Er 343 Link
Er 343 Link
Er 343 Link
Earthquake-induced Natural
Terrain Landslide Hazard
Assessment
Arup
Arup
Prepared by:
Preface
W.K. Pun
Head, Geotechnical Engineering Office
December 2018
4
Foreword
Executive Summary
This study is a regional assessment to estimate the order of magnitude of the likelihood
of earthquake-induced natural terrain landslides in Hong Kong. The earthquake-induced
landslide susceptibility has been examined relative to the rainfall-induced landslide
susceptibility. The adopted regional assessment approach is consistent with the regional
perspective considered by a study undertaken previously by the Geotechnical Engineering
Office on the global quantitative risk assessment of earthquake-induced landslides of man-made
slopes in Hong Kong. Such regional perspective provides insights on the overall scale of the
problem. This study is a hazard assessment and the direct consequences of earthquake-induced
natural terrain landslides are not considered.
Based on a literature review of some local and overseas state-of-the-art papers, the
methodology for landslide and boulder fall hazard assessments has been formed. For
earthquake-induced landslides, the slope displacements have been assessed using both
numerical modelling (by Oasys SIREN and Dynamic FLAC) and empirical correlations
(Jibson et al, 1998; Hsieh & Lee, 2011) for the seismic ground motions with 2% and 10%
probabilities of being exceeded in the next 50 years. Based on the literature review, it is found
that the vertical seismic ground acceleration would not have any significant effect on the slope
displacements. On the other hand, the topographic effects of 30 m, 100 m and 300 m ridge
heights have been considered in this study. It is found that the topographic effects of higher
ridges could significantly increase the induced slope displacements. The results have been
interpreted in terms of a static factor of safety corresponding to a threshold displacement value
of 100 mm, above which earthquake-induced landslides are likely to be triggered according to
the literature review. It is concluded that the slopes need to be close to failure at the time of
earthquake should there be any significant earthquake-induced landslide hazards.
A seismic microzonation map has been produced for natural terrain landslide
susceptibility taking into account the combined effects of seismic ground motions and
rainstorms. This is done by considering how much time the natural terrain is close to failure
as a result of rainstorms when an earthquake event takes place. The microzonation map is
formed based on a landslide susceptibility map derived from past landslides (up to 2009)
recorded in the Enhanced Natural Terrain Landslide Inventory. Overall, the effect of
earthquakes on triggering natural terrain landslides is small. There could be an increase of the
failure probability of being less than one tenth of that expected from rainstorms alone. To cater
6
for those uncertainties resulting from different assumptions made, a sensitivity study has been
carried out to illustrate the maximum credible effect of seismic ground motions on the natural
terrain landslide hazards.
A rock fall and boulder fall susceptibility map has also been produced in qualitative
terms based on the boulder density and the overall slope angle. This map has not been changed
by considering the effects of seismic ground motions.
The implication of distress of natural hillsides, which may be caused by earthquakes and
would increase the rainfall-induced landslide susceptibility, has not been considered in this
study. Furthermore, debris resulting from earthquake-induced landslides may be deposited on
natural hillsides and it is liable to be re-mobilised under subsequent rainstorms. Thus, the
potential increase of natural terrain landslide hazards after a large earthquake event should not
be neglected. It is believed that a ‘reactive approach’, including inspection and mapping of
potential slope distress as well as major debris deposits close to facilities, could be a pragmatic
approach to deal with post-earthquake landslide hazards. If a major earthquake occurs in Hong
Kong, slope inspections and mapping of vulnerable areas on the natural hillsides should be
carried out in the same way as the inspections that would be performed for buildings and
infrastructure.
7
Contents
Page
No.
Title Page 1
Preface 3
Foreword 4
Executive Summary 5
Contents 7
List of Tables 11
List of Figures 13
1 Introduction 26
1.1 Background 26
1.2 Scope of this Report 26
1.2.1 Literature Review 26
1.2.2 Development of Bedrock Seismic Ground Motion 27
Time Histories and Arias Intensity
1.2.3 Methodology for the Analysis of Earthquake-induced 27
Landslide
1.2.4 Methodology for the Analysis of Earthquake-induced 27
Boulder Fall
1.2.5 Landslide Susceptibility 27
2 Literature Review 28
2.1 General 28
2.2 Methodology for Seismic Stability Analysis of Slopes 29
in Natural Terrain
2.2.1 Ground Motion Parameters 29
2.2.2 Pseudo-static Approach 30
2.2.3 Newmark’s Method 31
2.2.4 Influence of Vertical Acceleration 36
2.2.4.1 Yang (2007) 36
2.2.4.2 Commentary 39
8
Page
No.
Page
No.
Page
No.
7 Conclusions 200
8 References 201
11
List of Tables
Table Page
No. No.
Table Page
No. No.
4.1 Spring (K) and Dashpot (D) Coefficients for Boulder 138
Seismic Stability Analysis
List of Figures
Figure Page
No. No.
Figure Page
No. No.
Figure Page
No. No.
2.27 Success Rate Curves for the Chi-Chi Earthquake Event in the 69
Kaohsiung Quadrangle: (a) Hilly Terrain; (b) Mountainous
Terrain
Figure Page
No. No.
Figure Page
No. No.
3.31 Arias Intensity for Time Histories for the 10% in 50 Years 106
Ground Motion
3.32 Arias Intensity for Time Histories for the 2% in 50 Years 107
Ground Motion
3.34 Arias Intensity for 100 m Crest Spectrally Matched Time 109
Histories of 1.4 Amplification Factor for 10% in 50 Years
Ground Motion
3.35 Arias Intensity for 300 m Crest Spectrally Matched Time 110
Histories of 1.4 Amplification Factor for 10% in 50 Years
Ground Motion
Figure Page
No. No.
3.37 Arias Intensity for 100 m Crest Spectrally Matched Time 112
Histories of 1.4 Amplification Factor for 2% in 50 Years
Ground Motion
3.38 Arias Intensity for 300 m Crest Spectrally Matched Time 113
Histories of 1.4 Amplification Factor for 2% in 50 Years
Ground Motion
3.40 Arias Intensity for 100 m Crest Spectrally Matched Time 115
Histories of 1.2 Amplification Factor for 2% in 50 Years
Ground Motion
3.41 Arias Intensity for 300 m Crest Spectrally Matched Time 116
Histories of 1.2 Amplification Factor for 2% in 50 Years
Ground Motion
3.42 Arias Intensity for (a) 2% and (b) 10% in 50 Years 118
Ground Motions (with No Topographic Effect and 30 m,
100 m, 300 m Ridge Heights with a 1.4 PGA
Amplification Factor)
3.43 Arias Intensity for (a) 1.4 PGA Amplification Factor and 119
(b) 1.2 PGA Amplification Factor (with No Topographic
Effect and 30 m, 100 m, 300 m Ridge Heights) for 2% in
50 Years Ground Motions
Figure Page
No. No.
Figure Page
No. No.
4.3 Limiting Eccentricity Required for the Boulder to Fall, for 140
2% in 50 Years Ground Motion
4.4 Limiting Eccentricity Required for the Boulder to Fall, for 140
10% in 50 Years Ground Motion
21
Figure Page
No. No.
4.5 Limiting Eccentricity Required for the Boulder to Fall, for 141
63% in 50 Years Ground Motion
Figure Page
No. No.
Figure Page
No. No.
Figure Page
No. No.
6.3 Factor of Safety and Percentage Time Relationship for the 191
Susceptibility Classes
Figure Page
No. No.
1 Introduction
1.1 Background
The Geotechnical Engineering Office (GEO) of the Civil Engineering and Development
Department (CEDD) has previously assessed the potential effects of earthquakes on the stability
of man-made slopes and retaining walls in Hong Kong (e.g. Wong & Ho, 2000; Au-Yeung &
Ho, 1995). As the management of natural terrain landslide hazards is a focus of the post-2010
Landslip Prevention and Mitigation Programme, there is a need to examine the responses and
stability of natural terrain in Hong Kong under an earthquake event.
On 1st April 2009, GEO commissioned Ove Arup & Partners Hong Kong Limited (Arup)
to undertake the Investigation Assignment of “Pilot Seismic Microzonation Study in North-
west New Territories for the Study of Potential Effect of Earthquake on Natural Terrain” under
Agreement No. CE 49/2008 (GE). The Guangdong Engineering Earthquake Resistance
Research Institute (GEERRI) is a sub-consultant to Arup on this Assignment.
The above study involves (a) an overall seismic hazard assessment of Hong Kong; (b) an
area-specific seismic microzonation assessment for a Study Area in the North-west New
Territories; and (c) an evaluation of the potential effects of earthquakes on the natural terrain,
with an account taken of local geology, topography and envisaged ground responses. The
study evaluates the overall seismic hazard assessment of Hong Kong and produces maps
delineating zones with similar predicted seismic intensity and motion at the ground surface in
the Study Area. The potential natural terrain landslide hazards associated with the seismic
ground motions are also evaluated. Findings of parts (a) and (b) of the study have been
documented in Arup (2015) and Arup (2018) respectively.
This report documents part (c) of the study i.e. an evaluation of the potential effects of
earthquakes on the natural terrain. The results from parts (a) and (b) of the study such as the
response spectra for Hong Kong, bedrock seismic ground motion time histories, topographic
effects, etc. would be used to evaluate the potential effects of earthquake on the stability of
natural terrain and assess the earthquake-induced natural terrain landslide hazards in the Study
Area.
The earthquake-induced natural terrain landslide hazard assessment for a Study Area in
the North-west New Territories of Hong Kong has been carried out in this study. Figure 1.2
of Arup (2018) shows the location of the Study Area. Key items of the work completed under
this study are listed as follows.
1.2.2 Development of Bedrock Seismic Ground Motion Time Histories and Arias
Intensity
Horizontal time histories of seismic ground motions have been derived to be compatible
with the horizontal uniform hazard response spectra (UHRS) having 10% and 2% probabilities
of being exceeded in the next 50 years (or, for simplicity, 10% and 2% in 50 years ground
motions) (see Arup (2015) for the UHRS). Analyses of vertical ground motions and the
topographic amplification effects have been considered in generating the time histories. The
Arias Intensity, which is found to correlate well with the distribution of earthquake-induced
landslides (e.g. Jibson, 1993), has been calculated from the time histories.
A one-dimensional dynamic stability analysis program has been developed for this study
to assess the boulder fall hazard under seismic loadings. Spherical boulders of 100 kg,
1,000 kg and 10,000 kg have been considered. The results have been presented in terms of the
limiting eccentricity necessary for a boulder to fall under a given seismic ground motion. The
horizontal ground acceleration time histories derived for site response analyses in Arup (2018)
as well as that catered for topographic effects (for 30 m, 100 m and 300 m ridge heights) have
been considered for 63%, 10% and 2% in 50 years ground motions. In addition, vertical
ground acceleration has been considered for the 2% in 50 years ground motion. The results
have been compared with that evaluated using an energy approach, which considers the amount
of energy required for a boulder to tip.
A landslide susceptibility analysis has been conducted for the Study Area considering
the existing geological conditions, topography, drainage, geomorphology and past instability
records (up to 2009) obtained from the Enhanced Natural Terrain Landslide Inventory (ENTLI).
Various susceptibility factors such as slope angle, geology, slope morphology, drainage have
been analysed and compared with the landslide density deduced from the ENTLI data.
Susceptibility classes have been defined and the probability of slope failures at any location has
been determined for each class. The increase in the probability of slope failures due to
earthquake ground motions has been determined together with the topographic amplification
28
effects. These combined failure probabilities have been used to generate landslide
susceptibility maps, which delineate and rank areas of potential landslide hazards within the
Study Area.
The implication of distress of natural hillsides, which may be caused by earthquakes and
would increase the rainfall-induced landslide susceptibility, has not been considered in this
study. Furthermore, debris resulted from earthquake-induced landslides may be deposited on
natural hillsides and it is liable to be re-mobilised by subsequent rainstorms. Thus, the
potential increase of natural terrain landslide hazards after a large earthquake event should not
be neglected. It is believed that a ‘reactive approach’, including inspection and mapping of
potential slope distress as well as major debris deposits close to facilities, could be a pragmatic
approach to deal with post-earthquake landslide hazards. If a major earthquake occurs in Hong
Kong, slope inspections and mapping of vulnerable areas on the natural hillsides should be
carried out in the same way as the inspections that would be performed for buildings and
infrastructure.
2 Literature Review
2.1 General
The review covers the methodologies for assessing the seismic displacements of slopes
in natural terrain, the landslide susceptibility factors and mapping as well as seismic stability
of boulders. This section summarises the findings of this literature review, on which the
assessment of the earthquake-induced natural terrain landslide susceptibility has been based.
There were some local studies pertaining to natural terrain landslides. For example,
Ding (2008) and Halcrow (2009) indicated that if Hong Kong was subjected to a magnitude 7.0
earthquake located within 50 km (e.g. at the Dangan Islands), there would be significant areas
of shallow landslides and rock falls. However, such an extreme event may not be considered
29
as a reasonable design event for Hong Kong. Also, no indication had been given to the
probability of such an extreme event in their studies.
Some studies (e.g. Tang et al, 2009 & 2010; Wong & Ding, 2010; Wong et al, 2010)
examined the extent of possible neotectonic movements of major faults in Hong Kong as a
result of earthquakes and their potential correlations with natural terrain landslides. The
studies involved geomorphological assessments, field observations, ground investigation and
dating of superficial deposits. It was found that there were two clusters of territory-wide deep-
seated landslide aged at around 50,000 and 30,000 years ago that could had been possibly
induced by seismic ground motions. Nevertheless, the potential triggering cause due to high
intensity rainstorms could not be ruled out. Hence, the studies could not conclude that
neotectonic movements had occurred simply by referring to the evidence of historical landslides.
Apart from natural terrain landslide studies, Wong & Ho (2000) presented a quantitative
risk assessment of earthquake-induced landslides of man-made slopes in Hong Kong. They
provided a very useful methodology for combining the hazards of rainfall-induced landslides
with that induced by seismic ground motions. The likelihood of low, moderate and high
degrees of soil saturation estimated in that report has been used to estimate the combined effects
of rainstorms and earthquakes on the natural terrain landslide hazard in this study (see
Section 6).
The Newmark’s method (Newmark, 1965) has been commonly used to assess the slope
movements under an earthquake event. Jibson (1993) described a state-of-the-art approach for
analysing the seismic slope stability using a simplified method to calculate Newmark
displacement. Blake et al (2002) described the ground motion parameters for seismic slope
stability analyses. They also recommended the use of pseudo-static seismic loadings in a
conventional limit equilibrium analysis as a screening process and a displacement analysis
based on an analogy of a rigid block on an inclined plane. The susceptibility of slope failures
to the accompanying vertical ground motions as well as the earthquake-induced reactivation of
previous landslides were also discussed.
Blake et al (2002) concluded that the ground motion parameters from Abrahamson &
Silva (1996) and Rathje et al (1998) used in the recommended seismic slope displacement
analysis procedures pertaining to the maximum horizontal acceleration, the duration of strong
shaking and the mean period of ground motion (Tm). Duration is typically quantified for this
purpose as the time span in which 90% of the energy in an earthquake accelerogram is released,
or more specifically, as the time between the 5% and 95% normalised Arias Intensity (D5-95).
D5-95 and Tm are mainly a function of earthquake magnitude (M), site-source distance (r), site
conditions (i.e. rock vs. soil), and the type of faulting.
30
-1/3
exp (A1 + A2 (M - 6))
[ 1.5M + 16.05 ]
ln (D5-95 ) med = ln [ 10
6 ] + ln (A4 ) ............................ (2.2)
15.7 × 10
where the coefficients A1, A2, A3 and A4 are listed in Table 2.1.
A1 A2 A3 A4 C1 C2 C3 V1 V2
(s) (s) (m/s) (s) (s) (s) (m/s) (m/s) (m/s)
5.204 0.851 15.873 2.378 0.411 0.0837 0.00208 1.87 3.477
Rathje et al (1998) defined the mean period (Tm) as the inverse of the weighted average
frequency, with weights defined from the Fourier amplitude spectrum over a frequency range
from 0.25 to 20 Hz. For practical application, the authors suggested to estimate the parameter
as:
where parameters C1, C2 and C3 should be selected for a rock site condition using the values
presented in Table 2.1.
centroid of the mass. If FoS > 1, the site passes the screening. If FoS < 1, the site fails and
should be subsequently analysed for slope deformation.
where MHAr is the maximum horizontal acceleration at a soft rock site, g = acceleration of
gravity, and feq is a factor related to the seismicity of the site, as described below.
Previous pseudo-static procedures for seismic slope stability have specified a single
value for feq, and thus have made implicit and usually very conservative assumptions about the
magnitude of earthquakes causing the design-basis MHAr. The paper suggested a method to
reduce any unnecessary conservatism by developing a range of feq values as a function of
magnitude (M) and site-source distance (r). Calculations were performed to evaluate, for
various combinations of MHAr, M and r , the feq values that lead to a 50% probability of the
seismic slope displacement exceeding the chosen threshold Newmark displacements of 5 cm or
15 cm. The results of the calculations are shown in Figure 2.1, and correspond to the following
equation:
u NRF
feq = [V1 - log10 ( MHAr )] ................................. (2.6)
( ) × NRF × D5-95 V2
g
where the coefficients V1 and V2 are defined in Table 2.1, u (cm) is the threshold displacement,
D5-95 is the median duration from the Abrahamson & Silva (1996) relationship, and NRF is a
factor to allow for the non-linear response of the material above the failure surface, which is
defined as:
- MHAr /g
NRF ≈ 0.6225 + 0.9196 × Exp ( )
0.4449
Figure 2.1 Required Values of feq as Function of MHAr and Seismological Condition
for Threshold Displacements of (a) 5 cm and (b) 15 cm (Blake et al, 2002)
Originally intended for dams and embankments, the Newmark’s method was proved to
be applicable to natural slopes by Wilson & Keefer (1983). This method assumes that the
33
landslide mass is treated as a rigid-plastic body. As such, the mass is assumed not to deform
internally, nor to experience any permanent displacement at a seismic acceleration below the
yield level. It deforms plastically along a discrete basal shear surface when the critical
acceleration is exceeded.
The author thus highlights that the Newmark’s method is the best applied to translational
block slides, especially infinite-slope models that are applicable to landslides on natural terrain,
and rotational slumps.
It should be noted that the Newmark’s method is also not applicable to certain slope
materials, which shear strength reduces with increasing strain. In this case, the Newmark’s
method would underestimate the slope displacements. On the other hand, for visco-elastic
materials, it would overestimate the slope displacements due to the viscous damping of the
seismic response of the soil.
The thrust angle () is the direction in which the centre of gravity of a sliding mass
moves when the displacement first occurs. For an infinite slope, this is the slope angle. For
a simple planar block, it is the basal shear surface inclination angle. Newmark (1965) defined
the thrust angle for a circular rotational movement as the angle between the vertical and a line
segment connecting the centre of gravity of the slide mass and the centre of the slip circle (see
Figure 2.2). For irregular shapes, an “equivalent” circular slip surface should be estimated.
The critical acceleration (Ac) is the pseudo-static acceleration that leads to a factor of
safety (FoS) of unity for a slope or a sliding block. For a rigid block,
The earthquake acceleration time histories are selected using PGA, Arias Intensity and
duration as the ground shaking intensity indexes.
Two approaches were suggested by Newmark (1965) for selecting an input ground
motion for displacement analyses:
34
These methods are now considered to be oversimplified to select time histories given
the current digitised acceleration time histories possessing a broad range of attributes, including
the magnitude and source distance that significantly influence the duration and predominant
periods of shaking. Jibson (1993) suggested the use of Arias Intensity, which was found to
correlate well with the distribution of earthquake-induced landslides.
