Susceptibility assessment of landslides under extreme-rainfall events

Abstract. The influence of climate change on patterns has the potential to alter stability of partially saturated soil slopes. Changes in rainfall patterns have a strong influence on stability of partially saturated soil slopes, which recently have resulted in shallow landslides. In this paper, a comprehensive case study on the 2011 Umyeonsan (Mt.) landslides was highlighted. The incident involves the collapse of a soil slope and the debris flow under extreme-rainfall event, causing 16 fatalities and serious damaged to 146 housings. A fundamental study was carried out on the cause and mechanism of landslide/debris flow. An analytical method is developed for determining the failure mechanism of unsaturated soil slopes under extreme-rainfall, the effect of groundwater flow; the downward velocity of wetting front, and the upward velocity of groundwater level. Based on this, we propose the conceptual methodology of landslide design based on experimental tests and numerical analyses which consider the important mechanism of the combined effects of both groundwater flow and rainfall infiltration into the slope.


Introduction
Global climate change led to fluctuations in rainfall pattern.In South Korea, the frequency and intensity of heavy rainfall have increased during the summer seasons, and as a result to trigger more landslides, which are characterised by a relatively shallow failure of surfaces that develop parallel to the original slope (Jeong et al., 2008;Kim et al., 2012).
A record heavy rainfall hits the whole land area, Korea on June and July of 2011.Early in the morning of 27 July, about 150 small and large landslides occurred in 13 villages all over Umyeonsan (Mt.)where is located in downtown of Seoul metropolis with the height of 312.6 m a.s.l.Because the mountain is located at the centre of a dense residential area, the hazard had a great impact on the society compared to debris flows that occur in rural areas, and it led to careful scrutiny of the hazard area.Severe Introduction

Conclusions References
Tables Figures

Back Close
Full rainfall-induced landslides with slope failure and debris flow occurred in four villages of Raemian, Shindonga, Jeonwon, and Hyeongchon at Umyeonsan (Mt.).Some of the landslides hit the villages near the lower part of Umyeonsan (Mt.), causing the most casualties.Some landslides evolved from debris flows, flowing into the roads, victimizing the people passing the streets.The landslides rapidly expanded into a fast debris flow spreading throughout the narrow and sloping roads, invading many villages and roads.In many cases the debris flow overflowed into the outer part of Umyeonsan (Mt.).This paper aims to investigate the cause of the landslides of Umyeonsan (Mt.) 2011, and to propose the conceptual methodology of landslide design based on lab and field tests and numerical analyses which consider the important mechanism of groundwater flow and rainfall infiltration into the slope using a new approach on the source of rainfallinduced landslide.

Climate change
A broad consensus that the Earth's surface and oceans are warming at rates unprecedented in the historical record prevails in the scientific community.Global warming will also affect the hydrologic cycle by directly increasing evaporation of available surface water and vegetation transpiration (IPCC, 2007).It is affecting regional weather patterns, rainfall trends, and their variability.These changes potentially have a strong on stability of partially saturated soil slopes.These changes could trigger deformation of the Umyeonsan (Mt.) soil slopes resulting in slope instability within many mountainous areas of South Korea (Jeong et al., 2008).Hence, it is needed to investigate the influence of changes in trend and frequency of rainfall patterns observed through the rainfall data collected over a relatively long period on the stability analysis of partially saturated soil slopes.The quantitative increase and frequency change of recent rainfall patterns often cause shallow soil slope failures in comparison with past rainfall patterns.
In South Korea, global climate change which led to fluctuations in rainfall pattern like sub-tropical regions.Figure 1 shows 4 sites among 56 sites where continuous Introduction

