Download as pdf or txt
Download as pdf or txt
You are on page 1of 38

GJ

Article
Vol. 21, No. 6, p. 10331070, December 2017
http://dx.doi.org/10.1007/s12303-017-0034-4
pISSN 1226-4806 eISSN 1598-7477 Geosciences Journal

Landslide prediction, monitoring and early warning: a


concise review of state-of-the-art
Byung-Gon Chae1, Hyuck-Jin Park2*, Filippo Catani3, Alessandro Simoni4, and Matteo Berti4
1
Planning and Coordination Division, Korea Institute of Geoscience and Mineral Resources, Daejeon 34131, Republic of Korea
2
Department of Geoinformation Engineering, Sejong University, Seoul 05006, Republic of Korea
3
Department of Earth Sciences, University of Florence, Firenze 50121, Italy
4
Department of Biological, Geological and Environmental Sciences, University of Bologna, Bologna 40126, Italy

ABSTRACT: Landslide is one of the repeated geological hazards during rainy season, which causes fatalities, damage to property
and economic losses in Korea. Landslides are responsible for at least 17% of all fatalities from natural hazards worldwide, and nearly
25% of annual casualties caused by natural hazards in Korea. Due to global climate change, the frequency of landslide occurrence
has been increased and subsequently, the losses and damages associated with landslides also have been increased. Therefore, accu-
rate prediction of landslide occurrence, and monitoring and early warning for ground movements are very important tasks to
reduce the damages and losses caused by landslides. Various studies on landslide prediction and reduction in landslide damage have
been performed and consequently, much of the recent progress has been in these areas. In particular, the application of information
and geospatial technologies such as remote sensing and geographic information systems (GIS) has greatly contributed to landslide
hazard assessment studies over recent years. In this paper, the recent advances and the state-of-the-art in the essential components
of the landslide hazard assessment, such as landslide susceptibility analysis, runout modeling, landslide monitoring and early warn-
ing, were reviewed. Especially, this paper focused on the evaluation of the landslide susceptibility using probabilistic approach and
physically based method, runout evaluation using volume based model and dynamic model, in situ ground based monitoring tech-
niques, remote sensing techniques for landslide monitoring, and landslide early warning using rainfall and physical thresholds.
Key words: landslide susceptibility, runout, landslide monitoring, early warning

Manuscript received May 7, 2017; Manuscript accepted July 7, 2017

1. INTRODUCTION metropolitan area. Various studies on landslide prediction and


reduction in landslide damage have been performed. In recent
Landslides represent a major threat to human life, properties, years, landslide hazard analysis and risk assessment have
infrastructures, and natural environments in most mountainous become a major subject in landslide studies and consequently,
regions of the world. Statistics from the Centre for Research on much of the recent progress has been in these areas. In particular,
Epidemiology of Disasters (CRED) (Brussels, Belgium) show the application of information and geospatial technologies such
that landslides are responsible for at least 17% of all fatalities as remote sensing and geographic information systems (GIS)
from natural hazards worldwide (Lacasse and Nadim, 2009). In has greatly contributed to landslide hazard assessment studies
the case of Korea, landslides cause the loss of 23 lives each year, over recent years. In addition, the importance of a quantitative
which accounts for nearly 25% of annual casualties caused by assessment for landslide hazard has been recognized, as it
natural disasters. In 2002, 75 people were killed in landslide- provides the basis for management decisions and mitigation
related incidents throughout the country, and in 2011, 18 people actions and the associated allocation of resources. Landslide
were killed by rainfall-induced landslides in the central Seoul hazard assessment, including hazard investigation and analysis,
refers to identifying and quantitatively describing the potential
*Corresponding author: landslide hazards, and consequently, evaluating the probability
Hyuck-Jin Park
Department of Geoinformation Engineering, Sejong University, Seoul of occurrence of landslides within a specified period of time.
05006, Republic of Korea Because landslide hazards have important spatial components
Tel: +82-2-3408-3965, Fax: +82-2-3408-4341, E-mail: hjpark@sejong.ac.kr related to the initiation of the hazard and the areas affected
The Association of Korean Geoscience Societies and Springer 2017 by landslides (van Westen et al., 2008), landslide initiation
1034 Byung-Gon Chae, Hyuck-Jin Park, Filippo Catani, Alessandro Simoni, and Matteo Berti

susceptibility analysis and runout modelling are the essential Landslide susceptibility assessment predicts the potential of
components of hazard assessment. Landslide initiation susceptibility landslide occurrence by considering landslide-inducing factors
analysis is the measurement of the probability of landslide of a and analysing the spatial distribution of these factors. Landslide
particular type being initiated at certain locations. In addition, susceptibility assessment methods can be classified into qualitative
assessing landslide runout (travel distance) is another essential (knowledge-driven) and quantitative (data-driven and physically
step for landslide hazard assessment in order to delineate based) methods, depending on the way they treat landslide-
potentially hazardous areas. inducing factors and models (Fig. 1).
Landslide monitoring is also a key component of most landslide Generally, knowledge-driven approaches are based entirely
hazard assessments, often with the objective of providing early on the judgement of the experts who conduct the susceptibility
warning of an impending failure where lives, communities, or assessment (van Westen et al., 1999). Therefore, knowledge-
infrastructure may be at risk (Eberhardt, 2012). Because we driven methods are seldom used for susceptibility assessment
cannot avoid landslide risk and must live with it, it is important over large areas because they lack a concrete physical concept of
to understand and predict landslide behaviour. The role of slope failure (Xie et al., 2004). Quantitative analyses can be divided
landslide monitoring and warning is to gather usable information into data-driven methods and physically based approaches. In
for avoiding or reducing the impact of landslide activity. After data-driven methods, the statistical relationships between the
recent catastrophic landslides around the world, landslide locations of landslides that have occurred in the past and
monitoring and especially early warning, have gained enormous landslide-inducing factors are evaluated, and then quantitative
interest. Therefore, in this paper, the essential components of predictions are made for landslide-free areas with similar
landslide hazard assessment, such as landslide susceptibility conditions. These methods constitute the so-called data-driven
analysis, runout modelling, landslide monitoring, and early approach because the data on past landslides is used to obtain
warning, will be discussed. Considerable advances have been information on the relative importance of each factor. This
made in landslide hazard assessment techniques in the last method assumes that conditions that have led to landslides in
decades, and this manuscript presents the progress and the the past will do so in future (Safeland, 2011). The three main
state-of-the-art of landslide hazard assessment methods. data-driven approaches commonly used are bivariate statistical
methods, multivariate methods, and artificial neural network
2. LANDSLIDE SUSCEPTIBILITY ANALYSIS analysis. In bivariate statistical analysis, each conditioning factor
such as slope, geology, or land use is combined with the
2.1. Landslide Initiation Susceptibility Assessment landslide occurrence locations, and weight values are calculated
for each parameter class. In multivariate statistical methods, the
The occurrence of landslides is controlled by various spatial combined relationship between a dependent variable (landslide
and climatic factors, such as geology, topography, hydrogeological occurrence) and a series of independent variables (conditioning
conditions, vegetation, and rainfall. This makes landslide factors such as slope, geology) are evaluated. An artificial neural
susceptibility assessment difficult, because diverse and large network offers a computational mechanism that is able to
amounts of spatial data must be acquired from the regional area acquire, represent, and compute a map from one multivariate
and considered in the analysis procedure. space of information to another given set of data representing

Fig. 1. Landslide susceptibility analysis.

http://dx.doi.org/10.1007/s12303-017-0034-4 http://www.springer.com/journal/12303
Landslide prediction, monitoring, and early warning: a concise review of state-of-the-art 1035

the relationships (Lu and Rosenbaum, 2003). An artificial neural infinite slope model has been used to analyse susceptibility in
network is trained by the use of a set of associated input and numerous previous studies of shallow landslides (Kamai, 1991;
output values. As the data-driven methods have been used Terlien, 1996; Pack et al., 1998; Frattini et al., 2004; Huang et al.,
widely, the details are not covered in this review. They are effective 2006; Rosso et al., 2006; Godt et al., 2008; D’Amato Avanzi et al.,
tools for evaluating various spatially distributed landslide- 2009; Apip et al., 2010; Griffiths et al., 2011; Santoso et al., 2011;
inducing factors over large areas. The most significant challenge Ho et al., 2012; Park et al., 2013; Ali et al., 2014; Alvioli et al.,
in using data-driven methods is the collection of data regarding 2014; Tsai et al., 2015). The infinite slope model has been used
the landslide distribution and factor maps over large areas (van previously in site-specific studies, but adoption of GIS methods
Westen, 2000). In addition, data-driven landslide susceptibility enables the use of this simple physical model for analysis and
assessment methods consider only the relationships between modelling of slope stability conditions over broad areas (Park et
landslides and related factors, not the failure mechanism (Park al., 2013).
et al., 2013). Moreover, the statistical models generally ignore The infinite slope model is based on a limit equilibrium
the temporal aspects of landslides, and are unable to predict the analysis that determines the balance between shear stress, which
impact of changes in landslide-controlling conditions (e.g., water is the forces tending towards movement along the supposed
table fluctuations and land use changes) (van Westen, 2004). failure plane, and shear strength, which is the forces resisting it,
Physically based landslide susceptibility assessment methods and calculates a factor of safety (FS). That is, assuming that the
are based on modelling the process of landslide occurrence groundwater is located at a distance from the failure surface and
using physical models of landslides. The methods estimate slope that groundwater flow is parallel to the slope (Fig. 2), FS can be
instability by taking geometrical and geotechnical characteristics expressed by Equation (1) (Coduto et al., 2010):
into account. Thus, unlike data-driven methods, the physically
c +  D –  w z w cos  tan
2
based approach can analyse slope stability regardless of landslide FS = -----------------------------------------------------------
-, (1)
 D sin cos 
occurrence by using physical slope models and on-site or
laboratory test results. The physically based method is a promising where is the depth of the failure plane, is the unit weight of the
approach for susceptibility analysis of shallow landslides because soil, is the unit weight of water, is the slope, is the effective soil
of its capacity to reproduce the physical processes governing cohesion, and is the effective friction angle.
landslide occurrence (Fell et al., 2008). Moreover, the general In this model, pore water pressure reduces the effective normal
grid-based structure and wide availability of GIS provide a stress and the shear strength of soil, ultimately causing slope
convenient framework allowing analysis over wide areas (Sorbino failure; thus, knowledge of the groundwater level is important
et al., 2010). Thus, physically based approaches were used to for predicting and preventing slope instability. It is practically
analyse landslide susceptibility over large areas using GIS (Luzi impossible, however, to measure the groundwater levels over an
and Pergalani, 1996; Zhou et al., 2003; Xie et al., 2004; Chen et extensive area. Therefore, previous studies have used a constant
al., 2005; Salciarini et al., 2006; Huang et al., 2007; Godt et al., or a randomly selected value for the groundwater level for the
2008; Liu and Wu, 2008; Ho et al., 2012; Park et al., 2013). entire study area (van Westen and Terlien, 1996; Zhou et al., 2003;
Recently, physically based models have been used widely because Xie et al., 2004; Griffiths et al., 2011). However, the groundwater
of their higher predictive capability and because they are the
most suitable for quantitative assessment of the effects of the
individual parameters that contribute to landslide initiation
(Corominas et al., 2014).
In the physically based approach, a physical slope model such
as an infinite slope model is used to determine the slope stability
by analysing the forces applied to the slope. The infinite slope
model (Fig. 1) is a simple but very useful model for shallow
sliding on a slip surface parallel to the slope of the ground. This
model assumes that landslides are infinitely long but have a
small depth compared with their length and width, and therefore,
this model is most appropriate for the analysis of shallow landslides
with planar failure surfaces. Because the failure surfaces for
rainfall-induced landslides are often shallow (the top few meters)
and are parallel to the ground surface (Lu and Godt, 2013), the Fig. 2. Infinite slope model.

http://www.springer.com/journal/12303 http://dx.doi.org/10.1007/s12303-017-0034-4
1036 Byung-Gon Chae, Hyuck-Jin Park, Filippo Catani, Alessandro Simoni, and Matteo Berti

level depends on the soil type, rainfall intensity, and hydraulic infiltration is in equilibrium with the steady-state, saturated
conductivity of the soil, so the use of a constant or randomly water flow parallel to the slope surface, above an impervious
selected value for a large area is inappropriate. boundary. For each cell, it considers the steady-state discharge
Physically based models generally couple hydrogeological as a function of the infiltration rate and a “contributing area”,
models to evaluate the effects of pore water pressure with the representing the upslope area that determines the subsurface
infinite slope stability model for the computation of the FS, flux through the considered cell. The steady-state discharge is
because hydrogeological models can evaluate an increase in combined with a general form for groundwater flow parallel to
pore water pressure, which may be caused by rainfall infiltration. the slope to estimate the relative water table depth and, as a
Consequently, physically based models coupled with a consequence, the relative pore water pressure. Because of its
hydrogeological model can predict distributed shallow landslide simple hydrogeological model for the generation of steady-state
initiation locations by evaluating a reduction in the shear strength pore water pressure, SHALSTAB has been used widely in
of the soil caused by increasing pore water pressure. numerous studies (Guimaraes et al., 2003; Fernandes et al., 2004;
The hydrogeological models can be classified into two Gorsevski et al., 2006; Huang and Kao, 2006; Rosso et al., 2006;
categories on the basis of the simplifying assumption: steady- Sorbino et al., 2010; Listo and Vieira, 2012; Zizioli et al., 2013;
state and transient-state models (Montgomery and Dietrich, Dal Sasso et al., 2014; Pradhan and Kim, 2015; Teixeira et al.,
1994; Terlien et al., 1995; Wu and Sidle, 1995; Pack et al., 1998; 2015; Kim et al., 2016).
Baum et al., 2002; Crosta and Frattini, 2003; Savage et al., 2004; SINMAP is based on the infinite slope stability model with
Godt et al., 2008). The commonly used hydrological model in groundwater pore pressures obtained from a topographically
slope stability analysis is the steady-state shallow subsurface based steady-state model of hydrogeology (Pack et al., 1998,
flow model that is described in TOPMODEL (Beven and Kirky, 2001). SINMAP derives its terrain stability classification from
1979) and TOPOG (O’Loughlin, 1986). The steady-state model topographic, hydrological, and soil characteristics. The input
assumes that rainfall infiltration is at its steady state, and saturated information (slope and specific catchment area) is obtained
water flows parallel to the slope surface (Montgomery and from the analysis of digital elevation models (DEMs). These
Dietrich, 1994; Wu and Sidle, 1995). Thus, this model assumes a parameters can be adjusted and calibrated with an interactive
uniform recharge state that simulates the spatial groundwater visual procedure that adjusts them based upon observed
level variation as a function of groundwater flow and rainfall landslides. SINMAP allows uncertainty in the input parameters
intensity over a long period rather than during a rainfall event. through the specification of lower and upper bounds that define
However, this model cannot evaluate the short-term temporal the uniform probability distributions. Between these boundaries,
variation of pore pressure and temporal changes in FS because the parameters are assumed to vary at random with respect to
of constant rainfall infiltration. Iverson (2000) attempted to the probability distribution. By assuming uniform distributions
overcome the restrictive assumptions of steady water table level of the parameters over uncertainty ranges, this model calculates
by introducing the infiltration process in an infinite slope using the stability index (SI), which is defined as the probability that a
an approximated form of Richards’ equation (Arnone et al., location is stable.
2011). The transient-state model performs transient seepage Both SINMAP and SHALSTAB predict that shallow landslides
analysis using the linearized solution of Richards’ equation (Iverson, will concentrate in areas of topographic convergence because
2000; Baum et al., 2002) and provides more realistic analysis the ratio of the contributing area to the cross-sectional width
results. However, one of the difficulties in applying the transient will be considered. Details, such as the flow-routing methods,
hydrologic model in physically based analysis approach is that the treatment of material properties, and use of results of the FS
transient hydrologic model requires abundant spatial information. calculations in identifying unstable slopes vary between the two
A large number of physically based models have been developed programmes, but their overall approach to assessing slope
in recent years. SHALSTAB (shallow landsliding stability model), stability is similar (Savage et al., 2004). SINMAP and SHALSTAB
SINMAP (stability index mapping) and TRIGRS (transient rainfall are useful for preliminary assessment of stability over broad areas
infiltration and grid-based regional slope stability analysis) are where the assumptions of the underlying models are adequately
the most widely distributed slope stability analysis methods satisfied. Both models consider a simple steady-state hydrogeological
based on the physically based model. SHALSTAB is a coupled process under constant rainfall. This means that both models
model that combines a stability model based on the infinite are used to predict a spatially distributed slope stability, but are
slope equation with a steady-state hydrological model, assuming limited to the temporal prediction of the slope stability because
that the subsurface flow is parallel to the slope (Beven and of the steady-state description of hydrological fluxes. In addition,
Kirkby, 1979; O’Loughlin, 1986). This model assumes that rainfall steady-state models are limited to few unrealistic situations

http://dx.doi.org/10.1007/s12303-017-0034-4 http://www.springer.com/journal/12303
Landslide prediction, monitoring, and early warning: a concise review of state-of-the-art 1037

related to both rainfall characteristics and in situ conditions SINMAP and SHALSTAB by mapping areas susceptible to
(Iverson, 2000). landslides in Brazil. The comparison between the different methods
The TRIGRS model is designed for modelling the potential aided in the identification of the most could be identified.
occurrence of shallow landslides by incorporating transient In recent years, several new physically based landslide
pressure responses to rainfall and downward infiltration (Baum susceptibility analysis methods were proposed. Arnone et al.
et al., 2008). TRIGRS combines an infinite slope stability calculation (2011) and Lepore et al. (2013) developed the triangulated irregular
and an analytic, one-dimensional solution for pore-pressure network (TIN)-based real-time integrated basin simulator (tRIBS)
diffusion in a soil layer of finite depth in response to time- model that allows the simulation of most spatial-temporal
varying rainfall (Baum et al., 2002; Savage et al., 2003, 2004). hydrologic processes (infiltration, evapotranspiration, groundwater
The infiltration models in TRIGRS are based on Iverson’s (2000) dynamics, and soil moisture conditions) that can influence
solution of Richards’ equation for an infinitely deep impervious landslide occurrence. This model considers the spatial variability
basal boundary and the surface condition of constant flux for a in precipitation fields, land-surface descriptors, and the corresponding
specified time and zero flux thereafter (Salciarini et al., 2006). moisture to account for the effects of heterogeneous and
Iverson (2000) provides a theoretical framework for understanding anisotropic soil. Montrasio (2000) and Montrasio and Valentino
how hydrologic processes affect the location, timing, and rate of (2008) proposed a SLIP model that translates the physical
landslide occurrence, based on a solution to the boundary problem phenomenon of the rainfall-induced process and focuses on the
posed by Richards’ equation. Baum et al. (2002) generalized destabilizing forces caused by the water downflow and the
Iverson’s original infiltration model solution for cases of unsaturated contribution of partial saturation to the shear strength of the soil
conditions and variable rainfall intensity and duration to develop (Montrasio et al., 2011). In this model, the partial saturation
the TRIGRS software package. TRIGRS has been adopted to contribution to the soil shear strength is included using the total
evaluate the spatio-temporal prediction of landslide occurrence soil cohesion. Total cohesion includes the effective and apparent
in many studies (Baum et al., 2005; Salciarini et al., 2006; Sorbino cohesion related to the matrix suction as proposed by Fredlund
et al., 2007; Godt et al., 2008; Vieira et al., 2010; Liao et al., 2011; et al. (1996). Montrasio and Valentino (2008) approximated the
Zizioli et al., 2013; Raia et al., 2014; Bordoni et al., 2015b; Lee suction-related cohesion as a mathematical function of the
and Park, 2016). degree of saturation based on experimental data. Chae et al.
Transient models are able to improve the effectiveness of (2015) proposed the modified infinite slope model based on the
susceptibility analysis, accounting for the transient effects of concept of the saturation depth ratio to analyse the slope stability
varying rainfall on slope stability conditions, but they generally change associated with rainfall on a slope. A rainfall infiltration
need abundant and accurate spatial information (Sorbino et al., test in unsaturated soil was performed using a column to
2010). Moreover, they are sensitive to some required input data develop an understanding of the effect of the saturation depth
such as hydraulic properties of soils, initial steady-state groundwater ratio following rainfall infiltration. The proposed model was applied
conditions, and soil depths, whose correct evaluation is often to practical examples to assess its feasibility and to develop a
possible only using empirical models or inverse deterministic regional landslide susceptibility map using a GIS.
analyses (Salciarini et al., 2006; Sorbino et al., 2007; Godt et al., In the physically based model, sufficient and accurate information
2008). Because different physically based models have been is required to obtain an accurate landslide susceptibility assessment.
suggested and used to obtain shallow landslide susceptibility Therefore, the quality and quantity of information for the input
maps, comparative analyses using different physical slope models parameters have been the main concerns in landslide susceptibility
were performed. Meisina and Scarabelli (2007) applied the analysis. This is because in practice, such data are often limited
SINMAP and SHALSTAB models for slope stability analysis in extent and have imperfections or variable quality (van Westen
and compared the results of these analyses with the locations of et al., 2006). In particular, the strength parameters of slope
the shallow landslides that occurred on November 2002 in the materials, such as cohesion and friction angle, are inherently
area of Oltrepo Pavese (Northern Apennines), Italy. Sorbino et spatially heterogeneous, as the slope material is produced by a
al. (2010) applied SHALSTAB and TRIGRS to analyse the natural process (Baecher and Christian, 2003; Carrara et al.,
source areas of huge rainfall-induced, shallow landslides that 2008; Chowdhury et al., 2010). In addition, the input parameters
occurred in May 1998 in the Campania Region, Italy. Zizioli et used in the physically based approach should be obtained from
al. (2013) used SINMAP, SHALSTAB and TRIGRS to obtain a very wide study area, likely with limited sampling; thus,
landslide susceptibility maps of an area in the northern Apennines uncertainties are inevitably involved in the physically based
where landslides are the cause of a large amount of damage to model analysis. Uncertainty in model parameter evaluation has
infrastructure and agriculture. Michel et al. (2014) compared been recognized as an important cause of mismatch between

http://www.springer.com/journal/12303 http://dx.doi.org/10.1007/s12303-017-0034-4
1038 Byung-Gon Chae, Hyuck-Jin Park, Filippo Catani, Alessandro Simoni, and Matteo Berti

simulated and observed distributions of landslide occurrence in an explicit form (Baecher and Christian, 2003). The detailed
(Burton et al., 1998). Therefore, when performing susceptibility explanations for the different probabilistic analysis are provided
analysis with physically based models, spatial variability and in several studies (Harr, 1987; Baecher and Christian, 2003;
uncertainties in ground conditions must be taken into account. Chowdhury et al., 2010).
However, most previous physically based model studies have As many reports have pointed out, probabilistic analysis has
used a deterministic analysis to estimate the potential or relative been adopted in site-specific slope stability analyses at hillslope
instability of slopes, over a large area, without considering the scale (Christian et al., 1994; Gokceoglu et al., 2000; Park and
uncertainties in the input parameters. The deterministic analysis West, 2001; El-Ramly et al., 2002; Park et al., 2005; Zhang et al.,
is not appropriate for considering uncertainties and variability 2005; Cho, 2009; Zhu et al., 2013; Ali et al., 2014; Li et al., 2014;
because only a single fixed value is given for an uncertain Dou et al., 2014). Recently, the adoption of GIS methods has
parameter. In particular, applying the deterministic approach to enabled the use of the probabilistic approach for spatially distributed
an extensive study area can be particularly difficult or impossible analysis and physically based modelling of landslide susceptibility
because of the uncertainties and difficulties in obtaining, checking, over broad areas. Santoso et al. (2011) applied the probabilistic
and processing large spatial data sets (Zhou et al., 2003). The analysis approach in the physically based model approach, but
variability and uncertainty can be quantitatively accounted for the strength parameters were not considered as random variables.
by the application of probabilistic analyses. Therefore, probabilistic Some studies used probabilistic analysis approaches in physically
analysis has been recognized as a formidable tool for dealing based model analyses, but the hydrogeological model was not
with uncertainty. adopted, and subsequently a constant groundwater level was
In probabilistic analysis, strength parameters should be assumed for the entire study area in these analyses (Zhou et al.,
considered as random variables to account for the uncertainties 2003; Shou and Chen, 2005; Shou et al., 2009). Other researchers
involved in their determination. In addition, the statistical have modelled the uncertainties in input parameters using
parameters (such as mean and standard deviation) and probability uniformly distributed strength parameters with upper and lower
density function (PDF) of uncertain variables are determined bounds, which is the simplest form of probabilistic analysis
from the available field or laboratory data. Then, using the (Dietrich et al., 2001; Meisina and Scarabelli 2007; Terhorst and
statistical parameters and the PDF of uncertain parameters with Kreja 2009; Yilmaz and Keskin 2009).
the performance function (i.e., a physically based model in this More complete probabilistic approaches for spatially distributed
study), the probabilistic analysis is carried out. and physically based landslide susceptibility analysis have been
There are three most commonly used probabilistic analysis proposed lately. Raia et al. (2014) proposed a probabilistic version
methods: the first-order second-moment method (FOSM), the of the TRIGRS code using Monte Carlo simulation. In this study,
point estimate method (PEM), and Monte Carlo simulations. the slope materials’ properties, such as strength and hydraulic
FOSM and PEM offer the advantage of allowing approximate parameters, are considered as uniformly distributed random
estimation of the probability of failure using only the means and variables. Then the proposed approach was applied to a test area
standard deviations even when additional information regarding and the results of the probabilistic and deterministic analyses
the random variables is lacking. However, the calculations become were compared. The study concluded that the predictive power
impossible when the performance functions are complex. of the probabilistic analysis was improved by about 10% over the
Moreover, because these methods can only be used to obtain deterministic analysis. Rossi et al. (2013) proposed a physically
approximate values for the probability of failure, they cannot be based slope stability simulator with Monte Carlo simulation,
used to determine the distributions for FSs, which are also called HIRESSS (high resolution slope stability simulator). In
treated as random variables (Harr 1987; Park and West 2001). In this study, the probabilistic approach was used to manage
contrast, Monte Carlo simulations are one of the most widely uncertainties of the typical geotechnical parameters, which is a
used methods of probabilistic analysis that in principle can be common weak point of the deterministic model. The proposed
applied to all models for which deterministic analysis is possible. simulator was applied to three test areas, and the results showed
Monte Carlo simulations are considered as the most complete good reliability with managing very uncertain input data on a
probabilistic analysis method because all random variables and large area and at a large scale. Park et al. (2013) and Lee and Park
the probability of failure that results from the reliability analysis (2016) adopted the Monte Carlo simulation as the probabilistic
are represented by their PDFs through repeated calculations approach in physically based and spatially distributed landslide
(Park et al., 2013). Monte Carlo simulations are relatively easy to susceptibility analysis coupled with the hydrological infiltration
implement on a computer and can accommodate a wide range model. The proposed model was applied to study areas that
of functions, including those that cannot be expressed conveniently experienced a large number of landslides and the results of the

