Recent rainfall-induced landslides and debris flow in
northern Taiwan
Hongey Chen
a,⁎
, Simon Dadson
b, Yi-Guan Chi
ca
Department of Geosciences, National Taiwan University, No. 1, Section 4, Roosevelt Road, Taipei, Taiwan
bDepartment of Earth Sciences, Cambridge University, Downing Street, Cambridge CB2 3EQ, United Kingdom cSyzygy Information Service Company, Taipei, Taiwan
Received 19 February 2004; received in revised form 23 September 2005; accepted 4 January 2006 Available online 7 February 2006
Abstract
This paper discusses the factors initiating landslides and a large-scale debris flow in the Chonho area of Taipei County, northern Taiwan, which resulted from heavy rainfall during Typhoon Xangsane in November 2000. The morphology and triggering mechanism of the landslides and the debris flow were investigated using aerial photographs taken over the past 20 years, field measurement of channel cross-sections, laboratory assessment of slope material properties, and slope stability analysis. A large landslide that occurred in the source area was the initial source of material for the debris flow. The majority of the landslide debris entered the main stream, where it mixed with water and became a debris flow. Eroding the sidewalls of the stream, the debris flow entrained additional material and traveled downstream into Chonho Village. Our findings indicate that the November 2000 landslides and the debris flow were strongly influenced by the physical properties of the underlying pyroclastic substrate and the pattern of bedrock discontinuities within the valleys.
Published by Elsevier B.V.
Keywords: Debris flow; Bedrock discontinuity; Landslide; Volcanic rocks; Taiwan
1. Introduction
Debris flows and landslides may lead to large-scale natural hazards, and may contribute to a large fraction of long-term sediment yields from mountain areas like Taiwan (Dadson et al., 2004). It is widely recognized that hillslope instability can be caused by increased subsurface pore pressures during periods of intense rainfall (Anderson and Sitar, 1995; Iverson et al., 1997), which reduce the shear strength of hillslope materials
(Keefer et al., 1984; Chen et al., 1995). A number of recent studies have demonstrated that rainfall-induced landslides can be transformed into debris flows as they move downslope (e.g.,Fleming et al., 1989; Dai et al., 1999; Montgomery et al., 2000; Marchi et al., 2002; Guzzetti et al., 2004). Landslide-induced debris flows, therefore, present a hazard that is being increasingly recognized, and geological and geomorphological information on such mass movements and their causing or contributing factors form a vital component of natural hazard planning (Rib and Liang, 1978).
Until now, few detailed case studies of landslide-triggered debris flows have been presented in the literature. In this paper, we present a detailed account
⁎ Corresponding author. Fax: +886 2 23636095. E-mail address: hchen@ntu.edu.tw (H. Chen). 0169-555X/$ - see front matter. Published by Elsevier B.V. doi:10.1016/j.geomorph.2006.01.002
of a recent debris flow in Taiwan and describe quantitative morphological and geotechnical observa-tions that reveal the triggering mechanism and geomor-phological evolution of the landslide and debris flow. Taiwan has undergone frequent destructive typhoons throughout its history (Chang, 1996), many of which have caused landslides and debris flows (Chen et al., 1999; Chen and Su, 2001; Hung et al., 2002). During the 1990s, several typhoons caused a number of landslides and debris flows, which led to serious damage. Hazardous landslides and debris flow were triggered by (1) Typhoon Ophelia (June 1990) in eastern Taiwan, (2) Typhoon Herb (August 1996) in central Taiwan, (3) Typhoon Zebert (October 1998) in northern Taiwan, and (4) Typhoon Xangsane (November 2000) also in northern Taiwan (Central Weather Bureau, 2003).
The landslides and debris flow that forms the subject of the present paper occurred on November 1, 2000, when 774 mm of rain fell in a single day in the Chonho area of the Chingshan district, in the eastern part of
Taipei County, northern Taiwan (Fig. 1). The rainfall was particularly intense in northern Taipei, with a maximum hourly precipitation of 70 mm. The Chonho debris flow, which resulted from one of the landslides, traveled over 1.5 km down the narrow Sanho stream, leaving an obvious disruption surface that ranged from 8 to 45 m in width. Finally, debris covered the entire area of Chonho Village, destroyed around thirty houses, and left the village isolated for approximately 30 h.
