Landslides and slope instabilities are major risks for human activities which often lead to economic losses and human fatali- ties all over the world. The main purpose of this study is to evaluate and compare the results of Landslide Nominal Risk Factor (LNRF), Frequency Ratio (FR), and Analytical Hierarchy Process (AHP) models in mapping Landslide Susceptibil- ity Index (LSI). The study case, Nojian watershed with an area of 344.91 km , is located in Lorestan province of Iran. The procedure was as follows: first, the effective factors of the landslide basin were prepared for each layer in the GIS software. Then, the layers and the landslides of the basin were also prepared using aerial photographs, satellite images, and fieldwork. Next, the effective factors of the layers were overlapped with the map of landslide distribution to specify the role of units in such distribution. Finally, nine factors including lithology, slope, aspect, altitude, distance from the fault, distance from river, fault land use, rainfall, and altitude were found to be effective elements in landslide occurrence of the basin. The final maps of LSI were prepared based on seven factors using LNRF, FR, and AHP models in GIS. The index of the quality sum (Qs) was also used to assess the accuracy of the LSI maps. The results of the three models with LNRF (40%), FR (39%), and AHP (44%) indicated that the whole study area was located in the classes of high to very high hazard. The Qs values for the three models above were also found to be 0.51, 0.70 and 0.70, respectively. In comparison, according to the amount of Qs, the results of AHP and FR models have slightly better performed than the LNRF model in determining the LSI maps in the study area. Finally, the study watershed was classified into five classes based on LSI as very low, low, moderate, high, and very high. The landslide susceptibility maps can be helpful to select sites and mitigate landslide hazards in the study area and the regions with similar conditions. Keywords Landslide susceptibility index · Nojian watershed · AHP model · FR model · LNRF model · GIS Introduction Nojian watershed in Lorestan province of Iran is located in Zagros tectonically active folded mountains across Landslides cause great damages in residential areas, roads, the Zagros Big Fault and the earthquake-prone belt of facilities, agricultural fields, gardens, grasslands, etc. Recog- Alpine–Himalayan zone. Therefore, potentially there are nizing the mechanism and zoning of prone areas to landslide suitable conditions for slope instability related to the alter- occurrence play a crucial role in disaster management. It can nating layers of limestone and thin layers of the marl and be employed as a standard tool to support decision-making deeply weathered formation on the slopes. Due to geomor- in different areas (Bui et al. 2016 ). According to its defi - phological properties such as lithology, tectonics activity, nition by engineering geology, landslide is the downward altitude, seismicity, slope, climatic conditions, deeply weath- movement of a mass of material on a slope. ered formation and human impact, the study watershed was potentially highly prone to landslide occurrence. Further- more, conditions of topography and geomorphology, clima- tology and human impact of the study area have provided the * M. Abedini best position for sliding small or large masses of materials abedini@uma.ac.ir of the slopes. University of Mohaghegh Ardabili, Ardabili, Iran Vol.:(0123456789) 1 3 405 Page 2 of 13 Environmental Earth Sciences (2018) 77:405 According to the reports of the global organization of 2016), comparison between bivariate statistical and multi- natural dangers in 2012, landslide was among the seven most variate adaptive regression (Wang et al. 2015), two-class dangerous natural disasters. Landslides, due to their nature kernel logistic regression (Hong et al. 2015), support vector and diversity and because of their potential hazards and machine, artificial neural network kernel, logistic regression the catastrophic effects for human’s life and property, have and logistic tree, Bui et al. (2016), frequency ratio, statisti- always been examined by scholars of various disciplines of cal index and certainty factor (CF):Wu et al. (2016a, b), earth sciences including engineering geology, geomorphol- landslide susceptibility mapping by support vector machine ogy, and watershed management. Therefore, planning to methods (Ballabio and Sterlacchini 2012; Gravina et al. control and mitigate its damages seems important and nec- 2016; Shirzadi et al. 2017a, b), random forest decision tree essary. Using different models of LSI mapping is one of the methods (Zhang et al. 2017), shallow landslide susceptibil- methods to identify the prone areas for future landslides and ity assessment using a novel hybrid model (Shirzadi et al. preventive activities should be taken to prevent or decrease 2017a, b), frequency ratio, statistical index, and weights their damages in those areas. (Razavizadeh et al. 2017), etc., has been done. The United Nations proclaimed the 1990s as the Inter- In addition, review of the literature indicated some national Decade for Natural Disaster Reduction (IDNDR); researches results such as Gupta and Joshi (1990); Shad- UNESCO named 1990s as the decade of dealing with far et al. (2011); Mohammadi et al. (2014) and Malik et al. natural disasters (Erich and Wolfgang 1994; Geeraay and (2016) indicated LNRF method has suitable method in land- Kariemi 2010). Therefore, these days in different centers of slide susceptibility mapping. academia, extensive research is being carried out on land- Our review of the literature of landslide-related research slides as an instance of natural disaster by many scholars revealed that no research has ever been done on the study worldwide. As the aim of this research is an evaluation of watershed with comparative models. On the other hand, LNRF, FR and AHP models in mapping (LSI) in Nojian this area is highly prone to slope instability, therefore, this watershed; therefore, we will focus on the related literature research the main purpose of comparison and performance having the purpose of the current study in mind. Although of the three models based on GIS has been done. In addition, the mechanism of landslide occurrence has been studied based on the results of some studies in Iran, these methods by many geologists and geomorphologists at