http://dx.doi.org/10.1007/s12303-017-0034-4 http://www.springer.com/journal/12303
Landslide prediction, monitoring, and early warning: a concise review of state-of-the-art 1039

probabilistic analysis showed superior performance over the may become extremely rapid following sudden failure. In general,
deterministic analysis. when the rock mass become fragmented the movement may
become flow-like, forming extremely rapid rock avalanches
2.2. Landslide Runout (Melosh, 1987). Hungr et al. (2005) describe the typical failure
behavior for various types of landslides that can be used as a
Landslide runout is generally intended as the overall distance guideline.
covered by a landslide in the case of a catastrophic failure Whatever the material involved, rapid gravitational processes
(McDougall, 2016). It may refer to the entire horizontal path are frequent and are the most dangerous type of landslides. The
length measured on a map (travel distance) or to distance measured ability to predict their travel distance, and possibly other parameters
downstream a defined point (runout length) (Rickenmann, like velocity or peak discharge, entails substantial advantages for
2005). In both cases, it indicates the ability of a landslide to travel hazard assessment and for the design of possible mitigation
much further than expected from simple frictional models measures.
(Corominas, 1996; Davies T.R. and McSaveney, 2012; Johnson Several methods have been developed to analyze the travel
et al., 2016). Although many landslide in many different materials distance and velocities of fast-moving landslides. They range
can exhibit a flow-like behavior and move over long distance from empirical-statistical methods to dynamic methods (Rickenmann,
(Hungr et al., 2001), the term “runout” is commonly used to 2005) that are continuum based models allowing the simulation
describe the mobility of fast landslides that travel at extremely of the flow and related parameters along the slope, including
rapid velocity (> 5 m/s; Cruden and Varnes, 1996). Fast landslides deposition.
such as rockfalls, debris flows and rock/soil avalanches show a Since early attempts to constrain the relationship between
sudden acceleration of the initial failure leading to high travel distance (L) of landslides and the elevation difference
displacements. Because of their speed and low predictability, between the starting point and the ending point of deposition
fast landslides have strong destructive effects and can cause a (H), it appeared clear that the volume of the mass movement (V)
large number of causalities (Guzzetti, 2000; Petley, 2012). The significantly influence its mobility (Corominas, 1996; Hunter
prediction of landslide runout is therefore essential for all levels and Fell, 2003; Legros, 2002; Rickenmann, 1999). A plot of the
of hazard and risk zoning (Dai et al., 2002; Glade et al., 2005). tangent of the travel angle (H/L) against landslide volume
How do we recognize whether a given potential landslide can generally shows that large landslides exhibit lower travel angles
become extremely rapid? Mostly based on experience and than smaller landslides (Fig. 3). The large scatter usually associated
comparison with precedents. Most shallow slides occurring on to such relationships derives from differences in material properties
steep slopes can become extremely rapid. They usually involve and motion mechanisms, and morphology of the deposition
loose granular regolith overlying stable substrate. Such failures areas or presence of obstacles. Their application has to consider
nearly always begin during heavy rain, ensuring perched saturation an homogeneous set of data that is representative of the specific
of the loose layer. As fast movement occurs, soil situated landslide type (Corominas, 1996; Hurlimann et al., 2015; Rickenmann,
downslope of the initial failure is over-ridden, liquefied by rapid 1999). However the use of mean values can give optimistic
undrained loading and incorporated in a growing debris results and it is advisable to use the lower envelope or the line
avalanche (Sassa, 1985). Debris mobilization may also occur along corresponding to a certain confidence level (e.g., 90%) (Hungr
established steep stream channels whenever loose sediment is et al., 2005).
abundant (Berti and Simoni, 2005, Kean, 2013). The flowing Most empirical methods are simple enough to be implemented
mixture of water and debris entrain further material and turn in GIS to delineate the extent of potentially affected areas (Toyos
into a surging, extremely rapid debris flow. Debris flows are et al., 2007) for preliminary hazard mapping purposes. In the
often recurrent phenomena that contribute to build a fan with case of debris flows, Zimmermann (1997) defined a lower envelope
their deposits. of tan (H/L) as a function of the catchment area, that can be
Soils dominated by fine clayey material are usually involved in used for preliminary assessments when volume estimates are
slides that do not evolve in faster than rapid (3 m/min) movement. lacking. Alternatively, Prochaska et al. (2008) propose to use the
Extra-sensitive (“quick”) clay flow slides are a notable and well- average channel slope to make preliminary runout predictions
known exception (Ter-Stepanian, 2000). However, also clay for small and medium-sized debris flows that occur along
slopes may produce shallow failures evolving in extremely rapid confined channels.
mudflows when mixing with surface water is sufficiently rapid Recognizing the importance of sediment entrainment for
(Hungr et al., 2001). debris flows, Fannin and Wise (2001) proposed to use a sediment
Rock slides, rock falls and rock topples in good-quality rock budget along the flow path to estimate the total travel distance.

http://www.springer.com/journal/12303 http://dx.doi.org/10.1007/s12303-017-0034-4
1040 Byung-Gon Chae, Hyuck-Jin Park, Filippo Catani, Alessandro Simoni, and Matteo Berti

Fig. 3. Relationship between landslide


mobility (H/L) and volume as described
by various authors for different types of
mass movements. *H ranging from 200
to 2000 m.

Their method can be used regardless of the knowledge of the empirical relationships describing the scaling relationships
expected debris flow volume and heavily relies on the quality of between volume V, B, and A are adapted. Many researchers have
field data (Miller and Burnett, 2008). demonstrated its applicability to debris flows (Crosta et al., 2003;
The approach proposed for lahars by Iverson et al. (1998) Berti and Simoni, 2007; D’Agostino et al., 2008; Simoni et al.,
correlates event volume (V) with inundated area (B) and cross- 2011) and other phenomena such as rock avalanches (Griswold
sectional flow area (A) (Fig. 4). It is based on scaling arguments and Iverson, 2008). Some useful modification have been proposed
and statistical analysis of past events data and can be used for to enhance the capabilities of the method or to address specific
different types of fast-moving landslides, provided the semi- situations. Scheidl and Rickenmann (2010) added a module to
their TopRunDF for automated prediction of the flow path on
the fan while expressing the mobility coefficient (constant
describing the V-B relationship) as a function of the gradient of
the deposition area. The prediction model DFLOWZ (Berti and
Simoni, 2014) allows to simulate the propagation of both
confined and unconfined flows and to take into account the
uncertainties associated to the scaling relationships and their
impact on results. Griswold and Iverson (2008) complemented
the GIS tool LAHARZ (Schilling, 2014) with a module for
delineating possible debris flow source areas and, recently, Reid
et al. (2016) embedded into the same method the effects of
debris-flow growth along the channel due to entrainment.
Empirical, volume-based models easily allow the likelihood
of different debris flow runout lengths or inundated areas to be
estimated through frequency-magnitude relationships. While it
would be beneficial to map hazard zones as a function of event
volume, difficulties can arise in the estimation of a probable range
of flow volumes for a given channel or a given area (Corominas
et al., 2014; Jakob and Friele, 2010; Jakob and Jordan, 2001; Riley
et al., 2013; Stoffel, 2010). However, due to their simplicity,
Fig. 4. Definition of empirical scaling relationships describing the relation speed and ease of application, empirical methods maintain their
between event volume (V), inundated area (B) and cross-sectional relevance and can be conveniently used for preliminary hazard
flow area (A). Mobility coefficients ka and kb can be obtained based
on statistical analysis of past events. assessment.

http://dx.doi.org/10.1007/s12303-017-0034-4 http://www.springer.com/journal/12303
Landslide prediction, monitoring, and early warning: a concise review of state-of-the-art 1041

Dynamic methods consider mass, momentum, and energy events that take place within the same catchment (Kean et al.,
conservation of a debris flow to simulate the propagation of the 2013; Marchi et al., 2002; Navratil et al., 2013). Rheological
flow using one-, two-, or three-dimensional models (O’Brien et parameters derived from back-analysis of documented past
al., 1993; Hungr, 1995; Laigle et al., 2003; McDougall and Hungr, events are commonly used for prediction purposes (Jakob et al.,
2003). Most methods treat the debris flow as a continuum 2013; Hurlimann et al., 2008; Pirulli, 2010). Though such
assigning simple constitutive relationships that are capable to strategy aims to effectively address the choice of rheological
reproduce the macroscopic behaviour of a given material. Several parameters, it necessarily adds uncertainty to the results because
single-phase rheological models have been proposed for simulating of the inter-event variability mentioned above. To date, no
the behavior of a propagating gravitational flow despite this systematic effort has been done to constrain the ranges of
latter being almost invariably a non homogeneous and multiphase rheological parameters as a function of the many variables that
mixture (Iverson, 2007; McArdell et al., 2007; Wang et al., 2008). influence the flow behavior (e.g., volume, grain size, solid
More rigorous, physically-based multi-phase models have been concentration), therefore back-analysis still appears as the only
proposed (Iverson and Denlinger, 2001) but their application is viable option.
still limited (Liu et al., 2016) due to complexity and abundance In general, the predictive use of dynamic methods requires
of required parameters. Determining the rheological behavior knowledge of the physical phenomena to be reproduced and a
of natural materials remains difficult because they encompass good understanding of the rheological model. Similarly to
coarse, irregular particles over a very wide range of size and a empirical methods, the design volume or, better, the volume-
variable proportion of water (Ancey, 2007). Probably we will frequency relationship is required to associate hazard values to
assist to the progression of multi-phase models in the future but, simulation results of a given catchment (Jakob, 2005). Alternatively,
at the time of writing, mono-phase rheological models are the erosion/deposition dynamics of the flow have to be taken
largely dominant in practical application and we will limit to into account. Attempts to simulate the entrainment process
them our concise treatment. have been made by Hungr and Evans (2004) by introducing a
Dynamic methods commonly describe gravitational flows as user-specified quantity of entrainment (e.g., bed-normal depth
a non-Newtonian fluid moving in a three dimensional terrain. eroded per unit flow and unit displacement) along the path, and
The flow resistance term is often described as a Bingham fluid other examples recently appeared in the literature (Hussin et al.,
or, more generally, as a Coulomb viscous (Johnson, 1984) or 2012; Frank et al., 2015). When including entrainment, models
Herschel-Bulkley fluid (Coussot, 1997). These models are become more difficult to constrain, as the entrainment area and
theoretically more appropriate for relatively fine-grained “mud the entrainment depth, become additional input parameters.
flows” (Costa, 1984) as opposed to “stony debris flows” (Takahashi, Field-based geological observations can be used to estimate
1991) that require the dilatant or inertial grain shearing model these parameters, but more work is required to effectively
(Bagnold, 1954; Takashi, 1991) to take into account the grain simulate the entrainment process with dynamic methods (van
collisions that dominate the flow behavior. Intermediate models Asch et al., 2008).
attempting to describe the combined effect of viscous, inertial Amongst the most widely applied dynamic models, FLOW-
and turbulent flow regimes also exist (O’Brien et al., 1993). 2D (3-D integrated Eulerian model) is designed for the analysis
Despite being originally developed for snow avalanches, the of debris flows and mud flows on colluvial fans (Sosio et al.,
Voellmy frictional model has been successfully and widely 2007; Marchi et al., 2010; Quan Luna et al., 2014). The model is
applied to debris flows (Revellino et al., 2004; Naef et al., 2015). based on the quadratic rheological approach as described earlier
It includes a Coulomb-type friction term which scales with the in this section (O’Brien et al., 1993). The inflow volume is
normal stress, and a turbulent drag coefficient that scales with described by an entering water hydrograph coupled to a time
the velocity squared. dependent sediment-concentration graph. The Manning coefficient
However, given that mono-phase models do not describe the should be assigned to each grid element to account for the hydraulic
physics of flowing materials, and that full-scale direct measurements roughness of the terrain surface to account for differences in
is impossible, the parameters of a given rheological model are surface coverage.
usually determined by back-analysis of events that are similar to Several other debris-flow simulation models were applied in
the case under consideration (Sosio et al., 2007). This is a major case studies and compared to real debris flows. They include
limitation of dynamic methods in that actual debris flows or DAN-3D (Hungr, 1995; McDougall and Hungr, 2004; 2005),
mud flows exhibit considerable variability in behavior that occurs RAMMS (Christen et al., 2010; 2012; Hussin et al., 2012), RASH-3D
during successive surges of a single event (Berti et al., 2000; (Pirulli and Sorbino, 2008), and TRENT-2D (Armanini et al.,
McCoy et al., 2010) and, even more so, between the different 2009). Dynamic numerical models can be used to investigate

http://www.springer.com/journal/12303 http://dx.doi.org/10.1007/s12303-017-0034-4
1042 Byung-Gon Chae, Hyuck-Jin Park, Filippo Catani, Alessandro Simoni, and Matteo Berti

runout distance and inundation patterns as well as flow heights There are three major categories of landslide monitoring. The
and flow velocities. Many authors agree on the importance of an first one is field observations of changing features of topography,
accurate digital elevation model of the potential depositional cracks, and water flow on the surface. This is a conventional
area given that the runout pattern is largely determined by the monitoring method in which geologists measure the changing
local topography (Hurlimann et al., 2008; Rickenmann, 2016). features at regular time intervals in the field. However, the
Comparative studies of the performance of simulation models method has a limitation in providing information on impending
include those made by Rickenmann et al. (2006) and Schraml et landslides or slope failures, because measuring changes at the
al. (2015). Making generalizations is not possible due to the site in a short time interval such as a minute or an hour scale is
wide range of physical processes that can be simulated. A impossible. The second category is an in situ ground-based
simulation model may better reproduce the behavior of some observation of slope displacement, hydrological and physical
types of flow (e.g., stony turbulent debris flows) while reproducing properties in the soil, as well as rainfall using various instruments.
less accurately others (e.g., muddy viscous flows). Anyway, due The main instruments that are installed in the field include rain
to the fact that most rheologies used by simulation models do gauges to measure precipitation; extensometers, inclinometers,
not actually describe the physics of the process, boundaries are tiltmeters, and GPS to observe slope displacement and deformation.
difficult to trace. It is not surprising, for example, that the Voellmy Total stations, laser scanners, and acoustic emission (AE) sensors
rheology, initially developed for snow avalanches (Christen et are also used to measure slope deformation in the field, as well
al., 2010) can be successfully used to simulate debris flows as time domain reflectometers (TDR), tensiometers, and piezometers
(Hussin et al., 2012) and also rock avalanches (Sosio et al., 2008; to measure changes of hydrological and physical properties of
Deline et al., 2011). Rheological relationships, in fact, should be soils. The third category consists of remote sensing methods
simple with few, easily constrained parameters (Hungr et al., 2005) including satellite image analysis, synthetic aperture radar (SAR)
to facilitate successful calibration on the basis of a sufficient interferometry, optical reflectometry, and light detection and
number of control parameters (e.g., inundated area, velocities, ranging (LiDAR). These methods are popular for monitoring
depths) and, possibly, more than one event. landslides in recent years because they can measure slope
Probably, the most desirable development in the research displacement with high resolution in wide target areas in the
field of fast-moving landslide propagation is the collection of field. They also can measure ground deformation under bad
field observations that can be used to systematically back-analyze weather conditions and even on dark nights. Recently, monitoring
past events. Only this way, much needed guidelines for the by SAR interferometry and LiDAR can be performed by using
selection of rheological parameters can be established and the ground-based SAR (GB-SAR) and LiDAR (GB-LiDAR), as well
performance of the models can be evaluated with respect to as by airborne satellites and planes. In this paper, in situ ground-
landslide typology. based monitoring methods and remote sensing methods are
reviewed to understand current state-of-the-art of landslide
3. LANDSLIDE MONITORING monitoring technologies.

Landslide monitoring is essential for recognition of landslide 3.1. In Situ Ground-based Monitoring Techniques
occurrence as well as for early warning. It observes slope
displacements in potential landslide sites and measures changes In situ ground-based monitoring usually observes precipitation,
in attribute values of landslide triggering factors to reduce slope displacement, and hydrological / physical properties of
landslide-induced damage. Monitoring of kinematic, hydrological, soils such as the groundwater level, volumetric water content
and climatic parameters plays a significant role in supporting (VWC), pore water pressure, and matric suction. The monitored
the development of slope stability models. Without understanding items are measured by instruments that are installed in the
movement patterns and responses to climate events, forecasting boreholes and on the surface of potential landslide sites. The
is not possible (Angeli et al., 2000). This requires not only instruments can measure the changes in observation factors in a
monitoring of actual movements, but also environmental factors regular time interval and/or in real time. Because the measurement
and geotechnical parameters (Uhlemann et al., 2016). Landslide interval can be easily controlled by the geologist, the changes in
monitoring helps to understand the landslide mechanism and observation factors can be measured in both short and long
to suggest reliable thresholds for landslide forecasting and early time intervals.
warning. Studies for landslide forecasting and early warning The monitoring data are analysed basically to recognize slope
have been performed actively in recent years, and the importance displacements. It is possible to determine the depth of a sliding
of landslide monitoring is emphasized. surface and to understand sliding rates of the slopes based on

http://dx.doi.org/10.1007/s12303-017-0034-4 http://www.springer.com/journal/12303
Landslide prediction, monitoring, and early warning: a concise review of state-of-the-art 1043

the monitoring data. Moreover, the analysis results give information of two or three of the mentioned conventional deformation
on the relationship between slope movements and hydrological monitoring techniques, showing good correlation between surface
or physical conditions of the slope under various geological and and subsurface deformations in terms of movement occurrences.
meteorological conditions. Therefore, in situ ground-based Chelli et al. (2006) and Konak etl al. (2004) installed a GPS,
monitoring enables decision-making on triggering factors of tiltmeter, and wire extensometer to determine the shape of the
landslides and suggesting thresholds of landslide triggering factors sliding masses and identify the depth of the sliding masses in the
such as rainfall, VWC, and suction stress at the monitoring site. upper part of the landslide. They also attempted to determine
However, ground-based monitoring is site-specific monitoring the absolute lateral and vertical movements, and the rate of
because it can observe landslides within the areas where the movement (velocity) of the different parts of the complex
monitoring sensors are installed. There is a limitation for effective landslide based on the displacement monitoring. Several studies
landslide monitoring extending over a wide area simultaneously. applied inclinometers to monitor landslide movement.
In situ ground-based monitoring can be grouped into three Stevens and Zehrbach (2000) interpreted 15 inclinometer
categories, namely monitoring of rainfall, slope displacement, datasets that were monitored for six years. The purpose of
and hydrological and mechanical properties in soils. Because monitoring was to determine the existence and depth of shear
rainfall represents the most common triggering factor, it is horizons and the velocity of the shear movement. In recent
important to analyse the relationship between rainfall condition years, movement measurements with high resolution and high
and landslide triggering based on rainfall measurement results. accuracy are required to provide data that may be used to verify
Most previous studies suggested empirical or statistical rainfall theories and models of landslide process as well as landslide
thresholds by rainfall measurement, using rain gauges in the warning (Xie et al., 2004). High resolution tiltmeters provide the
field (Caine, 1980; Wieczorek, 1996; Aleotti, 2004; Guzzetti et al., same type of data as inclinometers, but at significantly higher
2008; Brunetti et al., 2010; Martelloni et al., 2012). It is emphasized resolution, i.e., between 1 mrad (≈1 mm/m) and 1 nrad (≈1 nm/
that a landslide induced by rainfall is controlled by the locality m), depending on the tiltmeter model. Tiltmeters measure the
and the time of rainfall (Franks, 1999; Tsaparas et al., 2002). angle between the instrument body and the plumb line in two
Corominas (1999) completed a systematic study on the relationship horizontally perpendicular directions (Garcia et al., 2010). Chae
between landslides in the upper basin of the Llobregat River, et al. (2015) installed tiltmeters to observe slope displacements
Eastern Pyrenees and rainfall, and found that there were two affected by rainfall infiltration into the soil. Because the sensors
different rainfall patterns that were related to landslide occurrence were installed in a grid pattern both horizontally and vertically
(He et al., 2010). According to Guzzetti et al. (2005) the joint in the slope, they can measure fine movements of the soil layer
US-Japanese Tropical Rainfall Measuring Mission (TRMM) with changes in the wetting front behaviour by rainfall infiltration
launched in November 1997 (Kummerow et al., 2000) and the along the soil depth. The purpose of landslide monitoring in
planned co-operative United States National Aeronautics and Korea is to suggest a threshold of influence factor to trigger
Space Administration (NASA) and Japan Aerospace and Exploration landslides considering both geologic conditions and rainfall on
Agency (JAXA) Global Precipitation Mission (GPM) might natural terrain. Yin et al. (2010) performed a real-time monitoring
provide spatial and temporal precipitation information with of the Yuhuangge landslide at the Three Gorges Reservoir area.
sufficient detail to forecast the possible occurrence of landslides. The monitoring system included GPS with high-accuracy
In recent years, most landslide monitoring involved rainfall double frequency to monitor ground displacement, TDR, and a
measurements integrated with slope displacement, and hydrological borehole inclinometer to monitor deep displacement, a piezometer
and physical properties of soils (Luo et al., 2009; Yin et al., 2010; to monitor pore water pressure, precipitation, and reservoir water
Bittelli et al., 2012; Chae et al., 2012; Palis et al., 2016). The level monitoring. Based on the landslide monitoring experience
integrated monitoring enables understanding of the relationships at the Three Gorges Reservoir area, the early warning criteria on
between the multiple triggering parameters of landslides and landslides were established, in which a critical situation was classified
more reliable warning thresholds of landslide before landslide into four levels (Fig. 5). Uhlemann et al. (2016) combined
occurrence or at the very early stage of landslide movement conventional techniques such as GPS, inclinometer, and tiltmeter
(Fig. 5). with newly-applied deformation monitoring techniques such as
In the concept of displacement monitoring, GPS, a tiltmeter, AE monitoring using active waveguides (AEWG), and shape
an inclinometer, and an extensometer are generally installed on acceleration array (SAA). High temporal resolution information
the surface or in the subsurface of a slope to study the deformation allowed the capture of S-shaped slope displacement time
behaviour of rainfall-triggered landslides. Malet et al. (2002), behaviour, meaning phases of slope acceleration, deceleration,
Corsini et al. (2005), and Brückl et al. (2013) compared measurements and stability, in response to elevations in pore water pressures.

http://www.springer.com/journal/12303 http://dx.doi.org/10.1007/s12303-017-0034-4
1044 Byung-Gon Chae, Hyuck-Jin Park, Filippo Catani, Alessandro Simoni, and Matteo Berti

Macciotta et al. (2016) monitored the Ripley landslide by an


array of displacement measurement systems including GPS
units, a ShapeAccelArray (SAA), satellite (InSAR), and crack
extensometers, as well as an array of piezometers targeting pore
water pressures in the vicinity of the shear surface (Fig. 6). The
displacement monitoring system shows the annual cycle of
slope deformations and the average horizontal velocities during
the active displacement period. The purpose of the system is to
provide timely notice for landslide experts to evaluate the
hazard level associated with the Ripley landslide. Analyses of the
landslide displacement patterns and near-real-time monitoring
data are combined with previous studies of landslide-induced
railway track deflections and track quality standards to develop
early warning displacement thresholds and associated hazard
management protocols.
The third monitoring concept of rainfall-induced shallow
landslides is linked with a continuous monitoring of hydrological
and mechanical properties of the soil. Many previous studies
have demonstrated the possibility of monitoring techniques to
identify the soil’s hydrological and mechanical conditions during
the triggering of shallow landslides (Simoni et al., 2004; Matsushi
Fig. 5. An example of integrated ground-based monitoring of land-
slides which has been installed at Yuhuangge landslide, the Three et al., 2006; Godt et al., 2008, Godt et al., 2009; Baum et al., 2010;
Gorges Reservoir area, China. 1. Piezometer and TDR; 2. Borehole Bittelli et al., 2012; Damiano et al., 2012; Leung and Ng, 2013;
inclinometer; 3. GPS; 4. Field station (after Yin et al., 2010).
Springman et al., 2013; Bordoni et al., 2015b).

