mechanisms of slope failure in volcanic soils during earthquakes

Published by Elsevier Science Ltd. All rights reserved12 th European Conference on Earthquake EngineeringPaper Reference 782MECHANISMS OF SLOPE FAILURE IN VOLCANIC SOILSDURING EARTHQUAKESW. Murphy 1 , J. Bommer 2 and J. M. Mankelow 3 .1The School of Earth Sciences, University of Leeds, Leeds, LS2 9JT.2 Department of Civil & Environmental EngineeringImperial College London SW7 2BU, UK3 British Geological Survey, Kingsley Dunham Centre,Keyworth, Nottingham. NG12 5GGABSTRACTThe 13 th of January 2001 earthquake resulted in considerable slope instability. The majorityof the landslides were either small volume rock and debris falls or large volume debris flows.The majority of Rock and debris falls were in cut slopes in pyroclastic ashfall deposits,especially, a unit known at the Tierra Blanca. The larger volume slope failures occurred asdebris flows, again in the Tierra Blanca, but appeared to have been the result of a complexcollapse phenomenon related to the properties of the soil. The location of these slidesappears to have been dominated by topographic and geological conditions rather than thedistance between source and site.Keywords: Landslides, debris flows, volcanic soilsINTRODUCTIONThe El Salvador earthquake of January 2001 resulted in significant loss of life and damageover a large area. The majority of the deaths were the results of two large landslides that werethe result of the main shock at 14:33 UTC (table 1). These landslides at Las Colinas (in SantaTecla) and Las Barrioleras (west of Santa Tecla) resulted in c. 540 of the 870 deaths. Thispaper discusses the debris flows triggered by the main earthquake, and addresses theimplications for the behaviour of pyroclastic ashfall deposits during strong shaking.TABLE 1.SOURCE PARAMETERS OF THE TWO MAIN EL SALVADOR EARTHQUAKES.Time Epicentre Depth Magnitude Agency13 January 200117:33:32 13.049 o N 88.660 o W 60 km M W 7.7, M S 7.8, m b 6.4 NEIC17:33:46 12.97 o N 89.13 o W 56 km M W 7.7, M S 7.8, m b 6.4 HRV17:33:30 12.868 o N 88.767 o W 60 km M W 7.7 CASC13 February 200114:22:06 13.671 o N 88.938 o W 10 km M W 6.5, M S 7.5, m b 5.5 NEIC14:22:16 13.980 o N 88.970 o W 15 km M W 6.6, M S 6.5, m b 5.5 HRV

During January and February 2001 the Central American republic of El Salvador was shakenby a succession of strong earthquakes. Two significant shocks occurred which showedmarkedly different characteristics. The first of these two events had an epicentre within thesubducted slab of the Coccos plate at a depth of approximately 60 km, while the second had afocal depth of between 10 and 15 km. The mechanisms of these two events were notablydifferent. The former was an extensional event, the latter was almost pure strike-slipdeformation. The main earthquake of 13 January generated a peak horizontal groundacceleration in a North-South direction of 1.1 g in the town of La Libertad approximately 10km from the projected fault rupture. The second large event of 13 February generated a peakhorizontal ground acceleration of 0.41g recorded at Zacatecoluca approximately 9 km fromthe projected fault rupture.Numerous landslides were triggered by the main earthquake, many of which expanded due tothe second large earthquake of 13 February. Numerous large landslides which can beclassified as debris flows (Varnes [1], Dikau [2]) affected the Balsamo Cordillera. Themajority of these slope failures occurred in natural vegetated slopes in areas underlain byolder volcanic rocks.GEOLOGYThe geology of San Salvador is entirely volcanic (Schmidt-Thomé [3]). Three broadgeological units can be identified which are the Balsamo Formation; the CuscatalanFormation and the San Salvador Formation. Of these three geological formations, the mostimportant to this study are the Pliocene-Pleistocene rocks of the Balsamo Formation and aHolocene member of the San Salvador formation called the Tierra Blanca.Figure 1. The rocks of the Balsamo Formation exposed in the head of thelandslide at Las ColinasThe Balsamo formation is composed dominantly of felsic igneous rocks. These often forminterlayered rhyolites, dacites and pyroclastic deposits and occasional palaeosols. Field

