PRELIMINARY LANDSLIDE SUSCEPTIBILITY MAPPING IN THE NOLICHUCKY WATERSHED USING DEEP-LEARNING TECHNIQUES: IMPACTS OF HURRICANE HELENE

Hurricane Helene’s aftereffect as tropical rainfall led to significant geomorphic changes across the southeastern United States, including widespread landslides in the Appalachian region. This study investigates landslide susceptibility along the Nolichucky River of East Tennessee, an area that experienced slope failures in the storm’s aftermath. Among the wide range of slope failures, including small slumps, and rockfall, debris flows were common in the hillslopes. By integrating high-resolution LiDAR data with environmental variables such as lithology, soil type, elevation, slope angle, annual precipitation, drainage characteristics, topographic indices such as the Topographic Wetness Index (TWI) and Topographic Position Index (TPI), we applied deep-learning techniques to identify areas of heightened debris flow susceptibility. Multispectral imagery with unsupervised machine learning and a Recurrent Neural Network (RNN) and Decision Trees (DT) model, were used to predict landslide hazard in response to events like Hurricane Helene. Preliminary results suggest that deep-learning methods can effectively delineate zones of increased landslide hazard, offering insights into the interaction between extreme rainfall events and terrain susceptibility. The findings aim to inform risk mitigation strategies for communities vulnerable to future extreme weather events in the Appalachian region.