Development of Landslide Susceptibility Maps Using GIS and Remote Sensing
Geographic Information Systems and Science (GEO 565) Spring 2016,
Oregon State University
By: Nasim Adami
In this study, the identification of landslide hazard was performed by using GIS technology to create multiple datasets / layers and analyzing this dataset through discriminant analysis. Carrara et al. conducted their study on a small basin in Central Italy, which has Mediterranean climate. The data acquisition was performed through a series of existing topographic maps, aerial photographs, and field surveys. The final dataset included 40 morphological, geological, and vegetation attributes for each slope-unit. This dataset was analyzed by stepwise discriminant analysis, using several scenarios, including different number of variables (attributes). Finally, they found that including about 15 important variables, as representations of a physical characteristic of domain, was sufficient to produce acceptable results, although they mentioned that providing more detailed historical datasets could improve the overall accuracy. One of the most interesting aspects of this study is not including some significant parameters such as slope gradient, height and length of the area, due to the specific characteristics of the study domain.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
One of the most tedious and time/cost consuming procedures in developing a landslide hazard map is the data assembly, including geotechnical, groundwater, etc. Barredo et al. in this study implemented and compared two different approaches for assessment of landslide hazard in an area with relatively small measured database, due to the recurrences of large volume volcanic remain landslides. Two approaches were; a direct method that determines the degree and type of hazard by including detailed geomorphological mapping for individual uniquely coded polygons, and indirect method by indexing approach. The parameters that they included in their study were slope angle, landslide activity, landslide phases, material, proximity to reservoirs, discharge channels, and land use change. The indirect method was found to be too generalized for some locations, due to assigning the weight values to the locations with similar factors. They concluded that although the direct method is more time-consuming, but the detailed results enables the individual evaluation of locations. Also, the indirect method found to be suitable for the areas with limited availability of robust sampling strategies, despite the fact that this method is highly sensitive to the allocation of the parameter weighting values.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Dhakal, Amod S., Takaaki Amada, and Masamu Aniya. "Landslide hazard mapping and its evaluation using GIS: an investigation of sampling schemes for a grid-cell based quantitative method." Photogrammetric Engineering and Remote Sensing 66.8 (2000): 981-989.
Determination of influence and importance of each class and each factor is one of the most complicated and sensitive stages of hazard map development and it can dramatically alter the results in some cases. To evaluate the influence of each class and factor through a proper sampling procedure, Dhakal et al. created a landslide activity map by examination of aerial photos and field examination. The factors that they included in their study were topographic factors derived from DEM, geology, and land use. They did their analysis using a quantification scaling type II (QS II) scheme and examined multiple combinations of factors and classes to determine the best sampling technique and the influence of each factor on each class, which later they used to develop hazard maps and evaluation of their findings. They concluded that the most important factor for landslide hazard is geology and also suggested that unaligned stratified random sampling for selecting the non-landslide group could create acceptable results.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Pradhan, Biswajeet, Saro Lee, and Manfred F. Buchroithner. "A GIS-based back-propagation neural network model and its cross-application and validation for landslide susceptibility analyses." Computers, Environment and Urban Systems 34.3 (2010): 216-235.
In this study Pradhan et al. investigated the implementation of artificial neural network method in calculation of the weights for parameters, along with assessment of cross-application of these weights between the different study areas. The database created by authors included factors as slope, aspect, curvature, altitude, stream power index, topographic wetness index, topography, distance from drainage, distance from the road, land cover, soil texture and types, geology and distance from lineament, and the differentiated normalized vegetation index. After training and validation of their models, they deduced that the weights derived from the same study area resulted in more accurate landslide susceptibility maps than the case where they used the weights derived from cross-applied areas.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Floris, Mario, et al. "Using online databases for landslide susceptibility assessment: an example from the Veneto Region (northeastern Italy)."Natural Hazards and Earth System Science 11.7 (2011): 1915-1925.
The quality and availability of data is one of the most important factor in generating landslide susceptibility maps. These data sets are usually used in determination of activity of landslides and derivation of the influential factors. Floris et al. investigated the development of landslide susceptibility maps by using the available online databases. They started by reviewing WebGIS portals around the world and evaluating the existing DEMs and historic landslide events. Later, they derived the influential parameters of elevation, slope, curvature, profile, plan curvature, land use, distance to roads and distance to rivers, and performed their analysis using simple probabilistic approach. After validation of their results, they concluded that, although with the existing database development of landslide susceptibility map and its spatial distribution was possible, but, improvement was needed on expanding landslide databases infrastructure, especially on activity aspect. Also they suggested that other models and methods, such as heuristic, deterministic and statistical, could be used for achieving better results.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Sujatha, Evangelin Ramani, G. Victor Rajamanickam, and P. Kumaravel. "Landslide susceptibility analysis using Probabilistic Certainty Factor Approach: A case study on Tevankarai stream watershed, India." Journal of earth system science 121.5 (2012): 1337-1350.