Before this debris flow, the underlying Pleistocene volcanic bedrock had been considered stable, even under intense precipitation. The main aim of the present study is to discuss factors controlling the location and initiation of the landslides and the debris flow based on aerial photographs taken over the past 20 years, field surveys on geology, field measurement of channel cross-sections, laboratory assessment of slope material properties, and slope stability analysis. Such a combined analysis yields a more robust understanding of the triggering mechanism of landslides and debris flows
than any single approach (Hansen, 1984; Petley et al., 2002).
2. Study area
The Chonho area has a subtropical climate with average annual precipitation of ca. 4000 mm. On average, rainfall occurs on 190 days per year, and approximately four typhoons hit the area per year. Most typhoons occur during the period between July and October. The heaviest rainfall typically occurs in September and October, with a monthly average of 720 mm. For most of the summer, the weather is dry and the temperature ranges from 23 to 30 °C. In winter, the northeasterly monsoons bring in frontal depressions with intermittent rainfall and temperatures ranging between 8 and 16 °C. In December and January, the monthly rainfall is about 365 mm. The precipitation in the study area comes mostly from typhoons in early autumn and northeasterly monsoons in winter.
The catchment studied has a drainage area of about 1.4 km2. Its typical hillslope gradients are between 20°
and 45°. Chonho Village is located at an elevation of 50 m a.s.l. and on an alluvial deposit which covers an area of 0.3 km2. The surface gradient of the alluvial deposit ranges from 2° to 8°. The Sanho stream above Chonho Village has an average gradient of 12°.
The headwaters of the study area are located in warm temperate rain forest. The upper 60% of the catchment is covered with lauraceae forest composed of incense machilus, camphor trees and large-leaved machilus. Approximately 25% of the catchment with elevations between 110 and 180 m is covered by bamboo and bluestem. The downstream reaches below 110 m in elevation are located in farmland. The narrow zone along the main channel consists of bare rock slopes.
The northern and western parts of the study area are underlain by the Tatun volcanic series composed of lava flow, volcanic ash and pyroclastic bedrock (Fig. 2), which are thought to have erupted from Tatun volcano at 2.5–2.8 Myr BP (Chen and Wu, 1971). The Tatun volcanic series has been divided into fifteen lava flow units and two tuffaceous conglomerate units (Ho, 1986). The tuffaceous conglomerate consists of andesite
fragments with unconsolidated matrices. This weakly massive and intact conglomerate was subjected to loose cementation. This volcanic material unconformably overlies Miocene sedimentary rocks.
The Sanho stream of the study area drains a region with about 2 m sediment yields and has a braided pattern along its downstream half. The upper catchment of the stream contains the contact between the lava flow and tuffaceous conglomerate (Fig. 2), and the largest landslide due to Typhoon Xangsane in November 2000 is located on this contact (Fig. 3). About 50% of bedrock outcrops in the catchment are located along the sidewalls of the stream; the remainders of the outcrops are associated with landslides and gullies on hillslopes. Another bedrock unconformity appears in the middle of the study area where the tuffaceous conglomerate is in contact with the muddy sandstone of the Miocene Kueichulin Formation (Fig. 2). Most of the debris
transported by the Sanho stream has been supplied from the upper part of the study area.
Typhoon Xangsane passed through Taiwan from the southeast to the northwest between October 30 and November 1, 2000. This typhoon caused two periods of intense precipitation in the study area: (1) 90 mm of rainfall between 5 AM on October 30 and 3 AM on October 31, and (2) 685 mm of rainfall between 11 AM on October 31 and 2 AM on November 2 (Fig. 4). The debris flow occurred late in the night of November 1. The length of the debris flow is 1.75 km; its initiation point (Fig. 3A) is located on a hillslope at an elevation of 425 m, where the ground slope is approximately 45°. 3. Recent change in watershed geomorphology
In this work, panchromatic black-and-white aerial photographs taken in 1979, 1986, 1994 and 2001 were
used to observe topographical changes in the stream beds and mountain slopes over a twenty-year period before and after the debris flow occurred. The scale of the photographs is 1 : 17,000.
