Development of landslide susceptibility maps of Tripura, India using GIS and analytical hierarchy

16 Landslides are one of the most extensive and destructive geological hazards in the globe. Tripura, a north- 17 eastern hilly state of India experiences landslides almost each year during monsoon season causing casualty and 18 huge economic losses. Hence it is required to assess the landslide susceptibility of the area that would support in 19 short and long term planning and mitigation. Analytic hierarchy process (AHP) integrated with geospatial 20 technology has been adopted for landslide susceptibility mapping in the state. Eight influencing factors such as 21 slope, lithology, drainage density, rainfall, land use land cover, distance from riversand roads, and soil type were 22 selected to map the landslide susceptibility. Landslide susceptibility index (LSI) was found to vary from6.205 23 during monsoon to 1.427 during post-monsoon season. The LSI values were classified intovery high, high, 24 moderate, low and very low susceptibility. Landslide susceptibility maps for three different seasons, namely, 25 pre-monsoon, monsoon and post-monsoon were prepared. The study showed that most of the areas of the state 26 come under very low to moderate landslide susceptibility zones. Around 73.2% area of the state is found to be 27 under low landslide susceptible zones during the pre-monsoon season, around 62% area is prone to landslides 28 with moderate susceptibility during monsoon season and 68.5% area comes under landslides with low 29 susceptibility zones during the post-monsoon season. The output of this study may be referred by the engineers 30 and planners for the assessment, control and mitigation of landslides and development of basic infrastructure in 31 the state. 32

According to the CRED (2009), landslides accounted for about 4.4% of global natural disasters during 1990 -be achieved only if accurate information on risk management is available. Landslide susceptibility mapsgenerated using remote sensing data has been proven to be an important source of information for the (Razak et  Occurrences of new landslides and reactivation of the old landslides are mainly responsible for frequent disturbancesin the National Highway of Tripura, which is considered to be the lifeline of the state (Ghosh et al. of the landslide incidents in the state. Hence, it is essential to identify vulnerable areas prone to landslides and 100 prepare landslide susceptibility maps, which will help the scientists, researchers and the decision makers to 101 undertake suitable measures to cope up with this natural cause. Keeping the above facts in view the present 102 study has been undertaken to identify the critical factors causing landslides and to develop landslide susceptibility maps of the state of Tripura (India) using remote sensing, GIS and analytical hierarchy process soils, younger alluvial soils, and lateritic soils occupy less than 10%area each. In the study area, the main cause of landslides is the failure of slopesdue to high intensity rainfall during the monsoon season.

Sources of data and Preparation of thematic maps
Different types of primary and secondary data were used to prepare the seasonal landslide susceptibility maps of Tripura state. Some of the data were generated from satellite imagery while some were developed or collected 122 from potential sources. The development of landslide susceptibility maps of the study area involves the 123 preparation of thematic layers of the landslide causative factors. Eight factors such as land slope, drainage density, distance from rivers, distance from roads, land use land cover (LULC), lithology, rainfall and soil types

128
Convectional/existing layers such as lithology and soilmaps used in the study were obtained from different 129 sources (Table 1). These maps were scanned and introduced into ArcGIS 10.6.1 for generation of thematic 130 layers by properly following the rectification and digitalization process. LULC map was taken from the National  Table 1. All the thematic layers were prepared using ArcGIS software.

134
The slope map anddrainage density map of the area understudy were generated from the SRTM DEM using 135 spatial analyst tool in ArcGIS.Mapsshowing the distances from rivers and roads were generated from the DEM 136 using the Euclidean distance tool in ArcGIS. Average rainfall maps for three different seasons such as pre-137 monsoon (March to May), monsoon (June to October) and post-monsoon (November to February) were 138 prepared from the daily rainfall data of seven rain gauge stations (Agartala, Amarpur, Dharmanagar, Khowai,

139
Sabroom, Sakhan, and Sonamura) within the study area. The rainfall data for a period of 22 years (1998 to 140 2019) was downloaded from the NASA Power Data Access Viewer website. The daily data was converted to 141 seasonal data considering three seasons as stated above.Inverse distance weighting (IDW) interpolation 142 technique in ArcGIS was used to prepare the seasonal rainfall maps.  assigning values between 1 and 9. The reciprocal weights 1/2 to 1/9 were also used for inverse comparison. Table 2 shows the measurement scale of AHP as suggested by Saaty (1980). One activity is strongly favored over another and its dominance is shown in practice Extremely 9

161
The evidence of favoring one activity over another is of the highest degree possible of an affirmation Intermediate values 2, 4, 6, 8 Used to represent compromises between the preferences in weights 1, 3, 5, 7, and 9 respectively Reciprocals Opposites Used for inverse comparison In this hierarchical classification approach, the consistency of weights assigned to different layers was checked by calculating the consistency ratio (CR). This step was used to detect any inconsistencies in the comparison of the importance of each pair of criteria. The CR can be expressed as: is the principal eigenvalue and nrepresents the number of criteria or factors.
The RI is a value that depends on the size of matrix or the number of parameters used for pair-wise comparison consistent.Otherwise, re-evaluation of the corresponding weights should be done to avoid inconsistency.

