Introduction

Shallow landslides, the destabilization of soil-mantled hillslopes due to heavy rainfall and/or earthquakes, occur worldwide (Jia et al. 2012). Shallow landslides cause debris flows and simultaneously introduce various hazards (Dietrich et al. 2007; Perrone et al. 2021; Petley 2012). The sediment yielded by shallow landslides is partly washed out of the catchment, whereas the remnant introduces various residual hazards and risks. For example, residual soils on sliding surfaces may consist of sliding materials that lead to the re-occurrence of shallow landslides. Furthermore, landslide masses, or landslide deposits, lead to multiple hazards, such as the destruction of infrastructure (Highland and Bobrowsky 2008), turbid stream water (El-Raouf et al. 2021; Kobiyama et al. 2011), water pollution (Geertsema et al. 2009; Göransson et al. 2009, 2012, 2014, 2018), and flooding due to the formation and breach of landslide dams (Fan et al. 2019, 2020; Highland and Bobrowsky 2008; Zhang et al. 2016). Therefore, residual risks in the aftermath of shallow landslides should be assessed for sustainable human activities and urban development. In particular, detecting residual soils on sliding surfaces and landslide deposits on valley floors and understanding their internal structures and moisture conditions are essential first steps in robust assessment of residual risks following shallow landslides.

Despite the importance of risk assessment associated with landslides (Aleotti and Chowdhury 1999; Wang and Nam 2022), including residual risks (Perrone 2021; Perrone et al. 2021), few case studies have attempted to visualize the spatial distribution of residual soils and landslide deposits on shallow landslide scars. This omission suggests that the potential environmental and socio-economic impacts of shallow landslide re-occurrences are likely underappreciated, with the result that no effective method for assessing the residual risks of shallow landslides has yet been established. If residual soil layers and landslide deposits are unstable at a given site, destructive methods, such as drilling, should be avoided when seeking to understand the subsurface structure of the landslide scar (El-Raouf et al. 2021). Hence, a methodology involving minimal disturbance is required to enable the effective and safe investigation of subsurface structures that may involve residual risks.

Electrical resistivity tomography (ERT) is a non-destructive geophysical method that can effectively reveal subsurface structures on a hillslope scale. ERT can provide 2D and 3D images of electrical resistivity, reflecting the variation in lithology and distribution of moisture conditions below the ground surface (Pazzi et al. 2019; Perrone et al. 2014; Schrott and Sass 2008). This has enabled previous studies to utilize ERT to identify the thickness of weak soil layers in landslide-prone areas and the location of potential sliding surfaces for shallow landslides (e.g., Crawford et al. 2018; Yamakawa et al. 2012; Yoshihara and Hattanji 2022). However, there are few cases where ERT has been applied to visualize the spatial distribution of residual soils on a sliding surface caused by a shallow landslide or to landslide deposits on a valley floor, where residual risks may exist. We expect ERT to offer an effective means of assessing risks such as the re-occurrence of shallow landslides and the breach of landslide dams.

In this study, we identified the spatial distribution of residual soils on a sliding surface and landslide deposits on a valley floor by combining ERT and geotechnical methods to assess the residual risks of shallow landslide scars. We conducted ERT and portable cone penetration test (PCPT) along three survey lines on a shallow landslide scar resulting from the 2018 Hokkaido Eastern Iburi earthquake (HEIE) in Japan. Trench observations at a landslide scarp and bedrock outcrop and PCPT immediately above the trenches helped estimate the subsurface structure beneath the survey lines. A suite of results was interpreted to visualize the spatial distribution of residual soils, landslide deposits, and the location of the groundwater table. This study demonstrates that obtaining subsurface information on landslide-prone and/or disaster-stricken areas by combining PCPT and ERT may contribute to appropriate decision making for mitigating disasters related to shallow landslides.

Study area

The southern part of Hokkaido, Japan (42.820–42.891°N, 141.877–141.983°E; Fig. 1a) has a subarctic climate. According to the Japan Meteorological Agency station of Atsuma (42.73°N, 141.88°E; 20 m asl), mean annual temperature and precipitation from 2000 to 2021 were 7.1 °C and 1016 mm, respectively. The major basement lithology is the Neogene sedimentary rocks of the Kawabata Formation (conglomerate, sandstone, mudstone, and shale). Syncline and anticline axes exist in these sedimentary rock regions (Matsuno and Hata 1960; Matsuno and Ishida 1960) (Fig. 2). The volcanic soils overlying the basement rocks are derived from the volcanoes Tarumae (Ta-d; < 9000 years ago), Eniwa (En-a; < 20,000 years ago), and Shikotsu Caldera (Spfa-1; < 40,000 years ago) (Geological Survey of Japan 2023; Kawamura et al. 2019; Machida and Arai 2003; Matsuno and Hata 1960; Matsuno and Ishida 1960). The natural vegetation of beech, managed deciduous broad-leaved trees, and managed conifers dominate this study region (Biodiversity Center of Japan 1984a, b).

