Environmental Earth Sciences

Environmental Earth Sciences Landscape evolution analysis of deep-seated landslides at Donao Peak, Taiwan --Manuscript Draft-Manuscript Number:

ENGE-D-14-01552R1

Full Title:

Landscape evolution analysis of deep-seated landslides at Donao Peak, Taiwan

Article Type:

Original Manuscript

Corresponding Author:

Chia-Ming Lo, Ph.D. Chienkuo Technology University Changhua City, TAIWAN

Corresponding Author Secondary Information: Corresponding Author's Institution:

Chienkuo Technology University

Corresponding Author's Secondary Institution: First Author:

Ching-Fang Lee, Ph.D.

First Author Secondary Information: Order of Authors:

Ching-Fang Lee, Ph.D. Chia-Ming Lo, Ph.D. Hsien-Ter Chou, Ph.D. Shu-Yeong Chi, Ph.D.

Order of Authors Secondary Information: Funding Information: Abstract:

Typhoon Megi (2010) and the co-movement of the concurrent northeast monsoon brought massive rainfall to the Suao area of Yilan County, Taiwan, causing clusters of sediment-related landslide disasters on Provincial Highway No. 9. The most notable of these events was the large-scale landslide on the upper slope at 115.9 km near DonAo Peak, which dumped 2.1 million cubic meters of sediment into the streambed. Rainfall runoff turned this into a debris flow forming an alluvial fan at the river mouth. This study analyzed the evolution of landscapes in the area through a field investigation, disaster-causing mechanisms, image interpretation, and airborne LiDAR. Our results indicate that the landslide was associated with its location at a lithological junction as well as local geological structures. Interpretation of micro-photography revealed that the topographical changes in landslide areas in the Dakeng Stream catchment are controlled by the headward development of erosion gullies and the concave shape of the slopes. Previous earthquakes and rainfall exceeding that of a 200-year event were the external precipitating factors. Key Words: Don-Ao Peak, large-scale landslide, debris flow, landslide microphotography interpretation, shear zone.

Response to Reviewers:

RESPONSE TO REVIEWER’S COMMENTS Journal: Environmental Earth Sciences Manuscript #: ENGEO-6284 Title: Landscape evolution analysis of deep-seated landslides at Donao Peak Authors: Ching-Fang Lee, Chia-Ming Lo, Hsien-Ter Chou, Shu-Yeong Chi The manuscript has been revised according to the chief editor and reviewers’ comments. The detailed modification and response are described as follows: Editor's summary: Reviews have been received on your manuscript listed above which you submitted to the journal of Environmental Earth Sciences for consideration. The comments of the

Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation

reviewers are included at the bottom of this email. Some major revisions to your paper have been recommended prior to its reconsideration for publication. If you are prepared to undertake the work required, we would be pleased to evaluate your revised paper. I invite you to respond to these comments and revise your manuscript accordingly. Response: The authors are grateful for the comments and suggestions of the editor and reviewers. The manuscript has been revised according to the reviewers’ kind advices. A list of point-to-point replies to reviewers’ comments is attached according to your request.

Reviewer #1: (1)Label the symbol "R" in Fig.2. How are the transect section and Response: Thank you for the comment. We have labeled the symbol “R” and “I” in Fig.2. R and I represent the effective cumulative precipitation and rainfall intensity at the time for occurrence of landslide, respectively. Fig. 2 Rainfall histogram of landslide disasters: (a) Typhoon Megi (2010), (b) Typhoon Nalgae (2011), and (c) the 0512 rainfall event (2012)

(2)Expound that how to determine the effective cumulative precipitation. Response: Thank you for the comment. The criterion (or called segment principle) of total effective cumulative precipitation (Rtotal in Fig.2(a)) of rainfall event is defined as the rule of 4 mm (beginning)-6 hr(duration)-4 mm(ending). It means the rainfall value starts at the any rainfall events which are greater than 4 mm. The whole cumulative rainfall value ends once it is lower than 4 mm during a successive 6 hours (Lee, (2006)). Accordingly, the effective cumulative precipitation for triggering landslide (R) is calculated from the time of rainfall start to occurrence of landslide (Figs. 2(a)-(c)). ※The above sentences have been added on Page 5.

(3) The lines of landslide zone, tension crack and erosion gully are not clear in the Fig.3. Bold the lines. Response: Thank you for the comment. We have enhanced the outline of landslide zone, tension crack, and erosion gully in Figs. 3(a)-(f). All of the landform mappings in Figs.3 are improved to get a better visualization. Fig. 3 Comparison of remote sensing images at 115.9K on Route.9 between 20042013 (FS-2, SPOT 5)

(4)Supplement the influence of earthquake in the large-scale landslide at Don-Ao Peak in detail. Response: Thank you for the comment. We have added the description of the influence for early earthquake and Fig. 3 as below. “In order to understand the influence of early earthquake on the landslide events around the study area, this work collects the earthquake inventory from 1900 to 2011 (Central Weather Bureau). We extract the data which the Richter magnitude scale (ML) is greater than 5.0 from the surrounding area. The maximum magnitude of earthquake recorded here is approximately to 6.5. Accordingly, the landslide on the unstable hillslope can be triggered directly by earthquake while ML equals to or exceeds 5 (peak ground acceleration >250 gal) (Lee, 2014). The landslide record of the study area (2008/7-2012/12) is compared with the both daily rainfall and earthquake inventory (ML Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation

>5.0) and shown in Fig. 3. The result indicates that there are three important earthquake events happened during 2010/7-2011/5 but they don’t induce any landslide disasters immediately. It is worth to mention that the earthquake event occurred on Oct 3, 2010 before the rage of Typhoon Megi in Taiwan (No. 109, ML =5.09, epicenter: located on south-east area of Donao and the distance between Donao and epicenter is 20 km; Central Weather Bureau). We suppose the seismic intensity will affect slope stability of hillslope along Route 9 although such magnitude of earthquake doesn’t cause regional landslides at the same time (seismic intensity of this case ranges from IV to V grade in the study area). Great ground vibration by earthquake can enlarge the existing crack and decrease the stability of natural slope, as a result of which several acute landslide disasters were induced and brought casualties along the Route 9 after Typhoon Megi on Oct 22, 2010 (7 major landslide events happened, 29 people missing in 1 day). This case highlights the influence of post-earthquake on the vulnerable slope and steep valley covered with loose colluvium in Route 9. The ground motion of earthquake was the occasion of many apparent landslides when torrent rainfalls drop and infiltrate into surface crack or soil. For the study case, one can conclude that the both pre-earthquake and heavy rainfall may help triggering and expanding the regional landslide disasters on the coastal mountain highway.” Fig. 3 Time relationship among the magnitude of earthquake, cumulative rainfall, and number of landslide (2008/07/05-2013/01/05). ※The above sentences have been added on Page 6-7. A new figure (Fig. 3) was also added.

Reviewer #2 The authors present a case study dealing with the investigation of the occurrence of large mass movements. The article presents a lot of data and information. The content of the paper can be considered in the range of the journal of Environmental Earth Sciences. However, the paper needs many important revisions before its publication. (1)The reviewed paper needs a general reorganization. The topics of the paper need to be rewriting to express better the main steps of a scientific research (introduction/objectives, previous works, method, results and conclusions). Response: Thank you for the comment and suggestion. The topics of paper have been reorganized substantially to publish as a scientific paper. The outline has rewritten into the form as follows: (1) Introduction; (2) study area; (3) methodology; (4) results and discussions; and (5) conclusions. Please check the framework in the revised version.

(2)The current wording of the paper is very descriptive and tiring. I suggest that the authors use more tables to synthesize data and information. Response: Thank you for the comment and suggestion. In order to clarify the results we analyzed in the study, the new tables have added for summarized the important results. Table 3 Typhoon scenarios for landslide events at Don-Ao Peak from 2009-2012 Table 4 Irregularities associated with gravitation deformation areas (From Soeters and van Westen, 1996; Chigira, 2014) ※The above tables and corresponding descriptions have been added on Page 6, 7. The new tables (Tables 3 and 4) were also added.


(3)The authors seem to make a conceptual confusion between micro-photography and micro-topography. The surveys with LIDAR provide detailed topography. Authors need to better highlight the methodological advances in the case study presented. The use of surveys with LIDAR in the investigation of slope movements is already quite common, even in less developed countries economically and scientifically. Response: Thank you for the comment. This study adopts the shading relief with sky view factor and references features associated with gravitation deformation areas described in Soeters & van Westen (1996) and Chigira (2014) (Table 4) to identify the boundary of the gravitation deformation area in the Dakeng bridge watershed. The identified area is then entered into the landslide area-volume empirical equation to compute the probable landslide volume. That volume is then input into the numerical model to estimate the zone affected by the landslide. Interpretation of the gravitational deformation areas is shown in Fig. 15. ※The above tables and corresponding descriptions have been added on Page 19-20. The new tables (Table 4) were also added.

(4)The uniaxial compression test is not a good parameter for slope stability analysis. Shear tests or triaxial compression tests are most appropriate in this case (line 88-89, page 4). Amphibolites and schists are anisotropic rocks. Authors need to clarify whether this anisotropy influences or not the stability of slopes in the study area. Response: Thank you for your suggestion. We totally agree the rock anisotropy will influence the stability of natural slopes. The sample used for direct shear test in laboratory was acquired from barrel core. The sentence in the original paper is revised and added some physical properties of rock mass as “water content: 0.4~0.7%; specific gravity: 2.9~3.0; porosity: 0.02~0.04; water absorption: 0.6~1.7%; the peak strength of shear test is approximately 0.25 kg/cm2 on joint plane, φp: 25o; slake durability: 97.9~99.7%” ※The above sentences have been added on Page 5.

(5)The references need revision. Some articles cited in the text do not appear in the references (Central Geology Survey, 2010 for example, line 39 - page 2). The references are not in alphabetical order. Response: Thank you for your suggestion. The reference has been checked and revised again (including add several new citations). The reference is also sorted in alphabetical order.

(6)The kilometer abbreviation is km and not K. Response: Thank you for your suggestion. We have revised the K instead of km for specific milepost we mentioned in the paper.

(7)The article needs a general review of English. Response: Thank you for the comment. The version of revised paper has been submitted to revision with an English speaking editor.


Reviewer #3 The authors present an important case history from Taiwan and development of the mass movements over time due to various reasons. However, the manuscript is limited to the observations and measurements. It would be quite desirable if some materials could be added in regard with the mechanics of mass movements over time and their analytical and numerical modeling. Response: Thank you for providing those excellent publications. To explore the processes involved in the debris slide triggered by Typhoon Megi in 2010 at Su-Hua Highway 115.9k, we monitored the simulated movement speeds throughout the entire process as follows. 4.4 numerical simulation of landslide process “Simulation of landslide movement process Simulation was used to investigate the process underlying the movements involved in the large-scale debris slide triggered by Typhoon Megi in 2010 at 115.9 km of Provincial Highway No. 9. (Lo et al., 2014) To explain the processes involved, we monitored the movement speeds throughout the entire process. This revealed variations in speed during different stages and clarified the key processes involved in the overall movement. The results of numerical simulation are outlined as follows. (a) Fig. 13 presents the simulation results of the movement process and variations in speed associated with the landslide at 115.9 km. The numerical model was based on the results of subtracting pre-event from postevent DEM (pre-event: 2004; post-event: 2010). From upstream to downstream, we designated five primary landslide masses. Two of these areas were situated in the upstream source area (the sliding mass A and B), one in the midstream section (the sliding mass C), and two in the curve of the valley at the location of the collapse (the sliding mass D and E). The sliding masses A and B in the source area slumped directly downstream as a result of gravity, whereas The sliding masses C, D, and E maintained higher self supportability due to their larger coefficient of friction (equal to 0.6) and greater bonding strength. When the sliding masses A and B came into contact with the sliding masses C, D, and E downstream, the lack of support from the slope toe, which had been hollowed out by side erosion, caused them to collapse. This conforms with the landslide mechanisms derived from the onsite survey. Furthermore, this study used a PFC3D basic fluid option to install a water body downstream where the sea would be. When the particle elements encountered the body of water, their movement speed was reduced through fluid resistance, causing them to accumulate in an alluvial fan. This also enhanced the applicability of the landslide simulation. (b) These simulation results (Fig. 13) indicate that the entire landslide process progressed through seven key phases from sliding, accelerating, and decelerating to a final stop. In the first phase, the sliding masses A and B in the source area began moving (at the 5 sec point in Fig. 13). The speed of the landslide in this phase ranged between 5 m/sec and 20 m/sec (approximately 18~72 km/hr) with an average speed of 8.96 m/sec (approximately 32.3 km/hr, as shown in Fig. 15). The moving masses underwent collision interaction, in which the mass in the rear transferred energy to the mass in front. This increase in kinematic energy caused the mass in front to accelerate in its downward motion. Therefore, the masses moving the fastest were for the most part in the middle and front portions of Masses A and B, while those in the rear portions moved more slowly. (c) Second phase (15 sec point in Fig. 13): The sliding masses A and B migrated past the steepest terrain in the region, converged at the Southern Unnamed Creek, and reached the slope toe of the sliding mass C. At this point, the front portions of the sliding masses A and B were moving at the highest speed of 52.2 m/sec with an overall average of 20.4 m/sec, as shown in Fig. 15. The rear portions were also beginning their accelerated descent. (d) Third phase (22.5 sec point in Fig. 13): The front portions of the sliding masses A and B arrived at the highway where the Chuang-Yi tour group bus was parked roadside. Because the terrain in the valley turns gradually from southeast to south, the front portion of the sliding mass began slowing down (the average speed reduced to 10.36 m/sec). Here, the landslide gradually transformed from a large-scale debris slide to debris flow, and due to side erosion in the slope toe. The sliding mass C became unstable and began collapsing in the downstream direction. (e) Fourth phase (40 sec point in Fig. 13): By this time, the sliding mass C had completely collapsed, and the front portions of the sliding masses A and B had reached the sliding mass D. In terms of terrain, the sliding mass D was situated on an undercut slope at the bend of the watercourse. This caused the front Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation

portions of the sliding masses A and B to accelerate again (increasing to 27.5 m/sec) and quicken the erosion at the slope toe of Mass D. (f) Fifth phase (60 sec point in Fig. 13): Due to side erosion, the sliding mass D collapsed, and the front portions of the sliding masses A and B had passed another bend in the creek past the toe of the sliding mass E. Severe side erosion caused by the sliding masses from upstream severely destabilized Mass E, which was located on the lower slope of Su-Hua Highway 115.8 km. Sixth phase (70 sec point in Fig. 13): By this time, the sliding mass E had completely collapsed, and a portion of the sliding masses were beginning to pass the second bend before the reaching the sea. Due to the curve of the valley, the masses accelerated once more, reaching average speeds of 12.40-12.57 m/sec during the period between 70 sec and 77.5 sec. At approximately 77.5 sec, the speeds began gradually decelerating. At 70 sec, the front portions of the sliding masses A and B had reached the sea and were also slowing down in this body of water. In the seventh phase (220 sec point in Fig. 13), an alluvial fan formed, ending the movement of the landslide event at approximately 220 sec. Fig. 13 Movement process and speed variations in the simulated landslide at 115.9 km of Provincial Highway No. 9. ※The above sentences have been added on Page 17-19. A new figure (Fig. 13) was also added.

Some of additional comments on this manuscript for improvements are as follow: (1) Literature survey is insufficient. The authors are recommended to check some publications on the failure mechanism of slopes and mass movements such as Aydan et al. (1989), Aydan and Kawamoto (1992), Tokashiki and Aydan (2010), Aydan et al. (2009, 2011), Ulusay et al. (2007) etc.. Response: Thank you for providing those excellent publications. We have read and added some literature including Aydan et al. (1989), Aydan and Kawamoto (1992), Aydan et al. (2009, 2011), Ulusay et al. (2007), Sassa et al. (2005), Peyret et al. (2008) into introduction as follows. The corresponding publications are also cited in the reference. “Extreme changes in the climate have increased the number of Typhoons bearing torrential rains, which have caused major disasters in the mountainous areas of Taiwan, in the form of landslide and mass movement. These events lead to slope erosion, large-scale landslides, and debris flows within erosion gullies. With respect to the failure mechanism of landslide, Aydan et al. (1989) proposed limiting-equilibrium approach to examine the validity of stability in discontinuous rock mass. The study discussed theoretically three common failure types of rocky slope such as sliding, toppling, and combined sliding and toppling failure. The relationship among lower slope angle, friction, and inclination of the throughgoing discontinuity set were presented to evaluate the instability of rock mass by experiments in laboratory. Aydan and Kawamoto (1992) analyzed the flexural toppling failures behavior for surface slope and underground opening cases. The research presented (cantilever beams model) which suggests crack length dealing with suspended and overhanging behavior for actual rock slope. Adayn and Amini(2009) studied flexural toppling failure under dynamic loading in a layered rock mass by adopting horizontal shaking table. It found that flexural toppling failure type can be classified into active and passive modes. The seismic coefficient plays a dominant role on the initiation failure behavior (Aydan et al. (2011)). The landslide activity in past landslide sites were discussed widely for mobilized unstable mass remaining in the source area or upstream reaches (Sassa et al. (2005), Ulusay et al. (2007), Peyret et al. (2008)). Landslides and debris flow triggered by heavy rains or long storms affect the stability of slopes around mountain roads and jeopardize the safety of road users and the integrity of public facilities. Provincial Highway No. 9 (also known as the Suhua Highway) is currently the only road connecting the northern and eastern regions of Taiwan. The highway is lined with steep terrain and often runs through coastal areas. Despite on-going efforts to improve the highway, it remains pivotal route for transportation in the area. The heavy rains brought by Typhoon Parma (2009), Typhoon Megi (2010), the outer circulation of Typhoon Nalgae (2011) and the concurrent northeast monsoon, the 512 rainfall event Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation

(2012), and Typhoon Saola (2012) have caused clusters of sediment-related disasters on the slopes along Provincial Highway No. 9 between Suao (104.7 km; km is the condensation of kilometer) and Don-Ao (120.0 km). This has led to highway closures several times, and even produced in a major slope collapse that pushed a bus into the ocean causing numerous casualties (Central Geology Survey, 2011; Lee, 2010). Among the common geological disasters associated with mountain roads, Hearn (2011) listed rockfall, debris slides, deep-seated landslides, debris flow, slope cutting, toe erosion, and slope filling. Lee (2010) investigated disasters caused by Typhoon Megi in the Suao area and along the Suhua Highway, observing that the Typhoon had triggered many debris slides, only one of which was a rockslide (116.1 km, Fig. 1). The area with the greatest number of landslides was found along the upper slope at 115.9 km near Dakeng Bridge (Don-Ao Peak). Chou et al. (2012) examined the debris slide/debris flow triggered by Typhoon Megi between 115.9 km and 116.4 km on Provincial Highway No. 9 as well as the development of the coastal alluvial fan that it produced. They concluded that rainfall runoff had infiltrated the upper slopes, which caused the slope collapse and gully-type debris flow resulting in a Gilbert-type fan delta in the ocean. Lo et al. (2014) also focused on the landslide event at 115.9 km on Suhua Highway, using a three-dimensional discrete element method (Particle Flow Code in 3 Dimensions; PFC 3D) to perform numerical simulations of slope failure and discuss the kinematic processes and deposition characteristics of the debris slide. Lin et al. (2013) analyzed recent failure mechanisms and relief processes associated with roadbed collapses between 115.7 km and 116.1 km of the Suhua Highway with an assessment of the shear zone outcrops in the catchment area of the Dakeng Stream and predictions of how they may develop in the future. Lee et al. (2013) employed the potential disaster maps created by the Central Geology Survey and the Rock Mass Rating (RMR) System to section the area between Suao and Nanao of Provincial Highway No. 9 into zones of various potential for disaster. They applied slope units for zoning the highway and found that 44.0 % of the entire section is susceptible to landslides. Statistics on the number and landslide occurring in the Suao-Nanao section of Provincial Highway No. 9 between 2007 and 2012 indicate that rockfall and debris slides occur most frequent near Jiugongli (112.0 km) and Don-Ao Peak (115.9 km). Since Typhoon Megi in 2010, heavy rains from Typhoons continue to cause sedimentrelated disasters, indicating that these disasters are closely associated with local geological structures. In this paper, we analyzed firstly a series of satellite images and in-situ survey to clarify the characteristics of landslide site along a steep coastal mountain highway. Secondly, the potential landslide activity in Don-Ao Peak was examined by specific shading relief and micro-topography interpretation (Sekiguchi and Sato, 2004; Paolo Tarolli, 2014). The major objectives of this work were: (1) to explore the temporal landscape evolution and large scale landslide and investigate failure mechanism in initiation zone; (2) to map the high susceptibility landslide zone for landslide enlargement by combining high resolution terrain model (1 m) and openness shading relief technique.” ※The above sentences have been added on Page 2-4.

(2)On the basis of the observations of the reviewer on various large scale mass movements such as Kuzulu (Turkey), Kitauebaru, Oya-kuzure, Kitamatado (Japan), Muzaffarabad, Hattian (Kashmir), the gully formation and surfacial and internal erosion as well as the degradation of rocks and discontinuities such as bedding planes, whose properties to susceptible to water over time content play important roles on the eventual mass movements. It would be quite useful if the authors add some experimental data on the susceptibility of rock units and discontinuities to water content. The geological formation is likely to be similar to Shimanto-formation in Japan. Response: Thank you for your comment and suggestion. We agree the water content of rock mass will influence the stability of natural slopes. Some physical properties of rock mass was added as “water content: 0.4~0.7%; specific gravity: 2.9~3.0; porosity: 0.02~0.04; water absorption: 0.6~1.7%; the peak strength of shear test is approximately 0.25 kg/cm2 on joint plane, φp: 25o; slake durability: 97.9~99.7%.” For the amphibolite Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation

distributed in study area, the surface rock mass indicates a good durability but shows a highly fragmentation due to remarkable weathering effect. This is why the landslide area on upstream reach still continues to expand over time. ※The above sentences have been added on Page 5.

(3)It should be also noted that the internal structure of rock mass and undulations of layered rock masses also play an important role in combined complex modes of several failure mechanism such as, flexural toppling, buckling, sliding along beddings and shearing through the layers (Aydan and Kawamoto, Aydan et al. 1989, Aydan and Amini 2010; Aydan et al. 1994, Shimizu et al. 1993 etc.). It could be informative for readers to see some discussions and information on such features and how they influence overall mass movements in this particular case. Response: Thank you for providing those excellent publications to read. We have added description to discuss the influence of internal structure on mass movement in the study area. “Figures 8(a) and 8(b) illustrate the gradual process from stable to unstable in the slope body of the left bank of Dakeng Stream, with the strata of Don-Ao schist undergoing sediment erosion and weathering. In general, weathered and foliated rock presents flexural toppling failure along the foliation plane. An intersecting cross-joint set in the aforementioned rock will generally present a blocky, toppling failure or block-flexural toppling failure when the rock is unable to withstand the tensile bending stress (Amini et al., 2012). On the basis of geologic drilling (depth (h): 15 m, located on the ridge of Don-Ao Peak), the stratigraphic column is consists of surface soil (h=0-1.6 m), amphibolite with highly weathering (h=1.6-13.2 m), and slight fractured amphibolite bedrock (h=13.2-15.0 m). The result of mechanics test points out that the rock mass is characterized by a fragmental condition and poor engineering application. The rainfall can infiltrate easily into rock mass along the surface crack. Hence, the weight of rock mass rises gradually while increasing the saturation degree of rock. The unstable rock mas on the source zonation or deposited colluvium on the hillslope begin to collapse (or sliding) once they exceed the critical condition of gravitational deformation. Additionally, toe erosion effect of debris slide enhances the volume of mass movement. It also contributes a great amount of high mobility sliding mass transforming into valley debris flow.” ※The above sentences have been added on Page 13.

