Characterization of flow dynamics around highly-utilized agricultural wells in a fractured-rock aquifer: Assessment of uncertainties lying on groundwater age-dating
Abstract
When environmental tracers are used to determine groundwater age distributions, various factors including measurement error, model structural error, and sample contamination can affect the outcome. In this study, we tried to untangle and explain the uncertainties associated with the inference of age distribution for a set of highly-utilized production wells in a fractured-rock aquifer system. A total of 17 samples were collected at various depths (22-150 m) from production wells. These samples were analyzed for six age tracers, including 3H, tritiogenic 3He, radiogenic 4He, CFC-11, CFC-12, and CFC-113. A Bayesian approach was used for evaluating age distribution for the samples based on several presumed lumped parameter models (LPM)s and shape free models. It was found that the measured data in most wells could be explained by traditional LPMs reasonably well. However, for five of the 17 wells it was found that the simple LPMs were not adequate and the more complex shape-free approaches resulted better based on the Deviance Information Criteria (DIC) measure. This looked attributed to 1) competition between densely located wells, 2) a thick vadose zone layer possibly causing preferential flows 3) fractures allowing a localized production. Furthermore, it was shown that three other factors including tracer contamination, non-uniform 4He production, and the heterogeneity in the travel time in vadose zone added uncertainties to groundwater age-dating. Overall results showed that the combination of Bayesian approach with DIC criterion can help to identify uncertainties associated with the age distribution for hydrologically complicated systems.
- Publication:
-
Journal of Hydrology
- Pub Date:
- July 2020
- DOI:
- 10.1016/j.jhydrol.2020.124885
- Bibcode:
- 2020JHyd..58624885J
- Keywords:
-
- Fractured aquifer;
- Age tracers;
- Bayesian inference;
- Uncertainty analysis;
- Lumped parameter models;
- Residence time distribution