Fig. 6. Array of monitoring instruments measuring landslide displacement. GSC is a steel corner reflector for satellite InSAR; MSE is a crack
extension meter by manual measurement (after Macciotta et al., 2016).

http://dx.doi.org/10.1007/s12303-017-0034-4 http://www.springer.com/journal/12303
Landslide prediction, monitoring, and early warning: a concise review of state-of-the-art 1045

The concept of hydrological and mechanical monitoring of Berti and Simoni (2005) investigated mechanisms and prediction
landslide was also implemented in other previous studies. The methods for debris flow initiation by channel bed mobilization,
unsaturated soil hydrology has been shown to be a key factor for by monitoring rainfall and pore water pressure with sensors
shallow landslide triggering (Lu and Godt, 2008; Tsai and Chen, buried in a loose channel bed. Comegna et al. (2007) attempted
2010). Several authors maintain that the most important cause long-term monitoring of landslide movements and pore water
for shallow landslides is the decrease in matric suction after a pressure to analyse the mechanisms of mudslides in Italy.
rainfall and the development of positive pressures above the
groundwater table. Lu and Godt (2008) described the stability 3.2. Remote Sensing Techniques for Landslide Mon-
analysis of an infinite slope, taking into account the suction itoring
stress concept proposed by Lu and Likos (2006). Previous
studies also suggested that rainfall-induced slope failures can be 3.2.1. General framework
triggered by two potential mechanisms, namely: (i) reduction of Traditional direct field monitoring techniques have the major
matric suction (Ng and Shi 1998; Collins and Znidarcic, 2004; drawback of being in physical contact with the monitored
Rahardjo et al., 2007), and (ii) rise of the water table (Cho and Lee, object. For this reason, the measurement systems may incur two
2002; Crosta and Frattini, 2003). Through analytical analysis, Li problems: i) they may influence the system and, therefore, the
et al. (2013) proved that slopes with an effective friction angle measured quantities; ii) they may be influenced by the system.
greater than the slope angle are unlikely to be triggered by the Remote sensing methods, which adopt sensors able to measure
loss of matric suction. Water table may rise when the infiltrated geophysical variables of the ground from distance, without direct
rainfall reaches an impermeable layer or the main groundwater contact, almost completely overcome this limitation. This, together
table. In addition, several researchers found that the groundwater with the huge advancement lately experienced by space- or air-
flow regime in a highly permeable layer overlain by a less permeable borne sensing platforms, has produced a notable increase in the
layer may result in an exceptionally higher rise in pore water contribution of remote sensing to landslide hazard assessment,
pressure than under normal conditions (Montgomery et al., 2002; monitoring and early warning. As recently verified by a review
Jiao et al., 2005; Jiao et al., 2006). Li et al. (2016) installed the work by Tofani et al. (2013a), based on a large survey carried out
instrumentation of piezometers on the slope of the Yangbaodi in Europe, in 83% of the cases remote sensing is being used as a
landslide in China to correlate the change in hydraulic head with routine tool for landslide detection, mapping and monitoring.
rainfall and to analyse the effect of hydrological and mechanical The study also reveals that 75% of the users takes advantage of a
conditions that trigger landslides. To identify the depth of the combination of 2 or more techniques.
failure plane, as well as other landslide features, and to test the Despite the advantages, some major problems remain in the
validity of a specific modelling approach, it is necessary to monitor adoption of remote sensing methods for landslide monitoring,
unsaturated soil properties and slope stability. Several techniques i.e., i) the lack of underground penetration capability, ii) the
have been proposed to monitor landslides and various experiments lower frequency of acquisition with respect to direct automated
have been performed to assess their reliability (Rahardjo et al., methods and iii) the atmospheric disturbance effects.
2011). As the study of Tofani et al. (2013a) reveals, the main advantages
Among the available techniques, the use of TDR for mass and limitations of the remote sensing methods are driven by the
deformation studies is increasingly applied for landslides studies. landslide typology, and in particular by two important factors
O’Connor and Dowding (1999) presented a description of the which derive from the latter: landslide velocity and dimension.
technique, and several works have been published describing According to this assumption, we may classify remote sensing
applications in different settings. The TDR technique is more methods for landslide detection and monitoring based on the
commonly used for measuring soil water content and electrical distance between sensor and target, a parameter which strongly
conductivity (Topp et al., 1980; Robinson et al., 2003; Bittelli et influences both the scale and the frequency of the acquisition.
al., 2012). Chae et al. (2015) installed TDRs and tensiometers to Along this line, we traditionally distinguish among ground-
observe slope displacements affected by rainfall infiltration into based, air-borne and space-borne sensors, the main difference
the soil. Because the sensors were installed in the slope in a grid being the platform used to bring the sensor in position and,
pattern both horizontally and vertically, they can measure the thus, the measurement frequency and the distance from the
velocity of rainfall infiltration into the soil. They also analyse target. From a functional and operational standpoint another
changes in FS of the monitoring slope related to the rainfall distinction can be made on the basis of the type of sensor,
infiltration in the soil. In terms of relationship between rainfall between passive and active RS tools. Passive methods rely on the
and pore water pressure to occur landslides and debris flows, measurement of natural energy reflected by the target, which is,

http://www.springer.com/journal/12303 http://dx.doi.org/10.1007/s12303-017-0034-4
1046 Byung-Gon Chae, Hyuck-Jin Park, Filippo Catani, Alessandro Simoni, and Matteo Berti

in most of the cases, sun energy. Active sensors, instead, produce are easier to measure and predict, such as the rainfall amount.
their own energy by emitting electromagnetic signals and We may distinguish, then, between triggering monitoring and
registering the part which is reflected by the ground surface. displacement monitoring.
The difference is relevant as it will be detailed in the following. The triggering monitoring, i.e., the ability to measure with a
suitable frequency an initiation event, has mainly been concentrated
3.2.2. Passive sensors on rainfall. The most notable example is provided by Hong et al.
The first RS tool used in geomorphological studies, with (2006a) and following papers (Hong et al., 2006b, 2007; Hong
special reference to landslide recognition and mapping, was the and Adler, 2007; Kirschbaum et al., 2009) which exploit the
aerial photo-interpretation, which was traditionally carried out TRMM-based Multi-Satellite Precipitation Analysis (TMPA,
by means of calibrated optical cameras, flown on airplanes, relying Adler et al., 2000) product to provide a model for landslide
high-resolution pictures of the ground (both in grayscale and triggering worldwide. The authors use the precipitation data,
RGB), used since the end of WWII. In landslide studies, the usual together with the activation time of a global database of landslides,
method of mapping was, firstly, manual photo-interpretation by to set rainfall thresholds to be adopted in back-analysis,
experienced geomorphologists which, after the mapping, added susceptibility studies and, with some limitations, early-warning.
geographic positioning by ground control point (GCP) calibration After the expected end of the TRMM mission, the future of such
in a following phase. After the development of modern GIS type of applications will be based on new global weather satellite
systems, though, the geo-registration of the aerial images is carried missions such as the Global Precipitation Measurement (GPM)
out with a seamless procedure from the aerial acquisition to the mission (http://gpm.gsfc.nasa.gov) or the European Sentinel
digital positioning by onboard GNSSs down to the upload of constellation. As an example, ERS-1/2, ENVISAT and Sentinel-1
the final adjusted data in a GIS for vector digital mapping where synthetic aperture radar (SAR) images are being used to calibrate
the contribution of the photo-interpreter is still fundamental inversion models able to provide Earth surface moisture at high
(Mantovani et al., 1996; Canuti et al., 2000). resolution (20 m), a product which can be directly uploaded in
In the ‘70s space-borne sensors became available for the numerical and statistical triggering models for shallow or deep-
public and the research as well, with the start of the first multi- seated landslides (Paloscia et al., 2013; Rossi et al., 2013; Santi et
spectral satellite missions, such as the Landsat (Sauchyn and al., 2015). In other cases, rainfall patterns able to initiate mass
Trench, 1978; Temesgen et al., 2001; Barlow et al., 2003). This movements may also be anticipated in time (forward analysis)
represented a true milestone in geological and geophysical by exploiting satellite meteorological forecasts, numerical weather
sciences because, for the first time, data were being acquired models and downscaling tools as in the case of the prototype
routinely at global scale and with a spectral information content application proposed by Mercogliano et al. (2013). Since direct
higher than simple RGB photography. From that point on, the slope monitoring is not feasible, the ability to monitor and
evolution of remote sensing science has been overwhelming forecast rainfall dynamics is fundamental for the early warning
and nowadays the number of technical solutions for landslide of shallow landslides over medium to large scale areas (Baum et
studies is vast. al., 2010). Shallow landslides, with the specific reference to
Since, in general, the greater the observation distance, the larger debris-flow, in fact, may hardly be monitored by using remote
the area that can be monitored or mapped, whilst, conversely, sensing because of their: i) velocity, which would overcome the
the smaller the resolution of the measurement, in terms of maximum time resolution of the RS tool, ii) absence of progressive
either or both spectral and spatial accuracy, then, space-borne accelerating behavior, iii) pseudo-random space occurrence,
sensors are recommended for working at regional scales whilst which implies that we do not know in advance the exact
ground-based approaches are best when the focus is on a single location to monitor. When one of those three conditions is not
unstable slope. This rule of thumb, however, is not always true met, RS applications start to become possible for landslide early
because, with the advancement of new technologies, the accuracy warning based on displacement monitoring instead of triggering
of long-distance acquisitions is becoming higher every year. measurement. Table 1 gives a general overview of the different
In this general framework, the use of RS for landslide RS methods that can fit landslide types, according to the studies
monitoring has specific challenges because it requires revisiting cited herein and based on type, material and expected velocity.
a given site with a frequency driven by the landslide dynamics, Displacement monitoring can be, in turn, subdivided in
which is not constant and may evolve progressively or measurement of change (change/target detection), measurement
discontinuously (Corominas et al., 2014). The object of RS of surface displacement and measurement of underground
monitoring may be the displacement field of the mass movement displacement. The latter is at the moment not feasible by using
itself or, in some cases, some external triggering factors which traditional or innovative remote sensing techniques, mostly due

http://dx.doi.org/10.1007/s12303-017-0034-4 http://www.springer.com/journal/12303
Landslide prediction, monitoring, and early warning: a concise review of state-of-the-art 1047

to the lack of ground penetration capability by the available time monitoring tool, mainly because the latter would require
sensors. Only microwaves have a little potential of penetrating the measurement of displacement velocity and the quantification
the soil, which is roughly proportional to the wavelength, but of volumes, difficult to implement in a seamless operational
this only concerns a few centimeters of depth which is obviously chain based on photogrammetry. A possible exception to that is
not useful for displacement measurement and modeling. Radar offered by UAV-photogrammetry based on structure from
sensors are able to give a deeper underground reading only motion.
when used at direct contact with the surface, as in the case of GPR Optical correlation methods, also referred to as structure-
systems, but they are usually considered geophysical methods from-motion SfM techniques, are the modern development of
and will not be discussed in this review. For an overview of GPR traditional photogrammetric tools, enhanced by digital imaging
capabilities in landslide studies, which in any case do not include and numerical computation (Kaab, 2000; Delacourt et al., 2004;
actual monitoring, we suggest to refer to Van Dam (2012). Debella-Gilo et al., 2011; Travelletti et al., 2012; Westoby et al.,
Remote sensing methods have, conversely, a very high 2012; Lucieer et al., 2014; Stumpf et al., 2015). Nowadays, many
effectiveness for the exploration of surface changes of any type. authors have started adopting SfM as tools to easily measure
If the detection may be carried out repeatedly with the suitable and model surface displacements induced by mass movements
frequency in time, then the RS method may also be used as an as well as for activating monitoring procedures in cases where
actual monitoring tool. the continuous measurement is possible.
Space-borne optical image analysis has been the first tool to The ability to build accurate 3D models of a topographic
overcome traditional air-borne photogrammetry, essentially for surface by using simple handheld camera methods extends the
reasons of costs. A mandatory requirement for the use of optical potential of photogrammetry to near-vertical slope geometries
imagery in landslide monitoring is the shift from the traditional which were traditionally difficult to measure by airborne cameras.
manual interpretation and mapping to a semi-automated or An additional step in photogrammetry automation is offered by
fully automated one. In fact, if the revisiting time is small enough the possibility of flying SfM cameras on UAVs with the specific
to be compatible with monitoring, then the image elaboration use of small, low-cost drones. An example of this type of application
has to be similarly fast. The rapid mapping of landslides through to quickly obtain accurate DEMs over medium-sized areas is
change or target detection has been recently obtained by Stumpf presented by Lucieer et al. (2014) that use SfM and image
and Kerle (2011) that used a supervised classification of IKONOS, correlation techniques to generate terrain models from UAV
Quickbird, Geoeye-1 and aerial images to map landslides. The photography. The authors use multi-temporal data to analyze
classification method was based on the joint application of object- and monitor the landslide behavior and find that the
oriented analysis (OOA) and random forests and attained an methodology has a positioning accuracy in the order of 0.1 m,
accuracy up to 87% in a supervised approach. A multi-temporal the same as Westoby et al. (2012) found in their work. A simple
approach of OOA on high resolution data has been proposed by monitoring by UAV-photogrammetry has been carried out by
Lu et al. (2011), which have employed advanced image segmentation Niethammer et al. (2012), which used high resolution DEMs of
techniques for a fast landslide identification and mapping. This the Super-Sauze landslide (South French Alps) taken in 2007
method has the potential to become an actual monitoring and 2008 to analyze the areas affected by surface modification
technique when applied routinely at revisiting time compatible and verified horizontal displacements between 7 and 55 m. A
with the landslide development (see also Table 1). There are at recent review of UAV-based photogrammetric methods, including
present several VHR satellites operating for Earth observation a panoramic on regulations and regulatory bodies, is provided
that can be suitable for rapid mapping and monitoring by using by Colomina and Molina (2014).
change and target detection algorithms. Among them IKONOS The main advantages of terrestrial and UAV photogrammetry
(DigitalGlobe; panchromatic 0.8 m, multispectral 3.3 m), Quickbird are the low-cost and the easiness of use, that extend the capability of
(DigitalGlobe; panchromatic 0.6 m, multispectral 2.8 m), World using optical devices to monitoring purposes for small areas
View-3 (DigitalGlobe; panchromatic 0.31 m, multispectral 1.24 m), with complex topography. They are not usable, however, for
GeoEye-1 (DigitalGlobe; panchromatic 0.41 m, multispectral large areas. In such cases, the adoption of airborne or spaceborne
2.0 m), Pleiades (ORFEO; panchromatic 0.5 m, multispectral optical cameras is preferred, either in the visible or in a larger
2.0 m), EROS-B (Israel Space Agency; 0.7 m panchromatic only). spectral window including infrared. As an example, Forzieri et
It must be said, however, that despite the increase in acquisition al. (2013) used a multi-sensor airborne approach to monitor an
frequency and in the number of available post-processing tools, area across the Northern Apennines (Italy) and found that each
the use of optical data for landslide analysis it is still more a acquisition costs around 250 $/km2 including pre-processing
target/change detection technique rather than an actual real- and georegistration of data. If we consider economies of scale, in

http://www.springer.com/journal/12303 http://dx.doi.org/10.1007/s12303-017-0034-4
1048 Byung-Gon Chae, Hyuck-Jin Park, Filippo Catani, Alessandro Simoni, and Matteo Berti

Table 1. Remote sensing methods and main references for different types of landslides
Typology Material Velocity RS monitoring tools Notes
Basin scale application possible by using spaceborne
ES-VS PSI, DSI, GB-InSAR, AP(c), UAV-P sensors.
Earth
DInSAR(c), TLS, ALS, GB-InSAR, SOCD(c), SRM(b), Basin scale application feasible only through trigger
S-M GB-WR(b), AP(c), UAV-P monitoring (GB-WR, SRM) + numerical modeling.
(c)
S-M ALS, TLS, GB-InSAR, AP , UAV-P
Slides Debris
R-VR-ER GB-nonSAR, AP(c), UAV-P(c), SRM(b), GB-WR(b)
ES-VS PSI, DSI, GB-InSAR Non-vegetated areas only.
S-M GB-InSAR, AP(c), UAV-P
Rock
Usage of GB-nonSAR subject to runout lead time limita-
R-VR-ER GB-InSAR(a), GB-nonSAR, AP(c), UAV-P(c) tions.
S-M GB-InSAR, DInSAR(a)(c), DSI(a), UAV-P Mainly precursor monitoring.
Earth Usage of GB-nonSAR subject to runout lead time limita-
R-VR-ER TLS(a)(c), GB-InSAR(a), GB-nonSAR, UAV-P(c) tions.
Falls/Topples (a)(c) (a) (c) Usage of GB-nonSAR subject to runout lead time limita-
Debris R-VR-ER TLS , GB-InSAR , GB-nonSAR, UAV-P tions.
Usage of GB-nonSAR subject to runout lead time limita-
Rock R-VR-ER TLS(a)(c), GB-InSAR(a), GB-nonSAR, UAV-P(c) tions.
ES-VS PSI, DSI, GB-InSAR, UAV-P, AP
S-M DInSAR(c), GB-InSAR, SRM(b), GB-WR(b), UAV-P, AP(c) Usage of GB-InSAR with sub-apertures (Bardi et al., 2017).
Earth
(c)
, GB-InSAR(a), GB-nonSAR, SRM(b),
R-VR-ER DInSAR Usage of GB-nonSAR subject to runout lead time limita-
GB-WR(b), AP(c), UAV-P(c) tions.
Monitoring of slow homogeneous displacements of
ES-VS DSI, UAV-P debris cones and talus slopes. Non-vegetated areas only.
Flows (b) (b) Basin scale application feasible only through trigger
S-M GB-InSAR, SRM , GB-WR , UAV-P monitoring (GB-WR, SRM) + numerical modeling.
Debris
Basin scale application feasible only through trigger
R-VR-ER GB-nonSAR, SOCD(c), SRM(b), GB-WR(b), UAV-P§ monitoring (GB-WR, SRM) + numerical modeling. At
slope-scale, usage of GB-nonSAR subject to runout lead
time limitations.
ES-VS PSI, DSI Non-vegetated areas only.
Rock
S-M GB-InSAR, DInSAR(c), ALS(c), TLS(c), AP(c), UAV-P(c)
ES-VS PSI, DSI Non-vegetated areas only.
Earth S-M DInSAR(c), DSI(c), UAV-P, AP
Spreads R-VR-ER SOCD(c), SRM(b), GB-WR(b), UAV-P(c), AP(c)
ES-VS PSI, DSI Non-vegetated areas only.
Rock
S-M DInSAR(c), UAV-P(c), AP(c)
ES-VS PSI, DSI
Earth Basin scale application feasible only through trigger
S-M SRM(b), GB-WR(b)
monitoring (GB-WR, SRM) + numerical modeling.
Soil slips, ES-VS PSI, DSI Non-vegetated areas only.
solifluction (b) (b) Basin scale application feasible only through trigger
and soil creep Debris S-M SRM , GB-WR monitoring (GB-WR, SRM) + numerical modeling.
Basin scale application feasible only through trigger
R-VR-ER SOCD(c), SRM(b), GB-WR(b), UAV-P(c), AP(c) monitoring (GB-WR, SRM) + numerical modeling.
Rock ES-VS PSI, DSI, DInSAR(c) Non-vegetated areas only.
The classification is derived from Cruden and Varnes (1996) with some modifications to take into account lumped velocity classes, where
ES-VS = extremely slow, very slow (v < 1.6 m/yr); S-M = slow, moderate (1.6 m/yr < v < 1.8 m/hr), R-VR-ER = rapid, very rapid, extremely
rapid (v > 1.8 m/hr).
For complex landslides, please join single methods as required. ALS: aerial laser scanner; TLS: terrestrial laser scanner; UAV-P: drone photogram-
metry; AP: aerial photogrammetry; SOCD: satellite optical change detection; SRM: satellite rainfall measurement; GB-WR: ground-based weather
radar; DInSAR: satellite interferometry; PSI: persistent scatterers satellite multi-interferometry; DSI: distributed scatterers satellite multi-interfer-
ometry; GB-InSAR: ground-based SAR; GB-nonSAR: ground-based non-strictly SAR radar.
(a)
usage is only for precursor monitoring and inverse-velocity methods; (b)usage is only for external triggering monitoring; (c)usage is only for
delayed-time monitoring and long-term monitoring (change detection).

