1 Introduction

Debris slide is a common and destructive natural hazard in mountainous regions of many countries. In India, natural hazards like landslides affect almost 15% of the land area exceeding 0.49 million km2 (Kanungo et al. 2013; Sarkar et al. 2016). Tectonically active regions of India, especially the Himalayan region, witness recurring landslides where the problem is magnified by fragile lithology, high-intensity rainfall from the Indian summer monsoon (ISM), and often by enhanced developmental activities. More particularly, the Arunachal Himalaya in the North-Eastern (NE) part of India, which presents a geodynamically very unstable terrane with high relief and focused rainfall in its southern slopes, experiences recurring landslides. Many of these landslides occur along the newly constructed as well as existing roads, frequently disrupting surface communications in various parts of the state. Numerous landslides in Arunachal Himalaya generate a humongous amount of sediments that find their way into the south-flowing rivers that drain into the Brahmaputra valley. This sediment flux triggers another disaster in the lower riparian areas by way of enhanced flooding and erosion. Recent studies (e.g., Katzenberger et al. 2021) predict increasing but erratic monsoon rainfall in the North-Eastern Himalayan region in the twenty-first century, and in such a scenario, the intricate rainfall-landslide relationship needs to be given due attention considering its multiplier effects. Therefore, an assessment of sediment generation from landslides is called for which will contribute towards better management of flood and erosion in the downstream areas.

Debris slide is the downslope movement of broken rock masses, natural soil, and unconsolidated sedimentary materials along with a surface under the deposit (Hu and Bürgmann 2020). This process may become more frequent due to slope modification during highway construction or urbanization. It is further aided by underlying geological conditions like the presence of highly fractured or deeply weathered rocks in the slopes. Extreme climatic conditions like heavy precipitation also provide strong control for this type of hazard (Lee et al. 2007; Crosta and Frattini 2008; Tang et al. 2011; Jeong et al. 2017). Pore water pressure triggers initiation of debris slide (Varnes 1978; Caine 1980), once saturation is achieved following continued precipitation. Studies by Wen and Aydin (2005) in a slip zone of a rainfall-induced debris slide composed of volcanic saprolitic soil showed that during heavy rainfall, high pore water pressure (u) was generated at the tip of tension crack(s) that gradually propagated down the slope resulting in a significant increase of “u” within the toe section and a sudden release of the strain energy.

In the Arunachal Himalaya, although the fragile lithology and geodynamics are the preparatory factors for landslides, their frequency is significantly skewed towards the monsoon season, comprising the months of June, July, August, and September (JJAS), thus providing a triggering factor. In this work, we try to highlight this linkage between the rainfall and debris slide from a representative road section through field observation and simulation of one of the most damaging slide using finite element method (FEM) to analyze its mechanism of failure. The FEM results are then validated with the limit equilibrium method (LEM—Bishop’s simplified method).

2 Description of the study area

The selected study area is a 33 km long road section between Doimukh and Haj along the national highway 229 (NH-229) in Papum Pare district of Arunachal Pradesh (Fig. 1). The road is strategically important for the movement of defense personnel and civilians. This road connects the Ziro town with the state capital Itanagar and the rest of the country through Assam. This important road link is often disrupted due to landslides, particularly during the rainy season. As many as 30 debris flow gullies are present in this segment with the largest landslide located at 27° 11′ 38.31″N and 93° 46′ 48.64″ E in the western bank of the Dikrong river, a south-flowing Himalayan tributary of the Brahmaputra. The study area presents high relief and rugged topography with NE-SW trending structural hills developing cuesta ridges and hogbacks. Deep erosional scarps are also found dotting the area, mostly along the Dikrong river. The average height of the hills increases from south to north, ranging from about 300 m to 2700 m above mean sea level (AMSL), while most parts of the district have an altitudinal range of 1000 m to 2000 m AMSL (Fig. 2). The details of the datasets used in this study are given in Table 1.

