Modeling seawater intrusion under uncertainty: Application to the coastal aquifer of Kuwait City

Seawater intrusion (SWI) poses a critical threat to groundwater resources in arid coastal regions, particularly in countries with limited freshwater availability such as Kuwait. Groundwater in Kuwait is intensively exploited to supplement desalinated water, leading to progressive salinization and the closure of several production wells. Despite the severity of the problem, previous studies in Kuwait have mainly focused on water quality observations, with a lack of predictive, field-scale modeling frameworks that account for uncertainty.

 

This study develops a variable-density numerical model to simulate seawater intrusion in the coastal aquifer system of Kuwait City, focusing on the Al-Sulaibiya and Atraf aquifers. A two-dimensional vertical cross-section extending 8 km inland and 2 km offshore is considered. The model couples groundwater flow and salt transport, with fluid density represented as a linear function of salinity. The system is simulated under pre-development conditions and historical pumping scenarios from 1954 to 2021, followed by future projections up to 2050 assuming constant pumping rates.

 

Given the scarcity and uncertainty of hydrogeological data, an efficient uncertainty analysis strategy is proposed and applied at real field scale. The approach combines 3 steps; Plackett–Burman screening to identify the most significant parameters, global sensitivity analysis using Sobol indices to rank the parameters by order of importance, and uncertainty quantification based on Polynomial Chaos Expansion surrogate modeling to evaluate uncertainty of the output (salinity) assuming 10% uncertainty in the input parameters. The results highlight the most dominant parameters on the extent of the saltwater wedge, the aquifer permeability, and pumping rates. In some locations in the aquifer the 10% uncertainty in the input parameters can lead to more than 50% uncertainty in the salinity predictions.

 

The proposed framework significantly reduces computational cost while enhancing the model reliability estimates of uncertainty in SWI predictions. The results offer valuable insights for groundwater management and demonstrate the necessity of uncertainty analysis in modeling to support sustainable water resources planning in arid coastal environments.