Elsevier

Computers & Geosciences

Volume 45, August 2012, Pages 199-211
Computers & Geosciences

Landslide susceptibility mapping at Hoa Binh province (Vietnam) using an adaptive neuro-fuzzy inference system and GIS

https://doi.org/10.1016/j.cageo.2011.10.031Get rights and content

Abstract

The objective of this study is to investigate a potential application of the Adaptive Neuro-Fuzzy Inference System (ANFIS) and the Geographic Information System (GIS) as a relatively new approach for landslide susceptibility mapping in the Hoa Binh province of Vietnam. Firstly, a landslide inventory map with a total of 118 landslide locations was constructed from various sources. Then the landslide inventory was randomly split into a testing dataset 70% (82 landslide locations) for training the models and the remaining 30% (36 landslides locations) was used for validation purpose. Ten landslide conditioning factors such as slope, aspect, curvature, lithology, land use, soil type, rainfall, distance to roads, distance to rivers, and distance to faults were considered in the analysis. The hybrid learning algorithm and six different membership functions (Gaussmf, Gauss2mf, Gbellmf, Sigmf, Dsigmf, Psigmf) were applied to generate the landslide susceptibility maps. The validation dataset, which was not considered in the ANFIS modeling process, was used to validate the landslide susceptibility maps using the prediction rate method. The validation results showed that the area under the curve (AUC) for six ANFIS models vary from 0.739 to 0.848. It indicates that the prediction capability depends on the membership functions used in the ANFIS. The models with Sigmf (0.848) and Gaussmf (0.825) have shown the highest prediction capability. The results of this study show that landslide susceptibility mapping in the Hoa Binh province of Vietnam using the ANFIS approach is viable. As far as the performance of the ANFIS approach is concerned, the results appeared to be quite satisfactory, the zones determined on the map being zones of relative susceptibility.

Highlights

► Landslide susceptibility mapping was carried out using ANFIS. ► ANFIS models with the Sigmf and Gaussmf have a good prediction capability. ► ANFIS is robust and efficient method for landslide susceptibility mapping.

Introduction

In recent years, the occurrences of natural disasters in Vietnam have increased significantly mostly due to the effect of the climate change. The northwest mountainous area of Vietnam is one of the regions, which is heavily affected by landslide activities and flooding events. In most of the cases, landslides occurred following heavy rainfalls and especially during tropical rainstorms. However, to date, little effort has been made to assess or forecast these events. Through scientific analysis of landslides, geoscientists and civil engineers can assess and delineate landslide-susceptible areas, offering the potential for a decrease in landslide damage through proper slope management (Pradhan, 2011). So research on landslide susceptibility mapping for the Northwest area, including the Hoa Binh province, is an urgent task in Vietnam. The result of landslide research may provide valuable information that helps to forecast such events as well as to find measures to mitigate subsequent losses to future landslides.

Many techniques and methods have been proposed and developed in the landslide literature. Basically, they can be divided into direct and indirect methods (Guzzetti et al., 1999). Direct methods can be the methods of using total stations, global satellite navigation systems, or other methods for landslide surveying (Van Westen et al., 2003). Through direct methods, using experiences and knowledge of experts, the degree of susceptibility can be directly determined. However, in such cases fieldworks are extensively required, so they are time-consuming and not cost effective. Whereas in the case of the indirect methods, usually a landslide inventory map is used in conjunction with the landslide conditioning factors (Van Westen et al., 2003). Generally, the landslide susceptibility levels are determined based on the correlation between the existing landslide inventory and the conditioning factors as mentioned above. However, the accuracy of landslide susceptibility maps depends on the level of accuracy of mapping of different conditioning factors and landslide inventory.

In recent years, some newer approaches for landslide analysis have been carried out such as artificial neural networks (Ercanoglu, 2005, Ermini et al., 2005, Lee et al., 2003, Lee et al., 2004, Nefeslioglu et al., 2008, Pradhan and Lee, 2010a, Pradhan and Lee, 2010c, Pradhan et al., 2010a), fuzzy logic (Akgun et al., 2011, Ercanoglu and Gokceoglu, 2002, Ercanoglu and Gokceoglu, 2004, Kanungo et al., 2008, Pradhan, 2010a), decision tree (Nefeslioglu et al., 2010), and neuro-fuzzy (Kanungo et al., 2006, Oh and Pradhan, 2011, Pradhan et al., 2010b, Sezer et al., 2011, Vahidnia et al., 2010). Generally, these approaches give rise to qualitative and quantitative maps of the landslide hazard areas, and the spatial results are appealing (Pradhan, 2010b).