Fig. 1
figure 1

Location map of the study area. The road section from Duimukh to Haj measuring ~ 33 km overlaid on the LANDSAT ETM + (2000) FCC. The landslide location shown is the one used in the detailed study and modeling

Full size image
Fig. 2
figure 2

a SRTM-DEM derived NW–SE transverse topographic profiles across the landslide zone, b A-A′ cross-section, the slope trails into the Dikrong river, which flows SW in the study area

Full size image
Table 1 Details of datasets used in this study
Full size table

Geologically, the study area is part of the Sub-Himalayan domain composed of Neogene clastics termed as the Siwalik Group (Kumar 1997) which outcrops in a belt of about 15 km (Karunakaran and Rangarao 1976) within two boundary thrusts, i.e., the Himalayan frontal thrust (HFT) and the main boundary thrust (MBT) towards south and north, respectively. Rangarao (1983) estimated a thickness of about 6500 m for this sequence that overlies Gondwana sediments across an unconformity. The HFT marks the boundary between the Brahmaputra plain towards the south and the Neogene sediments towards north across a topographic break. These rocks are lithostratigraphically divided into the oldest Dafla Formation followed by the Subansiri Formation and Kimin Formation in stratigraphic order (Rangarao 1983; Kumar 1997). While Dafla Formation is primarily massive to thickly bedded, fine to medium sandstone and shale, Subansiri Formation is composed of bedded sandstone, siltstone, and shale/carbonaceous shale, and Kimin Formation is a thick pile of coarse to gritty sandstone and conglomerate. The rocks are often friable and easily eroded by surface runoff. Quaternary gravel deposits of variable thickness are found in many localities (e.g., in the Rono Hills area) unconformably overlying the deformed Neogene sequence forming terraces at various topographic levels.

3 Materials and methods

3.1 On-site investigation

The area of investigation witnesses frequent landslides during the rainy season. However, some attained catastrophic dimensions as it happened in July 2013. A post-disaster field study was carried out along the Doimukh-Haj national highway to assess the physical site factors of these landslides. All along the road hill slopes are steep, ranging between 55° and 65° exposing moderately strong sandstones with shale intercalations, which are covered with low strength and highly weathered debris material.

Boulders and clay pockets are found embedded within the debris. This assorted mixture of unsorted, highly porous debris came cascading down following prolonged rainfall. The absence of channelized flow of surface runoff resulted in very high percolation through the debris with a resultant increase in load and development of positive pore pressure (Fig. 3). It was also observed that for the water flowing over the slope, there were no proper drainage channels. The road was cleared after the landslide, but the slope remained in a very critical condition (Fig. 3b). The dimension of the landslide and runout length of the debris material is illustrated in Fig. 4. The failed Gabion wall made of stone with steel mesh (Fig. 5) shows inadequacy of the adopted mitigation measures.

Fig. 3
figure 3

Field photographs showing the post-failure condition of the debris slide, a front view, b lateral view

Full size image
Fig. 4
figure 4

Illustration of the failed slope geometry

Full size image
Fig. 5
figure 5

Field photographs showing a failed gabion wall made of stones with steel mesh

Full size image

An SRTM-DEM derived slope map (Fig. 6) shows that in most of the eastern parts of the road between Doimukh and Haj slope is > 25°, while a very gentle slope is found in the river valley west of the road. This makes the eastern slope highly vulnerable, more so because in many localities the river continues toe-cutting with consequent development of shear planes in the overlying bank materials.

Fig. 6
figure 6

Slope map of the area derived from 30-m resolution SRTM-DEM

Full size image

3.2 Debris material fragmentation analysis

The studied debris slide was a complex mixture of broken rock pieces, boulders, natural soil, and unconsolidated sedimentary materials. In this study, a representative section was chosen, and a fragmentation analysis was carried out for the rough estimation of frequency distributions of size, volume, and mass percentage of debris material constituents (Fig. 7a). The unconsolidated debris material consisting of large boulders was deposited at the base of the slope creating a heap. The heap was then digitized (Fig. 7b, c) to obtain the equivalent diameter of each block size followed by the determination of frequency distribution based on the gradation curve of the measured rock blocks (Fig. 7d). The maximum frequency of equivalent diameter ranges from 600 to 100 mm, with an approximate passing percentage range from 80 to 20%. The D01, D20, D50, D80, and D99 are 7.08 mm, 103.15 mm, 241.05 mm, 534.85 mm, and 1018.73 mm, respectively.

