Geomorphology 76 (2006) 92 – 108 www.elsevier.com/locate/geomorph
Mechanisms of shallow landslides on soil-mantled hillslopes with permeable and impermeable bedrocks in the Boso Peninsula, Japan Yuki Matsushi *, Tsuyoshi Hattanji, Yukinori Matsukura Graduate School of Life and Environmental Sciences, University of Tsukuba, Tenno-dai 1-1-1, Tsukuba, Ibaraki 305-8572, Japan Received 5 July 2005; received in revised form 19 October 2005; accepted 20 October 2005 Available online 13 December 2005
Abstract Rainfall-induced shallow landslides play a vital role in hillslope denudation processes in humid temperate regions. This study demonstrates the contrasting mechanisms of landslides in adjoining hills with permeable (sandstone) and impermeable (mudstone) bedrock in the Boso Peninsula, Japan. The characteristics of slope hydrology were inferred from pressure-head monitoring and rainfall–runoff observations. An analysis of slope stability provided critical conditions for several previously occurring landslides. The results are as follows. (1) In slopes with the permeable sandstone, infiltrated rainwater percolates through the bedrock as an unsaturated gravitational flow. The wetting front migration results in a decrease of soil cohesion and causes landsliding at the steep lower part of the hillslopes. (2) In contrast, the impermeable mudstone beneath a thin soil layer causes a transient positive pressure head that generates a saturated subsurface storm flow. The reduction in effective normal stress triggers shallow soil slipping at the uppermost part of a hollow. D 2005 Elsevier B.V. All rights reserved. Keywords: Landslides; Hillslope hydrology; Slope stability; Permeability; Rock control
1. Introduction Rainfall-induced shallow landslides frequently occur on soil-mantled steep hillslopes in humid temperate regions as a dominant denudation process. Repetition of such landslides deform hillslopes and in turn drainage basins. Several studies have established that hydrological processes affect the landslide initiation on hillslopes (Onda, 1992, 1994; Van Asch et al., 1996; Terlien, 1997; Onda et al., 2004). These studies emphasized the importance of a comprehensive understanding of the relation between subsurface water behavior and slope destabilization mechanisms. * Corresponding author. Tel.: +81 29 853 5691; fax: +81 29 853 6879. E-mail address:
[email protected] (Y. Matsushi). 0169-555X/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.geomorph.2005.10.003
Physically motivated models were developed in the late 1980s and 1990s to explain landslide initiation and distribution in mountain drainage basins. The models coupled an infinite slope stability analysis with the concept of steady-state saturated throughflow (Montgomery and Dietrich, 1989, 1994; Wu and Sidle, 1995). These steady-state models achieved success in assessing topographic control on landslide susceptibility (Iida, 1999; Borga et al., 2002). However, two issues remain unsolved regarding steadystate models. One is a time scale discrepancy in the supposed hydrological process. The concept of steady groundwater flow parallel to the slope above an impermeable bed can predict only the long-term distribution of groundwater pressure, which should be identified as a predisposition to landsliding. The second concern is that the model cannot apply to hill-
Y. Matsushi et al. / Geomorphology 76 (2006) 92–108
slopes underlain by highly permeable bedrocks, where the near-surface lateral water movement becomes an unfeasible proposition. The aforementioned shortcomings prevent steady-state models from providing a rationale for simultaneous landslide occurrence at a rainfall peak associated with intense rainwater infiltration and transient pore-pressure generation (Iverson, 2000; D’Odorico et al., 2005). Recent slope stability analyses have examined such a realistic hydrological response and proved rain infiltration and subsequent redistribution of groundwater pressure are sufficiently competent to trigger landslides (Gasmo et al., 2000; Cho and Lee, 2001; Collins and Znidarcic, 2004; Kim et al., 2004). However, despite these improved models and simulations there are still few quantitative studies of hillslope hydrological processes and landslide mechanisms based on field observations. We cannot expect more pertinent discussion about the evolution of landscapes in steep terrains unless the actual slope hydrology is linked with the critical condition for landsliding. Certain field evidence is needed to understand how subsurface-water dynamics influence the critical condition for landsliding.
93
The present study focuses on slope hydrological processes and landslide mechanisms in two hilly locations with contrasting morphology and bedrock. We measured geotechnical and hydraulic properties of slope materials. Pressure head fluctuations in hillslopes as well as rainfall–runoff responses in small watersheds were also monitored. Slope stability analysis revealed critical conditions for several past landslides. The results indicate different triggering mechanisms induce landsliding on hillslopes with disparate lithology. This study also postulates that the difference in the landslide mechanisms account for varying hill morphologies in this region. 2. Study area 2.1. Tectonic, climatic and geologic settings The study area is in the south-west Boso Peninsula, central Japan. The area is located on the edge of the North American plate, bordered on the south by the Philippine Sea plate (Fig. 1). The area has been uplifted rapidly at a rate of N 1 m/kyr throughout the Quaternary (Kaizuka et al., 2000).
Fig. 1. Location of the study area and shaded relief map of the hilly lands studied. Broken lines indicate geological boundaries.
94
Y. Matsushi et al. / Geomorphology 76 (2006) 92–108
The study area has a humid temperate climate with a mean temperature of 158C and annual rainfall of 1500– 2000 mm (1490 mm in Kisarazu; 1760 mm in Sakuma; and 2050 mm in Sakahata; the locations of these climatological stations are shown in Fig. 1). Seasonal fronts in early summer and late fall as well as occasional typhoons contribute to 40–50% of the total rainfall. The most intensive 1-day rainfall over the last 30 years is 296 mm in Kisarazu, 426 mm in Sakuma, and 364 mm in Sakahata. Brown forest soil covers the hillslopes, except for narrow ridges or steep slip scars with bedrock exposures. The majority of the study area is covered with a planted forest of cypress (Chamaecyparis obtusa) and cedar (Cryptomeria japonica), although hardwood and various understory species also coexist within the conifer stands. The forest age varies from several years to several decades. No large-scale timber harvesting has been conducted in recent decades. Bedrock in this area belongs to the middle Pleistocene forearc basin fill, referred to as the Kazusa Group (Ito, 1995, 1998). Submarine fan successions generated by glacioeustasy cycles under the influence of paleoocean currents produced repeated sandy and muddy depositional sequences (Ito, 1998; Ito and Horikawa, 2000). The emergence of these submarine deposits in the last 500 ka has led to the development of small hills ranging in elevation from 100–300 m.
