Mechanism analysis of landslide of a layered slope induced by drawdown of water level

136Science in China Ser. E Engineering & Materials Science 2005 Vol.48 Supp. 136--145Mechanism analysis of landslide of a layeredslope induced by drawdown of water levelZHANG Junfeng, LI Zhengguo & QI TaoInstitute of Mechanics, Chinese Academy of Sciences, Bejing 100080, ChinaCorrespondence should be addressed to Zhang Junfeng (email: zhangjf@ imech.ac.cn)Received September 29, 2004Abstract The frequent drawdown of water level of Yangtze River will greatly infuencethe stability of the widely existing slopes in the Three Gorges reservoir zone, especiallythose layered ones. Apart from the fluctuating speed of water level, the differentgeological materials will also play important roles in the failure of slopes. Thus, it must befirst to study the mechanism of such a landslide caused by drawdown of water level.A new experimental setup is designed to study the performance of a layered slopeunder the drawdown of water level. The pattern of landslide of a layered slope induced bydrawdown of water level has been explored by means of simulating experiments. Theinfluence of fluctuating speed of water level on the stability of the layered slope is probed,especially the whole process of deformation and development of landslide of the slopeversus time. The experimental results show that the slope is stable during the water levelrising, and the sliding body occurs in the upper layer of the slope under a certaindrawdown speed of water level. In the process of slope failure, some new small slidingbody will develop on the main sliding body, and the result is that they speed up thedisassembly of the whole slope.Based on the simulating experiment on landslide of a layered slope induced bydrawdown of water level, the stress and displacement field of the slope are calculated.The seepage velocity, the pore water pressure, and the gradient of pore water head arealso calculated for the whole process of drawdown of water level. The computing resultsare in good agreement with the experimental results. Accordingly, the mechanism ofdeformation and landslide of the layered slope induced by drawdown of water level isanalyzed. It may provide basis for treating this kind of layered slopes in practicalengineering.Keywords: layered slope, landslide, drawdown of water level.DOI: 10.1360/04zze241 Introduction中国煤化工The instability of slopes on the river banks andMHCN M H Gnt slopes, in-duced by drawdown of water level, are great danger for the local residents and for theCopynight by Science in China Press 2005Mechanism analysis of landslide of a layered slope induced by drawdown of water leve!137hydroelectric engineering. A lot of accidents or disasters relating to the drawdown ofwater level can be found in literatures. In October 1963, the landslide, near the recentlyconstructed Vajont Reservoir dam at that time, caused the failure of the toppest hyper-bolic arching dam in the world, ruined one city and several towns in the lower reach, and1926 persons perished in the resulting flooding". In July 1981, a flood that might occurnearly once every 100 years led to deformation of more than 100 loosely stacked geo-logical layers from Chongqiong to Yichang, China. From 1965 to 1969, the deformationup to 10 m occurred in several sliding bodies in the upper reach of the Cepatsch dam (inAustria) during the early stage of the running of the reservoir. The Grand Coulee reser-voir (in U.S.) also caused about 500 slopes to lose stability from 1942 to 1953. In China,the Three Gorges reservoir started conserving water from 2003. The water level of thereservoir would periodically fluctuate ranging from 135 to 175 m. The stability of theloosely stacked layers along the reservoir bank would be directly affected by the peri-odical drawdown of water level in addition to the other factors such as rainfall infiltra-tion, the specific geomorphological and geological environment, the nonuniformity andstratification of rock soil, the faults and joints within the slopes, and the seepage andsoakage in the geological materialsl2 0.Maoping sliding body, close to Geheyan Reservoir in Qingiang, Hebei Province, China,began deforming after the water conservation in the reservoir in April 1993. Up to Sep-tember 2001, the maximum displacement for the slope reached 2.1 m. About 3/4 of thefirst flat stair (about 180 m long, mainly composed of loose rock soil) was submerged inthe water. According to the in situ monitoring results, the deformation responded quicklyto the fall of water level but relatively slow to the rise of the water level. This means thatthe deformation is mainly affected by the fall of the water level. Though