INCORPORATING HUMAN IMPACTS AND NATURAL PROCESSES TO ASSESS SINKHOLE RISK

Current sinkhole hazard maps in Kentucky are based solely on geology, but the USGS estimates that over 80% of sinkhole collapses are anthropogenic in origin. A reliable sinkhole risk map is needed by land use planners, government agencies, and other stakeholders. We develop a new map of sinkhole hazard probabilities utilizing the random forests method and high precision sinkhole data from the Floyds Fork Watershed in North Central Kentucky. In applying the random forests method, we evaluate land use, geology, hydrogeology, land cover, topography, and soils to predict sinkhole risks. This work will identify the most common risk factors based on anthropogenic and natural impacts on a karst environment. Ultimately, applying the random forests classifier we generated to other areas, we intend to create more accurate sinkhole risk maps in other karst environments.