advanced seismic slope stability analysis

8
Landslides (2013) 10:729736 DOI 10.1007/s10346-012-0360-6 Received: 4 June 2012 Accepted: 2 October 2012 Published online: 19 October 2012 © Springer-Verlag Berlin Heidelberg 2012 Mihail Garevski I Zeljko Zugic I Vlatko Sesov Advanced seismic slope stability analysis Abstract The objective of this study was to present an advanced methodology for assessing seismic slope stability by taking into account the uncertainties related to the main input parameters. The methodology was applied on a real landslide in order to show the advantages of using the proposed procedure and establish the baseline trends of dynamic response and calculated permanent seismic displacements. It involves the following steps: preliminary analysis, probabilistic static and seismic factor of safety analysis, and permanent seismic displacement analysis. Estimating post- failure maximum seismic deformation of landslide mass and sounding properties is the most important part of this study. It involves both Newmark sliding block method and continuum mechanics approach, applied for characteristic set of input values in order to have more accurate assessment of slope performance and determine the relative importance of input parameters. The results of the analysis showed the benefits of using the proposed step-by-step methodology. The obtained difference in the results between the two methods depends strongly on the set input data for a particular analysis. Keywords Seismic slope stability . Uncertainties . Probabilistic approach . Permanent displacement . Sensitivity analysis Introduction The uncertainties related to input parameters that one faces while analyzing seismic slope stability make sensitivity and reliability methods very suitable for application in assessment of active and potential landslides as well as evaluation of design solutions projected to prevent sliding processes. There are different methods for probabilistic and pseudo-probabilistic seismic slope analysis and only a limited number of them deal with seismic permanent displacement. Most of the displace- ment-based methods are adopted and verified for regions of high seismicity. They are based on significant seismic data and require a large number of simulations (Rathje and Saygili 2009; Bray and Rathje 1998). Some of them require advanced knowl- edge of the probabilistic theory (Kim 2001). Nowadays, most of the probabilistic displacement assessment procedures are based on the sliding block (Newmark 1965) procedure. Continuum modelling is still not widely used due to its complexity and time required for analysis (Rathje and Bray 2000). 2D model assumption is quite common even for important projects. Therefore, in addition to all other uncertainties, one of the biggest sources of uncertainty is the slope model. Just a few researchers take into account the uncertainties related to soil properties (Rathje and Saygili 2009; Murphy and Mankelow 2004; Kim 2001). However, there is still no consensus in engi- neering community and therefore probabilistic methods are still not widely used in practice. The challenge is to develop a methodology that will be complex enough to take into account the uncertainties associated with the main input parameters and simple enough to provide results within a reasonable time, without complex probabilistic computation, being applicable in case of having an average amount and quality of seismic and geotechnical data. Developing a procedure to perform seismic slope stability analysis, learning about the relative importance of input parameters and comparison of results obtained using different methods for seismic slope deformation assessment, have been the objectives of this study. Motivation for research According to both general and preliminary designs of the mo- torway running from Belgrade to the South Adriatic, i.e. E-763, at the exit from Belgradethe capital of Serbia, the road facility corridor is located on the right bank of Sava river, at the mean- dering apex (Fig. 1). Along a length of 3 km, it crosses Umka- Duboko, the large active landslide with a depth of 1026 m, with dominant presence of Marly clays, covering an area of 1.8 km 2 (Fig. 2). Having in mind the importance of the project and the increased risk in the landslide area, extensive geotechnical inves- tigations were conducted in 2005. Some of the results important for this paper are summarised in Fig. 3. where one can notice quite a large scatter of soil parameters. They are related for the most critical slope Duboko(Fig. 4). Based on analyses and a series of iterative procedures, it has been decided to widen up the Sava river channel on the left bank, build a parallel protectiveretaining structure made of crushed stone on the right bank, and set the motorway road base on a high embankment (made of dredged sand) behind the mentioned structure (Fig. 5). In addi- tion, it has been envisioned to carry out works for drainage, levelling and aforestation of the unstable terrains. Considering that a motorway of such an importance is going to be built above the presented landslide, the necessity for assessing the static and seismic performance of such repair solution using advanced methods became obvious. The landslide is located in a seismic active area and the impact of possible seismic movement of the landslide on the structure of the future motorway is a very important issue of this project. Proposed methodology The general rule which is followed while defining the methodology is going from simple to complex analysis and not to combine the complex probabilistic framework with computation demanding seismic analysis. Therefore, it includes three steps, by increasing complexity and the final goal is to assess the slope permanent seismic displacement. The proposed methodology can be imple- mented in any commercial software. For analysis of the case study presented in this paper, commercially available programs SLOPE/ W, QUAKE/W (Krahn 2004) and FLAC (Itasca Consulting Group 2000) are used. The first step (Fig. 6) is evaluating the most critical failure mechanism by solving equilibrium and constitutive equations for full solution of the coupled stress/displacement state in the slope corresponding to the point of instability while performing a series of calculations with different prop- erties. It should be taken into account that, in some cases, the Landslides 10 & (2013) 729 Original Paper

