Landslide susceptibility mapping in the Penang Island, Malaysia

말레이시아 페낭섬의 산사태 민감도 지도작성

  • cc icon
  • ABSTRACT

    The purpose of this research is to identify and classify the Penang Island according to the grade of present or potential danger of land failure happening by generating landslide susceptibility zonation map. This work presents Analytical Hierarchy Process(AHP) method to identify the landslide prone areas. Seven factors were considered in this analysis. Sensitivity analysis was performed in detail by varying the contributing factors weights and their effects on defined landslide susceptible locations(alternatives).

    In other part of the project, ordinary least square(OLS) technique was used to estimate weights of point parameters then its result compared with AHP technique result. Then, to verify the practicality of susceptibility maps, they were compared with a landslide inventory map. The outcome was that the 75% of occurred land failures fit into the very high and high susceptibility class of AHP map(using seven parameters), while this is 73.8% in the case of AHP with point parameters(using four parameters), and 65.8% for the OLS map. As conclusion, the AHP method yields reasonable results which make it reliable and credible approach in comparison with OLS, especially in the case of using large number of landslide contributing factors.


  • KEYWORD

    Landslide , Susceptibility , GIS(Geographic Information System) , AHP(Analytical Hierarchy Process) , OLS(Ordinary Least Square) , Sensitivity Analysis

  • Ⅰ. Introduction

    Landslide as a natural hazard is the main problem of many countries since it can be resulted in loss of lives and extensive property damages. Penang Island in Malaysia country is the area which frequently experiences heavy rainfall and subsequently much landslide harm. Thus it has been selected as suitable region to study landslide susceptibility(Lee and Pradhan, 2006). Beside the tropical rainfall as triggering factor, uncontrolled urbanization and deforestation play an effective role to aggravate slope destabilization in this island. So, it is required to forecast and specify the region where future land failure is likely to happen, in order to restrict construction activities on that zone. To facilitate this process, it is necessary to mapping the landslide prone areas(Althuwaynee et al., 2012). Actually, reliable and accurate landslide susceptibility map shows the zonal distribution of slopes prone to slip, thus, it can be helpful for land planners, decision makers, and also risk assessment.

    Geographic Information System(GIS) has been almost a compulsory tool in landslide hazard and risk assessment(Matori et al., 2011). Over the last few decades, many landslide susceptibility maps have been produced using different GIS-based methods including the analytical hierarchy process(AHP), frequency ratio, bivariate, multivariate, Logistics regression, fuzzy logic, and artificial neural network. Although, all techniques have advantages, there are two issues that influence the landslide susceptibility zonation(LSZ) mapping: 1) Using qualitative methods, the incomplete knowledge makes the expert decisions inaccurate or wrong, 2) Using quantitative approaches, the imprecise or inaccurate data have the similar impact(Vahidnia et al., 2010). Therefore, the results from the different mixture of qualitative and quantitative techniques that merge ranking and weighting may be more credible(Ayalew and Yamagishi, 2005). These mixture methods are considered as semi-quantitative approaches. The analytic hierarchy process (AHP) (Saaty, 1980), and analytic network process(ANP) (Saaty, 1999), weighted linear combination(WLC) (Ayalew et al., 2004), and fuzzy logic theory(Zadeh, 1965), are the examples of semi-quantitative techniques.

    In this research the AHP and Ordinary Least Square (OLS) approaches have been chosen for making landslide susceptibility mapping. AHP is popular and reliable method in landslide susceptibility mapping, especially when it is integrated with GIS which results in precise values for criteria. The ordinary least squares approach is considered as a basic prediction technique which is commonly used in various application fields. This method can be very quickly applied even to problems with hundreds of features and thousands of data points. However, linear models are not perfect since it is not reasonable to fit a simple line or planes to real world relationships. Although OLS has some weaknesses but it has been chosen for this research as a method to compare with AHP, because some distance factors have been discussed in this study which are suitable for OLS approach.

    Finally, it is noticeable that the quality and validity of landslide susceptibility zonation depend on the used methods and also the followed sampling strategies (Yilmaz, 2010). Thus, it is necessary to do comparative studies in order to find the best method for LSZ mapping(Tien Bui et al., 2012).

