Landslide Hazard Zonation Using Multi-Criteria Decision Analysis: An Integrated GIS Approach

Abstract

For landslide hazard zonation in the Himalaya, integration of terrain, geological, hydrological, and anthropogenic controls is required. This paper describes a geospatial reanalysis of a verified landslide dataset across Uttarakhand, and describes Chauhan et al.’s fuzzy analytical hierarchy process (fuzzy-AHP) outputs. The original dataset on which the reanalysis is based, contains 7,182 mapped landslides and 16 factors. The fuzzy AHP scores were reanalyzed to understand rainfall, geology and slope account for 47.23% of total criterion weight, and the top five factors account for 63.95% of total criterion weight. With a report of 53,400 km2 state area, the fuzzy-AHP zoning separated 38.00% of Uttarakhand to very low/low susceptibility, 33.48% to moderate susceptibility and 28.52% to high/very high susceptibility, which equals to approximately 20,292, 17,878 and 15,230 km2, respectively. An independent validation revealed that fuzzy-AHP has AUC 55.49%, accuracy of 52 % and F1-score of 0.33, while random forest and XGBoost both achieved more than 90% AUC. The results revealed that fuzzy AHP provides a good framework for interpretable, prioritized spatial approaches, but fuzzy AHP models require both, calibrated inventories and external validations. Thus, a GIS integrated framework that offers fuzzy AHP with a benchmark of predictive model is recommended.