Upload: vlatko

Post on 23-Dec-2016

232 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Advanced seismic slope stability analysis

Landslides (2013) 10:729–736DOI 10.1007/s10346-012-0360-6Received: 4 June 2012Accepted: 2 October 2012Published online: 19 October 2012© Springer-Verlag Berlin Heidelberg 2012

Mihail Garevski I Zeljko Zugic I Vlatko Sesov

Advanced seismic slope stability analysis

Abstract The objective of this study was to present an advancedmethodology for assessing seismic slope stability by taking intoaccount the uncertainties related to the main input parameters.The methodology was applied on a real landslide in order to showthe advantages of using the proposed procedure and establish thebaseline trends of dynamic response and calculated permanentseismic displacements. It involves the following steps: preliminaryanalysis, probabilistic static and seismic factor of safety analysis,and permanent seismic displacement analysis. Estimating post-failure maximum seismic deformation of landslide mass andsounding properties is the most important part of this study. Itinvolves both Newmark sliding block method and continuummechanics approach, applied for characteristic set of input valuesin order to have more accurate assessment of slope performanceand determine the relative importance of input parameters. Theresults of the analysis showed the benefits of using the proposedstep-by-step methodology. The obtained difference in the resultsbetween the two methods depends strongly on the set input datafor a particular analysis.

Keywords Seismic slope stability . Uncertainties . Probabilisticapproach . Permanent displacement . Sensitivity analysis

IntroductionThe uncertainties related to input parameters that one faceswhile analyzing seismic slope stability make sensitivity andreliability methods very suitable for application in assessmentof active and potential landslides as well as evaluation ofdesign solutions projected to prevent sliding processes. Thereare different methods for probabilistic and pseudo-probabilisticseismic slope analysis and only a limited number of them dealwith seismic permanent displacement. Most of the displace-ment-based methods are adopted and verified for regions ofhigh seismicity. They are based on significant seismic data andrequire a large number of simulations (Rathje and Saygili 2009;Bray and Rathje 1998). Some of them require advanced knowl-edge of the probabilistic theory (Kim 2001). Nowadays, most ofthe probabilistic displacement assessment procedures are basedon the sliding block (Newmark 1965) procedure. Continuummodelling is still not widely used due to its complexity andtime required for analysis (Rathje and Bray 2000). 2D modelassumption is quite common even for important projects.Therefore, in addition to all other uncertainties, one of thebiggest sources of uncertainty is the slope model. Just a fewresearchers take into account the uncertainties related to soilproperties (Rathje and Saygili 2009; Murphy and Mankelow2004; Kim 2001). However, there is still no consensus in engi-neering community and therefore probabilistic methods are stillnot widely used in practice. The challenge is to develop amethodology that will be complex enough to take into accountthe uncertainties associated with the main input parametersand simple enough to provide results within a reasonable time,without complex probabilistic computation, being applicable in

case of having an average amount and quality of seismic andgeotechnical data. Developing a procedure to perform seismicslope stability analysis, learning about the relative importanceof input parameters and comparison of results obtained usingdifferent methods for seismic slope deformation assessment,have been the objectives of this study.

