Introduction

Background

Early warning is defined as “the set of capacities needed to generate and disseminate timely and meaningful warning information to enable individuals, communities and organizations threatened by a hazard to prepare and to act appropriately and in sufficient time to reduce the possibility of harm or loss” (UNISDR 2009). Effective early warning systems (EWSs) should encompass four main aspects: risk knowledge, monitoring and warning service, dissemination and communication and response capability (UNISDR 2009). A weakness or failure in any one aspect could result in failure of the whole system (UNISDR 2009). The first aspect regards the knowledge of the physical mechanisms that cause the hazard, and of the exposed elements at risk, with their level of vulnerability. The second aspect includes the technical capacity to continuously monitor the hazards and to develop evolutionary scenarios and to issue warnings. The third aspect regards the communication of these warnings. Finally, the last aspect includes the capacity to timely translate the predictions into a warning and action plan.

EWSs for landslides have been deployed since the 1970s in Hong Kong (Chan et al. 2003) and USA (Keefer et al. 1987), and more recently in other countries. However, these systems were mainly developed for the early warning of shallow landslides and debris flows at regional scale, based on rainfall thresholds and meteorological monitoring systems (Aleotti 2004; Baum and Godt 2010; Jakob et al. 2012).

For large slides that are known to be active (herein called “active slides”), a local EWS needs to be deployed, based on a detailed knowledge of the landslide and on monitoring different parameters that can act as precursors, such as superficial and/or deep displacement rate, groundwater pressures and seismic noise (Angeli et al. 2000; Zan et al. 2002; Froese et al. 2006; Froese and Moreno 2014; Blikra 2008, 2012; Casagli et al. 2010; Yin et al. 2010; Intrieri et al. 2012; Michoud et al. 2013).

In this paper, we present the findings and recommendations of the First International Workshop on Warning Criteria for Active Slides (IWWCAS) that took place in Courmayeur, Italy, from June 10 to 12, 2013. The main idea of this paper, as well as the one of the workshop, is to stimulate discussion and collaboration between organizations dealing with the complex task of managing hazard and risk related to active slides. This report contributes to the activities supported by the International Consortium on Landslides (ICL).

Objective of the IWWCAS

The aim of this workshop was to provide a unique opportunity to share experiences about the challenges, problems and available tools to determine warning criteria. In particular, the main issues addressed by the workshop were on how to use available data, at different sites and stages of the studied problem, to choose indicators, to define threshold values and to update them with the evolution of the phenomenon in order to set up a reliable and shared EWS. Technological solutions were then considered as tools but not as a substantial aspect to the problem. The event schedule is presented in Table 1, including the titles and authors of the 22 talks that were given, and the participants are identified in Fig. 1. In addition, further details on the workshop organization are provided in Appendix.

Table 1 Talks given at the First International Workshop on Warning Criteria for Active Slides
Full size table
Fig. 1
figure 1

Group picture. Seating down (left to right): Hideaki Marui, Marco Vagliasindi, Jacques Locat, Giovanni Crosta and Oldrich Hungr. First row standing up: Mario Lovisolo, Corey Froese, Nicoletta Negro, Denis Jongmans, Igor Bravo, Christian Zangerl, Paolo Fratinni, Carlo Troisi and Daniele Giordan. Second row standing up: Marc-Henri Derron, Jean-Philippe Mallet, Federico Agliardi, Michel Jaboyedoff, Lars Blikra. Last row: Clément Michoud, Carlo Rivolta, Catherine Cloutier, Mauro Rossi and André Stumpf

Full size image

Another objective was to find a venue that would provide an opportunity to visit at least one site where an active landslide is being investigated and ideally where an EWS is operating. Courmayeur was very suited for this purpose. The participants visited the Mont de La Saxe rockslide (Fig. 2), an active landslide under continuous monitoring since 2009 (Fig. 3) (Crosta et al. 2013). The rockslide monitoring data were used to run an exercise with involvement of all the contributors. The objective of this exercise was to walk as an expert group through all the steps leading from the planning to the design and management of a warning system, through the definition of shared scenarios. This exercise was tentatively steered on the basis of the information available at different stages since the beginning of the investigations and monitoring.