    In summery the results of this study suggest that the AHP method is the reliable method that can be used for landslide susceptibility mapping. And also, using more number of landslide contributing parameters increases the validation and reliability of susceptibility map.

    II. Study Area

    Penang Island is located in the west coast of Peninsular Malaysia(Figure 1). The Penang Island with the area of 293 Sq.km is the fourth-largest in the Malaysia country and the most populated as well. It is located almost between the latitude of 5°15´N to 5°30´N and longitudes 100°10´E to 100°21´E.

    The altitude of the Island changes from 0 to 817 meters above sea level and slope degree varies from 0° to 61.598°. The climate is tropical and the mean value of annual rainfall varies from about 2400 to 2700 mm (according to the meteorological records collected from 2006 to 2011). There are three main lithological formations in the study area which are alluvium, granite and micro-granite(Pradhan and Youssef, 2009). The terrain of the Island is mainly represented by coastal plains, hills and mountains with much developed lowland areas.

    III. Materials and Method

    Since 1970’s, many researchers have applied GISbased techniques to depict the spatial dispensation of landslide-prone areas. In general, all the models generated from all the methods rely on two assumptions: 1) future slope failures in a particular area will occur in the approximately same situation in which previous landslides have happened. 2) Casual factors that exist in GIS database in the form of spatial information can be utilized for evaluating the future landslides(Tangestani, 2009).

    The process of landsliding is relied on many causative factors such as geology, geomorphology, and hydrology. Meanwhile, causative factors which considered in landslide susceptibility mapping in Penang Island had been selected based on availability, relevance, and scales. Availability refers to related factors which are readily available, relevance to main factor of land failure happening in a particular area, and scale to existing maps of factors that can be used in making a slope stability map. Therefore, in this study seven causative factors have been verified in order to mapping the susceptible zones to land failure in Penang Island; Slope degree, distance to rivers, distance to roads, distance to faults, lithology, land use/land cover, and precipitation.

       1. landslide causative factors

    1) Slope

    Since, the slope degree is straightly related to landslide, it is frequently applied in providing landslide susceptibility maps(Dia et al., 2001; Cevik and Topal, 2003; Lee, 2005; Yalcin, 2008; Yalcin et al., 2011). Hence, the slope map of the study area was derived from digital elevation model(DEM) with a pixel size of 30 m by 30 m, and the terrain classified to five different zones according to the slope angle(Figure 2).

    2) Distance to Road

    Distance to roads is another parameter that has been considered in many researches because landslide may fall out on the road and on the slopes sides resulted from road constructing activities(Pachari et al., 1998; Ayalew and Yamagishi, 2005; Pourghasemi et al., 2012). In fact, constructing roads beside a slope results in some tensions, load decreasing on the slope heels, and landslide happening eventually. Therefore, the road proximity has been taken in account as a casual factor for landsliding.

    Road lines have been digitized from the hard copy of road network map that had been provided from Department of Survey and Mapping Malaysia(JUPEM), in the scale of 1:50,000. Then Straight Line or Euclidean Distance approach was applied to make different classes over the study area. The closest regions to the roads were classified to 5 zones based on the distance(Figure 3).

    3) Distance to Rivers

    Hydrological condition of the area or the saturation degree of the soil on slopes has been defined as a stability contributing parameter in landslide initiation in Penang Island. Water infiltration in soil, runoff, and groundwater flows were all happened under the hydrological circumstance(Thanh and De Smedt, 2011). Closeness of the slopes to river lines may affect its stability, because the proximity to rivers would activate the erosion process along the slope(Mancini et al., 2010). Therefore, the area close to rivers has been classified as high susceptible area to land failure. In this project five different zones have been buffered around river lines with 50 and 100 meters interval distance.

    The rivers map required for this part of the project had been provided from Department of Survey and Mapping Malaysia(JUPEM), in the scale of 1:50,000 (Figure 4).

