Impact of DEMderived factors and analytical hierarchy process on landslide susceptibility mapping in the region of Rożnów Lake, Poland
 1.7k Downloads
 13 Citations
Abstract
Choosing appropriate landslidecontrolling factors (LCFs) in landslide susceptibility mapping (LSM) is a challenging task and depends on the nature of terrain and expert knowledge and experience. Nowadays, it is very common to use digital elevation model (DEM) and DEMderivatives, as a representation of the topographic conditions. The objective of this study is to explore topography in depth and simultaneously reduce redundant information within DEMderivatives using principal component analysis. Moreover, this study investigates the impact of DEMderived factors on LSM. Therefore, three various strategies were tested. The first strategy included a set of LCFs created from the four initial principal components, which were provided from DEMderived factors. The second strategy included a set of parameters which contained additional lithological and environmental factors. The third strategy utilises the analytical hierarchy process (AHP) to assign weights to each LCF. The LSM was performed based on landslide susceptibility index. Obtained results show that 60% of existing landslides fell into high and very high susceptibility zones using first and second strategies. It proves that topographic factors play a significant role in LSM. Adding additional lithological and environmental factors to the set of LCFs did not improve the results significantly, unless the AHP was used in the third strategy. It improved results significantly; up to 70%. Results from second and third strategies highlight utility of AHP in LSM. Presented studies were performed on the area very prone to landslide occurrence in the region of Rożnów Lake, Poland.
Keywords
Landslide susceptibility mapping Landslideconditioning factors Principal component analysis Analytical hierarchy process1 Introduction
Landslides are generally defined as unexpected movements of soil, rock and organic material under the effect of gravity (Highland and Bobrowsky 2008). Landslides are a natural hazard that cause damage to the environment in many areas of the world. The slope failures can be fatal and also can destroy or damage residential and industrial facilities, as well as agricultural and forest areas. Moreover, landslides have negative effects on the quality of water in rivers and streams (Schuster and Fleming 1986). For the theory of landslide and additional background information, the reader is referred to the literature, e.g. Highland and Bobrowsky (2008). Considering landslide impact on development and urbanisation, effective landslide assessment is required (Aleotti and Chowdhury 1999). The increasing awareness of the socioeconomic significance of landslides provides motivation to develop appropriate landslide risk zonation (Aleotti and Chowdhury 1999).
The first step for hazard and risk prediction is landslide susceptibility mapping (LSM). LSM is the evaluation of the ground’s proneness to landslides and the possibility that landslide might occur at a specific terrain or under the influence of certain factors (Pourghasemi et al. 2013). It shows the spatial distribution of landslideprone areas, usually as landslide occurrence probabilities distributed across grid cells (Goetz et al. 2015). Different methods of LSM have been broadly examined and analysed in the past decades (Mohammady et al. 2012; Goetz et al. 2015; Bai et al. 2010; Mashari et al. 2012; Tien Bui et al. 2011, Feizizadeh et al. 2014; Dimri et al. 2007; Kanungo et al. 2008). Moreover, numerous comparisons of LSM methods have been evaluated and still no single best method has been selected (Goetz et al. 2015). In general, methods of LSM may be qualitative or quantitative (Aleotti and Chowdhury 1999). Qualitative approaches are entirely based on the perception and experience of the person or persons who carry out the susceptibility assessment. The quantitative methods are considered as more objective than qualitative approaches due to their datadependent characteristic. These methods cover a broad spectrum of geotechnical engineering approaches, statistical approaches, artificial neural network or fuzzy logic methods (Aleotti and Chowdhury 1999). Statistical approaches applied for modelling of the landslide susceptibility are based on the assumption that factors which caused landslides in the past are the same or similar as those which will create landslide in the future (Guzzetti et al. 1999). Therefore, these approaches concentrate on relations between landslidecontrolling parameters and the location of existing landslides from landslide inventory map (Saadatkhah et al. 2014; Aleotti and Chowdhury 1999). Within statistical approaches, there are either bivariate or multivariate analyses (Chalkias et al. 2014). Using bivariate statistical analyses, each of factors is individually compared to the landslide inventory map. These techniques apply primarylevel weights, which are commonly based on certain rules. The widely used rule is landslide density, which is calculated as relation between the area affected by landslide pixels on a class of a specific factor and the total area of that class; expressed as a percentage (Ayalew et al. 2004; Akgun et al. 2008). The most frequently used bivariate methods are frequency ratio calculation, also called the landslide susceptibility index, probabilistic likelihood ratio (PLR), weight of evidence and statistical index (Kavzoglu et al. 2015a, b; Thiery et al. 2007; Mezughi et al. 2011; Yalcin et al. 2011; Abay and Barbieri 2012; Constantin et al. 2011). Among multivariate methods, logistic regression is widely applied (Bai et al. 2010; Mashari et al. 2012). The comparison of bivariate (statistical index) and multivariate (logistic regression) methods was performed by Tien Bui et al. (2011). This comparison indicates an almost equal predicting capacity between these two methods.
