RAINFALL-INDUCED FAILURES OF NATURAL SLOPES IN ......Rainfall-induced slope failures are one of the most damaging natural hazards in the world, with slope failures occurring every

RAINFALL-INDUCED FAILURES OF NATURAL SLOPES

IN TROPICAL REGIONS by

MUHAMMAD SURADI

This thesis is presented for the degree of

Doctor of Philosophy

of

The University of Western Australia

School of Civil, Environmental and Mining Engineering

March 2015

i

Declaration

“I hereby certify that the work embodied in this Thesis is the result of original

research and has not been submitted for a higher degree to any other University or

Institute”

Muhammad Suradi

October 2014

ii

ABSTRACT

Rainfall-induced slope failures are one of the most damaging natural hazards in the

world, with slope failures occurring every rainy season throughout the world. The

occurrences often cause tremendous losses and show an increasing frequency in the

last decade. These slope failures are likely to occur on natural slopes, which are

heterogeneous due to the processes of natural soil formation, unlike constructed

slopes that are designed and mechanically placed and compacted to withstand

specified loads. Simplifications are commonly applied in slope stability analyses, such

as the use of deterministic analysis methods, to make analyses of stability problems

tractable. Such simplifications ignore the real nature of many controlling factors, such

as rainfall intensity fluctuation and patterns as well as spatial variability of soil

properties, potentially leading to inaccurate results. In this thesis, numerical modelling

was used for coupled seepage and slope stability analyses to consider more realistic

representations of the controlling factors in evaluating rainfall-induced slope failures.

The majority of occurrences are shallow slope failures, so only shallow failure

mechanisms were considered throughout this thesis.

The Jabiru landslide, which occurred during extreme rainfall in February and March

2007 in the region of Arnhem Land, Northern Territory of Australia, was selected as a

case study for this thesis. High resolution (hourly) rainfall data was applied in slope

stability analyses to take account of fluctuating intensities of rainfall. In-situ hydraulic

conductivity obtained from field tests using a tension infiltrometer/disc permeameter

was also considered in these analyses. Parametric studies were carried out to

investigate the effect of variation of the controlling factors, both rainfall events and

soil properties, on rainfall-induced slope failures. Results indicated that the Jabiru

landslide could have been predicted to occur due to the extreme rainfall. This study

also highlighted the role of various aspects of the controlling factors, including

rainfall intensity, duration, resolution and pattern as well as hydraulic and shear

strength properties of soils, in the slope failures.

Analyses were performed to take account of spatial variability of soil properties,

particularly hydraulic conductivity which significantly contributes to the magnitude

and rate of infiltration into a slope. A roadside slope of the Great Eastern Highway at

Sawyer’s Valley site (about 40 km northeast of Perth) was selected to characterise

iii

spatial variability of hydraulic conductivity. Numerous in-situ permeability tests were

conducted, and the spatial variability characterised using statistical techniques. A

parametric study was carried out to investigate the effect of spatial variability on

rainfall-induced slope failures for different slope inclinations and hydraulic

conductivities and other general cases. The results indicated that the spatial variability

generally causes a delay in failures of slopes triggered by rainfall, which was not

expected, based on previously published studies. The result is attributed to the

processes by which residual soils form (e.g. chemical weathering), which produce a

heterogeneous distribution of properties such as density (and hence permeability).

This differs from previous studies where rapid generation of positive pore pressure in

slopes during rainfall were predicted, due to deeper soil layers having lower hydraulic

conductivity (which may be a reasonable assumption for transported soils). This study

also showed that the effect of spatial variability of hydraulic properties on slope

failures is insignificant for different slope inclinations and high-conductivity slopes

but is much more significant for low-conductivity slopes.

To develop robust and inexpensive landslide prediction methods for areas where

numerous steep natural slopes occur, existing methods were evaluated, and where

appropriate, modified. The evaluated landslide prediction methods included the

rainfall intensity-duration-based method, the antecedent rainfall-based method and

rainfall intensity-frequency-duration (IFD) method. The landslide prediction method

based on antecedent rainfall was shown to hold much promise, particularly when

tailored to a particular region using typical soil strength and hydraulic parameters. A

simple screening tool that includes monitoring of daily rainfall events and occasional

measurement of in-situ water content could be invaluable to provide an early warning

system for prediction likelihood of landslides in areas of the world where use of

sophisticated instrumentation is not possible.

iv

ACKNOWLEDGEMENT

First and foremost, I would like to express my deepest gratitude to my supervisor,

Winthrop Professor Andy Fourie, for his guidance, inspiration, generous support,

constant encouragement and consistent feedback throughout my four-year study.

Moreover, his attitude towards research will be an invaluable stimulation for the rest

of my life.

I would also like to thank my co-supervisor, Professor Richard Durham, for his

invaluable feedback on my writing and his assistance of review comments during the

drafting of my PhD thesis.

I specially want to thank my former co-supervisor, Winthrop Professor Martin Fahey,

who invited me to the University of Western Australia and provided assistance for

initial settlement. His friendly style makes me easy to communicate with him.

I appreciate the financial support from the General Directorate of Higher Education

(GDHE) of Indonesia and Ujung Pandang State Polytechnic (UPSP) for me to pursue

a PhD study at the University of Western Australia (UWA).

I would like to thank Dr Mike J. Saynor for his wonderful assistance during fieldwork

at the Jabiru site, Northern Territory of Australia. I would also like to thank Dr Joanne

Edmondston for taking time to proofread initial draft of this thesis. Special thanks also

go to Dr Alsidqi Hasan, Dr Tutun Nugraha and Dr Agus S. Muntohar for their fruitful

discussion and friendship, as well as my relative Dr Andi Shiddiq Yunus for his great

support and companionship during the first half of my study period.

I am indebted to Keith Russell and his team for their IT support and our wonderful

admin staff Charlie Askew. My acknowledgement also goes to Alex Duff, Claire

Bearman, Usha Mani, Behnaz Abdollahzadeh, Masoomeh Lorestani for their support

and friendship during laboratory tests.

I am grateful to my dear friends and colleagues, Jun Li, Yanyan Sha, Amin

Rismanchian, Megan Walske, Dezheng Lao, Yusuke Suzuki, David Reid, Jinglong

Gao, Gonzalo Souza, Stefanus Safinus, Neyamat Ullah, Azrul Muttalib, Fauzan

Sahdi, Bassem Youssef, for their warm friendship.

v

Last but not least, infinite thanks and appreciation to my family: my respected and

beloved father Andi Mappasulle and mother Andi Dinar for their genuine love and

invaluable and perpetual support, my sisters and brothers for their encouragement and

prayers, my dearest wife Andi Nurhayati and children Andi Fazlul Nurhadi, Andi

Sofyan Hadi, Andi Nurdian Musfira and Andi Aisya Haliza for their wonderful love,

support and understanding during all this time.

vi

Table of Contents

Declaration ................................................................................................................................ i

Abstract .................................................................................................................................... ii

Acknowledgement .................................................................................................................. iv

Table of Contents ................................................................................................................... vi

List of Figures .......................................................................................................................... x

List of Tables ......................................................................................................................... xx

List of Symbols ..................................................................................................................... xxi

Chapter 1: INTRODUCTION ............................................................................................... 1

1.1 Research background .................................................................................................... 1

1.2 Occurrences of slope failures triggered by rainfall ....................................................... 5

1.3 Aims and scope of the research .................................................................................... 9

1.4 Thesis outline ............................................................................................................... 10

1.5 Publications .................................................................................................................. 11

Chapter 2: LITERATURE REVIEW .................................................................................. 13

2.1 Introduction ................................................................................................................ 14

2.2 Slope failure mechanisms .......................................................................................... 14

2.2.1 Characterising residual soil slopes in tropical regions .................................... 14

2.2.2 Contribution of controlling factors to slope failure ........................................ 19

2.2.3 Types of slope failure mechanisms ................................................................. 22

2.3 Seepage analysis ......................................................................................................... 23

2.4 Slope stability analysis ............................................................................................... 27

2.4.1 Limit equilibrium method (LEM) ................................................................... 28

2.4.2 Finite element method ..................................................................................... 31

2.4.3 Probabilistic method ....................................................................................... 34

2.5 Characterising the spatial variability of soil properties .............................................. 35

2.5.1 Classical statistical measures of soil properties ............................................. 36

2.5.2 Spatial variability of soil properties ................................................................ 38

2.5.3 Published data of inherent soil variability ....................................................... 41

2.6 Prediction of rainfall-induced slope failures ............................................................... 43

2.6.1 Empirical correlation between rainfall intensity-duration and landslide occurrences ..................................................................................................... 43

vii

2.6.2 Empirical correlation between antecedent-main rainfall and landslide occurrences ..................................................................................................... 44