    4) Lithology

    Different lithological specifications of an area lead to different strength and penetrance of rocks. Each particular lithological unit has its specific susceptibility value. For this reason, the soft copy of geological map in .pdf format and with the scale of 1:63,300 had been collected from Mineral and Geoscience Department of Malaysia(Figure 5). Then, it has been converted to IMAGINE Image format for further analysis in ArcGIS software. The lithology map of the study area was prepared including six lithological units; 1) Biotite micro granite, 2) Medium to coarse-grained granite, 3) Finegrained mega crystal biotite granite, 4) Fine-grained biotite-muscovite granite, 5) Medium to coarse-grained biotite-muscovite granite, and 6) Alluvium(marine deposits consist of sand, silt, and clay). Since the first 5 units generally known as granite type, the area was classified to two unit according to their grain size; Micro granite and Granite. So, the six units were compressed to following three lithological units; Alluvium, Micro granite, and Granite. In igneous rock areas(Granite areas), the probability of landslide occurrence is higher versus Alluvium. So, the sequence of the classes related to landslide susceptibility, was generated as bellow; 1) Alluvium, 2) Micro granite, 3) Granite. The region corresponding to class number 3 was found the high prone area to land failure(Figure 6).

    5) Distance to Faults

    The fractured rocks exist in faults form week zones or lines that present favorable conditions for landslides. It is accepted that the risk of land failure decreases in the further distance from tectonic structures. Landslides promotes in study area due to flow of water along fault planes and the erosion occurrence consequently. At first, the soft copy of geological map including fault lines in .pdf format(scale 1:63,300) and had been prepared from Mineral and Geoscience Department of Malaysia (Figure 7). Then, it has been converted to IMAGINE Image format in order to be used in ArcGIS software. In this study the distance from faults was calculated at 100 meter interval. The classified fault distance map has been extracted through Euclidean Distance method.

    6) Land use/ Land cover

    In this study the image of Penang Island downloaded from USGS website has been clipped by the boundary shapefile. Actually, the image firstly prepared in ENVI 4.7 software and then classified to four classes; water bodies, clear land, vegetation, and urban area. The Maximum Likelihood classification method has been used for this image preparation. Then classified image converted to shapefile and finally exported to ArcGIS software(Figure 8).

    7) Precipitation

    Precipitation is initiation factor in landslide occurrence due to its major influence on runoff and pore water pressure. Total rainfall, short-term intensity, antecedent precipitation, and storm duration are the main rainfall measurements considered as effective factor in landslide initiation. Choosing from these factors for subsequent analyzing usually depends on rainfall records availability. In this study the average monthly rainfall data from three existing weather stations, was collected for the past ten years (2003 to 2012).

    The stations located in northwest, northeast, and southeast of Island have recorded the annual average rainfall of 2608 mm, 2677 mm, and 2381 mm respectively. Then, an annual precipitation map of Penang Island was produced by Inverse Distance weighting(IDW) method over the three stations. This method is one of the spatial interpolation methods that are available in ArcGIS software spatial analyst tools.

       2. AHP analysis

    AHP is widely applied in regional planning, resource allocation, routing modeling, suitability analysis, site selection, and landslide susceptibility analysis(Ayalew et al., 2005). Actually, AHP uses pair-wise comparison to define weights or values for factors. Through pair-wise comparison each element compared to one another, and then the numerical weights are assigned to each factor due to their relative importance. Standard preference or ratio scale is 1-9 scale(Table 1).

    In this study the landslide causative parameters were put in sequence of importance. The relative importance of each parameter has been evaluated according to scientist researches especially for Penang Island(Lee and Pradhan, 2006; Pradhan et al., 2010; Oh and Pradhan, 2011). The following sequence indicates the priority of each parameter compare to others: Slope, Precipitation, Distance to roads, Distance to rivers, Land use/ Land cover, Lithology, Distance to faults. Then the preference matrix(Pair-wise comparison matrix) was created and AHP analysis was executed. Table 2 illustrates the weights calculated by AHP technique.

