Arab J Geosci DOI 10.1007/s12517-015-1891-7
ORIGINAL PAPER
A comparative study of statistical index and certainty factor models in landslide susceptibility mapping: a case study for the Shangzhou District, Shaanxi Province, China Chengxi Zhao 1 & Wei Chen 2 & Qiqing Wang 1 & Yanli Wu 1 & Bo Yang 3
Received: 15 December 2014 / Accepted: 16 March 2015 # Saudi Society for Geosciences 2015
Abstract Landslide susceptibility maps are vital for planning development activities in the mountainous areas in China. The main goal of this study was to produce landslide susceptibility mapping by statistical index (SI) and certainty factor (CF) models for the Shangzhou District of Shangluo City, China. For this purpose, a landslide inventory map with a total of 145 landslide locations was compiled from various sources such as aerial photographs and field surveys, out of which 101 (70 %) were randomly selected for training the models, while the remaining 44 (30 %) were used for validating the models. In this case study, the following landslide conditioning factors were evaluated: slope angle, slope aspect, curvature, elevation, lithology, distance to faults, distance to rivers, distance to roads, precipitation, and peak ground acceleration were considered in this study. The validation of landslide susceptibility maps were carried out using areas under the curve (AUC). From the analysis, it is seen that the CF model with a training accuracy of 70.48 % and predictive accuracy of 68.86 % performs slightly better than SI model (training accuracy, 70.19 %; predictive accuracy, 68.67 %). Overall, both of these two models showed almost similar results. The resultant susceptibility maps can be useful for general land use planning for the study area and other similar areas in the world.
* Wei Chen
[email protected] 1
China University of Mining and Technology, Xuzhou 221116, China
2
School of Geology and Environment, Xi’an University of Science and Technology, Xi’an 710054, China
3
Shaanxi Guangxin Mining Development Co., LTD, Xi’an 710077, China
Keywords Landslide . Susceptibility mapping . Statistical index (SI) . Certainty factor (CF) . Geographic information system (GIS)
Introduction In recent years, growing population and development of settlement, infrastructures, and life-lines have largely increased the impact of natural hazards both in industrialized and developing countries (Guzzetti 2005). The study of landslides has drawn worldwide attention mainly due to its increasing awareness of socio-economic impact. In China, a significant number of landslides occur each year, causing large numbers of casualties and huge economic losses, e.g., 15,403 landslides occurred in 2013, causing a total of 676 people dead or missing, 264 people injured, and a direct economic loss of 10.2 billion CNY. Landslide susceptibility is defined as a quantitative and qualitative assessment of the classification, volume (or area), and spatial distribution of landslides which exist or potentially may occur in an area (Fell et al. 2008; Grozavu et al. 2013). Landslide susceptibility mapping relies on understanding complex mass movement processes and their controlling factors (Ayalew and Yamagishi 2005). An accurate and reliable landslide susceptibility map requires high-quality data to make useful decisions, and an appropriate methodology for analysis and modeling. Statistical methods have become well established in landslide susceptibility studies particularly with the increasing sophistication of the Geographic Information Systems (GIS), allowing integration of data collected from various sources and methods and at different scales. Over the decades, various attempts have been made on spatial prediction of landslides using statistical models and landslides data. These statistical models used in landslide susceptibility
Arab J Geosci
mapping include frequency ratio (FR) model (Pradhan 2010; Ozdemir and Altural 2013; Park et al. 2013; Regmi et al. 2014), certainty factor (CF) model (Devkota et al. 2013; Pourghasemi et al. 2013c), weights-of-evidence model (WoE) (Pradhan et al. 2010; Regmi et al. 2010, 2014; Ozdemir and Altural 2013; Pourghasemi et al. 2013b, c), statistical index (SI) model (Bui et al. 2011; Pourghasemi et al. 2013a; Regmi et al. 2014; Chen et al. 2014;), and logistic regression (Nefeslioglu et al. 2008; Akgun 2012; Xu et al. 2012; Solaimani et al. 2013; Devkota et al. 2013; Ozdemir and Altural 2013; Kundu et al. 2013; Park et al. 2013; Grozavu et al. 2013). All these models provide solutions for integrating
Fig. 1 Location of the study area
information levels and mapping the outputs. Generally, although many different models and techniques for landslide susceptibility mapping have been proposed and implemented, no agreement has so far been reached on which model and technique are the best for landslide susceptibility mapping (Wang and Sassa 2005). The main purpose of this paper is to develop landslide susceptibility maps of the Shangzhou District of Shangluo City, China (Fig. 1), using two statistical models: statistical index (SI) and certainty factor (CF) models, in order to find the better model that is more accurate in landslide susceptibility mapping for the study area.