The major difficulties in quantitative methods mentioned above are the assessment of the factors related to landslide occurrence and the assignment of appropriate weights to these factors (Carrara 1988; Bui et al. 2016; Kavzoglu et al. 2015a, b). Hence, many researchers tested different approaches, which analyse the spatial distribution of landslides with different LCFs (Bui et al. 2016; Mahalingam et al. 2016). Nowadays, the significance of each LCF can be easily validated using geographic information system (GIS). In addition, GIS provides a powerful tool for multicriteria decision analysis (GISMCDA) which became more popular over the past years (Ahmed 2015; Aleotti and Chowdhury 1999). The concept of MCDA assumes that each LCF can be combined by applying primary and secondarylevel weights. Primarylevel weights follow the same rule as bivariate approaches. However, secondarylevel weights are expert opinionbased weights (Ayalew et al. 2004; Ahmed 2015). Among expert opinion methods for weights assignment, an analytic hierarchy process (AHP) is a technique which has been successfully applied to many decision maker systems (Kayastha et al. 2013; Ayalew et al. 2005). The AHP technique uses a pairwise relative comparison between each LCF (Saaty 1980). Due to this fact, the weights assignment is getting more complicated and timeconsuming if the quantity of LCF increases. Except for statistical methods, data mining using fuzzy logic (Feizizadeh et al. 2014; Dimri et al. 2007; Kanungo et al. 2008) and artificial neural network models (Ermini et al. 2005; Lee and Evangelista 2006; Kanungo et al. 2006) have also been applied to the LSM using GIS. Furthermore, applications of data mining and soft computing methods are increasing rapidly. Among these approaches also decisions trees, Bayesian networks, etc., are characterised by effectiveness in LSM (Bui et al. 2016).
Various natural and manmade factors can be considered as the LCFs. For that reason, the selection of the appropriate LCFs is a challenging task. LSM requires that topographic, environmental, geological and hydrological parameters should be taken into account. Some researchers assume that the accuracy of the created susceptibility map increases proportionally with the quantity of LCFs used (Jebur et al. 2014). Other scientists state that a small number of LCFs is satisfactory to produce landslide susceptibility maps with a reasonable quality (Jebur et al. 2014; Mahalingam et al. 2016). The investigations of Kingsbury et al. (1992) present that the additional factors (soil type, land use, slope aspect, proximity to watercourses) did not increase the reliability of the susceptibility maps and are suitable to a particular study area only. Therefore, no specific rule exists to define how many conditioning factors are sufficient for the susceptibility analysis on a given study area (Pourghasemi et al. 2013; Mahalingam et al. 2016). Moreover, various factors have a different impact on landslide occurrence; therefore, MCDA provides the possibility to include expert opinion to describe their impact.
 1.
deeply explore topographic information delivered from DEM by calculating ten different DEMderived LCFs,
 2.
reduce redundant information within DEMderived factors by applying PCA technique,
 3.
reduce the pairwise combinations within AHP by applying PCA technique,
 4.
investigate the impact of DEMderived factors and AHP technique for LSM on the study area by applying three strategies with different data set and methods,
 5.
produce a landslide susceptibility map at the regional scale of the study area using probabilistic likelihood ratio method (PLR).
2 Study area characteristic
3 Data used
Data used to create thematic layers
Data  Source  Type 

Landslide inventory map  SOPO project  Raster 
DEM  ISOK project  Point cloud 
Land use and drainage maps  Topographic object database  Vector 
Roads map  Open street map  Vector 
Lithology map  Polish geological institute  Raster 
3.1 Landslide inventory map
Due to the problem related to landslide occurrences and activities, Polish Geological Institute created a “Landslide Counteracting System” called SOPO (Borkowski et al. 2011). The aim of this system is to collect landslide inventory maps of all existing landslides in Poland and put them into one database. This database stores information about active, inactive and landslideprone areas. The online database content is available to the public to browse and is free of charge. The SOPO database showed more than 250 landslides for the study area. The landslideaffected areas cover 6.51 km^{2}, which means that 25% of total area is affected by landslides. It proves that the study area is very susceptible to the landslide activity and needs efficient landslide susceptibility assessment. Therefore, for modelling, 70% of randomly selected landslides was used and 30% was used for validation.
3.2 DEMderived LCFs
All DEMdelivered layers are in GRID format with the cell size equal to 5 m × 5 m that was generated from point cloud with resolution of 4–6 points/m^{2}. Point cloud was obtained from airborne laser scanning in the framework of the ISOK project (Pawłuszek et al. 2014). According to Pawłuszek et al. (2014), the height component accuracy of the ISOK data does not exceed 23 cm for forested areas. Based on the DEM, the presented above, geomorphological and hydrological thematic data layers were computed. Since the mostused LCFs are commonly known, we omitted background relations needed to calculate them. We provided equations below only for new and rarely used LCFs.
3.2.1 Elevation
3.2.2 Slope
3.2.3 Morphological gradient
3.2.4 Aspect
3.2.5 Area solar radiation (ASR)
3.2.6 Roughness
3.2.7 Topographic position index (TPI)
3.2.8 Topographic wetness index (TWI)
TWI index was calculated using script from Geomorphometry and Gradient metrics written by Jeffrey Evans, which is applicable in ArcGIS (Evans et al. 2014).