2.6.3 Approximate method ...................................................................................... 46

2.7 Research hypotheses ................................................................................................... 51

2.8 Summary ..................................................................................................................... 53

Chapter 3: METHODOLOGY OF RESEARCH .............................................................. 54

3.1 Introduction ................................................................................................................. 54

3.2 Study areas .................................................................................................................. 55

3.2.1 Jabiru ............................................................................................................... 55

3.2.2 Sawyer’s Valley .............................................................................................. 58

3.2.3 Boddington ...................................................................................................... 60

3.3 Research procedure ..................................................................................................... 61

3.3.1 Preparation ...................................................................................................... 63

3.3.2 Data collection ................................................................................................ 64

3.3.3 Data analysis and verification ......................................................................... 66

3.3.4 Analysis modelling ......................................................................................... 67

3.4 Field and laboratory investigations ............................................................................ 68

3.4.1 Soil-sampling and field tests ........................................................................... 68

3.4.2 Laboratory tests ............................................................................................... 73

3.5 Statistical analysis ....................................................................................................... 75

3.6 Analysis modelling ..................................................................................................... 76

3.7 Conclusions ................................................................................................................. 77

Chapter 4: APPLICATION OF COUPLED ANALYSES OF SEEPAGE AND STABILITY OF SLOPES SUBJECTED TO EXTREME RAINFALL ...... 78

4.1 Introduction ................................................................................................................. 78

4.2 Controlling factors of slope stability ........................................................................... 79

4.2.1 Rainfall data .................................................................................................... 79

4.2.2 Soil properties ................................................................................................. 79

4.3 Overview of analysis design ....................................................................................... 86

4.3.1 Analysis modelling ......................................................................................... 87

4.3.2 Parametric study .............................................................................................. 89

4.4 Results and discussion ................................................................................................ 97

4.4.1 The effect of rainfall events ............................................................................ 97

4.4.2 The effect of rainfall resolutions ................................................................... 103

viii

4.4.3 The effect of rainfall patterns ........................................................................ 109

4.4.4 The effect of soil properties .......................................................................... 112

4.5 Conclusions ............................................................................................................... 117

Chapter 5: APPLICATION OF COMPLEX ANALYSIS INCORPORATING SPATIAL VARIABILITY OF SOIL PROPERTIES .................................. 120

5.1 Introduction ............................................................................................................... 120

5.2 Slope stability analysis accounting for spatial variability of hydraulic conductivity 121

5.3 Determination of spatial variability parameters ........................................................ 122

5.4 Evaluation of slope stability accounting for variability of hydraulic conductivity .. 123

5.4.1 Analysis data ................................................................................................. 124

5.4.2 Modelling procedure .................................................................................... 125

5.4.3 Parametric study ............................................................................................ 127

5.5 Results and discussion .............................................................................................. 128

5.5.1 The effect of spatial variability of hydraulic conductivity on rainfall-induced slope failures for different slope inclinations ..................... 128

5.5.2 The effect of spatial variability of hydraulic conductivity on rainfall-induced slope failure for different soil hydraulic conductivities .... 136

5.5.3 The effect of spatial variability of hydraulic conductivity on rainfall-induced slope failures for general cases of slopes ........................... 147

5.6 Conclusions ............................................................................................................... 165

Chapter 6: SLOPE FAILURE PREDICTION ................................................................ 166

6.1 Introduction ............................................................................................................... 166

6.2 Overview of proposed approach ............................................................................... 168

6.3 Evaluation of existing methods for predicting rainfall-induced shallow slope failure ........................................................................................................................ 168

6.3.1 Evaluation of slope failure prediction method based on correlation between rainfall intensity and duration ........................................................ 168

6.3.2 Evaluation of slope failure prediction method based on antecedent rainfall .......................................................................................................... 171

6.3.3 Evaluation of slope failure prediction method based on rainfall IFD-soil interaction ..................................................................................................... 178

6.4 Rainfall thresholds for shallow slope failure ............................................................ 180

6.4.1 Rainfall thresholds based on rainfall intensity-duration ............................... 180

6.4.2 Rainfall thresholds based on antecedent rainfall .......................................... 184

6.4.3 Rainfall thresholds based on rainfall-soil interaction ................................... 188

ix

6.5 Development of a simple screening tool for anticipating slope failures ................... 191

6.6 Conclusions ............................................................................................................... 193

Chapter 7: CONCLUDING REMARKS .......................................................................... 195

7.1 Summary ................................................................................................................... 195

7.2 Extreme rainfall significantly contributes to rainfall-induced slope failures ........... 196

7.3 Rainfall events and soil properties are both primary controlling factors in rainfall-induced slope failures ................................................................................... 197

7.4 Spatial variability analysis is important when evaluating rainfall-induced failures of natural slopes ....................................................................................................... 197

7.5 Numerical analysis accounting for characteristics of rainfall events and unsaturated soil mechanics principles is necessary for more accurate and widespread use in rainfall-induced landslide prediction .......................................... 198

7.6 Recommendations for future work .......................................................................... 199

7.6.1 Slope stability analyses incorporating evaporation, transpiration, and root effects ............................................................................................................ 199

7.6.2 Evaluating rainwater infiltration using tipping bucket rain gauges .............. 199

7.6.3 Three-dimensional slope stability analyses .................................................. 199

7.6.4 Comprehensive spatial variability analyses .................................................. 199

7.6.5 Further development of a simple screening tool for landslide prediction .... 200

References ............................................................................................................................ 201

Appendices ........................................................................................................................... 216

Appendix A Calibration of the SVFLUX and SVSLOPE software .................................. 216

A.1 Seepage analyses ........................................................................................... 216

A.2 Slope stability analyses ................................................................................. 219

Appendix B Stability analyses .......................................................................................... 223

B.1 Methods of analysis ...................................................................................... 223

B.2 Modelling procedures of seepage analyses with SVFLUX (Thode and Gitirana, 2012) ........................................................................... 224

B.3 Modelling procedures of slope stability analyses with SVSLOPE (Fredlund et al., 2008) ...................................................................................233

Appendix C Shear box test results ....................................................................................236

x

LIST OF FIGURES

Figure Description Page

Figure 1.1 Worldwide slope failures triggered by rainfall in 2003-2010

(after Petley, 2014)

6

Figure 2.1 Seasonal variation in water table and pore pressure due to

climatic effects (after Wesley, 2010)

15

Figure 2.2 Typical SWCC for different types of soil (after Fredlund and

Xing, 1994)

16

Figure 2.3 Hydraulic conductivity function for unsaturated soils (after

Rahardjo et al., 2007)

17

Figure 2.4 Progress of infiltration through initially unsaturated soils

during rainfall (after Tholin and Kiefer, 1959)

18

Figure 2.5 Variability of a parameter illustrated by: (a) two different

coefficients of variation and (b) two different types of data

distribution (after Fenton and Griffiths, 2011)

37

Figure 2.6 Spatial variability of a parameter t in a slope geometry with

two different correlation lengths: (a) low correlation length

and (b) high correlation length (after Fenton and Griffiths,

2011)

39

Figure 2.7 Slope with two different correlation lengths: (a) low

correlation length and (b) high correlation length (after

Griffiths et al., 2007)

40

Figure 2.8 Thresholds of rainfall intensity-duration of landslide

occurrences obtained from many sites all over the world (after

Guzzetti et al., 2007)

44

Figure 2.9 Threshold line for landslide probability (after Crozier and

Eyles, 1980)

46

Figure 2.10 Cross-section of wetted zone of surficial soils due to rain

infiltration

47

Figure 2.11 The rainfall intensity-frequency-duration (IFD) curves for the

Jabiru site recorded at Gulungul Creek in 2007 (after Moliere

et al., 2007)

49

xi

Figure 2.12 The contribution of suction to shallow slope failures (after

Fourie, 1996)

50

Figure 3.1 Location of the Jabiru site 56

Figure 3.2 A characteristic Jabiru landslide: (a) front view and (b) side

view

57

Figure 3.3 Simplified geometry of the Jabiru slope 58

Figure 3.4 Location of the Sawyer’s Valley site 59

Figure 3.5 View of the Sawyer’s Valley site 60

Figure 3.6 Location of the Boddington site 61

Figure 3.7 Scheme of research procedures 62

Figure 3.8 Hourly extreme rainfall data obtained from Jabiru Airport

Station 014198, 24 to 28 February 2007 (Australian

Government, 2012)