Fast urbanization and high frequency heavy rainfalls are some of the most significant parameters in triggering landslides. Sujatha et al. studied the accuracy and applicability of Probabilistic Certainty Factor method in such regions. In their study, they assumed relief, slope, aspect, curvature, weathering, soil, land use, proximity to road and proximity to drainage as influential parameters. Their study resulted in several conclusions, most important of which was that, the certainty factor method was found to be an appropriate tool in assessment of landslide susceptibility in such regions and useful in determination of relationships between the individual parameters. They also deduced that the most influential factors in such regions were slope, aspect, soil and proximity to roads, and also due to importance of roads, they suggested additional local slope stability analysis for main roads.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Xu, Chong, et al. "The 2010 Yushu earthquake triggered landslide hazard mapping using GIS and weight of evidence modeling." Environmental Earth Sciences 66.6 (2012): 1603-1616.
Earthquakes are one of the most important triggers for landslides. After China’s earthquake in 2010, a total number of 2036 landslides were identified through high-resolution aerial photographs and multi-source satellite images by Xu et al.. Then, they generated landslide susceptibility map by using weight of evidence modeling, which uses Bayesian probability modeling applied to landslide hazard mapping, and assumed 12 influential parameters as elevation, slope gradient, slope aspect, slope curvature, topographic position, distance from main surface ruptures, peak ground acceleration, distance from roads, normalized difference vegetation index, distance from drainage, lithology, and distance from all faults. After validation of their results, they found that weight of evidence model can reproduce the historic landslides with high accuracy and also it can provide acceptable predictions for the “future” events, which were assumed to occur under same conditions.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Yilmaz, Cagatay, Tamer Topal, and Mehmet Lütfi Süzen. "GIS-based landslide susceptibility mapping using bivariate statistical analysis in Devrek (Zonguldak-Turkey)." Environmental earth sciences 65.7 (2012): 2161-2178.
In this study, Yilmaz et al. investigated the role of the seed cell selection procedure on the quality of the resulting landslide susceptibility maps. During the generation of data layers, they used three different criteria to identify the seed cell from different parts of landslide event, namely crown and flanks, crown only, and flank only. They included a total number of 10 parameters in their study as: elevation, slope, aspect, profile curvature, plan curvature, distance to streams, drainage density, distance to ridges, distance to road, power line network, and lithology. They used 26 historic landslide database for training and validation of their model, which was generated using the statistical index method. After comparison and validation of the generated landslide susceptibility map, they concluded that all three seed cells generated acceptable landslide susceptibility maps and specifically, the seed cell including the data from crowns only, presented the highest accuracy among all. Also, they found that the most influential parameters in this study were elevation, lithology, slope, aspect and drainage density.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Liu, Chun, et al. "Susceptibility evaluation and mapping of China’s landslides based on multi-source data." Natural hazards 69.3 (2013): 1477-1495.
Changes in climate and rainfall can significantly change the triggering process of landslides and eventually result in larger number of landslides in a region. To address the need of an accurate and large scale landslide susceptibility map for such regions, Liu et al. investigated the applicability and accuracy of developing an empirical model using back-propagation artificial neural network method. They collected and analyzed an amazing database including historic landslide events from past 60 years in China. The parameters that they included in development of their empirical model were lithology, convexity, slope gradient, slope aspect, elevation, soil property, vegetation coverage, flow, and fracture. They claimed that back-propagation artificial neural network method has the ability to merge nonlinear elements by qualitative and quantitative indices. Finally, after validating their results, they concluded that their empirical model can provide accurate and reliable results, in addition to its applicability to create large scale models.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Pardeshi, Sudhakar D., Sumant E. Autade, and Suchitra S. Pardeshi. "Landslide hazard assessment: recent trends and techniques." SpringerPlus2.1 (2013): 523.
Usually when several methods are developed in one field to address the same problem, the decision on which method is more accurate or efficient for a certain region is inevitable. To answer this question and also create a guide line for future researchers in landslide susceptibility area, Pardeshi et al. studied several methods and comparisons between their results in the recent years published articles. The methods they covered in this study were heuristic, semi quantitative, quantitative, probabilistic and multicriteria approaches. They found that the multivariate methods can generally provide more reliable and accurate results.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Pourghasemi, Hamid Reza, et al. "Landslide susceptibility mapping using support vector machine and GIS at the Golestan Province, Iran." Journal of earth system science 122.2 (2013): 349-369.
