Introduction to Hydrology - Solutions Manual, 5th Edition Warren Viessman Jr., Gary L. Lewis

Full text

(1)

‐‐

methods

methods

A

A literature

literature review

review towards

towards

reliable

reliable prediction

prediction of 

of time

time to

to failure

failure

Version:

(2)
(3)

Contents

Contents

Abstract Abstract ...33 Introduction Introduction...44 Characterizing

Characterizing aa landslidelandslide ...55 Lead

Lead timetime andand accuracyaccuracy of of predictionprediction ... 66 Long

Long‐‐termterm forecastforecast methodsmethods......77

Mid

Mid‐‐andand shortshort‐‐termterm forecastingforecasting...88

Forecasts

Forecasts basedbased onon inin‐‐situsitu measurementsmeasurements...... 88

Deformation

Deformation basedbased forecastsforecasts ...88 Pore

Pore‐‐waterwater pressurepressure basedbased forecastsforecasts ...... 1515

Water

Water contentcontent basedbased forecastsforecasts ... 1515 Micro

Micro seismicseismic basedbased forecastsforecasts ... 1616 Forecasts

Forecasts basedbased onon climaticclimatic conditionsconditions... 1717 Rainfall

Rainfall thresholdthreshold basedbased forecastsforecasts ... 1717 State

State‐‐of of ‐‐thethe‐‐artart andand futurefuture trendstrends ......2020

Measuring

Measuring devicesdevices...2020 Multi

Multi‐‐parameter basedparameterbased forecastsforecasts...... 2020

Wireless

Wireless SensorSensor NetworksNetworks WSNWSN... 2121 Wireless

Wireless UndergroundUnderground SensorSensor NetworksNetworks WUSNWUSN... 2424 Conclusion

Conclusion andand futurefuture needs...needs... 2525 Role

Role of of GeotechnicalGeotechnical ExpertsExperts ...2525 Selection

Selection of of appropriateappropriate monitoringmonitoring parameterparameter... 2525 Next

Next stepssteps towardstowards reliablereliable forecastsforecasts ... 2626 References

(4)

Abstract

Abstract

The

The followingfollowing articlearticle isis thethe resultresult of of anan extensiveextensive literatureliterature reviewreview aboutabout timetime forecastsforecasts of of landslides.landslides. A

A shortshort overviewoverview aboutabout longlong‐‐termterm forecastsforecasts isis given,given, butbut thethe focusfocus isis onon midmid‐‐ andand shortshort‐‐termterm

forecasts.

forecasts. TheThe methodsmethods areare classifiedclassified byby thethe parametersparameters requiredrequired toto makemake aa predictionprediction (i.e.(i.e. inin‐‐situsitu

measurements

measurements andand climaticclimatic conditions).conditions). TheThe methodsmethods areare summarizedsummarized andand referencesreferences toto detaileddetailed articles

articles areare givengiven forfor eacheach method.method. TheThe aimaim of of thisthis reviewreview isis toto givegive anan overviewoverview of of  previouslypreviously mademade attempts

attempts toto predictpredict landslideslandslides andand futurefuture researchresearch needsneeds andand challengeschallenges areare addressed.addressed. This

This reviewreview hashas shownshown aa strongstrong needneed forfor lowlow‐‐costcost warningwarning systemssystems forfor landslides.landslides. Therefore,Therefore, currentcurrent

research

research activitiesactivities inin thethe fieldfield of of wirelesswireless sensorsensor networksnetworks areare presentedpresented asas well.well. In

In additionaddition toto thisthis article,article, aa summarysummary of of  thethe presentedpresented methodsmethods isis givengiven inin aa separateseparate schemaschema calledcalled “Landslide

“Landslide ForecastForecast Toolbox”.Toolbox”. ForFor thethe futurefuture itit isis plannedplanned toto programprogram aa wiki,wiki, toto makemake thisthis literatureliterature review

(5)

Introduction

The term landslide denotes “the movement of  a mass of  rock, debris or earth down a slope”. Landslides are a regional and site‐specific problem. The occurrence of  landslides depends on

topography, geology, groundwater, weather, vibrations and human causes. Landslides range in many orders of  magnitude in size, from small boulders to several cubic kilometers of  mass. Speeds vary from extremely slow (mm/y) to extremely rapid movements (several 100km/h) (Cruden and Varnes 1996). Landslide processes are very complex and present challenges in the development of  early warning systems. Nevertheless, several successful attempts have been made to forecast landslides. This literature review gives an overview about landslide forecast methods with a focus on real‐time

warning. An outlook about future needs and research trends is given as well.

Besides rigorous avoiding of landslide prone areas, successful early warning is one of the most cost‐

effective ways of  disaster prevention. Real‐time warning systems are suitable for communication

routes and life lines, such as railroads and highways where hazard zones can not be bypassed. The United Nation International Strategy for Disaster Reduction ISDR divides an early warning system into four elements (web ISDR):

The Four Elements of Effective Early Warning Systems

Risk knowledge

Systematically  collect  data and undertake risk assessments Are the hazards and the vulnerabilities well known? What are the patterns and trends in these factors? Are risk maps and data widely available? Monitoring and warning service Develop hazard  monitoring and early  warning services Are the right parameters being monitored? Is there a sound scientific basis for making forecasts? Can accurate and timely warnings be generated? Dissemination and communication Communicate risk  information and early  warnings

Do warnings reach all of those at risk? Are the risks and the warnings understood? Is the warning

information clear and useable

Response capability

Build national and  community  response capabilities

Are response plans up to date and tested? Are local capacities and

knowledge made use of? Are people prepared and ready to react to warnings?

(6)

A general early warning system for landslides is not given up to this day. The types of movement (i.e. fall, topple, glide…) are very different processes and therefore it is difficult to develop a general landslide warning system (Baehr, 2004). However, it is possible to monitor instable slopes according to their local properties. This literature review gives an overview about the methods developed so far and intends to inspire engineers to develop new techniques and combine different methods.

Characterizing a landslide

Cruden and Varnes 1996 reviewed the range of  landslide processes and provided a vocabulary for describing the features of  landslides relevant to their classification. A nomenclature for the observable landslide features is illustrated in Fig. 2 below.

Fig. 2 Block diagram of idealized complex earth slideearth flow (Cruden and Varnes 1996)

Any landslide can be classified and described by two nouns: the first describes the material and the second describes the type of movement.

The material can be divided into either rock, a hard or firm mass that was intact in its natural place before the initiation of  movement, or soil, an aggregate of  solid particles, generally of  minerals and rocks, that either was transported or was formed by the weathering of rock in place.