(4)Ground shaking due to earthquakes over a long-term also plays an important role in the development of mass movements (Aydan et al. 2011). Earthquakes also cause passive type flexural toppling, block toppling, sliding etc, which are sometimes wrongly interpreted by geomorphologist such as Varnes, Chigira etc. as creep, sagging. Therefore, it would be quite useful for the readers to see some mechanical models on the overall mass movements of rock slopes over time in this manuscript. Response: Thank you for providing useful literature. We have added the description of the influence for early earthquake and Fig. 3 as below. “In order to understand the influence of early earthquake on the landslide events around the study area, this work collects the earthquake inventory from 1900 to 2011 (data collected form Central Weather Bureau; http://www.cwb.gov.tw/). We extract the data which the Richter magnitude scale (ML) is greater than 5.0 from the surrounding area. The maximum magnitude of earthquake recorded here is approximately to 6.5. Accordingly, the landslide on the unstable hillslope can be triggered directly by earthquake while ML equals to or exceeds 5 (peak ground acceleration >250 gal) (Lee, 2014). The landslide record of the study area (2008/7-2012/12) is compared with the both daily rainfall and earthquake inventory (ML >5.0) and shown in Fig. 3. The result indicates that there are three important earthquake events happened during 2010/72011/5 but they don’t induce any landslide disasters immediately. It is worth to mention Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation

that the earthquake event occurred on Oct 3, 2010 before the rage of Typhoon Megi in Taiwan (No. 109, ML =5.09, epicenter: located on south-east area of Donao and the distance between Donao and epicenter is 20 km; Central Weather Bureau). We suppose the seismic intensity will affect slope stability of hillslope along Route 9 although such magnitude of earthquake doesn’t cause regional landslides at the same time (seismic intensity of this case ranges from IV to V grade in the study area). Great ground vibration by earthquake can enlarge the existing crack and decrease the stability of natural slope, as a result of which several acute landslide disasters were induced and brought casualties along the Route 9 after Typhoon Megi on Oct 22, 2010 (7 major landslide events happened, 29 people missing in 1 day). This case highlights the influence of post-earthquake on the vulnerable slope and steep valley covered with loose colluvium in Route 9. The ground motion of earthquake was the occasion of many apparent landslides when torrent rainfalls drop and infiltrate into surface crack or soil. For the study case, one can conclude that the both pre-earthquake and heavy rainfall may help triggering and expanding the regional landslide disasters on the coastal mountain highway.” Fig. 3 Time relationship among the magnitude of earthquake, cumulative rainfall, and number of landslide (2008/07/05-2013/01/05). ※The above sentences have been added on Page 6-7. A new figure (Fig. 3) was also added.

# Editor-in-Chief's comments: References: Please be sure that all the references cited in the manuscript are also included in the reference list and vice versa with matching spellings and dates. Please see “Instructions for Authors” on this website for information about references and their proper journal format. When citing references by the same author(s) in a sequence, only use the year for the second, third, etc. The entries to the reference list must be alphabetized according to the last (family) name of the first author in each entry. The authors are grateful for the kind advices and helps from the reviewers. Response: Thank you for the comment. The reference has been checked and revised again (including add several new citations). The reference is also sorted in alphabetical order according to the “Instructions for Authors”.


Cover Letter

Chia-Ming Lo Associate Professor Dept. of Civil Engineering Chienkuo Technology University 1, Chieh Shou N. Road Changhua City 500, TAIWAN, R.O.C

20 June 2015 Editors-in-chief Editorial Office, Environmental Earth Sciences

Dear Sir,

I am submitting a manuscript entitled "Landscape evolution analysis of deep-seated landslides at Donao Peak, Taiwan", which has been major revised according to the reviewers’ kind advices. Meanwhile, the English of this paper has been polished by a native English speaker. A list of point-to-point replies to reviewers’ comments is attached according to your request.

If you have any questions, please do not hesitate to contact me. Many thanks for your attention and kindness in reviewing this manuscript.

Yours sincerely,

Chia-Ming Lo

Title of Manuscript: Landscape evolution analysis of deep-seated landslides at Donao Peak Correspondence: Contact Person: Associate Prof. Chia-Ming Lo Postal Address:

Department of Civil Engineering, Chienkuo Technology University. 1, Chieh Shou N. Road Changhua City 500, TAIWAN, R.O.C Telephone: +886-4-711-1111 ext. 3429 Fax: +886-4-7111165 E-mail: [email protected] or [email protected]

Authors' Response to Reviewers' Comments Click here to download Authors' Response to Reviewers' Comments: 02_Revision-notes_20150519.doc

RESPONSE TO REVIEWER’S COMMENTS Journal: Environmental Earth Sciences Manuscript #: ENGEO-6284 Title: Landscape evolution analysis of deep-seated landslides at Donao Peak Authors: Ching-Fang Lee, Chia-Ming Lo, Hsien-Ter Chou, Shu-Yeong Chi The manuscript has been revised according to the chief editor and reviewers’ comments. The detailed modification and response are described as follows:

Editor's summary: Reviews have been received on your manuscript listed above which you submitted to the journal of Environmental Earth Sciences for consideration. The comments of the reviewers are included at the bottom of this email. Some major revisions to your paper have been recommended prior to its reconsideration for publication. If you are prepared to undertake the work required, we would be pleased to evaluate your revised paper. I invite you to respond to these comments and revise your manuscript accordingly. Response: The authors are grateful for the comments and suggestions of the editor and reviewers. The manuscript has been revised according to the reviewers’ kind advices. A list of point-to-point replies to reviewers’ comments is attached according to your request.

Reviewer #1: (1) Label the symbol "R" in Fig.2. How are the transect section and Response: Thank you for the comment. We have labeled the symbol “R” and “I” in Fig.2. R and I represent the effective cumulative precipitation and rainfall intensity at the time for occurrence of landslide, respectively.

(a) Typhoon Megi(2010) - Donao station 1000

rainfall intensity, I [mm/hr]

120

approximate time while landslide occurring I=121.0 mm/hr, R=704.7 mm

precipitation effective cumulative precipitation

140

800

2010/10/19

R

100

600

R total 80

400

60 40

200 20 0

0 18

0

6

12

18

0

6

12

18

effective cumulative precipitation, R [mm]

160

24

Time [hr]

(b) Typhoon Nalgae(2011) - Donao Peak station


120

1000



800

R

100

600

80 60

400

2011/9/30 40

200 20 0

0 0

6

12

18

0

6

12

18


140

0

Time [hr]

(c) Rainfall event (2012/05/12) - Donao Peak station

precipitation [mm]



600

R

60

400 2012/05/13 40 2012/05/12 200 20

0

effective cumulative precipitation [mm]

80

0 22 23 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 11

Time [hr]

Fig. 2 Rainfall histogram of landslide disasters: (a) Typhoon Megi (2010), (b) Typhoon Nalgae (2011), and (c) the 0512 rainfall event (2012)

(2) Expound that how to determine the effective cumulative precipitation. Response: Thank you for the comment. The criterion (or called segment principle) of total effective cumulative precipitation (Rtotal in Fig.2(a)) of rainfall event is defined as the rule of 4 mm (beginning)-6 hr(duration)-4 mm(ending). It means the rainfall value starts at the any rainfall events which are greater than 4 mm. The whole cumulative rainfall value ends once it is lower than 4 mm during a successive 6 hours (Lee, (2006)). Accordingly, the effective cumulative precipitation for triggering landslide (R) is calculated from the time of rainfall start to occurrence of landslide (Figs. 2(a)-(c)). ※The above sentences have been added on Page 5.

(3) The lines of landslide zone, tension crack and erosion gully are not clear in the Fig.3. Bold the lines. Response: Thank you for the comment. We have enhanced the outline of landslide zone, tension crack, and erosion gully in Figs. 3(a)-(f). All of the landform mappings in Figs.3 are improved to get a better visualization.

Fig. 3 Comparison of remote sensing images at 115.9K on Route.9 between 20042013 (FS-2, SPOT 5)

(4) Supplement the influence of earthquake in the large-scale landslide at Don-Ao Peak in detail. Response: Thank you for the comment. We have added the description of the influence for early earthquake and Fig. 3 as below. “In order to understand the influence of early earthquake on the landslide events around the study area, this work collects the earthquake inventory from 1900 to 2011 (Central Weather Bureau). We extract the data which the Richter magnitude scale (ML) is greater than 5.0 from the surrounding area. The maximum magnitude of earthquake recorded here is approximately to 6.5. Accordingly, the landslide on the unstable hillslope can be triggered directly by earthquake while ML equals to or exceeds 5 (peak ground acceleration >250 gal) (Lee, 2014). The landslide record of the study area (2008/7-2012/12) is compared with the both daily rainfall and earthquake inventory (ML >5.0) and shown in Fig. 3. The result indicates that there are three important earthquake events happened during 2010/7-2011/5 but they don’t induce any landslide disasters immediately. It is worth to mention that the earthquake event occurred on Oct 3, 2010 before the rage of Typhoon Megi in Taiwan (No. 109, ML =5.09, epicenter: located on south-east area of Donao and the distance between Donao and epicenter is 20 km; Central Weather Bureau). We suppose the seismic intensity will affect slope stability of hillslope along Route 9 although such magnitude of earthquake doesn’t cause regional landslides at the same time (seismic intensity of this case ranges from IV to V grade in the study area). Great ground vibration by earthquake can enlarge the existing crack and decrease the stability of natural slope, as a result of which several acute landslide disasters were induced and brought casualties along the Route 9 after Typhoon Megi on Oct 22, 2010 (7 major landslide events happened, 29 people missing in 1 day). This case highlights the influence of post-earthquake on the vulnerable slope and steep valley covered with loose colluvium in Route 9. The ground motion of earthquake was the occasion of many apparent landslides when torrent rainfalls drop and infiltrate into surface crack or soil. For the study case, one can conclude that the both pre-earthquake and heavy rainfall may help triggering and expanding the regional landslide disasters on the coastal mountain highway.”

2008/7/05

Num. of landslides

Cumulative rainfall [mm]

Earthquake magnitude, M

7 6 5 4 3 2 1 0 1600 1400 1200 1000 800 600 400 200 0 8 6

2009/02/05

2009/09/05

2010/03/05

2010/10/05

2011/05/05

2011/12/05

2012/06/05

2013/01/05

5.81

Earthquake

5.25 5.09

typhoon Megi

Rainfall

typhoon Nalgae

typhoon Parma

Rainfall 512

Landslides

4 2 0 2008/7/05 2008070101

2008070102 2009/02/05

2008070103 2009/09/05

2008070104 2010/03/05

2008070105 2010/10/05

2008070106 2011/05/05

2008070107 2011/12/05

2008070108 2012/06/05

2013/01/05 2008070109

Date

Fig. 3 Time relationship among the magnitude of earthquake, cumulative rainfall, and number of landslide (2008/07/05-2013/01/05). ※The above sentences have been added on Page 6-7. A new figure (Fig. 3) was also added.

Reviewer #2 The authors present a case study dealing with the investigation of the occurrence of large mass movements. The article presents a lot of data and information. The content of the paper can be considered in the range of the journal of Environmental Earth Sciences. However, the paper needs many important revisions before its publication. (1) The reviewed paper needs a general reorganization. The topics of the paper need to be rewriting to express better the main steps of a scientific research (introduction/objectives, previous works, method, results and conclusions). Response: Thank you for the comment and suggestion. The topics of paper have been reorganized substantially to publish as a scientific paper. The outline has rewritten into the form as follows: (1) Introduction; (2) study area; (3) methodology; (4) results and discussions; and (5) conclusions. Please check the framework in the revised version.

(2) The current wording of the paper is very descriptive and tiring. I suggest that the authors use more tables to synthesize data and information. Response: Thank you for the comment and suggestion. In order to clarify the results we analyzed in the study, the new tables have added for summarized the important results. Table 3 Typhoon scenarios for landslide events at Don-Ao Peak from 2009-2012 typhoon event

Parma

Megi

Nalgae

0512 rainfall

occurrence time of landslide

2009/10/05

2010/10/21

2011/10/02

2012/05/12

typhoon track*

special track

9

-

-

accompanied effect

-

northeast monsoon

-

intensity (mm/hr)**

110.0

121.0

78.5

32.5

effective rainfall (mm)**

527.5

704.7

645.5

428.5

*

It refers to the definition of typhoon track which published by Central Weather Bureau. Both rainfall intensity and effective rainfall are calculated form the temporal value of rainfall event at occurrence of landslides. **

Table 4 Irregularities associated with gravitation deformation areas (From Soeters and van Westen, 1996; Chigira, 2014) Irregularity

Description

Step-like morphology

Step-like morphology usually forms after the landslide body has mobilized. Material upslope of the landslide scar moves downslope into the scar, forming the step like topography at the upper edge of the landslide

Semicircular niches

Semicircular niches form in the head of the landslide.

Back tilting of slope faces

Hummocky relief

Formation of cracks

Steeping of slopes

Illustration

The occurrence of back tilting slope faces indicates rotational failure. This feature is often oval or square in the horizontal surface and generally occurs in poorly drained material. Irregular topography that generally indicates instability and past movement and may be related to shallow failures or small failures moving towards an existing landslide scar. Fresh, open cracks that are evidence of recent movement and indicate the landslide body is sliding or tipping. Often the cracks parallel the edge of a scar. Once the landslide has occurred, an over-steepened slope, or scar remains, severing the head of the landslide.

※The above tables and corresponding descriptions have been added on Page 6, 7. The new tables (Tables 3 and 4) were also added.

(3)

The authors seem to make a conceptual confusion between micro-photography and micro-topography. The surveys with LIDAR provide detailed topography. Authors need to better highlight the methodological advances in the case study presented. The use of surveys with LIDAR in the investigation of slope movements is already quite common, even in less developed countries economically and scientifically.

Response: Thank you for the comment. This study adopts the shading relief with sky view factor and references features associated with gravitation deformation areas described in Soeters & van Westen (1996) and Chigira (2014) (Table 4) to identify the boundary of the gravitation deformation area in the Dakeng bridge watershed. The identified

area is then entered into the landslide area-volume empirical equation to compute the probable landslide volume. That volume is then input into the numerical model to estimate the zone affected by the landslide. Interpretation of the gravitational deformation areas is shown in Fig. 15. ※The above tables and corresponding descriptions have been added on Page 19-20. The new tables (Table 4) were also added.

(4) The uniaxial compression test is not a good parameter for slope stability analysis. Shear tests or triaxial compression tests are most appropriate in this case (line 88-89, page 4). Amphibolites and schists are anisotropic rocks. Authors need to clarify whether this anisotropy influences or not the stability of slopes in the study area. Response: Thank you for your suggestion. We totally agree the rock anisotropy will influence the stability of natural slopes. The sample used for direct shear test in laboratory was acquired from barrel core. The sentence in the original paper is revised and added some physical properties of rock mass as “water content: 0.4~0.7%; specific gravity: 2.9~3.0; porosity: 0.02~0.04; water absorption: 0.6~1.7%; the peak strength of shear test is approximately 0.25 kg/cm2 on joint plane, φp: 25o; slake durability: 97.9~99.7%” ※The above sentences have been added on Page 5.

(5) The references need revision. Some articles cited in the text do not appear in the references (Central Geology Survey, 2010 for example, line 39 - page 2). The references are not in alphabetical order. Response: Thank you for your suggestion. The reference has been checked and revised again (including add several new citations). The reference is also sorted in alphabetical order.

(6) The kilometer abbreviation is km and not K. Response: Thank you for your suggestion. We have revised the K instead of km for specific milepost we mentioned in the paper.

(7) The article needs a general review of English. Response: Thank you for the comment. The version of revised paper has been submitted to revision with an English speaking editor. ※The above sentences have been added on Page 13. Four new figures (Fig. 15-18) were also added.

Reviewer #3 The authors present an important case history from Taiwan and development of the mass movements over time due to various reasons. However, the manuscript is limited to the observations and measurements. It would be quite desirable if some materials could be added in regard with the mechanics of mass movements over time and their analytical and numerical modeling. Response: Thank you for providing those excellent publications. To explore the processes involved in the debris slide triggered by Typhoon Megi in 2010 at Su-Hua Highway 115.9k, we monitored the simulated movement speeds throughout the entire process as follows.

4.4 numerical simulation of landslide process “Simulation of landslide movement process Simulation was used to investigate the process underlying the movements involved in the large-scale debris slide triggered by Typhoon Megi in 2010 at 115.9 km of Provincial Highway No. 9. (Lo et al., 2014) To explain the processes involved, we monitored the movement speeds throughout the entire process. This revealed variations in speed during different stages and clarified the key processes involved in the overall movement. The results of numerical simulation are outlined as follows. (a) Fig. 13 presents the simulation results of the movement process and variations in speed associated with the landslide at 115.9 km. The numerical model was based on the results of subtracting pre-event from post-event DEM (pre-event: 2004; post-event: 2010). From upstream to downstream, we designated five primary landslide masses. Two of these areas were situated in the upstream source area (the sliding mass A and B), one in the midstream section (the sliding mass C), and two in the curve of the valley at the location of the collapse (the sliding mass D and E). The sliding masses A and B in the source area slumped directly downstream as a result of gravity, whereas The sliding masses C, D, and E maintained higher self supportability due to their larger coefficient of friction

(equal to 0.6) and greater bonding strength. When the sliding masses A and B came into contact with the sliding masses C, D, and E downstream, the lack of support from the slope toe, which had been hollowed out by side erosion, caused them to collapse. This conforms with the landslide mechanisms derived from the onsite survey. Furthermore, this study used a PFC3D basic fluid option to install a water body downstream where the sea would be. When the particle elements encountered the body of water, their movement speed was reduced through fluid resistance, causing them to accumulate in an alluvial fan. This also enhanced the applicability of the landslide simulation. (b) These simulation results (Fig. 13) indicate that the entire landslide process progressed through seven key phases from sliding, accelerating, and decelerating to a final stop. In the first phase, the sliding masses A and B in the source area began moving (at the 5 sec point in Fig. 13). The speed of the landslide in this phase ranged between 5 m/sec and 20 m/sec (approximately 18~72 km/hr) with an average speed of 8.96 m/sec (approximately 32.3 km/hr, as shown in Fig. 15). The moving masses underwent collision interaction, in which the mass in the rear transferred energy to the mass in front. This increase in kinematic energy caused the mass in front to accelerate in its downward motion. Therefore, the masses moving the fastest were for the most part in the middle and front portions of Masses A and B, while those in the rear portions moved more slowly. (c) Second phase (15 sec point in Fig. 13): The sliding masses A and B migrated past the steepest terrain in the region, converged at the Southern Unnamed Creek, and reached the slope toe of the sliding mass C. At this point, the front portions of the sliding masses A and B were moving at the highest speed of 52.2 m/sec with an overall average of 20.4 m/sec, as shown in Fig. 15. The rear portions were also beginning their accelerated descent. (d) Third phase (22.5 sec point in Fig. 13): The front portions of the sliding masses A and B arrived at the highway where the Chuang-Yi tour group bus was parked roadside. Because the

terrain in the valley turns gradually from southeast to south, the front portion of the sliding mass began slowing down (the average speed reduced to 10.36 m/sec). Here, the landslide gradually transformed from a large-scale debris slide to debris flow, and due to side erosion in the slope toe. The sliding mass C became unstable and began collapsing in the downstream direction. (e) Fourth phase (40 sec point in Fig. 13): By this time, the sliding mass C had completely collapsed, and the front portions of the sliding masses A and B had reached the sliding mass D. In terms of terrain, the sliding mass D was situated on an undercut slope at the bend of the watercourse. This caused the front portions of the sliding masses A and B to accelerate again (increasing to 27.5 m/sec) and quicken the erosion at the slope toe of Mass D. (f) Fifth phase (60 sec point in Fig. 13): Due to side erosion, the sliding mass D collapsed, and the front portions of the sliding masses A and B had passed another bend in the creek past the toe of the sliding mass E. Severe side erosion caused by the sliding masses from upstream severely destabilized Mass E, which was located on the lower slope of Su-Hua Highway 115.8 km. Sixth phase (70 sec point in Fig. 13): By this time, the sliding mass E had completely collapsed, and a portion of the sliding masses were beginning to pass the second bend before the reaching the sea. Due to the curve of the valley, the masses accelerated once more, reaching average speeds of 12.40-12.57 m/sec during the period between 70 sec and 77.5 sec. At approximately 77.5 sec, the speeds began gradually decelerating. At 70 sec, the front portions of the sliding masses A and B had reached the sea and were also slowing down in this body of water. In the seventh phase (220 sec point in Fig. 13), an alluvial fan formed, ending the movement of the landslide event at approximately 220 sec.

Fig. 13 Movement process and speed variations in the simulated landslide at 115.9 km of Provincial Highway No. 9. ※The above sentences have been added on Page 17-19. A new figure (Fig. 13) was also added.

Some of additional comments on this manuscript for improvements are as follow: (1) Literature survey is insufficient. The authors are recommended to check some publications on the failure mechanism of slopes and mass movements such as Aydan et al. (1989), Aydan and Kawamoto (1992), Tokashiki and Aydan (2010), Aydan et al. (2009, 2011), Ulusay et al. (2007) etc.. Response: Thank you for providing those excellent publications. We have read and added some literature including Aydan et al. (1989), Aydan and Kawamoto (1992), Aydan et al. (2009, 2011), Ulusay et al. (2007), Sassa et al. (2005), Peyret et al. (2008) into introduction as follows. The corresponding publications are also cited in the reference.

“Extreme changes in the climate have increased the number of Typhoons bearing torrential rains, which have caused major disasters in the mountainous areas of Taiwan, in the form of landslide and mass movement. These events lead to slope erosion, large-scale landslides, and debris flows within erosion gullies. With respect to the failure mechanism of landslide, Aydan et al. (1989) proposed limiting-equilibrium approach to examine the validity of stability in discontinuous rock mass. The study discussed theoretically three common failure types of rocky slope such as sliding, toppling, and combined sliding and toppling failure. The relationship among lower slope angle, friction, and inclination of the throughgoing discontinuity set were presented to evaluate the instability of rock mass by experiments in laboratory. Aydan and Kawamoto (1992) analyzed the flexural toppling failures behavior for surface slope and underground opening cases. The research presented (cantilever beams model) which suggests crack length dealing with suspended and overhanging behavior for actual rock slope. Adayn and Amini(2009) studied flexural toppling failure under dynamic loading in a layered rock mass by adopting horizontal shaking table. It found that flexural toppling failure type can be classified into active and passive modes. The seismic coefficient plays a dominant role on the initiation failure behavior (Aydan et al. (2011)). The landslide activity in past landslide sites were discussed widely for mobilized unstable mass remaining in the source area or upstream reaches (Sassa et al. (2005), Ulusay et al. (2007), Peyret et al. (2008)). Landslides and debris flow triggered by heavy rains or long storms affect the stability of slopes around mountain roads and jeopardize the safety of road users and the integrity of public facilities. Provincial Highway No. 9 (also known as the Suhua Highway) is currently the only road connecting the northern and eastern regions of Taiwan. The highway is lined with steep terrain and often runs through coastal areas. Despite on-going efforts to improve the highway, it remains pivotal route for transportation in the area. The heavy rains brought by Typhoon Parma (2009), Typhoon Megi (2010), the outer circulation of Typhoon Nalgae (2011) and the concurrent northeast monsoon, the 512 rainfall event (2012), and Typhoon Saola (2012) have caused clusters of sediment-related disasters on the slopes along Provincial Highway No. 9 between Suao (104.7 km; km is the condensation of kilometer) and Don-Ao (120.0 km). This has led to highway closures several times, and even produced in a major slope collapse that pushed a bus into the ocean causing numerous casualties (Central Geology Survey, 2011; Lee, 2010). Among the common geological disasters associated with mountain roads, Hearn (2011) listed rockfall, debris slides, deep-seated landslides, debris flow, slope cutting, toe erosion, and slope filling. Lee (2010) investigated disasters caused by Typhoon Megi in the Suao area and along the Suhua Highway, observing that the Typhoon had triggered many debris slides, only one of which was a rockslide (116.1 km, Fig. 1).

