Assessment of sudden sediment source areas incurred by extreme rainfall in a mountainous environment: Approach using a subsurface hydrologic concept

Quaternary International xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Quaternary International journal homepage: www.elsevier.com/locate/quaint

Assessment of sudden sediment source areas incurred by extreme rainfall in a mountainous environment: Approach using a subsurface hydrologic concept Minseok Kima, Hyunuk Anb,∗, Jinkwan Kimc, Sukwoo Kimd, Hyun-Joo Oha, Young-Suk Songa a

Geologic Environment Division, Korea Institute of Geoscience and Mineral Resources, 124 Gwahak-ro, Yuseong-gu, Daejeon, 34132, Republic of Korea Department of Agricultural and Rural Engineering, Chungnam National University, 99 Daehak-ro, Yuseong-gu, Daejeon, 34134, Republic of Korea c Department of Geography Education, Chonnam National University, Gwangju, 500-757, Republic of Korea d Division of Forest Science, Kangwon National University, 1 Kangwondaehakgil, Chuncheon-si, Gangwon-do, 24341, Republic of Korea b

A R T I C LE I N FO

A B S T R A C T

Keywords: Sudden sediment source area Shallow landslide Soil strength Soil depth Soil test methods Subsurface hydrologic concept

In mechanism studies on sudden sediment source areas (SSAs), such as those of shallow landslides, area-specific soil strength information at the soil layer between saturated and unsaturated soil is an important factor. Although various soil-testing methods have been developed, it is difficult to measure soil strength directly due to local variation. Therefore, the purpose of this study is to estimate SSAs of shallow landslides by considering the soil strength attributes reflected in a simple subsurface hydrologic concept (SHC). The selected study area is a granitic region in Jinbu, eastern South Korea, where shallow landslides occurred at more than 1200 locations in 2006. To estimate SSAs, soil strength was generated using several methods (i.e., direct shear tests [DSTs], Monte Carlo-simulated direct shear tests [MSDTs], and triaxial compression test [TCTs]). In addition, to investigate the effects of soil attributes in saturated/unsaturated soil on the SHC, soil depth was surveyed at a small hillslope site within the large study area. A topographically driven physical shallow landslide stability model (SHALSTAB) was utilized to investigate the performance based on the various soil strength attributes, and the results were analyzed using receiver operating characteristic (ROC) analyses. In the evaluation using the three soil strength measures (Case I), the accuracy assessment of the estimated SSAs was relatively low (range of 0.45–0.71). Using the soil strength attribute reflecting SHC (Case II), the results showed a reasonable range of accuracy (0.84–0.85). In addition, using the effects of soil strength reflected by SHC and average soil depth (Case III) also showed good predictive accuracy (0.81–0.82), but it was lower than that of Case II. This indicates that combining the hydraulic properties related to soil strength facilitates more accurate identification of SSAs causing geomorphic surface changes.

1. Introduction In mountainous countries such as South Korea (where mountains occupy more than 70% of the land area), rainfall-induced shallow landslides are important morphological phenomena that occur via various hydrologic processes in soil mantle or regolith with a depth of a few meters (Caine, 1980; Caine and Swanson, 1989; Hovius et al., 1997; Trustrum et al., 1999; Godt et al., 2009; An et al., 2016). For instance, on July 16, 2006, Jinbu-myeon, eastern South Korea, which is underlain by granite, had rainfall of approximately 500 mm/day (total rainfall during a day) and 80 mm/h (maximum rainfall intensity), which caused considerable damage, including loss of human life, collapsed

embankments, and flooding, from sediment transported by more than 1200 shallow landslides. Source areas of sudden sediment (sudden sediment source areas, SSAs) such as shallow landslides are affected by complex processes that involve different hydrologic, topographic, and geomechanical variables in the soil mantle (D’Odorico and Fagherazzi, 2003; Giannecchini et al., 2007; Mondini et al., 2011). Thus, many approaches have been proposed and applied to estimate these source areas (Carrara et al., 1991; Claessens et al., 2005). In particular, many recent studies have focused on approaches using physical-based models, combined with infinite slope models and hydrologic models (Montgomery and Dietrich, 1994; Wu and Sidle, 1995; Iverson, 2000; Godt et al., 2008; Simoni et al.,

∗ Corresponding author. Department of Agricultural and Rural Engineering, Chungnam National University, Daejeon, Republic of Korea. Tel.: +82 42 821 5797; fax: +82 42 821 5791. E-mail address: [email protected] (H. An).

https://doi.org/10.1016/j.quaint.2018.10.031 Received 29 May 2018; Received in revised form 23 October 2018; Accepted 23 October 2018 1040-6182/ © 2018 Elsevier Ltd and INQUA. All rights reserved.

Please cite this article as: Kim, M., Quaternary International, https://doi.org/10.1016/j.quaint.2018.10.031

Quaternary International xxx (xxxx) xxx–xxx

M. Kim et al.

Fig. 1. Shallow landslide locations in the study area.

addition, Park et al. (2013) used the soil strength attribute calculated by Monte Carlo simulation using 80ea soil samples, and they also mentioned the importance of soil strength. There are many methods for measuring soil strength, including the DST, triaxial compression test, and field tests to obtain more accurate measures of soil strength variables. However, it is difficult to measure soil attributes in the field (D’Odorico and Fagherazzi, 2003) because the sampling location and methods used to determine the physical properties of soil may affect these measurements (Kim et al., 2016)). In addition, soil properties can be affected by subsurface flow in the soil mantle during rainfall (Ferrer et al., 2003; Talebi et al., 2008; Lanni et al., 2013). Thus, some researchers (Rosso et al., 2006; Minder et al., 2009; Uchida et al., 2011) have attempted to understand hydrological mechanisms in relation to soil strength. They have evaluated the sensitivity of slope stability to hydrological effects in relation to intrinsic and extrinsic factors but have not considered hydrological effects on soil strength in the soil mantle. Therefore, more accurate analyses of SSAs, such as for shallow landslides, should consider the hydrological

2008; Baum et al., 2010; Uchida et al., 2011; An et al., 2016; Kim et al., 2016). Some researchers (Zizioli et al., 2013; Kim et al., 2016) have noted that applying models is useful for assessment of SSAs. Several previous studies (Milledge et al., 2012; Zizioli et al., 2013; Kim et al., 2015; Viet et al., 2017) have focused on estimating the spatial patterns of SSAs, such as shallow landslide areas, as soil properties play a critical role in assessing such areas. Zizioli et al. (2013) evaluated four landslide regions using four physical models (SINMAP: Pack et al.,1998; SHALSTAB: Montgomery and Dietrich, 1994; TRIGRS: Baum et al., 2008; and the SLIP model: Montrasio et al., 2011, 2012) with a 10 m resolution DEM and average soil values from soil samples of 80ea using the direct shear test (DST) method. The four models had similar degrees of success because the effective cohesion derived from the DST, which has high importance as a strength attribute in stability analyses, can be overestimated for shallow soils. Kim et al. (2013) and Viet et al. (2016) applied the TRIGRS model to obtain the relationship between soil attributes and vegetation properties, and they also emphasized the importance of soil strength in the occurrence of SSAs. In 2