In 1979, sixteen landslides were visible in the photograph, only three of which (L06, L10 and L13; L stands for landslide) were near the main stream (Fig. 5A). These landslides do not appear to be fresh and their material appears to have remained on hillslopes and did not flow into the main stream or major tributaries. The main stream looks narrow and appeared to contain only a little debris. The drainage system appeared to be stable as a whole.
The aerial photographs indicate little topographic change between 1979 and 1994. Some old landslides (L03, L04, L05, L10 and L16) were not visible any more as a result of plant recovery (Fig. 5B). Sediment along the main channel and tributaries was not abundant.
The aerial photograph taken after the passage of Typhoon Xangsane shows 11 new landslides (L17 to L27) including the large landslide in the upper part of the main stream (L22), which was the initial source of the debris flow (Fig. 6A). The run-out deposit of the landslide was about 120 m long and 20 m wide.
The majority of the new landslides were located at channel heads. Some of the new landslides (L23, L26 and L27) formed on the sidewalls of the main stream and some of the resultant debris entered the stream. The debris was partly removed by surface water during the
typhoon. Some northern tributaries also transported large amounts of debris supplied from the new or reactivated landslides (L09, L15, L24 and L25).
The sidewalls of the trunk stream without landslides were also eroded to some extent as a result of typhoon-induced flooding. The number of non-vegetated gullies increased and some gullies were extended. Some farmlands were covered with debris and some paved roads were destroyed.
In summary, the study area during the few decades before Typhoon Xangsane on November 1, 2000 was characterized by very limited geomorphological changes. However, Typhoon Xangsane suddenly trig-gered rapid erosion with landslides and gullying, yielding a large amount of sediment.
4. Field surveys of geology and topographic change Detailed field observations of geology and geomor-phology in relation to the November 2000 landslides and debris flow were performed. Fig. 6B shows geomorphological characteristics of the study area soon after the hazard. Symbols G and A stand for gullies and artificial features such as constructions, public roads and farmland.
The area affected by the landslides and debris flow can be divided into three sub-areas (Fig. 7A). In the source area, landslides resulted in a series of scars. In the flow area, material from channel sidewalls had first slid into the valley bottom, creating many small lateral lobes,
Fig. 4. Distribution of hourly and accumulated precipitation between Oct. 30 and Nov. 2, 2000. The debris flow in the Chonho area occurred after a severe rainstorm of 370 mm.
and they were subsequently removed by the debris flow. The depositional area was located downstream, along the north part of the Sanho school. In the following three
subsections, geomorphological and geological descrip-tions of the three areas will be given based on the results of our field surveys.
Fig. 5. (A) Aerial photograph of the study area taken in 1979. Most old landslides are away from streams. (B) Aerial photograph taken in 1994 showing few topographic changes from 1979.
Fig. 6. (A) Aerial photograph taken in 2001 showing devastation caused by Typhoon Xangsane. A large landslide appeared in the upper part of the main stream (L22). (B) Geomorphological map of the study area based on the aerial photograph of (A) and field surveys along the stream banks and trails.
Fig. 7. (A) Location of 71 cross-sections perpendicular to the stream with a 25-m interval. The sections were used to calculate the volume of erosion and deposition due mainly to the debris flow. (B) Locations of open pits and bedrocks where samples for laboratory tests were taken.
4.1. Source area
The source area of the debris flow ranges from 300 to 425 m in elevation and is about 200 m in length. The scar of the landslide that caused the debris flow is a spoon-shaped depression (Fig. 3A), on which almost no debris has remained. The landslide was around 10 m deep, 30 m wide and 95 m long. The angle of the landslide gradually shifts from 40° at the top to 20° at the bottom.
As noted before, the unconformity between the light grey lava (Fig. 3B) and the tuffaceous conglomerate (Fig. 3C) occurs in the landslide scar. The orientation of the unconformity is N70°E/15°S. The lava flow displays rhyolitic textures with major joint directions of N6°E/ 15°S, N71°E/75°S, and N75°E/82°N.
4.2. Flow area
The flow area ranges from 110 to 300 m in elevation and is 1100 m in length. The upper part of the flow area consists mainly of tuffaceous conglomerate, which is composed mostly of andesite rock fragments and a dark reddish matrix of fine sand, while the lower part below ca. 180 m a.s.l. consists of muddy sandstone.
