An ensemble machine learning method for δ18O prediction in groundwater has revealed significant mixing between modern Nile River water and ancient paleowater from the Nubian Sandstone Aquifer in Egypt’s Eocene limestone aquifer. This research highlights the complex hydrochemical heterogeneity of groundwater sources, driven by various geochemical processes and distinct recharge pathways. The findings provide a framework for understanding isotopic dynamics within the aquifer system.
The study utilized Python 3.9 within Visual Studio Code, employing core libraries such as NumPy and Pandas for data handling, Scikit-learn for model development, and SciPy for statistical analysis. Matplotlib and Seaborn were used for data visualization. This computational approach enabled detailed analysis of hydrochemical and isotopic data.
Groundwater samples exhibited wide concentration ranges and high coefficients of variation for major ions and total dissolved solids (TDS), indicating substantial hydrochemical heterogeneity. This variability stems from processes like salinization through rock-water interactions and the mixing of recharge waters with differing chemical compositions. The observed heterogeneity in major ion chemistry mirrors the variability in δO and δH values, confirming that both reflect the same recharge pathways and geochemical processes.
Measured δO values in groundwater samples ranged from -10.18‰ to +2.88‰, while δH values ranged from -78.48‰ to +25.29‰. These ranges encompass the isotopic signatures of Egypt’s two primary recharge sources: modern Nile River water and deep-seated paleowater stored in the Nubian Sandstone Aquifer. The relationship between δH and δO indicates that the Eocene limestone aquifer receives recharge from both sources, with isotopic compositions reflecting a blend of recent infiltration and upward leakage from deeper formations.
To quantify the relative contribution of paleowater, researchers applied a two-endmember mixing model. This model calculates the proportion of Nubian aquifer water in a sample based on its measured isotopic value, the Nubian Sandstone endmember, and the Nile water signature. The contribution from Nile water is then derived from this calculation.
Results show substantial spatial variation in recharge composition across the study area. In zones adjacent to the Nile floodplain, modern water predominates, with Nile contributions reaching up to 99%. Conversely, distal areas on the elevated limestone plateau show minimal influence from recent recharge, indicating a much lower proportion of Nile water.
The precise long-term implications of varying recharge compositions for water resource management in Egypt remain an area of ongoing study. Future research will likely focus on how climate shifts and human activities might alter these mixing dynamics, potentially impacting water availability and quality. Monitoring these isotopic signatures will be crucial for sustainable aquifer management.
Further investigation into the specific geochemical processes driving salinization and their interaction with recharge patterns could refine predictive models. Understanding the resilience of the Eocene limestone aquifer to changes in recharge sources will be key for regional water security planning.