Motivation for researchAccording to both general and preliminary designs of the mo-torway running from Belgrade to the South Adriatic, i.e. E-763, atthe exit from Belgrade—the capital of Serbia, the road facilitycorridor is located on the right bank of Sava river, at the mean-dering apex (Fig. 1). Along a length of 3 km, it crosses “Umka-Duboko”, the large active landslide with a depth of 10–26 m, withdominant presence of Marly clays, covering an area of 1.8 km2

(Fig. 2). Having in mind the importance of the project and theincreased risk in the landslide area, extensive geotechnical inves-tigations were conducted in 2005. Some of the results importantfor this paper are summarised in Fig. 3. where one can noticequite a large scatter of soil parameters. They are related for themost critical slope “Duboko” (Fig. 4). Based on analyses and aseries of iterative procedures, it has been decided to widen up theSava river channel on the left bank, build a parallel protective–retaining structure made of crushed stone on the right bank, andset the motorway road base on a high embankment (made ofdredged sand) behind the mentioned structure (Fig. 5). In addi-tion, it has been envisioned to carry out works for drainage,levelling and aforestation of the unstable terrains. Consideringthat a motorway of such an importance is going to be built abovethe presented landslide, the necessity for assessing the static andseismic performance of such repair solution using advancedmethods became obvious. The landslide is located in a seismicactive area and the impact of possible seismic movement of thelandslide on the structure of the future motorway is a veryimportant issue of this project.

Proposed methodologyThe general rule which is followed while defining the methodologyis going from simple to complex analysis and not to combine thecomplex probabilistic framework with computation demandingseismic analysis. Therefore, it includes three steps, by increasingcomplexity and the final goal is to assess the slope permanentseismic displacement. The proposed methodology can be imple-mented in any commercial software. For analysis of the case studypresented in this paper, commercially available programs SLOPE/W, QUAKE/W (Krahn 2004) and FLAC (Itasca Consulting Group2000) are used.

The first step (Fig. 6) is evaluating the most criticalfailure mechanism by solving equilibrium and constitutiveequations for full solution of the coupled stress/displacementstate in the slope corresponding to the point of instabilitywhile performing a series of calculations with different prop-erties. It should be taken into account that, in some cases, the

Landslides 10 & (2013) 729

Original Paper

Page 2: Advanced seismic slope stability analysis

mechanism for horizontal dynamic loads can be differentfrom the failure mode induced by static loading. There arestrong reasons why such an analysis is recommended in theearly stage of analysis of complex slopes like possibility todetect simultaneous failures, soil structure interaction, etc. Allof these are explained in literature (Dawson et al. 1999). Forpreliminary analysis, deterministic input parameters (best es-timate values) should be used. Nevertheless, it is recommen-ded to have an insight into the main uncertain parametersand their upper and lower bound values because this isnecessary for the next two steps. The second step is theprobabilistic limit equilibrium factor of safety analysis usingthe Monte Carlo simulation for the sliding mechanism definedas the most critical in the preliminary analysis. One of themain recommendations in this stage analysis is sampling thesoil properties of the sliding mass in several ways. The soilproperties can be sampled: for each slice along the slip sur-face, at a specified distance along the slip surface, or onlyonce for each soil for the entire slip surface. This could be avery important issue for huge landslides and according togeotechnical investigations, different sampling can be used.The obtained results (first of all, the probability of failure)provide an answer as to whether there is a necessity to

perform a displacement analysis. As the third step, slidingdisplacement analysis is proposed to be made using bothsliding block (Newmark 1965) and continuum mechanics ap-proach. The first benefit of this analysis is comparison of theresults by using different approaches. Another advantage isrelated to the time necessary to perform sensitivity analysisfor both methods. For many slope cases, the Newmark slidingblock procedure is just an index of slope behaviour, but veryuseful in the case when analysis should be run many timesfor various sets of input values. The results from the New-mark sliding block analysis help us detect which of the un-certain parameters are less important so that they can beexcluded from the continuum approach sensitivity analysis.Plenty of time can be saved if consideration is employed.The analysis, similar to the one performed by Chugh andStark (2006), is proposed to be repeated for a predefined setof inputs in order to perform sensitivity analysis. Sensitivityanalysis is a technique for systematically changing parametersin a model to determine the effects of such changes. Tornadodiagrams, also called tornado plots or tornado charts, areused for visualisation of the results from the sensitivity anal-ysis. This technique was applied in earthquake engineering byPorter et al. (2002). The first input is then set to its bestestimate value and the process is repeated for the next inputin order to determine the swing associated with the variabilityof that input. One can then rank the input variables accordingto their swing. A larger swing reflects a more important inputuncertainty.