The Arias Intensity (Ia) has an unit of velocity and is defined as:
where g is gravity, and a(t) is the time history of the ground acceleration. Ia can be estimated
based on the following relationships if the time history is not available:
where M and R are the moment magnitude of the earthquake and earthquake source distance in
kilometres respectively, and T is the duration in seconds.
In addition to the integration of the time history, there are numerous empirical
correlations to calculate Newmark displacement based on Ia (Jibson, 1993), horizontal
acceleration and the median duration (D5-95 (s)) (Blake et al, 2002).
Table 2.2 BASIC Program for Calculating Newmark Displacement (Jibson, 1993)
Yang (2007) studied the possible effects of vertical acceleration on the seismic landslide
hazard considering the effects of varying the groundwater level. He assumed that the
horizontal and vertical seismic coefficients (kh and kv respectively) were applied simultaneously
and studied the ratios of kv/kh = 0.5, 1.0 and 1.5.
Based on an infinite slope model (Figure 2.4), Yang (2007) calculated the vertical
acceleration-dependent factor of safety (FS) as:
where c' is the effective cohesion, ϕ' is the effective friction angle, γ is the unit weight of the
soil, σ is the total normal stress, and u is the pore water pressure on the failure plane.
Or further as:
c' + {[(1+ kv) cos β - kh sin β] γ - (1+ kv) γw m cos β} z cos β tan ϕ'
FS = [(1+ kv) sin β + kh cos β] γz cos β
...................... (2.10)
37
where γw is the unit weight of water, and m is the ratio of zw to z (see Figure 2.4) being 0 for a
dry slope and 1 for a saturated slope.
The earthquake-induced excess pore water pressure was not considered. The yield
acceleration could then be expressed as:
a - a4 a - a6
khy = (a1 ) kv + (a3 )............................................. (2.11)
5 - a2 5 - a2
Figure 2.5 shows that the effect of vertical acceleration is negligible when kh < 0.2, but
tends to become significant at larger kh values. Regarding the influence of the vertical
acceleration on slope displacements, Figure 2.6 illustrates how the vertical acceleration
amplifies the slope displacement, which is calculated as:
2.53
d 1 - ry (a'y /amax )
Ad = = (r-1.09
y )[ ] ..................................... (2.12)
d' 1 - (a'y /amax )
with
khy 1
ry = =
k'hy 1- χp
where Ad is the displacement amplification factor, χ = (a1 - a4) / (a5 - a2), p = kv/kh and a'y is the
yield acceleration excluding the effect of vertical acceleration.
38
Figure 2.5 Factor of Safety of a Saturated Slope under Various Combinations of kh and kv
(Yang, 2007)
2.2.4.2 Commentary
This paper sets out to maximise the effects of vertical ground acceleration on slope
movements. Figure 2.6 shows amplifications of slope displacement up to 100 times due to
vertical acceleration but in reality these large factors only apply to extremely small
displacements assessed without considering the vertical seismic ground motion. For a typical
case where the peak vertical acceleration is similar to the peak horizontal acceleration, it is most
unlikely that the vertical acceleration will be larger than 40% of the horizontal acceleration at
the time when the peak horizontal acceleration occurs. It can be concluded from this paper
that consideration of vertical acceleration will rarely lead to a significant increase in the
estimated downslope displacements.
The two cases studied were characterised through thorough investigation comprising
geological mapping, subsurface logging and sampling, and laboratory direct shear testing. The
actual landslide behaviour was then compared with the results of two analytical seismic stability
analyses, following the pseudo-static method and cumulative-displacement analysis (i.e.
Newmark method). This study provided a good opportunity to test the dynamic slope stability
methods currently used in practice, especially for the case of earthquake-induced reactivation
of landslides in natural terrain.
The Newmark displacement can be estimated by different methods. Hsieh & Lee (2011)
summarised the latest Arias Intensity-based empirical correlations for assessing earthquake-
induced slope displacements. Blake et al (2002) also recommended an empirical formula to
evaluate the Newmark displacement using the horizontal ground acceleration and the median
duration (D5-95 (s)), which is also obtained from the build-up of Arias Intensity (Ia) during an
earthqauake event. The most recent empirical correlations proposed by Hsieh & Lee (2011) are
discussed below. In addition, the threshold displacement for landsliding in natural terrain is
also discussed.
40
Hsieh & Lee (2011) employed strong-motion data from the 1999 Chi-Chi earthquake, the
1999 Kocaeli earthquake, the 1999 Duzce earthquake, the 1995 Kobe earthquake, the 1994
Northridge earthquake and the 1989 Loma Prieta earthquake to study the worldwide
applicability of previous relationships among critical acceleration, Arias Intensity, and
Newmark displacement, and refined them according to the new findings.
For example, it was found that the relationships derived by Jibson (1993) and Jibson et
al (1998) (see equations below) were not applicable to the 1999 Chi-Chi earthquake dataset.
where Ac is critical acceleration (in g), Ia is the Arias Intensity (in m/s), Dn is the downslope
displacement (in cm), and the last term is the estimation error of the models.
Table 2.3 Goodness of Fit to a Line in Different Relationships between Dn and Ia for Ac
from 0.01 to 0.4 g (Hsieh & Lee, 2011)
Based on this study, the authors derived two new forms of relationships (referred as
“New Form I” and “New Form II”) between log Dn, log Ia and Ac from the 1999 Chi-Chi
earthquake dataset, and the worldwide dataset, for both soil and rock site conditions. The
generic forms of New Form I and New Form II are shown as follows:
New Form I:
The dataset that established the following equations are shown in Figure 2.7.
New Form I:
R2 = 0.804
R2 = 0.837
Figure 2.7 Results of the Chi-Chi Dataset Fitting to (a) New Form I, (b) New Form II
(Hsieh & Lee, 2011)
Worldwide relationships
The dataset that established the following equations are shown in Figure 2.8.
42
New Form I:
R2 = 0.84
R2 = 0.89
Figure 2.8 Results of Worldwide Dataset Fitting to (a) New Form I, (b) New Form II
(Hsieh & Lee, 2011)
The results that led to the following equations are shown in Figure 2.9.
R2 = 0.875
R2 = 0.843
43
Figure 2.9 Results of the Chi-Chi Dataset Fitting to New Form II (Hsieh & Lee, 2011)
The results that led to the following equations are shown in Figure 2.10.
R2 = 0.884
R2 = 0.905
Figure 2.10 Results of Worldwide Dataset Fitting to New Form II (Hsieh & Lee, 2011)
44
Commentary
Any level of a threshold displacement can be used based on the ground conditions
considered by a specific landslide case study. For example, highly ductile materials may
accommodate more displacements without general failure, while brittle materials might
accommodate less. What constitutes failure may vary according to the experience of the user
(Jibson, 1993). For the present study, the threshold displacement is set as 100 mm, which is
an average value suggested in the above literatures.
In addition to the methods introduced above, there are other issues related to the subject
of earthquake-induced landslides such as susceptibility factors and others. As such, a review
of a Special Issue of Engineering Geology issued in 2011 has been undertaken. This Special
Issue was based on the outcomes of an international conference “The Next Generation of
Research on Earthquake-Induced Landslides” held in 2009. Its recommendations, particularly
for those relevant to Hong Kong, are briefly discussed below.
45
For the purpose of the present study, only Grade 1 and Grade 2 zonation approaches
have been reviewed.
(b) Category II: slumps, block slides and earth flows; and
Among these three classes, the events classified under Category I are associated with
shallow-seated landslides and thus are the major concern for the present study.
Tamura (1978) conducted case studies on slope failures which occurred during
37 earthquakes in Japan over the past 100 years. Four kinds of distances were adopted in the
study as follows:
(i) Df: distance from a fault to an outer boundary of the zone where
many slope failures occurred.
(ii) df: distance from a fault to an outer boundary of the zone where a
few slope failures occurred.
The magnitude-distance results for this study are shown in Figure 2.11.
Figure 2.11 Relationships between Magnitude and Distance to Slope Failure in Japan
(after Tamura, 1978, from ISSMFE, 1993)
Yasuda & Sugitani (1988) identified 105 landslide cases that occurred in Japan in the
past 100 years and classified them into two groups: surface slides and deep slides. The
magnitude-distance relationships and the lower bound for slope failure for both groups are
illustrated in Figure 2.12.
Keefer & Wilson (1989) conducted a statistical analysis for 47 typical events which have
occurred since 1811 worldwide. Those slope failures were classified in Categories I to III as
listed above. Figure 2.13 illustrates the results of this study.
The findings of this study were based on slope failures that occurred in Japan during
three large earthquakes: the 1974 Izuhanto-Oki earthquake; the 1978 Izuoshima-Kinkai
earthquake; and the 1984 Naganoken-Seibu earthquake. The susceptibility to slope failure in
each mesh of the map was calculated as the sum of the weighting factors listed in Table 2.5.
The classification of the topography of slopes is given in Figure 2.16.
49
Table 2.5 Weighting Factors Related to Slope Failure (after Kanagawa Prefectural
Government, 1986, from ISSMFE, 1993)
Mora & Vahrson (1994) conducted case studies of slope failures in historic earthquakes
as well as those induced by heavy rainfall in Central America. Five factors: relative relief,
lithology, soil moisture, seismicity and rainfall intensity were considered as triggering factors
for slope failure. The degree of slope failure hazard is defined as follows:
Table 2.6 Classes of the Potential Landslide Hazards (after Mora & Vahrson, from
ISSMFE, 1993)
I Negligible
II Low
III Moderate
IV Medium
V High
VI Very high
Table 2.7 Relative Relief (Sr) Values and their Classes of Influence in Landslide
Susceptibility (after Mora & Vahrson, from ISSMFE, 1993)
0 - 75 Very low 0
76 - 175 Low 1
176 - 300 Moderate 2
301 - 500 Medium 3
501 - 800 High 4
> 800 Very high 5
51
Table 2.9 Classes of Average Monthly Precipitation (after Mora & Vahrson, from
ISSMFE, 1993)
< 125 0
125 - 250 1
> 250 2
52
Table 2.10 Weighting for Annual Precipitation (after Mora & Vahrson, from ISSMFE,
1993)
5-9 Low 2
10 - 14 Medium 3
15 - 19 High 4
20 - 24 Very high 5
Intensities (MM)
Susceptibility Value, Ts
Tr = 100 Years
III Slight 1
IV Very low 2
V Low 3
VI Moderate 4
VII Medium 5
VIII Considerable 6
IX Important 7
X Strong 8
XI Very strong 9
Tables 2.13 and 2.14 summarise the characteristics of the landslides investigated, based
on Keefer (1984 & 1999). The types of slopes and geological environments that resulted in
earthquake-induced landslides varied, ranging from overhanging slopes in well-indurated
bedrock to unconsolidated sediments with nearly level surfaces. The minimum slope gradients
for various types of landslides were found to range from 0.3° to 40° (Table 2.13).
54
Table 2.13 Characteristics of Earthquake-induced Landslides (Keefer, 2002) (Sheet 2 of 2)
55
56
It was found that the approximate magnitudes of the smallest earthquakes that triggered
landslides were ∼4.0 for rock falls, rock slides, soil falls, and disrupted soil slides; ∼4.5 for soil
slumps and soil block slides; ∼5.0 for soil lateral spreads, rapid soil flows, subaqueous
landslides, rock slumps, rock block slides, and slow earth flows; ∼6.0 for rock avalanches; and
∼6.5 for soil avalanches. As these landslides could also be triggered by other non-seismic
agents, Keefer (1984) denoted that the possibility of earthquake-induced landslides could be
smaller than those indicated above.
The results from Rodriguez et al (1999) and Hancox et al (1997) were also interpreted
in terms of the relationship between the area affected by landslides (km2) and the moment
magnitude, as illustrated in Figure 2.17.
Based on the landslide dataset listed in Table 2.15 (1957-1999), a semi-log relationship
for the number of landslides that was expected under a given earthquake magnitude was derived,
as illustrated in Figure 2.18.
57
Note: Circles are data from earthquakes discussed by Rodriguez et al (1999), plotted using moment
magnitude (M); open circle is 1988 Saguenay, Quebec earthquake. Solid line is upper bound
of Keefer & Wilson (1989), see Figure 2.12. Dashed line is upper bound of Rodriguez et al
(1999). Triangle is datum from 1963 Peria, New Zealand, earthquake, for which area exceeds
upper bounds, plotted using Richter local magnitude (ML), from Hancox et al (1997; 2002).
Figure 2.17 Relations between Area Affected by Landslides and Earthquake Moment
Magnitude (Keefer, 2002)
Figure 2.18 Relation between Total Number of Reported Landslides and Earthquake
Magnitude for Earthquakes with Comprehensive Inventories of Landslides
(Keefer, 2002)
58
Approximate percentage by
category
Earthquake Earthquake Number of
Earthquake location Reference
date magnitude* landslides Disrupted Coherent Lateral
spreads and
(%) (%)
flows (%)
Daly City, Calif. USA 22 May 1957 5.3 23 48 30 22 Bonilla, 1960
Guatemala 4 Feb 1976 7.5 ~ 50,000 Most ? ? Harp et al., 1981; E.L. Harp,
unpublished data
Mt. Diablo, Calif. USA 24 Jan 1980 5.8 103 83 17 0 Wilson et al., 1985; Keefer and
Wilson, 1989
Mammoth Lakes, Calif. 25 May 1980 6.2 5,253 > 98 <1 <1 Harp et al., 1984; Keefer and
USA Wilison, 1989;
Wieczorek and Jäger, 1996
Coalinga, Calif. USA 2 May 1983 6.5 9,389 > 97 <2 <1 Keefer and Wilson (1989); Harp
and Keefer (1990)
San Salvador, EI Salvador 10 Oct 1986 5.7 > 216 > 93 <5 <2 Rymer, 1987; Rymer and White,
1989
Loma Prieta, Calif. USA 17 Oct 1989 6.9 ~ 1,500 74# 26# 0# Keefer and Manson, 1998
Northridge, Calif. USA 17 Jan 1994 6.7 > 11,000 > 90 <9 <1 Harp and Jibson, 1995, 1996
Hygoken-Nanbu, Japan 17 Jan 1995 6.9 674 - 747 81 to 83 13 to 15 3 to 4 Sassa et al., 1995; Fukuoka et al.,
1997; Okimura and Torii, 1999
Umbria-Marche, Italy 26 Sep 1997 6.0 100 - 124 61 34 5 Bozzano et al., 1998; Esposito et
al., 2000; I. LaRosa, unpublished
data
Chi-Chi, Taiwan 21 Sep 1999 7.7 22,000 > 85 11 to 15 <4 Lin and others, 2001; Sitar and
Bardet, 2001
*Magnitudes in bold are moment magnitudes; others are Richer surface-wave (Mt. Diablo) or Richer local (Daly City) magnitude.
#Percentages determined in central area only.
Finally, Keefer (2002) reviewed the threshold values of Arias Intensity for the
occurrence of earthquake-induced landslides. Keefer & Wilson (1989) suggested threshold
values of 0.11 m/s for disrupted landslides, 0.32 m/s for coherent slides and 0.54 m/s for lateral
spreads and flows, whereas Wilson (1993) found the best-fit threshold value for disrupted
landslides to be 0.1 m/s (Table 2.16).
Regarding the minimum slope gradients mentioned in Keefer (2002), the area of < 5°
was generally considered as the depositional area rather than the erosion area in natural terrain
landslides. Therefore, the probability of landslide occurrence in such a gentle slope was very
low. In Keefer (2002), 0.3° was referred as the minimum slope angle for soil lateral spreading,
which comprised less than 6% of the total number of landslides reported in the study.
The epicentre of the September 21, 1999 Chi-Chi earthquake (Mw 7.6) was located about
15 km south of the study area (Ma et al, 1999; Kao & Chen, 2000). Over 9,000 large landslides
of various types (with areas greater than 625 m2), with a total area of 128 km2 were triggered
by the Chi-Chi earthquake (Liao & Lee, 2000). Of these, 1,316 landslides, with a total area of
22 km2, were triggered in the Kaohsiung quadrangle, and 1,623 landslides, with a total area of
12 km2, were triggered in the Tungshih quadrangle.
The Chi-Chi earthquake event-based landslide inventory for the Kaohsiung quadrangle
is shown in Figure 2.19. Only shallow landslides (including rock falls), which included all of
the landslides in the study area, were used in the susceptibility analysis. Deep-seated
landslides, rock avalanches (located outside the study area) and debris flows were studied in a
separate project.
The working procedure for the event-based landslide susceptibility analysis, derived
from high resolution satellite images taken before and soon after the triggering event, is
illustrated in Figure 2.20. This procedure must be performed separately for each type of terrain
because of the differences in their geomorphic and geologic characteristics. Hilly terrain and
mountainous terrain were thus considered separately in the study. The basic data used included
a 40 m × 40 m grid digital elevation model (DEM), SPOT images, 1:5,000 photo-based contour
maps, 1:50,000 geologic maps, and earthquake strong-motion records.
60
Strong-motion seismograms in and around the study area were collected by the Central
Weather Bureau, Taiwan. Base-line correction and filtering of the data were performed
according to the standard procedure suggested by the Pacific Earthquake Engineering Research
Center (PEER) (Darragh et al, 2004). The Arias Intensity was then calculated from each
corrected seismogram. The arithmetical means of the Arias Intensities of the NS and EW
components were used to represent the earthquake intensity for each of the strong-motion
station sites. These values were interpolated on each grid point in the study area using the
Kriging method (Goovaerts, 1997).
61
The Arias Intensity (Ia) for each station was determined by correcting and processing the
strong-motion records of the main shock of the Chi-Chi earthquake and six major aftershocks.
Figure 2.21 shows the distribution of Ia for the main shock (stations located on top of ridges, as
indicated in the figure, were not included in the interpolation). Each grid point was checked
to find the maximum Ia among the seven earthquake events, and these values were adopted as
the intensity that triggered the landslides.
Topographic effect
The authors further considered topographic effects in relation to the earthquake intensity.
The height of the grid point above the riverbed was found to be a good factor for making
corrections, following the empirical formula proposed by Lin & Lee (2003).
62
Figure 2.21 Arias Intensity Isopleth of the Main Shock of the Chi-Chi Earthquake.
Strong-motion Stations with Indication of Ridge Top Stations, Earthquake
Epicentre, and the Study Area are Shown (Lee et al, 2008)
h
for f =√ + 0.287 + 0.464
93.8
where Ia is the Arias Intensity (m/s), I'a is the corrected Arias Intensity (m/s), f is the
amplification factor, and h is the height relative to the riverbed (m).
(a) Original arias intensity (b) Distribution of landslide and (c) Landslide ratio vs. original
non-landslide groups with arias intensity
original arias intensity
(d) Height of slope relative to (e) Distribution of landslide and (f) Landslide ratio vs. height
riverbed non-landslide groups with relative to riverbed
relative height above riverbed
(g) Corrected arias intensity (h) Distribution of landslide and (i) Landslide ratio vs. corrected
non-landslide groups with arias intensity
corrected arias intensity
Figure 2.22 Spatial Distribution of the Arias Intensity and Frequency Distribution of the
Landslide and Non-landslide Groups and the Landslide Ratio for the Chi-
Chi Earthquake Event in the Kaohsiung Quadrangle (Lee et al, 2008)
64
The amplification factors for heights of 30 m, 100 m and 300 m were calculated using
the above empirical formula, as listed in Table 2.17. These are compared with the topographic
amplification factors calculated in the present study. Details will be presented later in
Section 3.4.2.1.
