ARTICLE IN PRESS

Journal of Environmental Management ] (]]]]) ]]]–]]]

www.elsevier.com/locate/jenvman

Assessing slope stability in unplanned settlements

in developing countries
Malcolm G. Anderson, Liz Holcombe, Jean-Philippe Renaud
School of Geographical Sciences, University of Bristol, University Road, Bristol BS8 1SS, UK
Received 30 November 2005; received in revised form 3 May 2006; accepted 15 August 2006

Abstract

Unplanned housing in developing countries is often located on steep slopes. Frequently no building code is enforced for such housing
and mains water is provided with no drainage provision. Both of these factors can be particularly significant in terms of landslide risk if,
as is so often the case, such slopes lack any planned drainage provision. There is thus a need to develop a model that facilitates the
assessment of slope stability in an holistic context, incorporating a wide range of factors (including surface cover, soil water topographic
convergence, slope loading and point source water leakage) in order that appropriate advice can be given as to the general controls on
slope stability in such circumstances. This paper outlines a model configured for this specific purpose and describes an application to a
site in St. Lucia, West Indies, where there is active slope movement in an unplanned housing development on relatively steep topography.
The model findings are in accord with the nature of the current failure at the site, provide guidance as to the significance of slope drainage
and correspond to inferences drawn from an application of resistance envelope methods to the site. In being able to scenario test a
uniquely wide range of combinations of factors, the model structure is shown to be highly valuable in assessing dominant slope stability
process controls in such complex environments.
r 2006 Elsevier Ltd. All rights reserved.

Keywords: Slope stability; Unplanned housing; Developing countries

1. Introduction slopes and so intrinsically there is already a high risk of

The increasing level of construction activity (both
domestic and infrastructural) in developing countries is
giving rise to a growing frequency of small- and medium-
scale disasters related to patterns of human environmental
intervention, indicating that disaster risk is accumulating
These disasters may have a larger cumulative impact than
the spectacular but occasional large catastrophes (Tear
Fund, 2004).
Unplanned housing development built on steep slopes
frequently lacks any form of adequate slope drainage and
hence makes steeper slopes particularly vulnerable to
landslides. The problem, typical of those faced by many
of the most vulnerable communities, is illustrated in Fig. 1.
Of course, the majority of unplanned or squatter type
settlements are to be found on just such steep, marginal,
Corresponding author. Tel.: +44 117 9288682; fax: +44 117 9287878.
E-mail address: m.g.anderson@bris.ac.uk (M.G. Anderson).
slope instability. Such communities are also among the
most vulnerable in terms of social needs and housing
conditions. A relatively recent major landslide which
occurred in an area of unplanned housing in Castries, St.
Lucia, West Indies involved the relocation of numerous
residents and a cost to Government of US$8 million
funded in the form of a loan from the regional banking
system (Anthony, 2001).
Two key contributory factors here are that the Car-
ibbean region is on the Atlantic hurricane track and the
relative strength of the local institutions means that
building controls and planning processes are not always
as robust as they might be. In many developing countries,
such as those in the Caribbean, an appropriate way to
approach landslide risk reduction is through a programme
of preventative measures developed in partnership with
local communities and cross ministry initiatives specifically
designed to build capacity within vulnerable communities
(Anderson and Holcombe, 2004a).

0301-4797/$ - see front matter r 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jenvman.2006.08.005

