Landslide Susceptibility Analysis in Samanlı Mountains Massif by Frequency Ratio and Artificial Neural Networks Methods
Susceptibility analyses of landslides, which cause loss of life and property, have a high frequency of occurrence and are affected by many factors, play an important role in the prediction of possible landslides with the help of Geographic Information Systems (GIS). In this study, the landslide susc...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Artvin Coruh University
2025-07-01
|
| Series: | Doğal Afetler ve Çevre Dergisi |
| Subjects: | |
| Online Access: | http://dacd.artvin.edu.tr/tr/download/article-file/4716599 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Susceptibility analyses of landslides, which cause loss of life and property, have a high frequency of occurrence and are affected by many factors, play an important role in the prediction of possible landslides with the help of Geographic Information Systems (GIS). In this study, the landslide susceptibility analysis of the Samanlı Mountains massif in the eastern part of the Marmara Sea was created using Frequency Ratio (FR) and Artificial Neural Networks (ANN) methods. Firstly, landslide inventory was produced by using information obtained from different landslide databases, satellite images and field studies. Then, landslide susceptibility analysis of the site was carried out with FR and ANN methods using geology, slope, relative relief, topographic moisture index (TWI), distance to faults, distance to streams, Normalised Difference Vegetation Index (NDVI), topographic roughness index (TRI), rainfall and distance to roads parameters. Samanlı Mountains landslide susceptibility results were produced with 5 levels in both methods. Receiver operating characteristic (ROC) was used for accuracy analysis of the models. According to the Frequency Ratio method, 38% of the study area has very low-low landslide susceptibility, 43% has medium landslide susceptibility and 19% has high-high landslide susceptibility. According to ANN model, 27% of the study area has very low-low landslide susceptibility, 45% has medium landslide susceptibility, and 27% has high-high landslide susceptibility. In both models, the areas with high landslide susceptibility are concentrated in the basins between Çınarcık-Yalova-Karamürsel, Gölcük-Karamürsel coastal belt and İznik-Pamukova-Geyve-Arifiye line. The main parameters that trigger landslides in the study area are sedimentary lithological units, proximity to faults and roads and slope values. In the study, according to the ROC accuracy test, ANN was found to be more successful than FO among the landslide susceptibility models within the scope of the Samanlı Mountains mass, which is the sample area.
|
|---|---|
| ISSN: | 2528-9640 |