The area with the greatest number of landslides was found along the upper slope at 115.9 km near Dakeng Bridge (Don-Ao Peak). Chou et al. (2012) examined the debris slide/debris flow triggered by Typhoon Megi between 115.9 km and 116.4 km on Provincial Highway No. 9 as well as the development of the coastal alluvial fan that it produced. They concluded that rainfall runoff had infiltrated the upper slopes, which caused the slope collapse and gully-type debris flow resulting in a Gilbert-type fan delta in the ocean. Lo et al. (2014) also focused on the landslide event at 115.9 km on Suhua Highway, using a three-dimensional discrete element method (Particle Flow Code in 3 Dimensions; PFC 3D) to perform numerical simulations of slope failure and discuss the kinematic processes and deposition characteristics of the debris slide. Lin et al. (2013) analyzed recent failure mechanisms and relief processes associated with roadbed collapses between 115.7 km and 116.1 km of the Suhua Highway with an assessment of the shear zone outcrops in the catchment area of the Dakeng Stream and predictions of how they may develop in the future. Lee et al. (2013) employed the potential disaster maps created by the Central Geology Survey and the Rock Mass Rating (RMR) System to section the area between Suao and Nanao of Provincial Highway No. 9 into zones of various potential for disaster. They applied slope units for zoning the highway and found that 44.0 % of the entire section is susceptible to landslides. Statistics on the number and landslide occurring in the Suao-Nanao section of Provincial Highway No. 9 between 2007 and 2012 indicate that rockfall and debris slides occur most frequent near Jiugongli (112.0 km) and Don-Ao Peak (115.9 km). Since Typhoon Megi in 2010, heavy rains from Typhoons continue to cause sediment-related disasters, indicating that these disasters are closely associated with local geological structures. In this paper, we analyzed firstly a series of satellite images and in-situ survey to clarify the characteristics of landslide site along a steep coastal mountain highway. Secondly, the potential landslide activity in Don-Ao Peak was examined by specific shading relief and micro-topography interpretation (Sekiguchi and Sato, 2004; Paolo Tarolli, 2014). The major objectives of this work were: (1) to explore the temporal landscape evolution and large scale landslide and investigate failure mechanism in initiation zone; (2) to map the high susceptibility landslide zone for landslide enlargement by combining high resolution terrain model (1 m) and openness shading relief technique.” ※The above sentences have been added on Page 2-4.

(2) On the basis of the observations of the reviewer on various large scale mass movements such as Kuzulu (Turkey), Kitauebaru, Oya-kuzure, Kitamatado (Japan), Muzaffarabad, Hattian (Kashmir), the gully formation and surfacial and internal erosion as well as the degradation of rocks and discontinuities such as bedding planes, whose properties to susceptible to water over time content play important roles on the eventual mass movements. It would be quite useful if the authors add some experimental data on the susceptibility of rock units and discontinuities to water content. The geological formation is likely to be similar to Shimanto-formation in Japan. Response: Thank you for your comment and suggestion. We agree the water content of rock mass will influence the stability of natural slopes. Some physical properties of rock mass was added as “water content: 0.4~0.7%; specific gravity: 2.9~3.0; porosity: 0.02~0.04; water absorption: 0.6~1.7%; the peak strength of shear test is approximately 0.25 kg/cm2 on joint plane, φp: 25o; slake durability: 97.9~99.7%.” For the amphibolite distributed in study area, the surface rock mass indicates a good durability but shows a highly fragmentation due to remarkable weathering effect. This is why the landslide area on upstream reach still continues to expand over time. ※The above sentences have been added on Page 5.

(3) It should be also noted that the internal structure of rock mass and undulations of layered rock masses also play an important role in combined complex modes of several failure mechanism such as, flexural toppling, buckling, sliding along beddings and shearing through the layers (Aydan and Kawamoto, Aydan et al. 1989, Aydan and Amini 2010; Aydan et al. 1994, Shimizu et al. 1993 etc.). It could be informative for readers to see some discussions and information on such features and how they influence overall mass movements in this particular case. Response: Thank you for providing those excellent publications to read. We have added description to discuss the influence of internal structure on mass movement in the study area. “Figures 8(a) and 8(b) illustrate the gradual process from stable to unstable in the slope body of the left bank of Dakeng Stream, with the strata of Don-Ao schist undergoing sediment erosion and weathering. In general, weathered and foliated rock

presents flexural toppling failure along the foliation plane. An intersecting cross-joint set in the aforementioned rock will generally present a blocky, toppling failure or block-flexural toppling failure when the rock is unable to withstand the tensile bending stress (Amini et al., 2012). On the basis of geologic drilling (depth (h): 15 m, located on the ridge of Don-Ao Peak), the stratigraphic column is consists of surface soil (h=0-1.6 m), amphibolite with highly weathering (h=1.6-13.2 m), and slight fractured amphibolite bedrock (h=13.2-15.0 m). The result of mechanics test points out that the rock mass is characterized by a fragmental condition and poor engineering application. The rainfall can infiltrate easily into rock mass along the surface crack. Hence, the weight of rock mass rises gradually while increasing the saturation degree of rock. The unstable rock mas on the source zonation or deposited colluvium on the hillslope begin to collapse (or sliding) once they exceed the critical condition of gravitational deformation. Additionally, toe erosion effect of debris slide enhances the volume of mass movement. It also contributes a great amount of high mobility sliding mass transforming into valley debris flow.” ※The above sentences have been added on Page 13.

(4) Ground shaking due to earthquakes over a long-term also plays an important role in the development of mass movements (Aydan et al. 2011). Earthquakes also cause passive type flexural toppling, block toppling, sliding etc, which are sometimes wrongly interpreted by geomorphologist such as Varnes, Chigira etc. as creep, sagging. Therefore, it would be quite useful for the readers to see some mechanical models on the overall mass movements of rock slopes over time in this manuscript. Response: Thank you for providing useful literature. We have added the description of the influence for early earthquake and Fig. 3 as below. “In order to understand the influence of early earthquake on the landslide events around the study area, this work collects the earthquake inventory from 1900 to 2011 (data collected form Central Weather Bureau; http://www.cwb.gov.tw/). We extract the data which the Richter magnitude scale (ML) is greater than 5.0 from the surrounding area. The maximum magnitude of earthquake recorded here is approximately to 6.5. Accordingly, the landslide on the unstable hillslope can be triggered directly by earthquake while ML equals to or exceeds 5 (peak ground acceleration >250 gal) (Lee, 2014). The landslide record of the study area (2008/7-2012/12) is compared with the both daily rainfall and earthquake inventory (ML >5.0) and shown in Fig. 3. The result

indicates that there are three important earthquake events happened during 2010/7-2011/5 but they don’t induce any landslide disasters immediately. It is worth to mention that the earthquake event occurred on Oct 3, 2010 before the rage of Typhoon Megi in Taiwan (No. 109, ML =5.09, epicenter: located on south-east area of Donao and the distance between Donao and epicenter is 20 km; Central Weather Bureau). We suppose the seismic intensity will affect slope stability of hillslope along Route 9 although such magnitude of earthquake doesn’t cause regional landslides at the same time (seismic intensity of this case ranges from IV to V grade in the study area). Great ground vibration by earthquake can enlarge the existing crack and decrease the stability of natural slope, as a result of which several acute landslide disasters were induced and brought casualties along the Route 9 after Typhoon Megi on Oct 22, 2010 (7 major landslide events happened, 29 people missing in 1 day). This case highlights the influence of post-earthquake on the vulnerable slope and steep valley covered with loose colluvium in Route 9. The ground motion of earthquake was the occasion of many apparent landslides when torrent rainfalls drop and infiltrate into surface crack or soil. For the study case, one can conclude that the both pre-earthquake and heavy rainfall may help triggering and expanding the regional landslide disasters on the coastal mountain highway.” 2008/7/05

Num. of landslides



7 6 5 4 3 2 1 0 1600 1400 1200 1000 800 600 400 200 0 8 6

2009/02/05

2009/09/05

2010/03/05

2010/10/05

2011/05/05

2011/12/05

2012/06/05

2013/01/05

5.81

Earthquake

5.25 5.09

typhoon Megi

Rainfall

typhoon Nalgae

typhoon Parma

Rainfall 512

Landslides

4 2 0 2008/7/05 2008070101

2008070102 2009/02/05

2008070103 2009/09/05

2008070104 2010/03/05

2008070105 2010/10/05

2008070106 2011/05/05

2008070107 2011/12/05

2008070108 2012/06/05

2013/01/05 2008070109

Date

Fig. 3 Time relationship among the magnitude of earthquake, cumulative rainfall, and number of landslide (2008/07/05-2013/01/05). ※The above sentences have been added on Page 6-7. A new figure (Fig. 3) was also added.

# Editor-in-Chief's comments: References: Please be sure that all the references cited in the manuscript are also included in the reference list and vice versa with matching spellings and dates. Please see “Instructions for Authors” on this website for information about references and their proper journal format. When citing references by the same author(s) in a sequence, only use the year for the second, third, etc. The entries to the reference list must be alphabetized according to the last (family) name of the first author in each entry. The authors are grateful for the kind advices and helps from the reviewers. Response: Thank you for the comment. The reference has been checked and revised again (including add several new citations). The reference is also sorted in alphabetical order according to the “Instructions for Authors”.

English revised marks Click here to download Authors' Response to Reviewers' Comments: 03_English revised_Manuscript_20150619.doc

1

Landscape evolution analysis of large scale landslides at Don-Ao Peak,

2

Taiwan

3

Ching-Fang Lee1, Chia-Ming Lo2*, Hsien-Ter Chou3, Shu-Yeong Chi1

4

1

Disaster Prevention Technology Research Center, Sinotech Engineering Consultants, Inc., Taipei 105, Taiwan, R.O.C.

5

2

Department of Civil Engineering, Chienkuo Technology University, Changhua 500, Taiwan, R.O.C.

6

3

Department of Civil Engineering, National Central University, Chung-Li 320, Taiwan, , R.O.C.

7

*Corresponding Author: E-mail: [email protected] or [email protected]

8

Abstract

9

Typhoon Megi (2010) and the co-movement of the concurrent northeast monsoon brought

10

massive rainfalls to the Suao area of Yilan County, Taiwan, causing clusters of sediment-related

11

landslide disasters on Provincial Highway No. 9. The most notable of these events was the large-

12

scale landslide on the upper slope at 115.9 km near Don-Ao Peak, which dumped 2.1 million cubic

13

meters of sediment into the streambed. Rainfall runoff turned this into a debris flow forming an

14

alluvial fan at the river mouth. This study analyzed the evolution of landscapes in the area

15

bythrough a field investigation, disaster-causing mechanisms, image interpretation, and airborne

16

LiDAR. Our results indicate that the landslide was associated with its location at a lithological

17

junction as well as local geological structures. Interpretation of micro-photography revealed that the

18

topographical changes in landslide areas in the Dakeng Stream catchment are controlled by the

19

headward development of erosion gullies and the concave shape of the slopes. Previous earthquakes

20

and rainfall exceeding than that of a 200-year event were the external precipitating factors.

21

Key Words: Don-Ao Peak, large-scale landslide, debris flow, landslide micro-photography

22

interpretation, shear zone.

1

23

1. Introduction

24

Extreme changes in the climate have increased the number of tTyphoons bringingearing

25

torrential rains, which have frequently caused major disasters in the form of landslides and mass

26

movement in the mountainous areas of Taiwan, in the form of landslide and mass movement. These

27

events can lead to slope erosion, large-scale landslides, and debris flows within erosion gullies. In

28

effortsWith respect to understand the failure mechanism of landslides, Aydan et al. (1989) proposed

29

the limiting-equilibrium approach in anto examinatione the validity of stability in discontinuous

30

rock masses. Thise current study theoretically discussesd theoretically three common failure types

31

of rocky slope failure, such as sliding, toppling, and combined sliding and toppling failure. The

32

relationship amongbetween the angle of the lower slope angle, friction, and the inclination of the

33

through-going discontinuity set awere used in anpresented to evaluation ofe the instability of rock

34

mass bythrough laboratory experiments in a laboratory. Aydan and Kawamoto (1992) analyzed the

35

flexural toppling failures behavior offor surface slopes and underground opening cases. The

36

research presented here based on the (cantilever beams model) which suggests crack length dealing

37

with suspended and overhanging behavior for actual rock slopes. Adayn and Amini (2009) studied

38

flexural toppling failure under dynamic loading in a layered rock mass in experiments withby

39

adopting horizontal shaking tables. ItThey found that the flexural toppling failure type can be

40

classified into active and passive modes. The seismic coefficient plays a dominant role on the

41

initiation of the failure behavior (Aydan et al. (2011)).

42

The landslide activity in sites where past landslide have occurredsites has beenwere discussed

43

widely regardingfor mobilized unstable mass remaining in the source area or upstream reaches

44

(Sassa et al. (2005), Ulusay et al. (2007), Peyret et al. (2008)). Landslides and debris flow triggered

45

by heavy rains or long storms affect the stability of slopes around mountain roads and can

46

jeopardize the safety of road users and the integrity of public facilities. Taiwan’s Provincial

47

Highway No. 9 (also known as the Suhua Highway), is currently the only road connecting the

48

northern and eastern regions of Taiwan,. The highway is lined with steep terrain and often runs 2

49

through coastal areas. ODespite on-going efforts are being made to improve the highway, whichit

50

remains a pivotal route for transportation in the area. The heavy rains brought by Typhoon Parma

51

(2009), Typhoon Megi (2010), the outer circulation of Typhoon Nalgae (2011) and the concurrent

52

northeast monsoon, the 512 rainfall event (2012), and Typhoon Saola (2012) have caused clusters of

53

sediment-related disasters on the slopes along Provincial Highway No. 9 between Suao (104.7 km;

54

km is the condensation of kilometer) and Don-Ao (120.0 km). This frequentlyhas leads to highway

55

closures several times, and has even producedcaused in a major slope collapse that pushed a bus full

56

of passengers into the ocean causing numerous casualties (Central Geology Survey, 2011; Lee,

57

2010).

58

Among the more common geological disasters associated with mountain roads are, Hearn

59

(2011) listed rockfalls, debris slides, deep-seated landslides, debris flow, slope cutting, toe erosion,

60

and slope filling (Hearn, 2011). Lee (2010) investigated disasters caused by Typhoon Megi in the

61

Suao area and along the Suhua Highway, observing that the tTyphoon had triggered many debris

62

slides, only one of which was a rockslide (116.1 km, Fig. 1). The area with the greatest number of

63

landslides was found along the upper slope at 115.9 km near Dakeng Bridge (Don-Ao Peak). Chou

64

et al. (2012) examined the debris slide/debris flow triggered by Typhoon Megi between 115.9 km

65

and 116.4 km on Provincial Highway No. 9 as well as the development of the coastal alluvial fan

66

that it produced. They concluded that rainfall runoff had infiltrated the upper slopes, which caused

67

the slope collapse and the gully-type debris flow resulting in a Gilbert-type fan delta in the ocean.

68

Lo et al. (2014) also focused on the landslide event that occurred at 115.9 km on the Suhua

69

Highway., They useding a three-dimensional discrete element method (Particle Flow Code in 3

70

Dimensions; PFC 3D) to perform numerical simulations of slope failure and discuss the kinematic

71

processes and deposition characteristics of the debris slide. Lin et al. (2013) analyzed recent failure

72

mechanisms and relief processes associated with roadbed collapses between 115.7 km and 116.1 km

73

of the Suhua Highway providing with an assessment of the shear zone outcrops in the catchment

74

area of the Dakeng Stream and predictions of how they mightay develop in the future. Lee et al. 3

75

(2013) employed the potential disaster maps created by the Central Geology Survey and the Rock

76

Mass Rating (RMR) System to examinesection the area between Suao and Nanao onf Provincial

77

Highway No. 9 sectioning it into zones withof various potentials for disaster. They applied slope

78

units for zoning the highway and found that 44.0 % of the entire section is susceptible to landslides.

79

Statistics on the number andof landslide occurring in the Suao-Nanao section of Provincial

80

Highway No. 9 between 2007 and 2012 indicate that rockfall and debris slides occur most

81

frequently near Jiugongli (112.0 km) and Don-Ao Peak (115.9 km). Since Typhoon Megi in 2010,

82

heavy rains from tTyphoons have continued to cause sediment-related disasters., indicating that

83

Tthese disasters are closely associated with the local geological structures. In this studypaper, we

84

first analyzed firstly a series of satellite images and in-situ surveys to clarify the characteristics of

85

landslide sites along thea steep coastal mountain highway. Secondly, the potential landslide activity

86

in the Don-Ao Peak area was examined by specific shading relief and micro-topography

87

interpretation (Sekiguchi and Sato, 2004; Paolo Tarolli, 2014). The major objectives of this work

88

awere: (1) to explore the evolution of the temporal landscape evolution and large scale landslides

89

and investigate the failure mechanism in the initiation zone; (2) to map the high susceptibility

90

landslide zones for landslide enlargement by combining a high resolution terrain model (1 m) and

91

the openness shading relief technique.

4

92

2. Study area

93

2.1 Geography and geological sSetting

94

The study area (the catchment area of the wild stream under the Dakeng Bridge) is locateds in

95

the eastern Suao, Yilan County (seein the red square in Fig. 1 (red square), bordering the Central

96

Mountain Range to the west and the Pacific Ocean to the east. Provincial Highway No. 9 winds

97

along the steep coastal foothills at elevations between 220 m and 280 m. The slope of the terrain

98

inclines to the east and the ground elevation becomes higher whileas the highway

99

approachingapproaches to the southern region. The study area is situated between 115.8 km and

100

116.4 km onf Provincial Highway No. 9 (Fig. 1). Major collapses occurred upstream of Dakeng

101

Bridge and Don-Ao Peak. Don-Ao Peak, which displays steep drops and reaches a heightpeak of

102

800 m just 1.2 km from the ocean. A 1:50,000 geological map of Suao and Nanao by the Central

103

Geology Survey indicates that the landslide area is located in an area of amphibolite outcrops

104

belonging to the Don-Ao schist stratum. The attitude of the schistosity trends to N45∘W and dips

105

to 65∘SW. In this area, the strength of the amphibolite is high (water content: 0.4-~0.7%; specific

106

gravity: 2.9-~3.0; porosity: 0.02-~0.04; water absorption: 0.6-~1.7%; the peak strength of the shear

107

test is approximately 0.25 kg/cm2 on the joint plane, φp: 25o; slake durability: 97.9-~99.7%).;

108

Hhowever, severe weathering has resulted in extreme fragmentation. Disaster potential maps by the

109

Central Geology Survey indicate a high potential for debris slides and rockfalls in the area. In the

110

upstream area of Dakeng Stream, the slide-prone segment contains graphite mica schist belonging

111

to the Don-Ao schist stratum. The terrain here presents concave-shaped slopes that often collect

112

rainfall runoff.

113 114

Fig. 1 Geological settings forof the study area (115.9 km onf Provincial Highway No. 9).

115

2.2 Hydrometeorological conditions

116

The study area is facesing windward during the East Asia tropical monsoon, and haswith an 5

117

average annual rainfall of 4,480 mm (1971-2009). Statistics from the Suao Meteorological Station

118

(1971-2009, Table 1) indicate that the temperature is highest in July (32.0 C) and lowest in January

119

(12.9 C) with an average of 210. The number of rainy days per year averages 210., Thewith a

120

higher distribution of precipitation is higher on the windward slopes than on the leeward slopes,

121

mostly between October and December. The northeast monsoon lasts from October to January and

122

generally brings extended periods of light rain. Typhoons with heavy rain are most common in

123

summer. In July, 2011, the Central Weather Bureau added a new meteorological station, the Don-Ao

124

Peak station, on the upper slope at the 115.9 km mark onf Provincial Highway No. 9. Rainfall

125

statistics in the following two years revealed a staggering average annual rainfall of 6,558 mm

126

(Table 1), which is 1.5 times that of the rainfall at the Suao station. This is an indication that rather

127

than distinct flood/drought periods, this area has stable, plentiful precipitation during the rainy

128

season (in May and June), Typhoon season (from August to October), and during the northeast

129

monsoon (from October to February next year).

130

Table 1 Mmonthly rainfall statistics from Suao and Don-Ao Peak meteorological stations (from the

131

CWB, Taiwan).

132

2.3 Triggering factors for landslide events

133

According to historical disaster records, major sediment-related disasters in the catchment area

134

of Don-Ao Peak began with Typhoon Parma in 2009 (R=527mm, I=110 mm on October 5 , 2009).

135

We therefore compiled a list of major rainfall events that caused local slope collapses between 2009

136

and 2013. Fig. 2 presents the histograms and accumulated rainfall accumulations forof the

137

individual events. The criterion (or called segment principle) of total effective cumulative

138

precipitation (Rtotal in Fig. 2(a)) of a rainfall event is defined as the rule of 4 mm (beginning) - 6 hr

139

(duration) - 4 mm (ending). ItThis means that the rainfall value starts withat the any rainfall events

140

which are greater than 4 mm. The whole cumulative rainfall value ends once it is lower than 4 mm

141

during a successive 6 hours (Lee, (2006)). Accordingly, the effective cumulative precipitation for

142

triggering landslides (R) is calculated from the time ofthe rainfall starts to the occurrence of a 6

143

landslide (Figs. 2(a)-(c)).

144

Among the examples examined these, Typhoon Megi (2010) and Typhoon Saola (2012)

145

broughtincurred torrential rains and landslides. Typhoon Megi is considered an important disaster-

146

causing rainfall event with the accumulation atin which Don-Ao accumulated a total rainfall of 939

147

mm between October 19 and 21 and the intensity of the rainfall reached 121.0 mm/hr (13:00-14:00,

148

October 21), breaking the historical rainfall records in Taiwan. Onsite surveys revealed that the

149

large-scale landslide caused by Typhoon Megi at Don-Ao Peak began sometime between 12:30 and

150

13:30 (Fig. 2(a)). Shortly thereafter, rainfall had reached that of the local 200-yr return period

151

(R>600 mm, 24 hr), further inducing regional flooding around Yilan and the Suhua Highway as

152

well as debris flows and collapses. Typhoon Megi caused 21 fatalities along the Suhua Highway.

153

The following year, Typhoon Nalgae (October 2, 2011; Fig. 2(b)), the following year, did not make

154

landfall on Taiwan, however, the effects of its outer circulation and the concurrent northeast

155

monsoon brought torrential rains with an effective accumulated rainfall of 513 mm in the study

156

area. The rain peaked in intensity of the rain peaked at 78.5 mm/hr and expanded the Don-Aoed

157

landslide-prone area toon the west and east sides of Don-Ao Peak (Fig.1(c)). The 512 rainfall event

158

(Fig. 2(c)) refers to the torrential rains in the Nanao area (located about 10 km to the south of Don-

159

Ao) caused by convective clouds on the afternoon of May 12, 2012, which is during the rainy

160

season. The Don-Ao Peak station measured a total rainfall amount of 428.5 mm with an intensity of

161

32.5 mm/hr, which caused sinking displacement in slope-top scarps (Fig.3 (f)). Rainfall intensity-

162

duration-frequency analysis, revealed that the long-term rainfall intensities triggered by Typhoon

163

Megi (2010) and Typhoon Nalgae (2011) exceeded than that of the 200-yr return period.