Quaternary International xxx (xxxx) xxx–xxx

M. Kim et al.

vegetation; we supplemented these areas with blueprints for erosion control provided by the Korea Forest Service and field surveys. Fig. 1 displays the SSAs of shallow landslides in the study area, constructed using more than 1200 locations.

response of the soil and the soil's geotechnical properties (Duan and Grant, 2000; Schmidt et al., 2001; Giannecchini et al., 2007; Minder et al., 2009; Uchida et al., 2011; Lanni et al., 2013). Several recent studies (Iida, 1999; D’Odorico and Fragherazzi, 2003; Talebi et al., 2008) have presented a hydro-geomorphological method that considers a simple hydrological mechanism and bedrock topography within the soil mantle (depth from surface to bedrock). The authors suggested that the limited number of attributes required by this simplified framework favors analyses of how different hydrologic, geomorphic, and geomechanical variables affect landsliding. These studies calculated the return period of shallow landslides over a long timescale using the relationships among slope degree, soil depth, soil strength, and rainfall; however, they did not consider the range of soil strength and calibration values with respect to SSA. As mentioned above, it is highly important to understand the hydrological mechanism to estimate SSAs incurred by extreme rainfall. However, many uncertainties still remain in slope degree, soil depth, soil strength, and so forth. In particular, there are several measurement or estimation methods for soil strength that can have critical effects on the estimation of sudden sediment source area process. Therefore, the aims of this study are to analysis the effects of soil strength as measured using different methods of SSA estimation and an estimation method of soil strength based on a hydrologic concept. The topography-based SHALSTAB model (D’Odorico and Fagherazzi, 2003) was used to achieve the objectives of this study. The practicality of this approach and its various adaptations have been demonstrated over the last decade, and it has been found to perform well for many applications (Dietrich et al., 1995; Wu and Sidle, 1995; Dietrich and Montgomery, 1998; Montgomery et al., 1998; Borga et al., 2002; Vanacker et al., 2003; Fernandes et al., 2004).

2.2. Survey of soil depth at a small hillslope site To survey the soil depth at a small hillslope site, we used a dynamic cone penetrometer (diameter: 25 mm; tip angle: 60°) consisting of several 0.5 m sections of 15 mm diameter stainless steel rods with gradations etched at every 10 cm using 5 kg weigh. Then the soil distribution was calculated using the Nd values (10/penetrated depth) (Yoshinaga and Ohnuki, 1995). Soil depth in the small study site was measured in the vertical gravitational direction, and measurements were performed at 120 points (10–15 m intervals) along the slope (Fig. 3). As the soil depth was measured using the knocking pole test, the location of each measurement point was recorded using a real-time kinematic global positioning system (RTK-GPS: GPT-7001L, Topcon, Tokyo, Japan). Uchida et al. (2009) and Kim et al. (2015) compared vertical number of drops per 10 cm distributions (Nds) among locations outside and inside areas of slope failure and found that soil layers with Nd values of 5–20 did not occur at locations within areas of slope failure. They suggested that Nd values lower than 20 can be used to define soil layers with failure potential, and conversely, that Nd values greater than 20 indicate bedrock layers not prone to failure. 2.3. Determination of soil strength attribute To facilitate estimation of SSAs by shallow landslides, we collected soil strength attributes measured using different approaches: the DST method (Lee et al., 2009), Monte Carlo simulated data (MSD; Park et al., 2013), and triaxial compression test (TST; Kim et al., 2015). Lee et al. (2009) collected 40 soil samples within the study area, as in the present study, and analyzed them using DST. Park et al. (2013) collected 80 soil samples and also analyzed them using DST. After that, they estimated the physical soil strength values using MSD. These values were used as model input attributes to cover the entire study area. Kim et al. (2015) collected soil samples from the study area and tested them using the consolidated drained (CD) test of the TST to determine the model input attributes (Table 1)

2. Materials and methods 2.1. Study area and shallow landslides map The study area, approximately 70 km2, is located in Jinbu-Myeon, Pyeongchang-gun, Kangwon Prefecture, Republic of Korea, and has its center at 37°37′49″ N, 128°33′29″ E (Fig. 1). The annual mean precipitation in this area during the last 30 years (1978–2008) was about 1400 mm (Park et al., 2013). The region has a temperate climate with year-round precipitation. However, rainfall occurs primarily during the summer, from June to September, as part of the East Asian Monsoon. On July 16, 2006, more than 1200 shallow landslides occurred in the Jinbu region as a result of typhoon rain. The total measured rainfall and maximum rainfall intensity of the triggering events were 500 mm/day and 80 mm/h, respectively. The prevalent lithological unit exposed in the study area is granite. However, there is also sedimentary rock composed of fine sandstone with gray sandy shale originating from thick clastic successions of marginal marine to non-marine environments. The granite exists mainly as a large batholith trending NW–SE and as small stocks consisting of granite with minor syenite and diorite distributed along the Ogcheon Belt. All of the previous geological units were intensely deformed as a result of this intense orogenic event (Geological Society of Korea, 1962; Park et al., 2013). Although the study area included two distinct geologies (i.e., granite and sedimentary rock), we considered only the area underlain by granite because most of the shallow landslides occurred there (Fig. 2). The average soil depth in the study area is approximately 1 m, and shallow landslide scars in the soil range from 9 to 30 m in width and from 10 to 200 m in length. To obtain SSAs of the shallow landslides, digital aerial photographs with a ground resolution of 0.5 m were collected and used. Imagery before the shallow landsliding was taken on May 27, 2006, and imagery after it was taken on September 6, 2010. Using these two different sets of images, a map of SSAs was constructed. Some shallow landslide areas could not be captured by aerial imagery because of the presence of

2.4. Model description 2.4.1. SHALSTAB model To estimate SSAs, the SHALSTAB model combined with an infinite slope stability equation and a simple subsurface flow model was used (Fig. 4a). The calculation of Rcri (slope stability index; Eq. (1)) is based on the infinite slope form of the Mohr–Coulomb failure law, which is expressed by following equation (Montgomery and Dietrich, 1994; Dietrich and Montgomery, 1998; Zizioli et al., 2013).

