DELINEATING SATURATED CONDUIT PATTERNS AND DIMENSIONS IN THE UPPER FLORIDAN AQUIFER THROUGH NUMERICAL GROUNDWATER FLOW MODELING

A dual-permeability steady-state model of groundwater flow through the western Santa Fe River Basin, Florida has been used to delineate a 381 km network of saturated conduits through calibration to heads and spring discharges. The model was setup using FEFLOW™ where conduits were described as discrete features embedded in a porous matrix. Flow in the conduits was described by the Manning-Strickler equation where variables for conduit area and roughness were used to adjust the volume and velocity of simulated spring flows. Matrix flow was described by Darcy’s law where hydraulic conductivity variations were limited to three geologically defined internally homogeneous zones. Recharge for both the high-water and low-water periods was determined through a water budget analysis where variations were restricted to nine zones defined by land-use. All remaining variations in observed head were then assumed to be due to conduits.

Two calibration datasets were compiled from 30 years of data describing average high-water and average low-water conditions based on heads at 188 wells and discharge from 18 springs. The model was iteratively calibrated to the two datasets wherein the location, size and roughness of the conduits were assigned as needed to accurately simulate observed heads and spring discharges while bounding simulated velocities by the tracer test results. The final model calibrated to within 5% of the total head change across the model region at 99% and 94% of the 188 wells in the high-water low-water scenarios respectively. Simulated spring discharges fell within 13% and 100% of the observed range in the high-water and low-water scenarios respectively. Simulated velocities ranged from as low as 10^-4 m/day in the matrix to as high as 10⁺³ m/day in the largest conduits.

The significance of these results that we emphasize here is two-fold. First, plausible karstic groundwater flow conditions can be reasonably simulated if adequate efforts are made to include springs, swallets, caves, and traced flow paths. And second, detailed saturated conduit networks can be delineated from careful evaluation of hydraulic head data particularly when dense datasets can be constructed by correlating values obtained from different wells under similar hydraulic periods.