Bilge International Journal of Science and Technology Research
Yazarlar: Senem TEKİN, Tolga ÇAN
Konular:Yerbilimleri, Ortak Disiplinler
DOI:10.30516/bilgesci.525438
Anahtar Kelimeler:Ermenek watershed,Landslide,Artificial neural network,Landslide susceptibility
Özet: Ermenek River is one of the major tributary of the Göksu river with a watershed area of 4020 km2 .The lateral and vertical transitional Miocene clastic and carbonate units are widely exposed in the area along the deeply incised valleys with elevation range of more than 1000 m. Deep-seated slides on the valley sides and the rock fall events along the steep slopes on the edge of the reefal limestone platforms are abundant in the area. In this study susceptibility assessments for slide type landslides were evaluated using artificial neural network method. 302 landslides covering an area of 161 km2 were identified in the study area. Geology, digital elevation model, slope, roughness index, tangential, plan and profile curvatures, topographic wetness index, mean slope, surface-relief ratio were used for the landslide preparatory factors during the susceptibility assessments. The data base used in susceptibility models were randomly separated into three of which, 70 % for the analysis and 15 % for the test and validation sets. Landslide susceptibility map has been classified into five susceptibility classes from very low to very high. The validation of the landslide susceptibility map was evaluated by the prediction-success rate and receiver operator characteristics curve. As a consequence, it has seen that the produced susceptibility map has high prediction capacity where 77 % of the substantial landslides were located in the high and very high susceptiblity classes corresponding 29 % of the study area with the area under the receiver operator characterisctic curve of 0.893.
Dergi editörleri editör girişini kullanarak sisteme giriş yapabilirler. Editör girişi için tıklayınız.