ρ (sinβ − Ct ) ⎤ b R cri = Tsinβ ⎛ ⎞ ⎜⎛ s ⎟⎞ ⎡1 − (cosβtan∅) ⎥ a ρ ⎝ ⎠⎝ w ⎠⎢ ⎣ ⎦

(1)

where T is the saturated soil transmissivity (m /h), β is the local slope angle (°), ∅ is the internal friction angle of the soil (°), a is the upslope contributing area (m2), b is the unit contour length (m), ρs is the wet soil bulk density (g cm−3), and ρw is the density of water (g cm−3). C is the combined cohesion term (−), which is made dimensionless relative to the vertical soil depth and is defined as 2

C=

Cr + Cs hρs g

where Cr is the root cohesion (Nm 3

(2) −2

), Cs is the soil cohesion (Nm

−2

), h

Quaternary International xxx (xxxx) xxx–xxx

M. Kim et al.

Fig. 2. Geological map of the study area.

Dcr = Ct /cos2β{γsat (tanβ −tan∅) + γw tan∅}

is the vertical soil depth (m), and g is the constant of gravitational acceleration (9.81 ms−1). The stability index is expressed in mm/day of critical rainfall and is of variable scale, where lower values indicate a higher probability of instability and higher values indicate a greater probability of stability. This scale also encompasses areas identified as unconditionally stable and unconditionally unstable based on the estimated rainfall value (Zizioli et al., 2013).

When the soil depth D is equal to Dmax (completely saturated soil), the critical soil depth Dmax can be expressed as follows:

Dmax = Ct /γunsatcos2β(tanβ −tan∅)

Ct − γunsatcos2β(tanβ −tan∅)D 2 cos β{(γsat − γunsat)(tanβ −tan∅) + γw tan∅}

(5)

The modeling of soil evolution is important, because without cohesion, soils could never form on slopes greater than ∅, and even thin soils on slopes in the range β < ∅ could be extremely unstable because light rainfall could provide a sufficient saturated water depth (H) to cause landslides. These scenarios are contrary to our observations, which suggests that soil cohesion must be considered in slope stability models (Iida, 1999). Using equations (4) and (5), the SSAs of shallow landslides were restricted to the area within the curves of Dcr and Dmax shown in Fig. 5. The Dcr curve of the completely unsaturated state and the Dmax curve of the completely saturated state (Fig. 5) are integral for determining the optimal values of cohesion and internal friction angle. We can control the Dcr and Dmax curves by controlling the internal friction angle, particularly cohesion. When controlling the curves of Dcr and Dmax, the distribution of soil depth against slope angle is a very important input attribute for calculating the two soil attributes.

2.4.2. Soil strength using a subsurface hydrologic concept (SHC) Topography influences the initiation of SSAs through both the level of subsurface flow and the effects of slope gradient on slope stability (Montgomery and Dietrich, 1994; Talebi et al., 2008). Slope failure often occurs in areas of convergent topography, where subsurface soil water-flow paths increase the excess pore water pressure downslope (Anderson and Reimer, 1995; Wilkinson et al., 2002; Talebi et al., 2008). Planar infinite slope analyses have been applied widely to the evaluation of natural slope stability, particularly where the depth of the soil mantle is small relative to the slope length and where SSAs have occurred due to the failure of a soil mantle overlying a sloping drainage barrier (Talebi et al., 2008). Iida (1999) suggested that the two-layer model of soil and bedrock, which assumes a potential soil failure layer, is suitable for slope stability analyses of SSAs (Fig. 4b). Several researchers (Iida, 1999; D’Odorico and Fragherazzi, 2003; Talebi et al., 2008) have applied the same approach in a stochastic HGM model for shallow landslides resulting from rainstorms, and the failure condition can be expressed as

Hcr =

(4)

2.4.3. Evaluation of predictive accuracy Each simulated result was evaluated using receiver operating characteristic (ROC) analyses because the accuracy of estimation of regional SSAs of shallow landslides is typically evaluated by comparing the locations of known SSAs with simulation results from the model (Montgomery et al., 1998, 2001; Godt et al., 2008). The ROCs, which are used in various types of research (such as weather forecasting and landslide susceptibility mapping), provide a technique for comparing the performance of models whose results can be assigned to one of two classes or states (Swets, 1988; Fawcett, 2006; Van Den Eeckhaut et al., 2006; Godt et al., 2008). Fig. 6 shows the equations for calculating model performance accuracy. Comparisons of model performance in an ROC graph show the ratio of successes (true positives) to overpredictions (false positives). Points located toward the upper left of the graph are generally considered superior, with a perfect prediction being located in the upper left corner (0, 1). Prediction results that fall along the dashed line in Fig. 6 are considered random because they predict instability within and outside

(3)

where Hcr is the critical depth H of the saturated throughflow depth for shallow landsliding, β is the local slope angle (°), Ct is the cohesion (kN/ cm3), D is the soil depth (m), ∅ is the internal friction angle (°), γunsat is the weight per unit volume of unsaturated soil (g/cm3), γsat is the weight per unit volume of saturated soil (g/cm3), and γw is the weight of water per unit volume (g/cm3). In this model, shallow landslides occur when the soil depth D is between Dcr and Dmax. When the soil depth D is equal to Dcr (completely unsaturated soil), the critical soil depth Dcr can be expressed as 4

Quaternary International xxx (xxxx) xxx–xxx

M. Kim et al.

Fig. 3. Soil depth survey in a small hillslope area within the study area.

because the mapped SSAs of shallow landslides occurred only in mountainous areas (Godt et al., 2008).

Table 1 Soil attributes for shallow landslide modeling. Model input parameters

Case I-a

Case I-b

Case I-c

Saturated soil weight (kg/m3) Dry density (kg/m3) Water density (kg/m3) Hydraulic conductivity (m/h) Cohesion (kPa) Internal Friction Angle (°) Slope degree (°) Average soil depth (m)

1790 1550 1000 0.08 3.8 35.2 calculated by DEM 1m

1960 1510 1000 0.04 4 34

1740 1490 1000 0.05 1.6 36.5

1m

1m

3. Results 3.1. SSA using experimental dataset (case I) We analyzed the three soil attributes (Table 1) used in the shallow landslide model to quantify the spatial discrepancy between the landslides triggered by the July 2006 rainfall event and the prediction results for the critical rainfall level that would trigger a shallow landslide in the area (Fig. 7). The resulting map of the steady state critical rainfall (mm/day) that could trigger shallow landslides in the study area is shown in Fig. 7, where the shallow landslide-prone areas predicted by equation (1) are delineated by the steady state rainfall intensity (mm/ day) necessary for slope instability in each topographic element. We compared this prediction map with the shallow landslides observed through aerial photography. In the maps shown in panels (a) and (b) of Fig. 7, which are based on DST and MST, respectively, the steady state critical rainfall varies between 50 and 500 mm/day (except in the stable area) and shallow

the mapped landslides at the same rate. An acceptable prediction requires TPR/FPR > 1, where TPR and FPR represents the true positive rate and false positive rate (Fawcett, 2006; Baum et al., 2010; Rossi et al., 2010; Raia et al., 2014). More details of acceptable prediction rate can be referred to Fawcett (2006) or Baum et al. (2010). The least critical test of prediction accuracy would be to count a successful prediction when a single grid cell is located within mapped polygons of SSAs of shallow landslides. Relatively flat areas, such as rivers, alluvium, and rice paddies, were excluded from the analyses 5

Quaternary International xxx (xxxx) xxx–xxx

M. Kim et al.

Fig. 4. ROC analyses method for determining the accuracy of shallow landslide prediction in this study: SHALSTAB (a) and a hydro-geomorphology conceptual model (b).