Please cite this article as: Malcolm G. Anderson et al., Assessing slope stability in unplanned settlements in developing countries, Journal of
Environmental Management (2006), doi:10.1016/j.jenvman.2006.08.005
 ARTICLE IN PRESS
2 M.G. Anderson et al. / Journal of Environmental Management ] (]]]]) ]]]–]]]
Fig. 1. Site of major landslide in unplanned housing community Skate
Town, Castries, St. Lucia, West Indies. Note collapsed retaining wall at
the toe of the slide and potential undermining of house foundations at
uppermost extent of landslide.
For such an environment this paper seeks to
 Illustrate the nature and range of slope stability control
factors that need to be taken into account.
 Demonstrate that an appropriate hydrology-slope sta-
bility model can be designed and suitably parameterised
for the determination of the dominant instability process
controls.
 Undertake an application of the software to a currently
unstable site, to ascertain the appropriateness of the
scheme and to establish whether model outcomes are
deemed reliable and consistent with other approaches to
stability control identification.
2. An holistic view of landslide hazard
There are a number of factors that require integration in
order to develop an appropriate approach to landslide risk
in unplanned housing areas. Such communities typically
have mains water supply, but no planned drainage from
properties. Residents may well recognise the risks, but lack
either appropriate design or building skills to implement
structures such as drainage ditches or retaining walls. More
complex issues such as topographic convergence—leading
to high pore water pressures and high instability risk
areas—may not be evident to either planners or residents.
Vegetation on slopes is sometimes grown for food as a
Please cite this article as: Malcolm G. Anderson et al., Assessing slope stability in unplanned settlements in developing countries, Journal of
Environmental Management (2006), doi:10.1016/j.jenvman.2006.08.005
matter of necessity without consideration being given to
the impact on slope stability. Fig. 1 shows a landslide on a
551 slope, threatening unplanned housing in the immediate
vicinity. Here topographic convergence and poor drainage
from properties upslope resulted in locally high pore
pressures and the overturning of the retaining wall.
To communicate landslide risk reduction in this setting,
a model needs to be configured that incorporates water
supply to the slope from properties at specific slope
locations, vegetation surface cover, and allows for topo-
graphic convergence/divergence effects and the weight of
buildings.
Generalised schemes which address bioengineering have
been developed in the past (Greenwood et al., 2004).
However, these schemes do not have parsimonious
parameter requirements combined with the inclusion of
the wide range of controls required for the assessment
needed here (since they lack fully coupled dynamic
hydrology and the holistic inclusion of key topographic
elements and other factors). Therefore, for the research
reported here, we use an integrated hydrology and stability
model (Anderson et al., 1996), CHASM, which was
extended for this specific problem in order to provide the
following specific capabilities:
 point sources of water to represent leakage from
properties,
 influence of surface cover on rainfall interception and
soil cohesion,
 influence of hillslope topography,
 surface loading by buildings,
 unsaturated/saturated soil water modelling.
The outline of the scheme is detailed in the following
section. Importantly, all of these components can be driven
from data that is available from sources that are readily
available within the context of this application; the model
structure is thus not over-specified and likely to infer false
resolution (Anderson and Bates, 2001).
3. Modelling instability risk
3.1. Model description
To make an assessment of landslide risk with a view to
developing appropriate intervention methods, a software
approach that incorporates the dominant elements of the
controlling instability processes is required. The CHASM
software is a physically based combined soil hydrology and
slope stability model that comprises fully integrated
hydrology, surface cover (vegetation) and stability compo-
nents as illustrated in Fig. 2. We have described certain of
the main features of the model elsewhere (Anderson et al.,
1997; Bishop, 1955; Wilkinson et al., 2000; Premchitt et al.,
1986; Wu and Sidle, 1995; Janbu, 1977), but it is appropriate
to provide an outline description here, together with the
extensions implemented in the code for this application.
 ARTICLE IN PRESS
M.G. Anderson et al. / Journal of Environmental Management ] (]]]]) ]]]–]]]
Fig. 2. Outline of the core processes incorporated in CHASM.
3.2. Modelling soil water conditions
The procedure adopted to model the hydrological system
is a finite difference scheme in which the slope is divided
into a series of rectangular columns, each subdivided into
regular cells. The model simulates detention storage,
infiltration, evapo-transpiration, and unsaturated and
saturated flow regimes. Rainfall is allowed to infiltrate
the top cells governed by the infiltration capacity. Within
CHASM, infiltration during rainfall is calculated using
Darcy’s law and vertical flow in the unsaturated zone is
computed using Richards’ Eq. (1) solved in explicit form.
 