Conclusions References
Tables Figures

Back Close
Full daily data are collected for peak rainfalls over 33 year period.Geochang, Yeongju, Inje, and Jecheon areas are the 4 selected mountainous sites in South Korea as indicated by circles in the right of Fig. 1a. Figure 1b demonstrate an increasing trend in rainfall intensity for these 4 sites.The results show annual daily maximum (in mm) for the whole observation period from 1973 to 2006.The plots illustrate that steadily increasing trend in rainfall intensity was observed throughout the 33 year period.
According to report from Korean Meteorological Administration (KMA), the seasonal intense rainfall on the study area, the Umyeonsan (Mt.), prolonged for two months (1105 mm), and 304 mm of heavy rain fell the last one day (KMA, 2011).It reported that the cumulative rainfall for 2 months before landslides event in 2011 was unprecedented in the last 10 years.
3 Umyeonsan (Mt.) slope hazard sites Umyeonsan (Mt.) lies at 37 • 28 2 N latitude and 127 • 0 25 E longitude with a height of 312.6 m.It has a main ridge formed from the northeast to the southwest, including six valleys perpendicular to the main ridge.Most of the mountain consists of lacustrine biotite gneiss.Two typhoons, i.e.Typhoon Meari (22-27 June) and Typhoon Muifa (28 July to 9 August) occurred in 2011, and the seasonal rain front remained in the middle of the Korean peninsula including the Umyeonsan (Mt.).They produced much more rainfall compared to the average rainfall and eventually led to landslides and debris flows around the Umyeonsan (Mt.).
A large amount of data, such as longitude/latitude and topographical/geotechnical characteristics in places where landslides had occurred, was collected by field surveys using a portable GPS, a laser ranger and a clinometer.Based on this, the coordinates regarding landslides were indicated on an aerial photograph of the area around Mt. Umyeonsan, as shown in Fig. 2a.The landslide locations are marked by coloured dots, and the debris flow routes are marked by sky-blue lines.Most of the debris flows resulted from landslides and flowed downward to the stream.They occurred Introduction

Conclusions References
Tables Figures

Back Close
Full simultaneously over the whole Umyeonsan (Mt.) area.Among these, two test sites were selected for landslide analysis in this study.The sites are the Raemian watershed (test site 1) and the Dukwooam watershed (test site 2) shown in Fig. 2b.The surface areas of the two watersheds are 128 700 and 98 700 m 2 , respectively.Under heavy rainfall, in test sites 1 and 2, 6 and 4 landslides (46 and 44 landslides per km 2 with an average of 34 landslides km 2 ) were generated with 685 and 49 m 2 (0.53 and 0.05 % with an average of 0.04 %), respectively.
4 Umyeonsan (Mt.) slope hazard sites The better way for identifying the sloping top soil behaviour is to collect the material properties of unsaturated soils through the laboratory and field tests.That is why a matric suction that reflects rainfall infiltration into soils is very important to analyse rainfallinduced landslides.Therefore, the volumetric water content is estimated as a function of the matric suction of soil by laboratory test and the initial matric suction is measured in the field.

The mathematics of development
The SWCC is can be used to derive permeability functions for use in unsaturated groundwater flow problems and to establish unsaturated shear strength parameters (Vanapalli et al., 1996).To obtain the SWCC of top soil on the Raemian watershed of Umyeonsan (Mt.), Geotechnical Consulting and Testing Systems (GCTS) pressure plate extractor (Pham and Fredlund, 2004) and filter paper test were performed.The GCTS pressure plate consists of two main parts which comprise a pressure chamber and a loading system.The pressure chamber can be subjected to extremely high air pressure (i.e. from 0.1 to 1500 kPa).Filter paper test as specified in ASTM D5298-10 (ASTM, 2012) was used for defining the SWCC for the condition of high matric suction Introduction

Conclusions References
Tables Figures

Back Close
Full (more than 1000 kPa of matric suction).The filter paper of Whatman No. 42 was used (Kim et al., 2004).The results of these tests are given in Table 1.According to the Unified Soil Classification System (USCS), the top soil can be classified as SM (silty sand).It is found that the tests done by GTCS pressure plate extractor showed a higher air-entry value, saturated-and residual volumetric water content.From this SWCC tests, the volumetric water-contents corresponding to matric suction by field measurement were adopted for calculating wetting band depth, infiltration depth and recharge depth in this study.