Fig. 2
figure 2

View of the Mont de la Saxe and its rockslide. The village of Entrèves, where the participants were staying, is in the valley

Full size image
Fig. 3
figure 3

A visit of the Mont de la Saxe rockslide took place on the last day of the workshop. The village of Entrèves and the entrance to the Mont Blanc tunnel can be seen in the valley

Full size image

Findings and recommendations

The following presents various comments, conclusions and recommendations obtained from talks, discussions and from the exercise on the Mont de La Saxe rockslide. Most of the comments are presented as bullet points. It can be noticed that many questions are still open.

What are the concerns of scientists managing EWS?

  • General concerns were evoked about the civil responsibility of the scientist/consultant designing EWSs, providing the scenarios and defining the alert thresholds (Malone 2008; Jordan 2013; Alexander 2014). For example, the definition of warning criteria helps to reduce the responsibility of the person taking the decision of evacuation, because the decision is based on a pre-defined criterion. However, in cases of false alarm or if a slide happens without an alarm, what is the responsibility of the scientist/consultant who defined the threshold value? So, is it possible to define shared indicators and threshold values to share also the responsibilities? How uncertainty can be considered in the management of an EWS? Is it possible and advisable to communicate this uncertainty to the public? What are the differences involved in managing an EWS for landslides in natural or artificial environments?

  • When studying an active landslide with potentially catastrophic consequences, the scientist/consultant might be working under psychological or political pressure. This kind of situation that threatens human life requires to quickly provide recommendations regarding safety measures once the landslide is recognized. The importance is to avoid an emotional decision that could be taken by the scientist/consultant or by the technical staff and to communicate properly the decision to the population, so to make it acceptable as much as possible.

  • Uncertainty involves the limited knowledge relative to the immediate or future evolution of a landslide, or the change in its properties and consequently in the sensitivity to external perturbations of different sequence and intensity. How should we communicate the state of an active landslide and the related uncertainties to the authorities that do not have a robust understanding or scientific background? How can we convey the uncertainty to the public and to the managers or decision makers and stakeholders maintaining their confidence in the system and in the group of experts?

  • Uncertainties are greater at the beginning of an investigation and should decrease with time and a better understanding. So, at the beginning of an investigation, should the scientist present all the scenarios to the authorities including the very pessimistic ones?

  • Communication of threshold values and their successive readjustment can be of help or convey a feeling of inadequate knowledge and understanding. Is it reasonable to have multiple groups working independently at the same site and presenting different models explaining the landslide behaviour or diverse EWS solutions? At least for these last questions, we agree that a common understanding and set of threshold values and mitigation solutions should be reached and presented. In fact, any deficiency in sharing data and model results or scenarios should be considered as the worst managing procedure when a collective understanding is required.

Are EWS suited for long term operation?

  • EWS maintenance is expensive and requires time and effort to keep the system updated and a 100 % functional (e.g. Wilson 2004). The maintenance is complicated by quickly evolving technologies leading to incompatibility between older and newer systems. Keeping an EWS operational also implies conducting simulation exercises to train staff and update evacuation plans. Finally, a certain rotation within the personnel is recommended to avoid that know-how and system knowledge get lost when a key person is missing or more simply retire.

  • The dilemma of calling or not an evacuation might come back year after year. This can be the consequence of different constrains which should be evaluated, and it is also related to the need for updating the threshold values adopted for alert and alarm phases. In fact, system managers often forget to consider a landslide as an evolving process, which change in properties and sensitivity and as such can present different response in time. For the same reason, it is important to make aware the stakeholders, population and administrators that a minimum monitoring time is requested for the experts to reach a consensus about the type of behaviour, expected evolution, possible scenarios as well as to reformulate them after each major reactivation.

