The Value of the Groundwater Age Observation in Characterization of Local and Regional Groundwater Systems: An Inverse Model Study Performed in San Joaquin Valley, CA

The purpose of this study was to understand how the addition of observation data density and type (e.g., head and age) can improve the inverse solution in general and in particular setting of a regional groundwater model, and potentially reduce its inherent non-uniqueness. To this aim an inverse modeling study was performed using groundwater head and age observations. The study was applied into two system settings; the first one was an idealized 2-D aquifer system, and the second one was a real-world aquifer located in San Joaquin Valley, CA. The forward modeling of groundwater flow and transport was performed using a finite element solver called COMSOL for the idealized study and MODFLOW/MODPATH for the real-world aquifer. The inverse modeling was performed using UCODE 2005.;The results of this study showed that the addition of groundwater age data could improve the parameter estimation in both idealized and real-world aquifer settings. The sensitivity analysis performed in the real-world aquifer indicated that the head data acts as a local observation for the estimation of hydraulic conductivity. In contrary, the inferred age observations could provide information useful to the aquifer characterization even beyond their local vicinity. The implication is that for inverse calibration of hydraulic condition, age, that reflects advective-dispersive transport and thus information along streamlines, may serve as a more powerful data type than head, which reflects local conditions through the diffusion operator. Nevertheless, the location of the observation points along the streamlines is of high importance to the ultimate value of age data. The parameter estimation results indicated that the addition of age data resulted in a slight increase in the calculated variance. This is expected because the age observations carry a much higher uncertainty than the head observations, thus the calculated error variance associated to the age observations is higher in nature than those associated with head observations. However, the sum of absolute residuals in head observations in most cases decreased as a result of addition of age observation to the optimization. These results are not typically expected and tell about the potential power of age data in flow model inversion.