1 Introduction

A landslide with a volume of 800,000 m3 occurred on 12th March 2001 at a cut-slope in a slate quarry at Shitagura of Soja City in Okayama Prefecture, Japan, killing three workers of the quarry plant (Figs. 1, 2). It is well established that large landslides such as this show precursory phenomena associated with the accumulation of strain prior to failure due to progressive failure processes (Saito 1965; Saito and Uezawa 1966; Bjerrum 1967; Suwa 1991; Hungr and Kent 1995; Petley et al. 2002; Kilburn and Petley 2003; Petley et al. 2005). In the case of a rock slope failure such as this, the rock-fall frequency increases in the final stages prior to failure (Hirano et al. 1990; Suwa et al. 1991). These precursory signals provide an indication of potential danger after which evacuation of people from dangerous places can be effected. Unfortunately in this case the workers failed to evacuate, with fatal consequences.

Fig. 1
figure 1

The location of the study slope

Full size image
Fig. 2
figure 2

A picture of the study slope taken on the afternoon of 12 March 2001 after the landslide, courtesy of Soja City

Full size image

It is important to clarify the following points in this study to contribute to mitigation of similar disasters in the future. Firstly, the actual processes of the landslide should be determined from available data. Secondly, the causes of the landslide should be explored through field surveys and inspection of available data from the view points of geology, geomorphology, hydrology, and so on. Thirdly, the processes that lead to the climax of the landslide and those that follow the climax should be determined from the available data, including interviews not only with local peoples but also with workers who were present and those that previously worked at and have retired from the quarry.

2 Geologic and geomorphic conditions

We have used a field survey to determine the geologic structure and the geomorphic conditions of the study slope (Fig. 3). The study site is situated on the Kibi plateau, which was formed by the Pliocene drainage system through dissection of the top Miocene strata. The study slope is located on the outside of a meander of the Takahashi River, meaning that the slope has been intensively undercut, producing a slope angle of 40°.

Fig. 3
figure 3

A topographic and geologic map of study slope prior to the landslide

Full size image

The basement of the study area is composed of the Permian Maizuru group (Suzuki 1987), Cretaceous granodiorite, Paleogene andesitic dykes and alluvial deposits. The Maizuru group is mainly composed of mudstone that has been affected by contact metamorphism associated with granodiorite intrusion. Unconsolidated, loose alluvial deposits have accumulated in the channel and flood plains of the Takahasi River.

During the field survey, the cross-sectional structure was also determined (Fig. 4). The cross section in the figure is drawn as a projection of the western periphery of the landslide scarp on to the vertical plane that goes through the line A–A′ in Fig. 3. The slope is located in the geologic zone of the Maizuru group that is composed of mudstone, alternations of sandstone and mudstone, and alternations of acidic tuff and mudstone. The mudstone is characterized by a texture showing a week sedimentary structure. The alternations of sandstone and mudstone consist of parallel-layered, fine-grained rocks. The alternations of acidic tuff and mudstone consist of parallel-layered, siliceous, very fine-grained tuff, fine-grained tuffaceous sandstone, and mudstone. They have been subjected to contact metamorphism that has generally left them harder and with a more-homogeneous texture.

Fig. 4
figure 4

Cross section along the line A–A′ in Fig. 3 showing the geologic structure of the study slope. The cross section in the figure is drawn as a projection of the western periphery of the landslide scarp onto the vertical plane that goes through the line A–A′ in Fig. 3

Full size image

As the bedding plains of these strata have an inclination of 30–50°, the study slope is a typical dip-slope, and thus the slope was prone to bedding plane slip. This is illustrated in the stereographic projection of the structural plane data (Fig. 5). A large portion of the plots shows that strike and dip of the joints and the faults permit a rock block on the cut-slope to be detached northward, namely in the direction in which the cut-slope of the quarry is facing. The strikes and dips of the faults especially appear to have strongly promoted this condition.

Fig. 5
figure 5

Wulff’s net showing equal-angle projection of poles to the lower hemisphere. A large portion of the plots shows that strike and dip of the joints and the faults permit a rock block on the cut-slope to be detached northward, namely in the direction in which the cut-slope of the quarry is facing. The strikes and dips of the faults especially appear to have strongly promoted this condition

Full size image

3 The movement of the landslide

The movement of the 12 March 2001 landslide was observed by chance by an eyewitness located on a gas stand on the opposite side of the Takahashi River. The account obtained from an interview with him suggests that first, the lower part of the cut-slop started to collapse, thereafter the top of the cut-slope and upper slope collapsed successively, presumably as the upper parts lost the support provided by the slope toe. Overall the landslide was 220 m long, 200 m wide and a maximum of 50 m deep (Fig. 6). It had a volume of about 800,000 m3.

