Landslide hazard assessment using UAV imagery and GIS for road planning and development in Chure area: Sindhuli-Hetauda Section.


 Chure hills are formed with the highly fragile, weak young sedimentary rocks and are environmentally sensitive. Road construction in this region is a big challenge to conserve the Chure from landslide. Occurrence of landslide hazard along the highway is the threat to the objective of timely, efficient and qualitative construction of highway. Landslide hazard map can greatly help in fixing and shifting of the alignment to reduce the loss of life and property. This study “Landslide hazard assessment using UAV imagery and GIS for road planning and development in Chure area: Sindhuli-Hetauda section” aims at creating the hazard map, landslide inventory map and designation of hazard levels in one of the sensitive areas: Chure section. With the use of the Unmanned Aerial Vehicles (UAVs) as the primary means of carrying out the topographical surveying, the study used the Digital Surface Model (DSM) and the orthomosaic map produced from the UAV survey in acquiring the relevant results for fulfilling the study objectives. The survey area being ~100Km in length along the road alignment, four of the most crucial sites on the basis of existing landslide area and impact of those landslide in road, cultivation and settlement in the study area were selected for surveying. The study concluded that use of UAV for hazard mapping has good, accuracies and high resolution data. The Root Mean Square Error (RMSE) of the survey for the individual sites were found to be 0.001m, 0.045 m, 0.044 m and 0.804 m, respectively. A detailed topographical map of the area was created, along with the hazard map, including the factors such as slope, aspect, curvature, elevation, lithology, distance to road, distance to river and soil type. Furthermore, the hazard levels for the surveyed area were also obtained: the largest area being medium 60.68%, 57.45%, 71.21% and 71.16 %, respectively, followed by high 32.59%, 18.91%, 17.03% and 11.54% respectively and low 6.73%, 23.64%, 11.76% and 17.30%, respectively in Chattiwan, Bhawanchuli, Gurji and Hakpara. It was also concluded that the forest area is at high risk followed by the bush and the settlement area in the Chattiwan, Bhawanchuli and Hakpara site and the cultivable land followed by the bush and the settlement area were found to be in high risks in the fourth site (Gurji).


Background
Landslides are gravity-dominated mass movement that transport soils, rock, dumped waste material and artificial fills from elevated position to the down slope. Landslide generally occurs when extreme events like heavy rainfall, seismic activity, movement of mass of ground which is marginally stable take place. Landslides are common natural hazards in mountain region which are accelerated by the development processes.
Landslide hazard mapping is a basic tool which can be used to represent the hazard area with the magnitude of hazard (High, Moderate, Low). It is important to assess and ensure the year round functioning of the mountainous road for a country like Nepal where most of the terrain is occupied by hills and mountain (Pathak, 2015). Landslide hazard map indicate the areas susceptible to landslide which requires careful evaluation of hazard condition and its possible impact before road alignment is finalized. During construction of highway, alignment fixation plays important role. Occurrence of landslide hazard along the highway is the threat to the objective of timely, efficient and qualitative construction of highway (Pathak, 2015). Landslide also increases the maintenance and operation cost of highway. The ultimate aim of any road construction is to avoid landslide prone area for which landslide hazard map of that area is absolutely necessary.
Chure region is environmentally sensitive due to its loose structural nature of stone, gravel, coarse sands, mudstone and sandstones (Singh, 2017). Recently, Nepal Government announced to develop the alternative highway to connect east and west which passes through the Chure region, named Madan Bhandari Highway. This project will drastically change the area to urbanization which might lead to increase in pressure on the available resources in the Chure region. Land use pattern will be changed that may cause negative impact on the land. The improper land management along the Chure might aggravate the chances of landslide in already fragile landscape. The president of Chure-Terai Madhesh Conservation Development Board, which was formed in 2016, takes the highway construction as a challenge to conserve Chure from the aforementioned disasters (PCTMCD/GoN, 2018).