64

Figure 3.9 Hourly extreme rainfall data obtained from Sembawang

Station 80, December 2006 (Singaporean Government, 2011)

65

Figure 3.10 Hourly extreme rainfall data obtained from Brisbane Station

040913, January 2011 (Australian Government, 2011)

65

Figure 3.11 Layout of soil-sampling and field tests at the Jabiru site 69

Figure 3.12 Layout of soil-sampling and field tests at the Sawyer’s Valley

site

69

Figure 3.13 Layout of soil-sampling and field tests at the Boddington site 70

Figure 3.14 The 1988 CSIRO Disc Permeameter used in field tests 71

Figure 3.15 The Eijkelkamp Tension Infiltrometer used in field tests 72

Figure 4.1 Particle size distribution of slope soils for: (a) the Jabiru site

and (b) the Sawyer’s Valley site

81

Figure 4.2 Plasticity of slope soils for: (a) the Jabiru site and (b) the

Sawyer’s Valley site

82

Figure 4.3 Soil-water characteristic curve (SWCC) for: (a) the Jabiru

site, (b) the Sawyer’s Valley site and (c) the Boddington site

85

Figure 4.4

Figure 4.5

Slope geometry and boundary conditions applied in the

seepage analysis

Hydraulic conductivity function (average values) of the soils

from the Jabiru, Sawyer’s Valley and Boddington sites

88

88

xii

Figure 4.6

Simulated rainfall with 64 mm/h major intensities occurring

every 20 h and various minor intensities with an average

value of 3.13 mm/h between the major intensities

93

Figure 4.7 Simulated rainfall with various intensities and time intervals

for major rainfall and constant intensity for minor rainfall

(I = 0.5 mm/h, much lower than ks)

94

Figure 4.8 Simulated rainfall with 24-h cyclic pattern occurring every

2 h and 0.5 mm/h minor intensity occurring between the

major intensities

95

Figure 4.9 Simulated rainfall with three different patterns: (a) delayed

pattern, (b) advanced pattern and (c) normal pattern (after

Rahimi et al., 2011 and Muntohar et al., 2013) (all the three

rainfall patterns had the same rainfall amount)

96

Figure 4.10 Stages for the effect of rainfall on slope instability 97

Figure 4.11 Variation in factors of safety after the start of rainfall for the

Jabiru slope with different rainfall intensity (all other material

parameters kept constant)

98

Figure 4.12 Variation in factors of safety after the start of rainfall for the

Jabiru slope with different rainfall volume in mm3/mm2 (all

other material parameters kept constant)

99

Figure 4.13 Variation in factors of safety for the Jabiru slope after the start

of rainfall data from February 2007 recorded at Jabiru Airport

(the closest station to the site), starting with different initial

suctions

101

Figure 4.14 Variation in pore-water pressures with time at 3 locations on

the slope (top, midway and toe) and at three different depths,

where position 3 is at the soil surface, position 1 at the base of

the weathered soil and position 2 midway between positions 1

and 3

101

Figure 4.15 Pore-water pressure contours at the surface soil for two

different rainfall time (t): (a) t = 6 h and (b) t = 84 h

102

xiii

Figure 4.16

Variation in factors of safety for the Jabiru slope using the

application of three rainfall data sets with various resolutions:

(a) Jabiru rainfall, (b) Singapore rainfall and (c) Brisbane

rainfall

104

Figure 4.17

Variation in factors of safety for the Jabiru slope with the

application of three simulated rainfall scenarios with various

resolutions (dt): (a) rainfall pattern with high intensity

fluctuation presented in Figure 4.5, (b) rainfall pattern with

medium intensity fluctuation presented in Figure 4.6(a) and

(c) rainfall pattern with slight intensity fluctuation presented

in Figure 4.7

108

Figure 4.18 Variation in factors of safety after the start of rainfall for the

Jabiru slope with various values of uniform rainfall intensities

110

Figure 4.19 Variation in factors of safety after the start of rainfall for the

Jabiru slope with various fluctuating intensities of rainfall

111

Figure 4.20 Variation in factors of safety after the start of rainfall for the

Jabiru slope with various rainfall patterns of smooth intensity

change

112

Figure 4.21 Variation in factors of safety after the start of rainfall for the

Jabiru slope with different hydraulic conductivity (all other

material parameters kept constant)

113

Figure 4.22 Variation in factors of safety after the start of rainfall for the

Jabiru slope with different initial suctions (all other material

parameters kept constant)

114

Figure 4.23 Variation in factors of safety after the start of rainfall for the

Jabiru slope with different apparent cohesion values (all other

material parameters kept constant)

115

Figure 4.24 Variation in factors of safety after the start of rainfall for the

Jabiru slope with different internal friction angles (all other

material parameters kept constant)

116

xiv

Figure 4.25 Variation in factors of safety after the start of rainfall for the

Jabiru slope with different actual shear strength parameters

obtained from laboratory tests for different soil samples (all

other material parameters kept constant)

116

Figure 5.1 Determination of correlation length of soil hydraulic

conductivity along: (a) Line 1 and (b) Line 2

124

Figure 5.2 Slope geometry, boundary conditions and variability of

hydraulic conductivity applied in the seepage analysis

126

Figure 5.3 Factor of safety of the Jabiru slope (β = ) with rainfall

time for various coefficients of variation (CV) of hydraulic

conductivity and different correlation lengths (θln k):

(a) θln k = m, (b) θln k = 3 m and (c) θln k = 5 m

130

Figure 5.4 Factor of safety of the Jabiru slope (β = ) with rainfall

time for various correlation lengths (θln k) of hydraulic

conductivity and different coefficients of variation (CV):

(a) CV = 10%, (b) CV = 100% and (c) CV = 1000%

131

Figure 5.5 Factor of safety of a steeper slope (β=30 ) with rainfall time

for various coefficients of variation (CV) of hydraulic

conductivity and different correlation lengths (θln k):

(a) θln k = m, (b) θln k = 3 m and (c) θln k = 5 m

132

Figure 5.6 Factor of safety of a steeper slope (β=30 ) with rainfall time

for various correlation lengths (θln k) of hydraulic conductivity

and different coefficients of variation (CV): (a) CV = 10%,

(b) CV = 100% and (c) CV = 1000%

133

Figure 5.7 The effect of different coefficients of variation of hydraulic

conductivity (CV) on seepage flow in the slope for the same

rainfall time and θln k = 1 with various CVs: (a) CV = 10%,

(b) CV = 100% and (c) CV = 1000%

135

Figure 5.8 The effect of different correlation lengths (θln k) of hydraulic

conductivity on seepage flow in the slope for the same rainfall

time and CV = 500% with various θln k: (a) θln k = 1,

(b) θln k = 3 and (c) θln k = 5

135

xv

Figure 5.9 Factor of safety of the Jabiru slope (ks = 80 mm/h) with

rainfall time for various coefficients of variation (CV) of

hydraulic conductivity and different correlation lengths (θln k):

(a) θln k = m, (b) θln k = 3 m, and (c) θln k = 5 m

137

Figure 5.10 Factor of safety of the Jabiru slope (ks = 80 mm/h) with

rainfall time for various correlation lengths (θln k) of hydraulic

conductivity and different coefficients of variation (CV):

(a) CV = 10%, (b) CV = 100%, and (c) CV = 1000%

138

Figure 5.11 Seepage propagation in the low-conductivity slope (average

ks = 8 mm/h and β = ) with high spatial variability of

hydraulic conductivity (CV = 000% and θlnk = 1 m) exposed

to uniform rainfall intensity (I = 8 mm/h) from rainfall time:

(a) t = 0 h to (b) t = 48 h and (c) t = 96 h

139

Figure 5.12 Seepage propagation in the high-conductivity slope (average

ks = 80 mm/h and β = ) with high spatial variability of

hydraulic conductivity (CV = 000% and θlnk = 1 m) exposed

to uniform rainfall intensity (I = 8 mm/h) from rainfall time:

(a) t = 0 h to (b) t = 48 h and (c) t = 96 h

140

Figure 5.13 Variation in factors of safety with rainfall time for two

different slope inclinations and various coefficients of

variation for each slope inclination

141

Figure 5.14 Variation in factors of safety with rainfall time for two

different values of hydraulic conductivity and various

coefficients of variation for each slope inclination

142

Figure 5.15 The effect of spatial variability of hydraulic conductivity on

the amount of runoff in low-conductivity (ks = 8 mm/h) and

high-conductivity (ks = 80 mm/h) slopes exposed to rainfall

for deterministic (det) and spatial variability (sv) analyses

143

xvi

Figure 5.16 The effect of spatial variability of hydraulic conductivity on

pore-water pressures generated along the layer intercept (2 m

deep) in low-conductivity (ks = 8 mm/h) and high-

conductivity (ks = 80 mm/h) slopes after 90-hour rainfall (I =

8 mm/h) for deterministic (det) and spatial variability (sv)

analyses (the effect is also shown for slopes without an

impermeable layer at shallow depths, called ‘deep’ for

comparison)

144

Figure 5.17 Minimum factors of safety (Fm) of the Jabiru slope with

various spatial variability of hydraulic conductivity for:

(a) ks = 8 mm/h and β = , (b) ks = 8 mm/h and β=30 and

(c) ks = 80 mm/h and β=

146

Figure 5.18 Factor of safety of slopes (β = ) with rainfall time

resulting from deterministic (det) and spatial variability (sv)

analyses for 3 different soil stratifications: impermeable layer

at a shallow depth (imp), heterogeneous slopes (het) and

homogeneous slopes (hom)

149

Figure 5.19 Pore-water pressure contours of slopes (ks = 8 mm/h and

β = ) resulting from deterministic analyses at t = 90 h for:

(a) slopes with impermeable layer at a shallow depth, (b)

heterogeneous slopes and (c) homogeneous slopes

150

Figure 5.20 Pore-water pressure contours of slopes (ks = 8 mm/h and β =

) resulting from spatial variability analyses at t = 90 h for:

(a) slopes with impermeable layer at a shallow depth,

(b) heterogeneous slopes and (c) homogeneous slopes

151

Figure 5.21 Pore-water pressures along the base of surface soils (at 2 m

depth) resulting from deterministic analyses with different

time after the start of rainfall for shallow inclination slopes

(β = ) with 3 different soil stratifications: (a) impermeable

layer at a shallow depth (imp-19-det), (b) heterogeneous

slopes (het-19-det) and (c) homogeneous slopes (hom-19-det)

153

xvii

Figure 5.22 Pore-water pressures along the base of surface soils (at m

depth) resulting from spatial variability analyses with

different time after the start of rainfall for shallow inclination

slopes (β = ) with 3 different soil stratifications:

(a) impermeable layer at a shallow depth (imp-19-sv), (b)

heterogeneous slopes (het-19-sv) and (c) homogeneous slopes

(hom-19-sv)

154

Figure 5.23 Factor of safety of slopes (β = 40 ) with rainfall time

resulting from deterministic (det) and spatial variability (sv)

analyses for 3 different soil stratifications: impermeable layer

at 2 m depth (imp), heterogeneous slopes with less permeable

layer at 2 m depth (het) and homogeneous slopes (hom)

155

Figure 5.24 Pore-water pressure contours of slopes (ks = 8 mm/h and

β = 40 ) resulting from deterministic analyses at t = 90 h for:

(a) slopes with impermeable layer at 2 m depth,

(b) heterogeneous slopes with less permeable layer at 2 m

depth and (c) homogeneous slopes

156

Figure 5.25 Pore-water pressure contours of slopes (ks = 8 mm/h and

β = 40 ) resulting from spatial variability analyses at t = 90 h

for: (a) slopes with impermeable layer at 2 m depth,

(b) heterogeneous slopes with less permeable layer at 2 m

depth and (c) homogeneous slopes

157

Figure 5.26 Pore-water pressures resulting from deterministic analyses

along slope base (β = 40 ) at 2 m depth of surface soils

(ks = 8 mm/h) with different rainfall time for 3 different soil

stratifications: (a) impermeable layer at 2 m depth, (b)

heterogeneous slopes (less permeable layer at 2 m depth), and

(c) homogeneous slopes

158

Figure 5.27 Pore-water pressures resulting from spatial variability

analyses along slope base (β = 40 ) at 2 m depth of surface

soils (ks = 8 mm/h) with different rainfall time for 3 different

soil stratifications: (a) impermeable layer at a shallow depth,

(b) heterogeneous slopes and (c) homogeneous slopes

159

xviii

Figure 5.28 Factor of safety of slopes with rainfall time resulting from

deterministic (det) analyses for 2 different inclinations:

β = and 40 , and 3 different soil stratifications:

impermeable layer at a shallow depth (imp), heterogeneous

slopes (het) and homogeneous slopes (hom)

160

Figure 5.29 Factor of safety of slopes (β = ) with rainfall time

resulting from spatial variability (sv) analyses for different

inclinations: β = and 40 , and 3 different soil

stratifications: impermeable layer at a shallow depth (imp),

heterogeneous slopes (het) and homogeneous slopes (hom)

161

Figure 5.30 Pore-water pressure contours for slopes (ks = 8 mm/h and

β = 40 ) with impermeable layer at a shallow depth resulting

from deterministic analyses at: (a) t = 90 h and (b) t = 120 h

and spatial variability analyses at: (c) t = 90 h and (d) t = 120h

162

Figure 5.31 Pore-water pressure contours for heterogeneous slopes

(ks = 8 mm/h and β = 40 ) resulting from deterministic

analyses at: (a) t = 90 h and (b) t = 120 h and spatial

variability analyses at: (c) t = 90 h and (d) t = 120 h

163

Figure 5.32 Pore-water pressure contours for homogeneous slopes

(ks = 8 mm/h and β = 40 ) resulting from deterministic

analyses at: (a) t = 90 h and (b) t = 120 h and spatial

variability analyses at: (c) t = 90 h and (d) t = 120 h

164

Figure 6.1 Factor of safety for the Jabiru slope, with rainfall duration for

various rainfall intensities, and in-situ slope parameters

(ks = 8 mm/h and β = )

169

Figure 6.2 Threshold of rainfall intensity-duration triggering slope

failure (ks = 8 mm/h and β = ), where in this case ‘failure’

is deemed to be the minimum factor of safety achieved, i.e.

Fm = 1.1

170

Figure 6.3 Factor of safety for the Jabiru slope, with rainfall time related

to various initial suction measures (ks = 8 mm/h and β = )

171

Figure 6.4 Hypothetical variation of factor of safety with rainfall time

resulting from a slope stability analysis

172

xix

Figure 6.5 The effect of various combinations of antecedent and main

rainfall on slope stability for ks = 8 mm/h and β =

174

Figure 6.6 Rainfall threshold line for slope failure for ks = 8 mm/h and

β =

174

Figure 6.7 Rainfall threshold line for slope failure with ks = 8 mm/h and

β = verified by several rainfall events

177

Figure 6.8 The effect of various rainfall patterns and events on the

stability of the slope with ks = 8 mm/h and β =

178

Figure 6.9 Rainfall threshold for shallow slope failure (the Jabiru slope)

based on the modified rainfall IFD method

179

Figure 6.10 Comparison of rainfall time required to develop a wetting

front with various initial suction levels using both the

approximation and numerical methods of analysis

180

Figure 6.11 Rainfall threshold lines for slope failure based on rainfall

intensity-duration for a slope with two different hydraulic

conductivities (ks = 8 mm/h and 80 mm/h) and three slope

inclinations: (a) β = , (b) β = 30 and (c) β = 40

182

Figure 6.12 ainfall threshold lines for slope failure based on rainfall

intensity-duration for a slope with three inclinations (β = ,

30 , and 40 ) and two different hydraulic conductivities: (a)

ks = 8 mm/h and (b) ks = 80 mm/h

183

Figure 6.13 Rainfall threshold lines for slope failure based on antecedent

rainfall for a slope with two different hydraulic conductivities

(ks = 8 mm/h and 80 mm/h) and three slope inclinations: (a) β

= , (b) β = 30 and (c) β = 40

185

Figure 6.14 ainfall threshold lines for slope failure based on antecedent

rainfall for a slope with three different slope inclinations (β =

, 30 , and 40 ) and two different hydraulic conductivities:

(a) ks = 8 mm/h and (b) ks = 80 mm/h

187

Figure 6.15 Monitoring system using a simple screening tool to anticipate

slope failures

192

xx

LIST OF TABLES

Table Description Page

Table 2.1 Static equilibrium conditions satisfied by limit equilibrium

methods (after Abramson et al., 2002)

29

Table 2.2 The coefficients of variation for various soil properties (after

Lacasse and Nadim, 1996)

41

Table 2.3 Statistics of the hydraulic conductivity of compacted soil

liners (after Benson, 1993)