a large scale are almost slightly better in providing and forecasting the through quantitative methods (statistical), multi-criteria landslide susceptibility maps. Therefore, to comparatively decision-making (MCDM), support vector machine, logistic assess landslide susceptibility, three models of LNRF, FR, regression, artificial neural network (ANN), frequency ratio and AHP were selected for study area. (FR), including: It is noted that data preparation and processing were GIS-based landslide susceptibility mapping using analyti- carried out using ArcGIS Ver. 10.3. Although nowadays cal hierarchy process (AHP) and bivariate statistics (Zhu and many methods have been employed for landslide suscepti- Huang 2006; Fanyu Liu 2007; Yalcin 2008; Gupta and Joshi bility mapping around the world, the results of this study 1990; Akgun and Turk 2010; Bhatt et al. 2013; Kayastha can practically be used in land use planning and identifying et al. 2013; Dai and Lee 2002; Abedini et al. 2017; Gravina areas prone to landslide. However, there is an essential need et al. 2017), Landslide Susceptibility Index (LSI), Logistic to using methods and techniques to improve the accuracy Regression (LR), and Artificial Neural Network (ANN), of landslide prediction on a regional scale. Therefore, the Frequency Ratio (FR) (Romer and Ferentinou 2016), land- major aims of this study are to select the most important and slide susceptibility mapping based on Frequency Ratio and effective factors in landslide susceptibility and assess them Logistic Regression models (Ayalew and Yamagishi 2005; with LNRF, FR, and AHP models by Mapping Landslide Lee and Sambath 2006; Lee and Pradhan 2007; Solaimani Susceptibility Index in the Nojian watershed in the Zagros et al. 2013; Umar et al. 2014; Shahabi et al. 2014; Pham Mountains of Iran. In addition, comparison of the results of et al. 2015; Shirzadi et al. 2017b; Costanzo et al. 2014), LNRF, FR, and AHP models is made for study area. It should comparison of logistic regression and Naive Bayes classi- be mentioned that data preparation and processing wee done fier (Tsangaratos and Ilia 2016), landslides susceptibility using Arc GIS 10.3, SPSS and Expert Choice (EC) software. by dynamic landslide run-out model (Byron et al. 2016), multivariate and bivariate (Kavzoglu et al. 2015), auto logistic molding (Atkinson and Massari 2011; Costanzo The study area et al. 2014), artificial neural network and comparison with frequency ratio and bivariate logistic regression (Pradhan Nojian watershed, with an area of about 344.91 km , is and Lee 2010), static methods (Shirani and Seif 2012), ana- located in the 30 km Southeast of Khorramabad city, lytical hierarchy process and multivariate statistics (Komac Lorestan province, Iran. This basin is located between 2006; Pourghasemi et al. 2013; Amir Yazdadi and Ghanavati 48°23′E and 48°40′E longitude and 33°06′N and 33°17′N 1 3 Environmental Earth Sciences (2018) 77:405 Page 3 of 13 405 Fig. 1 The map of the geo- graphic location of Nojian watershed latitude (Fig. 1). Its average annual rainfall is 686 mm, According to the basic assumption, future landslides will and its climate is semi-moist with very cold winters. The occur under the same conditions (Lee and Talib 2005). average altitude is 1635 meters above the sea level. Due In this research, to identify the relationship between the to geological properties such as lithology, tectonics, seis- landslide distribution and the relevant conditioning factors, micity, and specific climatic conditions, Iran has areas of the existing landslide inventory map is necessary. Therefore, potential landslide. Located on the earthquake-prone belt to come up with a detailed and reliable inventory map for of Alpine–Himalayan passing Zagros Big Fault, Lorestan the study area, two processes were utilized including field province has plausible conditions for instability in large sec- surveys and laboratory interpretations. Therefore, the dis- tions of the mountains slopes due to the alternating layers tribution location of landslides was collected from the For- of massive limestone and thin embedded layers of the marl, ests, Rangelands and Watershed Management Organization shale, gypsum and conglomerate formation. of Iran. Then, using field surveys, this location was checked According to field and geology maps, the lithology of this and some characteristics of each landslide were recorded basin is more diverse: massive limestone, shale and sand- including length, width, and surface area of landslides using stone, dolomite, marly limestone, gypsum and sandstone, aerial photographs (1:40,000 scale), GPS and TM satel- conglomerate with sand stone and red marl, quaternary allu- lite images interpretation. In addition, we utilized SPSS to vial and sediment. perform the statistical calculations in the models, Expert Choice (EC) software to determine the relative weighting of the effective factors, aerial photographs with the scales of 1:50,000, also topographic map with the scale of 1:50,000 Methodology and geological map with the scale of 1:100,000 and Isohyet is a line on map connecting which are the regions having Landslide inventory map same amount of rainfall. Landslide zoning includes dividing the land into separate regions and ranking them according In all versions of susceptibility mapping models, preparing to real degree or sensibility potential caused by landslide on landslide inventory map is the first step of data production. hillside slopes. Due to the fact that the main purpose of this 1 3 405 Page 4 of 13 Environmental Earth Sciences (2018) 77:405 study is the evaluation of LNRF, FR and AHP models in erate are dominant. Therefore, to clarify landslide ten- mapping LSI, the research method involved attritional form, dency of the structures, the density of existing landslides morphometry, field work and experimental manipulation. In in different structures must be assessed (Fig. 2d). doing so, we pursued the following four steps: 5. Distance from fault Especially active faults play an important role in landslides from two points of view. 1. Collecting and reviewing the relevant books, theses, arti- First, they are the origin of earthquake occurrence. Sec- cles, and any source related to the topic of the study, ond, active faults have important roles in cracking stones 2. observing the field, and creating instability. This discontinuity of geology 3. doing the calculations and processing the data, formation leads to decreases in shear resistance of the 4. presenting the final report of the research. slide and preparation of landslide occurrence (Fig. 2e). 