Table 2.17 Arias Intensity Amplification Factor for Topographic Effect Derived in
Lee et al (2008)
There are more than 50 different landslide-related causative factors commonly used
(both in Taiwan and worldwide) for landslide susceptibility analyses (Lin, 2003). The authors
selected 14 of the most frequently used, based on both abundance and accessibility. These
were the lithology, slope gradient, slope aspect, terrain roughness, slope roughness, total
curvature (Wilson & Gallant, 2000), local slope height, total slope height, topographic index
(Kirkby, 1975), distance from a road, distance from a fault, distance from a river head, distance
from a river bend, and the normalised differential vegetation index (NDVI, Paruelo et al, 2004).
All these factors were processed by a raster GIS - ERDAS IMAGINE system. The
effectiveness of the factors as discriminators was evaluated by computing the standardised
differences for each factor (Davis, 2002) as:
Aj - Bj
Dj = ........................................................ (2.26)
Spj
where Aj is the mean of factor j for group A (landslide), Bj is the mean of factor j for group B
(non-landslide), Spj is the pooled standard deviation of factor j and Dj is standardised difference
of factor j.
The arger the standardised difference, the more effective the factor is for differentiating
between landslide and non-landslide groups. On this basis, of the fourteen causative factors,
six of them (lithology, slope gradient, slope aspect, terrain roughness, slope roughness, and total
curvature) were finally selected for the event-based landslide susceptibility analysis (EB-LSA)
of the earthquake-induced landslides.
The spatial distribution of the values of these six causative factors is shown in
Figure 2.23. Terrain roughness at a given point was defined as the standard deviation of
elevations within a certain distance (Wilson & Gallant, 2000), i.e. a radius of three pixels in this
case. The data were high-pass filtered before calculation of the terrain roughness. Slope
roughness was defined as the standard deviation of slope gradients within a radius of three
pixels. The definition of total curvature was the same as that used by Wilson & Gallant (2000).
65
Note: Lithologic unit: 1. Alluvium; 2. Terrace Deposits; 3. Lateritic Terrace Deposits; 4. Toukoshan
Formation (conglomerate dominate); 5. Toukoshan Formation (sandstone and shale dominate);
6. Cholan Formation; 7. Kueichulin Formation; 8. Fulungyuan Formation; 9. Hourdonqkeng
Formation; 10. Shihmentsum Formation; 11. Takeng Formation, Tanliaoti Member; 12. Takeng
Formation, Shihszeku Member; 13. Shuichangliu Formation; 14. Paileng Formation, Meitzulin
Member; 15. Paileng Formation, Tungmou Member.
Slope gradient, terrain roughness, and Arias Intensity were found to be the predominant
factors for the landslide susceptibility assessment of both hilly and mountainous terrains, but
lithology was also found to be determinant in hilly terrain.
The spatial probability of landslide occurrence was indicated by the relationhsip between
the landslide ratio (probability of failure) and the landslide susceptibility index (LSI) (see
Figure 2.24). The spatial probability of a landslide was then used to map the susceptibility
classes, as shown in Figures 2.25 and 2.26. The authors then compared actual landslides that
occurred during the Chi-Chi earthquake with those on the susceptibility map (Figures 2.25 and
2.26), and observed a general agreement between the landslide pattern and areas of high
susceptibility.
Figure 2.25 Landslide Susceptibility Map of the Kaohsiung Quadrangle Developed Using
a Susceptibility Model Trained with the Chi-Chi Landslide Inventory and
the Earthquake Shaking Intensity in that Quadrangle. Landslides
Triggered by the Chi-Chi Earthquake are Shown (Lee et al, 2008)
68
A prediction rate curve method (Chung & Fabbri, 2003) was then used to examine how
well the classification results fit the data. The success rate curves for the two terrains are
shown in Figures 2.27 and 2.28. The area under the curve is between 0 and 1; a higher value
indicates a higher prediction rate, whereas a value near 0.5 means the prediction is no better
than a random guess (Chung & Fabbri, 2003).
Figure 2.27 Success Rate Curves for the Chi-Chi Earthquake Event in the Kaohsiung
Quadrangle: (a) Hilly Terrain; (b) Mountainous Terrain (Lee et al, 2008)
Figure 2.28 Prediction Rate Curves for the Chi-Chi Earthquake Event in the Tungshih
Quadrangle: (a) Hilly Terrain; (b) Mountainous Terrain (Lee et al, 2008)
70
Based on this study, Lee et al (2008) identified the slope gradient, terrain roughness and
Arias Intensity as three major factors affecting earthquake-induced landslides. As these
factors are site-specific, they may not be applicable to other sites like Hong Kong. However,
this study provided an indication to different potential parameters that are worth to be
considered.
Sewell & Tang (2012) integrated the findings of previous studies (Tang et al, 2009 &
2010; Wong & Ding, 2010; Wong et al, 2010) to examine the possible evidence that
earthquakes could have triggered landslides in four areas in Hong Kong i.e. Ho Lek Pui Area,
Tung Chung East Area, Wong Chuk Yeung Area and Nam Shan/Pui O Area. They also
attempted to evaluate possible seismic relationships with sizable natural terrain landslides. In
addition, this study reviewed and discussed other possible evidences for recent fault activities
in Hong Kong.
According to Sewell & Tang (2012), although a time period of possible fault movement
may overlap with a pulse of landslide activity by dating evidence, there was no strong evidence
for faults displacing the quaternary deposit in Hong Kong. It was therefore considered that
the local neotectonic fault movement had no major influence on relict natural terrain large
landslides.
The study also found that the sources of seismic shaking that could trigger landslides in
Hong Kong were most likely located outside Hong Kong. This was deduced by the dating
results of the faults in Hong Kong that showed little correlation with the pulses of landslide
activities.
Parker et al (2011) examined whether the 2008 Wenchuan earthquake led to a net gain
in orogen volume (via rock uplift) or a new loss (via landsliding). It was found that the
landslides produced more erodible material than new volume added to the orogen by co-seismic
rock uplift. In an accompanying presentation, the factors that potentially contribute to
landslides during an earthquake were discussed. It was shown that the fault rupture distance
was a dominant factor whilst the hillslope gradient and geology were other potential factors.
and slope dynamic response. Other considerations included the impact of earthquakes on
mountain slope evolution and sediment yields, including enhanced landslide generation by
precipitation events following earthquakes, and the uncertainty in ground motion inputs and
outputs from seismic slope stability analysis. Recent developments in the regional scale
analysis and the seismic landslide hazard evaluation, such as Lee et al (2008), Hsieh & Lee
(2011), were also reviewed.
Jibson (2011) reviewed the latest methods to assess the stability or performance of slopes
during earthquakes. They generally comprise three categories:
Despite the pros and cons of each analysis, they are suitable for different situations.
The pseudo-static analysis is easy to use, and provides a simple, scalar index of stability.
However, this simplicity stems from a rather crude characterisation of the physical processes
of dynamic slope behaviour that produces several drawbacks, including the difficulty in
rationally selecting a pseudo-static coefficient and assessing the likelihood or results of failure.
The stress-deformation modelling gives more accurate picture of what actually happens
in the slope during an earthquake. Although models that account for the complexity of spatial
variability of properties and the stress-strain behaviour of slope materials may yield more
reliable results, stress-deformation modelling has drawbacks and can be quite challenging
because of:
The study concluded that the permanent-displacement method such as the Newmark’s
rigid-block analysis is the best suited to earthquake-induced landslides in natural slopes. More
sophisticated coupled analyses were available to account for the fact that a landslide mass is
not a rigid body but could deform internally under seismic shaking. In general, simplified
decoupled analyses could yield more conservative results, which are within about 20% of the
coupled results. Coupled analyses are appropriate for deeper landslides in softer materials,
which could include large earth structures and deep landslides. It is, therefore, concluded that
72
decoupled analyses are sufficient for the present study, which deals with shallow natural terrain
failures.
Rathje & Antonakos (2011) evaluated the seismic performance of slopes by predicting
the sliding block displacements for critical sliding masses. The current practice adopted a rigid
sliding block approach for shallow sliding masses and a decoupled, flexible sliding block
approach for deeper/softer sliding masses. This paper presented a unified framework that
extended the empirical displacement models for application to flexible sliding masses, where
the dynamic response of the sliding mass is important. This framework included predicting
the seismic loading for the sliding mass in terms of the maximum seismic coefficient (kmax) and
the maximum velocity of the seismic coefficient-time history (k-velmax). The predictive
models were a function of the peak ground acceleration (PGA), peak ground velocity (PGV),
the natural period of the sliding mass (Ts), and the mean period of the earthquake motion (Tm).
This unified framework provided a consistent approach for predicting the sliding displacement
of both rigid and flexible slopes. This modification was a function of Ts and increased the
predicted displacement due to the nonlinear response of the soil and the reductions in the
amplitude of the seismic loading (i.e. kmax). Modification for both the (PGA, M) and (PGA,
PGV) models were developed, but the (PGA, PGV) model was recommended because of the
significant frequency content information provided by PGV (for rigid sliding) and by k-velmax
(for flexible sliding). It was also recommended that when the Ts /Tm ratio was less than or
equal to 0.1, the response of slope could be considered as rigid.
Gaudio & Wasowski (2011) studied the site effects based on the long term
accelerometric monitoring conducted on a landslide-prone test area in the Apennine Mountains,
Italy, where the presence of site effects enhancing the seismic susceptibility of local slopes had
been invoked on the basis of historic accounts of landslides triggered at a large epicentral
distance. The recordings relative to low-to-moderate magnitude earthquakes showed
significant amplifications affecting hillslope portions covered by a thick layer of colluvium
(> 5 m) and pronounced amplification maxima oriented along the local maximum slope
direction on a recent deep-seated landslide. The amplifications of the colluvium could be due
to the high impedance contrast between surface materials and the underlying substratum.
The GEERRI methodology is a statistical approach based on data obtained from > 100
historical earthquakes which induced landslides in China. These data were used to generate
different factors that possibly contributed to landslides as a result of seismic ground motions.
The importance of each factor was assessed by the Certainty-Factor (CF) model proposed by
Heckerman (1986). GIS was used to give the ranking spatially by considering the factors at
different locations.
The GEERRI concluded that the major factor leading to earthquake-induced landslides
was the seismic ground motion. The proximity to a fault on which the earthquake occurred
was also a major contributor but this was probably due to the fact that larger ground motions
occurred in a closer proximity to the fault.
2.4.12 Commentary
The Arias Intensity has been used as the triggering factor in the susceptibility analysis,
and this factor makes the susceptibility model temporally significant. It is also found to be the
most effective factor for interpreting landslide distribution.
Even though the topographic effect has been taken into account in the analyses, it is not
very accurately representative of the phenomena as the topographic amplification is not simply
proportional to the local height relative to riverbed, but can also be affected by factors such as
lithology, geologic structure, and ridge sharpness (Celebi, 1987; Jibson, 1987; Geli et al, 1988;
74
Ashford et al, 1997; Murphy, 2006). Simulations and the application of multivariate
geostatistical methods may be needed to improve the results in the future. Finally, it should
be noted that the authors mentioned that extreme condition of earthquake during a storm is not
valid for their study model, but the probability for those two events to happen simultaneously
is very low.
The seismic stability of boulders depends on many parameters, including the geometry,
material properties of the boulder and its embedment as well as the geometry of the slope. Two
scenarios could be considered for the potential boulder instability i.e. sliding or toppling as a
result of an imposed vibration. A review of the literature about seismic stability assessment of
boulders has been carried out. The papers studied different simplified approaches for assessing
the boulder stability and the pertinent parameters.
The effects of blast-induced ground vibrations were analysed by Wong & Pang (1991).
Their findings have been used as the basis for assessing the potential effects of blasting on rock
slopes and soil slopes. For the stability of rock blocks on a slope, an energy approach, which
took into account the blasting vibration energy transmission to the potential rock block, was
recommended. Figure 2.29 illustrates the rock block system subjected to blasting vibration.
Based on the principle of conservation of energy, the following equation was derived:
U
f 1W 2
∫0 τAb dδ = (Vg ) + W Uf sin β ..................................... (2.27)
2 g
Figure 2.29 Rock Block System Subjected to Blasting Vibration (Wong & Pang, 1991)
This relationship may still be conservative, but was considered to be good enough as it
took into account the fact that PPVx, PPVy and PPVz had very remote chance to occur at the
same time. The fundamental energy equation therefore became:
U W
∫0 τAb dδ = 0.91 g (PPV)2 + W Uf sin β ................................. (2.29)
f
76
Commentary
This energy approach, considering the energy required to move a boulder, can be used
for the analysis of the seismic stability of boulders. This approach assumes that the boulder
sits on a smooth surface with no embedment in the ground. In the present study, the energy
approach will be compared with more complex dynamic stability analysis. Also, the effect of
vertical motion will be assessed using the equation suggested by Wong & Pang (1991) for
multi-directional vibration. Details are presented in Section 4 of this report.
Haneberg (2009) described a simplified approach for the preliminary assessment of rock
toppling due to earthquakes, blasting, or other vibrations. It was a simplified dynamic method
that incorporated information about vibration frequency in addition to acceleration magnitude
but had assumed a single-frequency vibration of a rectangular rock. As such, it was not a
complete dynamic analysis for an irregular three-dimensional rock with a complicated basal
geometry and a full spectrum of frequencies.
Based on the analyses, it was shown that high-frequency vibrations caused by typical
blasting operations were unlikely to pose problems even if the peak acceleration values were
relatively high. Moderate to large earthquakes generating lower frequency vibrations might,
however, topple some rocks at the same or lower accelerations.
Although the method lacked the sophistication of those dynamic or multiple block
analyses undertaken by others in site-specific studies, it had an advantage that the input
parameters can be easily obtained. For example, the geometric variables can be readily
estimated in the field and those vibration information such as the horizontal acceleration and
the vibration frequency can be obtained from blasting records or earthquake seismograms.
Therefore, this method had the potential of being used for screening or reconnaissance purpose.
Haneberg (2009) suggested that the method be used within a geographic information systems
(GIS) framework to create rock-toppling potential maps based on the regional seismic hazards
and slope-angle maps. This approach has been adopted in the present study.
(a) The pseudo-static analysis led to the following relationship between the horizontal
acceleration required for toppling of the rock and the slope angle for a certain block
geometry:
Figure 2.31 illustrates the correlation between the ratio b/h, the peak horizontal
acceleration and the slope angle.
77
Figure 2.31 Maximum Stable b/h Ratios as a Function of Slope Angle (θ) and Peak
Horizontal Acceleration (aH) as Given by the Pseudo-static Factor of
Safety (Haneberg, 2009)
(b) The dynamic stability analysis assumed a sinusoidal acceleration time history of the form:
aH = ap sin (2 f t), where f is the frequency in Hz, and ap is the peak ground horizontal
acceleration. This led to the following relationship between the frequency of the time-
dependent shaking, the peak ground horizontal acceleration, the slope angle and the
geometry of the boulder, as:
1/2f 1/2f
∫0 ∫0 ap sin(2fπt) dt dt = Lω ........................................ (2.31)
78
where
π√h2 + b2 b
Lω = [arctan ( ) - θ]
360 h
Figure 2.32 illustrates the correlation between the slope angle, the frequency of the
ground motion that causes toppling, and the peak ground acceleration, for different geometries.
The static toppling stability of a rock block was a function of its shape (i.e. its b/h ratio)
but not its size (i.e. the individual values of b and h). Its dynamic toppling stability, however,
was shown to be a function of both shape and size. As the rock size increased, the centre of
mass must rotate through a longer linear (but not angular) distance to reach its tipping point,
and therefore, lower frequency vibration was required. Thus, with all other things being equal,
short rocks would always be more stable than tall rocks with regard to the vibration-induced
toppling.
79
The literature review presented in Section 2 covers methodologies for assessing the
stability and the displacement of slopes in the event that the ground motion is large enough to
result in momentary slope failure. Based on the literature review, the methodology adopted in
the present study for earthquake-induced landslide assessment in natural terrain is introduced
in this section.
The assessment adopts numerical models and empirical correlations to evaluate the slope
movement (permanent displacement) induced by seismic ground motions. The methods are
essentially an extension of the Newmark’s theory, which determines the amount of movement
of a rigid block sliding down a frictional surface. The block only moves when the applied
downslope acceleration force exceeds the shear resistance given by the frictional surface. The
acceleration required to match the shear resistance is referred to as the critical acceleration, Ac.
The resulted movement is a function of the magnitude and duration over which the applied
acceleration exceeds this critical acceleration. The numerical models require a seismic ground
motion in form of a time history. The empirical models correlate the permanent slope
displacement with the Arias Intensity.
Although the seismic ground motion may lead to downslope accelerations greater than
the critical acceleration, the downslope movement may still be very small. It is observed that
the downslope movement needs to exceed a threshold limit before the slope fails. This
threshold limit has been set as 100 mm for this study based on the literature review.
It should be noted that the critical acceleration, Ac, of a slope is a direct function of the
shear strength of the soil. Therefore, it will vary with time as the strength varies as a result of
seasonal variations of groundwater level.
The time histories of seismic ground motions with different likelihood of being exceeded
in the next 50 years have been developed for input into the numerical analyses. The derivation
of time histories is discussed as follows. Apart from the horizontal ground motion, the effects
of the vertical ground motion and topographic amplification have also been considered for the
most critical scenarios. The Arias Intensity, which is being used to represent the ground
motion in the empirical correlations, is derived from the time histories (see Section 3.4).
Time histories of the horizontal ground motion developed in the seismic microzonation
study (Arup, 2018) have been used as the basis for the slope stability analyses. The time
histories have been derived to be compatible with the uniform hazard response spectra for
bedrock as derived in the seismic hazard assessment (Arup, 2015). The time histories have
been considered for scenarios with different combinations of magnitude and distance,
representative of short period (0.2 s) and long period (1 and 5 s) ground motions. Table 3.1
80
lists the magnitude-distance combinations that have been considered. Table 3.2 lists the actual
earthquake records used as a basis for deriving the time histories in this study. They are the
same as those used for the microzonation study (Arup, 2018) except for the 2% in 50 years
ground motion (1 s, far-field event), which is based on a record from Alaska.
Ground motions having 10% and 2% probabilities of being exceeded in the next 50 years
(in short, 10% and 2% in 50 years ground motions) have been used for the slope stability
analyses. Figures 3.1 and 3.2 show the spectrally-matched target rock site response spectra
(Arup, 2015) for 10% and 2% in 50 years ground motions respectively. It should be noted that
for the 2% in 50 years, 1 s, far-field ground motion (resulting from a magnitude 8 event at
250 km), the target response spectrum has been derived for the expected response spectrum for
such an event scaled to match the uniform hazard spectrum at 1 s rather than to match the
uniform hazard spectrum at all periods as shown in Figure 3.2. It is because the expected
spectrum does not match the uniform hazard spectrum at low periods as can be seen from the
figure. This poor match had been noted in the microzonation study (Arup, 2018) but for the
simplicity of interpretation, it was not specifically considered in the site response analyses at
that time. For the present slope displacement studies, this poor match could significantly affect
the results as the Arias Intensity would be unrealistically increased by matching the uniform
hazard spectrum rather than the expected spectrum from such an event.
81
The generated time histories for 10% and 2% in 50 years ground motions are shown in
Figures 3.3 and 3.4 respectively.
