Backward Probability Model for Identifying Multiple Contaminant Source Zones Under Transient Variably Saturated Flow Conditions

Abstract

Groundwater contamination is one of the major concerns to public health and water resource management. Identification of contamination sources is essential for designing cost-effective groundwater remediation systems and minimizing risks of further contamination. Various numerical approaches have been applied to reconstruct spatiotemporal distributions of contaminant sources. Due to difficulties in obtaining source information, the backward probability model is computationally efficient; however, its applications have largely been limited to relatively simple problems such as two-dimensional or simplified three-dimensional steady-state saturated flow conditions with a single contaminant source. More importantly, the traditional backward probability model requires some a priori knowledge on the suspected source locations or source release times to quantify source area location probabilities. Here we improve the capability of the backward probability model for solving real-world problems by implementing it into a three-dimensional transient variably saturated flow model. To evaluate the applicability of the model, we apply it combined with the shift-average method to a field-scale problem for identifying multiple contaminant sources occurring in the vadose zone that release contaminants to the saturated zone through infiltration. Compared with the suspected potential sources reported through a site characterization study, the enhanced backward probability model reasonably captures most of the reported potential sources under various hydrologic/hydrogeological/dispersion conditions, which demonstrates that the approach developed here can overcome limitations of the backward probability theory previously reported in the literature.

A Decision Support Framework for Pollution Source Detection via Coupled Forward‐Inverse Optimization and Multi‐Information Fusion

“…In multi‐source pollution scenarios, the sampled concentration is the merged concentration of several distinct point sources and it is difficult to discriminate among contaminant plumes (X. Wang & Zhang, 2021). Previous studies usually assumed that some pollution sources' characteristics are known priori, such as pollution source location or the number of pollution sources, to avoid discussing the pollution decomposition (Hamdi, 2017; Hwang et al., 2020; Singh & Datta, 2006). But in practical applications, they are as unknown as the other characteristics of unknown sources.…”

“…Wang & Zhang, 2021). Previous studies usually assumed that some pollution sources' characteristics are known priori, such as pollution source location or the number of pollution sources, to avoid discussing the pollution decomposition (Hamdi, 2017;Hwang et al, 2020;Singh & Datta, 2006). But in practical applications, they are as unknown as the other characteristics of unknown sources.…”

General Backward Model to Identify the Source for Contaminants Undergoing Non-Fickian Diffusion in Water

“…1,2 PSI has remained a research topic and has been applied extensively in hydrology for four decades, including delineation of groundwater protection zones, 3,4 identification of responsible parties, 5,6 assessment of aquifer vulnerability, 7 recovery of the contaminant history, 8 calculation of groundwater ages, 9,10 and identification of pollutant sources in water 11,12 or soil. 13,14 Source-identification problems have also been popular in other disciplines related to water and environments, such as oceanic sciences where backward-intime models were used to backtrack moving sea ice, ocean plankton, oil slicks, and marine debris, 15−17 atmospheric sciences where the models were used to track the source for airborne pollutants, 18,19 and other applications such as to track heat conduction or fish sources. 20,21 PSI usually requires quantitative analyses, which involve chemical techniques (such as isotope signatures 22,23 and molecular markers 24,25 applicable for specific contaminants), statistical analysis, 26,27 or process-based physical/mathematical techniques (such as forward-or backward-in-time modeling of dissolved contaminant transport).…”

“…Pollutant source identification (PSI) in surface and subsurface water focuses on using “results” (i.e., observed pollutant distributions) to find “causes” (such as the pollutant source location or release time). , PSI has remained a research topic and has been applied extensively in hydrology for four decades, including delineation of groundwater protection zones, , identification of responsible parties, , assessment of aquifer vulnerability, recovery of the contaminant history, calculation of groundwater ages, , and identification of pollutant sources in water , or soil. , Source-identification problems have also been popular in other disciplines related to water and environments, such as oceanic sciences where backward-in-time models were used to backtrack moving sea ice, ocean plankton, oil slicks, and marine debris, − atmospheric sciences where the models were used to track the source for airborne pollutants, , and other applications such as to track heat conduction or fish sources. , …”