qy q qc
¼ K 1 , (1)
qt qz qz
where c is the pore pressure head (m), K the soil
conductivity (m s1), y the moisture content (dimension-
less), z the elevation and t the time. To account for
unsaturated conditions, the soil conductivity K is adjusted
according to the pore pressure head using the Millington
and Quirk (1959) procedure. Below the water table, the
flow between columns is modelled using the Darcy law, i.e.
the gradient of the water table is used to determine the
quantity of water moving from one column to the next
between time steps.
The numerical stability of the solution to the Richards
equation is dependent on the time step. Hornberger et al.
(1985) provide a method of deriving the required time step
Please cite this article as: Malcolm G. Anderson et al., Assessing slope stability in unplanned settlements in developing countries, Journal of
Environmental Management (2006), doi:10.1016/j.jenvman.2006.08.005
3
to maintain numerical stability based on the distance
between the computational nodes, the gradient of the
suction–moisture curve at a given suction value and the
flow velocity. For most applications the commonly
recommended spatial and temporal resolutions are an
individual cell size of 1  1 m and 60 s time step.
3.3. Modelling three-dimensional topography
Within the integrated model structure (Fig. 2), the
hydrology scheme represents slope plan curvature (con-
vexity and concavity) by varying the column breadth. The
effect of the three-dimensional topography on water fluxes
can thus be investigated (in a suitably approximate
manner) and its impact on stability estimated. This
approach requires identification of the zero flux boundaries
at either side of the slope—the degree of convergence or
divergence of which defines the downslope change in
column breadth (Geotechnical Control Office, 1980). The
generated pore pressure field is then used as input to
standard two-dimensional stability analyses where the slip
surface is located within the mid-plane of the three-
dimensional structure (Fig. 2).
3.4. Slope stability analysis
At each hour of the simulation, the predicted hydro-
logical conditions are exported to a limit equilibrium
 ARTICLE IN PRESS
4 M.G. Anderson et al. / Journal of Environmental Management ] (]]]]) ]]]–]]]
method for determining slope stability. CHASM uses
Bishop’s (1955) simplified circular method for estimation
of the slope’s factor of safety (FOS) with an automated
search procedure (Wilkinson et al., 2000). Pore pressures,
both negative and positive, are incorporated directly into
the effective stress determination of the Mohr–Coulomb
equation for soil shear strength. This allows derivation of
the minimum FOS, with temporal variations arising from
hydrodynamic responses and changes in the position of the
critical slip surface (Wilkinson, 2001).
3.5. Modelling surface cover
The numerical scheme includes a surface cover model,
which allows comprehensive investigation into the hydro-
logical and geotechnical effects of vegetation on slope
stability. Vegetation affects slope stability through the
following principal mechanisms; interception, evapo-tran-
spiration, change in hydraulic conductivity, root reinforce-
ment and surface loading. All of these process elements are
represented in the model.
The procedure adopted in CHASM to represent inter-
ception depends on the type of vegetation. For example,
dense stands of tall grass can be flattened by intense rainfall
to form a semi-permeable barrier. This process, the ‘thatch
effect’, is incorporated into CHASM by reducing the
effective rainfall (the amount of rainfall allowed to
infiltrate), for the grid columns containing vegetation
according to:
Re ¼ sv R,
where R is the rainfall rate, Re the effective rainfall
available for infiltration and sv a shedding factor which
depends on rainfall intensity. Premchitt et al. (1986) and
Lamb and Premchitt (1990) provide graphical results from
both grass and tree covered slopes to facilitate an
estimation of the shedding factor.
The impact of vegetation on soil hydraulic conductivity
as a result of the presence of a root network is modelled in
CHASM by increasing the soil conductivity according to
an equation relating the root area ratio to the saturated
hydraulic conductivity
DK ¼ 2  106 þ 3  106 Rv ,
1
where DK is the increase of conductivity in m s due to the
root network and Rv is the root area ratio. Collison (1993)
undertook a major field programme to establish this
relationship for residual volcanic soils; soils of the same
type as those at the Trou Rouge site.
Mechanically, vegetation affects slope stability through
the process of root reinforcement and vegetation sur-
charge. The interaction between roots and soil can be
quantified using a simple root model (Wu and Sidle, 1995).
The vegetation components discussed above are incorpo-
rated directly into the limit equilibrium method for
deriving the FOS.
Please cite this article as: Malcolm G. Anderson et al., Assessing slope stability in unplanned settlements in developing countries, Journal of
Environmental Management (2006), doi:10.1016/j.jenvman.2006.08.005
Following Wilkinson et al. (1998) and Wu et al., 1979, cv
is calculated with the following formula:
cv ¼ 1:2T v Rv , (4)
where Tv is the root tensile strength and Rv is the root area
ratio.
3.6. Modelling point sources
On slopes of the type shown in Fig. 1, mains metered
water is supplied to the many unplanned houses. Grey
water from sinks and bathrooms is often discharged
directly to the slope. Foul water drainage goes to a septic
tank or pit latrine usually within a few metres of the
property, the outflow from which returns flow to the slope.
It is possible within CHASM to assign leakage at defined
points on the slope surface, with specified flux rates by
increasing the effective rainfall to the grid columns where
water leakage into the slope has been identified.
3.7. Modelling structure loading
On slopes typified by those shown in Fig. 1, housing
density can approach 70%. The loading of the slope due to
the weight of buildings therefore needs to be taken into
account when establishing the comparative influences on
slope stability. This is done by increasing the weight of the
slices used by Bishop’s method of slices if buildings are
located on the slice.
(2) 3.8. Accounting for earthquake triggering of landslides
CHASM can estimate the Peak Ground Acceleration
(PGA) required to reduce a given factor of safety to 1 using
the following empirical equation established by Charalam-
bous (2003):
FOS  1
PGA ¼ , (5)
0:94 þ 0:84= tan a
where PGA is the peak ground acceleration (m s2)
required to reduce the FOS to a value of 1 and a is the
slope angle (degrees). This is only an approximate method
but nonetheless provides an additional method of assessing
the performance and appropriateness of the model for
(3)
comparison with detailed tabulations of regional probabil-
istic peak ground acceleration.
4. Model limitations
In accounting for topographic divergence and conver-
gence, surface cover effects and metered water being
returned to the slope from households, CHASM represents
an acceptable platform to determine the dominant
influences on slope stability in vulnerable areas of
unplanned housing. However, inherent within the model
structure and approach are limitations that relate to
process representation and numerical implementation. It
 ARTICLE IN PRESS
M.G. Anderson et al. / Journal of Environmental Management ] (]]]]) ]]]–]]] 5