Fig. 7
figure 7

a Representative field view of debris heap, b block edge detection, c digitization of failed rock blocks, d gradation curve obtained from the fragmentation analysis, the diameter of equivalent spheres and % passing is plotted in the x-axis and y-axis respectively. The white patches in section b and c represents the finer particles

Full size image

3.3 Precipitation characteristics

Arunachal Pradesh receives rainfall of the order of 2000 to 5000 mm with a small zone of heavy rainfall of the order of 3000 to 5000 mm along the foothills of the Himalayas, north of the Brahmaputra river (Dhar and Nandargi 2004). The annual average precipitation of the Papum Pare district ranges between 400 and 1200 mm, with about 60% of the average rainfall during JJAS. Frequent heavy rainfall in a geologically young and steep mountainous region may cause shallow slope failures in which the failed material loses its strength and flows (Anderson and Sitar 1995; Cha et al. 2018; Cuomo 2020). Though rainfall-induced shallow landslide is an obvious phenomenon, it is very difficult to accurately determine the slope stability. This is because rainfall indirectly influences slope stability through its effect on pore water pressure on slope material (Blong and Dunkerley 1976). In a recent study by Gogoi et al., (2020), it has been stated that it is because of the highly friable sandstones which get easily eroded by the rivers in the Himalayan foothills recharged during JJAS, thereby causing sand aggradation further downstream destroying the fertility of the cultivated lands. Hence, the type of slope material is an important factor for causing a landslide.

3.4 Effect of precipitation

During JJAS, the debris material gets enough water to become saturated. Infiltration of rainwater in pore spaces increases the weight of the unconsolidated material. The pore water exerts pressure on the surrounding grains due to which steep tension cracks are formed (Wang and Sassa 2003). The pore pressure is higher at the toe of the slope as compared to the upper part of the slope. The cracks appeared at the toe of the slope after the saturation indicating the initiation of slope failure (Anderson and Sitar 1995; Hakro and Harahap 2015). The cracks on upper part of the slope facilitate expansion of the sliding surface affecting slope stability.

The fully saturated debris material exhibits positive pore water pressure which reduces the shear strength of the material, and if a shear failure begins, the loose material abruptly loses its strength and starts flowing (Fang et al. 2012). Above the water table, the pore water pressure becomes negative (see Eq. (1)) as the voids are partly filled with water. In this case, the surface-tension forces operate to achieve a suction effect to hold the surrounding grains together, and as a result, the shear strength of the soil increases (Lu and Likos 2006) (Fig. 8).

$$u \, = \, - \gamma_{w} h_{w}$$
(1)

where u = pore water pressure, γw = unit weight of water, and hw = height of the water column above the water table.

Fig. 8
figure 8

Ideal scenario for positive (+ u) and negative pore water Pressure (−u) variation with depth (h) (modified figure from http://environment.uwe.ac.uk/)

Full size image

The shear strength of soil increases with an increase in negative pore water pressure, and it is a function of effective stress (σeff) instead of total stress (σ). Mathematically effective stress is written as:

$$\sigma_{{{\text{eff}}}} = \sigma - \, u$$
(2)

The negative pore water pressure scenario generally exists during each monsoon interval when the material may reach its highest strength at the optimum moisture content (OMC).

3.5 Volume estimation

Landslides generate an enormous volume of sediments, most of which are transported to the mountainous rivers and their associated tributaries. However, quantifying the downstream transport of landslide-derived sediment remains a challenge till today. In landslide deposits, soil and gravel fragments generally co-exist and their proportions influence the hydrological and erosion processes on the steep slopes of the deposit surface (Korup 2005; Korup et al. 2010; Gan et al. 2018). The steep landslide generating sediments finds its way to the south-flowing Dikrong river with an erosional scarp in its right bank. This river enters the Brahmaputra plain which is ~ 40 km south of Doimukh releasing the sediment load as the river loses its competence and capacity downstream of its mountain exit. Heavy sediment flux has resulted in the active valley and river bed aggradation, flooding, and erosion in the downstream areas besides inducing widespread lateral channel changes. It is, therefore, imperative to estimate the possible quantity of sediments generated from the landslides in Arunachal Himalaya that form the upper catchment of these rivers.