Fig. 1
figure 1

Location of the study area. a Simplified geology and distribution of landslides caused by the 2018 Hokkaido Eastern Iburi earthquake (HEIE). Geology is based on the Geological Survey of Japan (2023), Matsuno and Hata (1960), and Matsuno and Ishida (1960). The distribution of landslides was determined by the Geospatial Information Authority of Japan. The HEIE epicenter is shown as an open circle. The rain gauges of Japan Meteorological Agency stations are shown as orange triangles. b Study site with survey lines for electrical resistivity tomography. The distribution of a landslide scarp, landslide scars, and landslide deposits is shown. The contour interval is 1 m

Full size image
Fig. 2
figure 2

Simplified geological map with the strike and dip of the basement lithology, along with the distribution of shallow landslides caused by the 2018 Hokkaido Eastern Iburi earthquake (HEIE). The geological information is based on the Geological Survey of Japan (2023), Matsuno and Hata (1960), and Matsuno and Ishida (1960). Strikes and dips of basement geology are extracted from Matsuno and Hata (1960) and Matsuno and Ishida (1960)

Full size image

The Hokkaido Eastern Iburi earthquake (HEIE; 6.6 Mw) occurred at 3:08 JST on September 6, 2018. The focal depth was 37 km, and the epicenter (42.690°N, 142.007°E) was located 20 km south of the study area (Japan Meteorological Agency 2018a, b). The HEIE occurred a day after heavy rainfall (due to Typhoon Jebi) and triggered more than 6000 shallow landslides across the region (Ito et al. 2021; Yamagishi and Yamazaki 2018). The thickness of volcanic soil layers, which served as sliding materials during the occurrence of HEIE, typically varied between 1 and 3 m (Ito et al. 2021; Kawamura et al. 2019). Following HEIE and consequent shallow landslides, the vegetation that had previously stood within the landslide scar fell and/or was buried (Figs. 3 and 4). The spatial density of these shallow landslides was affected by various factors such as ground shaking, slope angle, topographic wetness, and geological features (Wang et al. 2019).

Fig. 3
figure 3

Aerial photographs of the study site before and after the 2018 Hokkaido Eastern Iburi earthquake (HEIE) and subsequent shallow landslides. a An orthophoto constructed in 2016 by the Geospatial Information Authority of Japan. b Compilation of aerial photographs taken in 2018 by the Hokkaido Government Department of Fisheries and Forestry

Full size image
Fig. 4
figure 4

Locations of landslide scars (units A, B, C, and G) and landslide deposits (units D, E, F, and H) in slope map generated from 1-m resolution DEM constructed after the HEIE (a) and 0.3-m resolution aerial photographs taken after the HEIE (b). The DEM data and aerial photographs were provided by the Hokkaido Government Department of Fisheries and Forestry

Full size image

The study site selected for detailed PCPT and ERT surveys comprises a low-relief headwater catchment that spans 4.6 × 104 m2 and ranges in altitude from 85 to 166 m asl (Fig. 3). An artificial water reservoir (Mizuho Reservoir), which is primarily used for irrigation, is located at the lower end of the study catchment. In the present study, three survey lines for PCPT and ERT were established on a landslide scar, and a detailed survey was conducted on a sliding surface with a width of 75 m, slope distance of 80 m, and mean slope angle of 25° and on a downstream valley floor covered with a landslide deposit (Fig. 1b). Some shallow landslide scars occurred upstream of the landslide deposits. Perennial seepage occurred at the toe of the landslide deposits (Fig. 5). A transmission tower is situated on a ridge in the study catchment (55 m west of the north end of NS1) and power lines follow the north–south orientation (Fig. 3). These power lines and tower constitute the only human-made structures close to the ERT survey lines.

Fig. 5
figure 5

a A photograph showing ERT lines and a seepage at the toe of a landslide deposit. The seepage formed a gully with a perennial water flow, indicated by an arrow in dark blue. A light blue shaded area indicates the location where a saturated soil layer with groundwater seepage was observed at the ground surface. b A photograph showing ERT lines and the Mizuho Reservoir. The green shaded area indicates a water reservoir without a water table during the ERT. The blue shaded area indicates a water table in an area of the Mizuho Reservoir

Full size image

The 1-m resolution DEM and 0.3-m resolution aerial photographs taken after the HEIE show topographic changes in the study catchment. Immediately after the HEIE, shallow landslides (units A, B, and C in Fig. 4) occurred contemporaneously in the study catchment. The landslide masses or debris flowed along the slopes and deposited on the valley floor, forming units D and E. Subsequently, the landslide mass that originated from the upstream zone (unit D) flowed along the valley course, cutting through the slope of unit B and the landslide deposit of unit E. As a result, the landslide deposits of units D and E merged to enter the Mizuho Reservoir, forming unit F. Unit F reached the southern edge of the Mizuho Reservoir and temporarily dammed the river on merging with the landslide mass (unit H) that flowed from the southern slope (unit G). However, it is noteworthy that a portion of unit F was subsequently removed by repeated fluctuation in the water level of Mizuho Reservoir and through erosion by the inflow of upstream water. A portion of unit F was also removed by a disaster response measure prior to our PCPT and ERT in 2021 (see Fig. 5b).

