RAPID AND FLEXIBLE LANDSLIDE INVENTORY MAPPING WITH CCM FLOW

Building landslide inventories is an important first step toward understanding the geologic conditions and physical mechanisms that govern future slope failures. The most reliable and informative inventories are produced when experts scour lidar-derived high resolution topographic data, manually outlining and annotating landslide extents. While this process is highly effective, it is also time-consuming and subjective. In an attempt to overcome these limitations, numerous researchers have proposed automated or semi-automated alternatives that use machine learning and statistical analysis techniques to inventory landslides. Several of these alternatives compare impressively to expert-produced inventories, but their complexity and need for fine-tuning renders them computationally expensive and impractical to adapt for diverse geologic settings and digital elevation (DEM) model quality. Further, the requirement of an existing manual inventory for training prevents immediate use in areas where such datasets do not exist. The recent Contour Connection Method (CCM) developed by Leshchinsky et al. (2015) has been modified in context of these challenges. The new methodology, called CCM Flow, operates as a three step procedure. First, a DEM is refined to a standard input quality. Second, the refined DEM is used to identify landslide headscarps. Lastly, the DEM and headscarps are provided as inputs to the CCM algorithm, which essentially delineates flow paths of landslide material downslope to identify the deposits. This flexible approach allows for semi-automated landslide mapping in varied geologies, using any DEM where landslide features are visible, lidar or not, and with or without the benefit of information provided by a manual training inventory. Hence, CCM Flow’s capabilities range from rapid development of landslide inventories in regions stricken by natural hazards to producing a preliminary inventory to inform detailed, manual mapping efforts. The effectiveness of the CCM Flow procedure will be expressed through a comparison to several existing expert-derived manual landslide inventories.