Statement of the Problem
Chure hills are formed with the highly fragile, weak and young sedimentary rocks which are highly weathered and deformed, and inter-bedding of soft mudstone and hard sandstone beds provide differential weathering providing plenty of options for slope instabilities and occurrence of different types of landslides (Dhakal, 2016). Landslides in Chure are caused by both natural and human interferences. Change of land use system and land cover due to deforestation, haphazard road construction and over exploitation of raw materials have triggered landslide in Chure area. Madan Bhandari Highway, a new road project from east to west is proposed for construction in Chure area, having length of 1200km, connecting Jhapa of eastern Nepal to Dadeldhura of western Nepal (Poudel, 2018). The new highway proposed on geologically fragile Chure area will create adverse impact on the environment (Acharya, 2018). Sindhuli-Hetauda section is one section of Madan Bhandari Highway which is under construction without hazard mapping and risk assessment. Road contributes to the largest surface erosion and landslide losses (per unit area disturbed) compared to other land use (Sidle, et al., 2006). Land use has direct impact on the soil erosion and slope stability in tropical mountain including timber harvesting, road trails, various agro-forestry practices, and conversion of forest to agricultural land and to grazing land.
There is lack of detail study of landslides and related phenomena in Chure area that creates dilemma about the landslide hazard, its distribution, and possibility of future occurrences, landslide risks and mitigation measures (Dhakal, 2016). Occurrence of landslide hazard along the highway is the threat to the objective of timely, efficient and qualitative construction of highway (Pathak, 2015). Landslide also increases the maintenance and operation cost of highway. The ultimate aim of any road construction is to avoid landslide prone area for which landslide hazard map of that area is absolutely necessary.

Significance of the study
Unmanned Aerial Vehicle (UAV) was used to collect high resolution aerial images, which portray the features that can be obtained from the satellite imagery but with higher detail of data, to produce highly accurate Digital Elevation Model (DEM) and Ortho-photo maps. The Ortho-photo map can be used for inventory mapping of landslide and the Digital Elevation Model (DEM) for landslide hazard map. Use of UAV for collection of aerial imagery to produce DEM is cost-effective as well as time efficient. Landslide hazard is a map that includes topographic and geographic information that can be used to compare against the inventory. This can greatly help in fixing and shifting of the alignment to reduce the loss of life and property. Landslide risk analysis has been used to quantify landslide risk along the road surrounding areas. As a final output, direct risk can be quantified for properties (alignments, vehicles, buildings, and plantations) and people (commuters and residents) (Jaiswal & Westen, 2012).

Scope of the Study
Landslide hazard map along Sindhuli-Hetauda road section of the Chure Area (Sindhuli-Hetauda Section) has been prepared using GIS. Weight of Evidence Model has been used during preparation of landslide hazard map with different parameters such as: Lithology, Land Use, Topography, Drainage, and Road Construction. UAV and SfM technology have been used to get the efficient DEM of the study area along with the ortho-photo map which is useful to identify the actual condition of road construction of the area. Landslide hazard map, together with information on existing or expected vulnerability, have been used to estimate the risk associated with the road networks.

