Optimization for LNAPL remediation with considering multiphase flow behavior at Deok-so site, Korea

Non-aqueous phase liquids (NAPLs) are typical organic contaminants for soil and groundwater that do not dissolve in water. Especially Light non-aqueous phase liquids (LNAPLs), which has lower density than water, are focused in this study. LNAPLS are composed of volatile organic compounds (VOCs) such as benzene, toluene, and other aromatic hydrocarbons. For removing volatile compounds, soil vapor extraction (SVE) method is commonly engaged. Implementation plan of the SVE has a significant influence on the efficiency of contamination removal. Therefore, in this study, we proposed an optimization workflow by combining numerical and statistical surrogate modeling approach. For the numerical modeling, TOUGH2/TMVOC simulator, which can simulate the three-phase flow system (gas-water-NAPL), was used as a full-physics model for predicting the transport of LNAPLs. Hydrogeological properties of the model were characterized based on the Deok-so experimental Farm, Korea, where a field-property survey is progressed simultaneously. However, this full-physics model is computationally expansive for conducting the optimization analysis, which needs to simulate a large number of cases. Then, the statistical surrogate model was developed by using of design of experiment (DoE) and response surface methodology. The Box-Behnken design was engaged to generating design table based on six input factors, which took into account the locations of two wells (vapor injection well and extraction well for remediation), the bottom-hole pressure of extraction well, the well screen depth of two wells, and the steam injection rate. With the simulation results based on the design table, two response surface equations (RSEs) for the residual VOCs mass and the Gas/Oil phase ratio, which can indicate the removal of the LNAPL, are generated. Two RSEs showed sufficient predictability with R² of 0.99 and 0.97, respectively. For expanded research, these RSEs will be used as objective functions for determining the optimum implementation plan with considering model heterogeneity and fracture distribution based on the measured data from Deok-so site.

This research was supported by The SEM projects through the Korea Environmental Industry and Technology Institute (KEITI) funded by the Ministry of Environment (Grant Number: 2018002440003)