Southeastern Section - 64th Annual Meeting (19–20 March 2015)

Paper No. 1
Presentation Time: 8:05 AM


O'REILLY, Andrew M., Geology & Geological Engineering, University of Mississippi, P.O. Box 1848, University, MS 38677-1848 and SEPÚLVEDA, Nicasio, U.S. Geological Survey, Caribbean and Florida Water Science Center, 12703 Research Parkway, Orlando, FL 32826,

In the karst terrain of central Florida, variations in both rainfall and groundwater use may affect surface and subsurface water-level and flow conditions, potentially affecting the ability of the hydrologic system to meet both human and environmental needs. Increasing development of central Florida and associated groundwater withdrawals have led to decreased lake water levels, groundwater levels, and spring flows in some areas and raised concerns about future groundwater availability. To address groundwater management needs, a finite-difference physics-based groundwater flow model was recently developed for central Florida and calibrated to conditions during a 12-year period. However, complex interactions between the surface and subsurface environments in karst terrains are difficult to simulate with regional-scale physics-based models. Alternatively, many long-term records of historical hydrologic data for central Florida are available in the databases of local, State, and Federal agencies, which are well suited for empirical modeling. On the basis of these data, the response of lake water levels, groundwater levels, and spring flows to changing rainfall and groundwater-use conditions over a multidecadal period was analyzed using artificial neural network (ANN) and other data-mining techniques. Given the unique opportunity of having two different modeling approaches applied to the same area, predictions at 48 sites (19 lakes, 23 wells, and 6 springs) simulated by both models were analyzed to assess the vulnerability of the karstic hydrologic system to changing environmental and human stresses. This comparison shows how different science-based approaches provide different yet complementary information on the behavior of the hydrologic system. The statistics-based ANN has the flexibility of considering a larger number of degrees of freedom and generally better explains variability in the hydrologic system’s response. Other differences are attributable in part to different conceptual assumptions of each approach. The merits and limitations of each modeling approach must be recognized and appropriately applied to yield a more comprehensive assessment and a more robust understanding of the hydrologic system and to optimize the effectiveness of groundwater management practices.