Arab J Geosci
The study area
Conditioning factors
The study area is located almost in the middle part of China, between the longitude 109° 30′ E and 110° 16′ E, and between the latitude 33° 36′ N and 34° 12′ N, covering a surface area of 2672.0 km2 (Fig. 1). The geological, geomorphological, and hydrogeological environment of the study area is favorable to landslide activity, and many landslides have occurred in the past. The altitude ranges from 600 to 2060 m above the sea level. The study area belongs to the typical subtropical continental monsoon climate. The mean annual temperature is 12.9 °C while the maximum and minimum mean annual temperatures range between 39.8 and −14.8 °C, respectively. According to observations over the last decade, the maximum mean daily rainfall in the area varies from 53 to 106 mm. The main rainy months are from June to September. The increasing population and new residential areas with inappropriate land use of the study are the main factors for the increasing frequency of landslides.
To conduct any landslide susceptibility assessment accurately, researchers must evaluate landslide conditioning factors. In this study, various thematic data layers representing landslide conditioning factors namely slope angle, slope aspect, curvature, elevation, lithology, distance to faults, distance to rivers, distance to roads, precipitation, and peak ground acceleration, a total of ten potential predisposing factors, were considered in this study. A digital elevation model (DEM), with a resolution of 50× 50 m, was generated from topographic maps in 1:50,000 scale with a contour interval of 20 m. Using this DEM, slope angle, slope aspect, curvature, and elevation were produced. The slope angle map was reclassified into seven classes (Fig. 2a): (1) 0–8°, (2) 8–16°, (3) 16–24°, (4) 24–32°, (5) 32–40°, (6) 40–48°, and (7) 48–59°, respectively. The slope aspects are grouped into nine classes such as flat (−1), north (337.5°–360°, 0°–22.5°), northeast (22.5°–67.5°), east (67.5°–112.5°), southeast (112.5°–157.5°), south (157.5°–202.5°), southwest (202.5°–247.5°), west (247.5°– 292.5°), and northwest (292.5°–337.5°), respectively (Fig. 2b). The curvature map was produced using ArcGIS 10.0 software and was reclassified into three categories (Fig. 2c): (1) concave, (2) flat, and (3) convex, respectively. An elevation map was prepared from the 50×50 m digital elevation model and grouped into seven categories (Fig. 2d): (1) 600–800 m, (2) 800–1000 m, (3) 1000–1200 m, (4) 1200– 1400 m, (5) 1400–1600 m, (6) 1600–1800 m, and (7) 1800– 2060 m, respectively. The geological map of the study area was prepared by the China Geological Survey at 1:100,000 scale, and digitized in ArcGIS 10.0 software. The study area is covered with various types of lithological formations. The geological map was also used to produce the lithology map, which was divided into five categories (Fig. 2e): (1) hard intrusive rocks, (2) hard limestone, (3) soft metamorphic rocks, (4) sandstone and mudstone, and (5) clay, sand, and gravel, respectively. Using the geological map of the study area, four different buffer zones (<2 km, 2–4 km, 4–6 km, and >6 km) were created within the study area to determine the degree to which the faults affected the slopes (Fig. 2f). In order to assess the degree of stream lines on landslide occurrences, four different buffer zones were created within the study area (Fig. 2g). The rivers buffer categories were defined as follows: (1) <500 m, (2) 500–1000 m, (3) 1000– 1500 m, and (4) >1500 m, respectively. Five different buffer zones were created within the study area to determine the degree to which the road-cuts affected the slopes (Fig. 2h): (1) <500 m, (2) 500–1000 m, (3) 1000– 1500 m, (4) 1500–2000 m, and (5) >2000 m, respectively. Rainfall is also widely considered as a main triggering factor of landslides. In this study, the maximum mean daily rainfall in the area was used as a conditioning factor for
Data preparation For the landslide susceptibility mapping, the main steps were data collection and construction of spatial database from which relevant factors were extracted, followed by assessment of the landslide susceptibility using the relationship between landslide and landslide predictive factors, and validation of results (Ercanoglu and Gokceoglu 2004; Regmi et al. 2014). Various thematic data layers representing landslide conditioning factors, such as slope angle, slope aspect, curvature, elevation, lithology, distance to faults, distance to rivers, distance to roads, precipitation, and peak ground acceleration were prepared. These conditioning factors are extracted from pertinent data, which can be collected from available resources and from the field, using ArcGIS 10.0 software. A landslide distribution map was also prepared using earlier reports, aerial photographs, and field surveys. Landslide inventory A landslide inventory map is one that identifies the definite location of the existing landslides along with its type and the time of occurrence (Wieczorek 1984; Einstein 1988; Soeters and van Westen 1996). In order to produce a detailed and reliable landslide inventory map, extensive field surveys and observations were performed in the study area. A total of 145 landslides were identified and mapped by evaluating aerial photos in 1:50,000 scale with well supported by field surveys and subsequently digitized for further analysis. The locations (centroid) of 145 landslides are mapped in Fig. 1. From these landslides, 101 (70 %) randomly selected were taken for making landslide susceptibility models and 44 (30 %) were used for validating the models.