       3. Sensitivity Analysis

    Sensitivity analysis through simulation models helps researchers to specify the credibility of the result in cases where multi-criteria weights have been investigated(Chen et al., 2009; Abdullahi et al., 2013). It aims to identify the effects of changes in the inputs which are geographical data and the priorities on the outputs. If the changes do not result in significant effect on outputs, then the ranking is considered as robust and satisfactory. For this reason sensitivity analyses were performed to explore the response of the overall priority of landslide susceptible locations as alternatives to change in the relative synthesis value of each parameter. To perform sensitivity analysis three random points were chosen to act as alternatives. The overall susceptibility class of each point is shown in Table 3.

       4. Ordinary Least Square Technique

    Ordinary least square method is used in order to investigate the relation of current landslides and defined criteria, such as high slope regions, areas with high road density, area with high river density and so on. In fact the main aim of this method is to investigate the current situation. By using this method the effect of various criteria on existing landslides were found, and the coefficients for each landslide occurrence parameter is shown in a table generated by OLS function(Table 4). These coefficients calculated by OLS, are based on the average distance between each happened landslides to parameters. For this reason, the point layers were created for landslide contributing factors. So, the land use/ land cover, lithology, and rainfall were eliminated, because point layers resulted from these factors were not reliable. Thus, the four independent factors used in OLS approach were slope, distance to road, distance to river, and distance to fault. The inventory map which had been produced by scientists through interpretation of aerial photographs, reports and field surveys, has been used for digitizing the location points of land failures occurred in Penang Island. Then the OLS measurements between existing landslides to road network density and rivers, high slope regions, and regions on faults were produced. Calculated coefficients of each parameter has been normalized and converted to Satty scale in order to apply in ArcGIS software(Table 4).

       5. Validation

    Validation of landslide susceptibility maps was an absolutely essential component in this study. Without a validation, the maps are useless and have no value. For this reason, landslide inventory map which shows the previous landslide locations, were used to validate the landslide susceptibility maps. In order to verify the output maps, they were first converted to vector format then overlaid with the landslide inventory map by means of GIS software. From attribute table, the number of previous landslides that fell into each class of susceptibility was calculated easily. This procedure was done for three susceptibility maps generated in this research.

    IV. Result and Discussion

    In first part of the research, AHP method was applied using seven factors. Figure 10 shows the output map.

    Sensitivity analysis was performed by Expert choice software. Figure 11-A illustrates the numerical original condition of parameters and alternatives in Dynamic graph of Expert Choice software. Point 1 is 42.1% susceptible, point 2 is 30.7% and point 3 is 27.2% susceptible, with slope, rainfall, distance to road, distance to river, land use/cover, lithology, and distance to fault criteria priorities of 34.2%, 22.5%, 13.7%, 12%, 6.9%, 5.6%, and 5.1% respectively. As shown in figure 11-B, -C, -D, -E, -F, -G, and -H, the priority of each criterion is increased up to 50%; while susceptibility and priority of alternatives(locations) does not change significantly.

    In another part of the research, the weights calculated by OLS approach for four factors were used to make landslide susceptibility map. The map executed by OLS technique is shown in figure 12. In order to do practical comparison between AHP and OLS, a landslide susceptibility map has been made by AHP with the same factors used in OLS approach(Figure 13).

    According to the validation result summarized in table 5, many landslides that have occurred in the past fall into the very high and high susceptibility class. The outcome was that the 75% of occurred land failures fit into the very high and high susceptibility class of AHP map(using seven parameters), while this is 73.8% in the case of AHP with point parameters, and 65.8% for the OLS map. The weakness of OLS technique is that, only point parameters can be evaluated. The reason is that the evaluation of criteria by this technique is based on average distance of point parameters(as independent variable) to existing landslides(as dependent variable). This imperfection of OLS technique, decrease the validity of the outcome. However, the aim of performing this technique was to make challenge for AHP weighting technique with other approaches.

    V. Conclusion

    As conclusion the AHP method is the reliable method that can be used for landslide susceptibility mapping. And also, using more number of parameters contributing in landslide happening, increase the validation and reliability of susceptibility map.

    For comparing the results from OLS and AHP using same parameters(four parameters), AHP shows the better result in very high susceptible class(52%), but from low susceptible to moderate class, the number of landslide decreases from 21.8% to 9.6% which shows the invalid result. Actually this problem may be solved by considering more number of factors in landslide susceptibility analysis, like what has been done in the first part of the study applying AHP using seven factors.

