Comparing Groundwater Model Calibration Approaches: Does an Optimized Model Better Reflect Reality When Both Flow and Transport Observations are Applied?

This research compares two approaches to the application of formal calibration methods to groundwater modeling: sequential and combined. There exists research, both theoretical work and field studies, that point to improved estimation of hydraulic parameters when multiple types of observations, flow and transport, are applied simultaneously. There also exist theories about why combined calibration might not be compatible with the differing mathematical basis of flow compared to transport. This research has taken a closer look at the mechanisms at work by comparing these approaches when the input parameters are known. Using stochastic methods instead of field data, the original input parameters are specified. A synthetic heterogeneous K field is simulated using SGeMS Sequential Gaussian Simulation. This K-field, derived zones of porosity and defined boundary conditions comprise the simulation of a large confined aquifer. Two sets of synthetic observations obtained from forward models MODFLOW and MODPATH, heads and travel times, guide the calibration. Applying PEST++, the sequential approach performs two calibrations: a flow calibration using heads followed by a transport calibration using travel times. Both sets of observations are applied simultaneously in a single calibration run in the combined approach. Comparing final parameter estimates of each approach to the initial synthetic reality values shows that better results were achieved for both hydraulic conductivity and porosity with the combined approach. However, the sequential approach performed well with results falling within one standard deviation of the true value