Some debris flow material was deposited along the stream. The deposited material is composed of various loose rock fragments, sand, and mud. The observed andesite fragments range from 10 to 30 cm in diameter and are round to subangular in shape.
4.3. Depositional area
The depositional area ranges from 50 to 110 m in elevation and 450 m in length. The orientation of partly exposed bedrocks bedding ranges from N35° to N50° and it dips 52° towards the north. Two sets of observed joints are N85°E/60°S and N10°E/63°N, providing the potential geometry of wedge and plane failure models (Hoek and Brown, 1997). Most of the deposited debris flow materials are andesite boulders and tuffaceous debris. The boulders are subround to round in shape, and the debris is subangular to subround in shape.
4.4. Cross-section surveys
River cross-sections were surveyed after the debris flow in 2001, and they were compared with those surveyed in 1994, using topographic maps compiled by the Taiwan Agricultural and Forestry Aerial Survey Bureau (TAFASB) with a contour interval of 5 m. We
selected the maps of 1994 to represent the situation soon before the debris flow, because the TAFASB and local experts advised that there had been little topographic change in the study area between 1994 and 2000. In each survey, 71 cross-sections perpendicular to the channel were measured with a 25-m interval, from the source area at 450 m in elevation to the deposition area at 50 m in elevation (Fig. 7A). The total volume of deposition between 1994 and 2001 was 105,970 m3and that of erosion was 122,720 m3, showing that in total nearly 16,750 m3of sediment had left the stream (Table 1). However, not only the depositional area but also the flow area was characterized by net deposition.
5. Geotechnical investigation
Geotechnical data about the debris flow and the source area were also collected in the field. The distribution of bedrock discontinuities (e.g., joints and bedding planes) was investigated in details, and then the physical properties of the bedrock material were analyzed in the laboratory.
5.1. Physical properties of sediment
Sediment samples for particle size analysis were taken from twelve open pits in the source, flow and depositional areas of the debris flow (Fig. 7B). Two pits were dug in the source area, five pits in the flow area, and five pits in the depositional area. Each pit was a cube with a side length of 1 m. Most of the observed soil type was GP (poorly graded gravel) according to the Unified Soil Classification System.
The majority of material deposited in the source and flow areas was non-plastic silty gravel, whereas plastic silt formed the major component of sediment in the depositional area (Fig. 8). Overall, the fine silt and clay fractions of the deposits comprise less than 10% of the sample, and the plasticity index ranges from 2.28 to 8.40 (Table 2). More than 80% of sediment particles have diameters larger than 2 mm.
Table 1
Volume of erosion and deposition in different sections of the study area Source area Flow area Depositional area Total Erosion (m3) 68,890 53,410 420 122,720 Deposition (m3) 750 67,900 37,320 105,970 Deposition–Erosion (m3) −68,140 14,490 36,900 −16,750
5.2. Physical properties of bedrock
Bedrock samples were collected from fifteen out-crops along the stream sidewalls and tests were undertaken to determine the physical properties and shear strength using the methods of the International Society of Rock Mechanics (Brown, 1981). The rock samples R01 to R06 are sandstones; R07 to R13 are tuffaceous conglomerates; R14 to R15 are lavas.Table 3
shows that the saturation of tuffaceous conglomerate exceeds 40%, and its porosity is less than 10%.
Experimental results also show that the friction angle of the muddy sandstone ranges from 30° to 36° in dry conditions (Table 3). The samples of tuffaceous conglomerate have friction angles ranging from 32° to
39° in dry conditions and 28° to 33° in saturated conditions. Friction angles for the lava samples range from 34° to 49° in dry conditions and 23° to 38° when saturated.
5.3. Bedrock discontinuity analysis
Bedrock discontinuity analysis was undertaken for some rock slopes on both sides of the trunk stream based on a semi-quantitative set of parameters (Selby, 1987; Priest, 1992). The geological orientations of bedding, joints and diverse slope surfaces were measured at 15 locations (Ocp01–Ocp15) where the rock samples were collected, and were plotted on a stereonet to assess the likelihood of slope failure. The
Fig. 8. Particle size distributions of river sediment from different locations.