Landslide Inventory in the study area
Landslide inventory map along the Hetauda -Sindhuli Road section contains the information about landslides occured in that study area, such as landslide phenomena locations, type, volume and damages. Landslides are presented as area coverage by landslide. 100km of the Hetauda-Sidhuli road section was visited during site visit. Road alignment passes through plain as well as hill side. Mainly road contains the cut area at one side, left side as well as right side as shown in figure 2.1. Length of road containing cut at right side was 15.1km and at left side was 40.78km. Road has box cutting in some point. In cut slope along the road alignment, there are landslide visible in slope which exhibit both shallow as well as deep slide. During the field visit, SW Map was used to track the alignment which was also helpful to mark the cut slope in different location. Cut slope was separated by marking start and end of the cut slope for the remarks to know the point of change in cut and fill. During the field visit, major landslides were marked by GPS in which some were active. The main reason of those slides was construction of new road alignment in fragile geology.
The major landslides available in the study area are shown in table 2.1. In preparing this table, a deliberate effort has been made to set it up according to features that are observed at the time of site visit. Mainly location, soil type, type of landslide, state of activity etc. was tabulated to provide the information of landslides which were caused by the road construction (Sidhuli-Hetauda Road Section).  (Dahal, 2006) Types of landslides observed at study area were classified on the basis of type of landslide proposed by Varnes (1978). The type of landslides was determined by a short period of observation or by the shape of the slide and arrangement of debris. Landslide types are shown in figure 2.1.
The states of activity of landslides in the study area were classified on the basis of moving nature observed. An active landslide is the one which currently moving whereas dormant landslides are inactive landslide which can be reactivated by its original causes or other causes. In study area, major landslides observed were active in state due to the higher cut slope of road in hill side and lack of protective measures in those area.
The location of the landslides in the study area were surveyed with GPS which is presented in table 2.3. In table 2.3, landslide status has been categorized into small and big on the basis of area covered by landslide on that location. Big landslide has two or more than two landslides near about the point Land use available near those existing landslides in road alignment has been presented as the vulnerable land use. The thirteen landslides areas are shown along the proposed highway in figure 2.2.

Figure 2-2 Landslide inventory map
Considering the size, type and impact of the landslide in the study area, four sites were selected as the critical section for the study in which UAV was used to survey the area. The orthomosaic, output of the UAV survey, was used to identify the existing landslide in the selected specific study area of the road section. Major land use of area was also digitized to identify the impact of landslide in those areas.

Chattiwan site
Chattiwan site was selected on the basis of existing landslide area along the road alignment and junction of the two major road (Proposed Madan Bhandari Highway and Fast Track). The study mainly focused on the Madan Bhandari highway section but newly cut area of fast track also has seen impact on the section. Orthomosaic in figure 2.3 provides the picture of the site at the time of survey. Land use of that area was obtained by digitizing the orthomosaic presented in figure 2.4. The major land use has been found to be forest which covers the 0.725 sq. km. Other land use on Chattiwan site were cultivation, settlement, stream, wall, landslide, fill, road etc. The land use, landslide indicates the existing landslide in the study area which covers the 0.883 sq.km. Table 2.5 provided information of land use with area cover by individual land use in the study area.

Bhawanchuli site
Bhuwnchuli site was selected on the basis of existing landslide area along the road alignment and impact of landslide on settlement and cultivable land. Orthomosaic in figure 2.5 provides the picture of the site at the time of survey. By using orthomosaic, land use of the study area was digitized which is presented in figure 2.6. The major land use was found out to be forest which covers the 0.298 sq. km. Other land use on Bhawanchuli site were cultivation, settlement, stream, bagar, Barren land, river, landslide, fill, road etc. The parameter landslide indicates the existing landslide in the study area which covers the 0.229 sq.km. Table 2.6 provides the area covered by the land use pattern in surveyed area.

Gurji site
Gurji site was selected on the basis of existing landslide area along the road alignment and impact of landslide on settlement and cultivable land. Orthomosaic in figure 2.7 provides the picture of the site at the time of survey. By using orthomosaic, land use of the study area was digitized which is presented in figure 2.8. The major land use was cultivation which covers the 0.520 sq. km. Other land use on Gurji site were cultivation, settlement, stream, bagar, Barren land, river, landslide, fill, road etc.

Hakpara site
Hakpara site was selected on the basis of existing landslide area along the road alignment and impact of landslide on road. Orthomosaic in figure 2.9 provides the picture of the site at the time of survey. By using orthomosaic, land use of the study area was digitized which is presented in figure 2.10. The major land use was forest which covers the 0.452 sq. km. Other land use on Gurji site were cultivation, settlement, stream, bagar, Barren land, river, landslide, fill, road, pond etc. The parameter landslide indicates the existing landslide in the study area which covers the 0.451 sq.km. Table 2.8 provides the area covered by the different land use pattern in surveyed section i.e. site IV.

