Landslide susceptibility prediction in the loess tableland considering geomorphic evolution

Abstract

To explore the relationship between the loess tableland geomorphic evolution and landslide susceptibility, a loess tableland area in the northern of Xi’an City was selected in this study. The study area was divided into four areas according to the type of loess tableland. A complete historical landslide inventory was created and 11 landslide factors were selected. Landslide susceptibility prediction was performed for the subarea and the whole area based on the SVM and RF models, respectively, and several landslide susceptibility maps were generated. The results show that the prediction of landslide susceptibility in loess tableland areas needs to take into account the influence of its landscape evolution, and more accurate results can be obtained by calculating individually for different sub-areas. The RF model has strong performance in landslide susceptibility prediction when considering the tableland evolution, while the SVM model is better when considering the whole region as a whole. Multiple factors, such as human activities and river erosion, have increased the landslide susceptibility of the residual tableland phase, and repeated landslides along the tableland margins have led to a continuous reduction of the loess tableland surface. Given the importance of tableland landforms in the Loess Plateau region, a framework for susceptibility mapping and zonation was proposed, which is suitable for loess tableland regions where multiple developmental phases co-exist. This study strengthens the understanding of landslide development and susceptibility in response to the landscape evolution of the loess tableland and provides support for disaster prevention and mitigation and land planning.