examination indicates that the rock masses formed from rocks of this part of the successionshow layers of medium thickness, which are dominantly horizontal. Discontinuities showmedium spaced. Making use of Geological Strength Index (GSI) proposed by Hoek [4], theserock masses could be split into two categories of materials. The first category is thecompetent rocks with GSI between 50 and 60. These are normally rhyolites and dacites andshow low degrees of alteration. The second group of rocks is substantially weaker showingGSI values of 20-30 and generally consists of weak pyroclastic ashfall deposits with veryclosely spaced discontinuities. Material properties of these two groups of materials aregenerally moderately to extremely STRONG, while the latter group is dominated bymaterials which are WEAK. Additionally, palaeosols formed from tropical weathering ofvolcanic materials has resulted in the formation of local aquitards. Figure 1 shows rocks ofthe Balsamo formation exposed in the rear scarp of the landslide at Las Colinas. Figure 2shows a typical type of slope failure in the Tierra Blanca.Figure 2. The Tierra Blanca exposed in aroad cutting showing characteristic rock/debris fall type failureDEBRIS FLOWS TRIGGERED BY THE EARTHQUAKEThe earthquake of 13, January 2001 resulted in the initiation of a significant number of debrisflows. The most infamous of these was the landslide at Las Colinas. However, this was notthe only such failure; debris flows occurred extensively throughout the Balsamo Cordillera.The question of whether such landslides were the results of strong shaking, topographicamplification or some related site effect was investigated.

Data on the location of landslides were collected using SPOT monospectral satellite imagerycollected on 28 January 2001 (figure 3). These data have a spatial resolution of 10m.Additional geographic data on debris flows were collected by field survey. The majority ofthe debris flows was large, and was easily observed on SPOT data.SPATIAL DISTRIBUTION OF LANDSLIDESThe majority of landslides of all types occurred in the Balsamo Cordillera. The ridge itselfwas strongly affected either by rockfall and debris fall type landslides that resulted in seriousblockages due to loss of a section of the road, or, because of landslide debris blocking thecarriageway. Due to the metastable nature of the material that failed it, was not uncommon tofind the road surface buried in approximately 0.1 to 0.2 m of silty sand.Landslide densities in the order 100-200 failures per kilometre of road were observed alongthe Balsamo Ridge between Santa Tecla and the town of Comosagua. The majority of thesefailures were of the rockfall and debris fall varieties, with occasional translational slides.Where rocks of the Balsamo Formation were exposed at crest of the slope larger, rock blockfalls occurred.NDendritic drainage patternformed on terrain dominatedby the Balsamo FormationThe Balsiamo R dgeLas BarriolerasLas ColinasComosaguaFigure 3. SPOT monochromatic image (near infra-red) of the area west of SanSalvador. Landslides appear as white areas on the image.The incidence of debris flows appears to be unrelated to the distance between the epicentreand the site of slope failure. Based on the simple attenuation of seismic energy between thesource and a potential landslide, it would seem logical that a decrease in landslide incidencewould occur with distance. Such a relationship is in fact supported by Bommer [5] who bothobserve that to trigger landslides at a larger epicentral distance, a greater energy release isrequired. Equally therefore, in terms of work done, it would be expected that the size of thelandslide would be related to the energy arriving at the site.

Figure 4 shows a plot of the area of landslide area against the epicentral distance. The area ofthe landslide was measured from SPOT data, and therefore concentrates on the larger failures(smaller landslides, less than about 30 m, being beneath effectively below the resolution ofthe SPOT data). It can be seen that there is no relationship between the epicentral distanceand the size of the landslide. There does however appear to be an upper boundary, thereforewhile there is no simple relationship between the energy arriving at the site, there is clearly apoint where there is insufficient energy to induce further slope deformation.25area of landslide (acres)2015105090 95 100 105 110epicentral distance (km)Figure 4. Graph showing the relationship between the sizeof landslide measured as an area and the distance to theepicentre of the 13 January 2001 earthquake.landslide enlargement (%)16014012010080604020030 35 40 45 50 55 60epicentral distance (km)Figure 5. Graph showing landslide enlargement after the13 February 2001 earthquake related to epicentralBecause of the second large earthquake of 13 th February 2001, many landslides increased involume. Again, dealing only with large debris flows triggered by the earthquake, it can beseen that there is no correlation between the enlargement of a given landslide and the distancebetween the epicentre and the landslide (figure 5). This again tends to suggest that the