Topographical map preparation for hazard mapping
For topographic map preparation, GIS based technique was used to process the DEM generated by the PixdD mapper. PixdD Mapper uses Structure from Motion algorithm (SFM) to reconstruct the actual surface from a large number of overlapping photos. The software locates matching features on each image and uses iterative bundle block adjustment to estimate image orientation, exterior orientation parameters and building model geometry. The GCPs are entered to aero triangulation which enables precise calculation of the exterior orientation parameters and improves spatial geo-referencing accuracy. The final step generates the digital surface model (DSM) by building the model texture and exporting a 3D model (mesh) or orthomosaic.
Photogrammetry produces only a surface envelope and cannot capture terrain under vegetation cover. So, UAV point cloud data were manually classified in different classes for filtering vegetation cover and buildings. The elevation model generated from all point cloud data was labelled the digital surface model (DSM) and the model from ground point cloud data was termed the digital terrain model (DTM).
Plate 2: Data recording using DGPS Plate 1: GCP setup From the finding, it can be observed that the accuracy of the topographical survey is formidable, with the Root Mean Square Error (RMSE) of the survey for the individual sites as 0.001m, 0.045 m, 0.044 m and 0.408 m respectively. Meanwhile, the errors in the X, Y and Z of the sites are good enough, and the survey conducted can hence be said to be of the surveygrade accuracy. For the absolute accuracy of the survey, the Differential GPS (DGPS) played the vital role. Table 2.9 shows the co-ordinate of ground control point used during process of UAV imagery. After completion of processing of image PixdD mapper provided the mean root mean square error (RMS) of the geo-refrencing as presented in figure 2.11 and error calculation also presented in figure 2.12.

Elevation
Topographic relief of the study area divided into different classes is often used to describe and reflect the macroscopic features of the terrain surface, which is of great significance for landslide sensitivity analysis. Elevation model is one of the major outputs of this study. Whole study area can be generalized with reference of the elevation model. The study area lies in the range of 95-533 m from the mean sea level. The resolution of output is based on the ground sampling distance (GSD). In this study resolution of the orthomosaic is one times of GSD and DTM/DEM is three times of GSD. Because of the unequal elevation difference in four different sites, they have different GSD as well as resolution of output. GSD of first, second, third and fourth site is 6.08 cm/pixel, 5.6 cm/pixel, 4.42 cm/pixel, 7.25 cm/pixel respectively.
Elevation model is often used to describe and reflect the macroscopic features of the terrain surface, which is of great significance for landslide sensitivity analysis (Wang, et al., 2017).
DEMs of different site of study generated by PixdD software were reclassified to derive the elevation model.

Slope
In our study area, slope was classified in six categories as very gentle, gentle, moderate, moderately steep, steep, and very steep are shown in figure 2.14 (a), 2.14 (b),2.14 (c) and 2.14 (d) respectively, for site Chattiwan, Bhawanchuli, Gurji and Hakpara respectively. Due to the steep road side slope, there were many landslides active in the study area. In Chattiwan site around 2 km of road section has the steep roadside slope and Bhawanchuli, Gurji and Hakpara also have steep slope. The slope angle map was extracted from the DEM.

Aspect
The aspect includes nine faces, which are flat, north, northeast, east, southeast, south, southwest, west, and northwest. The aspect map is derived by using the spatial tool of GIS software which classified the aspect on the basis of slope with respect to the north. Aspect (slope orientation) affects the exposure to sunlight, wind and precipitation thereby indirectly affecting other factors that contribute to landslides such as soil moisture, vegetation cover and soil thickness (Meten & Bhandary, 2015).