The volume of a potential landslide can be estimated by the friction coefficient that is defined as the ratio of the vertical drop height (H), to the horizontal runout distance (L), measured from the crest of the scarp to the tip of the deposit (Scheidegger 1973) (see Fig. 9). The fast motion of a soil mass along a slope is governed by the laws of dry friction, and the friction coefficient can also be used to evaluate landslide mobility (Heim 1932). The coefficient of friction (f) is the tangent of the angle of repose of the sliding debris mass which is equal to the angle of the internal angle of friction, ϕ (Mohr–Coulomb) of the material.

Fig. 9
figure 9

Geometry of a typical landslide (modified after Scheidegger 1973)

Full size image

In Fig. 9, if we consider “m” as the sliding mass through the distance ΔS, and g is the acceleration due to gravity, then the kinetic energy will be

$$\left\{ {\Delta \left( {\frac{1}{2}mv^{2} } \right) = mg\,\Delta H - mgf\Delta L\;{\text{where}},\;\,\Delta H = \Delta S\sin \beta \;{\text{and}}\;\Delta L = \Delta S\cos \beta } \right\}$$
(3)

Deducing Eq. (3), we get,

$$\Delta H - f\Delta L = \frac{{\Delta \left( {\frac{{v^{2} }}{2}} \right)}}{g}$$
(4)

If we assume that the slide started from rest and is at rest at the end and then integrate Eq. (4) from S = 0 to S = total slide length, we will get

$$\left\{ {0 = H - fL,\,\;{\text{or}}\;f = \frac{H}{L} = \tan \alpha } \right\}$$
(5)

Based on 33 landslide case histories reported in the literature, Scheidegger, (1973) proposed an empirical correlation (Eq. (6)) between the landslide volume (V) and the coefficient of friction (see Fig. 10). Therefore, if the crown scarp height and the total travel distance of the slide are known, then the landslide volume can be estimated. The geometrical parameters of the landslide with friction coefficient are shown in Table 2.

$$\log f = - 0.15666\log V + 0.62419,\quad R^{2} = 0.82$$
(6)
Fig. 10
figure 10

Empirical correlation between landslide volume V (in m3) and the coefficient of friction f (H/L) (modified after Scheidegger 1973)

Full size image
Table 2 Geometrical parameters of the studied landslide with friction coefficient (f)
Full size table

Recently, based on 930 landslide case histories, Cha et al. (2018) proposed an empirical relation (Eq. (7)) between landslide volume (V, in m3) and landslide area (A, in m2). In their data, the largest landslide volume was 1165 m3 and the smallest was 2.5 m3 with a mean and standard deviation of 220 m3 and 186 m3, respectively.

$$V = 0.596A^{1.02} ,\quad R^{2} = 0.89$$
(7)

In the present study, considering the approximate scar area of the landslide as 18,343 m2, vertical height (H) of 82 m, and the total horizontal travel distance (L) of 90 m, using Eqs. (6) and (7), the volume of the debris material is calculated as 17,475 m3 and 13,230 m3, respectively (average ~ 15,352 m3). It is important to note that the landslide-generated sediment may not be completely dumped in the valley. Around 30 to 50% of the sediment volume may rest on the slope. The fluidization of the material depends on the amount and duration of rainfall. Also, if clay or soil proportion is more in the debris material then the deposit length would have been reached up to the calculated value provided the valley width is broad enough to accommodate the sediment.