Pourghasemi et al. investigated the accuracy and efficiency of landslide susceptibility map development by using Support Vector Machine with several kernel classifiers. The kernel classifiers used in this study were linear, polynomial degree of 2, polynomial degree of 3, polynomial degree of 4, radial basis function, and sigmoid. The support vector machine analysis with different kernel classifiers was performed in ENVI software. The factors included in this study were slope degree, slope aspect, altitude, plan curvature, profile curvature, tangential curvature, surface area ratio, lithology, land use, distance from faults, distance from rivers, distance from roads, topographic wetness index, and stream power index. They created a dataset of 81 historic landslide locations by using aerial photography and field survey, which later was used in training of the model and validation of the results. After validation of the results they concluded that, although radial basis function and polynomial degree of 3 models performed slightly better, but there are no significant difference or improvement in using different types of kernel classifiers. Also, they argued that the uniqueness of the results, obtained from support vector machine, could be another advantage of using this method.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Chalkias, Christos, Maria Ferentinou, and Christos Polykretis. "GIS-Based Landslide Susceptibility Mapping on the Peloponnese Peninsula, Greece."Geosciences 4.3 (2014): 176-190.
In this study Chalkias et al. used a bivariate statistical analysis (in comparison to multivariate statistical analysis) to develop landslide susceptibility maps on a regional scale. They incorporated a total number of 7 factors as: elevation, slope, aspect, lithology, land cover, mean annual precipitation, and peak ground acceleration. After validation of their results, they concluded that this method could produce more reliable results in a regional scale than previously done analysis by semi-quantitative (expert-based fuzzy weighting—EFW). They also found the most important factors to be land cover, elevation and mean annual precipitation.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Chandio, Imtiaz Ahmed, et al. "GIS-basedland suitability analysis of sustainable hillside development." Procedia Engineering 77 (2014): 87-94.
One of the applications of landslide susceptibility models is to create slope stability maps. Chandio et al. investigated the use of landslide susceptibility model of analytical hierarchical process in determination of suitability of hillside fields for development. In their approach, they used Expert Choice software to determine the ranking of the weights of the factors. Finally they validated their results though sensitivity analysis for proving the robustness of the results. Based on their validation, authors concluded that this method can be used in preparing the land use and land development plannings.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Feizizadeh, Bakhtiar, et al. "A GIS-based extended fuzzy multi-criteria evaluation for landslide susceptibility mapping." Computers & geosciences73 (2014): 208-221.
Feizizadeh et al. tried to improve the accuracy and reliability of Analytical Hierarchical Process by approaching it from a decision making point of view. They included the knowledge of the local landslide experts on the causal factors and further complemented their study by using fuzzy membership functions. In this research, fuzzy membership functions were used both in ranking the predisposing factors and also development of landslide susceptibility map. Finally, from the validation and comparison of results with historic landslide database, they concluded that this coupling between fuzzy membership functions and Analytical Hierarchical Process could significantly improve the results.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Feizizadeh, Bakhtiar, Piotr Jankowski, and Thomas Blaschke. "A GIS based spatially-explicit sensitivity and uncertainty analysis approach for multi-criteria decision analysis." Computers & geosciences 64 (2014): 81-95.
Any data driven model has an inheriting uncertainties from the original data and also it contains certain sensitivity to the implemented database. To address this problem in generation of landslide susceptibility maps, Feizizadeh et al. investigated the uncertainty and sensitivity of two major models of Analytical Hierarchical Process and Ordered Weighted Averaging. They evaluated the uncertainties and sensitivity of the models as a function of weights using Monte Carlo Simulation and Global Sensitivity Analysis, and for validation of their results they used Dempster–Shafer Theory. After validation and comparison of uncertainties they founded that, although Analytical Hierarchical Process provided more accurate results than Ordered Weighted Averaging, but Ordered Weighted Averaging had lower uncertainties in generated landslide susceptibility maps.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Yusof, Norbazlan M., and Biswajeet Pradhan. "Landslide susceptibility mapping along PLUS expressways in Malaysia using probabilistic based model in GIS." IOP Conference Series: Earth and Environmental Science. Vol. 20. No. 1. IOP Publishing, 2014.
Yusof et al. evaluated the validity and accuracy of the frequency ratio model on development of landslide susceptibility maps for two types of critical slopes, soil and rock slopes. They also increased their available database by including the data obtained from remote sensing satellites such as IFSAR (interoferometric synthetic aperture radar). Eventually, they concluded that frequency ratio model is an appropriate model to create landslide susceptibility maps for both types of slopes and also they found the important factors to be slope angle, geomorphology, soil type and precipitation.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Zhu, A-Xing, et al. "An expert knowledge-based approach to landslide susceptibility mapping using GIS and fuzzy logic." Geomorphology 214 (2014): 128-138.