1
0.8
0.6
Acceleration (m/s/s)
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1
0 10 20 30 40 50 60 70 80 90
Time (s)
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1
0 10 20 30 40 50 60 70 80 90
Time (s)
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1
0 10 20 30 40 50 60 70 80 90
Time (s)
Figure 3.3 Spectrally Matched Rock Input Time Histories for 10% in 50 Years Ground
Motion
83
2.5
2
1.5
Acceleration (m/s/s)
1
0.5
0
-0.5
-1
-1.5
-2
-2.5
0 10 20 30 40 50 60 70 80 90
Time (s)
Short Period Motion (0.2 s)
2.5
2
1.5
Acceleration (m/s/s)
1
0.5
0
-0.5
-1
-1.5
-2
-2.5
0 10 20 30 40 50 60 70 80 90
Time (s)
1
0.5
0
-0.5
-1
-1.5
-2
-2.5
0 10 20 30 40 50 60 70 80 90
Time (s)
1
0.5
0
-0.5
-1
-1.5
-2
-2.5
0 10 20 30 40 50 60 70 80 90
Time (s)
Figure 3.4 Spectrally Matched Rock Input Time Histories for 2% in 50 Years Ground
Motion
84
To take into account the effect of topographic amplification, the spectral ratios derived
in the seismic microzonation study (Arup, 2018) have been adopted to modify the horizontal
response spectra.
At low structural periods, the amplification factor generally increases from up to 1.4 at
zero period to 2 at the natural period of the crest, which is a function of the crest height. At
longer structural periods, the amplification factor reduces back to unity. The spectral ratios for
PGA amplification factors of 1.2 and 1.4 are plotted in Figure 3.5. Crest heights of 30 m,
100 m and 300 m have been considered. The target response spectra derived to account for
the topographic effect on the horizontal ground motions are plotted in Figures 3.6 to 3.8. As
the effect of amplification factor of 1.2 is relatively small, the target response spectra are only
plotted for the 2% in 50 years ground motions (see Figure 3.8).
1.6
Spectral Ratio
1.5
1.4
1.3
1.2
1.1
1
0.01 0.1 1 10
Period (s)
Figure 3.5 Spectral Ratios with Amplification Factors of 1.4 and 1.2 at Zero Period for
Ridge 30 m, 100 m and 300 m
85
Figure 3.6 Response Spectrum with Amplification Factor of 1.4 at Zero Period for 10%
in 50 Years Ground Motion
Figure 3.7 Response Spectrum with Amplification Factor of 1.4 at Zero Period for 2%
in 50 Years Ground Motion
86
Figure 3.8 Response Spectrum with Amplification Factor of 1.2 at Zero Period for 2%
in 50 Years Ground Motion
The time histories have been generated by modifying the same selected time histories as
the ones used to derive the horizontal ground motion time histories without topographic effect
in the time domain using RSPMATCH (Hancock et al, 2006), so that their response spectra
would closely match the targets as shown Figures 3.6 to 3.8. The spectrally matched target
response spectra for a PGA amplification factor of 1.4 for 10% and 2% in 50 years ground
motions are presented in Figures 3.9 to 3.14. Figures 3.15 to 3.17 illustrate the target matched
response spectra for a PGA amplification factor of 1.2. Figures 3.18 to 3.20 show the resulted
time histories for a PGA amplification factor of 1.4 for the 10% in 50 years ground motions.
Figures 3.21 to 3.23 show the corresponding generated time histories for the 2% in 50 years
ground motions. Figures 3.24 to 3.26 show the resulted time histories for a PGA amplification
factor of 1.2 for the 2% in 50 years ground motions.
87
Figure 3.9 30 m Crest Height Target Matched Spectra of 1.4 Amplification Factor for
10% in 50 Years Ground Motion
Figure 3.10 100 m Crest Height Target Matched Spectra of 1.4 Amplification Factor
for 10% in 50 Years Ground Motion
88
Figure 3.11 300 m Crest Height Target Matched Spectra of 1.4 Amplification Factor
for 10% in 50 Years Ground Motion
Figure 3.12 30 m Crest Height Target Matched Spectra of 1.4 Amplification Factor for
2% in 50 Years Ground Motion
89
Figure 3.13 100 m Crest Height Target Matched Spectra of 1.4 Amplification Factor
for 2% in 50 Years Ground Motion
Figure 3.14 300 m Crest Height Target Matched Spectra of 1.4 Amplification Factor
for 2% in 50 Years Ground Motion
90
Figure 3.15 30 m Crest Height Target Matched Spectra of 1.2 Amplification Factor for
2% in 50 Years Ground Motion
Figure 3.16 100 m Crest Height Target Matched Spectra of 1.2 Amplification Factor
for 2% in 50 Years Ground Motion
91
Figure 3.17 300 m Crest Height Target Matched Spectra of 1.2 Amplification Factor
for 2% in 50 Years Ground Motion
92
1.5
1
Acceleration (m/s/s)
0.5
-0.5
-1
-1.5
0 10 20 30 40 50 60 70 80 90
Time (s)
Short Period Motion (0.2 s)
1.5
1
Acceleration (m/s/s)
0.5
-0.5
-1
-1.5
0 10 20 30 40 50 60 70 80 90
Time (s)
1.5
1
Acceleration (m/s/s)
0.5
-0.5
-1
-1.5
0 10 20 30 40 50 60 70 80 90
Time (s)
Figure 3.18 30 m Crest Spectrally Matched Time Histories of 1.4 Amplification Factor
for 10% in 50 Years Ground Motion
93
1.5
1
Acceleration (m/s/s)
0.5
-0.5
-1
-1.5
0 10 20 30 40 50 60 70 80 90
Time (s)
1.5
1
Acceleration (m/s/s)
0.5
-0.5
-1
-1.5
0 10 20 30 40 50 60 70 80 90
Time (s)
1.5
1
Acceleration (m/s/s)
0.5
-0.5
-1
-1.5
0 10 20 30 40 50 60 70 80 90
Time (s)
Figure 3.19 100 m Crest Spectrally Matched Time Histories of 1.4 Amplification Factor
for 10% in 50 Years Ground Motion
94
1.5
1
Acceleration (m/s/s)
0.5
-0.5
-1
-1.5
0 10 20 30 40 50 60 70 80 90
Time (s)
1.5
1
Acceleration (m/s/s)
0.5
-0.5
-1
-1.5
0 10 20 30 40 50 60 70 80 90
Time (s)
1.5
1
Acceleration (m/s/s)
0.5
-0.5
-1
-1.5
0 10 20 30 40 50 60 70 80 90
Time (s)
Figure 3.20 300 m Crest Spectrally Matched Time Histories of 1.4 Amplification Factor
for 10% in 50 Years Ground Motion
95
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
Short Period Motion (0.2 s)
3
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
Long Period Motion (1 s) Near-field Event
3
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
Long Period Motion (1 s) Far-field Event
3
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
Long Period Motion (5 s)
Figure 3.21 30 m Crest Spectrally Matched Time Histories of 1.4 Amplification Factor
for 2% in 50 Years Ground Motion
96
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
Figure 3.22 100 m Crest Spectrally Matched Time Histories of 1.4 Amplification Factor
for 2% in 50 Years Ground Motion
97
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
Short Period Motion (0.2 s)
3
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
Long Period Motion (1 s) Near-field Event
3
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
Long Period Motion (1 s) Far-field Event
3
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
Long Period Motion (5 s)
Figure 3.23 300 m Crest Spectrally Matched Time Histories of 1.4 Amplification Factor
for 2% in 50 Years Ground Motion
98
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
Short Period Motion (0.2 s)
3
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
Figure 3.24 30 m Crest Spectrally Matched Time Histories of 1.2 Amplification Factor
for 2% in 50 Years Ground Motion
99
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
Short Period Motion (0.2 s)
3
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
Figure 3.25 100 m Crest Spectrally Matched Time Histories of 1.2 Amplification Factor
for 2% in 50 Years Ground Motion
100
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
Short Period Motion (0.2 s)
3
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
2
Acceleration (m/s/s)
-1
-2
-3
0 10 20 30 40 50 60 70 80 90
Time (s)
Figure 3.26 300 m Crest Spectrally Matched Time Histories of 1.2 Amplification Factor
for 2% in 50 Years Ground Motion
101
Attenuation equations specific for vertical ground motions are relatively scarce. Two
recently published period-dependent spectral ratios between vertical and horizontal ground
motions (V/H) are shown in Figure 3.27. The V/H ratio used as a basis to develop the vertical
target response spectrum has been derived as the mean curve of those two curves plotted in
Figure 3.27. The derived vertical target response spectrum is illustrated in Figure 3.28. The
corresponding spectrally matched time histories (see Figure 3.29) have been generated from the
vertical acceleration records of the same earthquake events as the ones selected for the 2% in
50 years horizontal ground motions for the periods of 0.2 s, 1 s (near-field) and 5 s.
0.8
V/H Ratio
0.6
0.4
0.2
0.0
0.01 0.1 1
Period (s)
Figure 3.27 Spectral Ratio of Horizontal Ground Motion to Vertical Ground Motion
horizontal spectrum
vertical spectrum
Spectral acceleration (m/s/s)
0.1
0.01
0.01 0.1 1 10
Period (s)
Figure 3.28 Target Response Spectra for Horizontal and Vertical 2% in 50 Years
Ground Motion
102
2
1.5
1
Acceleration (m/s/s)
0.5
0
-0.5
-1
-1.5
-2
0 10 20 30 40 50 60 70 80 90
Time (s)
Short Period Motion (0.2 s)
2
1.5
1
Acceleration (m/s/s)
0.5
0
-0.5
-1
-1.5
-2
0 10 20 30 40 50 60 70 80 90
Time (s)
0.5
0
-0.5
-1
-1.5
-2
0 10 20 30 40 50 60 70 80 90
Time (s)
Figure 3.29 Spectrally Matched Time Histories for Vertical 2% in 50 Years Ground
Motion
One-dimensional and two-dimensional numerical modelling have been carried out using
Oasys SIREN and dynamic FLAC respectively to calculate the downslope displacements of a
shallow slope subjected to seismic ground motions. The time histories described in Section
3.2 have been used as the input ground motions. The shear strength of soils has also been
considered. Both analyses have adopted a high small strain shear stiffness that degrades with
cyclic shear strain and generates hysteretic damping. A small amount of viscous damping
(0.5%) has been assigned to the soils in the analyses. More details are described as follows.
103
The program Oasys SIREN has been used to perform a one-dimensional dynamic slope
stability analysis. It is a finite difference program that analyses the horizontal response of a
one-dimensional soil column subjected to a seismic ground motion at its base. The program
has the advantage of modelling a system that is stronger in one direction than the other. As
such, it has been used to simulate an infinite slope.
In this method, the maximum undrained shear strength, c, of the slope (with a given
static FoS) is set by limiting the maximum stress level in the shear modulus degradation curve
in the soil layer at a shallow depth. The back-calculated c-value is used to match a target static
FoS for CDG at 2.5 m below ground level. The strength in the opposite direction is set to be
ten times of this value. By inputting the earthquake time history in the model, permanent
downslope displacements would automatically occur when the soil strength is exceeded.
The following equation shows the correlation between the maximum undrained shear
strength of a slope and the static FoS for different soil thicknesses and slope angles:
c
FoS = (γh cos α ) .................................................... (3.1)
sin α
In the analysis, a soil thickness of 2.5 m with slope angles of 20°, 30° and 40° have been
modelled using different undrained shear strengths. The critical acceleration corresponding to
this FoS is:
The topographic effect has also been considered using the time histories with
topographic amplification described in Section 3.2.2.
The dynamic analysis option of FLAC Version 6.0 has been used to undertake two-
dimensional fully non-linear dynamic analyses. It adopts explicit finite differences to solve
equations of motion using lumped grid point masses derived from the density of surrounding
zones. FLAC assumes a free-field boundary to prevent waves being reflected from the lateral
boundaries of the model. The hysteretic damping model has also been used to account for the
non-linear shear modulus degradation effect. A SHAKE type de-convolution analysis has
been carried out to determine suitable input ground motions at depth within the bedrock.
In the model, a 30° soil slope with 10 m slope height is covered by a 2.5 m thick layer
104
of CDG (Figure 3.30). The CDG has been assigned with a shear modulus degradation curve
according to Arup (2018). A shear wave velocity (VS) of 300 m/s, which is equivalent to a
SPT-N value of about 50 (based on the empirical correlation between VS and SPT-N value of
CDG presented in Arup (2018), has been used. It is also considered as a typical value for CDG
in Hong Kong. The slopes having a static FoS of 1.04 and 1.10 have been modelled by
assuming the shear strength of CDG as 15.8 kPa and 14.8 kPa respectively. The topographic
amplification has not been considered when using this slope geometry.
d ro c k J O B TITL E : B R 2 -g e o m e try ( *1 0 ^2 ) ( *1 0 ^2 )
F L A C (V e r s i o n 5 .0 0 )
2 .0 0 0 2 .0 0 0
LE GE ND
2 0 -J un-0 8 0 :0 3
s te p 903304
D yna m ic Tim e 4 .8 1 6 3 E +0 1
1 .5 0 0
CDG
CDG 1 .5 0 0
10m CDG
-1 .6 1 2 E +0 1 <x< 3 .0 6 1 E +0 2
10m 10m
-8 .9 1 2 E +0 1 <y< 2 .3 3 1 E +0 2
30o (2.5m thick of soil)
G rid p lo t 1 .0 0 0 1 .0 0 0
MDG MDG
0
s he a r_ m o d
5E 1
Rock
7 .6 0 0 E +0 7
0 .5 0 0 0 .5 0 0
2 .4 0 0 E +0 9
B o und a ry p lo t
0 5E 1
0 .0 0 0 0 .0 0 0
- 0 .5 0 0 - 0 .5 0 0
Empirical correlations based on the simplified Newmark’s method have been used to
calculate the downslope displacements. In this method, time histories, which have been used
to define seismic ground motions in the numerical models, could not be used. As discussed in
Section 2, the Arias Intensity (Ia) has been found to correlate well to the downslope movements.
This parameter has thus been adopted to represent the seismic ground motion. In addition to
the simplified Newmark’s method, correlations between the earthquake magnitudes and
distances as well as the landslide occurrence have been used to evaluate the landslide likelihood
in the Study Area for the earthquake scenarios derived from the probabilistic seismic hazard
assessment (PSHA) as presented in Arup (2015).
6.5 and 7.5 could be envisaged, with an epicentre at the Dangan Islands (around 30 km away
from Hong Kong). This event has been considered in the present analysis for the scenario
earthquake (short period of 0.2 s) for 10% in 50 years ground motion.
The results based on the magnitude-distance combinations from the PSHA of this study
or Dangan Islands scenario stated in Halcrow (2009) indicate that there is only a marginal
likelihood of landslides. Based on the displacement calculations, it has been found that the
static FoS is close to 1 if 100 mm is set as the threshold slope displacement. It means that
earthquake-induced landslides are unlikely to occur except when the slope, at its initial state, is
very close to failure.
log Dn = 0.847 log Ia - 10.62 Ac + 6.587 Ac log Ia + 1.84 ± 0.295 .............. (2.20)
log Dn = 0.788 log Ia - 10.166 Ac + 5.95 Ac log Ia + 1.779 ± 0.294 ............. (2.23)
Using the above correlations, the downslope displacements have been calculated and
compared for different seismic ground motions.
The Arias Intensity (Ia) is a quantitative measure of the total shaking intensity which has
been shown to possess remarkable correlation with the distribution of earthquake-induced
landslides. It is directly related to the integral over time of the square of the ground
acceleration as explained in Section 2.2.3.
Figures 3.31 and 3.32 show the Ia calculated using the original horizontal time histories
for 10% and 2% in 50 years ground motions respectively. The Ia plots corresponding to the
time histories modified for a PGA topographic amplification factor of 1.4 for 10% and 2% in
106
50 years ground motions are presented in Figures 3.33 to 3.35 and Figures 3.36 to 3.38
respectively. For the 2% in 50 years ground motions, the Ia plots corresponding to the time
histories modified by a PGA topographic amplification factor of 1.2 are illustrated in Figures
3.39 to 3.41.
0.12
0.1
Arias Intensity (m/s)
0.08
0.06
0.04
0.02
0
0 10 20 30 40 50 60 70 80 90
Time (s)
0.1
Arias Intensity (m/s)
0.08
0.06
0.04
0.02
0
0 10 20 30 40 50 60 70 80 90
Time (s)
0.1
Arias Intensity (m/s)
0.08
0.06
0.04
0.02
0
0 10 20 30 40 50 60 70 80 90
Time (s)
Figure 3.31 Arias Intensity for Time Histories for the 10% in 50 Years Ground Motion
107
0.35
0.3
Arias Intensity (m/s)
0.25
0.2
0.15
0.1
0.05
0
0 10 20 30 40 50 60 70 80 90
Time (s)
Short Period Motion (0.2 s)
0.35
0.3
Arias Intensity (m/s)
0.25
0.2
0.15
0.1
0.05
0
0 10 20 30 40 50 60 70 80 90
Time (s)
0.3
Arias Intensity (m/s)
0.25
0.2
0.15
0.1
0.05
0
0 10 20 30 40 50 60 70 80 90
Time (s)
0.3
Arias Intensity (m/s)
0.25
0.2
0.15
0.1
0.05
0
0 10 20 30 40 50 60 70 80 90
Time (s)
Figure 3.32 Arias Intensity for Time Histories for the 2% in 50 Years Ground Motion
108
0.25
0.2
Arias Intensity (m/s)
0.15
0.1
0.05
0
0 10 20 30 40 50 60 70 80 90
Time (s)
Short Period Motion (0.2 s)
0.25
0.2
Arias Intensity (m/s)
0.15
0.1
0.05
0
0 10 20 30 40 50 60 70 80 90
Time (s)
0.2
Arias Intensity (m/s)
0.15
0.1
0.05
0
0 10 20 30 40 50 60 70 80 90
Time (s)
Figure 3.33 Arias Intensity for 30 m Crest Spectrally Matched Time Histories of 1.4
Amplification Factor for 10% in 50 Years Ground Motion
109
0.25
0.2
Arias Intensity (m/s)
0.15
0.1
0.05
0
0 10 20 30 40 50 60 70 80 90
Time (s)
Short Period Motion (0.2 s)
0.25
0.2
Arias Intensity (m/s)
0.15
0.1
0.05
0
0 10 20 30 40 50 60 70 80 90
Time (s)
0.2
Arias Intensity (m/s)
0.15
0.1
0.05
0
0 10 20 30 40 50 60 70 80 90
Time (s)
Figure 3.34 Arias Intensity for 100 m Crest Spectrally Matched Time Histories of 1.4
Amplification Factor for 10% in 50 Years Ground Motion
110
0.25
0.2
Arias Intensity (m/s)
0.15
0.1
0.05
0
0 10 20 30 40 50 60 70 80 90
Time (s)
0.2
Arias Intensity (m/s)
0.15
0.1
0.05
0
0 10 20 30 40 50 60 70 80 90
Time (s)
0.2
Arias Intensity (m/s)
0.15
0.1
0.05
0
0 10 20 30 40 50 60 70 80 90
Time (s)
Figure 3.35 Arias Intensity for 300 m Crest Spectrally Matched Time Histories of 1.4
Amplification Factor for 10% in 50 Years Ground Motion
111
0.8
0.7
0.6
Arias Intensity (m/s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
Short Period Motion (0.2 s)
0.8
0.7
0.6
Arias Intensity (m/s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
Long Period Motion (1 s) Near-field Event
0.8
0.7
0.6
Arias Intensity (m/s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
Long Period Motion (1 s) Far-field Event
0.8
0.7
0.6
Arias Intensity (m/s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
Figure 3.36 Arias Intensity for 30 m Crest Spectrally Matched Time Histories of 1.4
Amplification Factor for 2% in 50 Years Ground Motion
112
0.8
0.7
0.6
Arias Intensity (m/s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
Figure 3.37 Arias Intensity for 100 m Crest Spectrally Matched Time Histories of 1.4
Amplification Factor for 2% in 50 Years Ground Motion
113
0.8
0.7
0.6
Arias Intensity (m/s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
Short Period Motion (0.2 s)
0.8
0.7
0.6
Arias Intensity (m/s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
Long Period Motion (1 s) Near-field Event
0.8
0.7
0.6
Arias Intensity (m/s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
Long Period Motion (1 s) Far-field Event
0.8
0.7
0.6
Arias Intensity (m/s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
Long Period Motion (5 s)
Figure 3.38 Arias Intensity for 300 m Crest Spectrally Matched Time Histories of 1.4
Amplification Factor for 2% in 50 Years Ground Motion
114
0.8
0.7
0.6
Arias Intensity (m/s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
Figure 3.39 Arias Intensity for 30 m Crest Spectrally Matched Time Histories of 1.2
Amplification Factor for 2% in 50 Years Ground Motion
115
0.8
0.7
0.6
Arias Intensity (m/s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
Figure 3.40 Arias Intensity for 100 m Crest Spectrally Matched Time Histories of 1.2
Amplification Factor for 2% in 50 Years Ground Motion
116
0.8
0.7
0.6
Arias Intensity (m/s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
Short Period Motion (0.2 s)
0.8
0.7
0.6
Arias Intensity (m/s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Time (s)
Figure 3.41 Arias Intensity for 300 m Crest Spectrally Matched Time Histories of 1.2
Amplification Factor for 2% in 50 Years Ground Motion
117
The variation of Ia has further been investigated for different scenarios. Figure 3.42
summarises the Ia values for 2% and 10% in 50 years ground motions respectively when
different ridge heights are considered for a 1.4 PGA amplification factor. The Ia values
increase with an increasing level of ground motions. For the 10% in 50 years ground motions,
Ia increases with longer period ground motions and also generally with increasing ridge height.
