Exploring Predictive Uncertainty at a Double‐Source Managed Aquifer Recharge Site via Stochastic Modeling

Abstract

AbstractIn Menashe Streams managed aquifer recharge (MAR) site (Israel) desalinated seawater (DSW) is recharged since 2015, alongside ephemeral stream flows. This study quantifies the uncertainty of predictions of the mixing of these two water sources in the aquifer. Mixing estimations are based on the significant difference in the content of stable water isotopes between the water sources. Uncertainties driven from aquifer heterogeneity and climate variability are compared. We use realistic flow and isotope‐transport models in a multiple‐realization stochastic approach considering space and time for the two drivers of uncertainty. Predictive uncertainty is evaluated via comparison of the temporal coefficient of variation of four realization ensembles. Results show that the impact of subsurface structure uncertainty on the predictions is small, compared to the uncertainty resulting from variability related to future hydrometeorological conditions. A generalized conclusion from this result is the difficulty to make long‐term predictions of the mixing ratios, regardless of the level of certainty in the subsurface structure interpretation, when one source of recharge water has significant annual fluctuations. Implication on prediction of magnesium concentration is demonstrated as an example for prediction uncertainty of concentration of a solute of interest in one of the MAR sources. Furthermore, we show that considering a single source MAR site, uncertainty in mixing of the MAR water and native groundwater in wells upgradient of the recharge location is higher than prediction uncertainty in wells located downgradient. Insights are also drawn regarding the change in uncertainty with distance from the recharge pond of the DSW.