3.2.9 Stream power index (SPI)
3.2.10 Shaded relief
3.3 Lithology
3.4 Environmental factors
3.4.1 Distance from roads
3.4.2 Distance from drainage
3.4.3 Land use
4 Methods
Differences between strategies applied for LSM
First strategy  Second strategy  Third strategy  

Data used  4 principal components  4 principal component, lithology, distance do drainage, distance to roads, land use  4 principal component, lithology, distance do drainage, distance to roads, land use 
Method  Probabilistic likelihood ratio (PLR)  Probabilistic likelihood ratio (PLR)  Multicriteria decision analysis (PLR + AHP) 
4.1 Reduction of DEMderived landslideconditioning factors
As mentioned earlier, no universal guidelines exist for selecting appropriate factors that affect landslides and can be used in the susceptibility mapping. Moreover, an excessive number of LCFs significantly extends computational time. Nevertheless, in the presented approach abundant topographic LCFs were considered. As mentioned previously, quantitative approaches demand weights to be assigned to each LCF. Hence, many researchers use AHP to determine weights for landslidecontrolling factors. This method requires pairwise comparisons of each LCF. The AHP is subjective approach, because it is based on the knowledge and an expert opinion. Moreover, if the number of used factors is bigger than the weights assigning process is the more complex. On the other hand, the selection of causal factors reflects the nature of the research area and has a certain degree of affinity with landslides (Ayalew et al. 2005). Since the DEMdelivered factors are the first or second order derivatives of the DEM, they contain redundant information. For that reason, it is rational to reduce such information. In order to extract as much information as possible from the DEMderived factors presented above and simultaneously reduce the number of components used, the PCA was applied.
4.2 Calculation of probabilistic likelihood ratio (PLR)

LCFs values categorisations.

Probabilistic likelihood ratio computation according to Eq. (3)

Final LSI calculation for each pixel of the study area according to the Eq. (4)
At this stage, the final LSI was produced for each pixel by summing the PLR values for all LCFs computed for that pixel (Chalkias et al. 2014).
4.3 Multicriteria decision analysis
4.3.1 Principal component physical representation
Correlation coefficients between PC components and DEMderived factors
DEMderived factor  Principal component  

1  2  3  4  
Aspect  1.00  0.00  −0.02  −0.01 
ASR  0.04  −0.20  0.48  0.79 
CTI  −0.01  −0.62  0.31  −0.38 
Elevation  0.02  0.84  0.53  −0.07 
Gradient  −0.03  0.68  −0.62  0.03 
Shaded relief  0.04  −0.58  0.66  0.04 
Slope  −0.03  0.63  −0.58  0.03 
SPI  0.00  0.09  0.00  −0.22 
Roughness  −0.02  0.28  −0.58  0.18 
TPI  0.00  0.01  0.14  0.19 
4.3.2 Secondarylevel weight assignment by AHP
The appropriate weight assignment is a challenging task. For that purpose, the AHP is used in many landslide studies in the world (Akgun et al. 2008; Kayastha et al. 2013; Ayalew et al. 2005; Feizizadeh et al. 2014). AHP is the multicriteria decision model, which uses pairwise comparisons of relative factors without inconsistencies in the decision process (Saaty 1980). The essential advantage of the AHP is involving the expert’s knowledge and experiences in the weights assigning process. This may improve the quality of the susceptible maps. On the other hand, the AHP is subjective and different researchers can achieve different results.
Scale of preference between two parameters in AHP (Saaty 1977)
Preference factor  Degree of preference  Explanation 

1  Equally  Two factors contribute equally to the objective 
3  Moderately  Experience and judgment slightly to moderately favour one factor over another 
5  Strongly  Experience and judgment strongly or essentially favour one factor over another 
7  Very strongly  A factor is strongly favoured over another and its dominance is showed in practice 
9  Extremely  The evidence of favouring one factor over another gives the highest degree of affirmation possible. 
2,4,6,8  Intermediate  Used to represent compromises between the preferences in weight 1,3,5,7 and 9 
Reciprocals  Opposites  Used for inverse comparison 
Pairwise comparison matrix and weights for landslide causative factors
LCFs (after PCA)  The first component  The second component  The third component  The fourth component  Geology  Distance from river  Distance from roads  Land use  Weights 

The first component  1  0.022  
The second component  9  1  0.332  
The third component  7  1/2  1  0.181  
The fourth component  7  1/4  1  1  0.129  
Lithology  4  1/6  1/3  1/4  1  0.050  
Distance from river  8  1  1/3  1  2  1  0.127  
Distance from roads  4  1/5  1/2  1  2  1  1  0.089  
Land use  2  1/8  1/3  1/2  2  1  1  1  0.07 
4.4 Isodata classification
In order to create landslide susceptibility zones, it is necessary to reclassify the continuous values of LSI in the final map. Various classification methods are available in GIS; however, four methods, namely: standard deviations, equal intervals, natural breaks and quantile, have been examined in the landslide studies (Ayalew and Yamagishi 2005). All mentioned methods of the classification depend on the statistical parameters. Previously, clustering was not commonly used to differentiate the susceptible classes. The natural breaks classification, the quantile classification and clustering were tested in this study to divide LSI in five susceptible classes. Based on the validation method presented in Sect. 4.5, the clustering provided the best performance. In this study, the isodata unsupervised classification was used to create classes of the landslide susceptibility. The isodata were performed in ArcGIS software (Ball and Hall 1965).
4.5 Validation of landslide susceptibility maps
Remondo et al. (2003) proposed a validation method for LSM, where the original landslide inventory map is randomly split in two parts: one for the susceptibility analysis and second for validation process. According to this concept, 70% of randomly selected landslide was used for modelling and 30% was used for validation in this study.
4.5.1 SCAI index
Moreover, the seed cell area index (SCAI) validation technique proposed by Süzen and Doyuran (2004) was implemented. The SCAI is calculated by dividing percentage of pixels of the specific landslide susceptibility class by percentage of existing landslides pixels in the specific landslide susceptibility zone. SCAI shows the density of landslides among the landslide susceptibility zones. It is expected that the high and very high susceptibility classes should have very small SCAI values and low, very low susceptibility zones should have higher SCAI values (Süzen and Doyuran 2004; Kıncal et al. 2009).
