Evaluating the use of data on canal seepage and solute concentration in aquifer parameter estimation

We evaluate the benefit of using field observations on groundwater flux to canals and pore-water solute concentration, in addition to hydraulic head, to estimate model parameters for groundwater flow. The field site was a 100 km2 area of the Biscayne Aquifer in Southeast Florida. A total of seven models were calibrated to different combinations of 196 observations of groundwater head, 56 measurements of seepage between the aquifer and canals, and 196 measurements of pore-water chloride concentration. The calculated initial parameter correlations indicated that the model with higher parameter freedom (six instead of five) and calibrated solely on hydraulic head or canal seepage estimated unreliable parameters due to the non-uniqueness of the solution. Calculated parameter sensitivities of the final values further suggested that the six-parameter models calibrated solely on hydraulic head did not have sufficient information to estimate the parameters independently. Each of the seven calibrated models was evaluated by comparing simulated (predicted) and measured values taken during a 10-day canal drawdown test and during a 4-month time period, neither of which were used in the calibration. The model evaluations demonstrate that (1) the models calibrated solely on head measurements predict poorly regardless of the number of parameters estimated; (2) the models calibrated solely on seepage fluxes predict accurately if the number of estimated parameters is restricted; and (3) the models calibrated on at least two types of data (e.g. hydraulic head and seepage fluxes) exhibit good predictive capabilities with both the five- and six-parameter models.