Arab J Geosci
Fig. 2 a Slope angle map of the study area. b Slope aspect map of the study area. c Curvature map of the study area. d Elevation map of the study area. e Lithology map of the study area. f Distance to faults map of
the study area. g Distance to rivers map of the study area. h Distance to roads map of the study area. i Precipitation map of the study area. j PGA map of the study area
Arab J Geosci
Fig. 2 (continued)
landsliding. The precipitation map was prepared using rainfall data from monitoring stations, and was divided into six categories (Fig. 2i): (1) <60 mm/day, (2) 60–70 mm/day, (3) 70– 80 mm/day, (4) 80–90 mm/day, (5) 90–100 mm/day, and (6) 100 mm/day, respectively. Earthquake is widely considered as a main triggering factor of landslides. In this study, the peak ground acceleration map was produced from the seismic ground motion parameter zonation map of China, and was divided into three categories (Fig. 2j): (1) 0.05 g, (2) 0.10 g, and (2) 0.15 g, respectively. The abovementioned ten factors were all converted into a raster grid with 50×50 m pixels in ArcGIS 10.0 software as input data for the landslide susceptibility mapping.
Modeling approach Statistical index The statistical index model is a bivariate statistical analysis for landslide susceptibility mapping, and it was proposed by van Westen et al. (1997). A weight value for each parameter class is defined as the natural logarithm of the landslide density in the class divided by the landslide density in the entire map (van Westen et al. 1997; Rautela and Lakhera 2000; Cevik and Topal 2003; Bui et al. 2011; Regmi et al. 2014; Pourghasemi et al. 2013a). This method is based upon the formula given by Van Westen as follows: . Di j Nij N W i j ¼ ln ¼ ln D Si j S
ð1Þ
where Wij is the weight given to a certain class i of parameter j. Dij is the landslide density within class i of parameter j. D is the total landslide density within the entire map. Nij is the number of landslides in a certain class i of parameter j. Sij is the number of pixels in a certain class i of parameter j. N is the total number of landslides in the entire map. S is the total pixels of the entire map.
The statistical index method is based on statistical correlation of the landslide inventory map with the explanatory attributes of the parameter maps. It means that the Wij is only calculated for landslide occurrence classes. If the parameter class contains no landslide occurrence, it will have no correlation with the landslide inventory (Pourghasemi et al. 2013a).
