The Literature Review

Introduction

Landslides are defined as the downward movement of slope-forming materials under the influence of gravity (Highland & Bobrowsky, 2008) which can be considered as a high negatively impacting geological hazard on the socio-economic conditions (Baum & Godt, 2010; Petley, 2012). Rainfall is the most common triggering mechanism of landslides (Song, Chae, & Lee, 2016; Tu, Kwong, Dai, Tham, & Min, 2009), hence it is of prime importance to develop real-time warning systems to manage risk and minimize the impact of landslides triggered by rainfall. However, the state of knowledge and resources available to issue real-time warnings of rain-induced landslides varies across the globe. Existing warning systems against rainfall-induced landslides are based on rainfall threshold for an area or deformation of a slope, in which only the latter is based on the characteristics of a slope such as geometry, rainfall infiltration characteristics, and mechanical characteristics of the slope (Sasahara, 2017), which varies more spatially. Therefore this study set out to investigate the usefulness of such physical parameters in producing a real-time warning for rain-induced individual landslides.

Landslides

Downslope mass movement in a natural or man-made slope is referred as landslides which can either be rotational or translational slides. This includes several typical failure processes such as slides, falls, and flows for which the basic triggering mechanisms be rainfall, earthquakes, snowmelt, vibration, and differential weather conditions (Highland & Bobrowsky, 2008).

Rain-induced Landslides

Rainfall is regarded as a predominant triggering mechanism of landslides (Lee, Gofar, & Rahardjo, 2009). Rainfall precipitation leads to a reduction of matric suction and hence the shear strength of the soil is reduced increasing the possibility of slope failure (Li, Tham, Yue, Lee, & Law, 2005). Apart from that, a positive pore water pressure can be developed under heavy rainfall conditions, that may also result in further reduced shear strength thus triggers slope failures landslides (Lee, et al., 2009).

There is a number of prevention and mitigation methodologies have been established against rainfall-induced landslides. Even though retaining walls and ground anchors are used in the prevention of slope failures, the application is limited to large-scale slopes. However, the historical data claims that majority of landslides take place on a small-scale slope, where the application of mechanical reinforcements have lesser adequacy (Towhata & Uchimura, 2013). In such circumstances, “non-structural countermeasures, such as landslide monitoring and early warning” have been employed (Towhata & Uchimura, 2013).

Landslide Monitoring

Landslide monitoring is the fundamental step in landslide risk reduction, in which the existing and susceptible landslides are observed to produce data in order to forecast landslides using various parameters and techniques. Forecasting landslides involve in-depth analysis of all the monitored parameters

Most of the existing monitoring systems against rainfall-induced landslides are based on rainfall data, where rainfall thresholds have been defined on diverse geological and climatic conditions (Martelloni, Segoni, Fanti, & Catani, 2012). In such studies, Researchers attempted to evaluate the application of real-time rainfall data along with defined rainfall thresholds to produce landslide warnings, as in the La Honda, California (Wilson & Wieczorek, 1995) , Malaysian Peninsula (Lee, et al., 2009) and  Emilia Romagna, Italy (Martelloni, et al., 2012). Such existing literature is extensively focused particularly on developing algorithms for landslide warning based on real-time rainfall data and statistically defined rainfall thresholds.

For the statistical definition of thresholds Martelloni, et al. (2012) used cumulative rainfall data collected up to three days period for shallow ground movements while up to 240 days cumulative rainfall data for deep-seated movements.  The study involves the initial development of prototype thresholds, which were calibrated using “past georegistered and dated landslides” (Martelloni, et al., 2012). Even though the study develops simple and rapid operational early warning system for all types for landslides incorporating only the precipitation values as the input data, in Martelloni, et al. (2012) found that accuracy is limited as most such statistical models. Depleted accuracy due to lack of precision and integrity of archived historical landslide data, is not only often resulted in false warnings but also adversely influenced on calibration and validation of the model (Martelloni, et al., 2012).

Lee, et al. (2009) developed a model incorporating statistical analysis along with intrinsic soil properties to determine the critical rainfall pattern of failure for four soil types. The analysis results showed that there exist “a unique relationship between rainfall Intensity-duration-frequency (IDF) curve, hydraulic conductivity function and Soil-water characteristic curve(SWCC) as the minimum suction value and the corresponding water content in soil under an extreme rainfall of any duration can be predicted through these correlations. The minimum suction value is an important input parameter in the computation of unsaturated soil shear strength” (Lee, et al., 2009). Also, the study concludes that the ratio of rainfall intensity to soil saturated permeability is the deciding factor in the determination of critical rainfall pattern of failure.

There is a considerable amount of literature has been published on numerical models that simulated rain-induced landslides. Tsaparas, Rahardjo, Toll, and Leong (2002) performed a numerical simulation to investigate the responses of a typical residual soil slope in Singapore to several hydrological parameters including rainfall distribution, the saturated permeability of soil, initial pore-water pressures and the groundwater table. Griffiths and Lu (2005) carried out an unsaturated slope stability analysis with steady infiltration using elastoplastic finite elements. A range of infiltration and evaporation rates were applied on two homogeneous slopes that consist of clay and silt. All the studies above concluded that the appropriate choice of rainfall patterns to be used for the design of an unsaturated soil slope has to be determined by taking into consideration soil characteristics and climatic conditions. Lu and Godt (2008) developed an analytical framework for the stability of infinite slopes under steady unsaturated seepage conditions. Two types of soil were considered in their framework, namely sandy soil and silty soil. They found that hillslope failures can occur above the water table under steady infiltration conditions for both sandy and silty soils. The most notable contributions in their framework were the inclusions of the suction stress and the changes in soil friction angle with depth, which were often omitted from the conventional slope stability analysis. It appears that an appropriate numerical or analytical model for the study of rainfall-induced slope failure should account for both unsaturated soil behavior and rainfall pattern parameters.

The rainfall pattern, however, is highly variable relative to geographical location and climatic condition. Under such circumstance, the rainfall-induced slope failure was commonly treated as localized problems in which studies carried out from different geographical regions might suggest different conclusions on the threshold rainfall condition for the slope failures. The threshold rainfall a