Inference of long-term groundwater age transience using environmental tracers (Invited)

The estimation of groundwater age has received increasing attention due to its applications in assessing the sustainability of water withdrawal from the aquifers and evaluating the vulnerability of groundwater resources to near surface or recharge water contamination. It has been recognized that due to hydrodynamic mixing, it is more appropriate to represent groundwater age as a distribution (referred to groundwater age distribution) rather than a single number. Environmental tracers, including anthropogenically released or cosmogenic radioactively decaying tracers have been widely used for inference of groundwater age distribution. This is typically done by presuming a mathematical form for the age distributions and then estimating the defining parameters using a maximum likelihood-based approaches. One of the potential applications of groundwater age distribution that has not received adequate attention in the past has been evaluating the long-term temporal changes in aquifer residence time distribution as a result of changes in land use or climate. This is partly due to the fact that measurement of environmental tracers in water samples collected at a single time can only provide information about the age distribution of the parcel of water represented by those samples. Under transient flow condition this age distribution does not represent the residence time distribution in the aquifer at the time of sampling since the parcel of water has experienced a non-steady flow pattern as it moved through the aquifer. However groundwater samples collected at different times can potentially contain information about the transience of groundwater age distribution. In this presentation the relationship between steady-state aquifer residence time distribution and the transient groundwater age distribution will be discussed. Also evaluation of the effectiveness of using multiple tracers collected at different times to infer groundwater age distribution and its temporal variations is demonstrated through a case study that used CFCs, SF6 and 85Kr in water samples collected at multiple times with certain intervals. The inference of groundwater age distribution and the temporal variation in the regional flow is conducted using Bayesian inverse modeling through Markov Chain Monte Carlo method. The information content of the tracers under various conditions and the effects of uncertainties associated with concentrations of tracers as a result of measurement error, spatial heterogeneity, short-term temporal variations and model structural error will also be discussed. Observed tracer concentration and modeled concentrations based on inferred age distributions and temporal trends in groundwater flow.