Performance evaluation for four GIS-based models purposed to predict and map landslide susceptibility: A case study at a World Heritage site in Southwest China

摘要: Abstract Landslide susceptibility mapping is a prerequisite for preventing and mitigating hazardous risks, especially in landslide-prone mountainous areas. This study applies two well-known models – the information value model (IVM) maximum entropy (MaxEnt) model, additional biological climatic (Bioclim) mean distance (Domain) to evaluate landslide of Honghe Hani Rice Terraces, World Heritage site Yuanyang County Southwest China. A spatial dataset comprising 235 historical locations were used at training-to-testing data ratio 75:25. Fifteen commonly environmental, geological, meteorological factors, as well three new factors including proximity settlements, human activity intensity, multi-year average temperature selected impacting occurance. Model validation comparison performed using area under receiver operating characteristic curve, several statistical indices (sensibility, specificity, accuracy, precision, recall, true skill statistic or TSS, F-measure). The results show that MaxEnt yields best overall performance analysis, followed by Domain IVM, Bioclim model. jackknife test result shows most important contributing hazard include roads, annual rainfall, settlements. Furthermore, only small proportion areas positioned along roads identified be high very risk landslides given fact located remote region with under-developed economy southwest Findings from this can facilitate mitigation sustainable conservation Terrace other similar