distance between source and site is a poor indicator of the ability of an earthquake to inducelandslides, except in a general manner. This observation does not improve substantially if thedistance between landslide and projected fault rupture is used in stead of epicentral distance.TERRAIN EFFECTSDue to the nature of the terrain in the Balsamo Cordillera, topographic amplification couldhave been the cause of site specific effects. Murphy [6] discusses the importance of slopeangle and slope length in the initiation of landslides during the Chi Chi, Taiwan earthquake in1999. It was demonstrated that the majority of landslides were initiated at breaks of slopewhen the slope facet was similar to the wavelength of the incident wave.In order to investigate the hypothesis a random selection of slopes was chosen from the areaunder investigation. These were then divided into a number of categories based ongeomorphology (i.e. the morphogenetic classification of a terrain facet), geology (was theground dominated by Tierra Blanca, the Balsamo Formation or the Cuscatalan Formation)and the geometry of the slope unit (length and orientation). The presence or absence oflandslides on each of these slopes was then noted.The use of the χ 2 test revealed that there was no statistical difference between the terrain unitswhich showed landslide activity and those which did not. This observation, combined withthe poor correlation between landslide size and epicentral distance, tends to suggest that thedevelopment of slope instability arose from problems related to soil behaviour as opposed tosite specific topographic effects.THE LAS COLINAS LANDSLIDEThe landslide at Las Colinas was examined in the field. Additionally, samples of the soil thatwas involved in this landslide were collected for laboratory analysis. Three block sampleswere taken from a trial pit off the site of the landslide. Due to the abundant evidence ofinstability at the site, sampling from around the landslide itself was considered hazardous.Figure 6 shows an outline geomorphological map of the landslide. The main observationsderived from field examination were:1. The rocks exposed in the scarp of the landslide were part of the Balsamo Formation.Examination of fragments of the landslide debris showed fragments of buff-white tephradeposits that were believed to be the Tierra Blanca. Failure appears to have occurred atthe junction between the Balsamo Formation and the mantling Tierra Blanca.2. Several components of the main sliding mass could be observed. Movement appears tohave been in two phases. An investigation of the geomorphology suggested that thelandslide was regressive in nature. The movement was dominantly translational without arotational component.3. Abundant evidence existed for slope instability developed along the ridges and close tothe main slide body. Tension cracks could be observed up to 23 m behind the crown ofthe landslide and near the flanks. Smaller scale translational (slumps) landslides could beobserved at the crest of the slope as well as adjacent to the main landslide track.4. No evidence of liquefaction was observed on low-lying ground. Sand volcanoes and sandboils were not evident on the crest or the top of the slope or at the flat ground at the footof the slope.5. Once initiated, the landslide moved on low slope angles. The trim line on buildingsadjacent to landslide suggested a partially fluidised material. The absence of splash marks

however suggested that the soils were probably not fully fluidised. Based on thisobservation, and the content of the material, the slope failure was categorised as a debrisflow (Dikau [2]).6. The debris flow extended approximately 730m downslope with approximately 300-350 mof movement occurring on slope angles of c 4-5 o . The landslide was estimated to beapproximately 240 m wide.7. The debris was eroded by a spring emerging from the Balsamo Formation (the presenceof a spring line was marked on topographic maps). The emergence of this seepage linewas above the landslide debris.8308256 o 3 o7 o10 o95026 o24 o 22 o1000105028 o90028 o33 o10 oNLandslide ToeLandslide TrackSlump BlockScarpTension CrackGullyContour (m)Dip of slope1085 0 500metres15 o69021 oFigure 6. Geomorphological map of the landslide at Las Colinas.These observations suggested that the landslide was initiated in the Tierra Blanca. Thisfailure is believed to have begun as a debris slide or slump that became a flow withincreasing strain. A rise in pore water pressures in these partially saturated soils resulted inthe sliding mass being able to move on low angle surface in a partially fluidised conditions.SOIL BEHAVIOUR AND LANDSLIDE MOVEMENTOther landslides examined elsewhere in the Balsamo Cordillera showed similarcharacteristics. Debris flows were associated with the presence of pyroclastic ashfall deposits.While the majority of these landslides moved on slopes that were steeper than the slide at LasColinas, the length of runout associated with these large failures were significant. In somecases, these displacements were in excess of a kilometre in length.Investigation of slopes formed in the Tierra Blanca throughout the El Salvador area indicatedan absence of landslide scars on slopes below 20-21 o . While the slopes around the Las