164

In order to understand the influence of earliery earthquakes on the landslide events in around

165

the study area, this work collects anthe earthquake inventory was collected from 1900 to 2011 (data

166

collected forom the Central Weather Bureau; http://www.cwb.gov.tw/). DWe extract the data were

167

extracted for events from the surrounding area with ain which the Richter magnitude scale (ML) is

168

greater than 5.0 from the surrounding area. The maximum magnitude of the earthquakes recorded 7

169

here is approximately to 6.5. LAccordingly, the landslides on the unstable hillslope can be triggered

170

directly by earthquakes whereile ML is equals to or exceeds 5 (peak ground acceleration > 250 gal)

171

(Lee, 2014). The landslide record of the study area (2008/7-2012/12) is compared with the both the

172

daily rainfall and earthquake inventory (ML >5.0) and is shown in Fig. 3. The results indicates that

173

there weare three important earthquake events which happened during 2010/7-2011/5 but they did

174

not immediatelyon’t induce any landslide disasters immediately. It is worth to mentioning that thean

175

earthquake event occurred on Oct 3, 2010 before the rage of Typhoon Megi struckin Taiwan (No.

176

109, ML =5.09, epicenter: located ion the area south-east area of Donao (and the distance between

177

Donao and the epicenter wais 20 km; Central Weather Bureau). We supposeassume that the seismic

178

intensity will affect the slope stability of hillslopes along Route 9 although such magnitude ofan

179

earthquake of such magnitude does notn’t necessarily cause regional landslides immediately at the

180

same time (in this case the seismic intensity was in the of this case ranges from grade IV to V grade

181

in the study area). SignificantGreat ground vibration by earthquakes can enlarge the existing cracks

182

and decrease the stability of natural slopes, as a result of which several significant acute landslide

183

disasters were induced and brought casualties along the Route 9 after Typhoon Megi on Oct 22,

184

2010 (7 major landslide events occurred withhappened, 29 people missing in 1 day). This case

185

highlights the influence of post-earthquakes on the vulnerable slopes and steep valleys covered with

186

loose colluvium alongin Route 9. The ground motion of the earthquake led towas the occasion of

187

many apparent landslides with the dropping ofhen torrential rainfalls drop thatand infiltrate into

188

surface cracks andor soil. Fromor the studiedy case, one can conclude that the combination of the

189

both pre-earthquakes and heavy rainfall may help triggering and worsenexpanding the regional

190

landslide disasters on the coastal mountain highway.

191

Fig. 2 Rainfall histograms of landslide disasters: (a) Typhoon Megi (2010);, (b) Typhoon Nalgae

192

(2011);, and (c) the 0512 rainfall event (2012).

193

Fig. 3 Time relationship amongbetween the magnitude of earthquake magnitude, cumulative

194

rainfall, and number of landslides (2008/07/05-2013/01/05). 8

195

2.4 Geomorphological interpretation

196

FThis work collected four remotely senseding satellite images and orthophotos taken between

197

2003 and 2012 were collected for this study (Fig. 4)., This time periodwhich encompasses

198

severalthe major tTyphoon events that affected the study area: Typhoon Parma (2009), Typhoon

199

Megi (2010), and Typhoon Saola (2012). The remote sensing images indicate that early in this

200

period (2003-2009), the study area was comprised of highly vegetated forest compartments (Fig.

201

4(a)). However, after Typhoon Parma (2009), a landslide area appeared emerged on the upper slope

202

at 115.9 km of Provincial Highway No. 9 (marked by thewith A in Fig. 4(b)), onat the midstream

203

section of the Dakeng Stream. Prior to Typhoon Megi, erosion gullies near the landslide areas in the

204

midstream section of Dakeng Stream displayed significant erosion, and the left upstream tributary

205

also showed signs of headward erosion. The erosion gullies on the lower slopes at 116.1 km and

206

116.8 km were progresseding towards the road, an indications of the imminent major roadbed

207

collapses between 115.8 km and 115.9 km (2011/10/27, 2012/10/24), which obstructed traffic in

208

both directions. After Typhoon Megi (2010) hit the eastern region of Taiwan, landslide areas and

209

erosion gullies became more obvious., and Tthe ditches on both sides of the lower slopes at Dakeng

210

Bridge (115.9 km) showed signs of severe lateral erosion and bed erosion, which ultimately induced

211

a roadbed collapse at 116.1 km (Fig. 4(c)). On the east side, the debris flow produced a symmetrical

212

alluvial fan in the ocean. In July, 2011, the upper rims of the two lobe-shaped scarps below Don-Ao

213

Peak showed the development of several significant tension cracks developing (to the northwest of

214

B and C in Fig. 4(d)). The lengths of these cracks ranged in length from dozens to a hundred meters,

215

and as they were scoured by heavy debris flow, the landslide on the side of the lower slope at 115.9

216

km became increasingly severe (marked by D, E, I, and J in Fig. 4(d)). Clear signs of erosion gully

217

development couldan also be seen in the upstream catchment area of the wild stream to the

218

northeast. In 2011, after the rains of Typhoon Nalgae and its outer circulation struck the study area,

219

the lobe-shaped landslide area at the bottom edge of Don-Ao Peak continued to develop headward

220

along the tension cracks (to the north of B and C in Fig. 4(e)). New tension cracks and lateral 9

221

erosion gullies also appeared along the expanding landslide area. Following the 512 rainfall event in

222

2012, the landslide area to the left of the upstream section of Dakeng Stream again presented

223

headward erosion (to the northeast of B in Fig. 4(f)). Downcutting of the erosion gullies where the

224

two lobe-shaped landslide areas met also caused a portion of the left bank to collapse. Table 2 lists

225

the areas of landslides and alluvial fans at 115.9 km of Provincial Highway No. 9, resulting from

226

the various Typhoon and rainfall events. Correlations are outlined in the subsequent analysis.

227

Table 2 Variations in landslide area on the upslope aton 115.9 km of Provincial Highway No. 9.

228

Fig. 4 Comparison of remote sensing images at 115.9 km on Provincial Highway No. 9 between

229

2004- 2013 (FS-2, SPOT 5).

10

230

3. Methodology

231

The aim of this study is to investigate and explore landslide characteristics in areas with high

232

landslide susceptibility along the coastal mountain highway based on ain terms of combination

233

ofing the historical landslide records, field surveys, remote sensing interpretation, monitoring data

234

(rainfall and GPS ground displacement), and LiDAR measurements. tTo shed light on the potential

235

for disaster and the evolution of post-disaster landscapes along the slopes at Don-Ao Peak (115.8 -

236

116.4 km)., WeThe study performed analysis of the rainfall associated with recent disasters in

237

conjunction with high-precision digital mapping of the terrain (resolution: 1 m) using full waveform

238

airborne light detection and ranging (LiDAR, Riegl LMS-Q680i, 2011/7) as well as ground-based

239

LiDAR (Riegl VZ1000, VZ620, 2013/8-10). For the purpose of understanding the long-term

240

landform evolution, we interpreted the micro-topography in the landslide areas of landslides

241

fromusing multiple remote sensing images from 2003-2012 and to analyzed the evolution of the

242

landscape and the mechanisms by which the debris slide transformed into a debris flow. Post-

243

disaster field surveys and displacement monitoring data werewere also examinedadopted to

244

characterize the features of deep-seated landslides in the catchment area near the Dakeng Bridge in

245

order to provide a reference for future monitoring processes and the implementation of measures to

246

avoid disasters. Information from aUsing a dual-band high-precision GPS receiver (Trimble 5700)

247

at the landslide monitoring and warning system in the Don-Ao Peak landslide area (at 115.9 km on

248

Provincial Highway No. 9), wase analyzed to understand the correlation between rainfall data and

249

the degree of displacement and settlement oin the upper slopes and to investigate topographical

250

changes in the landslide area. The GPS receiver (the base station of which is located in downtown

251

Suao, and the distance is 6.6 km toin the north offrom the study area) includes the Trimble 5700

252

base station and Trimble 4D control displacement monitoring system (4D indicates the E-W, N-S,

253

vertical, and t dimensions). Automated real-time monitoring (with a renewal frequency of 1 Hz) of

254

the landslide area alongin the upper slopes with a renewal frequency of 1 Hz in the monitoring 11

255

system enables the calculation of the bearings and distance at various points. In addition, an

256

openness visualization technique which called the ‘sky view factor’ was applied to enhance the

257

terrain features of the potential landslide (Zakšek et al., 2011; Doneus, 2013). This approach allows

258

efficient highlighting of thes relief characteristics on a two-dimensional map efficiently, so the sub-

259

scarp and tension cracks can emerge from the ridge or hillslope.

260

4. Results and discussions

261

4.1 Post-disaster physiographic characteristics of landslide areas

262

SThis study conducted several field surveys were conducted after Typhoons Megi (2010),

263

Typhoon Nalgae (2011), Typhoon and Saola (2012), and the 11/11 rainfall event. In addition to

264

observing the conditions inof the area near the 115.9 km mark following the sediment-related

265

disasters, we investigated the deposition characteristics and the erosion induced by rainfall in the

266

channel of the Dakeng Stream. The study areaAs shown in Fig. 5, the study area can be divided into

267

three sections following Typhoon Megi can be divided into three sections, according to the location

268

of the disaster: 1) the source area upstream of the landslide (the two lobe-shaped source areas below

269

Don-Ao Peak);, 2) the transportation zone (the channel of the Dakeng Stream);, and 3) the

270

deposition zone (the river mouth-coastal line alluvial fan). Previous post-disaster field surveys (Fig.

271

6) revealed that the depth of the landslides in the two lobe-shaped source areas reached nearly 30 m

272

(measured forom the digital elevation model (DEM) by airborne LiDAR), which means they can be

273

categorizeds them as large-scale, deep-seated landslides. The main scarp continued to expand

274

headward, serving as the main source of sediment in the area. However, the sliding of the mass of

275

the landslide sliding into the channel of the Dakeng Stream, due to gravity or rainfall, resulted in

276

blockages as a consequence of due to a narrowing of the channel section. Two blockages occurred

277

in the transportation zone: one at the old check dam near the junction of different lithological

278

distributions in Fig. 6 (a) and the other at the Dakeng Bridge on Provincial Highway No. 9, where

279

the existing culverts under the bridge were insufficient. When the debris flow was obstructed by the 12

280

flood releasing cross-section under Dakeng Bridge, a portion of the boulders andor sediment spilled

281

over the surface of Highway No. 9, thereby obstructing traffic (Fig. 6(b)). Field surveys revealed

282

that the key strata distributed within the source area originated from amphibolite outcrops of the

283

Don-Ao schist stratum (Fig. 7). Despite the high compression strength of the rock, the extreme

284

weathering produced multiple sets of fractures and weak planes within itthe rock (Fig. 7(a)). Thus,

285

the rock exposed below the scarp was comprised of extremely fragmented rock materials (Figs. 7(c)

286

and 7(d)). Furthermore, following Typhoon Megi, a groundwater exfiltration point appeared near

287

the lithological junction (amphibolite/schist) in the midstream section of the Dakeng Stream

288

(marked as E1 in Figs. 7(b)), which became a source of runoff inintofor the Dakeng Stream. It is

289

worth mentioning that this formsis the upper edge of the earliest landslide after Typhoon Parma, and

290

still presents the formation of a path of seepage path for groundwater. The continuous runoff

291

supplied by the seepage of groundwater was sufficient to initiate the fluidization of the colluvial soil

292

in the Dakeng Stream during rain events. With regards to the slope characteristics, the study area is

293

located within a concave section of the road. The geomorphology of the catchment area of the

294

Dakeng Stream is characterized by concave-shaped slopes, which facilitate the convergence of

295

debris accumulation and water flow. Physiographically speaking, the area is ideally suited to such

296

disasters.

297

Fig. 5 Three-dimensional terrain map of landslide hazards at aton 115.9K of Provincial Highway

298

No. 9.

299

Fig. 6 Dakeng Bbridge at 115.9 km of Provincial Highway No. 9 was blocked with debris flows

300

after (a) Typhoon Megi; and (b) Typhoon Nalage.

301

Fig. 7 Onsite survey in the upstream landslide area of the Dakeng gully catchment.

302

4.2 Geological structures and failure mechanisms

303

With regards to the local geological structures, Yeh (1998) examined the distribution of

304

foliation in the Suao-Nanao area and reported that most of the secondary foliation in the strata on 13

305

the north side dips towards the south, whereas in the strata on the south side slope it slopes toward

306

the north (presenting a fan structure) with amphibolite as its core. The geology of the study area

307

exhibits the features of ductile deformation structures. In terms of theAs for geological zoneing, the

308

study area belongs to the east wing of the Central Mountain Range, which has been compressed by

309

tectonic stress (from east to west) from plate collision. Thus, most failure patterns in the area are

310

comprised of slate or schist that has broken along foliation planes. As shown in Fig. 8, most of the

311

failure patterns on the upper slopes of the catchment area of Dakeng Stream are gravitationally

312

deformed schist. Figures 8(a) and 8(b) illustrate Tthe gradual process from stable to unstable in the

313

slope body of the left bank of the Dakeng Stream, with the strata of Don-Ao schist undergoing

314

sediment erosion and weathering are illustrated in Fig. 8(a) and 8(b). In general, in weathered and

315

foliated rock, presents flexural toppling leads to failure along the foliation plane. An intersecting

316

cross-joint set in this type ofe aforementioned rock will generally be the location ofpresent a blocky,

317

toppling failure or block-flexural toppling failure when the rock is unable to withstand the tensile

318

bending stress (Amini et al., 2012). On the basis ofThe stratigraphic column obtained from g

319

geological drilling (depth (h): 15 m, located on the ridge of the Don-Ao Peak), the stratigraphic

320

column is consists of surface soil (h=0-1.6 m), amphibolite with highly weathering (h=1.6-13.2 m),

321

and slightly fractured amphibolite bedrock (h=13.2-15.0 m). The results of mechanics testing points

322

out that the rock mass is characterized by a fragmental condition and poor engineering applications.

323

RThe rainfall can easily infiltrate easily theinto rock mass throughalong the surface cracks. Hence,

324

the weight of the rock mass increasesrises gradually with thehile increase in the degree ofing the

325

saturation degree of the rock. The unstable rock mass onin the source zonation or deposited

326

colluvium on the hillslope begins to collapse (or sliding) once they exceed the critical conditions

327

forof gravitational deformation are exceeded. Additionally, the toe erosion effect of the debris slide

328

enhances the volume of mass movement. It also greatly contributes to thea great amount of high

329

mobility of the sliding mass transforming into the valley debris flow that moves down the valley.

330

Primary failures oin the lower slopes of the Dakeng Stream between 115.7 km and 115.9 km 14

331

included debris sliding in the colluvium and the lateral and streambed erosion of the debris flow in

332

the channel due to entrainment. Field survey results (Section A-A’ in Fig. 9) revealed that in

333

addition to the exposure of fault shear zones on both sides of the stream (the features of which are

334

described in the next section), the rocky slope of the both banks also exhibited several tension

335

cracks (exfoliation), which was parallel to the slope surface and the stream (Fig. 9(a)). These

336

tension cracks were also observed to developing inalong the upper top of the rocky layer (Fig. 4(d)-

337

(f)), which wouldare likely not be conducive to maintaining the stability of the rock slope and was

338

thus the main causes of collapses on the both banks. Furthermore, buckling due to gravitational

339

deformation can be seen in the black schist cleavage wherewith the Dakeng Stream and wedge

340

failure in the lower slope at 116.1 K adjoin (Fig. 9(d)), as well as in the rock slope of the bank near

341

the channel.

342

In summaryTo summarize, the geological structures and mechanisms underlying the occurrence

343

of disasters in the study area were the result of previously existing headward development of the

344

erosion gullies along the upper slopes atof the Don-Ao Peak. This induced a large-scale landslide

345

with the collapse of massive quantities of sediment into the channel of the Dakeng Stream.

346

Subsequent rainfall producedformed runoff that changed the sediment into debris flows that

347

gradually progressed downstream to the coast. Onsite measurements of the attitude of exposed

348

strata in the upstream landslide area and the attitude of other strata along the highway (Fig. 10(a)),

349

in conjunction with the distributions of joint density (Fig. 10(b), where the shaft diameter is the

350

number of attitude distributions) indicate that the attitude of most of the cracks exposed in the

351

landslide area is close to that of the local geological structures (N70

352

demonstrates that the large-scale landslides caused by tTyphoons and heavy rains in this area are

353

primarily controlled primarily by the geological conditions. Runoff produced by heavy rainfall and

354

the seepage of groundwater are considered triggering factors. The extent of exfoliation and the fact

355

that the amphibolite in the source area is harder and more fragmented than the graphite schist made

356

it easier for surface water to infiltrate the slope from the crest of the slope or fissures in the slope 15

W/38

S). These resultsis

357

face. The groundwater then travelsed along the permeable strataum to beand was exuded from the

358

impermeable stratum near the slope toe. The slope toe was easily damaged uUnder the double

359

influence of side erosion from the groundwater and debris flow in the valley, the slope toe was

360

easily damaged. A loss of support from beneath caused the area over the slope toe to deform, which

361

caused slippage and creeping between the strata of the slope surface and intensified fragmentation.

362

Increasingly severe fragmentation in the body of the slope enabled substantial water infiltration into

363

the permeable stratum. The groundwater percolated through the fragmented amphibolite rock strata

364

and along the less permeable schist stratum beneath it to accumulate at the impermeable stratum

365

near the toe of the slope, undermining the stability of the entire slope. The torrential rains brought

366

by Typhoon Megi caused the groundwater levels within the slopes in the source area to rise rapidly,

367

thereby expediting erosion in the slope toe and destabilizing the entire slope. Therefore, during this

368

landslide event, the infiltration of substantial quantities of rain and increased side erosion in the

369

valley caused the section between the slope toe and the slope face to collapse first. In that instant,

370

the near-saturated sliding mass behind became unstable and slid downstream. The mechanisms

371

underlying this collapse should be taken into consideration in subsequent slope monitoring or

372

remediation projects along the Su-Hua Highway (Lo et al., 2014). HBesides, historical earthquake

373

records also show that prior to Typhoon Megi, the No. 109 earthquake on October 3, 2010, which

374

measured 5 on the Shindo? seismic intensity scale at the Nanao meteorological station, generated

375

north-south ground acceleration (195.6 gal) close to the threshold required to cause a landslide (250

376

gal; Lee, 2012). This is another crucial factor in the large-scale landslide at Don-Ao Peak.

377

Fig. 8 Schematic sketch showing the gravitational deformation of foliated rock.

378

Fig. 9 Schematic diagram of geological structure in the downstream landslide area of the Dakeng

379

gully catchment.

380

Fig. 10 (a) Mapping of rock attitude; and (b) joint density in the study area.

381

4.3 Displacement monitoring and analysis of landslide area 16

382

Precipitation is often the primary cause of landslides., in which Rrainfall saturates the surface

383

layer of the slopes, whereupon the wetting front and gradual rise in groundwater eventually overlap

384

and trigger slope failure mechanisms. Theis study proposed the relationship between the hourly

385

rainfall and vertical ground displacement ion the source area around the crest line proposed in this

386

paper ias discussed below.

387

(1) Typhoon Nalgae in 2011

388

Although Typhoon Nalgae did not make direct landfall on Taiwan, its outer circulation reacted

389

with the northeast monsoon from the cold high pressure of Mainland China resulting in torrential

390

rains in Yilan and Hualien. Post-disaster field surveys indicated that this Typhoon event caused an

391

expansion ofded the central portions of the west and east landslide areas inat the upstream reaches

392

onf Don-Ao Peak. The rainfall-induced sliding of debris tended to follow the development of

393

tension cracks formed after Typhoon Megi, resulting in headward erosion ofin the erosion gullies

394

inof the landslide area. The largest displacement of 140 cm was obtained at GPS05 (Fig. 1(c)). Fig.

395

11 presents the vertical displacement and settlement with corresponding rainfall. This event

396

exhibited a rainfall histogram with a peak at the center. Debris slides were triggered when hourly

397

precipitation was highest (11: 00, 03 Oct. 2011). Time lag characteristics associated with rainfall

398

and subsequent infiltration (the response to infiltration is slower, and, as the weight of the sliding

399

mass increases, it may continue to slide even after the rain has stopped), resulted in athe slide

400

lasting approximately 42 hrs (forom 05:00 03 Oct. to 23:00 04 Oct). The iInitial sliding speeds

401

obtained from monitoring of the displacement oin the vertical axis of the GPS receiver wasere

402

higher (4.4 cm/hr), gradually decreasing to 1.3 cm/hr. This slope failure can thus be categorized as

403

slow landslide deformation.

404

Fig. 11 The Rrelationship between displacement (vertical) of GPS05 and hourly precipitation at

405

Don-Ao Peak.

406

(2) 0512 rainfall event in 2012 17

407

At 15:00 on May 12, 2012, torrential rains struck the Suao and Nanao areas in Northeastern

408

Taiwan (R>500 mm). At the Don-Ao Peak station, the event resulted in the total accumulation of

409

452 mm of rainfall. Hourly precipitation peaked at 62 mm/hr, causing the bare ground at the top of

410

the upper slopes (right side of the Dakeng Stream) to slide downstream along the direction of the

411

tension cracks that formed southwest of GPS05 after Typhoon Nalgae (Fig. 1(c)). The greatest

412

displacement was measured at GPS04 on the right wing of the landslide area, where the amount of

413

settlement in the north-south direction was approximately 64 cm. Fig. 12 shows a comparieson of

414

the maximum displacement and settlement values with hourly precipitation. The displacement curve

415

shows that the landslide event associated with the in 0512 rainfall was similar to that caused by

416

Typhoon Nalgae in 2011. In addition to a direct association with the intensive rainfall and the

417

histogram characterized by a peak at the center, when the great amount of runoff or debris flow

418

passed by downstream, the lower slope was subjected to severe side erosion and hollowing out,

419

which triggered multiple collapses in the lower slope of downstream.

420

Fig. 12 Relationship between displacement (north-south) of GPS04 and hourly precipitation at Don-

421

Ao Peak.

