A Kriging surrogate model coupled in simulation–optimization approach for identifying release history of groundwater sources

• Kriging model is proposed to replace the complex simulation model.
• Kriging model is applied in both homogenous and heterogeneous media, as well as steady state and transient flow conditions.
• Kriging model shows its high accuracy compared with ANN model.
• The computation time by Kriging model is 10 times faster than non-surrogate model.
• Kriging model shows its advantages using measurement error concentrations.

As the incidence frequency of groundwater pollution increases, many methods that identify source characteristics of pollutants are being developed. In this study, a simulation–optimization approach was applied to determine the duration and magnitude of pollutant sources. Such problems are time consuming because thousands of simulation models are required to run the optimization model. To address this challenge, the Kriging surrogate model was proposed to increase computational efficiency. Accuracy, time consumption, and the robustness of the Kriging model were tested on both homogenous and non-uniform media, as well as steady-state and transient flow and transport conditions. The results of three hypothetical cases demonstrate that the Kriging model has the ability to solve groundwater contaminant source problems that could occur during field site source identification problems with a high degree of accuracy and short computation times and is thus very robust.