Colinas site locally exceeded, it should be remembered that there was a significantcomponent of slope strength derived from the underlying Balsamo Formation.An estimate of friction coefficient of a landslide can be derived from a consideration of theheight of vertical fall compared with the runout length. For the majority of slide-type failures,this ratio will be relatively high (>0.5). In rock and debris avalanches, this value is normallylow (0.2 or lower) as a result of the complex fluidisation processes involved in theirmovements. Similar ratios calculated for the debris flows observed in the Balsamo Cordilleragave values as low as 0.17 indicating low frictional strength mobilised in the sliding mass.Observations of the Tierra Blanca in the field are not consistent with a material with a lowfrictional strength, therefore additional laboratory analysis was carried out.Analysis of samples of the Tierra Blanca collected in the field gave values of cohesion andfriction of 0-30 kPa and 34-39 o respectively (Bommer [7]) However, these data areinsufficient to describe the Tierra Blanca, as additional geotechnical data are required. Fieldand laboratory analysis of these materials indicates porosities of up to 50%. Additionally, ithas been noted that these materials are not fully saturated in the field and pore tensions havebeen developed. Laboratory testing of pore tensions from samples collected at the time of theearthquake indicated pressures of up to 700 kPa (Bommer [7]).Void Ratio1.201.151.101.051.000.950.900.850.800.750AD1Test carried out at field moisturecontent before the cell was floodedTest carried out on saturatedsampleB10 100 1000 10000Applied Pressure (kPa)CFigure 7. Diagram showing the collapse of the Tierra Blanca whensubjected to saturation (after Mavrommati [8])One-dimensional consolidation tests carried out in a oedometer cell are shown in figure 7. Anumber of distinct sections to this curve can be observed. The first part of the loading curve(A-B) shows the consolidation of the sample subject to applied load. At point B, the cell isflooded and the Tierra Blanca undergoes rapid collapse (B-C). The unloading phase of thecurve (C-D) shows that a non-recoverable consolidation of the soil structure has occurred. Itcan be seen that the magnitude of the collapse is large with up to a 23% decrease in voidratio.

The collapse of this structure has a number of implications for the behaviour of the TierraBlanca when subjected to shaking. It is likely that the collapse of the metastable soil structureoccurred during the movement of the debris flows in the Balsamo Cordillera. It is suggestedthat movement initially occurred as a rigid block (the presence of the slump blocks at thehead of the landslide indicates some form of 'brittle' failure). As a result of strain, either fromsliding or earthquake ground accelerations, the weak cement bonds in the Tierra Blancabroke, and the structure began to collapse. As collapse occurred pore water pressuresincreased and the landslide changed from a debris slide to a debris flow. It is difficult todetermine the role of ground motions in the generation of pore water pressures. While severalstudies (e.g. Holzer [9]) indicate that large increases in pore water pressure can be associatedwith shaking, many of these analyses were carried out in saturated soils. There is no doubtthat a significant proportion of the strength of these pyroclastic ashfall deposits stems fromthe pore tensions developed due to partial saturation. It is therefore, difficult to assess theimpact that a reduction of such pressures on the strength of the slope.Regardless of how pore water pressure increases occurred there is a clear effect in thefrictional strength of the soil. It can be observed that there is a significant drop in angle ofinternal friction of the Tierra Blanca in the laboratory between peak (φ p = 34-39 o ) andresidual (φ r c. 23 o ) conditions. However, based on analysis of landslide movement patterns, asignificantly lower apparent friction is observed (φ = 6-9 o ). This suggests that the effects ofpore water pressures were substantial.DEBRIS FLOWS TRIGGERED BY EARTHQUAKESWhile the landslide at Las Colinas was the most closely examined of the debris flowstriggered by January 13 earthquake, there were many more such failures. It can be seen thatmany of these landslides had a long runout. There are a number of common factors involvedin all of the slope failures that are worth noting.Firstly, all the observed debris flows were initiated as some other form of landslide. Thelandslide at Las Colinas was clearly a debris slide (slump). Other debris flows appear to haveoriginated as rock and debris falls. This clearly suggests that some form of strain is necessaryto lead to collapse and flow of the soil. In the case of debris flows which were initiated asrock or debris falls, there is field evidence to suggest the entrainment of unstable materialfrom further downslope. This latter mechanism does not appear to be true for failures thatoccurred as 'slide' type of movements. The difference between the initiation of rock/debrisfalls and translational / rotational slides is one of original slope angle. The former requiredslopes of greater than 40-42 o to occur, while the latter could develop on lower angle slopes.Secondly, the majority of debris flows are associated with the geological association of theTierra Blanca and the Balsamo Formation. Many of the landslides observed on figure 3which occurred north of the Pan American highway show a H/L ratio which does not suggest'flow' type landslides. Field examination of slope failures in the Tierra Blanca that did notdevelop into debris flows shows that these slides happened entirely within pyroclastic ashfalldeposits. It is suggested that the presence of the impermeable horizons within the BalsamoFormation (such as rhyolitic and dacitic lavas and palaeosols) provide an essentialhydrogeological pathway for water to reach, and be held in, the Tierra Blanca.CONCLUSIONSNumerous debris flow landslides were triggered by the M W = 7.6 earthquake of 13 January2001. These failures were initiated as either rockfalls, debris falls or debris slides and

