Logisnet: A tool for multimethod, multiple soil layers slope stability analysis

https://doi.org/10.1016/j.cageo.2008.04.003Get rights and content

Abstract

Shallow landslides and slope failures have been studied from several points of view (inventory, heuristic, statistic, and deterministic). In particular, numerous methods embedded in Geographic Information Systems (GIS) applications have been developed to assess slope stability. However, little work has been done on the systematic comparison of different techniques and the incorporation of vertical contrasts of geotechnical properties in multiple soil layers. In this research, stability is modeled by using LOGISNET, an acronym for Multiple Logistic Regression, Geographic Information System, and Neural Network. The main purpose of LOGISNET is to provide government planners and decision makers a tool to assess landslide susceptibility. The system is fully operational for models handling an enhanced cartographic–hydrologic model (SINMAP) and multiple logistic regression. The enhanced implementation of SINMAP was tested at regional scale in the Highway 101 corridor in Del Norte County, California, and its susceptibility map was found to have improved factor of safety estimates based on comparison with landslide inventory maps. The enhanced SINMAP and multiple logistic regression subsystems have functions that allow the user to include vertical variation in geotechnical properties through summation of forces in specific soil layers acting on failure planes for a local or regional-scale mapping. The working group of LOGISNET foresees the development of an integrated tool system to handle and support the prognostic studies of slope instability, and communicate the results to the public through maps.

Introduction

Triggered by extrinsic factors (such as earthquakes or intense rainfall) and intrinsic factors (such as geology, slope, vegetation, and geotechnical parameters), landslides can cause significant damage and potentially generate destructive debris flows (Ohlmacher and Davis, 2003; Dai and Lee, 2002; Atkinson and Massari, 1998). In spite of the efforts made by local authorities and scientists to monitor and forecast landslides, it has been difficult to accurately evaluate landslide potential or susceptibility due to large spatial and temporal variability. Several inventory, heuristic, statistical, and deterministic approaches to assess landslide susceptibility have been proposed (Castellanos Abella and Van Westen, 2008; Demoulin and Chung, 2007; Metternicht et al., 2005; Lee et al., 2004; Zhou et al., 2003; Bozzano et al., 2002; Qin et al., 2002a, Qin et al., 2002b; Guzzetti et al., 1999; Parise, 2001; Maceo-Giovanni et al., 2000; Lang et al., 1999; González-Díez et al., 1999; Di Gregorio et al., 1999; Miles and Ho, 1999; Pasuto and Soldati, 1999; Dehn and Buma, 1999; Atkinson and Massari, 1998; Jäger and Wieczorek, 1994) and their automatization and application are embedded in landslide GIS programs (Claessens et al., 2007; D’Ambrosio et al., 2007; Bursik et al., 2007; Xie et al., 2006; Lee et al., 2004; Zhou et al., 2003; Hammond et al., 1992; Pack et al., 1998; Montgomery and Dietrich, 1994); however, few studies take full account of topographic control through shallow subsurface water flow in landslide generation (Pack et al., 2001; Montgomery and Dietrich, 1994). Based on a hydrologic model developed by O’Loughlin (1986), Montgomery and Dietrich (1994) developed a model that combines topographic and hydrographic variables to predict potential landslide zones with sparse information. Using a similar approach, Pack et al. (1998) developed a cartographic/hydrologic model (Stability Index Mapping: SINMAP) in which poorly constrained parameters are incorporated through the use of uniform probability distributions. However, little work has been done to clarify the importance of geotechnical properties of material in different soil layers that produce landslides (Chigira et al., 2003; Chigira, 2001; Simon et al., 2000; Voight and Elsworth, 1997). Evaluation of landslides with different soil layers was done by Simon et al. (2000). Simon et al. (2000) developed a 2-D streambank failure model that allows for a slip plane crossing multiple soil layers with differing friction angle and cohesion. The model worked well in post-diction of failure events along Goodwin Creek, MS, USA.

Also, in spite of notable contributions to evaluate landslide models (Haneberg, 2005; Chinnayakanahalli et al., 2003; Borga et al., 2002; Morrissey et al., 2001; Guzzetti et al., 1999), the systematic comparison of models is lacking to outline advantages and limitations of the methods due to the difficulties of assessing a model in natural conditions at different cartographic scales, DEM resolutions, and sampling strategies.