Certainty factor Among the commonly used GIS analysis models, the CF model has been widely considered and experimentally investigated in the literature for landslide susceptibility (Lan et al. 2004; Pourghasemi et al. 2012, 2013c; Devkota et al. 2013). The CF approach is one of the favorable functions proposed for handling the problem of combination of different data layers and the heterogeneity and uncertainty of the input data. The main difference is the bivariate model with other models of how to combine the maps. Thus, the maps are classified and the weight of each pixel is obtained using Eq. 2: 8 PPa ‐PPs > > if PPa ≥ PPs < PPa ð1‐PPs Þ CF ¼ ð2Þ PPa ‐PPs > > if PPa < PPs : PPs ð1‐PPa Þ where PPa is the conditional probability of landslide event occurring in class a. PPs is the prior probability of total number of landslide events in the study area S. With the use of the CF model, each class or area is assigned to a value that varies within the interval [−1, 1]. A positive value means a growth in the certainty of the landslide occurrence, whereas a negative value coincides with a decrease in the certainty of landslide occurrences. A value close to 0 reflects the inadequacy of information available regarding the variable. Thus, it is difficult to give information about the certainty of landslide occurrences. The CF values are calculated for all condition factors by overlaying and calculating the landslide frequency as given. The CF values of all parameters in ten landslide conditioning factors were determined using Eq. 3. Next, the CF values
241,747 393,006 79,890 257,949 319,470 254,181 118,171 23,702 3281 191,634 37,066 180,980 509,111 137,924 484,364 237,004 174,624 160,723
Flat Convex 600–800 800–1000 1000–1200 1200–1400 1400–1600 1600–1800
1800–2060 Hard intrusive rocks Hard limestone Soft metamorphic rocks Sandstone and mudstone Clay, sand, and gravel <2 2–4 4–6 >6
Distance to faults (km)
Lithology
Elevation (m)
Curvature
Slope aspect
194,501 279,183 336,855 207,087 37,632 1336 50 12,950 118,944 123,395 153,143 128,232 117,766 121,974 153,014 127,226 421,890
0–8 8–16 16–24 24–32 32–40 40–48 48–59 Flat North Northeast East Southeast South Southwest West Northwest Concave
Slope angle (°)
No. of pixels in domain
Class
0.003 0.181 0.035 0.171 0.482 0.131 0.458 0.224 0.165 0.152
0.229 0.372 0.076 0.244 0.302 0.241 0.112 0.022
0.184 0.264 0.319 0.196 0.036 0.001 0.000 0.012 0.113 0.117 0.145 0.121 0.111 0.115 0.145 0.120 0.399
Percentage of domain
Spatial relationship between each landslide conditioning factor and landslide by SI and CF models
Factors
Table 1
0 7 2 21 52 19 54 21 13 13
25 32 13 38 31 17 2 0
26 21 27 19 8 0 0 1 14 10 14 13 9 14 17 9 44
No. of landslide
0.000 0.069 0.020 0.208 0.515 0.188 0.535 0.208 0.129 0.129
0.248 0.317 0.129 0.376 0.307 0.168 0.020 0.000
0.257 0.208 0.267 0.188 0.079 0.000 0.000 0.010 0.139 0.099 0.139 0.129 0.089 0.139 0.168 0.089 0.436
Percentage of landslide
0.000 −0.962 −0.572 0.194 0.066 0.366 0.154 −0.076 −0.250 −0.167
0.079 −0.160 0.532 0.433 0.015 −0.357 −1.731 0.000
0.335 −0.240 −0.176 −0.041 0.799 0.000 0.000 −0.213 0.208 −0.165 −0.045 0.059 −0.224 0.183 0.150 −0.301 0.087
Statistical index
−1.000 −1.000 −0.618 −0.436 0.176 0.064 0.306 0.143 −0.073 −0.221 −0.154
0.076 −0.148 0.413 0.351 0.015 −0.300 −0.823
0.285 −0.213 −0.161 −0.040 0.550 −1.000 −1.000 −0.192 0.188 −0.179 −0.044 0.057 −0.200 0.167 0.140 −0.260 0.083
Certainty factor
Arab J Geosci
186,473 121,071 96,300 87,009 79,243 673,092 18,176 102,870 220,174 257,075 216,540 241,880 379,408 573,543 103,764
>1500 <500 500–1000 1000–1500 1500–2000 >2000 <60 60–70 70–80 80–90 90–100 >100 0.05 g 0.10 g 0.15 g
PGA
Precipitation (mm/day)
Distance to roads (m)
369,360 305,345 195,535
<500 500–1000 1000–1500
Distance to rivers (m)
No. of pixels in domain
Class
Factors
Table 1 (continued)
0.176 0.115 0.091 0.082 0.075 0.637 0.017 0.097 0.208 0.243 0.205 0.229 0.359 0.543 0.098
0.350 0.289 0.185
Percentage of domain
9 16 15 6 10 54 0 13 15 18 34 21 42 54 5
54 25 13
No. of landslide
0.089 0.158 0.149 0.059 0.099 0.535 0.000 0.129 0.149 0.178 0.337 0.208 0.416 0.535 0.050
0.535 0.248 0.129
Percentage of landslide 0.346 −0.143 −0.304 −0.495 0.277 0.386 −0.279 0.243 −0.161 −1.000 0.244 −0.287 −0.267 0.391 −0.092 0.137 −0.015 −0.496
−0.683 0.324 0.488 −0.326 0.278 −0.175 0.000 0.279 −0.339 −0.311 0.496 −0.096 0.147 −0.015 −0.685
Certainty factor
0.425 −0.155 −0.363
Statistical index
Arab J Geosci
Arab J Geosci
of the landslide conditioning factor were pairwise combined using the CF combination rule. A combination of two CF values, X and Y from two different layers of information is a CF value Z obtained as follows (Lan et al. 2004; Pourghasemi et al. 2012, 2013c; Devkota et al. 2013): 8 X þ Y −X Y X ; Y ≥ 0 > < X þY X Y < 0 Z¼ ð3Þ > 1−min ðjX j; jY jÞ : X þ Y þ XY X;Y < 0 By using the integration rule of Eq. 3, the pairwise combination is repeatedly performed until all the CF layers are combined to obtain the landslide susceptibility.