  • When a warning system has been in place for some time, and no acceleration periods were measured and an important maintenance is needed, should new money keep being invested or should a program of seasonal survey techniques be put in place instead? What criteria should come into play in the decision? Which kind of study or analysis should be completed? How this decision should be presented to the population and administrators? How can we communicate to politicians and stakeholders that some EWS at specific study sites should be maintained active just to help improve our understanding and foresight in landslide behaviour?

  • New technologies can be of help when the last series of questions arise. In fact, it is sometime possible to choose a monitoring system with lower or higher acquisition frequency that will save money and make sufficiently accurate monitoring to initiate pre-warning.

What should we think about when designing a warning system?

  • A conceptual revisable model of the instability should be created before designing the EWS, as such a model is mandatory to choose the location of devices. At the same time, an EWS can grow by progressively implementing a monitoring network initially installed for the completion of a revisable conceptual model. In this way, the monitoring network and EWS can be improved in successive steps when understanding is also enhanced The quality of the monitoring data depends on the knowledge of the instability geometry, failure mechanism and possible evolution in time and space, and as such will improve with time.

  • From experience, the use of multiple redundant monitoring techniques is advantageous for landslide characterization. For example, it is helpful to ensure the relevance of displacement measurements used to interpret the kinematic behaviour and to remain operational also under critical weather conditions or very complex phases of evolution of the landslide.

  • Active landslides with large cumulative displacements are extreme environments to be studied because the type of evolution causes short operational life for most of the monitoring equipment and especially for deep monitoring instrumentation. This is a major problem when aimed at the understanding of the role of some controlling factors such as groundwater recharge and pore pressure distribution, or displacements at depth. This is for example the case of the La Saxe rockslide (Crosta et al. 2013).

  • Instruments, proper to conduct surveillance, should be installed from the beginning of the investigation, because they will be useful (1) to follow the slide more or less continuously in case the interpreted hazard level is preoccupying, (2) to improve the general safety (workers on site and other people at risk), as the evolution of the slide can be followed and (3) to evaluate the efficiency of countermeasures that could be installed following the analysis. This also requires the capability to interact with the system during the various phases of evolution so to adapt frequency of measurements, derived variables and type of indicators and threshold values.

  • The remote location of some EWS limits their reliability as it hinders troubleshooting. For example, some agencies are now trying to rely less on private carrier (e.g. cellular network) and more on radio, because they can better control this type of communication channel maintaining operative also under critical conditions.

  • Data processing was improved and simplified by research and application of monitoring techniques at an increasing number of sites. There is still room for improvement, for example to develop cheaper monitoring systems, or to optimize communication systems, data file size and transmission protocols.

  • The analysis of monitored seismic noise showed that at some experimental site and in lab experiments the signal changes prior to an acceleration, so even before a minor displacement is captured by the monitoring system. Thus, seismic noise could be used to define threshold values for certain kinds of landslide (Mainsant et al. 2012). No operational warning system relies on this technique, yet but it could be a promising tool to experiment at sites involving different types of movement and affected materials.

How to determine warning criteria?

Discussion during the workshop emphasized the lack of tools to define warning criteria for active slides which could be established ahead of the collapse time and do not require continuous updating simultaneously to the event evolution. Here are some of the points that were raised by the participants:

  • Most of the time, the threshold values are defined empirically, sometimes based only on literature values and at the very early stages of the investigations and monitoring. Nevertheless, it has been shown that the range of displacement rate typical of the pre-collapse phase can be quite broad.

  • To define quantitative-physically based thresholds, we need a data set spanning over a long period of time and for which some of the most relevant variables are made available or collected. For systems close to collapse minimal changes in one of the controlling factors could bring to failure or catastrophic collapse, but following different evolutionary paths.

  • The difficulties related to criteria definition are, in good part, due to our misunderstanding of mechanisms controlling rockslides.