Methods

Determination of subsurface materials

To determine the subsurface materials at the study site, trench observations were conducted at a landslide scarp near the survey lines (PIT-1; 42.851°N, 141.911°E) and at a bedrock outcrop 3.5 km east of the study site (PIT-2; 42.844°N, 141.943°E) (Fig. 1). PIT-1 was created by re-excavating a soil pit previously created by Yoshihara et al. (2022), after which the soil structure was confirmed for the present study (Fig. 6a). PIT-2 was part of a road cut excavated by forest-road construction (Fig. 6d). A portable cone penetration test (PCPT), which is described in detail in the next section, was also conducted immediately above these trenches to determine the ranges of material hardness (Nc value) for the subsurface materials.

Fig. 6
figure 6

Results of trench observation and portable cone penetration test (PCPT) at two locations. a A view of the environment surrounding PIT-1. b A depth profile of Nc values at PIT-1. c A photograph of PIT-1. d A view of the environment surrounding PIT-2. e A depth profile of Nc values at PIT-2. f A photograph of PIT-2

Full size image

Portable cone penetration test

To determine the subsurface structure beneath the ERT survey lines, we conducted a portable cone penetration test (PCPT) immediately above PIT-1 and PIT-2 and at 17 points along the survey lines. The portable cone penetrometer is composed of a cone with a 60° tip angle and a 2.5-cm diameter, a 1.6-cm diameter rod, and a 5-kg weight. The cone-tipped rod is positioned vertically and driven into the ground by the impact of the weight falling from a height of 50 cm. The Nc value, an indicator of penetration resistance, is defined as the number of weight-drops required for the rod to penetrate 10 cm into the ground. The ranges of Nc values for subsurface materials were defined based on the relationship between the subsurface materials observed at PIT-1 and PIT-2 and the corresponding Nc values. Assuming that the correspondence between the Nc value and subsurface materials measured at the two trenches can be applied to the PCPT results along the survey lines, we determined the subsurface structure below 17 PCPT points along the three survey lines.

Electrical resistivity tomography

Electrical resistivity tomography (ERT) was conducted along three survey lines (EW1, EW2, and NS1) on November 1–4, 2021. The survey lines were set up at a shallow landslide scar (Figs. 1b, 3b, and 5). EW1 was set along a sliding surface on which the HEIE triggered a shallow landslide. EW2 crosses the middle section of NS1. NS1 passes through a valley floor covered by landslide deposits.

Apparent resistivity data were obtained using Syscal Pro and Switch Pro (IRIS Instruments, France), which provided a multi-electrode configuration. Table 1 lists the general settings and antecedent precipitation values for each ERT. A minor precipitation event (6 mm/day) occurred 1 day before the ERT along line EW1. During the field measurement, stainless steel electrodes with a length of 60 cm and a diameter of 0.6 cm were inserted into the ground surface. The number of channels (i.e., steel electrodes) utilized for ERT along lines EW1, EW2, and NS1 was 96, 47, and 56, respectively. Electrode spacing was 1.0 m along all three survey lines. Thus, the lengths of the survey lines were 95, 46, and 55 m for lines EW1, EW2, and NS1, respectively. These lengths were selected based on the accessibility and coverage of the area of interest. The ERT lines were arranged to overlay the slope sections and/or landslide deposits while avoiding thickets and fallen trees which may induce measurement errors due to the abnormally high ground resistance of electrodes.

Table 1 Settings and antecedent precipitation for each electrical resistivity tomography
Full size table

A 12-V car battery was used as the external electric power source. The voltage was set to a maximum of 800 V and the duration of the current injection was set to 1 s. Before measuring apparent resistivity, the ground resistance of all electrodes was lowered by pouring water on the electrodes. The average ground resistance and the corresponding standard deviation between the adjacent electrode pairs were 1.8 ± 1.6 kΩ for EW1, 4.3 ± 2.8 kΩ for EW2, and 3.6 ± 2.7 kΩ for NS1. It was presumed that these ranges of values were reasonable under the condition that the electrodes were inserted into a volcanic soil layer containing pumice, which exhibits high porosity.

The electric field induced by the transmission tower and power lines may affect the ERT results along NS1, which extends parallel to the power lines. However, these artifacts are unlikely to affect the ERT results along EW1 and EW2, which run approximately orthogonal to the power lines. To reduce the influence of these artifacts on the resistivity distribution, four stacks were employed in each measurement to improve the signal-to-noise (S/N) ratio. Furthermore, acquired data with negative voltages and standard deviations larger than 0.03 were excluded from subsequent data processing and analyses.

All survey lines used a mixture of dipole–dipole, Wenner, Schlumberger, and multi-gradient electrode array configurations. The apparent resistivity was determined from the current and voltage obtained from measurements using these four electrode arrays. This mixed electrode array configuration compensates for the drawback of each array. The Wenner array displays relatively greater sensitivity to vertical changes in the subsurface resistivity and exhibits a high S/N ratio; the Schlumberger array allows for deeper investigations with a greater sensitivity to horizontal changes in resistivity compared to the Wenner array; and the dipole–dipole array yields a better coverage of horizontal resistivity data with a higher spatial density and greater sensitivity than the Wenner array, although its S/N ratio is relatively weaker than the former two electrode array configurations (Loke 2004). The multi-gradient array enables us to obtain high-resolution resistivity data in a shorter time compared to the Wenner array (Aizebeokhai and Oyeyemi 2014; Dahlin and Zhou 2006). The dipole–dipole and multi-gradient arrays both offer simultaneous, multi-channel data acquisition (Dahlin and Zhou 2006).