Application of methodologyBefore the application of the above-described procedure, pre-vious geotechnical analyses and selection of design parameterswere consulted. For the established landslide models, along 15geotechnical profiles, the natural stability of the slope inter-cepted by the landslip has been analysed by Mitrovic andJelisavac (2006) in a recurrent mode; for the design condi-tions of equilibrium Fs01, a tentative angle has been lookedfor at average water of Sava the river and maximum watersaturation of the slope. The laboratory residual resistance andtentative angles for the design conditions of equilibrium arecorrelative. It has been recommended for the purpose ofchecking up the effect of repair to utilise the design residualparameters for the Duboko landslide which are φr011 degreesand cr00 kPa. In the past, analysis was performed on all the

Fig. 1 Location map of study area

Fig. 2 Umka–Duboko landslide

Original Paper

Landslides 10 & (2013)730

Page 3: Advanced seismic slope stability analysis

defined profiles. It was detected that cross-section 12–12 of the“Duboko” slope located at 9+611 m (Fig. 5) gives the lowestfactor of safety wherefore the proposed methodology wasapplied on this particular slope. Slope “Duboko” is of afrontal type, with a length of 1.45 km along the river, whereasalong the slope, it is 300 m long, thus amounting approxi-mately to 40 ha (Fig. 2). The volume of the landslide is6,000,000 m3 with an average depth of 15 m. So far, observa-tions have been made at the installed 19 inclinometers, 15piezometers, and 3 exploratory shafts. In accordance withthe morphology and the sliding mechanism, three blocks havebeen singled out: D, E and F. The length of the blocks alongthe river is 350–550 m, whereas the maximum depths ofsliding were recorded in block D—up to 25 m, in block E—up to 16.5 m, and in block F—up to 10 m. The analyses werefocused on block D presented in Figs. 3 and 5. The soilproperties evaluated from the geological investigations, thedesigned repair measures and iterative calculations are pre-sented in Table 1. More information can be found in thereferences (Mitrovic and Jelisavac 2006).

Definition of uncertain input parametersThe most of important soil parameters are given in Table 1.The uncertainties of soil properties, water level and slopeangle are defined by probability distribution functions (Ta-ble 2). The mean values are determined according to themeasurements from the site and laboratory tests. The stan-dard deviations and types of probability distributions aredefined according to the detected scatter of the data andrecommendations presented by other authors (Jones et al.2002; Lacasse and Nadim 1996) except the value of the waterlevel which is adopted according to the available hydro-geo-logical measurements. Geometry changing is achieved simplyby moving the material boundary up and down in order toachieve the lower and upper bound of the slip surface anglein the observed models.

Definition of seismic inputOn the basis of the available seismic data from the seismichazard maps, the location of Umka–Duboko is within thezone of maximum intensity of VII MSK PGA00.1 g, T0475 years. Figure 7 shows the hazard curve approximatedaccording to seismic maps and historical data. More detailsabout seismic hazard assessment for this region are given in

Fig. 5 Remedial design solution—the most critical engineering geologic profile 12(block D)

Newmark method (Slope/W, Quake/W)

Preliminary analysis (FLAC/Slope) Verification of most critical sliding mechanisms

Probabilistic (Monte Carlo Simulation)

FOS analysis (SLOPE/W)

Sensitivity analysis of

Slope seismic permanent displacement

Continuum modeling

(FLAC 5.0)

Fig. 6 Proposed methodology for seismic slope stability assessment

Fig. 3 Parameters obtained byinvestigation works and previousanalysis

Fig. 4 Slope Duboko

Landslides 10 & (2013) 731

Page 4: Advanced seismic slope stability analysis

literature (Mitrovic and Jelisavac 2006). Considering there isno valuable data, a set of accelerograms for dynamic anal-yses should be spectrum compatible, meaning that the averageresponse spectrum computed from all the accelerogramsshould match a target spectrum prescribed by a seismiccode Eurocode 8 EN 1998–5:2005 (2005) within a certaintolerance, over a specified range of periods. Spectral com-patibility is not simply achieved when recorded accelero-grams are used; however, it is an important requirementin order to avoid using records that are inconsistent withcode prescriptions.