Curvature
The plan curvature is defined as the rate of change of the slope angle, which directly affects surface runoff and the development of landslides (Wang, et al., 2017). Plan curvature is classified in three categories such as concave, flat and convex. In this study Arc GIS with its spatial analysis tool was applied to derive plan curvature in which negative curvatures value represents concave, around zero (-0.1 -0.1) curvatures value represent flat and positive curvatures value represent convex surface. Concave and convex surfaces of the earth are more influential in boosting landslide to occur.
DEM produced by PixdD mapper was used to derive plan curvature of the study area.

Distance from road
The study area encompasses Madan Bhandari Highway which passes through the Chure area. Chure is the most fragile area. Changing of land use system and land cover due road construction have triggered landslide in Chure area. Due to rugged terrain, roads in the chure are characterized by high gradient, steep slopes, sharp curves including unstable geological features, mass movements, debris flow, etc. There are frequent provision/occurrence of heavy cut and fills in the hill roads which are not mechanically compacted. These loose materials cause numerous landslides in the hills. Equally, the excessive cut areas are exposed to heat and rain effects which causes slope instability, ultimately landslides are inevitable.
The layer of the road from the land use map was used to generate the distance from the road. The distance from road was classified into the separate classes to identify the impact of the road in the landslide. Figure 2.17 (a), (b), (c) and (d) respectively shows the distance from road map for Chattiwan, Bhawanchuli, Gurji and Hakpara respectively. Road network layer was developed by digitizing the road from orthomosaic. The vector layer of road was used to create Euclidean distance raster. In this study Arc GIS with its spatial analysis tool was applied to derive the Euclidean distance. The distance from road was reclassified into six classes which represent the range of distance from road. In the study area, it was observed that the landslides generally occur near the road. In Chattiwan, Bhawanchuli, Gurji and Hakpara, higher percentage of landslides were seen near the road (0-50m

Distance from stream
The Madan Bhandari Highway crosses the lot of stream in sindhuli-Hetauda section. In this study, stream layer was developed by digitizing the stream and river from orthomosaic. The vector layer of stream and river was used to create Euclidean stream distance raster. In this study Arc GIS with its spatial analysis tool was applied to derive the Euclidean distance. The distance from stream was reclassified into six classes which represent the range of distance from stream.

Geology Map
The lithology of the study area comprises three lithological units. These are  Quaternary (Q): Alluvial boulders, gravels, sands and clay  Middle Siwaliks (MS1): Fine to medium grained friable, arkosic sandstones and hard, compact massive sandstone intercalated with green to greenish grey clays, thin bands of pseudo-conglomerates and mudstones, plants and animal's fossils are present in clay.  Lower Siwaliks (LS): Fine grained hard, grey, sandstones interbedded with purple and chocolate coloured shades, nodular, maroon clays and psedo conglomerates.

Rainfall Map
The rainfall data required for the analysis was obtained from Department of Hydrology and Meteorology. The monthly average precipitation of the collected data around area varies between 69.862 mm (Nepalthok) and 206.645 mm (MakawanpurGadi).
The rainfall data of eight stations available in Sindhuli and Hetauda district were collected from Department of Hydrology and Meteorology. The average monthly data of those stations were used in preparation of isohyets map known as rainfall map. The rainfall map was

