Landslide Distribution and Processes in the Hills of Central Nepal: Geomorphic and Statistical Approach to Susceptibility Assessment

The study examined the landslide distribution, processes, and susceptibility of the Lalbakaiya watershed using GIS and remote sensing technology. Inventory of landslides was done using high-resolution satellite imagery available on Google Earth and was verified and further investigated during the field visit. Geomorphic as well as statistical approaches were applied to assess landslides susceptibility and the significance of their outputs was discussed. Map layers representing conditioning and triggering factors of landslide occurrence were produced from various spatial data sources. The study found that the landslide of the Lalbakaiya watershed is primarily controlled by geology representing young, weak, fragile, and weathered sedimentary rocks. Besides, the role of topography such as steep slope, high relative relief, and land use and land cover played an important role in determining the landslide susceptibility. These processes are triggered by monsoon precipitation, seismicity, and land use change in addition to other factors. The geomorphic approach produces a reliable landslide susceptible map as evidenced by past and present (active) failures on a landscape unit, but this map has low predictability of the landslides occurrence. In contrast, the landslide susceptibility map derived from the landslide index method fairly conforms with that derived from the geomorphic approach. Susceptibility calculated by landslide index map is represented by a pixel value that indicates a probability of landslides occurrence, and is amenable to group into various susceptible classes. The model can predict areas of landslides based on quantitative relation between landslides and and can also overlook highly erodible areas where landslides are not apparent despite severe erosion and numerous minor failures. The study confirms that both geomorphic and statistical approaches can be complementarily integrated to produce predictable, reliable, and applicable landslide susceptibility maps that can make a plausible planning tool for conservation, development, and disaster risk reduction in the populated slopes of the Himalayas and like.


