Characterization of groundwater contaminant sources by utilizing MARS based surrogate model linked to optimization model

Abstract

Unknown groundwater contaminant source characterization is the first necessary step in the contamination remediation process. Although the remediation of a contaminated aquifer needs precise information of contaminant sources, usually only sparse and limited data are available. Therefore, often the process of remediation of contaminated groundwater is difficult and inefficient. This study utilizes Multivariate Adaptive Regression Splines (MARS) algorithm to develop an efficient Surrogate Models based Optimization (SMO) for source characterizing. Genetic Algorithm (GA) is also applied as the optimization algorithm in this methodology. This study addresses groundwater source characterizations with respect to the contaminant locations, magnitudes, and time release in a heterogeneous multilayered contaminated aquifer site. In this study, it is specified that only limited concentration measurement values are available. Also, the contaminant concentration data were collected a long time after the start of first potential contaminant source(s) activities. The hydraulic conductivity values are available at limited locations. The performance evaluation solution results of the developed MARS based SMO for source characterizing in a heterogeneous aquifer site with limited concentration measurement, parameter values, and under hydraulic conductivity uncertainties are shown to be satisfactory in terms of source characterization accuracy.