In each case, the numerical simulations of groundwater flow were calibrated to available data describing heads, spring discharges, stream flow gains and losses, and tracer-defined conduit flow velocities by modifying the extent, pattern, and size of conduits along with matrix permeabilities during model calibration. The broader significance of these results is three-fold. First, we learned that successful simulation of groundwater flow in karst aquifers is less limited by numerical capabilities or by direct knowledge of conduit patterns than by the explicit inclusion of and/or calibration to hydrologic data such as spring flows, swallet inputs, and tracer-defined velocities. Second, it is clear that where data is lacking, it is critical to include reasoned estimates of individual spring discharges and swallet inflow rates because omitting that data and/or estimates of those values will lead to a dramatically different depiction of groundwater flow patterns and velocities. Finally, the ability of these models to calibrate to multiple lines of hydrologic data demonstrates both the efficacy and utility of numerical modeling in karst aquifers if it is done correctly.
