Earth-Science Reviews 179 (2018) 228–247

Contents lists available at ScienceDirect

Earth-Science Reviews
journal homepage: www.elsevier.com/locate/earscirev

Territorial early warning systems for rainfall-induced landslides T

a,⁎ a b
Luca Piciullo , Michele Calvello , José Mauricio Cepeda
a
University of Salerno, Via Giovanni Paolo II 132, 84084 Fisciano, Italy
b
Norwegian Geotechnical Institute, Postboks 3930, Ulleval Stadion, 0806 Oslo, Norway

A R T I C L E I N F O A B S T R A C T

Keywords: Among the many mitigation measures available for reducing the risk to life related to landslides, early warning
Hazard systems certainly constitute a significant option available to the authorities in charge of risk management and
Risk governance. Two categories of landslide early warning systems (LEWSs) can be defined as a function of the scale
LEWS of analysis. Systems addressing single landslides at slope scale can be named local LEWSs (Lo-LEWSs), systems
Warning management
operating over wide areas at regional scale are herein referred to as territorial systems (Te-LEWSs). In the
Performance
literature there are several proposals schematizing the structure of LEWSs. They highlight the importance of the
Education
interconnection among different know-how and system components, as well as the key role played by the actors
involved in the design and deployment of these systems. This worldwide review is organized describing and
discussing the main components of 24 Te-LEWSs, following an original conceptual model based on four main
tiles: setting, modelling, warning and response. Te-LEWSs are predominantly managed by governmental in-
stitutions, thus information is often difficult to find in the literature and, when available, it is not always
complete and thorough. The information considered herein has been retrieved from different sources: articles
published in the scientific literature, grey literature, personal contacts with system managers, and web pages. Te-
LEWSs mainly deal with rainfall-induced landslides, thus pluviometers are the main monitoring instruments.
Intensity duration thresholds are typically employed and meteorological modelling is often used to forecast the
expected amount of rainfall in order to issue a warning with a given lead time. Public or internal statements are
disseminated for increasing the preparedness of both the public and institutions or agencies. Since the beginning
of the 21st century, Te-LEWSs are slowly becoming a commonly used risk mitigation option, employed
worldwide, for landslide risk management over wide areas. Considerations and insights on key-points for the
success or the failure of Te-LEWSs are presented, differentiating among issues related to the efficiency and the
effectiveness of the system. Among them, the important role played by performance analyses of the warning
model for increasing the system efficiency is thoroughly discussed.

1. Introduction populated areas characterized, at specific times, by an intolerable level

of landslide hazard (Calvello, 2017). Within the landslide risk man-
Over the last decades, significant consequences in terms of eco- agement framework proposed by Fell et al. (2005), landslide early
nomic losses and fatalities have been caused by natural hazards warning systems (LEWSs) may be considered a non-structural passive
worldwide (Barredo, 2009; European Environment Agency, 2010; mitigation measure. LEWSs differ widely depending on: the type of
CRED, 2011; Alfieri et al., 2012). Many natural disasters are related to landslides and their predisposing and triggering factors; the scale of
extreme weather events, which are continuously increasing in many operation—i.e. the size of the area covered by the system. The scale of
parts of the world due to climate change and global warning (Easterling operation influences several aspects connected to the design and em-
et al., 2000; Morss et al., 2011; IPCC, 2014). Landslides are natural ployment of a LEWS, such as: stakeholders involved, monitoring net-
hazardous phenomena often connected with severe social and economic work, type of landslides addressed, variables to be considered for the
consequences. The geographical pattern of fatal landslides and ha- warning model, the process of disseminating information, emergency
zardous landslide areas depend largely on relief, precipitation and the plan, education activity. Two categories of LEWSs can be defined con-
distribution and abundance of the population (Gariano and Guzzetti, sidering the scale of analysis (Bazin, 2012; Thiebes et al., 2012; Calvello
2016). Operational landslide early warning systems (LEWSs) aim at et al., 2015). Systems addressing single landslides at slope scale are
reducing the loss-of-life probability by inviting people to act properly in herein referred to as local LEWSs (Lo-LEWSs), and systems dealing with

⁎
Corresponding author.
E-mail addresses: [email protected] (L. Piciullo), [email protected] (M. Calvello), [email protected] (J.M. Cepeda).

https://doi.org/10.1016/j.earscirev.2018.02.013
Received 23 April 2017; Received in revised form 8 February 2018; Accepted 16 February 2018
Available online 21 February 2018
0012-8252/ © 2018 Elsevier B.V. All rights reserved.
L. Piciullo et al. Earth-Science Reviews 179 (2018) 228–247

the possible occurrence of multiple landslides at regional scale are requires many systematic approaches and diverse activities spanning
named territorial systems (Te-LEWSs). The adjective “territorial” is the four elements previously described. These activities need to be or-
herein preferred over the most commonly used adjective “regional” to iented firstly at identifying target populations potentially at risk, then at
provide a more general name for all the LEWSs employed over a wide increasing the human understanding of warnings and, finally, at gen-
area, e.g. a nation, a region, a municipal territory, a river catchment. erating public information tailored to target groups and making in-
In recent years scientists, governmental agencies and NGOs have novative use of the media and education systems. The term “early” in
shown an increasing interest on Te-LEWSs and several cooperative EWSs does not simply mean doing things faster but, just as importantly,
projects have been funded worldwide on this topic. However, at the doing things effectively (Hall, 2007), i.e. the elements at risk need to
present stage, a state of art paper on Te-LEWSs is still missing from the react to a warning appropriately and timely. Many reports of the World
literature. In the first part, this review provides a summary of the Bank (WDR, 2014; World Bank and GFDRR, 2013) outline the im-
schemes proposed in literature to describe structure and elements of a portance of EWS in reducing fatalities and providing cost-effective
LEWS. From this synthesis, an original conceptual model of the main means of mitigating the damage from natural hazards. The benefits may
tiles necessary to develop and manage an operational system is in- exceed costs by a margin of four to one at the global level (Rogers and
troduced. Then, the review summarizes all the available information Tsirkunov, 2010; Teisberg and Weiher, 2009). In the past decades,
collected on Te-LEWSs operational worldwide, in relation to the four EWSs have been developed around the world for a wide range of nat-
main tiles of the conceptual model proposed: setting, modelling, ural hazards, such as: extreme weather events, earthquakes, tsunamis,
warning and response. The material herein reviewed is presented by floods, volcanic eruptions, droughts, snow avalanches, and landslides.
means of summary tables in order to provide easily accessible in-
formation to stakeholders interested or involved in different aspects of 2.2. Proposed schemes on the structure of landslide early warning systems
Te-LEWSs. Finally, considerations on key-points of success and/or
failure of these systems, together with some insights on how to design The continuous urbanization process in landslide prone areas and
and operate a reliable LEWS, are provided. the increasing number of extreme atmospheric phenomena have dras-
tically raised, worldwide, the exposure of people affected by rainfall-
2. Landslide risk management with early warning systems induced landslides. A variety of options are available to mitigate
landslide risk: active measures addressing the reduction of the prob-
2.1. Warning systems as people-oriented risk mitigation measures ability of occurrence of landslides (e.g., modification of slope profile;
lowering of the water level and pore water pressure, reinforcements);
In general terms, an early warning systems (EWS) is an important, structural engineering works designed to decrease the vulnerability of
dynamic and non-structural mitigation alternative, upgradable over elements at risk (e.g., barriers, basins, protections); non-structural risk
time to reduce the risk for human life associated to the occurrence of mitigation measures (e.g., landslide early warning systems, land-use
hazardous events. Early warning constitutes a process where informa- planning, awareness, acceptance). Among the many mitigation mea-
tion generated from tailored observations of natural phenomena is sures available for reducing the risk to life related to rainfall-induced
provided to communities at risk, or to institutions which are involved in landslides, landslide EWSs (LEWSs) certainly constitute a significant
emergency response operations, so that certain tasks may be executed option available to the authorities in charge of risk management and
before a catastrophic event impacts such communities (Villagrán de governance.
León et al., 2013). EWSs can be defined as the set of capacities needed The first scheme of the structure of LEWSs presented in the litera-
to generate and disseminate timely and meaningful warning informa- ture (Di Biagio and Kjekstad, 2007) employs a flow chart to outline four
tion to enable individuals, communities and organizations threatened main sequential activities for such systems: monitoring, analysis and
by a hazard to prepare and to act appropriately and in sufficient time to forecasting, warning, and response. According to this scheme, the key
reduce the possibility of harm or loss (UNISDR, 2009). This definition is technical issue for the operation of an effective LEWS is the identifi-
rather concise yet it highlights the importance assumed, within such cation, measurement and monitoring of precursors of the occurrence of
systems, by the people as elements at risk. In the Hyogo framework for landslides. The choice of precursors varies with the type of landslides to
Action 2005–2015 (UNISDR, 2005), EWSs were recognized as im- be monitored (Lacasse and Nadim, 2009) and with the system objec-
portant tools for disaster risk reduction and for achieving sustainable tives. Typical examples of precursors for rainfall-induced landslides are
development and livelihoods. The succeeding Sendai Framework for heavy rains, rapid increase of pore water pressures, displacements,
Disaster Risk Reduction 2015-2030 (UN, 2015) corroborates this idea velocities and accelerations of existing phenomena. Depending on the
by defining one of its seven global targets as follows: “substantially type of precursor, typical instruments used within the monitoring net-
increase the availability and access to multi-hazard early warning sys- work of a landslide warning system include: pluviometers, in-
tems and disaster risk information and assessments to the people by clinometers, extensometers and other devices measuring ground or
2030.” subsurface movements, geophones, piezometers and water-content
According to UNISDR (2006), people-oriented EWSs always com- gauges.
prise, independently from the type of threat, few essential interrelated The importance attributed to people in LEWSs emerged in the
elements: i) knowledge of risks; ii) monitoring, analysis and forecasting ILEWS project (Integrative Landslide Early Warning Systems), funded
of hazards; iii) communication and dissemination of alerts and warn- by the German Federal Ministry of Education and Research in the
ings; and iv) local capabilities to respond to warnings. Knowledge of period 2007–2010 (Bell et al., 2009; Thiebes et al., 2012). The overall
risks means the study of hazards and vulnerabilities in a given area goal of the project was to develop and implement a transferable early
aimed at defining a level of risk. Monitoring deals with the collection of warning concept starting with sensors in the field and modelling of
data necessary to control, in time, the trend of variables which sig- early warning, and ending with user-optimized action advice embedded
nificantly affect the hazard and the risk level. To this end, the equip- in a holistic risk management strategy. The general structure of ILEWS
ment used can be very different depending on the purpose, the char- clearly distinguished natural-scientific interrelations from social sys-
acteristics and the scale of the warning system. Communication and tems and it was based on three modules: monitoring, modelling and
dissemination of warnings aims at informing people at risk. Finally, implementation. The main aim of the first module was the definition
response capability may be associated to the education of the popula- and operation of a monitoring system at the Swabian Alb, south-west
tion, to the information provided on how to act in areas at risk and to Germany, for hydrological and slope movement purposes. The model-
specific procedures adopted for handling emergency situations. ling module focused on data analysis providing reliable information on
Basher (2006) outlines that the setup of a people-oriented EWS future slope behaviour based on a range of modelling approaches. The

229
L. Piciullo et al. Earth-Science Reviews 179 (2018) 228–247

main objective of the implementation module was cooperative risk 2.3. Scale of analysis
management and end-user optimized warning communication.
Intrieri et al. (2012) slightly modified the diagram proposed by Di LEWSs can be designed and employed at slope scale (Lo-LEWSs) or
Biagio and Kjekstad (2007), adding, to the four main activities, risk at regional scale (Te-LEWSs). An initial key difference between Lo-
knowledge as first fundamental step. The knowledge of the geological, LEWSs and Te-LEWSs is the knowledge “a priori” of the areas affected
geomorphological, hydrogeological and geotechnical features of the by future landsliding. When the exact location of future landslides is
landslide-prone areas is necessary to identify the most critical parameters unknown and the area of interest extends beyond a single slope, only
to be monitored. Subsequently, by elaborating the definitions provided by Te-LEWS can be used. Conversely, Lo-LEWSs are typically adopted to
UNISDR (2006), Intrieri et al. (2013) highlighted the main elements of cope with the risk related to one or more known well-identified land-
LEWS as a balanced combination of the following four components: slides.
planning, monitoring, forecasting, and education. The planning compo- The main aim of a Lo-LEWS is the implementation of protective
nent is mainly focused on defining: needs and vulnerability of exposed actions in defined areas where, at specific times, the risk level to which
people; identification of constraints that people can encounter when acting people are exposed is intolerably high. Depending on the features of
in response to an alert; characterization of geological and meteorological both the landslides and the elements at risk, different protective actions
conditions that contribute to trigger landslides; definition of geo-in- may be employed, including road closures or the temporary evacuation
dicators. The monitoring, which includes instruments selection and in- of people from given zones. In a Lo-LEWS many relevant parameters,
stallation, is a crucial activity to gather data on factors triggering land- either related to the activity of an existing landslide system or to the
slides within the area of operation of the LEWS. The forecasting, which is slope susceptibility to new instability processes, are typically mon-
considered by Intrieri et al. (2013) the main process of a LEWS, includes itored. The monitoring network may include geodetic, geotechnical,
the definition of thresholds, models and other elements necessary to issue geophysical and remote sensing technologies. The characteristics of Lo-
a warning. Finally, the education activities aim at making people aware LEWSs are strongly affected by numerous constraints and factors, from
about the risk to which they are exposed and at clearly explaining to them time to time different, related to the specific problem to address. An
the behaviour to adopt during emergencies. interesting contribution aiming at providing guidance for the design of
Calvello et al. (2015) combined the different contributions, earlier such systems is proposed within a deliverable of the SafeLand European
presented, for designing and managing LEWSs to propose a schematic project (Bazin, 2012). The deliverable mainly addresses the technical
representation based on four levels of analysis, shown as concentric and practical issues related to the monitoring phase, identifying the best
rings within a wheel scheme. The scheme acknowledges that the op- technologies available for landslide monitoring in the context of both
eration of an effective LEWS necessarily implies synergy between hazard assessment and system design. Many other examples of land-
technical and social skills (Glade et al., 2008; Bell et al., 2009). The slides monitored at slope scale within a Lo-LEWS are available in the
main activities of the second level of analysis include: monitoring, literature (Clark et al., 1996; Lollino et al., 2002; Flentje et al., 2005; Di
modelling, warning, emergency, education and decision making. The Biagio and Kjekstad, 2007; Blikra, 2008; Cardellini and Osimani, 2008;
third level highlights the means necessary to accomplish the above- Badoux et al., 2009; Yin et al., 2010; Froese and Moreno, 2014; Jakob
mentioned activities, i.e. instruments, thresholds definition, warning et al., 2012; Intrieri et al., 2012; Olivieri et al., 2012; Michoud et al.,
levels and procedures. The fourth level of analysis, located at the centre 2013; Thiebes et al., 2013; Loew et al., 2016; among others).
of the scheme, highlights four fundamental elements of LEWSs: data, Early warning systems for rainfall-induced landslides operating at
thresholds, alerts, and people response. A final important aspect in- regional scale (Te-LEWSs) are used to assess the probability of occur-
troduced within this scheme is the temporal continuity of the activities rence of multiple landslides over appropriately-defined warning zones,
to be undertaken for updating the system during its operational life. typically through the prediction and monitoring of meteorological
A recent scheme of the structure of a LEWS has been presented by variables. The exact location of landslides is not known a priori. They
Fathani et al. (2016), combining the four interrelated elements pro- typically operate over wide areas affected by diffuse instability pro-
posed by UNISDR (2006) with the hybrid sociotechnical approach for blems (Huggel et al., 2008), such as a catchment, a municipality
disaster risk reduction of Karnawati et al. (2013a, 2013b). Basically the (Cheung et al., 2006; Calvello et al., 2015), a region (Baum and Godt,
main four elements from UNISDR (2006) have been detailed in seven 2010) or an entire country (Rossi et al., 2012; Piciullo et al., 2017a,
sub-systems: risk assessment and mapping; dissemination and commu- 2017b). Their purpose is to provide generalized warnings to authorities,
nication; establishment of disaster preparedness and response team; civil protection personnel and the population. Te-LEWSs principally
evacuation map; standard operating procedures for evacuation; hazard rely on empirical models, often implementing rainfall thresholds to
monitoring and warning services. The main purpose of this scheme was define the different warning levels of the system. Further references and
the proposal of an integrated method to develop a standard for com- a detailed comparison of Te-LEWSs employed at regional scale in many
munity-based early warning systems, including education activities areas of the world are provided in the following sections.
oriented at informing people about the risk to which they are exposed.
Calvello (2017) propose a scheme of the components of early 3. Components of territorial early warning systems for rainfall-
warning systems for weather-induced landslides based on a clear dis- induced landslides
tinction among landslide models, warning models and warning systems,
wherein a landslide model is one of the components of a warning model The different schemes describing the structure of a LEWS presented
and the latter is one of the components of a warning system. Within this in Section 2.2, clearly highlight the importance of both social and
framework, the main components of a landslide model are four: technical aspects in the design and management of such systems. Risk
weather, monitoring, geo characterization, and landslide event. The knowledge, monitoring, forecasting/modelling are common elements of
other two components of the warning model are: warning criteria, and each proposed scheme (see Table 1). They all belong to the technical
the warning event. Finally, four components complete the warning field, whereas communication and dissemination of warnings, cap-
system: warning dissemination, communication and education, com- abilities to respond and education programs concern the social one.
munity involvement, and an emergency plan. Most of the schemes and components recently proposed in the literature
Similarities and differences of the schemes proposed in literature to to describe the structure of a LEWS are incomplete (Table 1), thus a
describe the structure of LEWSs, are highlighted in Table 1. The ele- more detailed and focused conceptual model can be provided for a
ments referring to the same processes are placed on the same rows. The territorial EWS for rainfall-induced landslides. To this aim Fig. 1 shows
columns are organized according to the year of the publication where an original scheme, derived from a similar framework proposed by
the scheme is provided. Calvello and Piciullo (2016), highlighting the main components

230
L. Piciullo et al. Earth-Science Reviews 179 (2018) 228–247

Table 1
Comparison between UNISDR (2006) and schemes of LEWS proposed in literature, organized in terms of similarities and differences.