is of vital importance, as for any software model, to material overlying weathered andesite (grade 5). The slope
understand the model constraints and limitations for the is currently experiencing instability. Fig. 3 shows a cross-
purpose of parameterisation and directing the scope and section through a currently active slide area which has an
nature of the intended application. approximate housing density of 25%. The houses in the
Within CHASM, the limitations relate to the assump- unplanned developments on such slopes have metered
tions made within the hydrology scheme and the method of mains water supply, but no drainage provision. There is no
stability analysis. For the hydrology model the simplifying drainage provision of any significance on the slope. Fig. 4
assumptions primarily relate to flow representation. Soils shows conditions at the upslope and downslope areas of
exhibiting strong anisotropy in hydraulic conductivity or the sliding mass. A slope area immediately adjacent to the
those dominated by macropore flow are not capable of Trou Rouge community was the subject of a major slide in
being modelled by CHASM. For the stability assessment October 1999 which necessitated the relocation of families
component the use of limit equilibrium methods in from approximately 40 properties (Teams, 1999). The very
CHASM results in the imposition of the failure mechan- significant relocation costs of US$ 8 million add urgency to
ism. Thus, stress strain and progressive failure mechanisms the provision of a comprehensive model to evaluate the
cannot be accommodated. In addition the real kinematics potential in such environments for inexpensive remedial
of slope failure, being three-dimensional, are simplified by action to avoid the very significant financial and social costs
consideration of only 2 dimensions. of relocation.
Notwithstanding these issues, the integrated model aims
to improve on the present widely applied limit equilibrium
5.2. Model parameterisation
methods of analysis in defining the functional stability of a
slope by providing an holistic approach driven by
Appropriate model parameterisation is important if
acceptable and robust parameterisation needs. In this
dominant stability influences are to be identified. Central
context the limitations outlined above are not considered
to this question is the reliability of strength testing to
restrictive, but should be noted in order that the process
deliver geotechnical material properties. Whilst developing
mechanics in any application of the model are not ‘over
counties may have certain equipment, it may not be
described’. A further final issue is that with development of
maintained to appropriate standards and there is the
integrated models of the type outlined here, there is an
possibility of operating methods not being formally
increasing awareness that modellers fail to declare models
adhered to. However, it is possible to calibrate equipment
‘invalid’ largely because there are so many variables
with results reported from well documented testing
capable of adjustment to limited field data. In particular,
laboratories.
a critical general issue is one of distinguishing a good fit of
In this study, we undertook a set of direct shear tests
an over-parameterised model from a good fit based on
using Leighton Buzzard sand in order that the results could
accurate process description (Anderson and Bates, 2001). It
be compared with those obtained by Lings and Dietz
is therefore important to seek confirmation of the inferred
(2004) under differing levels of compaction. The tests
process controls with other related evidence. It should
undertaken by Lings and Dietz allowed the performance of
always be borne in mind that whilst such results may be
the direct shear apparatus (DSA) to be optimised by
encouraging, they are not definitive (Anderson and Bates,
exploring different modifications of the shear box. The
2001). In the application that follows we seek such process
modified direct shear apparatus (MDSA), using two states
corroboration of CHASM from seismic trigger estimation,
of compaction of Leighton Buzzard sand, was taken as the
resistance envelope methods of estimating the role of pore
benchmark against which the direct shear apparatus and
pressures in controlling stability, and of landslide runout
estimation methods. By seeking such process confirmation,
and not simply determining apparent agreement between
FOS and observed instability, greater weight can poten-
tially be accorded to the robustness of the model and the
inferences drawn (Anderson and Bates, 2001).

5. St. Lucia case study

5.1. Site description

Surrounding the city of Castries, St. Lucia West Indies is

a range of steep hills many of which have been the subject
of both unplanned housing and significant instability both
in the past and currently. One such community within the
Castries basin area is Trou Rouge. This community is
located on slopes that range from 251 to 601 on grade 6 Fig. 3. Cross-section of Trou Rouge site, Castries, St. Lucia, West Indies.

Please cite this article as: Malcolm G. Anderson et al., Assessing slope stability in unplanned settlements in developing countries, Journal of
Environmental Management (2006), doi:10.1016/j.jenvman.2006.08.005
 ARTICLE IN PRESS
6 M.G. Anderson et al. / Journal of Environmental Management ] (]]]]) ]]]–]]]
Fig. 4. (a) Failure towards top of landslide area caused the collapse of
concrete house structure (see Fig. 3). (b) Failure towards toe of landslide
area causing severe structural damage (see Fig. 3).
testing procedure employed in St. Lucia were compared.
The results of this inter-comparison revealed angles of
internal friction for loose samples of 291 (St. Lucia) and 311
(MDSA), whilst for dense samples the corresponding
results were 401 and 421. Within the limits of testing it
was concluded that on the basis of this inter-comparison
there was no evidence to refute the direct shear results
obtained in St. Lucia.
Such comparative testing is important in the context of
the interpretation of dominant controls on instability as we
shall report below. Assembling strength data from a variety
of local sources in St. Lucia, with the majority of
determinations having been made on the direct shear box
for which the comparative analysis was undertaken, thus
has reasonable validity. The results of such a compilation
show material strength properties for different material
grades (Anderson and Holcombe, 2004b).
Typical water consumption figures for unplanned hous-
ing are available and suggest consumption of approxi-
Please cite this article as: Malcolm G. Anderson et al., Assessing slope stability in unplanned settlements in developing countries, Journal of
Environmental Management (2006), doi:10.1016/j.jenvman.2006.08.005
mately 15 000 l per month per property. Using leakage rates
determined by the local water company (WASCO) we can
further assume leakage from the supply at about 20% and
for the sake of analysis, that all water supplied is returned
to the slope. Water supplied to the slope can equate to a
significant percentage of the rainfall rate. For example,
assuming a housing density of 70% and an average rainfall
of 3000 mm, then the water supply yields an equivalent rate
of 6.9  108 m s1 compared to a rainfall rate of
9.5  108 m s1, i.e. the rainfall rate is effectively increased
by 173%.
Slope loading can, in certain cases, contribute signifi-
cantly to the factor of safety. For unplanned housing the
form of construction can vary significantly from concrete
structures to all wood housing. British Standards (1964)
provide a basis for estimating the complete weight of a
building, from whatever construction materials are used.
The Organisation of East Caribbean States (OECS) has
provided a refinement of that document insofar as it relates
to the use of local building materials in the Caribbean.
Using BS 648 and the local building code refinements
(Commonwealth of Dominica, 1996) individual building
weights can be calculated to estimate total slope loading.
A model of the form outlined here can be relatively
demanding in terms of total parameterisation needs; a
balance has to be achieved between data availability and
model algorithm efficiency to avoid issues of false
resolution in terms of model output. The model we believe
can be supported by data drivers that are in this sense ‘fit
for purpose’ and which have consistency that will allow
instability process dominance to be identified. For the
purposes of the trial application reported in this paper, the
model was parameterised as defined in Table 1. Fig. 5
shows the main dialogue window for a sample application.
Note that the slope section shows marginal downslope
topographic convergence in plan and provides an indica-
tion of where, for this particular simulation, vegetation in
the form of uncropped vetiver grass was included. The
primary initial objective is to ascertain whether the model
structure, incorporating all the site conditions specified in
Table 1, predicts instability; indicates an instability zone
that accords with the currently active landslide area; and
suggests the process to be driven by individual storm events
which currently are observed to trigger movement.
6. Simulations
In vulnerable unplanned housing areas, housing density
plays a potentially important role in the context of
hydrological impact (through metered water supply and
the likely absence of drainage). It is appropriate to design a
set of simulations which reflect housing density variation
and allow for the presence or absence of drains (though the
removal of roof water), whilst allowing the inclusion of
‘leakage’ to reflect what amounts to the metered water
supply into the slope. This is a realistic scenario since the
cost of connecting a complete, unplanned, housing
 ARTICLE IN PRESS
M.G. Anderson et al. / Journal of Environmental Management ] (]]]]) ]]]–]]] 7