In this work, the 20 accelerograms to be adopted fordynamic analyses were selected from strong motion recorddatabases (Fig. 8), by imposing the constraint of spectrumcompatibility with a code-based target spectrum. The spec-tral shape of Eurocode 8 type 2 (Fig. 9) was used (M<5.5),considering ground type A (rock site Vs30≥800 m/s). Thedeconvolved accelerograms are calculated by using the com-puter program SHAKE91 (Idriss and Sun 1992) based onone-dimensional wave propagation theory and applied tothe bottom of the model.

Definition of inputs for tornado diagramsIn order to construct a tornado diagram, the range of seismicevents needed to be defined from seismic hazard curve (Fig. 7).Considering that the serviceability of the motorway has beenadapted to 100 years, the used values of seismic intensity havebeen: 10 % in 100 years corresponding to T043 years, PGA00.04 g50 % in 100 years corresponding to T0144 years PGA00.064 g90 % in 100 years corresponding to T0950 years, PGA00.13 g(Table 3).

For defining the input for seismic records, all therecords have been scaled to the best estimation PGA valueof 0.064 g. Having 20 records and 20 calculated displace-ments simply by sorting in a decreasing array, we havedefined 2/20 value as a 90 % percentile and 18/20 as 10 %value. These are record no. 3 and record no. 17. (see Table 3).The best estimation (10th largest displacement) has beenobtained for record no. 10 and has been used as the bestestimation value for ground motion record. A similar ap-proach was used by Porter et al. (2002) related to damagefactor in structural engineering. Another remark should begiven that the meaning of the upper bound and lower boundvalues of the variables in Table 3, addresses the upper boundvalues of the calculated displacement. All other inputs pres-ent in Table 3 are determined as mean values and 90 and10 % percentiles of the variables defined in Table 2. Uncer-tainties of the mass and damping parameters are not includ-ed. It was detected that mass variation has no significantinfluence. Damping variation was excluded because different

Table 1 Soil strength and stiffness parameters

Soil Density[kg/m3]

Material strength Stiffness parametersGmax/damping curve(shear and normalstiffness for interface)

Name Class c′ [kPa] ø′ [deg.]

Colluvium Marly clay 1,790.0 30 23.0 6,200 kPa (Hardin and Black 1968)

Rock Marlstone 1,890.0 290 25.2 1.2 GPa—constant

Mini dam Crushed stone 2,600.0 0 45.0 13,000 kPa (Kokusho and Esachi 1981)

Embankment Refuelled sand 1,700.0 0 25.0 8,400 kPa (Kokusho and Esachi 1981)

Slip surface Marly clay 1,790.0 0 11 Kn03×10e5kPa/m

Ks011×10e3kPa/m

Table 2 Definition of uncertain parameters

Parameter Meanvalue (μ)

Standarddeviation (σ)

Probabilitydistribution

Residual shearstrength alonginterfaces(degrees)

11 2 Normal

Water level(metres over sea)

71.7 m 1.25 m Normal

Shear modulus(Gmax) ofcolluvium (kPa)

6,200 1,600 Normal

Slope angle,θ (degrees)

4.57 1 Lognormal

0.0000

0.0001

0.0010

0.0100

0.1000

1.0000

0.01 0.1 1

PGA, g

Exci

danc

e fr

eque

ncy

1/ye

ar

Fig. 7 Site hazard curve

Original Paper

Landslides 10 & (2013)732

Page 5: Advanced seismic slope stability analysis

types of damping are considered in FLAC and Quake/Wanalysis, therefore the comparison of results won’t bestraightforward.

Results of analysisDuring the application of proposed methodology, numerousanalysis and different variants were observed. The most sig-nificant outcomes of this study are presented below.

