A Comparative Study of Karst Sinkhole Hazard Mapping Using Frequency Ratio and Artificial Neural Network for East Central Florida

Sinkholes are one of the most common geohazards occurring in East Central Florida (ECF). Identifying areas prone to sinkholes is vital for land use planning in the ECF area, and thus, sinkhole hazard mapping plays a critical role. The present study presents (1) sinkhole hazard maps of ECF by using frequency ratio (FR) and artificial neural network (ANN) models and (2) a validation and comparison of the performance of two models. An inventory map with a total of 757 sinkhole locations was prepared from Florida subsidence incident reports (FSIR). 70% (530 sinkholes) were randomly selected to calibrate the sinkhole hazard models, and the remaining 30% (227 sinkholes) were used for the model validation. Five sinkhole contributing factors were considered including age of sediment deposition, hydraulic head difference, groundwater recharge rate, overburden thickness, and proximity to karst features. The relationship between sinkhole occurrence and sinkhole contributing factors was investigated through a GIS-based statistical analysis