Table 1
Model parameterisation of the slope site (see Figs. 3 and 4) used for the simulation scenario tests

Variable Parametisation Source

Rainfall 1 in 5 years, 24 h storm
Saturated Hydraulic conductivity Soil 1:5  106 ms1 Soil
2:1  106 ms1
Geotechnical properties Soil 1:c ¼ 12 kPa; + ¼ 241
Soil 2:c ¼ 18 kPa; + ¼ 301
Vegetation cover Uncropped vetiver
Property loading British Standard BS 648
Point source ‘leakage’ WASCO metered supply
3-D Topography Topographic convergence
shown from Land Registry
Topographic Plans
Moisture retention curves Derived curves and
Millington-Quirk method
Seismicity Generalised empirical
method
Landslide runout tan b (slope angle), D
(depth to slip surface) and
H (height of landslide)
required
Boundary conditions
Water and Sewerage Company of St. Lucia [WASCO] Dam
data (Klohn-Crippen, 1995) provides recurrence interval
rainfall data. 1 in 5 years used in this paper for all simulations
Modal value for grade 6 and grade 5 material as determined
from the compilation of all test results in St. Lucia (Anderson
and Holcombe, 2004b)
Modal value for grade 6 and grade 5 material as determined
from the compilation of all test results in St. Lucia (Anderson
and Holcombe, 2004b). In agreement with F values in Gray
and Rogers (1999)
Hong Kong Special projects Division report SPR 5/86 and
Lamb and Premchitt (1990) report
Concrete structures surveyed on site. Loading determined from
British Standard, augmented by Commonwealth of Dominica
Building code in relation to local materials
House locations determined from site survey. Assuming 3500
gallons per month as determined by WASCO for communities
of this type and a further 20% supply leakage
Survey indicated small downslope topographic convergence;
represented in CHASM (Fig. 6)
Moisture retention curves used for grade 6 and 5 residual
materials Hong Kong (Anderson, 1983). Unsaturated
hydraulic conductivity estimation undertaken by the
Millington and Quirk method (1959)
Peak Ground Acceleration for FOS ¼ 1 determined with the
Charalambous (2003) method and exceedence probability
estimated by Aspinall et al. (1994) tabulated data.
b and H obtained from specification of slope geometry
configured in the model (Fig. 5) and D from critical slip surface
output (Fig. 6). Equation given by Finlay et al. (1999)
Upslope Inflow flux of 5 m s1
Water table at downslope location fixed at 1 m above slope
base
Initial groundwater estimated from field conditions and Gray
and Rogers (1999)

community on a steep slope vulnerable slope to a complete
drainage system for both household waste water and
sewage and roof runoff is prohibitive. Estimation of slope
loading from existing structures on the slope can be readily
undertaken using British Standard 648. Given that
drainage provision is a key practical option, then it is
reasonable to develop a scenario suite that makes provision
for differing initial soil surface pore water pressures.
It should be stressed that the purposes of the simulations
undertaken in this paper are threefold:
 to establish whether the model accords with the given
known conditions of instability at the site,
 to illustrate the relative prioritisation that can be
afforded to modifications that may improve stability
given the range of factors included in the model,
 to seek confirmation of any apparent model agreement
by the use additional alternative approaches (resistance
envelope method).
Specifically, the novel model structure affords the
opportunity to rank hydrological factors, construction
practice, and surface cover in order of potential remedia-
tion impact.
Table 2 shows the core set of results obtained. It is
noteworthy that the model predicts a factor of safety of
0.96 for the current conditions, namely concrete housing
construction, vetiver grass vegetation, no surface drainage,
point source metered water supply to the slope, and 25%
housing density. This result is in accord with the fact that
the slope is currently active in instability terms, and
continues to move following heavy precipitation. The
removal of surface water generated from roof water
sources and grey water waste from kitchens configured
from the metered supply (representing the installation of
surface drains) yields an increased factor of safety (1.00),
increasing to 1.12 if drier initial soil surface conditions
(4 m) are assumed. By way of further (but more minor)
confirmation of the appropriateness of the predicted

Please cite this article as: Malcolm G. Anderson et al., Assessing slope stability in unplanned settlements in developing countries, Journal of
Environmental Management (2006), doi:10.1016/j.jenvman.2006.08.005
 ARTICLE IN PRESS
8 M.G. Anderson et al. / Journal of Environmental Management ] (]]]]) ]]]–]]]

Fig. 5. Main dialogue window for CHASM software illustrating surface vegetation cover and three-dimensional topographic representation ‘switched on’.