3.6 Numerical modeling

Landsliding is an important mechanism of mass wasting in mountainous regions. Most landslide studies have been driven by an applied viewpoint; therefore, slope stability analyses and hazard prediction have been a major focus of study (Martin et al. 2002). There are several numerical techniques available (continuum method, discontinuum methods, etc.) and are being widely used in slope stability analysis (Kanungo et al. 2013; Kainthola et al. 2014; Stead 2016; Singh et al. 2016; Fathani et al. 2017; Pradhan et al. 2019; Sanzeni et al. 2019). In this research, the pre-event slope condition is simulated under the 2D finite element method (FEM) using Phase2 (Rocscience Inc. 2016) under both dry and saturated conditions to determine the corresponding critical strength reduction factor (SRF). The factor of safety (FoS) calculation using FEM is different from the conventional limit equilibrium method. In FEM, the shear strength reduction (SSR) technique is applied to determine the FoS by progressively reducing the shear strength of the material to bring the slope to a state of limiting equilibrium (Matsui and San 1992; Griffiths and Lane 1999; Dawson et al. 1999). The basic principle behind the SSR technique in the FE slope stability analysis is to reduce cohesion (c) and friction angle (ϕ) until the slope fails, in which the failure of the slope is defined as the failure shear strain that develops from the toe to the top of the slope. Since Mohr–Coulomb failure criterion is defined using the c and ϕ, the SSR technique is commonly used with the Mohr–Coulomb failure criterion. The Mohr–Coulomb failure criterion suggests that the shear strength of the ground is proportional to the normal effective stress as:

$$\tau = c + \sigma^{\prime } \tan \phi$$
(8)

where τ is the shear strength, σ′ is the effective normal stress. The water in pore spaces of material reduces the shear strength, and it decreases the effective normal stress that acts between the grains and, ultimately, reducing the frictional resistance. The shear failure in masses of unlithified loose regolith (debris) occurs if stresses obey the Mohr–Coulomb failure criterion (Iverson et al. 1997). Hammah et al. (2004) had examined the difficulties of application of the SSR method to the generalized Hoek–Brown material model in stability analysis using FEM. They also suggested a solution approach that uses the equivalent Mohr–Coulomb model instead of the Hoek–Brown failure criterion. In FEM, the FoS (F) is calculated as per the following equations

$$c^{trial} = \frac{c}{{F^{trial} }}$$
(9)
$$\phi^{trial} = \arctan \left( {\frac{\tan \phi }{{F^{trial} }}} \right)$$
(10)

A sequence of iterations was made using trial values of the strength reduction factor (Ftrial) to reduce the c, and ϕ until the slope fails.

3.6.1 Model discretization, meshing, boundary condition, and loading

To model the ground, three-node triangular elements were used. Almost any plane structural shape can be discretized with triangular finite elements, though individual triangular finite elements may be different in size and shape (Carroll 1998; Das and Singh 2021b). Graded mesh type was used with a gradation factor of 0.1. The boundaries of the model are fixed (no movement in x and y directions) along the left side and at the base of the model. The slope surface is kept free. As Zhang (2005) stated, for complex topographic conditions such as hills, the rock mass can be assumed to be under gravitational load. Therefore, gravitational load is applied to the model with a field stress ratio (k) of 1 (Das and Singh 2021a).

3.6.2 Material property

In a recent study, Sharma et al., (2018) characterized the mountain soil (upper weathered layer) along the Orang-Kalaktang-Shergaon-Rupa-Tenga (OKSRT) road, Arunachal Pradesh, India, which is very near to the present study area. The grain size analysis suggests that the soil contain 23.2% gravel, 64.5% sand, 12.3% fines (silt and clay). The maximum dry density was 1.731 ± 0.26 (g/cm3). They reported similar slope failure conditions along the OKSRT road cut slopes. The geomechanical properties of debris material used in the numerical simulation are taken from the published literature with similar slope material and similar slope failure condition (Kanungo et al. 2013; Guo et al. 2016; Sarkar et al. 2016; Sharma et al. 2018; Pradhan et al. 2019) and given in Table 3. Mohr–Coulomb failure criterion is used to model the rock mass (Eq. (8)).