6. Distance from road Roads have the most effective role Using aerial photographs, satellite imagery, and exten- in the concentration of runoff, therefore experience and sive fieldwork, landslide areas were identified. Then, the col- existing statistics about landslide during reconstruction lected data were digitized using GIS to produce the landslide and road widening shows the necessity of using this fac- distribution map of the basin. Figure 2j shows the map of tor in landslide sensitivity zoning. So, this layer consists landslide distribution of the area. of five classes (Fig. 2f). 7. Distance from drainage network The drainage density is 1. Elevation Altitude change of each region is one of the one of the factors that have a determining role in mass very effective factors in the creation of soil erosion and movements as in some hillsides with low slope, water- slope mass movement. This factor intensively controls way, and landslide aggression are high. This shows the runoff direction and rate of drainage density (Hossein - importance of waterway aggression in landslide occur- zadeh et al. 2009: 29). The maximum altitude of the rence. In this study, the map of distance from drainage region is 2848 m and the minimum altitude is 762 m. In system was drawn in five classes (Hosseinzadeh et al. addition, the general altitude variance will be 2086 m. 2009: 32) (Fig. 2j). For this reason, the map of height was assorted in seven 8. Land use The form of land use is an important index in classes (Fig. 2a). inconstancy of the slopes that affects the earth’s traits 2. Slope In most studies related to landslide sensibility zon- and results in changes in its action. This map was drawn ing, slope percentage is one of the most important fac- in five classes (Fig. 2h). tors (Abedini et al. 2017). Theoretically, with increasing 9. Rainfall Climate of the basin and rainfall intensity and the slope, shear stress is augmented and consequently its durability play an important role in landslide creation increasing slope instability potential is expected. Slope which depends on climate factors, topography, geology, map of the study region is prepared in six classes and slope (Lydia and Daniel 2002: 183). Campbell also (Fig. 2b). believes that precipitation over 254 mm and rainfall 3. Aspect This factor has an important role in utilizing intensity about 6.35 mm causes landslide. In this study, the amount of rainfall, sun energy and kind wind blow- precipitation map of the study area was drawn in five ing in the any area and reflecting the influence of soil classes (Fig. 2i). thickness, vegetation, moisture, etc. The study area is located in the middle attitude, therefore, in whole direc- tion of slopes of landslide susceptibility is different. The Description of methods aspect map of the study area is prepared in eight classes (Fig. 2c). Analytical hierarchy process (AHP) method 4. Lithology Lithological conditions of the basin are very effective in slope instability and landslide occurrence. The model of analytical hierarchy process (AHP) was On other hand, the stronger rocks give more resistance to proposed by Saaty (1980) for the first time. This model the driving forces in compared to the weaker rocks, and measurement through pairwise comparisons of effective hence are less prone to landslides and vice versa (Yalcin factors in study and relies on the judgments of experts to 2008). In determining the lithological resistance of the derive priority scales. In addition, pairwise comparisons study area, the type of rocks, the tectonic condition, and can allow decision makers to weigh coefficients and com- their resistance to weathering factors were considered as pare alternatives with relative case or factors in obtaining the most important factors. From lithological perspec- the suitable results (Saaty 2008). It works first with being tive, the most susceptible area to landslide occurrence weighted to every effective factor considered for zoning, in the study watershed is where the formations contain and alternatives to be selected (Abedini et al. 2017). In marl and limestone, shale, gypsum and marly conglom- AHP model, GIS is used for zoning. This section includes 1 3 Environmental Earth Sciences (2018) 77:405 Page 5 of 13 405 Fig. 2 The contributing factor maps in landslides of the basin (a–i) and the landslide distribution map of the area (j) entering the data into a GIS, then analyzing the production 2. production of the pairwise comparison matrix based on of the layers of information. The mapping of LSI using expert opinion, AHP model and EC and GIS was done as follows: 3. calculation of relative weight and the inconsistency rate determination using the EC software, 1. prioritization of the effective factors, 1 3 405 Page 6 of 13 Environmental Earth Sciences (2018) 77:405 Fig. 2 (continued) and then points to any of classes corresponding to each Numerical risk factor (LNRF) model of the factors, acquired coefficients based on them provide the final model. The main advantage of this method is that Landslide susceptibility map was integrated to compute the it helps the decisions to a complex problem be open to numeric value of each factor with the help of a Landslide the hierarchical structure, and then solves it. Furthermore, Numerical Risk Factor (LNRF) model. AHP is a decision support system to seek optimum con- Landslide Numerical Risk Factor (LNRF) model that is ditions for a complex circumstance through hierarchical a suitable model especially in the mountainous regions to structures, which comprised targets to be attained, various landslide susceptibility mapping (Gupta and Joshi 1990). criteria for decision In this model, (LNRF > 1) indicates that the special cate- gory of that geo-environmental factor is more susceptible to 4. calculation of the final weight of effective factors and landslides than the average. The lower value of (LNRF < 1) prioritizing them, means that the particular geo-environmental category is 5. preparation of the weighted maps based on the values of associated with more stable slopes and less susceptible to the final weight of units, landside occurrence (Gupta and Joshi 1990), (Mohammadi 6. preparation of the zonation map through the sum of the et al. 2014). This model is calculated from Eq. (1): weighted