Landslide Hazard Map
For hazard map preparation, WoE method was used in this study, positive weight (W+), negative weight(W-) and contrast value were calculated. All values were joined in each class of influencing factor and the weight map of each influencing factor were developed by using the spatial analysis tool in GIS. During preparation of weight of each influencing factor of contrast value of each classes were used. Prepared weight maps were combined with raster calculator in GIS to develop landslide hazard map. The magnitude of the contrast C was determined from the difference, W+ and W-, and provided a measure of the spatial association between a set of points and the patterns. C is positive for a positive spatial association, and negative for a negative spatial association.
In   The maximum area of Chattiwan site is formed with alluvial boulders, gravel, sands and clay. From figure 2.22, maximum area of high level hazard is near about the road. Due to the presence of slope cut more than 35 degrees with south and southeast facing aspect has more influence to cause landslide near the newly constructed road alignment.
Hazard Map of CHATTIWAN   The maximum area of Bhawanchuli site is formed with fine grained hard, grey, sandstones interbedded with purple and chocolate coloured shades, nodular, maroon clays and psedo conglomerates. From figure, maximum area of high level hazard is near about the road. Due to the presence very steep slope has more influence to cause landslide near the newly constructed road alignment.   The maximum area of Gurji site is formed with fine to medium grained friable, arkosic sandstones and hard, compact massive sandstone intercalated with green to greenish grey clays, thin bands of pseudo-conglomerates and mudstones, plants and animal's fossils are present in clay. From figure, maximum area of high level hazard is near about the road. Due to the presence very steep slope and south facing aspect has more influence to cause landslide near the newly constructed road alignment. Presence of Bagmai river near this study area also has significant impact on the landslide hazard.   The maximum area of Hakpara site is formed with fine grained hard, grey, sandstones interbedded with purple and chocolate colored shades, nodular, maroon clays and psedo conglomerates but also significant part have alluvial boulders, gravel, sands and clay. In both geological formations, there are landslides. From Figure, maximum area of high level hazard is near about the road. Due to the presence very steep slope and south facing aspect has more influence to cause landslide near the newly constructed road alignment.
Presence of Marin Khola near by the study area also possesses significant impact on the landslide hazard.

Risk Assessment
For risk assessment, hazard map was combined with land use raster map by using combine tool of GIS software which is very useful for calculating the presence of hazard level in existing land use. Land use map obtained from digitizing the orthomosaic was used to produce raster map with feature to raster tools in GIS is combined. The areas under the high and medium hazard were calculated to find the risk level on existing land use.

Chattiwan
The high risk and medium risk area of Chattiwan site on the basis of presence of hazard level on land use as shown in table 2.45. The forest area of 0.143 sq. km was found to be in high risk while 0.408 sq. km of the forest areas was found to be in medium risk of landslide hazard in the Chhattiwan site. The 0.0210 sq. km of road was observed to be high risk while 0.0089 sq. km of road was observed to be in medium risk. The settlement area of .00247 sq. km was found to be in high risk and 0.000812 sq. km was found to be in medium risk. The cultivable area of 0.0232 sq. km was prone to high risk of landslide hazard and 0.0176 sq. km of cultivable areas prone to the medium risk. The big portion of the forest area in the study section is prone to the high risk of landslide hazard. The greater portion of the total settlement area is prone to the high risk of landslide hazard as compared to area bearing medium risk. The larger portion of the total cultivation area is prone to the medium risk of landslide hazard in the study area. High Risk Medium Risk

Bhawanchuli
The high risk and medium risk area of Bhawanchuli site on the basis of presence of hazard level on land use is shown in table 2.46. The forest area of 0.082982 sq. km was observed to be in medium risk of landslide hazard while the forest area of 0.183614 sq. km was in medium risk. The cultivation area of 0.065519 sq. km was in high risk and 0.317649 sq. km was in medium risk. The road area of 0.011587 sq. km was in high risk and 0.01374 sq. km in medium risk. The river area of 0.000397 sq. km was in high risk while 0.018258 sq. km in medium risk. The settlement area of 0.00599251 sq. km was found to be in high risk whereas the area of 0.011025 sq. km was in medium risk. The greater portion of total cultivation, settlement, road and river area are in medium risk while the greater portion of forest area in this section is in high risk. High Risk Medium Risk

Gurji
The high risk and medium risk area of Gurji site on the basis of presence of hazard level on land use is shown in table 2.47. In the Gurji study section, 0.004078 sq. km. of the road area was found to be in high risk and 0.021193 sq. km was found to be in medium risk. 0.081081 sq. km of total cultivation area in the Gurji study section was found to be in high risk while 0.359573 sq. km of it was in medium risk. The river area of 0.000992 sq. km was in high risk and 0.031197 sq. km was in medium risk.