http://dx.doi.org/10.1007/s12303-017-0034-4 http://www.springer.com/journal/12303
Landslide prediction, monitoring, and early warning: a concise review of state-of-the-art 1049

case of monthly monitoring frequency, the costs could raise at (Jaboyedoff et al., 2012). Conversely, if the area to be scanned is
about 1000 $/km2 per year or higher. Even though the spatial almost flat or has a notable extension, TLS has a limited usability
resolution obtained in the case of Forzieri et al. (2013) was very and ALS must be used.
high (0.25 m horizontal resolution for the ADS40 4-band The repetition of the same measurements at different times
camera, 1 m DTM from LiDAR, 4 m for MIVIS hyperspectral) allows for the use of ALS and TLS systems in landslide monitoring.
this cost may well be unsustainable for routine monitoring The most frequent application of ALS/TLS, however, concerns
purposes, if compared to the use of very-high resolution (VHR) the mapping and characterization of mass movements. This is
satellite optical data (such as IKONOS, QuickBird, GeoEye, due to the fact that generating a full 3D model from multiple
WorldView, Pleiades, EROS) which have costs in the order of laser-scans in complex topography needs millions of points and
15–50 $/km2 (depending on resolution, acquisition mode and a lot of post-processing, issues that hamper an easy use for real-
level or pre-processing) and rather regular revisiting times even time displacement control. Glenn et al. (2006) used ALS to analyze
in non-tasking mode. the morphology of 2 large landslides in Idaho (USA) at different
spatial scales and concluded that geomorphometry applied to
3.2.3. Active sensors high resolution DEMs such as those generated by laser scanners
Optical cameras, on whatever support platform, are passive may help in reconstructing the kinematics and the geometry of
sensors, requiring solar radiation in order to work. Furthermore, a mass movement. The accuracy of DEMs and 3D models derived
they exploit a range of wavelengths (usually from 0.3 mm to 1.1 from ALS depends on a number of factors, such as the flight
mm) which is particularly sensitive to atmospheric disturbance, altitude, the quality of LiDAR used, the configuration of the
such as fog, mist, clouds or rain. In specific environments, such topographic surface and the vegetation effects. In an interesting
as volcanoes and large mining sites, the optical methods may work, Dewitte et al. (2008) compared the RMSE vertical accuracy
also be hampered by the presence of ash and dust in the air. of DEMs generated by multi-temporal stereo-pairs to that of an
Some of those limitations are overcome by the use of active ALS-based DEM for the long-term monitoring of landslides in
sensors, which generate their own energy instead of relying on West Belgium. They found that the total RMSEZstereo~0.4 m while
the reflected sun light. Among them, in landslide studies, lasers RMSEZALS~0.3 m, figures that are suitable for most of landslide
and radars are the most used. As we have already introduced, monitoring purposes on the long term. A similar accuracy is
lasers are mostly used through scanners, both air-borne (ALS) found by Chen et al. (2006) that use a LiDAR generated DEM to
and terrestrial (TLS), which generate a multiple laser beam and compute displaced mass volume at the Tsaoling landslide (China)
measure the energy which is reflected back from the target after the Chi-Chi earthquake of 1999. The reported accuracy is
(Jaboyedoff et al., 2012). ALS/TLS systems offer a 3D scan of the 0.2 m in open terrain and 0.2–0.5 m in vegetated areas. Recently,
target by providing distances measured along the Line-of-Sight Ghuffar et al. (2013) exploited the combination of both ALS and
(LOS) that can be translated into [x,y,z] relative or absolute TLS to study the dynamics of a landslide in Austria over a 10-
triplets by simple topographical computations (see e.g., Petrie year time span. They make use of the range flow method to
and Toth, 2008). Multiple flights from different LOS in ALS or compute the 3D motion (Spies et al., 2002), a novel application
from different observation positions in TLS allows for the in quantitative geomorphology that yields a dense 3D field-of-
resolution of shadow areas and even of hanging walls in rock motion vectors for the entire available time series. In most of the
slopes (Gigli and Casagli, 2011; Jaboyedoff et al., 2012; Gigli et cases, though, ALS scans are carried out once, without multi-
al., 2013). The main differences between TLS and ALS (also temporal capability. In such a case, they are very useful for modeling
referred to as LiDAR) refer to the acquisition geometry and and characterization as well as for detection and mapping (Glenn
distance. In the case of ALS the LOS is almost vertical and there et al., 2006; Schulz, 2007; Kasai et al., 2009) or for volume
is the need of fast acquisition rates during airplane routing that measurement, fundamental for the quantitative assessment of
limit ground resolution. This leads to the inability to measure landslide intensity (Catani et al., 2016).
steep slopes, hanging slopes and shadowed areas and to a point Due to its higher spatial resolution and to the ability of
density in the range 0.1–100 pts/m2. In the case of TLS, instead, covering near-vertical cliff faces, the use of TLS is much more
both the distance of acquisition and the LOS are very flexible common than ALS for the study of rock masses affected by rock
and the problem of reproducing a model of a near-vertical cliff falls (Abellan et al., 2006, 2009, 2010; Gigli and Casagli, 2011;
may easily be solved. Point density is much higher, ranging from Gigli et al., 2013). In particular, Abellan et al. (2006) applied a
50 to 10,000 pts/m2. As a consequence, also the nominal accuracy TLS analysis to a rock cliff affected by falls in Northern Spain
is different and is, in average, of about one order of magnitude and used the measured data to characterize discontinuities and
higher in TLS than in ALS (RMSEZTLS~10–2m; RMSEZALS~10–1m) to feed high-precision topography data to rockfall simulation

http://www.springer.com/journal/12303 http://dx.doi.org/10.1007/s12303-017-0034-4
1050 Byung-Gon Chae, Hyuck-Jin Park, Filippo Catani, Alessandro Simoni, and Matteo Berti

codes. Afterwards, in a notably important work, Abellan et al. depth investigation concerning the displacement time series
(2009) demonstrated that, even though raw multi-temporal TLS was carried out later on by Berardino et al. (2003) that exploited
scans are not capable of detecting millimeter-scale precursor 7 ERS-2 images to build interferograms that, together with GPS
displacements in rock masses, this is feasible in post-processing, and EDM measurements, allowed for the reconstruction of 3
by resorting to data filtering with nearest neighbor averaging. years of activity of the Maratea landslide (Italy). A similar approach,
For what concerns quantitative geomechanical modeling, Gigli based on ERS-1 data, was used by Singhroy et al. (2004) to map
and Casagli (2011) devised a semi-automated method for displacements occurred between 1993 and 1997 at the Frank
extracting rock slope discontinuities and other mechanical slide in Alberta. In the same paper, the authors also attempt a
parameters from TLS scans that may be used for improving the texture characterization of the debris by using a RADARSAT
spatial prediction of rock falls, toppling, wedge and block sliding. image. A more detailed investigation on the geomorphometric
The application, though, has no practical use for monitoring in parameters of a landslide that can be extracted from DInSAR is
real time. provided by Catani et al. (2005) that use JERS-1 radar data to
The main limitations of LiDAR systems for monitoring, then, directly extract topographic derivatives (slope and curvature)
are: i) the lack of relative millimeter-accuracy capability when from the interferograms instead of recurring to a DEM. They also
comparing different acquisitions at different times, which hampers measure deformations in a rockslide in the Italian Alps by using
slow displacement detection, unless permanent reflectors are the same JERS-1 SAR interferograms and find that the method
installed in situ; ii) the lack of high-frequency acquisition, due to is able to measure ±5 cm displacements occurred in a time interval
the complex pre- and post-processing needs and to the relevant of 44 days. The use of the L-band JERS radar, with a wavelength
volume of point clouds in terms of data storage, which hampers longer than the C-band of ERS-1 and ERS-2 reduces spatial
fast displacement monitoring; iii) the lack of a robust 24/7 operability, accuracy but increases coherence in areas affected by rapid
due to the low signal to noise ratio generated by atmospheric displacements and covered by low-stand vegetation (Catani et al.,
effects such as rain, fog, mist, clouds and dust. 2005; Strozzi et al., 2005). This is particularly important when
Some of those limitations are avoided by resorting to a using DInSAR for landslide studies in mountainous areas and
different wavelength and a different processing technique, by natural landscapes. A compendium on how to best use L-band
using radars and radar interferometry (see Ouchi, 2013 for a DInSAR products for a correct radar-interpretation of ground
complete technical review on radar and SAR systems). Microwaves, deformations in mapping and monitoring is provided by a
in fact, are able to penetrate fog, clouds, light rain and even dust recent work of Schlogel et al. (2014) in which the authors detail
in most of the cases. Furthermore, by using radar interferometry, the different patterns that can be expected in the interferograms
the analysis of data can be automated and fast with the for rapid landslides. ERS satellite data are also used by Rott and
displacement measured at every new reading with a frequency Nagler (2006) that use DInSAR to measure slope deformations
as low as 0.5 min (Bardi et al., 2017). Radar interferometry was at two different sites in the Swiss Alps and in Lichtenstein and find
first used in geophysics by Massonnet et al. (1993, 1994) and RMSE of about 4mm/yr with respect to GPS measurements.
Massonnet and Feigl (1998) for the case of ground deformations However, DInSAR is only usable for detecting movements of
induced by earthquakes. The authors found the method capable the ground that occurred between the two acquisitions and
of detecting displacements over large areas (> 1000 km2) with subject to the condition that the field of motion must be well
centimeter-scale accuracy. Since then, many other applications autocorrelated in space and have a smooth gradient in time.
have been found for mapping and measuring soil subsidence, Another limitation is that the maximum accuracy of DInSAR is
landslides, structure deformation, river and coastal dynamics. at the centimeter-scale whilst most of deep-seated landslides
The principle of differential radar interferometry (DInSAR) is move very slowly, at rates of a few millimeters per year (see
the comparison in range (sensor-target distance) between couples Table 1). In this range, DInSAR cannot be used as a monitoring
of radar images acquired at different times over the same area. tool and it is necessary to resort to time-series analysis of InSAR
The phase shift, suitably filtered, is proportional to the ground (TSA-InSAR).
displacement component parallel to the LOS of the satellite. There are two main types of TSA-InSAR, according to whether
One of the first reports on landslide mapping through DInSAR they are based on the measurement of phase differences on
is provided by Fruneau et al. (1996) which used a series of SAR stable point reflectors (PSI methods) or on distributed reflectors
images acquired in France by the ERS-1 satellite to detect a (DSI methods). A common characteristic of all TSA-InSAR
landslide with 1-km2 size. The application is essentially about methods is that they exploit multiple interferograms to generate
mapping and detection rather than kinematics but can still be long time series of ground displacement that cannot be produced
considered a breakthrough in landslide science. A more in- with single-interferogram DInSAR. PSI methods provide very

http://dx.doi.org/10.1007/s12303-017-0034-4 http://www.springer.com/journal/12303
Landslide prediction, monitoring, and early warning: a concise review of state-of-the-art 1051

accurate displacement measurements over specific sub-pixel buildings so that only about 8% of the mapped landslides have
objects that can be identified one by one on the ground. This is PS information. A much higher density can be obtained by
very helpful in urban areas and to study the development of resorting to X-band satellites such as TerraSAR-X and Cosmo-
ground deformation affecting man-made structures. DSI methods SkyMed (range and azimuth resolution of 1 m) or by mixing L,
deliver slightly less accurate measurements, with respect to PSI, C and X-band scenes as in Herrera et al. (2013) in which the
and are usually related to multiple pixel clusters in areas with or authors, by combining ALOS-PALSAR (L-band), ERS-ENVISAT
without strong reflectors. For this reason, they offer a higher (C-band) and TerraSAR-X (X-band), increased the percentage
information density in non-urban areas and are less sensitive to of landslides with PS information from 19% to 38%. A combination
temporal decorrelation. Stable point methods (PSI) are the of PS-InSAR and traditional underground displacement
PSInSARTM (Ferretti et al., 2000, 2001), the Interferometric Point monitoring systems for the large landslide of S. Stefano d’Aveto
Target Analysis (IPTA, Werner et al., 2003), the Stable Point (Italy) is presented by Tofani et al. (2013b). The authors compare
Network (SPN, Crosetto et al., 2008), the Persistent Scatterer Pairs the measurements and propose a technique for the LOS
(PSP, Costantini et al., 2008), the Stanford Method for Persistent component projection on the real 3D displacement vector
Scatterers (StaMPS, Hooper et al., 2004; Hooper, 2008). Distributed measured by inclinometers. An interesting technique to analyze
scatterers methods (DSI) available on the market are the Small PS time series is developed by Berti et al. (2013) that provide a
Baseline Subsets (SBS, Berardino et al., 2002; Lanari et al., 2007), tool to classify displacement trends and the deviation from
the Coherent Pixels Technique (CPT, Blanco-Sanchez et al., expectation. When ground scatterers are not dense but the
2008), the SqeeSARTM (Ferretti et al., 2009, 2011), the Quasi-PS movements are well autocorrelated in space, the SAR displacement
(QPS, Perissin and Wang, 2012). field may also be mapped by using a geostatistical interpolation
A more detailed report on the techniques and algorithms used of PS data, as suggested by Lu et al. (2012) that use Getis-Ord
by the various TSA-InSAR methods is provided by Osmanoglu statistics for the detection of slope instability hot-spots. In other
et al. (2016). cases, PS information may be used to define the state of activity
The first example of application of TSA-InSAR to a landslide of slow moving landslides and to attempt quantitative hazard
is the PS-InSARTM methodological paper by Ferretti et al. (2001) and risk analysis as proposed by Lu et al. (2014) and Ciampalini
where a time series relative to the Ancona landslide (Italy) is et al. (2016). A review and comparison of spaceborne SAR time-
obtained by applying the PSI technique to a series of 34 C-band series analysis methods is provided by Crosetto et al. (2016).
ERS SAR images from 1993 to 1998. Displacements on the order Despite all those important achievements, classical satellite
of 5 ± 0.4 mm/yr are measured on specific PS targets inside the DInSAR and multi-temporal PS-InSAR remain of a limited
landslide boundaries. The Ancona landslide is studied with the usefulness in the monitoring of rapid landslides. This is due
same method also by Colesanti et al. (2003). In this case, to the long revisiting time of SAR satellites. ERS-1/2, JERS-1,
displacement time series in the period of 1992–2000 (61 ERS RADARSAT-1 and ENVISAT only ensured, in optimal
images) are compared to GPS and EDM data with good agreement. conditions, temporal baselines of about 1 month. Present-day
The attained PS density is around 5 points/km2 in non urban satellites such as TerraSAR-X (11 days), TanDEM-L (8 days),
areas. By resorting to a different method (full-resolution SBAS) Sentinel-1 (6 days), Cosmo-SkyMed (1 day) have better capability
Lanari et al. (2004) obtained a much more continuous land but nonetheless are not suitable for the monitoring of rapid
coverage of the city of Naples (Italy) with a density which is landslides. Furthermore, the near-vertical LOS of SAR satellites
inversely proportional to the baseline and that is around 250 is not always the best to measure displacements occurring on steep
points/km2 in urban areas. Accuracy, compared to GPS stations, slopes. Future improvements in spaceborne SAR interferometry
is lower than in PS-InSAR being it around ±2 cm. The first for landslide monitoring could be offered by the development of
large-scale application of PSI data for landslide monitoring is constellations of a few SAR satellites in geostationary bistatic
offered by Farina et al. (2005) in an important study funded by configuration, where the spaceborne transmission-only radars
ESA in which 350 ERS-1/2 scenes from 1992 to 2002 are used to are complemented by low-altitude air-borne reception-only SARs
produce more than 590,000 PSI time series over an area of about to cover large areas with a revisiting time of about 10 seconds
9000 km2 in central Italy. The accuracy of displacement monitoring (Ouchi, 2013).
found by Farina et al. (2005) is around ±4 mm/yr, very good for At present, though, the best solution for slope-scale monitoring
monitoring extremely slow landslides, and the average PS density is the adoption of terrestrial interferometers, usually referred to
is of about 65 points/km2, decreasing to 37 points/km2 in hilly as ground-based radars, with (GB-InSAR) or without synthetic
and mountain areas. However, the distribution of stable scatterers aperture (GB-nonSAR). For SAR systems, the principle of use is
on the ground is not regular and most of them are located on almost the same as satellites, with the main difference that the

http://www.springer.com/journal/12303 http://dx.doi.org/10.1007/s12303-017-0034-4
1052 Byung-Gon Chae, Hyuck-Jin Park, Filippo Catani, Alessandro Simoni, and Matteo Berti

synthetic aperture is carried out by exploiting sub-scenes collected slope geometries which are not perfectly suitable for near-
by the radar while moving along a linear rail. So, while in space horizontal LOS, such as terraced slopes or open pit benches, a
the revisiting of the same area is only possible after one entire solution is offered by the combination of terrestrial and spaceborne
orbit has been completed, in GB-InSAR it depends on the SAR, where the latter contributes to the definition of sub-
length of the linear rail and the cumulated acquisition time, a vertical vector components within the displacement mass. One
matter of minutes. This gives terrestrial systems a much higher such cases is reported in Bardi et al. (2016) for the large rock
capability for monitoring rapid landslides than space-borne instability at Aknes (Norway).
SARs. Furthermore, they may be installed for optimizing the
viewing geometry while satellite InSAR have incidence angle 4. LANDSLIDE EARLY WARNING
and orbit that allow an acceptable sensitivity only towards vertical
and across-azimuth (roughly E-W) displacement components. Landslide early warning is essential for early or advance
Detailed descriptions of terrestrial SAR systems used for ground recognition of landslide indicators, so that residents can be
displacement measurement are given by Montserrat et al. (2014) evacuated from potential landslide areas to reduce the damage
and Caduff et al. (2015). caused by landslides. Early detection of landslides in a broad
Applications of GB-InSAR to landslide monitoring are as old natural terrain can be possible by the monitoring of rainfall and
as those of space-borne InSAR. Two of the first such studies are changes in the soil physical properties in real time or near-real
provided by Tarchi et al. (2003) and Pieraccini et al. (2003). In time. Most landslide warning methods use triggering thresholds
both studies the authors use radar interferometers sliding along that are determined based on rainfall and the soil physical
linear rails to monitor landslide displacements in real-time. properties. A threshold is the minimum or maximum level of
Tarchi et al. (2003) measured 15 days of slow displacement of some quantity needed for a process to take place or a state to
the Tessina landslide with velocities in the order of 30 mm/hr by change (White et al., 1996). Therefore, it is important to secure
using a frequency of acquisition of 14 minutes and a spatial practical and reliable triggering thresholds for landslide early
resolution of 2 m in both range and cross range. A similar warning.
approach was then repeated in 6 different sites in Italy by Even when slow landslides are associated with low destructive
Antonello et al. (2004) that used the same hardware based on significance, monitoring and early warning are essential for
the Linear SAR (LISA) apparatus developed by JRC. A more hazard management. Recent studies have reported the development
detailed discussion on the implication of the use of GB-InSAR of early warning systems for natural and engineered slopes (Vaziri
for the monitoring of unstable slopes in various geomorphological et al., 2010; Intrieri et al., 2012; Thiebes et al., 2014; Atzeni et al.,
environments is provided later on by Casagli et al. (2010) that 2015), and an international workshop held in Italy in 2013 focused
apply an early warning scheme for the prediction of a landslide on the challenges, problems, and available tools to determine
in the Italian Alps and a large slope instability on the Stromboli warning criteria (Cloutier et al., 2014). In addition, many studies
volcano. More recently, the use of GB-InSAR has evolved dealt with early warning, instrumentation, and monitoring during
towards a semi-automated use in early warning threshold and the Second World Landslide Forum (Margottini et al., 2013).
alert setting by inverse velocity methods. The latter, despite being These recent efforts highlight the need for reliable monitoring
suitable only for progressive failure modes, are increasingly used and early warning systems as a key part of cost-effective hazard
in large open pit mining operations and for large rock instabilities. management (Macciotta et al., 2016).
Dick et al. (2014), in an important paper, analyze the application As landslides occur under heavy rainfall condition in many
of pixel-based displacement field as measured by slope-stability cases, a triggering threshold of landslide is related to the rainfall
SAR to apply real-time time-of-failure (TOF) prediction schemes and the hydrological conditions in soil. In general, a rainfall
in open pit mines. For natural landslides, a detailed methodology to threshold can be determined empirically (statistically) or physically
implement early warning systems based on GB-InSAR integrated (deterministically) (Corominas, 2000; Aleotti, 2004; Wieczorek
to traditional monitoring tools is presented by Intrieri et al. and Glade, 2005; Guzzetti et al., 2007). Empirical thresholds are
(2012). Finally, new approaches to overcome known problem of defined by collecting rainfall data for meteorological events both
terrestrial SAR are already available. A reduction in the time with and without landslides, whereas physical thresholds are
delay between acquisitions for linear SARs can be obtained by based on numerical models considering the relationship among
exploiting sub-apertures to focus on the most unstable parts of a rainfall, pore water pressure, VWC, suction stress, and slope stability.
slope, as in the case reported by Bardi et al. (2017). This technique Guzzetti et al. (2008) grouped empirical rainfall thresholds into
can be useful to maintain all the advantages of synthetic three broad categories: (i) thresholds that combine precipitation
aperture whilst increasing the acquisition frequency. In case of measurements obtained for specific rainfall events, (ii) thresholds

http://dx.doi.org/10.1007/s12303-017-0034-4 http://www.springer.com/journal/12303
Landslide prediction, monitoring, and early warning: a concise review of state-of-the-art 1053

that include the antecedent conditions (Terlien, 1998; Crozier, independent of local morphological, lithological, and land use
1999; Glade et al., 2000; Chleborad, 2003; Aleotti, 2004), and (iii) conditions and of local or regional rainfall pattern and history.
other thresholds, including hydrological (Reichenbach et al., Global thresholds have been proposed by Caine (1980), Crosta
1998; Jakob and Weatherly, 2003). Thresholds established using and Frattini (2001), and Cannon and Gartner (2005). Regional
precipitation measurements obtained from individual or multiple thresholds are potentially suited for landslide warning systems
rainfall events can be further subdivided into intensity-duration based on quantitative spatial rainfall forecasts, estimates, or
(ID) thresholds, thresholds based on the total event rainfall (E), measurements and are defined for areas extending from a few to
rainfall event-duration thresholds (ED), and rainfall event-intensity several thousand square kilometres of similar climatic and
(EI) thresholds (Guzzetti et al., 2007). These thresholds are generally physiographic characteristics. Regional thresholds have been
obtained by drawing lower bound lines to the rainfall conditions proposed by Salciarini et al. (2006) and Melchiorre and Frattini
that resulted in landslides, plotted in Cartesian, semi-logarithmic, or (2012). Local thresholds are applicable to single landslides or to
logarithmic co-ordinates. Deterministic-based models attempt a group of landslides in areas extending from a few to some
to spatially extend the slope stability models adopted in geotechnical hundreds of square kilometres and explicitly or implicitly consider
engineering (Iverson, 2000; Godt, 2004; Capparelli and Versace, the local climatic regime, geomorphologic setting, and geological
2011). Deterministic-based models can be applied to global, regional, conditions.
or local thresholds. Moreover, deterministic-based models can Guzzetti et al. (2008) suggested a global ID threshold for the
determine the amount of precipitation needed to trigger slope initiation of shallow landslides and debris flows based on worldwide
failures, and the location and time of the expected landslides, rainfall and landslide information. To account for climatic effects,
making them of interest for landslide warning systems (Chung the rainfall intensity was normalized by the mean annual
et al., 2017). precipitation (MAP) and the rainy-day normal (RDN, Wilson
Rainfall thresholds for the initiation of landslides have been and Jayko, 1997), and separate thresholds were established for
defined as global, regional, or local. A global threshold provides the normalized rainfall data. They compared the new single and
a general minimum level below which landslides do not occur, composite thresholds with five published global ID thresholds,

Fig. 7. Comparison between the global


ID thresholds defined by Guzzetti et al.
(2008) and published global (world-
wide) ID rainfall thresholds. 1. Caine
(1980); 2. Innes (1983); 3. Clarizia et al.
(1996); 4. Crosta and Frattini (2001); 5.
Cannon and Gartner (2005); 6. threshold
inferred from the entire set of ID rainfall
data (Guzzetti et al., 2008); 7. thresholds
inferred from the probability estimates
of the rainfall conditions, for two differ-
ent rainfall periods (D < 48 h, and D ≥ 48
h) (Guzzetti et al., 2008). Dashed line
shows 0.25 mm h-1 rainfall intensity
(after Guzzetti et al., 2008).

http://www.springer.com/journal/12303 http://dx.doi.org/10.1007/s12303-017-0034-4
1054 Byung-Gon Chae, Hyuck-Jin Park, Filippo Catani, Alessandro Simoni, and Matteo Berti

including those proposed by Caine (1980); Innes (1983); Crosta established between the maxima cumulative rainfall values and
and Frattini (2001), and Cannon and Gartner (2005). The suggested the soil moisture prior to the triggering of landslides would make
single and composite thresholds were lower or significantly lower it possible to dynamically adjust the rainfall thresholds. Piciullo
than the other published global thresholds, for most rainfall et al. (2016) proposed a method based on a process chain to realize
durations (Fig. 7). an objective procedure for the definition and the evaluation of a
Aleotti (2004) studied the critical limit described by the reliable threshold-based operational early warning system in the
equation: I = 19D0.50 (where I is the rainfall intensity expressed Campania region, southern Italy. First, they adopted a consolidated
in mm/h and D is the rainfall duration expressed in h). Such a approach (Brunetti et al., 2010; Peruccacci et al., 2012; Gariano
limit is traced to envelop 90% of the points on the graph. In the et al., 2015; Melillo et al., 2015; Melillo et al., 2016) to define and
NI-D (normalized intensity-duration) diagram, the triggering validate empirical, cumulated event rainfall-rainfall duration
thresholds are given by the equations NI = 0.76D0.33 and NI = (ED) thresholds for possible landslide occurrences. Afterwards,
4.62D0.79 (where NI is the normalized intensity with respect to they proposed a method for issuing warning levels, as a result of
the annual precipitation, and MAP is expressed in %, [(mm/h)/ the comparison between measured rainfall and established
PMA] 100). Finally, the diagram normalized intensity-normalized thresholds. Finally, the authors assessed the performance of the
cumulative critical (NI-NCR) is proposed; the triggering threshold regional landslide early warning model (ReLWaM) employing
is given by the expression: NI = 0.09ln[NCR] + 0.54 (where the event, duration matrix, performance (EDuMaP) method,
NCR is the normalized cumulative critical rainfall, [mm/PMA] proposed by Calvello and Piciullo (2016).
100) (Fig. 8). Considering the relationship between rainfall and debris flow,
Several previous studies have suggested a regional threshold Jan et al. (2004) suggested a rainfall-based debris flow warning
of rainfall. Ponziani et al. (2012) discussed rainfall thresholds model. A method to define a rainfall event and its antecedent
and soil moisture modelling for operational hydrogeological rainfall used to define the rainfall parameters such as the rainfall
risk prevention in terms of early warning. Comparison with intensity, the rainfall duration, the accumulated rainfall, and the
rainfall thresholds used by the Umbria Region Civil Protection effective accumulated rainfall was proposed to analyse the risk
Decentrate Functional Centre in Italy was also carried out to of debris flows caused by rainstorms. The product of the hourly
evaluate the reliability of the procedures employed for landslide rainfall intensity and the effective accumulated rainfall was defined
warning. The main result of the analysis is the quantification of as the rainfall triggering index (RTI) that was used to set up a
the decreasing linear trend between the maximum cumulated rainfall-based debris flow warning model. They proposed a
rainfall values over 24, 36, and 48 h and the soil moisture conditions method to determine the lower critical RTI value and the upper
prior to landslide events. A linear trend between the rainfall critical RTI value, based on the RTI values of historical rainfall
thresholds and the initial soil moisture conditions was found, events. After the two critical RTI values are determined in a rainfall
with correlation coefficients up to 0.60, showing the key role of event, one can estimate the instant debris flow occurrence
soil moisture on landslide triggering. Therefore, the correlation potential, with a diagram having instant RTI values at time t on
the ordinate and the variation of time t on the abscissa.
Based on the rainfall thresholds, some landslide early warning
systems have been developed. The Research Institute for Geo-
Hydrological Protection (IRPI), of the Italian National Research
Council (CNR), are implementing SANF, an early warning system
for forecasting possible rainfall-induced landslides in Italy (Rossi
et al., 2012). The system is based on two components. The first
component consists of: (i) a set of national, regional, and local
rainfall thresholds (of the ID type) for possible landslide imitation,
(ii) a database of sub-hourly rainfall measurements obtained by
a network of 1950 rain gauges in Italy, and (iii) quantitative
rainfall forecasts acquired through numerical modelling. Every
day, and for each individual rain gauge, the system compares the
measured and the forecast rainfall amounts against pre-defined
thresholds, and assigns a probability for possible landslide occurrence
Fig. 8. Definition of critical triggering thresholds in NI-NCR graph.
Curve [100] envelops all the available data; curve [90] encompasses
to each rain gauge. This information is used to prepare synoptic-
90% of the points in the graph (Aleotti, 2004). scale maps showing where rainfall-induced landslides are expected,

http://dx.doi.org/10.1007/s12303-017-0034-4 http://www.springer.com/journal/12303
Landslide prediction, monitoring, and early warning: a concise review of state-of-the-art 1055

recorded by an automated regional network of rain gauges.