Soil is divided into earth and debris (Tab. 1). Earth describes material in which 80 percent or more of  the particles are smaller than 2 mm, the upper limit of  sand‐size particles recognized by most

geologists. Debris contains a significant proportion of  coarse material; 20 to 80 percent of  the particles are larger than 2 mm, and the remainder are less than 2 mm.

(7)

Abbreviated Classification of Slope Movements

(After: Cruden and Varnes 1996)

Type of Material

Engineering Soils Type of 

Movement Bedrock Coarse Fine Fall Rock fall Debris fall Earth fall

Topple Rock topple Debris topple Earth topple

Slide Rock slide Debris slide Earth slide

Spread Rock spread Debris spread Earth spread

Flow Rock flow Debris flow Earth flow

Tab. 1 Names to describe landslides are listed (e.g.,rock  fall, debris flow). After (Cruden and Varnes 1996)

Lead time versus accuracy of  prediction

In terms of time, forecasts can be roughly divided into three classes of lead time. Long‐term forecasts

mostly indicate a potential hazard within a certain region, years before they actually occur. Mid‐term

forecasts predict failures several months ahead. And finally short‐term predictions have a lead time

of months to days.

As a rule of  thumb we can say: “Longer lead time allows more preventive actions against landslide disasters. But on the other hand longer lead time comes often with less accuracy, in terms of  time and location.”

This review focuses mainly on mid‐and short‐term predictions. Nevertheless, a short overview about

(8)

Long

term forecast methods

The first step towards a landslide forecast is often a systematic collection of  data in a landslide hazard zonation. An ideal map of  slope instability hazard should provide information on the spatial probability, type, magnitude, velocity, runout distance and retrogression limit of  the mass movements predicted in a certain area (Soeters and van Westen 1996). These landslide inventories are mainly made by geologists. The maps can be interpreted as a first long‐term forecast. Although,

they can not predict the exact time of an event and cover a region, rather than a specific slope, they indicate a potential hazard several years in advance of an event.

The last few decades have shown very rapid development of  the application of  digital tools such as Geographic Information Systems (GIS), Digital Image Processing, Digital Photogrammetry and Global Positioning Systems (GPS). In landslide risk assessment at scales of  1:10’000 or smaller, GIS has become the standard tool.

Much progress has been made in the generation of  Digital Elevation Models (DEM) obtained from different sources like Synthetic Aperture Radar (SAR) or Light Detection and Ranging (LIDAR). DEM are used to generate landslide inventories. Landslide inventories can now make use of  a variety of  approaches, ranging from digital stereo image interpretation to automatic classification, based either on spectral or altitude differences, or a combination of both. Landslide inventory databases become available to more countries and several are now also available through the Internet. A comprehensive landslide inventory is a basic requirement in order to be able to quantify both landslide hazard and risk (van Westen 2007). Soeters and van Westen (1996) as well as Guzzetti et al. (1999) presented very good and structured reviews of current techniques to obtain landslide hazard maps. The results of  these assessments result in hazard maps and are used by regional planners to avoid endangered zones in public planning or take according measures.

(9)

Mid

and short

term forecasting

Routine surveys of landslide prone areas or slopes provide information about the progress of instable masses. Usually, unstable locations or even specific slopes can be identified. Depending on the method, a failure forecast can be made. Routine surveys allow monitoring with lead times of years to months for mid‐term forecasts.

Real‐time monitoring allows the most accurate prediction of  landslides, within several months to

days. For example Xiaoping et al. (1996) have forecasted a slope failure at Yellow River in Gansu Province, China on the 30. January 1995 at with an accuracy of one day!

The following chapter gives an overview about different mid‐ and shortterm forecast methods. The

methods are structured into two groups. The first group contains forecast methods based on in‐situ

measurements of different parameters in the slope. In the second group, methods based on climatic conditions, are presented. All the following methods require a careful assessment of  the instable slope in order to place the measuring devices at characteristic points.

Forecasts based on in

situ measurements

Deformation based forecasts

For deformation based forecasts the soil must have a plastic behavior in order to observe any deformations prior fracture.

Deformations of  soil under a constant load can be plotted in a time vs. strain diagram (Fig. 3). The strain rateε 

&

is defined as the derivative of  strainε , with respect to time. In the initial stage, known

as primary creep, the strain rate is relatively high, but slows with increasing strain. The strain rate eventually reaches a minimum and becomes near‐constant. This is known as secondary or steady

state creep and depends on creep mechanism and soil properties. In tertiary creep, the strain‐rate

(10)

secondary strain rate tertiary creep secondary creep primary creep Strain e Fracture time t Initial Load minimun

secondary creep rate Strain rate

e

Fracture

time t

ÿ

Fig. 3 In the secondary creep stage the strainrate is constant and thereafter increases until failure.

Based on measurements in the secondary creep range, Saito (1965) proposed an empirical formula to predict the time of  slope failure. The relationship between constant strain rate and rupture life (time to failure) was successfully applied to forecast landslides:

59 . 0 log 916 . 0 33 . 2 log10 = − 10ε 

&

± r  t  Eq. 1

tr: creep rupture life (min), i.e. total time from the beginning of movement until failure

ε 

&

 : constant strain rate (in 10‐4mm)

The relationship seems to be independent of  the type of  soil or testing method. Measurements of  relative displacement should be taken continuously. Forecasting the time of  slope failure is done by the following procedure:

1. Measurement of  the relative displacements of  a slope across tension cracks or along the centre line, depending on field conditions.

2. Determination of the beginning of the unstable state of the slope through the relative displacement curve.

(11)

4. Estimation of creep rupture life corresponding to the strain rate, using the relationship between strain rate and creep rupture life.

Saito (1969) extended his theory to the tertiary creep range in order to obtain more accurate forecasts for the time close to failure. The above mentioned relationship was adapted to the

transient strain rate. He presented a graphical and a numerical solution. The empirical formula for the tertiary creep range relates the time left before failure (tr‐t) to the displacement as follows:

t  t  t  t  a l l r  r  − − = Δ 0 log 0 Eq. 2

Δl: relative displacement between two measured points

l0: initial distance between two measured points

a: constant

tr: creep rupture life (min), i.e. total time from the beginning of movement until failure

t0: time when movement begins, Δl=0

t: optional time

The equation contains three unknown: a or l0∙a, trand t0. The remaining time to failure (tr‐t) can be

obtained with three or more points properly selected on the creep curve. The best way to have a good estimation is to begin displacement measurements as early as possible. The nearer failure comes the more reliable the forecast. It is advisable to roughly estimate time of  failure with steady state strain rate in the secondary creep range (Saito 1965) and predict precisely with data from the third creep range (Saito 1969). Hayashi et al. (1988) improved the Saito (1969) prediction in the tertiary creep range.