Preliminary analysis of the above presented slope hasbeen conducted using code Flac/Slope 5.0. The deterministicmean values of the input parameters listed in Table 2 havebeen used. For assessment of the seismic behaviour of theslope, horizontal force of 0.05 g (obtained as 0.5 of the 0.1 g)for a return period of T0100 years, the most critical slidingmechanism for horizontal force is the lifting of the roadembankment presented in Fig. 10. This “failure mode” wasthe lowest safety factor obtained (around 0.9), wherefore thepresented sliding mechanism was to be followed in furtheranalysis. The probabilistic factor of safety analysis involvedthe distribution of input values from Table 3 and also theconstant seismic force of 0.05 g for seismic analysis, same as

in the preliminary analysis. Static and pseudo-static seismicanalysis has been conducted using slope/W. Stability factoranalysis proposed by Krahn (2004) was done for the assess-ment of the static slope stability. For seismic (pseudostatic)is employed the GLEM formulation developed by Fredlund etal. (1981) that encompasses the key elements of all the other meth-ods. It is based on two factors of safety equations and allows for arange of interslice shear–normal force conditions. The model ispresented in Fig. 11. Sampling each slice for shear strength of a slidingsurface is employed while cohesion is zero for all slices. Theresults of the probabilistic analysis are presented in Table 4.The mean value of the seismic factor of safety for 2,000simulations of 1.05 suggests that the available slope shearstrength is near equilibrium with shear forces so that seismic

0 0.5 1 1.5 20

1

2

3

4

5

Period (s)

Res

pons

e Sp

ectr

a (m

/s2 )

Fig. 8 All 20 selected records

Fig. 9 Average value of 20. selectedrecords

Table 3 Input values for Tornado diagrams

Uncertainvariables:

Upperbound90 %percentile

Bestestimation50 %percentile

Lowerbound10 %percentile

Residual shearstrength value(degrees)

8.5 11 13.5

Water level ofSava river(meters over sea)

73.3 71.7 70.1

Slope angle(degrees)

5.9 4.57 3.4

Gmax—colluviumshear modulus(kPa)

3,600 6,200 8,300

Ground motionrecord

Recordno. 3

Recordno. 10

Recordno. 17

Peak groundaccelerationT0100 years

0.13 g 0.064 g 0.04 g

Landslides 10 & (2013) 733

Page 6: Advanced seismic slope stability analysis

instabilities are possible. The obtained value of the probabilityof failure value that is actually a percent of simulations thatgives a factor of safety of less than 1.0 is 0.08.

Permanent seismic displacement sensitivity analysis Having inmind that the accent of the study was on the step-by-stepprocedure and the relative importance of input variables,some important issues about modelling, especially about thecontinuum mechanic approach, are not motioned. Most ofthem can be found in literature, namely Chugh and Stark(2006), Itasca Consulting Group (2000) for the continuummodelling approach and Krahn (2004) for the sliding blockprocedure. The soil was modelled using the linear elasticperfectly plastic constitutive law, with the Mohr–Coulombyield criterion and the non-associated flow rule. The geotech-nical model was built step-by-step, starting from a simplemodel and gradually increasing the degree of complexity afterconsistent and stable results were obtained for the previousphase of the analysis. The following damping parameters werechosen: 0.01 % Rayleigh damping centred at 1 Hz, plus thehysteretic damping option included in FLAC. It is importantto be noted that, in the FLAC analysis, 14 points were fol-lowed. The results for two characteristic points are presentedhere. The first one (M1 centre of the mass of the slope) inorder to be compared with the Newmark sliding block method andanother (M2 middle of the motorway) valuable for assessment ofpossible damage on the designed road structure above (Fig. 13).

The range of displacements obtained by use of the FLAC analysis(1–15.5 cm) is less scattered beside the fact that it is more accuratethan the one determined by the slide block analysis (0–60 cm). Thesummary of the results is given in Table 5. The best estimate valuesagree very well with the results from Chugh and Stark (2006) givingapproximately two times larger displacement by the sliding block

method. By comparing the upper bound values, the differenceincreases—the displacements obtained with the Newmark methodare four times larger.

