Spatio-temporal prediction and mapping of landslides using MTInSAR-learning logistic regression and weight of evidence modeling in urban environments: a case study for the Büyükçekmece–Küçükçekmece region, İstanbul, Türkiye

The megacity İstanbul (Türkiye) is rapidly urbanizing with a high population growth. Similar to this, the region of İstanbul Metropole between the Büyükçekmece and Küçükçekmece lakes is being built up to a large extent. Landslides in this region, however, cause major socio-economic damage due to geological, geomorphological, tectonic, climatic, groundwater, and anthropological factors. The aim of the study is to carry out a spatio-temporal susceptibility modeling of landslides using the multi-temporal InSAR (MTInSAR)-learned logistic regression (LR) and weight of evidence (WofE) methods. First, spatial and temporal change information in landslide motions was learned using MTInSAR analysis of ALOS Palsar-1 SAR pictures for this purpose. Then, LR and WofE susceptibility modeling were performed using the previously learned information. The accuracy, model fit, and prediction power of susceptibility modeling and assessment findings were compared and evaluated using the receiver operating characteristic (ROC) curves and the area under the ROC curve (AUC), statistical metrics, and spatial percentage proportions. Furthermore, the effects of geospatial predictor variables on the occurrence of landslides were investigated. The obtained findings demonstrated that the applied MTInSAR-based susceptibility modeling approach has a good forecast capability of landslide probability. Landslide susceptibility maps have a high degree of goodness-of-fit at the MTInSAR-learned training dataset and landslide-based validation dataset. The AUC, statistical metrics, and spatial percentage proportions have demonstrated the usefulness and effectiveness of the suggested technique in urban contexts and landslide-prone locations.