Fig. 6
figure 6

Cross sections along the line A–A′ in Fig. 3 showing the profile before and after the slide

Full size image

A network of seismic instruments recorded the ground movements induced by the landslide. These tremors were detected at distances of up to 200 km from the location of landslide. Figure 7 shows a set of seismic records from observation points within 40 km of the slide. In this diagram the record sections are arrayed at their corresponding distance to the wave source at the quarry. These data indicate that the landslide was initiated at 10:13:42 local time and that it developed into discrete phases in which the movement of the major slide began 13 second after the onset of primary slide. These seismic data coincide well with the eyewitness accounts, which suggested that firstly the lower part of the cut-slope started to collapse, followed by the top of the cut-slope and upper slope.

Fig. 7
figure 7

Seismic record sections of Hi-net points within a distance of 40 km. The record sections are arrayed at their corresponding distance from the wave source at the quarry in Okayama

Full size image

4 Chronology of events before the slide

The quarry opened in 1971, since when it has experienced many rock falls and landslides. Some of these have been documented through interviews with workers and local residents (Fig. 8). Notably, three landslides occurred on 23 June 2000, with a total volume of about 50,000 m3. The rainfall conditions prior to the slide were 73 mm/7 days, 195 mm/30 days, and 350 mm/120 days. These rainfall data are moderate in this area. Thus it does not seem certain that this landslide sequence was triggered by elevations in groundwater from infiltration of rainfall. The weekly rainfall of more than 73 mm occurs several times a year in this area. The mean monthly rainfall for June, derived from the data between 1979 and 2000, is 196 mm. The mean annual precipitation is about 1.2 m and there is rainfall every month throughout the year in this area. Japan has a temperate humid climate. On the other hand, the rainfall conditions before the 12 March 2001 landslide were 5 mm/7 days, 52 mm/30 days, and 204 mm/120 days. Thus, the landslide occurred in comparatively dry conditions, suggesting that the groundwater system did not play a key role in the triggering of this failure.

Fig. 8
figure 8

The history of landslides from the cut-slope since the start of quarrying in 1971

Full size image

The precursory failures, which occurred as rockfalls and small slides, were observed by several residents living near the quarry from early in the Monday morning of 12 March 2001. Interviews with them suggest a sequential change in the failure frequency and volume of landslide debris (Fig. 9). These eyewitness accounts suggest that the location of these rockfall was distributed over the whole width of the cut-slope. The accounts also suggest that the frequency of rockfalls increased notably after 09:00. This is consistent with earlier observations that, as the time of the ultimate collapse approaches, the frequency of rockfall and slides increases markedly (Hirano et al. 1990). The time period and magnitude of these precursory phenomena appears to be proportional to the magnitude of the landslide (Suwa et al. 1991).

Fig. 9
figure 9

The history of precursory landslides in the morning of 12 March 2001. These data for the volume and time of the slides were estimated based on data from interviews with local residents and one of the surviving workers from the quarry

Full size image

5 Causes of the slide

The large failure of 12 March 2001 occurred without any clear trigger mechanism: the area did not experience exceptional rainfall, there were no recorded seismic events, and there was no large-scale blasting in the quarry immediately prior to the failure. However, three key conditions at this site probably promoted instability. Firstly, the slope was located on the outside of a meandering river channel, where the erosive action of the stream flow has permitted oversteepening (Fig. 1). Slopes on the outsides of a meander are prone to failure due to this steepness (For instance, Sah and Mazari 1998). Secondly the slope was prone to deep-seated landslides due to the geologic structure of the dip-slope (Fig. 4). Discontinuous dips are known to be able to form potential slip surfaces (For instance, Verstappen 1983). Thirdly, the configuration of joints and faults permitted the release of the landslide (Fig. 5). Undercutting of the steep slope by quarrying must have heightened the potential for failure.

The observed increase in rockfall frequency with time described in the previous section suggests that the slope was accumulating strain prior to the final failure, and that the rate of strain must have been increasing with time. Increased strain was probably brought about by the growth and propagation of cracks in the rock. We adopt the interpretation that the landslide occurred as a first-time failure at a very stiff, brittle rock. This interpretation coincides with the idea that the progressive failure of a stiff, brittle material due to a driving shear stress would accompany the interaction, coalescence, and propagation of cracks to create a sliding slip surface prior to ultimate failure (see, for instance, Saito 1965; Saito and Uezawa 1966; Petley et al. 2002; Kilburn and Petley 2003; Petley et al. 2005). Under this interpretation, a shear surface for the landslide might gradually form along one of the bedding planes by way of crack propagation. The final failure of the landslide probably occurred when the formation of the slip plane was completed.