Table 2
Physical properties of river sediment Pit no #10 passing (%) #200 passing (%) D50 (mm) Cu (mm) Wn (%) γint (t/m3) γsat (t/m3) e n (%) Sat (%) LL (%) PL (%) PI Cf Remarks
Up01 27.54 2.32 55 450 22.45 1.44 1.62 0.78 43.91 77.87 – – NP GP Source area
Up02 23.26 1.60 90 478 21.96 1.38 1.57 0.77 43.36 78.68 – – NP GP Mp01 20.82 2.95 43 412 13.91 1.75 1.93 0.66 39.69 56.87 38.14 35.86 2.28 GP Flow area Mp02 17.87 1.32 51 192 17.1 1.76 1.93 0.76 43.16 59.39 40.57 35.62 4.95 GP Mp03 19.07 1.67 83 367 10.44 1.81 2.02 0.61 37.86 46.34 – – NP GP Mp04 13.28 1.51 60 123 9.26 1.75 1.94 0.51 33.64 50.59 36.43 29.67 6.76 GP Mp05 14.91 0.95 61 126 8.74 1.85 2.03 0.48 32.47 49.56 – – NP GP Dp01 17.40 1.47 38 375 11.49 1.83 2.03 0.63 38.52 49.83 41.12 32.72 8.40 GP Depositional area Dp02 19.93 1.90 80 440 16.55 1.75 1.92 0.71 41.57 62.20 33.88 27.12 6.76 GP Dp03 15.60 0.76 68 148 18.79 1.78 1.93 0.76 43.19 67.02 37.91 32.27 5.25 GP Dp04 11.65 1.03 58 73 10.68 1.85 2.08 0.68 40.59 42.78 35.69 31.97 3.27 GP Dp05 16.78 1.46 50 115 11.5 1.87 2.10 0.73 42.36 43.16 35.46 30.15 5.31 GP
D50: soil diameter at 50% weight is finer, Cu: coefficient of uniformity, Wn: water content,γint: in situ unit weight,γsat: saturated unit weight, e: void
ratio, n: porosity, Sat: degree of saturation, LL: liquid limit, PL: plastic limit, NP: non-plasticity, PI: plasiticity index, Cf: classification of soil type, GP: poorly graded gravel.
rock types on sites Ocp01–Ocp06 are mainly composed of muddy sandstone, while those on Ocp07–Ocp13 consist of tuffaceous conglomerate, and Ocp14 and Ocp15 compose of the lava flow. These three groups also correspond to the depositional, flow and source areas, respectively.
The orientation of the bedding plane and the dip angle were measured at each outcrop. The constructed stereonets (Fig. 9) indicate that Ocp01, Ocp11 and Ocp13 have a potential of plane failure; Ocp01 and Ocp15 have a possibility of wedge failure; and Ocp02, Ocp04 and Ocp09 have a potential of topple failure.
5.4. Safety factor analysis
The safety factor represents the average ability of a slope profile to resist gravitational forces and other external forces. The safety factors for the three types of failure (Hoek and Bray, 1981) were computed for the bedrock outcrops along the stream.
For plane slope failure:
F¼c l þ ðWcoswq−U−VsinwqÞ tan/q W sinwqþ Vcoswq
ð1Þ where F is the safety factor, c is the cohesion (kN/m2), ψρis the failure angle (degrees), l is the length of the slip
surface (m), W is the weight of the failure mass (kN/m),
U is the uplift force (kN/m), V is the lateral water pressure (kN/m), andϕ is the friction angle (degrees).
For wedge slope failure: F ¼ðRaþ RbÞtan/ W sinui ¼ sinb tan/ sin1 2n tanui ð2Þ where Raand Rbare the normal stress on the two masses
a and b (kN/m) which correspond to two planar discontinuities in the wedge failure,ξ is the intersection angle between the masses a and b (degrees), φ1is the
dip angle of the line of intersection between the two discontinuities (degrees), and β is the tilt angle of the wedge (degrees).
For topple slope failure: F ¼ tan/
tan½a−ð90−dÞ ð3Þ
whereα is the dip angle of the slope surface (degrees), and δ is the dip angle of the bedrock discontinuity (degrees).