the deformationof Maoping sliding body is the synthetical result under both rainfall and drawdown ofwater level of Qingjiang river, the hydraulic effect and the reduction of strength of slopemedia are among the main factors to cause the occurrence of the landslide'”8.In Ref. [8], the authors considered the speed and the range of the fall of water level asthe main factors to cause slopes sliding in the upper reach of the embankment. In orderto explore the mechanism of landslide caused by drawdown of water level, the wholeprocess of landslide (from deformation to sliding) for a layered slope was studied byexperiment. The qualitative explanation for experimental phenomena had been givens.Numerical simulation was also conducted to reveal the cbange of some important pa-rameters. The analysis in this paper is concentrated on the stress and displacement fieldof the slope, the seepage velocity, the pore water pressure, and the gradient of pore waterhead in order to probe the hydraulic effect during the change of water level. The result-ing change of material strength for rock soil by water is not discussed here.2 Experimental setup and layered slope中国煤化工2.1 Expenimental setupMYHCNMHGIn Ref. [10], a box, made of plexiglass material, with additional specific setup to supplywww.scichina.com138Science in China Ser. E Engineering & Materials Science 2005 Vol.48 Supp. 136- -145water in the upper reach of a slope model300 mn|was used to study the slope failure byinfiltration of rainwater. We also designedan experimental setup adopting high-tran-sparent plexiglass material (Fig. 1). Thethickness of the wall of the plexiglass boxis 20 mm, and the inner size of the box isgContptits2000 mmx800 mmx1100 mm (lengthX三王Water Clay羽Macadamwidthx height). The bottom of the boxFig. 1. Main experimental box for landslide study.was designed with a 100 mm sandwichlayer and was covered with 5 mm steelplate in which equidistant holes (diameter 20 mm and space 50 mm) were arrayed. Thesteel plate was covered with a stainless steel net with 3 mmx3 mm mesh. The change ofwater level is controlled by injecting water into or drawing water out of the sandwichlayer through a controlling valve. A smaller box (800 mmx300 mmx1 100 mm) was alsodesigned at one end in the main box to supply water in the upper reach of the slopemodel (see Fig. 1). Only on the side of the small box adjacent to the slope are there holesto permnit supplying water to pass through.2.2 Layered slope and parameters of the rock-soil materialBy considering the geological configurations of the layered slopes, the experimentalslope model was simplified but the main characteristics were retained: (a) The upperlayer is of clay/silt to denote the weathered rock and soil (possibly mixed with crushedrock); (b) the second layer is of macadam to simulate the scree, gravel, or detritus, whichindicates the unweathered stacked layer;, (C) the fixed stainless steel net and plate in thebottom indicate the base rock. The initial size of the slope model is showed in Fig.1. Inorder to have a good look at the deformation, some tracking point was set between theside of the slope model and the wall of the plexiglass box. The lattice designed with seg-ment of thin cotton thread was also laid on the top surface of the slope model to measurethe initiation of landslide and the sliding displacement conveniently.The silt sample was adopted for the upper layer of the slope model because of its lowcohesion (defined as c). Its main physical and mechanical properties were given here:the material density Ps= 2.75x 10 kg/m', the ratio of void e = 1.8, the natural dry densitypa= 0.98x10' kg/m', the saturated density ρ= 1.625x10 kg/m', the coefficient of per-meability K = 3.20x 10° m/s, the compressible coefficient 1.78 MPa，the modulus ofcompression 1.935x10 N/m', index of liquidity 1.51, liquid limit 50.9, plastic limit27.2, plastic index 23.7, coefficient of consolidation 7.2x 10~7 cm/s; (a) quick shear test:c= 10.4 KPa, φ= 0.95°; (b) consolidated quick shear test: Cq= 14.5 KPa,中= 13°. Tri-axial test: (a) unconsolidated and undrained test: (中国煤化工) consolidatedand drained test: Ccu= 11.2 KPa,央u= 13.39. Und| YHCNMHGength:(a)un-disturbed qu = 20.45 KPa; (b) recomposed Qu = 5.4 KPa.Copyright by Science in China Press 2005Mechanism analysis of landslide of a layered slope induced by drawdown of water leve}1393 Experimental phenomena and resultsBy means of controlling the speed of injection and draining through the valve at thebottom of the main box, the change of water level with time in the lower reach isshown in Fig. 2. The slope was then stable after the process of several times consolida-tion and deformation at the initial slope angle 12° . The