A comparative study for landslide susceptibility mapping using GIS-based multi-criteria decision analysis (MCDA), logistic regression (LR) and association rule mining (ARM)

ERENER A., MUTLU A., Duzgun H. S.

ENGINEERING GEOLOGY, vol.203, pp.45-55, 2016 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 203
  • Publication Date: 2016
  • Doi Number: 10.1016/j.enggeo.2015.09.007
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Page Numbers: pp.45-55
  • Kocaeli University Affiliated: Yes


Landslide susceptibility mapping is one of the crucial stages of landslide hazard and risk assessment. Moreover, the susceptibility maps assist planners, local administrations, and decision makers in disaster planning. Various approaches have been applied in the literature to increase the accuracy of landslide susceptibility maps. The determination of suitable susceptibility mapping method plays critical role for obtaining required accuracy. In this study, the performances of three quantitative susceptibility mapping methods are evaluated. The logistic regression (LR) analysis is the typical example of statistical methods, while GIS-based multi-criteria decision analyses (MCDA) and association rule mining (ARM) are the examples of heuristic and data mining methods, respectively. The susceptibility maps based on the three methods are obtained for Savsat in Artvin province (NE Turkey) where the region has intense landslides. The landslide influencing parameters for the study area are lithology, land use/land cover, soil type, erosion, altitude, slope, aspect, distance to drainage, soil depth, distance to fault, distance to road. The models for the three methods are then compared and evaluated by using pixel-based evaluation metrics. Results showed that ARM provides a higher quality percent (QP), whereas percent of damage detection (PDD) is higher for LR method. The lowest QP is obtained by GIS-based MCDA. It is found that LR and ARM methods are better than GIS-based MCDA in modeling the landslide susceptibility and they can be integrated to obtain better performance. (C) 2015 Elsevier B.V. All rights reserved.