Zhu et al. claimed that data driven multivariate statistical models for defining the relationship between the influential factors and generating landslide susceptibility maps have some shortcomings, most important of which is their dependency on the training data quality. To overcome this flaw, they proposed an expert knowledge-based approach, in which they defined the relationship between the predisposing factors not by mathematical approaches, but by expert’s knowledge on the matter. Then by using the predisposing factors calculated from existing data/surveys, they used a fuzzy landslide susceptibility inference to predict landslide susceptibility. After the validation of the generated landslide susceptibility maps, the authors concluded that their proposed method was able to generate reliable maps and moreover, by collecting knowledge from local experts of any study area, their method could be portable to any desired region.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Meten, Matebie, Netra PrakashBhandary, and Ryuichi Yatabe. "Effect of Landslide Factor Combinations on the Prediction Accuracy of Landslide Susceptibility Maps in the Blue Nile Gorge of Central Ethiopia."Geoenvironmental Disasters 2.1 (2015): 1-17.
Meten et al. addressed two main issues in their study. First they investigated the independence of influential factors in landslide by using the correlation between them and determined that this correlation between factors was not present or it was ignorable. Second, they conducted a set of investigations to determine the best combination of factors in generating landslide susceptibility maps by using a frequency ratio model. They included a total number of 8 factors as: slope, aspect, profile curvature, plan curvature, lithology, land use, distance from lineament, and distance from river. Eventually, by using mathematical combination theory, they concluded that the most important factors in generating landslide susceptibility maps are slope and lithology, and an optimum map can be produced by using only four factors of slope, lithology, land use and distance from lineament.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
One of the major and difficult steps in development of a landslide susceptibility map is collecting the appropriate database. This is most difficult in regions with high vegetation and cloudy weather, such as tropical environment. Shahabi et al. investigated a region with such environment by using remote sensing data to identify the occurrence of landslides. The radar data enables the DEM derivation even with cloudy weather. By using this method, an inventory of 92 landslides were created and implementation of remote sensing satellites were proven to be beneficial. Total of 10 parameters were included in analysis of this study as: slope, aspect, soil, lithology, NDVI, land cover, precipitation, distance to fault, distance to drainage, and distance to road. Three models were used in analysis of the data and generation of landslide susceptibility map, which were analytical hierarchy process, weighted linear combination, and spatial multi-criteria evaluation. After validation of the landslide susceptibility maps, they concluded that the spatial multi-criteria evaluation model was the most accurate model for use in such regions. Also all three models identified the precipitation as the most important factor in occurrence of landslide in this study area.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Wang, Qiqing, et al. "GIS-based assessment of landslide susceptibility using certainty factor and index of entropy models for the Qianyang County of Baoji city, China." Journal of Earth System Science 124.7 (2015): 1399-1415.
In some parts of the world, landslide claims more lives and causes more property damages than other natural disasters. In such regions, the reliability and accuracy of the landslide susceptibility models are of high importance. Wang et al. assessed two approaches of landslide susceptibility calculations through certainty factor and index of entropy models. They collected 81 historic landslide events and used 70% of them for model training and model parameters calculation and the rest for validation. They included a total number of 15 parameters in their analysis as: slope, angle, slope aspect, general curvature, plan curvature, profile curvature, altitude, distance to faults, distance to rivers, distance to roads, the sediment transport index, the stream power index, the topographic wetness index, geomorphology, lithology, and rainfall. After comparison and validation of their results with historic landslide events, they concluded that both models can create reliable results and can be used in land use planning and hazard mitigation.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Zhou, Suhua, et al. "GIS-Based Integration of Subjective and Objective Weighting Methods for Regional Landslides Susceptibility Mapping."Sustainability 8.4 (2016): 334.
Currently, the fast grow in population makes the urbanization procedure more rapid than ever before. The use of the new lands call for an adequate method to both include the recent human land use effects and generate landslide susceptible maps. To this end, Zhou et al. proposed a new approach in developing the landslide susceptibility maps. In this new approach they combined the subjective weight, which is analytical hierarchy process in this study and was used to weight predictive factor’s contribution to landslide occurrence, and objective weight, which is the frequency ratio in this study and was used to calculate sub-class frequency ratio with respect to landslides within each predictive factor, to develop a more accurate landslide map. The predictive factors that they include in their method were: elevation, slope, aspect, terrain roughness index, lithology, land cover and mean annual precipitation. They used a database of 534 historic landslides in the study region to develop the predictive factors and the corresponding sub-classes. After validation of the landslide susceptibility maps developed by the proposed method, they concluded that their method can present more accurate results than those generated solely by analytical hierarchy process or frequency ratio.
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Further articles that can provide more details on this topic :