Fig. 5. Relationship between soil depth and slope angle using a simple hydrologic concept model (equations (6) and (7)). Shallow landslides can occur between Dcr and Dmax. White circles indicate the soil depth as measured using knocking pole tests outside of the shallow landslide area, and black circles indicate soil depth as measured using knocking pole tests inside the shallow landslide scar.

Fig. 6. Calculation of accuracy using ROC analyses for model performance (modified from Godt et al., 2008).

ROC analyses were performed to evaluate the accuracy of the modeling results (Fig. 7d). The ROC analyses results are presented in Table 2. The calculated accuracy values for Cases I-a, I-b, and I-c were 0.71, 0.73, and 0.48, respectively. The TPR/FPR ratios, for which a larger value indicates a better predictive performance, were 1.34, 1.43, and 1.13. The ROC anlyses results are presented in Table 2. The ROC analyses indicated low overall model accuracy, with the accuracy of Case I-c (0.48) being the lowest.

landslides occur near the high mountains in the simulated area. The landslide areas surveyed in 2006 showed a pattern similar to the critical rainfall simulation results, but the simulation was prone to overprediction, particularly in high mountain areas. Landslide occurrence based on TCT data also indicated critical rainfall levels of 50–500 mm/ day, except in the stable area (Fig. 7c). However, landslide occurrence was sensitive to rainfall levels of 50–200 mm/day, and most areas could be exposed to landslide risk according to comparison with mapped shallow landslides. Furthermore, in all three cases, the steep slope of the study area indicated a relatively high possibility of a landslide being triggered by a critical rainfall level of 0–50 mm/day. The steady state critical rainfall (mm/day) data from the simulated grid data (Fig. 8) were separated to obtain the distribution of steady state critical rainfall levels (mm/day) from the three types of experimental soil data (Case I). Cases I-a and I-b show a large stable area (stable cells and cells > 500 mm/day), while Case I-c shows less than half the number of stable cells. The low values for Case I-c indicate that the TCT soil data were more sensitive than the DST soil data.

3.2. Estimation of soil strength attributes The shallow landslide distribution in the study area was determined using equations (4) and (5) to calculate physical soil attributes such as cohesion and internal friction angle. Equation (5) assumes that shallow landslides only occur when the soil is completely saturated. These equations together set the limits on the depth–slope space in which landslides can occur (Fig. 9). Fig. 9 shows the results of the soil strength estimations, which reflect the effects of soil depth (i.e., the measured soil depth in Fig. 3 and 6

Quaternary International xxx (xxxx) xxx–xxx

M. Kim et al.

Fig. 7. Maps of shallow landslide prediction using three soil strength attributes.

Fig. 8. Number of simulated cells and distribution of critical rainfall intensity (mm/day) for Case I, Case II, and Case III based on the three soil attributes in Table 1. 7

Quaternary International xxx (xxxx) xxx–xxx

M. Kim et al.

than those listed in Table 1. When the measured soil distribution was applied to the HGM model with each soil strength shown in Table 1, the soil strength values were increased relative to the average soil distribution. In particular, the soil cohesion value determined by TCT (Table 1) was almost doubled due to the influence of soil depth (Table 3). Based on this result, we separated the modeling cases based on the effects of soil strength in relation to various simple hydrological processes that were assigned according to changes in soil depth. Case II reflected the measured soil depth distribution and Case III reflected the average soil depth distribution. Because a soil depth of 1 m was used as a topographic input attribute in this study, it is necessary to evaluate the effects of soil depth. We carried out shallow landslide prediction using the soil strength value of each case and evaluated each prediction result using ROC analyses.

Table 2 Accuracy analyses of shallow landslide prediction using ROC analyses. Model Class

True Positive Rate

False Positive Rate

TPR/FPR

Accuracy

Case I-a Case I-b Case I-c

0.37 0.35 0.65

0.28 0.26 0.52

1.34 1.34 1.13

0.71 0.73 0.48

the average soil depth of 1 m), obtained using equation (4) (Dcr) and equation (5) (Dmax) in the HGM model derived from the soil attributes in Table 1. Fig. 9a (Case I) shows an overlap with the distribution of measured soil depth (y-axis) against slope angle (x-axis) (black circles are located within the shallow landslide scar and white circles are located outside the shallow landslide scar). The curves Dcr and Dmax were calculated using the data in Table 1. However, some black circles are located outside Dcr and Dmax, which indicates that the cohesion and internal friction values used in the Case I simulation of shallow landslide prediction in Fig. 7 contain errors. Accordingly, we changed the internal friction angle and cohesion values, resulting in values nested between Dcr and Dmax in Fig. 9 for Cases II and III. Case II shows an overlap with the distribution of measured soil depth (y-axis) against slope angle (x-axis), and the curves were controlled via adjustments made to the internal friction angle and cohesion. Case III also shows an overlap with the distribution of the average soil depth (1 m) (y-axis) against the slope angle (x-axis), and the curves were again controlled via adjustments to the internal friction angle and cohesion. All values calculated by the HGM model are presented in Table 3. The internal friction angle and cohesion values in Table 3 are greater

3.3. SSA using estimated soil strength (case II) To evaluate the effects of soil strength in relation to hydrological processes and measured soil depth, we applied altered soil attributes to the SHALSTAB model and recalculated the steady state critical rainfall (mm/day) for shallow landslide prediction. We also evaluated the accuracy of this shallow landslide prediction using ROC analyses to determine any improvement in prediction accuracy compared to the results for Case I. Fig. 10 shows the distribution of the areas with rainfall values critical for the occurrence of shallow landslides. Case II-a and Case II-b indicate the critical rainfall distribution using soil attributes from DST and MST, respectively, and incorporate the effects of the measured soil

Fig. 9. Estimation of soil strength using a simple subsurface hydrologic concept. 8

Quaternary International xxx (xxxx) xxx–xxx

M. Kim et al.

Table 3 Soil property calculation results using a subsurface hydrologic concept. Model input parameters