maps, Npix (Si) 7. calculation of the area and the percentage of the area at LNRF = A B or LNRF = � �� , (1) risk zones. Npix(Si) n i=1 where A is the landslide area in every unit, and E is the mean area of landslide in the whole unit. 1 3 Environmental Earth Sciences (2018) 77:405 Page 7 of 13 405 Table 1 The weight related to 3. overlapping the raster layers of the factors with the raster Number Amounts Weight the amount of LNRF (Gupta layers of the basin landslides, LNRF and Joshi 1990) 4. calculation of the landslide cells and the no-slip cells, 1 0 > 1 5. calculation of the FR values based on Eqs. (2–4), 2 1 1–2 6. preparation of the weighted maps based on the FR values 3 2 < 2 obtained for the units, 7. preparation of the zonation map through the sum of the weighted maps, In this method, which is called the credibility factor of 8. calculation of the area and percentage of the area at risk landslide risk, using the occurred slip surface in a unit than zones. mean occurred slip in the whole unit, the index is prepared (Mohammadi et al. 2014) (Table 1). Landslide conditioning factors Frequency ratio (FR) method The most common method of identifying effective factors is the use of questionnaire and morphometry of the landslides The Frequency Ratio (FR) model is one of the important available inside the basin and using fieldwork which we did probability methods in landslide susceptibility mapping for Nojian watershed. The first important step in the zoning based on the observed relationship between distribution of of landslide susceptibility is to identify locations of landslide landslides and each landslide-related factor. Put simply, the occurrence in the past and present (Jiménez-Perálvarez et al. frequency ratio (FR) is the ratio of the study area where 2011). One of the basic steps in zoning landslide suscepti- landslides occurred in the whole study area. It is also the bility is creating a dataset and collecting the required data ratio of the probabilities of a landslide occurrence to a non- (Kavzoglu et al. 2015). In fieldwork, cases such as the land- occurrence for a given attribute (Lee and Talib 2005). This slides location, structure lithology, soil type, land use in the model is a complete model including more independent vari- area of landslides, vegetation, slope, aspect, linear elements ables that play an important role to determine the dependent (roads, drainages, and faults) are considered. The height was variables (Abedini et al. 2017). As one of the serial gen- necessary and the relevant expert could study these to some eralized linear models, it is very useful for analyzing the extent to specify the effective factors in landslides (Abedini existence of dependent variables and anticipating hillside et al. 2017). Based on a review done on the basin, parameters instability (Dai and Lee 2002). The frequency ratio (FR) such as land use, distance from faults, rainfall, lithology, model is calculated through Eqs. (2–4). Every unit of each slope, elevation, distance from the main drainage and the factor is the percentage of the landslide cells (A) divided distance from the road, landslide zone were identified as by the percentage of no slip cells (B). The percentage of effective factors in the occurrence of the landslide. The maps the landslide cells is the number of slide cells of each unit related to each factor were prepared by GIS. Using modula- (A) divided by the total landslide cells of the basin (B) mul- tion of nine effective layers in landslide, drawing the final tiplied by 100. The percentage of the no-slide cells is the map of landslide sensibility zoning in the study area is the number of cells of each unit (C) divided by the total cells main objective of this study. Figure 2a–i shows the map of of the basin (D) multiplied by 100 (Solaimani et al. 2013; the factors influencing landslides in the study area. Razavizadeh et al. 2017). A(%) FR = Discussion and results (2) B(%) Weighting the effective factors A(%) = × 100 (3) In AHP model, according to the different degrees of the effective factors in creating landslides, it is necessary to B(%) = × 100 identify and prioritize the correct factors. Part of this is done (4) by questionnaire and the other part by comparing each fac- Mapping of LSI using FR model and GIS software was tor. To compare the paired items and determine the amount as follows: of the priority of the different factors relative to each other, a major scale was used with values from 1 to 9. In this matrix, 1. preparation of the factor maps and their classification, two by two effective factors in landslide occurrence of the 2. preparation of the landslides maps and the factor maps basin were compared and scored. In the combined weighting of the basin (cell size: 5 m × 5 m), method, the effective factors were prioritized according to 1 3 405 Page 8 of 13 Environmental Earth Sciences (2018) 77:405 Table 2 The final weighting of the effective factors and their priority Priorities 1 2 3 4 5 6 7 8 9 Effective factors Lithology Slope Distance from Land use Rainfall Aspect Altitude Distance from Distance from the fault the road the main drainage The final weight 82.76 45.90 44.54 35.86 22.45 14.17 8.72 5.79 3.12 the judgment of experts and then each effective factor was calculated. Finally, using the equation corresponding to the classified in criteria’s and weighted. The weights of crite- model, the amount of LNRF was obtained for every unit of ria are calculated from the pairwise comparison matrix. To each factor, and the values were standardized. calculate the relative weight for effective factors, the values The values obtained for every unit of each factor showed were entered into Expert Choice (EC) Ver. 10.3, and finally that in all three models, two factors of the distance from the relative weight of each factor was obtained. Therefore, the road and distance from the main drainage due to lack of based on the relative weight, the factor of lithology (0.242), logical relation with landslide distribution did not influence slope (0.197), land use (0.183), distance from the fault the landslide of the basin in the sense that with increasing (0.174), rainfall (0.103), altitude (0.043), aspect (0.026), distance from the road and distance from drainage, the per- distance from the road (0.018), and distance from the main centage of landslides increased which was not logical. So drainage (0.014) in the order of priority were identified as