For the 2% in 50 years ground motion, Ia also generally increases with longer period ground
motions. However, the 1 s far-field ground motions show almost no increase of Ia for 30 m
and 100 m ridge heights but a significant amplification for the 300 m ridge height. As shown
in Figure 3.5, only 300 m ridge height would amplify the longer period ground motion, which
is expected to be present in this far-field ground motion. This is consistent with the trend
observed in Figure 3.42.
The effect of varying the PGA amplification factor on Ia has also been investigated.
The Ia values for the 2% in 50 years ground motions for the cases with no topographic effects
and with PGA amplification factors of 1.2 and 1.4 respectively are plotted in Figure 3.43.
The Ia values subject to topographic effects have been compared with the empirical
formula proposed by Lee et al (2008) (see Section 2.4.3). Equation 2.25 corrects Ia for the
topographic amplification effect as a function of height and Table 2.17 shows the Arias
Intensity amplification factor calculated for 30 m, 100 m and 300 m using this method. For
the present study, the amplification factors of Ia have been calculated using Ia of 30 m, 100 m
and 300 m ridge height ground motions divided by Ia without any topographic effects. The
results are summarised in Table 3.3, which are similar to those derived by Lee et al (2008).
They match especially well for the time histories obtained from the 1999 Chi-Chi earthquake
(1 s for 10% and 5 s for 2% in 50 year ground motions). This comparison suggests that the
Arias Intensities developed in the present study are reasonable.
It is well understood that a uniform hazard spectrum and an Arias Intensity are not
directly related (i.e. two earthquake time histories with the same response spectrum could have
very different values of Arias Intensity). For this reason, the time histories have been selected
based on the earthquake scenario magnitude – distance combinations that are most likely to
lead to various parts of the uniform hazard spectrum. These time histories, which should have
reasonable durations appropriate to the hazard, have then been used to derive Ia as plotted in
Figures 3.31 and 3.32 and again on Figures 3.42 and 3.43.
As an additional check that the ranges of values arising from the various earthquake
scenarios are reasonable, PSHA has been used to directly derive Arias Intensity. The
attenuation relationship by Travasarou et al (2003) has been used in this additional PSHA. The
resulting hazard curve is shown in Figure 3.44. It shows a 2% in 50 years (0.0004 annual
probability) Arias Intensity of about 0.2 m/s and a 10% in 50 year (0.0021 annual probability)
Arias Intensity of about 0.06 m/s. These values compare well with the range of values used in
the report of 0.14 to 0.32 m/s for the 2% in 50 years ground motions, and 0.06 to 0.104 m/s for
the 10% in 50 years ground motions.
118
(a)
(b)
Figure 3.42 Arias Intensity for (a) 2% and (b) 10% in 50 Years Ground Motions (with
No Topographic Effect and 30 m, 100 m, 300 m Ridge Heights with a 1.4
PGA Amplification Factor)
119
(a)
(b)
Figure 3.43 Arias Intensity for (a) 1.4 PGA Amplification Factor and (b) 1.2 PGA
Amplification Factor (with No Topographic Effect and 30 m, 100 m,
300 m Ridge Heights) for 2% in 50 Years Ground Motions
120
Probability of
PGA
Being Exceeded Original Time
Amplification Scenario 30 m 100 m 300 m
in the Next History
Factor
50 Years
Figure 3.44 PSHA Results for Arias Intensity Using Travasarou et al (2003)
121
3.5 Results of the Slope Stability Analyses in Natural Terrain under Seismic Loading
3.5.1 Magnitude-distance Landslide Susceptibility
Using the findings of the literature review presented in Section 2 together with the
combinations of magnitude and distance to epicentre for the probable ground motions derived
from the PSHA in Arup (2015), the overall landslide susceptibility in the Study Area has been
assessed.
Figures 3.45 to 3.48 show the comparison between the study dataset and the correlations
from Keefer & Wilson (1989), Tamura (1978), Yasuda & Sagitani (1988) and ISSMFE (1993)
respectively.
100
10% 0.2s
10% 0.5s-1s
10 10% 5s
2% 0.2s
2% 0.5s-1s near field
2% 0.5s-5s far field
1
2% 5s near field
0.1
4 5 6 7 8 9
Magnitude (M)
Note: Category I: falls, disrupted slides, and avalanches; Category II: slumps, block
slides, and earth flows; Category III: lateral spreads and flows.
Figure 3.45 Magnitude-distance Comparison from Study Dataset to Keefer & Wilson
(1989)
122
10% 0.2s
10% 0.5s-1s
30 10% 5s
2% 0.2s
2% 0.5s-1s near field
2% 0.5s-5s far field
2% 5s near field
3
5.5 8
Magnitude (M)
7.5
7
10% 0.2s
Magnitude, M
4.5
4
1 10 100 1000
Maximum epicentral distance of slope failure sites (km)
Figure 3.47 Magnitude-distance Comparison from Study Dataset to Yasuda & Sugitani
(1988)
123
8.5
10% 0.5s-1s
7 10% 5s
2% 0.2s
6.5 2% 0.5s-1s near field
2% 0.5s-5s far field
6
2% 5s near field
5.5
5
5 50 500
Epicentral distance (km)
The natural terrain landslide hazard has been assessed in terms of the induced downslope
displacements using both numerical and empirical methods as discussed in Sections 3.3 and 3.4
respectively. Both 10% and 2% in 50 years ground motions have been considered.
Additionally, calculations have been carried out for cases with topographic effects (i.e. ridge
heights of 30 m, 100 m, and 300 m), in which the PGA amplification factors of 1.2 and 1.4 have
been considered. As regards the vertical ground acceleration, given that its effect on the
seismic slope displacements is considered negligible for the ground motions appropriate to
Hong Kong (see Section 2.2.4), this factor has not been considered in the analyses.
Figure 3.49 shows the overall envelope (lower and upper bounds) of all the results for
the critical acceleration, Ac, corresponding to the displacement required to trigger a landslide
for the 10% and 2% in 50 years ground motions with and without topographic effects. It can
be seen that the Ac values are significantly lower for the 10% in 50 years ground motions than
those for the 2% in 50 years ground motions. The average and inferred Ac are found by
approximately averaging the values of Ac at 100 mm displacement calculated using different
empirical and numerical methods for all the scenario earthquakes. The Ac is judged graphically
for better consideration for all cases rather than calculated using the lower and upper bound
values. For the 10% in 50 years ground motions, the average FoS value corresponding to the
triggering of landslides lies between 1 and about 1.05. It implies that such earthquake ground
motions have barely any effect on the triggering of landslides compared to a static failure. For
the 2% in 50 years ground motions, the average FoS corresponding to landslide triggering lies
between 1.05 and 1.10 for 20° slopes, and around 1.05 for 30° to 40° slopes. When no
topographic effect is considered, the 2% in 50 years ground motions only have limited effect
on the slope stability.
Figure 3.49 Effect of Ground Motion on Induced Downslope Displacement for 2% and
10% in 50 Year Ground Motions
125
The detailed results for the 2% and 10% in 50 years ground motions (with no
consideration for topographic effects) are presented in Figures 3.50 and 3.51 respectively.
As can be seen, the 5 s long period ground motion gives rise to the largest downslope
displacements whereas the 0.2 s short period ground motion results in the least displacements.
For the 2% in 50 years ground motions, the Oasys SIREN analyses give lower displacement
values than the dynamic FLAC analyses, which are in good agreement with the results of
empirical correlations.
Detailed plots for the 2% in 50 years ground motions (1 s far-field and 5 s) are shown
for topographic effects of 30 m, 100 m and 300 m crest heights in Figures 3.52 to 3.57
respectively. The topographic effect is reflected on the calculated downslope displacements.
More discussions are provided in the following section.
The topographic effects on the induced seismic slope displacements have been studied
for the 10% and 2% in 50 years ground motions respectively.
Figures 3.58 to 3.60 show the summary plots of the results from the empirical
correlations as well as from Oasys SIREN analyses for the 2% in 50 years ground motions,
considering topographic effects of 30 m, 100 m and 300 m crest heights (for a PGA
amplification factor of 1.4) respectively. Overall, it is found that the higher the crest is, the
greater the induced displacements would be. Therefore, a 300 m crest height is considered the
most critical case. For a 30 m crest height, the average FoS corresponding to the threshold
displacement for landslide triggering lies between 1.06 for a 20° slope, to 1.03 to 1.04 for 30°
to 40° slopes. For 100 m and 300 m crest heights, the average FoS lies between 1.10 for 20°
slopes, to 1.06 for 30° slopes and 1.05 for 40° slopes. The effect of earthquakes on the natural
terrain slope stability is therefore not negligible when considering topographic effects. Such
an effect becomes stronger with larger crest heights.
130
Figure 3.59 Downslope Displacements Calculated for 2% in 50 Years, and 100 m Crest
Height Effect (1.4 PGA Amplification Factor)
131
Figure 3.60 Downslope Displacements Calculated for 2% in 50 Years, and 300 m Crest
Height Effect (1.4 PGA Amplification Factor)
Figures 3.61 to 3.63 show the same results for the 10% in 50 years ground motions.
It can be seen that the topographic effects are greatly reduced. Even for the largest crest
height, the static FoS to prevent a slope failure is less than about 1.05.
Figure 3.61 Downslope Displacements Calculated for 10% in 50 Years, and 30 m Crest
Height Effect (1.4 PGA Amplification Factor)
Figure 3.62 Downslope Displacements Calculated for 10% in 50 Years, and 100 m
Crest Height Effect (1.4 PGA Amplification Factor)
133
Figure 3.63 Downslope Displacements Calculated for 10% in 50 Years, and 300 m
Crest Height Effect (1.4 PGA Amplification Factor)
Figure 3.65 Downslope Displacements Calculated for 2% in 50 Years, and 100 m Crest
Height Effect (1.2 PGA Amplification Factor)
Figure 3.66 Downslope Displacements Calculated for 2% in 50 Years, and 300 m Crest
Height Effect (1.2 PGA Amplification Factor)
135
Summary
Based on the above seismic slope displacement assessments, it is found that both
numerical models (Oasys SIREN and Dynamic FLAC) and empirical correlations produce
comparable results. The findings help identify the most critical scenario for earthquake-
induced landslides i.e. a long period (1 s far-field or 5 s) event of the 2% in 50 years seismic
ground motions. The topographic effect significantly affects the induced slope displacements,
especially for a ridge higher than 100 m. However, the threshold displacement of 100 mm
required to trigger landslides is reached for a relatively low static FoS, implying that
earthquakes likely to happen in Hong Kong will generally have a limited effect on landslide
triggering even under 2% in 50 years ground motions. In high ridge areas, the topographic
amplification effects will be more significant. More discussions will be provided in the
susceptibility microzonation mapping in Section 5.
136
The results of the dynamic stability analysis have been compared with those obtained
using an energy approach. This approach considers the amount of energy required for a
boulder to fall under a certain peak ground velocity. The equation used for the energy
approach that accounts for the effects of vertical ground motions is based on Wong & Pang
(1991) for multi-directional blasting vibration (see Section 2).
The stability of boulders is dependent on the geometry of the boulder, the characteristics
of its position on the ground, and the surrounding soil/rock conditions. In this study, spherical
boulders of 100 kg, 1,000 kg and 10,000 kg have been considered. It is further assumed that
the boulders are sitting on rock and the primary falling mode of the boulder is by tipping.
The results of the analyses are presented in terms of the limiting eccentricity necessary
for the boulder to fall under a given acceleration time history. The eccentricity of a boulder
sitting on a slope is defined in Figure 4.1. It corresponds to the horizontal distance between
the centre of gravity of the boulder and the point around which it will rotate or tip.
The 1-D dynamic stability analysis program that has been used to assess the limiting
eccentricity necessary for a boulder to fall under seismic loadings is based on the calculation of
the relative displacement of the boulder to that of the ground at each time step of 0.001 s of the
seismic accelerations time histories:
dg0 = 0m
db0 = 0m
abi = Σ Fi / mb
where Fi = the forces applied to the boulder at the time step i (in N)
mb = the mass of the boulder (in kg).
or else Fs = 0 N
K being the spring coefficient (in N/m) to give a small displacement under
the static restoring of the boulder and has the values listed in Table 4.1.
or else Fd = 0 N
138
D being the dashpot coefficient (in N/m/s) to obtain critical damping of the
boulder when it reconnects with the spring after lifting off. It has been
assigned the values listed in Table 4.1.
Table 4.1 Spring (K) and Dashpot (D) Coefficients for Boulder Seismic Stability Analysis
Fg = 9.81 × mb × (e + db - dg) / R
where avi (in m/s2) is the imposed vertical acceleration at the ith time step.
The limiting eccentricity of the boulder required for it to move under a certain ground
motion has been deduced using the above equations. It corresponds to the eccentricity at which
the relative displacement of the boulder diverges to a very large value with time as illustrated
in an example shown in Figure 4.2.
500
400
Relative displacement (mm)
300
200
100
-100
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (s)
Figure 4.2 Calculated Relative Displacement when the Limit Eccentricity is Reached
for a 100 kg Spherical Boulder, under the 2% in 50 Years, 0.2 s Horizontal
Ground Motion
139
The results presented above have been compared with the results of an energy approach
that considers how much energy delivered by the ground motion is required to move a boulder.
Et = Fh × dt ................................................................................... (4.2)
where Fh (N) is the average horizontal force applied to the boulder, and dt is the distance that
the boulder has to move for it to tip ( = e ).
Fh = 0.5 mb × g × e / R
The eccentricity can therefore be derived from the fundamental energy equilibrium
equation as:
0.5 mb V2 = 0.5 mb g e2 / R
i.e.
e = V (R / g)0.5
where V (m/s) is defined as the peak horizontal particle velocity of the ground motion PPV.
When considering the effects of the vertical ground acceleration on the seismic stability
of boulders, the energy equation can be modified as recommended by Wong & Pang (1991) by
taking:
It should be noted that the horizontal PPV exceeds the vertical PPV as expected from
the response spectra shown in Figure 3.28.
Based on the methodology described above, a boulder seismic stability analysis has been
carried out using the 1-D dynamic stability method and the energy approach. The results of
both methods have been compared in terms of the limiting eccentricity of the boulder, that is
required for it to tilt under a given seismic ground motion.
The 1-D dynamic stability analysis has been carried out for 2%, 10% and 63% in
50 years seismic ground motions. The topographic effects have also been taken into account.
Figures 4.3 to 4.5 show the calculated limiting eccentricity for different ground motions for a
spherical boulder with different masses (100 kg, 1,000 kg and 10,000 kg). The results show
140
that the limiting eccentricity required for the boulder to tip under seismic loadings significantly
increases with the level of ground motion. However, even for a 1 s near-field, 2% in 50 years
ground motion, the highest limiting eccentricity for a 10,000 kg boulder is only 27 mm, which
is considered as a small value. The results indicate that under the probable earthquake ground
motions in Hong Kong, boulders do not have any likelihood to fall unless they are already very
close to toppling.
Figure 4.3 Limiting Eccentricity Required for the Boulder to Fall, for 2% in 50 Years
Ground Motion
Figure 4.4 Limiting Eccentricity Required for the Boulder to Fall, for 10% in 50 Years
Ground Motion
141
Figure 4.5 Limiting Eccentricity Required for the Boulder to Fall, for 63% in 50 Years
Ground Motion
In order to investigate the topographic effects on the seismic stability of boulders, the 1-
D dynamic stability analysis has been carried out for a 10,000 kg boulder with the 2% in 50
year time histories that account for the topographic effects for 30 m, 100 m, and 300 m ridge
heights (see Figure 4.6). It shows that higher crests tend to result in higher limiting eccentricity
values. As the crest height increase from 30 m to 300 m, the limiting eccentricity increases to
31 mm, 24 mm, 36 mm and 40 mm for the 0.2 s short period, 1 s far-field, 5 s near-field , and
1 s near-field ground motions respectively. For a 10,000 kg spherical granitic boulder of about
a metre radius, theses values considered as very small.
Figure 4.6 Limiting Eccentricity Required for the Boulder of 10,000 kg to fall for 2% in
50 Years Ground Motion, with Topographic Effect (30 m, 100 m, and 300 m
Crest Height)
142
In addition, the effect of the vertical acceleration has been checked. Figure 4.7 shows
the comparison between the horizontal ground motion results (without topographic effect) and
the ones calculated by taking the vertical acceleration into account. Clearly, the effect of the
vertical acceleration is negligible.
Figure 4.7 Effect of the Vertical Motion on the Limiting Eccentricity Necessary for
Boulders of 100 kg, 1,000 kg and 10,000 kg to Fall Under a 2% in
50 Years Ground Motion
The results between 1-D dynamic stability analysis and the energy approach have been
compared. Figures 4.8 to 4.11 compare the eccentricity calculated by the energy approach
and that calculated by the dynamic analysis for cases with no topographic amplification effect,
and with 30 m, 100 m, and 300 m ridge height effects considered respectively. In general, the
correlations are good, although there is a consistent trend that the energy approach would
predict a larger eccentricity to retain equilibrium.
When comparing the results of both methods modified to account for the vertical
acceleration effect, Figure 4.12 shows that the energy approach gives a slightly higher required
eccentricity than the dynamic analysis. The results are consistent with the cases in which
vertical acceleration has been ignored. This implies that the recommendations of Wong &
Pang (1991) who suggested that the velocity would be increased by 35% (i.e. V = 1.35 max
(PPVx, PPVy )) would give much higher limit eccentricity values than the dynamic stability
approach. Therefore, this additional coefficient appears not appropriate to account for the
vertical ground motion effect in the seismic stability assessment for boulders.