Based on large scale laboratory experiments Fukuzono (1985) presented a new method for predicting the failure time using the inverse number of  surface displacement velocity (1/v). If  the displacement velocity v at a slope surface increases over time, its inverse number (1/v) decreases. When (1/v) approaches zero, failure occurs.

(12)

time t t Failure 1 < a   < 2 (   c o n c  a v e  )  1 = {a (a-1)} ÿ (tr -t)    I  n  v   e   r   s   e   n   u   m    b  e   r   o    f  v   e    l  o  c    i    t  y    1    /  v a   = 2 (   l  i  n e a  r   )  a   > 2 (   c o n  v e x  )  v 1/(a-1) 1/(a-1) tr 

Fig. 4 Typical figures for changes of the Inverse number of velocity in surface displacement just before the failure. After (Fukuzono 1985)

The curve is described by following equation:

{

}

1 1 1 1 ) ( ) 1 ( 1 − ⋅ − − = α  α  α  t  t  a v r  Eq. 3

v: velocity of surface displacement a: constant

α: constant

tr: creep rupture life (min), i.e. total time from the beginning of movement until failure

t: optional time

Fukuzono distinguishes three cases, depending on the shape of the graph. For a linear graph α=2 and

the failure time trcan be predicted using two points. This time is equal to the time predicted by Saito

(1965): 2 1 2 1 1 2 ) / 1 ( ) / 1 ( ) / 1 ( ) / 1 ( v v v t  v t  t  − − = Eq. 4

In the case of α≠2, the failure time can be determined by differentiating Eq. 3 by t and solving for the

(13)

) )( 1 ( / 1 ) / 1 ( t  t  v dt  v d  r − − − = α  Eq. 5

A rough estimation can also be made by the point at which the tangent line of the curve crosses the axis of  abscissa. Therefore two graphical methods are presented in Fukuzono (1985). Azimi et al. (1988) proposed a graphical method that is equivalent to the linear case α=2 in Fukuzono (1985) and

similar to Saito (1969) (see Fukuzono 1990).

Rose and Hungr (2007) successfully applied the inverse velocity method in open pit mines to predict accurately three large failures ranging in size from 1 to 18 million cubic metres. They provide a nice discussion about the limitations of the inverse velocity method and suggest following general rules:

• The method must not be applied in isolation, and displacement monitoring is only one component of a complex process that comprises slope stability management.

• The method cannot be used for rock slides dominated by brittle failure, and particular care should be exercised when dealing with relatively small failures in strong rock.

• The monitoring data must be processed to remove the effects if  instrument error and eliminate records distorted by local movements.

• Failure forecasting relies on the identification of  consistent trends. The possibility of  trend changes, driven by observable or unknown factors, must always be kept in mind. Monitoring must be continued as long as possible prior to failure. The results must be constantly re‐

evaluated and any established best‐fit functions must be revised in view of the latest data.

• The treatment of cyclic changes depends on the magnitude of their amplitude, relative to the distance from the horizontal axis on the inverse‐velocity plot. If the ratio between these two

quantities is high, it may be necessary to assume that the low point of  any given cycle may produce sudden rupture.

• Data fitting using non‐linear inversevelocity trend lines may provide a more accurate

assessment of some longer‐term trends, but is more complex, which may limit practical use.

The authors recommend the use of linear fits, updated on an ongoing basis to identify trend curvature or to signal the onset of trend changes.

(14)

 f 

t  : Time at failure

*

: Optional time

*

Ω

&

: Displacement rate at*

 f 

Ω

&

: Displacement rate at t  f 

α , A: Constants of experience (commonly α = 2 ± 0.3)

Voight considers an estimated displacement rate at failure. Practical estimates are made under consideration of  slope geometry, materials and groundwater conditions. Failure occurs when the inverse rate value of  Ω

&

 f  is reached; this is not precisely the point of  intersection with the time abscissa in Fukuzono (1985). In contrast, Fukuzono assumes the displacement at failure to be infinite. Therewith, the predicted time to failure using Voigths equation is shorter and more conservative. Voights model is more ‘‘physically based’’ with respect to the empirical models proposed by Saito (1965, 1969) and provides, in suitable conditions, a very good forecast of  time to failure. As a drawback, the model is only applicable to data characterised by continuous acceleration and constant external conditions. The model fails, or becomes less accurate, when external conditions are not time invariant and deviations induced by seasonal variations in temperature and rainfall regime take place (Crosta and Agliardi 2002).

Kawamura (1985) presented a methodology, not only to predict when a landslide occurs, but also to characterize the type of  movement. The technique indicates whether a moving slope is going to fail or reaching a stable condition.

Another attempt to survey an unstable slope is to define alert velocity thresholds. Crosta and Agliardi (2002) proposed a method to obtain alert velocity thresholds for large rockslides. The method is based on Voights (1988) accelerating creep theory, where failure is assumed to occur at the time corresponding to a particular rate of  displacement. Velocity thresholds can be obtained and incorporated in an emergency concept. Like (Voight 1988), this method is only applicable to cases with invariant external conditions.

Xiaoping et al. (1996) proposed a very interesting approach. Instead of performing a classical stability analysis with force or moment equilibrium, their method analyzes the deformation power required for a slope failure. The deformation power of  a landslide at any time is never larger than at failure state, the so called failure power. Unfortunately, the authors do not give detailed information on how to obtain the failure power threshold. Nevertheless, their concept can be summarized as follows:

1. Arrange the geological survey reasonably to master the geological background and the environmental conditions for a landslide.

2. Monitor systematically with various measures to obtain the real and feasible data for the landslide deformation, abstract the valid information to control the law of  active deformation for a landslide.

(15)

3. Calculate the deformation power and the failure power of  a landslide, and check the relative level between them. Meantime, consider the acoustic emission reaction state and the macro traces of  landslide deformation to  justify the deformation phases for a landslide comprehensively.

4. During the monitoring of  landslide deformations, either the monitoring network or the monitoring line consists of  monitoring points, which are able to react and feel the deformation and active principle for the local masses of a landslide respectively. Therefore, it is necessary to analyse individual points by using the regressing method based on their own deformation information.

5. Because of the local active traits in the individual points, the results from analyzing individual points might be divergent and do not represent the deformation state of  the whole landslide. So the individual measurements have to be adapted to calculate and determine the failure time of the whole landslide.