The chemistry of development
The values for the matric suction were measured at test site 1 from 29 June to the 19 July 2012.Three tensiometers depending on shallow-and middle-and deep-depths were installed and then measured matric suctions at the total of six different locations (T1-T6), as shown in Fig. 3a.Typically monitored results at T2 and T5 have been selected and are shown in Fig. 3b and c.The initial matric suction measured at the site on 29 June before the rainy season was approximately 75-85 kPa.After the ranfall began the matric suctions of 0.2-0.6 m depths rapidly decreased to 0-16 kPa and the matric suctions of 0.6-1.3m gradually decreased to 20 kPa.For the depth of 1.4 m, the matric suction did not change during the six days, since the infiltrated rainwater did not reach to this depth, whereas, it tends to decrease after the 6th day.Since then, all tensiometers were maintained approximately at 10-20 kPa without further additional rainfall.Therefore, in the landslide analysis, the matric suctions of 80 and 20 kPa were adopted as an initial matric suctions for dry-and unsaturated soils, respectively.

Proposed methodology: YS-slope model
The proposed model (referred to as YS-slope model) is to simulate the potential occurrences of shallow landslides on unsaturated soil by rainfall-infiltration, storage, Introduction

Conclusions References
Tables Figures

Back Close
Full recharge and groundwater flow processes.This physically-based prediction model was based on geographic information system (GIS) (Soller et al., 1999(Soller et al., , 2005)), all of data sets, topographic, soil depth, precipitation and geotechnical parameters were built to raster data of matrix structure.

Geotechnical model
Landslide analysis considers the infiltration by rainfall, which can be classified into three mechanisms: (1) a mechanism that considers the downward velocity of the wetting front, (2) a mechanism that considers the upward velocity of the groundwater level, and (3) a mechanism that considers both of these factors.YS-Slope model improves infinite slope model for the first or second case to interpret the landslide susceptibility for the third of the mechanisms.The Mohr-Coulomb criteria for simulating the unsaturated soil strength that was adopted in the stability analysis has been improved by considering the pore water pressure and pore-air pressure (Pradel and Raad, 1993).The uniform load from the vegetation and constant number of additional shear strengths from the roots of vegetation are considered as described by Hammond (Hammond et al., 1992).As shown in Fig. 4a, the safety factor (FS) for the infinite slope is calculated from the ratio of the resisting stress on a slip surface to the gravitationally induced downward slope driving stress: where c s is the cohesion of the soil, c r is the constant value related to additional shear strengths from the roots of trees, q 0 is the uniform load by trees, D w is the depth of the wetting band (= D wm + D wn ), D s is the depth of the unsaturated soil (= D mn ), γ t is the total unit weight of the soil, γ sat is the saturated unit weight of the soil, γ w is the unit weight of water, and β is the angle of the slope.Introduction

Conclusions References
Tables Figures

Back Close
Full

Hydrological model
The hydrologic model has been improved by incorporating the combined effects of groundwater flow and rainfall infiltration to the raster model (Soller et al., 1999;Smit, 2000).This model was developed by considering the unsaturated soil behaviour to estimate the rainfall-infiltration (I R ) and the recharge (R).The basic assumptions applied to the hydrologic model are that the aquifer is homogeneous, the recharge is non-uniform and varies over time, and the groundwater flows only below the groundwater table.
The soil profile is ideally subdivided into three zones, i.e. the wetting band zone, partially saturated zone, and fully saturated zone, as shown in Fig. 4b.The vertical infiltration of water from the surface into the ground is modelled in the wetting band and partially saturated zones, and the storage time is also considered.The storage time is defined as elapsed time at which rainwater infiltrates into the groundwater through the unsaturated zone.The horizontal movement of the groundwater is generated when the infiltrated water recharges into the fully saturated zone (Table 2).

Rainfall-infiltration model
Commonly used Green-Ampt model (Green and Ampt, 1911) for one-dimensional rainfall-infiltration was modified and used to calculate infiltration depth and recharge depth.The proposed model assumed that the volumetric water content and deficit the water content remain constant above or below the wetting front, as shown in Fig. 4b.The rainfall infiltration (I R ) could be represented as Eq. ( 2), and it is estimated by the trial and error method with a variable of cumulative infiltration.The ponding time (Eq.3) was considered to estimate the relationship between the rainfall-infiltration and runoff (Mein and Larson, 1973).The depth of the wetting front (D wn ) in the vadose zone is defined as the ratio of the rainfall-infiltration (I R ) and the deficit water content (∆θ) in Eq. ( 4): Introduction

Conclusions References
Tables Figures

Back Close
Full where I is the rainfall intensity, t p is the time for ponding, t w is the rainfall duration, K s is the saturated permeability, ψ f is the head of metric suction, F is the cumulative infiltration depth and ∆θ is the deficit water content.