Case II-a

Case II-b

Case II-c

Case III-a

Case III-b

Case III-c

Saturated soil density (kg/cm3) Dry density (kg/cm3) Water density (kg/cm3) Hydraulic conductivity (m/h) Cohesion (kPa) Internal Friction Angle (°) Slope degree (°) soil depth (m)

1790 1550 1000 0.08 4.5 37 calculated by DEM Measured

1960 1510 1000 0.08 3.2 34

1740 1490 1000 0.04 5.2 37.2

1790 1550 1000 0.04 3.5 37

1960 1510 1000 0.05 3.2 34

1740 1490 1000 0.05 3.2 37.2

Measured

Measured

Average 1 m

Average 1 m

Average 1 m

Case II: Measured soil depth data by knocking pole test in Fig. 3. Case III: Average soil depth data using measured soil thickness by knocking pole test in Fig. 3.

shows the distribution of simulated critical rainfall (mm/day) based on the results for Case II-c, which follows a pattern similar to the aerial image grid (red color). However, the critical rainfall simulation results differ in that the critical rainfall value increased to more than 500 mm/ day. The distribution of critical rainfall matched well with the shallow landslide locations of 2006. The critical rainfall simulation distributions factoring in soil attributes reflected the effects of measured soil depth in Case II. The distribution patterns of critical rainfall cells in Cases II-a and II-b were similar, and the number of stable cells increased (in contrast to Cases I-a and I-b) when using the DST and MST methods. The distribution of critical rainfall cells in Case II-c showed that stable cells and cells >

depth distribution. Case II-c depicts the critical rainfall distribution using soil attributes from TCT and incorporates the effects of the measured soil depth distribution (Table 3). The critical rainfall simulation results in Fig. 8 were compared to the shallow landslide grid derived from aerial images taken during a post-event survey of the study area in 2006. The distribution of the simulated critical rainfall (mm/day) for Cases II-a and II-b followed a pattern similar to the aerial image grid (red color). The simulated critical rainfall values show an increase in the area of unconditionally stable cells. The critical rainfall values were 0–50 mm/day, which showed decreased sensitivity to shallow landslide occurrence compared to the cases depicted in Fig. 8a and b. Fig. 10c

Fig. 10. Maps of the simulation results showing the shallow landslide-prone area based on the steady state rainfall intensity (mm/day) necessary for slope instability using the soil strength attributes reflecting the distribution of measured soil depth. 9

Quaternary International xxx (xxxx) xxx–xxx

M. Kim et al.

shallow landslides are highly related to rainfall and soil strength via subsurface flow mechanisms. There are many methods for measuring soil strength for use as an input attribute in SSA area of shallow landslide analyses. In the present study, we evaluated several types of soil strength measurements (Table 1), as soil strength is sensitive to many factors and is a most difficult attributes to quantify in shallow landslide prediction. The accuracy of the prediction results, as determined by ROC analyses, was characterized by overprediction and excessively low values (0.45–0.71) in Case I. The low accuracy was due mainly to the internal friction angle, particularly cohesion. As explained above, soil strength is affected by subsurface flow in the saturated soil mantle. Therefore, we considered SHC and calculated two soil strength values, which were based on the distribution of soil depth and the average soil depth measured at a small hillslope site within the study area. Based on SHC, the internal friction angle, particularly the cohesion, is critical for estimating SSAs of shallow landslides. Furthermore, when considering the measured soil depth, the soil strength values were higher than those for average soil depth. This may be related to the range of saturated and unsaturated zones within the soil depth. Some researchers (Pellet et al., 2013; Wen and Yan, 2014; Li, 2018) mentioned that the strength parameter c (cohesion) is more sensitive than φ (internal friction angle) to change in moisture content due to increasing water content. Epecially, Pellet et al. (2013) mentioned the shear strength can be considerably reduced to approximately 50% of its original value. It is likely that with increasing moisture content both influence the strength parameters to different degrees at different moisture contents (Goldwater, 1990; Wen and Yan, 2014). Thus, studies examining the relationship between loess shear strength and infiltration capacity by rainfall has highlighted the significant influence of reduction of soil suction on shear strength (Tsai and Wang, 2011; Tu et al., 2009; Zhou et al., 2012; Yates et al., 2018). In Case II (Fig. 10), soil strength, in relation to the distribution of soil depth against slope, had higher accuracy (0.84–0.85) than in Case I. In the evaluation of soil strength in relation to average soil depth (Case III; Fig. 11), the results were also more accurate (0.81–0.82) than those of Case I, but not Case II. In both cases (Cases II and III), the use of soil strength calculated via SHC improved shallow landslide prediction. This indicates that estimating soil strength using soil depth plays an important role in improving the accuracy of SSA estimation; therefore, soil attributes (i.e., internal friction angle and soil cohesion) are important in the estimation of SSAs of shallow landslides.

Table 4 Accuracy analyses of shallow landslide prediction using estimated soil strength. Soil thickness data

Simulation case

True Positive Rate

False Positive Rate

TPR/FPR

Accuracy

measured soil thickness (m) average soil thickness (1 m)

CASE CASE CASE CASE CASE CASE

0.3 0.16 0.32 0.27 0.3 0.34

0.16 0.13 0.14 0.15 0.2 0.21

1.95 1.28 2.37 1.78 1.5 1.6

0.83 0.86 0.85 0.84 0.79 0.78

II-a II-b II-c III-a III-b III-c

500 mm/day increased (in contrast to Case I-c) when using TCT. This indicates that the overprediction of shallow landslide occurrence under light rainfall conditions decreased. The accuracy of the simulated critical rainfall prediction for Case II was evaluated by ROC analyses. The ROC accuracy values for Cases II-a, II-b, and II-c were 0.83, 0.86, and 0.85, respectively. The TPR/FPR ratios were 1.95, 1.28, and 2.37 for Cases I-a, I-b, and I-c, respectively. All simulation cases were clearly improved over those of Case I (Fig. 7). The ROC analyses results are presented in Table 4. The increased ROC values for Case II indicate improved predictive accuracy for the critical rainfall simulation. This also demonstrates that soil cohesion and internal friction angle are important factors for shallow landslide modeling. 3.4. SSA using estimated soil strength (case III) We used average soil depth distribution as an input attribute for shallow landslide prediction across the entire study area. Soil strength calculated using a constant soil depth of 1 m was used as an input attribute to the SHALSTAB model (Case III). The critical rainfall simulation distribution for the effects of a soil depth of 1 m on shallow landslide occurrence is shown in Fig. 11. The distribution of the critical rainfall values in Case III-a is similar to that of Case II-a, but it differs from the other cases. The values of < 50 mm/day and 100–200 mm/ day were low, different from the sensitivity shown in Fig. 8. The low sensitivities of Cases III-b and III-c indicate that soil cohesion and internal friction angle can still influence the outcome of the critical rainfall simulation, despite a constant average soil depth of 1 m. The distribution of simulated critical rainfall cells is presented in Fig. 8. The distribution patterns for Cases III-a and III-c are similar. However, the critical rainfall cells in Case III-b show that the distribution of < 50 mm/day cells increased while that of stable cells decreased compared to the other simulations. The critical rainfall simulation accuracy of the cases shown in Fig. 11 was evaluated by ROC analyses, and the accuracy values for Cases II-Ia, III-b, and III-c were 0.84, 0.79, and 0.78, respectively. The TPR/FPR ratios for Cases I-a, I-b, and I-c were 1.78, 1.50, and 1.60, respectively (Fig. 7d). The accuracy values for Case III were higher than those for Case I (Fig. 7) but lower than those for Case II (Fig. 10). Therefore, the accuracy of the shallow landslide prediction based on ROC analyses were in the order Case II > Case III > Case I.