UNISDR (2006) Di Biagio and Bell et al. (2009) Intrieri et al. Calvello et al. Fathani et al. (2016) Calvello (2017) This paper
Kjekstad (2007) (2013) (2015)

Risk knowledge Planning Decision making Risk knowledge Components of landslide model Setting
(part of warning model and
Monitoring and warning Monitoring Monitoring Monitoring Monitoring Monitoring and warning system) Modelling
service Warning
Analysis and Modelling Forecasting Modelling
forecasting

Warning Warning Components of warning model Warning

(part of warning system)

Dissemination and Implementation Dissemination and Other components of warning Response

communication communication system

Response capability Emergency plan Emergency plan Response capability

Education Education

landslide events as single slope failures, yet the focus of Te-LEWSs is the
occurrence of multiple landslides over a warning area. Therefore in
such cases, a landslide event should more correctly be defined as a
series of landslides grouped on the basis of their characteristics, so as to
implicitly evaluate and classify the magnitude of a set of multiple
phenomena occurring in a given area within a given time period
(Calvello and Piciullo, 2016). The third module of the proposed scheme
of structure for a territorial EWS for rainfall-induced landslides is the
warning strategy, necessary for defining the number of warning levels
to be employed in the system and the decision making procedures
needed for issuing warnings in a certain warning zone, i.e. the portion
of territory alerted with the same level of warning. The majority of
systems employ warning zones that are fixed in space; yet they can also
be variable, when they are formed grouping a certain number of pre-
defined territorial units (Piciullo et al., 2017a, 2017b). In the latter case
these territorial units represent the minimum spatial discretization
adopted for warnings, i.e. the portion of the territory for which a certain
Fig. 1. Scheme of the components of a territorial EWS for rainfall-induced landslides. level of warning can be independently issued. The fourth and final
module is the response strategy. It includes the education and the
necessary to design and manage a Te-LEWS. The conceptual model is communication strategies, the emergency plan and every other element
structured as a jigsaw puzzle because a weakness or failure in any one needed to effectively manage these systems, i.e. to reduce the risk to life
of the components could result in a loss of reliability or failure of the from rainfall-induced landslides in a given area at a given time.
whole system. The scheme is based on four main tiles, i.e. main modules The modelling and warning strategy, considered together, constitute
of the warning system: i) setting, ii) modelling, iii) warning strategy, the warning model, i.e. the selection of rainfall thresholds, and/or
and iv) response. The first three modules refer to technical aspects, thresholds considering other variables, to be employed for issuing dif-
whereas the last one to social aspects. Each tile may include several ferent levels of warning, the procedures to activate/deactivate a level of
components, but only the most important ones are reported in the warning, and the warning zones considered in the system. Finally, the
Figure. A component has links with either one or more of the other four tiles together constitute the complete scheme of structure for EWS
components inside a tile. for rainfall-induced landslides.
The setting module includes all the preliminary actions and choices
needed to start the design of a Te-LEWS, such as: the delimitation of the 4. Territorial landslide early warning systems operational
area where the system will operate, the types of landslides to address worldwide
and the choice of the monitoring instruments to use. For instance, early
warning systems need to employ near-real-time monitoring networks Over the last two decades many Te-LEWSs have been designed as
(Reid et al., 2008). The monitoring is the main link to the second landslide risk mitigation measures all over the world. Many reasons
module of the system: the modelling. For instance, the time-dependent may be identified for that: perceived cost-effectiveness of these non-
information on variables considered relevant for triggering landslide is structural risk mitigation measures in comparison to structural ones;
the main input to define rainfall thresholds. In the case of Te-LEWSs easy applicability over large and densely populated areas, where the
designed for rainfall-induced landslides, the thresholds definition is risk to people is widespread; improved knowledge on rainfall-landslide
typically based on frequent rainfall measures from pluviometers. In the correlations; increasing reliability in weather nowcasting; technologic
literature, several studies deal with the assessment of rainfall thresholds developments. A summary of the period of activity and the location of
responsible for landslide phenomena (Tiranti and Rabuffetti, 2010; the 24 Te-LEWSs operational worldwide considered in this review is
Giannecchini et al., 2012; Martelloni et al., 2012; Gariano et al., 2015; shown in Fig. 2. Before 2005 only few experiences on systems at re-
Segoni et al., 2014; Staley et al., 2013). Except for Giannecchini et al. gional scale have been carried out (AS1, AS2, NA1, NA2, NA3, NA4,
(2012), the other studies are based on correlations considering SA1) and the majority of them were designed and managed in USA.
Since the beginning of the 21th century, many new Te-LEWSs had been

231
L. Piciullo et al. Earth-Science Reviews 179 (2018) 228–247

Fig. 2. Territorial landslide early warning systems reviewed: location and period of activity. Legend: red squares = dates of catastrophic landslide events; dark blue = period of activity,
retrieved from reliable references; light blue = period of activity, assumed by Authors. (For interpretation of the references to colour in this figure legend, the reader is referred to the web
version of this article.)

emerging around the world, in Asia and Europe, particularly in Italy. In the system was found. Only two of the reported systems are surely not
9 cases out of 24 the system was designed and employed following active anymore, both of them located in USA, respectively employed in
catastrophic landslide events that caused many victims and significant the San Francisco Bay, California (NA1), and in the city of Seattle,
economic losses (see red square in Fig. 2). A significant example is re- Washington (NA4). The first terminated in 1995, mainly because the
presented by the system developed and still operational in Hong Kong, managing office, i.e. the National Weather Service, relocated and un-
China. In the late 1960s and early 1970s a high number of disastrous derwent a net staff reduction (Wilson, 2004); the latter was employed
landslides affected Hong Kong. In particular in 1972, 67 and 71 people under an informal agreement between USGS, NWS and the city of
died as a result of two different landslides, respectively occurring at Po Seattle only for four years, in the period 2002–2005 (Baum and Godt,
Shan Road and Sau Mau Ping (Cheung et al., 2006). The victims and the 2010).
economic losses of these events led to the establishment of the muni- Table 2 presents information on the institution in charge of the
cipal Geotechnical Control Office (now the Geotechnical Engineering LEWS, on the types of hazard addressed and on the source of the data
Office, GEO), with the main aims of enforcing slope safety and miti- used for analysing the system. In the majority of cases, the LEWSs are
gating landslide risk by alerting the public of the possible occurrence of managed by governmental agencies, often directly involved in civil
landslides during prolonged heavy rainfall. A key component of the defence and landslide risk mitigation measures. Many systems deal with
landslide risk mitigation strategy employed by the Hong Kong GEO in multiple natural hazards, such as: hurricanes, floods, typhoons and
the early years of its operations was the creation, in 1977, of the mu- snow avalanches. For instance, the Alerta-Rio system in Rio de Janeiro
nicipal Landslip Warning System. Since 1984 the system is jointly op- (SA1) issues two distinct warnings, one for extreme rainfall events and
erated by the GEO and the Hong Kong Observatory. one for landslides, based on different rainfall thresholds. A similar si-
The information on the period of activity of the Te-LEWSs results tuation occurs for all the regional systems operational in Italy (EU1,
from different sources: articles published in the scientific literature, EU2, EU3, EU4, EU5), designed to provide warnings for heavy rainfall,
grey literature, web pages and, in some cases, personal contacts of the floods and landslides. In Norway (EU7), the hydrological HBV-model
authors with system managers. The state of activity of a given LEWS (Beldring et al., 2003), i.e. a key component of the LEWS, is also used
was considered as certain (dark blue bars in the Figure) up to the date for issuing floods and snow warnings. In Eastern USA, the EWS for the
of the latest source of information, and uncertain (light blue bars in the Appalachians (NA3) is principally designed to issue hurricane warnings
Figure) for the following years unless information on the termination of and, as a consequence, also to inform the public on the possible

232
L. Piciullo et al. Earth-Science Reviews 179 (2018) 228–247

Table 2
Territorial landslide early warning systems reviewed: managing institution, types of hazard addressed, sources of information, date of most recent information. Legend: HK-GEO, Hong
Kong Geotechnical Engineering Office; PLUS, Expressway Malaysia; MILIT, Ministry of Land, Infrastructure, Transport and Tourism; JMA, Japan Meteorological Agency; DGH, Directorate
General of Highways; NCDR, National Science and Technology Center for Disaster Reduction; NBRO, National Building Research Organization; BUET-JIDPUS, Bangladesh University of
Engineering and Technology-Japan Institute of Disaster Prevention and Urban Safety; CNR-IRPI, National Research Centre; NVE, Norwegian Water Resources and Energy Directorate;
USGS, US Geological Survey; NWS, National Weather Service; ODOT, Oregon Department of Transportation; DOGAMI, Oregon Department of Geology and Mineral Industries; NOOA,
National Oceanic and Atmospheric Administration; Co-Rio, Municipal Operations Center; SDC, Swiss Agency for Development and Cooperation.

CODE Institution Type of hazard Sources of information Latest

information

AS1 HK-GEO Landslides Brand et al., 1984; Chan et al., 2003; Cheung et al., 2006; Cheung et al., 2006; 2016
Information Note, 2009; Lumb, 1975; Massey et al., 2001; Pang et al., 2000; Pun et al.,
2003; Wong, 2006; Wong et al., 2006; Yu et al., 2004; Yu et al., 2004; personal contacts
AS2 No info Landslides Yin et al., 2007; Zhang et al., 2011 2011
AS3 No info Landslides Yin et al., 2007; Zhong et al., 2009 2009
AS4 PLUS Landslides Huat et al., 2012; Lloyd et al., 2001 2012
AS5 MLIT and JMA Landslides Kuramoto et al., 2001; Kuramoto et al., 2005; Okada, 2005; Osanai et al., 2010 2010
AS6 DGH Landslides Chen et al., 2007; Wu et al., 2011 2010
AS7 NCDR Typhoons-induced landslides Huang and Hong, 2010; Su et al., 2010 2010
AS8 NBRO Landslides Web page from NBRO 2017 (last access: April 2017) 2016
AS9 BUET-JIDPUS Landslides Ahmed and Murillo, 2015; Web page from Bayes Ahmed 2015 (last access: April 2017); 2016
Web page from ICIMOD 2017 (last access: April 2017)
EU1 Civil defence Heavy rainfall, landslides Segoni et al., 2015 2016
EU2 Civil defence Heavy rainfall, floods and landslides DPRC 299-30/06/2005; personal contacts 2016
EU3 Civil defence Heavy rainfall, floods and landslides Lagomarsino et al., 2013, 2015; Martelloni et al., 2012 2016
EU4 Civil defence Heavy rainfall, floods and landslides Tiranti et al., 2014; Tiranti and Rabuffetti, 2010; Web page from ARPA (last access: April 2016
2017); personal contacts
EU5 Civil defence Heavy rainfall, floods and landslides Brocca et al., 2008; DGR, 2007; Ponziani et al., 2013; Web page from CFU Umbria (last 2016
access: April 2017)
EU6 CNR-IRPI Landslides Brunetti et al., 2010; Rossi et al., 2012 2016
EU7 NVE Landslides, floods, snow avalanches Beldring et al., 2003; Boje et al., 2014; Colleuille et al., 2010; Devoli et al., 2014; Piciullo 2016
et al., 2017a, 2017b; personal contacts
NA1 USGS and NWS Landslides Cannon, 1988; Cannon and Ellen, 1985a, 1985b; Ellen and Wieczorek, 1988; Wieczorek, 2016
1987; Wilson, 2004
NA2 ODOT and DOGAM Landslides Baum and Godt, 2010; DOGAMI, 2005; Jibson, 1989; Larsen and Simon, 1993; Mills, 2016
2002; Montgomery et al., 2000; Wieczorek and Morgan, 2008; Wiley, 2000; Wilson and
Wieczorek, 1995; Web page from DOGAMI (last access: April 2017)
NA3 USGS Hurricanes, landslides Baum, 2007; Wieczorek et al., 2000 2016
NA4 USGS, NWS, City of Seattle Landslides Baum and Godt, 2010; Baum et al., 1998, 2005; Chleborad, 2000, 2003; Chleborad et al., 2016
2006, 2008; Gerstel et al., 1997; Godt et al., 2006; Laprade et al., 2000; Tubbs, 1974
NA5 NOAA and USGS Landslides and floods Campbell, 1975; Cannon and Gartner, 2005; Cannon et al., 2008, 2011; NOAA-USGS 2016
Debris-Flow Task Force, 2005; Restrepo, 2008; Restrepo et al., 2009; Staley et al., 2013;
USGS, 2005
SA1 GEO-Rio Landslides, heavy rainfall Calvello and Piciullo, 2016; Calvello et al., 2015; d'Orsi et al., 1997, 2000, 2004, 2013; 2016
d'Orsi, 2012; Tatizana et al., 1987; personal contacts
SA2 SDC Landslides Huggel et al., 2008, 2010; Huggel, 2012; Terlien, 1998 2012
SA3 GEO-Rio Landslides Calvello et al., 2014; d'Orsi, 2012; personal contacts 2016

occurrence of landslides if certain rainfall thresholds are exceeded. instruments adopted for providing information on precipitation in near
It is safe to assume that other Te-LEWSs are either operational or at real-time. Only the national LEWS operating in Japan (AS5) does not
a prototype stage around the world. If they are not reported herein, it is employ a network of pluviometers as main tool for rainfall monitoring.
either because the Authors did not find any published resource dealing In this case, the rainfall data used in the system are rainfall intensities
those systems or because the reported information was too scarce, like estimated by a Radar Automated Meteorological Data Acquisition
in the case of the national systems for Iceland and England, respectively System and distributed by the Japan Meteorological Agency. Since
managed by the Icelandic Met Office (IMO) and the British Geological 2001 near real-time rainfall intensities and short-term 1 to 6 h rainfall
Survey (BGS). Indeed, the aim of these two systems is principally or- forecasts are available for the whole country over a regular 2.5-km grid
iented at weather forecast and, only recently, different probabilities of mesh (5-km grid mesh since 1988). In addition to rainfall monitoring
landslide occurrence have been linked to rainfall thresholds for warning some systems employ weather forecasts, mainly using nowcasting es-
purposes. In the following section, all the Te-LEWSs reported in Table 2 timates provided by different numerical meteorological models, typi-
will be described and discussed considering the four main components cally developed and deployed at national level.
defined in Section 3: setting, modelling, warning and response.
5.1.2. Landslide types
5. Discussion The majority of the reviewed systems has been set-up to forecast the
possible occurrence of landslides on natural slopes. In this case, the
5.1. Setting landslides considered are mostly debris flows and shallow landslides
reaching extremely rapid velocities (Cruden and Varnes, 1996). Only in
5.1.1. Monitoring systems and weather forecast few cases (AS8, AS9, SA1, SA3) both natural and man-made slopes are
According to the scheme proposed in Fig. 1, the system setting is the considered. As an example, the LEWS operating in Rio the Janeiro
first step needed to design a Te-LEWS. Table 3 reports the main char- (SA1) considers a wide range of well-defined landslide typologies. The
acteristics of the setting for the 24 systems reviewed herein. As ex- system managers—i.e. the GEO-Rio Foundation—defined 13 typologies
pected for LEWSs operating at regional scale, rainfall is the main of landslides: earth slide on excavated artificial slope; earth and rock
monitored variable. Automatic pluviometers are the most common slide on excavated artificial slope; rock slide on excavated artificial

233
 L. Piciullo et al.

Table 3
Te-LEWS reviewed: general information on the setting. Legend: SWIRL, Short-range Warning of Intense Rainstorms in Localized Systems; RBFN Radial Basis Function Network, AMeDAS Automated Meteorological Data Acquisition System,
QPESUMS Quantitative Precipitation Estimation-Segregation Using Multiple Sensor, WRF Weather Research & Forecasting model; LAMI, Limited Area Model Italia; COSMO, Consortium for Small-scale Modeling TREC Tracking Radar Echoes by
Correlation, QPF Quantitative Precipitation Forecast; NWS National Meteorological Center, FFMP Flash Flood Monitoring and Prediction, PMAR high-resolution weather forecasting program ECMWF European Centre for Medium-Range Weather
Forecasts.