To obtain 2D electrical resistivity distributions below the survey lines, the apparent electrical resistivity data were inverted using RES2DINV (Loke 2004). Parameters for inversion processing were summarized in Table 2. The smoothness-constrained least-squares inversion (L2-norm) method was adopted in the inversion process, and five to seven iterations were performed. The results of the inversion were visualized using ERTplot, a Python code developed by Halloran (2018).

Table 2 Parameters for inversion processing
Full size table

Immediately after the completion of ERT along the three survey lines, the electrical conductivity (EC) and temperature of the perennial seepage at the toe of the landslide deposits were measured with a portable meter (WQ-330 J, Horiba Ltd., Japan).

Infinite-slope stability analyses

To assess the potential re-activation of residual soils on the east-side slope of EW1 (i.e., unit B in Fig. 4), the slope stability was calculated based on two infinite-slope stability equations. A translational shallow landslide occurred on the target slope. This slope has long and linear sliding surfaces relative to their slip depth. Here, we assume that the target slope can be regarded as a planar and semi-infinite slope, based on the topography and distribution of residual soil along EW1. First, to assess the change in slope stability in response to the horizontal acceleration caused by an earthquake (Fs-earth), we applied the following equations used by Kawamura et al. (2019):

$${F}_{\mathrm{s}\,-\,\mathrm{earth}}= \frac{c\mathop{+} {\gamma }_{\mathrm{s}}z\cos\beta \left(\cos\beta \mathop{-}{k}_{\mathrm{h}}\sin\beta \right)\tan\varphi }{{\gamma }_{\mathrm{s}}z\cos\beta \left(\sin\beta \mathop{+}{k}_{\mathrm{h}}\cos\beta \right)}$$
(1)

where c is the cohesion, \(\varphi\) is the shear resistance angle, γs is the unit wet weight of the soil, z is the thickness of the soil layer, \(\beta\)  is the slope angle, and kh is the coefficient of horizontal seismic intensity (= a/g, where a is the acceleration and g is 980 Gal). Note that Eq. (1) does not include the coefficient of vertical seismic intensity, assuming that this is negligibly smaller than kh.

In addition, we applied the following equation proposed by Skempton and DeLory (1957) to assess how rainfall (Fs-rain) may affect slope stability at this site:

$${F}_{\mathrm{s}\,-\,\mathrm{rain}}=\frac{c\mathop{+}({\gamma }_{\mathrm{s}}\mathop{-}m{\gamma }_{\mathrm{w}})z\cos^{2}\beta \tan\varphi }{{\gamma }_{\mathrm{s}}z\cos\beta \sin\beta }$$
(2)

where γw (kN/m3) is the unit weight of water and m, ranging from 0 to 1, represents the groundwater conditions in the soil layer. A state where m = 0 indicates that no groundwater table exists above the potential sliding surface. When m is equal to 1, the groundwater table reaches the ground surface, indicating that the soil layer above the potential sliding surface is entirely saturated.

Results and discussion

Subsurface structure

Trench observations at the landslide scarp (PIT-1) and bedrock outcrop (PIT-2) revealed that the subsurface component consisted of three types: weak volcanic soil, highly weathered sedimentary rocks, and less-weathered sedimentary rocks as bedrock. At PIT-1, highly weathered volcanic soil layers were identified (Fig. 6b, c). A clay-rich soil layer was found at 1.7–2.1-m depth, corresponding to the depth of the sliding surface. These soil layers were classified as weak volcanic soils that mobilized when HEIE occurred, as the highly weathered volcanic soil layers were found above the slip depth, and the clay-rich soil layer corresponded to the sliding surface. PIT-2 exhibited boulders and sedimentary rocks of conglomerate and sandstone underlying a 0.8-m-thick weak volcanic soil layer containing sand and gravel (Fig. 6e, f).

By conducting PCPT immediately above PIT-1 and PIT-2, we confirmed that the Nc value differentiates the weak volcanic soils and weathered sedimentary rocks in the target area. At PIT-1, the Nc value was less than 2 for the weak volcanic soil layer and clay-rich soil layer, which corresponded to the sliding surface (Fig. 6b, c). At PIT-2, the Nc value ranged from 3 to 6 in the weak volcanic soil layer and exceeded 8 in sedimentary rocks (Fig. 6e, f). For consistency with Nc values observed at the trenches, we defined the soil thickness (i.e., the interface between the weak volcanic soil layer and the sedimentary rocks) as the depth at which Nc = 7. A layer composed of less-weathered boulders showed Nc ≥ 50 (Fig. 6e). Many studies have demonstrated that interfaces between weathered materials (e.g., soil and highly weathered bedrock) and fresh bedrock exhibit Nc ranging from 30 to 50 (e.g., Kosugi et al. 2006; Wakatsuki and Matsukura 2008; Watakabe and Matsushi 2019; Yamashita et al. 2017; Yoshihara et al. 2020). Thus, we defined the surface of the less-weathered sedimentary rocks as the depth at which Nc exceeds 50 (Table 3).