Table 2
Factors of safety for different slope conditions

Factors of safety

Housing density 75% Housing density 25% Housing density

Pore pressure at the soil surface 2 m 4 m 2 m 4 m
Vegetation Veg included No veg Veg included No veg Veg included No veg Veg included No veg
With drains installed 1.02 1.02 1.12 1.11 1.00 0.99 1.12 1.11
With no drains installed 0.93 0.93 1.11 1.11 0.96a 0.95 1.11 1.10
a
Existing slope condition.

Factor of Safety (indicating failure at the Trou Rouge site)
it is to be noted that the precipitation used in the model
simulations (Table 2) is above the empirically determined
rainfall threshold for landslide triggering determined for
similar material in Jamaica (Ahmad, 2003).
There is a wide range of compensatory processes that can
be introduced in situations capable of being modelled by
the scheme we have outlined. For the particular slope
modelled here we can draw the following generalised
conclusions from the model scenarios:
Factors leading to an improvement in the factor of
safety:
Drainage: Achieving full drainage of the slope through
the provision of surface drains is important (factor of
safety change from 0.96 to 1.12).
Initial conditions: Achieving reasonable levels of soil
surface initial conditions (4 m) is important and to some
degree of course is a manifestation of surface drainage
provision.
Factors of no apparent relevance in improving the factor
of safety:
Housing density: Higher housing density with surface
drainage implemented makes no effective difference to the
otherwise improved factor of safety. Similarly, a reduction
in slope loading (both lighter materials and lower density)
results in no improvement in the factor of safety.
Vegetation: Changing the vegetation cover from un-
cropped vetiver to either vegetation removal or enhance-
ment to shrubs makes no appreciable difference to the
factor of safety for any of the combinations shown in
Table 1.
The apparent significance of slope drainage can be
corroborated using resistance envelopes to identify the
controls on slope stability (Janbu, 1977; Kenny, 1967;
Fredlund, 1980). The principle of being able to determine
the average unit shear strength required for equilibrium
using a resistance envelope has been clearly established and
is well documented (Anderson et al., 1997). The determina-
tion of the resistance envelope is based on the use of the
friction circle method (Bishop, 1955) with numerous
circular failure analyses undertaken on an idealised slope,
in which the average normal stress is calculated for an

Please cite this article as: Malcolm G. Anderson et al., Assessing slope stability in unplanned settlements in developing countries, Journal of
Environmental Management (2006), doi:10.1016/j.jenvman.2006.08.005
 ARTICLE IN PRESS
M.G. Anderson et al. / Journal of Environmental Management ] (]]]]) ]]]–]]] 9

equivalent system. If negative pore pressures can be

assumed to act directly in effective stress terms, resistance
envelopes can be used to provide an indication of the
average negative pore pressure required for the main-
tenance of stability (Anderson et al., 1997) or conversely
that the slope may be expected to be stable under saturated
conditions. Resistance envelope techniques can thus be
used to corroborate the findings from stability analysis in
the context of the attribution of the dominant factor in
promoting slope instability where this is perhaps consid-
ered to be related to rainfall infiltration and resultant pore
pressure changes. Application of the method to the Trou
Rouge site shows that for two values of the key
geotechnical properties, the slope must be maintained at
either marginal negative pore pressure (Fig. 7a; for normal
loads in excess of 50 kPa, the resistance envelope shows
marginally greater shear strength is required for stability
than can be mobilised by the slope material as indicated by
the laboratory shear strength values used) or very low
positive pressures (Fig. 7b) to ensure stability. It is to be
inferred that significant rainstorm events will, through lack
of drainage provision on the slope, increase pore pressures
beyond those limits, thus confirming that instability can
reasonably be attributed to infiltration controls, in
accordance with the inferred process dominance of the
results shown in Table 2.
7. Earthquake triggering of landslides
The model can estimate the Peak Ground Acceleration
(PGA) required to reduce a given factor of safety to 1 using
a method by Charalambous (2003). In the Caribbean
region detailed PGA estimates associated with probability
of exceedance are available (Aspinall et al., 1994). The
results in Table 1 indicate a PGA of 0.06 g associated with
a predicted FOS of 1.12. Regional exceedance probability
data for PGA shows a PGA of 0.06 to be associated with
an annual probability of exceedance of approximately 0.08
(i.e. less than once in 10 years). This predicted frequency of
occurrence of FOS reduced to 1 would seem consistent
with the known occurrence of earthquakes and associated
landslides in the region.
The nature of this analysis, as with that of the resistance
envelope method, is indicative in broad measure of the
consistency of the model and provides further support for
the robustness of the approach as defined.
8. Landslide runout
Landslide travel distance is of importance in a number of
applications relating to slope design and maintenance. This
is a complex area of research requiring detailed rheology
and material parameterisation that is rarely if ever
available. Nonetheless it is highly relevant to attempt to
include some estimation methodology within the model
reported here, and which is parsimonious in accord with
the other parameter inclusions outlined above, to provide
Fig. 6. Illustration of the failure surface predicted for the Trou Rouge site.
Note the accordance with the site cross-section (Fig. 3).
indication of likely landslide impact as far as runout
distance is concerned. Finlay et al. (1999) used data from
the Geotechnical Engineering Office in Hong Kong to
carry out an empirical analysis of recorded landslides. In
that study 1100 landslides were included, out of a possible
3000, which had occurred between 1984 and 1993 in slopes
of weathered granite and other volcanics. Most of these
landslides had volumes of less than 1000 m3 and debris ran
out onto a near horizontal surface below the slope.
Multiple regression models based on slope geometry were
developed for the prediction of landslide travel across
horizontal surfaces at the base of slopes. This simplified,
but empirically extensive approach was incorporated
within the model structure. Fig. 6 shows the results for
Trou Rouge site indicate a runout of some 3 m; again this
accords with the observed site behaviour, in that in the
vicinity of the house at the toe of the slip (Fig. 4b), material
has been displaced some 3–5 m (Figs. 3 and 8).
9. Conclusions
A model which includes a wide range of factors relevant
to the modelling of slope stability in areas of unplanned
housing on steep tropical and subtropical slopes has been
outlined. The coupled stability—dynamic hydrology
scheme facilitates the modelling of the effects of surface
cover, convergent/divergent topography, surface loading
and point source of water discharge to the slope. As such
the scheme is comprehensive for the specific application
reported here and is parameterised from data sources that
are both robust and readily available. The model, in being
applied to the Trou Rouge site in St. Lucia has demon-
strated four important attributes:
 The model results for the site apparently agree with the
observed conditions in respect of the broad extent of