2.2. Topography The study site exhibits distinct topographic characteristics that vary from hills in the northwest to hills in the southeast. The central dashed line in Fig. 1 demarcates this boundary between the varying morphometric characteristics. Hills in the northwestern section exhibit relatively high rounded crests (relative relief of 150–200 m) with low drainage density (5–8 km 1). The southeastern hills display low rugged ridges (relative relief of 50–100 m) with high drainage density (15–22 km 1). The topographic difference corresponds to the geology of the area (Fig. 2). Coarse sandstone and conglomerates (the Ichijuku and Nagahama Formations; ca. 600–700 ka) comprise the northwest high terrain, whereas muddy sandstone and sandy mudstone (the Iwasaka and Awakura Formations; ca. 700–800 ka) make up the southeast lower terrain. Overall, this landscape comprises cuesta-like landforms with backslopes facing northwest. Hereafter, we call these two areas the sandstone area and the mudstone area (Figs. 1 and 2). 2.3. Shallow landslide Rainfall-induced shallow landslides in the two areas are of particular interest to this study. Torrential rainfall
Fig. 2. Geologic map and topographic cross-sections of the hilly lands studied. Nh: Nagahama Formation; Ij: Ichijuku Formation; Is: Iwasaka Formation; Ak: Awakura Formation. The geological map corresponds to the area of Fig. 1.
Y. Matsushi et al. / Geomorphology 76 (2006) 92–108
95
Fig. 3. Landslide density for the past two decades in the hills studied. Bar charts represent number of newly emerged landslide scars on vertical air photographs taken in 1981, 1984, 1990, 1995 and 2000. The area underwent a torrential rainfall on August 1, 1989 (total rainfall N300 mm).
on August 1, 1989 (total rainfall N 300 mm) produced many slides along the mudstone slopes, but only a few landslides occurred in the sandstone area (Furuya and Ohkura, 1992). Landslides occurred only on steep lower hillslopes adjacent to major valleys in the sandstone area. While the landslides in the mudstone area took place mainly on uppermost hollows near slope crests (Matsushi and Matsukura, 2004).
The number of the landslide scars was identified from vertical air photographs with various scales from 1 : 20 000 to 1 : 40 000 for the following years: 1981, 1984, 1990, 1995 and 2000 (Fig. 3). The landslide density over the past two decades in the mudstone area (127.6 km 2) is about 22 times larger than that in the sandstone area (5.7 km 2). The adjoining hills have no great variation in climatic
Fig. 4. Locations of the selected slopes including landslide scars (S-1, 2, 3 for the sandstone area (A) and M-1, 2, 3 for the mudstone area (C)), and detailed topographic maps of the experimental watersheds (S-watershed for the sandstone (B) and M-watershed for the mudstone (D)). See Figs. 1 and 2 for locations. Contours are in meters.
96
Y. Matsushi et al. / Geomorphology 76 (2006) 92–108
conditions, vegetation cover, and tectonic settings. Therefore, the difference in the density and locations of the landslides must reflect hydrological processes and critical limits for landsliding, controlled by the hydraulic and mechanical properties of the hillslope materials. This paper focuses on the validation of this hypothesis. 3. Methods 3.1. Selection of slopes and watersheds Field and laboratory experiments were designed to understand hillslope hydrological processes and obtain parameter values relevant to slope stability analyses. For these purposes, three slipped slopes and a small watershed were selected from each area (Fig. 4). We assumed that hillslope materials on the same substrates have similar homogeneous geotechnical properties. The sandstone watershed (S-watershed) maintains a stable base flow of 0.2–0.3 L/s welling from a spring at
5 m upstream from the catchment outlet (Fig. 4B). The channel floor is inundated with saturated sediment resulting from the groundwater seepage. Behind the spring, soil-mantled slopes with no slip scars form an unchanneled dry hollow. In the mudstone watershed (M-watershed), a channel with well-defined banks incises the bottom of the watershed and is dry except for a short period during rainfall events (Fig. 4D). The upstream end of the channel branches into tributaries and they extend to the upper convergent hollows. Several hillslopes near the deep channel underwent slipping, exposing the mudstone bedrock. Most slopes in the watershed have thin soil cover, and some have only organic deposits on the bedrock. Fig. 5 shows profiles of the six selected slopes, as well as the plan views and cross-sections of the slip scars on the slopes. All the scars have a shallow platy form bounded by a small scar step, showing general geometry of a translational landslide (Selby, 1993). Those landslides seem to have resulted from the
Fig. 5. Profiles of the selected slopes (S-1, 2, 3 and M-1, 2, 3), and plan views and cross-sections of slip scars. The subsurface layers of S-1 and M-1 are inferred from soil soundings. Enlarged views of the upper parts of S-1 and M-1 show tensiometer nests for pressure head monitoring (ST1, 2 and MT1, 2, 3). The schematic illustration in the lower right shows soil pit scraping at scar heads.
