Polynomial Chaos Expansion-based Global Sensitivity Analysis Workflow for LNAPLs Remediation Efficiency
Abstract
Soil and groundwater contamination problems incurred by light non-aqueous phase liquids (LNAPLs) spilled into the subsurface have been consistently raised. To deal with such problems, analysis through a simplified numerical model for the contaminated sites has been widely conducted. However, considering complex hydrogeologic properties, dynamic systems governing the LNAPLs behavior, and 3-dimension, the computational power and CPU time required for the analysis through numerical models increased explosively. Hence, the conventional numerical approaches were limited to conduct quantitative analysis necessitating a large amount of data. In this study, Sobol global sensitivity analysis, one of the quantitative analyses, was enabled by adopting the surrogate polynomial chaos expansion (PCE) method which expresses a mathematical formula from estimating the results to be calculated through a numerical model. In this study, a conceptual model was built assuming that the remediation through multi-phase extraction (MPE) and steam injecting (SI) well was operated after benzene, toluene, ethylbenzene, and xylene-p (BTEX) was spilled, and then, the removal efficiencies were computed. Sequentially, the developed surrogate PCE model estimated the removal efficiency with high predictability substituting the numerical model. Finally, the quantitative analysis was conducted from a large amount of data estimated from the surrogate model. As a result, regardless of the magnitude of permeability, the depth of the MPE well was the most important factor determining the removal efficiency. The removal efficiency was increased as the depth of the MPE well was installed below the groundwater. In addition, at low permeability, the removal efficiency did not change significantly depending on the well configuration factors, showing that most of the estimated removal efficiencies were close to zero. This study demonstrated the possibility to quantitatively evaluate factors influencing the improvement of LNAPLs removal efficiency and estimate the results without iterative runs of complex numerical models.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2021
- Bibcode:
- 2021AGUFM.H15R1256K