Heyelan Duyarlılık Haritalarının Üretilmesinde Örneklem ve Doğrulama Stratejilerinin Değerlendirilmesi (Gelibolu Yarımadası'nın Doğu Kesimi)

Abstract

The purposes of this study are to evaluate the sampling strategies used in landslide susceptibility mapping studies in the literature and to assess validation indices. In addition, validation indices of the models created by using seed cell sampling strategy and whole landslide mass (displacement + accumulation area) sample strategy have been analysed and compared. In line with this purpose, the east of Canakkale Gallipoli Peninsula has been selected as the study area. The research in the scope of this study has been conducted in 6 phases. These are; 1) mapping of the landslides and compilation of inventory map of the study area by using the field studies, the aerial photos and the Google Earth satellite images; 2) evaluation of conditioning factors to be employed in landslide susceptibility analyses; 3) development of landslide susceptibility models using seed cell samples established by buffer areas defined with different d distances and samples from displacement and accumulation areas; 4) mapping of landslide susceptibility for both models constructed; 5) calculation and comparisons of validation indices of the models; 6) the modification of distance used to define the buffer area suggested by the seed cell approach. Validation indices of both models have been analyzed under three groups. Calculated using the train data set on dependent threshold value (confusion matrix and Kappa index) and independent threshold value (ROC curve and AUC) and calculated the test data set on other validation indices (RMSE).