4.2. Other factors related to SSAs and further study Figs. 10 and 11 show the estimation results of SSAs of shallow landslides obtained using the calculated soil strength representing the effects of measured (Case II) and average (Case III) soil depth, respectively. The estimation of SSA performed using the estimated soil strength was more accurate than that based on various other soil strength values (Case I) (Fig. 7). The order of shallow landslide prediction accuracy, based on ROC analyses, was Case II > Case III > Case I. Although we performed soil attributization for shallow landslide prediction, overprediction was still detected (Figs. 10 and 11). We used a 5-m-resolution DEM with a 5 m contour digital map to estimate SSAs of shallow landslides, but we were unable to investigate the effects of DEM resolution. However, several authors (Zhang and Montgomery, 1994; Claessens et al., 2005; Penna et al., 2014) have previously analyzed the effects of DEM on modeling results. Claessens et al. (2005) calculated steady state critical rainfall using different DEM resolutions. They observed that models could not describe the characteristics of the natural hillslope in detail, even when using high-resolution DEMs, such as LIDAR data, for shallow landslide prediction modeling. Many previous studies (Claessens et al., 2005; Tarolli and Tarboton, 2006) have concluded that complete prediction of the shallow landslide area is difficult, even if high-resolution DEMs are

4. Discussion 4.1. Effects of soil strength in relation to subsurface hydrological mechanisms Shallow landslides occur when the strength of soil fails to resist the acting stress developed by increasing groundwater levels in the soil mantle during extreme rainfall events. Soil strength is inherently necessary for soil to accumulate in steep hollows; otherwise, landslides could occur even with light rainfall (Schmidt et al., 2001; D’Odorico and Fagherazzi, 2003; Rosso et al., 2006). Consequently, SSAs of 10

Quaternary International xxx (xxxx) xxx–xxx

M. Kim et al.

Fig. 11. Maps of the simulation results showing the shallow landslide-prone area based on the steady state rainfall intensity (mm/day) necessary for slope instability using the soil strength attributes reflecting the average soil depth (1 m).

5. Conclusions

used to analyze shallow landslides. In addition, Milledge et al. (2012) discussed the infinite length assumption for DEM resolution, which incorporates a length/depth (L/H) ratio using the safety factor from the finite element method. They also used the SHALSTAB model and compared it with a coarse DEM (10 m) and a fine DEM (1 m) for verification of their assumptions. They explained that the infinite length assumption is valid for most modeling studies, which have used relatively coarse resolutions, but noted that using finer-grid resolutions dramatically reduces the predicted width of landslides. Thus, the limited validity of the assumption may cause underprediction of upslope landslides, which generally have a somewhat smaller L/H ratio. The L/H ratios of shallow landslides in our study site were in the range of 10–200; the shortest length was 10 m, the longest length was 200 m, and the failure depth was approximately 1 m. Our results were obtained under the assumption of infinite slope length; thus, large watersheds with long slopes may perform better with coarser resolutions than small watersheds with short slopes. Although the hydrological attributes in a slope stability model used for analyzing SSAs can be obtained from field measurements or laboratory analyses, some of these attributes are difficult to define in practice, particularly with regard to their spatial variation (Dietrich et al., 1995). Therefore, further research on spatial variation in slope stability and alternative mechanisms for estimating SSAs is required.

SSAs of shallow landslides are very important in many fields for analyzing landscape development, geomorphic development, Quaternary environment, and so on because SSAs causes continuous topographic changes. The soil strength is one of the most fundamental elements for analyses of occurrence of SSAs, such as those of shallow landslides; however, it is typically necessary to carry out soil strength tests in soil layers with conditions between saturated and unsaturated due to local conditions. The present study considered the effects of soil strength, tested by several methods, in relation to shallow landslide prediction. In evaluating the effects of soil strength (Case I), predictive accuracy (as determined based on ROC) was relatively low. This was mainly due to the misinterpretation of local conditions, such as internal friction angle, and cohesion in particular. Therefore, we considered a subsurface hydrological mechanism for the soil strength data used in Case I. We applied these attributes to a simple hydrologic concept model using two soil depths (i.e., measured soil depth and average soil depth) and slope angle. Considering the subsurface flow mechanism using the measured soil depth (Case II), the soil strength values showed strong predictive results. In Case III, the soil strength calculated for average soil depth (1 m) also generated good predictive results, but it did not outperform Case II. Therefore, combining field survey data with the hydraulic properties related to soil strength may be useful for estimating SSAs of shallow landslides. 11

