Landslide Susceptibility Assessment Based on Geological Data and Remote Sensing Processed Data Using Bi-Variate Statistical Model and GIS: A Case Study on Rangamati District, Bangladesh


 Landslide is a major hazard all over the world and a very frequent phenomenon in Southeast Asian mountainous terrain. It has become a frequent phenomenon in the hilly regions of Bangladesh and the Rangamati district is the worst victims. This study aims to identify the landslide susceptible area of Rangamati district according to geological and satellite image processed data of the previously occurred landslide. Geological map of Rangamati District, Landsat 8 OLI images of October-2019 and SRTM images were used for landslide factor data mining. In present study Surface geology, Distance from fault line, Slope Angle, Slope Aspect, NDVI, BSI and NDCI are considered as the major factors related to landslide occurrence. Effects of the selected factors were calculated by using bivariate statistical approach (frequency ratio method) to examine the correlation between landslide occurrence and the influencing factors. Results of the landslide susceptibility show that surface geology and fault line has a strong correlation with landslide occurrence in Rangamati district. Analysis also shows that medium to high clay content availability with bare soil surface (no or low vegetation) that have a low slope angle (3 to 16 degree) are closely correlated to landslide occurrence. In the rainy season clay content saturated by heavy rainfall produce mud that flows according to very low slope cause landslide in this region. Due to geological characteristics this landslide can be defined as mud flow. Landslide susceptible map shows that the area near water bodies and lower part of hill slope are more susceptible to landslide occurrence.


Introduction
Landslides and associated slope failure phenomena (e.g. debris ows and mudslides) are a major hazard around the world (Chapagai 2011). Cited in Farahmand and AghaKouchak (2013), each year, it cause thousands of casualties and billions of dollars in damages across the world and results in 10 of deaths and over 1-2 billion USD in property damages (USGS 2006). For example in 1982-1983 millions of dollars loss occurred in USA due to several storm triggered landslide and in 1999 thousands of people died due to landslide in Vargas, Venezuela (Spiker and Gori 2003;Hong et al. 2006). Also in South and South-east Asia landslides are most widespread and frequent disaster because of climatic condition (rainfall), mountainous terrain and socioeconomic conditions (Apip et al. 2010;Chapagai, 2011).
In the hilly region of Bangladesh, especially in Chattagram region, during monsoon, landslides has become a very common natural hazard and are occurring The higher hill ranges in the Chattagram Hill Tracts, Chattagram and Sylhet regions are underlain by rocks of the Surma and Tipam formation (Brammer 1996;BBS 2009a, b). The composition of soil of these hilly regions is complex and the young rocks have higher contents of easily-weatherable feldspars [1].
Therefore, the soil of this hilly region is very much susceptible to the landslide risks during heavy rain carried by monsoon wind system (Sarker and Rashid 2013). Now a day's Landsat images are easily available and are used to explore various indices and characteristics of land surface. In this study Landsat images are used to explore the topographic and lithologic condition of the study area. Different bands of remote sensing image can provide information about the chemical composition of rocks and minerals that area on the Earth's surface and not completely covered by dense vegetation (Clark et al. 2003;Clark et al. 1990).
To take any action plan for a landslide affected region it is important to know the type of land and characteristics of land slide of that region (Chapagai 2011). Since landslides generally occur in existing slide areas, it is vital to know the locations of previously occurred landslides, the size of the landslides, and their related geomorphological factors (Chen et. al. 2017). Many study conducted for assessing landslide susceptibility of Rangamati Sadar Upazila by using statistical method and GIS where satellite image derived landslide triggering data were used. The major satellite derived data used in this study are land cover and NDVI and some common morphological data such as DEM, elevation, slop, aspect plan curvature etc. Considering the above aspect, this research attempts to explore the geomorphic and geologic status of the landslide occurrence area to produce Landslide Susceptibility Map (LSM) by GIS and remote sensing based approach. At rst various topographic characteristics of the study area derived from Landsat imageries. Factors responsible for landslide occurrence such as vegetation, bareness, slope angle and aspect, clay content surface geology and fault line distance are analysed to ful ll rst objective of the research. These factors are related to the lithologic and geologic condition of the study area. For the second objective, correlation between landslide occurrence and the selected factors are calculated.
Footnote: [1] Feldspars are group of rock forming minerals make up as much as 60 % of the earth crust.

Study Area
Rangamati district has an area of 6116.13 sq km and is located in between 22°27' and 23°44' north latitudes and in between 91°56' and 92°33' east Flow chart of the broad steps followed in this research for deriving results are given in Figure 2.