High Risk Medium Risk
The settlement area of 0.009976 sq. km was in high risk of the landslide hazard in the Gurji section while the remaining 0.01885 sq. km was in medium risk of the landslide hazard. The larger portion of total area remained in medium risk of landslide hazard in Gurji section of the study area.

Hakpara
The high risk and medium risk area of Hakpara site on the basis of presence of hazard level on land use is shown in table 2.48. In Hapkara section of the study area, the road area of 0.001503 sq. km was found to be in high risk of landslide hazard and 0.019407 sq. km was in medium risk.

High Risk Medium Risk
The forest area of 0.107249 sq. km was prone to high risk of landslide hazard and 0.352208 sq. km was prone to medium risk. The cultivation area of 0.008841 sq. km was in high risk and remaining 0.175020 sq. km was in medium risk. The river area of 0.000176 sq. km was in high risk and 0.035897 was in medium risk. The pond area of 0.000397 was prone to the landslide with high risk in the study area while 0.001597 sq. km was in medium risk. The settlement area of 0.000972 sq. km was in high risk whereas 0.013704 sq. km area of the settlement zone was in medium risk of landslide hazard in the Hapkara section of the study area.

Discussion on results
Although this study has not explored and discussed the different topographic data collection method for landslide hazard mapping method, it shows that perhaps, photogrammetric techniques and aerial photographs promise to be more efficient for discerning boundaries of recently active landslides and prepare a topographic factor map (slope, aspect, curvature and elevation. Classical surveying techniques can only provide data measurements with a very low sampling, and may not provide detailed information for deformation description in the landslide area along the road. UAV have alternative advantages in capturing high-density 3D point cloud data that opens substantial potential for the applications of natural hazards assessment like landslides (Pirasteh & Li, 2017). Pirasteh & Li, 2017 also shows that 3D with high-resolution data with centimeter in pixel size and detailed information can perform a better quality of generating DEM derivatives and further landslide susceptibility maps and motivates to prepare hazard assessment via '3D point clouds' that enables us to generate detail information about topography and geomorphology.
Landslide hazard map was developed by weight of evidence method which calculated the positive weight(W+), negative weight(W-) and contrast value (C) to show the existence or non-existence of influencing factor in the landslide and correlated power of the factor. As higher value for W+, as stronger is the positive correlation. High positive correlations to landslides are obtained for slope more the 35-degree slope and nearer from road alignment. This category has also a greater predictive power (C>1). Hazard map was classified in three class high medium and low with equal interval classification basis.
Land use map of the study are shows that the road alignment mainly passes through forest, cultivable land and some settlement area. In Chattiwan and Hakpara site the there was a forest in hill side which was under the high risk where in Bhawanchuli and Gurji site forest was lied in higher risk followed by cultivation and settlement area. In this study, risk assessment carried out with overlaying the landslide hazard map and current land use map. The major factor of landslide in the study area was road construction in highly fragile area with making high slope cut that leads to the high risk in surrounding road alignment. One of the most important ways that can minimize landslide risk is to minimize road mileage and to the fullest extent possible, locate necessary roads on the more stable portions of the landscape. Structural support such as masonry retaining wall, gabion wall can be used to along the hill slope for the toe protection. In Hakpara site, there is vertical cliff in several locations that can be avoid by shifting the alignment.