CODE Monitoring system Additional tools Weather forecast Variable evaluated Covered area Landslide type Slopes

AS1 110 Pluviometers, radars None SWIRL mm/1 h–3 h Municipality of Hong Kong Cut slopes, rock slopes, fill slopes, Man-made
retaining walls
AS2 Pluviometers WebGIS No info 6 h, 12 h, 24 h, 48 h Zhejiang Province Rainfall-induced landslides Natural
AS3 no info WebGIS No info 24 h Hubei Province Rainfall-induced landslides Natural
AS4 70 Pluviometers, piezometers WebGIS None None North-South Expressway, Debris flow Natural
Malaysia
AS5 Radar RBFN Radar AMeDAS analytical mm/1 h–6 h Japan Debris flows, shallow landslides Natural
rainfall
AS6 Pluviometers, radar none QPESUMS No info Areas at risk along highway in Shallow landslides and debris flows Natural
Taiwan
AS7 Pluviometers, radar GIS WRF No info Taiwan Shallow landslides and debris flows Natural
AS8 Pluviometers, radar None None No info Ten districts Rainfall-induced landslides Natural/man-
made
AS9 34 pluviometers WebGIS UK-based private company No info Chittagong Metropolitan area Rainfall-induced landslides Natural/man-
made
EU1 332 Pluviometers, radar LOKAL-LAPS, webGIS LAMI model mm/6, 12, 24, 48 h Toscana region Shallow landslides, debris slides/ Natural

234
flows
EU2 147 Pluviometers, radar LOKAL-LAPS LAMI model mm/6, 12, 24 h Campania region Shallow landslides, debris slides/ Natural
flows
EU3 Pluviometers, radar LOKAL-LAPS LAMI model mm/6, 12, 24 h Emilia Romagna region Shallow landslides, debris slides/ Natural
flows
EU4 Pluviometers, radar LOKAL-LAPS, GIS LAMI model, TREC technique mm/1, 6, 12, 24 h Piemonte Region Shallow landslides Natural
EU5 Pluviometers, radar web-GIS COSMO ME mm/24, 72 h Umbria region Shallow landslides, debris slides/ Natural
flows
EU6 1950 pluviometers WebGIS LAMI model QPF twice a day Italy Rainfall-induced landslides Natural
(3 days)
EU7 pluviometers, piezometers, tensiometers, Precipitation-runoff HBV- AROME, EC, HIRLAM QPF (2 + 4 days) Norway Debris flows, shallow landslides, slush Natural
radar model, webGIS flows, debris avalanches
NA1 Pluviometers, transportable weather radar None Long-term weather trends QPF twice a day San Francisco Bay region Debris flows and flash-floods Natural
and NWS network radars from the NWS
NA2 Pluviometers No info No info No info West Oregon Landslide/Debris flow Natural
NA3 Pluviometers No info No info No info Appalachians Shallow landslides Natural
NA4 Pluviometer, piezometer, radar None NWS quantitative QPF Municipality of Seattle Shallow landslides and debris flows Natural
precipitation forecasts
NA5 Pluviometer, radar No info NWS's FFMP system QPF twice a day Southern California Debris flows and flash-floods Natural
SA1 33 Pluviometers, two radars No info PMAR mm/24 h Municipality of Rio de Janeiro 13 Natural/man-
made
SA2 Pluviometers None ECMWF mm/6 h Combeima valley Rainfall-induced landslides No info
SA3 86 Pluviometers None None mm/6, 12, 24 h 113 communities (i.e. Favelas) of 13 Natural/man-
Rio de Janeiro made
Earth-Science Reviews 179 (2018) 228–247
L. Piciullo et al. Earth-Science Reviews 179 (2018) 228–247

Fig. 3. Distribution of reviewed Te-LEWSs for: a) territorial level; b) area covered by the system.

slope; earth work failure; earth slide on natural slope; earth and rock- investigation on rainfall thresholds for the initiation of landslides is pre-
slide on natural slope; rock slide on natural slope; rock fall and block sented in Guzzetti et al. (2007), wherein the authors identify three main
sand slabs; talus movements; failure of slope stabilisation work; waste categories: A) thresholds that combine precipitation measurements ob-
material slide; flow; erosion; landfill. Landslides of waste material, al- tained from specific rainfall events; B) thresholds that consider antecedent
though not common among the LEWSs reviewed herein, constitute a rainfall conditions; and C) other thresholds. For the first category four sub-
significant risk to human lives in many informal urban communities categories are further defined, depending on the considered rainfall
(i.e. favelas) of developing countries, as demonstrated by deadly land- measurements used to characterize a rainfall event: A1) intensity-duration;
slide events registered, in the last decades, in garbage dumps in the A2) rainfall event-duration; A3) total event rainfall; and A4) rainfall event-
Philippines, China, Colombia, Indonesia and Guatemala (Web page intensity. Following this schematization, Table 4 and Fig. 4 report the
from Petley, 2008). A particular example of system exclusively designed main characteristics of the thresholds adopted within the Te-LEWSs re-
for sizeable man-made slopes is the one operational in Hong Kong viewed herein. In 7 cases, the Te-LEWSs employ two rainfall thresholds for
(AS1). The slopes are inventoried by the system managers—i.e. the warnings issuing to be evaluated simultaneously, whereas 10 of them are
Geotechnical Engineering Office of Hong Kong—according to the fol- based on one rainfall threshold. In 7 cases the systems adopt other or more
lowing four categories (Cheung et al., 2006): cut slopes, rock slopes, fill complex rainfall thresholds. Most of the rainfall thresholds employed in
slopes, and retaining walls. these systems—17 out of 24—are intensity-duration thresholds (A1), fol-
lowed by rainfall thresholds considering antecedent rainfall conditions (B)
5.1.3. Covered area and other thresholds (C).
The information on the area covered by each system is summarized In the system employed in the San Francisco Bay (NA1) two
in Fig. 3. The reviewed systems are mostly designed to operate at three thresholds were considered to evaluate the chance for an approaching
different territorial levels: national (AS4, AS5, AS7, EU6, EU7), regional storm to trigger hazardous debris-flow activity: the accumulation of
(AS2, AS3, AS6, AS8, AS9, EU1, EU2, EU3, EU4, EU5, NA1, NA2, NA3, antecedent seasonal rainfall and the combinations of rainfall intensity
NA5, SA2), and municipal (AS1, NA4, SA1). The warning area varies by and duration forecast for the approaching storm. In order to evaluate
orders of magnitude (Fig. 3), even among systems designed to operate the seasonal progress of soil moisture towards the antecedent rainfall
at the same territorial level. The smallest and largest warning areas are threshold, the USGS Landslide Working Group installed and monitored
covered by the LEWSs operating, respectively, in: 103 informal com- shallow piezometers at the La Honda study area, which served as a
munities—i.e. favelas—of Rio de Janeiro (SA3); and in Norway (EU7). benchmark site for the region (Wilson, 2004). The Cannon and Ellen
The first one is a peculiar Te-LEWS system, as each community is (1985a, 1985b) intensity-duration threshold formed the basis for the
treated almost as a separate entity and warnings are based on local debris-flow warning system when the LEWS formally started, in Feb-
rainfall measures, coming from automatic pluviometers transmitting ruary 1986. By 1989, the USGS developed a pair of cumulative rainfall/
data at regular intervals of 15 min. Within each community, the aim of duration relationships for a spectrum of size and frequency of debris
the system is to temporary evacuate the population, by means of sirens, flows (Wilson et al., 1993). The lower “safety” threshold was adapted
from areas mapped at high risk whenever the rainfall exceeds pre- from Wieczorek's (1987) threshold for the initiation of individual debris
defined thresholds. The second example refers to a rather typical na- flows in the La Honda study area and represented a rainfall level below
tional LEWS, employed in Norway to inform in advance the public and which significant debris flow hazards were considered unlikely. The
local emergency authorities about the possible occurrence of cata- upper “danger” threshold was adapted from Cannon and Ellen (1988)
strophic events due to debris flows, debris slides, debris avalanches and and was intended to represent a rainfall level above which abundant
slush flows. Two particular cases are AS4 and AS6. These systems are debris flows were likely to occur across broad areas in the San Francisco
employed along two highways at high risk of landslide, respectively in Bay region. In the Te-LEWS developed in Seattle (NA4), the prototype
Malaysia and Taiwan. The aim is the reduction of the risk for drivers, system was based on two different thresholds: i) an antecedent rainfall
informing and imposing road closures if necessary. threshold (Chleborad, 2000, 2003) considering two variables, the 3-day
precipitation immediately prior to the landslide event and the ante-
cedent precipitation that occurred prior to the 3 days (Chleborad et al.,
5.2. Modelling
2006); ii) a rainfall intensity–duration threshold, considering the
average rainfall intensity for the entire storm and its duration (Godt,
The warning model of a Te-LEWS for rainfall-induced landslides most
2004; Godt et al., 2006). An antecedent water index, representing the
typically consists in defining one or more thresholds for landslide occur-
depth of water above or below the amount required to bring a 2-m-deep
rence in a certain area of interest. The rainfall thresholds can be either
column of soil to “field capacity” (Godt et al., 2006) was also con-
based on a physically-based schematization of the causal relationship be-
sidered. In this system, the monitoring network operated by the USGS
tween rainfall and landslides or on empirical criteria derived from a sta-
provides information in real-time for rainfall, soil moisture, and pore
tistical analysis of historical data. Concerning the latter, a comprehensive

235
L. Piciullo et al. Earth-Science Reviews 179 (2018) 228–247

Table 4
Thresholds considered in the reviewed Te-LEWSs. Rainfall thresholds: A1 — intensity-duration; A2 — rainfall event-duration; B — antecedent condition; C — other or more complex
thresholds.

CODE Rainfall threshold Code Measured variable Other thresholds

AS1 Failure frequency and rolling 24 h rainfall in semi-log plot C mm/24 h None
AS2 Landslide density and effective rainfall C Effective rainfall mm: function of daily rainfall None
AS3 Intensity-duration; antecedent condition A1/B mm/1 d, 2 d, 4 d, 7 d, 10 d, 15 d None
AS4 Intensity-duration A1 mm/3 d and mm/6 d, mm/0, 5 h or mm/1 h or mm/2 h Pore pressure
AS5 Cumulative rainfall and soil-water index for a 5 km grid mesh C mm/60 min Soil water content
AS6 Intensity-duration; event duration A1/A2 mm/d None
AS7 Intensity-duration A1 mm/h None
AS8 Intensity-duration A1 mm/h None
AS9 Intensity-duration A1 mm/1 d, 3 d None
EU1 Intensity-duration A1 mm/h None
EU2 Probabilistic approach considering different return periods 2, 5, and C mm/24 h, mm/48 h, mm/72 h Hydrometric parameters
10 years
EU3 Multiples of the standard deviation of cumulative rainfall distributions C mm/d None
EU4 Intensity-duration A1 mm/h None
EU5 Probabilistic approach considering different return periods 2, 5, and C mm/1, 3, 6, 12, 24, 36, 48 h Soil water content
10 years
EU6 Intensity-duration A1 mm/duration None
EU7 Water supply rain and snowmelt and soil water content C Rain and snowmelt, soil saturation/groundwater None
NA1 Intensity-duration; antecedent condition A1/B mm/h None
NA2 Intensity-duration A1 mm/h None
NA3 Intensity-duration A1 mm/h None
NA4 Intensity duration; antecedent condition A1/B mm/3d and mm/15d prior; mm/h Soil water content
NA5 Peak intensity-duration A1 mm/h None
SA1 Intensity duration; antecedent condition A1/B mm/1 h, mm/24 h, mm/24 and mm/96 h None
SA2 Intensity-duration frequency curves; antecedent condition A1/B mm/15, 30, 60, 120, 360 min None
SA3 Intensity duration; antecedent condition A1/B mm/1 h, mm/12 h, mm/96 h None

pressure, while the National Weather Service provides quantitative LEWS deployed in Japan (AS5) the territory is divided into a regular
precipitation forecasts. One more example of two thresholds definition grid with a cell size of 5 km by 5 km. The rainfall data has been col-
employed for early warning purposes is provided by the LEWS opera- lected, for each cell, at a time resolution of 1 h for > 10 years. Non-
tional in Rio de Janeiro (SA1). A relationship between rainfall and occurrence rainfall probabilities defining a series of “critical lines” have
landslides was established in 1997 based on 65 past events and rainfall been evaluated considering two rainfall indicators: 60-min cumulative
data from a set of five pluviometers (d'Orsi et al., 1997). This pre- rainfall and a soil-water index (Kuramoto et al., 2005). The system
liminary study led to the first criteria for landslide warning adopted by applies a method, called Radial Basis Function Network, to auto-
GEO-Rio which considered a combination of antecedent condition of matically set and refine the critical lines at each cell of the grid. For the
rainfall through the measurement of 24-h and 96-h cumulated rainfall final selection of the most appropriate critical lines, records on debris-
(Calvello et al., 2015). In 2004, a third rainfall variable considering the flow and slope-failure occurrences have been used (Osanai et al., 2010).
intensity duration, i.e. the monitored hourly rainfall, was added to the In Norway (EU7) the thresholds currently used in the system operated
previous two, following a detailed analysis of data from about 800 by NVE have been derived from empirical tree-classification using 206
landslides of different typologies (d'Orsi et al., 2004). Since then, dif- landslide events in different parts of the country (Colleuille et al.,
ferent thresholds were associated with warning levels and a series of 2010), as a function of two variables: normalised rainfall and snowmelt,
either/or rules were established to issuing a certain level of warning and normalised soil water content (Colleuille et al., 2010; Boje et al.,
(Calvello et al., 2015). Also for the LEWS employed in the Combeima 2014). In Italy, as a response to a catastrophic landslide event that
valley (SA2), both intensity-duration and antecedent rainfall thresholds occurred in Sarno in 1998 (Cascini, 2004), a national law has been
are available. In this case, however, the incompleteness of the landslide written on landslide and flood risk management (DPCM, 2005). Fol-
database together with the large variability of antecedent rainfall ob- lowing this law, all the 21 regional departments for Civil Protection
served for past landslide makes the definition of warning thresholds have been asked to design and manage an early warning system for
particularly challenging (Huggel et al., 2008, 2010). rainfall-induced landslides. An interesting example of thresholds defi-
7 Te-LEWSs employ other or more complex thresholds. Interesting nition is provided in the LEWS operational in the Emilia Romagna re-
examples of such LEWSs have been developed in Hong Kong (AS1), gion. The model is called SIGMA (Galliani et al., 2001; Martelloni et al.,
Japan (AS5), Norway (EU7) and in several Italian regions. In Hong 2012) and the name reflects the central role played by standard de-
Kong (AS1) different landslide-rainfall correlations have been devel- viations in the proposed methodology. The SIGMA model originates
oped and used over the years (Lumb, 1975; Brand et al., 1984; Pun from the a.s.c.a.v method (Galliani et al., 2001), which is based on a
et al., 2003; Yu et al., 2004). Currently, following Yu et al. (2004), the statistical analysis of the cumulative rainfalls with an n-day wide
municipal territory is divided into 40 × 40 grid cells, each having a moving window shifting at 1-day time steps along the whole rainfall
planar area of 1.5 km by 1.2 km. The spatial distribution of different record. The area of the region susceptible to landslide events is divided
types of man-made slopes in each cell is known thanks to a “catalogue in a number of territorial units, each one associated to a reference
of slopes” managed by the Hong Kong GEO. A set of bi-linear correla- pluviometer. Starting from daily precipitation data (from 1951 to
tions between landslides frequency and maximum rolling 24-h rainfall 2009), the time series of cumulated rainfall from 1 to 365 days have
exists for the following 4 common types of slopes: soil cut slopes, rock been derived for each pluviometer. The cumulative rainfall series are
cut slopes, fill slopes and retaining walls. The correlations were de- approximated by a standard Gaussian distribution through a target
termined considering the slope failure caused by about 118 rainstorms function: y = α · σ, where α is a constant and σ is the standard devia-
in the period 1984–2001, for which the maximum rolling 24-h rainfall tion. From a particular value of σ the corresponding cumulative fre-
always exceeded 50 mm (Cheung et al., 2006). Also in the national quency sample is calculated and from this value the precipitation (in

236
L. Piciullo et al. Earth-Science Reviews 179 (2018) 228–247

5.3. Warning strategy

Once the thresholds for one or more monitored variables are de-
fined, it is important to combine them in rules for issuing different
warning levels. Through a decisional algorithm it is possible to choose
the thresholds to employ for the activation of a level of warning in a
certain warning zone. Then, each warning level may activate one or
more procedures that agencies, authorities and people need to follow
with the aim of reducing the level of risk to life in the warned area.