4.5.2 Difference image analysis
In order to compare maps from three strategies, difference image analysis was applied. Difference image analysis provides information how maps are different from each other. By comparing two landslide susceptibility maps with different susceptibility zones, the socalled residual map is received (Gupta et al. 2008). This map elucidates how pixels shift from one landslide susceptibility zone to another zone between two maps. Therefore, a residual map can have a maximum five different classes: no difference, onezone difference, twozone difference, threezone difference and fourzone difference. The best performance of LSM is presented by third strategy, where AHP method was used in order to assign weights to LCF.
5 Obtained landslide susceptibility maps and discussion
Landslide density among the landslide susceptibility classes, the first strategy
Susceptibility  Area (%)  Seed (%)  SCAI 

Very low  0.20  0.02  9.28 
Low  0.20  0.09  2.25 
Moderate  0.27  0.30  0.89 
High  0.21  0.35  0.60 
Very high  0.13  0.25  0.51 
Landslide density among the landslide susceptibility classes, the second strategy
Susceptibility  Area (%)  Seed (%)  SCAI 

Very low  0.16  0.02  7.12 
Low  0.25  0.10  2.40 
Moderate  0.26  0.27  0.94 
High  0.19  0.33  0.59 
Very high  0.14  0.27  0.52 
Landslides density among the landslide susceptibility classes, the third strategy
Susceptibility  Area (%)  Seed (%)  SCAI 

Very low  0.15  0.02  6.30 
Low  0.17  0.06  3.04 
Moderate  0.26  0.22  1.21 
High  0.26  0.41  0.64 
Very high  0.16  0.29  0.53 
According to the first strategy, very high and high susceptibility classes contain 60% of existing landslides area used for validation. According to the second strategy, the set of parameters was extended by lithological and environmental factors. Environmental factors include distance to drainage, distance to roads and land use. Using the second strategy, very high and high susceptibility class contains also 60% of the existing landslide area used for validation.
Comparing the results to the first strategy, it can be seen that the same percentage of landslides areas fell into high and very high susceptibility class. However, SCAI index is higher for very low and low classes for strategy using only DEMderived factors. According to that, it can be concluded that LSM using only the four principal components obtained from DEMderived factors provides slightly higher performance than LSM that uses all factors. It is supposed that environmental factors should increase the performance of LSM. The reason for that could be that extended set of DEMderived factors was taken into account. For instance SPI or CTI, which are hydrological factors derived from DEM, contains information, which can be also contained in distance to drainage factor.
In third strategyMCDA, full set of LCFs was applied with weights assigned using AHP (Sect. 4.3). The SCAI index is smaller for very low susceptible class. Very high and high susceptibility classes contain 70% of the existing landslides areas used for validation. It means that susceptibility mapping using full data set with weights exhibits the highest performance.
6 Summary and conclusion
In presented study, three various strategies were applied to create landslide susceptibility maps for the area of Rożnów Lake, Poland. The first strategy used only DEMderived conditioning factors reduced to four uncorrelated principal components. The second strategy used the full set of LCFs that included four principal components, lithological and environmental factors. The third strategy utilised the full set of LCFs with weights assigned using AHP. The produced susceptibility maps were compared with 30% of randomly selected landslides for validation, and the effectiveness of these three strategies was tested. Based on the achieved results, the third strategy exhibits the best performance. According to our results, the only way to achieve improved performance of LSM is to assign appropriate weights to LCFs, e.g. deploying AHP.
Besides producing the landslide susceptibility map for the study area, the main objective of this study was to investigate the impact of DEMderived and environmental conditioning parameters for LSM in the area of Rożnów Lake. The difference image analysis between first and second strategies demonstrated the usefulness factors delivered from DEM. Comparing the SCAI, it can be concluded that LSM using only the four principal components obtained from DEMderived factors provides slightly higher performance than LSM that uses all factors. Based on achieved results, it can be stated that landslide susceptibility maps, created using only DEMdelivered factors, provide the possibility to produce reasonable landslide susceptible zones in areas where full data collection is complicated and timeconsuming. Approximately the same content of landslide areas (60%) selected for validation fell into high and very high susceptible zones in the first and the second strategies. The reason for that could be that nonDEMdelivered factors do not provide additional information, because of so deep exploring of the DEM. For instance, landslides often occur close to rivers, which can be indirectly represented by slope, roughness index or stream power index. Another reason for than could be that buffer classes of distance from drainage or roads were not chosen appropriately. It could be also that no relationship exists between landslides and land cover or lithology. It means that these LCF are not suitable to this particular study area.
After applying the AHP to assign weights to the LCFs, the effectiveness of the LSM increased up to 70%. Based on the weights assigned to the LCFs, it can be concluded that in the LSM the most important LCFs are the principal components two, three and four. They mostly correspond to the slope, elevation, roughness and ASR. Moreover, results indicated that distance to rivers is also a relevant factor in LSM. However, lithology does not have significant impact on LSM. Based on PLR obtained for each lithological class, it can be seen that the landslide occurrence in each lithological category is very similar. The reason for that could be the geological structure in the study area. Each lithological unit exhibits the same proneness for landslide occurrence. Moreover, the weights indicated that land use and aspect are not significant LCFs in LSM.
