Data Space Inversion for efficient and uncertainty-informed contaminant source identification in aquifers.

Area of Science:

• Environmental Science
• Hydrogeology
• Data Science

Background:

• Groundwater is a vital drinking water source for over a third of the global population, crucial for urban water security.
• Urban aquifers face increasing threats from subsurface contamination, leading to significant environmental, societal, and economic costs.
• Identifying contaminant sources and their discharge rates in urban settings is complex due to overlapping plumes, heterogeneous geology, and sparse monitoring data.

Purpose of the Study:

• To develop an efficient and reliable method for contaminant source identification in groundwater under hydrogeological uncertainty.
• To address the computational challenges of conventional inverse modeling in high-dimensional urban aquifer systems.
• To enable uncertainty-aware decision support for managing groundwater contamination.

Main Methods:

• Integration of Data Space Inversion (DSI) with an iterative Ensemble Smoother (iES) for contaminant source characterization.
• Application of the DSI+iES workflow to a synthetic heterogeneous aquifer simulating an urban worst-case scenario with scarce data.
• Comparison of DSI+iES performance against conventional parameter-space inversions and physics-based transport inversions.

Main Results:

• The DSI+iES method achieved probabilistic discrimination of active and inactive contaminant sources and estimated source mass discharge distributions.
• A computational gain of up to 15 was observed compared to conventional parameter-space inversions, enabling retention of subsurface heterogeneity and uncertainty propagation.
• Data-worth analyses revealed that strategically placed monitoring wells in areas of high predictive uncertainty provide the most significant information gains.

Conclusions:

• The DSI+iES workflow offers a computationally efficient and reliable approach for contaminant source identification in data-limited urban environments.
• This method enhances the practical applicability of subsurface contamination management by coupling uncertainty-aware inference with high-dimensional system representation.
• The study provides a foundation for field-scale source identification and adaptive monitoring design in settings requiring timely, cost-efficient decisions.