Elsevier

Geomorphology

Volume 95, Issues 3–4, 15 March 2008, Pages 172-191
Geomorphology

Susceptibility assessment of earthquake-triggered landslides in El Salvador using logistic regression

https://doi.org/10.1016/j.geomorph.2007.06.001Get rights and content

Abstract

This work has evaluated the probability of earthquake-triggered landslide occurrence in the whole of El Salvador, with a Geographic Information System (GIS) and a logistic regression model. Slope gradient, elevation, aspect, mean annual precipitation, lithology, land use, and terrain roughness are the predictor variables used to determine the dependent variable of occurrence or non-occurrence of landslides within an individual grid cell. The results illustrate the importance of terrain roughness and soil type as key factors within the model — using only these two variables the analysis returned a significance level of 89.4%. The results obtained from the model within the GIS were then used to produce a map of relative landslide susceptibility.

Introduction

An earthquake is a major natural process of high destructive potential, often resulting in both human and material losses as the direct consequence of the seismic phenomenon. However, some processes derived from an earthquake such as liquefaction, landslides, and tsunamis can often be more dangerous than the initial earthquake. The 2004 Southeast Asian tsunami and the 2001 landslides in El Salvador represent good examples. One of the earliest known studies on earthquake-induced landslide hazards was conducted by Keefer (1984), who analysed the types and magnitude of mass movements in tectonically active regions.

Landslides are significant natural hazards in many areas of the world. Each year they cause more than a 100 000 deaths and injuries, with damage costing more than a 1billion USD (Schuster, 1996). In many countries, the economic losses and casualties due to landslides are greater than commonly recognized, and landslides generate a yearly loss of property larger than that from any other natural disaster including earthquakes, floods and windstorms. Generally, landslides are triggered by seismicity or heavy rains. Other possible causes are anthropogenic, including deforestation, road cutting, and mining. The study of earthquake-induced landslides plays an important role in determining seismic risk, as earthquakes and landslides can result in considerable damage to infrastructure, in addition to a massive loss of life (Marzorati et al., 2002). In January and February of 2001, El Salvador experienced several destructive earthquakes, which caused hundreds of landslides of various sizes. In this study, we have used a logistic regression model to assess the susceptibility of earthquake-induced landslides for the whole country of El Salvador.

Two factors are important when modelling any natural phenomena in experimental science: data quality and the choice of scientific models. When data are incomplete or inaccurate, natural phenomena are usually analysed intuitively with ad hoc methods (e.g., Anbalagan, 1992, Anbalagan and Singh, 1996). When studying landslides in small geographical areas, methods most often used are GPS measurements, photogrammetry, or detailed field surveys (Casson et al., 2003, Agnesi et al., 2005); however, for larger geographical areas such as an entire country, methods usually used include remote sensing and thematic cartography. Some studies have used satellite imagery as a substitute for large- to medium-scale aerial photography of landslides (Nichol and Wong, 2005, Nichol et al., 2006). The scale of the model depends on the purpose of the investigation and the specifications of the user. Hazard assessment of earthquake-triggered landslides may be developed at different scales or detail levels, ranging from site-specific evaluation to regional studies (Bommer and Rodríguez, 2002). The framework for our study is classified as Grade 2 with a scale of 1:10 000–1:100 000 (ISSMGE, 1999). Our aim is to produce an earthquake-triggered landslide susceptibility map for the entire country of El Salvador, which requires certain data approximations and generalisations. The available data include topographical maps, geological maps (1:100 000), digital cartography (1:25 000), landslide inventories, and the rainfall database. The data were provided by the Servicio Nacional de Estudios Territoriales de El Salvador (SNET), and the Universidad Centroamericana Simeón Cañas (UCA), whose databases are well-documented and useful for both landslide hazard evaluation and model definition.

Section snippets

Summary of previous studies

A variety of approaches have been used in mapping slope instability, and they can be classified into qualitative and quantitative methods. Most of qualitative methods tend to be subjective, since they depend on expert opinions and portray hazard levels in descriptive terms (Anbalagan, 1992). Quantitative methods are based on the numerical expression of the relationship between instability factors and landslides, which can be divided into deterministic and statistical. Deterministic methods

Description of the study area

El Salvador is one of the smallest but most densely populated countries in Central America, with an area of just over 20 000km2. The country is located on the Pacific coast and bordered by Guatemala to the west and Honduras to the north and east (Fig. 1), and is affected by earthquakes from two main sources of seismicity. The largest shocks are generated in the Benioff-Wadati zones of the subducted Cocos plate, which is converging with the Caribbean plate in the Middle America Trench at an

Data sources

A Geographical Information System (GIS) database with different layers or coverages was compiled. The seven landslide-influencing parameters studied were: lithology (bedrock and soil), elevation, slope gradient, slope aspect, terrain roughness, mean annual precipitation, and land use.