Table 3 Material input parameters for the dry and saturated condition
Full size table

4 Results and discussion

4.1 Correlation between precipitation and landslide in the study area

In this research, the authors have analyzed and correlated the monthly average rainfall data and landslide data for the Papum Pare district of Arunachal Pradesh (NE Himalaya) for the period 2004–2010 and 2014–2018. The authors have observed that there is a sharp increase in rainfall in April–May after the post-monsoon dry season and reaches its peak during June (Figs. 11, 12). After June, the average rainfall data shows a decreasing pattern up to July end or beginning of August. The average rainfall then starts rising again and attains its peak in September and decreases afterward. Thus, a bimodal distribution of rainfall is observed with the primary peak in June and a reduced peak in September. A comparison of landslide incidence versus rainfall (Table 4) shows that all these events happened during JJAS. The results signify that there is a sharp and accurate correlation between precipitation and landslide. It is deduced that a minimum threshold value of 433 mm of monthly average rainfall can be considered as an early warning for landslide-triggering events for this region. The threshold is chosen by analyzing the lower limit of rainfall conditions that resulted in landslides in that region. The threshold value applies only during the monsoon season (JJAS) in this region. During the increased intensity of rainfall, both infiltration as well as surface runoff increases, and as a result, there is a decrease in soil suction and cohesion. The increased weight of the loose slope material increases shear stress and decreases the shear strength of the slip surface. Though in May, the average rainfall crosses the threshold, the landslide occurrence is insignificant, because the slope material does not reach its full saturation to develop positive pore pressure. Apart from rainfall, the slope angle plays a key role in landslide (Singh et al. 2014, 2016; Ansari et al. 2015) such that infiltration takes a longer duration along steeper slopes.

Fig. 11
figure 11

Monthly average rainfall (in mm) of stations under the Papum Pare district of Arunachal Pradesh, data collected from the Indian Meteorological Department and India Water Portal

Full size image
Fig. 12
figure 12

Plot of monthly average rainfall with standard deviations for Papum Pare district (for the year 2004–2010 and 2014–2018), the highlighted red zone represents landslide-prone rainfall intensities

Full size image
Table 4 Few major rainfall-induced landslides in the Papum Pare district. (
Full size table

A similar study was conducted by Sengupta et al. (2010) in a 15 km road stretch of the North Sikkim Highway, Sikkim, India (Eastern Himalaya), which is susceptible to shallow debris slides. Based on their results, 250 mm of cumulative rainfall for 15 days (i.e., average 0.69 mm/h) of continuous rainfall was considered as a triggering factor for landslide occurrence in that region. But recently, Teja et al. (2019) presented a rainfall threshold for landslide using data for 2010–2016 in the Darjeeling Himalayas, which is very near to Sikkim Himalaya, and concluded that cumulated event rainfall of 36.7 mm for 2 days (i.e., average 0.76 mm/h) is required for landslide occurrence. Our observation also depicts a very close threshold value of cumulative rainfall, i.e., 433 mm/month (average 0.59 mm/h), which is quite reasonable for Arunachal Sub-Himalaya, considering other geological factors. Considering the average rainfall intensity of 0.59 mm/h for the study area, the approximate time of landslides can be estimated through various rainfall intensity (I)—duration (D) equations proposed by various researchers (Table 5 and Fig. 13).

Table 5 Notable global rainfall intensity (I) (in mm/h)—duration (D) (in hours) threshold equations proposed by various researchers
Full size table
Fig. 13
figure 13

Sensitivity analysis, comparison, and applicability of rainfall intensity (I) vs rainfall duration (D) threshold equations (from Table 5) for debris slide, proposed by various researchers across the world

Full size image

For the sake of comparison in Fig. 13, and with the data available for the study location, it is clear that the global equations proposed by Caine (1980), Moser and Hohensinn (1983), and Cancelli and Nova (1985) are not suitable for the Arunachal Sub-Himalayan region. The rest of the equations were quite consistent and applicable for the study area. A horizontal threshold line is drawn at 0.59 mm/h intensity in Fig. 13 to estimate the rainfall duration required to trigger landslide in the study area. This implies that it will take at least 36 to 48 h after the onset of rainfall for the landslide to occur at such rainfall intensity. Guzzetti et al. (2008) also confirmed that a minimum of ~ 48 h is required to initiate shallow slope failures. Here we have calculated 0.59 mm/h (or 14.33 mm/day) as threshold rainfall intensity from the available data, notwithstanding the fact that rainfall rarely remains constant throughout a day. Therefore, shallow debris slides may occur if the rainfall intensity is skewed to a maximum for a particular time period of a day.