The presented results from the Newmark sliding blockanalysis (Fig. 12) can be considered as a good reference fromtwo reasons, namely (a) the repair solution does not includeany pile or shaft along the slide mass. In that case, a complexsoil structure phenomenon should include (b) a well-definedsliding surface along the material discontinuity can make thesliding block assumption reliable enough. The necessity forusing the continuum mechanics approach arises from the factthat there is a road structure above. The displacement of themiddle of the road (point M2, d03.4 cm) is a little bit smallerthan the one obtained for the centre of the mass (point M1,d03.1 cm). For the upper bound, the difference is againaround 10% (15.5–17.3 cm). These values from FLAC analysis includeboth x and y components, same as results from Table 5.

Fig. 11 Landslide “Duboko” Slope/W model

Table 4 Summary of results of probabilistic static and seismic analysis (2,000 trials)

Result Static analysisstability factor(FEM)

Seismic(T0475 years)factor of safety(GLEM)

Mean value (μ) 1.71 1.05

Standard deviation (σ) 0.09 0.104

Reliability indexβ0 (μ−1)/σ

7.914 0.568

Probability of failure 0 0.08

Fig. 10 Critical sliding mechanismdetected by preliminary analysis

Table 5 Summary of the results

Method Best estimationD (cm)

Maximal expectedD (cm)

Continuum approach(point M1)

3.1 15.5

Newmark sliding block 6 60

Original Paper

Landslides 10 & (2013)734

Page 7: Advanced seismic slope stability analysis

Obtained result represents a very good basis for damage assess-ment of the road structure due to seismic movement. The results oftime history analyses have shown that the amount of final displace-ment can be very different for different accelerograms, despite thefact that they are scaled to the same PGA. Therefore, any designprocedure relying only on a PGA-derived parameter will impreciselypredict the actual behaviour of the slope. This emphasizes the im-portance of proper selection of seismic intensity measure for perfor-mance-based design of slopes. The actual approaches developingduration depending parameters give very good results (Rathje andSaygili 2009). It was confirmed in the finding of Bray and Rathje(1998) that tje Newmark method can be reliable if the fundamentalperiod of the slope is sufficiently low (i.e. shallow soil deposits withT<0.2 s). Observed slope is quite shallow, therefore, the obtaineddifference in the results (two to four times larger displacementobtained by Newmark method) was expected. Having in mind thefrequency content of the employed seismic records (Fig. 9) and thatnatural period of the slope is around 0.2 s, one can expect thatNewmark sliding block will underestimate the seismic deformation,but it did not happen, probably due to small yield acceleration of theactive landslide analysed with residual soil parameters. Comparisonof the results in Figs. 12 and 13 (left) show that resultsobtained by Newmark method can be unconservative in casesof very low seismic excitation (lower bound values); but inthat case, the difference in the results is insignificant from theengineering viewpoint. A relatively big result sensitivity on theshear modulus of the sliding mass has been detected, but thestiffness parameters, by their definition, should influence theslope deformation in cases of either static or dynamic loadingmore than the strength parameters. The using of simplified

methods for slope stability analyses is the reason why thisaspect was not researched intensively. The results proved that,for the low level of seismic excitation, the variability of soilproperties can have a significant effect on the slope displace-ment performance that corresponds with the results obtainedby Kim (2001).ConclusionsThe obtained values of displacement according to the CaliforniaGeological Survey classifications (Table 6) fall into the class of lowlandslide hazards (5 cm<D<15 cm); therefore, the road project isfeasible from the viewpoint of seismic performance. Designing ofstructures over active or potential landslides definitely requiresusing advanced procedures and consideration of the uncertaintiesrelated to input parameters. Permanent seismic deformation is avaluable parameter; in most cases, directly correlated with damageto the observed structure and can be evaluated by the two pre-sented methods. In this study, the Newmark method basicallyprovided the larger estimates of displacements. Those results wereexpected having in mind that observed slope is shallow and thenatural period of sliding mass is small. The reliability of Newmarkmethod results is commented in terms of other factors like level ofseismic loading, frequency content of the input motion, yield accel-eration, and sliding surface precision. It was shown that the differ-ence between the results obtained by these two methods depends onthe particular set of input data (best estimate values, upper or lowerbound). This fact emphasises the significance of performing thesensitivity analysis. Performing sensitivity analysis for a continuummodelling approach requires a lot of time but the benefits are verybig; therefore, it should be done for important projects.