143
As discussed in Section 2.5.2, Haneberg (2009) showed that boulders with higher aspect
ratios would require lower eccentricities to topple. The energy approach gives a direct
indication to the effect of aspect ratios in that the equation shows that the limiting eccentricity
is directly related to the square root of the height of the centre of mass of the boulder above its
base. Therefore, for example, for a prismatic boulder having a height four times its width, the
height of its centre of mass would double that of a spherical boulder. It means that the
eccentricity required would be about 1.4 times that that required for a sphere.
It is therefore concluded that while boulders with high vertical to horizontal aspect ratios
will require larger eccentricities to limit falling, the eccentricity is not likely to be more than
about 1.5 times that of a spherical boulder.
Apart from rolling and falling, sliding of a boulder, which is initially resting on an
inclined surface, is another possible failure mode under earthquake shaking. In principle, this
mode of failure is essentially the same as the sliding failure of a slope, where the failure mass
is assumed as a rigid block. Thus, the same conclusions reached regarding the natural terrain
slope failure (e.g. the critical acceleration values shown in Figure 3.67) should also be
applicable to boulders when sliding failure is concerned. In addition, there is essentially no
chance of pore water pressure build up in boulders during rainstorms. Therefore, for a boulder
to fail in sliding mode, the static FoS against sliding will have to be very low at the time of
earthquake loading.
146
A regional landslide susceptibility model has been developed for the natural terrain areas
within the Study Area of this study. The susceptibility model adopts a landslide hazard
zonation approach to identify and document those areas considered the most susceptible to the
initiation of landslides. Varnes (1984) recommended three key principles for consideration in
the landslide hazard zonation process:
(b) The main conditions and processes that cause the landslides
(i.e. triggering factors) should be identified and understood.
To conduct the landslide susceptibility analysis with reference to the three principles, a
statistical approach has been adopted for this study. The Enhanced Natural Terrain Landslide
Inventory (ENTLI) has been used to represent the past instability locations. Some key general
possible susceptibility factors have been identified and the landslide density has been calculated
within each particular factor. The landslide density for each class of the initiation factors has
been compared to assess the importance of those factors which contribute to landslides. The
landslide susceptibility map has subsequently been generated by a matrix method which
included the selection and combination of different rankings of the selected susceptible factors.
Geographical Information System (GIS) has been used as a tool to analyse different
initiation factors based on the existing information. Each initiation factor has been compared
with the landslide density, which is considered as a direct measure of the probability of landslide
occurrence, to give the susceptibility ranking of each factor. The ranking of each factor has
been overlaid in the GIS to classify the susceptibility when combining all factors. The
reliability of the final susceptibility class has also been checked by the landslide density to
ensure that the classification could measure the expected probability of landslide occurrence.
The susceptibility map produced by analysing the ENTLI data only presents the
susceptibility of landslides triggered by rainfall. Over most of the area, this susceptibility is
also directly indicative of the susceptibility of the natural terrain to earthquake-induced
landslides as will be discussed in Section 5.5. For areas where the topographic effects are
applicable, there would be a slight increase in the earthquake-induced landslide susceptibility.
As a result, the overall susceptibility map has been modified.
147
In view of the regional scale of the present study and in order to develop a systematic
approach that could readily be adopted elsewhere within the territory of Hong Kong, the
developed approach has made use of existing available digital datasets. These include:
(i) Information on the location, size (plan area only) and age of
landslides from the GEO Enhanced Natural Terrain
Landslide Inventory (ENTLI), up to 2009 only.
(k) Rain gauge information for Gauge Nos. N07, N12, N28, N29,
N31, N32 and N49 for the years 1984 to 2010.
The above information provides a wealth of information concerning the location and the
scale of past instability in the Study Area as well as information on the topography, geology,
148
hydrology, geomorphology and extent of anthropogenic activity. Whilst these datasets are of
a variety of scales and resolutions, they are considered sufficiently accurate for a regional scale
assessment, the main purpose of which is identifying those areas of high susceptibility from
those with significantly lower susceptibility.
In addition, a natural terrain hazard study previously carried out for the Tsing Shan
Foothill Area by MFJV (2003) has been used as a further reference as their study area is a part
of the Study Area being assessed in the present study. Their work included a susceptibility
analysis and generation of site-specific susceptibility maps.
In order to identify the susceptibility of natural terrain landslides within the Study Area,
it is necessary to differentiate between those areas that comprise predominantly natural terrain
hillside (natural terrain) and those areas that have been significantly affected by anthropogenic
activities (urban area). The methodology adopted to differentiate between these two areas is
summarised below and presented graphically on Figures 5.1 to 5.6.
Geoprocessing Geoprocessing
LandsD B5000 Create 100m Buffer Buffer Create 100m inner Buffer Natural Hillside
Contour Line from Contour Line Polygon from buffered polygon Polygon
Group contoured area into one large polygon area with 100m buffer from 5 m contour line. The
100m buffer eliminates most gaps between contours and forms a continuous polygon of natural
terrain. Then create a 100m inner buffer to define a more precise outline for natural terrain.
Natural
Terrain Build-up
area
Geoprocessing
Create 100m Buffer from
Contour Line
GEO
features Road
Building Polygon
LandsD B1000
Building Polygon
50m Buffer
75m Buffer
Buffer
Create buffer from Building
and inner buffer from the
buffered polygon
LandsD B1000
Building Polygon
GEO Features
Catalogues
Remove man-made
features in the natural
terrain area Group
Form a
“Developed Area”
Buffer
Create 75m buffer
from Building
polygon and create The 75m buffer group all building
inner 75m buffer polygon together, the inner 75m
from the buffered buffer define a more precise outline
polygon for developed area
Selected Build-
up Area
Isolated small huts Select
(Area < 300m2) Select the build-up area
greater than 300m2 in area
Group
Form a “Developed Area”
Developed Area
Geoprocessing
Clip “Developed area”
“Developed Area”
within the natural
hillside polygon
Geoprocessing
Clip “Developed area”
Natural Hillside
Polygon
Before Before
Preliminary Natural
Terrain Polygon
Overlay
After After
Natural Terrain
Boundary
Manual Edit
Figure 5.6 Flow Chart of Generating Final Natural Terrain Boundary by Manual
Editing
The final natural terrain boundary presented in Figure 5.7 would be used for further
analyses of natural terrain landslide hazards. The data for those areas outside the boundary,
including those ENTLI data, would be discarded.
It should be noted that the natural terrain boundary is not the same as that for the
boundary of the rock site class defined in the seismic microzonation study (Arup, 2018) as their
definitions are different. Natural terrain is defined as the area that has not been developed,
whereas the rock site class is defined mainly based on slope gradient.
The landslide records within the ENTLI have been broadly classified into two groups
with regard to the date of occurrence, namely, recent landslides and relict landslides, defined
as follows.
Relict landslides are those that occurred earlier than the time
scale of the available aerial photographs. In general, they
refer to landslides that occurred before 1962.
To ensure a high degree of confidence in the susceptibility models derived for the Study
Area, calibration against the available landslide data has been carried out with reference to
following three groupings of data from the ENTLI dataset:
In general, there are some intrinsic differences between “Recent” and “Relict” landslides
in terms of the scale and correlation with geomorphological settings, especially when the highly
degraded Relict Class B and Relict Class C landslide features are concerned. However, the
typically sharp and well-defined nature of the back scarps for Relict A landslides suggests that
these features are quite young in terms of landform development and thus should be occurred
under similar climatic conditions to those present nowadays. In order to increase the number
of landslides for statistical analyses, both “Recent” and “Relict Class A” landslides have been
regarded as reliable. Based on the previous experience in reviewing the ENTLI data for natural
terrain hazard studies, these two datasets could typically be relied upon as being reasonably
accurate. On the other hand, Relict Class B and Relict Class C landslides could often be
questionable and could represent geomorphic features rather than actual landslide scars.
When reviewing the ENTLI data, it has been noted that the landslide source location
recorded in the dataset comprises point information (i.e. grid coordinates) at the crown of the
landslide scars. This means that the actual landslide scar itself is located downslope of the
coordinates recorded in the dataset and that direct adoption of this information to represent
landslide locations may result in inaccuracy and misinterpretation. In order to overcome this
problem, the source location for the landslide scar has been shifted 5 m downslope from the
crown and along the runout trail recorded in the ENTLI. This results in a modified ENTLI
dataset in GIS that is considered to more accurately represent the conditions of the area in which
actual landslides had taken place. The value of 5 m has been chosen because the average
landslide width is about 10 m to 12 m among the > 8,000 ENTLI records in the Study Area.
The radius of the landslide could generally be considered as being half of the width, i.e. about
155
5 m. Nevertheless, the results for both shifted and non-shifted data are presented in this study
for the sake of comparisons.
A total of 8,475 Recent and Relict (all classes) landslides within the natural terrain
boundary has been identified in the ENTLI, including 1,980 Recent landslides and 999 Relict
Class A landslides. As previously discussed, the remaining 5,496 ENTLI landslides, classified
as Relict Class B or Class C, are considered to be of lower reliability. As such, these landslides
have been excluded from the calibration exercise. The modified locations of the 2,979 Recent
and Relict Class A landslides recorded in the ENTLI up to 2009 are shown in Figure 5.9.
Figure 5.9 Location of 5 m down of the Crown in ENTLI Record up to 2009 within the
Natural Boundary
In addition to the location of landslides, the year in which the landslides were first
identified has also been reviewed. It is found that significant landslide swarms occurred within
the Study Area in 1982 (208 landslides), 1989 (284 landslides) and 2000 (730 landslides). A
plot of the total number of landslides occurring annually since 1963 is shown in Figure 5.10.
156
800
700
600
No. of landlside
500
400
300
200
100
0
1964
1968
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Figure 5.10 Annual Record of Number of Landslides from ENTLI Data
The data contained within the ENTLI, with modifications and groupings as discussed in
the preceding sections, has been used to determine the landslide density (in terms of the number
of landslides per unit area in km2) for various susceptibility factors. This information has been
used as a measure of the probability of landslide occurrence for rating various parameters to the
landslide susceptibility. Similar approach was used by MFJV (2003) to identify the key
initiation factors for landslides.
In the present study, the landslide density is expressed in terms of the area affected by
landslides per unit are in km2. This area is calculated using the landslide length and width
recorded in the ENTLI.
MFJV (2003) considered slope angle, regolith types, lithological boundaries, regolith
downslope of rock outcrop and the head of a drainage line as potential factors that affect the
landslide occurrence. With respect to the Study Area and considering the results of MFJV
(2003), the key initiation factors with the potential to predispose an area to landsliding mainly
include the geology, geological structure, topography, geomorphology, hydrology and
hydrogeology. The regolith types and lithological boundaries have not been considered
because they require detailed field mapping to obtain the information. While this was
applicable for the site-specific Tsing Shan Foothill Area, it is not appropriate for the present
study, which is of a regional scale. These two factors have been analysed indirectly through
the geological condition assessment.
157
These various parameters have largely been accounted for by consideration of the
following aspects within the susceptibility model.
The gradients of the natural terrain hillside within the Study Area have been calculated
in GIS based on a Triangulated Irregular Network (TIN) model generated from the
1:5,000 scale LIC Maps. The slope angles within the TIN model have subsequently been
“smoothed” to overcome potential problems associated with localised irregularities. The
“smoothing” process is presented as follows:
(i) A slope angle has been calculated for a 1 m cell size across
the whole Study Area.
(ii) A circular area within a radius of five cells has been used to
calculate the average value of each cell. This averaged slope
angle has then been adopted in that grid cell to provide a
smoothed slope angle map with a 1 m resolution.
The results of the slope angle smoothing exercise are presented in Figure 5.11. The
smoothing attempts to reduce the sensitivity of the slope angle determination arising from those
contours being extremely close together horizontally (note that the contour spacing of 1: 5,000
LIC map is 10 m vertically). Hence, it makes only a minor difference to the resulting slope
angle map with the smoothing only applied to very localised irregularities.
There are limitations in the use of the 1:5,000 LIC maps as they may not be as accurate
as the 1:1,000 LIC maps when the generation of slope angle maps is concerned. In order to
review the implications, a comparison of the landslide density for a number of key slope classes
using slope angle maps generated from both 1:1,000 LIC and 1:5,000 LIC maps have been
carried out for a small test area (about 3 km x 3 km) as shown in Figure 5.12. The results of
this review indicate that the densities computed using both LIC maps are largely comparable,
provided that the data for any areas less than 0.1 km2 have been screened out (see Figure 5.13).
158
Area
2
1.8 1k TIN 5k TIN
1.6
Area
Landslide density (per km2) 1.4
1.2
1
0.8
0.6
0.4
0.2
0
0 - 15 15 - 20 20 - 25 25 - 30 30 - 35 Recent
35 - 40 Landslide
40 - 45 45 - 50 50 - 55 55 - 60 60 - 65 65 - 70 70 - 75 80 - 85
400 Slope angle
1k tin 5k tin
350
Landslide density (per km2)
300
250 Recent
200
150
100
50
0
0 - 15 15 - 20 20 - 25 25 - 30 30 - 35 35Recent
- 40 40 -+45Relict A 50 - 55 55 - 60 60 - 65 65 - 70 70 - 75 75 - 80 80 - 85
45 - 50
600 Slope angle
1k tin 5k tin
500
Landslide density (per km2)
400
Recent +
Relict A
300
200
100
0
0 - 15 15 - 20 20 - 25 25 - 30 30 - 35 35All
- 40landslides
40 - 45 45 - 50 50 - 55 55 - 60 60 - 65 65 - 70 70 - 75 75 - 80 80 - 85
7000 Slope angle
1k tin 5k tin
6000
Landslide density (per km2)
5000
4000
Total
3000
2000
1000
0
0 - 15 15 - 20 20 - 25 25 - 30 30 - 35 35 - 40 40 - 45 45 - 50 50 - 55 55 - 60 60 - 65 65 - 70 70 - 75 75 - 80 80 - 85
Slope angle
Figure 5.13 Plots of ENTLI of Non-shifted Landslide Crown up to 2009 against Slope
Angle for TIN Model Generated from 1:1,000 (1 k) and 1:5,000 (5 k) LIC
Maps
With the above considerations, the use of the 1:5,000 LIC dataset is considered
acceptable for the present regional-scale assessment. However, for a detailed site-specific
study, the use of the 1:1,000 LIC or LiDAR datasets is recommended.
The landslide density of shifted and non-shifted landslide crown has then been reviewed
for each 5° slope angle increment within the natural terrain area (see Figures 5.14 and 5.15
160
respectively). The density of “Recent and Relict Class A” landslides is relatively low within
areas with slope gradients less than 30° and increases significantly for areas with gradients
above 40°. Whilst there is an apparent drop in the density of Recent and Relict Class A
landslides above slope angles of 50°, this can largely be attributed to small areas (< 1 km2)
occupied by these slope angle classes. Table 5.1 presents the susceptibility factor classes
recommended with regard to the influence of the slope angles.
Figure 5.14 Plots of ENTLI of Shifted Landslide Crown up to 2009 against Slope Angle
161
Figure 5.15 Plots of ENTLI of Non-shifted Landslide Crown up to 2009 against Slope
Angle
162
Figure 5.16 shows the distribution of different slope classes described above within the
Study Area.
The distribution of past landslides within the Study Area has been reviewed against the
published geological data at the landslide source area based on the 1:20,000 scale Solid and
Superficial Geological Map (Figure 5.17). The geological map shows geological boundaries,
different types of Quaternary deposits and rocks (which can be implied as different saprolites).
It is understood that the map could be oversimplified as compared with information obtained
from field mapping but it is still considered applicable for this regional study.
163
Figure 5.17 1:20,000 Scale Solid and Superficial Geology Map (GCO, 1987a & 1987b)
164
The results of this review indicate that the landslide densities are higher within areas
recorded as fine-grained granite (gf), fine-to-medium grained granite (gfm), block bearing tuff
and tuffite (bt), undifferentiated tuff and tuffite (t), siltstone with thin bedded limestone (Cmp)
and sandstone, siltstone and mudstone with conglomerate and tuff (JTS) (see Figure 5.18).
Figure 5.19 shows the equivalent non-shifted landslide results.
Area
250
200
150
km2
100
Area
50
rq
Cmp
ap
Qt
b
JTS
q
Cts
gfm
s
Qb
Qd
gf
JTU
gc
JSM
rf
t
gm
br
bt
Qpd
Qa
cg
KKO
JTM
Qpa
Jnl
14
20k Geological Unit
12
Landsldie dentisty (per km 2)
10
Recent + 6
Relict A 4
0 Recent landslides
Qb
Qd
JTU
gc
gf
gm
rf
t
Qpd
bt
rq
Cmp
Qa
cg
ap
JTM
Qt
b
JTS
Cts
Qpa
gfm
s
Jnl
JSM
br
KKO
7
20k Geological Unit
6
Landsldie dentisty (per km 2)
Recent 4
0 Total landslides
Qb
Qd
JTU
gc
gf
gm
rf
t
bt
Qpd
Cmp
rq
Qa
cg
ap
JTM
Qt
JTS
b
q
Cts
Qpa
gfm
s
Jnl
JSM
br
KKO
30
20k Geological Unit
25
Total
Landsldie dentisty (per km 2)
20
15
10
0
Qb
Qd
JTU
gc
gf
gm
rf
t
bt
Qpd
rq
Cmp
cg
Qa
JTM
ap
Qt
JTS
b
Cts
Qpa
gfm
s
Jnl
JSM
br
KKO
Area
250
200
150
km2
100
Area
50
0
Recent + Relict A landslides
rq
Cmp
ap
Qt
b
Cts
JTS
q
gfm
s
14
Qb
Qd
gf
JTU
gc
JSM
rf
t
gm
br
bt
Qpd
Qa
cg
KKO
JTM
Qpa
Jnl
12 20k Geological Unit
Landsldie dentisty (per km 2)
10
Recent + 6
Relict A 4
2
Recent landslides
0
7
Qb
Qd
JTU
gc
gf
rf
t
gm
Qpd
bt
rq
Cmp
Qa
cg
ap
JTM
Qt
JTS
b
Cts
Qpa
gfm
s
Jnl
JSM
br
KKO
6 20k Geological Unit
Landsldie dentisty (per km 2)
Recent 4
1
Total landslides
0
30
Qb
Qd
JTU
gc
gf
gm
rf
t
Qpd
bt
rq
Cmp
Qa
cg
ap
JTM
Qt
JTS
b
q
Cts
Qpa
s
gfm
Jnl
JSM
br
KKO
Total 20
15
10
0
Qb
Qd
JTU
gc
gf
gm
rf
t
Qpd
bt
rq
Cmp
Qa
cg
ap
JTM
Qt
JTS
b
Cts
Qpa
gfm
s
Jnl
JSM
br
KKO
Apparently, these geological units give rise to a higher predisposition to landsliding than
others. However, it has been found that the locations of these units are coincident with other
topographic and geomorphological indicators of landslide susceptibility. It is considered
reasonable as the development of the landforms is related to the underlying geological
characteristics. As such, other factors considered in the susceptibility model, in particular,
Slope Angle (see Section 5.5.1) and Terrain Unit (see Section 5.5.3), have already accounted
for the variance resulted from the solid geological conditions. On the other hand, the Study
Area is mainly exposed with granite (about 80% within the natural terrain boundary) and
166
slightly with tuff (only occupied about 5%). The weak sedimentary and volcanic rocks, which
have been considered to be more prone to landslide occurrence, are not considered in the present
study as the geology does not vary a lot spatially. To conclude, it is considered not necessary
to specifically include the geological strata in the susceptibility model.