Table 3 Description and ranges of Nc and electrical resistivity for each subsurface material (defined based on trench observations)
Full size table

Assuming that the ranges of Nc values for weak volcanic soils and weathered sedimentary rocks observed at PIT-1 and PIT-2 (Table 3) can be applied to the Nc profiles along the survey lines, we determined the depths of the interface between the weak volcanic soils and weathered sedimentary rocks at 17 PCPT points (Figs. 7 and 8). The thickness of the weak volcanic soil layer (i.e., the depth where Nc = 7) ranged from 0.3 to 4.1 m. The bedrock depth (i.e., the depth where Nc exceeds 50) ranged from 1.3 to 4.4 m. The bedrock depths at three points within the landslide deposit (NS1-1, NS1-2, and NS1-3; Fig. 8) could not be determined because the bedrock surface was located deeper than 5 m, making it impossible to measure using the PCPT method.

Fig. 7
figure 7

Interpretation of portable cone penetration test (PCPT) and electrical resistivity tomography (ERT) results along EW1 (a), EW2 (b), and NS1 (c) with distributions of residual soils, landslide deposits, and groundwater table. The dashed and solid black lines denote the lower boundary of residual soil layers and landslide deposits, respectively. The black arrows indicate the location of PCPT points

Full size image
Fig. 8
figure 8

Vertical profiles of Nc and true electrical resistivity. Locations where portable cone penetration test (PCPT) was conducted are shown in Fig. 7. The electrical resistivity data were extracted from corresponding inversion results shown in Fig. 7

Full size image

The landslide deposit on the valley floor had a thickness of more than 2.0 m (Figs. 7 and 8). In addition, the Nc value at the valley floor increased with increasing depth and finally exceeded 50, except for the three points along NS1 (i.e., NS1-1, NS1-2, and NS1-3), where the bedrock depth could not be determined (Fig. 8). These results suggest that the landslide mass overlies the basement of the weathered sedimentary rocks.

Resistivity distribution

At this study site, the resistivity distribution was characterized by high resistivity in the shallow zone, with an abrupt change to low resistivity in the deep zones (Fig. 7). The inversion results with 5–7 iterations and within 9% RMS error showed resistivity values ranging from 4 to 3063 Ωm. A resistivity higher than 256 Ωm was observed only from the ground surface to a depth of 2 m. Beneath these resistive shallow zones, the resistivity abruptly decreased with depth. Around the downstream section of NS1, where seepage was found, the shallowest resistivity data fell below 64 Ωm, which was 2 orders of magnitude lower than the resistivity at the upstream ground surface (Fig. 7c). Moreover, a conductive zone less than 64 Ωm dominated the area deeper than 3.0 m beneath the ground surface and around the toe of the landslide deposit, where seepage was found (below 82 m asl; Fig. 7c). This conductive zone showed no abrupt spatial changes in resistivity. These results suggest that the saturated zone at this site has resistivity less than 64 Ωm, irrespective of the subsurface material.

The EC and temperature of seepage found at the toe of the landslide deposit were 138 μS/cm and 11.7 °C, respectively (Fig. 5a). This EC value has a reciprocal resistivity value of 72.5 Ωm.

Interpretation of the results

We interpreted the results of PCPT and ERT to estimate the distribution of subsurface materials (i.e., weak volcanic soils, highly weathered sedimentary rocks, and less-weathered sedimentary rocks) at the target site (Fig. 7). Here, we estimated that the groundwater table is located around 82 m asl in the study catchment, based on the fact that seepage was found at the toe of the landslide deposit and the occurrence of a horizontal low-resistivity layer showing less than 64 Ωm in the vicinity of the seepage (Fig. 7c). The water level of Mizuho Reservoir during the ERT was assumed to be lower than 79 m asl since the water surface was intermittently observed in an area of the reservoir close to the study catchment, and ERT line NS1 did not enter the reservoir. Moreover, we assumed that adjacent soil depths (i.e., depths where Nc = 7) and surfaces of less-weathered sedimentary rocks (i.e., depths where Nc ≥ 50) above the groundwater table which were detected by PCPT show a similar range of electrical resistivity. The potential impact of the induced electric fields generated by the transmission tower and power lines on the resistivity distributions was deemed negligible in this study. This determination was based on the absence across all true resistivity distributions of any detected anomalies that could be reasonably attributed to artifacts (Fig. 7).

A comparison of vertical changes in Nc and true electrical resistivity at 17 PCPT points along the survey lines showed that the average true resistivity value for the weak volcanic soil layer, highly weathered sedimentary rocks, and less-weathered sedimentary rocks observed just beneath the PCPT points were 350, 106, and 26 Ωm, respectively (Table 3). The abrupt spatial change in electrical resistivity at the bottom of the weak volcanic soil layer reflects the differing physical characteristics between the weak volcanic soil and weathered sedimentary rocks. Electrical resistivity is affected by several parameters, such as the type, saturation degree, and porosity of rocks and electrical conductivity of pore fluids (Archie 1942; Lesmes and Friedman 2005). In unsaturated zones, weak volcanic soils exhibit higher electrical resistivity than weathered sedimentary rocks, probably because the soils show relatively high porosity, high permeability, and low water saturation. Moreover, in a saturated zone (below 82 m asl in this study), true resistivity is also strongly influenced by the resistivity of pore water (Archie 1942).