In the case of the neuro-fuzzy, which is a combination of fuzzy logic and neural networks, Kanungo et al. (2006) used the weights of the layers obtained from the trained neural network integrated with the ratings obtained from fuzzy logic to obtain a landslide susceptibility index. The membership degrees of each layer class were determined based on the relationship of existing landslides with the classes. Vahidnia et al. (2010) used the output of a fuzzy inference system as target for a neural network. There is no doubt that expert knowledge played an important role to obtain the accuracy of these results. And the subjectivity is not easy to eliminate.

Another combined method, developed by Jang (1993), is the Adaptive Neuro-Fuzzy Inference System (ANFIS). This method, using the Takagi–Sugeno rule format, is the combination of an optimized premise membership function (gradient descent) with an optimized consequent equation (linear least squares estimator). Based on a given input and target, ANFIS can construct a fuzzy inference system where their membership function parameters are to be adjusted using the hybrid learning algorithm. ANFIS has been widely used in modeling complex systems (Soyguder and Alli, 2009). However, its application in landslide studies is still limited to a very few cases. Pradhan et al. (2010b) used eight landslide conditioning factors for the ANFIS model for landslide susceptibility mapping in a study area in Malaysia. Their results indicated a very high prediction accuracy of 97%. Oh and Pradhan (2011) and Sezer et al. (2011) have also used an ANFIS based model for landslide susceptibility mapping in different parts of Malaysia. However, the disadvantage of these approaches is that it is difficult to objectively determine the epoch where the landslide model starts over-fitting in the training phase. Expert opinions are suggested used to determine the number of membership functions of the inputs, the physical meanings of the inputs, and the number of training epochs for preventing over-learning. Oh and Pradhan (2011) also asserted that, in order to check the performance of the neuron-fuzzy model in landslide susceptibility mapping, and for the method to be more generally applied, more case studies should be conducted.

The main difference between this study and the aforementioned literature is that the ANFIS was applied for landslide susceptibility assessment in the Hoa Binh province of Vietnam, using the subtractive clustering method proposed by Chiu (1994). This is the automated data-driven based method for constructing the primary fuzzy models. The main advantages of this method are that it can process a large number of input observations and avoid the explosion of the rule base (Eftekhari and Katebi, 2008). In addition, the method proposed by Jang et al. (1997) was used to control over-fitting of the landslide model by testing the fuzzy inference system (FIS), trained on the training data against the checking data. Using the six different membership functions, six ANFIS models were constructed. Finally, the comparison was carried out to find the most suitable model for the study area.

Section snippets

Study area

The Hoa Binh province is situated at the monsoonal region in the North West part of Vietnam. It covers an area of about 4660 km2 (Fig. 1) between the longitude 104°48′E and 105°50′E, and between the latitude 20°17′N and 21°08′N. The altitude decreases from Northwest to Southeast and varies in the range from 0 to 1510 m. This province is a “transition” area situated between the Northwest Mountains and the Red River delta.

The climate of this region is characterized by high temperatures and high

Data used

Since landslide occurrences in the past and present are keys to future spatial prediction (Guzzetti et al., 1999), a landslide inventory map is a pre-requisite for such a study. The landslide inventory map of the study area was compiled by inheriting the landslide locations from three projects: (1) the landslide inventory map in the Northern mountainous province (Hue et al., 2004); (2) the landslide inventory map 2005 (Thinh et al., 2005); and (3) the landslide inventory map 2007 (My, 2007).

Fuzzy inference system

Fuzzy inference is the process of using fuzzy logic to formulate a non-linear mapping from input data to an output. A fuzzy inference system (FIS) consists of three components: (1) a rule base, which contains a selection of fuzzy rules, (2) a data base, which defines the membership functions used in the fuzzy rules, and (3) a reasoning mechanism that performs the inference procedure on the rules and given facts (Alakhras, 2005, Jang et al., 1997, Kumanan et al., 2008).