Table 6 Landslide hazard criteria used by CGS

Landslide hazard Sliding displacement

Very low D<5 cm

Low 5 cm<D<15 cm

Moderate 15 cm<D<30 cm

High D>30 cm

Fig. 13 Tornado diagrams for permanent displacement obtained from continuum approach

Fig. 12 Tornado diagram—Newmark method

Landslides 10 & (2013) 735

Page 8: Advanced seismic slope stability analysis

References

Bray JD, Rathje EM (1998) Earthquake-induced displacements of solid-waste landfills. JGeotech Geoenviron Eng ASCE 124(3):242–253

Chugh A, Stark TD (2006) Permanent seismic deformation analysis of a landslide. J IntConsort Landslides 3(1):2–12

Dawson EM, Roth WH, Drescher A (1999) Slope stability analysis by strength reduction.Geotechnique 40(6):835–840

EN 1998–5:2005 (2005) Eurocode 8—design of structures for earthquake resis-tance—part 5: foundations, retaining structures and geotechnical aspects. CEN,Bruxelles

Fredlund DG, Krahn J, Pufahl DE (1981) The relationship between limit equilib-rium slope stability methods. Proc. of the 10th ICSMFE Stockholm, Sweden,vol. 3. pp 409–16

Hardin BO, Black WL (1968) Vibration modulus of normally consolidated clay: designequations and curves. J Soil Mech Found Eng Div ASCE 94(2):353–369

Idriss IM, Sun JI (1992) SHAKE91: A Computer program for conducting equivalentlinear seismic response analyses of horizontally layered soil deposits. Depart-ment of Civil and Environmental Engineering, University of California, Davis, 65pp

Itasca Consulting Group (2000) FLAC—fast Lagrangian analysis of continua. ItascaConsulting Group, Minneapolis

Jones AL, Kramer SL, Arduino P (2002) Estimation of uncertainty in geotechnicalproperties for performance-based earthquake engineering. PEER Report 2002/16, Pacific Earthquake Engineering Research Center, University of California,Berkeley

Kim J (2001) Probabilistic approach to evaluation of earthquake induced permanentdeformation of slopes. PhD dissertation. Dept. of Civil and Environmental Engineering,Univ. of California, Berkeley, California

Kokusho T, Esachi Y (1981) Cyclic triaxial test on sands and coarse materials. Proceedingsof the 10th International Conference on Soil Mechanics and Foundation Engineering,Stockholm

Krahn J (2004) Stability modelling with slope/W, dynamic modelling with quake/Wengineering methodology books, first edition GEO-SLOPE/W International, ltd

Lacasse S, Nadim F (1996) Uncertainties in characterizing soil properties. Proceedings,uncertainty in the geologic environment: from theory to practice. Geotechnical specialpublication no. 58, ASCE, vol. 1: 49–75

Mitrovic P, Jelisavac B (2006) Geotechnical conditions for E-763 motorway constructionfrom Belgrade to South Adriatic across the landslide Umka—Duboko 8. Slovenskikongres o cestah in prometu, Portorož,. Slovenia

Murphy W, Mankelow JM (2004) Obtaining probabilistic estimates of displacements on alandslide during future earthquakes. J Earthq Eng 8(1):133–157

Newmark NM (1965) Effect of earthquakes on dams and embankments. The Rankinelecture. Geotechnique 15(2):139–159

Porter KA, Beck JL, Shaikhutdinov RV (2002) Sensitivity of building loss estimates tomajor uncertain variables. Earthquake Spectra 18(4):719–743

Rathje EM, Bray JD (2000) Nonlinear coupled seismic sliding analysis of earth structures. JGeotech Geoenviron Eng 126(11):1002–1014

Rathje EM, Saygili G (2009) Probabilistic assessment of earthquake-induced slidingdisplacements of natural slopes. Bull N Z Soc Earthq Eng 42:18–27

M. Garevski : Z. Zugic ()) : V. SesovInstitute of Earthquake Engineering and Engineering Seismology, IZIIS,University “Ss. Cyril and Methodius”,73, Salvador Aljende Str. P.O. Box 101, 1000 Skopje, Macedoniae-mail: [email protected]

Original Paper

Landslides 10 & (2013)736