Predicting the performance of residual soil slopes

Prédiction du fonctionnement des talus des sols résiduels

EWBRAND, Head, Geotechnical Control Office, Engineering Development Department, Hong Kong

SYNOPSIS Residual soil slopes are com on in many parts of the world, but little attention
has been paid to appropriate methods of analysis and design. The basic approaches available for
dealing with these heterogeneous materials are reviewed in this Paper and details are given of some
recent developments in design methods, with particular reference to the design and performance of
residual soil cut slopes in Hong Kong. Examination of the relationship between rainfall and land­
slides for Hong Kong shows that short-term intensity is almost solely responsible for the many slope
failures, antecedent rainfall being of little account. The state-of-the-art with respect to the
application of soil mechanics methods to the analysis of residual soil slopes is reviewed, and the
main difficulties are examined with reference to a number of Hong Kong case histories. It is con­
cluded that our soil mechanics predictive tools are far from adequate for analysing residual soil
slopes, largely because of the difficulties of predicting pore pressures and of modelling geological
detail which often controls the mode of slope failure.

1, INTRODUCTION (b) they are generally very heterogeneous, which

makes them difficult to sample and test,
This Theme Lecture was originally conceived as
covering the broad field encompassed by the (c) they are nearly always unsaturated, which
title "Comparison of Prediction and Performance poses considerable difficulties for shear
of Earth Structures". During the preparatory strength measurements, and
stages, however, the Author became aware that
this was too large an undertaking for either the (d) they invariably have high permeabilities,
oral or written version of the Lecture. The which makes them subject to rapid changes
scope was therefore dramatically reduced to in material properties because of external
enable concentration on a single aspect of earth hydraulic influences.
structures which is of great importance and
which has been much neglected by geotechnical Experience has shown that engineering problems
engineers, namely the performance of residual in residual soils are not usually amenable to
soil slopes. the principles of soil mechanics alone, but that
these must be combined with the appropriate ele­
Residual soils are widespread, and they represent ments of 'geology', 'geomorpholopy' and 'hydro­
the most common type of material in those parts logy' in much greater measure than is necessary
of the world where slope failures are very pre­ for dealing with sedimented soils. Geotechnical
valent. This is particularly so in tropical engineering in residual soils therefore spans
countries, where deep weathering of the bedrock the narrowly separated fields of soil mechanics,
has invariably resulted in thick residual soil rock mechanics and engineering geology, and the
mantles, and where heavy rainfall is usually the engineering geological approach is generally the
main agent of slope failures. Apart from their most satisfactory one for these materials.
economic importance, rain-induced landslides
often present a significant hazard to life and From the point of view of analysis and design,
property. residual soils undoubtedly represent the ‘diffi­
cult' end of the broad spectrum of engineering
Residual soils have hitherto received very little soils, as opposed to saturated soft clays, which
attention from the soil mechanics com unity. represent the 'easy' end. Whereas a great deal
This neglect, and the resulting paucity of rele­ has been published about our ability to predict
vant literature, stems largely from the fact that the behaviour of soft clays, a negligible amount
the very nature of residual soils makes the has been written to enable an evaluation of soil
application of soil mechanics principles problem­ mechanics methods of prediction as applied to
atical and renders the materials difficult to residual soils.
model for the purposes of engineering analysis Mention must be made of the well-known writings
and design. The main characteristics of resi­ of Lambe (1973, 1975), which have given us a
dual soils responsible for this are : framework by which to judge our soil mechanics
(a) they cannot usually be considered in iso­ predictions, and which have done so much to focus
lation from the complete weathered rock attention on the need for more good measurements
profile, of which they are one component of performance. Lambe's predictive efforts have
part, for the most part, however, been concentrated on

2541
the 'easy' soils, and even for these, it has it is sometimes difficult to distinguish between
been found that the careful application of soil colluvium and the insitu material, particularly
mechanics methods does not always lead to accu­ if only drillhole samples are available for
rate predictions of field behaviour. A great examination. For geotechnical engineering pur­
deal of experience and engineering judgement is poses, colluvium can therefore be grouped with
therefore often necessary for making economical­ residual soils.
ly viable and acceptably safe predictions where
less predictable soils are involved. In this A recent publication contained a summary of the
context, the 'observational' philosophy con­ types of material regarded as ‘residual soil' in
tinually expounded by Peck (1969, 1975, 1980) eighteen countries, as revealed by a collection
is central to good geotechnical engineering. of review papers (Brand & Phillipson, 1985). No
universally accepted definition of 'residual
The material in this Paper is based on the soil' emerged, but the following general state­
Author‘s experiences with the performance of ments can be made :
residual soil slopes in Hong Kong, which is
prone to landslides that occur in the steep (a) residual soils can result from the weather­
granitic and volcanic residual soil slopes dur­ ing of any parent rock type,
ing times of heavy rainfall. These failures are
of major social and economic significance. Some (b) under the category of 'residual soil' are
severe rainfall events have in the past resulted included materials which are not completely
in many failures which have caused large numbers weathered and which retain the original
of casualties and widespread damage. In recent structure of the parent rock (saprolite),
years, therefore, a great deal of effort has and
been devoted to the prediction and prevention
of landslides. (c) colluvium is often categorised as 'residual
soil' for engineering purposes.
In part, this Paper draws heavily on two previous
publications by the Author (Brand, 1982, l985a) For the purpose of this Paper, therefore, 'resi­
which reviewed the various approaches to analysis dual soi1‘ will be defined as that broad group
and design in residual soils with particular of materials, formed by the insitu weathering of
emphasis on slope stability. any rock type, which exhibits engineering beha­
viour that is considered to be governed mainly
by the principles of soil mechanics. In terms
2. RESIDUAL SOILS
of the six-grade weathering classification system
given below (Table 1), these materials are grades
2.1 Definition IV, V & VI. Colluvium is included with these.
No universally accepted definition of 'residual 2.2 Published Literature
soils' exists. These materials are products of
insitu weathering of rocks, the degree of wea­ There is only a small amount of published litera­
thering and extent to which the original struc­ ture on the engineering properties and behaviour
ture of the rock mass is destroyed varying with of residual soils. As a starting point, reference
depth from the ground surface. This process should be made to the proceedings of the Specialty
gives rise to weathering profiles which contain Session on Lateritic Soils held in Mexico City
material 'grades' from fresh rock to completely (Moh, 1969) and to the book by Gidigasu (1975),
weathered material, the latter usually being which relies heavily on these proceedings. Also
described by geotechnical engineers as 'soil'. of importance are the proceedings of the con­
For engineering purposes, it is difficult to ferences on residual soils held in Hawaii (ASCE,
separate the 'soil' from the rest of the weather­ 1982) and Brazil (ABMS, l985a).
ing profile, and the whole profile is therefore
best treated as a single entity. The proceedings of the ten International Con­
ferences on Soil Mechanics and Foundation Engin­
For the purposes of tropically weathered pro­ eering each contain some papers directly relevant
files, the earth materials are sometimes cate­ to residual soils, as do the proceedings of the
gorised simply as 'laterite', 'saprolite' and many ISSMTE Regional Conferences held to-date;
the session on slope stability in residual soils
'rock'. The engineerin behaviours of laterite
and saprolite are usually considered to be at the Fourth Panamerican Conference is particu­
governed by the principles of soil mechanics, larly noteworthy (ASCE, 1971). To a lesser ex­
and they are therefore the materials of most tent, there are also specifically relevant papers
importance here. in the proceedings of the four Congresses of the
International Association of Engineering Geology
Large deposits of colluvium often exist in con­ and the five Congresses of the International
junction with residual materials, particularly Society for Rock Mechanics. There is a wealth
as colluvial fans on footslopes of hillsides. of residual soil literature in the proceedings
Colluvium is material derived from the weathering of the seven Brazilian Conferences on Soil Mecha­
of any parent rock which has been transported nics and Foundation Engineering, but unfortunate­
downhill by the agencies of gravity and water. ly this is virtually all in Portugese.
It can range in general composition from a col­
lection of matrixless boulders at one extreme to A few key publications address themselves to the
a fine slopewash material at the other. It pos­ general philosophy and approach to engineering
in residual materials. The first of these is the
sesses many of the same general characteristics
as residual soil, particularly in the context of state-of-the-art report on slope stability in
engineering behaviour. Because it is commonly residual soils by Deere & Patton (1971). The
found as slope cover over weathered rock profiles, others are the review papers by de Mello (1972)

2542
and Brand (1982, l985a), which all deal with the valuable comparative summary of the classifica­
subject of residual materials in fairly broad tion systems available at that time, but there
terms. In addition, the comprehensive report have been some important subsequent developments.
produced very recently by the ISSMTE Technical The Geological Society of London (1977) produced
Committee on Tropical Soils (ABMS, 1985b) is of a Working Party Report on the subject of the des­
importance. cription of weathered rock masses for engineering
purposes, the main elements of which have now
Especially worthwhile contributions to the liter­ been incorporated into the British Code of Prac­
ature on the engineering properties of residual tice for Site Investigation (British Standards
soils have been made by Vargas (1953), Lumb Institution, 1981). Other important papers on
(l962a, 1965), Sowers (1963), Little (1967, this general topic are those by Little (1967,
1969), De Graft-Johnson & Bhatia (1969) and Dear­ 1969), Dearman (1974), IAEG (1979, 1981), Hencher
man et al (1976). Only Vaughan & Kwan (1984) and & Martin (1982) and Martin & Hencher (1984).
Vaughan (l985a), however, appear to have attemp­
ted to provide a theoretical framework for resi­ Any weathering description and classification
dual soil behaviour. system must be suitable for the particular geo­
logical conditions and engineering purpose to
Noteworthy publications specifically on the slope which it is applied. In Hong Kong, where site
stability aspects of residual soils have been formation and slope stability are the main geo­
written by Vargas (1967), Patton & Hendron (1974) technical engineering problems (see Section 3
Morgenstern & de Matos (1975) and Blight (1977). below), the Geotechnical Control Office (1984)
Good reviews of the landslide problems in these has adopted a system for the granites and volca­
materials have been made by Vargas & Pichler nic rocks which is based on the original work by
(1957), Da Costa Nunes (1969) and Jones (1973) Moye (1955) and Ruxton & Berry (1957), but which
for Brazil, Page & James (1981) for Colombia, has several important additions suggested by
Sowers (1971) for Puerto Rico, Lu b (1975) and Hencher & Martin (1982). A profile is logged
Brand (l985b) for Hong Kong, Oyagi (1984) for according to the six material grades given in
Japan, and Brand (1984) for Southeast Asia. Table I and the four profile zones described in
Table II. It should be noted that weathering
Colluvium features hardly at all in the techni­ zones A and B comprise those materials referred
cal literature, even though it is a fairly common to in some countries as 'laterite' and 'saprolite
engineering soil type. Good geomorphological respectively.
textbooks, such as that by Young (1972), provide
a general description of the formation of collu­ For the purposes of geotechnical analysis and
vium, but few attempts appear to have been made design, the following should be noted :
to devise an engineering classification system,
nor to investigate its wide range of material (a) grades I to III material are usually treated
properties. Although colluviu features fairly as 'rock', and grades IV to VI material as
prominently in some published descriptions of 'soil', and
mass movements,
which there are only
describe engineering a fewin
designs papers
t is mat­ (b) the engineering behaviour of weathering
erial, the most useful of which is probably that zones A and B is broadly considered to be
by D'Appolonia et al (1966). governed by the principles of soil mechanics.
There exists no engineering description and clas­
2.3 Weathering Profiles sification system for colluvium, although one is
badly needed. An attempt has been made in Hong
The accurate logging of weathering profiles is Kong to provide a framework for such a system
fundamental to successful design and construction (Huntley & Randall, 1981), but this is entirely
in residual profiles. These often contain a descriptive in character and requires a great
whole range of materials from an engineering deal of further development.
point of view from 'soil' to 'rock‘. The wea­
thering profile is therefore of great importance
for the stability of slopes, because it usually 2.4 Investigation
controls
For projects in residual soils, site investigation
(a) the potential failure surface, and there­ must generally be more extensive and more expan­
fore the 'mode' of failure for analysis sive than for more homogeneous earth materials.
and design, and The emphasis must be on the
the fo
engineering geologi­
cal approach, for which lowing are the
(b) the groundwater hydrology, and therefore main elements :
the critical pore pressure distribution in
the slope. (a) execution of adequate surface and subsurface
exploration to establish the site engineer­
There is no universally accepted system for des­ ing geology. C0 define the 'engineering'
cribing and classifying the component parts of materials involved, and to retrieve good
a weathering profile. The classification sys­ quality samples for laboratory testing,
tems commonly used in soil mechanics have very
limited application to weathered rock profiles. (b) identification of especially significant
Classification in terms of weathering 'zones' geological, geotechnical and hydrological
and weathering 'grades' is essential for en­ features,
gineering design, and there have been several
major attempts to provide a satisfactory des­ (c) study of existing local and other 'case
cription and classification system for engineer­ histories' of similar projects and of sites
ing purposes. Deere E Patton (1971) gave a with similar geology,

2543
(d) routine insitu and laboratory measurement TABLE I
of selected engineering properties of mat­ Weathering Grade Classification System
erials to establish lower bound values, and Recommended by the Geotechnical Control Office
comparison of these with generalised para­ (1984) for Use in Hong Kong
meters developed from existing data, and
(e) continuous reappraisal of the site investi­
gation results throughout the period of
construction.
Grade Typical Characteristics
The importance of the study of good case histo­ VI Residual
soil
Soil formed by weathering Ln place but with
ries cannot be overemphasized. Particularly original texture of rock completely destroyed
important also is the continuous reappraisal Rock wholly decomposed but rock texture preserved
process, since site investigations carried out Completely No rebound from N Schmidt hammer
in residual soils before construction com ences V decomposed
rock Slakes readily ln water
can in most cases only be regarded as 'initial' Geological pick easily indents surface when pushed
investigations, and the geotechnical engineer
must be prepared to modify his design as exca­ Rock weakened - large pieces can be broken by hand
vations reveal much fuller subsurface information Positive N Schmidt rebound value up to 25
than was available prior to the construction IV
Highly
decomposed Does not slake readily in water
phase. This is the basis of the 'observational rock Geological pick cannot be pushed into surface
method' of design expounded by Peck (1969). Hand penetrometer strength index >Z50 kPa
Individual grains may be plucked from surface
Nhereas the use of generalised soil properties Completely discoloured
is not unsatisfactory where the soils are such
that these can be established within a suffi­ Ill Moderately Considerably weathered but possessing strength
decomposed such that pieces S5 mm diameter cannot be
ciently sensible range, this is not so for resi­ rock broken by hand
dual soils, because of their variability and in­ N Schmidt rebou d value 25 to B5
homogeneity. Generalised parameters for resi­ Rock material not friable
dual soils are difficult to obtain, and insuffi­
cient basic research has been carried out on most Slightly
Dlscoloured along discontinuities
residual materials to provide even approximate ll decomposed Strength approaches that of fresh rock
rock N Schmidt rebound value greater than LS
values. This is an unsatisfactory situation More than one blow of hsnmer to break specimen
which confronts everyone who works with residual
soils, and it enhances the importance of the I Fresh rock No visible signs of weathering; not discoloured
non-analytical methods of design described below.
Drilling and sampling in residual soil profiles
can be difficult because of the vertical varia­
bility encountered, particularly since it is
essential to obtain a high recovery for profile
description purposes. A recent international TABLE II
review of the methods used worldwide for the Weathering Zone Classification System
sampling and testing of residual soils (Brand & Recommended by the Geotechnical Control Office
Phillipson, 1985) revealed that the sampling (l984) for Use in Hong Kong
practice in most countries leaves much to be
desired, heavy reliance often being placed on
methods developed for sedimentary soils. In
this context, the drilling and sampling methods Zone Description
used for good quality site investigations in the
weathered granite and volcanic profiles of Hong
Kong may be of interest elsewhere (Brand & Phil­
\
Structurelesa sand, silt and clay.
Hay have boulder concentration at
lipson, 1984); triple-tube rotary core barrels A
the surface.
are used for sampling, and air-foam is sometimes
employed as the flushing medium.
Residual material with corestones
B Rock percentage is less than 501,

3. INTRODUCTION TO HONG KONG

lnterlocked.
and corestones are rounded and not \\\\€ii>
/
Since there will be continual reference in this <><>’®§
Paper to the slope stability problems in the
residual soils of Hong Kong, an introduction to
the Territory is appropriate at this stage. Corestones with residual materials. b@<>©®
C Rock percentage is 50 to 901, and
corestones are rectangular and
Lnterlocked. ®§©
3.1 General Description
>\\
The Territory of Hong Kong has a land area of
only l 050sq. km and a population of nearly six
million. It consists of Hong Kong Island, a
/"\ \`
piece of the Chinese mainland and more than 200 D
More than 901 rock. Minor residual // >({
material slang maaor
discontinuities structural
w Lch may be con- /
small islands, the largest of which is Lantau
slderably iron stained. // //// \><; //
(Fig. l). The population is concentrated in a
number of distinct geographical locations ///(YH
2544
 ~. _,9-\",_30
1*

ri-
CHINA ,_,.._.

. ‘ .
.§, _
~ a . 1~ 5
­
if?_ rf.;:}_
` f.,j Q ' wmv.
§
=¢~:=-‘

L‘ Y ‘ D 5'/'Az
~'-`-PH'-'-' ' 5. If-2' .f
LANTAU3ISLAND
W .U ~lit:
$2 "" _ ' '
L ;2:I;_'KOWLOON ­

_~1~2:2:2~ Q 2 . G =
5' oA 2 JI
km‘Q° ‘Howe
ui__aaJ
10
¢u V( u dé~

¢3 Kowe
ISLAND
Alluvium - Fleclamatlon
E Sedimentary Rock (Volcanic Rock
Rhyoliiic )
Inuuswe Rock
Metamorphic Rock ( Mainly Granite )

Fig. l Map of the Territory of Hong

Kong Showing the Geology

dictated to a great extent by the terrain. High

concentrations of building development and popu­
lation exist all along the north side of Hong Fig. 2 View Looking East along the North Coast of Hong
Kong Island (Fig. 2) and over the entire Kowloon Kong Island Showing the Central District and the
peninsula, as well as in the five new towns which Mid-levels Area (Note Po Shan Landslide Scar)
are currently under construction north of Kowloon.
Apart from Hong Kong Island, the islands are 4o
largely undeveloped and sparsely populated.
a o -40°
The terrain of Hong Kong is very hilly. The
land rises from sea level to 550
Island in a distance of onl 1.5 km Most of
the Kowloon peninsula has ngw been levelled,
m on Hong Kong lu :1:5:3:3:5
< ao o - 5° 2222522222
5 252225252 15-30° 2222222222
along with reclamation of adjacent sea areas, Q zo 222222222 2222222222 2222222222
2 3:3:§:§:§: :1:§:§:§:§ §:§:§:§:§:
but isolated hills of up to 100 m still exist. 5 2222222222 6-15° 2§5E5§3E5' 222222222
Immediately behind Kowloon, hills rise to over
450 m. In the New Territories, very little low­ a 10 2222222222 2222222222 222222222 222222222
lying land exists, and peaks of over 400 m are 5§§§ iii? E523 525 40-eo'
common. Natural slopes throughout the Territory EES
0 :;:;:;:;: 'EQE
;.;.;.;.; fin iii
-:-:-:-:- F53;:;.;.;.;
I;:;:;.;. >&6
are steep, typically with upper slopes steeper
than 35 degrees, midslopes of 25 to 30 degrees TERRAW ANGLE (degwea)
and footslopes of l5 degrees. Cut slopes formed
for roads and site development works are common­ Fig. 3 Distribution of Terrain Angles over
ly 40 to 80 degrees, and fill slopes are 30 to
35 degrees. Figure 3 gives the distribution of Hong Kong Island and Kowloon
terrain angles for Hong Kong Island and Kowloon.
About two-thirds of the Territory's small land annually, and more than BOZ of this falls during
area is considered undevelopable. The usable the period May to September (Fig. 4). Intensi­
land is therefore an exceptionally valuable com­ ties can be high, with 50 mm per hour and 200 mm
modity. In the absence of flat land, highrise in Zh hours being not uncommon.
buildings and other structures are increasingly
being built on the mid-slopes and upper slopes
of natural hillsides. In these circumstances, 3.2 Geology
very intensive use is made of all developable
land. The geology of Hong Kong, which is summarised in
Fig. l, has been described by Ruxton (1960) and
The rainfall in Hong Kong averages 2 225 mm Allen & Stephens (1971). The main rock types

