The details of this paper will be given in a manuscript to be submitted in the Bulletin of Engineering Geology and the Environment.

(Submitted to Guinsaugon Conference 02/24/2008)

Mechanism of the initiation and motion of the 2006 Leyte landslide, Philippines
Kyoji Sassa1, Hiroshi Fukuoka2, Renato Soridum3, Gonghui Wang2, Hideaki Marui3, Takashi Furumura5, Fawu Wang2 1 International Consortium on Landslides 2 Research Centre on Landslides, Disaster Prevention Research Institute, Kyoto University, Uji, Kyoto, Japan 3 Philippine Institute of Volcanology and Seismology, Diliman, Quezon City, Philippines 4 Research Center for Natural Hazards and Disaster Recovery, Niigata University 5 Earthquake Research Institute, University of Tokyo, Tokyo, Japan A rapid and long-traveling landslide occurred on 17 February 2006 in the southern part of Leyte Island, Philippines. The landslide resulted in 154 fatalities and 990 people disappeared in the debris. A Japanese and Philippine joint team investigated the site and took samples from the landslide. From eld investigation and dynamic-loading ring-shear tests on a sample taken from the landslide site, we report here: 1) A small (Ms 2.6), near-by earthquake was strong enough to trigger the initiation of the landslide which occurred after heavy rainfall. Thus, this landslide was rainfall and earthquake induced, and 2) The subsequent rapid motion of the landslide was a consequence of the sliding-surface liquefaction phenomenon. It was the result of the generation of high porewater pressure within the shear zone, which was caused by crushing of grains of the volcanoclastic debris at the site subjected to shearing under a high normal stress. This landslide on Leyte Island was the greatest single landslide disaster in the world since the 2001 Las Colinas landslide (about 200,000 m3) triggered by the El-Salvador earthquake (Ms 7.8), which killed 747 people (1, 2). A joint Japanese and Philippine team of 22 scientists and engineers (3) investigated the landslide on 19-26 March 2006 with support from Japanese and international organizations through the International Consortium on Landslides (ICL)(4). The team investigated the landslide from the ground and also from a helicopter. Fig.1 is a frontal view of the landslide. The slope had been covered by tropical forest. The landslide removed the forest-cover. Three linear concave ground forms (F1, F2, and F3) were identied by observation from the ground and also from the helicopter. These concave ground forms were the result of fracturing of rocks due to the Philippine fault and its sub-faults. Along one of these concave ground forms (F3), a secondary landslide, which occurred after the main landslide, was found at point A in Fig. 1. The occurrence of the landslide suggests lower shear strength along these fractured zones. The central line of the landslide seemed to bend at B and C; it is likely a combination of two parallel lines (Fig. 1). The section of the central line of the landslide was surveyed by a non-mirror total station and a groundbased laser scanner in the eld and compared to a SRTM (Shuttle Radar Topography Mission) map before the landslide (Fig. 2). The red-color part shows the initial landslide mass while the blue-color part presents the displaced landslide debris after deposition. The initial landslide mass shown in red seems to consist of two blocks (Block 1 and Block 2) corresponding to the two straight lines above/below the bending part in Fig. 1. Though exact verication is difcult, we can speculate a landslide development process from the photo and section as well as the eld observation: 1) The ground-water level rose upslope from point B in Fig. 1 because the ground-water ow probably was blocked at this bending point and likely dammed up upslope of this point; 2) This blockage necessarily lowered the stability of the lower part of Block 1 in Fig. 2; 3) The loss of stability in the lower part of Block 1 should have almost simultaneously caused the motion of the rest of Block 1, including the head scarp, due to loss of support

Extended Abstract

Fig. 1 Location of the 2006 Leyte landslide (bottom-left) and a frontal view from a helicopter. F1 was identied as the Philippine fault in the eld. It was also reported by others (5, 6). A: Secondary landslide, S: Sampling point. Photo by Kyoji Sassa.

at its bottom, and, at the same time, the movement of Block 1 provided additional forces onto the top of Block 2 initiating the movement of Block 2; and 4) Both landslide masses traveled onto the at residential and farming areas. The inclination connecting the top of the initial landslide and the toe of the displaced landslide deposit is approximately 10 degrees, which indicates the average apparent friction angle (7) mobilized during the whole travel distance. The value is much smaller than the usual friction angle of debris (sandy gravel) of 30-40 degrees. Therefore, it suggests that high excess pore-water pressure was generated during motion. Fig. 2B shows a ow mound that traveled from the initial slope to this at area without much disturbance. Movement without much disturbance is possible when the shear resistance on the sliding surface is very low; thus, movement of the material is like that of a sled.