Cumulative rainfall values are then compared with a series of
thresholds corresponding to increasing levels of rainstorm severity,
with which a warning level is associated. The system operates at
a spatial resolution of a few hundreds of square kilometres: the
hilly and mountainous territory of Emilia Romagna is partitioned
into 19 territorial units, each provided with a reference rain gauge
and its own rainfall thresholds.
As stated earlier, rainfall is the most important threshold of
landslide early warning. However, there can be differences in
landslide triggering under the same rainfall conditions. Early
warning for both shallow and deep landslides requires rainfall data
measured by rain gauges in landslide-prone areas. Nevertheless,
the effect of rainfall is difficult to assess quantitatively because it
depends on several factors, including the heterogeneity of the
soils. It is thus necessary to analyse the relationships between
soil properties and landslide triggering based on the physical
and mechanical thresholds in each case (Chae and Kim, 2012).
Physical thresholds have several limitations for application to
broad areas because of the variations in conditions pertaining to
each influential factor. To overcome these limitations, Glade (2000)
proposed a model of antecedent soil water status that considered
the factors related to both empirical thresholds, such as antecedent
rainfall, and physical properties, such as soil moisture and
Fig. 9. Logical framework adopted for the SANF early warning sys- potential evaporation. Huggel et al. (2010) also implemented an
tem for rainfall-induced landslides in Italy (Rossi et al., 2012). empirical rainfall threshold function in their numerical model
and considered a time period of 10 years divided into smaller
intervals of 6 h. Sirangelo and Versace (1996) developed the
within a period of time. The second component of the system forecasting of landslides induced by rainfalls (FLaIR) model to
consists of synoptic assessments of landslide hazard and risk in forecast landslide movements activated by rainfall within a
Italy, including small-scale zoning maps. The assessments are hydrological conceptual modelling framework. This model
obtained through statistical modelling of thematic and considers the relationship between the mobility function and
environmental information, including national catalogues of landslide probability dependent on the antecedent rainfall. Because
historical landslides and of historical landslides with human the mobility function is defined as a function of the infiltration
consequences in Italy, in the period 1900–2005 (Fig. 9). The rate, which is directly related to soil and topographical conditions,
combination of the hazard and risk zonation with the daily the FLaIR model is partly based on physical approaches. The
forecasts for possible landslide occurrence allows the Italian National system is composed of a monitoring module, taking into account
Department for Civil Protection to improve management of the recorded rainfall data, and a warning module, which employs
potential landslide risk conditions in Italy. the forecasted rainfall data. In particular, this latter module
A landslide warning system, SIGMA (Sistema Integrato Gestione considers the data provided by meteorological models which
Monitoraggio Allerta), is an integrated service for the management give rainfall forecasts for 6, 12, 18, and 24 h ahead. By using these
and monitoring of alerts. SIGMA is based on a decisional values, the system is able to indicate possible developments over the
algorithm that combines the overcoming of different rainfall coming 24 h (Capparelli and Tiranti, 2010). The FLaIR model was
thresholds and defines a corresponding alert level; the thresholds later extended to mudflow movement (Sirangelo and Versace,
are based on the rainfall amount accumulated considering time 1996) and so it led to an early warning system called MoniFLaIR
intervals ranging from 1 to 243 days and make it possible to take (Sirangelo and Braca, 2004; Capparelli and Tiranti, 2010).
into account both shallow and deep-seated landslides (Martelloni Chae and Kim (2012) analysed the relationship between slope
et al., 2012; Lagomarsino et al., 2013). The SIGMA warning failure and the VWC gradient that is dependent on rainfall
system combines rainfall forecasts with hourly rainfall measurements conditions. In cases of slope failure, the VWC gradient was distributed

http://www.springer.com/journal/12303 http://dx.doi.org/10.1007/s12303-017-0034-4
1056 Byung-Gon Chae, Hyuck-Jin Park, Filippo Catani, Alessandro Simoni, and Matteo Berti

in the range of 0.107–0.249, whereas field monitoring results analysis considering suction stress in unsaturated soil, the SWCC
without slope failure showed a gradient range of VWC between and the suction stress characteristic curve of the unsaturated
0.003 and 0.073 in the Deoksan study area, Korea. The results soil obtained from the study area were estimated using the van
demonstrate that a large amount of rainfall and a high gradient Genuchten (1980) and the Lu and Likos (2006) models. The FS
of VWC contributes to slope failure. Based on the results, they of the natural slope exhibited continuous fluctuations because
suggested a threshold for VWC gradient for early warning of of changes in the suction stress caused by the evaporation and
landslides that are triggered by rainfall. The results showed that infiltration of water in the unsaturated soil layer. Therefore, the
a landslide tends to be triggered at a VWC gradient of over 0.1 infinite slope stability of a slope in natural terrain can be
in the study area. evaluated in real time by estimating the suction stress caused by
Thiebes et al. (2014) integrated a limit equilibrium model into rainfall in the unsaturated soil, when the VWC or matric suction
a landslide early warning system. A physically based slope stability in the soil is being monitored in the field.
model, the combined hydrology and stability model (CHASM) Early warning systems for natural hazards are important tools
was initially applied to a reactivated landslide in the Swabian for disaster risk reduction. Early warning systems have been
Alps to assess stability conditions, and was subsequently integrated developed for a number of different hazards, including tsunamis,
into a prototype of a semi-automated landslide early warning volcanoes, snow avalanches, landslides, and others (Zschau and
system. For the assessment of slope stability and early warning Kuppers, 2003). Early warning systems commonly consist of
modelling, the CHASM combines the simulation of saturated different components, such as (i) sensors measuring geophysical,
and unsaturated hydrological processes to calculate pore water atmospheric, hydrodynamic, and soil-related parameters, (ii)
pressures, which are then incorporated into the computation of telecommunication equipment transmitting the data to a (iii)
slope stability by means of limit equilibrium analysis. The results of monitoring and analysis centre, (iv) decision procedures and
the CHASM application demonstrate that for several potential organizational structures that facilitate the translation of the
shear surfaces, the FS is relatively low, and subsequent rainfall technical data into commonly understood information, and (v)
events could cause instability. To integrate and automate CHASM response of the people who might be affected. The awareness
within an early warning system, international geospatial standards and knowledge of the people exposed to a hazard are very important
were used to ensure the interoperability of system components for their adequate response to an early warning. (Huggel et al.,
and the transferability of the implemented system as a whole. 2010).
The CHASM algorithm is automatically run as a web processing As the types and mechanisms of landslide early warning
service, using fixed, predetermined input data, and variable input systems vary, a universal standard needs to be developed so that
data including hydrological monitoring data and quantitative uniformity in the implementation of an early warning system
rainfall forecasts. When the pre-defined modelling or monitoring and improvement in the community’s and stakeholders’
thresholds are exceeded, a web notification service distributes preparedness in landslide-prone areas can be obtained (Jacks et
text and email messages to relevant experts, who then determine al., 2010). To ensure the widespread applicability of the early
whether to issue an early warning to local and regional stakeholders, warning systems, a standard should provide consistent and clear
as well as providing appropriate action advice. technical terms and definitions, as well as requirements and
The soil-water characteristic curve (SWCC), which is related general phases in the implementation of landslide early warning
to the pore water pressure and the water content, allows for the systems. It should also regulate the standardization of commonly
characterization of hydrological and mechanical behaviour of used monitoring equipment, warning criteria, the style of evacuation
unsaturated soils. The hysteretic nature of SWCC (Lu and Likos, process, and types of disaster preparedness and response organization.
2006; Fredlund et al., 2011; Likos et al., 2013; Lu et al., 2013), Although flexibility is required during the implementation, it
linked to in situ processes because of different drying and wetting should adhere to the key elements contained within the standard.
cycles that the soils suffer under natural conditions, determines Taking into account the four key elements of a people-centred early
the development of a main drying and a main wetting curve, warning system (UN-ISDR, 2006) and the hybrid sociotechnical
and thus it can have practical implications on water movement approach for disaster risk reduction (Karnawati et al., 2013), a
in soils and on the mechanical behaviour of unsaturated soils in universal standard for landslide early warning systems that
terms of deformation and shear strength (Wheeler et al., 2003; comprises seven sub-systems was proposed by Fathani et al.
Likos et al., 2013; Bordoni et al., 2015a). Song et al. (2016) evaluated (2016) (Fig. 10). It should be clarified that the monitoring and
the stability of an unsaturated slope on a natural terrain based warning services that to date are considered as the core of the early
on the variation in suction stress induced in the soil by rainfall warning systems will remain an important part of the disaster
infiltration with time. To apply the concept of slope stability management programme.

http://dx.doi.org/10.1007/s12303-017-0034-4 http://www.springer.com/journal/12303
Landslide prediction, monitoring, and early warning: a concise review of state-of-the-art 1057

Fig. 10. A proposed algorithm for a standard landslide early warning system (Fathani et al., 2016).

5. CONCLUSIONS is the collection of data regarding the landslide distribution and


factor maps over large areas. In addition, data-driven methods
In this paper, the authors reviewed the recent advances and consider only the statistical relationships between landslides
the state-of-the-art in the essential components of the landslide and related factors, not the failure mechanism. On the other hands,
hazard assessment such as landslide susceptibility analysis, runout physically based landslide susceptibility assessment methods
modeling, landslide monitoring and early warning. Landslide are based on the modelling the process of landslide occurrence
susceptibility assessment methods are classified into qualitative using physical models of landslides. The physically based approach
(knowledge-driven) and quantitative (data-driven and physically can analyze slope stability regardless of landslide occurrence by
based) methods, depending on the way they treat landslide- utilizing physical slope models. Therefore, physically based models
inducing factors and models. Knowledge-driven methods are have been widely used because they have a higher predictive
seldom used for susceptibility assessment over large areas because capability and they are the most suitable for quantitatively assessing
they lack a concrete physical concept of slope failure. Meanwhile, the influence of individual parameters that contribute to landslide
data-driven methods evaluate the statistical relationships between initiation. In the physically based model, however, sufficient
the locations of landslides occurred in the past and landslide- and accurate information is required to obtain an accurate
inducing factors, and then, quantitative predictions are made landslide susceptibility assessment. But in most practical cases,
for landslide free areas with similar conditions. Therefore, data- data are often limited in extent and have imperfections or
driven methods have been used for evaluating various spatially variable quality. Consequently, uncertainties are involved in the
distributed landslide-inducing factors and analysing landslide analysis procedure and they are recognized as an important
susceptibility. However, the difficulty in using data-driven methods cause of mismatch between simulated and observed distributions

http://www.springer.com/journal/12303 http://dx.doi.org/10.1007/s12303-017-0034-4
1058 Byung-Gon Chae, Hyuck-Jin Park, Filippo Catani, Alessandro Simoni, and Matteo Berti

of landslide occurrence. Therefore, the probabilistic analysis in potential landslides and measures changes to attribute values
method has been adopted to quantify the uncertainties. In this of landslide triggering factors with the aim of reducing damage
manuscript, the overviews of probabilistic analysis application by landslides. The monitoring of kinematic, hydrological, and
in landslide susceptibility analysis at catchment scale are discussed. climatic parameters plays a significant role in supporting the
When landslides transform in fast-moving flows, their ability development of slope stability models. In situ ground-based
to travel can substantially increase and the prediction of runout monitoring usually observes precipitation, slope displacement,
becomes essential to define hazard and risk zoning at all levels. and hydrological / physical properties of soils such as groundwater
A major challenge in hazard assessment is to define the expected level, VWC, pore water pressure, and matric suction. In situ
magnitude of the event. In fact, regardless of the methodology ground-based monitoring can be grouped into three categories:
that will be chosen to simulate the behavior of the flow, a “design” rainfall monitoring, slope displacement monitoring, and monitoring
event must be defined. Historical information and geological/ of hydrological and mechanical properties of soils. Because rainfall
geomorphological field surveys can be helpful to constrain a represents the most common triggering factor, it is important to
range of expected magnitude. Anyway, it is usually difficult to analyse the relationship between rainfall conditions and landslide
forecast how much material will transform into a flow and also triggering based on the results of rainfall measurements. In the
the potential to entrain material along the flow path should be concept of displacement monitoring, GPS, tiltmeters, inclinometers,
further investigated. and extensometers are generally installed on the surface or in
For preliminary estimates of runout, empirical-statistical the subsurface of a slope to study the deformation behaviour of
methods can be conveniently used due to their ease of application rainfall-triggered landslides. The third monitoring concept of
and possible implementation in GIS environment. Provided rainfall-induced shallow landslides is linked to continuous
that the adopted methodology is tailored to the specific type of monitoring of hydrological and mechanical properties of the soil.
landslide, results can be helpful to delineate the extent of potentially Many studies demonstrated the ability of monitoring techniques
affected areas. In some cases, they can also be associated to specific to identify the soil’s hydrological and mechanical conditions
confidence levels. during shallow landslide triggering.
Dynamic methods require the selection of an appropriate Remote sensing methods have a very high effectiveness for
flow regime capable to describe the behavior of the flow during the exploration of surface changes. If the detection may be
propagation and deposition. They generally have the ability to carried out repeatedly with the suitable frequency in time, then
predict other flow parameters beside the extent of runout. Flow the RS method may also be used as an actual monitoring tool.
velocities, flow heights and peak discharges can be used to improve The traditionally used methods are based on optical airborne
the assessment of hazard and can be useful to design possible photogrammetry and space-borne multispectral imaging. Since
risk mitigation measures. Anyway, specific criteria to distinguish then, several new tools and methods have become available that
appropriate flow regimes for the different types of fast-moving relate mainly to the three groups which were introduced at the
landslides and the different types of involved materials are beginning, i.e., ground-based, air-borne and space-borne. It must
generally lacking in the literature and trial and error procedures be said, however, that despite the increase in acquisition frequency
are usually adopted to compare the results with field evidences and in the number of available post-processing tools, the use of
and/or historical information. Also the parameters of the selected optical data for landslide analysis it is still more a target/change
rheological model cannot be chosen based on well defined “a detection technique rather than an actual real-time monitoring
priori” criteria, and calibration is required based on back-analysis tool, mainly because the latter would require the measurement
of past events. There is an urgent need to boost research efforts of displacement velocity and the quantification of volumes,
aimed to broaden the systematic documentation of events and difficult to implement in a seamless operational chain based on
their quantitative back-analysis using various rheological models. photogrammetry.
Only such efforts will provide much needed guidelines for the Lasers and radars are the most used as active sensors in
selection of appropriate rheologies and associated parameters landslide studies. Lasers are mostly used through scanners, both
for the different types of fast-moving landslides. In this perspective, air-borne (ALS) and terrestrial (TLS), which generate a multiple
the connection and the dialogue between field scientists and laser beam and measure the energy which is reflected back from
model developers assumes an increasing importance and must the target. The repetition of the same measurements at different
be used to advance our prediction skills and judge the reliability times allows for the use of ALS and TLS systems in landslide
of our results. monitoring. Some of monitoring limitations for laser scanning
Landslide monitoring is essential for the recognition of landslide are avoided by resorting to a different wavelength and a different
occurrence as well as early warning. It observes slope displacements processing technique, by using radars and radar interferometry.

http://dx.doi.org/10.1007/s12303-017-0034-4 http://www.springer.com/journal/12303
Landslide prediction, monitoring, and early warning: a concise review of state-of-the-art 1059

Microwaves are able to penetrate fog, clouds, light rain and even Therefore, physical thresholds are based on numerical models
dust in most of the cases. Furthermore, by using radar interferometry, considering the relationship among rainfall, pore water pressure,
the analysis of data can be automated and fast with the displacement VWC, suction stress, and slope stability.
measured at every new reading with a frequency as low as 1
image every 0.5 min. The potential of classical 2-scene DInSAR ACKNOWLEDGMENTS
has been proven over a large range of areas and geophysical
conditions worldwide. However, it is only usable for detecting This research was partly supported by National Research
movements of the ground that occurred between the two Foundation of Korea (NRF) grant funded by Korea government
acquisitions and subject to the condition that the field of motion (MSIP) (No. NRF-2016R1A2B4008963) and the Basic Research
must be well auto-correlated in space and have a smooth gradient Project (Grant No. 17-3413) of the Korea Institute of Geoscience
in time. Another limitation is that the maximum accuracy of and Mineral Resources (KIGAM) funded by the Ministry of
DInSAR is at the centimeter-scale whilst most of deep-seated Science, ICT & Future Planning of Korea.
landslides move very slowly, at rates of a few millimeters per
year. For slower movement detection, a multi-temporal PS-InSAR REFERENCES
technique has to be used, that relies on stable natural ground Abellán, A., Vilaplana, J.M., and Martínez, J., 2006, Application of a
reflectors. long-range Terrestrial Laser Scanner to a detailed rockfall study at
At present, the best solution for slope-scale monitoring is the Vall de Núria (Eastern Pyrenees, Spain). Engineering Geology, 88,
adoption of terrestrial interferometers, usually referred to as 136–148.
ground-based radars, with (GB-InSAR) or without synthetic Abellán, A., Calvet, J., Vilaplana, J.M., and Blanchard, J., 2010, Detec-
aperture. Terrestrial systems have a much higher capability for tion and spatial prediction of rockfalls by means of terrestrial laser
scanner monitoring. Geomorphology, 119, 162–171.
monitoring rapid landslides than space-borne SARs. Furthermore,
Abellán, A., Jaboyedoff, M., Oppikofer, T., and Vilaplana, J.M., 2009,
they may be installed for optimizing the viewing geometry
Detection of millimetric deformation using a terrestrial laser scan-
while satellite InSAR have incidence angle and orbit that allow ner: experiment and application to a rockfall event. Natural Haz-
an acceptable sensitivity only towards vertical and across-azimuth ards Earth System Sciences, 9, 365–372.
displacement components. Adler, R.F., Huffman, G.J., Bolvin, D.T., Curtis, S., and Nelkin, E.J.,
Landslide early warning is essential for recognizing landslide 2000, Tropical rainfall distributions determined using TRMM com-
indicators early or in advance and for evacuating residents from bined with other satellite and rain gauge Information. Journal of
potential landslide areas to reduce the damage caused. Early Applied Meteorology, 39, 2007–2023.
Aleotti, P., 2004, A warning system for rainfall-induced shallow fail-
detection of landslides in a broad natural terrain can be possible
ures. Engineering Geology, 73, 247–265.
by the monitoring of rainfall and changes in the soil physical Ali, A., Huang, J., Lyamin, A.V., Sloan, S.W., Griffiths, D.V., Cassidy,
properties in real time or near-real time. Most landslide warning M.J., and Li, J.H., 2014, Simplified quantitative risk assessment of
methods use triggering thresholds that are determined based on rainfall-induced landslides modelled by infinite slopes. Engineer-
rainfall and soil physical properties. A rainfall threshold can be ing Geology, 179, 102–116.
determined empirically or physically. Empirical rainfall thresholds Alvioli, M., Guzzetti, F., and Rossi, M., 2014, Scaling properties of rain-
are grouped into three broad categories, namely: (i) thresholds fall induced landslides predicted by a physically based model. Geo-
morphology, 213, 38–47.
that combine precipitation measurements obtained for specific
Angeli, M.G., Pausto, A., and Silvano, S., 2000, A critical review of land-
rainfall events, (ii) thresholds that include the antecedent conditions,
slide monitoring experiences. Engineering Geology, 55, 133–147.
and (iii) other thresholds, including hydrological thresholds. Antonello, G., Casagli, N., Farina, P., Leva, D., Nico, G., Sieber, A.J., and
Thresholds established using precipitation measurements obtained Tarchi, D., 2004, Ground-based SAR interferometry for monitor-
from individual or multiple rainfall events can be further subdivided ing mass movements. Landslides, 1, 21–28.
into ID thresholds, thresholds based on the total event rainfall Apip, T.K., Yamashiki, Y., Sassa, K., Ibrahim, A.B., and Fukuoka, H.,
(E), rainfall event-duration thresholds (ED), and rainfall event- 2010, A distributed hydrological-geotechnical model using satel-
intensity (EI) thresholds. lite-derived rainfall estimates for shallow landslide prediction sys-
tem at a catchment scale. Landslides, 7, 237–258.
Rainfall is the most important threshold of landslide early
Arnone, E., Noto, L.V., Lepore, C., and Bras, R.L., 2011, Physically-
warning. However, the effect of rainfall is difficult to assess
based and distributed approach to analyze rainfall-triggered land-
quantitatively because it depends on several factors, including slides at watershed scale. Geomorphology, 133, 121–131.
the heterogeneity of soils. It is thus necessary to analyse the Atzeni, C., Barla, M., Pieraccini, M., and Antolini, F., 2015, Early warn-
relationships between soil properties and landslide triggering ing monitoring of natural and engineered slopes with ground-
based on the physical and mechanical thresholds in each case. based synthetic-aperture radar. Rock Mechanics and Rock Engi-

http://www.springer.com/journal/12303 http://dx.doi.org/10.1007/s12303-017-0034-4
1060 Byung-Gon Chae, Hyuck-Jin Park, Filippo Catani, Alessandro Simoni, and Matteo Berti