As an attempt to develop a 3-D model allowing for soil layering, SINMAP, Multiple Logistic Regression (MLR), and Neural Network (NN) approaches are being modified to accept geotechnical parameters for multiple soil layers in order to assess landslide potential. A proposed computer system uses Geographic Information System (GIS) resources. The system, called LOGISNET is developed by using Arc Macro Language (AML) under ArcInfo GIS software. LOGISNET is used to host, compare, and visualize results after using the three approaches. This paper describes development, implementation, and testing of the first two approaches (SINMAP and MLR) as methods for delineating landslide potential. Pilot tests are performed using LOGISNET as a GIS application in geohazard for an area on Highway 101 in the northern Coast Ranges of California.

Section snippets

Background and physical basis

Two approaches for landslide susceptibility are used: (1) SINMAP, which expresses the stability of the slope in terms of a factor of safety, a ratio between the forces that prevent the slope from failing and those that make the slope fail and (2) MLR, which has the advantage over other multivariate statistical techniques for this application in that the dependent variable can have only two values—an event occurring or not occurring (Ohlmacher and Davis, 2003). The SINMAP and MLR approaches are

Pilot tests

The study area is near the Highway 101 corridor in Del Norte County, California (Fig. 5). In the area, landslides along the coast between Wilson Creek and Crescent City create a potentially hazardous situation for people and property. More than 200 landslides have been mapped in this area by the California Geological Survey and the California Department of Transportation. The study area is prone to landslides due to the combination of several factors, such as high precipitation (2583.94 mm/y),

Results

A comparison of outputs by using SINMAP with default system parameters, SINMAP with field data parameters, LOGISNET multiple soil layers, and MLR is done. Fig. 6 shows the results at 30 m pixel resolution. At this resolution, SINMAP, LOGISNET multiple soil layers, and MLR approaches appear to reflect topographic conditions more than landslide geomorphology. It seems the models just detect steep terrain rather than the landslide process.

The three SINMAP outputs (Fig. 6b–d) show high instability

Conclusions

We introduce the implementation of LOGISNET as a user interface and collection of functions in AML designed to facilitate the analysis and validation of cartographic–hydrologic (SINMAP) and MLR landslide models with multiple soil layers data. The set of functions provides a toolbox that allows evaluation and calculation for prediction of landslides. The tools and methods to assess the model quality at different cartographic scales and DEM resolutions are still under development and depend on

Acknowledgments

The authors wish to express their deepest appreciation to authorities from the Redwood National and State Parks, California for their approval and help in conducting the field sampling for this research. The inventory and elevation data used in this study was drawn from Dave Best (GIS Coordinator, Redwood National Park) and from C. J. Wills (Department Of Conservation California Geological Survey). This work benefited greatly from the comments and help of Vicky Ozaki, Brian R. Merrill, Andrew

References (53)

  • M. Dehn et al.

    Modelling future landslide activity based on general circulation models

    Geomorphology

    (1999)
  • A. Demoulin et al.

    Mapping landslide susceptibility from small datasets: a case study in the Pays de Herve (E Belgium)

    Geomorphology

    (2007)
  • S. Di Gregorio et al.

    Mount Ontake landslide simulation by the cellular automata model SCIDDICA-3

    Physics and Chemistry of the Earth, Part A: Solid Earth and Geodesy

    (1999)
  • M.J. García-Rodríguez et al.

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

    Geomorphology

    (2008)
  • A. González-Díez et al.

    A methodological approach for the analysis of the temporal occurrence and triggering factors of landslides

    Geomorphology

    (1999)
  • R. Greco et al.

    Logistic regression analysis in the evaluation of mass movements susceptibility: the Aspromonte case study, Calabria, Italy

    Engineering Geology

    (2007)
  • F. Guzzetti et al.

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

    Geomorphology

    (1999)
  • A. Lang et al.

    Classic and new dating methods for assessing the temporal occurrence of mass movements

    Geomorphology

    (1999)
  • S. Lee et al.

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

    Engineering Geology

    (2004)
  • G. Metternicht et al.

    Remote sensing of landslides: an analysis of the potential contribution to geo-spatial systems for hazard assessment in mountainous environments

    Remote Sensing of Environment

    (2005)
  • S.B. Miles et al.