Introduction
Landslide is a prominent geomorphic process that has shaped the landscape of the mountains (Korup et al., 2010;Dhakal, 2015). In the tectonically active Himalayas, landslides are the major landscape forming processes. In the populated parts of the mountain, landslides have caused severe damage and destruction to lives and property. In recent times, owing to haphazard infrastructure development, the incidences of landslides have increased, and thus increased the risk.
Landslide refers to the geological phenomena that involve downward and outward movements of slope materials by falling, sliding, and flowing under the influence of gravity (Varnes, 1978;Dikau et al., 1996;Cruden & Varnes, 1996). Normally, a landslide denotes any down-slope movement of soil and rock under the direct influence of gravity (Highland & Bobrowsky, 2008). For this study purpose, here it is defined as "the movement of a mass of rock, earth or debris down a slope" (Cruden, 1991).
The mechanism and the causative factors of these processes are attributable to varying characteristics of geology and structure, tectonics, topography, geomorphology, climate, and hydrology (Cruden & Varnes, 1996;Guzzetti et al., 1999, Dahal et al., 2008 Landslides are the product of a complex interplay of various triggering and conditioning in situ factors. An in-depth understanding of landslide relation to terrain and susceptibility thereby requires various methods, tools, and techniques developed in geomorphology, geology, and engineering. In Nepal, landslides have been the cause of the major geomorphic disasters that claim numbers of people and huge amounts of property loss. Landslides trigger floods downstream contributing to a huge sediment load in the river. These cascading events contributed to about 40% of the total population affected by all disasters in Nepal during 1971-2019. Landslide and erosion have been the major cause of land degradation in Nepal (Ghimire, 2020).
The basic requirement for minimizing landslide risk and controlling soil erosion and land degradation is to understand the types and processes of landslides and thereby identifying the areas of landslide susceptibility. Potential sites that are particularly prone to landslides therefore need to be identified. Hence, understanding the landslide processes in the local context and their relation to control- ling or triggering factors, and identification of spatially varying landslide susceptible areas has been one of the areas of research in geography and earth science. Landslide susceptibility map refers to the relative likelihood of future landsliding based solely on the intrinsic properties of a locale or site. It is also known as the "landslide potential map". Evaluation of landslide susceptibility has been carried out through geomorphic and quantitative approaches across the world. In the Himalayas, few studies have applied hazard or susceptibility assessment through geomorphic mapping or landslide inventory methods. Similarly, recently numerous studies have applied quantitative methods of landslide susceptibility and their reliability has been claimed. But the relative advantage in terms of accuracy, predictability, and practicability of landslides susceptibility maps obtained from these approaches to reduce landslides risk and vulnerability has been less studied in the Himalayas.
In this regard, the present study attempts to understand the landslides distribution and processes and analyze the causative factors of landslides in the Lal Bakaiya watershed, central Nepal. This study also intends to prepare a landslide susceptibility map adapting geomorphic and quantitative approaches. Lalbakiya watershed, a fairly densely populated area in the Siwaliks and adjacent slope of the Mahabharata Range in central Nepal is speculated to be a high risk of landslides and flash floods. Landslides disasters are frequently reported during monsoon (DWIDP, 2019). The Master Plan prepared by President Chure Terai Madhesh Conservation Development Board (PCTMCDB, 2017) has identified the Lalbakaiya watershed as one of the highly vulnerable ones in the Siwaliks (PCTMCDB, 2017). Hence, understanding the landslide process and susceptibility zonation of this watershed will be a step forward to identify and implement measures to control landslide and soil erosion processes and reduce disaster risk in the Himalayas. The findings of this study can be replicated to similar geo-ecological conditions elsewhere.
Studies on landslide susceptibility range from case studies at the local to the regional scale (Ghimire, 2011b;Youssef et al., 2015;Pradhan et al., 2019;Yang et al., 2019;Trigila et al., 2013). Landslide susceptibility assessment has been carried using two approaches, i.e., qualitative and the quantitative or semi-quantitative approach. Qualitative approaches are, in fact, very subjective insofar as they depend on the knowledge of the experts (Baum,Miyagi,Lee,& Trofymchuk,Figure 1. Location of the Lalbakaiya watershed. (mean relief 158 ± 73 m/km 2 ) relief and gentle slopes (mean 15 ± 10 degree) and highly dissected by streams and gullies. Middle Siwaliks is composed of a higher proportion of sandstone in a sequence of interbedded sandstone and mudstone.
The topography is relatively massive and sharp with an internal relief of 290 ± 84 m/km 2 (in the southern section) 381 ± 101 m/km 2 (northern section) and has generally steep slopes, i.e., mean 24 ± 10 and 29 ± 9 degree respectively. Lower Siwaliks is the oldest of Siwalik formation. It consists of interbedded mudstones and fine sandstones. The ratio of less resistant mudstone is greater than that of sandstone in this formation. The LS is delineated from alluvial deposits of the Tarai plain by the Main Frontal Thrust (MFT) in the south and in the north by Main Churia Thrust lying in the south. The LS, due to the greater proportion of the mudstone bed, has relatively lower internal relief (253 ± 74 and 205 ± 103 m/km 2 ) in the southern and northern sections respectively. The average slope is about 20 ± 10 degrees, i.e., less than that of the MS. Sandwiched between the Si- The slope in most of the watershed is built mainly on the homoclinal folds with bedrock dipping more or less towards the north in a low to moderate angle (25 -40 degrees) in the south, less than 30 degrees in the middle, and above 50 degrees in the north (Figure 3). On low dipping bedrocks, series of major to minor asymmetric ridges characterizing gentle dip slopes facing north and south-aspect cliffs are formed. These ridges are formed by the tributary streams developed along the strike direction (Ghimire, 2020 Climate is subtropical to warm temperate below or above 1200 masl. This watershed receives 85% of rainfall during the summer monsoon. The effect of climate on weathering is high, although the deep weathered layer and soil are not developed as they are washed away by a high rate of erosion on the slopes. Vegetation varies from subtropical evergreen to temperate deciduous forest towards the higher part of the watershed in the north. The population of the watershed is 168,000 (CBS, 2011). During the last six decades, the population density has increased by 6.8 times, which resulted in changes in land use and infrastructure development leading to the intensification of runoff and sediment transportation processes (Shrestha & Bajracharya, 2013). The north-south Fast Track connecting Kathmandu and Nijgadh and east-west Heatuda-Chatara road also traverse the watershed.