There are three types of

Validation and comparison of landslide susceptibility maps

Using the success rate and the prediction rate methods, the validation process was performed by comparing the existing landslide data with the six landslide susceptibility maps. (Brenning, 2005, Chung and Fabbri, 1999, Chung and Fabbri, 2003, Lee et al., 2007). The success rate results were obtained by comparing the landslide training pixels (684 landslide pixels) with the six landslide susceptibility maps. Fig. 8 shows the success rate curves for the six ANFIS landslide models. The model with

Concluding remarks

In this study, an adaptive neuro-fuzzy inference system (ANFIS) was applied to assess the regional landslide susceptibility for the Hoa Binh province of Vietnam. The landslide inventory, comprising all significant landslide events that occurred during the last ten years, was used in the analysis. Among the total of 118 landslide areas, 82 cases with 864 landslide grid cells were selected for the ANFIS training process and the remaining 36 cases with 315 landslide grid cells were used for the

Acknowledgment

This research was funded by the Norwegian Quota scholarship program. The data analysis and write-up were carried out as a part of the first author's Ph.D. studies at the Geomatics section, Department of Mathematical Sciences and Technology, Norwegian University of Life Sciences, Norway. Thanks to the anonymous reviewers for their valuable and constructive comments, which were useful to revise the manuscript.

References (58)

  • H.A. Nefeslioglu et al.

    An assessment on the use of logistic regression and artificial neural networks with different sampling strategies for the preparation of landslide susceptibility maps

    Engineering Geology

    (2008)
  • H.-J. Oh et al.

    Application of a neuro-fuzzy model to landslide-susceptibility mapping for shallow landslides in a tropical hilly area

    Computers & Geosciences

    (2011)
  • B. Pradhan

    Remote sensing and GIS-based landslide hazard analysis and cross-validation using multivariate logistic regression model on three test areas in Malaysia

    Advances in Space Research

    (2010)
  • B. Pradhan et al.

    Landslide susceptibility assessment and factor effect analysis: backpropagation artificial neural networks and their comparison with frequency ratio and bivariate logistic regression modelling

    Environmental Modelling & Software

    (2010)
  • B. Pradhan et al.

    A GIS-based back-propagation neural network model and its cross-application and validation for landslide susceptibility analyses

    Computers Environment and Urban Systems

    (2010)
  • A. Sengur

    Wavelet transform and adaptive neuro-fuzzy inference system for color texture classification

    Expert Systems with Applications

    (2008)
  • E.A. Sezer et al.

    Manifestation of an adaptive neuro-fuzzy model on landslide susceptibility mapping: Klang valley, Malaysia

    Expert Systems with Applications

    (2011)
  • S. Soyguder et al.

    An expert system for the humidity and temperature control in HVAC systems using ANFIS and optimization with fuzzy modeling approach

    Energy and Buildings

    (2009)
  • O. Taylan et al.

    An adaptive neuro-fuzzy model for prediction of student's academic performance

    Computers & Industrial Engineering

    (2009)
  • M.H. Vahidnia et al.

    A GIS-based neuro-fuzzy procedure for integrating knowledge and data in landslide susceptibility mapping

    Computers & Geosciences

    (2010)
  • I. Yilmaz et al.

    Multiple regression, ANN (RBF, MLP) and ANFIS models for prediction of swell potential of clayey soils

    Expert Systems with Applications

    (2011)
  • L.C. Ying et al.

    Using adaptive network based fuzzy inference system to forecast regional electricity loads

    Energy Conversion and Management

    (2008)
  • M. Abul Hasanat et al.

    Bayesian belief network learning algorithms for modeling contextual relationships in natural imagery: a comparative study

    Artificial Intelligence Review

    (2010)
  • A. Akgun et al.

    An easy-to-use MATLAB program (MamLand) for the assessment of landslide susceptibility using a Mamdani fuzzy algorithm

    Computers & Geosciences

    (2011)
  • M.N.Y. Alakhras

    Neural network-based fuzzy inference system for exchange rate prediction

    Journal of Computer Science

    (2005)
  • A. Brenning

    Spatial prediction models for landslide hazards: review, comparison and evaluation

    Natural Hazards and Earth System Sciences

    (2005)
  • D. Bui et al.

    Landslide susceptibility analysis in the Hoa Binh province of Vietnam using statistical index and logistic regression

    Natural Hazards

    (2011)
  • D. Caniani et al.

    Neural networks and landslide susceptibility: a case study of the urban area of Potenza

    Natural Hazards

    (2007)
  • S.L. Chiu

    Fuzzy model identification based on cluster estimation

    Journal of Intelligent and Fuzzy Systems

    (1994)
  • Cited by (344)

    View all citing articles on Scopus
    View full text