42

Table 3.1 Detailed activities during preparation phase 63

Table 3.2 Overview of data collection 66

Table 3.3 General overview of analysis modelling 67

Table 4.1 Basic and index soil properties and classifications 83

Table 4.2 Data analysis of hydraulic conductivity tests using the Disc

Permeameter or Tension Infiltrometer

84

Table 4.3 Shear strength parameters resulting from shear box test 86

Table 4.4 Summary of various parameters applied in parametric study 91

Table 4.5 Summary of variations of simulated rainfall with fluctuating

intensity

95

Table 5.1 Statistical parameters regarding the spatial variability of

hydraulic conductivity of slopes at the Sawyer’s Valley site

123

Table 5.2 Constant input parameters used in the analysis 127

Table 5.3 Variation of input parameters used in the analysis 128

Table 6.1 Determination of antecedent daily rainfall factors (Kn) based

on uniform rainfall intensity (I = 43 mm/day)

173

Table 6.2 Antecedent daily rainfall factors (Kn) for different hydraulic

conductivities (ks) and angles of slope inclination (β)

173

Table 6.3 Combination of antecedent and main rainfall for determining

the threshold line for slope failure (ks = 8 mm/h and β = )

174

Table 6.4 Various patterns of simulated rainfall and rainfall data used

for verification in slope failure probability (Pn refers to a

rainfall event that n-day before the main rainfall (P0))

176

Table 6.5 Risk analyses for the Jabiru slope 188

xxi

LIST OF SYMBOLS

a SWCC parameter

AE actual evaporation

B point where soil sampling and field tests were carried out in the

Boddington site

CP percentage of coarse particles

CV coefficient of variation

C(ψ) correction function for SWCC equation

c parameter for landslide prediction

c′ apparent cohesion

cp′ peak apparent cohesion

cr′ residual apparent cohesion

D rainfall duration

dt rainfall resolution

e the natural number

F factor of safety

Fi initial factor of safety

Fm minimum factor of safety

Fn factor of safety for slopes on the nth day before main rainfall

FP percentage of fine particles

h water tension

ht hydraulic or total head

I rainfall intensity

Imin minimum infiltration

IFD intensity-frequency-duration of rainfall

K factor indicating contribution of rainfall to antecedent rainfall index

Kn factor indicating contribution of antecedent rainfall on the nth day

before main rainfall

klim limiting value for saturated hydraulic conductivity required to saturate

surficial soils

kmin minimum hydraulic conductivity

ks saturated hydraulic conductivity

kw unsaturated hydraulic conductivity

xxii

L landslide

LL liquid limit

m SWCC parameter

mw slope of SWCC

n number of data

n SWCC parameter

NL no landslide

NP net percolation

Nx number of grids in X axis

Ny number of grids in Y axis

P point where soil sampling and field tests were carried out in the

Sawyer’s Valley site

P precipitation

P rainfall volume

p power factor adjusting prediction of infiltration

P0 main rainfall (volume)

Pa0 antecedent daily rainfall index

Pn antecedent rainfall on the nth day before main rainfall

PI plasticity index

PL plastic limit

Q infiltration discharge

q applied boundary flux

R infiltration rate under steady-state condition

Roff runoff

r inner radius of the water tower for tension infiltrometer and disc

permeameter

S wetting front capillary suction

S point where soil sampling and field tests were carried out in the Jabiru

site

SG specific gravity

T analysis duration

t elapsed rainfall time

Tw time required to saturate wetted zone

ua pore-air pressure

xxiii

uw pore-water pressure

(ua – uw) matric suction

V rainfall volume

vi infiltration rate

w water content

zw depth of wetting front

α infiltration factor

α parameter for landslide prediction

α slope angle

β parameter for landslide prediction

β slope angle

ϕ′ internal friction angle

ϕp′ peak internal friction angle

ϕr′ residual internal friction angle

ϕb the angle indicating the rate of increase in shear strength relative to

matric suction

μ change of volumetric water content from initial condition

μ mean value

σ normal stress

σ standard deviation

σn total normal stress

σ2 variance

(σ – ua) net normal stress

τ distance

τ shear strength

ψ matric suction

ψi initial suction

θ correlation length

θ volumetric water content

θs saturated volumetric water content

θw unsaturated volumetric water content

Θ normalized volumetric water content

γt total unit weight of soil

xxiv

γw unit weight of water

ρ bulk density

ρ correlation estimator

1

CHAPTER 1

INTRODUCTION

1.1 RESEARCH BACKGROUND

Rainfall-induced slope failures occur frequently all over the world during rainy

seasons. These types of slope failures have become one of the most disastrous natural

hazards worldwide (Alcantara-Ayala, 2002), usually causing economic loss and

sometimes even fatalities. These failures commonly occur in natural slopes,

particularly residual soil slopes (Campbell, 1975; Lumb, 1975; Morgestern and de

Matos, 1975; Fukuoka, 1980; Brand et al., 1984; Vargas et al., 1986; Kim et al., 1991;

Lacerda, 1997; Au, 1998; Franks, 1999; Rahardjo et al., 2009) and as infrastructure

develops around slopes the risk of damage to such infrastructure due to slope failure

increases. It is noted that the majority of rainfall-induced slope failures occur through

shallow failure mechanisms (Guzzetti et al., 2008), with the depth of failure usually

less than 2 m.

Both rainfall and soil properties have been widely accepted as primary controlling

factors in rainfall-induced slope failures (Brand et al., 1984; Rahardjo et al., 2007),

particularly in tropical regions where there are often high intensity rainfall and humid

conditions (i.e. low evaporation rates). Rainfall can also cause intense and deep

chemical weathering of slopes and the leaching of minerals from near-surface soils.

This may result in open structures of the soils near the slope surface, with high void

ratios not uncommon. Hydraulic properties of a soil, which are related to void ratio

2

(among other factors) determine the amount of rainwater infiltrating to slopes; this

infiltration may trigger slope failures.

Rainfall events, which may be quantified in terms of intensity, duration, antecedent

condition, resolution, and pattern play an important role in rainfall-induced slope

failure, as suggested by researchers such as Brand et al. (1984), Fourie (1996),

Rahardjo et al. (2001; 2007), Hearman and Hinz (2007), Rahimi et al. (2011) and

Muntohar et al. (2013). Intense rainfall has often been identified as a triggering factor

for many slope failures around the world (Fuchu et al., 1999; van Asch et al., 1999;

Olivares and Picarelli, 2003; Shaw-Shong, 2004; Huat et al., 2006; Guzzetti et al.,

2008) and it is accepted that there have been many slope failures during prolonged

rainfall (Petley, 2012). It is well recognised that antecedent rainfall significantly

contributes to rainfall-induced failures of low-conductivity slopes, but probably has

less significant contribution to those of high-conductivity slopes (Rahardjo et al.,

2008). As rainfall intensity usually fluctuates, rainfall resolution is often crucial in

determining the amount of rainwater infiltration which may lead to slope failures.

Thus, the use of high resolution rainfall data (hourly, rather than daily rainfall data) in

the analysis of rainfall-induced slope stability may produce more accurate results, as

suggested by Hearman and Hinz (2007) and Lowry et al. (2009). In addition, specific

rainfall patterns (high intensities in the beginning, followed by a consistent decrease

towards the end of the rainfall) produced the worst slope stability, the lowest

minimum factor of safety (indicator of slope stability) and the shortest time to reach

the minimum factor of safety (Rahimi et al., 2011; Muntohar et al., 2013).

The interaction between rainfall events and soil hydraulic properties essentially

determines the amount of rainwater infiltration required to reduce suction of surficial

soil, which can trigger a slope failure. Theoretically, the incident rainfall can be

totally infiltrated to soils when the rainfall intensity is about the same magnitude as

the soil hydraulic conductivity. In this case, rainwater infiltration is most likely to

reduce suction of the surficial soil to a critical condition. Rainfall with very low

intensity will infiltrate completely to the surficial soil but it may be insufficient to

reduce suction of the soil. In contrast, when rainfall has very high intensity, rainwater

will transfer partly to runoff. Thus, rainwater infiltration may also be insufficient to

reduce suction of the soil because intense rainfall is usually of shorter duration.