Materials and Methods
Data types and sources

Inventory of landslide
The past and present occurrences of landslide are keys to the spatial prediction of future events (Guzzetti et al., 1999). Therefore, the inventory of landslide is the entry point for evaluating the hazard and risk associated with it. In the present study, landslide detection was done by using high-resolution multi-temporal im- landslides were inventoried, out of which 798 were used for assessing susceptibility using a statistical model and the rest were used for validation.
Geomorphic assessment of the landslide and erosion susceptibility Geomorphic interpretation of landslides and erosion susceptibility was done following the basic principles: 1) The past and present is key to the future (Hutton, 1788;Varnes, 1978), 2) Past landslides and erosional processes leave discernible morphological features on the terrain, and 3) Similar biophysical conditions produce similar phenomena (Guzzetti et al., 2012). From the location of active landslides, similar landslides did occur or not in the past were investigated.
This approach relies on the geomorphic footprints of landslides and erosion both currently active and old types. It tries to identify the landscape unit where landslide occurrence was frequent in the past and present and are likely to occur in the future ( Figure 2). Such landscape units comprise of past and present failures, a similar rock or soil strength, slope steepness, and aspect. Inventory of the observed fresh and old landslide morphology, runout deposits or fan, and identification of the landscape units of repeating topography and geomorphology, where footprints of landslides are frequent are the main components of this approach.
Statistical approach of landslide susceptibility mapping The basic underlying principle of determining landslide susceptibility is "Landslides do not occur randomly, or by chance (Guzzetti et al., 2002;Crozier, 1986). Slope failures are the result of the interplay of physical processes, and mechanical laws against the controlling and triggering factors determined the stability or failure of a slope". These physical phenomena, which control the landslide occurrence in various size and types in space and time, can be applied to determined landslide susceptibility using statistically and other quantitative methods of derived relation.  that landslide susceptible is conditioned by inherent in situ terrain factors, which comprise geology, topography, soil, and hydrology. Besides, the rainfall and tectonic driven seismicity trigger landslides. Human manipulation of land use and land cover, mining and excavation, and haphazard construction of infrastructures also induce landslide occurrence (Ghimire, 2011b;Guzzetti et al., 1999;Deoja & Thapa, 1991;Dhital et al., 1991). Therefore, this study intends to incorporate the aforementioned factors that influence landslide occurrence. These factors are broadly categorized into four domains: i) Geology ii) Topography (aspect, relief range, slope gradient, solar illumination, slope shape, and topographic position index), iii) Hydrology(annual rainfall, drainage density, and topographic wetness index), iv) Anthropogenic (land use and land cover).
A statistical analysis, called the landslide index method, was used for calculating landslide susceptibility, which has proved to be very efficient in landslide prediction (Pradhan & Kim, 2014). In this method, the role of each landslide triggering and conditioning factors were analyzed with existing landslides, thereby landslide susceptibility weights for each factor class were calculated. The weightage might be different for different classes of a factor since they have a different impact on landslide occurrence. A weight-value for a factor class, such

Density class Density map ln
where, W i = Calculated weight of certain factor class.
Density Class = the landslide density of each factor class.
Density Map = the landslide density in the entire map.
Landslide susceptibility index (LSI) is determined by the summation of the weight of multiple class of each factor using an equation (Lee & Pradhan, 2006).
where W i = Weight calculated for classes of each i factor N = Total number of factors Then, LSI map was reclassified map into five categories of landslide susceptibility zones, i.e., very low, low, moderate, high, and very high.
The validation of the landslides map was done by evaluating the success rate concerning those landslides of the inventory that were not used in the model.

Result and Discussion
Landslides distribution, process, and mechanism The total number of landslides in the watershed is 828 with an average area of 0.6 ± 0.06 ha ( Similarly complex to rotational slides of moderate to small size (up to 3 ha Besides, the steep slope and high relief developed on resistant thick-bedded sandstone in the Middle Siwalik have increased shear stress against shear strength in the slopes (Terzaghi, 1950;Varnes, 1978;Dhakal, et.al, 2014;Dhital, 2015). Hence the slope failures such as rockslide, rockfall, and complex movements are common (Ghimire, 2011b).
Lower Siwalik consists dominantly of weak beds of mudstones that are highly weathered or have converted to residual soil after the complete weathering.
Weathered materials move easily downslope, especially during intense rainfall.
Owing to the low permeability of the rocks, water cannot easily infiltrate deeper, remain within the soil mass. The saturated soil mass exerts pore water pressure.  Apart from weathering, faults, intricate joints, and fractures also play a prominent role in the occurrence of landslides in the Siwaliks. Through the joints, water enters deeper into the rocks that exert pore pressure which reduces the sheer strength and triggers landslides through the plane of weakness (Selby, 1993).