6. According to the two methods presented in the paper “analysing with individual points” and “predicting with all points” failure time can be forecasted using polynomial regression.

A large landslide along Yellow River in Gansu Province of  China occurred on 30 January 1995. Using the approach of  Xiaoping et al. (1996), the landslide was predicted with only one day between forecast and effective failure.

TDR based deformation monitoring

Time Domain Reflectometry TDR was originally developed during the 1950 to locate and identify cable faults in the telecommunication and power industries. Later on, the application was extended to measurements in geomaterials and is now well recognized for soil moisture measurement in soils. With few modifications TDR can also be used for the monitoring of localized deformation in rock and soils. The comparably low installation costs and the possibility to perform continuous measurements make TDR an interesting alternative to inclinometers. TDR is capable of determining the exact depth of the observed deformation zone. Some semi‐quantitative statement can be made of the amount of 

movement. The orientation can not be determined at all. A landslide warning system based on TDR deformation detection is not known so far. Nevertheless, a few promising examples of  soil deformation monitoring, performed by the California Department of Transportation (CALTRANS), are presented in O’Connor and Dowding (1999). Thuro et al. (2007) are currently developing a Landslide Warning System and make use of TDR for deformation detection.

(16)

Fig. 5 Basic setup of a TDR measuring site. The coaxial cable is installed into an instable slope and connected to the TDR cable tester. As soon as the coaxial cable is deformed by the mass

movement a peak can be been in the reflected signal. Its amplitude is dependant to the amount of deformation taking place (Thuro et al. 2007)

Pore

water pressure based forecasts

Beside deformations, pore‐water pressures can be very helpful parameters to monitor unstable

slopes. Yagi and others carried out several laboratory experiments. For sandy slopes they’ve found measuring pore‐water pressures to be more effective, than monitoring deformation or strain of  the

slope surface (Yagi et al. 1985) (Yagi and Yatabe 1987).

Water content based forecasts

In order to develop a predictive warning system for rainfall‐induced slope failures, a number of 

studies have focused on monitoring the pore‐water pressures. However, present day sensors like

tensiometers and piezometers have some limitations due to their maintenance, performance and reliability. Pore‐water pressure changes in unsaturated soils go hand in hand with changes in water

content (Soil‐Water Characteristic Curve). Tohari et al. (2007) carried out several laboratory tests and

found monitoring of  moisture content in the soil to be very useful for prediction of  slope failures. Moister content measuring devices like, ThetaProbe, TDR waveguides and others overcome the disadvantages of  pore‐water pressure sensors. Tohari et al. (2007) proposed a warning algorithm

(17)

Fig. 6 The conceptual prediction methodology for rainfall induced slope failures based on moisture content measurements (Tohari et al. 2007)

1. It is crucial to understand the failure mechanism in order to install an appropriate monitoring system. The use of  a numerical analysis of  saturated–unsaturated flow can help to find the most appropriate monitoring locations to monitor the development of  the seepage area, which are often close to the face of the slope.

2. Prior installation, it is necessary to determine the in situ soil porosity at each monitoring point in order to calculate the values of  saturated volumetric moisture content, which are required to determine the degree of saturation.

3. The critical time for evacuation can be defined as the time required for the wetting front to travel from the sensor head to the impervious layer or initial groundwater level t1to t2.

4. The initiation of the second stage of increase in the moisture content of the near surface soil represents the hydrologic condition in which the groundwater table starts to rise towards the slope surface.

5. Thus, the corresponding time t2can be designated to trigger a final warning against the slope

failure hazard. The extent of the time for evacuation will depend on the depth of the sensor, with respect to the impervious layer or initial groundwater, as well as on the antecedent soil moisture condition. Consequently, the depth of the impermeable layer or groundwater level must also be determined through adequate subsurface investigations to determine an optimum depth for installation of the sensor for the effective prediction of failure initiation.

(18)

early warning system. The location of  AE observation stations at a study site is critical for recording representative data because acoustic emissions monitoring systems are extremely sensitive.

A typical acoustic emission system consists of a backfilled wave guide in a borehole with a transducer fixed to the free end, amplifiers, data acquisition and processing equipment (Fig. 7). Fine grained soils (i.e. clay and silt) generate low AE. Even coarse grained soils that are considered as relatively noisy, have a very high attenuation due to high energy loss as AE is transmitted from one particle to another. Backfill Data capture and processing  Amplifier  Cable Pre-amplifier Filter  Borehole Slip plane Transducer  Wave guide

Fig. 7 Typical devices of an acoustic emission system. After (Kousteni et al. 1999)

Depending on the backfill material two possible systems can be formed: Passive systems have to be backfilled with low AE activity material (i.e. clay) so the installation does not introduce any additional sources of AE into the waveguide. Active wave‐guide systems are installed when the monitoring site

consists of  cohesive material. The borehole is then filed with granular material which produces high AE. Acoustic emission wave guides in dry sands work in a range of 500Hz up to 16kHz (Kousteni et al 1999).

Dixon and Spriggs (2007) developed a real‐time slope monitoring system using AE to quantify pre‐

failure slope deformation rates.

Beside these buried systems there are surface micro seismic monitoring systems in use to detect AE in rock slopes with geophones. Movements in the slope are located by the geophones in a frequency spectrum of about 10‐650Hz (Stewart et al. 2004).

Forecasts based on climatic conditions

Rainfall threshold based forecasts

Compared with other methods, rainfall can be monitored with relatively low costs and therefore rainfall thresholds are a cost effective method to forecast landslides. Guzzetti et al. 2005 presented a worldwide literature review of  landslide triggering rainfall thresholds. They launched an informative homepage where the thresholds can be viewed (web Guzzetti). Following their work is summarized: Landslide‐triggering rainfall thresholds separate events that resulted, from those which failed to

trigger landslides. First of  all it must be stated, that rainfall threshold based forecasts only make sense where rainfall is a triggering factor. Most of  the proposed thresholds perform reasonably well in the region where they were developed, and cannot be exported to neighboring areas. Also, their

(19)

temporal accuracy remains largely untested. Below different rainfall thresholds are classified and some comments are given:

• Empirical thresholds are obtained studying rainfall conditions which resulted in slope failures. The main advantage of  empirical rainfall thresholds lays in the fact that rainfall is relatively simple and inexpensive to measure over large areas.