these two factors were removed and were not considered in the effective factors contributing to landslide occurrence in the preparation of the final maps of the zonation. the Nojian watershed. The amount of the incompatibility rate also was obtained to be 0.08, indicating the correctly Landslide susceptibility mapping paired comparisons of the effective factors. To calculate the final weight, effective factors in landslide After calculating the final weight by the models, raster occurrence of the basin were divided by the units. We over- maps were obtained based on these weights. The maps of lapped the landslide distribution map on each of the layers, the weighted raster and the map of LSI were obtained for and the percentage of the sliding surface was calculated for each model. Figure 3 shows these maps that have been clas- each unit. Finally, the percentage of the sliding surface of sie fi d into v fi e susceptibility classes. Area and percentage of each unit was allocated a rate from zero to 100. This way, area for each of the classes of risk in the zonation maps are different values were given to each unit that had the maxi- provided in Table 3. mum percentage of the sliding surface of the 100 value, and As shown in Table 3, the results of assessment and clas- for other units also proportional to the amount of the per- sification of landslide susceptibly in the study area have been centage of the sliding surface. The scores obtained for every determined using the three models. The whole watershed unit of each factor were multiplied by the relative weight of area according to the landslide susceptibly hazard is classi- its factor to calculate the ultimate weight for each unit. By fied into very low, low, high, medium, high and very high summing the final weights of the units together and the final zones. Based on the landslide susceptibly hazard classifi- weight for each factor was obtained (Table 2). cation in the study watershed, AHP (44%), LNRF (40%), In FR model, first, the map of the effective factors was and FR (39%) form the total area located in the high and prepared and classified (cell size 5 m × 5 m). Then, the very high hazard zones. Therefore, the comparative results overlapped raster maps of factors with the raster map of of the three models are very close together showing that the landslides of the basin and the number of slide cells and the watershed is potentially very prone to landslide occurrence. number of no-slide cells for every unit of each factor were obtained using GIS. Finally, using equations related to the Model validation and comparison model, the amount of the frequency ratio for every unit of each factor was calculated. For the assessment accuracy and comparison of the zon- In LNRF model, first, the landslide area for every unit ing maps, quality sum (Qs) index was used. The amount of of each factor was calculated. Then, their average was Qs index represents the desirability function of methods in predicting the landslide hazard in the study area. Therefore, in assessing the method, the greater the value of this index, the more desired utility and this method indicated accurate If the obtained inconsistency rate is more than 0.1, this represents prediction. To determine this index, first the density ratio that the paired comparison and ratings were not done correctly, and (DR) must be calculated. The amount of DR is obtained this operation should be performed again with changes in the points. 1 3 Environmental Earth Sciences (2018) 77:405 Page 9 of 13 405 Fig. 3 The LSI maps of Nojian watershed (A) AHP model (B) FR model (C) LNRF model through Eq. (5). In this equation, Si is the sum of the area of The amount of Qs index also is obtained through Eq. (6) the landslides in each class risk, Ai is the area in the class of in which, Qs is the quality sum index, DR is the density risk and n is the number of risk classes in a zonation map. ratio and S is the area ratio of each risk class to the total 1 3 405 Page 10 of 13 Environmental Earth Sciences (2018) 77:405 Table 3 The area and 2 2 Models Hazard zonesSi (km )Ai (km ) DR S Qs percentage of the area of the hazard zones in the zoning maps AHP model Very high 10.14 72.35 2.36 0.21 0.70 High 6.6 79.47 1.4 0.23 Medium 1.97 63.29 0.52 0.18 Low 1.43 68.82 0.35 0.2 Very low 0.05 57.34 0.01 0.16 LNRF model Very high 6.71 54.23 2.16 0.15 0.51 High 6.42 82.59 1.36 0.24 Medium 4.49 64.82 1.21 0.18 Low 1.7 76.57 0.38 0.22 Very low 0.15 63.05 0.04 0.18 FR model Very high 6.95 50.92 2.46 0.14 0.70 High 7.06 75.54 1.68 0.22 Medium 3.53 80.62 0.78 0.23 Low 1.19 65.42 0.33 0.19 Very low 0.18 68.79 0.04 0.20 Table 4 The assessment results 2 Parameters The area of the hazard zones (km ) Percentage of the area of the of the zoning maps of the basin, hazard zones (%) using the Qs method Models LNRF FR AHP LNRF FR AHP Susceptibility zones Very high 54.32 50.92 72.35 16 16 21 High 82.59 75.54 79.47 24 23 23 Medium 64.82 65.42 63.29 19 20 19 Low 76.57 65.42 68.82 23 20 20 Very low 63.05 68.76 57.34 18 21 17 area (David 1992; Amir Yazdadi and; Ghanavati 2016). Predicted accuracy values for FR (0.70), AHP (0.70) and Table 4 shows the assessment results of the zonation maps LNRF (0.51) models showed that the map obtained from of the basin using the Qs method. frequency ratio model and AHP is more accurate than the LNRF model. The results revealed the better performance Si of these two models (FR and AHP) than the LNRF model in Ai Dr = , n determining the zones susceptible to landslide occurrence in (5) Si the Nojian watershed. Ai Discussion and conclusion Qs = (Dr − 1) × S . (6) i=1 Landslide susceptibility mapping methods play an impor- tant role in providing a suitable approach to decision makers As shown in Table 4, in the hazard zones, the amount of and authorities, particularly in landslide prone areas. Essen- density ratio (DR) has increased commensurate with the tially, Landslide Susceptibility Mapping Index (LSMI) pro- increased risk which represents a high level of the accu- vides very fundamental knowledge of the effective factors racy of the risk zone in the zoning maps. Moreover, the Qs and causes of landslide occurrence. In addition, it can be values obtained for FR, AHP are 0.70, and 0.70 and 0.51 an effective method in hazard management and mitigation for LNRF models, respectively, representing the higher measures. performance of these two models than the LNRF model in In the present study, we attempted to