The pertinent geological structure mainly includes faults and joints. For the purpose of
a regional-scale assessment, the analysis of joint sets is considered not applicable. Instead, the
fault proximity has been checked against the landslide density to assess if geological structure is
a controlling factor. The faults, inferred faults and photo-lineaments shown in 1:100,000 Hong
Kong Geological map (GEO, 2000) and GASP report III and IV (GCO, 1987a & 1987b) have
been used to represent the geological structure (Figure 5.20) and buffering of 10 m, 25 m, 50 m
and 100 m has been carried out to review whether there exists any strong correlation between the
location of past landslides and the proximity of the geological structure.
Figure 5.20 Geological Structure Extracted from 1:100,000 Hong Kong Geological Map
and the GASP Reports
167
The results of shifted and non-shifted landslide crowns, as shown in Figures 5.21 and
5.22 respectively, show that the landslide densities for all buffer distances are generally
comparable. It indicates that the geological structure is unlikely a controlling factor in this
case.
Area
Recent +
Relict A
Recent
Total
Area
Recent +
Relict A
Recent
Total
The geomorphological setting of the natural terrain within the Study Area has been
characterised using the terrain component of the Landform Maps from the Geotechnical Area
Studies Programme (GASP) Reports (Figure 5.23). The various terrain components
documented in these maps, which are generally based on the slope form and process, have been
reviewed and grouped to develop a simplified terrain model based on the method developed by
Dalrymple et al (1968) (Figure 5.24).
Rock Outcrop
Crest or Ridge
Sideslope – Straight
Sideslope - Convex
Sideslope – Concave
Drainage Plain
Various Footslopes & plains
Figure 5.24 The Hypothetical Nine-unit Land Surface Model by Dalrymple et al (1968)
and the Simplified Terrain Model in Our Study
The terrain units developed based on the published GASP Maps have been validated
using ortho-rectified aerial photographs and Digital Terrain Models (DTM) of the Study Area.
This process has confirmed that the model is broadly acceptable, with an exception of several
areas, which have previously been classified in the GASP Maps as footslope terrain but is
considered more representative of either incising or interfluve terrain (Figure 5.25).
1963 photo
C
B
2010 photo G
E
D
It is noted that, in a number of areas, the extent of the Incising Terrain does not fully
capture the areas of incised and over-steepened channel banks and valley side slopes associated
with significant drainage line activities. These inaccuracies in the model have been overcome
by expanding the Incising Terrain area into the surrounding Interfluve Terrain areas through
the application of a 10 m buffer.
The validation exercise has confirmed that whilst the published GASP Maps provide a
valuable starting point for the assessment of Terrain Units, careful review by an experienced
engineering geologist or geomorphologist is still required to validate the model.
A review of the landslide densities of shifted and non-shifted landslide crown for various
terrain units derived by the model shows that the Incising Terrain has a relatively higher
landslide density than Interfluve Terrain, which in turn has a higher density than the Footslope
172
Terrain (see Figures 5.26 and 5.27 respectively). The areas recorded as Rock Outcrop are also
noted to result in high landslide densities. With due consideration given to the above points,
three key Terrain Classes related to landslide susceptibility have been derived as summarised
in Table 5.3.
Figure 5.26 Landslide Density of Shifted Landslide Crown with Terrain Unit
173
Figure 5.27 Landslide Density of Non-shifted Landslide Crown with Terrain Unit
Figure 5.28 shows the distribution of the susceptibility classes for terrain unit within the
Study Area.
5.5.5 Drainage
Taking cognisance of the fact that landslides often occur at the head or along the banks
of natural drainage lines, the location of such features within the Study Area have been extracted
from 1:5,000 scale LIC Maps (Figure 5.29). These drainage lines have then been buffered at
distances of 10 m, 25 m, 50 m and 100 m to review whether there exist any correlations between
the location of past landslides and the proximity of the terrain to the natural drainage system.
Apart from drainage lines, the heads of drainage have also been buffered at radii of 10 m,
25 m, 50 m and 100 m to review if there exist any correlations with landslide densities
(Figure 5.30).
Based on this review, there appears no specific correlation between the landslide
locations of shifted and non-shifted landslide crowns and the proximity to drainage lines
(Figures 5.31 and 5.32 respectively) and to heads of drainage (Figures 5.33 and 5.34
respectively), with the landslide densities for all buffer distances considered showing broadly
comparable results. As such, it is considered that the Incising Terrain Unit derived under the
geomorphological assessment provides a more reliable means of defining the influence of the
hydrological system.
175
Figure 5.31 Landslide Density of Shifted Landslide Crown for Different Proximities to
Drainage Lines
177
Figure 5.33 Landslide Density of Shifted Landslide Crown for Different Buffers to the
Head of the Drainage Line
179
Figure 5.34 Landslide Density of Non-shifted Landslide Crown for Different Buffers to
the Head of the Drainage Line
180
A preliminary susceptibility model has been developed in order to combine the influence
of various susceptibility factors identified in the previous section as key contributors to the
predisposition of hillsides to landsliding. Based on a statistical review of various datasets
discussed in Section 5.5 against the published ENTLI data, the Slope Angle and the Terrain
Unit are considered as the most relevant indicators for a regional-scale screening of potential
landslide hazards. The combined influence of these two parameters has been assessed using a
matrix approach to provide a qualitative indicator of the landslide susceptibility within the
Study Area. Using this matrix (as shown in Table 5.4), a susceptibility map for the Study Area
has been generated in GIS to graphically present those areas falling within different landslide
susceptibility classes as shown in Figure 5.35.
Table 5.4 Combined Influence of Slope Angle and Terrain Unit on Landslide
Susceptibility
Figure 5.35 Susceptibility Map Based on Terrain Unit and Slope Angle
181
The landslide susceptibility model has been checked and calibrated against the available
landslide information contained in the ENTLI. As indicated below and in Figures 5.36 and
5.37 for shifted and non-shifted landslide records, the model shows a good correlation between
the qualitative landslide susceptibility class and the corresponding density of past landslides
within those areas. The corresponding landslide densities are listed in Table 5.5.
Area
60
50
Area 40
km2
30
20
10
Recent
Low +Moderate
Relict A (ENTLI
High up to High
Very 2009)Overall
350 Susceptability class
Landslide density (per km2)
300
250
Recent + 200
Relict A 150
100
50
0
Low Recent (ENTLI High
Moderate up to 2009)
Very High Overall
180 Susceptibility class
Landslide density (per km2)
160
140
120
Recent 100
80
60
40
20
0
Low Total (ENTLI up
Moderate to 2009)
High Very High Overall
1200 Susceptibility class
Landslide density (per km 2)
1000
800
Total 600
400
200
0
Low Moderate High Very High Overall
Susceptibility class
Figure 5.36 Landslide Density of Shifted Landslide Crown with Susceptibility Class up
to 2009
182
Area
60
50
40
Area
km2
30
20
10
0
Recent
Low +Moderate
Relict A (ENTLI
High up to High
Very 2009) Overall
400 Susceptability class
Landslide density (per km2)
350
300
250
Recent + 200
Relict A 150
100
50
0
Low Moderate High Very High Overall
Susceptibility class
Recent
Total
Table 5.5 Susceptibility Class and the Corresponding Density of Past Landslides
Further validation of the model has also been carried out by examining the ENTLI data
within 10-year intervals. The results of the shifted and non-shifted landslide densities,
presented in Figures 5.40 and 5.41 respectively, show that the qualitative landslide
susceptibility class gives very similar and consistent trends for each decade and reinforces the
stability over time of the susceptibility classes.
The total landslide area of each susceptibility class has been normalised to the total area
of each class and the time period of the landslide record to give the landslide occurrence
probability. The landslide area has been estimated by multiplying the length and width
recorded in the ENTLI inventory. As the recent ENTLI records started from 1963 i.e. when
aerial photographs were available, and those landslide scars occurred one to two years before
1963 as observed in the aerial photographs were also classified as recent landslides, it follows
that the time period for recent landslides up to 2009 is approximately equal to 50 years.
Figures 5.42 and 5.43 present the landslide area as a fraction of the total area for each
susceptibility class for shifted and non-shifted landslides respectively for the ENTLI data
recorded in that 50-year period.
184
Figure 5.38 Landslide Density of Shifted Landslide Crown with Susceptibility Class in
Particular Years
185
1970 - 1979
1980 - 1989
1990 - 1999
2000 - 2009
Figure 5.40 Landslide Density of Shifted Landslide Crown with Susceptibility Class in
10 Year Intervals
187
1970 - 1979
1980 - 1989
1990 - 1999
2000 - 2009
Figure 5.42 Ratio of Landslide Area to Total Area of Shifted Landslide Crown with
Susceptibility Class for Recent Landslides up to 2009
Figure 5.43 Ratio of Landslide Area to Total Area of Non-shifted Landslide Crown
with Susceptibility Class for Recent Landslides up to 2009
189
Wong & Ho (2000) discussed the effects of earthquakes on man-made slopes and
addressed the issue of the occurrence of seismic ground motions concurrently with a slope being
in a weakened state due to precedent rainfall. Based on the wetting front theory, it was
concluded that if the daily rainfall exceeded 50 mm, there would be a moderate degree of
saturation, and if the daily rainfall exceeded 380 mm, there would be a high degree of saturation.
By studying the rain gauge data for Hong Kong, it was concluded that there would be a 4.5%
likelihood of moderate saturation, and a 0.5% likelihood of high saturation. In the remainder
of the time (i.e. 95% likelihood), the slope would be at a low degree of saturation. It was also
concluded that the static FoS of the slope would be enhanced by about 0.3 for low saturation,
and by 0.15 for moderate saturation.
190
Figure 6.1 Soil and Rock Distribution within Natural Terrain Boundary
In this study, the durations and the static FoS reported by Wong & Ho (2000) have been
combined with the observed failure rates from the ENTLI data as shown in Figure 6.3. For the
Average curve, it is assumed that the FoS reduces to 1.05 when the soil is at a high level of
saturation for 0.5% of the time, and the FoS is 1.2 and 1.35 at 5% and 20% of the time
respectively. For the remaining 80% of the time, the FoS is much larger and beyond the level
of concern. For the purpose of plotting the ENTLI data on a logarithmic scale, it is assumed
that the soil slopes actually fail when their FoS is 1.001. The “% time” values plotted on the
x axis are those observed for various susceptibility classes (see Figure 5.42) divided by the 50-
year observation period. These values along the x-axis (i.e. at a FoS of 1.001) is interpreted as
the annual chance that any particular location will fail due to rainfall. These points have then
been extrapolated over time to show the likely FoS as a function of duration for each of the
susceptibility class.
Figure 6.3 Factor of Safety and Percentage Time Relationship for the Susceptibility
Classes
Figure 6.3 can be used to estimate the additional likelihood of failure induced by seismic
ground motions. For example, for a 2% in 50 years ground motion with no topographic
amplification, Figure 3.50 implies that, on average, slopes with an Ac value of 0.15 m/s2 would
be at the onset of failure. This Ac value corresponds to a FoS of about 1.03 (see Figure 6.3).
It can be seen to have an average chance of occurrence for 0.3% of the time. It, therefore,
implies that over 50 years, the enhanced hazard of slope failure is 2% times 0.3% i.e. 0.006%.
192
The effect of how likely earthquakes would take place also needs to be considered.
Figure 3.51 shows that under the 10% in 50 years ground motion, a slope with a FoS of about
1.01 would be at the onset of failure. Based on Figure 6.3, this FoS is likely to exist over about
0.08% of the time. By approximately integrating the ground motions between 10% and 2% in
50 years, the overall hazard is increased by about 0.018% in the next 50 years. Given that the
average landslide area ratio hazard from rainstorms in the past 50 years is 0.25% at present (see
Figure 5.42), the overall hazard in the next 50 years is therefore likely to increase to 0.268%.
For areas assessed as being of very high susceptibility, using the same approach, it has
been evaluated that the landslide area ratio hazard is increased by about 4 times to a value of
0.07% in the next 50 years. As a result, the overall hazard increases from 1.08% to 1.15% in
the next 50 years. Similar assessments have been undertaken for other susceptibility classes
and the results are summarised in Figure 6.4. This figure also shows the additional influence
of topographic effects derived using the same procedures. For example, for the case of 1.4
PGA amplification, Figure 3.67 shows that the FoS needs to be about 1.07 for the onset of
failure for the 2% in 50 years ground motion. From Figure 6.3, thitaFoS would occur at about
4% of the time for very highly susceptible slopes leading to an enhanced landslide area ratio
hazard of 0.08%. For more frequent seismic ground motions, it is increased to about 0.24%
leading to a total value of 1.32% as shown in Figure 6.4.
Figure 6.4 Landslide Area Ratio in 50 Years for Different Amplification Factors and
Susceptibility Classes
These enhanced overall susceptibility values have been added to the susceptibility map
in Figure 5.23 and plotted as colours representing the likelihood of failure in the next 50 years
(in steps of 0.1%). The resulted natural terrain landslide microzonation map is shown in
Figure 6.5. For comparison, the same contours have been mapped for rainstorms only as
shown in Figure 6.6.
193
Figure 6.5 Susceptibility Map for Landslide Density Considering Earthquake and
Rainfall Effect
The increase in the overall landslide hazard has been determined by multiplying the
natural terrain hazard factors shown in Figure 6.4 by the areas associated with each
classification and the pertinent PGA amplification factor. Table 6.1 shows these areas of
natural terrain. Table 6.2 shows the total area of natural terrain that is likely to fail within the
Study Area together with the overall average hazard in the next 50 years for cases where seismic
loading is ignored, and where it is considered. It can be seen that the overall hazard in the next
50 years increases from the rainfall-induced value of 0.244% to 0.261% if seismic effects are
included. It should be noted that this modest increase is less than one tenth and within the error
of prediction for the amount of failure due to rainfall alone.
Table 6.1 Areas of Natural Terrain (m2) Associated with Susceptibility Class and with
PGA Amplification Factor
Table 6.2 Area of Natural Terrain Likely to Fail (m2) and Average % Area Ratio in the
Next 50 Years
Seismic Effect
Susceptibility Class
None Expected Worst credible
Low 8,574 9,163 10,337
Moderate 41,354 44,225 49,729
High 54,615 58,424 66,119
Very High 12,734 13,770 15,852
2
Total (m ) 117,277 125,582 142,037
Av. Area Ratio 0.244% 0.261% 0.296%
The microzonation map shown in Figure 6.5 can be considered to be the best estimate.
As shown in Figure 6.4, the effect of seismic ground motions, compared with the effect of
rainfall only, is relatively small and does not have a major effect on the overall zoning. It is,
however, admitted that many assumptions have been made in combining the effects of rainfall
and earthquake. There is also uncertainty in the derivation of the effects of seismic ground
motions.
195
To address this issue, the maximum credible effect has been assessed. This has been
considered by assuming that the FoS of an average slope reduced from 1.05 to 1.02 for 0.5% of
the time and that the 5% value has reduced from 1.2 to 1.1. The effect of this on the plot of
FoS against % time is shown in Figure 6.7. This adjustment effectively increases the effects
of the seismic ground motion by a factor of three. The resulted landslide area ratio hazard in
the next 50 years is shown in the lower diagram of Figure 6.8. The best estimate values derived
previously are shown in the upper diagram of the figure as a direct comparison. As can be
seen, the overall result is not greatly changed. However, for the peak values for areas of high
and very high susceptibility with topographic effects, some noticeable increases are observed.
The resulted seismic zonation map, drawn using the same contours as Figures 6.5 and 6.6, is
shown in Figure 6.9.
The area of natural terrain likely to fail and the overall average hazard in the next 50
years under this worst credible seismic effect is shown in Table 6.2. It can be seen that the
current expected hazard of 0.244% increases by 0.052% (or by about one fifth) to 0.296%.
Figure 6.7 Worst Credible Factor of Safety and Percentage Time Relationship
196
Figure 6.8 Landslide Area Ratio in 50 Years for the Maximum Credible Seismic Effects
Compared with the Best Estimate
197
Figure 6.9 Susceptibility Map Considering the Maximum Credible Effect from Seismic
Ground Motion
A large number of landslides occurred in steeply sloping terrain in Taiwan due to the
1999 Chi-Chi earthquake, which led to very large ground motions in the vicinity of the fault
rupture. This was also observed in the 2008 Sichuan earthquake and in the 1990 Philippines
fault event. Nevertheless, these levels of ground motion are too extreme to realistically
represent the situation in Hong Kong. Also, according to Arup (2015), the bedrock PGA cross
the Study Area ranged from about 0.89 m/s2 to 0.92 m/s2 and 1.94 m/s2 to 2 m/s2 for 10% and
2% probabilities of being exceeded in the next 50 years. This means that the variation of the
ground motion across the territory is small. Hence, both the ground motion and the distance
to the fault rupture, which have been observed to be major factors affecting earthquake-induced
landslides in past major events, are not a concern in the present study.
In this study, the 2% in 50 years ground motion is considered to be the most extreme
ground motion for Hong Kong. This ground motion corresponds to a magnitude 5.5
earthquake with its epicentre within the Study Area (Arup, 2015). Such an event is not likely
to cause a fault rupture at the surface. At the same time, there is no clear evidence of any active
faults that exist in Hong Kong (Sewell & Tang, 2012). Hence, even for such case, it is
considered that widespread landslides similar to those observed in the three afore-mentioned
major events will still not occur. The focus of this study should be remained on hillsides that
are already close to failure.
198
The present study addresses the direct effect of earthquake on the natural hillside
stability. Its effect on the post-earthquake susceptibility of rainfall-induced natural terrains
landslides due to degradation and distress of hillsides brought about by the earthquake has not
been examined. In addition, any debris resulted from earthquake-induced landslides that may
be deposited on natural hillsides is liable to re-mobilisation by subsequent rainstorms. As such,
the potential increase of landslide hazard after earthquake should not be neglected. Yin (2008)
described the post-earthquake situations on slopes with regard to the 2008 Wenchuan
Earthquake in China. He studied the number of events in the affected area before and after
the earthquake event and found that there was a large number of slopes considered to be
unstable as a consequence of the ground shaking, according to the Chinese Survey.
However, some studies suggested that the effects of earthquake disturbance on slopes
would be prolonged but in a diminishing rate. Ching et al (2012) studied the disturbance effect
of the 1999 Chi-Chi Earthquake on slopes by observing the number of landslides during several
major rainstorms after the earthquake event. It was found that although the amount of rainfall
in Typhoon Mindulle (July 2004) was much higher than that in Typhoon Toraji (July 2001) and
Nari (Sept. 2001), the number of rain-induced landslides in the former event was much less
than those in the latter. This implied that the earthquake disturbance effect could be transient
in time and diminish with time. It was concluded that for the purpose of assessing the
landslide susceptibility in the long term, such an effect could be ignored. Similar findings
were proposed by Lin et al (2008), who suggested that the effects of earthquake could be
prolonged but in a diminishing rate. Clearly, continual observations in future earthquake
events will enhance the understanding of how long such adverse effects on slopes may persist
after major earthquakes.
In addition to the potential of natural terrain landslides within the Study Area, it is
recognised that a number of surface boulders and rock outcrops are present within the natural
terrain areas. A screening of potential rock fall and boulder fall susceptibility has, therefore,
been carried out.
This assessment relied primarily on the information contained within the GEO’s
Boulder Inventory, which records the density of boulder coverage for hillside areas
throughout the whole territory of Hong Kong, and the Slope Angle maps developed as part
of the landslide susceptibility assessment. A detailed inventory of boulder fields on natural
slopes in Hong Kong was documented by Emery (1998). This inventory was prepared using
aerial photograph interpretation and documented the percentage area covered by boulders as
well as the approximate boulder types, typical sizes and shape.