5.3.1. Warning levels

The minimum number of warning levels for a Te-LEWS is two,
where the first level refers to a low-probability of landslide occurrences,
e.g. in case of good weather conditions or light rainfall, and the second
level to a likelihood of landslide occurrences. Systems with two levels of
warning simply switch between insignificant risk to potentially dan-
gerous situations. The use of more warning levels means the system
differentiates states with an increased probability of landslides, which
in turn activate different emergency procedures. The majority of Te-
LEWSs (15 out of 24) employs four warning levels (Fig. 5), with the
exception of: the National system operational in Italy (EU6), which has
five levels; the systems in Malaysia (AS4) and Sri Lanka (AS8), with
three warning levels; and the systems operational in Hong Kong (AS1),
Taiwan (AS7) and in the communities of Rio de Janeiro (SA3), based on
two warning levels. No information is available, on this issue, for the
municipal system deployed in Chittagong (AS9). A particular example is
provided by the system employed in Japan (AS5), where the warning
levels are not directly correlated to the exceedance of one or more
thresholds. On the contrary, the five warning levels considered in the
system are set as a function of different lead times before a pre-defined
critical line is exceeded in a certain grid cell.

5.3.2. Warning zones and territorial units

Besides defining and issuing a level of warning, the managers of the
system need to identify the zones where to issue the warnings. Warning
zones can be fixed or variable (Piciullo et al., 2017a, 2017b), in the
latter case they are composed by grouping together territorial units
alerted with the same warning level in a given temporal unit. In this
case, the extent of the warning zones changes in time as a function of
the number of territorial units alerted with the same warning level.
Almost all the Te-LEWSs herein reviewed employ fixed warning zones
with the exception of the systems operational in the Chinese Zhejiang
Fig. 4. a) Classification (following the schematization by Guzzetti et al., 2007) and b) province (AS2) and in Norway (EU7). The definition of the warning
number of thresholds employed in the reviewed Te-LEWSs. zones is generally based on the hydro-geomorphological conditions of
the area to be warned as well as on the characteristics of the employed
mm) of the original series is retrieved. Proceeding in the same way for monitoring network, considering factors such as: number of monitoring
the number of cumulative rainfalls between 1 and 365 days, it is pos- instruments per unit area, rainfall homogeneity, geology and geomor-
sible to build the precipitation curves (σ curves) associated with various phology, extent of the administrative boundaries. For instance, for the
probabilities of non-exceedance. Multiples of the standard deviation (σ)
are used as thresholds to discriminate between ordinary and extra-
ordinary rainfall events (Martelloni et al., 2012). Further proposals of
thresholds definition developed and tested in Italy, yet not officially
employed in any of the 21 operational regional Te-LEWSs, have been
presented in the literature. Among them it is worth mentioning: the
FLaIR (Sirangelo and Versace, 1996; Sirangelo et al., 2003) and SAKe
(Terranova et al., 2015) models tested in Calabria (Capparelli et al.,
2012); the Bayesian approach proposed by Berti et al. (2012), tested in
Emilia Romagna; the warning system for rainfall-induced shallow
landslide proposed by Aleotti (2004) and the MoniFLaIR (Capparelli
and Tiranti, 2010), tested in Piemonte; the SUSHI model (Capparelli
and Versace, 2011) tested on the Pizzo d'Alvano massif in Campania;
the warning model proposed by Piciullo et al. (2017b) for the Campania
region; the rainfall thresholds proposed for the Toscana (Rosi et al.,
2012; Segoni et al., 2014), Umbria (Brunetti et al., 2010), Sicilia
(Gariano et al., 2015) and Calabria (Vennari et al., 2014) regions. Fig. 5. Number of warning levels adopted in the reviewed Te-LEWS.

237
 L. Piciullo et al.

Table 5
Summary of the main components of the warning strategies for the reviewed Te-LEWSs. Legend: HKO, Hong Kong Observatory; GEO, Geotechnical Engineering Office; JMA, Japan Meteorological Agency; DGH, Directorate General of Highways;
CEOC, Central Emergency Operations Center; DPC, National Department for Civil Protection; HF, Section for forecast of flood and landslides hazards; USGS, US Geological Survey; NWS, National Weather Service; ODOT, Oregon Department of
Transportation; Co-Rio, Municipal Operations Center; CREPAD, Regional Emergency Committee of Tolima.

CODE Covered area Warning zones Territorial units Warning levels Decision on issuing/cancelling a warning Additional information

AS1 Municipality of Hong Kong Fixed: 1 Municipality 2 Director of the HKO and the Head of the GEO None
AS2 Zhejiang province Variable No info 4 Government No info
AS3 Hubei province Fixed: 6 Municipality 4 Wuhan meteorological center No info
AS4 Malaysia Fixed Specific tracks of North-South Highway 3 PLUS Headquarters No info
AS5 Japan Fixed: 1 5 km grid mesh 5 JMA and local government Landslide susceptibility map,
AS6 Highway Fixed Highways in mountainous areas 4 DGH Risk points, local observations of rivers and slopes
AS7 Taiwan Fixed Potential areas at risk of landslides 2 CEOC Landslide risk maps
AS8 Sri Lanka Fixed: 10 districts 50 communities 3 Disaster Management Centre Landslide hazard maps
AS9 Chittagong Metropolitan area Fixed Communities in vulnerable areas No info No info Landslide susceptibility maps
EU1 Toscana region Fixed: 25 25 Independent warning zones 4 Regional Civil Protection Headquarters No info

238
EU2 Campania region Fixed: 550 Municipalities 4 Regional Civil Protection Headquarters Meteorological parameters
EU3 Emilia Romagna region Fixed: 8 25 territorial units 4 Regional Civil Protection Headquarters Meteorological parameters
EU4 Piemonte region Fixed: 15 Thyssen polygons 4 Regional Civil Protection Headquarters Meteorological parameters, susceptibility maps
EU5 Umbria region Fixed: 6 No info 4 Regional Civil Protection Headquarters Meteorological parameters
EU6 Italy Fixed: 129 No info 5 DPC None
EU7 Norway Variable Municipalities 4 HF and NVE Landslide susceptibility maps
NA1 San Francisco Bay Region, CA Fixed: 1 San Francisco Bay Region, CA 4 USGS and NWS None
NA2 Oregon Fixed: 1 No info 4 ODOT No info
NA3 Appalachians Fixed: 1 No info No info USGS No info
NA4 Municipality of Seattle Fixed: 1 Municipality 4 USGS Pore water pressure, snowmelt
NA5 Southern California Fixed Burned areas 4 NWS Hazard maps, soil water content, local observations of
rivers and slopes
SA1 Municipality of Rio de Janeiro Fixed: 4 Guanabara, Zona Sul, Sepetiba, 4 Co-Rio Susceptibility and risk maps
Jacarepaguà
SA2 Combeima valley Fixed: 1 No info 4 CREPAD Local observations of rivers and slopes
SA3 113 Communities i.e. Favelas) of Rio de Fixed: 103 Communities 2 Co-Rio Coordinator and Sub-secretary of civil Risk maps
Janeiro defence
Earth-Science Reviews 179 (2018) 228–247
L. Piciullo et al. Earth-Science Reviews 179 (2018) 228–247

Table 6
Response characteristics for the reviewed Te-LEWSs. Legend: DPC, National Department for Civil Protection; NOAA, National Oceanic and Atmospheric Administration; CAP, Common
Alerting Protocol; DMIS, Disaster Management Information System.

CODE Warning methods Information through… Persons informed Public education

AS1 Public statements Internet, Television and radio The public Training, school education, seminars
AS2 Public statements Internet The public None
AS3 Public statements Internet The public None
AS4 Public and internal statements No info Drivers and users None
AS5 Public and internal statements TV, radio, and the Internet Residents and Mayor of local government No info
AS6 Public statements Local Broadcasting System, highway Drivers and residents within or near high risk None
Changeable Message Signs and text potential highway sections
messages
AS7 Internal statements Warning message, broadcasting Decision makers/analysts No info
AS8 Public and internal statements Media, sirens, SMS People living in vulnerable communities and Training, school education, seminars
technical agencies
AS9 Registered users Internet, messages People living in vulnerable communities Social awareness building program, flyers,
website
EU1 Internal statements FAX, webGIS Majors and public institutions None
EU2 Internal statements FAX Majors and public institutions None
EU3 Internal statements FAX Majors and public institutions None
EU4 Internal statements internet, FAX Majors and public institutions None
EU5 Internal statements internet, FAX Majors and public institutions None
EU6 Internal statements Synoptic-scale maps of critical levels DPC None
EU7 Public and internal statements Internet, email, radio and television, app Administrative region, Road and railway None
authorities, public institutions
NA1 Public statements Radio broadcast system, SMS Motorists, people leaving close to steep slopes None
NA2 Public and internal statements NOAA Weather Radio, Law Enforcement Motorists, people leaving close to steep slopes Outreach, technical assistance, and
Data System, television disclosure requirements
NA3 Public and internal statements Internet, CAP, DMIS The public No info
NA4 NWS weather statements Internet, radio, television, CAP, DMIS City officials and public None
NA5 Public and internal statements Internet, NOAA weather, radio, television, Emergency managers and the public No info
AWIPS, CAP, DMIS
SA1 Public and internal statements Television and radio, internet Emergency managers and the public None
SA2 Public and internal statements Emergency protocol Local population Preparedness and social programs
SA3 Public and internal statements Sirens Residents Training for residents, courses, simulations

Zhejiang province, rainfall monitoring and forecast provide data for different landslide susceptibility maps are also used as support data in
evaluating the “effective rainfall” map (Zhang et al., 2011), that is deciding daily warning levels. One map indicates the initiation and
jointly used with the landslide hazard map (Yin et al., 2007) for de- runout areas for debris flows at slope scale (Fischer et al., 2012), the
fining the zones and the level of warning. The Norwegian EWS uses second one zones susceptibility at catchment level (Bell et al., 2014).
municipalities as the minimum warning territorial unit. Then, munici-
palities alerted with the same warning level are grouped together to
define a warning zone for a given warning event (Piciullo et al., 2017a, 5.3.4. Decision on issuing/cancelling a warning level
2017b). In the majority of cases the adopted territorial units coincide For 22 out of 24 Te-LEWSs the warnings are not issued auto-
with administrative boundary (Table 5), mainly for reasons related to matically, especially the highest level of warning, but a consultation
the applicability of emergency plans and to responsibility assignments. among system managers, institutions (geotechincal and/or meteor-
In some cases, the territorial units coincide with areas pre-identified at ological) and state officials takes place. As an example, in the com-
“high risk”, such as: roads and highways in mountainous areas in munity-based alert system of Rio de Janeiro in Brazil (SA3), areas at
Taiwan (AS6), specific tracks of the North-South Highway in Malaysia high risk for rainfall-induced landslides are present in 103 informal
(AS4), favelas located on slopes in: Sri Lanka (AS8), Bangladesh (AS9), communities (i.e. favelas). When the recorded rainfall locally exceeds
Rio de Janeiro (SA3). the alert level, sirens prompting residents to evacuate these areas are
not activated automatically but only after a consultation between the
5.3.3. Additional information to be considered for issuing a warning sub-secretary of the municipal civil defence and the coordinator of the
Elements considered in the reviewed Te-LEWSs to define a warning municipal emergency operating centre, who also consider the updated
level include, besides the rainfall-based thresholds definition, other weather forecasts to take their decision (Calvello et al., 2014). Differ-
additional information such as: hazard or susceptibility maps (AS2, ently from all the other Te-LEWSs reviewed herein, in the system op-
AS3, EU4, EU5, EU7, NA1, NA2, NA4, SA3), soil water content (AS4, erational in Hong Kong (AS1) the issuing of the highest level of warning
AS5, NA5), hydrometric thresholds (EU2) and local observations of is related to the predicted number of landslides. The municipality is
rivers and slopes (AS6, NA5, SA2). As a relevant example, in the LEWS divided into grid cells, with a known number of man-made slopes in
operational in Norway (EU7) daily hydro-meteorological conditions each cell. Different rainfall thresholds are defined for four types of man-
such as rainfall, snowmelt, runoff, soil saturation, groundwater level made slopes as a function of the frequency associated to a given 24-h
and frost depth are obtained from a distributed version of the hydro- rolling rainfall. The rainfall thresholds exceeded are multiplied by the
logical HBV-model (Beldring et al., 2003). These data are displayed in related type and number of slopes in each cell (Wong et al., 2006;
real-time in an open access webpage, http://www.senorge.no (Engeset Cheung et al., 2006). The total number of predicted landslides for the
et al., 2004), which is used by NVE as a decision expert tool to forecast municipality of Hong Kong is obtained by summing the number of
various natural hazards, including floods, snow avalanches and land- expected landslides for each cell. The alert is emitted in the whole area
slides. The warnings are issued once per day with the possibility to of the municipality of Hong Kong, when the predicted number of
update them according to the information gathered by the monitoring landslides exceeds 15.
variables (Devoli et al., 2014; Piciullo et al., 2017a, 2017b). Two

239
L. Piciullo et al. Earth-Science Reviews 179 (2018) 228–247

Table 7
Features of the performance analysis for the Te-LEWSs for which relevant information is available.

CODE Performance evaluation method Through…

AS1 Contingency table Statistical indicators from contingency matrix

EU1 Comparison between predicted and reported landslides Time frame analysis
EU3 Contingency table Statistical indicators from contingency matrix
EU4 Comparison between predicted and reported landslides Time frame analysis
EU7 EDuMaP method Statistical indicators from duration matrix
NA5 Comparison between warning issued and reported landslides Statistical indicators from time frame analysis
SA1 EDuMaP method Statistical indicators from duration matrix
SA3 Comparison between alerts issued and reported landslides Time frame analysis

