BRANCHING BETWEEN GEOLOGICAL UNDERSTANDING AND MODEL PROCESS DESIGN OF THE BISCAYNE AQUIFER TO EXAMINE EFFECTS FROM MODEL UNCERTAINTY ON WELL-FIELD CAPTURE-ZONE ESTIMATES (Invited Presentation)

Determination of well-field recharge or capture zones is a common hydrological investigation in the field of groundwater hydraulics. Groundwater flow models are historically used, along with backward particle-tracking, to estimate subsurface areas that contribute water to wells and associated travel-time contours. Travel-time based capture-zones can be sensitive to the choice of conceptual model, parameters such as hydraulic conductivity and effective porosity, boundary conditions, sources and sinks. Historically, a model is developed then calibration is undertaken to make sure the model is effectively matching the observation dataset by selecting a single choice of input parameters that best matches this dataset. The calibrated model is then used to create a single deterministic estimate of the times of travel and capture zone. The process discussed here, on the other hand, uses 10,000 stochastically generated model-input realizations, or distributions, to examine uncertainty of these parameters on capture-zone results.

Increased understanding of the Biscayne aquifer has led to the need for updated capture-zone analysis of municipal supply well fields in Miami-Dade County, Florida. Although geophysics and well logs have provided an extensive amount of geologic data in the Biscayne aquifer, estimates of hydraulic parameters needed for model design and input are still not well understood and have an associated level of uncertainty, both at the borehole scale and between boreholes. To determine the validity of conceptualizing the Biscayne aquifer as a 2-D groundwater flow system on a regional scale, test models were developed to compare capture-zone results of a multi-layer model (with explicit properties for each layer) to those of a one-layer model (using effective model input parameters). Each test model was run using 10,000 stochastically generated model-input realizations and results were compiled and statistically analyzed to quantify uncertainty. Model input parameters that were varied stochastically include hydraulic conductivities, effective porosity estimates, and canal-bed conductances. Results suggested that for purposes of delineation of capture zones within the Biscayne aquifer, a 2-D one-layer groundwater flow model was sufficient for estimation of these subsurface areas.