Paper Type

Master's Thesis


College of Computing, Engineering & Construction

Degree Name

Master of Science in Civil Engineering (MSCE)



NACO controlled Corporate Body

University of North Florida. School of Engineering

First Advisor

Dr. Ryan Shamet

Second Advisor

Dr. Raphael Crowley

Third Advisor

Dr. Cigdem Akan

Department Chair

Dr. Alan Harris


Sinkholes are natural geohazards prevalent in Florida's distinctive karst geology, posing risks to public safety, infrastructure, and groundwater integrity. Considering the increasing frequency of sinkhole occurrences in Florida due to land development in vulnerable soils, understanding the potential for sinkhole occurrence can aid in the identification of high-risk locations and inform effective mitigation strategies. Researchers have developed three point-based indices to be used when assessing the vulnerability a project site has to sinkhole formation based on the encountered subsurface conditions. Engineers have historically used an index based on the subsurface layer thicknesses, called the Raveling Index (RI). In recent years, researchers have introduced more comprehensive point-based indices, the Sinkhole Resistance Ratio (SRR) and the Empirical Vulnerability Index (VIG + VIR), incorporating encountered soil strengths and other empirically based data fitting techniques. However, the indices, which were only validated for a dataset containing only three central Florida sites, lack calibration, which poses a challenge for engineers in interpreting the specific implications of index values and their connection to the likelihood of sinkhole formation. This study set out to validate the indices based on a more diverse database of 197 CPTs from 43 sites across Florida and develop practical techniques for the assessment of sinkhole potential in all Florida sites and for two specific sinkhole Areas I and III.

First, statistical regression was used to validate the indices and test their effectiveness in predicting the probability of historical sinkhole occurrence. The results, listed in increasing order of predictive capability for each area category, were as follows: RI and SRR for all sites; RI, SRR, VIG + VIR in Area III; and none of the indices in Area I. The probability regression functions for the indices showing evidence of correlation can thus be applied to preliminary sinkhole investigations in their respective area categories, to associate non-calibrated index values for each CPT to the likelihood of sinkhole occurrence based on historical occurrences. An example implementation of this technique is provided.

Second, to address the limitations in the first analysis for Area I, Finite Element Modeling (FEM) was used to simulate data representing the point at which failure occurs, to develop a stability chart for assessment of sinkhole potential based on encountered vulnerability for Area I sites. The simulated data was validated by a t-test affirming its utility in representing Area I field data. An analysis of the trend of simulated failure data in a stability scatter plot produced a failure threshold that separates the chart into two envelopes representing collapse conditions and non-collapse conditions for Area I sites. The resulting stability chart accurately predicted 70% of true collapse field data in the collapse envelope and 50% of non-collapse field data in the non-collapse envelope of the chart. Therefore, considering the simplistic nature of the numerical model, which assumes a homogenous clay soil whose stability is solely due to the soil strength and the size of the void, the proposed stability chart may be used by engineers to assess the potential of sinkhole occurrence based on where field data from preliminary sinkhole investigations plots with respect to the failure threshold.