  • In the workshop talks, the factors accounted for warning criteria definition were (1) the landslide characteristics and dynamics (e.g. type of involved material), (2) the position of the element at risk in relation to the slide (on or below), (3) the previously monitored activity of the landslide and (4) the history of the slope (e.g. recently excavated or natural long term evolution). The definition of warning criteria is based on scenarios of failure and run out models. The participants insisted on the importance of creating rapidly a first geomechanic and kinematic model of the slide to confront our understanding to incoming data and subsequently improve the model.

  • The participants discussed about the inverse velocity method as a prediction tool (Saito 1969; Fukuzono 1985, 1990; Crosta and Agliardi 2003; Rose and Hungr 2007; Federico et al. 2012) and successful applications of the technique were presented mainly for mine slopes. However, this technique is not totally safe from false alarm, as for a landslide undergoing seasonal variations, and its use is sometimes hampered by the availability only of superficial displacement data.

  • Operational thresholds presented through the talks were, for the majority, velocity thresholds, sometime applied in conjunction with the inverse velocity method. Thus, most criteria for active slides are based on displacements data even though the triggering agent is recognized to be groundwater recharge (Michoud et al. 2013) associated to snow melting or intense rainfall. The role of water can be incorporated by studying slide sensitivity to groundwater recharge and by the use of precipitation thresholds. This approach is widely used for regional-scale EWS for shallow landslides and debris flows (Baum and Godt 2010). Crosta et al. (2013) discuss the evolution of the La Saxe slide sensitivity to groundwater recharge during the evolution of the slide itself, with the progressive change in hydraulic and mechanical properties accompanying the accumulation of large displacements, the opening of fractures and the localization of shear at depth.

  • In the study cases presented at the workshop, the cumulative displacement was not a parameter directly used to define warning criteria. However, cumulative displacement has the advantage to be more stable than velocity, at least in most of the phases not immediately preceding the final collapse. The cumulative displacement has been sometime used to define warning criterion, for example, by fitting a curve to the evolution of displacements with time (Crosta and Agliardi 2002). It has been used also in the case of toppling failures (Zvelebil and Moser 2001).

  • It seems inadequate to fix warning levels based on a single criterion when coping with complex landslides especially in natural environments. The establishment of warning levels should take into account all parts of a system, such as displacement data, weather, season, consequences, groundwater recharge and other types of data. Data interpretation by an expert is often required prior to the initiation of emergency procedure and even more at the closure of the emergency phase.

  • During the exercise, the participants to the workshop did not reach a complete agreement on the type of thresholds to apply, neither on the values, nor the techniques on which to rely. This lack of general agreement points out to the need for EWS design guidelines and to the need of a longer and more informed discussion before the achievement of a consensus even within a community with a specific expertise on the subject. This can be surprising especially when considering that most of the participants recognized the level of knowledge and the relative abundance of monitoring data available for the La Saxe case study. Furthermore, this discussion emphasized the major difference between managing an EWS for natural slopes in highly populated areas and managing an EWS for slopes under artificial conditions and very specific type of occupancy, like mine slopes.

On what should focus future research?

  • When possible, we should opt for monitoring techniques enabling characterization and monitoring at the same time, ensuring robustness and continuity in data acquisition

  • There is a need for guidelines and for a tool box to help the scientists quickly provide answers to end-users (managers) (Intrieri et al. 2013).

  • Guidelines for stakeholders, EWS managers, professionals, technical staff and scientists should be prepared to help steering the main steps of an EWS development and especially to steer the group towards a shared set of procedures, indicators and threshold values. This type of guidance should help in developing a participative system to support decisions and to share the responsibilities.

  • As groundwater pressures and precipitations appear to be driving factors for many active slides, there should be more work about the hydrogeology of active slides, considering also their progressive evolution from initial failure to collapse through a series of successive evolutionary steps.