neering, 48, 235–246. Berti, M., Genevois, R., LaHusen, R., Simoni, A., and Tecca, P.R., 2000,
Baecher, G.B. and Christian, J.T., 2003, Reliability and Statistics in Geo- Debris flow monitoring in the Acquabona Watershed on the Dolo-
technical Engineering. Wiley, New York, 618 p. mites (Italian Alps). Physics and Chemistry of the Earth, Part B:
Bardi, F., Raspini, F., Ciampalini, A., Kristensen, L., Rouyet, L., Lauknes, Hydrology, Oceans and Atmosphere, 25, 707–715.
T.R., Frauenfelder, R., and Casagli, N., 2016, Space-borne and Beven, K.J. and Kirkby, M.J., 1979, A physically based, variable contrib-
ground-based InSAR data integration: the Åknes test site. Remote uting area model of basin hydrology. Hydrological Sciences Bulle-
Sensing, 8, 237. tin, 24, 43–69.
Bardi, F., Raspini, F., Frodella, W., Lombardi, L., Nocentini, M., Gigli, Bittelli, M., Valentino, R., Salvatorelli, F., and Pisa, P.R., 2012, Monitor-
G., Morelli, S., Corsini, A., and Casagli, N., 2017, Monitoring the ing soil-water and displacement conditions leading to landslide
rapid-moving reactivation of earth flows by means of GB-InSAR: occurrence in partially saturated clays. Geomorphology, 173, 161–
the April 2013 Capriglio landslide (Northern Appennines, Italy). 173.
Remote Sensing, 9, 165. Blanco-Sánchez, P., Mallorquí, J.J., Duque, S., and Monells, D., 2008,
Barlow, J., Martin, Y., and Franklin, S.E., 2003, Detecting translational The coherent pixels technique (CPT): an advanced DInSAR tech-
landslide scars using segmentation of Landsat ETM+ and DEM nique for nonlinear deformation monitoring. Pure and Applied
data in the northern Cascade Mountains, British Columbia. Cana- Geophysics, 165, 1167–1193.
dian Journal of Remote Sensing, 29, 510–517. Bordini, M., Meisina, C., Valentino, R., Lu, N., Bittelli, M., and Cher-
Baum, R.L. and Godt, J.W., 2010, Early warning of rainfall-induced sich, S., 2015a, Hydrological factors affecting rainfall-induced shal-
shallow landslides and debris flows in the USA. Landslides, 7, 259– low landslides: from the field monitoring to a simplified slope
272. stability analysis. Engineering Geology, 193, 19–37.
Baum, R.L., Godt, J.W., and Savage, W.Z., 2010, Estimating the timing Bordoni, M., Meisina, C., Valentino, R., Bittelli, M., and Chersich, S.,
and location of shallow rainfall-induced landslides using a model 2015b, Site-specific to local-scale shallow landslides triggering
for transient, unsaturated infiltration. Journal of Geophysical zones assessment using TRIGRS. Natural Hazards & Earth System
Research, 115, F03013. Sciences, 15, 1025–1050.
Baum, R.L., Savage, W.Z., and Godt, J.W., 2002, TRIGRS – a fortran Bruckl, E., Brunner, F.K., Lang, E., Mertl, S., Muller, M., and Stary, U.,
program for transient rainfall infiltration and grid-based regional 2013, The Gradenbach observatory – monitoring deep-seated
slope-stability analysis. U.S. Geological Survey Open-File Report gravitational slope deformation by geodetic, hydrological, and seis-
02-0424. http://pubs.usgs.gov/of/2002/ofr-02-424 mological methods. Landslides, 10, 815–829.
Baum, R.L., Savage, W.Z., and Godt, J.W., 2008, TRIGRS – a fortran Brunetti, M.T., Peruccacci, S., Rossi, M., Luciani, S., Valigi, D., and Guz-
program for transient rainfall infiltration and grid-based regional zetti, F., 2010, Rainfall thresholds for the possible occurrence of
slope-stability analysis, Version 2.0. U.S. Geological Survey Open- landslides in Italy. Natural Hazards and Earth System Sciences, 10,
File Report 2008-1159. http://pubs.usgs.gov/of/2008/1159 447–458.
Baum, R.L., Coe, J.A., Godt, J.W., Harp, E.L., Reid, M.E., Savage, W.Z., Burton, A., Arkell, T.J., and Bathurst, J.C., 1998, Field variability of
Schulz, W.H., Brien, D.L., Chleborad, A.F., McKenna, J.P., and landslide model parameters. Environmental Geology, 35, 100–114.
Michael, J.A., 2005, Regional landslide-hazard assessment for Seat- Caduff, R., Schlunegger, F., Kos, A., and Wiesmann, A., 2015, A review
tle, Washington, USA. Landslides, 2, 266–279. of terrestrial radar interferometry for measuring surface change in
Berardino, P., Fornaro, G., Lanari, R., and Sansosti, E., 2002, A new the geosciences. Earth Surface Processes and Landforms, 40, 208–
algorithm for surface deformation monitoring based on small 228.
baseline differential SAR interferograms. IEEE Transactions on Caine, N., 1980, The rainfall intensity-duration control of shallow land-
Geoscience and Remote Sensing, 40, 2375–2383. slides and debris flows. Geografiska Annaler Series A: Physical
Berardino, P., Costantini, M., Franceschetti, G., Iodice, A., Pietranera, Geography, 62, 23–27.
L., and Rizzo, V., 2003, Use of differential SAR interferometry in Calvello, M. and Piciullo, L., 2016, Assessing the performance of
monitoring and modelling large slope instability at Maratea (Basili- regional landslide early warning models: the EDuMaP method.
cata, Italy). Engineering Geology, 68, 31–51. Natural Hazards and Earth System Sciences, 16, 103–122.
Berti, M. and Simoni A., 2005, Experimental evidences and numerical Cannon, S.H. and Gartner, J.E., 2005, Wildfire-related debris flow from
modelling of debris flow initiated by channel runoff. Landslides, 2, a hazards perspective. In: Jakob, M. and Hungr, O. (eds.), Debris
171–182. Flow Hazards and Related Phenomena. Springer, Berlin, p. 363–
Berti, M. and Simoni A., 2007, Prediction of debris flow inundation 385.
areas using empirical mobility relationships. Geomorphology, 90, Canuti, P., Casagli, N., Catani, F., and Fanti, R., 2000, Hydrogeological
144–161. hazard and risk in archaeological sites: some case studies in Italy.
Berti, M. and Simoni, A., 2014, DFLOWZ: a free program to evaluate Journal of Cultural Heritage, 1, 117–125.
the area potentially inundated by a debris flow. Computers & Geo- Capparelli, G. and Tiranti, D., 2010, Application of the MoniFLaIR
sciences, 67, 14–23. early warning system for rainfall-induced landslides in Piedmont
Berti, M., Corsini, A., Franceschini, S., and Iannacone, J.P., 2013, Auto- region (Italy). Landslides, 7, 401–410.
mated classification of Persistent Scatterers Interferometry time Capparelli, G. and Versace, P., 2011, FLaIR and SUSHI: two mathemat-
series. Natural Hazards Earth System Sciences, 13, 1945–1958. ical models for early warning of landslides induced by rainfall.

http://dx.doi.org/10.1007/s12303-017-0034-4 http://www.springer.com/journal/12303
Landslide prediction, monitoring, and early warning: a concise review of state-of-the-art 1061

Landslides, 8, 67–79. to slope stability analysis. Journal of Geotechnical Engineering, 120,


Carrara, A., Crosta, G.B., and Frattini, P., 2008, Comparing models of 2180–2207.
debris-flow susceptibility in the alpine environment. Geomorphol- Chung, M.C., Tan, C.H., and Chen, C.H., 2017, Local rainfall thresh-
ogy, 94, 353–378. olds for forecasting landslide occurrence: Taipingshan landslide
Casagli, N., Catani, F., Del Ventisette, C., and Luzi, G., 2010, Monitor- triggered by Typhoon Saola. Landslides, 14, 19–33.
ing, prediction, and early warning using ground-based radar inter- Ciampalini, A., Raspini, F., Lagomarsino, D., Catani, F., and Casagli, N.,
ferometry. Landslides, 7, 291–301. 2016, Landslide susceptibility map refinement using PSInSAR data.
Catani, F., Tofani, V., and Lagomarsino, D., 2016, Spatial patterns of Remote Sensing of Environment, 184, 302–315.
landslide dimension: a tool for magnitude mapping. Geomorphol- Cloutier, C., Agliardi, F., Crosta, G.B., Frattini, P., Froese, C., Jaboyedoff,
ogy, 273, 361–373. M., Locat, J., Michoud, C., and Marui, H., 2014, The First Interna-
Catani, F., Farina, P., Moretti, S., Nico, G., and Strozzi, T., 2005, On the tional Workshop on Warning Criteria for Active Slides: technical
application of SAR interferometry to geomorphological studies: issues, problems and solutions for managing early warning systems.
estimation of landform attributes and mass movements. Geomor- Landslides, 12, 205–212.
phology, 66, 119–131. Coduto, D.P., Yeung, M.R., and Kitch, W.A., 2010, Geotechnical Engi-
Chae, B.G. and Kim, M.I., 2012, Suggestion of a method for landslide neering: Principles and Practices (2nd edition). Pearson, New York,
early warning using the change in the volumetric water content 794 p.
gradient due to rainfall infiltration. Environmental Earth Sciences, Colesanti, C., Ferretti, A., Prati, C., and Rocca, F., 2003, Monitoring
66, 1973–1986. landslides and tectonic motions with the Permanent Scatterers
Chae, B.G., Kim, K.S., Kim, J.K., Song, Y.S., Oh, H.J., Lee, C.O., Jeong, Technique. Engineering Geology, 68, 3–14.
S.W., Cho, Y.C., and Choi, J., 2015, Technology development of Collins, B.D. and Znidarcic, D., 2004, Stability analyses of rainfall
landslide rapid detection based on a real-time monitoring. Report induced landslides. Journal of Geotechnical and Geoenvironmen-
GP2015-024-2015(1), Korea Institute of Geoscience and Mineral tal Engineering, 130, 362–372.
Resources, Daejeon, 331 p. Colomina, I. and Molina, P., 2014, Unmanned aerial systems for photo-
Chelli, A., Mandrone, G., and Truffelli, G., 2006, Field investigations grammetry and remote sensing: A review. ISPRS Journal of Photo-
and monitoring as tools for modelling the Rossena castle landslide grammetry and Remote Sensing, 92, 79–97.
(Northern Appennines, Italy). Landslides, 3, 252–259. Comegna, L., Picarelli, L., and Urciuoli, G., 2007, The mechanics of
Chen, C.Y., Chen, T.C., Yu, F.C., and Lin, S.C., 2005, Analysis of time- mudslides as a cyclic undrained-drained process. Landslides, 4,
varying rainfall infiltration induced landslide. Environmental 217–232.
Geology, 48, 466–479. Corominas, J., 1996, The angle of reach as a mobility index for small
Chen, R.F., Chang, K.J., Angelier, J., Chan, Y.C., Deffontaines, B., Lee, and large landslides. Canadian Geotechnical Journal ,33, 260–271.
C.T., and Lin, M.L., 2006, Topographical changes revealed by high- Corominas, J., 1999, Reconstruction in recent landslide activity in rela-
resolution airborne LiDAR data: the 1999 Tsaoling landslide induced tion to rainfall in the Llobregat River basin, Eastern Pyrenees, Spain.
by the Chi-Chi earthquake. Engineering Geology, 88, 160–172. Geomorphology, 30, 79–93.
Chleborad, A.F., 2003, Preliminary evaluation of a precipitation thresh- Corominas, J., 2000, Landslides and climate. Proceedings of the 8th
old for anticipating the occurrence of landslides in the Seattle, International Symposium on Landslides, Cardiff, Jun. 26–30, 4, p.
Washington Area. US Geological Survey Open-File Report 03-463, 1–33.
USGS, Denver, 39 p. Corominas, J., van Westen, C., Frattini, P., Cascini, L., Malet, J.P.,
Cho, S.E., 2009, Probabilistic assessment of slope stability that consid- Fotopoulou, S., Catani, F., Van Den Eeckhaut, M., Mavrouli, O.,
ers the spatial variability of soil properties. Journal of Geotechnical Agliardi, F., Pitilakis, K., Winter, M.G., Pastor, M., Ferlisi, S., Tofani,
and Geoenvironmental Engineering, 136, 975–984. V., Hervas, J., and Smith, J.T., 2014, Recommendations for the
Cho, S.E. and Lee, S.R., 2002, Evaluation of surficial stability for homo- quantitative analysis of landslide risk. Bullettin of Engineering
geneous slopes considering rainfall characteristics. Journal of Geo- Geology and the Environment, 73, 209–263.
technical and Geoenvironmental Engineering, 128, 756–763. Corsini, A., Pausto, A., Soldati, M., and Zannoni, A., 2005, Field moni-
Chowdhury, R., Flentje, P., and Bhattacharya, G., 2010, Geotechnical toring of the Corvara landslide (Dolomites, Italy) and its relevance
Slope Analysis. CRC, New York, 738 p. for hazard assessment. Geomorphology, 66, 149–165.
Christen, M., Bartelt, P., and Kowalski, J., 2010, Back calculation of the Costantini, M., Falco, S., Malvarosa, F., and Minati, F., 2008, A new
In den Arelen avalanche with RAMMS: interpretation of model method for identification and analysis of Persistent Scatterers in
results. Annals of Glaciology, 51, 161–168. series of SAR images. Proceedings of the International Geoscience
Christen, M., Bühler, Y., Bartelt, P., Leine, R., Glover, J., Schweizer, A., and Remote Sensing Symposium (IGARSS 2008), Boston, Jul. 6–
Graf, C., McArdell, B.W., Gerber, W., Deubelbeiss, Y., Feistl, T., and 11, 2, II-449-II-452.
Volkwein, A., 2012, Integral hazard management using a unified Crosetto, M., Biescas, E., Duro, J., Closa, J., and Arnaud, A., 2008, Gen-
software environment: numerical simulation tool “RAMMS” for eration of advanced ERS and Envisat interferometric SAR prod-
gravitational natural hazards. Proceedings of the 12th Congress ucts using the stable point network technique. Photogrammetric
Interpraevent 2012, Grenoble, Apr. 23–26, 1, p. 77–86. Engineering & Remote Sensing, 74, 443–450.
Christian, J.T., Ladd, C.C., and Baecher, G.B., 1994, Reliability applied Crosetto, M., Monserrat, O., Cuevas-González, M., Devanthéry, N., and

http://www.springer.com/journal/12303 http://dx.doi.org/10.1007/s12303-017-0034-4
1062 Byung-Gon Chae, Hyuck-Jin Park, Filippo Catani, Alessandro Simoni, and Matteo Berti

Crippa, B., 2016, Persistent scatterer interferometry: a review. ISPRS neering Geology, 99, 11–22.
Journal of Photogrammetry and Remote Sensing, 115, 78–89. Dick, G.J., Eberhardt, E., Cabrejo-Liévano, A.G., Stead, D., and Rose,
Crosta, G.B. and Frattini, P., 2001, Rainfall thresholds for triggering soil N.D., 2015, Development of an early-warning time-of-failure anal-
slips and debris flow. Proceedings of 2nd EGS Plinius Conference ysis methodology for open-pit mine slopes utilizing ground-based
on Mediterranean Storms, Siena, Oct. 16–18, p. 463–487. slope stability radar monitoring data. Canadian Geotechnical Jour-
Crosta, G.B. and Frattini, P., 2003, Distributed modeling of shallow nal, 52, 515–529.
landslides triggered by intense rainfall. Natural Hazards and Earth Dietrich, W.E., Bellugi, D., and De Asua, R.R., 2001, Validation of the
System Science, 3, 81–93. shallow landslide model, SHALSTAB, for forest management. In:
Crosta, G.B., Cucchiaro, S., and Frattini, P., 2003, Validation of semiem- Wigmosta, M.S. and Burges, S.J. (eds.), Land Use and Watersheds:
pirical relationships for the definition of debris-flow behaviour in Human Influence on Hydrology and Geomorphology in Urban
granular materials. Proceedings of the 3rd International Confer- and Forest Areas. American Geophysical Union, Washington, p.
ence on Debris-Flow Hazards Mitigation: Mechanics, Prediction 195–227.
and Assessment, Davos, Sep. 10–12, p. 821–831. Dou, H.Q., Han, T.C., Gong, X.N., and Zhang, J., 2014, Probabilistic
Crozier, M.J., 1999, Prediction of rainfall-triggered landslides: a test of slope stability analysis considering the variability of hydraulic con-
the antecedent water status model. Earth Surface Processes and ductivity under rainfall infiltration-redistribution conditions. Engi-
Landforms, 24, 825–833. neering Geology, 183, 1–13.
Cruden, D.M. and Varnes, D.J., 1996, Landslide types and processes. In: El-Ramly, H., Morgenstern, N.R., and Cruden, D.M., 2002, Probabilis-
Turner, A.K. and Schuster, R.L. (eds.), Landslides – Investigation tic slope stability analysis for practice. Canadian Geotechnical Jour-
and Mitigation. Transportation Research Board Special Report 247, nal, 39, 665–683.
National Research Council, USA, p. 36–75. Fannin, R.J. and Wise, M.P., 2001, An empirical-statistical model for
D’Agostino, V., Cesca, M., and Marchi, L., 2010, Field and laboratory debris flow travel distance. Canadian Geotechnical Journal, 38,
investigations of runout distances of debris flows in the Dolomites 982–994.
(Eastern Italian Alps). Geomorphology 115, 294–304. Farina, P., Colombo, D., Fumagalli, A., Marks, F., and Moretti, S., 2006,
D’Amato Avanzi, G., Falaschi, F., Giannecchini, R., and Puccinelli, A., Permanent Scatterers for landslide investigations: outcomes from
2009, Soil slip susceptibility assessment using mechanical-hydro- the ESA-SLAM project. Engineering Geology, 88, 200–217.
logical approach and GIS techniques: an application in the Apuan Fathani, T.F., Karnawati, D., and Wilopo, W., 2016, An integrated meth-
Alps (Italy). Natural Hazards, 50, 591–603. odology to develop a standard for landslide early warning systems.
Dai, F.C., Lee, C.F., and Ngai, Y.Y., 2002, Landslide risk assessment and Natural Hazards and Earth System Sciences, 16, 2123–2135.
management: an overview. Engineering Geology, 64, 65–87. Fell, R., Corominas, J., Bonnard, C., Cascini, L., Leroi, E., and Savage,
Dal Sasso, S.F., Sole, A., Pascale, S., Sdao, F., Bateman Pinzón, A., and W.Z., 2008, Guidelines for landslide susceptibility, hazard and risk
Medina, V., 2014, Assessment methodology for the prediction of zoning for land use planning. Engineering Geology, 102, 85–98.
landslide dam hazard. Natural Hazards and Earth System Sciences, Fernandes, N.F., Guimarães, R.F., Gomes, R.A., Vieira, B.C., Montgom-
14, 557–567. ery, D.R., and Greenberg, H., 2004, Topographic controls of land-
Damiano, E., Olivares, L., and Picarelli, L., 2012, Steep-slope monitor- slides in Rio de Janeiro: field evidence and modeling. Catena, 55,
ing in unsaturated pyroclastic soils. Engineering Geology, 137, 1–12. 163–181.
Davies, T.R. and McSaveney, M.J., 2012, Mobility of long-runout rock Ferretti, A., Fumagalli, A., Novali, F., Prati, C., Rocca, F., and Rucci, A.,
avalanches. In: Clague, J.J. and Stead, D. (eds.), Landslides – Types, 2011, A new algorithm for processing interferometric data-stacks:
Mechanisms and Modeling. Cambridge University Press, Cambridge, SqueeSAR. IEEE Transactions on Geoscience and Remote Sens-
p. 50–58. ing, 49, 3460–3470.
Debella-Gilo, M. and Kääb, A., 2011, Sub-pixel precision image match- Ferretti, A., Prati, C., and Rocca, F., 2000, Nonlinear subsidence rate
ing for measuring surface displacements on mass movements using estimation using permanent scatterers in differential SAR interfer-
normalized cross-correlation. Remote Sensing of Environment, ometry. IEEE Transactions on Geoscience and Remote Sensing, 38,
115, 130–142. 2202–2212.
Delacourt, C., Allemand, P., Casson, B., and Vadon, H., 2004, Velocity Ferretti, A., Prati, C., and Rocca, F., 2001, Permanent scatterers in SAR
field of the “La Clapière” landslide measured by the correlation of interferometry. IEEE Transactions on Geoscience and Remote
aerial and QuickBird satellite images. Geophysical Research Let- Sensing, 39, 8–20.
ters, 31, L15619. Forzieri, G., Moser, G., and Catani, F., 2012, Assessment of hyperspec-
Deline, P., Alberto, W., Broccolato, M., Hungr, O., Noetzli, J., Ravanel, tral MIVIS sensor capability for heterogeneous landscape classifi-
L., and Tamburini, A., 2011, The December 2008 Crammont rock cation. ISPRS Journal of Photogrammetry and Remote Sensing, 74,
avalanche, Mont Blanc massif area, Italy. Natural Hazards and 175–184.
Earth System Sciences, 11, 3307–3318. Forzieri, G., Tanteri, L., Moser, G., and Catani, F., 2013, Mapping natural
Dewitte, O., Jasselette, J.C., Cornet, Y., Van Den Eeckhaut, M., Col- and urban environments using airborne multi-sensor ADS40-
lignon, A., Poesen, J., and Demoulin, A., 2008, Tracking landslide MIVIS-LiDAR synergies. International Journal of Applied Earth
displacements by multi-temporal DTMs: a combined aerial stereo- Observation and Geoinformation, 23, 313–323.
photogrammetric and LIDAR approach in western Belgium. Engi- Frank, F., McArdell, B.W., Huggel, C., and Vieli, A., 2015, The impor-

http://dx.doi.org/10.1007/s12303-017-0034-4 http://www.springer.com/journal/12303
Landslide prediction, monitoring, and early warning: a concise review of state-of-the-art 1063

tance of entrainment and bulking on ebris flow runout modeling: GIS framework. Engineering Geology, 102, 214–226.
examples from the Swiss Alps. Natural Hazards and Earth System Gokceoglu, C., Sonmez, H., and Ercanoglu, M., 2000, Discontinuity
Sciences, 15, 2569–2583. controlled probabilistic slope failure risk maps of the Altindag (set-
Franks, C.A.M., 1999, Characteristics of some rainfall-induced land- tlement) region in Turkey. Engineering Geology, 55, 277–296.
slides on natural slopes, Lantau Island, Hong Kong. Quaterly Jour- Gorsevski, P.V., Gessler, P.E., Boll, J., Elliot, W.J., and Foltz, R.B., 2006,
nal of Engineering Geology, 32, 247–259. Spatially and temporally distributed modeling of landslide suscep-
Frattini, P., Crosta, G.B., Fusi, N., and Dal Negro, P., 2004, Shallow tibility. Geomorphology, 80, 178–198.
landslides in pyroclastic soils: a distributed modelling approach for Griffiths, D.V., Huang, J.S., and Fenton, G.A., 2011, Probabilistic infinite
hazard assessment. Engineering Geology, 73, 277–295. slope analysis. Computers and Geotechnics, 38, 577–584.
Fredlund, D.G., Sheng, D., and Zhao, J., 2011, Estimation of soil suc- Griswold, J.P. and Iverson, R.M., 2008, Mobility statistics and auto-
tion from the soil-water characteristic curve. Canadian Geotechni- mated hazard mapping for debris flows and rock avalanches. U.S.
cal Journal, 48, 186–198. Geological Survey Scientific Investigations Report 5276, 59 p.
Fredlund, D.G., Xing, A., Fredlund, M.D., and Barbour, S.L., 1996, The Guimaraes, R.F., Montgomery, D.R., Greenberg, H.M., Fernandes, N.F.,
relationship of the unsaturated soil shear to the soil-water charac- Gomes, R.A.T., and de Carvalho, O.A., 2003, Parameterization of
teristic curve. Canadian Geotechnical Journal, 33, 440–448. soil properties for a model of topographic controls on shallow
Fruneau, B., Achache, J., and Delacourt, C., 1996, Observation and landsliding: application to Rio de Janeiro. Engineering Geology, 69,
modelling of the Saint-Étienne-de-Tinée landslide using SAR inter- 99–108.
ferometry. Tectonophysics, 265, 181–190. Guzzetti, F., 2000, Landslide fatalities and the evaluation of landslide
Garcia, A., Hordt, A., and Fabian, M., 2010, Landslide monitoring with risk in Italy. Engineering Geology, 58, 2, 89–107.
high resolution tilt measurements at the Dollendorfer Hardt land- Guzzetti, F., Peruccacci, S., Rossi, M., and Stark, C.P., 2007, Rainfall
slide, Germany. Geomorphology, 120, 16–25. thresholds for the initiation of landslides. Meteorology Atmospheric
Gariano, S.L., Brunetti, M.T., Iovine, G., Melillo, M., Peruccacci, S., Ter- Physics, 98, 239–267.
ranova, O.G., Vennari, C., and Guzzetti, F., 2015, Calibration and Guzzetti, F., Peruccacci, S., Rossi, M., and Stark, C.P., 2008, The rainfall
validation of rainfall thresholds for shallow landslide forecasting in intensity-duration control of shallow landslides and debris flows:
Sicily, Southern Italy. Geomorphology, 228, 653–665. an update. Landslides, 5, 3–17.
Ghuffar, S., Székely, B., Roncat, A., and Pfeifer, N., 2013, Landslide dis- Harr, M.E., 1987, Reliability-based Design in Civil Engineering. McGraw-
placement monitoring using 3D range flow on airborne and terres- Hill, New York, 291 p.
trial LiDAR data. Remote Sensing, 5, 2720–2745. He, K., Wang, S., Du, W., and Wang, S., 2010, Dynamic features and effects
Gigli, G. and Casagli, N., 2011, Semi-automatic extraction of rock mass of rainfall on landslides in the Three Gorges Reservoir region, China:
structural data from high resolution LIDAR point clouds. Interna- using the Xintan landslide and the large Huangya landslide as the
tional Journal of Rock Mechanics and Mining Sciences, 48, 187–198. examples. Environmental Earth Sciences, 59, 1267–1274.
Gigli, G., Morelli, S., Fornera, S., and Casagli, N., 2014, Terrestrial laser Herrera, G., Gutiérrez, F., García-Davalillo, J.C., Guerrero, J., Notti, D.,
scanner and geomechanical surveys for the rapid evaluation of Galve, J.P., Fernández-Merodo, J.A., and Cooksley, G., 2013, Multi-
rock fall susceptibility scenarios. Landslides, 11, 1–14. sensor advanced DInSAR monitoring of very slow landslides: the
Glade, T., 2000, Modeling landslide-triggering rainfalls in different Tena Valley case study (Central Spanish Pyrenees). Remote Sens-
regions of New Zealand-the soil water status model. Zeitschrift fur ing of Environment, 128, 31–43.
Geomorphologie NE, 122, 63–84. Ho, J.Y., Lee, K.T., Chang, T.C., Wang, Z.Y., and Liao, Y.H., 2012, Influ-
Glade, T., Anderson, M., and Crozier, M.J., 2005, Landslide Hazard and ences of spatial distribution of soil thickness on shallow landslide
Risk. Wiley, Chichester, 824 p. prediction. Engineering Geology, 124, 38–46.
Glade, T., Crozier, M.J., and Smith, P., 2000, Applying probability deter- Hong, Y. and Adler, R.F., 2007, Satellite remote sensing for global land-
mination to refine landslide-triggering rainfall thresholds using an slide monitoring. Eos Transactions AGU, 88, 357.
empirical “Antecedent Daily Rainfall Model”. Pure and Applied Hong, Y., Adler, R., and Huffman, G., 2006a, Use of satellite remote
Geophysics, 157, 1059–1079. sensing data in the mapping of global landslide susceptibility. Natu-
Glenn, N.F., Streutker, D.R., Chadwick, D.J., Thackray, G.D., and Dorsch, ral Hazards, 43, 245–256.
S.J., 2006, Analysis of LiDAR-derived topographic information for Hong, Y., Adler, R., and Huffman, G., 2006b, Evaluation of the poten-
characterizing and differentiating landslide morphology and activ- tial of NASA multi-satellite precipitation analysis in global land-
ity. Geomorphology, 73, 131–148. slide hazard assessment. Geophysical Research Letters, 33.
Godt, J.W., 2004, Observed and modeled conditions for shallow land- Hong, Y., Adler, R.F., and Huffman, G., 2007, An experimental global
sliding in the Seattle, Washington area. Ph.D. dissertation, Univer- prediction system for rainfall-triggered landslides using satellite remote
sity of Colorado, Boulder, 151 p. sensing and geospatial datasets. IEEE Transactions on Geosciences
Godt, J.W., Baum, R.L., and Lu, N., 2009, Landsliding in partially satu- and Remote Sensing, 45, 1671–1680.
rated materials. Geophysical Research Letters, 36, L02403. Hooper, A., 2008, A multi-temporal InSAR method incorporating both
Godt, J.W., Baum, R.L., Savage, W.Z., Salciarini, D., Schulz, W.H., and persistent scatterer and small baseline approaches. Geophysical
Harp, E.L., 2008, Transient deterministic shallow landslide model- Research Letters, 35.
ing: requirements for susceptibility and hazard assessments in a Hooper, A., Zebker, H., Segall, P., and Kampes, B., 2004, A new method