An open question is finding a proper criterion for choosing the optimal number of the principal components that represent DEMderivatives in order to reduce the computational effort and the complexity of LSM and simultaneously to achieve a seasonable accuracy of LSM. This issue should be addressed in the future research.
Based on achieved results, this approach can be applicable to the landslide susceptibility mapping in other regions in the world. However, it is important to assign appropriate weights into the specific landslidecontrolling factors, because it is mostly attributable to the nature of the terrain and type of landslide. On the other hand, most of the landslides located within the study area have different types (translational, rotational and combined rockdebris slides or debris slides); therefore, it suggests that methodology is more comprehensive and not narrowed into one type of landslide.
Akgun et al. (2008) obtain higher performance of the LSM using the same PLR method with weights, assigned from the AHP in the study area in Turkey. On the other hand, Komac (2012) applied other bivariate Monte Carlo approach in Slovenia achieving also 70% of correctness. Similar results can be found in work (Mashari et al. 2012; Akgun and Türk 2010; Ayalew et al. 2005; Kanungo et al. 2006). According to the results, the presented approach provides diverse results for diverse study areas in the world. Moreover, choosing the most appropriate method for the LSM is essential, because it provides different results depending on the nature of the terrain. For this reason, more studies have to be performed in the Carpathian Mountains in order to select suitable methods for the LSM in this region. It will be investigated by authors in a further work. Furthermore, many other LCFs have not been tested; therefore, the authors will test importance of other LCFs on LSM, for instance distance to faults, distance to lineament, NDVI and soil depth or texture.
References
 Abay A, Barbieri G (2012) Landslide susceptibility and causative factors evaluation of the landslide area of Debresina, in the southwestern Afar escarpment, Ethiopia. J Earth Sci Eng 2(3):133–144Google Scholar
 Abdi H, Williams LJ (2010) Principal component analysis. Wiley Interdiscip Rev Comput Stat 2(4):433–459. doi: 10.1002/wics.101 CrossRefGoogle Scholar
 Ahmed B (2015) Landslide susceptibility mapping using multicriteria evaluation techniques in Chittagong Metropolitan Area, Bangladesh. Landslides 12(6):1077–1095. doi: 10.1007/s103460140521x CrossRefGoogle Scholar
 Akgun A, Türk N (2010) Landslide susceptibility mapping for Ayvalik (Western Turkey) and its vicinity by multicriteria decision analysis. Environ Earth Sci 61(3):595–611. doi: 10.1007/s1266500903731 CrossRefGoogle Scholar
 Akgun A, Dag S, Bulut F (2008) Landslide susceptibility mapping for a landslideprone area (Findikli, NE of Turkey) by likelihoodfrequency ratio and weighted linear combination models. Environ Geol 54(6):1127–1143. doi: 10.1007/s0025400708828 CrossRefGoogle Scholar
 Aleotti P, Chowdhury R (1999) Landslide hazard assessment: summary review and new perspectives. Bull Eng Geol Environ 58(1):21–44. doi: 10.1007/s100640050066 CrossRefGoogle Scholar
 Ayalew L, Yamagishi H (2005) The application of GISbased logistic regression for landslide susceptibility mapping in the KakudaYahiko Mountains, Central Japan. Geomorphology 65(1):15–31. doi: 10.1016/j.geomorph.2004.06.010 CrossRefGoogle Scholar
 Ayalew L, Yamagishi H, Ugawa N (2004) Landslide susceptibility mapping using GISbased weighted linear combination, the case in Tsugawa area of Agano River, Niigata Prefecture, Japan. Landslides 1(1):73–81. doi: 10.1007/s1034600300069 CrossRefGoogle Scholar
 Ayalew L, Yamagishi H, Marui H, Kanno T (2005) Landslides in Sado Island of Japan: Part II. GISbased susceptibility mapping with comparisons of results from two methods and verifications. Eng Geol 81(4):432–445. doi: 10.1016/j.enggeo.2005.08.004 CrossRefGoogle Scholar
 Bai SB, Wang J, Lü GN, Zhou PG, Hou SS, Xu SN (2010) GISbased logistic regression for landslide susceptibility mapping of the Zhongxian segment in the Three Gorges area, China. Geomorphology 115(1):23–31CrossRefGoogle Scholar
 Ball GH, Hall DJ (1965) Isodata, a novel method of data analysis and pattern classification. Stanford Research Institute, Menlo Park. doi: 10.1016/j.geomorph.2009.09.025 Google Scholar
 Borkowski A, Perski Z, Wojciechowski T, Jóźków G, Wójcik A (2011) Landslides mapping in Rożnów Lake vicinity, Poland using airborne laser scanning data. Acta Geodyn Geomater 8(3):163Google Scholar
 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(3):1413–1444. doi: 10.1007/s1106901198442 CrossRefGoogle Scholar