Weak volcanic soils entirely covered the hillslope and valley floor, with resistivity values ranging from 20 to 64 Ωm in the saturated zone (i.e., below the groundwater table) and exceeded 64 Ωm in the unsaturated zone. Resistivity readings greater than 1057 Ωm were interpreted as indicating weak volcanic soil zones because they were only found in the soil zone (Table 3). The weathered sedimentary rocks in the saturated zone corresponded to a zone that mostly exhibited electrical resistivity of less than 100 Ωm. The interfaces between the weak volcanic soils and weathered sedimentary rocks generally exhibited abrupt spatial changes in true resistivity, ranging from 20 to 100 Ωm (Fig. 8). The resistivity range for the less-weathered sedimentary rocks was low (< 70 Ωm) in the deeper parts beneath the landslide deposits. In contrast, the resistivity range was high in the shallower part of the slope section, with a range of 64–256 Ωm around EW1-5, 1–6, and 1–7. This observation suggests that the true resistivity value for the saturated less-weathered sedimentary rock falls below 70 Ωm, while that for the unsaturated less-weathered sedimentary rocks exceeds 70 Ωm (Fig. 7a). In addition, although the observed highest true resistivity was 93 Ωm just beneath the PCPT points, resistivity readings in the unsaturated less-weathered sedimentary rocks reach as high as 256 Ωm in the slope section (Fig. 7). Based on these assumptions and interpretations, the 2D distributions of residual soils, landslide deposits, bedrock surfaces, and groundwater tables were identified (Fig. 7).

The electrical resistivity of the water seeping from the landslide deposit (72.5 Ωm) exceeds the minimum electrical resistivity observed in the inversion results around the toe of the landslide deposit. Therefore, we presume that the true resistivity of less than ~ 70 Ωm is substantially affected by the water saturation and clay content. This also implies that ERT could not identify the soil–bedrock interface in the saturated zone, probably due to low and similar resistivity ranges for weak volcanic soil zones and weathered sedimentary rock zones. In contrast, PCPT, which reflects only the material hardness, could reliably distinguish the soil–bedrock interface in the saturated zone.

Berm-like landforms were observed near EW1-1, between EW1-3 and EW1-4, and between EW2-1 and EW2-2 (Fig. 7a, b). Traces of soil movement imprinted on the ground surface (i.e., scars) indicate that the location of these berm-like landforms coincides with the intersection of the landslide mass from the upstream parts (i.e., unit A in Fig. 4) and the landslide deposits from the downward parts of the study catchment (i.e., unit B). Therefore, these berm-like landforms are deemed to dissect the residual soil layers on the slopes and the landslide deposits on the valley floor. Notably, unlike the three points mentioned earlier, no berm-like landforms were found between EW2-3 and EW2-4, where an interface of residual soils and landslide deposits should exist (Fig. 7b). This observation is likely attributable to (i) the small volume of landslide deposit supplied from the vicinity slope located downstream of unit B and (ii) the diffusive flow of thinned unit B.

A resistive zone with a resistivity greater than 512 Ωm was identified directly below EW2-3 (Fig. 7b). This zone was dissected by a resistivity zone ranging from 256 to 512 Ωm. This horizontal discontinuity in true resistivity, which coincided with an absence of vegetation, indicates the presence of two separate flow units of landslide materials within the landslide deposit. Additionally, the lower resistivity zone with no vegetation cover may form a preferential flow path for water, facilitating its downward infiltration.

A high-resistivity zone of thickness ~ 1.6 m and exhibiting more than 512-Ωm true resistivity was observed beneath a knickpoint located just upslope of NS1-3 (Fig. 7c). This zone is composed of weak volcanic soils with high porosity and high hydraulic conductivity, which are likely to facilitate the downward percolation and lateral flow of subsurface water. The high-resistivity zone thins in the downstream direction and disappears once at the downstream side of NS1-2 (Fig. 7c). Soils around the downstream side of NS1-2 are estimated to show high water content compared to other parts of the landslide deposits, potentially due to the contribution of subsurface water flow originating from upslope.

The distribution of vegetation cover on the landslide deposits and its burial within the deposits can influence the subsurface distribution of true resistivity. Vegetation and plant root systems affect subsurface electrical resistivity (Giambastiani et al. 2022). Thus, the presence and physicochemical characteristics of plant root systems around the ERT lines may give rise to spatial variation in the true resistivity distribution. For instance, vegetation cover and burial could influence the varying thicknesses of resistive zones at three knickpoints along line NS1, where vegetation cover is sparse (Fig. 7c). However, evaluation of the impact of vegetation on true resistivity is limited in the present study due to insufficient data on the subsurface burial of vegetation.

Based on the results of PCPT, we detected residual soil layers of 0.3–3.2 m in thickness on the sliding surface (Figs. 7 and 8). In addition, we found that the thickness of the landslide deposit on the valley floor generally exceeded 2.0 m. The proposed method successfully visualized the 2D distributions of the landslide deposit at the target site, and the results imply that landslide-prone areas exist even in shallow landslide scars. Our interpretation also suggests the presence of a groundwater table within the shallow weathered bedrock layer. In other words, in the study catchment, the groundwater table is located below the landslide deposit which is mainly composed of weak volcanic soils.