Y. Matsushi et al. / Geomorphology 76 (2006) 92–108
heavy rainfall on August 1, 1989, since they were first visible on the aerial photograph taken in 1990. Landslides on the same substrates have similar slip depths and slope angles. Table 1 describes the dimensions of the landslides. The listed slip depths measured vertically from the original surface as well as slope angles represent mean values estimated from the cross-sections and longitudinal profiles of the landslides. Volumes of the landslides were calculated from the mean slip depth and the sliding area. The landslides on the sandstone slopes are deeper, steeper, and larger than those on the mudstone slopes. 3.2. Investigation of soil thickness On the S-1 and M-1 slopes, we conducted soil soundings along the scar profiles using a simplified dynamic cone penetrometer. The penetrometer consists of a penetration rod with a cone tip (25 mm diameter with 608 tip angle), guide rods and a 5-kg weight. The weight drives the cone into subsoil falling from a height of 50 cm along the guide rod. Here, the resistance value for the cone penetration N c is defined as the number of impacts needed for every 10-cm penetration. Using N c, Wakatsuki et al. (2005) distinguished four subsurface layers: 0 V N c b 5 (upper); 5 V N c b 10 (middle); 10 V N c b 30 (lower) and N c z 30 (bedrock). Accordingly, the thickness of soil mantle is defined as the depth attained when the N c reaches 30. 3.3. Soil sampling In addition to the soil soundings, we scraped soil pits in order to inspect subsurface structure (Fig. 5; pits 1 and 2 on the S-1 slope, pit 3 on the M-1 slope). Subsurface soil profiles were recorded using the exposed upslope face of the pits. The soil profile at the scar head includes a potential failure plane that
Table 1 Dimensions of the landslides Slip depth (m)
Slope angle (degrees)
Sliding area (m2)
Volume (m3)
Sandstone S-1 1.6 S-2 1.7 S-3 1.4
38.4 37.1 38.4
620 570 230
990 980 330
Mudstone M-1 0.7 M-2 0.6 M-3 0.6
32.2 35.4 34.7
70 70 100
50 40 40
97
corresponds to the original height of the scar step (Fig. 5). Undisturbed soil cores of 5.0 cm diameter and 5.1 cm height were extracted at every 10–15 cm depth intervals. These samples provided the depth-averaged physical properties and hydraulic characteristics of the soil. Samples for soil strength determination were also obtained from the pits 1 and 3 using a trimming ring of 6.0 cm bore diameter and 2.0 cm height. The dimension of the soil specimens is identical to the shear instrument. Sampling depths are 70–80 cm for the pit 1 (24 cores) and 30–40 cm for the pit 3 (20 cores), which roughly correspond to the depth of the potential failure planes. 3.4. Soil tests The dry unit weight, porosity, grain-size distribution, and depth profiles of saturated hydraulic conductivity were obtained from the cored materials. In-situ testing of the infiltration capacity was also conducted at the bottom of the mudstone pit because of difficulty in undisturbed sampling. Drying soil–water characteristic curves were also measured for the soil cores from 30 cm depth in the pit 1 (sandstone) and 15 cm depth in the pit 3 (mudstone). Hysteresis between drying and wetting processes and spatial variance in soil–water retention characteristics were not taken into account. Therefore, only one soil–water characteristic curve could be defined for each soil. Basic box shear tests using the undisturbed soil specimens were performed to determine soil strength parameters. The moisture contents of the specimens were adjusted stepwise from an air-dried condition to a capillary saturated condition. These specimens were delicately placed into an isometric shear box after sealing them in a plastic bag for at least a week to allow moisture equilibration. Single stage direct shear tests at a strain rate of 1 mm/min were conducted under four different normal stresses (10, 20, 30 and 40 kPa). The volumetric water content of the specimens was calculated from the weight difference between the specimen just after the shear test and that in completely dried condition. 3.5. Hydrological observation Tensiometers were set-up within 10 m upslope from the scar head on S-1 and M-1 to monitor the pressure head (Fig. 5). The tensiometers were established at two nests on the S-1 slope (ST1 and ST 2, with ceramic cups at depth of 30, 75, and 120 cm), and at three nests
98
Y. Matsushi et al. / Geomorphology 76 (2006) 92–108
on the M-1 slope (MT1, MT 2 and MT 3; the depths of ceramic cups were 30 cm for MT1, 23 cm and 46 cm for MT2, and 30, 60, and 90 cm for MT3). A data logger connected to the tensiometers recorded the subsoil pressure head at every 10 min. Runoff from the S- and M-watersheds (Fig. 4B and D) was measured at the watershed outlet by means of a weir and a water-level gauge. The flow depths in the weir were recorded with a data logger at 10-min intervals. The observed water level was converted into a discharge using a calibration curve. Rainfall in 10-min
intervals was also observed at the points denoted with stars in Fig. 4A and C. The period of these observations was from May to October in 2004, including a respite during dry summer season. 4. Subsurface structure of the slopes 4.1. Soil thickness The sandstone and mudstone slopes have significantly differing soil thicknesses. The value of N c for
Fig. 6. Subsurface soil profiles with depth variations of the cone penetration resistance (N c values), in the sandstone slope (A), (B) and the mudstone slope (C). Horizontal narrow bands represent the depth of potential failure planes.