422

4.4 Nnumerical simulation of landslide process

423

A Ssimulation of the landslide movement process Simulation was used to investigate the

424

process underlying the movements involved in the large-scale debris slide triggered by Typhoon

425

Megi in 2010 at 115.9 km onf Provincial Highway No. 9. (Lo et al., 2014) To explain the processes

426

involved, we monitored the movement speeds throughout the entire process. This revealed

427

variations in speed during different stages and clarified the key processes involved in the overall

428

movement. The results of the numerical simulation are outlined belowas follows. (a) Fig. 13

429

presents the simulation results of the movement process and variations in speed associated with the

430

landslide at 115.9 km. The numerical model was designed based on the results obtained byf

431

subtracting the pre-event DEM from the post-event DEM (pre-event: 2004; post-event: 2010). Five

432

primary landslide masses were designated fFrom upstream to downstream, we designated five 18

433

primary landslide masses. Two of these areas were situated in the upstream source area (the sliding

434

mass A and B), one in the midstream section (the sliding mass C), and two in the curve of the valley

435

at the location of the collapse (the sliding masses D and E). SThe sliding masses A and B in the

436

source area slumped directly downstream as a result of gravity, whereas The sliding masses C, D,

437

and E maintained higher self- supportability due to their larger coefficient of friction (equal to 0.6)

438

and greater bonding strength. When the sliding masses A and B came into contact with the sliding

439

masses C, D, and E downstream, the lack of support from the slope toe, which had been hollowed

440

out by side erosion, caused them to collapse. This conformscoincides with the landslide

441

mechanisms derived from the onsite survey. Furthermore, in this study used a PFC3D basic fluid

442

option was used to install a water body downstream where the sea would be. When the particle

443

elements encountered the body of water, their movement speed was reduced through fluid

444

resistance, causing them to accumulate in an alluvial fan. This also enhanced the applicability of the

445

landslide simulation. (b) These simulation results (Fig. 13) indicate that the entire landslide process

446

progressed through seven key phases from sliding, accelerating, and decelerating to a final stop. In

447

the first phase, the sliding masses A and B in the source area began moving (at the 5 sec point in

448

Fig. 13). The speed of the landslide in this phase ranged between 5 m/sec and 20 m/sec

449

(approximately 18-~72 km/hr) with an average speed of 8.96 m/sec (approximately 32.3 km/hr, as

450

shown in Fig. 15). The moving masses underwent collision interaction, in which the mass in the

451

rear transferred energy to the mass in front. This increase in kinematic energy caused the downward

452

motion of the mass in front to accelerate in its downward motion. Therefore, the masses moving the

453

fastest were for the most part in the middle and front portions of Masses A and B, while those in the

454

rear portions moved more slowly. (c) Second phase (15 sec point in Fig. 13): The sliding masses A

455

and B migrated past the steepest terrain in the region, converged at the Southern Unnamed Creek,

456

and reached the slope toe of the sliding mass C. At this point, the front portions of the sliding

457

masses A and B were moving at the highest speed of 52.2 m/sec with an overall average speed of

458

20.4 m/sec, as shown in Fig. 15. The rear portions were also beginning their accelerated descent. (d) 19

459

Third phase (22.5 sec point in Fig. 13): tThe front portions of the sliding masses A and B arrived at

460

the highway where the Chuang-Yi tour group bus was parked on the roadside. Because the terrain in

461

the valley turns gradually from southeast to south, the front portion of the sliding mass began

462

slowing down (the average speed was reduced to 10.36 m/sec). Here, the landslide gradually

463

transformed from a large-scale debris slide to a debris flow, and due to side erosion in the slope toe.

464

SThe sliding mass C became unstable and began collapsing in the downstream direction. (e) Fourth

465

phase (40 sec point in Fig. 13): bBy this time, the sliding mass C had completely collapsed, and the

466

front portions of the sliding masses A and B had reached the sliding mass D. In terms of terrain, the

467

sliding mass D was situated on an undercut slope at the bend of the watercourse. This caused the

468

front portions of the sliding masses A and B to accelerate again (increasing to 27.5 m/sec) and

469

quickening the erosion at the slope toe of Mass D. (f) Fifth phase (60 sec point in Fig. 13): dDue to

470

side erosion, the sliding mass D collapsed, and the front portions of the sliding masses A and B had

471

passed another bend in the creek past the toe of the sliding mass E. Severe side erosion caused by

472

the sliding masses from upstream severely destabilized mMass E, which was located on the lower

473

slope of the Su-Hua Highway at 115.8 km. (g) Sixth phase (70 sec point in Fig. 13): bBy this time,

474

the sliding mass E had completely collapsed, and a portion of the sliding masses were beginning to

475

pass the second bend before they reacheding the sea. Due to the curve of the valley, the masses

476

accelerated once more, reaching average speeds of 12.40-12.57 m/sec during the period between 70

477

sec and 77.5 sec. At approximately 77.5 sec, they gradually decreased in speeds began gradually

478

decelerating. At 70 sec, the front portions of the sliding masses A and B had reached the sea and

479

were also sloweding down byin this body of water. In the seventh phase (157.5220 sec point in Fig.

480

13), an alluvial fan formed, ending the movement of the landslide event at approximately 157.5 sec.

481

Fig. 13 Movement process and speed variations in the simulated landslide at 115.9220 km of

482

Provincial Highway No. 9.

20

483 484

4.5 Interpretations of topographical changes and landslide micro-photographicy images

485

This study compared digital elevation models from 2004, 2011 (after Typhoon Megi), and 2013

486

(after Typhoon Saola) (Fig. 14). The locations of the cross-sections are presented in Fig. 14

487

(stretching from the landslide area on the right wing of the Don-Ao Peak along the Dakeng Stream

488

to the alluvial fan at the mouth of the river). The three cross-sections represent topographical

489

changes in the area, from before the disasters to after the two most recent tTyphoon events. The

490

figure shows that before and after Typhoon Megi (the black dashed lines and the blue bold lines in

491

Fig. 14, respectively), the primary landslide and erosion areas were at the bottom edge of the Don-

492

Ao Peak (at most, the depth of the landslide was approximately 30 m) and at the midstream section

493

of the Dakeng Stream where the two lobe-shaped landslide areas met (maximum erosion depth was

494

approximately 15 m). Outcrops could be seen throughout the streambed after the disaster, and the

495

two old check dams midstream at Dakeng Bridge had been completely silted up and destroyed,

496

forming a pileups under Dakeng Bridge. Deep streambed erosion (approximately 25 m) also

497

appeared at the bend in the Dakeng Stream (marked “curved channel” in Fig. 14) after Typhoon

498

Megi. This canould be attributed to scouring of the cutbank oin the bend and the side bank of the

499

debris flow. Furthermore, an alluvial fan with gentle changes in slope can be clearly seen at the

500

mouth of the river, descending into the ocean in the form of a Gilbert-type fan delta. According to

501

the underwater side-scan sonar results obtained by Chou et al. (2012), the alluvial fan extends

502

approximately 150 m along the seabed. A comparison of changes in the terrain profiles before and

503

after Typhoons Megi and Saola show that only the bare landslide area on the right wing upstream of

504

Don-Ao Peak presented sliding at the ground surface, which was comprised of collapsed scarps

505

resulting from old tension cracks. The newly constructed dam in the midstream section of the

506

Dakeng Stream also presented pileups from being silted up completely. Disturbance from

507

construction and the flow of the debris flow resulted in downcutting erosion in the channel 21

508

downstream, which was most pronounced under Dakeng Bridge and at the bend. The radial length

509

of the alluvial fan remained roughly the same with a slight retreat of approximately 7 m, and the

510

deposition zone comprised ofing finer particles expanded on both sides (in the east-west direction).

511

We speculate that this is associated with the partial erosion of erosion gullies near 115.9 km (at

512

116.8 km and 117.0 km). However, the main channel of the stream above the fan was altered by the

513

previous flow of debris and sediment congestion. Overall, the topographical changes matched the

514

field survey results.

515

Interpretation of high-precision micro-photographs of the landslide obtained frrom airborne

516

LiDAR (2011/7, 2013/8) and ground-based LiDAR (2013/9) enabled us to approximate the ground

517

surface conditions without vegetation and identify the potential areas of disaster in the study area.

518

We adopted a one-meter digital elevation model (DEM), digital surface models (DSM), and

519

orthophotos forof the same time period. These were used in conjunction with the spatial analysis

520

module from theof ArcGIS software to produce slope maps, aspect maps, contour maps, and eight-

521

directional sun shadow maps. This interpretation was also aided by the use of multiple satellite

522

images (FS-2, SPOT5) as well as geological maps and potential disaster maps published by the

523

Central Geology Survey. Our aim was to identify current landslide areas and disaster-prone areas.

524

For our methods of interpretation, we referred to The Investigation and Analysis Project of

525

Geologically Sensitive Areas for Homeland Conservation (Central Geology Survey, 2012) with

526

regard to the topographic characteristics of deep-seated landslides and large-scale landslides (Fig.

527

15), such as the scarps, crown tension fractures, side fissures, break lines, erosion gullies, bulging

528

slope toes, or sliding masses (Takasuke, 2000; Agliardi et al., 2001; Chigira, 2009). The msore

529

significant landslide terrain characteristics in the analysis' in this study included scarps, surface

530

ruptures surface, sliding ranges, and erosion gullies. The gully lines were categorized as either Type

531

A (headward erosion) or Type B (no headward erosion). SThis study adopts the shading relief with

532

a sky view factor and references features associated with gravitation deformation areas as described

533

in Soeters and& van Westen (1996) and Chigira (2014) (Table 4) were adopted to identify the 22

534

boundary of the gravitation deformation area in the Dakeng Bbridge watershed. The identified area

535

wais then entered into the landslide area-volume empirical equation to compute the probable

536

landslide volume. That volume wais then input into the numerical model to estimate the zone

537

affected by the landslide. The Iinterpretation of the gravitational deformation areas is shown in Fig.

538

16. The results of landslide micro-photography analysis in two catchment areas are shown belowas

539

follows.

540

Fig. 14 Surface profile of the main landslide and the corresponding geological setting in 2004,

541

2011, and 2013.

542

Fig. 15 Active landslide geomorphic evolution and deep-seated landslide characteristics (revised

543

from Takasuke, 2000).

544

Fig. 16 Results of micro-photographic interpretation of the landslide at the 115.9 km watershed on

545

Provincial Highway No. 9.

546

Table 4 Irregularities associated with gravitation deformation areas (fFrom Soeters and van Westen,

547

1996; Chigira, 2014).

548

Catchment area of the Dakeng Stream

549

The crown of the bare landslide on the southwest side of Don-Ao Peak at the source of the

550

Dakeng Stream still exhibits continuous tension cracks (Fig. 16), the progress of which haves

551

already traversed the ridge line. The depressed region at the top of the slope shows distinct cracking

552

in the sliding surfaces, which may progress and cause landslides in the future. Due to erosion from

553

rainfall, the bare ground also shows the continued headward development of erosion gullies.

554

Rainfall events of greater rainfall intensity (2013/11/11) have produced small-scale landslides on

555

the slope face. Some of the loose mass that has sliding from the scarp is still deposited on the slope

556

face (break line). Reinforcement projects at the top of the slope have decelerated the expansion of

557

tension cracks and lateral cracks at the upper edge of the landslide area on the left wing of Don-Ao

558

Peak. Most of the vegetation at the top of the slopes in the study area is comprised of herbaceous

559

plants with shallow roots, which makes the headward development of the erosion gullies more 23

560

severe. Deeper erosion gullies are clearly shown on bBare slopes clearly show deeper erosion

561

gullies. In addition to presenting more significant expansion in area on the left wing than onthat in

562

the right wing inover the last few years, the landslide area on the left wing is also gradually

563

progressing towards the catchment area of the wild stream. At the place where the two landslide

564

areas on the two wings meet (the lithological junction shown in Fig. 16), another two more

565

landslide areas (the restored ones formed downstream after Typhoon Parma) and erosion gulliesy

566

appears to be developing headward. The erosion gully cuts along the side boundary of a hummocky

567

sliding mass, which may induce a landslide in the future. The lower slope channel of the Dakeng

568

Stream between 116.1 km and 116.4 km encompasses an area whereith previous disasters have

569

occurred. In this area, headward progress is apparent whichand itsis likely to cause toe erosion and

570

sliding, due to the fact that it lies on the outside of the bend whereof the Dakeng Stream at 116.0 km

571

and where the downstream portions of the erosion gullies meet at 116.4 km. Furthermore, high

572

resolution terrain shows that the alluvial fan at the mouth of the river comprises coarser particles

573

provided by the landslide area upstream of the Dakeng Stream. In contrast with the terrain near the

574

two sides of the fan (in the east-west direction), the size of sediment particles deposited at the

575

mouth of the river are smaller due to the shorter erosion gullies and smaller source area (comprising

576

mostly shallow-seated landslides). This indicates that the composition of the materials in the

577

alluvial fan is determined by the scope of the upstream landslide and the depth of the slides.

578 579

Catchment area of the wild stream to the north

580

The catchment area of the wild stream containspresents concave-shaped forested terrain. The

581

vegetation at the top of the slopes is similar to that in the catchment area of the Dakeng Stream,

582

comprising herbaceous plants. Despite the lack of landslides and erosion gullies with headward

583

erosion in this area, several erosion gullies with vertical contours beganhave begun appearing along

584

the ridgeline after Typhoon Megi. The micro-photographic interpretation maps without vegetation

585

(Fig. 16) show distinctly show the expansion of several scarps at the ridgeline (curved scarp lines

586

with lengths reaching tens of meters). These scarps developmenedt of these scarps is similar to 24

587

those in the Dakeng Stream catchment, forming a landslide boundary in the shape of an inverted

588

cone. On the right side within the boundary are mound-shaped sliding masses clearly being cut by

589

erosion gullies. Comparisons of the 1936 and 2004 DEMs show the addition of a new tributary in

590

the midstream section of the wild stream as well as marked changes in the upstream channel near

591

the ridgeline. Prior to Typhoon Megi, Chu et al. (2013) reported the existence of numerous

592

hummocky surfaces around the Dakeng Stream and the upstream portion of the wild stream to the

593

north, which were shown to be the result of earlier debris slides or rock slope deformation. As can

594

be seenshown in Figs. 16(a) and 16(b), field surveys verified the existence of crest boulders

595

(approximately 4 m in size, mostly amphibolite) in the midstream and downstream sections of the

596

wild stream, possibly the remnants of a debris flow. In contrast with the area of Dakeng Stream

597

landslides, the catchment area of the wild stream is on the leeward side of the terrain (in terms of

598

the northeast monsoon), and therefore does not receive as much precipitation as the Dakeng Stream

599

catchment. Nevertheless, the progress of the upstream scarps should be monitored closely.

25

600

5. Conclusions

601

This study investigated the evolution of landscape featuress after major rainfall events in the

602

catchment areas surrounding the section of Provincial Highway No. 9 between 115.8 km and 116.4

603

km. Through hydrological analysis, we examined the role of precipitation in the occurrence of

604

landslides and estimated the deposition area (alluvial fan distribution) of resultant debris flows. We

605

also conducted interpretedations of topographical changes and analyzed landslide micro-

606

photographicy images analysis using onsite field surveys, the interpretation of remote sensing data,

607

and high-precision digital terrain.

608

Our interpretation of multiple remote sensing images shows distinctly that landslides in the

609

study area were initially caused by the headward development of erosion gullies. Since Typhoon

610

Megi, the expansion of the landslide area on the upper slopes of Don-Ao Peak tended to follow the

611

arced tension cracks that appeared at the top of slopes. Furthermore, the retreating landslide disaster

612

that affectedof the roadbed oin the lower slopes was associated with the generation of exfoliations

613

and the toe erosion after ofa the largegreat amount of runoff or debris flows. Analysis of the

614

intensity-duration-frequency of the rainfall revealed that most recent landslide-inducing rainfall

615

events near the Dakeng Stream were associated with rainfall amounts greater than that of a 200-year

616

event. This study presents the first application of high-precision digital terrain mapping derived

617

from airborne and ground-based LiDAR for the identification of potential areas of disaster in the

618

vicinity of Don-Ao Peak. Landslide micro-photography analysis indicates that topographical

619

changes in the landslide areas in the Dakeng Stream catchment are determined by local geological

620

structures (fault zones and lithological junctions), headward-developing erosion gullies, and the

621

concave, water-collecting shape of the slopes. Deposits of boulders from earlier debris flows remain

622

in the vegetated catchment of the wild stream to the north, and the ridgelines of the upper slopes

623

also reveal the formation of several curved scarps. Variations in strain rate in the body of the slope

624

should be monitored to further elucidate the occurrence of sediment-related disasters following 26

625

tTyphoons or rainfall events.

626

Acknowledgements

627

We thank the Fourth Maintenance Office, Directorate General of Highways, MOTC in Taiwan for

628

providing the valuable monitoring data for intense rainfall events. The authors would also like to

629

thank Prof. Ming-Lang Lin at the Department of Civil Engineering, National Taiwan University and

630

colleagues work in Sinotech ConsulatantsConsultants, INC. for their advices and assistances withon

631

thise study.

27

632

References

633

Agliardi F (2001) Structural constraints on deep-seated slope deformation kinematics. Engineering

634 635 636 637 638 639 640

Geology 59:83-102 Amini M, Majdi A, Veshadi MA (2012) Stability analysis of rock slopes against block-flexure toppling failure. Rock Mech Rock Eng 45:519-532 Aydan Ö, Shimizu Y, Ichikawa Y (1989) The effective failure modes and stability of slopes in rock mass with two discontinuity sets. Rock Mech Rock Eng 22:163-188 Aydan Ö, Kawamoto T. (1992) The stability of slopes and underground openings against flexural toppling and their stabilization. Rock Mech Rock Eng 25(3):143-165

641

Aydan Ö, Amini M (2009) An experimental study on rock slopes against flexural toppling failure

642

under dynamic loading and some theoretical considerations for its stability assessment. Journal

643

of the School of Marine Science and Technology, Tokai University, 7(2):25-40

644 645

Aydan Ö, Ohta Y, Hamada M (2009) Geotechnical evaluation of slope and ground failures during the 8 October 2005 Muzaffarabad earthquake, Pakistan. J Seismol 13:399-413

646

Aydan Ö, Ohta Y, Daido M, Kumsar H, Genis M, Tokashiki N, Ito T, Amini M (2011) Earthquakes

647

as a rock dynamic problem and their effects on rock engineering structures. Chapter 15:

648

Earthquakes as a rock dynamic problem and their effects on rock engineering structures.

649

Advances in Rock Dynamics and Applications, CRC Press, Taylor and Francis Group, 341-

650

422.

651 652 653 654

Central Geological Survey (2011) Disaster survey report for Typhoon Megi in Suao area and Suhua highway. Central Geological Survey (in Chinese) Central Geological Survey (2012) Investigation and analysis for geologically sensitive area in national preservation domain program. Central Geological Survey (in Chinese)

655

Chigira M (2009) September 2005 rain-induced catastrophic rockslides on slopes affected by deep-

656

seated gravitational deformations, Kyushu, southern Japan. Engineering Geology 108:1-15

657

Chigira M (2014) Geological and geomorphological features of deep-Seated catastrophic landslides 28

658

in tectonically active regions of asia and implications for hazard mapping. Episodes Journal of

659

International Geoscience 37(4):284-294

660

Chou HT, Lee CF, Lo CM, Lin CP. (2012) Landslide and alluvial fan caused by an extreme rainfall

661

in Suao, Taiwan. 11th International Symposium on Landslides (ISL) and the 2nd North

662

American Symposium on Landslides, Banff, Alberta, Canada pp 487-494

663 664 665 666 667 668 669 670 671 672

Chu HK, Lo CM, Chang YL (2013) Numerical analysis of slope stability at the 115.9k point of the Su-Hua Highway. Journal of Chinese Soil and Water Conservation 44(2):97-104 Doneus M (2013) Openness as Visualization Technique for Interpretative Mapping of Airborne Lidar Derived Digital Terrain Models. Remote Sens. 5:6427-6442 Hearn GJ (2011) Engineering Special Publication 24 - Slope Engineering for Mountain Roads. Geological Society of London pp 5-20 Lee CT (2012) Characteristics of earthquake-induced landslides and differences compared to storminduced landslides. EGU General Assembly 2012 pp 6937 Lee CT (2014) Statistical seismic landslide hazard analysis: An example from Taiwan. Engineering Geology, 182: 201-212

673

Lee CF, Huang WK, Chang YL, Chi SY, Lin WC, Chen HR, Lin CS, Chang SC, Liao WC, Chang

674

YH (2013) Establishing Landslide hazard index with landslide susceptibility and rock mass

675

rating system on Highway No.9 (Suao-Nanao), 2nd Conference for highway transportation

676

development, pp 108-121 (in Chinese)

677

Lee KT (2010) Hazard analysis at 104k~120k in Route.9 (Suhua highway, Suao-Don-Ao) during

678

Typhoon Megi. Fourth Maintenance Office Directorate General of Highways, MTOC (in

679

Chinese)

680 681 682 683

Lee MH (2006) The Rainfall threshold and analysis of Debris flows. Ph.D. Thesis, National Cheng Kung University, Taiwan, R.O.C. (in Chinese) Lin CW, Kao MC (2009) Suao-1/50,000 Taiwan geological map (2nd version), No.16, Central Geological Survey (in Chinese) 29

684

Lin ML, Chen RT., Yang DX, Zang PZ, Liu HC, Lin WK, Lo CM, Liao RT, Huang WC, Wang TT

685

(2013) Engineering geological assessment for failure mechanism at 115.7~116.1k on Suhua

686

highway. Geotechnical Engineer 6:6-13 (in Chinese)

687 688 689 690

Lo CM, Lee CF, Chou HT, Lin ML (2014) Landslide at Su-Hua Highway 115.9k triggered by Typhoon Megi in Taiwan. Landslides 11(2): 293-304 Lu RX, Chang YM, Lin XJ (2003) The rainfall intensity-duration equation analysis of rain gauge station in Taiwan. Water Resources Agency, MOEA (in Chinese)

691

Peyret M, Djamour Y, Rizza M, Ritz JF, Hurtrez HE, Goudarzi MA, Nankali H, Chéry J, Le Dortz

692

K, Uri F (2008) Monitoring of the large slow Kahrod landslide in Alborz mountain range (Iran)

693

by GPS and SAR interferometry. Engineering Geology 100: 131-141

694 695 696 697 698 699 700 701 702 703 704 705

Ulusay R, Aydan Ö, Kilic R (2007) Geotechnical assessment of the 2005 Kuzulu landslide (Turkey). Engineering Geology 89:112-128 Sassa K, Fukuoka H, Wang FW, Wang GH (2005) Dynamic properties of earthquake-induced large-scale rapid landslides within past landslide masses Landslides 2(2):125-134 Sekiguchi T, Sato HP (2004) Mapping of micro topography using airborne laser scanning. Landslides 1:195-202 Soeters R, van Westen, CJ (1996) Landslides: Investigation and Mitigation. Transportation Research Board Special Report 247:129-177 Takasuke S (2000), “Introduction of topographic map interpretation for construction engineer Vol.3: terrace, hill, and mountain. Kokon Publisher, pp 811-815 (in Japanese) Tarolli P (2014) High-resolution topography for understanding Earth surface processes: Opportunities and challenges. Geomorphology 216:295-312

706

Yeh EC (1998) The ductile shear deformation and structural evolution of the Penglai Orogency at

707

Suao-Don-Ao area in northeastern Taiwan. National Taiwan University, Masters thesis. (in

708

Chinese)

709

Zakšek K, Oštir K, Kokalj Ž (2011) Sky-View Factor as a Relief Visualization Technique. Remote 30

710

Sens. 3:398-415

31

Manuscript Click here to download Manuscript: 04_Manuscript.doc Click here to view linked References

1 1 2 32 4 5 63 7 8 94 10 11 5 12 13 6 14 15 7 16 17 18 19 8 20 21 22 9 23 24 2510 26 2711 28 29 12 30 31 3213 33 3414 35 36 3715 38 3916 40 41 4217 43 4418 45 46 4719 48 4920 50 5121 52 53 5422 55 56 57 58 59 60 61 62 63 64 65

Landscape evolution analysis of large scale landslides at Don-Ao Peak, Taiwan Ching-Fang Lee1, Chia-Ming Lo2*, Hsien-Ter Chou3, Shu-Yeong Chi1 1

Disaster Prevention Technology Research Center, Sinotech Engineering Consultants, Inc., Taipei 105, Taiwan, R.O.C.

2

Department of Civil Engineering, Chienkuo Technology University, Changhua 500, Taiwan, R.O.C.

3

Department of Civil Engineering, National Central University, Chung-Li 320, Taiwan, , R.O.C.