    Rigorous landslide hazard zonation using Newmark's method and stochastic ground motion simulation

    Soil Dynamics and Earthquake Engineering

    (1999)
  • G.C. Ohlmacher et al.

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

    Engineering Geology

    (2003)
  • M. Parise

    Landslide mapping techniques and their use in the assessment of the landslide hazard

    Physics and Chemistry of the Earth, Part C: Solar, Terrestrial and Planetary Science

    (2001)
  • A. Pasuto et al.

    The use of landslide units in geomorphological mapping: an example in the Italian Dolomites

    Geomorphology

    (1999)
  • S. Qin et al.

    A nonlinear dynamical model of landslide evolution

    Geomorphology

    (2002)
  • A. Simon et al.

    Bank and near-bank processes in an incised channel

    Geomorphology

    (2000)
  • Cited by (27)

    • Infinite slope stability model and steady-state hydrology-based shallow landslide susceptibility evaluations: The Guneysu catchment area (Rize, Turkey)

      2021, Catena
      Citation Excerpt :

      Nevertheless, even though they are the most suitable methods in shallow landslide susceptibility evaluation, physically-based model approaches rarely have been investigated in Turkey before (Yilmaz and Keskin, 2009; Akgun and Erkan, 2016). Additionally, the number of studies conducted in the World, most probably due to data limitations, are also limited (Paulin and Bursik, 2009; Bueechi et al., 2019; Kim et al., 2019). Different physically-based models have different levels of complexity, and the results depend on the quality and detail of the input data (Zizioli et al., 2013).

    • Application of data-driven evidential belief functions to landslide susceptibility mapping in Jinbu, Korea

      2013, Catena
      Citation Excerpt :

      Audisio et al., 2009; Dahal et al., 2008; Lee and Min, 2001; Lee and Pradhan, 2006; Lee et al., 2004a; Mousavi et al., 2009; Oh et al., 2009; Ozdemir, 2009; Pirasteh et al., 2009; Poudyal et al., 2010; Pradhan and Youssef, 2009; Regmi et al., 2010; Vahidnia et al., 2009; Yalcin, 2008; Yalcin et al., 2011; Yilmaz, 2009c,d, 2010). One of the statistical models available, the logistic regression model, has also been applied to landslide susceptibility mapping (Akgun and Bulut, 2009; Akgun et al., 2012; Bai et al., 2010, 2011; Chauhan et al., 2010; Das et al., 2010; Dominguez-Cuesta et al., 2010; Falaschi et al., 2009; Ghosh et al., 2011; Lee, 2005, 2007a; Paulin and Bursik, 2009; Mancini et al., 2010; Mathew et al., 2009; Nandi and Shakoor, 2010; Oh and Lee, 2010; Pradhan, 2010; Pradhan and Lee, 2010a,c; Pradhan et al., 2011). More sophisticated assessments involve applications of fuzzy logic, artificial neural network, analytical hierarchy process, Dempster–Shafer theory of evidence, and Monte Carlo analysis (Akgun and Türk, 2010; Akgun et al., 2011; Chen et al., 2009a,b; Choi et al., 2010; Ghosh and Carranza, 2010; He and Fu, 2009; Kanungo et al., 2008, 2009; Kawabata and Bandibas, 2009; Lee, 2007b; Lee and Evangelista, 2006; Lee et al., 2004b, 2006; Liu et al., 2009; Melchiorre et al., 2008; Miles and Keefer, 2009; Muthu et al., 2008; Oh and Pradhan, 2011; Park, 2011; Poudyal et al., 2010; Prabu and Ramakrishnan, 2009; Pradhan, 2011a,b; Pradhan and Lee, 2007, 2009, 2010b; Pradhan and Pirasteh, 2010; Pradhan et al., 2010b,c; Sezer et al., 2011; Shafri et al., 2010; Tangestani, 2009; Vahidnia et al., 2009; Wang et al., 2009).

    • Autologistic modelling of susceptibility to landsliding in the Central Apennines, Italy

      2011, Geomorphology
      Citation Excerpt :

      Nefeslioglu et al. (2008) compared logistic regression with an artificial neural network (ANN) and found that the ANN over-predicted whereas the logistic regression model under-predicted susceptibility. Paulin and Bursik (2009) combined the logistic regression model with an ANN in a GIS framework. It is not the intention here to promote one method over another.

    View all citing articles on Scopus
    View full text