Robustness evaluation of fuzzy expert system and extreme learning machine for geographic information system-based landslide susceptibility zonation: A case study from Indian Himalaya

In the last few decades, with the development of computers and geographic information system (GIS), a wide range of landslide susceptibility zonation (LSZ) techniques were orchestrated by various researchers around the globe. Among them, the artificial intelligence (AI) have been distinctly regarded as the most effective and suitable approach to part with GIS for LSZ. Though, suitability of AI for LSZ is well addressed in the landslide literature, noises of processing data, choice of causative factors and landslide density of study area are the number of hindrances that cause quandary over preference of ideal AI technique among many. The current study intends to analyse and compare the predictive performance of two entirely different AI techniques, fuzzy expert system (FES), a bivariate statistical technique, and extreme learning machine (ELM), a multivariate statistical technique for GIS based LSZ. The Mussoorie Township, a famous tourist destination in the Indian State of Uttarakhand was taken as the study area. Thematic layers of relevant causative factors and landslide inventory were prepared for the study area through field survey, remote sensing, and GIS. The resultant landslide susceptibility maps (LSM) of the study area, LSM-I of FES and LSM-II of ELM were critically evaluated and compared with the aid of landslide inventory of the study area.