2545
 400 in composition from a boulder field to a fine
slope-wash, but it most commonly consists of
boulders, cobbles and gravel in a matrix of sand,
E soo silt and clay. The colluvium is up to 30 m thick
5 in places. It is sometimes in a loose state,

E 10° §§$
j< 200
\L

< ;&&&&% £3 &5:&&&H£§£§

I NiH5Hw##M#§&¢§&&$&E&&§&&&&%m»m
with a high permeability, and it frequently gives
rise to perched water-table conditions. It is
also prone to the formation of 'pipes' or 'tun­
nels' as a result of internal erosion (Pierson,
1983; Nash & Dale, 1983; Brand et al, 1986), and
these features can be of major significance to
JFMAMJJAS0ND
0 .'~'.'.'.'.‘-'.'.'.'.‘.'.'.'.'.f~f~f~'-‘-f-f-f-f-f~:~f~_._-_._._-_._.>.>._._ ._._- _._._._._._._._.>._._._-&,_._._._._._._._._.Q ____ _ _
the hydrogeology of an area (Leach & Herbert,
MONTH
1982; Premchitt et al, 1985).
Some engineering properties of the fresh and de­
Fig. 4 Average Monthly Rainfalls in Hong Kong composed granite and volcanic rocks of Hong Kong
have been reported by Lumb (l962a, l962b, 1965,
1975, 1983).
are granite and acid volcanic rocks, which_to­
gether cover the major portion of the Territory
and are by far the most important from an en­ 3.3 Slope Stability
gineering point of view. The small amounts of
sedimentary and metamorphic rocks are of much Geotechnical engineering practice in Hong Kong
less importance, although some specific land­ tends to be dominated by the slope stability pro­
slide problems have been associated with these. blems brought about by the combination of the
Granite predominates in those areas of the Terri­ high intensity rainfall and the steep terrain.
tory where building development is densest. Several hundred failures may occur in a year,
most of which are not of great consequence, but
The granite varies somewhat in colour and compo­ severe effects are felt from some of these in
sition, but the fresh rock is fairly uniform in terms of casualties and damage. In fact, a sig­
its engineering properties. Of major signifi­ nificant landslide event, in which a large num­
cance for slope stability are the joint patterns ber of failures occur in one day causing consi­
and the extent of weathering. Joint spacings derable disruption and damage, can statistically
are typically 0.5 to 1 m, but they can be as be expected to take place in Hong Kong about
wide as 3 m. In a few locations, sheeting joints once every two years (Lumb, 1975; Brand et al,
occur almost parallel to the natural ground sur­ 1984).
face. The granite is extensively weathered al­
most everywhere, with depths of up to 60 m of The whole range of slope 'features' is prone to
silty-sandy residual soil, often with large landslides in Hong Kong, including natural slopes
corestones in the matrix or exposed on the sur­ soil cut slopes, rock cut slopes, earth fill
face. Below the residual soil mantle, a consi­ slopes, retaining walls and boulders. The majo­
derable depth of differentially decomposed mat­ rity of failures, and usually those with the most
erial extends to great depth in a weathered gra­ severe consequences, take place in man-made fea­
nite profile. The weathering process in Hong tures or are triggered by man-made features,
Kong granite has been described by Ruxton & particularly cut slopes in soil (weathering
Berry (1957). grades IV to VI).
The volcanic rocks consist mainly of coarse Cut slope failures in soil, or in mixed soil and
tuffs, fine tuffs and rhyolite,but some ignim­ rock,now constitute by far the most common form
brites and breccias also exist. These rocks are of landslide. The volcanic rock profiles are
usually fine grained and have a blocky structure, more susceptible to failure than the granite pro­
with close joint spacings of generally less than files, and colluvium is frequently involved. The
about 0.25 m. They are more resistant to wea­ failures nearly always occur suddenly during in­
thering than the granite, the residual soil man­ tense rain without prior warning (Brand, 1984,
tles being only up to 20 m thick. A volcanic l985b), and most slip surfaces are shallow, the
profile usually exhibits a steadily decreasing thickness of the failed zone usually being less
degree of decomposition with depth, the unwea­ than 3 m. A typical example is shown in Fig. 5.
thered rock being reached at a much shallower
depth than in a granite profile.
Steeply dipping dolerite dykes have been intruded 4. M THODS OF SLOPE FAILURE PREDICTION
in many places into the granite and volcanic
rocks. These dykes, which vary in thickness be­ The stability analysis of a slope is ipso facto
tween about 150 mm and 1.5 m, greatly complicate related directly to the prediction of the condi­
the engineering geology of many sites and are tions under which the slope could fail. Soil
often a significant feature in landslides (Hen­ mechanics methods of analysis have long been
cher et al, 1984; Hudson & Hencher, 1984). Grano used for this purpose, but it must be realised
diorite intrusions also exist in some places. that the majority of the world's man-made slopes
were in fact formed on the basis of experience
Of particular engineering geological significance and precedent. The vast majority of residual
in Hong Kong are the extensive bodies of collu­
vium which cover about 20Z of the total land
soil slopes certainly fall into the latter cate­
gory.
area. This accumulation of debris from old land­
slips and mass movements very often carpets the Apart from the application of sound judgement and
lower slopes of most of the hills. It varies experience alone, there are four basic methods

2546
 others with similar characteristics.
Whereas the vast majority of cut slopes in resi­
dual soils were not designed on the basis of
rigorous soil mechanics methods, the soil mecha­
nics approach is being increasingly adopted even
in this difficult soil type. This is certainly
true in Hong Kong, where design practice is
governed largely by the Geotechnical Manual for
Slopes (Geotechnical Control Office, 1980). The
degree of safety of a slope can only be quanti­
fied on the basis of an analytical method, where­
as methods (a). (b) and (c) provide no such quan­
tification.

5. RELATIONSHIP BETWEEN RAINFALL AND LANDSLIDES

IN HONG KONG
5.1 Background
The vast majority of failures in residual soil
slopes are caused by rainfall. In several coun­
tries of the world, studies have been carried
out to correlate slope failures with the pattern
of rainfall. At their best, such correlations
provide a broad basis for predicting widespread
slope failures,which can lead to the establish­
°-<.
`~'K
.' -~-~
ny”
\t~'
ment of a warning system for those whose lives
might be endangered. Rainfall-failure correla­
tions might therefore be regarded as 'bulk' pre­
dictions.

»~m,m“H
` "“-z-vs_. _ _. ` ­ _ ' »
-.,;;|.¥~ “_ _ sam - s<-»=»_~_,M _,
Simple direct rainfall-failure correlations have
been made for Brazil (Barata, 1969; Guidicini &
Iwasa, 1977), Italy (Rossetti 8 Ottone, 1979),
Japan (Onodera et al, 1974; Fukuoka, 1980) and
the United States (Campbell, 1975; Nilsen et al,
> - ' _ ~ i »\» l“':“~s­
¢-~~ 1, ` |` W , ,_ . gg" ... -` . \\;\`;;>:;_,>(% N"
;5<.t
.~ ._ *.\;- §7?T`WErw~*~»\‘H¥»'
-- ‘ _ ~ f W* gt _'a
s,
1976), whereas more sophisticated correlation
attempts have been undertaken in New Zealand
(Crozier, 1969; Eyles et al, 1978; Eyles, 1979;
Crozier & Eyles, 1980) and Hong Kong (Lumb, 1975,
1979; Brand et al, 1984).
Fig. 5 Typical Hong Kong Soil Cut Slope
Failure Caused by Heavy Rainfall It is very rare for detailed information to be
available anywhere in the world on the geographi­
cal distribution and short-term intensities of
available for the prediction of rain-induced rainfall, measurements of rainfall commonly being
failures in residual soil slopes. These are : made on a daily basis at stations which are far
apart. It is also rare for the precise times to
be known for the occurrence of landslides. All
(a) correlations between slope failures and
pattern of rainfall, previously published analyses of landslide data
have therefore been based upon regional correla­
(b) terrain evaluation, mainly on the basis of tions of landslides with one-day or longer dura­
geomorphological mapping, tion rainfalls, the more sophisticated correla­
tions taking account of the antecedent rainfalls
(c) semi-empirical (or modified precedent) for periods of up to several weeks.
approach, which is based on an examination
of the geomorphology and geology of stable Relationships between rainfall and landslides
and unstable slopes, and were first established for Hong Kong by Lumb
(1975, 1979), who classified landslide events
(d) soil mechanics analytical methods, usually into four categories, defined as follows :
in the form of limit equilibrium analysis.
(a) 'Disastrous event' - Territory-wide
The first of these two approaches can be consi­ damage, with more than 50 individual fail­
dered to be directly related, since they apply ures recorded in one day,
to the stability of a land area in one particu­
lar location. The last three methods can be re­ (b) 'Severe event' - widespread damage,
garded as methods of analysis and design; all with between 10 and 50 failures in one day,
three have been used extensively in residual
soils. The terrain evaluation and semi-empirical (c) 'Minor event' - localised damage, with
methods are closely related, in that both are less than 10 failures in one day, and
based on an explicit assumption that the stabili­
ty characteristics of a slope can be assessed on (d) 'Isolated
failure.
event' - a single individual
the basis of observations of the performance of

2547
1'3 ‘.
"E Mmon
Z
;m
400SEVERE
"E300
200|-
8
EJ DISASTER Q ,_._,~-" "" 5:4
g _ o°~g 0
jO
S O o E' ,,
O N -/""
o
°O
Oo v
+
*°
~
I"
° ° MWOR +‘ a.
0
I
*
*°§
Fig. 6Landslides
Relationship
‘
_
0
9
*g
||||||_='
Q 100 5 °°¢1 e H ° 1- 0 + °+ + + Q °
2% ° 0 8% ISOLATED <7 °§°°"°'°'f * 'Z ` 9°
0 100
15 -200
DAY300 400 500
RAINFALL ( mm 600
)A£%
between
Established
il ` G 1
Rainfall
by Lumb (l975) “'*'*“'*‘
On this basis, Lumb correlated the landslide
events for the period 1950 to 1973 with the one­
day rainfalls and the 15-day antecedent rainfalls
to obtain the interesting results shown in Fig.
5
and Jr 0 10 km ‘ao
I

+ Automatle gauges n Autographlc gauges

Fig. 7 Locations of Continually Recording

Raingauges in Hong Kong, July 1985
6.
gauges have been installed elsewhere in the
In incorporating antecedent rainfall into his Territory for many years.
correlations, Lumb (1975) in effect took account
of the well-known phenomenon of soil moisture Since 1978, a sophisticated system of automatical
deficit and its relationship to shear strength. ly recording raingauges has come into use to pro­
His one-day rainfalls, however, were taken as vide data for the Geotechnical Control 0ffice's
being the calendar-day rainfalls measured at the rainfall-landslide correlation programme. At
Royal Observatory (see below), and it is now present, 46 automatic gauges transmit rainfall
known that better correlations are possible if measurements continuously through telephone lines
account is taken of the localised variations in to a central micro-computer. The locations of
rainfall. A further correlation study was there­ these, and other continuously reading gauges,
fore undertaken recently in Hong Kong (Brand et are shown in Fig. 7.
al, 1984) which had the benefit of extremely
detailed data not previously available. The Large geographical variations in rainfall occur
results of this have greatly aided our under­ over the Territory during any rainstorm event,
standing of rain-induced slope failures, and as dramatically illustrated by rainfall records
may have application elsewhere. shown in Fig. 8 for three automatic gauges during
a very recent rainstorm. These variations are
The rainfall-landslide correlation programme, attributable to the effects of topography. A
carried out by the Geotechnical Control Office, common feature of a Hong Kong rainstorm is that
was based on the six main parameters : the highest rainfalls occur near the tops of
(a) one-hour rainfall, 250
(b) 24-hour rainfall,
(c) antecedent rainfall for periods of up to
30 days,
(d) number of landslides which caused casual­ 150­
ties, Kwa|Chmu (4km)
(e) number of casualties, and
100­
(f) times of landslide occurrence.
5.2 Rainstorm and Landslide Data
Excellent rainfall data is available for Hong
Kong. Rainfall records have been kept by the
Royal Observatory (RO) since 1884. Rainfall
statistics are referenced to the 'principal'
gauge situated at the Observatory on the Kowloon
peninsula (Fig. 7), at which hourly measurements
°I||l
50­
Royal Observatory
_._--------------1
2:00 4:00 6:00 8:00 10:00 12:00
TIME : 9 JUNE , 1985

Fig. B Example of Large Variations in Rainfall

are made. ln addition, several autographic over Short Distances in Hong Kong

2548
 formation on precise times of failure, however,
is invariably difficult to obtain. In Hong Kong,
N 1' 3”_` '12
f .1 a
,,j 51°' ''
there is often a delay of several hours between
a failure occurring and an engineer being called
out to inspect the failure; landslides in rural
»- 55
|

_- H., "
areas are sometimes not inspected for several
days. Information on times of failure is there­
fore often not reliable from this source.
)­
1.

ln collecting data for its landslide correlation

' ¥}:4}§EL'
f1'
<7*J* Gm3\ '~ study, the Geotechnical Control Office found
the Fire Services Department to be an extremely
reliable source of information on the occurrence
._ §|$.\
1 Fs .J __n tin`*»
A ¢"\
Aft
by-rkk Vg
ll;
of landslides in Hong Kong over the past twenty
years. Although the FSD is involved with only
a small proportion of the total nu ber of land­

J“ kU o
slides that occur, these comprise all the more
serious incidents, and their excellent records
ia 50 10
155mH
C]
can be considered to be representative of the
landslide events in Hong Kong as a whole. Their
data which relates to times of call-out is parti­
cularly good, and the fact that they are usually
- Landshde -f-Isohyet called to a serious incident immediately after
its occurrence enables the time of failure to be
fixed fairly accurately. The correlation studies
were therefore undertaken largely bn the basis of
Fig. 9 Locations of Landslides in Re1ation to the Fire Services data.
3-hour Isohyets in the May 1982 Rainstorm
The Geotechnical Control Office's rainfall-land­
slide correlation study was based on data avai­
large hills, and that the rainfall isohyets tend lable for the 20-year period 1963 to 1983, for
to follow the ground elevation contour lines. which the l3 severest rainstorm events are listed
These features are illustrated in Fig. 9, which in Table III, together with their hu an consequen
shows the contours of two separate three-hour ces. Seven of the rainstorms were associated
rainfalls during a severe storm in May 1982. with troughs of low pressure and six were caused
by tropical cyclones. For the majority of these
If meaningful rainfall-landslide correlations rainstorms, there are no adequate records of the
are to be made for any geographical location, actual numbers of landslides, but the small num­
accurate information is needed on the number of bers of Fire Services reports are indicative of
landslides and their times of occurrence. ln­ the numbers of particularly serious incidents.
TABLE III
The Thirteen Major Rainstorm Events in Hong Kong and Their Consequences during the Period 1963-1983

3.2 E
D t Maximum Rainfall, mm Landslide Consequences
Of Ti; Observatory Ogher Location No. Failures No. Fire No. People No. People
Rainstorm Storm24-hour
Reported in Services Killed or Permanently
1-hour 24-hour 1-hour Newspapers* Reports Injured Evecuated
24-25 August 1976 STS Ellen 416 52 500 82 314 23 57 2400
12 June 1966 trough 401 108 525 157 100 30 35 8500
29 May 1982 trough 394 44 430 111 498 15 48 8000
16-17 October 1978 STS Nina 380 37 380 38 15 1 1 no record
16 August 1982 STS Dot 362 68 370 95 62 6 9 1500
17 June 1983 trough 347 69 460 101 114 5 2 600
27 September 1965 TS Agnes 333 47 333 47 9 4 4 200
17 August 1971 Typhoon Rose 328 63 328 63 10 5 7 no record
12-13 October 1964 Typhoon Dot 304 60 375 94 8 10 39 8000
12-13 June 1968 trough 287 100 343 143 10 7 27 200
16-17 June 1972 trough 280 36 560 71 >l5 15 21 7800
17-18 June 1972 trough 275 99 300 98 "dozens" 14 229
17 May 1972 trough 271 79 377 92 "dozens" 2 0 6000
TS - Tropical Storm * The numbers of landslides reported in newspapers are for comparison only - they
STS - Severe Tropical Storm represent the lower limits of the numbers of landslides that actually occurred.

2549
However, detailed landslide information was ob­ 800
tained by the Geotechnical Control Office for l5°'JHourly
H Cumuhhve TJ '
EI|;fZ
the May and August 1982 rainstorms, which caused
over l 500 slope failures (observed from aerial
photographs), and the June 1983 rainstorm, which "E- (| Max
I __) rJ:2*Max
600
l-L 1
caused more than 150 failures. While the num­
bers of landslides reported in newspapers are -'J 100-
(|lO.)
Hourly V
( __ 1 FLO. 400
1 ._
well below the actual figures, they give some

I ,
relative measure of the number of failures that
occurred during each storm.
lt should be noted from Table III that the num­
ber of people killed and injured was not neces­
E 50-n I
> :: »: -200
Z

sarily proportional to the severity of a parti­ I5

Q \ -lv-E-:»:`:`
1 L .--:-’ _|": ' _
cular rainfall event, since a few major indivi­
dual landslides were responsible for large num­
bers of casualties. In particular, the June
1972 storm resulted in 224 casualties from the
two failures that occurred at Po Shan Road on
0 . __-1__23Q]
1- °
Landslides -> III II ll I I
Hong Kong Island and Sau Mau Ping in Kowloon
(Government of Hong Kong, l972a, l972b; Vail. 0 6 12 18 24
1984). Also, the vast majority of the people NME: 12JUNE.1966
who were permanently evacuated from their
homes in the 20-year period as a result of rain­ Fig. ll Occurrence of Landslides in Relation to
storms were people who occupied temporary squat­ Rainfall at the Observatory and at Any
ter dwellings on unformed hillside sites. Location during the June l966 Rainstorm
B00
5.3 Effect of Rainfall Intensity

-1
E
E 100- II
_ H //_-‘ __
_ fl
The available rainfall and landslide data was

u< _Q r1 -200
examined in detail for all thirteen rainstorm
-400 v

5 ri _
events listed in Table III. Figure 10 shows the
patterns of rainfall for five of the most serious
rainstorms (in terms of consequence). It will
be seen that these storms generally lasted for
a few days, but the majority of the rain fell
Q 1 Cumumuve
Z 50-
`~ Houdy
in a few hours in each case. All these storms I _ ' .Q ‘_V_V ///F­
were therefore associatedgwith short duration,
high intensity rainfalls. Where sufficient
data was available, a definite pattern emerged
from a close examination of this fact for each
of the thirteen cases.
I -I 11 1- °
5 __ ~ V
O 0 I' _,-.=.,:j2'; _______ '_I_1; ',_,,
L=~d=»1=@= -1 IIII |11 1
The storms of June 1966, October 1978, May 1982
and August 1982 will be used as examples. The
0 6 12 1B
last two are of particular importance, because TIME: 29 MAY, 1962
the Geotechnical Control Office obtained detailed
information on these two severe storms, which Fig. l2 Occurrence of Landslides in Relation to
Rainfall at the Observatory and at Any
Location during the May 1982 Rainstorm
' June 1972
,_ __,__..._,._...._ 150

600 - / May 1982

August 1976 / r/_
" I August19B2 EV' H°"'|V
“_ 100 r.:
- E
Cumulative Boo
- :LI
1/ ,__
400 .1 _:_ /
/ f/
»1
zoo­
//
Z
400° ~ / ' June 1966 Z 11 1
5 5° - £3 '“ fi' ia
1 __ F; 9¢§§ HE 5§ 200
1 0 . ,___, nl ..-- 1 ___.,._‘ amos; _____ ¢m¢ _ ,_ 0
Landandes--»|H|H

123456
°11111111111
DAY OF RANSTOFIM
0 6 12 1B 24
TIME: 16 AUGUST, 1962
Fig. 13 Occurrence of Landslides in Relation to
Fig. l0 Hong Kong's Five Severest Rainstorms in l963-B3 Rainfall at the Observatory and at Any
Location during the August l9B2 Rainsto PITI

