Groundwater Availability Modeling of the Barton Springs Segment of the Edwards Aquifer
Client: Texas Water Development Board
Reference: Dr. Robert Mace
Phone: (512) 936-0861
Email: Robert.Mace@twdb.state.tx.us
Project Completion Date: March 2003
Project Manager: Dr. Bridget Scanlon
The Bureau of Economic Geology and
subcontractor Barton Springs Edwards
Aquifer Conservation District developed
a groundwater flow model for the Barton
Springs segment of the Edwards Aquifer
to evaluate the effects of future pumpage
and potential future droughts on
groundwater availability (Fig. 1). The
model covers an area of ~ 260 square
miles and uses a 14,400 node grid (7,043
active nodes) with rectangular cells (500
ft x 1000 ft) to simulate the
potentiometric surface and spring
discharge in response to past, present,
and future pumpage and potential future
droughts.
Dr. Mace (TWDB) was
involved in the early development of this
model. The model was calibrated using a
combination of trial and error and an
inverse optimization code (UCODE).
Figure 1. Barton Springs model region.
The Barton Springs aquifer constitutes the sole source of water to about 45,000 residents. Barton
Springs pool also serves as a municipal swimming pool in Zilker Park, downtown Austin.
Increased population growth and recent droughts (1996) have focused attention on groundwater
resources and sustainability of spring flow. The primary management issue for this aquifer is
maintaining spring flow during drought periods and assessing current and future pumpage
effects on spring flow. Maintaining spring flow is a critical objective because the spring outlets
are the sole habitat of the Barton Springs salamander, which is listed as an endangered species.
Numerical modeling of groundwater flow in the Barton Springs Edwards Aquifer presented
many challenges because of the complex geology resulting from numerous faults with large
vertical offsets, and the extremely dynamic nature of the aquifer as a result of large conduits.
Unique aspects of the model that made it particularly suitable for modeling include:
 stream gauge data upstream of the outcrop areas for recharge estimation
 detailed pumping records reported by individual users of large wells collected by the
BSEACD since 1989
 daily spring flow records for  100 yr for Barton Springs and water level hydrographs
distributed throughout the aquifer for variable time periods up to 10 yr for comparison
with simulated values
 detailed synoptic water level maps developed by BSEACD for low and high flow
periods for comparison with model simulations
No other aquifer in the state has such detailed records for model input and evaluation of model
results. The information provided by these data was critical in developing a conceptual model of
flow for this Barton Springs Edwards Aquifer.
Model calibration was particularly challenging because of the complex geologic structure in the
model area and the dynamic nature of flow. Model parameterization included a trial and error
approach based on the distribution of hydraulic heads followed by automated inverse modeling
to further reduce the root mean square (RMS) error between simulated and measured heads.
Steep head gradients in the outcrop area were assigned low values of hydraulic conductivity and
shallow head gradients in the central and eastern part of the model were assigned high hydraulic
conductivities. Highest hydraulic conductivities were assigned to a zone surrounding Barton
Springs because this area should reflect the convergence of flow paths. This approach of
parameterising hydraulic conductivity resulted in a parsimonious distribution of hydraulic
conductivity and low RMS errors.
The transient model was run using data from 1989 through 1998 and generally reproduced the
temporal variability in spring discharge with no calibration. The challenge with the model was to
accurately
simulate
low
spring
discharges because these discharges are
critical for using the model to predict the
effect of increased pumpage and
potential future droughts on spring
discharge. By varying the specific yield
in the outcrop area, the low spring
discharges could be simulated fairly
accurately. High spring discharges were
generally overestimated because the
model did not include reported ungauged
springs that start flowing at high
discharges.
Figure 2. Simulated and measured discharge at
Barton Springs.
To assess the impact of future pumpage and potential future droughts on groundwater
availability, transient simulations were conducted using extrapolated pumpage for 10-yr periods
(2001 through 2050) and average recharge for a 3-yr period and recharge from the 1950’s
drought for the remaining 7 yr. Results of these simulations were compared with those using
average recharge and future pumpage. Simulated spring discharge in response to future pumpage
under average recharge decreased proportionally to future pumpage (2 cfs per decade), whereas
spring discharge decreased to 0 cfs in response to future pumpage under drought-of-record
conditions. Management of water resources under potential future drought conditions should
consider enhanced recharge and conservation measures.
In addition, simulations based on the distributed parameter modeling of the Barton Springs
Edwards Aquifer using MODFLOW were compared to those based on a much simpler lumped
parameter model developed by Barrett and Charbeneau, 1987. Results from this comparison
indicated that both the distributed and lumped parameter models could adequately simulate
spring discharge (Fig. 2) and well hydrographs, but that the distributed parameter model is
required to simulate the potentiometric surface and to evaluate aquifer response to future
pumpage.
The results of this study demonstrate the ability of equivalent porous media distributed and
lumped parameter models to simulate regional groundwater flow, which is critical for managing
water resources in karst aquifers and predicting the impact of future pumping and potential
future drought conditions on spring flow.