1. An Integrated Approach for Subsidence Monitoring and Sinkhole Formation in the Karst Terrain of Dougherty County, Georgia
Matthew Cahalan1 and Adam Milewski1
1 Department of Geology, University of Georgia, Athens, GA, 30602, USA
Abstract:
The Upper Floridan Aquifer (UFA) is an important water source for agricultural and municipal purposes in Dougherty County. However, prolonged water demand, dissolution of the carbonate aquifer (Ocala Limestone), and short and long-term fluctuations in precipitation, surface water discharge, and groundwater levels have placed many areas at risk for sinkhole formation and broader scale land deformation (i.e., subsidence) in the covered karst terrain. Evaluating the causes of sinkhole formation and subsidence is essential for reducing risk within karst settings.
In this study, we evaluate the controlling factors on subsidence feature formation within a 71 mi2 area of southern Dougherty County, Georgia, by applying a sinkhole delineation and regression analysis procedure. ArcGIS 10.1 software was used to process digital elevation models (DEMs) from 1999 (10m) and 2011 (1m and resampled 10m) to locate subsidence features that formed between 1999 and Different types of subsidence features were delineated based on area values from the high spatial resolution LiDAR DEM (1m). Results show that small-scale subsidence features (i.e., cover collapse sinkholes less than 100 m2) form close to streams in areas with relatively thin overburden thickness. This is most likely due to flooding and subsequent liquefaction processes transporting overburden material into subsurface voids.
Our conceptual model of subsidence feature controlling factors, or independent variables, included geologic, topographic, hydrologic, anthropogenic, and hydrogeologic variables. Regression analysis allows you to examine the influences of various factors driving observed spatial patterns. Ordinary least squares (OLS) and geographically weighted regression (GWR) statistical models were applied to the recently formed subsidence features between 1999 and 2011 to determine the factors that are most influential in subsidence feature development.
The OLS global regression tool was applied to examine if we had a properly specified model with limited biases. OLS results showed that distance to lineaments, elevation, aspect, distance to roads, and overburden thickness were statistically significant (p-value < 0.05), or the most influential controlling factors in subsidence feature density. The 11 included independent variables did not show any redundancy (Variance Inflation Factors < 7.5). Overall, the OLS model explained 42% of the spatial variation of subsidence feature density (R2 = 0.42).
GWR models, a local form of regression that illustrates how relationships vary across space by calculating a regression equation for each feature in the study area, were applied. GWR analysis resulted in a R2 value of 0.70 when 9 of the independent variables were included in the model (slope, aspect, land use, overburden thickness, distance to lineaments, distance to roads, distance to surface water features, distance to rivers, and distance to wetlands).
Subsidence Feature Development .
Subsidence Feature Density Maps
OLS Regression Results
Category
Description
Source
Method
OLS p-value
Elevation
NED 10m DEM
Raster to polygon *
Topography
Aspect
ArcMap Aspect tool *
Slope
ArcMap Slope tool
Distance to ponds
Georgia GIS Clearinghouse
ArcMap Generate Near Table tool
Hydrology
Distance to streams
National Hydrography Dataset
Distance to wetlands
National Wetland Inventory
Distance to lineaments
Brook and Allison (1986) *
Overburden thickness
Borehole data and USGS cross-sections (48 wells)
Empirical Bayesian Kriging interpolation (RMS = 2.42) *
Hydrogeology
Upper Floridan Aquifer fluctuations
USGS well data (12 wells)
Kriging interpolation (RMS = 3.22)
Land use
National Land Cover Dataset
Distance to roads *
Fig. 4 – 1999 (10m).
Fig. 5 – 2011 (10m).
Geology
Dobecki and Upchurch, 2006
Fig. 2 – Subsidence features can be triggered by flooding events (i.e., liquefaction) and drought or groundwater withdrawal (i.e., loss of hydrostatic support).
Density (features/km2)
Anthropogenic
Subsidence Feature Delineation and Regression Procedure
Fig. 8 – Independent variable descriptions. Statistically significant variables (p-values < 0.05) are in bold.
Subsidence Feature Types
Fig. 6 – Subsidence features that formed between 1999 and 2011.
Independent Variables
Geology
Hydrology
Fig. 9 – Lidar-derived map of subsidence features within study area. Cover subsidence features are > 100 m2, and cover collapse features are < 100 m2.
Objectives:
Apply statistical regression models (OLS and GWR) to evaluate the influences of topographic, hydrologic, geologic, hydrogeologic, and anthropogenic variables on subsidence feature formation.
Locate recently formed (1999 – 2011) subsidence features using a semi-automated mapping procedure.
Area of Investigation
Conclusions
Anthro
The proposed processing procedure was successful at delineating sinkholes temporally and attributing their formation to various factors spatially. Regression results showed that the increase in subsidence features between 1999 and 2011 was primarily influenced by geologic (i.e., distance to lineaments and overburden thickness), topographic (i.e., elevation and aspect), and anthropogenic (i.e., distance to roads) factors. The cover-collapse features’ distributions along streams is evidence of liquefaction processes occurring in areas with shallow overburden thickness being a primary control on their formation, while cover subsidence features form in areas with greater overburden thickness at higher elevations.
Hydrogeology
Fig. 1 – Outlined study area location in southern Dougherty County, GA (71 mi2).
Fig. 3 - Processing procedure for the and 2011 digital elevation models (DEMs), which were used to locate subsidence features and analyze the factors controlling their development in the covered karst setting of southern Dougherty County, Georgia.
Acknowledgements
Fig. 7 – Maps of independent variables used in the ordinary least squares and geographically weighted regression models.
Randy Weathersby – Albany GIS
UGA Provost Fund for financial support of the project