2550
(Table III). A E
resulted
ranked

fell on
third

Hong
in fifth
and

the
the ofloss

Kong
the

heaviest
Island,
of2027
last

rainfall
and the
lives,
years' 150

inhighest
and which .500
rainstorm events in terms of 24-hour rainfall »- A
In June 1966, E cumumuve (n.o.) -90° 5
memoryrain-
' _ ‘°“°m°l)
j
fall to-date was recorded at the Observatory. I 4oo Ioo ‘_ J
4
Widespread
throughout landslides
the and
Territory. flooding
For this occurred
rainstorm, Q E
- §
m
Fig. ll shows the hourly cumulative rainfall at > 5° w
the Observatory and atinthe location ofcompared
measured 4with Hmmy "2°° §j
maximum
the occurrence
rainfall each
of landslides
case,
as pattern
recorded
S if __
Fire Services Department. The ofby the 20Ji' ""`
cumu- `V,_"li<` _,_, _-."~0D5
`""'"“'
lative similar
quite rainfall measured at the Observatory
to that measured at the location is Lmwakh
_° 1 U
of maximum rainfall, although it is inevitably 12 18 o o 12 1a
somewhat lower. In contrast, the hourly maximum
rainfalls are much higher than the corresponding 'UMEI17/15 OCTOBER .1978
ones measured at the Observatory. Landslide
timings for the rainstorm show very good corre- _ _ _ _
lations with hourly maximum rainfalls, which F1g.l4 Octurrmme 0fljhdH1dESlh Rdat1onto
suggests that high intensity rainfall of short RHHWQH at@h€0b5UWitWW atA"Y_
sduration
1 es. is a major triggering factor for land- Locat1on dur1ng the October 1973 Ra1nstorm
For the May 1962 rainstorm, Fig. 12 shows the 30 1-20
rainfall and Fire Services landslide information. nnmr
It can be seen that virtually all the recorded "'91d°Y'
landslides
rainfall occurred
intensity. Also, at the time
although the of the maximum
maximum Q 0, l
hourly rainfalls are dramatically higher than ° 2Q :QQ
those measured at the Observatory, the pattern E QQQ 2o_4o Anas B
case. B §§§ _am Dayawnh
of cumulative rainfall is very similar in each é §§§ "nm, Y
Ifflfif §§;§§¢§§ )3 landslides
For the August 1982 rainstorm, the data plotted 5 1° §§§ §§§ 4o_ao
in Fig. 13 shows essentially the same pattern 3 ii jij num,
depicted in Figs.
In Fig, 14 is shown llforu the
the data 12.October
z eo_s°
1978 Iii 80-100
§§§ _y _m 100-120
"mm, hmm,
rainstorm. Although the 24-hour rainfall of 380 - QEQ ‘Nw Ill F'1 F'1
mm was the fourth highest in the 20-year period °
(and is similar to that for the May 1982 rain- MAXMUM|{1;1ygAgf;iL (mm)
storm, Fig. 12), only one landslide and one
casualty were reported for this event, for which _ _ _
the maximum hourly rainfall was only 38 mm_ F1g. l5 Correlatwn of Landsl1de_Occurrence
This lends confirmation to the dependence of dU[`l"'9 1932 Wltll the Maxlmum H0U"ly
landslide occurrence on the short duration rain- Ralnfam I"t9"9t¥ AnYWh@VE
fall intensity.
During 1982, 3 248 mm of rain fell in Hong Kong, 40 _____
the highest annual rainfall on record. The dis- £25
tributions of all rainy days and of days with Eff?
three or more landslides are shown in Fig. 15, w ao QEE
plotted against the maximum hourly rainfall any- W QQQ
where on each
fall was lessday.
thanWhen the maximum
20 mm, no slopehourly rain- m
failures g QQQ
Qu;
usually
always occurred.
occurred onIn contrast, failures
the few days on nearly
which 3the
2°<ff@f
Qjffjf
maximum hourly rainfall intensity exceeded 40 mm J QQQQQQ
per hour. The failures that occurred were also i 10 Eiiiiimuuw
found to be in the vicinity of the area that Eyiimiiiii
experienced the highest rainfall intensity'
For the 20-year
analysis of all theperiod
times from
of 1963 no 1983,
occurrence of an I0I 4I 5I I12
land- I I18 )1a
slides was carried out. The results presented
in Fis~ 16 Sh°w UMC the maj°riry Of the reported Houns AFTER MAx|MuM Houm_v INTENSITY
landslides occurred within four hours of peak
rainfall
than 16 hours intensity,
to occur, F1 with onlyof102
I6 T1mes takingofmore
Occurrence Ma_]or_Landshdes
___
as Related to the Maximum Hourly Rainfall
The dependence of landslide occurrence and Intensity, during the Period 1963-83

2551
casualties on hourly rainfall intensity is clear­ groundwater build up (Hudson & Hencher, 1984).

“3.o<2"§P‘2.BZ
ly shown in Fig. 17. Throughout the twenty-year
period, the number of landslides that caused The often large short-term differences in rain­

EDE<a
casualties was very small unless the maximum fall from place to place throughout Hong Kong
hourly rainfall anywhere in the Territory ap­ tend to even out over the longer rainfall
proached 70 mm. Above this figure, the number periods. This is demonstrated in Figs. ll to 14,
of landslides with casualties increased sharply where the 24-hour rainfalls at the Observatory
with increasing maximum hourly intensity. The are generally not greatly different from the
actual number of casualties caused by landslides maximum values measured anywhere. It is there­
shows a similar but even more dramatic correla­ fore useful to examine the relationship between
tion with hourly intensity. landslides in Hong Kong and the Observatory 24­
hour rainfall.
Correlations between landslides in Hong Kong and
the Observatory 24-hour rainfall are shown in

3 so 10 Q
E 40 Talal no. casualties 14 Q Fig.
A 19 onevent
minor' the basis offor
is one twowhich
categories
there of
areevent.
than ten recorded landslides in one day, and a
'major' event is one for which there are more
than ten landslides in one day. The 'minor'
less

gE§
2 20
_|
§¥
ifHE
1°
< E?
Qi
;-=- 4
__ 3
No.Caauallles
landslides
if H2
causng
2 §E< ;_;.;5;:;:;
_ 5<0
Q
category is identical to that used by Lumb (1975)
while the 'major'
combined 'severe'category corresponds
and disastrous' to his
categories.
Figure 19 demonstrates clearly that, where the
24-hour rainfall at the Observatory is below 100
m , only a few minor events and no major events
occur. With increasing 24-hour rainfall, the
0 ..._--_._-u:_'_‘I1BR 'ff' 140
20 40 60 B0 100 120 3~1'1160
proportion of landslide days increases, with
MAXIMUM HOURLY RAINFALL ( nlll ) major events increasingly dominating at higher
rainfalls. When more than 270 mm of rain falls
in 24 hours, every rainstorm event results in a
Fig. 17 Correlation of Landslides Causing Casualties major landslide event.
and Number of Casualties during 1963-83
with Maximum Hourly Rainfall Anywhere

5.4 Correlations with 24-hour Rainfalls

Although clear evidence has been presented that
Hong Kong landslides are dependent on short
duration rainfall intensity, it is worth exami­ II .-1-:
555555;
so
3:3152 I
CEI
zero
2.5
159-'_I3_°'8
II
>­
ldndsllden
40 5;§;§, 2.o ,,,
0 §:§:f: gi
ning whether any correlations exist between land­
slides and 24-hour rainfalls.
2 §3§5§5 5
*Q so ( 10 Iandslldea / day 1_5 E
II6 :§:§:§

32A A
For the May 1982 rainstorm, Fig. 18 shows the
daily rainfalls at the Observatory for the event,
together with the occurrence of all landslides e13£522 e2§2§2 l E°
2° gg; gg; >1o landslides / any ‘-° E

V
for which the dates of failure are known. This
shows that the large majority of landslides
10 riff; 325231 _ 0.5
5
Z §;§;§; ;@;@§a g§m 0 32

u
occurred on the 29th May, the day of the heaviest 0 ._100mu
200 mu ...M
250 500
8Z0 D
5
rainfall. The few failures that occurred up to
a week later were probably caused bv delayed
300 400
24-HOUR RAINFALL AT ROYAL OBSERVATORY (mm)

Lesd 4oo
_I zoo H" S' |`d E
E
Fig. 19 Correlation of Number of Landslides
with the 24-hour Rainfall at the

0
Observatory during the Period 1963-B3

§ 150
I-L aoo
Rainfall
0 100 200 E Q'
Z The relationship shown in Fig. 19 is not unlike
the distribution in Fig. 15 of landslide days
in relation to maximum hourly rainfall for 1982.

g 5°29 31100 Q
This is to be expected, since the heavier the
m 24-hour rainfall, the more likely it is that it
will include a short duration rainfall of high
intensity. The main importance of the corre­
lation between landslides and the 24-hour rain­
fall, however, is its usefulness as a landslide
246 warning criterion, since the accumulation of
rainfall to approach a 'trigger' value can be
MAY JUNE. 1982 sensibly anticipated a few hours in advance,
whereas the short duration intensity cannot
Fig. 1B Relationship between the Daily Rainfall possibly be predicted in advance with any preci­
at the Observatory and the Occurrence of sion. The use of the 24-hour rainfall also has
Landslides during the May 1982 Rainstorm the advantage that measurements from the Obser­

2552
vatory gauge alone can be used to represent the
situation throughout the whole Territory.
5.5 Effect of Antecedent Rainfall
The evidence presented here suggests that the
antecedent rainfall is not a significant factor
in the occurrence of major landslide events in
Hong Kong, despite the fact that it has been
found to be of major significance elsewhere.
E
0o o1
100
70'
v
E
300
A LANDSLIDE EVENT FREQUENCY E

200
"’ DISASTROUS 1 IN s YEARS j
SEVERE

MINOR 3 IN 1 YEAR
1 IN 2 YEARS

This was confirmed in the study by the poor cor­

relations obtained between landslides and ante­
cedent rainfalls for periods of up to 30 days.
In Hong Kong, failures appear to be directly
related to short duration rainfall intensity.
It is common, however, for such high intensity
I N|L i 2%. 100

rainfalls of short duration to occur during pro­ Fig. 20 Approximate Relationship between Rainfall
longed rainfall events. Where limited informa­
tion is available on rainfall intensities and Intensity and Landslide Events in Hong Kong
times of landslides, it is therefore not sur­
prising that a relationship can be found to one-hour and 24-hour rainfalls. This forms the
exist between antecedent rainfall and the occur­ basis of a landslide warning system currently
rence of a large number of landslides. used in Hong Kong for hillside squatters.
Of interest is the fact that the three- or four­
day antecedent rainfall appears to influence the 6. TERRAIN EVALUATION
occurrence of minor landslide events in Hong
Kong when short duration intensities do not 6.1 Background
dominate. This indicates that soil moisture
deficit might well be a factor in rain-induced Terrain evaluation is a powerful technique for
landslides in soil slopes, but that its effects categorising terrain for stability purposes. It
are secondary in situations of high intensity has hitherto been somewhat neglected by geotech­
rainfall. nical engineers, who could make much more use of
the technique for planning and development stu­
5.6 Conclusions dies, and to provide input for land use manage­
ment purposes.
From the rainfall-landslide correlation study The evaluation of terrain is based on a suitable
in Hong Kong, the following main conclusions can
be drawn :
'terrain classification' system,which provides
a means for classifying terrain into homogeneous
land units, largely on the basis of surface fea­
(a) The large majority of landslides are in­ tures. For this purpose, suitable geomorpholo­
duced by localised short duration rainfalls gical 'attributes must be chosen. Attributes
of high intensity, and these landslides for stability purposes com only include slope
take place at about the same time as the angle, land form, vegetation, surface hydrology
peak hourly rainfall. and erosion, each of which is divided into a
number of 'classes'. Once classified in this
(b) Antecedent rainfall is not a major factor way, the terrain evaluation is carried out by
in landslide occurrence, except in cases sorting the land units into categories on the
of minor landslide events which take place basis of predetermined combinations of the attri
under relatively low intensity rainfalls bute classes.
of short duration. In these circumstances,
only a few days antecedent rainfall appears An excellent example of the development of a
to be significant. comprehensive system of terrain evaluation for
engineering purposes is the Australian PUCE sys­
(c) A rainfall intensity of about 70 mm/hour tem described by Grant (l975a, 1975b). This has
appears to be the threshold value above been developed continuously over 20 years into
which landslides occur. The number of a highly complex system which requires the use
landslides and the severity of the conse­ of a large storage capacity computer (Grant &
quences increase dramatically as hourly Finlayson, 1978; Finlayson, 1984).
intensity increases above this level.
In addition to general engineering geological
(d) The 24-hour rainfall usually reflects short assessment maps, 'risk' or 'hazard' maps can be
constructed from terrain evaluation data. This
duration rainfalls of high intensity, and
this can therefore be used as an indicator has been done in some landslide-prone areas of
of the likelihood of landslides. A 24-hour a few countries, notably France (Porcher & Guil­
rainfall of less than 100 mm is very un­ lope, 1979), Italy (Carrara et al, 1978; Carrara
likely to result in a major landslide 1983) and the United States (Nilsen & Brabb,
event. 1977; Nilsen et al, 1979). Good reviews of the
various systems used have been made recently by
The results of the rainfall-landslide correlation
study are summarised broadly in Fig. 20, which (1 5).(1984), Varnes & Keaton (1984) and Varnes
Brggb
shows the approximate frequency of landslide
events (as defined by Lumb, 1975) in terms of The most advanced form of the landslide hazard

2553
map is the digitized form. The methodology used early stage that no existing system of terrain
is no different in principle from that for evaluation was ideally suited to Hong Kong's
manually-compiled maps, but the terrain classi­ peculiar conditions. The GASP approach differs
fication is digitized to facilitate ease of in several important aspects from any other known
storage, retrieval and sorting. The computer­ terrain evaluation or hazard mapping system. It
stored data also acts as a data bank of informa­ combines elements of both of these approaches,
tion that can be called upon at will. Hazard but the zonation framework employed is based on
maps have been prepared in this way in a few the overall geotechnical assessment of the land
countries, the most advanced system probably units, not only on the identification of 'hazards
being that developed by Carrara et al (1978) and from a stability viewpoint. As might be expected
Carrara (1983) in Italy. however, considerations of stability are dominant
A comprehensive report is produced for each Area
6.2 The GASP System in Hong Kong Study, an essential part of which is a base map
and a series of transparent overlay maps.
The Geotechnical Control Office in Hong Kong has
for several years placed considerable emphasis For a Regional Study (1:20 000), the overlay maps
on its Geotechnical Area Studies Programme (GASP) consist of :
to provide systematic geotechnical input for
land use management and development planning (a) Terrain Classification Map,
purposes. This programme and some aspects of (b) Landform Map,
the system of terrain evaluation employed have
been described in several publications (Brand (c) Erosion Map,
et al, l982a, l982b; Burnett & Styles, 1982; (d) Physical Constraints Map,
Burnett et al, l985).
(e) Engineering Geology Map, and
GASP was designed to be carried out in the fol­ (f) Geotechnical Land Use Map (GLUM).
lowing three phases :
(a) Regional Studies ­ Initial geotechnical For a District Study (l:2 500), the overlay maps
assessments (at a sca le of 1:20 000), are :
based entirely on aer ial photograph inter­ (a) Terrain Classification Map,
pretation, site reconnaissance and existing
geotechnical informat ion. (b) Surface Hydrology Map,
(b) District Studies: Sta e l - Initial geo­ (c) Vegetation Map,
tec nica assessments (at a scale of (d) Engineering Data Sheet,
l:2 500), based entir ely on aerial photo­ (e) Engineering Geology Map, and
graph interpretation, site reconnaissance
and existing geotechn ical information, to (f) Geotechnical Land Use Map (GLUM).
give more detailed as sessments of specific
areas identified from Regional Studies. The Engineering Geology Map indicates the broad
pattern of the geological materials and their
(c) District Studies: Sta e 2 - Expanded geo­ general engineering characteristics, and it iden­
tec ica assessments , ased on the results tifies features which are of engineering rele­
of Stage l Studies to gether with data ob­ vance. One of the major purposes of this map is
tained from planned p rogrammes of site in­ to present the geomorphological and engineering
vestigation. geological constraints that influence the allo­
cations of GLUM classes (see below). It is
Regional Studies, each of which covers an area designed for use by engineers and engineering
of 50 to 100 sq. km, provide geotechnical input geologists who require explanations for the
for outline and strategic planning within the nature of the geotechnical limitations which
Territory. In addition, they are designed to affect the terrain.
provide a comprehensive physical land resource
inventory through a computer-based Geotechnical For a Regional Study, the Physical Constraints
Terrain Classification System (GEOTECS). All Map sum arises the physical constraints to re­
eleven Regional Studies have now been completed gional development planning. For a District
or are in their final stages. Study, every item of available engineering geo­
logical information (including every borehole
District Studies are typically concerned with location) is plotted and referenced on the En­
areas of 2 to 4 sq. km. They provide information gineering Data Sheet to avoid confusion on the
suitable for local planning needs and are of Engineering Geology Map; this information is
assistance in the basic layout planning of large shown in a much more general way on the Engineer­
sites. For most localised areas, Stage l Dis­ ing Geology Map of a Regional Study. The En­
trict Studies provide adequate information, a gineering Geology Map and Engineering Data Sheet
Stage 2 Study being required only if very signi­ have been described in some detail by Burnett &
ficant geotechnical constraints are revealed by Styles (l982).
Stage l. Nine Stage l Studies have so far been
completed. As yet, no Stage 2 Studies have been The main findings of a Geotechnical Area Study
undertaken as an integral part of GASP, but a are engendered in the Geotechnical Land Use Map
related Stage 2 area study was completed several (GLUM). In this interpretative map, land units
years ago (Geotechnical Control Office, 1982; are classified into four GLUM 'classes' on the
Rodin et al, 1982). basis of combinations of attributes from the
terrain classification, with due account being
The Geotechnical Control Office realised at an taken of the other data collected during the

2554
°’
TABLE IV
The Four-class Classification System Used for the Geotechnical Land Use Map (GLUM) in Hong Kong

Characteristics of
GLUM Classes Class I Class II Class III Class IV
Geotechnical Limitations Low Moderate High Extreme
Suitability
Development for High Moderate Low Probably unsuitable
Development Low Normal High Very high
Engineering Cost for

Intensity of Site Normal Normal Intensive Very intensive

Investigation Required

Typical Terrain Insitu terrain with Insitu terrain with Insitu terrain with Very steep insitu
Characteristics gentle slopes (0­ slopes between 15° slopes between 30° slopes (>60°) and
l5°) without 5 30° without in­ G 60° without cliffs.
(Some, but not necessarily severe erosion or stability. severe erosion or
all, of the stated instability. instability. Steep to very steep
characteristics will Insitu terrain of insitu and collu­
Cut platforms in gentle slopes Insitu terrain less vial slopes with
occur in the respective insitu terrain. associated with than 15° with his­ history of insta­
GLUM class) bility.
drainage but no tory of landslips.
instability. Colluvial terrain Colluvial terrain
Colluvial terrain less than 15° with with gentle slopes,
with gentle slopes evidence of in­ but associated
(O-l5°) without stability. with instability
severe erosion or and drainage.
High to moderate

W&i@”s g
ih%\.€!lf{@b/hllég , 'EZ ` °:"`°`° °° °°"5
2* E:r;‘:;f.3;.:.2:';;;2.§‘::ii§;s:;1sms;, Ejjfjj
Fig. 22 Example of a Geotechnical Land Use Map
Study. Examples of a GASP terrain classifica­ (GLUM) for the Area Shown in Fig. 2l
tion map and a GLUM are given in Figs. 21 & 22.
The GLUM class is assigned to a unit on the
basis of the likely geotechnical limitations on
the land unit posed by the combination of its The results of GASP are intended for use largely
categories of the six attributes - slope grad­ by planners, but they also provide valuable data
ient, terrain component, terrain morphology, for engineering appraisal and feasibility studies
erosion (which includes instability), slope con­ A recent study carried out to compare the GLUM
class assessments in the Mid-levels area with
dition and hydrology. The four GLUM classes are
summarised in Table IV. the results of engineering stability analyses

2555
(Styles et al, l9Bh) indicated clearly that the
GASP system provides a reliable means of rapidly
delineating localised areas of potential insta­
bility. Q
300 . N i S/ G 3 y’ lu-Lsl
1 lrend
ol aolelylinee lor '4 . .
2 l»v= 5- zu' / ¢»
9 loclol

7. SEMI-EMTIRICAL DESIGN METHODS

vI_200 I Ci1l00lpl|q’
Ea\.l{Q_ O_ ///e
7.1 Background E , -1, /51 /1,.,L_ __ _
The majority of existing cut slopes in residual E
Iu.l/. 19°
/; //O/ /031.1-HI
Ln' /0"|l.l5-9§/
soils were not designed in the engineering sense D- O Nolurol/"L O slope-
r/ I Legend
alnble
but were formed on the basis of judgement or
precedent. In most cases, the precedents were 0 I Nolurol alone-eller failure
established essentially on the basis of common 5| 0 0 léogurpl alop=_&'::nre.:oilure
" a u a e- l a, c..
sense and were then gradually modified in the
light of experience with the performance of the
slopes. In a few instances, this 'precedent'
approach has evolved into a formulated semi­
0 0.5 1 5 21 2.5
I | I3 one ana ya
.ll 0 D glojectwzlopeis-llmmioryl