3

In recent decades here has been a change in the understanding of how rainfall-induced

shallow slope failure occurs. Initially, the mounding of groundwater tables in high

hydraulic conductivity soils and artesian uplift pressure in surface soils in low

hydraulic conductivity soils (Deere and Patton, 1971), were assumed to trigger

rainfall-induced slope failure, and were usually associated with deep-seated failure

mechanisms. These assumptions ignored matric suction above the water table. As a

result of investigations of slope failures in residual soils of Hong Kong for the period

of 1950 – 1973 (Lumb, 1975), rainwater infiltration was identified as the primary

cause of slope failures, rather than seepage from below. This is confirmed by the fact

that many residual soil slopes with deep groundwater tables and inclination angles

greater than the repose angle remain stable during the dry season, but fail when the

slopes are subject to prolonged intense rainfall. In these cases, the contribution of

matric suction to shear strength cannot be ignored (Fredlund and Rahardjo, 1993).

While matric suction increases the strength of unsaturated soils, this strength

decreases significantly as rainwater infiltrates the surficial soil of the slope.

A number of studies have confirmed that matric suction plays a key role in shallow

slope failures (Pradel and Raad, 1993; Rahardjo et al., 1995; Au, 1998; Fourie et al.,

1999). As rainwater infiltrates the slope surface, matric suction decreases and the

wetting front moves down until reaching a critical depth where the shear strength of

the soil cannot maintain slope stability (Fourie, 1996). This type of failure is more

likely to occur in slopes with relatively low hydraulic conductivity, than those with

high hydraulic conductivity, such as clean sands (Pradel and Raad, 1993). In the

former case, infiltration can only reduce suction of the surficial soils to shallow

depths, leading to a shallow slope failure mechanism.

Accepting that rainfall and soil properties are the controlling factors, coupled analyses

of seepage and slope stability are now commonly performed to evaluate rainfall-

induced slope instability. The deterministic approach is a common practice in both

seepage and slope stability analyses. In this approach, homogeneous (in terms of both

shear strength and hydraulic properties) slope soils are usually assumed, to simplify

the analysis problem. However, soils are rarely homogeneous in-situ and tend to be

spatially variable due to the changeable nature of soil formation (Vanmarcke, 1977a).

As a result, rainfall may produce infiltration rates at the site that are different from

simulations that assume soil homogeneity.

4

To overcome the limitations of slope stability assessments that assume soil

homogeneity, pre-defined failure plane and inter-slice forces, finite element and

probabilistic methods have been increasingly employed in slope stability analyses

(Fenton and Griffiths, 2005; Griffiths et al., 2011). With the finite element method, it

is possible to incorporate spatial variability of soil properties in seepage and stability

analysis of a slope. The probabilistic approach for slope stability analysis commonly

ignores spatial correlation of soil variability. Vanmarcke (1977a) indicated that in-

situ soil properties are inherently spatially variable. Christian (2004) suggested that

hydraulic conductivity was most variable (coefficient of variation, CV up to 767 %)

among engineering properties of soil. However, practicing engineers still rarely take

account of spatial variability of hydraulic conductivity, to avoid the increase of

complexity in the analysis due to nonlinearity of soil hydraulic properties, including

hydraulic conductivity. Moreover, other sources of soil variability can be minimised

by improved soil sampling, testing, and analysis. Spatial variability analysis may be

important for simulating in-situ soil properties, particularly hydraulic conductivity, to

evaluate the stability of slopes exposed to rainfall.

Due to the uncertainty of rainfall-induced failure of natural slopes, prediction of the

typical shallow slope failures is necessary to anticipate its consequences. Many

studies have established techniques for predicting landslide probability, starting from

traditional techniques (Vaughan, 1985; Nunes et al., 1989; Senanayaka et al., 1994) to

more quantitative approaches. Probably the most common method used to predict

rainfall-induced landslides in many different countries was the use of empirical

correlation between intensity and duration of rainfall leading to landslides (e.g. Caine,

1980; Kim et al., 1991; Larsen and Simon, 1993; Corominas et al., 2003; Guzzetti et

al., 2007; Dahal and Hasegawa, 2008). Another interesting approach to defining

rainfall threshold of landslide probability was illustrated by Crozier and Eyles (1980).

This approach was established based on empirical correlation between antecedent

conditions and a particular rainfall event leading to landslides. Both approaches rely

on landslide occurrences in the past and do not explicitly account for soil properties.

The significant role of both rainfall and soil properties was clearly indicated by Brand

et al. (1984) through a study on typical characteristics of slope failures in Hong Kong,

and Rahardjo et al. (2007) based on investigations of slope failures in Singapore.

5

It is now widely accepted that shallow slope failure mechanisms are triggered by

infiltration of rainwater to surficial soils as described previously. This mechanism has

been used to establish an approximate method to predict landslide probability based

on statistical rainfall data and soil properties (Pradel and Raad, 1993; Fourie, 1996).

However, this approximate method tends to produce conservative results because of

inherent simplifications.

The purpose of this chapter is to outline the importance of this project, the aim and

scope of this research, and provide an overview of the thesis.

1.2 OCCURRENCES OF SLOPE FAILURES TRIGGERED BY RAINFALL

Slope failures have become one of the most frequent natural hazards all over the

world, even recorded as the highest frequency in America for period of 1990-1999

(Alcantara-Ayala, 2002) among the other most frequent natural hazards such as

storms, volcano, earthquake, flood, and tsunami. In the period of 2003-2010,

worldwide rainfall-induced slope failures had generally shown an increasing trend

and 2010 was a bad year as shown in Figure 1.1. There were 6211 deaths recorded for

494 slope failures triggered by rainfall in 2010. The largest event in terms of lives lost

was the Gansu landslide in China on the 8th August, which killed 1765 people. Other

very large events were the 2nd March Bududa landslide, Uganda (358 deaths), the 6th

April Morrao de Bubma landslide in Niteroi, Brazil (196 deaths), the 7th August

debris flows in Leh, India (234 deaths); and the 4th October Wasior landslide in West

Papua, Indonesia (145 deaths).

6

Figure 1.1 Worldwide slope failures triggered by rainfall in 2003-2010 (after Petley,

2014)

In several regions, slope failures are commonplace and the occurrences occasionally

cause tremendous losses. For example, rainfall-induced slope failures in the Nepal

Himalaya region have caused huge damage to lives, property, infrastructure, and

environment particularly in the monsoon season (Dahal, 2012). The Nepal Himalayan

region is one of the most vulnerable zones of worldwide landslides, constituting about

30% of the world’s total landslide-related damage value (Li, 1990). A series of

landslides have occurred in this region with huge losses. For example, 50 people were

killed by landslides (in the half monsoon, 10 June – 15 August 2009) in Nepal. In

1988, a huge landslide at Darbang about 200 km west of Kathmandu, killed 109

people and temporarily blocked the Myagdi River. About 62 years before this

incident, a landslide had buried Darbang area, killing 500 people (Yagi et al., 1990).

This was the worst landslide disaster in the history of the Himalayan landslides.

Another landslide tragedy took place at Malpa Uttarakhand, India on 11 and 17

August 1998 resulting in the deaths of 380 people when massive landslides washed

away the entire village. Apart from such huge landslides, many small-scale landslides

were unreported when they occurred in remote areas of the Himalayas. Moreover, the

loss of productive lands in the hills due to landslides and related mass erosion

phenomena during rainy seasons, which are seldom reported unless they involve the

loss of life, seems to be so great that the economic loss, if quantified, would be no less

than that during any other big natural disasters. National infrastructures such as roads,

7

bridges, dams, hydropower stations, canals and buildings repeatedly suffer landslide

and flood damages. Similarly, due to a rapid increase in population over the

Himalayan hills in the last three decades, the landslides continuously cause

considerable loss of life, property, and significant damage to the vital economic

system of the nations in the Himalayan Region.

China is possibly another country with extremely serious geological disasters

including landslides triggered by rainfall. Every year, the direct economic losses of

geological disasters account for over 20% of the total losses from all natural disasters.

Nationwide, the landslide related direct and indirect economic losses account for

more than 20 billion Yuan (approximately 2 billion EUR) every year (Hu and Tang,

2005; Bai et al., 2011). According to the inventories of the China Institute of Geo-

Environment Monitoring, there were a total number of 102,804 geological disasters

nationwide in 2006, of which 86% were landslides. In 2007 there were 25,364 entries

nationwide, of which 61% were landslides. In 2008, 14,350 landslides were recorded

from a total number of 26,580 geological disasters, which accounts for 54%. In the

past 10 years, several large landslide disasters occurred. For example, the Gansu

landslide which took place on August 8, 2010, caused massive fatalities as mentioned

previously. These numbers underline the importance of disaster prevention and relief

for the reduction of economic losses. Therefore, landslide risk mapping and scientific

predictions are critical for disaster management agencies worldwide.

