Quantifying and Mapping Saltwater Intrusion in a Complex Urban Environment: A Universal Kriging Approach

Abstract

Unsustainable groundwater exploitation in coastal aquifers has resulted in significant drops in freshwater heads and the promotion of seawater intrusion (SWI) into affected coastal aquifers. With nearly two-thirds of the world’s population living in coastal cities, SWI is a global problem impacting many coastal aquifers worldwide. Several modelling frameworks have been developed over the years and implemented for mapping SWI; yet they have performed poorly in complex urban heterogeneous karst environments with limited data. In the current study, a Universal Kriging (UK) model is proposed and developed for the city of Beirut, Lebanon. The developed model was successfully able to capture the main socio-economical, geological/geographical, and land-use drivers of SWI in the study area, while at the same time explaining the spatial correlation structure in the sampled data. The developed UK outperformed a landuse regression (LUR) model across a set of selected model performance metrics (NSE 0.71 vs 0.54; RMSE 0.44 vs 0.93), while at the same time significantly reducing prediction uncertainties across the spatial domain by more than 20 %. Moreover, model validation results showed that both the UK and LUR models were significantly more robust and showed higher prediction skills as compared to previously developed kriging and GIS-based groundwater vulnerability models for the same study area. Overall, the predictive maps generated from the UK model clearly highlighted the severity of SWI in Beirut, whereby more than 55% of the groundwater aquifer in the city was predicted to hold brackish to saline waters. The model results also showed clear large-scale and fine-scale spatial variabilities in the predicted salinity levels, while also identifying regions suffering from upconing and those that appear to be benefiting from potential freshwater karst conduits or from local elevated recharge rates. The results of this study provide a promising modelling and mapping approach that can be implemented to map SWI in other poorly monitored urban coastal aquifers and to better understand their vulnerabilities to anthropogenic stressors.