SLOPE CONTANGENT,X
empirical design method in the form of a set of
design rules. Fig. 24 Design Guidelines Developed by Lane
A particular set of slope design rules can ob­ (1961) on the Basis of the Performance
viously only be applied to a specific geological of Clay Shale Slopes in Northern USA
formation at one geographical location, because
these rules have evolved from the topography, 1500
geology and climate of that location. It is
true, however, that design rules developed for ul N0m6I'\C|2fU|’E
one location may provide a good guide for the 190° 11-5111 Q) o'/. orobubilily
A 111-11 curve(|or cl peruod9' of
luilufe
\L
development of rules for a similar geological
formation elsewhere.
Semi-empirical design rules must be based on the
3 ®
2 IJ'/.years)
probubil`l oHo'l re
‘E 500 curve (ford
years of ‘lL
examination of stable and unstable slopes of
similar geology. This can in some cases result
in the establishment of simple relationships
I IL yeorsl
E ® Median l rm o period of
between slope height and slope angle. At the
other extreme, very complicated statistical
relationships can sometimes be devised among a
'1
Q 1
200 6 ,
` H1 ,L
'yearsl "
‘-‘I 250 2 ° (9:t£l:'r5lof§?zilizi§:1 of
6 150 ' _ =_ ` o l96L lopogrophic mop
large number of geological and geomorphological '\\ A Reconslrucilon lo 1958
parameters.
An example of the simplest type of semi-empirical
e3
23 k100
;===i1an. __ W`NQllll
-_ _ 1950
U Reconslruclionlo
P Illlllllllll H
design rules is illustrated in Fig. 23, which
shows the recommended slope angles for different § llllllllllllll=i`\§§&!lllI
heights of cut slopes on the basis of the degree
(LS50 1_ll_hh*
I-ll-11 _ _#_-l_
<1-ll
of weathering of Wellington greywacke (Grant­
Taylor, l96h; Taylor et al, 1977). Earlier,
Lane (1961) had correlated slope height and
slope angle for slopes in clay shales in Montana,
North Dakota and Kansas, to produce the plot
t5E25
-llll_l%Q\!ll
T1_l1-lll _00]
Hlll-Illll ‘Ill
\ \l
shown in Fig. 24. The work by Shuk (1965, 1968),
sum arised in Fig. 25, is of particular note,
I _llll _ .
o 0.15 0.25 0.5 1015 25

1
HEIGHT OF CUT
( 6m 6-10m 10-16m
Comblelely weathered ( yellow-brown; TANGENT OF SLOPE ANGLE,B->
eaelly eruehed In llngere ) on 1 n 2 0
Fig. 25 Design Guidelines Developed by Shuk (l965,
Hlghly weathered ( yelow-brown; 1968) for Clay Shale Slopes near Bogota
partly crushed In llngere )

Moderately wealhered ( brown; since he combined a probability of failure in

broken elally wllh hammer ) his log-log plot of height versus slope angle
for natural shale slopes near Bogota, Colombia.
Slghtly weathered ( grey, partly brown; Shuk's approach was later used by McMahon (1976)
broken wllh hammer) 2 on 2 on for data obtained on 233 slopes, for which he
found that the log-log straight line relation­
Freeh ( mnwealhered; grey;
broken only wllh hammer )
ship applied for any rock type.
Morphometric models have been used to establish
the attributes common to landslides. The work
Fig. 23 Simple Empirical Design Rules Developed for by Blong (1973) and Crozier (1973) are good
Wellington Greywacke by Grant-Taylor (1964) examples of this approach in its more descriptive

2556
 form. A more complex model was devised by Neu­
land (1976) using 31 variables, which included
morphometric, geological, stratification and
soil mechanics attributes. Equally sophisticated
are the terrain models being used in parts of
Italy (Carrara et al, 1977, 1978; Carrara, 1983).

7.2 The CHASE Study in Hong Kong

A major Study (CHASE) was carried out in Hong
Kong a few years ago (Brand, 1982; Brand E Hud­
son, l9B2) to examine the possibility of esta­
blishing semi-empirical guidelines for the design
of cut slopes in the residual soil profiles.
Detailed examinations were conducted on l77
stable and failed slopes, and the data was
analysed statistically in an attempt to obtain

A/
simple correlations among the geological, hydro­
logical and geometrical factors which control
cut slope stability. More than 200 items of

:~
data were collected for each slope to ensure
that any factors which influenced its stability
were recorded.
A typical CHASE slope is shown in Fig. 26. This
is A5 m high and is cut at an angle of about 60°
into a steep natural hillside covered with dense
Fig. 26 Typical Stable Hong Kong Cut Slope Included vegetation. The material of the slope comprises

501968
U-1 , 50 A
in the CHASE Study (Brand & Hudson, 1982) 5.5 m of colluvium overlying decomposed granite,

KEY: KEY:
40_0 O______
O ,\H.__ 5
° 0 ,_ osnats H A /_V
~ _ H .vse ~ =
1972
1 40 - n
`|STABLE|
~ ° ° °o-° O
°
xFJ ,
____ _M
~
_:
A
°=- :“,:::,~:‘ H
.`.`.
f'
/ f;f§_,_
.‘. |uNs‘rABLE| 'gy -`
""~~’--’ °°
"" ._
:":.2&°"“
\
H V
’ A
;»‘-fiif;
30- oo0 O
0 0CO
'O O
Q33Q-;;:;:;:;:;:;'-'-:-
§;§;§§;§z@a§ `»'f- 0 \ to1 .-»V-
O 1 '~~-­
Ooo
. O°°,
O. OE .§l§I§"'"`
. I 0 J ..`\\
\."
o oo 2 °0~§~t,f.
, 0¢ 3 I _§§2f§§§§§f§f;;;:;: cgi `\` \
O I |-U :if-f \‘.*z*¢`$
20~ ‘A 9 20~§§§¥ 3, E;
° I Eiéfiézéefai o`
*_ °"»
- “ Q O -2122 _ ,‘~~._"-F- 1:

\\ Q8 `
°.°
» an 9.Q0° °., I 1950 ° "*F.. ‘T-eq. -_=.5e§5§§§2§2§5§5§§§5§§;g; . 3 ?

E§E§E§E§E§E§§§E§E,,1}E§E§E§E§E§E; 1 ug \ \~~_
\ “°° ° ° ;~;-;"""5E5£5E5E§E5E5E5_\§§E§E§E§€§E_ ~:3°i1"*_.\;` ° "
\ °»7:7;5f5:3:1;-'-'-’
10 » | s 10 . .O°° ' 0 .O "" OZftfif'
i;§§§=a;;;z;' ""' `I-~I‘Q:
.__' 5:5:f:§;§1;._ Z\>\ \ ‘»_` O
\ ~ 0 °O'°-~ 0° Q§§§§"" t~~.j"`~2¥::;§~f‘.i~_°
‘S‘a°""Y a"a'YS'S’ ° E ` ~ _ c
. . . Le
_ 1982 S °.:iii-1-2-I ' _:Z:I:I:I:l:I;1:.:§:.:1'...
a - :§2§2§2é2é2§;§;%;§f§=;§ .U`'`
Stability
ao 50 70 90 30 so 70 90
0 1 ‘ K I I 1 O 1-ggzggg; 3;1;;§;§;:;§¢:;§;§;§;§;§;I;§;§;§:§?;2;1;I:?- 5:3:i:§:¥:7:5i:§:5:f’i:§:?'i'T:2:3:§tT:I-Q:1:25:31:§:§:-:§:§2;:3:»¢:- I

ANGLE(deg) ANGLE(ded
Fig. 27 Relationship between Height and Angle for Fig. 28 Relationship between Height and Angle for the
the l77 Hong Kon Cut Slopes Included in ll7 Hong Kong Granite Soil Cut Slopes Included
the CHASE Study ?Brand & Hudson, l982) in the CHASE Study (Brand & Hudson, l982)

2557
which varies from grade V at the top to grade IV
at road level. The surface is covered with a
protective coating of 'chunam' (lime stabilised
soil), which is com only used in Hong Kong to
prevent surface erosion and to minimise infil­ Ta
tration.
Despite the vast amount of data collected, and
the extensive statistical analyses undertaken,
the CHASE Study did not reveal the simple rela­
tionships hoped for, but it did result in the
establishment of some lower bound envelopes for
the relationships between height and slope angle
of the kind shown in Figs. 27 & 28. These have
.xg Budmg mass

proved to be useful for the initial design of

new cut slopes, and for assessments of the like­ Sllp surface
lihood of failure of existing slopes.
Fig. 29 Basis of the Limit Equilibrium Method
Figure 27 is of interest for the information it of Stability Analysis for Slopes
provides about empirical design rules adopted at
various times in Hong Kong's history. It can be
seen that none of these was meaningful in the
context of the results of the CHASE Study. It
is also evident that classical stability analyses Assessed State-of-the-Art of Stability Predictions
are probably very conservative for Hong Kong con­ for Residual Soil Slopes
ditions. The same conclusion is well supported
by Fig. 28, where the changes in slope height
and angle are shown for a large nu ber of grani­
tic soil cut slopes which had previously failed.
Component State-of-the-Art
Mode of failure Good to poor
8. SOIL MECHANICS PREDICTIONS Loading Very good
8.1 Summary Method of analysis Very good
Slope stability analysis is usually based on the
Shear strength Fair
limit equilibrium approach, for which the equili­ Pore pressure Very poor
brium of a sliding mass is examined (Fig. 29).
The degree of stability is quantified in terms Overall Fair to poor
of a 'factor of safety', which is most commonly
defined as the ratio between the average shear
resistance and the average shear stress along
the most critical slip surface, i.e. : (c) method of analysis,
1==§ (1)
s
3
(d) shear strength, and
(e) pore pressure distribution at failure
Table V summarises, in very broad terms, the
The determination of the F-value for a particular Author's assessment of the overall state-of-the­
slope therefore requires : art of stability predictions in residual soil
(a) prediction of the correct mode of failure slopes in terms of the five components of predic
(i.e. selection of the critical slip sur­ tion. The accurate determination of loads is
face), thought to present very little difficulty, the
dead weight of the sliding mass (Fig. 29) being
(b) prediction of the distribution of shear readily computed on the basis of data from bore­
stress over the critical slip surface, and hole samples. Neither are there reasons to be­
lieve that the available methods of analysis are
(c) prediction of the distribution of shear anything but very good, as long as these are
resistance over the critical slip surface. restricted to appropriate methods for the parti­
cular mode of failure. Our ability to select a
The distribution of shear stress along the criti­ correct mode of failure varies fairly widely
cal slip surface is dependent on the loading and depending on the geology and weathering profile
the method of analysis employed. The shear re­ for the slope. Because of the small amount of
sistance along the slip surface is governed by work that has been done to investigate the shear
the effective shear strengths of the materials strength properties of residual soils, our know­
of the slope and the normal effective stress ledge of this component is only fair. However,
distribution. The normal effective stress dis­ by far the major difficulty with stability pre­
tribution is, in turn, a function of the pore dictions in residual soil slopes is the poor
pressure distribution at failure and the method state-of-the-art with respect to the prediction
of analysis. The five 'components' of stability of pore pressure distribution at failure.
prediction are therefore :
(a) mode of failure, 8.2 Mode of Failure
(b) loading, The selection of a suitable mode of failure is

2558
crucial to accurate stability predictions in all usually inappropriate, and methods which apply to
soil types. It is particularly crucial for any shape of surface must be used. In this cate­
slopes in residual profiles, where the potential gory, are the well-known methods by Janbu (1954,
failure surface is often governed by geological 1973), Morgenstern & Price (1965) and Sarma (1973
detail. Some guidance on the accurate assess­ 1979). The first two of these are in very common
ment of failure modes for geologically complex usage. Although Morgenstern & Price made more
conditions was given many years ago by Terzaghi satisfactory assumptions than Janbu, the latter's
(1950). More recent important publications in method is much easier to programme and requires
this context are those by Skempton & Hutchinson much smaller computer capacity. For these rea­
(1969), Deere 8 Patton (1971), Patton & Hendron sons, it is more widely used for routine stabi­
(1974) and Coates (1977). lity calculations. Sarma's method is as rigorous
as that of Morgenstern & Price, but it requires
Whereas the pre-failure geometry of a slope is
easily defined, it is sometimes difficult to tlisa
me.iterations with consequent less computer
decide upon the critical potential failure sur­
face for design. Occasionally, even a post­ All the above methods of slope stability analysis
failure surface cannot readily be determined are based on a two-dimensional failure mode. In
because of the multiple-failure nature of some practice, however, slope failures are nearly al­
landslides in residual materials. ways three-dimensional, but it is rare for account
to be taken of this in design or stability assess­
As a general rule, the shear strength of a resi­ ment, or even during the back analyses of failures
dual soil increases with depth, and slope failures that have occurred. Three-dimensional effects on
can therefore be expected to occur on relatively slope stability analysis have been examined by
shallow slip surfaces. These surfaces are large­ Baligh & Azzouz (1975), Azzouz et al (1981),
ly controlled by the weathering profile. Fail­ Lovell (1984) and Leshchinsky et al (1985), but
ures most frequently occur along surfaces dic­ more work on this needs to be done.
tated largely by relict joints or by boundaries
between weathering zones. By the very nature A great deal of effort has been put into refining
of residual weathering profiles, non-circular the details of methods of stability analysis over
failure surfaces are the most common, and these the years, without sufficient attention being
are often almost planar over a major proportion paid to the wav in which these methods are used.
of their length. A small study carried out in Hong Kong (tumsdaine
& Tang, 1981, 1982) is revealing in this res­
Geological detail is often crucial to the loca­ pect. A large number of organisations intimately
tion of the critical slip surface. A complex involved with slope stability assessment in Hong
weathering profile can rarely be adequately in­ Kong on a regular basis were asked to compute
vestigated at the design stage. In those few the factors of safety for a number of problem
instances where the profile is determined in situations by means of the well-known methods of
considerable detail by extensive surface and analysis for non-circular slip surfaces. Many
subsurface investigations, the true geological incorrect answers were received, and this re­
situation usually does not lend itself to being flected the fact that many of the computer pro­
properly modelled for the purposes of meaningful grams being used had never been properly checked
engineering analysis. or documented. In addition, an alarming number
of errors had resulted from careless data input.
The importance of soil fabric to the mode of
slope failure cannot be over-emphasized. In
particular, relict joints, especially when slick­ 8.4 Shear Strength Measurement
ensided, can be instrumental in the initiation
of a failure, and their presence can appreciably Despite the obvious objections to the measure­
reduce the mass strength of the soil (see Section ment of shear strengths on residual soils by
8.4). means of laboratory tests, these still comprise
the most satisfactory means of establishing the
likely range of shear strengths on the softer
Some
detailexamples
to slope of the importance
failure modes inofresidua
geological
soils materials (grades V & VI). The effects of core­
are given later in Section 9. stones and other large-sized particles, however,
cannot be determined, and there is no doubt that
laboratory strength tests carried out on the
8.3 Methods of Analysis 'matrix' material of residual soils and colluvium
will usually underestimate the shear strength of
Although the state-of-the-art with respect to the insitu material because of the neglect of the
methods of stability analysis is thought to be boulder content.
very good, it is well-known that the available Residual soils and colluvium are invariably un­
methods produce different theoretical F-values
because of the different assumptions made in saturated and are of relatively high permeability.
their formulation. Many papers have been Stability computations must therefore always be
published which compare the various assumptions made in terms of effective stresses; analysis on
and which make comparisons of the numerical the basis of undrained strengths has no relevance.
factors of safety obtained for certain specific Because of their high permeabilities (usually
slope situations. For the up-to-date position, lO'4 to l0‘5 m/sec.), rainwater infiltrates with
reference should be made to the recent publica­ ease into most residual soils and colluvium, and
tions by Fredlund et al (1981), Fredlund (1984)
and Ching & Fredlund (1984). it is thought likely that saturation conditions
will be approached at shallow depths in the field
For residual soil slopes, methods of stability during the life of a slope. It is therefore
analysis which employ circular slip surfaces are generally felt to be appropriate to measure

2559
strength parameters on the basis of shear tests to the value of the average shear stress on that
carried out on saturated soil specimens. Al­ surface. Rain-induced slope failures, therefore
though this may often be a more severe condition take place under conditions of almost constant
than that experienced by the soil insitu, it iggil).stress and increasing pore pressure (Brand
remains the only certain means by which a 'base'
shear strength envelope can be established.
The triaxial test is most commonly conducted by
The triaxial test is the most widely used method increasing the axial stress, ol, to failure
for shear strength measurement on residual soils while
This isthe cell pressure,
carried out either03,asisa kept constant.
drained test
(Brand & Phillipson, 1985). Test specimens
should be as large as possible, full-diameter (Au==O), or as an undrained test (or constant
lengths of drillhole sample being ideal for water content test) with pore pressure measured
routine work, and they are usually saturated by throughout the shearing process. In contrast,
the application of a sufficiently high back pres for rain-induced failure, U1 and U3 are almost
sure prior to shear. Either drained tests (CD) constant, and the pore pressure increases to
or consolidated undrained tests with pore pres­ failure. The comparative stress paths for the
sure measurement (CU) can be used, but the common triaxial tests and for the real field
latter are much to be preferred, because they situation areasshown
are defined : in Fig. 31, in which p' & q
are quicker and provide much more information
about the stress-strain behaviour of the soil.
Cell pressures used for triaxial testing must
relate to the correct insitu stress range if P' =U1"'Y"`
+ Ga ‘<11
q =- U3
__'7`_` (2)
the measured strengths are to be meaningful.
Critical slip surfaces in residual soil slopes
are most commonly shallow, and the effective q Strength envelope
stresses on these are therefore low (typically
30 to 200 kPa). At such low effective cell
pressures, however, triaxial tests are difficult
to control satisfactorily, and this stress range
is not recommended for routine use. There is
S' _______ co
Sri! -'
some evidence to suggest that the strength
envelopes for some residual soils are curved at
low effective stresses, and that the straight­
line projection of strengths measured at high
stresses underestimates the strengths in the
low stress range (Fig. 30). It is almost cer­
tain that this is part of the explanation for
why many stable residual soil slopes have theo­
retical factors of safety of less than 1.0.
As discussed in Section 8.5 below, failures of
residual soil slopes are invariably caused sole­
Cu
!»Pg' of D
cu. |
( Compreeslve )

| Fwld I ( DHatant)
I

Correct
ly by pore pressure increases. In its simplest
form, the mechanism of failure is therefore that STFBSS I’8l'lQ8
the soil suction (i.e. negative pore pressure)
decreases (i.e. pore pressure increases), there­ Fig. 31 Comparison between Stress Paths for
by reducing the shear strength of the soil, as Rain-induced Slope Failure and Those
dictated by the effective stress principle. for Triaxial Tests (Brand, 1981)

I. I/ ,_
Failure will occur when the average shear resis­

§ .
tance on the critical slip surface has decreased

IP \
Figure 31 illustrates that the stress paths com­
monly followed in the triaxial test are quite
True
Testresuhs different from that which pertains in the field,
and that the stress ranges over which triaxial

_
,D strength tests are usually conducted are generally not
_ envelope -\ appropriate to the field stress conditions.
P Where the strength envelope of a soil is marked­
(D ly curved, this disparity will result in an
appreciable underestimate of the correct shear

I/
w

E _//" Assumed envempe

(D /'
a' range for o'fortests
range CV
strength for stability assessment (Fig. 30). It
is possible to follow the correct stress path in
the laboratory simply by decreasing the cell
pressure, but this does not simulate the correct
mechanism of failure in the field. This can
only be done by means of a constant load test
in which the pore pressure is increased from an
initial negative value until failure occurs.
Tests of this kind have been carried out by the
shaHow sudaces Geotechnical Control Office for some time, but
the experimental difficulties are considerable,
and this test procedure cannot be recommended
Fig. 30 Underestimate of Shear Strengths because of for routine use.
Incorrect Strength Envelope Deduced from
Data Obtained at Too High Nonnal Stresses Direct shear (shear box) testing is deservedly