Paul et al. (2000) conducted a study of a similar landslide that occurred on 18 August 1998 (monsoon period) in Malpa village, Kumaun Higher Himalaya region, along the Indo–Nepal border. It was a complex type of landslide (i.e., combination of rockfall and debris flow) triggered due to high precipitation (maximum 113 mm/day for the month of August 1998) which enhanced the pore-water pressure beyond the threshold limits.

4.2 Drainage density

Studies show that shallow landslides in soil and weathered rocks are often triggered on steep slopes during the more intense parts of a rainfall event, or moderate-intensity rainfall lasting several days (Anderson and Sitar 1995; Turner and Schuster 1996; Corominas and Moya 1999; Paul et al. 2000). The nature of the slope material and its geotechnical properties can be denoted by the drainage density, Dd, (ratio of the sum of the channel lengths per unit area) of a region which may be an important contributing factor for landslides (Bera et al. 2019). The Dd is inversely related to the function of infiltration such that less infiltration causes more favorable runoff condition, resulting in high Dd. Drainage density also affects the water transportation time through the landscape and thus exerts a primary control on catchment flood response (Clubb et al. 2016). For the Papum Pare district, the Dd has been classified into five different classes using spatial analyst tools of ArcGIS (Fig. 14). From the drainage density map, it is clear that the drainage density contours are closely spaced near Doimukh and Haj (Dd value 0.63—1.1), which implies presence of impermeable subsoil layer. However, at the location of the landslide a comparatively low drainage density (Dd value 0.26—0.5) suggests a relatively more permeable subsoil/debris layer. During rainfall, the permeable layer gets fully saturated and exhibits positive pore water pressure which reduces the shear strength of the material, and this becomes one of the primary causes for landslides in this region.

Fig. 14
figure 14

source: River network data derived from HydroSHEDS data at 15 arc-second resolution was collected from “https://www.hydrosheds.org/”, popularly known as “HydroRIVERS data” (Lehner and Grill 2013)

Drainage density map of Papum Pare district, Arunachal Pradesh. The red line segment denotes the Doimukh-Haj national highway, and the white square box represents the landslide study location. (Data

Full size image

4.3 Numerical results

The shear strength reduction (SSR) method is performed to analyze the stability of the slope. In this method, the strength parameters of a slope are reduced by a certain factor (SRF), and the finite element stress analysis is computed. This determines the critical strength reduction factor (critical SRF) or safety factor, of the slope. When the pre-event slope was simulated, the critical SRF value was found to be 1.12 and 0.72 for the dry and saturated conditions, respectively. The saturated slope condition shows a lower value of critical SRF. The decrease in the value of critical SRF in saturated condition indicates the enhanced vulnerability of the slope under the influence of rainfall saturation in that region.

In the model, the portion of the slope where sliding occurred is color-coded in Fig. 15a, b. The figures also indicate the total displacement along the slope for both conditions. The FE analysis results are validated with the post-failure zones observed in the field. The lower critical SRF value for the saturated slope supports the fact that in saturated conditions there is an increase in positive pore water pressure which ultimately reduces the shear strength of the debris material of the actual slope.

Fig. 15
figure 15

Total displacement in, a dry condition, and b saturated condition

Full size image

The finite element results were validated with the limit equilibrium analysis using Bishop’s simplified vertical slice method on SLIDE 6.0 (Rocscience Inc. 2012) (Fig. 16). The limit equilibrium methods (LEM) are approximate methods that assume the existence of a slip surface of various simple shapes. In this method, the rock/soil mass is divided into slices and calculates the most critical position for the slip surface. In Bishop’s Simplified method, the FoS is the ratio between the shear strength to shear stress acting on the failure surface (Bishop 1955). The FoS results for both dry and saturated condition are given in Table 6.