• Physically based thresholds attempt to link regional or local rainfall measurements to local terrain characteristics. For example the time needed for the wetting front to reach the slip surface, is related to the depth of  the slip surface and the permeability. Physically based thresholds are calibrated using rainfall events for which rainfall measurements and the location and time of slope failures are known.

• Intensityduration thresholdsmight be very helpful, but have also some disadvantages. They do not consider the antecedent moisture conditions. For this reason, they are less suited to predict the occurrence of  deep‐seated landslides, or of  slope failures triggered by low

intensity rainfall events. Further, intensity‐duration thresholds do not consider that

landslides can occur several hours after the end of  the rainfall event, and do not take into account site specific rainfall conditions.

• Normalized intensityduration thresholds implicitly consider climatic characteristic of  the area for which they are defined, and emphasise the regionalization of  the thresholds, since the calculation takes into account the climatic regimes of the study area.

• Thresholds considering the antecedent rainfall conditionsare mostly suited for deep‐seated

landslides. The antecedent rainfall condition thresholds require data of  lower resolution (daily rainfall data) which are available for longer periods (up to 120 years in Italy).

The forecasting phase of meteorological forecasts has improved considerably in recent years. Thanks to improving weather forecasts, landslide‐triggering rainfall thresholds will play a more significant

role in disaster management and become an important tool fore decision makers. Landslide Warning Systems based on Rainfall Thresholds

In a few places of  the world rainfall thresholds are part of  operational landslide warning systems, in which real‐time rainfall measurements are compared with established thresholds, and when pre

established values are exceeded alarm messages are issued.

(20)

Landslides in Hong Kong are heavily dependent on the magnitude of  the short period rainfall intensity, with a threshold value of  about 70 mm/hr. This relationship was adopted as the basis of  the Government’s Landslide Warning System that has been in operation since 1984.

Japan

The occurrence of  debris flows in Japan is closely related to rainfall characteristics, such as the amount, intensity and duration. To predict the standard amount of  rainfall for warning and evacuation from debris disaster, a model for the occurrence of debris flow was developed, based on the infiltration process to surface flow. An information system for warning and evacuation was necessary between governmental offices and inhabitants. This system was organized into three sections: observation, analyses and management. Rainfall was measured by rain gauges and the information to a local station where it was analyzed using the standard rainfall programs based on the static slope stability model. To improve the accuracy of rainfall prediction, combinations of radar rain gauge systems (MP‐X) and telemetering were used in order to analyze data in real‐time.

Landslide disaster prediction support system LAPSUS (Fukuzono et al.) is a software used by the National Research Institute for Earth Science and Disaster Prevention (NIED) in Japan to provide information about the potential for shallow‐landsides and the evaluation of  landslide risk. The

information is open to the public.

So far this is the one most sophisticated landslide prediction system found in this literature review.

Yangtze River, China

A regional warning system to monitor landslides and mudslides was built up and extended along the upper reaches of the Yangtze River in China. The system was set up in 1991 to monitor landslides by using 70 stations and employing over 300 professionals. The network protected a population of  300,000 people and had forecasted 217 landslides avoiding estimated economic losses of  US$ 27 million.

Rio de Janeiro, Brazil

1996 the Rio de Janeiro Geotechnical Engineering Office (GeoRio) installed a system for warning against landslides triggered by severe rainfall. The watch system was based on landslide‐rainfall

correlation and depended on confirmation from three sources of  information: (i) a short term (4‐

hour) weather forecast, obtained by meteorological radars, (ii) an automated rain gauge network of  30 rain gauges, and (iii) records of failures reported by the Civil Defence Brigade. Alarms were issued when: (i) the rain gauge network software indicated that the hourly or daily rainfall alarm level were reached at least in three rain gauges, and the weather forecast predicted rain in the successive hours, and (ii) when the hourly or daily rainfall thresholds levels were reached in three rain gauges, and the short term weather report indicated that heavy rains was expected. Once these conditions were reached, GeoRio contacted the Civil Defence Board of  the Rio Government to assess the situation and implement action (Ortigao 2000) (Ortigao et al.).

(21)

State

of 

the

art and future trends

Despite remarkable progress, substantial research deficits still exist in the development and use of  early warning systems for landslides. Therefore, future research and development projects aim on advancement of  early warning capacities by improvement of  scientific bases, development of  new technologies and prototypical implementation of early warning systems (Baehr 2004).

Measuring devices

Nowadays and in future satellites play an important role in landslide assessments. New techniques like differential Global Positioning System (dGPS) improved the accuracy of displacement monitoring up to a few millimeters. Aguado et al. (2006) designed a Low‐Cost, Low‐Power Galileo/GPS

Positioning System for monitoring landslides. Their research focuses mainly on electro technical and software specific concerns. The system has an accuracy of horizontally ± (5 mm + 0.5 ppm × baseline length) RMS and vertically ± (5 mm + 1 ppm × baseline length) RMS (TBC). Unfortunately, no detailed field application is mentioned so far.

Synthetic Aperture Radar (SAR) are usually space‐borne instruments that emit electromagnetic

radiation and then record the strength and time delay of  the returning signal to produce images of  the ground. The elevation of  the ground can be estimated using two antennas and applying interferometry (InSAR). Consecutive couples of  SAR images can be cross‐correlated to form

interferograms representing phase variations which can be directly related to ground displacement within millimeter‐level accuracy. Clear weather is necessary to operate spaceborne SAR. Therefore,

Antonello et al. (2004) developed a ground‐based InSAR system, called LISA, and applied it

successfully to monitor unstable slopes.

Light Detection and Ranging (LIDAR) is an analog method to Radiowave Detection and Ranging (Radar). Similar to the radar technology, light (laser pulses) are used instead of radiowaves, the range to an object is determined by measuring the time delay between transmission of  a pulse and detection of the reflected signal. LIDAR can also be used to monitor displacements of instable slopes. Beside these new methods to observe deformations, other measuring techniques like micro seismics have the potential to be used as indicators for landslide warning systems.

Multi 

 parameter based  forecasts

(22)

Baum et al. (2005) proposed a multi‐parameter based landslide warning system for the rail traffic

between Seattle and Everett in Washington USA. The system is based on continuous monitoring of  precipitation, soil water content and pore pressure, combined with an empirical rainfall threshold. Three levels of warning, with successively shorter lead times were projected.

 “Advisory” is the lowest level of  warning and given, if  field observations indicate a high degree of  soil saturation (>60‐80%), and therewith the material is wet enough to be

susceptible to landslide activity. In the event of prolonged, intense rainfall, shallow landslides are likely during this lowest level of warning.