Y. Matsushi et al. / Geomorphology 76 (2006) 92–108
the sandstone slope increases gradually and reaches 30 (bedrock) at depths of 6–7 m. The slip plane of the landslide perches on the boundary of the upper (0 V N c b 5) and middle (5 V N c b 10) layers. Beneath the scar head of the landslide, thick soil mantles (4–5 m of N c b 30 layers) remain on the bedrock (Fig. 5). In contrast, the soil thickness on the mudstone slope is up to a meter except at the bottom of the slope covered with colluvial deposits. On the intact part of the slope upward from the scar head, soft upper soil (0 V N c b 5) translates sharply into the bedrock (N c z 30) at the base of the soil layer. The slip plane of the landslides lies just above this soil–bedrock boundary. 4.2. Soil profiles The subsurface soil profiles of the slopes S-1 and M1 have contrasting appearances. Fig. 6 shows the soil profiles at the pits with depth variations of the N c values. Horizontal narrow bands on the diagrams represent the potential failure planes (~75 cm deep for S-1, and ~55 cm deep for M-1). The soil profile of the slope S-1 consists of homogeneous sand (Fig. 6A). The N c value in the pit 1 increases slightly with depth, but no apparent discontinuity is found around the potential failure depth. The sand at the bottom of the pit 2 on the landslide scar (Fig. 6B) remains the original beddings of the stratum (the Ichijuku Formation). This observation indicates the slope material in this part originates from in-situ weathering of the sandstone bedrock. The soil profile on slope M-1 (Fig. 6C) exhibits sharply demarcated weathering horizons that contrast the finding from the S-1 profile. Silty organic aggregates make up a soft loose soil (N c b 5) that includes scattered gravels at depths less than 0.3 m. From 0.3 to 0.6 m deep, interlocking rock fragments form a dense residual horizon (10 V N c b 20). The residual horizon sharply switches into the parent material (i.e., the silty mudstone of the Awakura Formation) below 0.6 m depth. 5. Slope materials 5.1. Physical properties The soils on the sandstone and mudstone slopes have different physical characteristics. Eight samples from the sandstone slope (pit 1) and five samples from the mudstone slope (pit 3) provided values of soil unit weight, porosity and grain-size distribution through the
99
soil profiles. Table 2 presents the depth-averaged values of these soil properties. The physical properties reflect the grain-size distribution of the soils. Soil originating from the sandstone has a large fraction of sand, whereas soil originating from the mudstone contains more fine materials. Correspondingly, the soil from the sandstone slope has a greater dry unit weight and a smaller porosity than the soil from the mudstone slope (Table 2). 5.2. Permeability Fig. 7 shows depth profiles of the saturated hydraulic conductivity of the soils. The box-plots in the upper diagram represent annual maximum 1-h rainfalls from 1979 to 2004 at the Kisarazu, Sakuma, and Sakahata climatological stations. Horizontal whiskers, shaded boxes, and vertical lines within the boxes represent the extremes, interquartile range, and median values, respectively. The intensity of annual maximum 1-h rainfalls ranges from 5 10 6 to 2 10 5 m/s (Fig. 7). The variation of the saturated hydraulic conductivity with depth strongly depends on geology. The hydraulic conductivity of the sandstone slope has a nearly straight vertical profile, with values around 10 4 m/s from the shallow soil layer (pit 1; including the potential failure plane) down to the bedrock (pit 2). In contrast, the hydraulic conductivity of the mudstone slope changes significantly with depth. The value on the order of 10 5 m/s near the surface falls abruptly to 8 10 7 m/s at a depth of 0.7 m. Field measurements of the infiltration capacity gave a minimum hydraulic conductivity of 5 10 8 m/s at a depth of 0.9 m. The hydraulic discontinuity in the profile at a depth of about 0.6 m corresponds to the potential failure plane. This implies a strong relation of rainwater accumulation upon the impervious bedrock with the conformation of the slip surface. 5.3. Soil–water retention characteristics The soils originating from the two parent materials have differing soil–water retention characteristics (Fig. 8). The volumetric water content of the sandstone slope Table 2 Depth-averaged physical properties of soils Specific Dry Porosity Grain-size distribution (%) gravity unit wt. (m3/m3) Clay Silt Sand () (kN/m3) Sandstone 2.69 Mudstone 2.68
12.7 11.1
0.52 0.58
5.6 12.4
10.1 43.7
84.3 43.9
100
Y. Matsushi et al. / Geomorphology 76 (2006) 92–108 Table 3 Optimum parameter values for the model fitted to soil–water characteristic curves Saturated v.w.c.* h s (m3/m3) Sandstone soil 0.465 Mudstone soil 0.530
Residual v.w.c.* h r (m3/m3)
Inflection Parameter r2 point w 0 m () (m)
0.171 0.061
0.186 0.001
0.569 0.085
0.97 0.81
*v.w.c. = volumetric water content.
pressure head decreases to 0.5 m. However the soil maintains relatively high volumetric water content (0.4 to 0.35) even when the pressure head is 4 m, since the small pores within the aggregates may retain immobile water. The broken and solid lines in Fig. 8 show the best-fit model curves. An equation of Van Genuchten’s (1980) model modified by Kosugi (1994) was adopted: " #m 1 1m w ð1Þ h ¼ hr þ ð hs hr Þ 1 þ m w0
Fig. 7. Depth profiles of saturated hydraulic conductivity in the sandstone and mudstone slopes. Box-plots in the upper diagram represent the annual maximum 1-h rainfalls from 1979 to 2004 at the Kisarazu, Sakuma and Sakahata climatological stations (see Fig. 1 for their locations).
soil falls smoothly from 0.45 to 0.2 as the pressure head decreases from 0 to 1 m and then reaches a residual state as the soil contains only suspended water. The soil on the mudstone slope rapidly drains gravitational water from its inter-aggregate macro pores before the
where h is the volumetric water content, h s is the saturated volumetric water content, h r is the residual volumetric water content, w is the pressure head, w 0 is the pressure head at the inflection point, and m is a constant (0 b m b 1). In this case, the air-entry value of the soil is assumed to be zero, because the specimens are undisturbed natural soils having sufficiently large maximum pore sizes to bubble in a near-zero negative pressure head. Eq. (1) was fitted to the data in Fig. 8 using the nonlinear least squares optimization of Marquardt (1963). The value of h s used was the volumetric water content at zero pressure head, and the other parameters (i.e., h r, w 0 and m) were estimated by the regression. The fitting analysis provided the optimum parameter values as in Table 3. The values accurately represent the soil–water
Fig. 8. Soil–water characteristic curves for soils originating from the sandstone (A) and the mudstone (B). The broken and solid curves are best-fit Kosugi’s (1994) model curves, whose optimum parameter values are listed in Table 3.