slope angle was slowly changedby lifting the right hand of the main box until it reached 23°. The water level waschanged again and the whole process of deformation and sliding was recorded by usinga digital camera.When the water level changed with time (see Fig. 3), the following phenomena couldbe observed during the experimental process: The tensile cracks initiated on the top sur-face of the slope when the water level reached 300 mm in the lower reach of the slope(Fig. 4). Two minutes later, the length and width of the tensile cracks were up to 155 mmand 4 mm respectively, while the water level in the lower reach was 236 mm. Com-panied with the fall of the water level in the lower reach of the slope, the length andwidth of the tensile cracks increased quickly, a part of the slope body began to slidedown and the sliding speed became quicker and quicker. A secondary sliding body sepa-rated from the main sliding body and it furher speeded up the movement of the latter(Fig. 5). lt was the secondary sliding body that caused the second peak value of speed inFig.6(b). As more and more sliding mass stacked at the foot of the slope, the sliding600 (500一- Experimental result400300Lower reach- - Upper reach星200200100050 100 1S0 200 250 300 350100 200 300 400500 600Time/minTime/sFig.2. Water level in the lower reach versus time.Fig. 3. Water level versus time.Tensile cracksThe main sliding body中国煤化工MHCNMHGFig. 4. Tensile cracks on the surface of slope.Fig.5. The main and secondary sliding bodies.www. scichina.com140Science in China Ser. E Engineering & Materials Science 2005 Vol.48 Supp. 136- -145body moved slower and slower and finally stopped. The whole process of deformationand landslide lasted about 20 min.The sliding displacement and speed of landslide were obtained by analyzing the digi-tal images. The results are shown in Fig. 6. The changes of the slope angle and the shapeof the sliding surface could be observed from the side of the main experimental box. Itcould be seen that the shape of the sliding surface was nearly circular, and the slope an-gle of the sliding debris decreased by 5° compared with the initial slope angle.旨14060 t一一 Experimental result120E 50-10040 t80 t-Experimental result30 t品60卜g 20宣40厉20-(a)(b),400 600 800 1000 1200 1400 160010121416182022!2426Time/sFig. 6. (a) Sliding displacement versus time; (b) sliding velocity versus time.4 Analysis of stress and displacement fieldsThe finite element method was adopted to analyze the static stress field and dis-placement field for the experimental slope model, and the size of the computing modelwas consistent with the experimental one (Fig. 1). Half of the full slope model was se-lected in analysis because the slope is geometrically symmetrical about the longitudinalcentral section, and this section was taken as symmetrical boundary. Contact elementwas used to simulate the interface between the up-per and the lower layers. The wall of the main boxRigid contactwas assumed as rigid shell and contact element wasboundariesalso used to simulate the interfaces between thewall of the box and the two soil layers. The 8-nodesolid element was adopted to divide the slope body,and the parameters of materials were identical tothat in section 2.2 above. The computational modelof mesh is shown in Fig. 7.The contours of Von Mises stress and the dis-placement field are shown in Fig. 8. It can be seenfrom Fig.中国煤化工displacementield are par:CS on the top0HCNMHGone of theFig. 7. Computational model of mesh.surface of th. orpJ u u oiiupr SfCopyright by Science in China Press 2005Mechanism analysis of landslide of a layered slope induced by drawdown of water leve!141contours of displacement is the same as that of the sliding body (Fig. 5). Some tensile :cracks also developed behind the sliding body in the area of relatively large deformation.B=.001769C=.002949D=.004128.005308=] 5899F\F19374007667G- 22849H-. 008846H=26324]=29799G一用Y游xAHHHGMXB的(8)(b)Fig. 8. (a) Contours of Von Mises stress (Pa); (b) contours of the displacement field (mm).5 Change of the fields of seepage and pore water pressureThe drawdown of water level is one of the main factors to cause landslides. The hy-draulic effect of the drawdown of water level is exhibited by means of the changes ofseepage field and the pore water pressure field within the slope body. For the experi-mental slope model, the permeability of the upper layer is low. In addition to the watersupply in the upper reach of the slope, the change of water table within the upper layerin a short time would be negligible. However, the rapid fall of the water level in thelower reach causes a sharp change for the gradient of the pore water pressure within theslope, especially