Residual risks

In this study, PCPT and ERT were used to estimate the subsurface structure of a hillslope and valley floor disturbed by earthquake-induced shallow landslides. The results helped to visualize the spatial distributions of residual soils that can be potential sliding materials and landslide deposits that may lead to multi-hazards such as debris flows, mudflows, and floods. Our findings also suggest that the combined use of PCPT and ERT is effective in imaging the subsurface structure of shallow landslide scars and detecting residual risks on sliding surfaces.

Our interpretation of the results indicates that the groundwater table was located below the landslide deposit a few days after minor rainfall. Therefore, the landslide deposit appears to be unsaturated and separated from the groundwater table, even after a small amount of precipitation. This estimation poses little risk of multiple hazards associated with landslide deposits under base-flow conditions. Conversely, it also suggests that soil particles and soluble ions may be released from the landslide deposit as the groundwater level fluctuates in response to heavier rainfall events and/or the change in the water level of the nearby water reservoir (in this case, Mizuho Reservoir), thereby increasing stream turbidity.

Potential for re-activation of residual soils

In this section, we present the results of the infinite-slope stability analyses (Fs-earth and Fs-rain). For the determination of Fs-earth, we solved Eq. (1) by assigning two kinds of combination of c, \(\varphi\) , and γs of a potential sliding surface as determined by Kawamura et al. (2019): c = 8.0 and 33.0 kPa, \(\varphi\)  = 29.0° and 42.5°, and γs = 9.09 kN/m3. Parameters \(\beta\)  and z were defined as the mean slope angle and mean thickness of residual soil on the east-side slope of EW1 (i.e., unit B in Fig. 4): \(\beta\) = 25° and z = 2.0 m. Analytical results are shown in Fig. 9a. Under the conditions of low cohesion (c = 8.0 kPa) and high-friction angle (\(\varphi\) = 42.5°), Fs-earth fell below 1 when kh exceeded 0.7. Under conditions of high cohesion (c = 33.0 kPa) and low-friction angle (\(\varphi\)  = 29.0°), Fs-earth fell below 1 when kh exceeded 1.8. Therefore, it is estimated that an earthquake with kh = 0.7 or more can induce the re-occurrence of shallow landslides on the east-side slope of EW1. The maximum seismic acceleration was 1004 Gal in the north–south direction and 904 Gal in the east–west direction at the nearby Oiwake K-NET station when the HEIE occurred (National Research Institute for Earth Science and Disaster Resilience 2022), in which kh was 1.02. Therefore, residual soils on sliding surfaces with low shear strength may mobilize owing to an earthquake with an intensity lower than that of HEIE in this region. Given that the HEIE caused more than 6000 shallow landslides in an area of more than 20 km × 20 km (Fig. 1a), we infer that there is still a regional-scale risk of the re-occurrence of shallow landslides and resultant disasters on gentle slopes that have relatively thicker (>~2 m) residual soil layers in this region.

Fig. 9
figure 9

Relationship between horizontal coefficient and Fs-earth (a) and relationship between soil thickness and Fs-rain (b). Parameter \(\beta\) is determined as the mean slope angle of the east-side slope of ERT line EW1 (i.e., unit B in Fig. 4). Other parameters were obtained from Kawamura et al. (2019). The kh value when HEIE occurred (1.02) is indicated by a dark red dashed line in Fig. 9a

Full size image

Next, we determined Fs-rain by solving Eq. (2). In cases of (i) c = 8.0 and \(\varphi\) = 42.5° and (ii) c = 33.0 kPa and \(\varphi\)  = 29.0° as the geotechnical parameters for the potential sliding surface, Fs-rain exceeded 1 even if the soil layer is fully saturated (m = 1) and its thickness exceeded 3 m (Fig. 9b). This result indicates that a rainfall-triggered shallow landslide is less likely to occur on the east-side slope of EW1 than an earthquake-induced shallow landslide. Furthermore, considering that the slope angle of the majority of sliding surfaces ranges from 25° to 30° in this region (Wang et al. 2019), we predict that residual soils on sliding surfaces formed by HEIE are less prone to mobilize when heavy rainfall occurs. Based on slope stability analyses using Eqs. (1) and (2), it is estimated that residual soil has the potential to mobilize mainly because of an earthquake, whereas it is more resistant to rainfall.

Limitations and implications

Although the method used in this study is effective for estimating subsurface structures and assessing residual risks after shallow landslides on a single-slope scale, the application of PCPT and ERT requires time-consuming and laborious work to obtain regional-scale information. To assess the residual risks for all landslide-prone or disaster-stricken areas in a given region, PCPT and ERT should be conducted on slopes with representative hydrogeomorphological conditions, and subsurface information must be expanded or extrapolated to a regional scale. This limitation may be overcome, for example, by relating topographic characteristics derived from high-resolution airborne or satellite data, such as airborne LiDAR DEM and remote sensing data, to estimate electrical resistivity and/or subsurface structure below the landslide scar.