Groundwater flow model
The groundwater flow was derived for a small elemental volume, for which the properties of the soil were assumed to be effectively constant.A mass balance was done on the water flowing in and out of this small volume, and the flux was calculated in terms of the head using constitutive equation of Darcy's law (Eq.5), which requires that the flow is slow (Fig. 5).Therefore, a form of the Laplace equation describes this steadystate flow of the groundwater as following Eq.( 14).The kinetic energy is neglected as described by Cedergren (1977).
where v is the seepage velocity, k is the permeability, i is hydraulic gradient, s is the distance through the soil, and the subscript x (or y, z) denotes the x (or y, z) direction.Introduction

Conclusions References
Tables Figures

Back Close
Full The variable h is the groundwater potential, equal to the sum of the kinetic energy, pressure and potential energy heads as in (Eq.7): The total volume of the flow leaving the element (Q x ) in the x direction for a unit time (∆t) can be expressed by assuming that the pore water pressure is constant over a small elementary volume of space.This relationship is expressed in Eq. ( 8): where sin β(= i = dh/ds) is the hydraulic gradient resulting from a difference in the groundwater potential across an element of the medium.To apply the raster model, the total volume of flow leaving the element (Q x ) in a cell can convert to the change in height (∆H) using the deficit water content (∆θ), and be expressed by where s is the distance between each cell (or size of a cell, = dx = dy), H w is the height of groundwater in a cell (= dz), and θ s is the volumetric water content.The total volume of flow leaving the element and its flow direction were estimated using the eight-flow method (Jenson and Domingue, 1988).The flow of groundwater was calculated using the slope and slope direction of the bedrock.Based on the assumption that the kinetic energy is negligible, the total volume of flow leaving a cell can only be affected by the neighbourhood cells.It noted that a vector analysis including the watershed and drainage network is not required.Therefore, the flow of groundwater can be calculated from a cell to a neighbourhood by shifting the change of groundwater Introduction

Conclusions References
Tables Figures

Back Close
Full height based on Eq. ( 9).The variations in the groundwater level in each cell over time are expressed as following Eqs.( 10) and ( 11): As mentioned above, the rainfall infiltration was completely recharged into the groundwater with the hydrometeorological model and calculated with Eq. ( 2).Equation ( 2) can be represented by Eq. ( 12) because of the recharge depth, which considers the storage time.Figure 6 shows the flow chart for groundwater flow, which considers the storage time in the YS-Slope.

Summary of the proposed methodology
We propose a conceptual methodology (Fig. 7) for the typical susceptibility assessment of shallow landslide based on hydro-geotechnical method.A computer program, YS-Slope has been developed using this methodology.The YS-Slope uses unsaturated soil parameters such as field matric suction and SWCC.Various types of rainfall events

Conclusions References
Tables Figures

Back Close
Full like real-time rainfall and probability rainfall are used.They were built to raster data of matrix structure.Using these data and DEM, the rainfall infiltration depth and recharge into the groundwater considering storage time were calculated for different times and locations of watershed.It was used for calculating factor of safety in second level.Finally, it can assess a rainfall-induced landslide depending on time varying by using groundwater table and wetting band depth.The case studies at the Raemian watershed and Dukwooam watershed described in this paper were conducted using this program.