5.4. Response 5.5. Qualitative and quantitative performance analyses

5.4.1. Communication strategy The system performance can be described as the system capability
The response module usually includes information, communication to forecast the occurrence of one or more landslides in a warning zone,
and emergency strategies, such as education activities and the appli- with a certain lead time. The presence of false and missed alerts reduces
cation of emergency plans. The actors involved and the activities un- the performance of LEWSs because, as stated by Wilson (2004), the
dertaken vary for each Te-LEWS and, within each system, they are a firsts create nuisances and erode credibility but, on the other hand, the
function of the level of warning issued. The method and the media absence of an advisory when landslides do cause death or destruction
employed to warn of a certain level of warning, as well as the public (i.e. missed alert) becomes a dereliction of duty. Among the Te-LEWSs
informed, also differ widely from system to system, mainly in relation reviewed herein, only in 8 cases out of 24 (Table 7) the performance of
to the aims for which the system is designed and employed (Table 6). In the system has been evaluated, essentially by computing the joint fre-
the regional LEWSs adopted in Italy, each civil defence regional de- quency distribution of landslides and warnings in three different ways.
partment manages its own network of pluviometers and provides a Four evaluations (EU1, EU4, NA5, SA3) have been carried out by
continuous service throughout the year, working 24/7 when appro- analysing time frames during which significant high-consequence
priate, in order to release statements to competent public author- landslides occurred in the test area (Keefer et al., 1987; Segoni et al.,
ities—e.g. mayors of municipalities—informing them of the warning 2015; Restrepo, 2008; Tiranti and Rabuffetti, 2010; Calvello et al.,
level issued. The mayors have the duty to decide whether to inform or 2014). Such an analysis allows the evaluation of the performance of the
not the population, considering local factors and expert judgements. adopted warning model with reference to correct and missed alerts in a
Also for the systems SATIS in Taiwan (AS7) and SANF in Italy (EU6), relative small time frame, yet it does not yield any information on the
the information is not directly spread to the public but it is managed possible occurrence of false alerts during periods when high-con-
through internal statements for decision makers, analysts, public au- sequence landslides did not occur in the test area. Based on analyses
thorities and political figures. Differently, the community-based LEWSs performed during two main critical events (i.e., in December 2013 and
operational in Sri Lanka (AS8), Bangladesh (AS9), Colombia (SA2), and March 2013), the performance of the system operational in Toscana
Rio de Janeiro (SA3), as well as the Te-LEWSs designed for highways in region, Italy (EU1), have been considered appropriate by the regional
Taiwan (AS6) and Malaysia (AS4), are structured to directly inform and civil defence. In Piemonte region, Italy (EU4), different thresholds to be
warn the population of the possible occurrence of landslides during employed in the LEWS were compared considering some indicators
heavy rainfall and to invite them to move to safer places. In Rio de derived from a 2 × 2 contingency matrix (hit rate, false alert rate and
Janeiro, for instance, when the evacuation order is issued, the sirens threat score). The only reported analysis refers to a single heavy rainfall
start issuing a pre-recorded warning message which includes a char- event at the end of May 2008 and allows the authors to highlight the
acteristic sound and a vocal message informing about the strong pos- best-performing rainfall thresholds for the warning model. In Southern
sibility of landslide occurrences and prompting the residents to evac- California (NA5) a performance evaluation was carried out for the
uate from areas at high risk and move to safer places. Within each winter season 2005/2006, during which below-normal precipitation
community a number of gathering places (Pontos de Apoio) exist. They and only a handful of storms threatened the area. A total of 39 warnings
are located in areas classified at low risk and, if possible, in communal were issued, 11 of which were verified to have generated debris flows.
places, such as schools, sports centres or churches (d'Orsi, 2012; In the community-based LEWS operating in the favelas of Rio de Ja-
Calvello et al., 2014). neiro, Brazil (SA3), even if the available data-set did not allow a direct
comparison between alarms and landslide occurrences, a significant
number of false alarms were issued in the period 2011–2012.
5.4.2. Public education Two evaluations (EU3, AS1) were based on the computation of
To operate effective Te-LEWSs it is really important that the public statistical indicators from 2 × 2 contingency matrices. In the literature
reacts properly and on time to landslide warnings. To this aim, the this method has been often used to validate rainfall thresholds (Staley
public needs to be educated on the meaning of the warnings and on the et al., 2013; Segoni et al., 2014; Gariano et al., 2015; Piciullo et al.,
appropriate actions to undertake in order to reduce the probability of 2017b). This elementary matrix format displays the interrelation be-
casualties. Training days, school educational programs and seminars tween events of landslides and warnings, both of them expressed in two
are some of the initiatives adopted to increase the effectiveness of the classes. The combinations of landslide occurrence and non-occurrence
warnings issued. Education programs are extremely important in par- with the warning issuing and non-issuing identify: true positives (i.e.
ticular for community-based early warning systems (AS8, AS9, SA2, correct predictions, CP; or correct alerts), true negatives (TN), type I
SA3), where people need to be informed on the procedures to undertake and type II errors. A type I error is the false detection of an effect that is
and on the safe places to reach in case the highest level of warning is not present (i.e. false positive, or false alert, FA), while a type II error is
issued. Only for 6 out of 24 systems, education programs have been the failure to detect an effect that is present (i.e. false negative, or
defined and carried out: Hong Kong (AS1), Sri Lanka (AS8), Bangladesh missed alert, MA). For the performance evaluation of the system in the
(AS9), Oregon (NA2), Colombia (SA2), and Rio de Janeiro (SA3). Emilia Romagna region (EU3), Italy, the main statistical indicators
considered were: the efficiency index, ((CP + TN)/

240
L. Piciullo et al. Earth-Science Reviews 179 (2018) 228–247

(TN + CP + MA + FA)); the sensitivity, (CP/(CP + MA)); the false regional scale. The reliability of a LEWS can be measured in terms of
positive rate, (FA/(FA + TN)); and the false negative rate, (MA/ efficiency and effectiveness. Efficiency, in a general sense, can be de-
(MA + CP)). In these statistics, landslides are considered predicted (CP) fined as the ability of being able to accomplish something at minimum
if they occurred in a day for which the model (see Section 5.2) leads to cost and in the least amount of time. Effectiveness is the capability of
any level of warning. The results showed, for the period 2008–2010 producing a desired result. In the case of LEWSs, the final expected
(Martelloni et al., 2012; Lagomarsino et al., 2013), a high value of the outcome is the reduction of the elements at risk. Erroneous predictions,
efficiency index (95%) and sensitivity (73%), even if 30% of the i.e. false and missed alerts, produce both economic and social costs.
warnings issued were MAs (i.e., false negative rate). In Hong Kong False alerts lead to economic losses, i.e. low efficiency of the system,
(AS1), the evaluation has been carried out for the period 2001–2005 and/or inconveniences for the public, for instance: a road closure or the
(Cheung et al., 2006). In this case, coherently with the adopted warning worthless deployment of civil protection agents and rescue teams. On
model that considers only two warning levels, a threshold number of 15 the other hand, missed alerts can generate a higher number of victims
landslides was used to identify what constitutes a landslide event. The and can evolve in unexpected damage to infrastructures, i.e. low ef-
predicted number of landslides constituting each landslide event for fectiveness of the system.
each warning phase was then compared with the reported number of Many aspects may influence the efficiency of a LEWS. For instance,
landslides during that phase to define the number of: “hits”, i.e. correct a comprehensive and detailed database of landslides is a useful tool for
predictions, “false alarms”, “missed alarms” and “true negatives”. The defining and refining the rainfall thresholds to be employed for issuing
system achieved, for the period 2001–2005, 78% of the critical success warnings. In the reviewed systems only 50% of them (Fig. 6a, Table 3)
index (CP / (CP + MA + FA)) with a perfect detection (100% prob- clearly define the type of landslide for which the thresholds and, con-
ability of detection, CP / (CP + MA) and very few false alarms (only 2 sequently, the system have been built. A constantly updated catalogue
out of 9 warnings issued). of landslides is also essential to operate a continuous validation and
The performance of the systems operational in Norway (EU7) and performance evaluation of the thresholds and of the warnings issued. As
Rio de Janeiro (SA1) was analysed applying the “event, duration ma- an example, the GEO in Hong Kong (AS1) and the GEO-Rio foundation
trix, performance” (EDuMaP) method (Calvello and Piciullo, 2016). (SA1, SA3) manage a pretty detailed catalogue, respectively, of man-
This method employs a n × m matrix that relates a number of classes of made slopes and of natural/artificial landslides. These institutions use
landslide events, n, function of the spatial density of landslide occur- the database to update the rainfall thresholds in order to increase the
rence in the warning area, to the number of warning levels, m, adopted efficiency of the warnings issued. Another important aspect, in pursuing
by the LEWS to be analysed. The EDuMaP method is able to explicitly this objective, is the definition of thresholds considering variables other
consider: the possible occurrence of multiple landslides in the warning than rainfall, such as: pore water pressure, soil water content, snow
zone; the duration of the warnings in relation to the time of occurrence melting. Only in 4 cases out of 24 (Fig. 6b, Table 4) other thresholds
of the landslides; the level of the issued warning in relation to the have been considered together with rainfall thresholds for issuing a
landslide spatial density in the warning zone; the relative importance level of warning. Local observations of slopes and additional informa-
system managers attribute to different types of errors. The performance tion on the triggering causes, landslide susceptibility and/or hazard
of the LEWS in Rio de Janeiro, Brazil, has been analysed for two maps and meteorological parameters, can be useful data to be con-
warning zones with the EDuMaP method for the 3-year period sidered in the decision of issuing/not-issuing a warning in a certain
2010–2012, also conducting a parametric analysis on the method's territorial unit. In 15 Te-LEWSs (Fig. 6c, Table 5) such information is
input parameters (Calvello and Piciullo, 2016). The performance ana- available and considered during operational processes. Finally, the
lysis for the period 2010–2012 highlighted different efficiency index performance evaluation is another important aspect influencing the
values (CP / (CP + MA + FA)) for the warning zones Baia de Guana- efficiency of a LEWS. As highlighted in the previous section this issue is
bara (75%) and Zona Sul (66%). The results for Zona Sul also highlight often overlooked, both by system managers and researchers dealing
a high value (91%) of the false alert rate (FA / (FA + CP)), probably due with LEWSs, and only for 8 systems out of 24 (33%) the performance of
to rainfall thresholds set inadequately low for this warning zone, to- the adopted warning model has been evaluated (Fig. 6d, Table 7).
gether with a low value of correct predictions. In Norway, the perfor- Many aspects of LEWSs may be associated to their effectiveness.
mance analysis has been carried out for 4 regions in the Western part of Radar and hydro/meteorological models to forecast landslide triggering
the Country in the period 2013–2014. In this case, the EDuMaP method variables are useful to increase the lead time of a warning, which in
has been adapted to evaluate the performance of the warnings issued on turn gives to the actors involved (e.g., the public, rescue teams and civil
variable warning zones (Piciullo et al., 2017a, 2017b). In this perfor- protection agencies) more chances to act adequately and promptly to
mance analysis the high values (> 95%) of the efficiency index (CP / reduce the number of victims and the economic losses, i.e. increase the
(CP + MA + FA)) and the odd ratio (CP / (MA + FA)) could be inter- effectiveness of the system. The majority of the Te-LEWSs (71%) take
preted as an excellent result but, in contrast, the high value (43%) of into account weather forecast for issuing a level of warning (Fig. 6e,
the missed and false alerts balance (MA / (MA + FA) highlights some Table 3). Also well-designed warning management processes tend to
issues related to the duration of MAs in relation to the total duration of increase the effectiveness of a territorial LEWS. For instance, timely and
wrong predictions. meaningful statements spread to the public are useful to directly inform
For the prototype system operational in Seattle, Washington (NA4), the elements at risk of a high probability of landslides. Among the
and for the Te-LEWSs of Southern-California some statistical analyses systems herein reviewed, 17 out of 24 (71%) produce public statements
on thresholds exceedance and landslide occurrences have been carried and, in 7 cases, internal statements to institutions, government officials
out respectively by Chleborad et al. (2006) and Staley et al. (2013). and agencies are exclusively provided (Fig. 6f, Table 6). Moreover,
As a final consideration it is important to highlight that the true information campaigns on the risk, the predisposing factors and the
negatives, representing the absence of both warnings or landslides, are triggering causes of landslides help the public to familiarize with the
generally orders of magnitude higher than other variables (i.e., CPs, system. Education programs and training days explaining the correct
FAs, MAs). Thus, considering them in the evaluation of statistical in- behaviour to adopt during a high level of warning may significantly
dicators can lead to a positive overestimation of the performance. reduce the number of victims, particularly in community-based LEWSs.
Only for 6 out of 24 Te-LEWSs (25%) education programs and cam-
5.6. Final considerations and insights paigns have been clearly defined (Fig. 6g, Table 6). For the majority of
cases education is not conceived as a part of the LEWS.
The knowledge acquired in this review allows to highlight some Cooperation among scientists, system managers and public admin-
key-points for the success and the failure of a LEWS operating at istrators is always necessary to design and manage efficient and

241
L. Piciullo et al. Earth-Science Reviews 179 (2018) 228–247

Fig. 6. Important aspects associated to the reliability of a

Te-LEWS: a) type of landslide; b) other thresholds con-
sidered for issuing warnings; c) additional information for
issuing warnings; d) performance evaluation; e) weather
forecast; f) public statements provided; and g) education
campaigns. The graphs show the number of reviewed sys-
tems where these aspects: have been considered (green);
have not been considered (red); or information is not
available (grey). (For interpretation of the references to
colour in this figure legend, the reader is referred to the
web version of this article.)

effective Te-LEWSs (Stähli et al., 2015). Indeed, sharing knowledge and when available, it is not always complete and thorough. In this paper,
data (i.e., weather forecasts, risk maps, rescue teams dislocation, the data on the reviewed systems were gathered from various sources,
number of people at risk, threatened infrastructures) reduces the pos- including scientific articles, technical reports, web pages and personal
sibility of erroneous predictions and contributes to the correct func- contacts. The review highlighted and discussed the main components of
tioning of a LEWS. As an example of the key role played by the team- 24 EWSs for rainfall-induced landslides operating at regional scale.
work of the actors involved in the decision-making process of the San
Francisco bay LEWS, Wilson et al. (2004) state that the end of the 6.1. Setting
system operations in 1995 was essentially caused by staff reduction and
relocation of NWS forecasters far from USGS offices, which complicated Rainfall is the most important triggering factor for landslides; in-
interactions during periods of crisis. deed it is the main variable monitored in Te-LEWSs. The monitoring
All the seven aspects previously described as important to increase instruments commonly used to this purpose are pluviometers (for 22
the reliability, i.e. efficiency and effectiveness, of LEWSs are resumed in systems out of 24). They typically provide near real-time information
Fig. 7. Specifically the figure highlights the “completeness” of the re- on rainfall in different locations of a warning zone, which is then used
viewed systems in relation to the abovementioned aspects. From the for evaluating the exceedance of predefined rainfall thresholds. Past
information and data gathered, none of the systems is considering all rainfall data may also be employed for the identification of the
the aspects. However, the sources analysed clearly describe a significant thresholds, either by means of statistical analyses of the rainfall series
number of aspects for a good number of the reviewed Te-LEWSs. For or in combination with landside data. When a landslide database is
instance 5 out of the 7 aspects are considered for: Hong Kong (AS1); missing or uncertainties on past landslide events are significant, the
Japan (AS5); Norway (EU7); Seattle, Washington, USA (NA4); Southern reliability of the employed rainfall thresholds must be questioned.
California, USA (NA5); and Rio de Janeiro, Brazil (SA1, SA3). With the Ideally, system reliability assessment should be performed con-
information at hand, these systems should be considered more reliable tinuously, although this is seldom the case. In 50% of the Te-LEWSs
than the systems for which only one or two of these aspects are de- reviewed, the low number of inventoried rainfall-induced landslides
scribed (AS2, AS3, EU1, EU6, NA2). induces system managers to define rainfall thresholds by considering all
the landslides together, irrespective of their typology or triggering
6. Conclusions mechanism. When detailed landslide catalogues are available, specific
correlation analyses and more reliable thresholds can be defined. Many
Te-LEWSs are predominantly managed by governmental institu- Te-LEWSs (17 out of 24) employ meteorological modelling to evaluate
tions, thus information is often difficult to find in the literature and, the expected amount of rain for the following hours/days in order to

Fig. 7. The 24 Te-LEWSs reviewed are grouped on the basis of the number of aspects that are important to increase the system reliability considered (according to the analysed sources):
type of landslides; other thresholds for issuing warnings; additional information for issuing warnings; performance evaluation; weather forecast; public statements; and education
campaigns.