compare the results determining the zones susceptible to landslide occurrence of landslide susceptibility mapping using three different in the study area. models, namely Landslide Nominal Risk Factor (LNRF), 1 3 Environmental Earth Sciences (2018) 77:405 Page 11 of 13 405 Frequency Ratio (FR), and Analytical Hierarchy Process et al. 2014), MCDA, SVM, and LR (Kavzoglu et al. 2015; (AHP). Abedini et al. 2017), heuristic and bivariate statistical The main objective of this study was to investigate poten- models (Bijukchhen et al. 2012), probabilistic, bivari- tial application of LNRF, FR, and AHP models in mapping ate and multivariate models (Pradhan and Youssef 2010; Landslide Susceptibility Index (LSI) especially compare Kevin et al. 2011; Ozdemir and Altural 2013; Shahabi each of them in the Nojian watershed in Lorestan province et al. 2014; Wu et al. (2016a, b). of Iran. On the other hand, a host of studies such as, Ayalew and In the study area, landslides are mainly affected by alti- Yamagishi (2005), Komac (2006), Yalcin (2008), Esmali tude, faulting, slope, rainfall, various geological formations, Ouri and Amirian (2009), Jin et al. (2010), Dehban and road construction, and land use changes. For the preparation Stakhri (2013), Bhatt et al. (2013), Pourghasemi et al. of the LSI map, there are numerous weight combining meth- (2013), Kumar and Annadurai (2015), Amir Yazdadi and ods. In the present comparative study, FR, AHP, and LNRF Ghanavati (2016), Romer and Ferentinou (2016), Wu et al. models were employed for landslide susceptibility mapping (2016a, b), Abedini et al. (2017), Shirzadi et al. (2017a, in Nojian watershed located in the Lorestan province of Iran. b), Razavizadeh et al. (2017) indicated that FR and AHP The results revealed that density ratio (DR) and area ratio models were more often better than the logistic regression S, variant weights, and data value were the most accurate in in Japan, Iran, China, respectively. dividing the classes of landslide. The preparation of the LSI In addition, some research results such as Gupta and map is important through which one can detect susceptible Joshi (1990); Shadfar et al. (2011); Mohammadi et al. areas for the future landslides and use these maps for the (2014) and Malik et al. (2016) indicated LNRF method future planning. The evaluation of the final weight of the has suitable method in landslide susceptibility mapping. effective factors and the resulting overlay of these layers The results based on LNRF, FR and AHP models in with the landslides layer area revealed the highest landslide study watershed, respectively, accounted for 40, 39 and due to the kind of geological formation. 44% of the area of the basin located in the classes of the From lithological perspective, the most susceptible area high to very high hazard zones due to the landslides sus- to landslide occurrence in the study watershed is where the ceptibly. The prediction performance of the susceptibility formations contain marl and limestone, shale, gypsum and map is checked by considering quality sum amount (Q ). marly conglomerate are dominant. The quality sum (Q ) values for the three models above The majority of landslides were in the class of slope more also were obtained to be 0.51, 0.70 and 0.70, respectively. than 30°, the class of altitude more than 2000 m, the dis- Therefore, according to the amount of Qs and comparison tance less than 2 km from the faults, the forest land use (due of the results, we conclude that the two models of AHP to deforestation and conversion to agricultural land), and and FR have better performance than the LNRF model the class of rainfall more than 850 mm. Two factors of the in determining the LSI maps in the Nojian watershed. In distance from the road and distance from the main drain- addition, comparison of the past landslide occurred map age were excluded due to the lack of reasonable relationship (Fig. 2j) with provided landslide susceptibility map by with the percentage of landslides, in the sense that, with an LNRF, FR, AHP models (Fig. 3a–c) was showed high con- increase in distance from the road and distance from the formity. In other words, this demonstrates the high capa- main drainage, the percentage of landslides increased, which bility of these models in the preparation and evaluation of was not reasonable. So, based on the above models, seven landslide susceptibility maps in the studied watersheds. factors of lithology, slope, distance from the fault, land use, Finally, the study watershed was classified into five sen- rainfall, slope and elevation, respectively, were identified as sitivity classes of very high, high, moderate, low and very the effective factors in the landslides of the basin. low. Based on the results, landslide hazard zoning maps Although the results of MCDM such as AHP, fuzzy and information can be useful for general development logic, LNRF, FR, AHP, support vector machine (SVM), planning, landslide risk management, selecting sites and artificial neural network (ANN), and logistic regression mitigating landslide hazards for decision makers in the (LR) models are slightly different from region to region, regions with similar conditions. but the accuracy of the models is very close in providing Acknowledgements The authors gratefully acknowledge of National the landslide susceptibility mapping. Meanwhile, many Resource Organization of Iran (NRO) for data supporting and are studies have found overall accuracy rate relatively similar thankful to the members of geomorphology department of Mohaghegh in some models such as LNRF, FR, AHP, SVM, ANN and Ardabil University and director of environmental management organ- ization of Ardabil. Ultimately, the authors would like to thank two LR (Jin et al. 2010; Park et al. 2012), Conditional proba- anonymous honor reviewers and the editor for their helpful comments bility (CP), LR, ANN, and SVM (Yilmaz 2010; Gupta and on the previous version of this manuscript. Joshi 1990; Akbari and Mashayekhan 2012; Mohammadi 1 3 405 Page 12 of 13 Environmental Earth Sciences (2018) 77:405 Open Access This article is distributed under the terms of the Crea- watershed, Ardabil, Iran. In: International conference on ACRS tive Commons Attribution 4.0 International License (http://creat iveco 2009, Beijing mmons.or g/licenses/b y/4.0/), which permits unrestricted use, distribu- Fanyu Liu Z (2007) Study on landslide susceptibility mapping based tion, and reproduction