 Bui DT, Tuan TA, Hoang ND, Thanh NQ, Nguyen DB, Van Liem N, Pradhan B (2016) Spatial prediction of rainfallinduced landslides for the Lao Cai area (Vietnam) using a hybrid intelligent approach of least squares support vector machines inference model and artificial bee colony optimization. Landslides. doi: 10.1007/s1034601607119 Google Scholar
 Carrara A (1988) Landslide hazard mapping by statistical methods. A black—box approach. In: The proceedings of the workshop on natural disasters in European Mediterranean Countries, Italy, pp 208–224Google Scholar
 Chalkias C, Ferentinou M, Polykretis C (2014) GISbased landslide susceptibility mapping on the Peloponnese Peninsula, Greece. Geosciences 4(3):176–190. doi: 10.3390/geosciences4030176 CrossRefGoogle Scholar
 Chen W, Li X, Wang Y, Liu S (2013) Landslide susceptibility mapping using LiDAR and DMC data: a case study in the Three Gorges area, China. Environ Earth Sci 70(2):673–685. doi: 10.1007/s1266501221518 CrossRefGoogle Scholar
 Constantin M, Bednarik M, Jurchescu MC, Vlaicu M (2011) Landslide susceptibility assessment using the bivariate statistical analysis and the index of entropy in the Sibiciu Basin (Romania). Environ Earth Sci 63(2):397–406. doi: 10.1007/s126650100724y CrossRefGoogle Scholar
 Dahal RK, Hasegawa S, Nonomura A, Yamanaka M, Masuda T, Nishino K (2008) GISbased weightsofevidence modelling of rainfallinduced landslides in small catchments for landslide susceptibility mapping. Environ Geol 54:311–324. doi: 10.1007/s0025400708183 CrossRefGoogle Scholar
 Dimri S, Lakhera RC, Sati S (2007) Fuzzybased method for landslide hazard assessment in active seismic zone of Himalaya. Landslides 4(2):101–111. doi: 10.1007/s1034600600686 CrossRefGoogle Scholar
 Donati L, Turrini MC (2002) An objective method to rank the importance of the factors predisposing to landslides with the GIS methodology: application to an area of the Apennines (Valnerina; Perugia, Italy). Eng Geol 63(3):277–289. doi: 10.1016/s00137952(01)000874 CrossRefGoogle Scholar
 Ermini L, Catani F, Casagli N (2005) Artificial neural networks applied to landslide susceptibility assessment. Geomorphology 66(1):327–343. doi: 10.1016/j.geomorph.2004.09.025 CrossRefGoogle Scholar
 Evans JS, Oakleaf J, Cushman SA, Theobald D (2014) An ArcGIS toolbox for surface gradient and geomorphometric modeling. Version 2.00. http://evansmurphy.wix.com/evansspatial. Accessed Dec 2015
 Feizizadeh B, Roodposhti MS, Jankowski P, Blaschke T (2014) A GISbased extended fuzzy multicriteria evaluation for landslide susceptibility mapping. Comput Geosci 73:208–221. doi: 10.1016/j.cageo.2014.08.001 CrossRefGoogle Scholar
 Gessler PE, Moore ID, McKenzie NJ, Ryan PJ (1995) Soillandscape modelling and spatial prediction of soil attributes. Int J GIS 9(4):421–432. doi: 10.1080/02693799508902047 Google Scholar
 Glenn NF, Streutker DR, Chadwick DJ, Thackray GD, Dorsch SJ (2006) Analysis of LiDARderived topographic information for characterizing and differentiating landslide morphology and activity. Geomorphology 73(1):131–148. doi: 10.1016/j.geomorph.2005.07.006 CrossRefGoogle Scholar
 Goetz JN, Brenning A, Petschko H, Leopold P (2015) Evaluating machine learning and statistical prediction techniques for landslide susceptibility modeling. Comput Geosci 81:1–11. doi: 10.1016/j.cageo.2015.04.007 CrossRefGoogle Scholar
 Gokceoglu C, Sonmez H, Nefeslioglu HA, Duman TY, Can T (2005) Kuzulu landslide (Sivas, Turkey) and landslidesusceptibility map of its near vicinity. Eng Geol 81:65–83. doi: 10.1016/j.enggeo.2005.07.011 CrossRefGoogle Scholar
 Gorczyca E, WrońskaWałach D, Długosz M. (2013) Landslide hazards in the Polish Flysch Carpathians: example of Łososina Dolna Commune. In: Geomorphological impacts of extreme weather, Springer, Netherlands, pp 237–250. doi: 10.1007/9789400763012_15
 Gupta RP, Kanungo DP, Arora MK, Sarkar S (2008) Approaches for comparative evaluation of raster GISbased landslide susceptibility zonation maps. Int J Appl Earth Obs Geoinf 10(3):330–341. doi: 10.1016/j.jag.2008.01.003 CrossRefGoogle Scholar
 Guzzetti F, Carrara A, Cardinali M, Reichenbach P (1999) Landslide hazard evaluation: a review of current techniques and their application in a multiscale study, Central Italy. Geomorphology 31(1):181–216. doi: 10.1016/s0169555x(99)000781 CrossRefGoogle Scholar
 Guzzetti F, Mondini AC, Cardinali M, Fiorucci F, Santangelo M, Chang KT (2012) Landslide inventory maps: new tools for an old problem. Earth Sci Rev 112(1):42–66. doi: 10.1016/j.earscirev.2012.02.001 CrossRefGoogle Scholar
 Hengl T, Reuter HI (2009) Geomorphometry: concepts, software, applications, developments in soil science 33. Elsevier, HungaryGoogle Scholar
 Highland L, Bobrowsky PT (2008) The landslide handbook: a guide to understanding landslides. US Geological Survey, Reston, p 129. doi: 10.1007/9783642220876_5 Google Scholar