179
The landslide susceptibility maps of Tripura state were prepared using AHP technique. After generating the

190
On the basis of LSI values, the study area was divided into five classes, namely very low, low, moderate, high,

191
and very high zones susceptible to landslides. The landslide susceptibility maps for winter, pre-monsoon, 192 monsoon and post-monsoon seasons were prepared using the weighted sum tool in ArcGIS software.

194
Thematic maps of Tripura State

195
The landslide susceptibility maps of Tripura State were prepared in the GIS platform considering eight landslide

205
The slope of any terrain makes the soil more vulnerable to landslides. In areas having gentle slope, the velocity 206 of flow is low, which allows the runoff water to get absorbed into the soil, whereas steep slope areas facilitate landslide susceptibility was found to vary from 46.2% for slopes more than 45% to 3.1% for slopes less than 5% (Table 4).   agriculture, scrub land, barren land and shifting cultivation were considered (Fig. 2d). The weights assigned to 253 different classes were used based past studies and the expert's opinions. The influence of shifting cultivation 254 towards landslide susceptibility was found to be the highest at 32.8%, whereas it was least at 2.1% only for 255 water bodies (Table 4).

256
Distance from roads map

257
The nature of topography is changed by construction of roads, and hence, the possibility of landslides increases based on their distances from the roads (Fig. 2e). Higher weights were assigned to the class nearer to the roads 265 and lower weights were giver to the class far from the roads. The influence of distance from roads of different 266 classes on landslide susceptibility was found to vary from 50.3% for distance lower than 50 m to 3.5% for 267 values exceeding 350 m (Table 4).

269
The presence of loose surface soils increases landslides susceptibility. Landslide is activated when rainwater  (Table 4).
that lithology significantly influences the occurrence of landslides. This is due to the fact that the variations in 279 lithology often lead to significant differences in the permeability and strength of rocks and soils.The map was classified into six classes, namely Tipam, Dupitila, Bokabil, Bhuban, Quaternary and Alluvium, which are the prominent classes available in the study area (Fig. 2g). The Bhuban lithological units were assigned maximum 282 weight because this formation is characterized by a thinly bedded moderate to highly weathered sandstone shale.

283
The weights were given to different formations based on the experts' opinions and past studies. The influence Rainfall is one of the important triggering factors for landslides. High intensity rainfall causes heavy runoff.

288
Changes in soil moisture regime produce hydrostatic pressure. The mobilized shear resistance decreases with 289 increase in pore water pressure, which is likely to cause shear instability. Average seasonal rainfall maps were 290 prepared from the daily rainfall data that was collected for the last 22 years (1998 to 2019). It was found that the 291 average seasonal rainfall ranges from 518.64 to 721.18 mm during pre-monsoon season (Fig. 2h). The study 292 area experiences an average rainfall of 1534.67 to 1889.19 mm during the monsoon season (Fig. 2i)and the post-293 monsoon season receivesan average rainfall of 91.26 to 92.15 mmonly (Fig. 2j). The average seasonal rainfall 294 was classified into nine classes such as <200, 200-400, 400-600, 600-800, 800-1000, 1000-1200, 1200-1400,

295
1400-1600 and >1600 mm, and weights were assigned to them (Table 4). The weights assigned were then 296 normalized. The influence of average annual rainfall exceeding 1600 mm on landslide susceptibility was found 297 to be maximum (30.7%). It reduced with decrease in average annual rainfall. The influence was only 1.9% at 298 rainfall below 200 mm per annum (Table 4).
All the maps (thematic layers) were then arranged according to their impact on landslide, which was decided on the basis of relative weights derived based on experts' opinions and past studies. A pair-wise comparison matrix was prepared to find the influence of different factors on landslide susceptibility (

337
The study also depicts that around 45% of the study area has Tipam soil formations (containing ferruginous 338 sandstone with siltstone and clay) and 30% area has Bokabil formations (containing sequence of sandy shale-