242
L. Piciullo et al. Earth-Science Reviews 179 (2018) 228–247

provide forecasts for weather-induced landslides with a short/medium to operate a reliable Te-LEWS. To this purpose the systematic collection
lead time. of landslide and warning data is one of the key aspects for assessing and
reviewing the warning criteria adopted by a system (e.g. rainfall
6.2. Modelling thresholds). Looking at the systems reviewed in this paper, the per-
formance of the warning model has been evaluated only for 33% of the
Even if pluviometers are generally used to monitor rainfall, often systems. For 4 systems (EU1, EU4, NA5, SA3) relatively small time
the location and/or the number of instruments installed are not suffi- frames have been considered for the analyses. In two cases (AS1, EU3)
cient to define reliable rainfall thresholds for warning purposes. For this 2 × 2 contingency tables have been employed. Both these methods
reason monitoring other parameters can be useful. Only 5 Te-LEWSs neglect the meaning associated to different levels of warning. The
(20%) defined and employed other thresholds considering variables multiple simultaneous occurrence of landslides was considered only in
different from rainfall, i.e. pore pressure and soil water content. In 10 one of the six cases, for the system employed in Hong Kong (AS1). A
systems (42%) only one rainfall threshold, in particular intensity- different approach has been used to evaluate the performance of the
duration, has been calibrated and employed for issuing the warnings. systems in Norway (EU7) and Rio de Janeiro, Brazil (SA1): the EDuMaP
Two rainfall thresholds are implemented in 7 systems (29%) in order to method (Calvello and Piciullo, 2016). The method is defined around the
increase the reliability and reduce the possibility to erroneously fore- computation of a n × m table, called “duration matrix,” quantifying the
cast landslides. time associated with the occurrence of a given number of landslides in
relation to the different warning levels adopted by the system. Overall,
6.3. Warning strategies the discussion on the performance evaluation issue indicated that there
is still significant room for improving the efficiency of the warning
The definition of the warning zones may be based on administrative models adopted in Te-LEWSs.
boundaries, on the hydro-geomorphological conditions of the area to be
warned, on the characteristics of the employed monitoring network or 6.6. Final comments
on a combination of all these criteria. The majority of the Te-LEWSs
reviewed employs fixed warning zones. In few cases the zones to be The review shows an increasing trend, in the last decade, in the
warned are varied in space and time by overlapping, in a GIS en- employment of Te-LEWSs in developing countries to reduce the risk to
vironment, the time-dependent information on the thresholds' ex- life associated with rainfall-induced landslides. In these cases, the at-
ceedance with hazard or susceptibility maps. When the latter occurs, tention of institutions, scientists and NGOs is typically focused on
systems managers have the possibility to issue more focused warnings, community-based systems. These systems are often perceived as a cost-
thus optimizing civil defence resources. However, in such cases, site effective landslide risk mitigation measure, also due to the relatively
specific correlations are necessary together with a higher density of simple monitoring network they deploy, and, sometimes, the only
monitoring instruments, possibly also gathering additional information possibility to reduce the risk to life in densely populated areas. An in-
different from rainfall. The number of warning levels employed in the creasing number of such systems is thus expected in the future.
24 reviewed Te-LEWSs varies from a minimum of two (i.e. no alert, Concluding, Te-LEWSs can be considered a useful landslide risk
alert) to a maximum of five. The majority of the systems employs four mitigation measure to employ over widespread areas. To design and
warning levels, considering the first level of warning as a no warning operate these systems it is important to clearly have in mind a con-
status. Different advantages and drawbacks are associated with the ceptual model of all the fundamental tiles that compose a LEWS. An
number of warning levels. For instance, two levels of warning could be omission or an underestimation of any component may indeed reduce
economically and socially demanding (e.g., high number of false alerts); the reliability of the system, eventually leading to its failure or termi-
on the contrary a high number of warning levels can be easily mis- nation. The reliability of a LEWS can be assessed as a function of the
understood. Adopting three levels of warning can be considered a good correctness (i.e. warning model efficiency) and effectiveness of the
compromise solution, with the intermediate level having the meaning warnings issued. This review highlights that there is still a need to in-
of a preparedness stage for system managers and civil defence per- crease the efficiency of such systems. To this aim, improving the
sonnel. landslide database can lead to the definition of more reliable rainfall
thresholds and defining thresholds with triggering variables different
6.4. Response from rainfall can be an important additional tool in the reduction of
erroneous predictions. Also the continuous performance evaluation of
In the majority of the reviewed Te-LEWSs (71%) public statements these systems contributes to increasing the warning model efficiency,
are issued. The tools most commonly used are: TV, radio, social media, for instance by allowing updates of the rainfall thresholds adopted to
SMS. In these cases, warning statements need to be clear and compre- issue the warnings. Evaluation of the public behaviour during emer-
hensible in order to be understood by ordinary citizens that need to gency phases, i.e. when high warning levels are issued, is paramount for
react, appropriately, according to the meaning of each warning level. assessing the system effectiveness. While the evaluation of the perfor-
Indeed, the effectiveness of a Te-LEWS can be ensured only if the public mance of a warning model is addressed by a number of recent studies
is aware of the correct actions to undertake in case of emergency. To and applications from several authors, there is still a lack of literature
this aim education program and campaign can be helpful, however for contributions dealing with education and people response to warnings
only 6 Te-LEWSs (25%) the educational activities deployed are clearly and, thus, with how to quantify the overall effectiveness of a LEWS.
presented. Often public and internal statements are issued together
(46%), but for the 29% of Te-LEWSs only internal statements are pro- References
vided. In the latter case the target of the warnings are, most often: of-
ficials, civil defence agencies, institutions in charge of public safety. The Ahmed, B., Murillo, C.A.O., 2015. Developing Dynamic Web-GIS Based Early Warning
warning recipients are typically informed by means of inter-institu- System for the Communities at landslide Risks in Chittagong Metropolitan Area,
Bangladesh. International Centre for Integrated Mountain Development Kathmandu,
tional or internal reports and they are required to prepare and act fol- Nepal Report. http://www.landslidebd.com/.
lowing predefined operational protocols. Aleotti, P., 2004. A warning system for rainfall-induced shallow failures. Eng. Geol. 73,
247–265. http://dx.doi.org/10.1016/j.enggeo.2004.01.007.
Alfieri, L., Salamon, P., Pappenberger, F., Wetterhall, F., Thielen, J., 2012. Operational
6.5. Performance early warning systems for water-related hazards in Europe. Environ. Sci. Pol. 15 (1),
35–49. http://dx.doi.org/10.1016/j.envsci.2012.01.008.
Periodical performance analyses of the warning model are necessary Badoux, A., Graf, C., Rhyner, J., Kuntner, R., McArdell, B.W., 2009. A debris-flow alarm

243
L. Piciullo et al. Earth-Science Reviews 179 (2018) 228–247

system for the Alpine Illgraben catchment: design and performance. Nat. Hazards 49, Cannon, S.H., Gartner, J.E., Wilson, R.C., Bowers, J.C., Laber, J.L., 2008. Storm rainfall
517–539. http://dx.doi.org/10.1007/s11069-008-9303-x. conditions for floods and debris flows from recently burned areas in southwestern
Barredo, J.I., 2009. Normalised flood losses in Europe: 1970–2006. Nat. Hazards Earth Colorado and southern California. Geomorphology 96, 250–269. http://dx.doi.org/
Syst. Sci. 9, 97–104. http://dx.doi.org/10.5194/nhess-9-97-2009. 10.1016/j.geomorph.2007.03.019.
Basher, R., 2006. Global early warning systems for natural hazards: systematic and Cannon, S., Boldt, E., Laber, J., Kean, J., Staley, D., 2011. Rainfall intensity–duration
people-centred. Phil. Trans. R. Soc. A 364, 2167–2182. http://dx.doi.org/10.1098/ thresholds for postfire debris-flow emergency-response planning. Nat. Hazards 59
rsta.2006.1819. (1), 209–236.
Baum, R.L., 2007. Landslide warning capabilities in the United States—2006. In: Capparelli, G., Tiranti, D., 2010. Application of the MoniFLaIR early warning system for
Proceedings of the 1st North America Landslide Conference, Vail, Colorado, June rainfall-induced landslides in piedmont region (Italy). Landslides. http://dx.doi.org/
3–8, 2007. Association of Engineering Geologists Special Publication 23. pp. 1–13 10.1007/s10346-009-0189-9.
CD-ROM. Capparelli, G., Versace, P., 2011. FLaIR and SUSHI: two mathematical models for early
Baum, R.L., Godt, J.W., 2010. Early warning of rainfall-induced shallow landslides and warning of landslides induced by rainfall. Landslides 8, 67–79. http://dx.doi.org/10.
debris flows in the USA. Landslides 7 (3), 259–272. http://dx.doi.org/10.1007/ 1007/s10346-010-0228-6.
s10346-009-0177-0. Capparelli, G., Iaquinta, P., Iovine, G.G.R., Terranova, O.G., Versace, P., 2012. Modelling
Baum, R.L., Chleborad, A.F., Schuster, R.L., 1998. Landslides triggered by the winter the rainfall-induced mobilization of a large slope movement in northern Calabria.
1996–97 storms in the Puget Lowland, Washington. In: U.S. Geological Survey Open- Nat. Hazards 61, 247–256. http://dx.doi.org/10.1007/s11069-010-9651-1.
File Report 98-239. Cardellini, S., Osimani, P., 2008. Living with landslides: the Ancona case history and early
Baum, R.L., Godt, J.W., Harp, E.L., McKenna, J.W., McMullen, S.R., 2005. Early warning warning system. In: Ber. Geol. B.-A. 82 ISSN 1017-8880 — Landslide Monitoring
of landslides for rail traffic between Seattle and Everett, Washington, USA. In: Hungr, Technologies & Early Warning Systems.
O., Fell, R., Couture, R., Eberhardt, E. (Eds.), Landslide Risk Management. Cascini, L., 2004. The flowslides of May 1998 in the Campania region, Italy: the scientific
Proceedings of the International Conference on Landslide Risk Management, emergency management. Italian Geotech. J. 38 (2), 11–44.
Vancouver, Canada, May 30–June 3, 2005. Balkema, New York, pp. 731–740. Chan, R.K.S., Pang, P.L.R., Pun, W.K., 2003. Recent developments in the Landslip
Bazin, S., 2012. SafeLand deliverable 4.8. Guidelines for Landslide Monitoring and Early Warning System in Hong Kong. In: Proceedings of the Fourteenth Southeast Asian
Warning Systems in Europe — Design and Required Technology. European Project Geotechnical Conference. Southeast Asian Geotechnical Society, Hong Kong, pp.
SafeLand, Grant Agreement No. 226479. Deliverable 4.8, 153 pp., available at: 219–224.
http://www.safeland-fp7 eu, 2012 https://www.ngi.no/eng/Projects/SafeLand (Last Chen, C.-Y., Lin, L.-Y., Yu, F.-C., Lee, C.-S., Tseng, C.-C., Wang, A.-H., Cheung, K.-W.,
access: April 2017). 2007. Improving debris flow monitoring in Taiwan by using high-resolution rainfall
Beldring, S., Engeland, K., Roald, L.A., Sælthun, N.R., Voksø, A., 2003. Estimation of products from QPESUMS. Nat. Hazards 40, 447–461. http://dx.doi.org/10.1007/
parameters in a distributed precipitation-runoff model for Norway. Hydrol. Earth s11069-006-9004-2.
Syst. Sci. 7, 304–316. Cheung, P.Y., Wong, M.C., Yeung, H.Y., 2006. Application of rainstorm Nowcast to real-
Bell, R., Glade, T., Thiebes, B., Jäger, S., Krummel, H., Janik, M., Holland, R., 2009. time warning of landslide hazards in Hong Kong. In: WMO PWS Workshop on
Modelling and web processing of early warning. In: Malet, J.-P., Remaître, A., Warnings of Real-Time Hazards by Using Nowcasting Technology, Sydney, Australia,
Bogaard, T. (Eds.), Landslide Processes, From Geomorphic Mapping to Dynamic 9–13 October 2006.
Modelling. CERG Editions, Strasbourg, France, pp. 249–252. Chleborad, A.F., 2000. A method for anticipating the occurrence of precipitation induced
Bell, R., Cepeda, J., Devoli, G., 2014. Landslide susceptibility modeling at catchment level landslides in Seattle, Washington. In: U.S. Geological Survey Open-File Report 00-
for improvement of the landslide early warning system in Norway. In: Proceedings of 469.
World Landslide Forum 3, 2–6 June 2014, Beijing. Chleborad, A.F., 2003. Preliminary evaluation of a precipitation threshold for antici-
Berti, M., Martina, M.L.V., Franceschini, S., Pignone, A., Simoni, A., Pizziolo, M., 2012. pating the occurrence of landslides in the Seattle, Washington, area. In: U.S.
Probabilistic rainfall thresholds for landslide occurrence using a Bayesian approach. Geological Survey Open-File Report 03-463.
J. Geophys. Res. 117, F04006. http://dx.doi.org/10.1029/2012JF002367. Chleborad, A.F., Baum, R.L., Godt, J.W., 2006. Rainfall thresholds for forecasting land-
Blikra, L.H., 2008. The Aknes rockslide; monitoring, threshold values and early warning. slides in the Seattle, Washington, area—exceedance and probability. In: U.S.
In: Landslides and Engineered Slopes. From the Past to the Future Proceedings of the Geological Survey Open-File Report 2006-1064.
10th International Symposium on Landslides and Engineered Slopes, pp. 1089–1094. Chleborad, A.F., Baum, R.L., Godt, J.W., 2008. A prototype system for forecasting land-
Boje, S., Colleuille, H., Cepeda, J., Devoli, G., 2014. Landslide thresholds at regional scale slides in the Seattle, Washington area. In: Baum, R.L., Godt, J.W., Highland, L.M.
for the early warning system in Norway. In: Proceedings of World Landslide Forum 3. (Eds.), Engineering Geology and Landslides of the Seattle, Washington Area:
June 2–6, 2014, Beijing. Geological Society of America Reviews in Engineering Geology. XX. Geological
Brand, E.W., Premchitt, J., Phillipson, H.B., 1984. Relationship between rainfall and Society of America, Boulder, pp. 103–120. http://dx.doi.org/10.1130/2008.
landslides in Hong Kong. Proceedings of the Fourth International Symposium on 4020(06).
Landslides, Toronto 1, 377–384. Slope monitoring and early warning systems: application to coastal landslides on the
Brocca, L., Melone, F., Moramarco, T., 2008. On the estimation of antecedent wetness south and east coast of England, UK. In: Clark, A.R., Moore, R., Palmer, I.S., Senneset
condition in rainfall-runoff modeling. Hydrol. Process. 22, 629–642. (Eds.), Landslides. Balkema, Rotterdam 1996. ISBN 90-5410-818-5.
Brunetti, M.T., Peruccacci, S., Rossi, M., Luciani, S., Valigi, D., Guzzetti, F., 2010. Rainfall Colleuille, H., Haugen, L.E., Beldring, S., 2010. A forecast analysis tool for extreme hy-
thresholds for the possible occurrence of landslides in Italy. Nat. Hazards Earth Syst. drological conditions in Norway. In: Poster Presented in Sixth World FRIEND 2010.
Sci. 10, 447–458. http://dx.doi.org/10.1007/s10346-015-0630-1. Flow Regime and International Experiment and Network Data. Fez, Morocco.
Calvello, M., 2017. Early warning strategies to cope with landslide risk. In: Rivista CRED, 2011. EM-DAT. In: The OFDA/CRED International Disaster Database. Université
italiana di geotecnica 2/2017. Patron editore, http://dx.doi.org/10.19199/2017.2. Catholique de Louvain, Brussels, Belgium. www.emdat.be.
0557-1405.063. Cruden, D.M., Varnes, D.J., 1996. Landslide types and processes. In: Turner, A.K.,
Calvello, M., Piciullo, L., 2016. Assessing the performance of regional landslide early Schuster, R.L. (Eds.), Landslides Investigation and Mitigation. Transportation
warning models: the EDuMaP method. Nat. Hazards Earth Syst. Sci. 16, 103–122. Research Board, US National Research Council. Special Report 247, Washington, DC,
http://dx.doi.org/10.5194/nhess-16-103-2016. Chapter 3, pp. 36–75.
Calvello, M., d'Orsi, R.N., Piciullo, L., Paes, N., Magalhaes, M., Coelho, R., Lacerda, W.A., Devoli, G., Ingeborg, K., Monica, S., Nils-Kristian, O., Ragnar, E., Erik, J., Hervé, C., 2014.
2014. The community-based alert and alarm system for rainfall induced landslides in Landslide early warning system and web tools for real-time scenarios and for dis-
Rio de Janeiro, Brazil. In: Engineering Geology for Society and Territory “Landslide tribution of warning messages in Norway. In: Engineering Geology for Society and
Processes”, Proc. XII Int. IAEG Congress, Torino, Italy. 2. pp. 653–657. http://dx.doi. Territory “Landslide Processes”, Proc. XII International IAEG Congress, Torino, Italy,
org/10.1007/978-3-319-09057-3_109. pp. 625–629. http://dx.doi.org/10.1007/978-3-319-09057-3_104.
Calvello, M., d'Orsi, R.N., Piciullo, L., Paes, N., Magalhaes, M., Lacerda, W.A., 2015. The DGR, 2007. n. 2312 del 27/12/2007, Deliberazione di Giunta Regionale dell'Umbria:
Rio de Janeiro early warning system for rainfall-induced landslides: analysis of per- Disposizioni in attuazione dell'art, 3 bis della Legge 100/2012 e della Direttiva del
formance for the years 2010–2013. Int. J. Disaster Risk Reduct. 12, 3–15. http://dx. Presidente del Consiglio dei Ministri del 27.02.2004 – Sistema di Allertamento
doi.org/10.1016/j.ijdrr.2014.10.005. Regionale e Centro Funzionale Regionale, Umbria, Italy.
Campbell, R.H., 1975. Soil slopes, debris flows, and rainstorms in the Santa Monica Di Biagio, E., Kjekstad, O., 2007. Early warning, instrumentation and monitoring land-
Mountains and vicinity, southern California. In: U.S. Geol. Surv. Prof. Pap. 851. slides. In: 2nd Regional Training Course, RECLAIM II, 29th January–3rd February
Cannon, S.H., 1988. Regional rainfall-threshold conditions for abundant debris-flow ac- 2007.
tivity. In: Ellen, S.D., Wieczorek, G.F. (Eds.), Landslides, Floods, and Marine Effects of DOGAMI, 2005. Oregon's Debris-Flow Warning System and How It Works. Oregon
the Storm of January 3–5, 1982, in the San Francisco Bay Region, California. U.S. Department of Geology and Mineral Industries. http://www.oregongeology.com/
Geological Survey Professional Paper 1434 USGS, Washington, DC. sub/Landslide/debrisflow.htm Accessed 8 Apr 2009.
Cannon, S.H., Ellen, S.D., 1985a. Rainfall conditions for abundant debris avalanches, San DPCM, 2005. D27/02/2004, Direttiva Presidente Consiglio dei Ministri: Indirizzi oper-
Francisco Bay region, California. Calif. Geol. 38 (12), 267–272. ativi per la gestione operativa e funzionale del sistema di allertamento nazionale,
Cannon, S.H., Ellen, S.D., 1985b. Rainfall conditions for abundant debris avalanches, San statale e regionale per il rischio idrogeologico ed idraulico ai fini di protezione civile,
Francisco Bay region, California. Calif. Geol. 38 (12), 267–272. in: Gazzetta Ufficiale n. 59 del 11/03/2004, Roma, Italy, 2004. DPGR: n. 299 del 30/
Cannon, S.H., Ellen, S.D., 1988. Rainfall that resulted in abundant debris-flow activity 06/2005, Decreto del Presidente della Giunta Regionale della Campania: Il Sistema di
during the storm. In: Ellen, S.D., Wieczorek, G.F. (Eds.), Landslides, Floods, and Allertamento Regionale per il rischio Idrogeologico e Idraulico ai fini di protezione
Marine Effects of the Storm of January 3–5, 1982, in the San Francisco Bay Region, civile, Bollettino ufficiale della regione Campania, n. speciale 01/08/2005,
California. US Geological Survey Professional Paper 1434. pp. 27–34. Campania, Italy. (In Italian).
Cannon, S.H., Gartner, J.E., 2005. Wildfire-related debris flow from a hazards perspec- Easterling, D.R., Meehl, G.A., Parmesan, C., Changnon, S.A., Karl, T.R., Mearns, L.O.,
tive: chapter 15. In: Jakob, M., Hungr, O. (Eds.), Debris Flow Hazards and Related 2000. Climate extremes: observations, modeling, and impacts. Science 289,
Phenomena: Praxis. Springer, Berlin, Heidelberg, pp. 363–385. 2068–2074.

