Application of analytic hierarchy process, frequency ratio, and statistical index to landslide susceptibility: an approach to endangered cultural heritage

This paper presents the applicability of the analytic hierarchy process, frequency ratio, and statistical index in landslide susceptibility mapping for an area from north-eastern Romania. The dependent factor (in this case the landslides) was determined by combining seven conditioning factors: elevation, slope angle, curvature, normalised difference vegetation index, roughness, distance to rivers, and landforms. A landslide inventory was prepared using topographic plans and orthophotos and by carrying out field surveys with the total station (Leica TCR1201) and geodetic GPS (Leica System 1200) in RTK mode. The final susceptibility maps were made with the help of GIS and validated by carrying out the receiver operating characteristics (ROC) curves. ROC plots for training data have shown that the susceptibility map generated using the analytic hierarchy process with AUC value of 79.46 has a better prediction accuracy compared with frequency ratio method (AUC = 75.24), and statistical index method, which had an AUC value of 61.46. The prediction rate of the three models is over 50% (79.64% for the analytic hierarchy process model, 75.65% for the frequency ratio model, and 66.36% for the statistical index model). More than a half of the cultural heritage sites are located in areas with a very high and high vulnerability to landslides. Taking into consideration that multi-stratified settlements are located in areas with low and medium landslide susceptibility, this fact can indicate a possible awareness of prehistoric people from natural hazards. The final susceptibility maps can be used for land use planning, risk reduction, hazard mitigation, to evaluate the present state of cultural heritage sites in danger and to predict which of the sites will be in more danger in the future.