Landslide susceptibility mapping Application of statistical index model To perform the statistical index modeling, every parameter map is crossed with the landslide inventory map using the ArcGIS 10.0 software, and the density of the landslide in each class is calculated. The weight is positive when the landslide density is higher than the normal and is negative when it is less than normal (Regmi et al. 2014). The resultant weights for each thematic map for the SI model are given in Table 1. These weights were then analyzed by using the weighted sum option in the spatial analyst tools of ArcGIS 10.0 to get the final Landslide Susceptibility Index (LSI) map (Fig. 3). The final calculated LSI values of the study area for SI model range from about −5.044 to 3.603. The index values were classified into four classes (low, moderate, high, and very high) using the natural break method. The areas in the low, moderate, high, and very high landslide susceptibility classes are 12.97, 29.24, 35.88, and 21.91 %, respectively. Fig. 3 Landslide susceptibility map derived from the SI model
Application of certainty factor model The correlation between the location of landslides and the landslide conditioning factors was performed. The CF values were reckoned for all conditioning factors by overlaying and calculating the landslide frequency (Table 1). Then, the CF values of ten landslide conditioning factors were determined using Eq. 2. The results of spatial relationship between landslide and conditioning factors using CF model are given in Table 1. The calculated LSI values for certainty factor model of the study area range from about −3.793 to 2.881. Obviously, larger LSI values indicate a higher susceptibility for landsliding. The pixel values obtained are then classified into four classes (low, moderate, high, and very high) based on natural break method to determine the class intervals in the landslide susceptibility map (Fig. 4). The areas in the low, moderate, high, and very high landslide susceptibility classes are 16.04, 31.48, 33.32, and 19.16 %, respectively.
Validation of the landslide susceptibility maps Landslide susceptibility maps without validation are of little meaning (Chung and Fabbri 1998). In the present study, the total landslides observed in the study area were split into two parts, 101 (70 %) landslides were randomly selected from the total 145 landslides as the training data and the remaining 44 (30 %) landslides were kept for validation propose. The spatial effectiveness of these susceptibility maps was checked by areas under the curve (AUC). Landslide susceptibility maps were verified by comparing the susceptibility map with both the training data that were used for building the models and with the landslide locations that were not used during the model building process. The cumulative area percentages of ordered index values in descending order were categorized
Arab J Geosci Fig. 4 Landslide susceptibility map derived from the CF model
into 100 classes with 1 % cumulative intervals as the horizontal axis. Then, the cumulative percentage of landslide numbers corresponding to landslide susceptibility index values range as the longitudinal axis. The AUC values were obtained for both the training and the validation data (Fig. 5a, b). The results showed that both of the two models used in this study exhibited similar performance. The training accuracy were 0.7019 (70.19 %) and 0.7048 (70.48 %) for SI and CF models, with prediction accuracy 0.6867 (68.67 %) and 0.6886 (68.86 %), respectively. According to the results of the AUC evaluation, the map produced by CF model exhibited the better one.