2560
bccoming more commonly used for the strength well-documented Hong Kong case (Hunt, 1982),
assessment of residual soils. Apart from its which was of major economic significance, led
relative simplicity, this method has obvious to an extensive investigation into the shear
merits over the triaxial test when determina­ strength of jointed sedimentary volcanic mat­
tions of strengths along relict joints are re­ erial (Koo, l9B2a, l982b). Persistent joint sur­
quired, because orientated specimens can be pre­ faces were coated with a thin black-brown deposit
pared. However, there is no certain method of believed to be the precipitation of iron and
saturation, and pore pressures cannot be measured manganese oxide products that filled the joints
during shear. Despite its lack of theoretical in the parent rock during the course of weather­
'purity', the direct shear test is seen by some ing. The results of laboratory direct shear and
as providing a means of obtaining shear strength triaxial tests showed that the effective strength
data under conditions which model those in the parameters on the joints were much lower than
field more closely than the triaxial test. It those measured on the intact material.
is also readily adaptable for insitu measure­
ments of shear strength on relatively large
masses of material. The equipment shown in Fig.
32 has been used successfully for testing 300 x
300 mm soil and soft rock specimens in Hong Kong
(Brand et al, l983b). A typical test specimen of
. ”‘\ <~ Q ­
grade V granitic soil is shown in Fig. 33.
Relict joints and other such discontinuities are 'Gk
worthy of special mention, because they frequent­
ly play a major part in slope instability (Hen­
cher et al, 1984). These are often slickensided
and sometimes coated with thin deposits of very
weak material. The significance of smooth slick­
ensides in decomposed igneous and metamorphic
rocks was fully discussed in a paper by St John
et al (1969). There have been some reported
failures of quite flat slopes attributable to
extremely low relict joint strengths. One such

Fig. 33 Block Sample of Grade V Hong Kong Granite

Used in the Apparatus Shown in Fig. 32

8.5 Pore Pressure Prediction

For a residual soil slope, the prediction of the
pore pressure distribution is by far the most
critical factor for stability analysis. This
is particularly so since most failures in resi­
dual soil slopes are caused by rainfall.
The hydrological effects of rainfall on a per­
meable slope are depicted in Fig. 34. Some of
the water runs off the slope and may cause sur­
face erosion if there is inadequate surface pro­
tection. Because of the high soil permeability,
however, the majority of the water infiltrates.
This causes the water table to rise, or it may
cause a perched water table to be formed at some
less permeable boundary, usually dictated by the
weathering profile. Above the water table, the
degree of saturation of the soil increases, and
v._ . the soil suction (i.e. negative pore pressure)
- ~,.~ \~‘vs_ therefore decreases.
Failures in residual cut slopes are thought to
be caused mostly by the 'wetting-up' process by
which the soil suction (and hence the soil
Fig. 32 insitu Direct Shear Test Apparatus Used on strength) is decreased, but there is some evi­
Hong Kong Residual Soi] (Brand et al, 1983b) dence to suggest that transient rises in ground­

2561
 rainfall, Lumb (1975) later showed that the
thickness, h, of the wetting front after time t
Ralnfall could be approximated by :
_ kt
Innnranon 1 1

§" \
Perched water table 1 Surface runoff
`
`¥§J\
i ` E J- Eroslon
"-~_______
Rise In wafer fable §
83
DEGREE OF SATURATION

o -:-: :-:-:-:- ' """" 2:1:5:f:?: 2!:1:¥:I:I:J ‘O

:, :;:;:;:;: ;:;:;:;:;:;:;:;: :;:;:-:-'

Fig. 34 Diagrammetic Representation of the h

Effects of Rainwater on a Slope

water tables are responsible for some rain­

induced landslides (Premchitt et al, 1985). =§:s:§:a=s:z:5:§:s:5=a=§2`e555_YY9_!?F"9 _fI9F!?_E;2i
Rain-induced slope stability failures thus occur
as a direct result of pore pressure increases
(see Section B.h), and pore pressure distribution
is therefore the variable of most concern. -=~-- ~ "T_
The pore pressure distribution in a residual soil
is dependent on the pattern of rainfall and the
hydrogeology of the slope. Pore pressure is
therefore a variable which is independent of soil
mechanics considerations, being imposed upon a
slope by external influences. For this reason, Representation of the Advance of the
it is extremely difficult to predict the appro­ wetting Front as Hater Infiltrates
priate pore pressures for slope design or stabi­ into Soil (Lumb, 1962b, 1975)
lity assessment. This fact is vividly illus­
trated in Fig. 35, which shows the rapid varia­
tion in piezometric head recorded in a Hong Kong Despite its shortcomings, equation (3) is fre­
slope during a rainstorm in June 1982. quently used for the prediction of design pore
pressures in Hong Kong (Beattie & Chau, 1976;
For the infiltration of water into the horizontal Beattie & Attewill, 1977; Geotechnical Control
surface of a porous medium, Lumb (l962b) derived Office, 1984) by the superimposition of the depth
an expression for the advance of the 'wetting of wetting front onto the groundwater table level
front' (Fig. 36). On the assumption that diffu­ at the end of the wet season. For this purpose,

nI 140
=.'
sion is negligible at the end of an intensive a rainfall return period of ten years is commonly
employed. This approach suffers from the fact
that the value of h is directly proportional to

5E eo ~
Sf-So, and the initial and final degrees of satu­

E
142
ration are somewhat speculative. In practice,
the Hong Kong wetting band thickness is usually

§1aa §m
g Plezometrlc head calculated as being about two metres. Such pre­

E 2040 E
dictions will generally lead to overly conserva­
tive slope design and stability assessments,
especially if it is assumed that no soil suction
exists anywhere in the soil profile.
Q

na|nfa|| For the analysis of slope stability in residual

soils and colluvium, measured pore pressure data
is much to be preferred to the application of
uncertain predictive methods. For such data to
136OB| 1-1 I* | i I i 1 0
2 1 B 24 OB 1 2 1 B 24
be meaningful, however, pore pressures must be
monitored for a sufficiently long period of time,
and piezometers must be installed at the appro­
28 May 1982 29 May 1982 priate depths at sufficient locations on the
DATE AND TME slope. In addition, an appropriate type of
instrument must be used which can respond rapidly
to pore pressure changes. In Hong Kong, numeri­
Fig. 35 Rapid Changes in the Groundwater Table cal modelling techniques have been applied in a
Measured in a Hong Kong Slope few cases (Leach & Herbert, 1982), but simple

2562
direct correlation and extrapolation methods are increasing the shear strength by (ua-uw)tan ¢b,
more commonly resorted to (Koo & Lumb, 1981; where ab is the angle of internal friction with
Endicott, 1982). respect to matric suction (Fredlund, 1981; Ho &
Fredlund, 1982; Fredlund & Rahardjo, 1985). It
Because of the rapid changes in pore pressure has been suggested (Sweeney & Robertson, 1982;
that occur with rainfall in many residual soil Boonsinsuk & Yong, 1982) that this increase in
slopes, the critical pore pressures are rarely soil strength should be permissible for the
obtained from normal piezometer measurements. design of Hong Kong slopes as long as insitu
In order to enable standpipe piezometers to measurements are available to substantiate that
record the maximum water levels attained during suctions continue to exist in the slopes through­
a given period, Halcrow 'buckets' (British Patent out the year.
No. l53B487)(Fig. 37) have become popular in
Hong Kong. These simple plastic devices are Much valuable data on soil suctions has been ob­
threaded onto a weighted nylon string at selected tained on some Hong Kong slopes over several
depth intervals above the normal base water level years. It has been found that the suctions on
in a standpipe. When the string of buckets is relatively shallow potential slip surfaces de­
withdrawn, the highest transient water level is crease everywhere close to zero during a wet
indicated by the upper limit of water-filled season, and that suction cannot therefore normal­
buckets. This system is cheap and simple, and ly be relied upon for slope stability purposes.
it can be used to provide important design in­ However, some measurements made at considerable
formation when expens ive automatic piezometer depth in Hong Kong granite (Sweeney, 1982;
systems,of the type described below, cannot be Sweeney & Robertson, 1982) suggest that not all
justified. soil suction is destroyed by infiltration. The
extent to which suctions play a significant part
in slope stability in residual soil profiles else
where in the world clearly depends on the pattern
O Strengthened of rainfall and the infiltration characteristics
of the material concerned.
Auxmary f_\, 'wng'ugs
drainage Much valuable data on soil suctions can be ob­
hom |­
¢ tained for shallow depths by means of simple
agricultural tensiometers (Anderson, 1984; Green­
;@-ilntake hole way et al, 1984). These instruments, which con­
Q sist of a small porous ceramic tip connected to
E a perspex tube and a vacuum gauge (Fig. 38), have
been found to perform satisfactorily. They suf­
E Plastic float for fer from the serious disadvantage, however, that
they must be read manually, and critical readings
-|-easy visual observation during intensive rainfall are frequently unable
0 10 20mm\`-'
I of waterlevelin bucket to be obtained.
L_L;J, C Some sophisticated, automatic reading equipment

_lill
has been developed to measure both positive and
negative pore pressures in Hong Kong conditions
FRONT VIEW SIDE VIEW
Fig. 37 Halcrow Plastic 'Buckets' Used for 17 Reeervoh bunon
Recording the Maximum Hater Elevation Reeervoh cover
Attained in a Standpipe Piezometer
Jet-Hhreservoh cap
During most of the year, suctions exist in many
ii
residual soil slopes. These suctions, which can ,. 'O' Hng cap seal
Itlllll
be of high magnitudes, are reduced dramatically Depth label
by the process of water infiltration during
rainfall. There are almost certainly situations Port molded lnto slde
where even the heaviest rainfall does not com­ wa" ofbodytube
pletely destroy the soil suctions, and these Ventsuew
continue to contribute to the stability of a Heavy clear pIasNc
slope. Whereas slope design is commonly based cover plate
on predictions of the maximum positive pore
pressures that are likely to develop in the Vacuum Malgauge
slope, no account is usually taken of negative
pore pressures that might be sustained under the Backup washer
worst rainfall conditions. This neglect of the
contribution of soil suction to the shear strength 'O' rlng stem seal
of the soil might be the main reason why stable
residual soil slopes often have theoretical fac­ Plaauc body tube
tors of safety of less than unity. Ceramic aenslng tlp
Programmes of laboratory testing carried out
over many years on Hong Kong residual soils have
proved that suction pressures act as 'modified' Fig. 38 Agricultural Tensiometer for the
Measurement of Insitu Soil Suction
effective stresses, a matric suction of ua -uw

2563
(Pope et al, 1982). This is based on similar 8.6 Factors of Safety
equipment designed in the United Kingdom (Ander­
son & Burt, 1977; Burt, 1978). Small diameter Residual soil slopes commonly occur in steep
tensiometer tips of high air-entry ceramic are natural terrain, and it is often necessary to
each connected by twin water-filled flexible work to very low factors of safety in order to
nylon tubes to a single transducer through a avoid the excessive costs of adopting higher
fluid scanning switch (Fig. 39). The switch is values. In these circumstances, it is logical
analogous to an electrical wafer switch, in that to relate design safety factors to the consequen­
it allows sequential measurement of a number of ces of failure in terms of risk of casualties
fluid pressure inputs. A rotating metal valve and risk of damage to property and services.
inside the water-filled wafer (Scanivalve) en­
ables the pressure transducer to be connected to Slope design practice in Hong Kong is governed
24 different inputs in turn, two of which are by the Geotechnical Manual for Slopes (Geotech­
used as reference reservoirs for calibration nical Control Office, 1984), which specifies
purposes. The measured pore pressures are re­ factors of safety. Table VI shows the recom­
corded on a variety of types of strip-chart re­ mended factors against loss of life and against
corder, the pressure-sensitive paper type now economic risk for new soil and rock cut slopes.
being regarded as the most reliable. The system These values are for groundwater conditions re­
is powered by a twelve volt battery, and all sulting from a ten-year return period rainfall.
measuring and recording components are housed in There are three risk categories in each case ­
a water-tight metal box. Although installation 'negligible', 'low' and 'high'. The risk-to-life
and maintenance costs are high, the systems have category reflects the likelihood of loss of life
been found to be reliable for periods of up to in the event of failure, while the economic risk
three years. category reflects the likely magnitude of econo­
mic loss. Typical of high risk-to-life slopes
are high cut slopes adjacent to occupied build­

_Datum
---J
Zora- '°°°'“°'
n¢|gn\
relerenct
" Gonnoctlonl to
ings. An example of a negligible risk-to-life
slope is one which threatens only a lightly
trafficked secondary road. Other examples of
the two risk categories for Hong Kong conditions
are given in Tables VII & VIII.

tonslometorn TABLE VI
1m Scanlvalvo
iluld
Factors of Safety Specified for New Cut Slopes in
Hong Kong (Geotechnical Control Office, 1984)
tcunnlng
switch
Risk to Life
,mf
Imolro
' lenslon Tlmlng
lor c|°'"° "1
clrcultScanlvelvo
(vnu) surface
reloronee
BI d b '
°'°‘"" _S
2
High Low Negligible
ee 'U e <-Connecting
Rubber bun lube 3 High 1.4 1.4 1.4
nr:

Perspel rlng
(glued to cup)
g Low 1.4 1.2 1.2
E

Negligible 1.4 1.2 >l.0

Porous cup
(water flllod) Note: these factors of safety _are for a rain­
fall return period of ten years

Fig. 39 Automatic Scanivalve System Used in Hong Hong Kong‘s Geotechnical Manual for Slopes also
Kong for the Measurement of Positive and gives guidance on preventive and remedial works
Negative Pore Pressures (Pope et al, 1982) to existing slopes. When analysing an existing
slope to determine the extent of any preventive
(i.e. before failure) or remedial (i.e. after
In the context of pore pressure prediction, men­ failure) works required, the performance history
tion must be made of soil 'pipes' (erosion tun­ of that slope can be of considerable assistance
nels) which frequently occur in residual and to the designer. There is, for example, an
colluvial terrain. These can vary in size from opportunity to examine the geology of the slope
a few millimetres to perhaps a metre in diameter, more closely than for an undeveloped site, and
and they are capable of very rapidly conveying to obtain more realistic information on ground­
large volumes of water downslope (Fig. 35). water. The Manual maintains that the designer
There is no doubt that they are frequently an is therefore able to adopt with confidence fac­
important factor in slope failures (Pierson, tors of safety for proposed preventive or reme­
1983). In Hong Kong, pipes can be of major sig­ dial works that are slightly lower than those
nificance to the hydrogeology of an area (Leach specified in Table VI for new works, as long as
& Herbert, 1982; Nash & Dale, 1983; Premchitt et rigorous geological and geotechnical investiga­
al, 1985). tions are conducted (which include a thorough

2564
 TABLE VII TABLE VIII
Examples of Risk-to-Life Categories for Slope Failures Examples of Economic Risk Categories for Slope Failures
in Hong Kong (Geotechnical Control Office, 1984) in Hong Kong (Geotechnical Control Office, 1984)

Examp le - Example _ Risk to Life

Neg. Low High

Economic Risk

Neg. Low High

(l) Fallures affecting country parks (l) Failures affecting country parks. °
and llghtly used open-air
recreation areas.
°
(2) Failures affecting rural (B),
(2) Failures affecting roads with low , feeder, district distributor and ,

parks. '
traffic density. local distributor roads which are
not sole accesses.
(J) Failures affecting storage com- , (J) Failures affecting open-air car
pounds (non-dangerous goods).
(4) Failures affecting densely used (5) Failures affecting rural (A) or
open
tles(e -g.spsces and
1s tt1ng ­out recreation faclll- ,
areas, primary distributor roads which *
playgrounds, car parks).
are not sole accesses.

(5) Failures affecting roads wlth high (S) Failures affecting essential
vehicular or pedestrian traffic *
density.
services, which could cause loss
of that service for a temporary '
period (e.g. power, wster and gas
mains).
(6) Failures affecting publlc waltlng
areas (e.g. railway platforms, * (6) Failures affectlng rural or urban
bus stops, petrol stations). trunk
importance.
roads or roads of strategic *
buildings (e.g._res1dent1sl, ,
(7) Failures affecting occupied
educational, commerical, (7) Failures affecting essential
services, which could cause loss
industrial). of that service for an extended
(B) Failures affecting buildings ,
storing dangerous goods.
period.

which could cause excessive i

(B) Failures affecting buildlngs,
structural damage.

examination of slope maintenance history, ground­

water records, rainfall records and any slope
monitoring records). These reduced factors of TABLE IX
safety are shown in Table IX.
Factors of Safety Specified for Existing Cut Slopes
mHmgmm(%MuMkMCmUMOHkLl%M
8.7 Conclusions
It is clear from the preceding brief review that
the application of rigorous soil mechanics me­
thods of slope stability analysis to residual Factor of Safety 1.2 1.1 >l.0
Risk to Life P High ( Low | Negligible'
profiles is extremely difficult. lt is prudent
to adopt this approach only in conjunction with
a thorough engineering geological assessment and Kong enables a more detailed state-of-the-art
with the liberal application of sound judgement. assessment to be made than that contained in
Analytical solutions alone cannot be relied upon, Table V. This has been done for Hong Kong‘s
as succinctly stated by Peck (1975), thus : residual soil slopes by Hencher et al (1986)
" Analytical procedures have been developed for cal­ (Table X), who also suggested where advances
culating the factors of safety of a slope under could be made in the two particularly poorly
various conditions ... The theories and their understood areas of shear strength and pore pres­
applications have met with some successes and more sure prediction (Table XI).
failures. Most of the failures are a consequence
of oversimplification. Valuable information for future design purposes
can often be obtained from back analyses of slope
Even the most complex theories are necessarily over­ failures. Some thorough analyses have been
simplifications of nature; they are general rather carried out for some Hong Kong landslides (Hen­
than specific. A theory that would take into cher et al, 1984). Such analyses can only be
account all the significant variables at a given meaningful, however, in circumstances where the
location would he far too complex for use. Further­ majority of factors which contributed to the
more the parameters that must be evaluated to use failure can be positively evaluated, so that the
the theory are not simple invariant quantities, results of the calculations provide an unambi­
few in number, but are instead complex, highly guous measure of shear strength at failure. All
variable, and usually not constant with respect to too often, back-analyses are conducted without
time. Frequently the range of variability of the adequate information being available on the mode
parameters remains extraordinarily large even after of failure or on the pore pressures that existed
completion of exploration and testing that costs as
much as the project can possibly afford. "
at the time of failure. It is suggested that
unsatisfactory isolated analyses of this kind can
be harmful to our understanding of slope failure
Experience with the application of soil mechanics mechanisms, since they purport to provide 'evi­
methods of analysis to slope failures in Hong dence' in support of a design method about which

2565
 TABLE X
Assessment of the State-of-Knowledge of the Various Aspects of Slope Stability Predictions
for Hong Kong Conditions (Updated from Hencher et al, l984)

Aspect Current State-of-Knowledge for Hong Kong Conditions o;;r;i;wT:§;;3

Stability factory' Very good

Methods of Janbu (1954, 1973) method of analysis for non-linear surfaces thought satis­
Analy sis Recommended
Computational datafactors of poorly
is often safety of 1.2 to 1.4
handled are satisfactory
(Lumsdaine 5 Tang,(GCO,
1981,1984).
1982).
Pre-failure geometry is easily defined.
Geometry Often difficult to decide critical potential failure surface for design, espe- Good to
of Failure cially where geology is complex (Hencher et al, 1984; Hencher 5 Martin, Very good
1984; Hudson 5 Hencher, 1984).

Site investigation procedures are adequate, but descriptions often poor.

Geology Complex weathering profiles are difficult to describe (Hencher 5 Martin, 1982) Fair
Understanding of influence of geological details on hydrogeology is poor.

Mass strength as distinct from sample strength is poorly understood.

Shear Laboratory tests are commonly used to determine saturated strengths of samples Fair to
Strength in terms of effective
results stress,
(Brand, 1981, 1982). but doubt exists about applicability of test Poor
Limited amount of insitu strength testing carried out (Brand et al, 1983a).
weakening effect of relict joints recognised (Koo, 1982s, 1982b).
Effects of boulder and corestone content unknown (Hencher 5 Martin, 1982).

Useful correlations available between landslides and rainfall (Lumb, 1975;

Brand et al, 1984).
Rapid changes in pore pressure with rainfall are very difficult to predict for
design (Anderson et al, 1983; Premchitt et al, 1985).
Groundwater Only limited attempts made to model groundwater (Leach 5 Herbert, 1982).
and Pore Extrapolation of insitu measurements seems best design approach (Koo 5 Lumb, Poor
Pressures 1981; Endicott, 1982).
Some progress made with field instrumentation (Pope et al, 1982; Brand et al,
l983a).
Erosion pipes are important in transmitting water (Nash 5 Dale, 1983; Brand
et al, 1986).

