Graduation Year

2013

Document Type

Thesis

Degree

M.S.

Degree Granting Department

Geology

Major Professor

Sarah E. Kruse

Keywords

anomaly, GIS, GPR, Logistic, SPT, Statistics

Abstract

Abstract

Sinkholes and sinkhole-related features in West-Central Florida (WCF) are commonly identified using geotechnical investigations such as standard penetration test (SPT) borings and geophysical methods such as ground penetrating radar (GPR) and electrical resistivity tomography (ERT). Geophysical investigation results can be used to locate drilling and field testing sites while geotechnical investigation can be used to ground truth geophysical results. Both methods can yield complementary information. Geotechnical investigations give important information about the type of soil, groundwater level and presence of low-density soils or voids at the test location, while geophysical investigations like GPR surveys have better spatial coverage and can resolve shallow stratigraphic indicators of subsidence.

In GPR profiles collected at 103 residential sites in covered-karst terrain in WCF, sinkhole-related anomalies are identified using GPR and SPT methods. We analyze the degree to which the shallow features imaged in GPR correlate spatially with the N-values (blow counts) derived from SPTs at the 103 residential sites. GPR anomalies indicating sinkhole activity are defined as zones where subsurface layers show local downwarping, discontinuities, or sudden increases in amplitude or penetration of the GPR signal. "Low SPT values" indicating sinkhole activity are defined using an optimization code that searched for threshold SPT value showing optimum correlation between GPR and SPT for different optimal depth ranges. We also compared these criteria with other commonly used geotechnical criteria such as weight of rod and weight of hammer conditions.

Geotechnical results were also used to filter the data based on site characteristics such as presence of shallow clay layers to study the effectiveness of GPR at different zones. Subsets of the dataset are further analyzed based on geotechnical results such as clay thickness, bedrock depth, groundwater conditions and other geological factors such as geomorphology, lithology, engineering soil type, soil thickness and prevalent sinkhole type. Results are used to examine (1) which SPT indicators show the strongest correlations with GPR anomalies, (2) the degree to which GPR surveys improve the placement of SPT borings, and (3) what these results indicate about the structure of sinkholes at these sites.

For the entire data set, we find a statistically significant correlation between GPR anomalies and low SPT N-values with a confidence level of 90%. Logistic regression analysis shows that the strongest correlations are between GPR anomalies and SPT values measured in the depth range of 0-4.5 m. The probability of observing a GPR anomaly on a site will decrease by up to 84% as the minimum SPT value increases from 0 to 20 in the general study area. Boreholes drilled on GPR anomalies are statistically significantly more likely to show zones of anomalously low SPT values than boreholes drilled off GPR anomalies. We also find that the optimum SPT criteria result in better correlation with GPR than other simple commonly used geotechnical criteria such as weight of rod and weight of hammer. Better correlations were found when sites with poor GPR penetrations are filtered out from the dataset. The odds ratio showed similar result while the result varied with the depth range, statistics and threshold SPT value (low N- value with optimum correlation), with a maximum observed odds ratio of 3.

Several statistical results suggest that raveling zones that connect voids to the surface may be inclined, so that shallow GPR anomalies are laterally offset from deeper zones of low N-values. Compared to the general study area, we found locally stronger correlation in some sub-regions. For example, the odds ratio found for tertiary hawthorn subgroup were 25 times higher than the odds ratio found for the general study area (WCF).

Share

COinS