Conclusions In this study, two statistical models such as statistical index (SI) and certainty factor (CF) models were used for landslide susceptibility mapping and their performances were compared. The results of each model were given. In order to prepare the susceptibility maps, ten landslide conditioning factors
a
b
100
100 90
80 70 60 50 40 30 20
SI model AUC=0.7019
10
CF model AUC=0.7048
0
Cumulative percentage of landslide occurrence (%)
90
Cumulative percentage of landslide occurrence (%)
Fig. 5 AUC representing quality of the model
such as slope angle, slope aspect, curvature, elevation, lithology, distance to faults, distance to rivers, distance to roads, precipitation, and peak ground acceleration were considered as the input data, for which maps were derived using various GIS tools. The landslide inventory map of the study area was prepared by evaluating aerial photographs and field surveys. In this process, a total of 145 landslides were identified and mapped. Out of which, 101 (70 %) were randomly selected as training data and the remaining 44 (30 %) were used for validation purposes. The validation results showed that the area under the curve for statistical index and certainty factor models were 0.7019 (70.19 %) and 0.7048 (70.48 %) with prediction accuracy 0.6867 (68.67 %) and 0.6886 (68.86 %), respectively. Thus, the performance of the produced susceptibility map by certainty factor model was slightly higher than that of the map produced by statistical index model. Landslide prediction map could be the basis. The information provided by these landslide susceptibility maps could be of help for decision makers to reduce losses caused by existing and future landslides during project implementation.
80 70 60 50 40 30 20
SI model AUC=0.6867
10
CF model AUC=0.6886
0 0
10
20 30 40 50 60 70 80 90 100
0
10
20 30 40 50 60 70 80 90 100
Arab J Geosci Acknowledgments The authors want to express their gratitude to everyone who provided assistance for the present study.
References Akgun A (2012) A comparison of landslide susceptibility maps produced by logistic regression, multi-criteria decision, and likelihood ratio methods: a case study at İzmir, Turkey. Landslides 9:93–106 Ayalew L, Yamagishi H (2005) The application of GIS-based logistic regression for landslide susceptibility mapping in the Kakuda Yahiko Mountains, Central Japan. Geomorphology 65:15–31 Bui DT, Lofman O, Revhaug I, Dick O (2011) Landslide susceptibility analysis in the Hoa Binh province of Vietnam using statistical index and logistic regression. Nat Hazards 59:1413–1444 Cevik E, Topal T (2003) GIS-based landslide susceptibility mapping for a problematic segment of the natural gas pipeline, Hendek (Turkey). Environ Geol 44:949–962 Chen W, Li WP, Hou EK, Zhao Z, Deng ND, Bai HY, Wang DZ (2014) Landslide susceptibility mapping based on GIS and information value model for the Chencang District of Baoji, China. Arab J Geosci 7:4499–4511 Chung CF, Fabbri AG (1998) Three Bayesian prediction models for landslide hazard. In: Buccianti A, Nardi G, Potenza R (eds) Proceedings of International Association for Mathematical Geology 1998 Annual Meeting (IAMG’98), Ischia, Italy, October 1998. pp. 204– 211 Devkota KC, Regmi AD, Pourghasemi HR, Yoshida K, Pradhan B, Ryu IC, Dhital MR, Althuwaynee OF (2013) Landslide susceptibility mapping using certainty factor, index of entropy and logistic regression models in GIS and their comparison at Mugling-Narayanghat road section in Nepal Himalaya. Nat Hazards 65:135–165 Einstein HH (1988) Special lecture: landslides risk assessment procedure. In: Proceedings of 5th symposium on landslides, Lausanne, vol 2, pp 1075–1090 Ercanoglu M, Gokceoglu C (2004) Use of fuzzy relation to produce landslide susceptibility map of a landslide prone area (West Black Sea Region, Turkey). Eng Geol 75:229–250 Fell R, Corominas J, Bonnard C, Cascini L, Leroi E, Savage WZ (2008) Guidelines for landslide susceptibility, hazard and risk zoning for land use planning. Eng Geol 102:85–98 Grozavu A, Plescan S, Patriche CV, Margarint MC, Rosca B (2013) Landslide susceptibility assessment: GIS application to a complex mountainous environment. The Carpathians: integrating nature and society towards sustainability, environmental