http://www.springer.com/journal/12303 http://dx.doi.org/10.1007/s12303-017-0034-4
1064 Byung-Gon Chae, Hyuck-Jin Park, Filippo Catani, Alessandro Simoni, and Matteo Berti

for measuring deformation on volcanoes and other natural terrains systems and applications of nowcasting and warning operations.
using InSAR persistent scatterers. Geophysical Research Letters, 31, WMO/TD No. 1559, World Meteorological Organization, Geneva,
L23611. 22 p.
Huang, J.C., Kao, S.J., Hsu, M.L., and Lin, J.C., 2006, Stochastic proce- Jakob, M., 2005, Debris flow hazard analysis. In: Jakob, M. and Hungr,
dure to extract and to integrate landslide susceptibility maps: an O. (eds.), Debris-flow Hazards and Related Phenomena. Springer,
example of mountainous watershed in Taiwan. Natural Hazards Berlin, p. 411–443.
and Earth System Science, 6, 803–815. Jakob, M. and Friele, P., 2010, Frequency and magnitude of debris flows
Huang, J.C., Kao, S.J., Hsu, M.L., and Liu, Y.A., 2007, Influence of spe- on Cheekye River, British Columbia. Geomorphology, 114, 382–395.
cific contributing area algorithms on slope failure prediction in Jakob, M. and Jordan, P., 2001, Design flood estimates in mountain
landslide modeling. Natural Hazards and Earth System Science, 7, streams the need for a geomorphic approach. Canadian Journal of
781–792. Civil Engineering, 28, 425–439.
Huggel, C., Khabarov, N., Obersteiner, M., and Ramirez, J.M., 2010, Jakob, M. and Weatherly, H., 2003, A hydroclimatic threshold for land-
Implementation and integrated numerical modeling of a landslide slide initiation on the north shore mountains of Vancouver, British
early warning system: a pilot study in Colombia. Natural Hazards, Columbia. Geomorphology, 54, 137–156.
52, 201–518. Jakob, M., McDougall, S., Weatherly, H., and Ripley, N., 2013, Debris-
Hungr, O., 1995, A model for the runout analysis of rapid flow slides, flow simulations on Cheekye River, British Columbia. Landslides,
debris flows and avalanches. Canadian Geotechnical Journal, 32, 10, 685–699.
610–623. Jan, C.D. and Lee, M.H., 2004, A debris-flow rainfall-based warning
Hungr, O., Corominas, J., and Eberhardt, E., 2005, Estimating landslide model. Journal of Chinese Soil and Water Conservation, 35, 273–
motion mechanism, travel distance and velocity. Landslide Risk 283.
Management, 99–128. Jiao, J.J., Ding, G., and Leung, C.M., 2006, Confined groundwater near
Hungr, O., Evans, S.G., Bovis, M., and Hutchinson, J.N., 2001, Review the rockhead in igneous rocks in the Mid-Levels area, Hong Kong,
of the classification of landslides of the flow type. Environmental China. Engineering Geology, 84, 207–219.
and Engineering Geoscience, 7, 221–238. Jiao, J.J., Wang, X.S., and Nandy, S., 2005, Confined groundwater zone
Hunter, G. and Fell, R., 2003, Travel distance angle for rapid landslides and slope instability in weathered igneous rocks in Hong Kong.
in constructed and natural soil slopes. Canadian Geotechnical Journal, Engineering Geology, 80, 71–92.
40, 1123–1141. Johnson, B.C., Campbell, C.S., and Melosh, H.J., 2016, The reduction of
Hurlimann, M., McArdell, B.W., and Rickli, C., 2015, Field and labora- friction in long runout landslides as an emergent phenomenon.
tory analysis of the runout characteristics of hillslope debris flows Journal of Geophysical Research: Earth Surface, 121, 881–889.
in Switzerland. Geomorphology, 232, 20–32. Kaab, A., 2000, Photogrammetry for early recognition of high moun-
Hurlimann, M., Rickenmann, D., Medina, V., and Bateman, A., 2008, tain hazards: new techniques and applications. Physics and Chem-
Evaluation of approaches to calculate debris-flow parameters for haz- istry of the Earth, Part B: Hydrology, Oceans and Atmosphere, 25,
ard assessment. Engineering Geology, 102, 152–163. 765–770.
Hussin, H.Y., Quan Luna, B., van Westen, C.J., Christen, M., Malet, J.P., Kamai, T., 1991, Slope stability assessment by using GIS. Research
and van Asch, T., 2012, Parameterization of a numerical 2-D debris Report, Science and Technology Agency of Japan. (in Japanese)
flow model with entrainment: a case study of the Faucon catchment, Karnawati, D., Fathani, T.F., Wilopo, W., and Andayani, B., 2013, Hybrid
Southern French Alps. Natural Hazards and Earth System Sci- socio-technical approach for landslide risk reduction in Indonesia.
ences, 12, 3075–3090. In: Wang, F., Miyajima, M., Li, T., Shan, W., and Fathani, T.F. (eds.),
Innes, J.L., 1983, Debris flows. Progress in Physical Geography, 7, 469– Progress of Geo-disaster Mitigation Technology in Asia. Springer,
501. Berlin, p. 157–169.
Intrieri, E., Gigli, G., Mugnai, F., Fanti, R., and Casagli, N., 2012, Design Kasai, M., Ikeda, M., Asahina, T., and Fujisawa, K., 2009, LiDAR-
and implementation of a landslide early warning system. Engineer- derived DEM evaluation of deep-seated landslides in a steep and
ing Geology, 147, 124–136. rocky region of Japan. Geomorphology, 113, 57–69.
Iverson, R.M., 2000, Landslide triggering by rain infiltration. Water Kean, J.W., McCoy, S.W., Tucker, G.E., Staley, D.M., and Coe, J.A., 2013,
Resources Research, 36, 1897–1910. Runoff-generated debris flows: observations and modeling of surge
Iverson, R.M. and Denlinger, R.P., 2001, Flow of variably fluidized gran- initiation, magnitude, and frequency. Journal of Geophysical Research:
ular masses across three-dimensional terrain, 1. Coulomb mixture Earth Surface, 118, 2190–2207.
theory. Journal of Geophysical Research: Solid Earth, 106, 537–552. Kim, M.S., Onda, Y., Uchida, T., and Kim, J.K., 2016, Effects of soil
Iverson, R.M., Schilling, S.P., and Vallance, J.W., 1998, Objective delin- depth and subsurface flow along the subsurface topography on
eation of lahar-inundation hazard zones. Geological Society of shallow landslide predictions at the site of a small granitic hillslope.
America Bulletin, 110, 972–984. Geomorphology, 271, 40–54.
Jaboyedoff, M., Oppikofer, T., Abellán, A., Derron, M.H., Loye, A., Kirschbaum, D.B., Adler, R.F., Hong, Y., and Lerner-Lam, A., 2009,
Metzger, R., and Pedrazzini, A., 2012, Use of LIDAR in landslide Evaluation of a preliminary satellite-based landslide hazard algorithm
investigations: a review. Natural Hazards, 61, 5–28. using global landslide inventories. Natural Hazards and Earth Sys-
Jacks, E., Davison, J., and Wai, H.G., 2010, Guidelines on early warning tem Sciences, 9, 673–686.

http://dx.doi.org/10.1007/s12303-017-0034-4 http://www.springer.com/journal/12303
Landslide prediction, monitoring, and early warning: a concise review of state-of-the-art 1065

Konak, G., Onur, A.H., Karakus, D., Köse, H., Koca, Y., and Yenice, H., soil-water retention curve parameters. Journal of Geotechnical and
2004, Slope stability analysis and slide monitoring by inclinome- Geoenvironmental Engineering, 140.
ters readings: Part 2. Mining Technology, 113, 171–180. Listo, F.D.L.R. and Vieira, B.C., 2012, Mapping of risk and susceptibility of
Kummerow, C., Simpson, J., Thiele, O., Barnes, W., Chang, A.T.C., shallow-landslide in the city of São Paulo, Brazil. Geomorphology,
Stocker, E., Adler, R.F., Hou, A., Kakar, R., Wentz, F., Ashcroft, P., Kozu, 169, 30–44.
T., Hong, Y., Okamoto, K., Iguchi, T., Kuroiwa, H., Im, E., Haddad, Liu, C.N. and Wu, C.C., 2008, Mapping susceptibility of rainfall-trig-
Z., Huffman, G., Ferrier, B., Olson, W.S., Zipser, E., Smith, E.A., gered shallow landslides using a probabilistic approach. Environ-
Wilheit, T.T., North, G., Krishnamurti, T., and Nakamura, K., 2000, mental Geology, 55, 907–915.
The status of the Tropical Rainfall Measuring Mission (TRMM) Liu, W., He, S., and Li, X., 2016, A finite volume method for two-phase
after two years in orbit. Journal of Applied Meteorology and Cli- debris flow simulation that accounts for the pore-fluid pressure
matology, 39, 1965–1982. evolution. Environmental Earth Sciences, 75, 1–10.
Lagios, E., Sakkas, V., Novali, F., Bellotti, F., Ferretti, A., Vlachou, K., Lu, N. and Godt, J., 2008, Infinite slope stability under steady unsatu-
and Dietrich, V., 2013, SqueeSARTM and GPS ground deformation rated seepage conditions. Water Resources Research, 44, W11404.
monitoring of Santorini Volcano (1992–2012): Tectonic implications. Lu, N. and Godt, J.W., 2013, Hillslope Hydrology and Stability. Cam-
Tectonophysics, 594, 38–59. bridge University Press, Cambridge, 437 p.
Lagomarsino, D., Segoni, S., Fanti, R., and Catani, F., 2013, Updating Lu, N. and Likos, W.J., 2006, Suction stress characteristic curve for
and tuning a regional-scale landslide early warning system. Land- unsaturated soil. Journal of Geotechnical and Geoenvironmental
slides, 10, 91–97. Engineering, 132, 131–142.
Lanari, R., Mora, O., Manunta, M., Mallorqui, J.J., Berardino, P., and Lu, N., Kaya, M., Collins, B.D., and Godt, J.W., 2013, Hysteresis of
Sansosti, E., 2004, A small-baseline approach for investigating unsaturated hydromechanical properties of a silty soil. Journal of
deformations on full-resolution differential SAR interferograms. Geotechnical and Geoenvironmental Engineering, 139, 507–510.
IEEE Transactions on Geoscience and Remote Sensing, 42, 1377– Lu, P. and Rosenbaum, M.S., 2003, Artificial neural networks and grey
1386. systems for the prediction of slope stability. Natural Hazards, 30,
Lee, J.H. and Park, H.J., 2016, Assessment of shallow landslide suscepti- 383–398.
bility using the transient infiltration flow model and GIS-based Lu, P., Casagli, N., Catani, F., and Tofani, V., 2012, Persistent Scatterers
probabilistic approach. Landslides, 13, 885–903. Interferometry Hotspot and Cluster Analysis (PSI-HCA) for detec-
Legros, F., 2002, The mobility of long-runout landslides. Engineering tion of extremely slow-moving landslides. International Journal of
Geology, 63, 301–331. Remote Sensing, 33, 466–489.
Lepore, C., Amone, E., Noto, L.V., Sivandran, G., and Bras, R.L., 2013, Lu, P., Catani, F., Tofani, V., and Casagli, N., 2014, Quantitative hazard
Physically based modeling of rainfall-triggered landslide: a case and risk assessment for slow-moving landslides from Persistent
study in the Luquillo forest, Puerto Rico. Hydrology and Earth sys- Scatterer Interferometry. Landslides, 11, 685–696.
tem Sciences, 17, 3371–2287. Lu, P., Stumpf, A., Kerle, N., and Casagli, N., 2011, Object-oriented
Leprince, S., Berthier, E., Ayoub, F., Delacourt, C., and Avouac, J.P., change detection for landslide rapid mapping. IEEE Geoscience
2008, Monitoring earth surface dynamics with optical imagery. EOS and Remote Sensing Letters, 8, 701–705.
Transactions AGU, 89, 1–2. Lucieer, A., de Jong, S.M., and Turner, D., 2014, Mapping landslide dis-
Leung, A.K. and Ng, C.W.W., 2013, Seasonal movement and ground- placements using Structure from Motion (SfM) and image correla-
water flow mechanism in an unsaturated saprolitic hillslope. Land- tion of multi-temporal UAV photography. Progress in Physical
slides, 10, 455–467. Geography, 38, 97–116.
Li, D.Q., Qi, X.H., Phoon, K.K., Zhang, L.M., and Zhou, C.B., 2014, Luo, X., Wang, F., Zhang, Z., and Che, A., 2009, Establishing a monitor-
Effect of spatially variable shear strength parameters with linearly ing network for an impoundment-induced landslide in Three
increasing mean trend on reliability of infinite slopes. Structural Gorges Reservoir Area, China. Landslides, 6, 27–37.
Safety, 49, 45–55. Luzi, L. and Pergalani, F., 1996, Applications of statistical and GIS tech-
Li, W.C., Dai, F.C., Wei, Y.Q., Wang, M.L., Minm, H., and Lee, L.M., niques to slope instability zonation (1:50,000 Fabriano geological
2016, Implication of subsurface flow on rainfall-induced landslide: map sheet). Soil Dynamics and Earthquake Engineering, 15, 83–94.
a case study. Landslides, 13, 1109–1123. Macciotta, R., Hendry, M., and Martin, C.D., 2016, Developing an early
Li, W.C., Lee, L.M., Cai, H., Li, H.J., Dai, F.C., and Wang, M.L., 2013, warning system for a very slow landslide based on displacement
Combined roles of saturated permeability and rainfall characteris- monitoring. Natural Hazards, 81, 887–907.
tics on surficial failure of homogeneous soil slope. Engineering Malet, J.P., Maquaire, O., and Calais, E., 2002, The use of Global Posi-
Geology, 153, 105–113. tioning System for the continuous monitoring of landslides, Appli-
Liao, Z., Hong, Y., Kirschbaum, D., Adler, R.F., Gourley, J.J., and Woo- cation to the Super-Sauze earthflow (Alpes-deHaute-Provence, France).
ten, R., 2011, Evaluation of TRIGRS (transient rainfall infiltration Geomorphology, 43, 33–54.
and grid-based regional slope-stability analysis)’s predictive skill for Mantovani, F., Soeters, R., and van Westen, C.J., 1996, Remote sensing
hurricane-triggered landslides: a case study in Macon County, North techniques for landslide studies and hazard zonation in Europe.
Carolina. Natural Hazards, 58, 325–339. Geomorphology, 15, 213–225.
Likos, W.J., Lu, N., and Godt, J.W., 2013, Hysteresis and uncertainty in Marchi, L., Arattano, M., and Deganutti, A.M., 2002, Ten years of

http://www.springer.com/journal/12303 http://dx.doi.org/10.1007/s12303-017-0034-4
1066 Byung-Gon Chae, Hyuck-Jin Park, Filippo Catani, Alessandro Simoni, and Matteo Berti

debris-flow monitoring in the Moscardo Torrent (Italian Alps). Catani, F., Casagli, N., and Tiranti, D., 2013, A prototype forecast-
Geomorphology, 46, 1–17. ing chain for rainfall induced shallow landslides. Natural Hazards
Marchi, L., Cavalli, M., and D’Agostino, V., 2010, Hydrogeomorphic & Earth System Sciences, 13, 771–777.
processes and torrent control works on a large alluvial fan in the Metternicht, G., Hurni, L., and Gogu, R., 2005, Remote sensing of land-
eastern Italian Alps. Natural Hazards and Earth System Sciences, slides: An analysis of the potential contribution to geo-spatial sys-
10, 547–558. tems for hazard assessment in mountainous environments. Remote
Margottini, C., Canuti, P., and Sassa, K., 2013, Landslide Science and Sensing of Environment, 98, 284–303.
Practice, Vol. 2: Early Warning, Instrumentation and Monitoring. Michel, G.P., Kobiyama, M., and Goerl, R.F., 2014, Comparative analy-
Springer, Berlin, 685 p. sis of SHALSTAB and SINMAP for landslide susceptibility map-
Martelloni, G., Segoni, S., Fanti, R., and Catani, F., 2012, Rainfall ping in the Cunha River basin, southem Brazil. Journal of soils and
thresholds for the forecasting of landslide occurrence at regional sediments, 14, 1266–1277.
scale. Landslides, 9, 485–495. Michoud, C., Carrea, D., Costa, S., Derron, M.H., Jaboyedoff, M., Dela-
Massonnet, D. and Feigl, K.L., 1998, Radar interferometry and its appli- court, C., Maquaire, O., Letortu, P., and Davidson, R., 2015, Land-
cation to changes in the Earth’s surface. Review of Geophysics, 36, slide detection and monitoring capability of boat-based mobile
441–500. laser scanning along Dieppe coastal cliffs, Normandy. Landslides,
Massonnet, D., Briole, P., and Arnaud, A., 1995, Deflation of Mount 12, 403–418.
Etna monitored by spaceborne radar interferometry. Nature, 375, Miller D.J. and Burnett K.M., 2008, A probabilistic model of debris-
567–570. flow delivery to stream channels, demonstrated for the Coast Range
Massonnet, D., Feigl, K., Rossi, M., and Adragna, F., 1994, Radar inter- of Oregon, USA. Geomorphology, 94, 184–205.
ferometric mapping of deformation in the year after the Landers Monserrat, O., Crosetto, M., and Luzi, G., 2014, A review of ground-
earthquake. Nature, 369, 227–230. based SAR interferometry for deformation measurement. ISPRS
Massonnet, D., Rossi, M., Carmona, C., Adragna, F., Peltzer, G., Feigl, Journal of Photogrammetry and Remote Sensing, 93, 40–48.
K., and Rabaute, T., 1993, The displacement field of the Landers Montgomery, D.R. and Dietrich, W.E., 1994, A physically based model
earthquake mapped by radar interferometry. Nature, 364, 138–142. for the topographic control on shallow landsliding. Water Resources
Matsushi, Y., Hattanji, T., and Matsukura, Y., 2006, Mechanisms of shal- Research, 30, 1153–1171.
low landslides on soil-mantled hillslopes with permeable and Montgomery, D.R., Dietrich, W.E., and Heffner, J.T., 2002, Piezometric
impermeable bedrocks in the Boso Peninsula, Japan. Geomorphol- response in shallow bedrock at CB1: implications for runoff gener-
ogy, 76, 92–108. ation and landsliding. Water Resources Research, 38, 12.
McCoy, S.W., Kean, J.W., Coe, J.A., Staley, M., Wasklewicz, T.A., and Montrasio, L., 2000, Stability analysis of soil-slip. In: Brebbia, C.A. (ed.),
Tucker, G.E., 2010, Evolution of a natural debris flow: in situ mea- Risk Analysis II. WIT Transactions on Ecology and the Environ-
surements of flow dynamics, video imagery, and terrestrial laser ment, 45, p. 357–366.
scanning. Geology, 38, 735–738. Montrasio, L. and Valentino, R., 2008, A model for triggering mecha-
McDougall, S., 2016, Landslide runout analysis – current practice and nisms of shallow landslides. Natural Hazards and Earth System Sci-
challenges. Canadian Geotechnical Journal, 54, 605–620. ences, 8, 1149–1159.
McDougall, S. and Hungr, O., 2004, A model for the analysis of rapid Montrasio, L., Valentino, R., and Losi, G.L., 2011, Towards a real-time
landslide motion across three-dimensional terrain. Canadian Geo- susceptibility assessment of rainfall-induced shallow landslides on
technical Journal, 41, 1084–1097. a regional scale. Natural Hazards and Earth System Sciences, 11,
McDougall, S. and Hungr, O., 2005, Dynamic modelling of entrainment in 1927.
rapid landslides, Canadian Geotechnical Journal, 42, 1437–1448. Naef, D., Rickenmann, D., Rutschmann, P., and McArdell, B.W., 2006,
Meisina, C. and Scarabelli, S., 2007, A comparative analysis of terrain Comparison of flow resistance relations for debris flows using a
stability models for predicting shallow landslides in colluvial soils. one-dimensional finite element simulation model. Natural Haz-
Geomorphology, 87, 207–223. ards and Earth System Science, 6, 155–165.
Melchiorre, C. and Frattini, P., 2012, Modelling probability of rainfall- Navratil, O., Liebault, F., Bellot, H., Travaglini, E., Theule, J., Chambon,
induced shallow landslides in a changing climate, Otta, Central G., and Laigle, D., 2013, High-frequency monitoring of debris-flow
Norway. Climatic Change, 113, 413–436. propagation along the Réal Torrent, Southern French Prealps. Geo-
Melillo, M., Brunetti, M.T., Peruccacci, S., Gariano, S.L., and Guzzetti, morphology, 201, 157–171.
F., 2015, An algorithm for the objective reconstruction of rainfall Ng, C.W.W. and Shi, Q., 1998, Influence of rainfall intensity and dura-
events responsible for landslides. Landslides, 12, 311–320. tion on slope stability in unsaturated soils. Quarterly Journal of
Melillo, M., Brunetti, M.T., Peruccacci, S., Gariano, S.L., and Guzzetti, Engineering Geology, 31, 105–113.
F., 2016, Rainfall thresholds for the possible landslide occurrence in Niethammer, U., James, M.R., Rothmund, S., Travelletti, J., and Joswig,
Sicily (Southern Italy) based on the automatic reconstruction of M., 2012, UAV-based remote sensing of the Super-Sauze landslide:
rainfall events. Landslides, 13, 165–172. evaluation and results. Engineering Geology, 128, 2–11.
Melosh, H.J., 1987, The mechanics of large rock avalanches. Reviews in O’Brien, J.S., Julien, P.Y., and Fullerton, W.T., 1993, Two dimensional
Engineering Geology, 7, 41–50. water flood and mudflow simulation. Journal of Hydraulic Engi-
Mercogliano, P., Segoni, S., Rossi, G., Sikorsky, B., Tofani, V., Schiano, P., neering, 119, 244–261.

http://dx.doi.org/10.1007/s12303-017-0034-4 http://www.springer.com/journal/12303
Landslide prediction, monitoring, and early warning: a concise review of state-of-the-art 1067