Y. Matsushi et al. / Geomorphology 76 (2006) 92–108
retention characteristics of each soil (r2 = 0.97 and 0.81). The values of h s and h r are similar to those in a previously published catalog (Mualem, 1976); w 0 and m for the soil from the sandstone are also similar to those in Kosugi (1994). However, w 0 and m for the soil from the mudstone are smaller than the previously reported values. This difference may be a result of the aggregated structure of the soil, which allows rapid water drainage from the inter-aggregate large pores. 5.4. Shear strength Shear tests revealed the characteristics of decline in shear strength with increasing volumetric water content (Fig. 9). An analysis of dry to saturated conditions provides evidence that the soils reduced their shear strength by 60–70% in the sandstone samples (Fig. 9A), and by 70–80% in the mudstone samples (Fig. 9B). The solid and broken lines in Fig. 9 represent the best-fit regression curves that express the relation between the shear strength, normal stresses, and volumetric water content based on the following formula: s ¼ Celh þ rtan/
ð2Þ
where, s is the shear stress, C is the ultimate increment of apparent soil cohesion (when h = 0), l is the reduction coefficient, r is the applied normal stress, and / is the apparent angle of shearing resistance. The equation involves an exponential function for the simplest regression of the change in cohesive component to the soil shear strength, assuming that the angle of shearing resistance takes a value content. Eq. (2) is consistent with the existing general remark that only the cohesive component decreases with in-
101
Table 4 Optimum parameter values for the regressions for reduction in soil shear strength Ultimate Reduction Angle of shearing r2 soil cohesion coefficient resistance C (kPa) l () / (degrees) Sandstone soil 35.8 Mudstone soil 192.9
4.81 6.94
28.3 27.7
0.88 0.93
creasing soil moisture (Fredlund and Raharjo, 1993). However, some previous studies employed a function of matric suction (soil pore-air pressure minus porewater pressure) as a parameter of soil moisture condition (Fredlund et al., 1978; Gan et al., 1988; Vanapalli et al., 1996), which cannot be determined accurately in our shear tests. Therefore, we use Eq. (2) as an approximate formula of soil shear strength to evaluate stability of unsaturated slopes. The unknowns in the equation (i.e., C, l and /) were determined by a multiple linear-regression analysis (Table 4). The variables in the equations reflect liquefaction characteristics of the soils and susceptibility to changes in pore-water pressure during shear deformation. Therefore, we regard these values as purely empirical rather than physically significant. Instead, the equations using these soil-specific parameters accurately represent the observed strength reduction characteristics of each soil (Fig. 9). 6. Subsurface water movement and runoff response 6.1. Long-term fluctuations in pressure head and rainfall–runoff characteristics The pressure heads in the sandstone slope show a small range of fluctuation throughout the observation
Fig. 9. Reduction of shear strength with respect to increasing soil volumetric water content, from air-dried to capillary saturated conditions. (A) Soil originating from the sandstone, (B) soil originating from the mudstone. The solid and broken lines represent best-fit regression curves for each soil, whose optimum parameter values are listed in Table 4.
102 Y. Matsushi et al. / Geomorphology 76 (2006) 92–108
Fig. 10. Pressure head fluctuations at ST2 and MT3 and runoff from the experimental watersheds in early summer and fall in 2004.
Y. Matsushi et al. / Geomorphology 76 (2006) 92–108
period (Fig. 10A). The pressure head at deeper points maintained a negative value even during intensive rainfall events. Pressure head at the shallowest monitoring depth (30 cm) momentarily exceeded zero only at rainfall peaks. Runoff from the S-watershed did not increase measurably in response to individual storms. The watershed maintained a stable discharge less than 0.5 L/s, with slight fluctuations on a rather long time scale. We verified the low discharge by field observations during a rainfall event. The pressure head on the mudstone slope varied over a wide range during the observation period from negative values of 2 m to positive values
103
up to 0.5 m (Fig. 10B). The pressure head displayed positive values during almost all distinct rainfall events at every depth. Then a marked decrease in the pressure heads continued until the next rainfall. The pressure heads between the events oscillate diurnally because of the evapotranspiration cycles. Runoff from the M-watersheds responded sharply to rainfall events. The accurate magnitudes of peak flows could not be measured because of overflow. Following these extremes the discharge fell rapidly, and then the watershed yielded a small discharge during the slope maintained the positive pressure head.
Fig. 11. Pressure head responses in the sandstone slope (A) and the mudstone slope (C) from May 19 to 21, 2004 (cf. Fig. 10). Slope profiles (B) and (D) illustrate the time-series distributions of the hydraulic head and of the saturated zone within the slopes. The numbers in boxes in (B) and (D) correspond to the arrows in (A) and (C), respectively.
