This thesis presents an integrated, effective approach to understanding sinkhole formation dynamics and provides a decision-making tool to mitigate risks associated with sinkhole formation. Comprehensive ordinary least squares (OLS) and geographically weighted regression (GWR) geostatistical models are developed for quantifying sinkhole formation mechanisms in the mantled karst terrain of Dougherty County, Georgia (area: 183 km2). Sinkhole density was determined using a GIS-based sinkhole mapping procedure with ten-meter resolution digital elevation models (DEMs) from 1999 and 2011 and a one-meter 2011 LiDAR DEM. Geostatistical models were performed on two sinkhole density datasets: 1) a spatiotemporal dataset representing newly formed and enlarged sinkholes between 1999 and 2011, and 2) LiDAR-derived sinkholes. Geostatistical results show that geologic, hydrologic, and hydrogeologic variables are most influential in sinkhole formation. Geostatistical prediction results were used with deterministic interpolators to assess the ability of statistically-based methods to predict sinkhole density.