Stability analysis using the plane failure model under normal dry conditions at Ocp01, Ocp11 and Ocp13 reveals a safety factor of 1.07, 1.20, and 1.16, respectively, indicating high stability. When these slopes are saturated, however, the safety factor falls below 1.0, which indicates that the slopes will fail (Table 4). The application of the wedge failure model to Ocp01 and Ocp15 yields safety factors of 1.06 and 1.41 under
Table 3
Physical properties of rock samples from the study area Sample location Rock type Wn (%) γint (t/m3) γ sat (t/m3) e n (%) Sat (%)
Dry condition Saturated condition c (kg/cm2) ϕ (°) c (kg/cm2) ϕ (°) R01 Muddy sandstone 3.86 2.33 2.47 0.16 14.11 54.80 0.1 34.2 – – R02 Muddy sandstone 4.15 2.41 2.55 0.15 13.31 65.12 0.1 34.3 – – R03 Muddy sandstone 3.32 2.41 2.54 0.15 13.32 52.09 0.1 35.3 – – R04 Muddy sandstone 12.50 1.92 2.21 0.41 29.17 57.64 0.2 30.2 – – R05 Muddy sandstone 11.79 1.95 2.32 0.57 36.44 39.82 0 32.6 – – R06 Muddy sandstone 5.24 2.29 2.57 0.38 27.61 31.50 0 36.4 – – R07 Tuffaceous conglomerate 1.15 2.62 2.68 0.07 6.18 45.57 0 35.5 0.2 31.5 R08 Tuffaceous conglomerate 2.16 2.32 2.41 0.09 8.67 52.70 0.1 31.5 0.2 28.2 R09 Tuffaceous conglomerate 2.55 2.74 2.83 0.10 9.41 67.42 0 36.5 0.5 33.0 R10 Tuffaceous conglomerate 1.53 2.62 2.71 0.10 9.15 39.75 0 35.0 0.5 30.5 R11 Tuffaceous conglomerate 1.09 2.74 2.78 0.05 4.45 64.44 0.1 39.0 0.7 32.5 R12 Tuffaceous conglomerate 1.10 2.73 2.78 0.05 4.75 60.18 0 38.0 0.3 31.8 R13 Tuffaceous conglomerate 2.61 2.68 2.77 0.09 8.41 76.25 0.2 38.3 0.7 31.0 R14 Lava 0.36 2.74 2.78 0.04 3.94 24.39 0 48.2 0.3 37.1 R15 Lava 0.36 2.75 2.78 0.04 3.67 26.26 0.2 37.1 0.2 32.5
“–” collapse after water emerging, Wn: water content,γint: in situ unit weight,γsat: saturated unit weight, e: void ratio, n: porosity, Sat: degree of
normal dry conditions, but the factor falls to 0.74 and 0.89 under saturated conditions. The application of the topple failure model to Ocp02, Ocp04 and Ocp09 yields safety factors of 1.09, 1.10 and 1.07 under normal dry conditions. When these slopes are saturated, the safety factor falls below 1.0. For the other cases, slopes were found to be stable even under saturated conditions.
6. Discussion
Our investigation indicates that the landslides and debris flow due to Typhoon Xangsane occurred during rainfall not only of high intensity (up to 70 mm/h) but also with an exceptionally long period of over 10 h (Fig. 4). Prolonged rainfall caused a saturated zone to
Fig. 9. Examples of stereonet analysis showing three types of potential unstable zones, illustrated by Ocp01, Ocp04 and Ocp15 in dry and saturated conditions.
develop, with elevated pore-water pressures in the substrate, and contributed directly to the occurrence of the landslide. Such a failure had not been expected at this location because the area is underlain by strong volcanic rocks. The relatively low strength of the tuffaceous conglomerate located below the lava flow (Fig. 3A) seems to have contributed to the large landslide that induced the debris flow.
Moreover, the geometry of bedrock discontinuities was found to exert an important control on the location of the landslides. Despite the high intact rock strength having friction angle more than 30°, the landslides occurred in areas where the entire rock mass had been weakened through the infiltration of water into regularly spaced bedrock discontinuities. Indeed, we have calculated that the safety factor for some saturated rock slopes fell below 1.0.