TABLE XI
Suggested Advances which Can Be Made in the State-of-Knowledge of Slope Stability Predictions
for Hong Kong Conditions (Updated from Hencher et al, l9B4)

Methods
Key Field T@==1“s
Observations ‘ Theoretical
Studies Back
Analysis
Areas Field Laboratory
Shear Advances can be Limited advances Considerable advan- Advances can be This is the only
Strength made by obeerva- can be made in ces can be made in made in establish- means of checking
tion of factors relating sample our understanding of ing models for the validity of rela­
which influence strengths to mass the shear strength mass behaviour of tionships between
the mass strengths strengths through behaviour of un- materials properties
of materials field testing saturated
soils
residual

Groundwater These are useful Case studies re- Not applicable Advances are pos- This is the only
and Pore for checking lating subsurface sible, e.g. deve- means for checking
Pressure theories and me- profiles to hydro lopment of more that the methods
thods of predic­ geology and infil sophisticated in­ of incorporating
tion tration character­ filtration and groundwater into
istics can lead hydrogeological design are satis­
to improved pre­ models factory
dictions

2566
the analyst often has preconceived ideas. It studied as part of a Territory-wide appraisal of
is important to guard against these dangers in­ existing cut slopes carried out between 1979 and
herent in the back-analysis of failures, as em­ 1981. Each slope was the subject of detailed
phasized by Leroueil & Tavernas (1981). There assessment, which included a review of the slope
is no doubt, however, that a large number of history, a detailed site survey, and an appro­
case histories of failures in a particular loca­ priate site investigation and testing programme.
tion provides a good basis for future design, in Holes were drilled to supplement the surface geo
that they can be used as the input to a semi­ logical examinations and to obtain samples, and
empirical method (see Section 7). piezometers were installed to monitor pore pres­
sures throughout one or two wet seasons. Tri­
axial tests were carried out in the laboratory
9. SOM HONG KONG CASES to obtain good quality shear strength data.
9.1 Introduction The three slopes were all analysed using the
Janbu routine method (Janbu, 1954, 1973) applied
In establishing his framework by which to judge to a large number of slip surfaces chosen on the
soil mechanics predictions, Lambe (1973) recog­ basis of the site geology as revealed by the
nised three main categories of prediction, name­ field studies. The minimum calculated factor of
ly: type A predictions (made before the event), safety was in each case significantly lower than
type B predictions (made durin the event), and 1.0. In May 1982, August 1982 and June 1983,
type C predictions (made after the event). For severe rainstorms occurred (Table III) which be­
obvious reasons, type A predictions are thought tween them caused well over 2 000 slope failures
to be the only true predictions, and these are but the three slopes in question did not fail,
certainly the most valuable kind for assessing and they remain stable to this day.
the reliability of our predictive methods.
Type C predictions are regrettably the most Case 1 slope is shown in the simplified cross­
common type reported in the published literature, section in Fig. 40. It has a height of about
and alone they provide unconvincing confirmation 36 m from the flat open space below. Deep exca­
of the accuracy of our predictive tools. How­ vations into the original ground surface left
ever, these 'post-mortem' predictions can be a relatively thin layer of granitic soil (grades
extremely valuable where they are accompanied IV, V & VI) over the granite bedrock. The slope
by full details of the engineering design carried surface is covered partly by vegetation and
out before construction. partly by lime-stabilised soil (‘chunam‘). The
water table, as determined by piezometer instal­
Mention has been made in Section 8.7 of the lations,is low even during the wet season, no
detailed back analyses carried out on some slope positive pore pressures having been measured in
failures in Hong Kong (Hencher et al, 1984). the soil.
These were part of a continuing programme to
evaluate the engineering design and analysis
methods currently in use and specified in the
Geotechnical Manual for Slopes (Geotechnical Sllp surface _ ‘om
Control Qffice, 1984). This continuing evalua­ ` (F=0B5)
tion process is vitally important in Hong Kong,
\\`* or/ _ao
w \`°
where the design factors of safety are necessarily
very low (Table VI). \ “h>`g57q/Pro
"` U17
In order to illustrate the state-of-the-art with
respect to predicting the performance of resi­
\`\ \\\ E i`fZ¥qbc
\\ Residual "~\ - zo
dual soil slopes, details will be given here of
the performance of six selected Hong Kong cut
slopes. These fall into the two categories :
(a) slopes which have remained stable even
though the present theoretical factors of
safety are less than 1.0 (Cases 1, 2 & 3),
and
\\ \-10
Granlte 'rock' \\ so", ( mlm ) `
<1.1un> \
gwgfe
§ EL `
(b) slopes which have failed, even though the
'design' factors of safety were above 1.0 Fig. 40 Case 1: Stable Cut Slope with F <l.0
(Cases 4, 5 & 6).
The slopes represented by Cases 1 to 3 were The measured shear strength parameters for the
formed some years ago on the basis of precedent, material of the Case l slope vary between c' =l4
no engineering design methods having been used. kPa and ¢' =45° for the grade IV material, and
Cases 4 to 6, on the other hand, involved the c' =0 and 6' =42° for the grade VI material. On
latest analysis and design procedures. All six the basis of these parameters, the factor of
cases depict common types of cut slope in Hong safety for the steep upper part of the slope,
Kong, and they together represent a small pro­ above the berm, is calculated to be 0.95, with
portion of the examples that could be used to the critical slip surface being very shallow and
illustrate the same points. passing
The largelycross-section
detailed through the grade
o the V material.
'unstable'
part of the slope (Fig. 41) clearly illustrates
9.2 Stable Slopes with F <l.0 the importance of the weathering profile on
stability predictions for slopes of this kind.
The three stable slopes (Cases 1 to 3) were

2567
"I, Q s""'5§5e=...
~t:i Oéfiia
"II.. O.; 'aaa
*_ik"30m
F=1.0
"ffl '” '=i= ;===.
. . . _ . . , <. ` oranne
(c&2kPa\ F = °~° `aN2_`\`
F=°3
. .. . <~
' =-<3 o' °'= 39') ` Colluvlum \\\ 2°

_. 0 ; 5 '10
* + w ¢'= 3B°)
~'B7::uE5"'¥ (d=0

_ _ _ _m ' ' "5§a=_

_ =» 4 ""5§=. s & Magonry we
-_-_-_~_-_-_-_ _ --_ <>
. . -'wg

_ Berm
-_-_~_~_~_~_'_'_'_ 5
I `-:-:-:-:~Z- 112-1-1-_ _ - - - ~

F..¢\l_" V -,Q ‘:-Y: '_-_~ ~ ~ - ~ ° ' Q °

Fig. 4l Detailed Weathering Profile for the Case l
Slope Showing the Grades IV, V & VI Soil
(HammQ|* ­|I­
5L
"- _

Fig. 43 Case 3: Stable Cut Slope with F <l.0

shallow, having a factor of safety of only 0.80.

Case 2 involves a cut slope in granitic soil
whiEh'was formed as part of a combined site
formation for a housing block and a road (Fig. 9.3 Unstable Slopes with F >l.0
42). The steep cut slope, which is about l8 m
high and coated with chunam, is retained at its Examples will be given of three cut slopes which
toe by a 3 m concrete retaining wall. The very were designed to be adequately stable and which
thick deposit of granitic soil is predominantly subsequently failed during the severe rainstorms
of grade V, with shear strength parameters of of 1982 and 1983 (Table Ill). These will be re­
c' =5 kPa and 6' =35°. The water table remains ferred to as Cases 4, 5 & 6.
low throughout the wet season. The calculated
factor of safety on the deep critical slip sur­ Case 4 cut slope, shown in Fig. 44, was designed
face shown in Fig. 42 is 0.82 and constructed in 1980-Bl as part of a new road
scheme. What was thought to be adequate site
\9 investigation was carried out at the time in the

;_
form of drilling, sampling and strength testing.
4u%,`_`
ROAD \ \ \ 20m This indicated that the granite rock was overlain
by a relatively thin layer of grades IV, V & VI
materials. Piezometers installed in drillholes
, 1-v \\
“$§E$i§Ei$§$? "\ \\
'1°°' . . ._.'11:': ' wa"
showed the water table to be quite close to the
ground surface in the upper part of the slope,
and two rows of horizontal drains were included
at mid-slope as part of the design. On the as­
sumption that the horizontal drains would control
Granhk
F=oa2 ‘° the maximum level of the water table, the criti­
cal slip surface was found to be as shown in the
( C' = 5 k PB Buudhg failure. The measured shear strengths of c' =lO
kPa and 6' =35° yielded a design factor of safety
..._Granite ¢'= 35') l` Ill
.__.._._.____ L
-_...___
QV/_ llll||
of 1.24, which was considered adequate for the
risk category concerned.

Fig. 42 Case 2: Stable Cut Slope with F <l.0

I

|11 //// { // I
4° Horlzontal dralns /’ Q
If

Case 3 involves a 30 m high slope (Fig. 43) cut

( F =124 ) /’ ,/ L_
30 Deslgn sllp aurlace

in granitic soil with a thin covering of collu­

vium over its lower half. At the toe of the
slope is a 5 m high masonry retaining wall, and
above the slope is a road. The surface of the
slope is protected partly by chunam and partly
by vegetation. The granitic soil varies from
grade IV to grade VI, the vast majority being
of grade V with strength parameters of c'= 2 kPa
and 6' =39°. At its highest, the water table is
located many metres below the top of the slope,
10
__,-»'
/
20 ,,," 4 lFallure

oa W/ Granlte
Rd
eurlace
(F= o.a9)
Granltlc (c' = 10 KF3/
but it is close to the surface of the bottom 0' = 35°)
part of the slope. The results of the stability
analyses are depicted in Fig. 43. Several slip
surfaces give factors of safety which are less Fig. 44 Case 4: Unstable Cut Slope Previously
than l.0, the critical surface, which is very Predicted as Being Stable (F= l.24?)

2568
After the heavy rainstorm at the end of May 1982,
failure of a large portion of the Case 4 slope
occurred, as can be seen in the aerial photograph
in Fig. 45. The failed mass measured approxi­
mately 70 m along the slope and 40 m from top to
bottom, the deduced failure surface being as
shown in Fig. 44. It should be noted that, un­
like the vast majority of Hong Kong landslides,
failure did not occur suddenly and dramatically,
som
,f 4 ,
so
CoHuWum _,ng
_ _ F = 1.3 , // ,2§§§f.§§§§5”""
///'IPM /
Desngn sllp F: 1 35 / ;£§2<~;_=-­
surfaces F=1'38 ’ _;f€==
/iff w/
but took place over several days, and the large ,’Eailure surface / Z
mass of failed material remained on the slope.
40 ,/’ ResMuaIsoH
/ (F = 1.3 7)
The failure was extensively investigated, many
boreholes being drilled to obtain a much more
detailed geological profile than that on which
the design was based. Reanalysis of the slope
on the basis of the actual failure surface, which
was much deeper than the design critical surface f
,Z /Granite
,eii
/
a/"¢ 3 lf& YI
(Fig. 44), gave a range of F-values from 0.82 to
1.0 depending on the water table level, with
the most probable value being F= 0.89. The cru­ Fig. 46 Case 5: Unstable Cut Slope Previously
cial design error, therefore, was a misinterpre­ Predicted as Being Stable (F= l.3?)
tation of the weathering profile because of in­
sufficient borehole data.
and ¢'= 40° for the grade V material. On the
basis of this data and the maximum groundwater
elevation recorded during 1981-82, stability
analyses gave a 'design' factor of safety in
excess of 1.3.
A portion of the slope, 30 m high and 15 m long,
failed during the rainstorm of August 1982, and
a further study was then carried out to investi­
gate the cause. The failure surface was accurate
ly determined to be as shown in Fig. 46, for
which the calculated factor of safety was 1.3,
unless considerably higher pore pressures than
those measured were assumed in the calculations.
Of major importance was the presence of almost
vertical slickensided relict joints visible at
the failure scarp, which could have resulted in
the joint-controlled failure mechanism depicted
in Fig. 47. In addition, soil pipes (see Section
8.5) discovered in the vicinity could have trans­
mitted large quantities of water to the slip sur­
face to bring about localised transient pore pres
Fig. 45 Aerial View of Case 4 Slope after Failure sures higher than those measured by the piezome­
ters. It is therefore difficult to envisage how
this slope failure could possibly have been pre­
Case 5 concerns a very large cut slope formed in dicted on the basis of available soil mechanics
the late 1960s for the construction of a road. methods.
At its highest point, the crest of the slope is Case 6 is a 70 m high cut slope in volcanic soil
90 m above the road level. The slope was ori­
ginally cut to a 'standard' angle without the constructed in 1974-77 as part of the Tuen Mun
benefit of investigation or design. A number of Highway, which features in Section 10 below. No
shallow failures occurred at various points on proper investigation or stability analysis was
this long slope during 1973, 1974 and 1976, as carried out before construction. Failure of the
a result of which a limited amount of investiga­ slope first occurred in 1975, and this led to
tion was conducted to enable remedial works to remedial works, which included cutting back the
be designed and carried out. These took the slope and the installation of some horizontal
form of cutting back together with the installa­ drains, to leave the slope as shown in the cross­
tion of some horizontal drains in the failed section in Fig. 48. On the basis of site investi
areas. gation data collected for the remedial works, a
full stability analysis was later carried out in
Figure 46 shows a section of the upper portion 1978, to give F`=l.3 on the 'design' critical
of the Case 5 slope as it was in 1981. Several slip surface shown in the figure. In September
metres of colluvium overlay about 12 m of grades 1983, an extremely large failure, about 10 000
V & V1 granitic soil. A comprehensive site in­ sq.m in area (Fig. 49), occurred on the slope.
vestigation on the slope was carried out in 1981­
82 to reassess its stability. This included the A very detailed geotechnical study was carried
installation of piezometers and a thorough pro­ out on the Case 6 failure. The many boreholes
gramme of laboratory shear strength testing. drilled (Fig. 49) indicated that the bedrock
The strength parameters were measured as c =4 (grades III or stronger) was much deeper over
kPa and ¢' =36° for the colluvium, c' =6 kPa and most of the slope than assumed in the 1978
¢' =40° for the grade V1 material, and c‘= 20 kPa analysis. lnspite of this, borehole data and

2569
 _ } Stress
Vertical } <_
Ulsplacement
along joints

fQQ§®
QD

_ _=---
Af . 1:2-;.
~:;~\ :I_._
_ ~¢;
~A

Horizontal
~|\°‘ 9
GQ/

.~;=a<e ;=a=
'=7*;?#15 iéi,1
f~ff“@.=§e¢r::f-'~
,/
Rotation

e-ii I/
,_ EB? \_.-'
,r

A,
\,f/I\-~­ __o'.o' t

g
»»"_..»_. _ ff.
__ af _
" H

s\\d\f\9 N-,¢;1:,g; =;‘~=2;§~.E gate ipaq; *<-ff-:sex ;;.;.

p;`;9;~/_...-ea'-!<_;`;f`._
__=_`,-
<1:1;;3=~5
_.-_.» . ---
_ \._=-
"- :\\.'~:.'../;;~ ’ 'F'
Shear
\_`_ . .» _-/f
,~a=`__\a-_`
" ;'”":"
.1.»;:<\if;mfE5§i#£:
\
surface
Y ,_\ -,J ~,_
Q&`7"`I<!.€~:\-
\;
s _-_,-~..-."..~
HOflZ0l1\a|

..-. < 1-:= =>:1 =a

f_¢~‘=-:;,;»1:=»-';§.:?f
.~,?P’ 'C-.¢:.23 _1\`=§

_ .1-' l=:§~ .f»i§@§7F;i?1er: e"'s'°"' Z°"'e if

displacement
between slices

- '-- 1-1>-1€*:~=;?=2Er1f
an - :‘:.-:L
C. _`\>`_.._»
-~:;€»

f./T»_\-71-`i"-~`
1€$.'%'-'§f1&€:Ls;=."=~'i:>
-=_._-:§-11:21-*kk=~>7:'5f¥f<?Z~25:t':1'eif5v2"Q;°55ei;€=;I;=;1:= Sas: ­

(b) Ingress of water through Open Joints

Fig. 47 Possible Failure Mechanism of
E
=J ‘I-':ii
-1

,V .'.»:::` - .~-»1.e..~-:___ _~ .. :~" 1 , _ 31; f;'1:~;, :>¢»_¢~5`:_-=_-:ans ; z`:'~

(a) Rotation and Displacement of Joints

Ingress of water

_-,. _
.gif/§:=,:
- - fQ15§'£=i
_._-1'/21; A 5; -‘7~‘;
. . »';:J;T:Sj';§'¥ fri*
§~:f`~&§'<..j:f;
el-`~T!<':/i\<-.~-2-w
._~§»- ,_ g_~ ;fr '
Buildings

4`FAlED'
1* x
wlss d
/P I//.»‘ ."
I/.Q`_
-°
1' ­,
‘\
o‘,-
-9_°0 sorenoles

`\`
if Q 5 00' . \ §
_ “ \
/ .‘;_,- I %__<_ l It 1503 * `
| _. _ -»_s-,> _A _‘l,_>._.-_, I -_

I/-___. l__.o ._ “QM

Fig. 49 Plan Showing the Extent
of the Case 6 Failure

inclinometer readings determined the failure sur­

face as being fairly close to that predicted in
1978 (Fig. 48), although the back scarp was ap­
preciably steeper. This fact appeared to be
related to the subvertical relict joints in the
decomposed volcanic soil which were crucial in
dictating the mode of failure. The computed F­
value for the actual failure surface, however,
was still about 1.3 on the basis of strength
data obtained from good quality laboratory tri­
axial tests. Like Case 5, it is thought that
gradual slope 'deterioration' caused by the in­
gress of water along the relict joints could
have been the main cause of failure. Piezometers
installed for the back analyses gave measured
groundwater levels very close to those assumed in
the original l97B stability calculations.
9.4 Discussion
The six Hong Kong examples described above illus­
Case 5 and Case 6 Slopes trate our apparent inability to accurately pre­
dict the failure of residual soil slopes. On
the one hand, the classical soil mechanics me­
thods of stability analysis would seem to be con­
- 80m servative, as evidenced by the Cases l to 3
slopes which have remained stable with factors
_ Design ellp surface of safety of less than 1.0. On the other hand,

_* ao= , "4
( F=
Z/ W4
1.3
4
Fallure surface ’/
,/
) ,/’ Cases A to 6 suggest that the same methods over­
estimate the F-values for some slopes which are
in fact unstable.