104
Y. Matsushi et al. / Geomorphology 76 (2006) 92–108
6.2. Short-term responses of the pressure head The fluctuations from May 19 to May 21 are similar to those observed during the other rainfall events (Fig. 11A, C). Fig. 11B and D illustrate the distributions of the hydraulic head and the saturated zone at the four moments shown by the arrows in Fig. 11A and C. The pressure heads in the sandstone slope responded to rainfall sequentially from shallow to deep monitoring depths (Fig. 11A). The time lag between the rising edges of the pressure head at shallow and deep locations was about 6 h for the first event on May 19, and became shorter for the subsequent rainfall on May 21. In addition, the pressure heads rose abruptly at shallower depths but more gradually at deeper points. The equipotential hydraulic head lines in the sandstone slope tended to be horizontal in the absence of significant saturated zones (Fig. 11B). The pressure heads rose nearly concurrently with rainfall and reach positive values at all the monitoring depths in the mudstone slope (Fig. 11C). The sharp increases in the pressure heads began from the deepest point and rapidly moved to shallower depths. The positive pressures persisted for several hours at 30 cm depth, but several days at 90 cm depth. The hydraulic head lines under an expanded saturated zone tended to be perpendicular to the slope surface (Fig. 11D). 7. Slope stability analysis 7.1. Formulation of the slope stability model We employed a limit equilibrium method to analyze the critical conditions for landsliding. In this method, materials on a hillslope are subject to two opposing forces: a downslope component of soil weight as a driving force, and shear strength of the soil providing a resisting force. The factor of safety FS is: FS ¼ ¼
Resisting force Driving force Shear strength of soil : Downslope component of soil weight
Assuming an equable wetting of homogeneous soil, i.e., uniform soil water content and pressure heads from the land surface to the slip plane, FS can be expressed as: FS ¼
tan/ Celh þ tanb ðcd þ hcw ÞZsinbcosb
maxðw; 0Þcw tan/ ðcd þ hcw ÞZsinbcosb
ð4Þ
where Z is vertical soil depth, b is slope angle, c d is the dry unit weight of soil, and c w is the unit weight of water (9.8 kN/m3). On the right side of Eq. (4), the first term represents the frictional component of slope stability; the second term expresses the change in cohesive resistance in an unsaturated state (on the basis of Eq. (2)); and the third term refers to slope instabilization by the positive pressure head in the saturated state (w N 0 m). Substitution of Eq. (1) into Eq. (4) generates the factor of safety mediated by the pressure head, and allows a simulation of slope instabilization for both unsaturated and saturated conditions. The moisture condition of soil between the land surface and the slip plane is not actually uniform. Soil moisture decreases with increasing depth because of rainwater infiltration from the land surface during an intensive storm event. Thus the weight of soil mass in the unsaturated condition may be slightly underestimated. However, this under estimate has much smaller influence on FS than that of the other parameters (Borga et al., 2002; Dykes, 2002). The effect of the moisture heterogeneity in the soil column was ignored based upon the findings from these previous studies. We also did not take into account the stabilizing effect of tree roots, because the potential failure plane in the actual soil profiles lies well below the major root zone (Fig. 6). 7.2. Critical conditions for past landslides
ð3Þ
If FS b 1, the slope fails along a potential failure plane. Three instability factors were taken into account: reduction of soil cohesion in response to soil wetting, weight increase of soil resulting from water absorption, and decrease in effective stress derived from positive pore-water pressure. The former two were considered only in an unsaturated state, because their values are fixed at the saturated condition.
The slope stability analysis was applied to specify the critical conditions of some past landslides whose scars are described in Fig. 5. We obtained slip depths and slope angles of the landslides (Table 1) and also input-variables for the analysis using Eqs. (1) and (4) (Tables 2–4). We can therefore simulate the factor of safety in those slipped slopes as a function of the pressure head. Fig. 12 juxtaposes curves showing the results of the simulation. FS for the sandstone slopes falls below unity just before the saturation (w = 0.10 m for S-1,
Y. Matsushi et al. / Geomorphology 76 (2006) 92–108
105
Fig. 12. Simulation of the reduction in the factor of safety with respect to the pressure head in hillslopes, using actual slip depths and slope angles of the six landslides.
0.05 m for S-2, and 0.03 m for S-3). FS for the mudstone slope becomes 1.7–1.8 at the saturation point (w = 0 m), and sinks below unity when the pressure head reaches approximately 0.7 m (w = 0.76 m for M1, 0.69 m for M-2, and 0.71 m for M-3).
drains rainwater through the soil layer out to channel systems, giving acute runoff peaks in response to every rainfall event (Fig. 10B). In fact, discharge from the watershed continues while the pressure head in the slope maintains a positive value, and declines as the pressure head reverts to a negative value.
8. Discussion 8.2. Hydrological triggering mechanisms of landslides 8.1. Slope hydrology Relations between pressure head fluctuations in the slopes and runoff from the watersheds in each area can be explained as follows. In the sandstone slope, infiltrated rainwater can seep down vertically though the soil layers into bedrock, because there is no hydraulic discontinuity in the subsurface profile (Fig. 7). Accordingly, equipotential lines of the hydraulic head in the sandstone slope were horizontal (Fig. 11B), indicating vertical rainwater percolation. This unsaturated gravitational water flow should recharge the deep groundwater body in this permeable bedrock. Indeed, no perceptible responses in runoff were observed (Fig. 10A). Infiltrated rainwater discharges to the stream through the deep aquifer on a long-term basis, giving the observed constant base flow. The impermeable bedrock beneath a rather permeable soil layer prevents the percolation of incoming rainwater in the mudstone slopes (Fig. 7). The resultant equipotential lines of the hydraulic head perpendicular to the slope and expansion of the saturated zone demonstrate saturated downslope water movement upon the impermeable bedrock (Fig. 11D). This movement
The thick soil layer on the permeable sandstone bedrock cannot reach a saturated state (Fig. 10A). In this case, wetting from land surface migrates into subsoil without a positive pressure head (Fig. 11A). This redistribution of incoming rainwater causes reduction in cohesive strength of unsaturated slope materials (Fig. 9). The critical pressure heads for the sandstone slopes were indeed calculated to be negative values (Fig. 12). Consequently, the wetting front migration by intense rainwater infiltration causes a landslide associated with reduction of soil cohesion (Fig. 13A). This concept of landslide initiation is supported by a simulation of porepressure diffusion associated with slope-normal rainwater infiltration (Collins and Znidarcic, 2004). Forefront of the down-seeping wetting band locates the slip plane of this type of landslide. The thin overburden of the impermeable mudstone becomes saturated with a relatively small amount of rainfall (Fig. 10B). This results in generation of positive pressure heads, building up a shallow transient groundwater table from soil bottom to upwards (Fig. 11D). The mudstone slopes became unstable when the positive pressure head reaches to ~0.7 m (Fig. 12). This
106
Y. Matsushi et al. / Geomorphology 76 (2006) 92–108
Fig. 13. Schematic illustrations showing time-domain alterations in moisture depth profile and subsurface stress states in the sandstone slope (A) and the mudstone slope (B). The left part of the diagrams indicates wetting processes of slope materials, and the right part represents the change in shear stress (broken lines) and shearing resistance (solid curves) within the slopes.