Data and Data Source
Historical landslide events of Rangamati district were considered in this research (Table 1). Geological map of Bangladesh was collected from Geological Survey of Bangladesh (GSB) server (Link: http://www.gsb.gov.bd/site/view/commondoc/Geo-scienti c%20Map/-?page=3&rows=20). Landsat-8 OLI images and SRTM elevation images of the study area were collected from USGS server (Link: https://earthexplorer.usgs.gov/). Landsat image of rainy season was collected to measure the soil condition of the study area because landslide occurs during monsoon season. Table 1 shows the historical landslide events of Rangamati district and their common cause, casualties and number of wounded people.
General assumption to predict landslide is that landslide occurrence is determined by landslide related factors and the future landslides will occur under the same conditions as the past landslide (Lee and Talib  where Wij shows the FR of class i of parameter j; FLij is the frequency of occurrence of landslides for class i of parameter j; and FNij is the frequency of nonoccurrence of landslides for class i of parameter j. Then the Probability Ratio of each factor was calculated and nally by using ArcMap 10.5 Landslide Susceptibility Map (LSM) of Rangamati district was produced.
Where, Wij is the weight of class i of parameter j, and n is the number of conditioning factors.

Result And Discussion
In generally landslide occurred in relatively high elevated area. This research attempts to explore the land characteristics of previously occurred landslide area and produce a LSM according to the result. Findings show that the major surface geology that responsible for landslide occurrence is Dihing and Dupitila Formation Undivided and Boka Bil Formation. 80% landslide occurred in his two geological formations where Boka Bil formation alone is responsible for 47% landslide (Table 4 and Fig. 3). Distance from fault line has major role in landslide occurrence and 93% of the landslide occurred within a distance of 11000 meter from fault line where 33% within a distance of 5500 meter (Table 4 and Fig. 4). Twelve out of fteen landslide event occurred where slope angle is between 3.71 and 16.85 degree and nine of them occurred where slope is below 12 degree (Fig. 6 Tables 3 and 4). Though aspect plays major role in weathering material for occurring landslide, in this area it is oriented according to water body (Kaptai Lake and Karnafuli river). In Rangamati, slope aspects that have similar orientation with fault line have most correlation with landslide occurrence. So that north, northeast, south and southwest slope aspect has major correlation with landslide occurrence.  that has partly mixture of unconsolidated rock and sandstone. Due to the constituting material this kind of landslide can be de ned as mud ow as we see this needs very low slope for triggering the landslide.

Distance from Fault Line
In Rangamati District about 93% landslides occurred within 11000 meters (11 Km) from the fault line (Table 4 and Figure 4). This indicates that like all over the world fault line plays a major role in occurring landslides in Rangamati District. Fault lines of Rangamati District are orientated southwest and north east direction ( Figure 3). Due to this orientation land slide occurrence have shows strong correlation with the slope that have north, northeast, south and southwest aspect (Table 3 and (Table 3 and Figure 5.C), which means landslide event occurred in the area where soil is composed with higher silt and clay content that can be easily saturated by rainwater. During heavy rainfall this saturated soil increases load and makes the mud eligible for ow. This mud load, by the pull of gravity, ows downward though it has not very steep slope ( Figure 6.B and Table 3).  (Table 3 and Figure 6.B) which are highly expose to sun so that weathering occurred in this site rapidly. Rainfall also plays role in weathering and this site has get substantial rainfall during rainy season that loosen the soil and make it susceptible to landslide. The resulted aspects are also facing towards water body such as Kaptai Lake and Karnafuli River.

Landslide Susceptible Area
Occurrence of landslide in an area depends largely on complex interaction among a large number of factors (Dai et al. 2002). So the most exact method to landslide susceptibility is through eld surveys (Catani et al. 2005;Nadim et al. 2006). But satellite remote sensing data can be a great information source for landslide susceptibility analysis. Data derived above are factors of landslide occurrence. By using frequency ratio method frequency ratio, relative frequency and predictability of each factor are identi ed. Frequency ratio indicates the correlation of landslide with the corresponding class. Results show that low slope has greater correlation with landslide occurrence (Figure 7 and Table 4). Northeast and southwest facing slopes also show same correlation (Figure 7 and Table 4). Greater correlation found for high to very high NDCI (Figure 7 and Table 4). Two major geological formations and low distance from fault has also grater correlation with landslide occurrence. According to this result weighted map of landslide susceptility of Rangamati district shows medium to high landslide susceptible areas are located near Kaptai Lake and Karnafuli river and in the slope of hills (Fig. 7). These areas are adjacent to fault line and have similar orientation with the slope aspect and have a low slope. There are many areas that are moderately susceptible to landslide during heavy rainfall.  Landslide susceptibility map of Rangamati District