//'
(F 137)
__4°f//// \’ /\ There are two possible contributory factors to
the incorrect predictions that the Cases l to 3

/mg’
_ 2° Volcanic ”, (c'=20kPa Volcanic tuft
slopes should have failed, namely :
(a) the insitu mass shear strengths are higher
than those measured in the laboratory, for
the reasons discussed earlier in Section
8.4, and
_E 5 1 :S
um
O 5) (b) soil suctions are sustained in the slopes
throughout the heaviest rainstorms which
are sufficient to render them stable.
Fig. 48 Case 6: Unstable Cut Slope Previously The usually dominant part played in slope stabi­
Predicted as Being Stable (F= l.3?) lity calculations by positive pore pressures

2570
cannot in any way be the cause of error in these It is noteworthy that none of the three slopes
three cases, since the pore pressures were zero represented by Cases 4, 5 & 6 failed dramatical­
everywhere on the determined critical slip sur­ ly, but all involved slow and relatively small
faces. displacements of slope material which in each
case remained on the slope surface. This suggests
For soil suction alone to be the reason why the that a calculated value of 1.3 might be the lower
three slopes have remained stable (i.e. F >l.O), bound F-value to ensure long-term stability in
suction values greater than about 2 kPa, 16 kPa all Hong Kong's large soil cut slopes.
and B kPa would need to be sustained in the three
slopes respectively throughout severe rainstorms. Of importance is the fact that there are no re­
In order for the slopes to possess factors of corded failures of Hong Kong slopes for which
safety of 1.4, as specified in Hong Kong for high the F-values predicted on the basis of state-of­
risk situations, the soil suctions required are the-art soil mechanics methods have exceeded 1.4.
about 8 kPa, 45 kPa and 28 kPa respectively. As This gives good support to the use of a minimum
mentioned earlier, it is unlikely that suctions factor of safety for high risk situations of 1.4,
of these large magnitudes could exist at such as recommended by the Geotechnical Manual for
shallow depths in these slopes during severe Slopes (Geotechnical Control Office, 1984)
rainstorms. However, it is probable that the (Table VI).
soil never becomes fully saturated insitu, in
contrast to the full saturation state usually
imposed for laboratory shear measurements. 10. PROBABILITY OF SLOPE FAILURE
Where slopes fail with 'design' factors of safety 10.1 Prediction Quality
which exceed 1.0 (Cases 4 to 6), the role of geo­
logical detail appears to be paramount,IOUt1D€
and this The calculated factor of safety for a slope is a
cannot be readily accounted for in our ` way of representing the probability of failure.
methods of slope analysis and design. It is Engineering design aims to produce a low probabi­
not uncommon in weathered profiles for the geo­ lity of failure combined with a high degree of
logical complexity to be such that our soil economy. A perfect prediction method would result
mechanics approach can only be applied SUCCESS' in the failure of all slopes with F <l.0 and no
fully with hindsight. An example of such a failures of slopes with F >l.0. In other words,
0 should
situation is shown in Fig. 50, for which back the probability of failure for F <1 of failure
analyses have been carried out by Hencher & ideally be l.0, and the probability
Martin (1984). for F >l.O should be O. This ideal situation is

‘ae
f /
-E use
T' /
, / \/
sleenly Q3
rsrsnaal,
repll lunlll

/ / lll nl vmlallen
I ` natural || lelilon
dinning /'_”
A .~' »:..
lalngnahuve ,A/10//V.. I / an made \ f
dolerile dyke , \ li, __o_ ,732 a umed lul nl
residual§soll
A
lllnllt .
ml_"\'‘,3_3
2 3
\_ `\/ 'I'
/f\.°`?‘°"“"""‘
_’' ’/1/V”
1/ll’
mlulale
/' ~‘
» _
"""
.' "I
‘ uc
lvvm
. f 'Ii ar
.'
highly and comnlelely [ / ,f ' 'Q' f’ ­
f'°°§‘”§°f f°,"'" I \< , / ” / ` ' llirl\s|;|J rm|;d|i\:l':;:»ip|e|e|y
/ , -"'.»,- gn ex - ll
in e f \ mljnl ical I /Sv- '/ t I n".i»y:§)éP ruoannslgd aglvulh 301
uri inal chuname f \ ' /f , ‘-vu) 1 \ jim, sm , ' / e fealpnes nl xlronge
I "slope
snir eul U /// - ff/ V.
5 /,////%
i , ,/-, 4//.dmnage, l hzoheq
/ , /' /éflllil3.|8130
/ IH]
supwl
dmnau :A
A-
;. / ''Nfl
J hanny/‘ `H, 'I , ,
, 1 Xf»/ gi. ,, Iith-
_, Ish:
_'I lov rllngung
channel 1 'I I,V/, ff/ " "'
_v_| , ," -
, ,i__ f;l ,f.- 1'
doleule dyke
f5
:i|:I/'I//l//1V/
/J,’fslanf
xnned
eh mxéaa.
heneh
I' ilu'
H). l //l _.'
ly, /’
*"u|.| 1/- //1
ne i ~'Zl»,,_,|'
-L* - I'
: nermanenl 'nlel
-Y,-gf"`L~ nmedv
lable I.r _ ' ./‘ I/'
3cl ' 4___ 14 _»|,

Fig. 50 Diagram owing

Sh ` the Com p gical Conditions at the Failure Site Studied by Hencher & Martin (1984)
lex Geolo

2571
depicted in Fig. 5l(a), which shows the probabi­ 10.2 A Hong Kong Example
lity of failure as a function of theoretical
factor of safety. In practice, the ideal pre­ In order to evaluate, in a general way, our
diction is not achievable, but predictions based ability to predict slope failures in residual
on the best slope design methods could be expected soils, the probability of slope failure will be
to yield results represented by the probability examined for the Tuen Mun Highway in Hong Kong.
curves A S B. While prediction A is clearly This major highway, which was constructed over a
superior to prediction B, both are 'balanced' number of years commencing in l97b, runs for
predictions, in the sense that each gives a pro­ about 15 km along the southwest coast of the
bability of failure of 0.5 for F= 1.0. mainland portion of the Territory. A large num­
ber of cut slopes were required to be formed in
steep natural hillsides, and this was done on
the basis of the standard 'design' geometry of
LO 5 vertical to 3 horizontal (about 60°), with 1.5
m wide berms at 7.5 m vertical intervals. Pro­
\ . prediction
Med vision was made for any slope to be flattened if
it proved to be unstable during construction.
Acwm \\\ The majority of the Tuen Mun Highway slopes are

enor , \ \
0.5 "'edktbn A Achml cut in weathered granite, but a few at the west­
prediction B ern end are in weathered volcanic rocks. Most
of the slopes have exposed rock (grades I to III)
at the toe with residual soil (grades IV to VI)

°1
Predictive above. Of the rest, a few are cut entirely in
rock, while the others are composed of residual
soil throughout. While the slopes can therefore
0 05 L0 FACTOR OF SAFETY
"‘|'
1 .5 2.0
be expected to vary considerably in their stabi­
lity, they have all been in existence for the
same length of time and they have all experienced
the same rainfall history.
The Territory-wide slope survey carried out in
(a) Ideal and Actual Balanced Predictions 1977-78 recorded 78 cut slopes higher than 5 m

\\ \\ \
along the Tuen Mun Highway. Fourteen of these
1.0 were selected for detailed geotechnical studies,
which included full stability analyses. The
distribution of the calculated factors of safety
Unsafe
for these is shown in Fig. 52. Half of the
slopes had factors of safety greater than 1.4,
Bahnced\ \ \ prediction while the others had F-values which varied from
1.1 to 1.4. No stability analyses were carried
0.5 Prediction
\ out on the other GA slopes along the Highway,
but the crude assumption will be made here that
\ \\
\ \\\\ \
°11
the distribution of F-values for these slopes
_ Conservative would be similar to the distribution for the
prediction fourteen shown in Fig. 52.
The Tuen Mun Highway slopes have endured six
major rainstorm events since they were constructed
0 0.5 1.0 1.5 2.0 (Table III). The rainstorms of 1982 and 1983 were
particularly severe tests of their stability.
Four slopes failed in the storm of May 1982 and
FACTOR OF SAFETY another (Case 6, Section 9.3) in September 1983,
1 b) Conservative and Unsafe Predictions
4
Fig. Sl Reliability of Slope Stability Predictions w
in terms of Probability of Failure and m
1
Theoretical Factor of Safety O3
_I
w
‘5 2
m

3E 1
53

:3:3:? ¥:3:3:H3152
z w§H§I§§§§§ 513: i3:3:
AEE?
° 011.0
F 11.2"iii 71 1 12.0
1.4 1.6 1.3 1
FACTOR OF SAFETY

Fig. 52 Distribution of Factors of Safety for the

l4 Tuen Mun Highway Slopes Studied in Detail

2572
which all caused disruption to traffic for many ll. CONCLUSIONS
days. During the highway's first seven years
of life, therefore, five major failures occurred It has been shown that there are several basic
in the 78 slopes, which gives a rough annual approaches that can be used for the design and
probability of failure of about l in l00 for stability assessment of residual soil slopes.
each slope. For most residual soil problems, the traditional
division between the methods developed separate­
The rough annual probability of slope failure on ly for soil mechanics, rock mechanics and engin­
the Tuen Mun Highway can be compared with the eering geology is not meaningful, and a unified
probabilities quoted by Whitman (1984) for other approach is necessary for good engineering design
types of geotechnical failure. Figure 53 shows Because of the nature of residual materials, the
Whitman's plot of probability of failure against application of soil mechanics methods alone will
consequence of failure for the situation in the generally not be adequate. The selected use of
United States. To this has been added a zone to terrain evaluation has much to recom end it,
represent the group of Tuen Mun Highway slopes. especially for planning and feasibility studies
Whitman's acceptability criteria, as represented of relatively large land areas. In addition,
by straight lines in the figure, suggest that the establishment of semi-empirical design guide­
the Tuen Mun Highway situation would represent lines will often be invaluable where there is
an acceptable risk in the USA and would be adequate information available on the performance
little different from the risk accepted for
foundation failure. of existing cut slopes over a sufficiently long
period of time.
The state-of-the-art with respect to the appli­
m 10° cation of soil mechanics methods to the analysis
g MARGNALLY ACCEPTED of residual soil slopes is generally poor.
Whereas some simple situations readily lend
f 10-1 Ph Merchantsmppmg themselves to solution by analytical means, the
1
Q ='°P°$ __ _` Mobile
10-2 '#,,,“~ drill rigs
often complex geological conditions which exist
in residual soil slopes usually preclude the
E Gawmb .~l>
m 10'° F _d tl ®
sensible application of the well-known methods
of stability analysis. Major difficulties inva­
riably arise with the selection of a suitable
< °"" B °"= Fixed arm nga 40 mode of failure, with the meaningful measurement
of shear strengths and, particularly, with the
3g HIGHWAY
10'4 Tuen MUN C5/°r5D
Z awahon
E 10“° l |
75, 10'° Commercial
UVES LOST 1 10 100 1000
prediction of the most unfavourable pore pres­
sures which are likely to occur during the life
of the slope.
There is considerable scope for improvements to
the application of analytical methods for the
prediction of slope performance. Long-term pro­
COST,US$ 1m 10m 100m 1,000m grammes of research and development are required
CONSEQUENCEOFFAlURE into the shear strength behaviour of residual
soils and into improved methods for predicting
the pore pressures induced by rainfall. Most
Fig. 53 Annual Probabilities and Consequences of Geo­ importantly, good case histories are badly needed
technical Failures in the USA (Whitman, 1984) which give detailed comparisons of stability
Compared with Those on the Tuen Mun Highway analyses and slope performance. There is no
doubt that the available analytical methods have
far out-striped our ability to observe, measure
The Geotechnical Control Office has established and interpret data. As Peck (1975, 1980) has
a comprehensive catalogue of approximately commented, analytical methods of design are
lO O00 large cut slopes in Hong Kong, the majo­ usually too heavily relied upon by the engineer
rity of which were constructed before engineer­ at the expense of the application of engineering
ing design methods were adopted. If the proba­ judgement based on sound experience.
bility of failure of l in lO0 for each Tuen Mun
Highway slope is representative of the slope
stability situation overall in Hong Kong, then
approximately lOO slope failures can be expected ACKNOWLEDGEMENTS
in an average year. This figure is in fact not
far in excess of the average number of signifi­ Acknowledgement is due to many of the Author's
cant failures recorded annually over the last colleagues in the Geotechnical Control Office
20-year period (Table III). The situation is who have collectively contributed significantly
improving dramatically, however, because of the to the progress made in the past few years in
high standard of geotechnical control which is geotechnical engineering practice in residual
now applied to the design and construction of soils. He is particularly grateful to Dr Jerasak
all new cut slopes, which are required to comply Premchitt, who gave generous assistance with the
with the Geotechnical Manual for Slopes (Geotech­ preparation of this Paper, and to Dr S.R. Hencher
nical Control Office, l98h). In time, it is who supervised many of the slope studies describ­
envisaged that the overall probability of failure ed. As usual, Miss Margaret Wong produced a
will tend towards a reductign bymuch
an order high quality manuscript in a very short time.
magnitude (from lO'2 to 10' ), in the of
same
way as Peck (1980) hoped for a reduction from This Paper is published with the permission of
lO'4 to lO'5 in the annual probability of failure the Director of Engineering Development of the
for the world‘s dams. Hong Kong Government.

2573
REFERENCES landslide hazard and risk mapping. Proc.
4th Int. S . Landslides, Toronto, (1),
ABMS (1985a). Proceedings of the First Inter­
national Conference on Geomechanics in Tro­ Brand, E.W. (1981). Some thoughts on rain-induced
pical Lateritic and Saprolitic Soils, slope failures. Proc. 10th Int. Conf. SPEE,
Brasilia, 4 vols, ABMS, Brazilian S5c. Soil Stockholm, (3), 373-376.
Mech., Sao Paulo, 1758 p,
ABMS (l985b). Peculiarities of Geotechnical Brand, E.W. (1982). Analysis and design in resi­
Behaviour of Tropical Lateritic and Sapro­ dual soils. Proc. ASCE Spec. Conf. Engin­
litic Soils. (ISSMTE Technical Committee eerin andSoils,
Residual Construction in Trppical
HonolE1u, 89- 3. an
on Tropical Soils, Progress Report 1982­
l985). ABMS, Brazilian Soc. Soil Mech., Brand, E.W. (1984). Landslides in Southeast Asia;
Sao Paulo, 470 p. a state-of-the-art report. Proc. 4th Int.
Allen, P.M. & Stephens, E.A. (1971). Report on S _ Landslides, Toronto, (1), 17-59.
(A en um, ( ), 105-106).
the Geological
Kong Surveyp.of&
Govt Press, Hgng Kong. Hong
maps. Brand, E.W. (1985a). Geotechnical engineering in
Anderson, M.G. (1984). Prediction of Soil Suc­ tropical residual soils. Proc. lst Int.
Conf. Geomechanics in Tropical Lateritic
tion for Slopes
Control Office, in Hong
Hong Kong.
Kong, 42Geotech.
p. and Saprolitic Soils, Brasilia, (3), 23799.
Anderson, M.G. & Burt, T.P. (1977). Automatic Brand, E.W. (l985b). Landslides in Hong Kong.
monitoring of soil mositure conditions in Proc. 8th S.E. Asian Geotech. Conf., Kuala
a hillslope spur and hollow. Jour. Hydrol., Lumpur, (2), in press.
(33). 27-36. Brand, E.W., Borrie, G.W. & Shen, J.M. (1983a).
Anderson, M.G., McNicho1l, D.P. & Shen, J.M. Field measurements in Hong Kong residual
(1983). On the effect of topography in soils. Proc. Int. Sym . Field Measurements
controlling soil water conditions, with in Geomechanics, Zurich, (1), 639-648.
specific regard to cut slope piezometric Brand, E.W., Burnett, A.D. & Styles, K.A. (l982a).
levels. Hong Kong Engr, (11), ll, 35-41. The Geotechnical Area Studies Programme in
ASCE (1971). Session ll: Slope Stability in Hong Kong. Proc. 7th S.E. Asian Geotech.
Residual Soils. Proc. 4th Panam. Conf. Conf., Hong Kong, (1), 107-123.
SMFE, San Juan, Puerto Rico, (2), 185-167, Brand, E.W., Dale, M.J. & Nash, J.M. (1986).
and (3), 125-146. Soil pipes and slope stability in Hong Kong.
ASCE (1982). Proceedings of the ASCE Specialty Qtrly Jour. Eng. Geol., in press.
Brand, E.W. & Hudson, R.R. (1982). CHASE - An
Conference on Engineering and Construction empirical approach to the design of cut
%26 Tropical
p. an Residual Soils, Hono1Elu, slopes in Hong Kong soils. Pggc. 7th S.E.
Azzouz, A.S., Baligh, M.M. & Ladd, C.C. (1981). Asian Geotech. Conf., Hong Kong. (I), I-IE.
Three dimensional stability analyses of (Discussion, (2), 61-72 and 77-79).
four embankment failures. Proc. 10th Int. Brand, E.W. & Phillipson, H.B. (1984). Site in­
Conf. SMFE, Stockholm, (3), - . vestigation and geotechnical engineering
Baligh, M.M. & Azzouz, A.S. (1975). End effects practice
97-153.
in Hong Kong. Geotech. Eng., (15),
on stability of cohesive slopes. Jour.
Geotech. Eng. Div., ASCE, (101), 1155-1117. Brand, E.W. & Phillipson, H.B. (Eds.)(l985).
Barata, F.E. (1969). Landslides in the tropical Sampling and Testing of Residual Soils - A
region of Rio de Janeiro. Proc. 7th Int. Review of International Practice. Scorpion
Conf. SMTE, Mexico City, (2), 587-516. Press, Hong Kong, 194 p.
Beattie, A.A. & Attewill, L.J.S. (1977). A land­ Brand, E.W., Phillipson, H.B., Borrie, G.W. &
slide study in the Hong Kong residual soils. Clover, A.W. (l9S3b). ln-situ direct shear
Proc. 5th S.E. Asian Conf. Soil Eng;, Bang­ tests on Hong Kong residual soils. Proc.
kok, 177-188. Int. Symp. Soil and Rock lnvesti ations by
In-situ Testin , Paris, (2), 13-17. (Dis­
Beattie, A.A. & Chau, E.P.Y. (1976). The assess­ cussion, , 5-56).
ment of landslide potential with recommen­ Brand, E.W., Premchitt, J. & Phillipson, H.B.
dations for future research. Hon Kon (1984). Relationship between rainfall and
En r, (4), 1, 27-44. (Discussion, (4), 2, landslides in Hong Kong. Proc. 4th Int.
55-62).
Symp. Landslides, Toronto, (1), 377-384.
Blight, G.E. (1977). Slopes and excavations in Brand, E.W., Styles, K.A. & Burnett, A.D. (1982).
residual soils. Proc. 9th Int. Conf. SMTE, Geotechnical land-use maps for planning in
Tokyo, (3), 582-5 0. Hong Kong. Proc. 4th Int. Con _ IAEG, New
Blong, R.J. (1973). Morphometric attributes of Delhi, (1), 5-153.
landslides. Zeitschrift fur Geomorphologie, British Standards Institution (1981). Code of
Suppl. Vol. 18, 66-77. Practice for Site Investigation BS 5935:
Boonsinsuk, P. & Yong, R.N. (1982). Some as­ 1981. Brit. Stan. Inst., London, 150 p.
pects of slope stability in Hong Kong resi­
dual soils. Proc. ASCE Spec. Conf. Engineer­ Burnett, A.D., Brand, E.W. & Styles, K.A. (1985).
Terrain evaluation mapping for a landslide
ingiand
Soi Construction
s, Hono1E1u, in Tropical and Residual
463-483. inventory in Hong Kong. Proc. 4th Int. Conf.
& Field Workshop on Landslides, Tokyo, 63­
Brabb, E.E. (1984). Innovative approaches to 68.