Please cite this article as: Malcolm G. Anderson et al., Assessing slope stability in unplanned settlements in developing countries, Journal of
Environmental Management (2006), doi:10.1016/j.jenvman.2006.08.005
 ARTICLE IN PRESS
10 M.G. Anderson et al. / Journal of Environmental Management ] (]]]]) ]]]–]]]

instability, the triggering mechanism and the low

positive pore pressures required for instability (Figs. 3
and 6). We should note that the purpose of this
application was not to undertake a detailed back
analysis of a particular event. Rather it was to use this
site, typical of many at risk slopes, to ascertain more
broadly the appropriateness of such a modelling scheme
for assisting in determinations of dominant process
controls on instability, and hence allow better informed
judgements to be made as to appropriate preventative
measures.
 The versatility of the scheme has been illustrated in that
a range of scenarios was developed (Table 2) to
determine the likely dominant controls on slope
instability.
 The implication of relatively low positive pore pressures
required for instability, is corroborated by the applica- Fig. 8. Landslide runout (some 3–5 m in extent) covering the ground floor
tion of resistance envelope methods to the site (Fig. 7). of a downslope property leaving only the upper floor visible. This runout
 The predicted runout extent of the slide (Fig. 6) accords distance accords with the model predicted value shown in Fig. 6. Arrow
with the material disturbance evident at the toe of the indicates height of ground-floor ceiling.
slide (Fig. 8).

The central conclusion that can be drawn is that the
success of the model application reported here, the
versatility of the scheme and the parsimonious data
requirements, can be taken to suggest that comprehensive
scenario testing could now be undertaken to systematically
explore dominant process controls for a wide range of
slope conditions. It is to be hoped that the outcomes of
such comprehensive modelling could contribute to im-
proved environmental management through the revision of
building codes and planning regulations, all which is so
urgently needed in vulnerable communities typified by that
shown in Fig. 1.

References
Lab results Resistance Envelope
Resistance Envelope Ahmad, R., 2003. Developing early warning systems in Jamaica: Rainfall
50 thresholds for hydrogeological hazards. National Disaster Manage-
ment Conference, Jamaica: /http://isis.uwimona.edu.jm/uds/rainha-
Shear Strength

40
zards_files/frame.htmS.
30 Anderson, M.G., 1983. The Prediction of Soil Suction for Slopes in Hong
Kong. Geotechnical Control Office, Hong Kong CE3/81, 244p.
20 Anderson, M.G., Bates, P.D. (Eds.), 2001. Model Validation: Perspectives
10 in Hydrological Science. Wiley, Chichester.
Anderson, M.G., Holcombe, E.A., 2004a. Management of slope stability
0 in communities. Insight 6, 15–17.
0 20 40 60 80 100 Anderson, M.G., Holcombe, E.A., 2004b. Ministerial briefing for the
(a) Normal Load Government of St. Lucia: Management of Slope Stability in
Communities, 27p.
Lab results Resistance Envelope Anderson, M.G., Collison, A.J.C., Hartshorne, J., Lloyd, D.M., Park, A.,
Resistance Envelope 1996. Developments in slope hydrology—stability modelling for
60 tropical slopes. In: Anderson, M.G., Brooks, S.M. (Eds.), Advances
Shear Strength

50 in Hillslope Processes, pp. 799–821.

Anderson, M.G., Kemp, M.J., Lloyd, D.M., 1997. Hydrological Design
40
Manual for Slope Stability in the Tropics. Transport Research
30 Laboratory, 58p.
20 Anthony, K.D., 2001. Strengthening modernizing and repositioning the
10 economy. Budget Speech by The Honourable Dr. Kenny D. Anthony,
0 Prime Minister and Minister for Finance, Economic Affairs and
0 20 40 60 80 100 Information. Government of St. Lucia.
(b) Normal Load Aspinall, W.P., Shepperd, G., Woo, M., Wightman, A., Rowley, K.C.,
Lynch, L.L., Ambeh, W.B., 1994. Seismic ground motion hazard
Fig. 7. (a) Resistance envelope + ¼ 241, c ¼ 12 kPa showing low positive assessment at a site near a segmented subduction zone: the Roseau
pore pressure dependency for stability. (b) Resistance envelope + ¼ 201 Dam, St. Lucia West Indies. Earthquake Spectra 30 (2), 259–292.
c ¼ 10 kPa showing marginally negative pore pressure dependency for Bishop, A.W., 1955. The use of the slip circle in the stability analysis of
stability. slopes. Geotechnique 5 (1), 7–77.

