Multi-hazard landslide dams on the global scale: distribution, characteristics, and formation

The landslide dam (LDam) is a complicated multi-hazard that can lead to enormous loss of life and properties worldwide. LDams have been frequently reported and recorded in the upstream mountainous areas, where the environment is prone to landslide occurrence and the rivers are narrow and steep. As a multi-hazard interaction between hillslope and fluvial systems, the study of landslide dams brings greater challenges compared to single hazards, especially in data collation and the integration of information from the two systems. Previous research has focused a large amount of effort on establishing LDam datasets, identifying the characteristics related to landslide dam formation and estimating the landslide dam formation probabilities through proposing geomorphological indices or dam-formation landslide probability mapping, based on the empirical relationships among related variables. However, the applicability of outputs for other study areas, particularly for larger-scales, and the limited use and application of fluvial information remain critical gaps in current research. This thesis proposes a new global-scale applicable framework for evaluating landslide dam formation susceptibility on river reaches. Contributing to this framework is the establishment of a global-scale LDam dataset named River Augmented Global Landslide Dams (RAGLAD) augmented by auxiliary global fluvial datasets, and a better understanding of the morphometric and spatial differences in characteristics between dam forming landslides and landslides more generally, based on multiple landslide and LDam datasets. This study expands the data collection method for enhancing the comprehensiveness of LDam records and shifts the focus from landsides to mapping the LDam formation probability on rivers instead of just the hillslopes. It also contributes a new entry point for integrating multiple data sources, such as global fluvial datasets and landslide datasets, which will enhance future LDam studies.