*Corresponding Author: E-mail: [email protected] or [email protected]

Abstract Typhoon Megi (2010) and the co-movement of the concurrent northeast monsoon brought massive rainfall to the Suao area of Yilan County, Taiwan, causing clusters of sediment-related landslide disasters on Provincial Highway No. 9. The most notable of these events was the largescale landslide on the upper slope at 115.9 km near Don-Ao Peak, which dumped 2.1 million cubic meters of sediment into the streambed. Rainfall runoff turned this into a debris flow forming an alluvial fan at the river mouth. This study analyzed the evolution of landscapes in the area through a field investigation, disaster-causing mechanisms, image interpretation, and airborne LiDAR. Our results indicate that the landslide was associated with its location at a lithological junction as well as local geological structures. Interpretation of micro-photography revealed that the topographical changes in landslide areas in the Dakeng Stream catchment are controlled by the headward development of erosion gullies and the concave shape of the slopes. Previous earthquakes and rainfall exceeding that of a 200-year event were the external precipitating factors. Key Words: Don-Ao Peak, large-scale landslide, debris flow, landslide micro-photography interpretation, shear zone.

1

23 1 224 3 425 5 6 26 7 8 927 10 1128 12 13 1429 15 1630 17 18 1931 20 2132 22 23 33 24 25 2634 27 2835 29 30 3136 32 3337 34 35 3638 37 3839 39 40 4140 42 4341 44 4542 46 47 4843 49 5044 51 52 5345 54 5546 56 57 5847 59 6048 61 62 63 64 65

1. Introduction Extreme changes in the climate have increased the number of typhoons bringing torrential rains, which have frequently caused major disasters in the form of landslides and mass movement in the mountainous areas of Taiwan. These events can lead to slope erosion, large-scale landslides, and debris flows within erosion gullies. In efforts to understand the failure mechanism of landslides, Aydan et al. (1989) proposed the limiting-equilibrium approach in an examination of stability in discontinuous rock masses. This current study theoretically discusses three common types of rocky slope failure, sliding, toppling, and combined sliding and toppling failure. The relationship between the angle of the lower slope, friction, and the inclination of the through-going discontinuity set are used in an evaluation of the instability of rock mass through laboratory experiments. Aydan and Kawamoto (1992) analyzed the flexural toppling failure behavior of surface slopes and underground openings. The research presented here based on the cantilever beam model suggests crack length dealing with suspended and overhanging behavior for actual rock slopes. Adayn and Amini (2009) studied flexural toppling failure under dynamic loading in a layered rock mass in experiments with horizontal shaking tables. They found that the flexural toppling failure type can be classified into active and passive modes. The seismic coefficient plays a dominant role on the initiation of the failure behavior (Aydan et al. (2011)). The landslide activity in sites where landslide have occurred has been discussed widely regarding mobilized unstable mass remaining in the source area or upstream reaches (Sassa et al. (2005), Ulusay et al. (2007), Peyret et al. (2008)). Landslides and debris flow triggered by heavy rains or long storms affect the stability of slopes around mountain roads and can jeopardize the safety of road users and the integrity of public facilities. Taiwan’s Provincial Highway No. 9 (also known as the Suhua Highway), currently the only road connecting the northern and eastern regions of Taiwan, is lined with steep terrain and runs through coastal areas. On-going efforts are being made to improve the highway, which remains a pivotal route for transportation in the area. The heavy rains brought by Typhoon Parma (2009), Typhoon Megi (2010), the outer circulation of 2

49

Typhoon Nalgae (2011) and the concurrent northeast monsoon, the 512 rainfall event (2012), and

150 2 3 451 5 652 7 8 53 9 10 1154 12 1355 14 15 1656 17 1857 19 20 2158 22 2359 24 25 60 26 27 2861 29 3062 31 32 3363 34 3564 36 37 3865 39 4066 41 42 4367 44 4568 46 47 69 48 49 5070 51 5271 53 54 5572 56 5773 58 59 6074 61 62 63 64 65

Typhoon Saola (2012) have caused clusters of sediment-related disasters on the slopes along Provincial Highway No. 9 between Suao (104.7 km) and Don-Ao (120.0 km). This frequently leads to highway closures, and has even caused a major slope collapse that pushed a bus full of passengers into the ocean causing numerous casualties (Central Geology Survey, 2011; Lee, 2010). Among the more common geological disasters associated with mountain roads are rockfalls, debris slides, deep-seated landslides, debris flow, slope cutting, toe erosion, and slope filling (Hearn, 2011). Lee (2010) investigated disasters caused by Typhoon Megi in the Suao area and along the Suhua Highway, observing that the typhoon triggered many debris slides, only one of which was a rockslide (116.1 km, Fig. 1). The area with the greatest number of landslides was found along the upper slope at 115.9 km near Dakeng Bridge (Don-Ao Peak). Chou et al. (2012) examined debris slide/debris flow triggered by Typhoon Megi between 115.9 km and 116.4 km on Provincial Highway No. 9 as well as the development of the coastal alluvial fan that it produced. They concluded that rainfall runoff had infiltrated the upper slopes, which caused the slope collapse and the gully-type debris flow resulting in a Gilbert-type fan delta in the ocean. Lo et al. (2014) also focused on the landslide event that occurred at 115.9 km on the Suhua Highway. They used a threedimensional discrete element method (Particle Flow Code in 3 Dimensions; PFC 3D) to perform numerical simulations of slope failure and discuss the kinematic processes and deposition characteristics of the debris slide. Lin et al. (2013) analyzed recent failure mechanisms and relief processes associated with roadbed collapses between 115.7 km and 116.1 km of the Suhua Highway providing an assessment of the shear zone outcrops in the catchment area of the Dakeng Stream and predictions of how they might develop in the future. Lee et al. (2013) employed the potential disaster maps created by the Central Geology Survey and the Rock Mass Rating (RMR) System to examine the area between Suao and Nanao on Provincial Highway No. 9 sectioning it into zones with various potentials for disaster. They applied slope units for zoning the highway and found that 44.0 % of the entire section is susceptible to landslides. 3

75

Statistics on the number of landslide occurring in the Suao-Nanao section of Provincial

176 2 3 77 4 5 678 7 879 9 10 1180 12 1381 14 15 1682 17 1883 19 20 84 21 22 2385 24 2586 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

Highway No. 9 between 2007 and 2012 indicate that rockfall and debris slides occur most frequently near Jiugongli (112.0 km) and Don-Ao Peak (115.9 km). Since Typhoon Megi in 2010, heavy rains from typhoons have continued to cause sediment-related disasters. These disasters are closely associated with the local geological structures. In this study, we first analyzed a series of satellite images and in-situ surveys to clarify the characteristics of landslide sites along the steep coastal mountain highway. Secondly, potential landslide activity in the Don-Ao Peak area was examined by specific shading relief and micro-topography interpretation (Sekiguchi and Sato, 2004; Paolo Tarolli, 2014). The major objectives of this work are: (1) to explore the evolution of the temporal landscape and large scale landslides and investigate the failure mechanism in the initiation zone; (2) to map the high susceptibility landslide zones for landslide enlargement by combining a high resolution terrain model (1 m) and the open shading relief technique.

4

87 1 2 388 4 5 689 7 890 9 10 1191 12 1392 14 15 1693 17 1894 19 20 95 21 22 2396 24 2597 26 27 2898 29 3099 31 32 33 100 34 35 101 36 37 102 38 39 40 103 41 42 104 43 44 45 105 46 47 106 48 49 50 107 51 108 52 53 54 55 109 56 57 110 58 59 60 111 61 62 63 64 65

2. Study area 2.1 Geography and geological setting The study area (the catchment area of the wild stream under the Dakeng Bridge) is located in eastern Suao, Yilan County (see the red square in Fig. 1), bordering the Central Mountain Range to the west and the Pacific Ocean to the east. Provincial Highway No. 9 winds along the steep coastal foothills at elevations between 220 m and 280 m. The slope of the terrain inclines to the east and the ground elevation becomes higher as the highway approaches the southern region. The study area is situated between 115.8 km and 116.4 km on Provincial Highway No. 9 (Fig. 1). Major collapses occurred upstream of Dakeng Bridge and Don-Ao Peak. Don-Ao Peak displays steep drops and reaches a height of 800 m just 1.2 km from the ocean. A 1:50,000 geological map of Suao and Nanao by the Central Geology Survey indicates that the landslide area is located in an area of amphibolite outcrops belonging to the Don-Ao schist stratum. The attitude of the schistosity trends to N45∘W and dips to 65∘SW. In this area, the strength of the amphibolite is high (water content: 0.4-0.7%; specific gravity: 2.9-3.0; porosity: 0.02-0.04; water absorption: 0.6-1.7%; peak strength of the shear test is approximately 0.25 kg/cm2 on the joint plane, φp: 25o; slake durability: 97.999.7%). However, severe weathering has resulted in extreme fragmentation. Disaster potential maps by the Central Geology Survey indicate a high potential for debris slides and rockfalls in the area. In the upstream area of Dakeng Stream, the slide-prone segment contains graphite mica schist belonging to the Don-Ao schist stratum. The terrain here presents concave-shaped slopes that often collect rainfall runoff.

Fig. 1 Geological settings for the study area (115.9 km on Provincial Highway No. 9).

2.2 Hydrometeorological conditions The study area faces windward during the East Asia tropical monsoon, and has an average annual rainfall of 4,480 mm (1971-2009). Statistics from the Suao Meteorological Station (19715

112

2009, Table 1) indicate that the temperature is highest in July (32.0 C) and lowest in January (12.9

1 113

C) with an average of 210 rainy days per year. The distribution of precipitation is higher on the

2 3 114 4 5 6 115 7 8 116 9 10 11 117 12 13 118 14 15 119 16 17 18 120 19 20 121 21 22 23 122 24 25 123 26 27 28 124 29 30 31 125 32 33 126 34 35 36 127 37 38 128 39 40 41 129 42 43 130 44 45 46 131 47 48 132 49 50 51 133 52 53 134 54 55 135 56 57 58 136 59 60 137 61 62 63 64 65

windward slopes than on the leeward slopes, mostly between October and December. The northeast monsoon lasts from October to January and generally brings extended periods of light rain. Typhoons with heavy rain are most common in summer. In July, 2011, the Central Weather Bureau added a new meteorological station, the Don-Ao Peak station, on the upper slope at the 115.9 km mark on Provincial Highway No. 9. Rainfall statistics in the following two years revealed a staggering average annual rainfall of 6,558 mm (Table 1), which is 1.5 times that of the rainfall at the Suao station. This is an indication that rather than distinct flood/drought periods, this area has stable, plentiful precipitation during the rainy season (in May and June), Typhoon season (from August to October), and during the northeast monsoon (from October to February next year). Table 1 Monthly rainfall statistics from Suao and Don-Ao Peak meteorological stations (from the CWB, Taiwan).

2.3 Triggering factors for landslide events According to historical disaster records, major sediment-related disasters in the catchment area of Don-Ao Peak began with Typhoon Parma in 2009 (R=527mm, I=110 mm on October 5 , 2009). We therefore compiled a list of major rainfall events that caused local slope collapses between 2009 and 2013. Fig. 2 presents the histograms and rainfall accumulations for the individual events. The criterion (or segment principle) of total effective cumulative precipitation (Rtotal in Fig. 2(a)) of a rainfall event is defined as the rule of 4 mm (beginning) - 6 hr (duration) - 4 mm (ending). This means that the rainfall value starts with any rainfall events greater than 4 mm. The whole cumulative rainfall value ends once it is lower than 4 mm during successive 6 hours (Lee, (2006)). Accordingly, the effective cumulative precipitation for triggering landslides (R) is calculated from the time the rainfall starts to the occurrence of a landslide (Figs. 2(a)-(c)). Among the examples examined, Typhoon Megi (2010) and Typhoon Saola (2012) brought torrential rains and landslides. Typhoon Megi is considered an important disaster-causing rainfall 6

138

event with the accumulation at Don-Ao a total rainfall of 939 mm between October 19 and 21 and

1 139

the intensity of the rainfall reached 121.0 mm/hr (13:00-14:00, October 21), breaking historical

2 3 140 4 5 141 6 7 8 142 9 10 143 11 12 13 144 14 15 145 16 17 18 146 19 20 147 21 22 148 23 24 25 149 26 27 150 28 29 30 151 31 32 152 33 34 35 153 36 37 154 38 39 40 155 41 42 156 43 44 157 45 46 47 158 48 49 159 50 51 52 160 53 54 161 55 56 57 162 58 59 163 60 61 62 63 64 65

rainfall records in Taiwan. Onsite surveys revealed that the large-scale landslide caused by Typhoon Megi at Don-Ao Peak began sometime between 12:30 and 13:30 (Fig. 2(a)). Shortly thereafter, rainfall had reached that of the local 200-yr return period (R>600 mm, 24 hr), further inducing regional flooding around Yilan and the Suhua Highway as well as debris flows and collapses. Typhoon Megi caused 21 fatalities along the Suhua Highway. Typhoon Nalgae (October 2, 2011; Fig. 2(b)), the following year, did not make landfall on Taiwan, however, the effects of its outer circulation and the concurrent northeast monsoon brought torrential rains with an effective accumulated rainfall of 513 mm in the study area. The rain peaked in intensity at 78.5 mm/hr and expanded the Don-Ao landslide-prone area to the west and east sides of Don-Ao Peak (Fig.1(c)). The 512 rainfall event (Fig. 2(c)) refers to torrential rains in the Nanao area (located about 10 km to the south of Don-Ao) caused by convective clouds on the afternoon of May 12, 2012, during the rainy season. The Don-Ao Peak station measured a total rainfall amount of 428.5 mm with an intensity of 32.5 mm/hr, which caused sinking displacement in slope-top scarps (Fig.3 (f)). Rainfall intensity-duration-frequency analysis, revealed that the long-term rainfall intensities triggered by Typhoon Megi (2010) and Typhoon Nalgae (2011) exceeded that of the 200-yr return period. In order to understand the influence of earlier earthquakes on the landslide events in the study area, an earthquake inventory was collected from 1900 to 2011 (data collected from the Central Weather Bureau; http://www.cwb.gov.tw/). Data were extracted for events from the surrounding area with a Richter magnitude scale (ML) greater than 5.0. The maximum magnitude of the earthquakes recorded here is approximately 6.5. Landslides on the unstable hillslope can be triggered directly by earthquakes where ML is equal to or exceeds 5 (peak ground acceleration > 250 gal) (Lee, 2014). The landslide record of the study area (2008/7-2012/12) is compared with both the daily rainfall and earthquake inventory (ML >5.0) and is shown in Fig. 3. The results indicate that there were three important earthquake events which happened during 2010/7-2011/5 but they did 7

164

not immediately induce any landslide disasters. It is worth mentioning that an earthquake event

1 165

occurred on Oct 3, 2010 before Typhoon Megi struck Taiwan (No. 109, ML =5.09, epicenter: located

2 3 166 4 5 167 6 7 8 168 9 10 169 11 12 13 170 14 15 171 16 17 18 172 19 20 173 21 22 174 23 24 25 175 26 27 176 28 29 30 177 31 32 178 33 34 35 179 36 37 180 38 39 40 181 41 42 182 43 44 45 183 46 47 48 184 49 50 185 51 52 53 186 54 55 187 56 57 58 188 59 60 189 61 62 63 64 65

in the area south-east of Donao (the distance between Donao and the epicenter was 20 km; Central Weather Bureau). We assume that the seismic intensity will affect the stability of hillslopes along Route 9 although an earthquake of such magnitude does not necessarily cause regional landslides immediately (in this case the seismic intensity was in the range from grade IV to V in the study area). Significant ground vibration by earthquakes can enlarge existing cracks and decrease the stability of natural slopes, as a result of which several significant landslides were induced and brought casualties along Route 9 after Typhoon Megi on Oct 22, 2010 (7 major landslide events occurred with 29 people missing in 1 day). This case highlights the influence of earthquakes on the vulnerable slopes and steep valleys covered with loose colluvium along Route 9. The ground motion of the earthquake led to many apparent landslides with the dropping of torrential rainfall that infiltrate into surface cracks and soil. From the studied case, one can conclude that the combination of both earthquakes and heavy rainfall may trigger and worsen regional landslide disasters on the coastal mountain highway. Fig. 2 Rainfall histograms of landslide disasters: (a) Typhoon Megi (2010); (b) Typhoon Nalgae (2011); and (c) the 0512 rainfall event (2012). Fig. 3 Time relationship between the earthquake magnitude, cumulative rainfall, and number of landslides (2008/07/05-2013/01/05).

2.4 Geomorphological interpretation Four remotely sensed satellite images and orthophotos taken between 2003 and 2012 were collected for this study (Fig. 4). This time period encompasses several major typhoon events that affected the study area: Typhoon Parma (2009), Typhoon Megi (2010), and Typhoon Saola (2012). The remote sensing images indicate that early in this period (2003-2009), the study area was comprised of highly vegetated forest compartments (Fig. 4(a)). However, after Typhoon Parma (2009), a landslide area appeared on the upper slope at 115.9 km of Provincial Highway No. 9 8

190

(marked by the A in Fig. 4(b)), on the midstream section of the Dakeng Stream. Prior to Typhoon

1 191

Megi, erosion gullies near the landslide areas in the midstream section of Dakeng Stream displayed

2 3 192 4 5 193 6 7 8 194 9 10 195 11 12 13 196 14 15 197 16 17 18 198 19 20 199 21 22 200 23 24 25 201 26 27 202 28 29 30 203 31 32 204 33 34 35 205 36 37 206 38 39 40 207 41 42 208 43 44 209 45 46 47 210 48 49 211 50 51 52 212 53 54 213 55 56 57 214 58 59 215 60 61 62 63 64 65

significant erosion, and the left upstream tributary also showed signs of headward erosion. The erosion gullies on the lower slopes at 116.1 km and 116.8 km progressed towards the road, an indication of the imminent major roadbed collapses between 115.8 km and 115.9 km (2011/10/27, 2012/10/24), which obstructed traffic in both directions. After Typhoon Megi (2010) hit the eastern region of Taiwan, landslide areas and erosion gullies became more obvious. The ditches on both sides of the lower slopes at Dakeng Bridge (115.9 km) showed signs of severe lateral erosion and bed erosion, which ultimately induced a roadbed collapse at 116.1 km (Fig. 4(c)). On the east side, the debris flow produced a symmetrical alluvial fan in the ocean. In July, 2011, the upper rims of the two lobe-shaped scarps below Don-Ao Peak showed the development of several significant tension cracks (to the northwest of B and C in Fig. 4(d)). These cracks ranged in length from dozens to a hundred meters, and as they were scoured by heavy debris flow, the landslide on the side of the lower slope at 115.9 km became increasingly severe (marked by D, E, I, and J in Fig. 4(d)). Clear signs of erosion gully development could also be seen in the upstream catchment area of the wild stream to the northeast. In 2011, after the rains of Typhoon Nalgae and its outer circulation struck the study area, the lobe-shaped landslide area at the bottom edge of Don-Ao Peak continued to develop headward along the tension cracks (to the north of B and C in Fig. 4(e)). New tension cracks and lateral erosion gullies also appeared along the expanding landslide area. Following the 512 rainfall event in 2012, the landslide area to the left of the upstream section of Dakeng Stream again presented headward erosion (to the northeast of B in Fig. 4(f)). Downcutting of the erosion gullies where the two lobe-shaped landslide areas met also caused a portion of the left bank to collapse. Table 2 lists the areas of landslides and alluvial fans at 115.9 km of Provincial Highway No. 9, resulting from the various Typhoon and rainfall events. Correlations are outlined in the subsequent analysis. Table 2 Variations in landslide area on the upslope at 115.9 km of Provincial Highway No. 9. 9

216

Fig. 4 Comparison of remote sensing images at 115.9 km on Provincial Highway No. 9 between

1 217

2004- 2013 (FS-2, SPOT 5).

2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

10

218

1 2 3 219 4 5 220 6 7 221 8 9 10 222 11 12 223 13 14 15 224 16 17 225 18 19 20 226 21 22 227 23 24 228 25 26 27 229 28 29 230 30 31 32 231 33 34 232 35 36 37 233 38 39 234 40 41 42 235 43 44 236 45 46 237 47 48 49 238 50 51 239 52 53 54 240 55 56 241 57 58 59 242 60 61 62 63 64 65

3. Methodology The aim of this study is to investigate and explore landslide characteristics in areas with high landslide susceptibility along the coastal mountain highway based on a combination of historical landslide records, field surveys, remote sensing interpretation, monitoring data (rainfall and GPS ground displacement), and LiDAR measurements to shed light on the potential for disaster and the evolution of post-disaster landscapes along the slopes at Don-Ao Peak (115.8 -116.4 km). We performed analysis of the rainfall associated with recent disasters in conjunction with high-precision digital mapping of the terrain (resolution: 1 m) using full waveform airborne light detection and ranging (LiDAR, Riegl LMS-Q680i, 2011/7) as well as ground-based LiDAR (Riegl VZ1000, VZ620, 2013/8-10). For the purpose of understanding the long-term landform evolution, we interpreted the micro-topography in the landslide areas from multiple remote sensing images from 2003-2012 and analyzed the evolution of the landscape and the mechanisms by which the debris slide transformed into a debris flow. Post-disaster field surveys and displacement monitoring data were also examined to characterize the features of deep-seated landslides in the catchment area near the Dakeng Bridge in order to provide a reference for future monitoring processes and the implementation of measures to avoid disasters. Information from a dual-band high-precision GPS receiver (Trimble 5700) at the landslide monitoring and warning system in the Don-Ao Peak landslide area (at 115.9 km on Provincial Highway No. 9) was analyzed to understand the correlation between rainfall data and the degree of displacement and settlement on the upper slopes and to investigate topographical changes in the landslide area. The GPS receiver (the base station of which is located in downtown Suao, 6.6 km to the north of the study area) includes the Trimble 5700 base station and Trimble 4D control displacement monitoring system (4D indicates the E-W, N-S, vertical, and t dimensions). Automated real-time monitoring (with a renewal frequency of 1 Hz) of the landslide area along the upper slopes enables the calculation of the bearings and distance at various points. In addition, an openness visualization technique called the ‘sky view factor’ was 11

243

applied to enhance the terrain features of the potential landslide (Zakšek et al., 2011; Doneus,

1 244

2013). This approach allows efficient highlighting of the relief characteristics on a two-dimensional

2 3 245 4 5 6 7 246 8 9 10 247 11 12 13 248 14 15 249 16 17 18 250 19 20 251 21 22 252 23 24 25 253 26 27 254 28 29 30 255 31 32 256 33 34 35 257 36 37 258 38 39 40 259 41 42 260 43 44 261 45 46 47 262 48 49 263 50 51 52 264 53 54 265 55 56 57 266 58 59 267 60 61 62 63 64 65

map, so the sub-scarp and tension cracks can emerge from the ridge or hillslope.