    Sensitivity analysis was done to evaluate the overall priority of the parameters or objectives. Results of sensitivity analysis for first part of research were acceptable and showed that by changing the degree of preferences for each objective or factor, no significant changes has been occurred on alternatives.

  • 1. Abdullahi S., Rodzi A., Pradhan B. 2013 Spatial modelling of site suitability assessment for hospitals using geographical information system-based multicriteria approach at Qazvin city, Iran [Geocarto International] P.1-21 google doi
  • 2. Althuwaynee O. F., Pradhan B., Lee S. 2012 Application of an evidential belief function model in landslide susceptibility mapping [Computers & Geosciences] Vol.44 P.120-135 google doi
  • 3. Ayalew L., Yamagishi H. 2005 The application of GIS-based logistic regression for landslide susceptibility mapping in the Kakuda-Yahiko Mountains, Central Japan [Geomorphology] Vol.65 P.15-31 google doi
  • 4. Ayalew L., Yamagishi H., Marui H., Kanno T. 2005 Landslides in Sado Island of Japan: Part II. GIS-based susceptibility mapping with comparisons of results from two methods and verifications. [Engineering Geology] Vol.81 P.432-445 google doi
  • 5. 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] Vol.1 P.73-81 google doi
  • 6. Pachari AK., Gupta PV., Chander R. 1998 Landslide zoning in a part of the Vol.36 P.325-334 google
  • 7. Chen Y., Yu J., Shahbaz K., Xevi E. July, 2009 A GIS-based sensitivity analysis of multi-criteria weights. [18th World IMACS/MODISM …] P.3137-3143 google
  • 8. Cevi E., Topa T. 2003 GIS-based landslide susceptibility mapping for a problematic segment of the natural gas pipeline, Hendek (Turkey) [Environmental Geology] Vol.44 P.949-962 google doi
  • 9. Dia F. C., Lee C .F., Li J., Xu Z. W. 2001 Assessment of landslide susceptibility on the natural terrain of Lantau Island, Hong Kong [Environmental Geology] Vol.40 P.381-391 google doi
  • 10. Lee S., Pradhan B. 2006 Probabilistic landslide hazards and risk mapping on Penang Island, Malaysia [Journal of Earth System Science] Vol.115 P.661-672 google doi
  • 11. Lee S. 2005 Application of logistic regression model and its validation for landslide susceptibility mapping using GIS and remote sensing data [International Journal of Remote Sensing] Vol.26 P.1477-1479 google doi
  • 12. Mancini F., Ceppi C., Ritrovato G. 2010 GIS and statistical analysis for landslide susceptibility mapping in the Daunia area, Italy [Natural Hazards and Earth System Science] Vol.10 P.1851-1864 google doi
  • 13. Matori A. N., Basith A., Harahap I. S. H. 2011 Study of regional monsoonal effects on landslide hazard zonation in Cameron Highlands, Malaysia [Arabian Journal of Geosciences] Vol.5 P.1069-1084 google doi
  • 14. Oh H. J., Pradhan B. 2011 Application of a neuro-fuzzy model to landslide-susceptibility mapping for shallow landslides in a tropical hilly area [Computers & Geosciences] Vol.37 P.1264-1276 google doi
  • 15. Pourghasemi H. R., Pradhan B., Gokceoglu C. 2012 Application of fuzzy logic and analytical hierarchy process (AHP) to landslide susceptibility mapping at Haraz watershed, Iran [Natural Hazards] Vol.63 P.965-996 google doi
  • 16. Pradhan B., Lee S., Buchroithner M. F. 2010 A GIS-based back-propagation neural network model and its cross-application and validation for landslide susceptibility analyses, Computers [Environment and Urban Systems] Vol.34 P.216-235 google doi
  • 17. Pradhan B., Youssef A. M. 2009 Manifestation of remote sensing data and GIS on landslide hazard analysis using spatial-based statistical models [Arabian Journal of Geosciences] Vol.3 P.319-326 google doi
  • 18. Satty T. L. 1980 The Analytical Hierarchy Process google