Results and discussion
A GIS-based landslide analysis was performed to find out causing and failure mechanism by using the YS-Slope model.A 5 m × 5 m gridded DEM converted from the 2009 LiDAR data, soil data obtained from detailed investigations and field measurements, and precipitation gauged by the auto weather station (AWS) of KMA (KMA, 2011) were used as the input data.Landslide inventory was acquired from comprehensive investigation such as a field observation and an aerial light detection and ranging (LiDAR) surveying.The soil investigation was performed to obtain the detailed information of soil properties.The soil properties were estimated and are summarised in Table 3, some soil properties were adopted from the literature (Clausnitzer et al., 1998;Norris, 2008;Smit, 2000).The soil layer depth is one of the important parameter in verifying the cause of a disaster.To obtain the soil layer thickness at the study area, a krigging method was performed by using data from 14 boreholes, seismic prospecting, and a DEM. Figure 8a and b shows the result of the landslide analysis using the proposed model for the test site 1, and the landslides locations investigated in field are marked by black dots.The landslides and debris flows led to erode approximately 67 % of the watershed.The area of erosion from these results was compared with a map of safety factors obtained by using the proposed model, and the result overlapped in the erosion map of 49.7 % of the area.The area of 49.7 % is associated with debris flows.Initial material volume triggered by landslide flowed down the valley, and eroded subsurface of these area.Nevertheless, all 6 landslide observations are located within the resultant unstable area by the proposed model.Figure 8b shows the result of rainfall-induced landslide as the types of failure.At the point of L3, L4 and L6 in the landslide inventory, it was assessed that the landslides were occurred under the groundwater, and L1, L2 and L5 on wetting front.Thus, two types of landslide were occurred in this study area, and the series of landslide were triggered by subsurface infiltration and rising groundwater-table.Figure 8c and d shows the result of the analysis using the predictive models for the test site 2. In this test site, the results of analysis for the both mechanisms reasonably agree with all of landslide observations.Because this site has generally a shallow soil depth of 2.0-3.0 m, it is sufficiently shallow with relation to wetting band depth of 2.0 m in the test site.Thus, this area was exposed to the same level of risk of landslide, which was triggered by subsurface infiltration and rising groundwater-table .The results show the primary triggering factors causing landslides, i.e. soil depth, and spatial and temporal distributions of groundwater variations.It is shown that the YS-Slope model which takes into these factors account can be an intermediate approach between the detailed and simple analytical methods.

Conclusions
The primary object of this study was to investigate a landslide on Umyeonsan (Mt.) and to develop the conceptual methodology of landslide design.A comprehensive study was carried out on the cause and mechanism of landslide.The proposed model was used to assess the susceptibility of landslide for two study areas.The following conclusions can be drawn from the findings of this study: Instability of soil slopes can be predicted by observing changing matric suction distributions spatially and temporally over the soil slope cross section.We develop Figures

Back Close
Full the conceptual methodology of landslide design based on in-situ tests and numerical analyses which consider the important mechanism of groundwater flow and rainfallinfiltration.It is noted that the proposed model which takes into these factors account can be more suitable than previous models (Baum et al., 2008;Montgomery et al., 1994;Pack et al., 1998;Tarboton et al., 1997) due to its capability to model combing 1-D-infilatriaon process in unsaturation conditions and 2-D-groundwater flow in steadystate conditions.Thus, despite its simplicity, this model is capable of adopting in large scale analysis.
The proposed model has been improved by incorporating the combined effects of groundwater flow and rainfall infiltration into the raster model.These elements were integrated into the infinite slope-stability model for a rainfall-induced landslide.The special attention is given to the consideration of groundwater flow, rainfall water infiltration into ground and groundwater rising in ground.Existing models do not consider the three factors simultaneously.
The proposed model has clearly demonstrated that two types of rainfall induced landslide were covered: in the study area 1, the series of landslide were triggered by rainfall water infiltration into ground and groundwater rising in ground.In the other hand in the study area 2, two types of landslide mechanism had all good agreement with landslide observation.It is indicated that a case of shallow depth of soil like as the study area 2 should be exposed to the same level of risk of landslide, which was triggered by subsurface infiltration and rising groundwater-table.
Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Fig. 8a.The proposed model has a limitation to predict landslides in approximately Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Figure 8 .Figure 8 .
Figure 8. Landslide case on the Raemian and Dukwooam watersheds in 2011.a: comparison between erosion depth from LiDARs and stability by using physically-based model (test site 1), b: type of slope failure (test site 1), c: slope failure under the groundwater table (test site 2), d: failure on wetting front (test site 2)