in any medium, provided you give appropriate GIS and with bivariate statistics a case study in Longnanarea credit to the original author(s) and the source, provide a link to the highway 212. Science (paper online) Creative Commons license, and indicate if changes were made. Geeraay P, Kariemi H (2010) The determination the most appropriate method the landslide hazard zonation in the Ilam dam watershed. J Geogr Res 96(24):101–128 Ghanavati EAYE. (2016) Landslide hazard zonation by using AHP References (analytical hierarchy process) model in GIS(geographic informa- tion system) environment (case study: Kordan watershed). Int J Abedini M, Ghasemyan B, Mogaddam MH (2017) Landslide suscep- Sci High Technol 2:24–39 tibility mapping in Bijar city, Kurdistan province, Iran: a com- Gravina TF, Figliozzi NF, Schinosa DT (2017) Landslide risk parative study by logistic regression and AHP models. Environ perception in Frosinone (Lazio, Central Italy). Landslides Earth Sci 76:308 14(4):1419–1429 Akbari M, Mashayekhan A (2012) The comparison of concep- Gupta RP, Joshi BC (1990) Landslide hazard zoning using the GIS tual model LNRF and WLC in zoning landslide risk (case approach—a case study from the Ramanga catchment, Himalayas. study:Akhlamad basin, Chenaran, Khorasan, Iran). J Iran Water- Eng Geol 28:119–131 shed Manag Sci Eng 6(18):1–10 Hong P, Pradhan B, Xu C, Tien Bui D (2015) Spatial prediction of Akgun A, Turk N (2010) Landslide susceptibility mapping for Ayva- landslide hazard at the Yihuang area (China) using two-class ker- lik (Western Turkey) and its vicinity by multi-criteria decision nel logistic regression, alternating decision tree and support vector analysis. Environ Earth Sci 61:595–611 machines. Geomorphology 133:266–281 Atkinson P, Massari R (2011) Autologistic modelling of susceptibil- Hosseinzadeh M, Servati M, Mansouri A, Mirbagheri B, Khezri S ity to landsliding in the Central Apennines, Italy. Geomorphol- (2009) Landslide hazard zonation using logistic regression, the ogy 130:55–64 way Sanandaj–Dehgolan, Iran. J Geol 11:27–37 Ayalew L, Yamagishi H (2005) The application of GIS-based logistic Jiménez-Perálvarez JD, Irigaray C, El Hamdouni R, Chacón J (2011) regression for landslide susceptibility mapping in the Kakuda– Landslide-susceptibility mapping in a semi-arid mountain envi- Yahiko mountains, central Japan. Geomorphology 65:15–31 ronment: an example from the southern slopes of Sierra Nevada Ballabio C, Sterlacchini S (2012) Support vector machines for land- (Granada, Spain). Bull Eng Geol Environ 70(2):265–277 slide susceptibility mapping: the Staffora river basin case study, Jin GC. Che OY. Choi CU (2010) The comparative research of land- Italy. Math Geosci 44:47–70 slide susceptibility mapping using FR, AHP, LR, ANN. Korean Bhatt BP. Awasthi KD. Heyojoo BP, Silwal T, Kafle G (2013) Using Soc Geosp Inf Syst 9:13–20 geographic information system and analytical hierarchy process Kavzoglu T, Sahin EK, Colkesen I (2015) An assessment of multivari- in landslide hazard zonation. Appl Ecol Environ Sci 2(1):14–22 ate and bivariate approaches in landslide susceptibility mapping: a Bijukchhen SM. Kayastha P, Dhital MR (2012) A comparative case study of Duzkoy district. Nat Hazards 76:471–496 evaluation of heuristic and bivariate statistical modelling for Kayastha P, Dhital M. De Smedt F (2013) Application of the analytical landslide susceptibility mappings in Ghurmi–Dhad Khola, hierarchy process (AHP) for landslide susceptibility mapping: a east Nepal. Arab J Geosci. h t t p s : / / do i . o r g / 1 0 .1 0 0 7 /s 1 2 5 1 case study from the Tinau watershed, west Nepal. Comput Geosci 7-012-0569-7 52:398–408 Bui DT. Tuan TA. Klempe H. Pradhan B, Revhaug I (2016) Spatial Kevin LKW, Tay LT, Lateh H (2011) Landslide hazard mapping of prediction models for shallow landslide hazards: a comparative Penang Island using probabilistic methods and logistic regres- assessment of the efficacy of support vector machines, artificial sion. Imaging systems techniques (IST) 2011 IEEE international neural networks, kernel logistic regression, and logistic model conference, pp 273–278 tree. Landslides 13:361–378 Komac M (2006) Landslide susceptibility model using the analytical Byron Q, Luna JB, Theo VA, Cees VW, Melanie K (2016) hierarchy process method and multivariate statistics in perialpine ASCHFLOW—a dynamic landslide run-out model for medium Slovenia. Geomorphology 74:17–28 scale hazard analysis, environmental disasters. Geoenviron Dis- Kumar K, Annadurai R (2015) Comparison of frequency ratio model asters 3:29. https ://doi.org/10.1186/s4067 7-016-0064-7 and analytic hierarchy process methods upon landslide suscep- Costanzo D. Chacón J, Conoscenti C, Irigaray C, Rotigliano E (2014) tibility mapping using geospatial techniques. Disaster Adv 8(5) Forward logistic regression for earth-flow landslide susceptibil- Lee S, Pradhan B (2007) Landslide hazard mapping at Selangor, ity assessment in the Platani river basin (southern Sicily, Italy). Malaysia using frequency ratio and logistic regression models. Landslides 11:639–653 Landslides 4(1):33–41 Dai FC, Lee CF (2002) Landslide characteristics and slope instability Lee S, Sambath T (2006) Landslide susceptibility mapping in the modeling using GIS, Lantau Island, Hong Kong. Geomorphol- Damrei Romel area, Cambodia using frequency ratio and logistic ogy 42:213–228 regression models. Environ Geol 50(6):847–855 David HB (1992) Landslides: proceedings of the sixth international Lee S, Talib JA (2005) Probabilistic landslide susceptibility and factor symposium, Christchurch, 10–14 February. Landslide investiga- effect analysis. Environ Geol 47(7):982–990 tions, stability analysis techniques, Rotterdam, Balkema Lydia EE, Daniel B (2002) Land slide hazard and risk zonation map- Dehban A, Stakhri M (2013) Landslide susceptibility mapping for ping in the Rio Grande basin, central Andes of Mendoza, Argen- Poulrood earth fill dam reservoir (the comparison of two meth- tina. Mt Res Dev 22(2):177–185 ods) Iran, Guilan province. Univ J Geosci 2(1):69–76 Malik R, Ghosh S. Dhar S, Singh P, Singh M (2016) GIS based Land- Erich JP, Wolfgang K (1994) International decade for natural disaster slide hazard from zonation along national highway AY-58, from reduction (IDNDR): a challenge to science. Soil Dyn Earthq Rishikesh to Joshimath, Uttarkhand, India. Int J Adv Remote Sens Engine 13(1):45–48 GIS Geogr 4(2):41–55 Esmali Ouri A, Amirian S (2009) Landslide