LANDSLIDE SUSCEPTIBILITY AND RISK MAPPING IN THE BIG SANDY AREA DEVELOPMENT DISTRICT, EASTERN KENTUCKY
Using LiDAR-derived datasets and landslide inventories, we developed landslide-susceptibility maps using a hybrid machine-learning approach based on hillslope morphology variables. Bagged trees, a machine-learning random-forest classifier, was used to evaluate geomorphic variables; 12 were identified as important: standard deviation of plan curvature, standard deviation of elevation, sum of plan curvature, minimum slope, mean plan curvature, range of elevation, sum of roughness, mean curvature, sum of curvature, mean roughness, minimum curvature, and standard deviation of curvature. These variables were further evaluated using logistic regression modeling to determine probability of landslide occurrence and create a landslide-susceptibility map. The logistic-regression model’s performance was evaluated by the area under the receiver operating characteristic curve, which was 0.83, indicating strong performance. The maps are divided into five classes to increase their utility: low (15.0 percent), low‒moderate (39.2 percent), moderate (23.4 percent), moderate‒high (13.9 percent), and high (3.7 percent).
The risk assessment is a static, socioeconomic approach including potential landslide effects on population, roads, railroads, buildings, and land. We combined vulnerability and consequence data with landslide susceptibility data to calculate a risk factor. Resulting risk factors were classified as low (70.7 percent), moderate (11.0 percent), or high (1.5 percent). The maps and data will be incorporated into hazard mitigation plans that implement measures related to improved spatial assessment of landslide hazards, land use and development, transportation, critical facilities, and safety.