In Japan, many recurring rainfall-induced landslides occurred during heavy rains over

the last 65 years, resulting in a total of more than 1000 casualties over the last 65

years (Chigira, 2001). Such recurring disasters are possible because the weathered

granite had the potential for repeated landslides since the failures exposed rock

having low shear strength and the depth of weathering stages could be long-standing

erosion base levels (Durgin, 1977). Such fast weathering phenomena and repeated

failure on granitic terrain was also studied by Chigira and Ito (1999) on artificial cut

slopes in Japan. In 2004, very intense rainfall (the highest rainfall in the previous 30

years) triggered more than 300 landslides in Moriyuki and Monnyu catchment area,

Shikoku Island of Japan (Dahal et al., 2008). Field observations indicated that the

slides occurred mainly in residual soils on forested or partly forested slopes. Most of

the slides were shallow and translational in nature with the failure surface located

along the contact between overlying residual soil and relatively less weathered

8

bedrock at varying depths. Not only in Japan, but also in other granitic terrains of the

humid and tropical regions, shallow failure phenomena are very common. In granite

and gneiss areas of Rio de Janeiro in 1966 and 1967, severe rainstorms resulted in

tens of thousands of landslides and about 1000 casualties (Durgin, 1977). During the

main rainfall months of May to September in Hong Kong, numerous landslides occur

in cut and natural slopes of soils formed by the residual soils over granite and

granodiorite of Jurassic to cretaceous age (Irfan, 1998; Dai et al., 2003). Moreover,

two thirds of the land area of the Korean peninsula is composed of soils formed by

weathered products of granite and gneiss. During heavy rainfall, many slope failures

in these weathered rocks are characterized by relatively shallow failure surfaces

(typically 2-3 m in depth) that develop parallel to the original slope (Kim et al.,

2004). Southern Italy has also suffered from landslides in weathered granite

(Calcaterra et al., 1996). A great number of landslides (2560 events) during 55 years

(1950-2005) were compiled through a thorough literature search worldwide and the

dominant modes of the landslides were recorded as shallow landslide (52.8%) and

debris flow (42.2%) (Guzzetti et al., 2008). Therefore, this thesis focused on shallow

landslide mechanisms triggered by rainfall.

Farahmand and Aghakouchak (2013) indicated that landslides cause thousands of

casualties and billions of dollars in damages across the world every year. According

to the US Geological Survey (USGS), landslides result in tens of deaths and over 1-2

billion USD in property damages (USGS, 2006) annually. For example, the Western

US has suffered from several storm-triggered landslides during the El-Nino seasons

of 1982-1983, resulting in millions of dollars in loss (Spiker and Gori, 2003; Hong et

al., 2006b). In several other landslide events, thousands of people died and

disappeared within a few minutes/hours, e.g. 1999 landslide in Vargas, Venezuela

(Larsen et al., 2000). Landslides in South-east Asia are also one of the most

widespread disasters mainly because of the climate condition, mountainous terrain

and socioeconomic conditions (Apip et al., 2010). For instance, in 2006, after a period

of heavy rainfall, a series of landslides on Leyte Island, Philippines caused over 1000

fatalities (Sassa et al., 2010) and the 4th October 2010 Wasior landslide in West

Papua, Indonesia claimed 145 lives.

9

1.3 AIMS AND SCOPE OF THE RESEARCH

This thesis investigates the effect of the main controlling factors on the mechanisms

of rainfall-induced shallow slope failures. In particular, the effect of spatial variability

of soil hydraulic properties on the slope failure is taken into consideration to account

for the in-situ condition of natural slopes. Risk analysis was also carried out to

develop insight into how the likelihood of slope failures can be determined,

particularly the mechanism and consequences of rainfall-induced shallow slope

failure.

Numerical modelling was used to carry out coupled analyses of seepage and slope

stability using the commercially available software SVFLUX and SVSLOPE. The

finite element method was employed to incorporate complex analysis modelling for

more visual and accurate results. The spatial variability method was specifically

utilised to take account of the inherent soil variability closer to in-situ condition for

more realistic results. All the analyses were referred to a landslide occurrence in 2007

at the Jabiru site in the Northern Territory, Australia. This study highlighted the effect

of soil hydraulic properties as controlling factors on rainfall-induced shallow

landslides.

In order to achieve the overall aim of this research, research areas were summarized

as follows:

1. The first area of research uses the landslides that occurred at Jabiru in the

Northern Territory, Australia in 2007 to develop insights into how the main

controlling factors, particularly soil hydraulic properties, such as initial suction,

hydraulic conductivity, soil water characteristic curve, and unsaturated shear

strength properties, and rainfall events (in terms of intensity, duration, resolution,

and pattern) determine the shallow failure mechanism of rainfall-induced slopes.

Parametric studies were carried out to cover not only the specific case of the

Jabiru landslide but also general cases of rainfall-induced shallow slope failures

possibly occurring in any other sites.

2. The second area of research investigated the effect of spatial variability of soil

hydraulic conductivities on rainfall-induced shallow slope failure mechanisms.

Another, more accessible site, Sawyer’s Valley, which is near Perth, was chosen

10

for collecting data from field tests required to characterise the spatial variability of

hydraulic conductivities. Spatial variability parameters, such as correlation length,

indicating correlation levels (strong or weak) of soil properties with distance, and

coefficient of variation, indicating distribution of soil properties variation, were

used to model variability of soil hydraulic properties in the slope that are more

representative of the in-situ condition, and the importance of this factor in the

evaluation of rainfall-induced failures.

3. The last area of research in this thesis investigated approaches to predict the

likelihood of rainfall-induced shallow slope failures based on soil hydraulic

properties and antecedent conditions. New approaches for risk analysis were

developed to determine rainfall thresholds of the probability of slope failures

based on existing approaches. This analysis could be used as a screening tool for

the slope failure probability based on rainfall data and unsaturated soil mechanics

principles.

1.4 THESIS OUTLINE

This thesis focuses on three research areas as follows:

1. Numerical modelling of controlling factors in the analyses of rainfall-induced

slope failures.

2. Spatial variability analyses of rainfall-induced slope failures.

3. Prediction of rainfall-induced slope failures.

The structure of the thesis reflects the three primary topics above and it is presented in

seven chapters. In the current chapter, the background, aims, scope and outline of the

research are presented. The literature review is presented in Chapter 2 to provide a

background to subsequent chapters. The characteristics of tropical residual soils in

natural slopes, the contribution of the controlling factors on slope instability, failure

mechanisms, seepage and slope stability analyses of rainfall-induced slopes, and

approaches for predicting rainfall-induced shallow slope failures, are reviewed.

The research methodology is discussed in Chapter 3. Site characteristics are described

and general modelling for seepage and stability analysis of rainfall-induced slopes are

presented.

11

Chapter 4 examines the effect of the main controlling factors of rainfall-induced

shallow slope failures. Parametric studies regarding the effect of the controlling

factors on the slope instability, such as rainfall events (in terms of intensity, duration,

resolution, pattern, and antecedent event), hydraulic conductivity, soil water

characteristic curve, and unsaturated shear strength parameters, were carried out.

Deterministic analyses were performed and discussed with respect to the Jabiru

landslide to provide a better understanding of the mechanisms involved in rainfall-

induced shallow slope failure. Results obtained from deterministic analyses based on

the limit equilibrium method (the most common analysis of slope stability) in this

chapter are used as benchmark for those in the next chapters.

Spatial variability analysis was performed in Chapter 5 to investigate the effect of soil

properties, particularly hydraulic conductivity and randomly distributed spatial

variables, on slope instability. The spatial variability of the soil hydraulic

conductivities was applied based on soil parameters determined from field

investigations. The results of this chapter are compared with the results of

deterministic analysis method presented in the previous chapter.

Risk analysis was performed in Chapter 6 to examine the likelihood of slope failure.

This analysis provides invaluable information for taking actions including early

warning to avoid the consequences of a slope failure.

Finally, concluding remarks and recommendations for future studies are presented in

Chapter 7. All key points described previously in the main body of the thesis were

briefly discussed in this closing chapter. The chapter highlights the main points of the

three study areas and points to new questions reflected from results of the thesis for

further research.

1.5 PUBLICATIONS

Publications based on this thesis are as follows:

Suradi, M., Fourie, A., Beckett, C., and Buzzi, O. (2014). Rainfall-induced landslides:

development of a simple screening tool based on rainfall data and unsaturated soil

mechanics principles. Proceedings of the Sixth International Conference on

Unsaturated Soils, 1-4 July, Sydney, Australia, 1459-1465.

