Comparison and evaluation of landslide susceptibility maps obtained from weight of evidence, logistic regression, and artificial neural network models

The main purpose of this study is to compare the performance of two statistical analysis models like weight of evidence and logistic regression (LR) with a soft computing model like artificial neural networks for landslide susceptibility assessment. These models were applied for the Selinous River d...

Full description

Saved in:
Bibliographic Details
Published inNatural hazards (Dordrecht) Vol. 93; no. 1; pp. 249 - 274
Main Authors Polykretis, Christos, Chalkias, Christos
Format Journal Article
LanguageEnglish
Published Dordrecht Springer Netherlands 01.08.2018
Springer Nature B.V
Subjects
Artificial neural networks
Civil Engineering
Distance
Drainage
Drainage basins
Drainage network
Earth and Environmental Science
Earth Sciences
Environmental Management
Evaluation
Geophysics/Geodesy
Geotechnical Engineering & Applied Earth Sciences
Hydrogeology
Imagery
Land cover
Landslides
Lithology
Mathematical models
Model accuracy
Model testing
Natural Hazards
Neural networks
Original Paper
Regression analysis
River basins
Rivers
Satellite imagery
Satellites
Soft computing
Spaceborne remote sensing
Statistical analysis
Statistical methods
Statistical models
Surveys
Weight
Greece
Logistic regression
Model comparison
Landslide susceptibility
Weight of evidence
Artificial neural networks
Online AccessGet full text

Cover

More Information
Summary:The main purpose of this study is to compare the performance of two statistical analysis models like weight of evidence and logistic regression (LR) with a soft computing model like artificial neural networks for landslide susceptibility assessment. These models were applied for the Selinous River drainage basin (northern Peloponnese, Greece) in order to map landslide susceptibility and rate the importance of landslide causal factors. A landslide inventory was prepared using satellite imagery interpretation and field surveys. Eight causal factors including altitude, slope angle, slope aspect, distance to road network, distance to drainage network, distance to tectonic elements, land cover, and lithology were considered. Model performance was tested with receiver operator characteristic analysis. The validation findings revealed that the three models show promising results since they give good accuracy values. However, the LR model proved to be relatively superior in estimating landslide susceptibility throughout the study area.
ISSN:0921-030X
1573-0840
DOI:10.1007/s11069-018-3299-7