Fig. 16
figure 16

Limit equilibrium analysis result for, a Dry condition, and b saturated debris slope

Full size image
Table 6 FoS calculated by FEM and LEM for dry and saturated condition
Full size table

It is important to note that, to accurately model the slope, the geometry of the slope and material property are the most crucial data. From the simulation result, it can be confirmed that FEM simulates a more realistic model of a slope than the LEM. This is because LEM has a limitation over the predefined failure surface, where the vertical surfaces failed on a circular slip surface, centered on a common point (Kainthola et al. 2013; Kanungo et al. 2013; Huang 2014; Singh et al. 2014) (Fig. 16). The shear stress at the base of each slice is shown in Fig. 17 for both dry and saturated conditions. As mentioned earlier, the increase in water content increases the volume weight of the unconsolidated soil mass. The vertical component of this increased weight further increases the shear stress at the base of the slip surface. This induces expansion of the sliding surface and affects the stability of slopes. Additionally, the fully saturated debris material exhibits positive pore water pressure, which reduces the shear strength of the material and thus triggering failure.

Fig. 17
figure 17

Shear stress vs. slice number (LEM: Bishop’s simplified method)

Full size image

5 Conclusion

Arunachal Himalaya witnesses recurring landslide events, most of which occur during the rainy season. Therefore, although geodynamics, lithology, and geomorphology contribute as preparatory factors, rainfall acts as the primary trigger. This, of course, excludes the catastrophic events due to seismicity, for example, the 1950 Assam earthquake (Poddar 1950) that triggered thousands of landslides in Arunachal Himalaya. Prolonged rainfall saturates the debris materials inducing positive pore pressure, which reduces the shear strength of the material. In this research, as a case study, we have investigated a 33 km long road stretch between Doimukh and Haj along the national highway 229 (NH-229) in Papum Pare district of Arunachal Pradesh. The field observation data of the studied landslide, systematic analysis of precipitation data, and data related to landslide incidents made us draw the following conclusions:

  • Based on a limited period rainfall study and its correlation with landslides, a minimum threshold rainfall value of 433 mm can be considered critical to induce landslides, and this threshold may be useful for developing a landslide early-warning system for the study area. Considering lithological, topographic, and climatic homogeneity, this threshold may be extrapolated to the other parts of Arunachal sub-Himalaya.

  • Estimation of sediment generation from a selected recurring landslide indicates at least ~ 15,352 m3 of sediments from a single event. This is a pointer to the enormous amount of sediment load being fed into the Himalayan rivers draining through this landslide-prone Sub-Himalayan domain. This added debris in the fluvial system triggers a series of devastating secondary impacts in the downstream areas by way of intense flooding, river bed and valley aggradation, bank erosion, and frequent channel changes in the northern Brahmaputra valley.

  • A finite element model validated with the limit equilibrium analysis shows that the studied slope was in critical condition in dry state. Slope failure was triggered by rainfall-induced positive pore pressure in the monsoon season. Pre-event simulation for dry and saturated conditions gives the critical SRF values of 1.12 and 0.72 in FEM and FoS of 1.009 and 0.786 in LEM, respectively. The critical SRF and FoS, both are significantly reduced due to the saturation of the debris material.

  • Flood, erosion, and sedimentation remain a hitherto untamed natural disaster in the Brahmaputra valley with consequent adverse fallout in the socio-economic life of millions of people. A holistic strategy to deal with this problem will require interventions in the upper catchment of the Himalayan rivers, which receive sediment flux primarily from landslides. This also calls for a better understanding of the space–time variability of the Indian Summer Monsoon to enable developing an early warning system for landslides in Arunachal Himalaya and, ultimately, to address the vexed problem of landslide, sedimentation aggradation, flood and erosion.