Modeling seawater intrusion with multiple sources of uncertainty: Application to Kuwait City

Variable-density flow models are widely used to simulate seawater intrusion (SWI), serving as essential tools for groundwater management in coastal regions. However, reliable analyses require addressing uncertainties in model inputs and their impact on outputs. Performing uncertainty analysis is computationally demanding due to multiple uncertainty sources and high parameter dimensionality. Existing applications and approaches of uncertainty for SWI are restricted to hypothetical scenarios. This paper proposes an efficient strategy for uncertainty analysis of SWI that can be applied at the field scale. This strategy begins with a preliminary screening technique to reduce the number of uncertain parameters by identifying the most influential ones. A global sensitivity analysis (GSA) is then conducted to rank these parameters by their significance. The GSA leverages sparse polynomial chaos expansion and Sobol indices. A specific method is introduced to handle uncertainty for time-dependent parameters such as pumping rate. Stochastic simulations are carried out using the surrogate model to quantify the uncertainties. The proposed methodology is applied to a coastal aquifer in Kuwait City. Results indicate that in natural conditions, the most influential parameters are aquifer permeability and the regional groundwater inflow, consistent with results from previous hypothetical studies. Results for both current and projected years, under uncertain pumping conditions, indicate that pumping rate and well locations are the most critical parameters. The stochastic simulations show that 10% uncertainty in the most influenceable parameters leads to more than 50% uncertainty on salinity predictions. This study provides an operational tool for managing groundwater resources in coastal areas.