The debris flow appears to have occurred through multiple processes. Runoff within the catchment was mixed with sediment from the large landslides, transforming them into the debris flow. This mechanism has been suggested elsewhere by Montgomery et al. (2000)andMarchi et al. (2002). Subsequently, the size of the debris flow was probably augmented by sediment supply from the collapse of the sidewalls along the debris flow channel. However, the sediment budget estimation inTable 1 suggests that the majority of the sediment deposited in the lowermost reach came from the upper reach rather than the middle reach. This suggests that not a set of small landslides but a
large-scale landslide is often needed to induce a catastrophic debris flow.
We also note that prior to the debris flow, there was only a small amount of sediment on the bed of the Sanho stream. This fact may have led any engineering assessment to conclude that the likelihood of a debris flow in the channel was low, since an insufficient amount of debris had accumulated. However, the single large landslide induced the debris flow that suddenly changed the river condition. Our study has revealed that, in order to quantify the risk and likely severity of mass movements in steep mountain regions, it is necessary to consider the stability of possible initial source regions and to assess the potential with which mass failures of one type (e.g., landslides) can be converted into those of another (e.g., debris flows).
7. Conclusions
The November 2000 landslides in the study area were triggered by a severe rainstorm of 370 mm in less than one day. Intense rainfall caused groundwater to infiltrate through the volcanic bedrock material along the fractures in the exfoliated zones. Consequently, the tuffaceous conglomerate and the weathered lava became saturated with water in some places and slopes there became unstable. The largest landslide in the upper reach mobilized debris as a landslide mass and it flowed down with water along the stream channel as a debris flow.
Table 4
Examples of safety factor analysis for slopes in the study area
Slope γt(t/m3) c (kg/cm2) ϕ (°) Geometric model Safety factor Condition Remark
Ocp01 (1) 2.42 0.1 34.2 Plane 1.07 Dry Critical stable
0.96 Saturated Fail (3.3 ma)
Ocp01 (2) 2.42 0.1 34.2 Wedge 1.06 Dry Critical stable
0.74 Saturated Fail (3.1 ma)
Ocp02 2.51 0.1 34.3 Topple 1.17 Dry Stable
0.98 Saturated Fail (3.1 ma)
Ocp04 2.16 0.2 30.2 Topple 1.10 Dry Stable
0.94 Saturated Fail (2.5 ma)
Ocp09 2.81 0 36.5 Topple 1.19 Dry Stable
0.98 Saturated Fail (2.5 ma)
Ocp11 2.77 0.1 39.0 Plane 1.20 Dry Stable
0.95 Saturated Fail (2.8 ma)
Ocp13 2.75 0.2 38.3 Plane 1.18 Dry Stable
0.98 Saturated Fail (8.3 ma)
Ocp14 2.74 0 48.2 Wedge 2.18 Dry Stable
1.47 Saturated Stable
Ocp15 2.76 0.2 37.1 Wedge 1.41 Dry Stable
0.89 Saturated Fail (14.7 ma)
a
The rapidly enlarging debris flow eroded the stream sidewalls but at the same time left some sediment on the riverbed, until it stopped in the downstream area. The accumulated deposits, in addition to those dislodged from valley sides during the storms, provided a source for future debris flows triggered by rainfall.
Acknowledgements
This work was support by the National Science Council, ROC. The authors would like to thank Prof. M. L. Lin, Dr. C.S. Wu and Mr. D.C. Lai for their critical suggestion, Prof. K. Thompson for revising the article and Ms. C.M. Yu for redrawing the diagram. Two reviewers, Prof. M. Soldati and Prof. G.R. Whittecar, provided constructive comments are gratefully appreci-ated. We thank Dr. T. Oguchi for many helpful suggestions which have improved the clarity of the paper.
References
Anderson, S.A., Sitar, N., 1995. Analysis of rainfall-induced debris flows. J. Geotech. Eng. 121, 544–552.
Brown, E.T., 1981. Rock Characterization, Testing, and Monitoring: ISRM Suggested Methods. Pergamon Press, Oxford. 211 pp. Central Weather Bureau, 2003. Historical typhoon data:http://www.
cwb.gov.tw/V4/typhoon/history/typhoon-history.htm.
Chang, J.C., 1996. Natural hazards in Taiwan. GeoJournal 38, 251–257.
