This research was primarily motivated by several major deficiencies in data-driven approaches for mapping landslide susceptibility, including strong sensitivity to training data, which leads to a lack of expandability and portability, and their unsuitability for large area applications. In this paper we presented an alternative approach: an expert knowledge-based approach to address the issues related to the data-driven approaches.
The proposed expert knowledge-based approach includes three general steps: (1) extraction of knowledge on relationships between landslide susceptibility and predisposing factors from local domain experts by using knowledge acquisition techniques, (2) characterization of the needed predisposing factors by using GIS techniques, and (3) fuzzy landslide susceptibility inference to predict landslide susceptibility.
The proposed approach was conducted and evaluated in two case study areas: Kaixian and Three Gorges. The Kaixian study area was used to develop and test the methodology. The Three Gorges study area was used to test the portability of the expert knowledge and the applicability of the developed methodology for large-scale study areas. From the results of the case study we conclude that the expert knowledge-based methodology is effective for mapping landslide susceptibility, and its performance was maintained when it was moved to a new and significantly larger area without changes to the knowledge base. This suggests that the knowledge-based approach is portable and is suitable for application over large areas.