O’Connor, K.H. and Dowding, C.H., 1999, Geomeasurements by Puls- Sciences, 10, 1009.
ing TDR Cables and Probes. CRC Press, Boca Raton, 424 p. Ponziani, F., Pandolfo, C., Stelluti, M., Berni, N., Brocca, L., and Mora-
O’Loughlin, E.M., 1986, Prediction of surface saturation zones in natu- marco, T., 2012, Assessment of rainfall thresholds and soil mois-
ral catchments by topographic analysis. Water Resources Research, ture modeling for operational hydrogeological risk prevention in
22, 794–804. the Umbria region (central Italy). Landslides, 9, 229–237.
Osmanoğlu, B., Sunar, F., Wdowinski, S., and Cabral-Cano, E., 2016, Pradhan, A.M.S. and Kim, Y.T., 2015, Application and comparison of
Time series analysis of InSAR data: methods and trends. ISPRS shallow landslide susceptibility models in weathered granite soil
Journal of Photogrammetry and Remote Sensing, 115, 90–102. under extreme rainfall events. Environmental Earth Sciences, 73,
Ouchi, K., 2013, Recent trend and advance of synthetic aperture radar 5761–5771.
with selected topics. Remote Sensing, 5, 716–807. Prochaska, A.B., Santi, P.M., Higgins, J., and Cannon, S.H., 2008,
Pack, R.T., Tarboton, D.G., and Goodwin, C.N., 1998, The SINMAP Debris-flow runout predictions based on the average channel slope
approach to terrain stability mapping. Proceedings of the 8th Congress (ACS). Engineering Geology, 98, 29–40.
of the International Association of Engineering Geology, Vancouver, Rahardjo, H., Santoso, V.A., Leong, E.C., Ng, Y.S., and Hua, C.J., 2011,
Sep. 21–25, p. 1157–1165. Numerical analyses and monitoring performance of residual soil
Pack, T.T., Tarboton, D.G., and Goodwin, C.N., 2001, Assessing terrain slopes. Soils and Foundations, 51, 471–482.
stability in a GIS using SINMAP. Proceedings of the 15th Annual Rahardjo, H., Ong, T.H., Rezaur, R.B., and Leong, E.C., 2007, Factors
GIS Conference (GIS 2001), Vancouver, Feb. 19–22. controlling instability of homogeneous soil slopes under rainfall.
Palis, E., Lebourg, T., Tric, E., Malet, J.P., and Vidal, M., 2016, Long-term Journal of Geotechnical and Geoenvironmental Engineering, 133,
monitoring of a large deep-seated landslide (La Clapiere, South-East 1532–1543.
French Alps): initial study. Landslides, 14, 155–170. Raia, S., Alvioli, M., Rossi, M., Baum, R.L., Godt, J.W., and Guzzetti, F.,
Paloscia, S., Pettinato, S., Santi, E., Notarnicola, C., Pasolli, L., and Rep- 2014, Improving predictive power of physically based rainfall-induced
pucci, A., 2013, Soil moisture mapping using Sentinel-1 images: shallow landslide models: a probabilistic approach. Geoscientific
algorithm and preliminary validation. Remote Sensing of Environ- Model Development, 7, 495–514.
ment, 134, 234–248. Reichenbach, P., Cardinali, M., De Vita, P., and Guzzetti, F., 1998,
Park, H.J. and West, T.R., 2001, Development of a probabilistic approach Regional hydrological thresholds for landslides and floods in the
for rock wedge failure. Engineering Geology, 59, 233–251. Tiber River Basin (Central Italy). Environmental Geology, 35, 146–159.
Park, H.J., Lee, J.H., and Woo, I., 2013, Assessment of rainfall-induced Reid, M.E., Coe, J.A., and Brien, D.L., 2016, Forecasting inundation
shallow landslide susceptibility using a GIS-based probabilistic approach. from debris flows that grow volumetrically during travel, with applica-
Engineering Geology, 161, 1–15. tion to the Oregon Coast Range, USA. Geomorphology, 273, 396–
Park, H.J., West, T.R., and Woo, I., 2005, Probabilistic analysis of rock 411.
slope stability and random properties of discontinuity parameters, Revellino, P., Hungr, O., Guadagno, F.M., and Evans, S.G., 2004, Velocity
Interstate Highway 40, Western North Carolina, USA. Engineering and runout simulation of destructive debris flows and debris avalanches
Geology, 79, 230–250. in pyroclastic deposits, Campania region, Italy. Environmental Geol-
Perissin, D. and Wang, T., 2012, Repeat-pass SAR Interferometry with ogy, 45, 295–311.
partially coherent targets. IEEE Transactions on Geoscience and Rickenmann, D., 1999, Empirical relationships for debris flows. Natu-
Remote Sensing, 50, 271–280. ral Hazards, 19, 47–77.
Peruccacci, S., Brunetti, M.T., Luciani, S., Vennari, C., and Guzzetti, F., Rickenmann, D., 2005, Runout prediction methods. In: Jakob, M. and
2012, Lithological and seasonal control of rainfall thresholds for the Hungr, O. (eds.), Debris-flow Hazards and Related Phenomena.
possible initiation of landslides in central Italy. Geomorphology, Springer, Berlin, p. 305–324.
139, 79–90. Rickenmann, D., 2016, Debris-flow hazard assessment and methods
Petley, D., 2012, Global patterns of loss of life from landslides. Geology, applied in engineering practice. International Journal of Erosion
40, 927–930. Control Engineering, 9, 80–90.
Petrie, G. and Toth, C.K., 2009, Terrestrial laser scanners. In: Shan, J. Rickenmann, D., Laigle, D.M.B.W, McArdell, B.W., and Hubl, J., 2006,
and Toth, C.K. (eds.), Topographic Laser Ranging and Scanning Comparison of 2D debris-flow simulation models with field events.
Principles and Processing. CRC Press, Boca Raton, p. 87–128. Computational Geosciences, 10, 241–264.
Piciullo, L., Gariano, S.L., Melillo, M., Brunetti, M.T., Peruccacci, S., Riley, K.L., Bendick, R., Hyde, K.D., and Gabet, E.J., 2013, Frequency-
Guzzetti, F., and Calvello, M., 2016, Definition and performance of magnitude distribution of debris flows compiled from global data,
a threshold-based regional early warning model for rainfall-induced and comparison with post-fire debris flows in the western US.
landslides. Landslides, 1–14. Geomorphology, 191, 118–128.
Pieraccini, M., Casagli, N., Luzi, G., Tarchi, D., Mecatti, D., Noferini, L., Robinson, D.A., Jones, S.B., Wraith, J.M., Or, D., and Friedman, S.P.,
and Atzeni, C., 2003, Landslide monitoring by ground-based radar 2003, A review of advances in dielectric and electrical conductivity
interferometry: a field test in Valdarno (Italy). International Jour- measurement in soils using time domain reflectometry. Vadose
nal of Remote Sensing, 24, 1385–1391. Zone Journal, 2, 444–475.
Pirulli, M., 2010, On the use of the calibration-based approach for Rossi, G., Catani, F., Leoni, L., Segoni, S., and Tofani, V., 2013, HIR-
debris-flow forward-analyses. Natural Hazards and Earth System ESSS: a physically based slope stability simulator for HPC applica-

http://www.springer.com/journal/12303 http://dx.doi.org/10.1007/s12303-017-0034-4
1068 Byung-Gon Chae, Hyuck-Jin Park, Filippo Catani, Alessandro Simoni, and Matteo Berti

tions. Natural Hazards and Earth System Sciences, 13, 151. ery and historical records, Seattle, Washington. Engineering Geol-
Rossi, M., Peruccacci, S., Brunetti, M.T., Marchesini, I., Luciani, S., ogy, 89, 67–87.
Ardizzone, F., Balducci, V., Bianchi, C., Cardinali, M., Fiorucci, F., Segoni, S., Rosi, A., Rossi, G., Catani, F., and Casagli, N., 2014, Analys-
Mondini, A.C., Reichenbach, P., Salvati, P., Santangelo, M., Barto- ing the relationship between rainfalls and landslides to define a
lini, D., Gariano, S.L., Palladino, M., Vessia, G., Viero, A., Antronico, L., mosaic of triggering thresholds for regional-scale warning systems.
Borselli, L., Deganutti, A.M., Iovine, G., Luino, F., Parise, M., Pole- Natural Hazards Earth System Science, 14, 2637.
mio, M., Guzzetti, F., Luciani, S., Fiorucci, F., Mondini, A.C., San- Shou, K.J. and Chen, Y.L., 2005, Spatial risk analysis of Li-shan land-
tangelo, M., and Tonelli, G., 2012, SANF: National warning system slide in Taiwan. Engineering Geology, 80, 199–213.
for rainfall-induced landslides in Italy. Landslide and Engineered Shou, K.J., Chen, Y.L., and Liu, H., 2009, Hazard analysis of Li-shan
Slopes: Protecting Society through Improved Understanding, 2, landslide in Taiwan. Geomorphology, 103, 143–153.
1895–1899. Simoni, A., Mammoliti, M., and Berti, M., 2011, Uncertainty of debris
Rosso, R., Rulli, M.C., and Vannucchi, G., 2006, A physically based flow mobility relationships and its influence on the prediction of
model for the hydrologic control on shallow landsliding. Water inundated areas. Geomorphology, 132, 249–259.
Resources Research, 42. Simoni, A., Berti, M., Generali, M., Elmi, C., and Ghirotti, M., 2004,
Rott, H. and Nagler, T., 2006, The contribution of radar interferometry Preliminary result from pore pressure monitoring on an unstable
to the assessment of landslide hazards. Advances in Space Research, clay slope. Engineering Geology, 73, 117–128.
37, 710–719. Singhroy, V. and Molch, K., 2004, Characterizing and monitoring rock-
Salciarini, D., Godt, J.W., Savage, W.Z., Conversini, P., Baum, R.L., and slides from SAR techniques. Advances in Space Research, 33, 290–
Michael, J.A., 2006, Modeling regional initiation of rainfall-induced 295.
shallow landslides in the eastern Umbria Region of central Italy. Sirangelo, B. and Braca, G., 2004, Identification of hazard conditions for
Landslides, 3, 181–194. mudflow occurrence by hydrological model: application of FLaIR
Santi, E., Paloscia, S., Pettinato, S., Notarnicola, C., Greifeneder, F., Hahn, S., model to Sarno warning system. Engineering Geology, 73, 267–276.
Wagner, W., Vreugdenhil, M., and Reimer, C., 2015, Developing an Sirangelo, B. and Versace, P., 1996, A real time forecasting model for
operational algorithm based on ANN for the retrieval of SMC from landslides triggered by rainfall. Meccanica, 31, 73–85.
the incoming metop SCA mission. 2015 IEEE International Geo- Song, Y.S., Chae, B.G., and Lee, J., 2016, A method for evaluating the
science and Remote Sensing Symposium (IGARSS 2015), Milan, stability of an unsaturated slope in natural terrain during rainfall.
Jul. 26–31, p. 2024–2027. Engineering Geology, 210, 84–92.
Santoso, A.M., Phoon, K.K., and Quek, S.T., 2011, Effects of soil spatial Sorbino, G., Sica, C., and Cascini, L., 2010, Susceptibility analysis of
variability on rainfall-induced landslides. Computers & Structures, shallow landslides source areas using physically based models. Nat-
89, 893–900. ural Hazards, 53, 313–332.
Sassa, K., 1985, The mechanism of debris flows. Proceedings of the 11th Sorbino, G., Sica, C., Cascini, L., and Cuomo, S., 2007, On the forecasting of
International Conference on Soil Mechanics and Foundation Engi- flowslide triggering areas using physically based models. Proceed-
neering, San Francisco, Aug. 12–16, 3, p. 1173–1176. ings of the 1st North American Landslides Conference on Land-
Sauchyn, D.J. and Trench, N.R., 1978, LANDSAT applied to landslide slides, Vail, Jun. 3–8, 23, p. 305–315.
mapping. Photogrammetric Engineering and Remote Sensing, 44, Sosio, R., Crosta, G.B., and Hungr, O., 2008, Complete dynamic model-
735–741. ing calibration for the Thurwieser rock avalanche (Italian Central
Savage, W.Z., Godt, J.W., and Baum, R.L., 2003, A model for spatially Alps). Engneering Geology, 100, 11–26.
and temporally distributed shallow landslide initiation by rainfall Sosio, R., Crosta, G.B., and Frattini P., 2007, Field observations, rheo-
infiltration. Proceedings of the 3rd International Conference on logical testing and numerical modeling of a debris-flow event. Earth
Debris-flow Hazards Mitigation: Mechanics, Prediction, and Assess- Surface Processes and Landforms, 32, 290–306.
ment, Davos, Sep. 10–12, 1, p. 179–187. Spies, H., Jähne, B., and Barron, J.L., 2002, Range flow estimation.
Savage, W.Z., Godt, J.W., and Baum, R.L., 2004, Modeling time depen- Computer Vision and Image Understanding, 85, 209–231.
dent areal slope stability. Proceedings of the 9th International Sym- Springman, S.M., Thielen, A., Kienzler, P., and Friedel, S., 2013, A long-
posium on Landslides, Rio de Janeiro, Jun. 28-Jul., 2, p. 23–36. term field study for the investigation of rainfall-induced landslides.
Scheidl, C. and Rickenmann, D., 2010, Empirical prediction of debris- Geotechnique, 63, 1177.
flow mobility and deposition on fans. Earth Surface Processes and Stevens, W.R. and Zehrbach, B.E., 2000, Inclinometer data analysis for
Landforms, 35, 157–173. remediated landslides. In: Marr, W.A. (ed.), Geotechnical Measure-
Schilling, S.P., 1998, LAHARZ; GIS programs for automated mapping ments: Lab and Field. American Society of Civil Engineers, Reston, p.
of lahar-inundation hazard zones. U.S. Geological Survey Open- 126–137.
File Report 98-638, 80 p. Stoffel, M., 2010, Magnitude-frequency relationships of debris flows – a
Schlögel, R., Doubre, C., Malet, J.-P., and Masson, F., 2015, Landslide case study based on field surveys and tree-ring records. Geomor-
deformation monitoring with ALOS/PALSAR imagery: a D-InSAR phology, 116, 67–76.
geomorphological interpretation method. Geomorphology, 231, Strozzi, T., Farina, P., Corsini, A., Ambrosi, C., Thüring, M., Zilger, J.,
314–330. Wiesmann, A., Wegmüller, U., and Werner, C., 2005, Survey and
Schulz, W.H., 2007, Landslide susceptibility revealed by LIDAR imag- monitoring of landslide displacements by means of L-band satellite

http://dx.doi.org/10.1007/s12303-017-0034-4 http://www.springer.com/journal/12303
Landslide prediction, monitoring, and early warning: a concise review of state-of-the-art 1069

SAR interferometry. Landslides, 2, 193–201. ral ground-based optical images for landslide monitoring: Applica-
Stumpf, A. and Kerle, N., 2011, Object-oriented mapping of landslides tion, potential and limitations. ISPRS Journal of Photogrammetry
using Random Forests. Remote Sensing of Environment, 115, and Remote Sensing, 70, 39–55.
2564–2577. Tsai, T.L. and Chen, H.F., 2010, Effects of degree of saturation on shal-
Stumpf, A., Malet, J.P., Allemand, P., Pierrot-Deseilligny, M., and low landslides triggered by rainfall. Environmental Earth Sciences,
Skupinski, G., 2015, Ground-based multi-view photogrammetry 59, 1285–1295.
for the monitoring of landslide deformation and erosion. Geomor- Tsai, T.L., Tsai, P.Y., and Yang, P.J., 2015, Probabilistic modelling of rain-
phology, 231, 130–145. fall-induced shallow landslide using a point estimate method. Envi-
Tarchi, D., Casagli, N., Fanti, R., Leva, D.D., Luzi, G., Pasuto, A., Pierac- ronmental Earth Sciences, 73, 4109–4117.
cini, M., and Silvano, S., 2003, Landslide monitoring by using Tsaparas, I., Rahardjo, H., Toll, D.G., and Leong, E.C., 2002, Con-
ground-based SAR interferometry: an example of application to trolling parameters for rainfall-induced landslides. Computers and
the Tessina landslide in Italy. Engineering Geology, 68, 15–30. Geotechnics, 29, 1–27.
Teixeira, M., Bateira, C., Marques, F., and Vieira, B., 2015, Physically Uhlemann, S., Smith, A., Chambers, J., Dixon, N., Dijkstra, T., Haslam,
based shallow translational landslide susceptibility analysis in Tibo E., Meldrum, P., Merritt, A., Gunn, D., and Mackay, J., 2016, Assessment
catchment, NW of Portugal. Landslides, 12, 455–468. of ground-based monitoring techniques applied to landslide inves-
Temesgen, B., Mohammed, M.U., and Korme, T., 2001, Natural hazard tigations. Geomorphology, 253, 438–451.
assessment using GIS and remote sensing methods, with particu- United Nations Office for Disaster Risk Reduction (UNISDR), 2006,
lar reference to the landslides in the Wondogenet area, Ethiopia. Developing early warning system: a checklist. Proceedings of the
Physics and Chemistry of the Earth, Part C: Solar, Terrestrial & 3rd International Conference on Early Warning: From Concept to
Planetary Science, 26, 665–675. Action (EWC III), Bonn, Mar. 27–29, 10 p.
Ter-Stepanian, G., 2000, Quick clay landslides: their enigmatic features van Asch, T.W., Malet, J.P., van Beek, L.P., and Amitrano, D., 2007,
and mechanism. Bulletin of Engineering Geology Environment, Techniques, issues and advances in numerical modelling of land-
59, 47–57. slide hazard. Bulletin de la Societe Géologique de France, 178, 65–88.
Terhorst, B. and Kreja, R., 2009, Slope stability modelling with SINMAP Van Dam, R.L., 2012, Landform characterization using geophysics – recent
in a settlement area of the Swabian Alb. Landslides, 6, 309–319. advances, applications, and emerging tools. Geomorphology, 137,
Terlien, M.T., 1996, Modelling spatial and temporal variations in rain- 57–73.
fall-triggered landslides: the integration of hydrologic models, slope Van Genuchten, M.T., 1980, A closed-form equation for predicting the
stability models and geographic information systems for the hazard hydraulic conductivity of unsaturated soils. Soil Science Society of
zonation of rainfall-triggered landslides with examples from Man- America Journal, 44, 892–898.
izales (Colombia). International Institute for Aerospace and Earth van Westen, C.J., 2000, The modelling of landslide hazards using GIS.
Sciences (ITC), Enschede, Netherlands, Publication No. 32. Surveys in Geophysics, 21, 241–255.
Terlien, M.T., 1998, The determination of statistical and deterministic van Westen, C.J., 2004, Geo-information tools for landslide risk assess-
hydrological landslide-triggering thresholds. Environmental Geol- ment – an overview of recent developments. Proceedings of the Pro-
ogy, 35, 124–130. ceedings of the 9th International Symposium on Landslides, Rio de
Terlien, M.T., van Westen, C.J., and van Asch, T.W., 1995, Determinis- Janeiro, Jun. 28-Jul., 2, p. 39–56.
tic modelling in GIS-based landslide hazard assessment. In: Carrara, van Westen, C.J. and Terlien, M.T.J., 1996, An approach towards deter-
A. and Guzzetti, F. (eds.), Geographical Information Systems in ministic landslide hazard analysis in GIS. A case study from Manizales
Assessing Natural Hazards. Springer Netherlands, p. 57–77. (Colombia). Earth Surface Processes and Landforms, 21, 853–868.
Thiebes, B., Bell, R., Glade, T., Jager, S., Mayer, J., Anderson, M., and van Westen, C.J., Asch, T.W.J., and Soeters, R., 2006, Landslide hazard
Holcombe, L., 2014, Integration of a limit equilibrium model into a and risk zonation – why is it still so difficult? Bulletin of Engineer-
landslide early warning system. Landslides, 11, 859–875. ing Geology and the Environment, 65, 67–184.
Tofani, V., Raspini, F., Catani, F., and Casagli, N., 2013, Persistent Scat- Van Westen, C.J., Castellanos, E., and Kuriakose, S.L., 2008, Spatial data
terer Interferometry (PSI) technique for landslide characterization for landslide susceptibility, hazard, and vulnerability assessment:
and monitoring. Remote Sensing, 5, 1045–1065. and overview. Engineering Geology, 102, 112-131.
Tofani, V., Segoni, S., Agostini, A., Catani, F., and Casagli, N., 2013°, van Westen, C.J., Seijmonsbergen, A.C., and Mantovani, F., 1999, Com-
Technical Note: Use of remote sensing for landslide studies in paring landslide hazard maps. Natural Hazards, 20, 137–158.
Europe. Natural Hazards and Earth System Sciences, 13, 299–309. Vaziri, A., Moore, L., and Ali, H., 2010, Monitoring systems for warn-
Topp, G.C., Annan, J.L., and Davis, A.P., 1980, Electromagnetic deter- ing impending failures in slopes and open pit mines. Natural Haz-
mination of soil water content: measurements in coaxial transmis- ards, 55, 501–512.
sion lines. Water Resources Research, 16, 574–582. Vieira, B.C. and Fernandes, N.F., 2010, Shallow landslide prediction in
Toyos, G.P., Cole, P.D., Felpeto, A., and Marti, J., 2007, A GIS-based the Serra do Mar, São Paulo, Brazil. Natural Hazards and Earth Sys-
methodology for hazard mapping of small volume pyroclastic den- tem Sciences, 10, 1829.
sity currents. Natural Hazards, 41, 99–112. Werner, C., Wegmuller, U., Strozzi, T., and Wiesmann, A., 2003, Inter-
Travelletti, J., Delacourt, C., Allemand, P., Malet, J.P., Schmittbuhl, J., ferometric point target analysis for deformation mapping. 2003
Toussaint, R., and Bastard, M., 2012, Correlation of multi-tempo- IEEE, International Geoscience and Remote Sensing Symposium

http://www.springer.com/journal/12303 http://dx.doi.org/10.1007/s12303-017-0034-4
1070 Byung-Gon Chae, Hyuck-Jin Park, Filippo Catani, Alessandro Simoni, and Matteo Berti

(IGARSS 2003), Toulouse, Jul. 21–25, 7, p. 4362–4364. ping of landslide hazard using a three-dimensional deterministic
Westoby, M.J., Brasington, J., Glasser, N.F., Hambrey, M.J., and Reyn- model. Natural Hazards, 33, 265–282.
olds, J.M., 2012, ‘Structure-from-Motion’ photogrammetry: a low-cost, Yilmaz, I. and Keskin, I., 2009, GIS based statistical and physical
effective tool for geoscience applications. Geomorphology, 179, approaches to landslide susceptibility mapping (Sebinkarahisar, Tur-
300–314. key). Bulletin of Engineering Geology and the Environment, 68,
Wheeler, S.J., Sharma, R.J., and Buisson, M.S.R., 2003, Coupling of 459–471.
hydraulic hysteresis and stress-strain behavior in unsaturated soils. Yin, Y., Wang, H., Gao, Y., and Li, X., 2010, Real-time monitoring and
Geotechnique, 53, 41–54. early warning of landslides at relocated Wushan Town, the Three
White, I.D., Mottershead, D.N., and Harrison, J.J., 1996, Environmen- Gorges Reservoir, China. Landslides, 7, 339–349.
tal Systems: An Introductory Text (2nd edition). Chapman & Hall, Zhang, L.L., Zhang, L.M., and Tang, W.H., 2005, Rainfall-induced slope
London, 616 p. failure considering variability of soil properties. In: Hicks, M.A.
Wieczorek, G.F., 1996, Landslide triggering mechanism. In: Turner, (ed.), Risk and Variability in Geotechnical Engineering. Thomas
A.K. and Schuster, R.L. (eds.), Landslides Investigation and Mitiga- Telford Ltd., London, p. 183–188.
tion, Special report. Transportation Research Board, National Acad- Zhou, G., Esaki, T., Mitani, Y., Xie, M., and Mori, J., 2003, Spatial prob-
emy Press, Washington, 247, p. 76–89. abilistic modeling of slope failure using an integrated GIS Monte
Wieczorek, G.F. and Glade, T., 2005, Climatic factors influencing occur- Carlo simulation approach. Engineering Geology, 68, 373–386.
rence of debris flows. In: Jakob, M. and Hungr, O. (eds.), Debris Zhu, H., Zhang, L.M., Zhang, L.L., and Zhou, C.B., 2013, Two-dimen-
Flow Hazards and Related Phenomena. Springer, Berlin, p. 325–362. sional probabilistic infiltration analysis with a spatially varying per-
Wilson, R.C. and Jayko, A.S., 1997, Preliminary maps showing rainfall meability function. Computers and Geotechnics, 48, 249–259.
thresholds for debris-flow activity, San Francisco Bay region, Cali- Zimmermann, M., Mani, P., and Romang, H., 1997, Magnitude-fre-
fornia. US Geological Survey Open-File Report 97-745F, USGS, quency aspects of Alpine debris flows. Eclogae Geologicae Helve-
Denver, 20 p. tiae, 90, 415–420.
Wu, W. and Sidle, R.C., 1995, A distributed slope stability model for Zizioli, D., Meisina, C., Valentino, R., and Montrasio, L., 2013, Com-
steep forested basins. Water Resources Research, 31, 2097–2110. parison between different approaches to modeling shallow land-
Xie, M., Esaki, T., and Cai, M., 2004, A time-space based approach for slide susceptibility: a case history in Oltrepo Pavese, Northern Italy.
mapping rainfall induced shallow landslide hazard. Environmental Natural Hazards and Earth System Sciences, 13, 559.
Geology, 46, 840–850. Zschau, J. and Kuppers, A.N., 2003, Early Warning Systems for Natural
Xie, M., Esaki, T., and Zhou, G., 2004, GIS-based probabilistic map- Disaster Reduction. Springer, Berlin, 834 p.

http://dx.doi.org/10.1007/s12303-017-0034-4 http://www.springer.com/journal/12303

You might also like