Landslide density analysis

The landslides inventory used consists of data on slope movement from the 2001 El Salvador earthquakes, compiled by the SNET. Description and classification of landslides was mainly based on the system of Cruden and Varnes (1996), which takes into consideration the type of movement, materials involved, and the state or activity of unstable slopes. This study separates debris flow from other types of mass movements such as rock falls and avalanches because of their significant differences, and

Logistic regression

In order to choose an appropriate statistical analytical technique for landslides investigation, we should take into account the categorical characteristics of independent variables. The logistic regression has the advantage of being less affected when the basic assumption of normality of the variables is not met (Hair et al., 1998). Other techniques used to solve this problem include neural networks (Lee et al., 2004, Gómez and Kavzoglu, 2005, Ermini et al., 2005).

Logistic regression is

Concluding remarks

In many other landslide studies using logistic regression, elevation and slope angle were the best predictor variables or factors for estimating the probability of landslide occurrences (Ohlmacher and Davis, 2003, Ayalew and Yamagishi, 2005). However, we found that terrain roughness and lithology are the best factors to estimate landslide susceptibility, while land use is the least significant factor. Maximum frequency of landslides occurs in the slope gradient range 73°–81°, with elevations

Acknowledgements

This research was developed within the framework of the projects financed by the Ministerio de Educación y Ciencia (CGL2005-07456-C03-03/BTE) and UPM-Relaciones con Latinoamérica (2006) of Spain. Cartographic data were provided by the Ministerio de Ambiente y Recursos Naturales (MARN) and Servicio Nacional de Estudios Territoriales (SNET) of El Salvador. The authors wish to express their sincere thanks for these contributions, in particular Carlos Pullinger and Giovanni Molina. The authors

References (67)

  • ErminiL. et al.

    Artificial Neural Networks applied to landslide susceptibility assessment

    Geomorphology

    (2005)
  • GómezH. et al.

    Assessment of shallow landslide susceptibility using artificial neural networks in Jabonosa River Basin, Venezuela

    Engineering Geology

    (2005)
  • GuzzettiF. et al.

    Landslide hazard evaluation: a review of current techniques and their application in a multi-scale study, Central Italy

    Geomorphology

    (1999)
  • LeeS. et al.

    Determination and application of the weights for landslide susceptibility mapping using an artificial neural network

    Engineering Geology

    (2004)
  • MarzoratiS. et al.

    Rock falls induced by earthquakes: a statistical approach

    Soil Dynamics and Earthquake Engineering

    (2002)
  • NicholJ.E. et al.

    Application of high-resolution stereo satellite images to detailed landslide hazard assessment

    Geomorphology

    (2006)
  • OhlmacherG.C. et al.

    Using multiple logistic regression and GIS technology to predict landslide hazard in northeast Kansas, USA

    Engineering Geology

    (2003)
  • ReficeA. et al.

    Probabilistic modeling of uncertainties in earthquake induced landslide hazard assessment

    Computers & Geosciences

    (2002)
  • RowbothamD.N. et al.

    GIS Modelling of regional slope stability in Phewa Tal Watershed, Nepal

    Geomorphology

    (1998)
  • Van Den EeckhautM. et al.

    Prediction of landslide susceptibility using rare events logistic regression: a case-study in the Flemish Ardennes (Belgium)

    Geomorphology

    (2006)
  • YesilnacarE. et al.

    Landslide susceptibility mapping: a comparison of logistic regression and neural networks methods in a medium scale study, Hendek region (Turkey)

    Engineering Geology

    (2005)
  • ZhouG. et al.

    Spatial probabilistic modeling of slope failure using an integrated GIS Monte Carlo simulation approach

    Engineering Geology

    (2003)
  • AshfordS.A. et al.

    Topographic effects on the seismic response of steep slopes

    Bulletin of the Seismological Society of America

    (1997)
  • AtakanK. et al.

    Local site response in San Salvador, El Salvador, based on the October 10, 1986 earthquake

    Reduction of Disasters in Central America Report No. 6

    (1994)
  • AtakanK. et al.

    Local site effects on microtremors, weak and strong ground motion in San Salvador, El Salvador

  • BandL.E.

    Spatial aggregation of complex terrain

    Geographical Analysis

    (1989)
  • BenitoB. et al.

    Analysis of the spatial and temporal distribution of the 2001 earthquakes in El Salvador

  • BommerJ.J. et al.

    Propiedades mecánicas de las Tierras Blancas y las inestabilidades de Taludes

    Revista ASIA

    (1998)
  • BommerJ.J. et al.

    Geotechnical properties and seismic slope stability of volcanic soils

  • BrenningA.

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

    Natural Hazards and Earth System Sciences

    (2005)
  • CarraraA. et al.

    GIS techniques and statistical models in evaluating landslide hazard

    Earth Surface Processes and Landform

    (1991)
  • CrudenD. et al.

    Landslide types and processes

  • DaiF.C. et al.

    A spatiotemporal probabilistic modelling of storm-induced shallow landsliding using aerial photographs and logistic regression

    Earth Surface Processes and Landforms

    (2003)
  • Cited by (178)

    View all citing articles on Scopus
    View full text