in the vicinity of the foot of the slope. Therefore, the hydraulic effect isthe main reason to cause the occurrence of landslide.The finite element meshing for computing the fields of seepage and pore water pres-sure is shown in Fig. 9. The coefficients of permeability used in computation are ks = :3.2x10* m/s for the upper layer and kg= 1.0x10 2 m/s for the lower layer. The knownwater heads for the upper and the lower reaches are provided and the bottom boundarycondition along with other boundaries is free. The change of the saturated water table(i.e. the water head being zero) with time is given in Fig. 10. It is shown that the changeof the saturated water table is small even if the water level in the lower reach of theslope fell rapidly, because of, just as the description in the above paragraph, the un-changed water level in the upper reach and the low permeability of the upper layer. Thedistributions of the pore water pressure, the seepa中国煤化工ient of pres-sure head are given for two cases. The first is for tlHCNMHGerlevelinthewww.scichina.com142Science in Chima Ser. E Engineering & Materials Science 2005 Vol.48 Supp. 136- -145lower reach (h = 46.8 cm) began to fall (t= 0) (Fig. 11). The second is for the momentwhen the water level in the lower reach reached 13.8 cm att = 600 s (Fig. 12). It shouldbe noted that the computing water tables in Pigs. 11 and 12 mean the conlours of thecomputing waler head being zero. The negative pore water head above the computingwater table indicated the unsaturated state in this area.Water tableime (s)一号20TE300600Fig. 9. Computational meshes for seepage.Fig. 10. Change of water table versus timne.Water tablc叶(8子0.L0e-7.2...-0.3-.0.4田)冬L9 03(0)(d)Fig. 11. Pressure head and seepage field (intial water level in the lower reach: 46.8 cm). (a) Conlours of waterhead (m); (b) conlours of seepage velocity (m/s); (c) seepage held of the slope; (d) contours of gradient of pressureComparing the computing results of the two cases with the water level difference be-ing the largest, there are obvious differences in the distributions of the pore water pres-sure (Figs. 11(d) and 12(d)) and the seepage velocity (Figs. 11(b) and 12(b)) respectively.In this experiment, the tensile cracks initiated when the hcight of water level in the lowerreach was 300 mm witht= 370 s. Hence, it is ne中国煤化工stributions ofthe pore water pressure and the seepage velocity at!YHCNMHGCopyight by Science in China Press 2005Mechanism analysis of landslide of a layered slope induced by drawdown of water }eve]143Water table1.06-6 ... 0折1.0e-60.20.31T.0e-5 .).4-号a)(b)￥(C)(d)Fig. 12. Pressure head and seepage field (water level in the lower reach: 13.8 cm). (a) Contours of water head (m);(b) contours of seepage velocity (m/s); (c) seepage field of the slope; (d) contours of gradient of pressure head.Water lable(t= 370 s)(1=370 s)τ47 cmTE亏土30cm之30cm(a)b)Fig. 13. Gradieni of pressure head and seepage field (waler level of low reach: 30 cm). (a) Con1ours of seepagevelocity (m/s); (b) contours of gradiert of pressure head.6 Mechanism of landslide of a layered slope by drawdown of water levelComparing the results in Figs. (11)- (13), the following conclusions can be drawn forthe simulating experiment: (1) The gradient of pore water pressure and the seepage ve-locity in the vicinity of the foot of the slope were always larger than those in the otherpart of the slope before and after the fall of the water level, especially when the waterlevel reached to a low level (Fig. 12(d)). The seepage pressure, a common saying for thegradient of pore water pressure in slope engineering, is one of the factors to cause land-slide. (2) During the process of the fall of the中国煤化工f pore waterpressure near the steep surface and the foot of thewhich meansthe dynamic pore water pressure taking effect. ThTYHC N M ! Guse landsidewww. scichina.com144Science in China Ser. E Engineering & Materials Science 2005 Vol.48 Supp. 136- -145(Figs. 11(d), 12(d), and 13(). (3) The contours of seepage velocity near the steep sur-face and the foot of the slope in Figs. 12(b) and 13(a) are obviously larger than those inFig.11(b). It indicates that the dynamic pore water pressure after the fall of the waterlevel is larger than that before the fall of the water level.In this experiment, the gradient of water pressure in the lower layer (macadam) wouldbe small for the permeability is relatively large during the change of the water level.Contrarily, for the upper layer with low permeability, water was continuously suppliedfrom the small box in the upper reach, and there was no time for the pore water to seepout during the rapid fall of the water level. Hence, the large gradient of pore water pres-sure