Conclusion
This study "Landslide hazard assessment using UAV imagery and GIS in the Sindhuli-Hetauda section of Chure Area" mainly aimed at finding the risk area in the Hetauda-Sindhuli road section of proposed Madan Bhandari Highway acts as one of the baselines in disaster management. Acting as a preventive medium to minimize landslide risks in the affected area, these processes could facilitate other steps in disaster management such as landslide risk mapping (which involves the assessment of loss of lives and property) and its integration with road planning and development. This study used Weight of Evidence (WOE) method to prepare landslide hazard map, and led to the following conclusions: The Ground Sampling Distance (GSD) for the respective sites is 6.08 cm/pixel, 5.6 cm/pixel, 4.42 cm/pixel, 7.25 cm/pixel. Since the RMSE is very less and resolution of the data is very high, the use of UAV was found to be fruitful in preparing the topographic map.
 Majority of landslide along the road in the study area were found to be active, with only one being dormant. The active landslides were mainly of the type fall owing to the steep road cut slope while small number of the landslide was slide and rotational.
 The study area (consisting of four individual sites within the area) was categorized into three landslide hazard levels as low, medium and high. In all the four sites, the obtained hazard levels are unique, owing to the uniqueness in the influencing factors. As per the results obtained from hazard map, the largest area being medium 46.39%, 57.45%, 71.21% and 71.16 % respectively, followed by high 28.47%, 18.91%, 17.03% and 11.54% respectively and low 25.14%, 23.64%, 11.76% and 17.30% respectively in Chattiwan, Bhawanchuli, Gurji and Hakpara.
 The land use map is combined with landslide hazard map to prepare inventory of the elements at risk. In all three sites Chattiwan, Bhawanchuli and Hakpara, forest is at high risk followed by cultivation and settlement, while in Gurji, mainly cultivable land is in risk followed by bush and settlement.
In order to maintain the balance between the development and the its impact in environment in Chure area of Nepal, in which land degradation is primarily contributed by different types of landslides and mass wasting phenomena which are collectively controlled by the lithospheric plate dynamics, geology, topography, intense precipitation and human interference, hazard assessment and proper planning before road construction is absolutely necessary.

Research Design
This study utilized the quantitative methodology. High resolution imagery were used to get topographic data, landslide data and land use data by using UAV and SfM software and some other data such as geological, rainfall, land use data were obtained from the Department of Mines and Geology(DMG), Department of Hydrology and Meteorology(DHM), SOTER. For landslide hazard assessment weight of evidence method (Statistical method) was used. In this method, each factor maps (slope, geology, land use etc.) was combined with the landslide distribution map, and weight values, based on landslide densities, was calculated for each parameter class (slope class, lithological unit, land use type, etc.). For risk assessment quantitative method was used. Risk area of the land use was quantified by combining hazard map with land use map. Risk over the settlement, forest, cultivation, road and other natural resources near the road have been identified with spatial analysis tool in GIS. Overall research design to obtain the research objective is shown in figure 4.1.

Hazard map preparation
In this study WoE method is used to prepare the landslide hazard map. The method was originally developed for mineral potential mapping but recently the method has been used in landslide susceptibility analysis by several researchers in which evidence maps are generally converted to binary predictor maps in order to apply the model (Mezughi, et al., 2011).Weights were calculated for the initial categorization of the parameters derived from DEM (Slope, Aspect, Curvature, Elevation) and for other parameter such as distance from road, distance from stream, geology, soil. As per following equation, positive weight (W+) as well as negative weights (W-) were calculated ). The contrast (C) is calculated as the difference between the two weights (C = W+ -W-) where C is positive for a positive spatial association indicating the factor is favorable for the landslides, but C is negative if the spatial association is negative indicating that the factor is unfavorable. The magnitude of the contrast indicates an overall of spatial association between the causative factor and landslides whereas C equal to zero when a class has no spatial relationship with landslides occurrence. The weights and contrast value of all influencing factor is shown in table below. In GIS, the calculated weights and contrast have been assigned with join and relate in the existing thematic layer to produce multi-class weighted maps for all evidences. Then, raster map of all influencing factor named as contrast class raster were generated with the lookup tools in spatial analysis tool. For the final hazard map raster calculator was used to combine all contrast class raster map. Figure 4.2 shows the flowchart of hazard map preparation in this study.