Please cite this article as: Malcolm G. Anderson et al., Assessing slope stability in unplanned settlements in developing countries, Journal of
Environmental Management (2006), doi:10.1016/j.jenvman.2006.08.005
 ARTICLE IN PRESS
M.G. Anderson et al. / Journal of Environmental Management ] (]]]]) ]]]–]]]
British Standard, 1964. Schedule of Weights of building materials, No.
648.
Charalambous, S., 2003. Methodology development for the stability
evaluation of natural and man-made slopes against static and seismic
loads in a GIS environment. Unpublished Dissertation, Technical
University of Athens.
Collison, A., 1993. Assessing the influence of vegetation on slope stability
in the tropics. Unpublished Ph.D. Thesis, University of Bristol.
Commonwealth of Dominica, 1996. Building Code.
Finlay, P.J., Mostyn, G.R., Fell, R., 1999. Landslide risk assessment:
prediction of travel distance. Canadian Geotechnical Journal 36,
556–562.
Fredlund, D.G., 1980. The shear strength of unsaturated soil and its
relationship to slope stability problems in Hong Kong. Hong Kong
Engineer 8, 57–59.
Gray, D., Rogers, C., 1999. Black Mallet Landslide: Preliminary
Geotechnical Evaluation, 9p. Unpublished Report.
Geotechnical Control Office, 1980. Cutslopes in Hong Kong: Assessment
of Stability by Empiricism [CHASE]. Public Works Department, Hong
Kong, 123p.
Greenwood, J.R., Norris, J.E., Wint, J., 2004. Assessing the contribution
of vegetation to slope stability. Geotechnical Engineering 157 (GE4),
199–207.
Hornberger, G.M., Beven, K.J., Cosby, B.J., Sappington, D.E., 1985.
Shenandoah Watershed Study: Calibration of a Topography-Based,
Variable Contributing Area Hydrological Model to a Small Forested
Catchment. Water Resources Research 21, 1841–1850.
Janbu, N., 1977. Slope and excavations in normally and lightly
consolidated clays. In: Proceedings of the Ninth International
Conference on Soil Mechanics and Foundation Engineering, vol. 2,
Tokyo, pp. 549–566.
Kenny, T.C., 1967. Slide behaviour and shear resistance of a quick clay
determined from a study of the landslide at Selnes, Norway. In:
Proceedings of the Geotechnical Conference on Shear Strength
Properties of Natural Soils and Rock, vol. 1, Oslo, pp. 57–64.
Please cite this article as: Malcolm G. Anderson et al., Assessing slope stability in unplanned settlements in developing countries, Journal of
Environmental Management (2006), doi:10.1016/j.jenvman.2006.08.005
11
Klohn-Crippen, 1995. Tropical Storm Debbie Final Report on hydrology:
Roseau Dam & Ancillary Works, St. Lucia, West Indies. Klohn-
Crippen Unpublished Report.
Lamb, T.S.K., Premchitt, J., 1990. Rainfall-runoff on slopes 1984–1988.
Civil Engineering Services Report Hong Kong 1990. Special Project
Report I 1/90.
Lings, M.L., Dietz, M.S., 2004. An improved direct shear apparatus for
sand. Geotechnique 54 (4), 245–256.
Millington, R.J., Quirk, J.P., 1959. Permeability of porous media. Nature
183, 387–388.
Premchitt, J., Lam, H.F., Shen, J.M., 1986. Rainstorm Runoff on Slopes.
Geotechnical Control Office, Hong Kong (Special Projects Division
Report SPR, 5/8699).
Teams, 1999. Position paper on the earth movement crisis of Black Mallet
and Maynard Hill, 7p.
Tear Fund, 2004. Before disaster strikes. Global Action, 8–9.
Wilkinson, P.L., Brooks, S.M., Anderson, M.G., 1998. Investigating the
effect of moisture extraction by vegetation upon slope stability: further
developments of a Combined Hydrology and Stability Model
(CHASM). In: Proceedings of the British Hydrological Society
International Symposium on Hydrology in a Changing Environment.
Theme 4: Hydrology of environmental hazards, pp. 165–178.
Wilkinson, P.L., Brooks, S.M., Anderson, M.G., 2000. Design and
application of an automated non-circular slip surface search within a
Combined Hydrology and Stability Model (CHASM). Hydrological
Processes 14, 2003–2017.
Wilkinson, P.W., 2001. Investigating the hydrological and geotechnical
effects of vegetation on slope stability: development of a fully
integrated numerical model. Unpublished Ph.D. Thesis, University
of Bristol.
Wu, W., Sidle, R.C., 1995. A distributed slope stability model for steep
forested basins. Water Resources Research 31, 1351–1375.
Wu, W., McKinnell, W.P., Swanston, D.N., 1979. Strength of tree roots
and landslides on Prince of Wales Island, Alaska. Canadian
Geotechnical Journal 16, 19–33.