2574
Burnett, A.D. & Styles, K.A. (1982). An approach Kong, 5-43.
to urban engineering geolo ical mapping as
used in Hong Kong. Proc. Eth Int. Con . Deere, D.U. & Patton, F.D. (1971). Slope stabi­
IAEG, New Delhi, (1), - . lity in residual soils. Proc. 4th Panam.
Conf. SMFE, San Juan, Puerto Rico, (1),
Burt, T.P. (1978). An automatic fluid-scanning
switch tensiometer system. British Geomor­ Endicott, L.J. (1982). Analysis of piezometer
pholog.
-30. Res. Group., Tech. Bu11. No. 21, data and rainfall records to determine
Campbell, R.H. (1975). Soil slips, debris flows
and rainstorms in Santa Monica mountains
(10), 9, 53-56. "'
groundwater conditions. Hong Kong Engr,

and vicinity, Southern California. US Geol. Eyles, R.J. (1979). Slip-triggering rainfalls
Survey, Prof. Paper 851, 55 p. in Wellington City, New Zealand. New Zea­
land Jour. Sc., (22), 117-121.
Carrara, A. (1983). Multivariate models for Eyles, R.J., Crozier, M.J. & Wheeler, R.H. (1978).
landslide hazard evaluation. Math. Geol., Landslips in Wellington City. New Zealand
(15), 403-426. Geographer, (34), 2, 58-74.
Carrara, A., Catalano, E., Sorriso Valvo, M., Finlayson, A.A. (1984). Land surface evaluation
Reali, C. & Ossi, I. (1978). Digital ter­ for engineering practice: applications of
rain analysis for land evaluation. Geolo­ the Australian PUCE System for terrain eva­
gia
& 3gpplicata
rgs. e Idrogeologia, (13), 69-I27 lggtion. Qtrly Jour. Eng. Geol., (17), 149­
Carrara, A., Catalano, E., Sorriso Valvo, M., Fredlund, D.G. (1931). The shear strength of
Reali, C., Merendi, L. & Rizzo, V. (1977). unsaturated soils and its relationship to
Landslides morphometry and topology in two slope stability problems in Hong Kong.
, - 3.
§2ne§,1Ca1abria, Italy. Bull. IAEG, No. Hong Kong Engr, (9), 4, 37-45.
Ching, R.K.H. & Fredlund, D.G. (1984). Quanti­ Fredlund, D.G. (1984). Analytical methods for
tative comparison of limit equilibrium slope stability analysis. Proc. 4th Int.
methods of slices. Proc. 4th Int. S . Symp. Landslides, Toronto, , - .
Landslides, Toronto,_T2), 7 -37 . Fredlund, D.G., Krahn, J. & Pufahl, D.E. (1981).
Coates, D.R. (1977). Landslides perspectives. The relationship between limit equilibrium
Landslides, ed. D.R. Coates. Geol. Soc. slope stability methods. Proc. 10th Int.
Conf. SMTE, Stockholm, (3)T_409- .
America, Reviews in Eng. Geol., (3), 3-28.
Crozier, M.J. (1969). Earthflow occurrence Fredlund, D.G. & Rahardjo, H. (1985). Theoretical
during high intensity rainfall in eastern context for understanding unsaturated resi­
dual soil behaviour. Proc. lst Int. Conf.
Otago. Eng. Geol., (3), 325-334. Geomechanics in Tropical Lateritic and
Crozier, M.J. (1973). Techniques for the mor­ Saprolitic Soils, Brasilia, (1), 295-306.
phometric analysis of landslips. Zeits­ Fukuoka, M. (1980). Landslides associated with
chrift fur Geomorphologie, (17), 78-IUI. rainfall. Geotech. Eng., (ll), 1-29.
Crozier, M.J. & Eyles, R.J. (1980). Assessing Geological Society of London (1977a). The log­
the probability of rapid mass movement. ging of cores for engineering purposes.
Proc. 3rd Aus.-NZ Conf. Geomechanics, (Second Rep. by Geol. Soc. ing. Group Work­
We1lington, (2), 47-51.
Da Costa Nunes, A.J. (1969). Landslides in soils ing Party). Qtrly Jour. Eng. Geol., (10),
5-52.
of decomposed rock due to intense rainstorms. Geological Society of London (1977b). The des­
Proc. 7th Int. Conf. SMTE, Mexico City, (2), cription of rock masses for engineering
5 7-55 . purposes. (Rep. by Geol. Soc. Eng. Group
D'Appolonia, E.R., Alperstein, R. & D'Appolonia, Working Party). Qtrly Jour. Eng. Geol.,
D.J. (1966). Behavior of a colluvial (10), 355-388.
slope. Proc. Conf. Stability and Perfor­ Geotechnical Control Office (1982). Mid-levels
mance of S10 es and Embankments, Berkeley,
Cal., 489-515. (Reprinted Jour. Soil Mech. Soil Pro
Study: erties.
Report Geotec .Hydrology
on Geologyi Contro Office,
and
Found. Div. ASCE, (93), SM4, 447-493, 1976). Hong ong, p. & 54 drgs.
Dearman, W.R. (1974). Weathering classification Geotechnical Control Office (1984). Geotechnical
in the characterization of rock for engin­ Manual for Slo es (2nd ed.). Geotech. Con­
eering purposes in British practice. Bull. tro O ice, Hong Kong, 295 p.
IAEG, No. 9. 33-42.
Dearman, W.R., Baynes, F.J. & Irfan, T.Y. (1976).
Gidigasu, M.D. (1975). Laterite Soil Engineering.
Elsevier, Amsterdam, 570 p.
Engineering properties of weathered granite. Government of Hong Kong (l972a). Interim Re ort
Eng. Geol., (12), 345-374. of the Commission of Inquiry into t e Rain­
De Graft-Johnson, J.W.S. & Bhatia, H.S. (1969). storm Disasters I972. Hong Kong Govt
Engineering characteristics (of lateritic Printer, p.
soils). Proc.
perties Spec. Sess.
of Lateritic Engineering
Soils Pro­
7t Int. Conf. Government of Hong Kong (1972b). Final Re ort of
SMTE, Mexico City, (2), 13-43. the Commission of Inquiry into t e Ra nstorm
Disasters 1972. Hong Kong Govt Printer, 91
De Mello, V.F.B. (1972). Thoughts on soil en­ p.
gineering applicable to residual soils. Grant, K. (1975a). The PUCE programme for terrain
Proc. 3rd S.E. Asian Conf. Soil Eng;, Hong

2575
 evaluation for engineering purposes: I. Embankment Dam En ineerin (Casa rande
Principles. CSIRO Australia, Div. App. Geo­ Vo ume , e . . . irsc e . . oulos,
47-107. Wiley, New York.
mech., Paper No. 11, 209 p.
Grant, K. (1975b). The PUCE programme for ter­ Jones, F.O. (1973). Landslides of Rio de Janeiro
rain evaluation for engineering purposes: and the Serra das Araras escarpment, Brazil.
II. Procedures for terrain classification. US Geol. Survey Prof. Paper 697, 48 p. & 2
CSIRO Australia Div. App. Geomech. Paper maps.
No. 19, 68 p. & 2 drgs. Jones J.A.A. (1981). The Nature of Soil Pi in
Grant, K. & Finlayson, A.A. (1978). The assess­ £_A Review of Research. Geo Books, NorwicE,
ment and evaluation of geotechnical resour­ UK, p.
ces in urban and regional environments. Koo, Y.C. (1982a). Relict joints in completely
Eng. Geol., (12), 219-293. decomposed Volcanics in Hong Kong. Canad.
Grant-Taylor, T.L. (1964). Stable angles in Geotech. Jour., (19), 117-123.
Wellington greywacke. New Zealand Engng, Koo, Y.C. (1982b). The mass strength of jointed
April 15, 129-130. residual soils. Canad. Geotech. Jour., (19),
Greenway, D.R., Anderson, M.G. & Brian-Boys, 225-231.
K.C. (1984). Influence of vegetation on Koo,Y.C. 8 Lumb, P. (1981). Correlation between
slope stability in Hong Kong. Proc. 4th rainfall intensities and groundwater levels
Int. S . Landslides, Toronto, (I), 399­ in the natural slopes of Hong Kong. Hong
4.
Kong Engr, (9), 10, 49-52.
Guidicini, G. & lwasa, O.Y. (1977). Tentative Lambe, T.W. (1973). Predictions in soil engineer­
correlation between rainfall and landslides ing. Géotechnigue, (23), 149-202.
in a humid tropical environment. Bull.
IAEG, No. 16, 13-20. Lambe, T.W. (1975). PRIME: Predict - instrument
Hencher, S.R. & Martin, R.P. (1982). The des­ - measure - evaluate. Proc._4§h S.E. Asian
cription and classification of weathered Conf. Soil Eng., Kuala Lumpur, G.1- G.6.
rocks in Hong Kong for engineering purposes. Lane, K.S. (1961). Field slope charts for stabi­
Proc. 7th S.E. Asian Geotech. Conf., Hong lity studies. Proc. 5th Int. Conf. SMFE,
Kong, (1), f25-f42. Paris, (2), 651- .
Hencher, S.R. & Martin, R.P. (1984). The failure Leach, B. 8 Herbert, R. (1982). The genesis of
of a cut slope on the Tuen Mun Road in Hong a numerical model for the study of the
Kong. Proc. Int. Conf. Case Histories in hydrogeology of a steep hillside in Hong
Geotechnical En ineerin , St Louis, Missouri Kong. Qtrly Jour. Eng. Geol., (15), 243­
( ), 6 3-6 . 259.
Hencher, S.R., Massey, J.B. & Brand, E.W. (1984). Leroueil, S. & Tavanas, F. (1981). Pitfalls
Application of back analysis to some Hong of back analyses. Proc. 10th Int;_Conf.
Kong landslides. Proc. 4th Int. S . SMFE, Stockholm, (1 , - _ (Discussion,
Landslides, Toronto, ( ). - 3 . (4), 603-604).
Hudson, R.R. & Hencher, S.R. (1984). The delayed Leshchinsky, D., Baker, R. & Silver, M.L. (1985).
failure of a large cutting in Hong Kong. Three dimensional analysis of slope stability
Proc. Int. Conf. Case Histories in Geotech­ Int. Jour. Numeric. & Analyt. Methods in
nical En ineerin , St Louis, Missouri, (2), Geomech., (9), 199-223.
7-. Little, A.L. (1967). Laterites. Proc. 3rd Asian
Ho, D.Y.F. E Fredlund, D.G. (1982). Increase in Reg. Conf. SMFE, Haifa, Israel, (2), 61-71.
strength due to suction for two Hong Kong Little, A.L. (1969). The engineering classifica­
soils. Proc. ASCE Spec. Conf. Engineering tion of residual tropical soils. Proc. Spec.
and Construction in Tropical and Residual Sess. Lateritic Soils 7th Int. Conff SMFE,
Soils, Honolulu, 263-295. Mexico City, (1), 1710.
Hunt, T. (1982). Slope failures in colluvium Lovell, C.W. (1984). Three dimensional analysis
overlying weak residual soils in Hong Kong. of landslides. Proc. 4th Int. S` . Land­
Proc. ASCE Spec.
Construction Conf. Engineering
in Tropica and
an Resi uaf_Soi1s, slides, Toronto, ( ), - 55.
Honolulu, 443-462. Lumb P. (1962a). The properties of decomposed
Huntley, S.L. & Randall, P.A. (1981). Recogni­ granite. Géotechnigue, (12), 226-243 (& 2
tion of colluvium in Hong Kong. Hong Kong plates).
Engr, (9), 12, 13-18. Lumb P. (l962b). Effect of rain storms on
IAEG (1979). Classification of rocks and soils slope stability. Proc. Symp. Hong Kong
for engineering geological mapping. Part Soils, Hong Kong, 73'B7.
I: rock and soil materials. Bull. IAEG, Lumb P. (1965). The residual soils of Hong
No. 19, 364-371 Kong. Géotechni ue, (15), 180-194. (Dis­
IAEG (1981). Report of the Commission on Site cussion, (16), 1 , 78-81 and 359-360).
Investigation. Bull. IAEG, No. 24, 185-226. Lumb P. (1972). Landslides in Hong Kong. Proc.
Janbu, N. (1954). Application of composite slip lst Int. Symp. Landslide Control, Kyoto,
91-93.
surfaces for stability analysis. Proc. Euro
Conf. Stability of Earth Slopes, Stoc holm, Lumb, P. (1975). Slope failures in Hong Kong.
13). 43-49. Qtrly Jour. Eng. Geol., (8), 31-65.
Janbu, N. (1973). Slope stability computations. Lumb P. (1979). Statistics of natural disasters

2576
 in Hong Kong, 1884-1976. Proc. 3rd Int. Page, W.D. & James, M.E. (1981). Landslides and
Conf. Applications of Statistics & Probabi­ mudflows, major contributors to the land­
lity in Soil & Structural Engineering, scape in Northwest Colombia. Revista CIAF,
Sydney, (1), 9-22. (6), 455-467.
Lumb, P. (1983). Engineering properties of fresh Patton, F.D. & Hendron, A.J. (1974). General
and decomposed igneous rocks from Hong Kong. report on "Mass Movements". Proc. 2nd Int.
Eng. Geol., (19), 81-94. Cong. IAEG, Sao Paulo, (2), V-GR.1 -V.GR.57.
Lumsdaine, R.W. & Tang, K.Y. (1981). On stabi­ Peck, R.B. (1969). Advantages and limitations
lity analysis in Hong Kong. Proc. 10th lnt. of the observational method in applied soil
Conf. SMFE, Stockholm, (4), 9 - . mechanics. Géotechnigue, (19), 171-187.
Lumsdaine, R.W. & Tang, K.Y. (1982). A compari­ Peck, R.B. (1975). General introduction to the
son of slope stability computations. Proc. session "Landslides and Their Prevention".
Zth S.E. Asian Geotech. Conf., Hong Kong, Proc. 4th Guel h Symp. Geomor holo , Guelph
(1), 31-38. Ontario, 133-136. P gl
Martin, R.P. & Hencher, S.R. (1984). Principles Peck, R.B. (1980). "Where has all the judgement
for description and classification of wea­ gone?". The fifth Lauritz Bjerrum memorial
thered rocks for engineering purposes. lecture. Canad. Geotech. Jour., (17), 584­
Proc. 20th Conf. Eng. Group Geol. Soc. 590. (Reprinted, Norweg. Geotech. Inst.
London, Guilford, 304-308. Ppp. No. 134, 1-5).
McMahon, B.1. (1976). Estimation of upper bounds Pierson, T.C. (1983). Soil pipes and slope stabi
to rock slopes by analysis of existing slope lity. Qtrly Jour. Eng. Geol., (16), 1-ll.
data. CANMET, Canada Centre Min. & Energy Pope, R.G., Weeks, R.C. & Chipp, P.N. (1982).
Tech., Rep. 76-14, 65 p. Automatic recording of standpipe piezometers
Moh, Z.C. (Ed.)(l969). Proceedings of the Spe­ Proc. 7th S.E. Asian Geotech. Conf., Hong
cialty Session on En ineering Properties of Kong, (1), 77789.
Lateritic Soils 7thgInt. Conf. SMFE, Mexico Porcher, M. E Guillope, P. (1979). Cartographic
City, 2 vols., 419 p. des risques ZERMOS appliquée a des plans
Morgenstern, N.R. & de Matos, M. (1975). Stabi­ d'occupation des sols en Normandie. Bull.
lity of slopes in residual soils. Proc. Liais. Lab. Ponts Chaus., No. 99, 43-54.
5th Panam. Conf. SMFE, Buenos Aires, (3), Premchitt, J., Brand, E.W. & Phillipson, H.B.
(1985). Landslides caused by rapid ground­
Morgenstern, N.R. & Price, V.E. (1965). The water changes. Proc. 21st Conf. En . Grou
analysis of the stability of general slip Geol. Soc. London, S e ie , - _
surfaces. Géotechnigue, (15), 79-83. Rodin, S., Henkel, D.J. & Brown, R.L. (1982).
Moye, D.G. (1955). Engineering geology for the Geotechnical study of a large hillside area
Snowy Mountains Scheme. Jour. Inst. Engrs in Hong Kong. Hong Kong Engr, (10), 5, 37­
Australia, (27), 287-298. 45.
Nash, J.M. & Dale, M.J. (1983). Geology and Rossetti, R. & Ottone, C. (1979). Esame pre­
hydrology of natural tunnel erosion in liminaire delle condizioni pluviometriche
superficial deposits in Hong Kong. Proc. dell'Oltrepo Pavese e dei valori critici
Conf. Geology of Surficial Deposits in Hong delle precipitazioni in relazione ai feno­
Kong, Hong Kong, 61-72. meni di dissesto franoso (Preliminary analy­
Neuland, H. (1976). A prediction of landslips. sis of rainfall in Oltrepb Pavese and the
critical rainfall values as related to
Catena, (3), 215-230. landslide failures). Geolo ia A licata e
Nilsen, T.H. & Brabb, E.E. (1977). Slope stabi­ Idrogeologia, (14), 3, - .
lity studies in the San Francisco Bay region, Ruxton, B.P. (1960). The geology of Hong Kong.
California. Geol. Soc. America, Reviews in trl Jour. En . Geol., (115), 233-260
Eng. Geol., (3), 235-243. (& p ates & map).
Nilsen, T., Taylor, F.A. & Brabb, E.E. (1976). Ruxton, B.P. & Berry, L. (1957). Weathering of
Recent landslides in Alameda County, Cali­ granite and associated erosional features
fornia (1940-71): An estimate of economic in Hong Kong. Bull. Geol. Soc. America,
losses and correlations with slope, rain­ (68). 1263-1291 (& T-Eiate).
fall and ancient landslide deposits. QS Sarma, S.K. (1973). Stability analysis of em­
Geol. Survey, Bull. 1398, 24 p. & l drg.
bankments and slopes. Géotechnigue, (23),
Nilsen, T.H., Wright, R.H., Vlasic, T. & Spangle, 423-433.
W.E. (1979). Relative slope stability and
land-use planning in the San Francisco Bay Sarma, S.K. (1979). Stability analysis of em­
region, California. US Geol. Survey, Prof. bankments and slopes. Jour. Geotech. En .
Paper 944, 104 p. & 3 drgs. Div. ASCE, (105), 1511- .
Onodera, T., Yoshinaka, R. E Kazama, H. (1974). Shuk, T. (1965). Discussion on "Some aspects of
Slope failures caused by heavy rainfall in the safety factor in soil mechanics" by A.
Japan. Proc. 2nd Int. Con _ IAEG, Sao Proc. 6th Int. Conf. SMFE, Mon­
Langejan.
Paulo, ( ), v.11. -v.I1. _ treal, (3), -.
Oyagi, N. (1984). Landslides in weathered rocks Shuk, T. (1968). Unmetodo sencillo de diseno
and residual soils in Japan and surrounding para minimizar el costo de taludes en roca
areas. Proc. 4th Int. Symp. Landslides, (A simple design method to minimize the
Toronto, (3), l-31. costs of slopes in rock). Proc. lst

2577
 Columbian S . En . Geol., Bogota, (2), Terzaghi, K. (1950). Mechanism of landslides.
7 p. (Eng is trans ation reprinted under Application of Geolo to En ineerin Prac­
Appendix B to "Estimation of upper bounds tice (Berke Volume), E3-123; Geo1TgSoc.
to rock slopes by analysis of existing America, Bou er, Col.
slope data" by B.K. McMahon. CANMET Canad. Vail, A.J. (1984). Two landslide disasters in
gintre Min. & Energy Tech., Rep; - , ­ Hong Kong. Proc. 4th Int. Svgp. Landslides,
Toronto, (1), 717-722.
Skempton, A.W. & Hutchinson, J. (1969). Stabi­ Vargas, M. (1953). Some engineering properties
lity of natural slopes and embankment foun­ of residual clay soils occurring in southern
dations. Proc. 7th Int. Conf. SM E, Mexico Brazil. Proc. 3rd Int. Conf. SMFE, Zurich,
City, State-of-the-Art vol., 291-340. (1), 67-7 .
So, C.L. (1971). Mass movements associated with Vargas, M. (1967). Design and construction of
the rainstorm of June 1966 in Hong Kong. large cuttings in residual soils. Proc.
-6 . Inst. British Geographers, No. 53,
gpang. 3rd Panam Conf. SMFE, Caracas, (2), 243-254.
Sowers, G.F. (1963). Engineering properties of Vargas, M. & Pichler, E. (1957). Residual soil
residual soils derived from igneous and and rock slides in Santos (Brazil). Proc.
metamorphic rocks. Proc. 2nd Panam. Conf. 4th Int. Conf. SMTE, London, (2), 394-398.
SMTE, Brazil, (1), 3 - . Varnes, D.J. (1985). Landslide Hazard Zonation:
Sowers, G.F. (1971). Landslides in weathered Review of Principles and Practice. UNESCU,
Volcanics in Puerto Rico. Proc. 4th Panam. Paris.
Conf. SMTE, San Juan, Puerto Rico, (2), Varnes, D.J. & Keaton, J.R. (1984). Trends in
engineering geological and related mapping
St John, B.J., Sowers, G.F. & Weaver, Ch. E. 1972-1983. Bull. Assoc. Eng. Geologists,
(1969). Slickensides in residual soils (21), 255-267.
and their engineering significance. Proc. Vaughan, P.R. (1985a). Mechanical and hydraulic
7th Int. Conf. SM E, Mexico City, (2), properties of in-situ residual soils, parti­
591-597. cularly as related to their structure and
Styles, K.A., Hansen, A., Dale, M.J. & Burnett, mineral components. Proc. lst Int. Conf.
A.D. (1984). Terrain classification methods Geomechanics in Tropical Lateritic and
for development planning and geotechnical Saprolitic Soils, Brasilia, (3), 231-263.
appraisal: a Hong Kong case. Proc. 4th Int. Vaughan, P.R. (l985b). Pore pressures due to
Symp. Landslides, Toronto, (2), 561-568. infiltration into partially saturated
Sweeney, D.J. (1982). Some in situ soil suction slopes. Proc. lst Int. Conf. Geomechanics
measurements in Hong Kong's residual soil in Tro ical Lateritic and Saprolitic Soils,
slopes. Proc. 7th S.E. Asian Geotech. Conf. Brasilga, (2), 61-71.
Hong Kong, (1), 91-106. (Discussion, (2), Vaughan, P.R. G Kwan, C.W. (1984). Weathering
93-96). structure and in situ stress in residual
Sweeney, D.J. & Robertson, P.K. (1982). Slope soils. Géotechnigue, (34), 43-59.
stability in residual soils in Hong Kong. Whitman, R.V. (1984). Evaluating calculated
Canad. Geotech. Jour., (19), 521-525. risk in geotechnical engineering (7th Ter­
Taylor, D.K., Hawley, J.G. & Riddolls, B.W. zaghi Lecture). Jour. Geotech. Eng., ASCE,
(1977). Slope stability in Urban Develop; (110), 145-158.
ment. New Zealand GeoméEh. Soc., Welling­ Young, S. (1972). Slopes. Longman, London,
ton, 71 p. 295 p.

2578
 General reports on recent failures
Flapports généraux sur des ruptures récentes