 Jebur MN, Pradhan B, Tehrany MS (2014) Optimization of landslide conditioning factors using very highresolution airborne laser scanning (LiDAR) data at catchment scale. Remote Sens Environ 152:150–165. doi: 10.1016/j.rse.2014.05.013 CrossRefGoogle Scholar
 Jenness J, Brost B, Beier P (2011). Land facet corridor designer: extension for ArcGIS. Jenness Enterprises. http://www.jennessent.com/downloads/Land_Facet_Tools_A4.pdf. Accessed Dec 2015
 Kanungo DP, Arora MK, Sarkar S, Gupta RP (2006) 0 A comparative study of conventional, ANN black box, fuzzy and combined neural and fuzzy weighting procedures for landslide susceptibility zonation in Darjeeling Himalayas. Eng Geol 85(3):347–366. doi: 10.1016/j.enggeo.2006.03.004 CrossRefGoogle Scholar
 Kanungo DP, Arora MK, Gupta RP, Sarkar S (2008) Landslide risk assessment using concepts of danger pixels and fuzzy set theory in Darjeeling Himalayas. Landslides 5(4):407–416. doi: 10.1007/s1034600801343 CrossRefGoogle Scholar
 Kavzoglu T, Sahin EK, Colkesen I (2015a) An assessment of multivariate and bivariate approaches in landslide susceptibility mapping: a case study of Duzkoy district. Nat Hazards 76(1):471–496. doi: 10.1007/s1106901415068 CrossRefGoogle Scholar
 Kavzoglu T, Sahin EK, Colkesen I (2015b) Selecting optimal conditioning factors in shallow translational landslide susceptibility mapping using genetic algorithm. Eng Geol 192:101–112. doi: 10.1016/j.enggeo.2015.04.004 CrossRefGoogle Scholar
 Kayastha P, Dhital MR, De Smedt F (2013) Application of the analytical hierarchy process (AHP) for landslide susceptibility mapping: a case study from the Tinau watershed, west Nepal. Comput Geosci 52:398–408. doi: 10.1016/j.cageo.2012.11.003 CrossRefGoogle Scholar
 Kıncal C, Akgun A, Koca MY (2009) Landslide susceptibility assessment in the Izmir (West Anatolia, Turkey) city center and its near vicinity by the logistic regression method. Environ Earth Sci 59(4):745–756. doi: 10.1007/s1266500900700 CrossRefGoogle Scholar
 Kingsbury PA, Hastie WJ, Harrington AJ (1992) Regional landslip hazard assessment using a geographic information system. In sixth international symposium on landslidesGoogle Scholar
 Komac M (2012) Regional landslide susceptibility model using the Monte Carlo approach–the case of Slovenia. Geol Q 56(1):41–54Google Scholar
 Lee S, Evangelista DG (2006) Earthquakeinduced landslidesusceptibility mapping using an artificial neural network. Nat Hazards Earth Syst Sci 6(5):687–695. doi: 10.5194/nhess66872006 CrossRefGoogle Scholar
 Mahalingam R, Olsen MJ, O’Banion MS (2016) Evaluation of landslide susceptibility mapping techniques using lidarderived conditioning factors (Oregon case study). Geomat Nat Hazards Risk. doi: 10.1080/19475705.2016.1172520 Google Scholar
 Mashari S, Solaimani K, Omidvar E (2012) Landslide susceptibility mapping using multiple regression and GIS tools in Tajan Basin, North of Iran. Environ Nat Resour Res 2(3):43. doi: 10.5539/enrr.v2n3p43 Google Scholar
 McKean J, Roering J (2004) Objective landslide detection and surface morphology mapping using highresolution airborne laser altimetry. Geomorphology 57(3):331–351. doi: 10.1016/s0169555x(03)001648 CrossRefGoogle Scholar
 Mezughi TH, Akhir JM, Rafek AGM, Abdullah I (2011) Landslide susceptibility assessment using frequency ratio model applied to an area along the EW highway (GerikJeli). Am J Environ Sci 7(1):43–50CrossRefGoogle Scholar
 Mohammady M, Pourghasemi HR, Pradhan B (2012) Landslide susceptibility mapping at Golestan Province, Iran: a comparison between frequency ratio, DempsterShafer, and weightsofevidence models. J Asian Earth Sci 61:221–236. doi: 10.1016/j.jseaes.2012.10.005 CrossRefGoogle Scholar
 Moore ID, Grayson RB, Ladson AR (1991) Digital terrain modelling: a review of hydrological, geomorphological, and biological applications. Hydrol Process 5:3–30. doi: 10.1002/hyp.3360050103 CrossRefGoogle Scholar
 Moore ID, Gessler PE, Nielsen GA, Petersen GA (1993) Terrain attributes: estimation methods and scale effects. In: Jakeman AJ, Beck MB, McAleer M (eds) Modeling change in environmental systems. Wiley, London, pp 189–214Google Scholar
 Ozturk U, Tarakegn YA, Longoni L, Brambilla D, Papini M, Jensen J (2016) A simplified earlywarning system for imminent landslide prediction based on failure index fragility curves developed through numerical analysis. Geomat Nat Hazards Risk 7(4):1406–1425. doi: 10.1080/19475705.2015.1058863 CrossRefGoogle Scholar
 Pawłuszek K, Borkowski A (2016) Landslides identification using Airborne Laser Scanning data derived topographic terrain attributes and support vector machine classification. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences  XXIII ISPRS Congress, Prague, Czech Republic, 12–19 July 2016, Vol. XLIB8 No. Comm. VIII, WG VIII/1, Göttingen, Germany, pp 145–149. doi: 10.5194/isprsarchivesXLIB81452016 CrossRefGoogle Scholar
