordan by Husein et al. (2000), Yongin, Korea by Lee and Min (2001), Bhagirathi Valley, Himalayas by Saha et al. (2002), Ramganga catchment, Himalayas by Gupta and Joshi (1990), Kulekhani watershed, Nepal by Dhakal et al. (1999), Phewa Tal watershed, Nepal by Rowbotham and Dudycha (1998), and Western Ghat, India by Nagarajan et al. (1998); in Europe: northern Spain by Duarte and Marquinez (2002), Tuscany, Italy by Luzi et al. (2000), Central Italy by Carrara et al. (1991), Southern Italy by Refice and Capolongo (2002), Fabriano, Italy by Luzi and Pergalani (1996), and Southeast England by Collison et al. (2000); and in Africa: Wondogenet area, Ethiopia by Temesgen et al. (2001).

In the case of Hong Kong, GIS-assisted landslide hazard analysis has only been proposed for Lantau Island, where the new International Airport is located (Dai et al (2001), Dai et al. (2002); Dai and Lee (2001b), Dai and Lee (2002a), Dai and Lee (2002b); Lee et al., 2001). However, most of the population of Hong Kong does not locate at Lantau Island, but instead on the Hong Kong Island. Therefore, it is meaningful to consider the landslide hazard as well as the risk for Hong Kong Island, on which most of the tall buildings are located.

Therefore, the main objective of the present study is to consider more comprehensively the landslide hazard of Hong Kong, paying particular attention to Hong Kong Island and combining historical landslide data with GIS. In particular, diurnal and seasonal distributions of landslides are established. Correlation between rainfall and fatalities and injuries caused by landslides is examined. Facilities on Hong Kong Island at the highest risk are identified. The correlation between landslide occurrence and geological formation, slope angle, and slope height are investigated. A landslide hazard zonation map is also proposed based on spatial distribution of historical data. The potential use of GIS technology in incorporating geological, geomorphological, and climatic data is illustrated using Hong Kong Island. The present discussion of the use of GIS technology in the hazard analysis on Hong Kong Island should shed new light on the future works in this direction. More detailed hazard analysis will be presented in our forthcoming papers.