Quaternary International xxx (xxxx) xxx–xxx

M. Kim et al.

Acknowledgments

Korean). Li, Y., 2018. A review of shear and tensile strengths of the Malan Loess in China. Eng. Geol. 236, 4–10. https://doi.org/10.1016/j.enggeo.2017.02.023. Milledge, D.G., Griffiths, D.V., Lane, S.N., Warburton, J., 2012. Limits on the validity of infinite length assumptions for modelling shallow landslides. Earth Surf. Process. Landforms 37, 1158–1166. Minder, J.R., Roe, G.H., Montgomery, D.R., 2009. Spatial patterns of rainfall and shallow landslide susceptibility. Water Resour. Res. 45, W04419. https://doi.org/10.1029/ 2008WR007027. Mondini, A.C., Guzzeti, F., Reichenbach, P., Rossi, M., Cardinali, M., Ardizzone, F., 2011. Semi-automatic recognition and mapping of rainfall induced shallow landslides using optical satellite images. Remote Sens. Environ. 115 (7), 1743–1757. Montgomery, D.R., Dietrich, W.E., 1994. A physically based model for the topographic control on shallow landsliding. Water Resour. Res. 30, 1153–1171. Montgomery, D.R., Sullivan, K., Greenberg, M.H., 1998. Regional test of a model for shallow landsliding. Hydrol. Process. 12, 943–955. Montgomery, D.R., Balco, G., Willett, S.D., 2001. Climate, tectonics, and the morphology of the Andes. Geology 29, 579–582. https://doi.org/10.1130/0091–7613(2001) 029< 0579:CTATMO>2.0.CO;2. Montrasio, L., Valentino, R., Losi, G.L., 2011. Towards a real-time susceptibility assessment of rainfall-induced shallow landslides on a regional scale. Nat. Hazards Earth Syst. Sci. 11https://doi.org/10.5194/nhess-11–1927–2011. 1927–1947. Montrasio, L., Valentino, R., Losi, G.L., 2012. Shallow landslides triggered by rainfalls: modeling of some case histories in the reggiano apennine (emilia romagna region, northern Italy). Nat. Hazards 60, 1231–1254. Pack, R.T., Tarboton, D.G., Goodwin, C.N., 1998. Terrain Stability Mapping with SINMAP, Technical Description and Users Guide for Version 1.00. Report and software available at: http://www.engineering.usu.edu/dtarb/ (last access: 15 June 2007). Park, H.J., Lee, J.H., Woo, I.K., 2013. Assessment of rainfall-induced shallow landslide susceptibility using a GIS-based probabilistic approach. Eng. Geol. 161, 1–15. Penna, D., Borgal, M., Aronica, G.T., Brigandì, G., Tarolli, P., 2014. The influence of grid resolution on the prediction of natural and road-related shallow landslides. Hydrol. Earth Syst. Sci. 18, 2127–2139. https://doi.org/10.5194/hess-18–2127–2014. Pellet, F.L., Keshavarz, M., Boulon, M., 2013. Influence of humidity conditions on shear strength of clay rock discontinuities. Eng. Geol. 157, 33–38. Raia, S., Alvioli, M., Rossi, M., Baum, R.L., Godt, J.W., Guzzetti, F., 2014. Improving predictive power of physically based rainfall-induced shallow landslide models: a probabilistic approach. Geosci. Model Dev. (GMD) 7, 495–514. Rossi, M., Guzzetti, F., Reichenbach, P., Mondini, A., Peruccacci, S., 2010. Optimal landslide susceptibility zonation based on multiple forecasts. Geomorphology 114, 129–142. Rosso, R., Rulli, M.C., Vannucchi, G., 2006. A physically based model for the hydrologic control on shallow landsliding. Water Resour. Res. 42, W06410. https://doi.org/10. 1029/2005WR004369. Schmidt, K.M., Roering, J.J., Stock, J.D., Dietrich, W.E., Montgomery, D.R., Shaub, T., 2001. Root cohesion variability and shallow landslide susceptibility in the Oregon Coast Range. Can. Geotech. J. 38, 995–1024. Simoni, S., Zanotti, F., Bertoldi, G., Rigon, R., 2008. Modeling the probability of occurrence of shallow landslides and channelized debris flows using GEOtop-FS. Hydrol. Process. 22, 532–545. Swets, J.A., 1988. Measuring the accuracy of diagnostic systems. Science 240, 1285–1293. Talebi, A., Troch, P.A., Uijlenhoet, R., 2008. A steady-state analytical hillslope stability model for complex hillslopes. Hydrol. Process. 22, 546–553. Tarolli, P., Tarboton, D.G., 2006. A new method for determination of most likely landslide initiation points and the evaluation of digital terrain model scale in terrain stability mapping. Hydrol. Earth Syst. Sci. 10, 663–677. Trustrum, N.A., Gomez, B., Page, M.J., Reid, L.M., Hicks, D.M., 1999. Sediment production, storage and output: the relative role of large magnitude events in steepland catchments. Z. Geomorphol. - Suppl. 115, 71–86. Tsai, T.L., Wang, J.K., 2011. Examination of influences of rainfall patterns on shallowlandslides due to dissipation of matric suction. Environ. Earth Sci. 63, 65–75. https:// doi.org/10.1007/s12665-010-0669-1. Tu, X.B., Kwong, A.K.L., Dai, F.C., Tham, L.G., Min, H., 2009. Field monitoring of rainfall infiltration in a loess slope and analysis of failure mechanism of rainfall-induced landslides. Eng. Geol. 105, 134–150. https://doi.org/10.1016/j.enggeo.2008.11.011. Uchida, T., Mori, N., Tamura, K., Takiguchi, S., Kamee, K., 2009. The role of data preparation on shallow landslide prediction. J the Jap Soc Eros Cont Engi 62, 23–31. Uchida, T., Tamur, K., Akiyama, K., 2011. The role of grid cell size, flow routing algorithm and spatial variability of soil depth on shallow landslide prediction. Ital. J Eng Geol and Environ 149–157. https://doi.org/10.4408/IJEGE.2011–03.B-018. Vanacker, V., Vanderschaeghe, M., Govers, G., Willems, E., Poesen, J., Deckers, J., De Bievre, B., 2003. Linking hydrological, infinite slope stability and land-use change models through GIS for assessing the impact of deforestation on slope stability in high Andean watersheds. Geomorphology 52, 299–315. Viet, T.T., Lee, G., An, H., Kim, M., 2016. Shallow landslide assessment considering the influence of vegetation cover. J. Kor. Geoenviron. Soc. 17 (4), 17–31. Viet, T.T., Lee, G., An, H., Kim, M., 2017. Comparing the performance of TRIGRS and TiVaSS in spatial and temporal prediction of rainfall-induced shallow landslides. Environ. Earth Sci. 76 (8), 315. Wen, B.P., Yan, Y.J., 2014. Influence of structure on shear characteristics of the unsaturated loess in Lanzhou, China. Eng. Geol. 168, 46–58. https://doi.org/10. 1016/j. enggeo.2013.10.023. Wilkinson, P.L., Anderson, M.G., Lloyda, D.M., Renaudb, J.P., 2002. Landslide hazard and bioengineering: towards providing improved decision support through integrated