Landslide conditioning and triggering factors
Landslide causes are diverse and have been recognized by several authors Terzaghi, 1950;Varnes, 1978;Crozier, 1986;Brunsden et al., 1975;Cruden & Varnes, 1996. Broadly, these factors can be divided into two types, i.e., a) conditioning factors and b) triggering factors determine the inherent shear strength of the slope. They control the binding and frictional force of the slope (Selby, 1993 The following conditioning and triggering factors are considered in the study (Table 2).

1) Topographic
Slope: The slope gradient represents the downhill component of the force.
This force is high on the steeper slope which induces gravitational shear stress on slope materials to induce slope failure (Dai et al., 2001;Chapin et al., 2002).
However, slopes in combination with the slope material cohesion, angle of respose, and moisture conditions normally determine slope stability conditions; therefore, the gentler slope may also render the landslides, particularly in colluvial or deeply weathered slopes (Selby, 1993). In the study area, landslide density has increased with slope steepness. The steep slopes of the Middle and Lower Siwaliks of the Lalbkaiya are related to the landslide occurrence and affecting both in area and frequency, which is also recognized in other parts of the Chure hills (Ghimire, 2001;Ghimire, 2011b).
Relative relief: It reflects the local difference in height within a unit area. As other parts of the Chure hills, relative relief has shown a significant association with the landslide in Lalbakaiya (Ghimire, 2011a). The threshold local height beyond where the probability of the observed landslide increases around 50 m/2.5ha (Ghimire, 2011b). Potential energy required for erosion and mass movement in general increases with increased local elevation. High relative relief can be both cause and consequence of the landslides and erosion in the sedimentary rocks of the Siwaliks (Table 2) Aspect: It is related to parameters such as the orientation of geological discontinuities controlling landslides, precipitation, wind impact, and sunlight exposure (Ercanoglu et al., 2004). Landslide distribution on various slope aspects is revealed in Table 2. Landslides seem to prevail in the south-facing slopes of the asymmetric ridges which are normally very steep and also on similar slopes in the northern part of the watershed.
Slope shape: It has a strong influence on slope stability. The shape of the slope act as a flow converging or dispersing surface, and a primary determinant of subsurface water in the hillslopes. There are three basic slope units: a) convex, b) planar, and c) concave (Table 2). Generally, convex slopes are more stable as they disperse the runoff more equally down the slope, whereas concave slopes are considered potentially unstable because they concentrate water at the lowest point and contribute to the buildup of adverse hydrostatic pressure (Stocking, 1972  Topographical position index (TPI): The TPI is derived from DEM using Jenness (2006) method. This index identifies topographic slope positions as a basis of landform classifications. Many physical and biological processes acting on these landforms influence landslide occurrence. These landforms are upper slope or ridge, middle slope, footslope, open slope or river valley, and incised river or stream. Landslide density was found higher in foot slopes and incised river in the Lalbakaiya watershed induced by toe erosion and undercutting.
Solar radiation: Solar radiation is the quantification of the light environment in the earth's surface, which is strongly influenced by elevation, surface orientation, slope gradient, and slope shape, shadows, and others (Cioban et al., 2013). It influences vegetation patterns and dynamics, and landscape morphology in the mountains. The landslide density was found in the areas of lower radiation, which is probably due to pre-existing high moisture content in the soils as well as joints, fissures, and fractures. This makes favorable for an early rise in pore pressure during a rainstorm which triggers slope failures.

2) Hydrological
Rainfall The models also predict that by the end of the century period there will be an increase in precipitation for all seasons except the pre-monsoon season. The monsoon rainfall is projected to increase by 27.1% (MoFE, 2019), hence more rainfall induced disasters are expected.
Topographic wetness index (TWI): The upslope contributing area calculated from the DEM can be a proxy expression of the ridgeline, site of residual soil, colluvium concentration and moisture availability, and drainage flow line (Beven & Kirkby, 1979;Dietrich et al., 1995;Pack et al., 2005). It determines where the slope materials and water is dispersed and where the flow of both water and slope material tends to concentrate. Convergent slopes can be potential areas of colluvium deposits, flow accumulation leading to saturation led debris flow and landslides. The landslide density was found to decreasing in higher wetness index, i.e., more on the divergent slope and less than convergent slope.
Note: The topographic indexes were developed in ArcGIS Environment and then were reclassified as natural break intervals (Table 2).
Drainage density: It indicates run-off conditions as well as the degree of dissection of the landscape. The dense network of drainage line indicates high runoff and low infiltration and vice versa. Some experience shows that pore pressure develops due to the high infiltration of rainwater and makes the slope potentially unstable (Doornkamp, 1974;Selby, 1993 (Dietrich & Dunne, 1978). Such cases are found in the Upper Siwaliks and quaternary parts of Lalbaikaiya watershed on deeply weathered rocks. The landslides incidences were observed high in drainage density above 2.5 km/km 2 .