developed into debris flows. Evidence suggests that the process of 'fluidisation' of the slidingmass was associated with the collapse of the soil structure of the Tierra Blanca observed inthe laboratory.The occurrence of debris flows appears to be controlled, at least partly, by the geologicalconditions. The majority of this type of slope failure had an origin where the Tierra Blancawas deposited on top of the older, stronger and less permeable Balsamo Formation. In theabsence of strong motion data, it is impossible to rule out significant differences in theseismic response of the two different geological units. However, it seems likely that theprincipal control was hydrogeological. The frictional strength of the Tierra Blanca at peak,and even residual, strength conditions is not consistent with the long runout conditions ofmany of the landslides observed in the field and on satellite imagery.The use of geological and geomorphological observations can be used as a 'first order'method of identifying future hazards.ACKNOWLEDGEMENTSThe authors wish to acknowledge the receipt of funding from the Natural EnvironmentResearch Council, UK (NER/A/S/00030) and the Royal Academy of Engineering. Theassistance of colleagues from PRISMA (especially Herman Rosa) and the University ofCentral America was invaluable during firledwork.REFERENCES1. Varnes, D. J. Slope movement type and processes. In: R. L. Schuster & R. J. Krizek,(eds), Landslides: analysis and control: 11-33, Washington. Transportation ResearchBoard Special Report 176.2. Dikau, R., Brunsden, D., Schrott, L. & Ibsen, M. L. Introduction. In Dikau, R., Brunsden,D., Schrott, L. & Ibsen, M. L. (Eds), Landslide Recognition: Identification, Movementand Causes. 1996, pp251, Wiley, Chichester, ISBN 0-471-96477-8.3. Schmidt-Thomé M. The geology in the San Salvador area (El Salvador, CentralAmerica), a basis for city development and planning. Geol. Jb. 1975; 13: 207-228.4. Hoek, E. The strength of rock and rock masses. News Journal, International Society ofRock Mechanics, 1994, 2, 4-16.5. Bommer JJ, Rodríguez CE. Earthquake- induced landslides in Central America.Engineering Geology 2002, 63(3/4).6. Murphy, W., Petley, D.N., Bommer, J.J. & Mankelow, J. M.. Uncertainty in groundmotion estimates for the evaluation of slope stability during earthquakes. QuarterlyJournal of Engineering Geology and Hydrogeology, 2002. In press.7. Bommer JJ, Rolo R, Mitroulia A, Berdousis P. Geotechnical properties and seismic slopestability of volcanic soils. In: Proceedings of Twelfth European Conference onEarthquake Engineering; UK: London, Paper No. 695, 2002.8. Mavrommati, Z. C. Seismic behaviour of slopes in an undersaturated volcanic soil. M.Sc.Dissertation, 2000. Imperial College London.9. Holzer, T. L., Youd, T.L. & Hanks, T.C. Dynamics of Liquefaction During the 1987Superstition Hills, California, Earthquake. Science, 1989,Vol. 244, pp. 56-59.