In 2006, a public-supply well in San Antonio, Texas, was selected for intensive study to assess the vulnerability of public-supply wells in the Edwards aquifer to contamination by a variety of compounds. A local-scale, steady-state, three-dimensional numerical groundwater-flow model was developed and used in this study to evaluate the movement of water and solutes from recharge areas to the selected public-supply well. Particle tracking was used to compute flow paths and advective traveltimes throughout the model area and to delineate the areas contributing recharge and zone of contribution for the selected public-supply well.

The local-scale model grid has a finer vertical discretization than do previous regional Edwards aquifer models and incorporates refined parameter zones corresponding with multiple (10) hydrogeologic units representing the Edwards aquifer. In the Edwards aquifer, high matrix porosity and permeability likely are overshadowed by high permeability developed in structurally influenced karstic conduit systems that transmit water into, through, and out of the aquifer system. The complexity of the aquifer system in the local-scale study area is further increased by numerous faults with varying vertical displacements. The extensive faulting results in the juxtaposition of hydrogeologic units with differing hydraulic properties and has appreciable effects on groundwater flow in the Edwards aquifer. The local-scale model simulations use the MODFLOW Hydrogeologic-Unit Flow Package and include two hydrogeologic units with high hydraulic conductivities (one or more orders of magnitude higher than for the other simulated hydrogeologic units) that are intended to simulate fast flow paths attributable to karst features. The two “conduit” hydrogeologic units of the Edwards aquifer represent the lower 8 meters of the leached and collapsed members and the Kirschberg evaporite member of the Edwards Group. The MODFLOW Horizontal-Flow Barrier Package was used to simulate faults in the local-scale model. The assumption was made that the degree to which a fault acts as a barrier to groundwater flow is proportional to the fault displacement. The final calibrated hydraulic-conductance values ranged from 0.01 to 0.2 per day for fault displacements ranging from 0 to more than 100 percent of the total aquifer thickness.

The calibrated steady-state simulation generally reproduces the spatial distribution of measured water-level altitudes. Simulated water-level altitudes were within 9.0 meters of measured water-level altitudes at 74 of the 84 wells used as targets for the local-scale model for the calibrated steady-state simulation. The overall mean absolute difference between simulated and measured water-level altitudes is 4.2 meters, and the mean algebraic difference is 1.9 meters. The simulated springflow for San Antonio Springs was 7.7 percent greater and for San Pedro Springs was 4.2 percent less than the median measured springflow. Simulated tritium concentrations were within 0.14 tritium units of measured tritium concentrations for 11 of the 13 local-scale study tritium observations from the 10 local-scale study wells used to calibrate the steady-state local-scale model, with a mean absolute difference between simulated and measured tritium concentrations of 0.11 tritium units and a mean algebraic difference of -0.04 tritium units. Simulated tritium concentrations in the selected public-supply well during November 2007 were within 0.09 tritium units of the measured concentrations, with the exception of the shallowest observation from the well.

The steady-state simulation water budget indicates that recharge occurring in the local-scale study area accounts for 31.8 percent of the sources of water to the Edwards aquifer in the local-scale model area and that inflow through the model boundaries contributes 68.2 percent. Most of the flow into the local-scale model area through the model boundaries occurs through the western and southern boundaries, 58.2 and 39.6 percent, respectively. The largest discharges from the Edwards aquifer in the local-scale model area are boundary outflow (71.4 percent) and withdrawals by wells (24.9 percent). Most of the flow out of the local-scale model area through the model boundaries occurs through the southern and eastern boundaries, 54.2 and 39.6 percent, respectively.

The simulated zones of contribution for the selected public-supply well, Timberhill well nest, and Zarzamora well nest extend to the north, northeast, and northwest from each site in the confined zone of the aquifer into the recharge zone, where all recharge to the aquifer occurs. The area contributing recharge for the selected public-supply well has the greatest extent. The area contributing recharge for the Timberhill well nest encompasses approximately the western one-half of the area contributing recharge for the selected public-supply well, and that for the Zarzamora well nest encompasses approximately the eastern two-thirds of the area contributing recharge for the selected public-supply well.

Simulated particle ages ranged from less than 1 day to more than 1,900 years in the 10 local-scale study wells (13 local-scale study tritium observations) used to calibrate the local-scale model. The simulated mean particle ages for the tritium observations representing selected well depths (shallow, intermediate, and deep) ranged from 2.5 to 15 years. The minimum (youngest) mean particle ages for the selected public-supply well and the Timberhill monitoring wells were for the intermediate well depth, while the youngest mean particle age for the Zarzamora monitoring wells was for the intermediate and deep well depth. The maximum (oldest) mean particle ages for the selected public-supply well and the Zarzamora monitoring wells were for the shallow well depth. The mean of simulated particle ages for tritium observations representing well depths open to the simulated conduit hydrogeologic units was 3.8 years, whereas the mean of simulated particle ages for tritium observations representing well depths not open to the simulated conduit hydrogeologic units was 9.6 years.

The effect of short-circuit pathways, for example karst conduits, in the flow system on the movement of young water to the selected public-supply well could greatly alter contaminant arrival times compared to what might be expected from advection in a system without short circuiting. In a forecasting exercise, the simulated concentrations showed rapid initial response at the beginning and end of chemical input, followed by more gradual response as older water moved through the system. The nature of karst groundwater flow, where flow predominantly occurs via conduit flow paths, could lead to relatively rapid water quality responses to land-use changes. Results from the forecasting exercise indicate that timescales for change in the quality of water from the selected public-supply well could be on the order of a few years to decades for land-use changes that occur over days to decades, which has implications for source-water protection strategies that rely on land-use change to achieve water-quality objectives.