2003 Seattle Annual Meeting (November 2–5, 2003)

Paper No. 6
Presentation Time: 8:00 AM-12:00 PM


CICHON, James R., BAKER, Alan E., ARTHUR, Jonathan D., WOOD, Henry A.R. and MEANS, Guy H., Florida Geological Survey, FDEP, 903 W. Tennessee St, Tallahassee, FL 32304-7700, guy.means@dep.state.fl.us

The relative vulnerability of an aquifer to contamination is dependent upon the thickness and composition of sediments overlying it and the rate at which contaminants travel through these sediments. To predict the vulnerability of Florida’s major aquifer systems to contamination the Florida Geological Survey is currently developing the Florida Aquifer Vulnerability Assessment (FAVA) model. FAVA differs from the Environmental Protection Agency’s DRASTIC model in that the newer technique is GIS based and accounts for Florida’s karst terrain. Current methods employed in FAVA model development include Weights of Evidence, Fuzzy Logic and Travel Time. Of these methods, Weights of Evidence best utilizes new and available data sets to predict relative vulnerability since it is statistically validated on the front end, easily updateable, uncertainties can be calculated and the model avoids preconceptions.

Weights of Evidence quantifies relationships between spatial layers with actual contaminant occurrences in order to assess a hypothesis. Contaminant source data (i.e., training points) is obtained from the Florida Department of Environmental Protection’s Background Water Quality Network of wells. Spatial layers (i.e., evidential themes) consist of existing and newly developed GIS data and include depth to water, soil drainage, distance to karst features, thickness of confinement and vertical leakage rates. Different evidential themes are utilized based on the aquifer being modeled. The evidential themes included in the Floridan aquifer system (FAS) model, for example, are thickness of confining unit, distance to karst features and soil drainage. To aid in the creation of these themes, data collected during geologic mapping projects (e.g., cores, well cuttings, and wireline logs) are utilized. By calculating the statistical significance between training points and evidential themes, interactions can be analyzed to yield a data-driven predictive model. The output is a grid-based probability map that can be used by environmental, regulatory and planning professionals to facilitate the protection of Florida’s ground-water resources.