Chen, H., Su, D.I., 2001. Geological factors for hazardous debris flows in Hoser, central Taiwan. Environ. Geol. 40, 1114–1124. Chen, C.H., Wu, Y.J., 1971. Volcanic geology of the Tatun geothermal
area, northern Taiwan. Proc. Geol. Soc. China 14, 5–20. Chen, R.H., Lin, M.L., Chen, H., 1995. Mechanism of initiation of
debris flow. In: Cheng, F.Y., Shew, F.S. (Eds.), Urban Disaster Mitigation. Elesevier, Oxford, pp. 231–243.
Chen, H., Chen, R.H., Lin, M.L., 1999. Initiation of the Tungmen debris flow, Eastern Taiwan. Environ. Eng. Geosci. 5, 459–473. Dadson, S.J., Hovius, N., Chen, H., Dade, B.W., Lin, J.C., Hsu, M.L.,
Lin, C.W., Horng, M.J., Chen, T.C., Milliman, J., Stark, C.P., 2004. Earthquake-triggered increase in sediment delivery from an active mountain belt. Geology 32, 733–736.
Dai, F., Lee, C.F., Wang, S., 1999. Analysis of rainstorm-induced slide-debris flows on natural terrain of Lantau Island, Hong Kong. Eng. Geol. 51, 279–290.
Fleming, R.W., Ellen, S.D., Algus, M.A., 1989. Transformation of dilative and contractive landslide debris into debris flows an example from Marin County, California. Eng. Geol. 27, 201–223. Guzzetti, F., Cardinali, M., Reichenbach, P., Cipolla, F., Sebastiani, C., Galli, M., Salvati, P., 2004. Landslides triggered by the 23 November 2000 rainfall event in the Imperia Province, Western Liguria, Italy. Eng. Geol. 73, 229–245.
Hansen, A., 1984. Landslide hazard analysis. In: Brunsden, D., Prior, D.B. (Eds.), Slope Instability. John Wiley & Sons, New York, pp. 523–602.
Ho, C.S., 1986. An Introduction to the Geology of Taiwan: Explanatory Text of the Geologic Map of Taiwan. The Ministry of Economic Affairs, Republic of China. 164 pp.
Hoek, E., Bray, J., 1981. Rock Slope Engineering. The Institution of Mining and Metallurgy, London, pp. 150–198.
Hoek, E., Brown, E.T., 1997. Practical estimates of rock mass strength. J. Rock Mech. Mining Sci. 34, 1165–1186.
Hung, J.J., Lee, C.C., Lin, M.L., 2002. Tsao-Ling rockslides, Taiwan. In: Evans, S.G., DeGraff, J.V. (Eds.), Catastrophic Landslides: Effects, Occurrence, and Mechanisms. Colorado Geol. Soc. Am. Rev. Eng. Geology, vol. XV, pp. 91–115.
Iverson, R.M., Reid, M.E., LaHusen, R.G., 1997. Debris flow mobilization from landslides. Annu. Rev. Earth Planet. Sci. 25, 85–138.
Keefer, D.K., Wilson, R.C., Mark, R.K., Brabb, E.E., Brown III, W. M., Ellen, S.D., Harp, E.L., Wieczorek, G.F., Alger, C.S, Zatkin, R. S., 1984. Real-time landslide warning during heavy rainfall. Science 238, 921–925.
Marchi, L., Arattano, M., Deganutti, A.M., 2002. Ten years of debris-flow monitoring in the Moscardo Torrent (Italian Alps). Geomor-phology 46, 1–17.
Montgomery, D.R., Schmidt, K.M., Greenberg, H.M., Dietrich, W.E., 2000. Forest clearing and regional landsliding. Geology 28, 311–314.
Petley, D.N., Bulmer, M.H., Murphy, W., 2002. Patterns of movement in rotational and translational landslides. Geology 30, 719–722. Priest, S.D., 1992. Discontinuity Analysis for Rock Engineering.
Chapman Hall, London. 420 pp.
Rib, H.T., Liang, T., 1978. Recognition and identification. In: Schuster, R.L., Krizek, R.L. (Eds.), Landslides: Analysis and Control. Spec. Rep., vol. 176. National Academy of Sciences, Washington, D.C, pp. 34–80.
Selby, M.J., 1987. Rock slopes. In: Anderson, M.G., Richards, K.S. (Eds.), Slope Stability. John Wiley & Sons, New York, pp. 475–504.