3) Land use and land cover
A large part of the study area is under forest, followed by cultivated land (Figure 4). Land use and land cover play an important role in the stability of the slopes. The slopes are mostly unstable on the barren areas provided the geology of the area is also unfavorable (LRMP, 1986). Forest cover prevents the rocks from being exposed to the sun and water, which ultimately reduces the grade of weathering in rocks. Vegetation grabs the topsoil and prevents the drop strike of intense raindrops on the earth's ground. They control slope stability by mechanically reinforcing slopes through plant roots, modifying soil moisture distribution, and pore water pressures (Bishop & Stevens, 1964;Gray, 1970  triggering landslides and erosion (Lavé & Avouac, 2001).

Landslide Susceptibility Mapping
Following the approaches mentioned above, landslides susceptible maps were prepared ( Figure 5). Geomorphic interpretation based susceptible maps reveal susceptible landscape units where past present landslides or erosional scars are frequent, and severe gully and river erosion is observed. The southern part of the watershed, which is an active fold zone (Lavé & Avouac, 2001 (Table 3). Thus the results are very much convincing for the mountainous terrain of the study area ( Figure 6 and Figure 7).    In contrast to the geomorphic approach, the landslide index approach is an objective method of assessing landslide potential areas considering observable Hence, both approaches can be integrated to produce predictable, reliable, and applicable landslide susceptibility maps.

Conclusion
Landslide of the Lalbakaiya watershed is primarily controlled by geology representing young, weak, and fragile sedimentary rocks. Besides the role of topography such as steep slope, high relative relief, and land use was found to be important in determining the landslide and erosion susceptibility. These processes are triggered by monsoon precipitation, seismicity, and land use change in addition to other factors. These geomorphic processes and their intensity in hill catchment have as implication on the flash flood, bank erosion, river shift, and course change, and inundation events in the low.
Landslide susceptibility assessment was performed using the geomorphic and statistical index and both approaches have produced similar results. The susceptible map produced from the Geomorphic approach is simple and easy to communicate with the community, practitioners, and planners thus can fruitfully be used in participatory planning such as disaster risk management. Prioritization for mitigation and conservation on a larger scale can be assessed efficiently through this technique. However, this approach is qualitative and may not render a con- The study confirms that both geomorphic and statistical approaches can be complementarily integrated to produce predictable, reliable, and applicable landslide susceptibility maps that can make a plausible planning tool for conservation, development, and disaster risk reduction in the populated slopes of the Himalayas and like.
Except for high erodible quarternary underlain areas, both susceptibility maps showed the approximately similar location and extent of high landslide susceptibility areas.
Lastly, landslide susceptibility maps revealed that the southern part of the watershed is highly vulnerable to landslides, which can be attributed to fragile geology, active tectonics, steep and rugged topography, and high rainfall.

Recommendations
• The geomorphic approach of landslide susceptibility assessment needs to incorporate geo-engineering properties of the hillslope materials and form, and geo-hydrological processes operating on the hillslopes. This will increase accuracy, and thus approval and recognition from the user and scientific community. • Similarly in the absence of geo-technical surveys, the simple and economic statistical method is a viable alternative approach to susceptibility assessment.
But it needs to be used in combination with that of the geomorphic approach.
• Such landslide susceptibility maps can be fruitfully used by the government agencies and various stakeholders working in soil conservation, watershed management, disaster risk management.
• Landslide mitigation actions such as bio-engineering along with land cover management approaches can be implemented. Some landcover management strategies are afforestation in degraded land, forest conservation, and less tillage farming on steep land, gully treatment on the hillslope, and many others.