 “Watch” is given for an "Advisory" combined with a forecast of  rainfall that is sufficient to exceed the empirical intensity‐duration threshold.

 “Warning” is the highest level of warning, given in case the near real‐time field data indicate that actual rainfall conditions are approaching the threshold and that soil wetness and pressure head are high, so that landslides are likely at any time.

Other multi‐parameter based warning systems are in operation at the south and east coast of 

England and described in Clark et al. (1996).

Wireless Sensor Networks WSN

Many of the warning systems described above are costly. There is a strong need for mobile, low‐cost

equipment that is quickly ready for use. A combination of different low cost sensors and monitoring methods allows making warning systems affordable.

Wireless Sensor Networks WSN match these requirements quite well. WSN consist of  spatially distributed autonomous devices using sensors to cooperatively monitor physical or environmental conditions, such as temperature, sound, vibration, pressure, motion or pollutants, at different locations. In addition to one or more sensors, each node in a sensor network is typically equipped with a radio transceiver or other wireless communications device, a small microcontroller, and an energy source, usually a battery. The envisaged size of  a single sensor node can vary from shoebox‐

sized nodes down to the size of  grain of  dust, although functioning 'motes' of  genuine microscopic dimensions have yet to be created. The cost of  sensor nodes is similarly variable, ranging from hundreds of  dollars to a few cents, depending on the size of  the sensor network and the complexity required of  individual sensor nodes. A sensor network normally constitutes a wireless ad‐hoc

network, meaning that each sensor supports a multi‐hop routing algorithm (several nodes may

(23)

Fig. 8 Schema of a Wireless Sensor based Landslide Early Warning System (Arnhardt et al. 2007)

SLEWS

The SLEWS project (Arnhardt et al. 2007) investigates the complete information chain, starting from data gathering using wireless sensor networks via information processing and analysis to information retrieval. This is demonstrated for landslides and mass movements. The proposed approach pays special attention to mobile, cost‐reduced and easy deployable measurement systems, as well as the

modern information systems under consideration of  interoperability and service orientated architecture concepts. The SLEWS project is still in the stage of development.

Senslide

Senslide is a low cost Landslide Prediction System based on Wireless Sensor Network WSN technique (Sheth et al.2007). The system consists of  single‐axis strain gauges connected to cheap nodes, each

with a CPU, battery and a wireless transmitter. Measurements are propagated to a set of  “base stations”. Sensors (600 to 900 sensors/km2) make point measurements at various parts of  the rock, but make no attempt at measuring the relative motion between rocks. Laboratory tests on rock samples from the field site show linear stress‐strain behaviour of  the rock until there is fracture.

(24)

Fig. 9 The Landslide Warning System is based on measurements in the sensor columns. Terzis et al. (2006)

Detection is performed through a three‐stage algorithm: First, sensors collectively detect small

movements consistent with the formation of a slip surface separating the sliding part of hill from the static one. Once the sensors agree on the presence of  such a surface, they conduct a distributed voting algorithm to separate the subset of  sensors that moved from the static ones. In the second phase, moved sensors self ‐localize through a trilateration mechanism and their displacements are

calculated. Finally, the directions of  the displacements, as well as the locations of  the moved nodes are used to estimate the position of the slip surface.

This information along with collected soil measurements (e.g. soil‐pore pressures) are subsequently

passed to a Finite Element Model that predicts whether and when a landslide will occur. Initial results from simulated landslides indicate that accuracy in the order of cm in the localization as well as the slip surface estimation seems to be possible. In a next step, the system will be tested on a variety of hill configurations.

alpEWAS

Thuro et al. (2007) are currently developing an integrative 3D early warning system for alpine instable slopes (alpEWAS). Surface movements will be detected punctiform and highly precise with the Global Navigation Satellite System (GNSS) as well as extensively in a large part of  the landslide area through reflector less tacheometry. The final aim of  reflector less tachoemetry is to develop a system, able to autonomously find and select suitable natural targets for deformation monitoring. Movements in the depth alongside boreholes will be detected by using newly adapted TDR. Hydrostatic pore pressures and climatic conditions will be measured as well. The researchers expect a relatively short time (6 to 9 months) to be necessary for calibration and definition of  critical thresholds. Wireless Local Area Network WLAN is used for communication.

ILEWS

The research project Integrative Landslides Early Warning Systems (ILEWS) aims to develop a system to predict landslides and debris flows (Glade et al. 2007). The project is split into subprojects, wherefrom a few aspects are indicated here. Measurements are taken using several different sensors

(25)

in a wireless sensor network (web ScatterWeb). Geoelectrical survey systems are applied to examine the landslide body.

In order to make predictions, the monitoring data are fed into three different models:

 The Movement Analysis Model performs trend analysis of  reciprocal landslide movement rates (inclinometer, inclinometer chain, tachymetry, GPS). Depending on the way how the movement rates change, the catastrophic failure of  a slope can be predicted. For example, linear trends give an early indication of time to a catastrophic event.

 The continuously measured field parameters are modeled in a permanent slope stability analysis in the “Near Real‐Time” Physically based Model. For example soil moisture and

rainfall will be integrated in equations to calculate slope stability. If  the factor of  safety gets lower than a specified threshold value, preliminary warnings will are provided through WebGIS and SMS are released.

 In the Real‐Time Model for Debris Flows forecasts are made under consideration of weather forecasts.

Data will be visualized in a web based geographical information system WebGIS. RockNet 

An interdisciplinary team of  geotechnical and electrical engineers at HSR University of  Applied Science in Rapperswil, Switzerland developed a rockfall detection system called RockNet. A self ‐

organizing wireless sensor network is used for rockfall surveillance and provides real‐time warning.

Wireless sensor nodes are equipped with accelerometers, temperature and other sensors. The nodes are distributed in a rockfall prone area. If a specified number of nodes recognize vibrations an alert is released and for traffic can be stopped outside the danger zone.

Wireless Underground Sensor Networks WUSN

Akyildiz and Stuntebeck (2006) presented a good overview of  Wireless Underground Sensor Networks WUSN. The current research challenges are addressed and examples of  applications of  landslide monitoring (soil movement) and mining (high –sensitivity microphone for rescue purposes). WUSN have a potential to be used for landslide monitoring in the future. WUSN are buried and therewith solutions for power conservation, charging methods as well as communication between nodes have to be found first.

(26)

Conclusion and future needs

This extensive literature review leads to the following main conclusions:

 There is a lack of appropriate warning systems. More warning system prototypes have to be developed and implemented in practice.

 In the future, landslide warning systems will play an important role to avoid loss and damages.

