|dc.description.abstract||Karst feature inventories provide essential information used to evaluate a site’s degree of hydrogeologic connectivity to local and regional flow systems, as well as its environmental and ecological sensitivity. For developments proposed on the Edwards Aquifer recharge zone, TCEQ rules require a full-coverage karst feature inventory of karst features during a geological assessment. However, visual surveys may be subjective, depending on the experience of the person performing a survey.
Considering this, my research focused on whether it is possible to develop an independent method for identifying the most sensitive recharge areas for visual surveys, when time and resources are limited, as well as provide a means for assessing the accuracy of surveys. The question motivating this research is: can relationships be identified between predictor variables and karst feature density that allow estimation of density without physical surveys?
A partial, statistically-designed, karst feature survey of the 17 km2 Freeman Center of Texas State University in San Marcos, Texas resulted in 60 documented karst features, including three sinkholes ground-truthed from a GIS-based sinkhole detection method. The survey design used for Freeman was then tested on Camp Bullis, near San Antonio, TX, an area with known karst feature density, revealing that random surveying does not yield representative karst feature density results, as karst features tend to cluster.
The entirety of Camp Bullis was analyzed for factors that influence karst feature density. An Ordinary Least Squares model determined that slope, distance to nearest flowline, lithology, and apparent resistivity were significant predictors of karst feature density (R= 0.30; p<0.01). A Geographically Weighted Regression was also used to visualize the nonstationarity of predictor variables (R= 0.81). However, both models resulted in spatial autocorrelation of residuals, indicating model misspecification. Despite concluding that karst features density is difficult to model, these methods offered a more nuanced understanding of factors controlling the distribution of karst features and the significance of these factors on Camp Bullis.||