  • Is the shear zone behaviour a key features for rapid failure? This has been suggested by many authors in the literature (Fukuoka et al. 2007; Pinyol and Alonso 2010; Kalenchuk et al. 2012), but its description and definition in many cases remains a difficult task which can require careful investigations and monitoring both in time and space.

  • Some new technologies must be more widely tested, such as seismic noise, in order to transfer these technologies from research to practice. More attention should be placed on data interpretation with respect to the simple acquisition by new technologies.

  • EWS should be limited to specific cases where countermeasures are not suitable.

A second workshop?

The main objective of the workshop was to focus on warning criteria definition, adoption and management. The workshop turned out to point out more to the problems related to their definition than to tools and solutions.

For a future second workshop, it is suggested to form small working groups prior to the event to gather suggestions and to make sure that the focus is on solutions. A few suggestions for working groups are proposed hereunder:

  • A grouping of similar case studies to create some sort of classification chart, including types of scenario or run out, in order to have examples on which to rely for new designs of EWS.

  • Guidelines for the application of different approaches such as the dated inverse velocity method on different types of landslide, including procedures, mathematical tools, precisions, types of landslide suited for the method, frequency of updating, etc.

  • A review on the responsibilities, legally talking, of designers of a warning system in the case of false alarm or in the case of failure to predict an event causing consequences. For example, recently in Italy, scientists have been sued for false alarm because they are considered responsible for the economical losses resulting from a misplaced evacuation or for an underestimation of the level of risk. This is an important point, because in case of an EWS, it hampers its efficiency. This problem can be overpassed only by an informed and shared decisional process, where the final decision is the result of a common path done by all the stakeholders, experts and representative of the population.

  • Define the requirements of land managers, in order to design warning systems that respond correctly to their needs and to those of the affected population and activities.

  • To avoid the confusion between scenarios and thresholds. Scenarios refer for example to the expected volume and runout behaviour, whereas thresholds are values for specific indicators representative of an expected change in behaviour and which can be associated to a scenario. The same scenario can develop in a very short time or in a much longer one, according to some local or environmental constraints (e.g. rainfall, snowmelt). Emergency or civil protection plans should consider this difference and evaluate the requirements for evacuation or activation of other emergency actions.

  • Apart from displacement, velocity, acceleration, precipitation and groundwater recharge criteria, what are the other possibilities?

  • When to choose EWS instead of active or other passive mitigation measures? Or when to combine them? By associating different probabilities of occurrence to the different scenarios?

Concluding remarks

The contained size of the First International Workshop on Warning Criteria for Active Slides (28 participants) enabled an open-minded discussion. This type of structure should be maintained. The venue, next to the active La Saxe rockslide, definitely immersed the participants into active landslides management and showed concretely the implications of EWS in terms of evacuation and repercussions on the population, as well as on the technical staff and the involved experts.

The participants shared their experience and issues related to the definition of warning criteria and to management of EWS. We realized that different organizations had common problems, especially in EWS management and maintenance. This observation leads to one of the conclusions of this workshop: EWSs are relatively new in natural hazard protection, and we are still learning how to make it right and the tools for warning criteria definition are limited. There are many benefits of landslide monitoring and surveillance; for example, it increases our understanding of the phenomenon and its possible changes and it shows the limited validity of some scenarios or the need for multiple scenarios both to decide about mitigations and to prepare suitable and alternative emergency plans.

Difficulties in forecasting the behaviour of a landslide are partly related to misunderstanding of complex active landslides physics. In fact, the group was constantly diverging towards a discussion about failure mechanism when trying to determine warning criteria.

The study cases presented at the workshop showed that, most of the time, characterization and surveillance of landslides are done synchronously and that for most of the cases, excepted for mining, the monitoring starts well after the beginning of the instability. It also showed that EWS requires regular analyses by an expert, or better an expert panel, to interpret the landslide behaviour, to detect any kind of evolution and to update the models.