hazard zonation Mohammadi AT, Irani A, Sorur J (2014) The preparation of land- using MR and AHP methods and GIS techniques in Langan slide map by landslide numerical risk factor (LNRF) model and 1 3 Environmental Earth Sciences (2018) 77:405 Page 13 of 13 405 geographic information system (GIS). Egypt J Remote Sens Space Shirzadi A, Bui DT, Pham BT, Solaimani K (2017a) Shallow land- 17(2):159–170 slide susceptibility assessment using a novel hybrid intelligence Ozdemir A, Altural T (2013) A comparative study of frequency ratio, approach. Environ Earth Sci 76:60 weights of evidence and logistic regression methods for landslide Shirzadi A, Chapi K, Shahabi H, Solaimani K, Kavian A, Ahmad BB susceptibility mapping: Sultan Mountains, SW Turkey. J Asian (2017b) Rock fall susceptibility assessment along a mountainous Earth Sci 64:180–197 road: an evaluation of bivariate statistic, analytical hierarchy pro- Park S, Choi C, Kim B, Kim J (2012) Landslide susceptibility mapping cess and frequency ratio. Environ Earth Sci 76:152 using frequency ratio, analytic hierarchy process, logistic regres- Solaimani K, Mousavi SZ, Kavian A (2013) Landslide susceptibility sion, and artificial neural network methods at the Inje area, Korea. mapping based on frequency ratio and logistic regression models. Environ Earth Sci. https ://doi.org/10.1007/s1266 5-012-1842-5 Arab J Geosci 6(7):2557–2569 Pham BT, Tien Bui D, Indra P, Dholakia M (2015) Landslide suscep- Tsangaratos P, Ilia I (2016) Comparison of a logistic regression and tibility assessment at a part of Uttarakhand Himalaya, India using Naïve Bayes classifier in landslide susceptibility assessments: the GIS-based statistical approach of frequency ratio method. Int J influence of models complexity and training dataset size. Catena Eng Res Technol 4:338–344 145:164–179 Pourghasemi HR, Moradi HR, Aghda SF (2013) Landslide suscep- Umar Z, Pradhan B, Ahmad A, Neamah Jebur M, Shafapour Tehrany tibility mapping by binary logistic regression, analytical hierar- M (2014) Earthquake induced landslide susceptibility map- chy process, and statistical index models and assessment of their ping using an integrated ensemble frequency ratio and logistic performances. Nat Hazards 69:749–779. https://doi.or g/10.1007/ regression models in West Sumatera province, Indonesia. Catena s1106 9-013-0728-5 118:124–135 Pradhan B, Lee S (2010) Landslide susceptibility assessment and fac- Wang L, Guo M, Sawada K, Lin J, Zhang J (2015) Landslide suscep- tor effect analysis: backpropagation artificial neural networks and tibility mapping in Mizunami city, Japan: a comparison between comparison with frequency ratio and bivariate logistic regression logistic regression, bivariate statistical analysis and multivariate modelling Klang valley. Geomorphology 25:747–759 adaptive regression spline models. Geomorphology 135:271–282 Pradhan B, Youssef AM (2010) Manifestation of remote sensing data Wu Y, Li W, Liu P, Bai H, Wang Q, He J. Liu Y, Suns (2016a) Appli- and GIS on landslide hazard analysis using spatial-based statisti- cation of analytic hierarchy process model for landslide suscep- cal models. Arab J Geosci 3:319–326 tibility mapping in the Gangu county, Gansu province. Environ Razavizadeh S, Solaimani K, Massironi, Kavian M (2017) Mapping Earth Sci 75:422 landslide susceptibility with frequency ratio, statistical index, and Wu Y, Li W, Wang Q, Liu Q, Yang D, Xing M, Pei Y, Yan S (2016b) weights of evidence models: a case study in northern Iran. Environ Landslide susceptibility assessment using frequency ratio, statisti- Earth Sci 76:499 cal index and certainty factor models for the Gangu county, China. Romer C, Ferentinou M (2016) Shallow landslide susceptibility assess- Arab J Geosci 9:84 ment in a semiarid environment—a quaternary catchment of Kwa- Yalcin A (2008) GIS-based landslide susceptibility mapping using Zulu-Natal, South Africa. Eng Geol 201:29–44 analytical hierarchy process and bivariate statistics in Ardesen Saaty T (1980) The analytic hierarchy process: planning, priority set- (Turkey). Catena 72(1):1–12 ting, resource allocation. McGraw-Hill, New York Yilmaz I (2010) Comparison of landslide susceptibility mapping meth- Saaty T (2008) Decision making with the analytic hierarchy process. odologies for Koyulhisar, Turkey: conditional probability, logistic Int J Serv Sci 1(1):83–98 regression, artificial neural networks, and support vector machine. Shadfar S, Yamani M. Namaki M (2011) Zoning land slide hazard by Environ Earth Sci 61:821–836 information value method (IVM), density area (DA) and landslide Zhang K, Wu X, Niu R, Yang K, Zhao L (2017) The assessment of numerical risk factor (LNRF) model in Chalkrood. J Watershed landslide susceptibility mapping using random forest and decision Manag Eng 3(1):40–47 tree methods in the three Gorges reservoir area, China. Environ Shahabi H, Khezri S, Ahmad BB, Hashim M (2014) Landslide suscep- Earth Sci 76:405 tibility mapping at central Zab basin, Iran: a comparison between Zhu L, Huang JF (2006) GIS-based logistic regression method for land- analytical hierarchy process, frequency ratio and logistic regres- slide susceptibility mapping on regional scale. J Zhejiang Univ sion models. Catena 115:55–70 Sci A 7(12):2001–2017 Shirani K. Seif A (2012) The landslide hazard zonation using the statis- tically methods: the Pishkuh region, Fereidoonshahr city. J Earth Sci 22:149–158 1 3
Environmental Earth Sciences – Springer Journals
Published: May 28, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”
Daniel C.
“Whoa! It’s like Spotify but for academic articles.”
@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”
@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”
@JoseServera
DeepDyve Freelancer | DeepDyve Pro | |
---|---|---|
Price | FREE | $49/month |
Save searches from | ||
Create lists to | ||
Export lists, citations | ||
Read DeepDyve articles | Abstract access only | Unlimited access to over |
20 pages / month | ||
PDF Discount | 20% off | |
Read and print from thousands of top scholarly journals.
Already have an account? Log in
Bookmark this article. You can see your Bookmarks on your DeepDyve Library.
To save an article, log in first, or sign up for a DeepDyve account if you don’t already have one.
All DeepDyve websites use cookies to improve your online experience. They were placed on your computer when you launched this website. You can change your cookie settings through your browser.
ok to continue