School of Mechanical Engineering Ziv Moreno
School of Mechanical Engineering Seminar
Monday, November 5, 2018 at 14:00
Wolfson Building of Mechanical Engineering, Room 206
Stochastic Modeling of Pollutant Transport under Remediation
PhD student of Dr. Amir Paster and Prof. Gedeon Dagan
Prediction of pollutant transport in aquifers is necessary for decision making regarding a suitable remediation strategy and risk assessment for production wells. Prediction is usually conducted based on a calibrated numerical model, which couples the water mass balance and Darcy's equations for water flow, and the advection dispersion equation for pollutant transport. However, due to the scarcity of data, the spatial variability of the hydraulic conductivity, which has a pronounced effect upon pollutant transport, is subject to uncertainty.
The aim of this study was to provide prediction of the pollutant plume evolution in space and time under remediation while considering the uncertainty which stems from the unknown structure of the aquifer. An actual polluted site, located at the Eastern borders of the Israeli Coastal Plain Aquifer, served as a case study.
The lithological units' spatial correlations were identified based on lithological data from surrounding boreholes. A Monte-Carlo approach was adopted and multiple realizations of the aquifer structure were constructed. Two geostatistical methods were considered: discrete (lithological unit); and continuous (hydraulic conductivity). Once constructed, realizations were implemented into a flow and transport model, allowing the quantification of the expected remediation efficiency and its associated uncertainty.
First, we considered a simple case of vertical heterogeneity. The two geostatistical methods were compared with a semi-analytical solution of the problem. The semi-analytical solution was in good agreement with the complex numerical simulations, therefore showing that it can be used for an initial estimation of the remediation efficiency.
For a more complex 3D heterogeneity case, the discrete approach was adopted. Here, a forward predictive model, based on measurements of hydraulic head and concentrations, was used to reconstruct the plume shape. Remediation efficiencies, as well as risk analysis for contamination, were evaluated from the ensemble of realizations. We noticed that by reconstructing the plume shape, the uncertainty regarding the pollutant fate was significantly reduced.
Finally, a new approach for stochastic inversion was developed, combining heuristic (genetic algorithm) and geostatistical methods. The approach was able to generate an ensemble of solutions with a better agreement with measurements while significantly reducing the computational effort.
Moreno, Z. and A. Paster (2017), Prediction of Remediation of a Heterogeneous Aquifer: A Case Study, Groundwater, 55, 428-439.
Moreno Z. and A. Paster (2018), Prediction of pollutant remediation in a heterogeneous aquifer in Israel: Reducing uncertainty by incorporating lithological, head and concentration data. Journal of Hydrology, doi: http://doi.org/10.1016/j.jhydrol.2018.07.012.
Moreno Z. and A. Paster (2018), A Genetic Algorithm for Stochastic Inversion in Contaminant Subsurface Hydrology, submitted to Water Resources Research (under review).