244
L. Piciullo et al. Earth-Science Reviews 179 (2018) 228–247

Ellen, S.D., Wieczorek, G.F., 1988. Landslides, flood and marine effects of the storm of (Eds.), Landslide Processes of the Eastern United States and Puerto Rico. Geological
January 3–5, 1982, in the San Francisco Bay region, California. In: USGS Professional Society of America Special Paper 236. Geological Society of America, Boulder, pp.
Paper No. 1434. United States Geological Survey, Washington D.C. 29–55.
Engeset, R., Tveito, O.E., Mengistu, Z., Udnæs, H.-C., Isaksen, K., Førland, E.J., 2004. Karnawati, D., Fathani, T.F., Wilopo, W., Andayani, B., 2013a. Hybrid socio-technical
Snow map system for Norway. In: XXIII Nordic Hydrological Conference, Tallin. NHP approach for landslide risk reduction in Indonesia. In: Wang, F., Miyajima, M., Li, T.,
Report No. 48, Tartu. Shan, W., Fathani, T.F. (Eds.), Progress of Geo-Disaster Mitigation Technology in
European Environment Agency, 2010. Mapping the impacts of natural hazards and Asia. Springer-Verlag, Berlin Heidelberg, pp. 157–169.
technological accidents in Europe — an overview of the last decade. In: EEA Karnawati, D., Ma'arif, S., Fathani, T. F., Wilopo, W., 2013b. Development of socio-
Technical Report No 13/2010, ISBN: 978-92-9213-168-5. technical approach for landslide mitigation and risk reduction program in Indonesia,
Fathani, T.F., Karnawati, D., Wilopo, W., 2016. An integrated methodology to develop a ASEAN Eng. J. Part C, 2, 22–49, 2013b.
standard for landslide early warning systems. Nat. Hazards Earth Syst. Sci. 16 (9), Keefer, D.K., Wilson, R.C., Mark, R.K., Brabb, E.E., Brown III, W.M., Ellen, S.D., Harp,
2123–2135. http://dx.doi.org/10.5194/nhess-16-2123-2016. E.L., Wieczorek, G.F., Alger, C.S., Zatkin, R.S., 1987. Real-time landslide warning
Fell, R., Ho, K.K.S., Lacasse, S., Leroi, E., 2005. A framework for landslide risk assessment during heavy rainfall. Science 238 (13), 921–925.
and management. In: Hungr, O., Fell, R., Couture, R., Eberhardt, E. (Eds.), Landslide Kuramoto, K., Tetsuga, H., Higashi, N., Arakawa, M., Nakayama, H., Furukawa, K., 2001.
Risk Management. Taylor and Francis, London, pp. 3–26. A study on a method for determining non-linear critical line of slope failures during
Fischer, L., Rubensdotter, L., Stalsberg, K., Melchiorre, C., Horton, P., Jaboyedoff, M., heavy rainfall based on RBF network. In: Doboku Gakkai Ronbunshu, No. 672, VI-50.
2012. Debris flow modeling for susceptibility at regional to national scale in Norway. Japan Society of Civil Engineers, pp. 117–132 (in Japanese with English Abstract).
In: Eberhardt (Ed.), Landslides and Engineered Slopes: Protecting Society Through Kuramoto, K., Noro, T., Osanai, N., Kobayashi, M., Okada, K., 2005. A study on rainfall
Improved Understanding, pp. 723–729. indexes for giving early warning information for sediment-related disasters. In:
Flentje, P.N., Chowdhury, R.N., Tobin, P., Brizga, V., 2005. Towards real-time landslide Proceedings of Annual Research Meeting in 2005. Japan Society of Erosion Control
risk management in an urban area. In: Hungr, O., Fell, R., Couture, R., Eberhardt, E. Engineering, pp. 186–187 (in Japanese).
(Eds.), International Conference on Landslide Risk Management. Taylor & Francis Landslide risk assessment and mitigation strategy. In: Lacasse, S., Nadim, F., Sassa, K.,
Ltd., Vancouver, Canada, pp. 741–751. Canuti, P. (Eds.), Landslides—Disaster Risk Reduction. Springer-Verlag, Berlin
Froese, C.R., Moreno, F., 2014. Structure and components for the emergency response and Heidelberg, pp. 31–61 (2009).
warning system on Turtle Mountain, Alberta, Canada. Nat. Hazards 70, 1689–1712. Lagomarsino, D., Segoni, S., Fanti, R., Catani, F., 2013. Updating and tuning a regional
http://dx.doi.org/10.1007/s11069-011-9714-y. scale landslide early warning system. Landslides 10, 91–97. http://dx.doi.org/10.
Galliani, G., Pomi, L., Zinoni, F., Casagli, N., 2001. Analisi meteoclimatologica e soglie di 1007/s10346-012-0376-y.
innesco delle frane nella regione Emilia-Romagna negli anni 1994–1996. Quad. Geol. Lagomarsino, D., Segoni, S., Rosi, A., Rossi, G., Battistini, A., Catani, F., Casagli, N., 2015.
Appl. 8, 75–91. Quantitative comparison between two different methodologies to define rainfall
Gariano, S.L., Guzzetti, F., 2016. Landslides in a changing climate. Earth-Sci. Rev. 162 thresholds for landslide forecasting. Nat. Hazards Earth Syst. Sci. 15 (2413–2423),
(2016), 227–252. http://dx.doi.org/10.1016/j.earscirev.2016.08.011. 2015. http://dx.doi.org/10.5194/nhess-15-2413-2015.
Gariano, S.L., Brunetti, M.T., Iovine, G., Melillo, M., Peruccacci, S., Terranova, O.G., Laprade, W.T., Kirkland, T.E., Nashem, W.D., Robertson, C.A., 2000. Seattle landslide
Vennari, C., Guzzetti, F., 2015. Calibration and validation of rainfall thresholds for study. In: Shannon & Wilson, Inc. Internal Report W-7992-01, . http://www.seattle.
shallow landslide forecasting in Sicily, southern Italy. Geomorphology 228, 653–665. gov/dpd/landslide/study/ Accessed 8 Apr 2009.
http://dx.doi.org/10.1007/s11069-014-1129-0. Larsen, M.C., Simon, A., 1993. A rainfall intensity–duration threshold relation for land-
Gerstel, W.J., Brunengo, M.J., Lingley, W.S., Logan, R.L., Shipman, H., Walsh, T.J., 1997. slides in a humid-tropical environment, Puerto Rico. Geogr. Ann. 75A (12), 13–23.
Puget Sound bluffs: the where, why, and when of landslides following the holiday Lloyd, D., Othman, M.A., Anderson, M.G., 2001. Predicting landslides: assessment of an
1996/97 storms. Wash. Geol. 25, 17–31. automated rainfall based landslide warning system. In: 14th South East Asian
Giannecchini, R., Galanti, Y., D'Amato Avanzi, G., 2012. Critical rainfall thresholds for Geotechnical Conference, Hong Kong.
triggering shallow landslides in the Serchio River Valley (Tuscany, Italy). Nat. Loew, S., Gschwind, S., Gischig, V., Keller-Signer, A., Valenti, G., 2016. Monitoring and
Hazards Earth Syst. Sci. 12, 829–842. http://dx.doi.org/10.5194/nhess-12-829- early warning of the 2012 Preonzo catastrophic rockslope failure. Landslides 14 (1),
2012. 141–154. http://dx.doi.org/10.1007/s10346-016-0701-y.
Glade, T., Nadim, F., Stötter, H., Guzzetti, F., 2008. Early warning systems and multi- Lollino, G., Arattano, M., Cuccureddu, M., 2002. The use of the automatic inclinometric
disciplinary approaches in natural hazards and risk assessments. In: Special Volume system for landslide early warning: the case of Cabella Ligure (North-Western Italy).
in GEORISK 2008, 2(4), 272. Phys. Chem. Earth 27 (36), 1545–1550. http://dx.doi.org/10.1016/S1474-7065(02)
Godt, J.W., 2004. Observed and Modeled Rainfall Conditions for Shallow Landsliding in 00175-4.
the Seattle, Washington, Area. PhD thesis. University of Colorado, Boulder. Lumb, P., 1975. Slope failures in Hong Kong. Q. J. Eng. Geol. 8, 21–65.
Godt, J.W., Baum, R.L., Chleborad, A.F., 2006. Rainfall characteristics for shallow land- Martelloni, G., Segoni, S., Fanti, R., Catani, F., 2012. Rainfall thresholds for the fore-
sliding in Seattle, Washington, USA. Earth Surf. Process. Landf. 31, 97–110. casting of landslide occurrence at regional scale. Landslides 9 (4), 485–495. http://
Guzzetti, F., Peruccacci, S., Rossi, M., Stark, C.P., 2007. Rainfall thresholds for the in- dx.doi.org/10.1007/s10346-011-0308-2.
itiation of landslides in central and southern Europe. Meteorog. Atmos. Phys. 98 (3), Massey, J.B., Mak, S.H., Yim, K.P., 2001. Community based approach to landslide risk
239–267. http://dx.doi.org/10.1007/s00703-007-0262-7. reduction. In: Proceedings of the Fourteenth Southeast Asian Geotechnical
Hall, P., 2007. Early warning systems: reframing the discussion. Austral. J. Emerg. Conference. Hong Kong, China, pp. 141–147.
Manag. 2 (22), 1324–1540. Michoud, C., Bazin, S., Blikra, L.H., Derron, M.-H., Jaboyedoff, M., 2013. Experiences
Huang, M.-L., Hong, J.-H., 2010. A geospatial service approach towards the development from site-specific landslide early warning systems. Nat. Hazards Earth Syst. Sci. 13,
of a debris flow early-warning systems. Adv. Inf. Sci. Serv. Sci. 2 (2). http://dx.doi. 2659–2673. http://dx.doi.org/10.5194/nhess-13-2659-2013.
org/10.4156/aiss.vol2.issue2.13. June 2010. Mills, K.A., 2002. Oregon's debris flow warning system. Geol. Soc. Am. Abstr. Programs
Huat, L.T., Ali, F., Osman, A.R., Rahman, N.A., 2012. Web Based Real Time Monitoring 34 (5), 25.
System Along North-South Expressway, Malaysia. 17. pp. 623–632. Montgomery, D.R., Schmidt, K.M., Greenberg, H.M., Dietrich, W.E., 2000. Forest clearing
Huggel, C., 2012. Early warning systems: the “last mile” of adaptation. EOS Trans. Am. and regional landsliding. Geology 28 (4), 311–314.
Geophys. Union 93 (22, 29), 209–216 May 2012. Morss, R., Wilhelmi, O., Meehl, G., Dilling, L., 2011. Improving societal outcomes of
Huggel, C., Ramirez, J.M., Calvache, M., González, H., Gutierrez, C., Krebs, R., 2008. A extreme weather in a changing climate: an integrated perspective. Annu. Rev.
landslide early warning system within an integral risk management strategy for the Environ. Resour. 36, 1–25.
Combeima-Tolima Region, Colombia. In: Proceedings of the International Disaster NOAA-USGS Debris-Flow Task Force, 2005. NOAA-USGS debris-flow warning sys-
and Risk Conference (IDRC), Davos, Switzerland, 25 August 2008, pp. 273–276. tem—final report. In: U.S. Geological Survey Circular 1283.
Huggel, C., Khabarov, N., Obersteiner, M., Ramirez, J.M., 2010. Implementation and Okada, K., 2005. Soil water index. Sokkoujihou No. 69 5 (5), 67–100 (in Japanese).
integrated numerical modelling of a landslide early warning system: a pilot study in Olivieri, W., Lovisolo, M., Crosta, G., 2012. Continuous geotechnical monitoring for alert
Colombia. Nat. Hazards 52, 501–518. http://dx.doi.org/10.1007/s11069-009- thresholds and hazard management. In: Landslides and Engineered Slopes. CRC Press
9393-0. Taylor and Francis Group, pp. 1929–1934.
Information Note, 2009. Landslip Warning System. Prepared by. Standards and Testing d'Orsi, R.N., 2012. Landslide risk reduction measures by the Rio de Janeiro city gov-
Division Geotechnical Engineering Office Civil Engineering and Development ernment. Improving the assessment of disaster risks to strengthen financial resilience.
Department March 2009. In: Special Joint G20 Publication, Government of Mexico and World Bank, pp. 77–91.
Intrieri, E., Gigli, G., Mugnai, F., Fanti, R., Casagli, N., 2012. Design and implementation open publication. https://www.gfdrr.org/G20DRM.
of a landslide early warning system. Eng. Geol. 147–148. http://dx.doi.org/10.1016/ d'Orsi, R.N., D'Avila, C., Ortigao, J.A.R., Moraes, L., Santos, M.D., 1997. Rio-Watch: the
j.enggeo.2012.07.017. Rio de Janeiro landslide watch. In: Proceedings of the 2nd PSL Pan-Am Symposium
Intrieri, E., Gigli, G., Casagli, N., Nadim, F., 2013. Brief communication “Landslide Early on Landslides, Rio de Janeiro, Brazil, pp. 21–30.
Warning System: toolbox and general concepts”. Nat. Hazards Earth Syst. Sci. 13, d'Orsi, R.N., Feijó, R.L., Paes, N.M., 2000. Rainfall and mass movements in Rio de Janeiro.
85–90. http://dx.doi.org/10.5194/nhess-13-85-2013. In: Proceedings of 31st International Geological Congress, Rio de Janeiro, Brasil.
IPCC — Intergovernmental Panel on Climate Change, 2014. Climate Change 2014: d'Orsi, R.N., Feijó, R.L., Paes, N.M., 2004. 2,500 Operational days of Alerta Rio system:
Synthesis Report. Contribution of Working Groups I, II and III to the Fifth Assessment history and technical improvements of Rio de Janeiro warning system for severe
Report of the Intergovernmental Panel on Climate Change, Geneva, Switzerland. pp. weather. Proc. IX Int. Symp. Landslides, Rio de Janeiro, Brazil, pp. 831–836.
151. d'Orsi, R.N., Magalhaes, M., Coelho, R.d.S., Costa, L.d.C., Paes, N., 2013. Análise dos
Jakob, M., Owen, T., Simpson, T., 2012. A regional real-time debris-flow warning system Escorregamentos em Encostas do Município do Rio de Janeiro Deflagrados por
for the District of North Vancouver, Canada. Landslides 9, 165–178. http://dx.doi. Chuvas Intensas no Ano de 2010. COBRAE (in Portuguese).
org/10.1007/s10346-011-0282-8. Osanai, N., Shimizu, T., Kuramoto, K., Kojima, S., Noro, T., 2010. Japanese early-warning
Jibson, R.W., 1989. Debris flows in southern Puerto Rico. In: Schultz, A.P., Jibson, R.W. for debris flows and slope failures using rainfall indices with radial basis function

245
L. Piciullo et al. Earth-Science Reviews 179 (2018) 228–247

network. Landslides 7, 325–338. http://dx.doi.org/10.1007/s10346-010-0229-5. Model Dev. 8, 1955–1978. http://dx.doi.org/10.5194/gmd-8-1955-2015.