4. Results and discussion 4.1 Post-disaster physiographic characteristics of landslide areas Several field surveys were conducted after Typhoons Megi (2010), Nalgae (2011) and Saola (2012), and the 11/11 rainfall event. In addition to observing the conditions in the area near the 115.9 km mark following sediment-related disasters, we investigated the deposition characteristics and the erosion induced by rainfall in the channel of the Dakeng Stream. The study area shown in Fig. 5, can be divided into three sections following Typhoon Megi, according to the location of the disaster: 1) the source area upstream of the landslide (the two lobe-shaped source areas below DonAo Peak); 2) the transportation zone (channel of the Dakeng Stream); and 3) the deposition zone (the river mouth-coastal line alluvial fan). Previous post-disaster field surveys (Fig. 6) revealed that the depth of the landslides in the two lobe-shaped source areas reached nearly 30 m (measured from the digital elevation model (DEM) by airborne LiDAR), which means they can be categorized as large-scale, deep-seated landslides. The main scarp continued to expand headward, serving as the main source of sediment in the area. However, the sliding of the mass of the landslide into the channel of the Dakeng Stream, due to gravity or rainfall, resulted in blockages as a consequence of a narrowing of the channel section. Two blockages occurred in the transportation zone: one at the old check dam near the junction of different lithological distributions in Fig. 6 (a) and the other at the Dakeng Bridge on Provincial Highway No. 9, where the existing culverts under the bridge were insufficient. When the debris flow was obstructed by the Dakeng Bridge, a portion of the boulders and sediment spilled over the surface of Highway No. 9, obstructing traffic (Fig. 6(b)). Field surveys revealed that the key strata distributed within the source area originated from amphibolite outcrops of the Don-Ao schist stratum (Fig. 7). Despite the high compression strength of the rock, 12

268

extreme weathering produced multiple sets of fractures and weak planes within it (Fig. 7(a)). Thus,

1 269

the rock exposed below the scarp was comprised of extremely fragmented rock materials (Figs. 7(c)

2 3 270 4 5 271 6 7 8 272 9 10 273 11 12 13 274 14 15 275 16 17 18 276 19 20 277 21 22 278 23 24 25 279 26 27 280 28 29 30 281 31 32 282 33 34 35 283 36 37 38 284 39 40 41 42 285 43 44 286 45 46 287 47 48 49 288 50 51 289 52 53 54 290 55 56 291 57 58 59 292 60 61 62 63 64 65

and 7(d)). Furthermore, following Typhoon Megi, a groundwater exfiltration point appeared near the lithological junction (amphibolite/schist) in the midstream section of the Dakeng Stream (marked as E1 in Fig. 7(b)), which became a source of runoff into the Dakeng Stream. It is worth mentioning that this forms the upper edge of the earliest landslide after Typhoon Parma, and still presents the formation of a path of seepage for groundwater. The continuous runoff supplied by the seepage of groundwater was sufficient to initiate the fluidization of the colluvial soil in the Dakeng Stream during rain events. With regards to the slope characteristics, the study area is located within a concave section of the road. The geomorphology of the catchment area of the Dakeng Stream is characterized by concave-shaped slopes, which facilitate the convergence of debris accumulation and water flow. Physiographically speaking, the area is ideally suited to such disasters. Fig. 5 Three-dimensional terrain map of landslide hazards at 115.9K of Provincial Highway No. 9. Fig. 6 Dakeng Bridge at 115.9 km of Provincial Highway No. 9 was blocked with debris flows after (a) Typhoon Megi; and (b) Typhoon Nalage. Fig. 7 Onsite survey in the upstream landslide area of the Dakeng gully catchment.

4.2 Geological structures and failure mechanisms With regards to the local geological structures, Yeh (1998) examined the distribution of foliation in the Suao-Nanao area and reported that most of the secondary foliation in the strata on the north side dips towards the south, whereas in the strata on the south side slope it slopes toward the north (presenting a fan structure) with amphibolite as its core. The geology of the study area exhibits the features of ductile deformation structures. In terms of the geological zone, the study area belongs to the east wing of the Central Mountain Range, which has been compressed by tectonic stress (from east to west) from plate collision. Thus, most failure patterns in the area are comprised of slate or schist that has broken along foliation planes. As shown in Fig. 8, most of the 13

293

failure patterns on the upper slopes of the catchment area of Dakeng Stream are gravitationally

1 294

deformed schist. The gradual process from stable to unstable in the slope body of the left bank of

2 3 295 4 5 296 6 7 8 297 9 10 298 11 12 13 299 14 15 300 16 17 18 301 19 20 302 21 22 303 23 24 25 304 26 27 305 28 29 30 306 31 32 307 33 34 35 308 36 37 309 38 39 40 310 41 42 311 43 44 312 45 46 47 313 48 49 314 50 51 52 315 53 54 316 55 56 57 317 58 59 318 60 61 62 63 64 65

the Dakeng Stream, with the strata of Don-Ao schist undergoing sediment erosion and weathering are illustrated in Fig. 8(a) and 8(b). In general, in weathered and foliated rock, flexural toppling leads to failure along the foliation plane. An intersecting cross-joint set in this type of rock will generally be the location of a blocky, toppling failure or block-flexural toppling failure when the rock is unable to withstand the tensile bending stress (Amini et al., 2012). The stratigraphic column obtained from geological drilling (depth (h): 15 m, on the ridge of the Don-Ao Peak) consists of surface soil (h=0-1.6 m), amphibolite with high weathering (h=1.6-13.2 m), and slightly fractured amphibolite bedrock (h=13.2-15.0 m). The results of mechanics testing point out that the rock mass is characterized by a fragmental condition and poor engineering applications. Rainfall can easily infiltrate the rock mass through surface cracks. Hence, the weight of the rock mass increases gradually with the increase in the degree of saturation of the rock. The unstable rock mass in the source zonation or deposited colluvium on the hillslope begins to collapse (or sliding) once the critical conditions for gravitational deformation are exceeded. Additionally, the toe erosion effect of the debris slide enhances the volume of mass movement. It also greatly contributes to the high mobility of the sliding mass transforming into the debris flow that moves down the valley. Primary failures on the lower slopes of the Dakeng Stream between 115.7 km and 115.9 km included debris sliding in the colluvium and lateral and streambed erosion of the debris flow in the channel due to entrainment. Field survey results (Section A-A’ in Fig. 9) revealed that in addition to the exposure of fault shear zones on both sides of the stream (the features of which are described in the next section), the rocky slope of both banks also exhibited several tension cracks (exfoliation) parallel to the slope surface and the stream (Fig. 9(a)). These tension cracks were also observed to develop in the upper rocky layer (Fig. 4(d)-(f)), which would likely not be conducive to maintaining the stability of the rock slope and was thus the main cause of collapses on both banks. Furthermore, buckling due to gravitational deformation can be seen in the black schist cleavage with the Dakeng 14

319

Stream and wedge failure in the lower slope at 116.1 K adjoin (Fig. 9(d)), as well as in the rock

1 320

slope of the bank near the channel.

2 3 321 4 5 322 6 7 8 323 9 10 324 11 12 13 325 14 15 326 16 17 18 327 19 20 328 21 22 329 23 24 25 330 26 27 28 331 29 30 332 31 32 33 333 34 35 334 36 37 335 38 39 40 336 41 42 337 43 44 45 338 46 47 339 48 49 50 340 51 52 341 53 54 342 55 56 57 343 58 59 344 60 61 62 63 64 65

In summary, the geological structures and mechanisms underlying the occurrence of disasters in the study area were the result of previously existing headward development of the erosion gullies along the upper slopes of the Don-Ao Peak. This induced a large-scale landslide with the collapse of massive quantities of sediment into the channel of the Dakeng Stream. Subsequent rainfall produced runoff that changed the sediment into debris flows that gradually progressed downstream to the coast. Onsite measurements of the attitude of exposed strata in the upstream landslide area and the attitude of other strata along the highway (Fig. 10(a)), in conjunction with the distributions of joint density (Fig. 10(b), where the shaft diameter is the number of attitude distributions) indicate that the attitude of most of the cracks exposed in the landslide area is close to that of the local geological structures (N70

W/38

S). These results demonstrate that the large-scale landslides caused by

typhoons and heavy rains in this area are primarily controlled by the geological conditions. Runoff produced by heavy rainfall and the seepage of groundwater are considered triggering factors. The extent of exfoliation and the fact that the amphibolite in the source area is harder and more fragmented than the graphite schist made it easier for surface water to infiltrate the slope from the crest of the slope or fissures in the slope face. The groundwater then travels along the permeable strata to be exuded from the impermeable stratum near the slope toe. The slope toe was easily damaged under the double influence of side erosion from the groundwater and debris flow in the valley. A loss of support from beneath caused the area over the slope toe to deform, which caused slippage and creeping between the strata of the slope surface and intensified fragmentation. Increasingly severe fragmentation in the body of the slope enabled substantial water infiltration into the permeable stratum. The groundwater percolated through the fragmented amphibolite rock strata and along the less permeable schist stratum beneath it to accumulate at the impermeable stratum near the toe of the slope, undermining the stability of the entire slope. The torrential rains brought by Typhoon Megi caused the groundwater levels within the slopes in the source area to rise rapidly, 15

345

thereby expediting erosion in the slope toe and destabilizing the entire slope. Therefore, during this

1 346

landslide event, the infiltration of substantial quantities of rain and increased side erosion in the

2 3 347 4 5 348 6 7 8 349 9 10 350 11 12 13 351 14 15 352 16 17 18 353 19 20 354 21 22 355 23 24 25 356 26 27 28 357 29 30 358 31 32 33 34 359 35 36 37 360 38 39 361 40 41 42 362 43 44 363 45 46 47 48 364 49 50 365 51 52 366 53 54 55 367 56 57 368 58 59 60 369 61 62 63 64 65

valley caused the section between the slope toe and the slope face to collapse first. In that instant, the near-saturated sliding mass behind became unstable and slid downstream. The mechanisms underlying this collapse should be taken into consideration in subsequent slope monitoring or remediation projects along the Su-Hua Highway (Lo et al., 2014). Historical earthquake records also show that prior to Typhoon Megi, the No. 109 earthquake on October 3, 2010, which measured 5 on the Shindo? seismic intensity scale at the Nanao meteorological station, generated north-south ground acceleration (195.6 gal) close to the threshold required to cause a landslide (250 gal; Lee, 2012). This is another crucial factor in the large-scale landslide at Don-Ao Peak. Fig. 8 Schematic sketch showing the gravitational deformation of foliated rock. Fig. 9 Schematic diagram of geological structure in the downstream landslide area of the Dakeng gully catchment. Fig. 10 (a) Mapping of rock attitude; and (b) joint density in the study area.

4.3 Displacement monitoring and analysis of landslide area Precipitation is often the primary cause of landslides. Rainfall saturates the surface layer of the slopes, whereupon the wetting front and gradual rise in groundwater eventually overlap and trigger slope failure mechanisms. The relationship between the hourly rainfall and vertical ground displacement in the source area around the crest line proposed in this paper is discussed below.

(1) Typhoon Nalgae in 2011 Although Typhoon Nalgae did not make direct landfall on Taiwan, its outer circulation reacted with the northeast monsoon from the cold high pressure of Mainland China resulting in torrential rains in Yilan and Hualien. Post-disaster field surveys indicated that this Typhoon event caused an expansion of the central portions of the west and east landslide areas in the upstream reaches on Don-Ao Peak. The rainfall-induced sliding of debris tended to follow the development of tension 16

370

cracks formed after Typhoon Megi, resulting in headward erosion of the erosion gullies in the

1 371

landslide area. The largest displacement of 140 cm was obtained at GPS05 (Fig. 1(c)). Fig. 11

2 3 372 4 5 373 6 7 8 374 9 10 375 11 12 13 376 14 15 377 16 17 18 378 19 20 379 21 22 380 23 24 25 381 26 27 28 29 382 30 31 383 32 33 384 34 35 36 385 37 38 386 39 40 41 387 42 43 388 44 45 389 46 47 48 390 49 50 391 51 52 53 392 54 55 393 56 57 58 394 59 60 395 61 62 63 64 65

presents the vertical displacement and settlement with corresponding rainfall. This event exhibited a rainfall histogram with a peak at the center. Debris slides were triggered when hourly precipitation was highest (11: 00, 03 Oct. 2011). Time lag characteristics associated with rainfall and subsequent infiltration (the response to infiltration is slower, and, as the weight of the sliding mass increases, it may continue to slide even after the rain has stopped) resulted in a slide lasting approximately 42 hrs (from 05:00 03 Oct. to 23:00 04 Oct). The initial sliding speed obtained from monitoring of the displacement on the vertical axis of the GPS receiver was higher (4.4 cm/hr), gradually decreasing to 1.3 cm/hr. This slope failure can thus be categorized as slow landslide deformation. Fig. 11 The relationship between displacement (vertical) of GPS05 and hourly precipitation at DonAo Peak.

(2) 0512 rainfall event in 2012 At 15:00 on May 12, 2012, torrential rains struck the Suao and Nanao areas in Northeastern Taiwan (R>500 mm). At the Don-Ao Peak station, the event resulted in the total accumulation of 452 mm of rainfall. Hourly precipitation peaked at 62 mm/hr, causing the bare ground at the top of the upper slopes (right side of the Dakeng Stream) to slide downstream along the direction of the tension cracks that formed southwest of GPS05 after Typhoon Nalgae (Fig. 1(c)). The greatest displacement was measured at GPS04 on the right wing of the landslide area, where the amount of settlement in the north-south direction was approximately 64 cm. Fig. 12 shows a comparison of the maximum displacement and settlement values with hourly precipitation. The displacement curve shows that the landslide event associated with the 0512 rainfall was similar to that caused by Typhoon Nalgae in 2011. In addition to a direct association with the intensive rainfall and the histogram characterized by a peak at the center, when the great amount of runoff or debris flow passed by downstream, the lower slope was subjected to severe side erosion and hollowing out, which triggered multiple collapses in the lower slope downstream. 17

396

Fig. 12 Relationship between displacement (north-south) of GPS04 and hourly precipitation at Don-

1 397

Ao Peak.

2 3 398 4 5 6 399 7 8 400 9 10 11 401 12 13 402 14 15 16 403 17 18 404 19 20 405 21 22 23 406 24 25 407 26 27 28 408 29 30 409 31 32 33 410 34 35 411 36 37 38 412 39 40 413 41 42 414 43 44 45 415 46 47 416 48 49 50 417 51 52 418 53 54 55 419 56 57 420 58 59 421 60 61 62 63 64 65

4.4 Numerical simulation of landslide process A simulation of the landslide movement process was used to investigate the process underlying the movements involved in the large-scale debris slide triggered by Typhoon Megi in 2010 at 115.9 km on Provincial Highway No. 9. (Lo et al., 2014) To explain the processes involved, we monitored the movement speeds throughout the entire process. This revealed variations in speed during different stages and clarified the key processes involved in the overall movement. The results of the numerical simulation are outlined below. (a) Fig. 13 presents the simulation results of the movement process and variations in speed associated with the landslide at 115.9 km. The numerical model was designed based on the results obtained by subtracting the pre-event DEM from the post-event DEM (pre-event: 2004; post-event: 2010). Five primary landslide masses were designated from upstream to downstream. Two of these areas were situated in the upstream source area (sliding mass A and B), one in the midstream section (sliding mass C), and two in the curve of the valley at the location of the collapse (sliding masses D and E). Sliding masses A and B in the source area slumped directly downstream as a result of gravity, whereas sliding masses C, D, and E maintained higher self-supportability due to their larger coefficient of friction (equal to 0.6) and greater bonding strength. When sliding masses A and B came into contact with sliding masses C, D, and E downstream, the lack of support from the slope toe, which had been hollowed out by side erosion, caused them to collapse. This coincides with the landslide mechanisms derived from the onsite survey. Furthermore, in this study a PFC3D basic fluid option was used to install a water body downstream where the sea would be. When the particle elements encountered the body of water, their movement speed was reduced through fluid resistance, causing them to accumulate in an alluvial fan. This also enhanced the applicability of the landslide simulation. (b) These simulation results (Fig. 13) indicate that the entire landslide process progressed through seven key phases from sliding, accelerating, and decelerating to a final stop. In the first phase, sliding masses A and B in 18

422

the source area began moving (at the 5 sec point in Fig. 13). The speed of the landslide in this phase

1 423

ranged between 5 m/sec and 20 m/sec (approximately 18-72 km/hr) with an average speed of 8.96

2 3 424 4 5 425 6 7 8 426 9 10 427 11 12 13 428 14 15 429 16 17 18 430 19 20 431 21 22 432 23 24 25 433 26 27 434 28 29 30 435 31 32 436 33 34 35 437 36 37 438 38 39 40 439 41 42 440 43 44 441 45 46 47 442 48 49 443 50 51 52 444 53 54 445 55 56 57 446 58 59 447 60 61 62 63 64 65

m/sec (approximately 32.3 km/hr, as shown in Fig. 15). The moving masses underwent collision interaction, in which the mass in the rear transferred energy to the mass in front. This increase in kinematic energy caused the downward motion of the mass in front to accelerate. Therefore, the masses moving the fastest were for the most part in the middle and front portions of Masses A and B, while those in the rear portions moved more slowly. (c) Second phase (15 sec point in Fig. 13): sliding masses A and B migrated past the steepest terrain in the region, converged at the Southern Unnamed Creek, and reached the slope toe of sliding mass C. At this point, the front portions of sliding masses A and B were moving at the highest speed of 52.2 m/sec with an overall average speed of 20.4 m/sec, as shown in Fig. 15. The rear portions were also beginning their accelerated descent. (d) Third phase (22.5 sec point in Fig. 13): the front portions of sliding masses A and B arrived at the highway where the Chuang-Yi tour group bus was parked on the roadside. Because the terrain in the valley turns gradually from southeast to south, the front portion of the sliding mass began slowing down (the average speed was reduced to 10.36 m/sec). Here, the landslide gradually transformed from a large-scale debris slide to a debris flow, due to side erosion in the slope toe. Sliding mass C became unstable and began collapsing in the downstream direction. (e) Fourth phase (40 sec point in Fig. 13): by this time, sliding mass C had completely collapsed, and the front portions of sliding masses A and B had reached sliding mass D. In terms of terrain, sliding mass D was situated on an undercut slope at the bend of the watercourse. This caused the front portions of sliding masses A and B to accelerate again (increasing to 27.5 m/sec) and quickening the erosion at the slope toe of Mass D. (f) Fifth phase (60 sec point in Fig. 13): due to side erosion, sliding mass D collapsed, and the front portions of sliding masses A and B passed another bend in the creek past the toe of sliding mass E. Severe side erosion caused by the sliding masses from upstream severely destabilized mass E, which was located on the lower slope of the Su-Hua Highway at 115.8 km. (g) Sixth phase (70 sec point in Fig. 13): by this time, sliding mass E had completely collapsed, and a 19

448

portion of the sliding masses were beginning to pass the second bend before they reached the sea.

1 449

Due to the curve of the valley, the masses accelerated once more, reaching average speeds of 12.40-

2 3 450 4 5 451 6 7 8 452 9 10 453 11 12 13 454 14 15 455 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

12.57 m/sec during the period between 70 sec and 77.5 sec. At approximately 77.5 sec, they gradually decreased in speed. At 70 sec, the front portions of sliding masses A and B had reached the sea and were also slowed down by this body of water. In the seventh phase (220 sec point in Fig. 13), an alluvial fan formed, ending the movement of the landslide event at approximately 220 sec. Fig. 13 Movement process and speed variations in the simulated landslide at 115.9 km of Provincial Highway No. 9.

20

456 1 2 457 3 4 5 458 6 7 8 459 9 10 460 11 12 13 461 14 15 462 16 17 18 463 19 20 464 21 22 465 23 24 25 466 26 27 467 28 29 30 468 31 32 469 33 34 35 470 36 37 471 38 39 40 472 41 42 473 43 44 474 45 46 47 475 48 49 476 50 51 52 477 53 54 478 55 56 57 479 58 59 480 60 61 62 63 64 65

4.5 Interpretation of topographical changes and landslide micro-photographic images This study compared digital elevation models from 2004, 2011 (after Typhoon Megi), and 2013 (after Typhoon Saola) (Fig. 14). The locations of the cross-sections are presented in Fig. 14 (stretching from the landslide area on the right wing of the Don-Ao Peak along the Dakeng Stream to the alluvial fan at the mouth of the river). The three cross-sections represent topographical changes in the area, from before the disasters to after the two most recent typhoon events. The figure shows that before and after Typhoon Megi (the black dashed lines and the blue bold lines in Fig. 14, respectively), the primary landslide and erosion areas were at the bottom edge of the DonAo Peak (at most, the depth of the landslide was approximately 30 m) and at the midstream section of the Dakeng Stream where the two lobe-shaped landslide areas met (maximum erosion depth was approximately 15 m). Outcrops could be seen throughout the streambed after the disaster, and the two old check dams midstream at Dakeng Bridge had been completely silted up and destroyed, forming a pileup under Dakeng Bridge. Deep streambed erosion (approximately 25 m) also appeared at the bend in the Dakeng Stream (marked “curved channel” in Fig. 14) after Typhoon Megi. This can be attributed to scouring of the cutbank on the bend and the side bank of the debris flow. Furthermore, an alluvial fan with gentle changes in slope can be clearly seen at the mouth of the river, descending into the ocean in the form of a Gilbert-type fan delta. According to the underwater side-scan sonar results obtained by Chou et al. (2012), the alluvial fan extends approximately 150 m along the seabed. A comparison of changes in the terrain profiles before and after Typhoons Megi and Saola show that only the bare landslide area on the right wing upstream of Don-Ao Peak presented sliding at the ground surface, which was comprised of collapsed scarps resulting from old tension cracks. The newly constructed dam in the midstream section of the Dakeng Stream also presented pileups from being silted up completely. Disturbance from construction and the flow of the debris flow resulted in downcutting erosion in the channel 21

481

downstream, which was most pronounced under Dakeng Bridge and at the bend. The radial length

1 482

of the alluvial fan remained roughly the same with a slight retreat of approximately 7 m, and the

2 3 483 4 5 484 6 7 8 485 9 10 486 11 12 13 487 14 15 488 16 17 18 489 19 20 490 21 22 491 23 24 25 492 26 27 493 28 29 30 494 31 32 495 33 34 35 496 36 37 497 38 39 40 498 41 42 499 43 44 500 45 46 47 501 48 49 502 50 51 52 503 53 54 504 55 56 57 505 58 59 506 60 61 62 63 64 65

deposition zone comprised of finer particles expanded on both sides (in the east-west direction). We speculate that this is associated with the partial erosion of erosion gullies near 115.9 km (at 116.8 km and 117.0 km). However, the main channel of the stream above the fan was altered by the previous flow of debris and sediment congestion. Overall, the topographical changes matched the field survey results. Interpretation of high-precision micro-photographs of the landslide obtained from airborne LiDAR (2011/7, 2013/8) and ground-based LiDAR (2013/9) enabled us to approximate the ground surface conditions without vegetation and identify potential areas of disaster in the study area. We adopted a one-meter digital elevation model (DEM), digital surface models (DSM), and orthophotos for the same time period. These were used in conjunction with the spatial analysis module from the ArcGIS software to produce slope maps, aspect maps, contour maps, and eight-directional sun shadow maps. This interpretation was also aided by the use of multiple satellite images (FS-2, SPOT5) as well as geological maps and potential disaster maps published by the Central Geology Survey. Our aim was to identify current landslide areas and disaster-prone areas. For our methods of interpretation, we referred to The Investigation and Analysis Project of Geologically Sensitive Areas for Homeland Conservation (Central Geology Survey, 2012) with regard to the topographic characteristics of deep-seated landslides and large-scale landslides (Fig. 15), such as the scarps, crown tension fractures, side fissures, break lines, erosion gullies, bulging slope toes, or sliding masses (Takasuke, 2000; Agliardi et al., 2001; Chigira, 2009). The more significant landslide terrain characteristics in the analysis included scarps, surface ruptures, sliding ranges, and erosion gullies. The gully lines were categorized as either Type A (headward erosion) or Type B (no headward erosion). Shading relief with a sky view factor and reference features associated with gravitation deformation areas as described in Soeters and van Westen (1996) and Chigira (2014) (Table 4) were adopted to identify the boundary of the gravitation deformation area in the Dakeng Bridge 22

507

watershed. The identified area was then entered into the landslide area-volume empirical equation

1 508

to compute the probable landslide volume. That volume was then input into the numerical model to

2 3 509 4 5 510 6 7 8 511 9 10 512 11 12 13 513 14 15 514 16 17 18 515 19 20 516 21 22 517 23 24 25 518 26 27 519 28 29 30 520 31 32 521 33 34 35 522 36 37 523 38 39 40 524 41 42 525 43 44 526 45 46 47 527 48 49 528 50 51 52 529 53 54 530 55 56 57 531 58 59 532 60 61 62 63 64 65

estimate the zone affected by the landslide. The interpretation of the gravitational deformation areas is shown in Fig. 16. The results of landslide micro-photography analysis in two catchment areas are shown below. Fig. 14 Surface profile of the main landslide and the corresponding geological setting in 2004, 2011, and 2013. Fig. 15 Active landslide geomorphic evolution and deep-seated landslide characteristics (revised from Takasuke, 2000). Fig. 16 Results of micro-photographic interpretation of the landslide at the 115.9 km watershed on Provincial Highway No. 9. Table 4 Irregularities associated with gravitation deformation areas (from Soeters and van Westen, 1996; Chigira, 2014). Catchment area of the Dakeng Stream The crown of the bare landslide on the southwest side of Don-Ao Peak at the source of the Dakeng Stream still exhibits continuous tension cracks (Fig. 16), which have already traversed the ridge line. The depressed region at the top of the slope shows distinct cracking in the sliding surfaces, which may progress and cause landslides in the future. Due to erosion from rainfall, the bare ground also shows the continued headward development of erosion gullies. Rainfall events of greater rainfall intensity (2013/11/11) have produced small-scale landslides on the slope face. Some of the loose mass that has slid from the scarp is still deposited on the slope face (break line). Reinforcement projects at the top of the slope have decelerated the expansion of tension cracks and lateral cracks at the upper edge of the landslide area on the left wing of Don-Ao Peak. Most of the vegetation at the top of the slopes in the study area is comprised of herbaceous plants with shallow roots, which makes the headward development of the erosion gullies more severe. Deeper erosion gullies are clearly shown on bare slopes. In addition to more significant expansion in area on the 23

533

left wing than on the right wing over the last few years, the landslide area on the left wing is also

1 534

gradually progressing towards the catchment area of the wild stream. At the place where the two

2 3 535 4 5 536 6 7 8 537 9 10 538 11 12 13 539 14 15 540 16 17 18 541 19 20 542 21 22 543 23 24 25 544 26 27 545 28 29 30 546 31 32 547 33 34 35 548 36 37 549 38 39 40 550 41 42 551 43 44 552 45 46 553 47 48 49 554 50 51 555 52 53 556 54 55 56 557 57 58 558 59 60 559 61 62 63 64 65

landslide areas on the two wings meet (the lithological junction shown in Fig. 16), two more landslide areas (the restored ones formed downstream after Typhoon Parma) and erosion gullies appear to be developing headward. The erosion gully cuts along the side boundary of a hummocky sliding mass, which may induce a landslide in the future. The lower slope channel of the Dakeng Stream between 116.1 km and 116.4 km encompasses an area where previous disasters have occurred. In this area, headward progress is apparent which is likely to cause toe erosion and sliding, due to the fact that it lies on the outside of the bend of the Dakeng Stream at 116.0 km and where the downstream portions of the erosion gullies meet at 116.4 km. Furthermore, high resolution terrain shows that the alluvial fan at the mouth of the river comprises coarser particles provided by the landslide area upstream of the Dakeng Stream. In contrast with the terrain near the two sides of the fan (in the east-west direction), the size of sediment particles deposited at the mouth of the river are smaller due to the shorter erosion gullies and smaller source area (comprising mostly shallow-seated landslides). This indicates that the composition of the materials in the alluvial fan is determined by the scope of the upstream landslide and the depth of the slides.

