APPLICATION OF REMOTE SENSING DATA TO LANDSLIDE
MAPPING IN HONG KONG
V. K. Vohora and S. L. Donoghue
Department of Earth Sciences, The University of Hong Kong, Pokfulam Road, Hong Kong, SAR, China
vijay@hkucc.hku.hk, donoghue@hkucc.hku.hk
Commission IV, WG IV/6
KEY WORDS: Landslides, Vegetation, Environment, Multispectral, High resolution, Correlation
ABSTRACT:
Rapid encroachment of Hong Kong’s urban areas into natural terrain creates a considerable potential risk from landslides. In recent years,
considerable effort has been directed towards identifying geological parameters that pre-dispose natural terrains to landsliding and the
assessments of landslide risk. There has, however, been little investigation of other possible causative factors, specifically, the role of
vegetation type and disturbance, particularly hill fires, in promoting shallow landsliding. Fire-denuded hill slopes may be subject to more
frequent landsliding events. Today, much of Hong Kong’s hill slopes are covered in fire-prone grasslands, and hill fires associated with
cultural festivals are common place. This paper presents an approach to using remote sensing techniques to model areas of disturbed
vegetation and detect associated shallow landslides in natural terrain. The research uses data from two satellites, LANDSAT TM and
IKONOS. LANDSAT TM is used to detect disturbed vegetation using visible, near and mid-infrared bands and to obtain vegetation
indices NDVI and NDMIDIR that indicate plant vigor. IKONOS high resolution multispectral data are used to prepare a color composite
image to locate shallow landslides, that are correlated to LANDSAT depicted vegetation indices. Results show that this approach can be
used to correlate landslides to areas of disturbed vegetation. Potentially this work has application in defining landslide prone regions.
1. INTRODUCTION
Hong Kong is one of the world’s pre-eminent cities with a present,
but rapidly expanding population of nearly 7.5 million. Its
population is mostly nestled on the lowland coastal fringe, giving
this coastal city one of the highest population densities in the
developed world. Within a few kilometres of its urban centres is
an expansive hilly terrain comprising deeply weathered Mesozoic
granites and volcanics. Steep slopes, deeply weathered profiles,
and a humid tropical climate predispose this terrain to landsliding
events. Heavily scarred hillsides are testimony to the rapidity and
severity of landslides in both natural and man-modified
landscapes. Increased demand for livable space now sees major
urban developments and infrastructural projects being located
nearer to natural terrain hill slopes. Such developments have been
affected by historical events and are at risk from future events (Ng
et. al., 2002). Historic landslides have been mapped from aerial
photographs in order to define landslide risk (Ng er. al., 2002).
The potential use of digital remotely sensed data to the mapping of
landslides and determination of temporal associations between
landsliding events and surface conditions has, however, not
previously been examined. This paper thus represents a first foray
into the application of digital remote sensing data to natural terrain
landslide studies in Hong Kong.
2. BACKGROUND
2.1 Natural Terrain Landslides in Hong Kong
‘Natural’ terrain that covers about 60% of the land area of Hong
Kong is defined as terrain that has not been substantially modified
489
by human activity but does include areas where grazing, hill fires
and deforestation have occurred. Much of the natural terrain is
only marginally stable over large areas with 27,000 landslides
recorded between 1945 and 1997. The Hong Kong Geological
Survey, Civil Engineering Department (GEO, CED) has compiled
a Natural Terrain Landslide Inventory (NTLI) for 1945-2000 from
high altitude aerial photographs. There is also a Large Landslides
Database (defined by source areas of >20 m wide) compiled from
low altitude aerial photographs.
Ng et. al., (2002) recognize that the NTLI suffers from certain
limitations (poor aerial photograph resolution, perspective
distortion, ground shadows and vegetation cover) inherited in the
coverage due to the quality of the high altitude aerial photographs.
Ng et. al., (2002) also recognize that for the purpose of an a
assessment of hazards the NTLI database, needs to be
supplemented with data on corresponding rainfall history and any
environmental changes (e.g. anthropogenic effects such as hill
fires) that have affected the mapped arcas.
Limitations of the existing natural terrain landslide mapping
programme potentially can be overcome through application of
remotely sensed data. Remotely sensed data has many advantages
over traditional aerial photography for mapping landslides and
other natural hazards. The capability of remote sensing data to
provide synoptic information over wide areas is a particular
advantage (Richards, 1993). From a single, high-resolution
satellite image in digital format, with multispectral information, a
range of ground surface features can be identified, classified, and
mapped directly. By comparison, a large number of aerial
photographs are required to view a similar area, and tedious image
rectification (geometric correction) is required before features can