Pang, P.L.R., Pun, W.K., Yu, Y.F., 2000. Estimation of failure frequency of soil cut slopes Thiebes, B., Glade, T., Bell, R., 2012. Landslide analysis and integrative early warning-
using rainfall and slope information. In: Proceedings of GeoEng2000 — International local and regional case studies. In: Eberhardt, E. (Ed.), Landslides and Engineered
Conference on Geotechnical and Geological Engineering, Melbourne, Australia, Slopes: Protecting Society Through Improved Understanding. Taylor & Francis
19–24 November 2000. Group, London, pp. 1915–1921.
Piciullo, L., Dahl, M.-P., Devoli, G., Colleuille, H., Calvello, M., 2017a. Adaptation of the Thiebes, B., Bell, R., Glade, T., Jäger, S., Mayer, J., Anderson, M., Holcombe, L., 2013.
EDuMaP method for the performance evaluation of the alerts issued on variable Integration of a limit-equilibrium model into a landslide early warning system.
warning zones. Nat. Hazards Earth Syst. Sci. 17 (6), 817–831. http://dx.doi.org/10. Landslides 11 (859–875), 30. http://dx.doi.org/10.1007/s10346-013-0416-2.
5194/nhess-17-817-2017. Tiranti, D., Rabuffetti, D., 2010. Estimation of rainfall thresholds triggering shallow
Piciullo, L., Gariano, S.L., Melillo, M., Brunetti, M.T., Peruccacci, S., Guzzetti, F., Calvello, landslides for an operational warning system implementation. Landslides 7, 471–481.
M., 2017b. Definition and performance of a threshold-based regional early warning http://dx.doi.org/10.1007/s10346-010-0198-8.
model for rainfall-induced landslides. Landslides 14 (3), 995–1008. http://dx.doi. Tiranti, D., Cremonini, R., Marco, F., Gaeta, A.R., Barbero, S., 2014. The DEFENSE (Debris
org/10.1007/s10346-016-0750-2. Flows triggered by storms-Nowcasting SystEm): an early warning system for tor-
Ponziani, F., Berni, N., Stelluti, M., Zauri, R., Pandolfo, C., Brocca, L., Moramarco, T., rential processes by radar storm tracking using a Geographic Information System
Salciarini, D., Tamagnini, C., 2013. Landwarn: an operative early warning system for (GIS). Comput. Geosci. 70, 96–109. http://dx.doi.org/10.1016/j.cageo.2014.05.004.
landslides forecasting based on rainfall thresholds and soil moisture. In: Margottini, Tubbs, D.W., 1974. Landslides in Seattle. In: Washington Division of Geology and Earth
C. (Ed.), Landslide Science and Practice. 2 Springer-Verlag, Berlin Heidelberg. Resources Information Circular 52.
http://dx.doi.org/10.1007/978-3-642-31445-2_82. 2013. United Nations Inter-Agency Secretariat of the International Strategy for Disaster
Pun, W.K., Wong, A.K.W., Pang, P.L.R., 2003. A review of the relationship between Reduction (UN ISDR), 2005. Hyogo Framework for Action 2005–2015: building the
rainfall and landslides in Hong Kong. In: Proceedings of the Asian Technical resilience of nations and communities to disasters. In: Extract from the Final Report of
Committee (ATC 3) Workshop on Rain-induced Landslides, Hong Kong. 3. pp. the World Conference on Disaster Reduction (A/CONF.206/6), pp. 22.
211–217. United Nations Inter-Agency Secretariat of the International Strategy for Disaster
Reid, M.E., Baum, R.L., LaHusen, R.G., Ellis, W.L., 2008. Capturing landslide dynamics Reduction (UN ISDR), 2006. Global Survey of Early Warning Systems: An Assessment
and hydrologic triggers using near-real-time monitoring. In: Chen, Z., Zhang, J., Li, of Capacities, Gaps and Opportunities Towards Building a Comprehensive Global
Z., Wu, F., Ho, K. (Eds.), Landslides and Engineered Slopes from the Past to the Early Warning System for All Natural Hazards. available at: http://www.unisdr.org/
Future. Proceedings of the Tenth International Symposium on Landslides and 2006/ppew/info-resources/ewc3/Global-Survey-of-Early-Warning-Systems.pdf (last
Engineered Slopes, 30 June–4 July 2008, Xi'an, China. 1. Taylor and Francis, access: November 2014).
London, pp. 179–191. United Nations International Strategy for Disaster Reduction (UN ISDR), 2009.
Restrepo, P., 2008. In: Jorgensen, D.D.P., Cannon, S.H., Costa, J., Laber, J., Major, J., Terminology on Disaster Risk Reduction. UNISDR, Geneva, Switzerland.
Martner, B., Purpura, J., Werner, K. (Eds.), Joint NOAA/NWS/USGS Prototype Debris UN, 2015. Transforming our world: the 2030 Agenda for Sustainable Development.
Flow Warning System for Recently Burned Areas in Southern California. American United Nations, A/RES/70/1. pp. 35. http://sustainabledevelopment.un.org/
Meteorological Society, pp. 1845–1851 December 2008. post2015/transformingourworld (last accessed: 30/01/2017).
Restrepo, P., Cannon, S.H., Laber, J., Jorgensen, D.P., Werner, K., 2009. NOAA/USGS USGS, 2005. A NOAA–USGS demonstration flash-flood and debris-flow early-warning
demonstration flash-flood and debris-flow early-warning system for recently burned system. In: U.S. Geological Survey Fact Sheet 2005-3104.
areas in Southern California, USA. In: 7th International Conference on Vennari, C., Gariano, S.L., Antronico, L., Brunetti, M.T., Iovine, G., Peruccacci, S.,
Geomorphology. Australia, July 2009. Terranova, O., Guzzetti, F., 2014. Rainfall thresholds for shallow landslide occur-
Rogers, D.P., Tsirkunov, V., 2010. Costs and Benefits of Early Warning Systems. Global rence in Calabria, southern Italy. Nat. Hazards Earth Syst. Sci. 14, 317–330. http://
Assessment Report on Disaster Risk Reduction. ISDR and World Bank 17 pp. http:// dx.doi.org/10.5194/nhess-14-317-2014.
www.preventionweb.net/english/hyogo/gar/2011/en/bgdocs/Rogers_&_Tsirkunov_ Villagrán de León, J.C., Pruessner, I., Breedlove, H., 2013. Alert and Warning Frameworks
2011.pdf. in the Context of Early Warning Systems, a Comparative Review, Intersections No.
Rosi, A., Segoni, S., Catani, F., Casagli, N., 2012. Statistical and environmental analyses 12. United Nations University Institute for Environment and Human Security, Bonn.
for the definition of a regional rainfall threshold system for landslide triggering in WDR—The World Development Report, 2014. Risk and Opportunity—Managing Risk for
Tuscany (Italy). J. Geogr. Sci. 22 (4), 617–629. Development. ISBN 978-0-8213-9903-3.
Rossi, M., Peruccacci, S., Brunetti, M.T., Marchesini, I., Luciani, S., Ardizzone, F., Wieczorek, G.F., 1987. Effect of rainfall intensity and duration on debris flows on central
Balducci, S.V., Bianchi, C., Cardinali, M., Fiorucci, F., Mondini, A.C., Reichenbach, P., Santa Cruz Mountains, California. In: Costa, J.E., Wieczorek, G.F. (Eds.), Debris-
Salvati, P., Santangelo, M., Bartolini, D., Gariano, S.L., Palladino, M., Vessia, G., Flows/Avalanches: Processes, Recognition, and Mitigation. Geological Society of
Viero, A., Antronico, L., Borselli, L., Deganutti, A.M., Iovine, G., Luino, F., Parise, M., America, Reviews in Engineering Geology 7. pp. 23–104.
Polemio, M., Guzzetti, F., 2012. SANF: national warning system for rainfall-induced Wieczorek, G.F., Morgan, B.A., 2008. Debris-flow hazards within the Appalachian
landslides in Italy. In: Eberhardt, E., Froese, C., Turner, K., Leroueil, S. (Eds.), Mountains of the Eastern United States. In: Internal Report U.S. Department of the
Landslides and Engineered Slopes: Protecting Society Through Improved Interior U.S. Geological Survey. Fact Sheet 2008-3070 August 2008.
Understanding. Taylor & Francis, London, pp. 1895–1899. Wieczorek, G.F., Morgan, B.A., Campbell, R.H., 2000. Debris-flow hazards in the blue
Segoni, S., Rossi, G., Rosi, A., Catani, F., 2014. Landslides triggered by rainfall: a semi- ridge of central Virginia. Environ. Eng. Geosci. 6 (1), 3–23. http://dx.doi.org/10.
automated procedure to define consistent intensity-duration thresholds. Comput. 2113/gseegeosci.6.1.3.
Geosci. 63, 123–131. http://dx.doi.org/10.1016/j.cageo.2013.10.009. Wiley, T.J., 2000. Relationship between rainfall and debris flows in western Oregon. Or.
Segoni, S., Battistini, A., Rossi, G., Rosi, A., Lagomarsino, D., Catani, F., Moretti, S., Geol. 62 (2), 27–43.
Casagli, N., 2015. Technical note: an operational landslide early warning system at Wilson, R.C., 2004. The Rise and Fall of a Debris-flow Warning System for the San
regional scale based on space–time-variable rainfall thresholds. Nat. Hazards Earth Francisco Bay Region, California. Landslide Hazard and Risk Edited by Thomas
Syst. Sci. 15, 853–861. http://dx.doi.org/10.5194/nhess-15-853-2015. Glade, Malcolm Anderson and Michael J. Crozier, John Wiley & Sons, Ltd. ISBN: 0-
Sirangelo, B., Versace, P., 1996. A real time forecasting model for landslides triggered by 471-48663-9.
rainfall. Meccanica 31, 73–85. http://dx.doi.org/10.1007/BF00444156. Wilson, R.C., Wieczorek, G.F., 1995. Rainfall thresholds for initiation of debris flows at La
Sirangelo, B., Capparelli, G., Versace, P., 2003. Forewarning model for landslides trig- Honda, California. Environ. Eng. Geosci. 1 (1), 11–27.
gered by rainfall based on the analysis of historical data file. In: Hydrology of the Wilson, R.C., Mark, R.K., Barbato, G., 1993. Operation of a real-time warning system for
Mediterranean and Semiarid Regions. Proceedings of an international symposium debris flows in the San Francisco Bay area, California. In: Shen, H.W., Su, S.T., Wen,
held al Montpellier, April 2003. Publ. No. 278. F. (Eds.), Hydraulic Engineering '93: Proceedings of the 1993 Conference, Hydraulics
Stähli, M., Sättele, M., Huggel, C., McArdell, B.W., Lehmann, P., Van Herwijnen, A., Division, American Society of Civil Engineers, San Francisco, CA, 25–30 July 1993.
Berne, A., Schleiss, M., Ferrari, A., Kos, A., Or, D., Springman, S.M., 2015. Monitoring 2. pp. 1908–1913.
and prediction in early warning systems for rapid mass movements. Nat. Hazards Wong, W.K., 2006. RAPIDS — operational blending of Nowcast and NWP QPF. In: The
Earth Syst. Sci. 15 (905–20), 917. http://dx.doi.org/10.5194/nhess-15-905-2015. 2nd International Symposium on Quantitative Precipitation Forecasting and
Staley, D.M., Kean, J.W., Cannon, S.H., Schmidt, K.M., Laber, J.L., 2013. Objective de- Hydrology, Boulder, Colorado, USA, 4–8 June 2006.
finition of rainfall intensity–duration thresholds for the initiation of post-fire debris Wong, M.C., Wong, W.K., Lai, E.S.T., 2006. From SWIRLS to RAPIDS: Nowcast applica-
flows in southern California. Landslides 10, 547–562. http://dx.doi.org/10.1007/ tions development in Hong Kong. In: WMO PWS Workshop on Warnings of Real-Time
s10346-012-0341-9. Hazards by Using Nowcasting Technology, Sydney, Australia, 9–13 October 2006.
Su, W.-R., Hsu, P.-H., Wu, S.-Y., Lin, F.-T., Chou, H.-C., 2010. Development of Safe World Bank and GFDRR (Global Facility for Disaster Reduction and Recovery), 2013.
Taiwan Information System (SATIS) for typhoon early warning in Taiwan. Syst. Weather, Climate and Water Hazards and Climate Resilience: Effective Preparedness
Cyber. Inform. 8 (4), 1690–4524. through National Meteorological and Hydrological Services. World Bank,
Tatizana, C., Ogura, A.T., Cerri, E.L.S., da Rocha, M.C.M., 1987. Análise de correlação Washington, DC.
entre chuvas e escorregamentos — Serra do Mar, Município de Cubatão. In: Wu, M.F., Chen, C.F., Lee, W.F., Ishihara, K., 2011. Development and Deployment of Early
Proceedings of 5th Congresso Brasileiro de Geologia de Engenharia, ABGE Vol I, Warning System for Highway Safety under Extreme Natural. Report. https://ssms.jp/
225:236, São Paulo - Brazil. wp-content/uploads/PDF/ssms_papers/sms12-4918_
Teisberg, T.J., Weiher, R.F., 2009. Background Paper on Benefits and Costs of Early 593a85ba6d960600113a720248bc18af.pdf.
Warning Systems for Major Natural Hazards. GFDRR, Washington D.C., USA. Yin, K., Chen, L., Zhang, G., 2007. Regional landslide hazard warning and risk assessment.
Terlien, M.T.J., 1998. The determination of statistical and deterministic hydrological Front. Earth Sci. 14 (6), 85–97. http://dx.doi.org/10.1016/S1872-5791(08)60005-6.
landslide triggering thresholds. Environ. Geol. 35, 124–130. http://dx.doi.org/10. Yin, Y., Wang, H., Gao, Y., Li, X., 2010. Real-time monitoring and early warning of
1007/s002540050299. landslides at relocated Wushan Town, the Three Gorges Reservoir, China. Landslides
Terranova, O.G., Gariano, S.L., Iaquinta, P., Iovine, G., 2015. GASAKe: forecasting 7 (3), 339–349. http://dx.doi.org/10.1007/s10346-010-0220-1.
landslide activations by a genetic-algorithms-based hydrological model. Geosci. Yu, Y.F., Lam, J.S., Siu, C.K., Pun, W.K., 2004. Recent advance in landslip warning

246
L. Piciullo et al. Earth-Science Reviews 179 (2018) 228–247

system. In: Proceedings of the HKIE Geotechnical Division 24th Annual Seminar, gov.it/rischinaturali/rischi/rischio-idrogeologico/frane-superficiali/scenario-
Hong Kong, pp. 139–147. attuale.html (last access: April 2017).
Zhang, G., Chen, L., Zhengxing, D., 2011. Real-time warning system of regional landslides Ahmed, Bayes, 2015. http://www.landslidebd.com (last access: April 2017).
supported by WEBGIS and its application in Zhejiang Province, China. Procedia Earth CFU — Centro Funzionale Umbria http://www.cfumbria.it/FraneMonitoraggio/
Planet. Sci. 2 (2011), 247–254. http://dx.doi.org/10.1016/j.proeps.2011.09.040. LandWarn/FraneMonDescrizione.php (last access: April 2017).
Zhong, L., Xiao, S., Zhou, Y., 2009. Research on the early warning and forecast system of DOGAMI — Oregon Department of Geology and Mineral Industries http://www.
geologic hazards in Hubei Province based on WEBGIS. In: First International oregongeology.org/sub/Landslide/debrisflow.htm (last access: April 2017).
Workshop on Education Technology and Computer Science. Wuhan, Hubei, China,
pp. 602–606. ICIMOD — International Centre for Integrated Mountain Development (2017). http://
www.icimod.org/?q=19755 (last access: April 2017).
NBRO — National Building Research Organisation, 2017. http://www.nbro.gov.lk (last
Web references access: April 2017).
Petley, 2008. http://blogs.agu.org/landslideblog/2008/06/22/garbage-dump-
ARPA — Agenzia Regionale per la Protezione Ambientale http://www.arpa.piemonte. landslides/ (last access: April 2017).

247