6 Chronology of events after the failure

Rescue activities to retrieve the three victims were continued until 26 May 2001. Monitoring of rockfalls and landslides was undertaken to reduce the risk to the rescue members. The recorded rockfall events indicate that the volume of failure gradually decreased with time (Fig. 10), although in a somewhat complex manner. This complexity might reflect the occurrence of the 24 March 2001 Geiyo earthquake (M6.4 on the Richter scale), and heavy rainfall on 25 March 2001. The seismic intensity at Soja due to the Geiyo earthquake was recorded as level 4 using the intensity scale of the Japan Meteorological Agency, which approximately corresponds to level 6 to 7 on the modified Mercalli intensity scale.

Fig. 10
figure 10

History of landslides after the 12 March 2001 event. Data were available only for 13 days from 14 March 2001

Full size image

Considering the existence of these slides before and after the landslide climax, the authors propose the new terms foreslide, mainslide and afterslide for these sequential landslides following the terms, foreshock, mainshock and aftershock used in seismology.

Earthquakes are phenomena of rock collapse in the Earth's crust far underground, while landslides are phenomena of rock and soil collapse on the Earth's surface. Therefore the analogical terms foreslides, mainslides, and afterslides may be applied for landslides. However, definitions should be provided to these new terms. Although foreslides should be defined as the slides that precede the mainslide, the landslides listed in Fig. 8 should not be included as foreslides. Foreslides should be confined to those that are generated directly by the progressive instability that forces the mainslide to occur. Thus the failures and landslides listed in Fig. 9 should be included in the group of foreslides.

On the other hand, the landslides listed in Fig. 10 should be included as afterslides, which must have occurred due to a new instability of the slope resulting from the mainslide. The time periods for the foreslides and the afterslides were limited to 4 h and 2 weeks, respectively, in this study due to the limitation of the time period of eyewitness reports for the foreslides and the duration of monitoring after the mainslide. The actual time periods for the foreslides and the afterslides in this case must have been longer than these periods. The time periods of the foreslides and afterslides would generally depend on the scale of the landslide. It is presumed that there is a tendency that, the larger the scale, the longer the time periods. The scales of the foreslides and afterslides is substantially smaller than that of mainslides. Although the volume of the mainslide of the 12 March 2001 slide was 800,000 m3, the volume of each foreslide and afterslide was less than several tens of cubic meters.

7 Concluding remarks

Our study has shown that the 12 March 2001 landslide was conditioned by: (1) the steepness of the slope due to undercutting by the river, (2) the presence of a dip-slope, and (3) the presence of numerous joints and faults. In addition, quarrying activity left the cut-slope too steep and high. The rockfall data suggest that rainfall and earthquakes did not play any role in the triggering of this slide. The increase in the frequency of the precursory failures strongly suggests that a stress state induced the propagation and extension of cracks in the rock. A slip plane in the slide may have gradually extended along one of the bedding planes due to the propagation of cracks, such that the slide occurred when the formation of the slip plane was completed. This raises the question of whether completion of this shear surface is not therefore the final trigger for this landslide. We like to reserve the use of the term landscape “trigger” for an external agent or condition. Completion of a shear surface is not an external condition but the final stage of an internal condition. It is a natural phenomenon that results from the propagation of cracks. Therefore we adopt the interpretation that this landslide occurred without any clear trigger.

Were these three tragic fatalities avoidable? It appears that the response to the precursory phenomena was poor. Our interviews revealed that one of the local residents observed the abnormal level of small rockfalls and landslides during his morning walk at about 06:15. He immediately telephoned the head of the quarry company to inform him of this abnormal activity. The quarry company head then visited the quarry and confirmed that new deposits of a few tens cubic meters were present, but he did not think that the situation was sufficiently serious to justify a countermeasure against a possible disaster.

Workers arrived at the quarry before 08:00, which was the start of the working day, to find that the frequency of rockfalls from the cut-slope was already high. As a results, work on the quarry was stopped, and the workers watched the rock-fall processes from vantage points inside the quarry yard. Clearly they did not expect a landslide as large as 800,000 m3, perhaps anticipating an event similar in size to that of the slide on 22 June 2000.

A proper analysis of the rockfall activity would have alerted the quarry management to the possibility of a large-scale failure across the full width of the excavation. It appears that there was a failure to appreciate that the quarry was excavating a very unstable slope consisting of a dip-slope with a certain level of systematic configuration of joints and faults that would easily allow release a landslide.

In order to mitigate disasters against similar hazards in the future, quarry companies should survey the geologic structure of the slope prior to excavation, and workers should understand the need for preparedness against slope hazards. When precursory phenomena are observed, workers should be evacuated to safe locations.