 Several attempts have been made to predict landslides by monitoring different parameters. These methods are summarized in this review. The inverse velocity method (Fukuzono 1985, Rose and Hungr 2007) has proven to be applicable in practice and is a relative simple approach to predict time to failure.

 There is a trend towards multi‐parameter based warning forecasts. The combination of a few independent indicators has the advantage to make more reliable predictions.

 Wireless Sensor Networks are expected to be the technological innovation towards low‐cost landslide warning systems.

 Landslides are very complex. In order to make forecasts, a careful assessment of the specific site is essential. The selection of  appropriate monitoring parameters is crucial for a reliable forecast. Appropriate parameters are site‐ and trigger‐specific. The selection has to be done

by geological and geotechnical experts.

Role of Geotechnical Experts

Landslides are a very complex phenomenon and controlled by many variable site specific conditions. This makes it impossible to have an overall technique to predict landslides. Therefore, warning systems have to be adapted to local conditions. Geotechnical experts are required to select the appropriate indicators. To make the systems applicable to different sites, the sensors should be scalable and alerting algorithms have to be adjustable.

Selection of appropriate monitoring parameter 

Deformation measurements as indicator for an emerging landslide can be very useful. However, they require knowledge of  the deformation behavior of  the material. Depending on the material properties (e.g. rock or soil, pore‐pressures, water content, permeability, temperature…),

deformations for a given stress state can vary by several magnitudes.

Soils with a plastic behavior can be conveniently surveyed by displacement observation. Stiff  materials experience relatively small deformations compared to the load. According deformation measurements are challenging. Special attention has to be paid to brittle materials, where deformation prior fracture is very small. For brittle materials other indicators (e.g. pore‐water

pressure, water content, acoustic emission…) might be more useful.

Intensive rainfall has a more significant influence near the slope surface. Monitoring rainfall for landslide prediction makes often more sense, if  the expected landslide is shallow rather than for deep seated slip surfaces. Nevertheless, karsts can act as a direct link between surface conditions and deeper layers.

(27)

Next steps towards reliable forecasts

A major outcome of  the International Decade for Natural Disaster Reduction was the recommendation of  early warning systems for loss reduction. As mentioned in the Report on Early Warning Capabilities for Geological Hazards of the United Nations (Hamilton 1997) and reiterated in the UN Global Survey of Early Warning Systems (UN 2006), there is still a lack of appropriate warning systems for landslides. Warning systems should be affordable, especially for third world countries who suffer the most under these hazards. Until today, such systems do not exist (Baehr 2004).

Several attempts have been made to predict landslides. Currently research teams develop complete landslide warning systems from measurement to alarm release. Many attempts make use of Wireless Sensor Networks WSN (Arnhardt et al. 2007, Sheth et al.2007, Terzis et al. 2006, Thuro et al. 2007, Glade et al. 2007). Most of  these systems are still under development and need to be tested in the field. Nevertheless, there is a strong need in implementation and building more warning system prototypes in different parts of  the world to make landslide hazards predictable and help us to live with natural hazards.

(28)

References

Aguado, L.E., O'Driscoll, C. Xia, P., Nurutdinov, K., Hill, C. and O'Breine, P. 2006. A Low Cost, Low  Power Galileo/GPS Positioning System for Monitoring Landslides. Navitec October 2006. http://www.ggphi.eu/monitoring_landslides.pdf 

Akyildiz, I.F. and Stuntebek, E.P. 2006. Wireless underground sensor networks: Research challenges. Ad Hoc Networks 4 (2006) pp.669–686.

http://www.ece.gatech.edu/research/labs/bwn/underground.pdf 

Antonello, G., Casagli, N., Farina, P., Leva, D., Nico, G., Sieber, A.J. and Tarchi, D. 2004. Ground based  SAR interferometry  for monitoring mass movements. Landslides (2004) 1: 21–28.

DOI: 10.1007/s10346‐00300096

Arnhardt, C., Asch, K., Azzam, R, Bill, R., Fernandez‐Steeger, T.M., Homfeld, S.D., Kallash, A.,

Niemeyer, F, Ritter, H., Toloczyki, M., and Walter, K. 2007.Sensor based Landslide Early 

Warning System ‐SLEWS. Development of a geoservice infrastructure as basis for early warning systems for landslides by integration of real time sensors. GEOTECHNOLOGIEN Science Report. Early Warning Systems in Earth Management. Kick‐Off Meeting 10 October 2007 Technical

University Karlsruhe, pp.75 ‐88.

http://www.slews.de/Development%20of%20a%20geoservice%20infrastructure%20as%20basi s%20for%20early%20warning%20systems%20for%20landslides.pdf 

http://www.geotechnologien.de/Download/pdf/Science_Report/SR10.pdf  Azimi et al. 1988.Time prediction method of  Azimi et al.In: Landslide News 4:pp.11. Baehr, H.‐P. 2004.Fruehwarnsysteme im Erdmanagement . Geotechnologien ‐Ein

geowissenschaftliches FuE‐Programm vom Bundesministerium für Bildung und Forschung

(BMBF) und der Deutschen Forschungsgemeinschaft (DFG). http://www.geotechnologien.de/oeffentlichkeit/oeffent6.html

Baum, R.L., Godt., J.W., Harp, E.L., McKenna, J.P., and McMullen, S.R. 2005.Early warning of  landslides for rail traffic between Seattle and Everett, Washington, USA. Landslide Risk

Management ‐Hungr, Fell, Couture & Eberhardt (eds), ©2005 Taylor & Francis Group, London,

pp. 731‐740.

ISBN: 04 1538 043 X

Clark, A.R., Moore, R. and Palmer, I.S. 1996.Slope monitoring and early warning systems: Application to coastal landslides on the south and east coast of England, UK . Landslides, Senneset (ed.) © 1996 Balkema, Rotterdam.

(29)

Crosta, G.B. and Agliardi, F. 2002.How to obtain alert velocity thresholds for large rockslides. Physics and Chemistry of the Earth 27 (2002) 1557–1565.

Cruden, D.M., and Varnes, D.J. 1996.Landslide Types and Processes. In: Landslides: Investigation and Mitigation, A.K. Turner and R.L. Schuster (eds.), Special Report 247, Transportation Research Board, pp. 36‐75.

ISBN: 0‐309‐06151‐2

http://www.geonet.org.nz/images/landslide/landslide_glossary_classification_lge.gif 

Dixon , N. and Spriggs, M. 2007.Quantification of slope displacement rates using acoustic emission monitoring. Can. Geotech. J 44: 966‐976.