The material of the ow mound is volcanoclastic debris, including sand and gravel. Though the sizes of ow mounds on this landslide were diverse, most of them consisted of volcanoclastic debris. We also observed from the surface and excavation that the valley-side slope in the source area consisted of volcanoclastic debris or strongly weathered volcanoclastic rocks. Therefore, we took a sample of about 100 kg from the base of the ow mound shown in Fig. 2 (S in Fig.1). We subjected this sample to dynamic-loading ring-shear tests assuming the sample to be the soil in which the sliding surface was formed. The initial dynamic-loading ring-shear apparatus (DPRI-3) was developed by K. Sassa in 1992 (8). Then it was improved and four versions with different specications (DPRI-4, 5, 6, 7) were developed one by one from 1994-2003 (9). These apparatuses have been used to investigate the mechanisms of earthquake-induced landslides (9, 10, 11) and rainfall-induced landslides (9, 12), debris ows transformed from slides (9, 12), and a landslide causing a tsunami (13). The basic concept of this apparatus is illustrated in Fig. 3. The major characteristics of this apparatus that differ from those of conventional apparatuses (14, 15, 16) are: 1) Pore-water pressure can be monitored during shearing by preventing any leakage of water from the contact between two rings during shearing, and 2) Dynamic loading, up to 5 Hz, and high-speed shearing, up to 224 cm/sec (DPRI-6), are possible. The testing procedure of this apparatus is as follows: 1) A sample is taken from the soil layer in which the sliding surface was or will be formed, 2) The sample is set in the ring-shaped shear box, saturated, and consolidated under the stress due to the self-weight of the soil layer, 3) Seismic stress or pore-pressure change due to rainfall will be provided, 4) If failure occurs, pore-water pressure and mobilized shear resistance with progress of shear displacement are monitored.

Extended Abstract

Fig. 2 Central section (A) of the landslide and a ow mound (B) at the sampling point. The central section was estimated by comparison of the topographical survey result in the eld to the SRTM map before the landslide. Many ow mounds that maintained their original structure from the source area were found in the deposition area. The photo shows one of the big ow mounds from which samples were taken at its base (S in Fig. 1).

We examined the triggering factors of the Leyte landslide. A small earthquake occurred near the site at the time of occurrence of the landslide, and the location of the hypocenter of this earthquake was estimated by the U.S. Geological Survey (USGS) and the Philippine Institute of Volcanology and Seismology (PHIVOLCS). According to PHIVOLCS, the earthquake occurred at a location 21 km west of the landslide, 8 km deep, with magnitude Ms 2.6, at 10:36 hrs on 17 February 2006. Using the standard attenuation function between peak ground acceleration and hypocentral distance (17), the peak ground acceleration at the landslide site was estimated at 10 gal for this magnitude. We then estimated the expected peak acceleration at the bottom of the landslide mass as about 60 gal. This was based on about three times ground accelerations at the sliding surface due to the difference in the compressional (P) wave speeds between soft volcanoclastic debris (Vp= 0.5 1.5 km/s) and hard volcanic bed rock (Vp= 2.5 5 km/s) that outcropped in the head scarp, because the amplication level is proportional to the velocity contrast between two layers. Though the shear (S) wave speeds of the volcanoclastic debris and the bed rock are unclear, similar level of velocity contrast to the P wave is expecting at the sediment/bed rock interface. An additional magnication of two times is also expecting in the landslide site due to the focusing of seismic waves on the steep mountain topography; this resulted in a total magnication of six times.

Extended Abstract

Heavy rainfall (459.2 mm for 3 days on 10-12 February and 571.2 mm for 5 days on 8-12 February 2006) occurred in this area before the day of the landslide (18). This rainfall should have increased the ground-water level and porewater pressure inside the slope. However, the peak ground-water level had likely passed before the occurrence of the landslide on 17 February because the rainfall on 13-17 February was small (total 99.0 mm for 5 days). We simulated the ground-water level using a tank model that had been developed to simulate the ground-water level in the Zentoku landslide, Japan (19), which had a depth and inclination similar to that of the Leyte landslide. When inputting 10 days precipitation records at the nearest monitoring station in Otikon (about 7 km west of the landslide) on 8-17 February, the peak groundwater level occurred on 13 February 2006. Because the peak ground-water level had already passed when the landslide occurred on 17 February, we deduced that a small earthquake was the trigger of the landslide. Based on this consideration, a dynamic-loading ring-shear test on the sample from the landslide was conducted as follows:

The sample was set in the shear box (250 mm inside diameter, 350 mm outside diameter) of DPRI-6, was fully saturated, and was consolidated under the stresses acting on the sliding surface in the lower part of Block 1. From Fig. 2 it can be seen that the depth of the sliding surface at the lower part of Block 1 is 50-70 m. However, a lesser soil depth was assumed because of the capacity of this apparatus: the sliding surface was assumed for the test to be 35 m deep and at an inclination of 25 degrees. The unit weight of the soil was assumed to be 20 kN/m3. In the preliminary test to increase pore-water pressure until failure, the critical groundwater level causing a landslide without earthquake loading in this slope was obtained. In the simulation test of a rain- and earthquake-induced landslide, the normal stress corresponding to that of 5 m lower than the critical ground-water level (i.e., further 5 m of ground-water level elevation will trigger the landslide) was rst loaded on the sample in the ring-shear apparatus and consolidated. Then, the shear stress due to the self-weight of the soil layer was loaded. Finally, a peak seismic stress corresponding to a seismic acceleration of 60 gal using the ground acceleration observed in Maasin (PHIVOLCS, Code number: MSLP, Latitude: 10.1340, Longitude: 124.8590, Elevation: 50.0) was loaded on the sample. The test result is presented in Fig. 4B. The black line shows the normal stress loaded on the sample, the red line shows the shear resistance mobilized at the sliding surface; the blue line shows the monitored pore-pressure change inside the sample; and the shear displacement at the center (300 mm diameter) of the ring-shape sample is shown in green. The static normal stress and shear stress working on the sliding surface before seismic loading are shown at Time = 0 T1. The seismic stress was loaded from T1. The shear displacement started at T2 after 19 seconds of loading when the peak of shaking had passed. With progress of shearing, the pore-water pressure increased while the shear resistance decreased. Then, the

Fig. 3 Concept of the dynamic-loading ring-shear apparatus. This device aims to geotechnically simulate the formation of a sliding surface and the resulting postfailure motion by reproducing the stresses acting on the potential sliding surface in the slope: the static stress due to the self-weight of the soil layer, seismic stress due to an earthquake, and pore-pressure increase due to rainfall.

Extended Abstract

Fig. 4 Landslide simulation test due to earthquake loading in addition to pore-pressure increase due to rainfall (B) and the illustration of sliding-surface liquefaction due to shear-induced grain crushing (A). The left diagram of Fig. 4A corresponds to the state of T = 0 T2 in Fig. 4B and the right diagram of A corresponds to the state after T3. The range between T2 and T3 is a state of progress of grain crushing.

pore pressure reached close to the normal stress and a steady-state of high-speed shearing proceeded at a very low shear resistance of only 7 kPa. This phenomenon is called sliding-surface liquefaction (12, 20, 21) The mechanism of sliding-surface liquefaction is illustrated in Fig. 4A. The volcanoclastic debris tends to dilate during seismic loading before the initiation of failure; so a negative pore-water pressure was monitored before the failure as shown in the left diagram of Fig. 4A and the values in the range of T= 0 T2 in Fig. 4B. The shear displacement started at T2, and the shear displacement caused breaking and crushing of grains of volcanoclastic debris under the normal stress, as shown in the right diagram of Fig. 4A. The grain crushing necessarily caused volume reduction, leading to a high pore-pressure generation as shown in the range of T2-T3 of Fig. 4B. Pore-pressure continued to increase until a certain value under which no further grain crushing occurred, at point T3. Every material has a certain value of effective normal stress (normal stress pore-water pressure) under which grains can be sheared without any crushing. The state was reached at point T3 in this soil. Thereafter, shear displacement continued to increase with a steady state of stress. The state of grain crushing during shearing was observed through the transparent shear box (DPRI-7). The photos were presented in (10). The change of grain-size distribution after shearing was presented in (12, 21).