developed in this layer as shown in Figs. 11(d), 12(d), and 13(b). The tensile de-formation initiated on the top surface of the slope under the action of gravity andstatic/dynamic pore water pressure, then the tensile deformation developed quickly intotensile cracks, and the earliest observed one was near the steep surface and continueddeveloping into the longest one. As new tensile cracks continued developing towards theinner of the slope, the slip surface developed, the lower part of the slope body graduallyseparated from the main slope body, and landslide occurred under the high hydraulicgradient of pressure.7 Concluding remarksThe change of water level can cause the increase of the gradient of pore water pres-sure, which induces the seepage within the slope body. The seepage velocity respondsdirectly to the hydraulic pressure. Based on the simulating experiment on drawdown ofwater level in the lower reach of a layered slope, the seepage field, the distribution of thepore water pressure, and the field of the gradient of water head within the slope modelwere computed. The computing results were compared with the results from experimen-tal observation and measurement, and the two results were essentially in agreement. Theresults showed that the slope would become instable after the fall of the water level inthe lower reach, because the high gradient of pore water pressure would develop near thefoot of the slope. Thus, as the part of the steep slope surface above water level was ex-posed more and more, the deformation initiated, the slip surface developed, and thelandslide occurred under the action of dynamic seepage pressure and the static waterpressure. The mechanism of landslide for such a kind of slopes is then probed and clari-fied. In practical slope engineering, the reasonable design for drainage system and rein-forcement for the foot of the slope are necessary to prevent collapse or landslide inducedby drawdown of water level.Acknowledgements The authors wish to thank all their colleagues in the Institute of Mechanics, CAS, for theirvaluable discussion on various parts of the work described in the paper. This work was supported by the NationalNatural Science Foundation of China (Grant No. 10372104), the Special Funds for the Major State Basic ResearchProject (Grant No. 2002CB412706), the Knowledge Innovation Project of the Chinese Academy of Sciences (GrantNo. KJCX2-SW-L1-2), the Special Research Project for Landslic中国煤化工ree Gorges Reser-voir Areas (Grant No.4- 5). .YHCNMHGCopyright by Science in China Press 2005Mechanism analysis of landslide of a layered slope induced by drawdown of water level145References1. Trollope, D. H., The Vaiont slope failure[J]. Rock Mechanics, 1980, 13(2):71- 88.2. Cui Zhengguan, Li Ning, Slope Engineering- New Advancement in Tbeory and Practice[M], Beijing: ChinaWater Power Press (in Chinese), 1999.3. Cai Yaojun, Cui Zhengquan, Cojean R., Mechanism on deformation and instabiliry of bank slope induced byreservoir{A], in Proceeding of the 6' A National Conference of Rock Mechanics and Engineening[C], Beijing:China Science and Technology Press (in Chinese), 2000, 618- 622.4. Ding Enbao, Problems of super-high and steep slopes in development of hydroelectric power on Jinshajiangriver[J], Joumal of Engineering Geology (in Chinese), 2000, 8(2): 131-135.5. Cheng Chiyuan, The analysis and treatment on the landslide of the dam of Dayuanhe reservoir[J], JiangxiHydraulic Science and Technology (in Chinese), 1992, 18(3): 207- 210.6. Tang Weizeng, The accident analysis that reservoir drainage result in sliding for earthfill dam's upstreamface[J], Jiangxi Hydraulic Science and Technology (in Chinese), 1992, 21(4): 197- 202.7. Deng Jianhui, Ma Shuishan, Zhang Baojun et al, Preliminary investigation on the reactivation of Maopinglandslide, Geheyan Reservoir, Qingjang River[J], Chinese Jourmal of Rock Mechanics and Engineering (inChinese), 2003. 22(10): 1730 - 1737.8. Wang Shangqing, Yi Qinglin, Yan Xueqing, The synthetic analysis on deformation characteristics and infiu-ential factors of Maoping landslide, Qingjiang[J], The Chinese Jourmal of Geological Hazard and Control (inChinese), 1999, 10(2): 40--44.9. Zhang Junfeng, Meng Xiangyue, Zhu Erqian, Testing stady on landslide of layered stope induced by fluctua-tion of water level[D], Chinese Joumal of Rock Mechanics and Engineering (in Chinese), 2004, 23(16): 2676-2680.10. Enoki, M, Kokubu, A. A., Ikeda, Y, Infiltration of rainwater and slope failure[A], in Proceedings of the NinthInternational Conference and Field Trip on Landslides[C] (eds.. James S. Griffiths, Martin R. Stokes, RobbertG Thomas), Rotterdam: A. A. Balkema, 1999, 27-35.中国煤化工MYHCNMHGwww.scichina.com