12

Suradi, M., and Fourie, A. (2014). The effect of rainfall patterns on the mechanisms

of shallow slope failure. Aceh International Journal of Science and Technology, 3(1):

1-18.

Suradi, M., Fourie, A., and Saynor, M.J. (2014). Rainfall-induced landslides: lessons

learned from an extreme rainfall event in northern Australia. Landslides (submitted).

13

CHAPTER 2

LITERATURE REVIEW

2.1 INTRODUCTION

Rainfall-induced slope failures have become a crucial issue in many countries all over

the world. Numerous failures have occurred in natural slopes, particularly those of the

residual soil-type, and shallow slope failure appears to be the most common

occurrence. Slope failure mechanisms appear to be related to engineering properties

that are inherent to residual soils (Fourie, 1997; Maail et al., 2004; Wesley, 2010).

Coupled analyses of seepage and slope stability are usually performed to evaluate the

level of risk regarding a slope failure. The deterministic approach, common practice in

seepage and slope stability analyses will be reviewed here as the theoretical basis for

the main analyses in the following chapters, and also used as a benchmark in this

study. The finite element approach is specifically considered in seepage analysis in

order to account for the spatial variability of hydraulic conductivity that is inherent to

the soil in natural slopes. Due to the high uncertainty of rainfall-induced failures in

natural slopes, many studies (Caine, 1980; Crozier and Eyles, 1980; Pradel and Raad,

1993; Fourie, 1996; Guzzetti et al., 2007) have attempted to predict landslide

probability, and therefore minimise the consequences. The theoretical background to

the above is reviewed in the following sections.

14

2.2 SLOPE FAILURE MECHANISMS

Slope failure mechanisms are governed by various controlling factors, and failure

occurs along the weakest paths in the slopes or more specifically the path where the

shear stress exceeds the shear strength. These paths may vary in their shape and

constitution, depending on the characteristics of the slope soil. These failure

mechanisms are briefly discussed below.

2.2.1 Characterising residual soil slopes in tropical regions

Residual soil is a term used to differentiate one type of soil from the more dominant

type, i.e. transported soil. Residual soil is created by the in-situ weathering and

decomposition of the soil’s original parent rock (Blight and Leong, 0 ) . Tropical

climates strongly influence the formation of residual soils (Morin and Ayetey, 1971;

Weinert, 1974), thus governing their characteristics. In tropical regions, the extremes

of alternation between intense rainfall and hot temperatures exert a rapid weathering

influence and cause leaching of the mobile constituents of the soil (Strakhov, 1967).

The combined effects of weathering and possible stress release from erosion can

expand and crack the weathered rock, producing small particles and clay minerals and

creating a system of interconnected voids. This condition makes residual soils both

more compressible and permeable to penetration by air and water than other types of

soil. As a result, residual soil slopes are susceptible to failure from prolonged heavy

rainfall, a typical scenario in the humid tropics. Unlike natural slopes, constructed soil

slopes such as embankments and reinforced soil slopes are designed to withstand

specified loads, usually include drainage, are compacted and essentially safer.

The groundwater table in residual soil slopes is often deep, located at depths of 5 m to

10 m below the slope surface (Blight and Leong, 2012), and is subject to fluctuations

from climatic effects (Wesley, 2010) as illustrated in Figure 2.1. In this situation, the

contribution of negative pore pressure or matric suction above the water table, in the

unsaturated soil zone, is significant to slope stability. The effects of unsaturated soils

should therefore be considered, along with slope stability, in geotechnical design.

There are many studies in relation to the effects of unsaturated conditions on soil

properties (Bishop and Blight, 1963; Blight, 1967; Fredlund and Morgestern, 1977;

Fredlund et al., 1978; Fredlund and Rahardjo, 1993). However, in the unsaturated

15

situations discussed here, the shear strength for saturated soils, which is usually

calculated utilising Equation 2.1 (below), (Terzaghi, 1950) does not apply.

Figure 2.1 Seasonal variation in water table and pore pressure due to climatic effects

(after Wesley, 2010)

τ = c′ + (σ – uw) tan ϕ′ (2.1)

where τ is shear strength, c′ is effective cohesion, σ is normal stress, uw is pore-water

pressure, and ϕ′ is the effective internal friction angle.

Unsaturated soils require additional parameters to calculate their shear strength, as

shown in Equation 2.2 (Fredlund and Rahardjo, 1993).

τ = c′ + (σ – ua) tan ϕ′ + (ua – uw) tan ϕb (2.2)

where (σ – ua) is net normal stress, ua is pore-air pressure, (ua – uw) is matric suction,

and ϕb is the angle indicating the rate of increase in shear strength relative to matric

suction.

The correlation between water content and matric suction provides significant data for

unsaturated soil characterisation. The curve illustrating this relationship is called a soil

water characteristic curve (SWCC), as shown in Figure 2.2. Errors or deviations in

laboratory tests may be accounted for by using the best-fit curve for the SWCC data,

16

as demonstrated in Equation 2.3 (below), (Fredlund and Xing, 1994). Significant

parameters associated with the SWCC are a, m and n, known as the SWCC

parameters. These parameters are usually used as inputs in geotechnical engineering

analysis, including slope stability analysis, when dealing with unsaturated soils.

Typical values of the parameters indicate types and characteristics of a soil, as

illustrated in Figure 2.2 (Fredlund and Xing, 1994). The SWCC indicates the water

storage capacity of a soil, and it can be used to determine the matric suction based on

the water content of the soil.

mn

s

ae

C

ln

)( (2.3)

where C(ψ) is a correction function, ranging from 1 for low suction to 0 for high

suction (ψ = 106 kPa), θs is the saturated volumetric water content, e is the natural

number (e = 2.71828), ψ is matric suction (kPa), and a, m, n are parameters

controlling the SWCC shape (indicating air-entry value, the shape near the air-entry

value, the slope of the SWCC and the residual water content, respectively).

Figure 2.2 Typical SWCC for different types of soil (after Fredlund and Xing, 1994)

17

The hydraulic conductivity of unsaturated soils varies with their degree of saturation.

The correlation between the degree of saturation, or water content, and unsaturated

hydraulic conductivity can be depicted using Equation 2.4, as proposed by Campbell

(1974), (illustrated in Figure 2.3). Soil types were indicated by f for fine-grained soils

with two soil parameters (a and ks) which directly affect the rainwater infiltration.

kw = (ks- kmin) (Θ)p + kmin (2.4)

where kw is the unsaturated hydraulic conductivity, ks is the saturated hydraulic

conductivity, kmin is the minimum hydraulic conductivity, Θ is the normalised

volumetric water content (= θw/θs), and p is the power factor for adjusting the

prediction (p = 4 is commonly used).

This correlation indicates that matric suction (ua – uw) is inversely proportional to

unsaturated hydraulic conductivity (kw). Hydraulic conductivity describes the ability

of a soil to transmit water through its voids.

Figure 2.3 Hydraulic conductivity function for unsaturated soils (after Rahardjo et al.,

2007)

18

Saturated hydraulic conductivity becomes a limiting value of the infiltration rate, as

illustrated in Figure 2.4. Initially, the infiltration rate in unsaturated soils is relatively

high, and then it decreases significantly as the degree of saturation increases up to the

lowest value in the saturated condition (Tholin and Kiefer, 1959).

Figure 2.4 Progress of infiltration through initially unsaturated soils during rainfall

(after Tholin and Kiefer, 1959)

Soils are naturally variable, due to the continuous and changeable nature of soil

formation. The inherent variation of soil properties from one point to another is not a

completely random process, rather it is spatially correlated, i.e., controlled by location

in space. The magnitude of soil properties at two close locations is likely to be

strongly correlated. This correlation weakens as the distance between the two

locations increases until no correlation can be made. Vanmarcke (1977a) suggested

that such spatial correlation should be considered in the modelling of soil properties.

It was revealed that the hydraulic properties of soil are the most variable, with its

coefficients of variation (CV) for hydraulic conductivity ranging from 27% to 767%

(Benson, 1993), while shear strength is the least variable, with a CV ≤ 45% obtained

from common in-situ tests (Kulhawy and Trautman, 1996).

0

1

2

3

4

5

6

0 0.5 1 1.5 2

Rat

io t

o in

filt

rati

on

cap

acit

y at

1 h

Time (h)

Saturated hydraulic conductivity

conductivity

19

It should be noted that tropical factors influence the characteristics of residual soil

slopes in relation to their vulnerability to shallow failure. Due to the deep