  • 19. Saaty T. L. 1999 Fundamentals of the analytic network process. [In: Proceedings International Symposium on Analytical Hierarchy Process] google
  • 20. Tangestani M. H. 2009 A comparative study of Dempster-Shafer and fuzzy models for landslide susceptibility mapping using a GIS: An experience from Zagros Mountains, SW Iran [Journal of Asian Earth Sciences] Vol.35 P.66-73 google doi
  • 21. Thanh L. N., De Smedt F. 2011 Application of an analytical hierarchical process approach for landslide susceptibility mapping in A Luoi district, Thua Thien Hue Province, Vietnam [Environmental Earth Sciences] Vol.66 P.1739-1752 google doi
  • 22. Tien Bui D., Pradhan B., Lofman O., Revhaug I., Dick O. B. 2012 Landslide susceptibility mapping at Hoa Binh province (Vietnam) using an adaptive neuro-fuzzy inference system and GIS [Computers & Geosciences] Vol.45 P.199-211 google doi
  • 23. Vahidnia M. H., Alesheikh A. a., Alimohammadi A., Hosseinali F. 2010 A GIS-based neuro-fuzzy procedure for integrating knowledge and data in landslide susceptibility mapping [Computers & Geosciences] Vol.36 P.1101-1114 google doi
  • 24. Yalcin a., Reis S., Aydinoglu a. C., Yomralioglu T. 2011 A GIS-based comparative study of frequency ratio, analytical hierarchy process, bivariate statistics and logistics regression methods for landslide susceptibility mapping in Trabzon, NE Turkey. [Catena] Vol.85 P.274-287 google doi
  • 25. Yalcin A. 2008 GIS-based landslide susceptibility mapping using analytical hierarchy process and bivariate statistics in Ardesen (Turkey): Comparisons of results and confirmations [Catena] Vol.72 P.1-12 google doi
  • 26. Yilmaz I. 2010 The effect of the sampling strategies on the landslide susceptibility mapping by conditional probability and artificial neural networks [Environmental Earth Sciences] Vol.60 P.505-519 google doi
  • 27. Zadeh L. A. 1965 Fuzzy Sets [Information and Control] Vol.8 P.338-353 google doi
  • [Figure 1.] Malaysia, Penang Island Map
    Malaysia, Penang Island Map
  • [Figure 2.] Penang Island Slope Map
    Penang Island Slope Map
  • [Figure 3.] Road Distance Map
    Road Distance Map
  • [Figure 4.] River Distance Map
    River Distance Map
  • [Figure 5.] Geological Map of Penang Island
    Geological Map of Penang Island
  • [Figure 6.] Classified Lithological Map of Penang Island
    Classified Lithological Map of Penang Island
  • [Figure 7.] Fault Distance Map of Penang Island
    Fault Distance Map of Penang Island
  • [Figure 8.] Land use/cover Map of Penang Island
    Land use/cover Map of Penang Island
  • [Figure 9.] Rainfall Map of Penang Island
    Rainfall Map of Penang Island
  • [Table 1.] Rating scale
    Rating scale
  • [Table 2.] Aggregate pair-wise comparison
    Aggregate pair-wise comparison
  • [Table 3.] Susceptibility of locations (Alternatives)
    Susceptibility of locations (Alternatives)
  • [Table 4.] Ordinary Least Square calculation
    Ordinary Least Square calculation
  • [Figure 10.] Landslide Susceptibility Map generated by AHP method
    Landslide Susceptibility Map generated by AHP method
  • [Figure 11.] Numerical assessment of sensitivity analysis(Dynamic graph), A) Present condition
    Numerical assessment of sensitivity analysis(Dynamic graph), A) Present condition
  • [Figure 12.] Landslide Susceptibility Map generated by OLS scales
    Landslide Susceptibility Map generated by OLS scales
  • [Figure 13.] Landslide Susceptibility Map resulted from AHP using four factors
    Landslide Susceptibility Map resulted from AHP using four factors
  • [Table 5.] Comparison of different susceptibility map
    Comparison of different susceptibility map