Probabilistic spatial susceptibility modeling of carbonate karst sinkhole.

Sinkhole is one of major geohazards in karst area where soluble carbonate bedrock is underlain by cover soils (e.g., sandy or clayey soils) and active groundwater circulation is present. Groundwater recharge easily erode the overburden soils away into cavities in the bedrock, and ultimately structural collapse in ground. The induced ground collapse/subsidence may pose a great threat to human safety, environment, and buildings and infrastructure. The goal of this study is to develop a method/tool to spatially assess the possibility of sinkhole occurrence. The paper presents two main taks: (1) development of a probabilistic spatial prediction model of sinkhole susceptibility and (2) a geographical information system (GIS)-based regional-scale sinkhole susceptibility map. The study area is the east central Florida (ECF) region that has been experiencing more abrupt and larger cover collapse sinkholes. The research methodology employs two statistical methods for the spatial prediction model, frequency ratio (FR) and logistic regression (LR), accounting for six criteria that are soil erosion, mechanical stability, distance, geology, human activity, and climate condition. The final section of factors includes head difference, soil permeability, thicknesses of aquifer systems (overburden, surficial/intermediate aquifer), distance to karst features, surficial geology, lithology, land use/land cover, and rainfall. Once the models were constructed, both models were validated and compared by using receiver operating characteristics (ROC) analysis. The results indicate that the LR model better delineates high susceptibility areas of sinkhole and hydrological factors (recharge and head difference) are more accurately reflected on the model of the LR. • Develop probabilistic spatial susceptibility model of carbonate karst sinkhole. • A regional-scale sinkhole susceptibility map for the east central Florida region. • Evaluate the performance of Frequency Ratio (FR) and Logistic Regression (LR) methods in the sinkhole susceptibility model. • Statistical analysis on the impact of sinkhole contributing factors. [ABSTRACT FROM AUTHOR]