Given that no details of boulder fixity or stability can be inferred from the Boulder
Inventory, the potential of boulder fall has been assessed on the basis of the density of boulders
present and the gradient of the hillside areas within which they are located. As for the landslide
susceptibility assessment, the combined influence of these two parameters has been assessed
following a matrix approach (Table 6.1).
199
Table 6.1 Combined Influence of Slope Angle and Boulder Density on Boulder Fall
Susceptibility
The boulder density is presented in Figure 6.10. The rock fall and boulder fall
susceptibility map for the Study Area has been generated in GIS to graphically present those
areas falling within different susceptibility classes as shown in Figure 6.11.
7 Conclusions
The earthquake-induced natural terrain landslide hazard assessment for a Study Area in
the North-west New Territories of Hong Kong has been carried out in this study. The findings
of the literature review on local and overseas state-of-the-art papers have been used to
consolidate the methodology applied in this study for both natural terrain landslide and boulder
fall hazard assessments.
For the seismic stability of boulder fall, a series of simplified one-dimensional dynamic
201
time stepping analyses have been carried out to determine how close to tipping a boulder (of
various sizes) needs to be to trigger instability under the 2%, 10% and 63% in 50 years ground
motions. The topographic effects of 30 m, 100 m, and 300 m ridge heights have also been
considered for the 2% and 10% in 50 years seismic ground motions. The analyses also cover
the combined effect of horizontal and vertical accelerations (without topographic effect) for 2%
in 50 years ground motions. It is found that the boulders have to be close to failure at the time
of earthquake should there be any significant earthquake-induced boulder fall hazard.
A seismic microzonation map has been prepared for natural terrain landslide
susceptibility to reflect the combined effects of rainfall and seismic ground motions. This
relies on a detailed landslide susceptibility mapping derived from past landslides based on the
ENTLI database. The observed effects from rainfall have been combined with the calculated
seismic ground motion effects by considering how much of the time the natural terrain is close
to failure as a result of rainstorms when the seismic ground motion occurs. The assessment is
based on the assumptions made by Wong & Ho (2000), who studied the effects of earthquakes
on man-made slopes. Overall, the effect of earthquakes on the triggering of landslides in
natural terrain is small. The increase in failure probability is found be to less than one tenth of
that expected from rainstorms alone. Apparently, there is uncertainty in some of those
assumptions made. Thus, a sensitivity study has been conducted to illustrate the maximum
credible effect of seismic ground motions on the natural terrain landslide hazard. On the basis
that the slopes must be very close to failure at the time of earthquake, it is found that the number
and size of landslides would probably be similar to those induced solely by rainfall. In the
worst scenario, the hazard would only be marginally larger.
A microzonation map has also been produced in qualitative terms for boulder fall and
rock fall susceptibility based on the boulder density and the overall slope angle. It is found
that this map is not changed even if seismic ground motions are considered.
The scope of the present study does not include the examination of the degradation of
slopes or hillsides, which do not collapse or detach. In addition, the debris of earthquake-
induced landslides that may be deposited on natural hillsides is liable to be re-mobilised by
subsequent rainfall. Therefore, the increase in the likelihood of rain-induced landslides after
an earthquake should not be neglected. It is believed that a ‘reactive approach’ including
inspection and mapping of potential slope distress and major debris deposits close to
developments could be a pragmatic approach to deal with the post-earthquake landslide hazard.
8 References
Abrahamson, N.A. & Silva, W.J. (1996). Empirical Ground Motion Models. Report
prepared for brookhaven national laboratory, New York, May, 144 p.
AGS (2007). Guideline for landslide susceptibility, hazard and risk zoning for land use
management. Australian geomechanics society landslide taskforce landslide zoning
working group. Australian Geomechanics, 42 (1), pp 13-36.
Arup (2015). Seismic Hazard Analysis of the Hong Kong Region (GEO Report No. 311).
Geotechnical Engineering Office, Hong Kong, 327 p.
202
Arup (2018). Report on the Seismic Microzonation Assessment of the North-west New
Territories (GEO Report No. 338). Geotechnical Engineering Office, Hong Kong,
583 p.
Ashford, S.A., Sitar, N., Lysmer, J. & Deng, N. (1997). Topographic effects on the seismic
response of steep slopes. Bulletin of the Seismological Society of America, vol. 87,
no. 3, pp 701-709.
Au-Yeung, Y.S. & Ho, K.K.S. (1995). Gravity Retaining Walls Subject to Seismic Loading
(GEO Report No. 45). Geotechnical Engineering Office, Hong Kong, 63 p.
Blake, T.F., Hollingsworth, R.A. & Stewart, J.P. (2002). Recommended Procedures for
Implementation of DGM Special Publication 117 Guidelines for Analyzing and
Mitigating Landslide Hazards in California. ASCE and Southern California
Earthquake Center publication.
Ching, J., Liao, H.J. & Lee, J.K. (2012). Prediction rainfall-induced landslide potential along
a mountain road in Taiwan. Geotechnique, vol. 62, issue 6, pp 555-561.
Chung, C.F. & Fabbri, A.G. (2003). Validation of spatial prediction models for landslide
hazard mapping. Natural Hazards, vol. 30, pp 451-472.
Cole, W.F., Marcum, D.R., Shires, P.O. & Clark, B.R. (1998). Analysis of
Earthquake-reactivated Landslides in the Epicentral Region, Central Santa Cruz
Mountains, California. U.S. Geological Survey Professional Paper 1551-c,
pp C165-C185.
Dalrymple, J.B., Blong, R.J. & Conacher, A.J. (1968). A hypothetical nine unit land surface
model. Zeitschrift fur Geomorphologie, vol. 12, pp 60-76.
Darragh, B., Silva, W. & Gregor, N. (2004). Strong motion record processing for the PEER
Center. Proceedings of COSMOS Invited Workshop on Strong-Motion Record
Processing, Richmond, California, USA, pp 26-27.
Davis, J.C. (2002). Statistics and Data Analysis in Geology. 3rd ed. John Wiley & Sons,
638 p.
Ding, Y.Z. (2008). Risk of Earthquake-induced Landslides in Hong Kong. Technical report.
(in Chinese)
Dobry, R., Idriss, I.M. & Ng, E. (1978). Duration characteristics of horizontal components of
strong motion earthquake records. Bulletin of the Seismological Society of America,
vol. 68, no. 5, pp 1487-1520.
203
Emery, K.A. (1998). Boulder Study of Hong Kong - Final Report. Maunsell Geotechnical
Services Ltd. Geotechnical Engineering Office, Hong Kong, 12 p.
Gaudio, V.D. & Wasowski, J. (2011). Advances and problems in understanding the seismic
response of potentially unstable slopes. Special Edition of Journal of Engineering
Geology, Toward the Next Generation of Research on Earthquake-induced Landslides:
Current Issues and Future Challenges, vol. 122, pp 73-83.
GCO (1987a). Geotechnical Area Studies Programme (GASP) Report III, West New
Territories. Geotechnical Control Office, Hong Kong, 155 p. plus 4 maps.
GCO (1987b). Geotechnical Area Studies Programme (GASP) Report IV, North West New
Territories. Geotechnical Control Office, Hong Kong, 120 p. plus 3 maps.
Geli, L., Bard, P.-Y. & Jullien, B. (1988). The effect of topography on earthquake ground
motion: a review and new results. Bulletin of the Seismological Society of America, vol.
78, no. 1, pp 42-63.
GEO (2000). Geological Map of Hong Kong - Millennium Edition, January. Geotechnical
Engineering Office, Hong Kong.
Halcrow (2009). Scoping Study of Natural Terrain Landslide Risk in Hong Kong from a
Maximum Credible Earthquake. Prepared by Halcrow China Limited for Geotechnical
Engineering Office, Hong Kong, 110 p.
Hancox, G.T., Perrin, N.D., Dellow, G.D. (1997). Earthquake-induced Landsliding in New
Zealand and Implications for MM Intensity and Seismic Hazard Assessment. GNS
Client Report 43601B.
Hancox, G.T., Perrin, N.D. & Dellow, G.D. (2002). Recent studies of historical earthquake-
induced landsliding, ground damage, and MM intensity in New Zealand. Bulletin of
the New Zealand Society for Earthquake Engineering, vol. 35, no. 2, pp 59-94.
Hancock, J., Watson-Lamprey, J., Abrahamson, N.A., Bommer, J.J., Markatis, A., McCoyh, E.
& Mendis, R. (2006) An Improved Method of Matching Response Spectra of Recorded
Earthquake Ground Motion Using Wavelets. Vol 10, Issue sup001, 67-89.
Hsieh, S.Y. & Lee, C.T. (2011). Empirical estimation of the Newmark displacement from the
arias intensity and critical acceleration. Special Edition of Journal of Engineering
Geology, Toward the Next Generation of Research on Earthquake-induced Landslides:
Current Issues and Future Challenges, vol. 122, pp 34-42.
ISSMFE (1993). Manual for Zonation on Seismic Geotechnical Hazards. The Japanese
Society of Soil Mechanics and Foundation Engineering.
Jibson, R.W. (2011). Methods for assessing the stability of slopes during earthquakes - a
retrospective. Special Edition of Journal of Engineering Geology, Toward the Next
Generation of Research on Earthquake-induced Landslides: Current Issues and Future
Challenges, vol. 122, pp 43-50.
Jibson, R.W., Harp, E.L. & Michael, J.M. (1998). A Method for Producing Digital
Probabilistic Seismic Landslide Hazard Maps: An Example from the Los Angeles,
California Area. U.S. Geological Survey Open-File Report 98-113, 17 p.
Jibson, R.W. & Keefer, D.K. (1993). Analysis of the seismic origin of landslides: examples
from the New Madrid seismic zone. Geological Society of America Bulletin, vol. 105,
pp 521-536.
Kao, H. & Chen, W.P. (2000). The Chi-Chi earthquake sequence: active out-of-sequence
thrust faulting in Taiwan. Science, vol. 288, pp 2346-2349.
Keefer, D.K. (1999). Earthquake-induced landslides and their effects on alluvial fans.
Journal of Sedimentary Research, vol. 69, pp 84-104.
Keefer, D.K. (2007). Landslides caused by recent earthquakes: anomalies and relations to
general trends. North American Conference on Landslides, Vail, Colorado, June 3-8,
pp 876-885.
205
Keefer, D.K. & Wilson, R.C. (1989). Predicting Earthquake-induced Landslides, with
Emphasis on Arid and Semi-arid Environments. Landslides in a semi-arid environment,
Inland Geological Survey Society 2, pp 118-149.
Kirkby, M.J. (1975). Hydrograph Modelling Strategies. In: Peel, R., Chisholm, M. &
Haggett, P. (Eds.), Process in Physical and Human Geography, Heinemann, London,
pp 69-90.
Lee, C.T., Huang, C.C., Lee, J.F., Pan, K.L., Lin, M.L. & Dong, J.J. (2008). Statistical
approach to earthquake-induced landslide susceptibility. Engineering Geology,
vol. 100, pp 43-58.
Liao, H.W. & Lee, C.T. (2000). Landsides triggered by the Chi-Chi earthquake. Proc. 21st
Asian Conference on Remote Sensing, vols. 1 & 2, pp 383-388.
Lin, M.L., Wang, K.L. & Kao, T.C. (2008). The effects of earthquake on landslides - a case
study of Chi-Chi earthquake, 1999. Special Lecture 6, 10th International Symposium
on Landslides and Engineered Slopes.
Lin, P.S. & Lee, C.T. (2003). Arias intensity in landslide susceptibility analysis. Proc. 2003
Annual Meeting of Chinese Geophysical Society (Taiwan), pp 91-96 (in Chinese with
English abstract) of Geotechnical Engineering, vol. 112, No. 1, pp 44-59.
Lin, Y.H. (2003). Application of Neural Networks to Landslide Susceptibility Analysis. M.S.
thesis of institute of applied geology, National Central University, 81 p. (in Chinese
with English abstract)
Ma, K.F., Lee, C.T. & Tsai, Y.B. (1999). The Chi-Chi, Taiwan Earthquake: Large Surface
Displacements on An Inland Fault. EOS Transactions, American Geophysical Union,
vol. 80 (50), pp 605-611.
MFJV (2003). Landslide susceptibility analysis. Agreement No. 47/2000. Natural Terrain
Hazard Study for Tsing Shan Foothill Area. Prepared by Maunsell-Fugro Joint
Venture. Geotechnical Engineering Office, Hong Kong.
Mora, S. & Vahrson, W-G. (1994). Macrozonation methodology for landslide hazard
determination. Bulletin of the Association of Engineering Geologists, vol. 31, no. 1,
pp 49-58.
Murphy, W. (2006). The role of topographic amplification on the initiation of rock slopes
failures during earthquakes. In: Evans, et al (Ed.), Landslides from Massive Rock Slope
Failure, pp 139-154.
Parker, R.N., Densmore, A.L., Rosser, N.J., Michele, M.D., Li, Y., Huang, R., Whadcoat, S. &
Petley, D.N. (2011). Mass wasting triggered by the 2008 Wenchuan earthquake is
greater than orogenic growth. Nature Geoscience, vol. 4, pp 449-452.
206
Paruelo, J.M., Garbulsky, M.F., Guerschman, J.P. & Jobbagy, E.G. (2004). Two decades of
normalized difference vegetation index changes in South America: identifying the
imprint of global change. International Journal of Remote Sensing, vol. 25, no. 14, pp
2793-2806.
Rathje, E.M., Abrahamson, N.A. & Bray, J.D. (1998). Simplified Frequency Content
Estimates of Earthquake Ground Motions. Journal of Geotechnical Engineering,
ASCE, vol. 124, no. 2, pp 150-159.
Rathje, E.M. & Antonakos, G. (2011). A unified model for predicting earthquake-induced
sliding displacement of rigid and flexible slopes. Special Edition of Journal of
Engineering Geology, Toward the Next Generation of Research on Earthquake-induced
Landslides: Current Issues and Future Challenges, vol. 122, pp 51-60.
Rodriguez, C.E., Bommer, J.J. & Chandler, R.J. (1999). Earthquake-induced landslides: 1980-
1997. Soil Dynamics and Earthquake Engineering, vol. 18, pp 122-127.
Sewell, R.J. & Tang, D.L.K. (2012). The Potential Evidence for Neotectonic Fault Movement
and Correlation with Natural Terrain Landslides in Hong Kong. (Geological Report No.
GR 4/2012). Geotechnical Engineering Office, Hong Kong, 33 p.
Tang, D.L.K., Ding, Y.Z., Lee, C.W., Wong, J.C.F. & Sewell, R.J. (2009). Study of the
Potential Evidence for Neotectonic Movement and Correlation with Natural Terrain
Landslides in Hong Kong, Part 1: Ho Lek Pui Area. (Geological Report No.
GR 1/2009). Geotechnical Engineering Office, Hong Kong, 44 p.
Tang, D.L.K., Sewell, R.J., Wong, J.C.F. & Ding, Y.Z. (2010). Study of the Potential Evidence
for Neotectonic Movement and Correlation with Natural Terrain Landslides in Hong
Kong, Part 3: Tung Chung East. (Geological Report No. GR 3/2010). Geotechnical
Engineering Office, Hong Kong, 31 p.
Travasarou, T., Bray, J.D. & Abrahamson, N.A. (2003). Empirical attenuation relationship for
arias intensity. Earthquake Engineering and Structural Dynamics, vol. 32, pp 1133-
1155.
Varnes, D.J. (1984). Landslide Hazard Zonation: A Review of Principals and Practice.
UNESCO Press, Paris, 63 p.
Wang, K-L. & Lin, M-L. (2010). Development of shallow seismic landslide potential map
based on Newmark’s displacement: the case of Chi-Chi earthquake, Taiwan.
Environment Earth Sciences, vol. 60, pp 775-785.
Wasowski, J., Keefer, D.K. & Lee, C.T. (2011). Toward the next generation of research on
earthquake-induced landslides: Current issues and future challenges. Engineering
Geology, vol. 122, pp 1-8.
207
Wieczorek, G.F., Wilson, R.C. & Harp, E.L. (1985). Map Showing Slope Stability During
Earthquakes in San Mateo County California. U.S. Geological Survey Miscellaneous
Investigations Map I-1257-E, scale 1:62,500.
Wilson, J.P. & Gallant, J.C. (2000). Terrain Analysis. John Wiley & Sons, Inc., 479 p.
Wilson, R.C. (1993). Relation of Arias Intensity to Magnitude and Distance in California.
U.S. Geological Survey Open-File Report 93-556.
Wilson, R.C. & Keefer, D.K. (1983). Dynamic analysis of a slope failure from the 6 August
1979 Coyote Lake, California, earthquake. Bulletin of the Seismological Society of
America, vol. 73, pp 863-877.
Wilson, R.C. & Keefer, D.K. (1985). Predicting areal limits of earthquake-induced landsliding.
In: Ziony, J.I. (Ed.), Evaluating Earthquake Hazards in the Los Angeles Region An
Earth-science Perspective: U.S. Geological Survey Professional Paper 1360, pp 316-
345.
Wong, H.N. & Ho, K.K.S. (2000). Preliminary Quantitative Risk Assessment of Earthquake-
induced Landslides at Man-made Slopes in Hong Kong. (GEO Report No. 98),
Geotechnical Engineering Office, Hong Kong, 69 p.
Wong, H.N. & Pang, P.L.R. (1991). Assessment of Stability of Slopes Subjected to Blasting
Vibration. (GEO Report No. 15). Geotechnical Engineering Office, Hong Kong,
112 p.
Wong, J.C.F. & Ding, Y.Z. (2010). Study of the Potential Evidence for Neotectonic Movement
and Correlation with Natural Terrain Landslides in Hong Kong, Part 2: Wong Chuk
Yeung Area. (Geological Report No. GR 1/2010). Geotechnical Engineering Office,
Hong Kong, 40 p.
Wong, J.C.F., Tang, D.L.K., Sewell, R.J. & Lee, C.W. (2010). Study of the Potential Evidence
for Neotectonic Movement and Correlation with Natural Terrain Landslides in Hong
Kong, Part 4: Nam Shan and Pui O Areas. (Geological Report No. GR 4/2010).
Geotechnical Engineering Office, Hong Kong, 24 p.
Yang, J. (2007). On seismic landslide hazard assessment. Geotechnique, vol. 57, no. 8,
pp 707-713.
Yasuda, S. & Sugitani, T. (1988). Case histories of slope failure during past earthquakes in
Japan. Proc. 23rd ISSMFE, pp 891-892. (in Japanese)
Writing to 書面訂購
Publications Sales Unit, 香港北角渣華道333號
Information Services Department, 北角政府合署6樓626室
Room 626, 6th Floor,
政府新聞處
North Point Government Offices,
刊物銷售組
333 Java Road, North Point, Hong Kong.
or 或
Calling the Publications Sales Section of Information Services 致電政府新聞處刊物銷售小組訂購 (電話:(852) 2537 1910)
Department (ISD) at (852) 2537 1910 進入網上「政府書店」選購,網址為
Visiting the online Government Bookstore at
http:// www.bookstore.gov.hk http://www.bookstore.gov.hk
Downloading the order form from the ISD website at 透過政府新聞處的網站 (http://www.isd.gov.hk) 於網上遞交
http://www.isd.gov.hk and submitting the order online or by 訂購表格,或將表格傳真至刊物銷售小組 (傳真:(852) 2523
fax to (852) 2523 7195 7195)
Placing order with ISD by e-mail at [email protected] 以電郵方式訂購 (電郵地址:[email protected])
1:100 000, 1:20 000 and 1:5 000 geological maps can be 讀者可於下列地點購買1:100 000、1:20 000及1:5 000地質圖:
purchased from: