Medium scale earthflow susceptibility modelling by remote sensing and geographical information systems based multivariate statistics approach: an example from Northeastern Turkey


DAĞ S., AKGÜN A., KAYA A. , ALEMDAĞ S., BOSTANCI H. T.

ENVIRONMENTAL EARTH SCIENCES, vol.79, no.19, 2020 (Journal Indexed in SCI) identifier identifier

  • Publication Type: Article / Article
  • Volume: 79 Issue: 19
  • Publication Date: 2020
  • Doi Number: 10.1007/s12665-020-09217-7
  • Title of Journal : ENVIRONMENTAL EARTH SCIENCES
  • Keywords: Earthflow, Susceptibility, Logistic regression, Remote sensing, Turkey, SHALLOW LANDSLIDE SUSCEPTIBILITY, LIKELIHOOD-FREQUENCY RATIO, LOGISTIC-REGRESSION, GIS TECHNOLOGY, MULTICRITERIA DECISION, HIERARCHY PROCESS, NEURAL-NETWORKS, AREA, HAZARD, NORTH

Abstract

The aim of the present study was to produce an earthflow susceptibility map for the city center and environs of the province of Rize located at the Northeastern part of Turkey. The study area is the rainiest region of Turkey, and due to the triggering effect of precipitation earthflows are frequently observed in and around the study area. Besides this point, weathered rock units and steep topography accompany with precipitation for the occurrence of earthflow cases. Considering this point, an earthflow susceptibility mapping was inferred to be a necessity for the area. Parameters, such as lithology, slope gradient, slope aspect, topographical wetness index (TWI), stream power index (SPI), slope curvature, and sediment capacity index (LS) were considered to be earthflow conditioning parameters. A multi-temporal earthflow inventory map for a period of 4 years (2011-2015) was initially generated by way of remote sensing approach and field surveys. Earthflow susceptibility map was produced using the logistic regression method after which the produced susceptibility map was validated. The mapped earthflows were separated into two groups prior to modelling and validation. The first group was for training and the second group was for validation steps. The accuracy of the model was measured by fitting them to a validation set of mapped earthflows. Area under curvature (AUC) approach was applied for validation purposes. The prediction capability of the earthflow susceptibility map produced can be regarded as acceptable in accordance with the AUC values of 0.62 for the logistic regression model. Based on these results, the obtained earthflow susceptibility map can be used to mitigate hazards related to landslides and to aid in land-use planning for the Rize city center.