indicates that a decrease in effective stress as well as reduction in soil cohesion triggers a soil slip upon the bedrock (Fig. 13B). The pressure head in the mudstone slope responds promptly to rainfall, reaching positive states in a short period (Fig. 11C). This lends support to the concept of efficiency of pore-pressure diffusion associated with slope-normal rainwater infiltration (Iverson, 2000). It also implies the concept of the steady-state groundwater flow is not appropriate to assess the direct triggering of landslides, even where the impermeable bedrock slopes. In fact, landslides in the mudstone area took place on uppermost hollows (Matsushi and Matsukura, 2004) without large contributing areas theorized in the steady-state model. 8.3. Landslide mechanisms, slope conditions, and landscape evolution Landslides on the sandstone slope are controlled whether wetting front reaches deep enough to form a failure plane. The highly permeable condition of the sandstone slopes disperses soil moisture and reduces
the potential for landsliding. Thus the sandstone area rarely experience landslide events as shown in Fig. 3. The less the landslides occur, the more thick and hardly saturated soil layers will be preserved on the hillslope (Fig. 5). The landsliding with negative pressure heads usually occur on hillslopes whose angles are steeper than the angle of shearing resistance of slope material (Rao, 1996). Slopes gentler than the angle of shearing resistance can slip only when the ascending force acts on the potential failure material. However, such a buoyancy effect is not expected for the sandstone slopes, because of the absence of the positive pressure head (Figs. 10A and 11A). The angle of shearing resistance is 28.38 for the soil from the sandstone (Table 4). Indeed, landslide locations in the sandstone area are confined to the steep (N308) lower parts of the slopes adjacent to main valleys (Matsushi and Matsukura, 2004). Thus, landslides do not erode the gentle (typically b208) upper slopes, resulting in the relatively high hillcrests in the sandstone area (Fig. 2). The subsurface water dynamics and frequency of landslide events on the mudstone slopes are in direct
Y. Matsushi et al. / Geomorphology 76 (2006) 92–108
opposition to these findings. The low permeability of the mudstone bedrock permits generation of positive pressure head that decreases shear resistance and promotes the frequent landsliding identified in Fig. 3. The critical pressure heads for landsliding (~0.7 m) roughly corresponds to thickness of the landslides (Table 1). This indicates that the maximum possible pressure head causes a landslide of minimum thickness. Consequently, the mudstone area has been eroded at a maximum possible rate maintaining only thin overburden on the hillslope (Fig. 5). The similar slope condition was reported in the undisturbed tropical rainforests in Brunei, where intensive rainstorms often hit hillslopes (Dykes, 2002). Rainfall often triggers landslides because of the low permeability of the bedrock, provided that adequately thick soil has developed upon the bedrock. Convergent hollows may accumulate sediment from surrounding ridges or side slopes, in addition to the in-situ disintegration of the bedrock. Readying sufficient soil thickness for a failure event, the hollows become the most feasible locations for landsliding. Landslides on the uppermost hollows probably spread their effect toward the hillcrest, evolving the low and rugged hilly landscapes (Figs. 1 and 2). 9. Conclusions This study has focused on landslide mechanisms on hillslopes underlain by permeable sandstone and impermeable mudstone in the Boso Peninsula, central Japan. Pressure head monitoring and rainfall–runoff observation revealed contrasting hydrological processes in hillslopes with the different substrates. Slope stability analysis, including strength reduction of the slope material, was also used to specify the critical conditions for landsliding. In the case of hillslopes with permeable sandstone, infiltrated rainwater percolates though the bedrock as an unsaturated gravitational flow. The reduction of soil cohesion resulting from wetting front migration causes landslides on the steep lower parts of the hillslopes when deep soil becomes sufficiently wet to form a failure plane. This type of landsliding contributes to the preservation of the relatively high hillcrests. The impermeable mudstone, by contrast, allows occurrence of saturated subsurface storm flow draining incoming rainwater though a thin permeable soil layer upon the bedrock. The resulting decrease in the effective stress at the soil–bedrock boundary causes a landslide. In this case, if soil is sufficiently thick, one can often expect rainfall-triggered landslides. The mudstone
107