In this study, sampling and geotechnical experiments were not conducted for soils at the depth of potential sliding surfaces. Instead, physical parameters were referenced from a case study conducted near the ERT site. To accurately determine the slope stability and evaluate the residual risks associated with shallow landslides caused by HEIE, it is necessary to employ site-specific data on the physical properties of sliding surfaces. However, in this study, we assume that the physical properties reported by Kawamura et al. (2019) are applicable to our infinite-slope stability analyses. That is, the values for cohesion (c), friction angle (\(\varphi\)), and soil unit weight (γs) for the study area were obtained from Kawamura et al. (2019) considering the similarity in origins of volcanic soils. Thus, the infinite-slope stability analyses are based on (i) site-specific value of soil depth and slope angle (i.e., z and \(\beta\)) and (ii) potentially spatially homogeneous physical properties (i.e., c, \(\varphi\), and γs). Although it is desirable to collect site-specific values of c, \(\varphi\), and γs to refine the residual risk assessment, we suggest that combining PCPT and ERT is advantageous in obtaining the spatial distribution of local values of z and \(\beta\), which are necessary for assessing residual risks.

In addition to the slope stability, assessing the short- to long-term stability of landslide dams is also important for predicting, preventing, and mitigating landslide hazards. Some landslide deposits form river-blocking or valley-blocking dams. The subsequent breaching of landslide dams can lead to outburst floods, causing extensive damage downstream (Fan et al. 2012, 2019). The stability of landslide dams depends on various factors, including their volume, materials, and geometry, as well as the hydrogeological conditions of the catchment area, hydraulic gradient, and dam permeability (Fan et al. 2020). Our findings disclose the internal composition of the landslide deposits, which is linked to the stability of landslide dams and the groundwater level within them. Given the factors that govern the stability of landslide dams as outlined above, it is expected that visualizing the internal structure of landside dams by PCPT and ERT will also aid in interpreting the risks associated with floods and/or mudflows resulting from the failure of landslide dams.

Despite the need to overcome some limitations of scale, the present findings suggest that the combination of PCPT and ERT is effective for assessing residual risks in the aftermath of shallow landslides. The combined approach was able to detect several meters of residual soils that may re-activate, and landslide deposits that can present a risk of multiple hazards. Our findings highlight the importance of assessing residual risks in landscapes previously disturbed by shallow landslides. Future research into the residual risks of shallow landslides should explore the site-specific relationship between the subsurface structure and geophysical and geotechnical characteristics. This type of research will lead to the precise determination of residual risks at a regional scale.

Conclusions

The present study combined portable cone penetration test (PCPT) and electrical resistivity tomography (ERT) to estimate the subsurface structure of landslide scar. The results and interpretation aided in assessing residual risks associated with sliding surfaces and landslide deposits resulting from shallow landslides triggered by the 2018 Hokkaido Eastern Iburi earthquake (HEIE). The results and interpretation indicate a risk of re-occurrence of shallow landslides in the study region. The vertical profiles of material hardness (Nc value) at 17 points along the three survey lines on a landslide scar were identified by applying PCPT. The correspondence between subsurface materials and the ranges of Nc values were verified from the results of PCPT at a soil pit (PIT-1) and a bedrock outcrop (PIT-2). The observed Nc and electrical resistivity ranges for each subsurface material were collated as follows: weak volcanic soils: Nc < 7 and resistivity of 20 to 1852 Ωm, highly weathered sedimentary rocks: 7 ≤ Nc < 50 and resistivity of 16 to 1057 Ωm, and less-weathered sedimentary rocks: Nc ≥ 50 and resistivity of 10 to 93 Ωm. Based on the Nc and electrical resistivity range for each subsurface material and the interface between the weak volcanic soils and weathered sedimentary rocks generally showing 20–100 Ωm, the 2D subsurface structure along the survey lines was estimated. As a result, the 2D distributions of residual soils on the sliding surface and valley-floor landslide deposit were identified. At this site, the thickness of the residual soil layer on the slope ranged from 0.3 to 3.2 m and that of the landslide deposit exceeded 2.0 m. A groundwater table, which was inferred from the location of seepage and the distribution of the horizontal low-resistivity zone (< 64 Ωm), was broadly located below the landslide deposit.

In addition, infinite-slope stability analyses were performed to assess the potential of earthquake-induced (Fs-earth) and rainfall-triggered (Fs-rain) shallow landslides, using the topographic characteristics of the study site together with the soil geotechnical properties determined by Kawamura et al. (2019). The stability analyses indicate that shallow landslides may re-occur due to an earthquake with smaller acceleration than that of the HEIE, while rainfall-triggered landslides are less likely to occur at this site. For instance, a coefficient of horizontal seismic intensity (kh) of 0.7, which is ~ 0.7 times smaller than that of HEIE, can be a critical condition at some gentle hillslopes with a 2.0-m-thick layer of volcanic residual soil (in this case, z = 2.0 m, c = 8.0 kPa, \(\varphi\) = 42.5°, \(\beta\) = 25°, and γs = 9.09 kN/m3). Given the similar topographic reliefs and soil characteristics throughout the study region, it is predicted that the risk of re-occurrence of shallow landslides extends over an area of 20 km × 20 km. Based on our findings, we suggest that a combination of PCPT and ERT is effective in inferring the distribution of landslide-prone materials and in assessing residual risks that emerged after shallow landslides. This combined method has several limitations related to data coverage, but these may be overcome by collating high-resolution topographic characteristics, subsurface structures, and electrical resistivity distributions. Future research is expected to enable regional-scale assessment of the residual risks of shallow landslides.