Extended Abstract

This study of the Leyte landslide showed that even a small earthquake can be the critical trigger of a landslide when the stability of the slope has already been reduced due to rainfall. Thus, the combined effect of rainfall and earthquake should be considered in landslide risk analysis. Another implication of this study is that a rapid and long-traveling landslide like the Leyte landslide, which may cause a catastrophic disaster can be predicted if the shear characteristics of the soil at the site are measured and the depth of the landslide is estimated in advance. References and Notes 1. K. Konagai, J. Johansson, P. Mayorca, T. Yamamoto, M. Miyajima, R. Uzuoka, N. E. Pulido, F. C. Duran, K. Sassa, H. Fukuoka. Proc. International Symposium Landslide Risk Mitigation and Protection of Cultural and Natural Heritage, Kyoto, 139-154. (2002). 2. H. Fukuoka, K. Sassa, Y. Okada. Proc. APERIF (Areal Prediction of earthquake and rain induced rapid and long-travelling ow phenomena) Symposium. 301-312 (2002) 3. The joint team includes the mechanical study sub-team writing this report, other sub-teams studying the topographical survey, geomorphic survey, and social impact survey, and liaison members. All cooperated from their specialties. We thank F. Castro (Ofce of Civil Defense, Government of Philippines), L. Liongson (University of the Philippines), A. Dagg and B. Bautista (Philippine Institute of Volcanology and Seismology), H. Marui (Niigata University, Japan), H. Suwa and C. Mondonedo (Disaster Prevention Research Institute, Kyoto University), K. Araiba (National Research Institute of Fire and Disaster, Japan), M. Murata (Asian Disaster Reduction Center, Japan), Y. Okada and Y.Adikari (Forestry and Forest Products Research Institute, Japan), S. Takahashi and S. Hanatani (Nippon Koei Co. Ltd., Japan), and H. Tamaki and others (Mathematical Assist Design Laboratory, Japan). 4. Support for the planning of this investigation was obtained from S. Nishikawa (Cabinet Ofce of Japan), M. Miyano (Ministry of Foreign Affairs of Japan), the M. Nagaishi (Japan International Cooperation Agency (JICA)), M. Nakagawa (United Nations Ofce for the Coordination of Humaritarian Affairs (OCHA)), and P. Basabe (United Nations Secretariat for the International Strategy for Disaster Reduction (UN/ISDR)) through the International Consortium on Landslides (ICL). R. Yniquez-Lerias (Governor of the Southern Leyte Province, Philippines), M. Lim (Mayor of Saint Bernard City), S. Hanatani (Nippon Koei Co. Ltd.), H. SATO (Geographical Survey Institute, Japan), R. Furuta (Japan Aerospace Exploration Agency, JAXA), H. J. Liao (Xian Jiaotong University) are appreciated for their logistic and technical cooperation. 5. A. A. Lagmay, J. B. T. Ong, D. F. Fernandez, M. R. Lapus, R. S. Rodolfo, A. P. Tengonciang, J. L. T. Soria, E. G. Baliatan, Z. P. Quimba, C. L. Uichanco, E. R. Paguican, A. R. C. Remedio, G.R. H. Lorenzo, F. B. Avila, and W. Valdibia. Eos, 84-12, 121 and 124 (2006). 6. S. G. Catane, H.B. Cabria, C. P. Tomarong, Jr., R. M. Saturay, Jr., M. A. H. Zarco and W. C. Pioquinto. Catastrophic rockslide-debris avalanche at St. Bernard, Southern Leyte, Philippines. Landslides: Journal of the International Consortium on Landslides (in press) 7. K. Sassa. Proc. 5th International Symposium on Landslides, Vol. 1, 37-55, (1988) 8. K. Sassa. Proc. 6th International Symposium on landslides, 1919-1937 (1992) 9. K. Sassa, H. Fukuoka, G. Wang, N. Ishikawa. Landslides, Landslides, 1-1, 7-19 (2004) 10. K. Sassa, G. Wang, H. Fukuoka,F. Wang, T. Ochiai, M. Sugiyama, T. Sekiguchi. Landslides, 1-3, 221-235 (2004) 11. K. Sassa, H. Fukuoka, F. Wang, G. Wang. Landslides, 2-2, 125-134 (2005) 12. K.Sassa. Proc. GeoEng2000, Vol.1, 1671-1702 (2000) 13. D. Boldini, F. Wang, K. Sassa, P. Tommasi. In Landslides-risk analysis and sustainable disaster management (Eds: K. Sassa, H. Fukuoka, F. Wang and G. Wang), Springer, 173-180 (2005) 14. A.W. Bishop, G. E. Green, V. K. Garga, A. Anderson, J. D. Brown. G.otechnique, 21-1, 273-328 (1971)

Extended Abstract

15. O. Hungr, N.R. Morgenstern. Gotechnique, 34-3, 415-421 (1984) 16. V.K. Garga, J.I. Sendano. Geotechnical Testing Journal, 25-4, 414-421 (2002) 17. Y. Fukushima, T. Tanaka. Bull. Seism. Soc. Am., 84, 757-783(1990) 18. Philippine Atmospheric, Geophysical and Astronomical Services Agency (PAGASA), Precipitation record at Otikon, Libagon (2006) 19. Y. Hong, H. Hiura, K. Shino, K. Sassa, H. Fukuoka. Landslides, 2-1, 31-41 (2005) 20. K. Sassa, H. Fukuoka, G. Scarascia-Mugnozza, S. Evans. Soils and Foundation, Special issue, 53-64 (1996) 21. K. Sassa. Proc. 7th International Symposium on Landslides. 115-132 (1996)