Froese, C.R., Murray, C., Cavers, D.S., Anderson,, W.S. and Bidwell, A.K. 2005.Development and  implementation of a warning system for the South Peak of Turtle Mountain. Landslide Risk Management ‐Hungr, Fell, Couture &Eberhardt (eds) ©2005 Taylor & Francis Group, London,.

ISBN 04 1538 043 X

Froese, C.R., Carter, G., Langenberg, W. and Moreno, F. 2006.Emergency Response Planning for a Second Catastrophic rock Slide at Turtle Mountain, Alberta. 1st Speciality Conference on Disaster Mitigation, Calgary, Alberta, Canada, 23‐26 May 2006.

http://www.ags.gov.ab.ca/activities/conferences/Emergency_Response_Turtle_Mountain.pdf  Fukuzono, T., Moriwaki, H., Inokuchi, T., Maki, M., Iwanami, K., Misumi, R., Takami,S., and Shikoku,T.

Landslide Disaster Prediction Support System based on Web GIS. . http://gisws.media.osaka‐cu.ac.jp/gisideas04/viewpaper.php?id=40

Fukuzono T. 1985. A new method  for  predicting the failure time of a slope. Proceedings of the 4th International Conference and Field Workshop on Landslides, Tokyo. Tokyo University Press, pp.145‐150.

Fukuzono T. 1990.Recent studies on time prediction of slope failure. Landslide News 4: pp.9–12. http://www.landslide‐soc.org/publications/lnews/04/0409.htm

(30)

Guzzetti, F., Peruccacci, S., and Rossi, M. 2005.RISK  AWARE Definition of critical threshold  for  different scenarios (WP 1.16).

http://www.smr.arpa.emr.it/riskaware/get.php?file=Report_WP1.16.pdf 

Hamilton R. 1997.Report on Early Warning Capabilities for Geological Hazards. IDNDR Secretariat, Geneva, October 1997.

http://www.unisdr.org/ppew/whats‐ew/pdf/reportonewcapabilitiesforgeological

hazards.pdf 

Hayashi et al. 1988.Time prediction method of  Azimi et al.. In Landslide News 4: pp.10.

Jurich, D.M. and Miller, R. J. 1987. Acoustic Monitoring of Landslides. Transportation Research Record No. 1119, Geotechnology, 1987, pp. 30‐38.

ISBN: 0‐309044723

Kawamura, K. 1985.Methodology  for landslide prediction. In Proceedings of the 11th International Conference on Soil Mechanics and Foundation Engineering, San Francisco, Calif., 12– 16 Aug. 1985. Edited by Publications Committee of the ICSMFE. A.A. Balkema, Rotterdam, The

Netherlands. Vol. 3, pp. 1155–1158.

Keefer, D.K., Wilson, R.C., Mark, R.K., Brabb, E.E., Brown, W.M., Ellen, S.D., Harp, E.L., Wieczorek, G.F., Alger, C.S., and Zatkin, R.S. 1987.Real time Landslide Warning During Heavy Rainfall . In: Science, Vol. 238: 921‐925.

Kousteni, A., Hill,R., Dixon, N. and Kavanagh, J. 1999. Acoustic emission technique for monitoring soil  and rock slope instability . In: Slope Stability Engineering, Yagi, Yamagami & Jiang, Balkema, Rotterdam.

ISBN: 90 5809 079 5

O'Connor, K.M. and Dowding, C.H. 1999.Monitoring Soil Deformation. In: GeoMeasurements by Pulsing TDR Cables and Probes, CRC Press, Boca Raton, Florida.

ISBN: 0‐8493‐0586‐1

Ortigao B. 2000.RioWatch: The Rio de Janeiro landslide watch. MonoSys Guide to Monitoring Quarter 1 2000.

http://www.terratek.com.br/downloads/Geotechnical%20Engineering%20papers%20English/2 000%20Ortigao%20on%20Landslide%20Watch%20Guide%20to%20Monitoring.pdf 

Ortigao et al.RioWatch 2001: The Rio de Janeiro landslide alarm system.

http://www.nucleocinco.com.br/terratek/downloads/eng_papers/2001%20Rio%20Watch%20 Hong%20Kong.pdf 

(31)

Read, R.S., Langenberg, W., Cruden et al. 2005.Frank Slide a century later: the Turtle Mountain monitoring project.In Landslide Risk Management ‐Hungr, Fell, Couture & Eberhardt (eds),

Taylor & Francis Group, London, ISBN 04 1538 043 X.

Rose, N.D. and Hungr,O. 2007.Forecasting potential rock slope failure in open pit mines using the inversevelocity method . International Journal of Rock Mechanics & Mining Sciences 44 (2007) 308–320.

Saito, M. 1965.Forecasting the time of occurrence of slope failure. In Proceedings of the 6th International Conference on Soil Mechanics and Foundation Engineering, Montréal, Que., University of Toronto Press, Toronto, Ont. Vol. 2, 537‐542.

Saito, M. 1969.Forecasting time of slope failure by tertiary creep. In Proceedings of the 7th

International Conference on Soil Mechanics and Foundation Engineering, Mexico City, 25–29 Aug. A.A. Balkema, Rotterdam, The Netherlands. Vol. 2, 677–683.

Sheth, A., Thekkath, C.A., Metha, P., Tejaswi, K., Parekh, C., Singh, T.N., and Desai, U.B. 2007.

Senslide: a distributed landslide prediction system. ACM SIGOPS Operating Systems Review, Vol. 41, Issue 2 (April 2007).

ISSN: 0163‐5980

http://doi.acm.org/10.1145/1243418.1243428

Soeters, R. and van Westen, C.J. 1996.Slope instability recognition, analysis and zonation. In: Landslides: Investigation and Mitigation, A.K. Turner and R.L. Schuster (eds.), Special Report 247, Transportation Research Board, pp. 129‐177.

Stewart, R., Bland, H., Thurston, J., and Hall, K. 2004.The surface microseismic monitoring system on Turtle Mountain, Alberta. CSEG RECORDER November 2004.

http://www.cseg.ca/publications/recorder/2004/11nov/nov04‐turtle‐mountain.pdf 

Terzis, A., Anandarajah, A., Moore, K., and Wang, I‐J., 2006.Slip Surface Localization in Wireless

Sensor Networks for Lanslide Prediction. IPSN’06, April 19–21, 2006, Nashville, Tennessee, USA. http://www.cs.jhu.edu/~terzis/landslide‐camera.pdf 

Figure

Updating...

References

Updating...

Related subjects :