area has then been eroded at a high rate, forming the low and rugged landscapes. Acknowledgments We thank Dr. Thad Wasklewicz for his technical support to improving the manuscript. This research was financially supported by the Science Research Fund of the Ministry of Education, Science and Culture (B-16300292, principal investigator: Y. Matsukura) and the JSPS Research Fellowship for Young Scientists in 2003–2004 held by T. Hattanji. References Borga, M., Fontana, G.D., Gregoretti, C., Marchi, L., 2002. Assessment of shallow landsliding by using a physically based model of hillslope stability. Hydrological Processes 16, 2833 – 2851. Cho, S.E., Lee, S.R., 2001. Instability of unsaturated soil slopes due to infiltration. Computers and Geotechnics 28, 185 – 208. Collins, B.D., Znidarcic, D., 2004. Stability analyses of rainfall induced landslides. Journal of Geotechnical and Geoenvironmental Engineering 130, 362 – 372. D’Odorico, P., Fagherazzi, S., Rigon, R., 2005. Potential for landsliding: dependence on hyetograph characteristics. Journal of Geophysical Research 110, F01007. doi:10.1029/2004JF000127. Dykes, A.P., 2002. Weathering-limited rainfall-triggered shallow mass movements in undisturbed steepland tropical rainforest. Geomorphology 46, 73 – 93. Fredlund, D.G., Raharjo, H., 1993. Soil Mechanics for Unsaturated Soils. John Wiley & Sons Inc, New York. Fredlund, D.G., Morgenstern, N.R., Widger, R.A., 1978. The shear strength of unsaturated soils. Canadian Geotechnical Journal 15, 313 – 321. Furuya, T., Ohkura, H., 1992. Some geological and geomorphological characteristics of the slope failures on the Mt. Kano-zan and its environs in the Boso Peninsula, Japan. Journal of Japan Landslide Society 28-4, 29 – 36 (in Japanese with English abstract). Gan, J.K.M., Fredlund, D.G., Rahardjo, H., 1988. Determination of the shear strength parameters of an unsaturated soil using the direct shear test. Canadian Geotechnical Journal 25, 500 – 510. Gasmo, J.M., Rahardjo, H., Leong, E.C., 2000. Infiltration effects on stability of a residual soil slope. Computers and Geotechnics 26, 145 – 165. Iida, T., 1999. A stochastic hydro-geomorphological model for shallow landsliding due to rainstorm. Catena 34, 293 – 313. Ito, M., 1995. Volcanic ash layers facilitate high-resolution sequence stratigraphy at convergent plate margins: an example from the Plio-Pleistocene forearc basin fill in the Boso Peninsula, Japan. Sedimentary Geology 95, 187 – 206. Ito, M., 1998. Submarine fan sequences of the lower Kazusa Group, a Plio-Pleistocene forearc basin fill in the Boso Peninsula, Japan. Sedimentary Geology 122, 69 – 93. Ito, M., Horikawa, K., 2000. Millennial- to decadal-scale fluctuation in the paleo-Kuroshio Current documented in the Middle Pleistocene shelf succession on the Boso Peninsula, Japan. Sedimentary Geology 137, 1 – 8. Iverson, R.M., 2000. Landslide triggering by rain infiltration. Water Resources Research 36, 1897 – 1910.
108
Y. Matsushi et al. / Geomorphology 76 (2006) 92–108
Kaizuka, S., Koike, K., Endo, K., Yamazaki, H., Suzuki, T., 2000. Regional geomorphology of the Japanese Islands. Geomorphology of Kanto and Izu-Ogasawara, vol. 4. University of Tokyo Press, Tokyo (in Japanese). Kim, J., Jeong, S., Park, S., Sharma, J., 2004. Influence of rainfallinduced wetting on the stability of slopes in weathered soils. Engineering Geology 75, 251 – 262. Kosugi, K., 1994. Three-parameter lognormal distribution model for soil water retention. Water Resources Research 30, 891 – 901. Marquardt, D.W., 1963. An algorithm for least-squares estimation of non-linear parameters. Journal of the Society for Industrial and Applied Mathematics 11, 431 – 441. Matsushi, Y., Matsukura, Y., 2004. Mechanism and location of slope failures in hilly terranes with different bedrock permeability. Transactions, Japanese Geomorphological Union 25, 139 – 159 (in Japanese with English abstract). Montgomery, D.R., Dietrich, W.E., 1989. Source areas, drainage density, and channel initiation. Water Resources Research 25, 1907 – 1918. Montgomery, D.R., Dietrich, W.E., 1994. A physically based model for the topographic control on shallow landsliding. Water Resources Research 30, 1153 – 1171. Mualem, Y., 1976. A new model for predicting the hydraulic conductivity of unsaturated porous media. Water Resources Research 12, 513 – 522. Onda, Y., 1992. Influence of water storage capacity in the regolith zone on hydrological characteristics, slope processes, and slope form. Zeitschrift fu¨r Geomorphologie N. F. 36, 165 – 178. Onda, Y., 1994. Contrasting hydrological characteristics, slope processes and topography underlain by Paleozoic sedimentary rocks
and granite. Transactions, Japanese Geomorphological Union 15, 49 – 65. Onda, Y., Tsujimura, M., Tabuchi, H., 2004. The role of subsurface water flow paths on hillslope hydrological processes, landslides and landform development in steep mountains of Japan. Hydrological Processes 18, 637 – 650. Rao, S.M., 1996. Role of apparent cohesion in the stability of Dominican allophane soil slopes. Engineering Geology 43, 265 – 279. Selby, M.J., 1993. Hillslope Materials and Processes, 2nd edition. Oxford University Press, Oxford. Terlien, M.T.J., 1997. Hydrological landslide triggering in ashcovered slopes of Manizales (Colombia). Geomorphology 20, 165 – 175. Vanapalli, S.K., Fredlund, D.G., Pufahl, D.E., Clifton, A.W., 1996. Model for the prediction of shear strength with respect to soil suction. Canadian Geotechnical Journal 33, 379 – 392. Van Asch, Th.W.J., Hendriks, M.R., Hessel, R., Rappange, F.E., 1996. Hydrological triggering conditions of landslides in varved clays in the French Alps. Engineering Geology 42, 239 – 251. Van Genuchten, M.Th., 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Science Society of America Journal 44, 892 – 898. Wakatsuki, T., Tanaka, Y., Matsukura, Y., 2005. Soil slips on weathering-limited slopes underlain by coarse-grained granite or fine-grained gneiss near Seoul, Republic of Korea. Catena 60, 181 – 203. Wu, W., Sidle, R.C., 1995. A distributed slope stability model for steep forested basins. Water Resources Research 31, 2097 – 2110.