This research was supported by a Basic Research Project of the Korea Institute of Geoscience and Mineral Resources (KIGAM) funded by the Ministry of Science and ICT (MSI) and, Korea Institute of Planning and Evaluation for Technology in Food, Agriculture and Forestry (IPET) through Advanced Production Technology Development Program, funded by Ministry of Agriculture, Food and Rural Affairs (MAFRA) (grant number: 318060-3). References An, H., Kim, M., Lee, G., The viet, T., 2016. Survey of spatial and temporal landslide prediction methods and techniques. Kor. J. Anim. Sci. 43, 507–521. https://doi.org/ 10.7744/kjoas.20160053. Anderson, S.A., Reimer, M.F., 1995. Collapse of saturated soil due to reduction in confinement. J. Geotech. Engi. 121, 216–220. Baum, R.L., Savage, W.Z., Godt, J.W., 2008. TRIGRS–A FORTRAN Program for Transient Rainfall Infiltration and Grid-based Regional Slope Stability Analysis, Version 2.0. U.S. Geological Survey Open-File Report 2008–1159. Baum, R.L., Godt, J.W., Savage, W.Z., 2010. Estimating the timing and location of shallow rainfall-induced landslides using a model for transient, unsaturated infiltration. J. Geophys. Res. 115. https://doi.org/10.1029/2009JF001321. Borga, M., Dalla Fontana, G., Gregoretti, C., Marchi, L., 2002. Assessment of shallow landsliding by using a physically based model of hillslope stability. Hydrol. Process. 16, 2833–2851. Caine, N., 1980. The rainfall intensity-duration control of shallow landslides and debris flows. Geog. Ann. Series A. 62, 23–27. Caine, N., Swanson, F.J., 1989. Geomorphic coupling of hillslope and channel systems in two small mountain basins. Z. Geomorphol. 33, 189–203. Carrara, A., Cardinali, M., Detti, R., Guzzetti, F., Pasqui, V., Reichenbach, P., 1991. GIS techniques and statistical models in evaluating landslide hazard. Earth Surf. Process. Landforms 16, 427–445. Claessens, L., Heuvelink, G.B.M., Schoorl, J.M., Veldkamp, A., 2005. DEM resolution effects on shallow landslide hazard and soil redistribution modeling. Earth Surf. Process. Landforms 30, 461–477. Dietrich, W.E., Montgomery, D.R., 1998. SHALSTAB- a Digital Terrain Model for Mapping Shallow Landslide Potential. National Council of the Paper Industry for Air and Stream Improvement, Technical Report, Berkeley February 29. Dietrich, W.E., Reiss, R., Hsu, M., Montgomery, D.R., 1995. A process-based model for colluvial soil depth and shallow landsliding using digital elevation data. Hydrol. Process. 9, 383–400. Duan, J., Grant, G.E., 2000. Shallow landslide delineation for steep forest watersheds based on topographic attributes and probability analysis. In: Wilson, J.P., Gallant, J.C. (Eds.), Terrain Analysis—principles and Applications. John Wiley & Sons, New York, pp. 311–329. D'Odorico, P., Fagherazzi, S., 2003. A probabilistic model of rainfall-triggered shallow landslides in hollows: a long-term analysis. Water Resour. Res. 39, 1262. Fawcett, T., 2006. An introduction to ROC analysis. Pattern Recogn. Lett. 27, 861–874. Fernandes, N.F., Guimarães, R.F., Gomes, R.A.T., Vieira, B.C., Montgomery, D.R., Greenberg, M.H., 2004. Topographic controls of landslides in Rio de Janeiro: field evidence and modeling. Catena 55, 163–181. Geological Society of Korea, 1962. Changdong–hajinburi Geological Map Sheet. KIGAM. Giannecchini, R., Naldini, D., D'Amato Avanzi, G., Puccinelli, A., 2007. Modelling of the initiation of rainfall-induced debris flows in the Cardoso basin (Apuan Alps, Italy). Quat. Int. 171, 108–117. Godt, J.W., Baum, R.L., Savage, W.Z., Salciarini, D., Schulz, W.H., Harp, E.L., 2008. Transient deterministic shallow landslide modeling: requirements for susceptibility and hazard assessments in a GIS framework. Eng. Geol. 102, 214–226. Godt, J.W., Baum, R.L., Lu, N., 2009. Landsliding in partially saturated materials. Geophys. Res. Lett. 36, L02403. https://doi.org/10.1029/2008GL035996. Goldwater, S., 1990. Slope Failure in Loess, a Detailed Investigation, Allendale, Banks Peninsula. MSc Thesis. Department of Geological Sciences, University of Canterbury. Hovius, N., Stark, C.P., Allen, P.A., 1997. Sediment flux from a mountain belt derived by landslide mapping. Geology 25, 231–234. Iida, T., 1999. A stochastic hydro-geomorphological model for shallow landsliding due to rainstorm. Catena 34, 293–313. Iverson, R.M., 2000. Landslide triggering by rain infiltration. Water Resour. Res. 36 (6), 1897–1910. Kim, D., Im, S., Lee, C., Woo, C., 2013. Modeling the contribution of trees to shallow landslide development in a steep, forested watershed. Ecol. Eng. 61, 658–668. Kim, M.S., Onda, Y., Kim, J.K., Kim, S.W., 2015. Effect of topography and soil parameterisation representing soil thicknesses on shallow landslide modelling. Quat. Int. 384, 91–106. Kim, M.S., Onda, Y., Uchida, T., Kim, J.K., 2016. Effects of soil depth and subsurface flow along the subsurface topography on shallow landslide predictions at the site of a small granitic hillslope. Geomorphology 271, 40–54. Lanni, C., McDonnell, J., Hopp, L., Rigon, R., 2013. Simulated effect of soil depth and bedrock topography on near-surface hydrologic response and slope stability. Earth Surf. Process. Landforms 38, 146–159. Lee, M.S., Ryu, J.C., Kim, K.S., 2009. Development of the linear regression analysis model to estimate the shear strength of soils. The J. Eng. Geol. in Korea 19, 177–189 (in

12

Quaternary International xxx (xxxx) xxx–xxx

M. Kim et al.

Zhang, W., Montgomery, D.R., 1994. Digital elevation model grid size, landscape representation, and hydrologic simulations. Water Resour. Res. 30, 1019–1028. Zhou, A.N., Daichao, S., Scott, W.S., Antonio, G., 2012. Interpretation of unsaturated soil behaviour in the stress-Saturation space, I: Volume change and water retention behavior. Comput. Geotech. 43, 178–187. Zizioli, D., Meisina, C., Valentino, R., Montrasio, L., 2013. Comparison between different approaches to modeling shallow landslide susceptibility. Nat. Hazards Earth Syst. Sci. 13, 559–573.

numerical model development. Environ. Model. Software 17, 333–344. Wu, W., Sidle, R.C., 1995. A distributed slope stability model for steep forested basins. Water Resour. Res. 31, 2097–2110. Yates, K., Fenton, C.H., Bell, D.H., 2018. A review of the geotechnical characteristics of loess and loess-derived soils from Canterbury, South Island, New Zealand. Eng. Geol. 236, 11–21. Yoshinaga, S., Ohnuki, Y., 1995. Estimation of soil physical properties from a handy dynamic cone penetrometer test. J of the Jap Soc of Eros Cont Engi 48, 200.

13