Interpolation-Aided Evaluation of Groundwater Model Calibration

Lack of sufficient observed data exerts challenges to proper calibration and evaluation of predictive ability of groundwater models. In this paper, the use of interpolation of observed hydraulic heads for improving the calibration and evaluation of groundwater models at two transects across the Indian River Lagoon Estuary, has been investigated. Additional synthetic head values were obtained from the interpolated head surfaces. Numerical models were developed using SEAWAT code. Statistical measures used to evaluate the calibration results included: 1) root mean square error (RMSE), 2) Nash-Sutcliffe efficiency NSE index, 3) two-sided t-test, and 4) visual comparisons head contours below the lagoon. Calibration by RMSE was found to be misleading since relatively small RMSE values were obtained even with poor calibration. The effect of adding the synthetic interpolated observations on improving the calibration results was more obvious in the trnasect that has smaller width compared to the other transect that has larger width and has the same number of actual observations used. Both NSE and t-test calibration statistics improved significantly by using the interpolated observations in the larger transect. However, these two measures did not have effect on the results of the smaller transect. Results of this work suggest using interpolation to provide calibration targets in situations where limited field data is available for model calibration and evaluation.