Catchment area of the wild stream to the north The catchment area of the wild stream contains concave-shaped forested terrain. The vegetation at the top of the slopes is similar to that in the catchment area of the Dakeng Stream, comprising herbaceous plants. Despite the lack of landslides and erosion gullies with headward erosion in this area, several erosion gullies with vertical contours began appearing along the ridgeline after Typhoon Megi. The micro-photographic interpretation maps without vegetation (Fig. 16) show distinct expansion of several scarps at the ridgeline (curved scarp lines with lengths reaching tens of meters). These scarps developed similar to those in the Dakeng Stream catchment, forming a landslide boundary in the shape of an inverted cone. On the right side within the boundary are mound-shaped sliding masses clearly cut by erosion gullies. Comparison of the 1936 and 2004 24

560

DEMs show the addition of a new tributary in the midstream section of the wild stream as well as

561 1

marked changes in the upstream channel near the ridgeline. Prior to Typhoon Megi, Chu et al.

2 3 562 4 5 563 6 7 564 8 9 565 10 11 12 566 13 14 567 15 16 568 17 18 569 19 20 21 570 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

(2013) reported the existence of numerous hummocky surfaces around the Dakeng Stream and the upstream portion of the wild stream to the north, shown to be the result of earlier debris slides or rock slope deformation. As can be seen in Figs. 16(a) and 16(b), field surveys verified the existence of crest boulders (approximately 4 m in size, mostly amphibolite) in the midstream and downstream sections of the wild stream, possibly the remnants of a debris flow. In contrast with the area of Dakeng Stream landslides, the catchment area of the wild stream is on the leeward side of the terrain (in terms of the northeast monsoon), and therefore does not receive as much precipitation as the Dakeng Stream catchment. Nevertheless, the progress of the upstream scarps should be monitored closely.

25

571 1

5. Conclusions

2 3 572 4 5 573 6 7 574 8 9 10 575 11 12 576 13 14 15 577 16 17 578 18 19 20 579 21 22 580 23 24 581 25 26 27 582 28 29 583 30 31 32 584 33 34 585 35 36 37 586 38 39 587 40 41 42 588 43 44 589 45 46 590 47 48 49 591 50 51 592 52 53 54 593 55 56 594 57 58 59 595 60 61 62 63 64 65

This study investigated the evolution of landscape features after major rainfall events in the catchment areas surrounding the section of Provincial Highway No. 9 between 115.8 km and 116.4 km. Through hydrological analysis, we examined the role of precipitation in the occurrence of landslides and estimated the deposition area (alluvial fan distribution) of resultant debris flows. We also interpreted topographical changes and analyzed micro-photographic images using onsite field surveys, the interpretation of remote sensing data, and high-precision digital terrain. Our interpretation of multiple remote sensing images shows distinctly that landslides in the study area were initially caused by the headward development of erosion gullies. Since Typhoon Megi, the expansion of the landslide area on the upper slopes of Don-Ao Peak tended to follow the arced tension cracks that appeared at the top of slopes. Furthermore, the retreating landslide that affected the roadbed on the lower slopes was associated with the generation of exfoliations and the toe erosion after a large amount of runoff or debris flows. Analysis of the intensity-durationfrequency of the rainfall revealed that most recent landslide-inducing rainfall events near the Dakeng Stream were associated with rainfall amounts greater than that of a 200-year event. This study presents the first application of high-precision digital terrain mapping derived from airborne and ground-based LiDAR for the identification of potential areas of disaster in the vicinity of DonAo Peak. Landslide micro-photography analysis indicates that topographical changes in the landslide areas in the Dakeng Stream catchment are determined by local geological structures (fault zones and lithological junctions), headward-developing erosion gullies, and the concave, watercollecting shape of the slopes. Deposits of boulders from earlier debris flows remain in the vegetated catchment of the wild stream to the north, and the ridgelines of the upper slopes also reveal the formation of several curved scarps. Variations in strain rate in the body of the slope should be monitored to further elucidate the occurrence of sediment-related disasters following typhoons or rainfall events. 26

596

Acknowledgements

1 597

We thank the Fourth Maintenance Office, Directorate General of Highways, MOTC in Taiwan for

2 3 598 4 5 6 599 7 8 600 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

providing the valuable monitoring data for intense rainfall events. The authors would also like to thank Prof. Ming-Lang Lin at the Department of Civil Engineering, National Taiwan University and colleagues work in Sinotech Consultants, INC. for their advice and assistance with this study.

27

601

References

1 602

Agliardi F (2001) Structural constraints on deep-seated slope deformation kinematics. Engineering

2 3 603 4 5 6 604 7 8 605 9 10 11 606 12 13 607 14 15 16 608 17 18 609 19 20 610 21 22 23 611 24 25 612 26 27 28 613 29 30 614 31 32 33 615 34 35 616 36 37 38 617 39 40 618 41 42 619 43 44 45 620 46 47 621 48 49 50 622 51 52 623 53 54 55 624 56 57 625 58 59 626 60 61 62 63 64 65

Geology 59:83-102 Amini M, Majdi A, Veshadi MA (2012) Stability analysis of rock slopes against block-flexure toppling failure. Rock Mech Rock Eng 45:519-532 Aydan Ö, Shimizu Y, Ichikawa Y (1989) The effective failure modes and stability of slopes in rock mass with two discontinuity sets. Rock Mech Rock Eng 22:163-188 Aydan Ö, Kawamoto T. (1992) The stability of slopes and underground openings against flexural toppling and their stabilization. Rock Mech Rock Eng 25(3):143-165 Aydan Ö, Amini M (2009) An experimental study on rock slopes against flexural toppling failure under dynamic loading and some theoretical considerations for its stability assessment. Journal of the School of Marine Science and Technology, Tokai University, 7(2):25-40 Aydan Ö, Ohta Y, Hamada M (2009) Geotechnical evaluation of slope and ground failures during the 8 October 2005 Muzaffarabad earthquake, Pakistan. J Seismol 13:399-413 Aydan Ö, Ohta Y, Daido M, Kumsar H, Genis M, Tokashiki N, Ito T, Amini M (2011) Earthquakes as a rock dynamic problem and their effects on rock engineering structures. Chapter 15: Earthquakes as a rock dynamic problem and their effects on rock engineering structures. Advances in Rock Dynamics and Applications, CRC Press, Taylor and Francis Group, 341422. Central Geological Survey (2011) Disaster survey report for Typhoon Megi in Suao area and Suhua highway. Central Geological Survey (in Chinese) Central Geological Survey (2012) Investigation and analysis for geologically sensitive area in national preservation domain program. Central Geological Survey (in Chinese) Chigira M (2009) September 2005 rain-induced catastrophic rockslides on slopes affected by deepseated gravitational deformations, Kyushu, southern Japan. Engineering Geology 108:1-15 Chigira M (2014) Geological and geomorphological features of deep-Seated catastrophic landslides 28

627

in tectonically active regions of asia and implications for hazard mapping. Episodes Journal of

1 628

International Geoscience 37(4):284-294

2 3 629 4 5 630 6 7 8 631 9 10 632 11 12 13 633 14 15 634 16 17 18 635 19 20 636 21 22 637 23 24 25 638 26 27 639 28 29 30 640 31 32 641 33 34 35 642 36 37 643 38 39 40 644 41 42 645 43 44 646 45 46 47 647 48 49 648 50 51 52 649 53 54 650 55 56 57 651 58 59 652 60 61 62 63 64 65

Chou HT, Lee CF, Lo CM, Lin CP. (2012) Landslide and alluvial fan caused by an extreme rainfall in Suao, Taiwan. 11th International Symposium on Landslides (ISL) and the 2nd North American Symposium on Landslides, Banff, Alberta, Canada pp 487-494 Chu HK, Lo CM, Chang YL (2013) Numerical analysis of slope stability at the 115.9k point of the Su-Hua Highway. Journal of Chinese Soil and Water Conservation 44(2):97-104 Doneus M (2013) Openness as Visualization Technique for Interpretative Mapping of Airborne Lidar Derived Digital Terrain Models. Remote Sens. 5:6427-6442 Hearn GJ (2011) Engineering Special Publication 24 - Slope Engineering for Mountain Roads. Geological Society of London pp 5-20 Lee CT (2012) Characteristics of earthquake-induced landslides and differences compared to storminduced landslides. EGU General Assembly 2012 pp 6937 Lee CT (2014) Statistical seismic landslide hazard analysis: An example from Taiwan. Engineering Geology, 182: 201-212 Lee CF, Huang WK, Chang YL, Chi SY, Lin WC, Chen HR, Lin CS, Chang SC, Liao WC, Chang YH (2013) Establishing Landslide hazard index with landslide susceptibility and rock mass rating system on Highway No.9 (Suao-Nanao), 2nd Conference for highway transportation development, pp 108-121 (in Chinese) Lee KT (2010) Hazard analysis at 104k~120k in Route.9 (Suhua highway, Suao-Don-Ao) during Typhoon Megi. Fourth Maintenance Office Directorate General of Highways, MTOC (in Chinese) Lee MH (2006) The Rainfall threshold and analysis of Debris flows. Ph.D. Thesis, National Cheng Kung University, Taiwan, R.O.C. (in Chinese) Lin CW, Kao MC (2009) Suao-1/50,000 Taiwan geological map (2nd version), No.16, Central Geological Survey (in Chinese) 29

653

Lin ML, Chen RT., Yang DX, Zang PZ, Liu HC, Lin WK, Lo CM, Liao RT, Huang WC, Wang TT

1 654

(2013) Engineering geological assessment for failure mechanism at 115.7~116.1k on Suhua

2 3 655 4 5 656 6 7 8 657 9 10 658 11 12 13 659 14 15 660 16 17 18 661 19 20 662 21 22 23 663 24 25 664 26 27 28 665 29 30 666 31 32 667 33 34 35 668 36 37 669 38 39 40 670 41 42 671 43 44 45 672 46 47 673 48 49 674 50 51 52 675 53 54 676 55 56 57 677 58 59 678 60 61 62 63 64 65

highway. Geotechnical Engineer 6:6-13 (in Chinese) Lo CM, Lee CF, Chou HT, Lin ML (2014) Landslide at Su-Hua Highway 115.9k triggered by Typhoon Megi in Taiwan. Landslides 11(2): 293-304 Lu RX, Chang YM, Lin XJ (2003) The rainfall intensity-duration equation analysis of rain gauge station in Taiwan. Water Resources Agency, MOEA (in Chinese) Peyret M, Djamour Y, Rizza M, Ritz JF, Hurtrez HE, Goudarzi MA, Nankali H, Chéry J, Le Dortz K, Uri F (2008) Monitoring of the large slow Kahrod landslide in Alborz mountain range (Iran) by GPS and SAR interferometry. Engineering Geology 100: 131-141

Ulusay R, Aydan Ö, Kilic R (2007) Geotechnical assessment of the 2005 Kuzulu landslide (Turkey). Engineering Geology 89:112-128 Sassa K, Fukuoka H, Wang FW, Wang GH (2005) Dynamic properties of earthquake-induced large-scale rapid landslides within past landslide masses Landslides 2(2):125-134 Sekiguchi T, Sato HP (2004) Mapping of micro topography using airborne laser scanning. Landslides 1:195-202 Soeters R, van Westen, CJ (1996) Landslides: Investigation and Mitigation. Transportation Research Board Special Report 247:129-177 Takasuke S (2000), “Introduction of topographic map interpretation for construction engineer Vol.3: terrace, hill, and mountain. Kokon Publisher, pp 811-815 (in Japanese) Tarolli P (2014) High-resolution topography for understanding Earth surface processes: Opportunities and challenges. Geomorphology 216:295-312 Yeh EC (1998) The ductile shear deformation and structural evolution of the Penglai Orogency at Suao-Don-Ao area in northeastern Taiwan. National Taiwan University, Masters thesis (in Chinese) Zakšek K, Oštir K, Kokalj Ž (2011) Sky-View Factor as a Relief Visualization Technique. Remote 30

679 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

Sens. 3:398-415

31

Figure

Figure index

1

Fig. 1 Geological settings of study area (115.9K of Provincial Highway No. 9) Fig. 2 Rainfall histogram of landslide disasters: (a) typhoon Megi (2010), (b) typhoon Nalgae (2011), and (c) the 0512 rainfall event (2012) Fig. 3 Time relationship among the magnitude of earthquake, cumulative rainfall, and number of landslide (2008/07/05-2013/01/05). Fig. 4 Comparison of remote sensing images at 115.9 km on Provincial Highway No. 9 between 2004- 2013 (FS-2, SPOT 5) Fig. 5 Three-dimensional terrain map of landslide hazard at on 115.9 km of Provincial Highway No. 9 Fig. 6 Dakeng bridge at 115.9 km of Provincial Highway No. 9 was blocked with debris flow after (a) typhoon Megi and (b) typhoon Nalage Fig. 7 Onsite survey in the upstream landslide area of Dakeng gully catchment Fig. 8 Schematic sketch showing the gravitational deformation of foliated rock Fig. 9 Schematic diagram of geological structure in the downstream landslide area of Dakeng gully catchment Fig. 10 (a) Mapping of rock attitude and (b) joint density in the study area Fig. 11 Relationship between displacement (vertical) of GPS05 and hourly precipitation at DonAo Peak Fig. 12 Relationship between displacement (north-south) of GPS04 and hourly precipitation at Don-Ao Peak Fig. 13 Movement process and speed variations in the simulated landslide at 115.9 km of Provincial Highway No. 9. Fig. 14 Surface profile of main landslide and approximate flow velocity; inset shows the path of travel from the source area to the alluvial fan

1

Fig. 15 Active landslide geomorphic evolution and deep-seated landslide characteristics (revised from Takasuke, 2000) Fig. 16 Results of micro-photographic interpretation of landslide at 115.9 km watershed on Provincial Highway No. 9 2

3

Fig. 1 Geological settings of study area (115.9 km of Provincial Highway No. 9)

4

2

(a) Typhoon Megi(2010) - Donao station 1000


120



140

800

2010/10/19

R

100

600

R total 80

400

60 40

200 20 0

0 18

0

6

12

18

0

6

12

18


160

24

Time [hr]

5

(b) Typhoon Nalgae(2011) - Donao Peak station


120

1000



800

R

100

600

80 60

400

2011/9/30 40

200 20 0

0 0

6

12

18

0

6

12

18


140

0

Time [hr]

6

(c) Rainfall event (2012/05/12) - Donao Peak station

precipitation [mm]



600

R

60

400 2012/05/13 40 2012/05/12 200 20

0

effective cumulative precipitation [mm]

80

0 22 23 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 11

7

Time [hr]

8

Fig. 2 Rainfall histogram of landslide disasters: (a) typhoon Megi (2010), (b) typhoon

9

Nalgae (2011), and (c) the 0512 rainfall event (2012)

10 11 12

3

2008/7/05

Num. of landslides



7 6 5 4 3 2 1 0 1600 1400 1200 1000 800 600 400 200 0 8 6

2009/02/05

2009/09/05

2010/03/05

2010/10/05

2011/05/05

2011/12/05

2012/06/05

2013/01/05

5.81

Earthquake

5.25 5.09

typhoon Megi

Rainfall

typhoon Nalgae

typhoon Parma

Rainfall 512

Landslides

4 2 0 2008/7/05 2008070101

2008070102 2009/02/05

2008070103 2009/09/05

2008070104 2010/03/05

2008070105 2010/10/05

13 14 15

2008070106 2011/05/05

2008070107 2011/12/05

2008070108 2012/06/05

2013/01/05 2008070109

Date

Fig. 3 Time relationship among the magnitude of earthquake, cumulative rainfall, and number

of

landslide

4

(2008/07/05-2013/01/05)

16

17 18

Fig. 4 Comparison of remote sensing images at 115.9 km on Provincial Highway No. 9 between 2004- 2013 (FS-2, SPOT 5)

19

5

20 21 22

Fig. 5 Three-dimensional terrain map of landslide hazard at on 115.9 km of Provincial Highway No. 9

23

6

24 25 26

Fig. 6 Dakeng bridge at 115.9 km of Provincial Highway No. 9 was blocked with debris flow after (a) typhoon Megi and (b) typhoon Nalage

27

7

28 29

Fig. 7 Onsite survey in the upstream landslide area of Dakeng gully catchment

30

8

31 32

Fig. 8 Schematic sketch showing the gravitational deformation of foliated rock

33

9

34 35 36

Fig. 9 Schematic diagram of geological structure in the downstream landslide area of Dakeng gully catchment

37

10

38

(b) (a) 39

Fig. 10 (a) Mapping of rock attitude and (b) joint density in the study area

40

11

41 42 43

Fig. 11 Relationship between displacement (vertical) of GPS05 and hourly precipitation at Don-Ao Peak

44

12

45 46 47

Fig. 12 Relationship between displacement (north-south) of GPS04 and hourly precipitation at Don-Ao Peak

48

13

49

50 51

Fig. 13 Movement process and speed variations in the simulated landslide at 115.9 km of Provincial Highway No. 9

52 53

14

54

55 56

Fig. 14 Surface profile of main landslide and approximate flow velocity; inset shows the path of travel from the source area to the alluvial fan

57 58

15

59 60 61

Fig. 15 Active landslide geomorphic evolution and deep-seated landslide characteristics (revised from Takasuke, 2000)

62 63

16

64

65 66

Fig. 16 Results of micro-photographic interpretation of landslide at 115.9 km watershed on Provincial Highway No. 9

17

Table

Table index

1

Table 1 monthly rainfall statistics from Suao and Don-Ao Peak meteorological stations (from CWB, Taiwan) Table 2 Variations in landslide area on the upslope on 115.9 km of Highway No.9 Table 3 Typhoon scenarios for landslide events at Don-Ao Peak from 2009-2012 Table 4 Irregularities associated with gravitation deformation areas (From Soeters and van Westen, 1996; Chigira, 2014) 2 3

1

4

Table 1 monthly rainfall statistics from Suao and Don-Ao Peak meteorological

5

stations (from CWB, Taiwan) month

1 364.9

2 339.7

3 211.2

4 192.8

5 267.4

6 248.8

687.8

659.3

328.0

215.0

797.0

363.5

7 8 9 10 Suao 178.7 280.7 536.9 719.6 rainfall Don-Ao [mm] 256.8 558.0 412.3 618.8 Peak Suao station： 1971/1~2009/12；Don-Ao Peak station：2012/1~2013/12.

11 696.6

12 422.7

846.0

815.8

rainfall [mm]

Suao Don-Ao Peak

month

6 7

2

8

Table 2 The variation of landslide area at the upslope on 115.9 km of Highway

9

No. 9

event

after Typhoon Parma (2009)

after Typhoon Megi (2010)

landslide area [m2]

5,638

89,574

119,449

123,387

alluvial area [m2]

-

49,040

63,370

65,920

10 11

3

after Typhoon Nalgae (2011)

after rainfall 512 (2012)

12

Table 3 Typhoon scenarios for landslide events at Don-Ao Peak from 2009-2012 typhoon event occurrence time of landslide typhoon track* accompanied effect intensity (mm/hr)** effective rainfall (mm)**

Parma

Megi

Nalgae

0512 rainfall

2009/10/05

2010/10/21

2011/10/02

2012/05/12

special track 110.0 527.5

9 northeast monsoon 121.0 78.5 704.7 645.5

*

32.5 428.5

It refers to the definition of typhoon track which published by Central Weather Bureau. Both rainfall intensity and effective rainfall are calculated form the temporal value of rainfall event at occurrence of landslides. **

13 14

4

15

Table 4 Irregularities associated with gravitation deformation areas (From Soeters and van Westen,

16

1996; Chigira, 2014) Irregularity

Description

Step-like morphology

Step-like morphology usually forms after the landslide body has mobilized. Material upslope of the landslide scar moves downslope into the scar, forming the step like topography at the upper edge of the landslide

Semicircular niches

Semicircular niches form in the head of the landslide.

Back tilting of slope faces

The occurrence of back tilting slope faces indicates rotational failure. This feature is often oval or square in the horizontal surface and generally occurs in poorly drained material.

Hummocky relief

Irregular topography that generally indicates instability and past movement and may be related to shallow failures or small failures moving towards an existing landslide scar.

Formation of cracks

Steeping of slopes

Fresh, open cracks that are evidence of recent movement and indicate the landslide body is sliding or tipping. Often the cracks parallel the edge of a scar. Once the landslide has occurred, an oversteepened slope, or scar remains, severing the head of the landslide.

17

5

Illustration