science and engineering, pp 31–44 Guzzetti F (2005) Landslide hazard and risk assessment. PhD Dissertation, Rheinischen Friedrich-Wilhelms-University Bonn, p 389 Kundu S, Saha AK, Sharma DC, Pant CC (2013) Remote sensing and GIS based landslide susceptibility assessment using binary logistic regression model: a case study in the Ganeshganga Watershed, Himalayas. J Indian Soc Remote Sens 41(3):697–709 Lan HX, Zhou CH, Wang LJ, Zhang HY, Li RH (2004) Landslide hazard spatial analysis and prediction using GIS in the Xiaojiang watershed, Yunnan, China. Eng Geol 76:109–128 Nefeslioglu HA, Gokceoglu C, Sonmez H (2008) An assessment on the use of logistic regression and artificial neural networks with different sampling strategies for the preparation of landslide susceptibility maps. Eng Geol 97:171–191
Ozdemir A, Altural T (2013) A comparative study of frequency ratio, weights of evidence and logistic regression methods for landslide susceptibility mapping: Sultan Mountains, SW Turkey. J Asian Earth Sci 64:180–197 Park S, Choi C, Kim B, Kim J (2013) Landslide susceptibility mapping using frequency ratio, analytic hierarchy process, logistic regression, and artificial neural network methods at the Inje area, Korea. Environ Earth Sci 68:1443–1464 Pourghasemi HR, Pradhan B, Gokceoglu C, Mohammadi M, Moradi HR (2012) Application of weights-of-evidence and certainty factor models and their comparison in landslide susceptibility mapping at Haraz watershed, Iran. Arab J Geosci 6:2351–2365 Pourghasemi HR, Moradi HR, Fatemi Aghda SM (2013a) Landslide susceptibility mapping by binary logistic regression, analytical hierarchy process, and statistical index models and assessment of their performances. Nat Hazards 69:749–779 Pourghasemi HR, Pradhan B, Gokceoglu C, Deylami Moezzi K (2013b) A comparative assessment of prediction capabilities of DempsterShafer and weights-of-evidence models in landslide susceptibility mapping using GIS. Geomatics Nat Hazards Risk 4(2):93–118 Pourghasemi HR, Pradhan B, Gokceoglu C, Mohammadi M, Moradi HR (2013c) Application of weights-of-evidence and certainty factor models and their comparison in landslide susceptibility mapping at Haraz watershed, Iran. Arab J Geosci 6:2351–2365 Pradhan B (2010) Landslide susceptibility mapping of a catchment area using frequency ratio, fuzzy logic and multivariate logistic regression approaches. J Indian Soc Remote Sens 38(2):301–320 Pradhan B, Oh HJ, Buchroithner M (2010) Weights-of-evidence model applied to landslide susceptibility mapping in a tropical hilly area. Geomatics Nat Hazards Risk 1(3):199–223 Rautela P, Lakhera RC (2000) Landslide risk analysis between Giri and Tons Rivers in Himachal Himalaya (India). Int J Appl Earth Obs Geoinf 2:153–160 Regmi NR, Giardino JR, Vitek JD (2010) Modeling susceptibility to landslides using the weight of evidence approach: Western Colorado, USA. Geomorphology 115:172–187 Regmi AD, Devkota KC, Yoshida K, Pradhan B, Pourghasemi HR, Kumamoto T, Akgun A (2014) Application of frequency ratio, statistical index, and weights-of-evidence models and their comparison in landslide susceptibility mapping in Central Nepal Himalaya. Arab J Geosci 7(2):725–742 Soeters R, van Westen CJ (1996) Slope stability recognition analysis and zonation. In: Turner AK, Schuster RL (eds) Landslides: investigation and mitigation, transportation research board special report 247. National Academy Press, Washington, pp 129–177 Solaimani K, Mousavi SZ, Kavian A (2013) Landslide susceptibility mapping based on frequency ratio and logistic regression models. Arab J Geosci 6:2557–2569 van Westen CJ, Rengers N, Terlien MTJ, Soeters R (1997) Prediction of the occurrence of slope instability phenomena through GIS based hazard zonation. Geol Rundsch 86:404–414 Wang HB, Sassa K (2005) Comparative evaluation of landslide susceptibility in Minamata area, Japan. Environ Geol 47:956–966 Wieczorek GF (1984) Preparing a detailed landslide-inventory map for hazard evaluation and reduction. Assoc Eng Geol Bull 21(3):337– 342 Xu C, Xu XW, Dai FC, Saraf AK (2012) Comparison of different models for susceptibility mapping of earthquake triggered landslides related with the 2008 Wenchuan earthquake in China. Comput Geosci 46: 317–329