 Pawłuszek K, Ziaja M, Borkowski A (2014) Accuracy assessment of the height component of the airborne laser scanning data collected in the ISOK system for the Widawa River Valley. Acta Sci Pol Geod Descr Terr 13(3–4):27–38Google Scholar
 Pourghasemi HR, Mohammady M, Pradhan B (2012) Landslide susceptibility mapping using index of entropy and conditional probability models in GIS: Safarood Basin, Iran. Catena 97:71–84. doi: 10.1016/j.catena.2012.05.005 CrossRefGoogle Scholar
 Pourghasemi HR, Pradhan B, Gokceoglu C, Mohammadi M, Moradi HR (2013) Application of weightsofevidence and certainty factor models and their comparison in landslide susceptibility mapping at Haraz watershed, Iran. Arab J Geosci 6(7):2351–2365. doi: 10.1007/s1251701205327 CrossRefGoogle Scholar
 Pourghasemi HR, Moradi HR, Aghda SF, Gokceoglu C, Pradhan B (2014) GISbased landslide susceptibility mapping with probabilistic likelihood ratio and spatial multicriteria evaluation models (North of Tehran, Iran). Arab J Geosci 7(5):1857–1878. doi: 10.1007/s125170120825x CrossRefGoogle Scholar
 Remondo J, González A, De Terán JRD, Cendrero A, Fabbri A, Chung CJF (2003) Validation of landslide susceptibility maps; examples and applications from a case study in Northern Spain. Nat Hazards 30(3):437–449. doi: 10.1023/b:nhaz.0000007201.80743.fc CrossRefGoogle Scholar
 Rivest JF, Soille P, Beucher S (1992) Morphological gradients. In SPIE/IS&T 1992 symposium on electronic imaging: science and technology. International Society for Optics and Photonics. pp 139–150. doi: 10.1117/12.58373
 Saadatkhah N, Azman K, Lee ML (2014) Qualitative and quantitative landslide susceptibility assessments in Hulu Kelang area, Malaysia. EJGE C 19(2014):545–563. doi: 10.1007/s1070601498188 Google Scholar
 Saaty TL (1977) A scaling method for priorities in hierarchical structures. J Math Psychol 15(3):234–281. doi: 10.1016/00222496(77)900335 CrossRefGoogle Scholar
 Saaty TL (1980) The analytic hierarchy process: planning, priority setting, resources allocation. McGraw, New YorkGoogle Scholar
 Saaty TL (2000) Fundamentals of decision making and priority theory with the analytic hierarchy process, vol 6. Rws Publications, Pittsburgh. doi: 10.1007/9789401597999_2 Google Scholar
 Sarkar S, Kanungo DP (2004) An integrated approach for landslide susceptibility mapping using remote sensing and GIS. Photogramm Eng Remote Sens 70(5):617–625. doi: 10.14358/pers.70.5.617 CrossRefGoogle Scholar
 Schulz WH (2004) Landslides mapped using LIDAR imagery, seattle, Washington. U.S. Geological Survey OpenFile Report 20041396Google Scholar
 Schuster RL, Fleming RW (1986) Economic losses and fatalities due to landslides. Bull As Eng Geol 23(1):11–28. doi: 10.2113/gseegeosci.xxiii.1.11 Google Scholar
 Soille P (2013) Morphological image analysis: principles and applications. Springer, BerlinGoogle Scholar
 Solanas Pérez A, Manolov R, Leiva Ureņa D, Richard’s MM (2011) Retaining principal components for discrete variables. Anu Psicol 41(1–3):33–50Google Scholar
 Starkel L (1972) An outline of the relief of the Polish Carpathians and its importance for human management. Probl Zagospod Ziem Gór 10:75–150Google Scholar
 Süzen ML, Doyuran V (2004) A comparison of the GIS based landslide susceptibility assessment methods: multivariate versus bivariate. Environ Geol 45(5):665–679. doi: 10.1007/s0025400309178 CrossRefGoogle Scholar
 Thiery Y, Malet JP, Sterlacchini S, Puissant A, Maquaire O (2007) Landslide susceptibility assessment by bivariate methods at large scales: application to a complex mountainous environment. Geomorphology 92(1):38–59. doi: 10.1016/j.geomorph.2007.02.020 CrossRefGoogle Scholar
 Van Den Eeckhaut M, Poesen J, Verstraeten G, Vanacker V, Moeyersons J, Nyssen J, van Beek LPH, Vandekerckhove L (2007) Use of LIDARderived images for mapping old landslides under forest. Earth Surf Proc Land 32:754–769. doi: 10.1002/esp.1417 CrossRefGoogle Scholar
 van Westen CJ, Castellanos E, Kuriakose SL (2008) Spatial data for landslide susceptibility, hazard, and vulnerability assessment: an overview. Eng Geol 102(3):112–131. doi: 10.1016/j.enggeo.2008.03.010 CrossRefGoogle Scholar
 Wilson JP, Gallant JC (2000) Terrain analysis principles and applications. Wiley and Sons, New YorkGoogle Scholar
 Woźniak A (2014) Anomalously high monthly precipitation totals in the Polish Carpathian Mountains and their foreland (1881–2010). Prace Geograficzne. 138:7–28 [in Polish] Google Scholar
 Yalcin A, Reis S, Aydinoglu AC, Yomralioglu T (2011) A GISbased comparative study of frequency ratio, analytical hierarchy process, bivariate statistics and logistics regression methods for landslide susceptibility mapping in Trabzon, NE Turkey. Catena 85(3):274–287. doi: 10.1016/j.catena.2011.01.014 CrossRefGoogle Scholar
Copyright information
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.