Application of the self-organizing map (SOM) to assess the heavy metal removal performance in experimental constructed wetlands

The self-organizing map (SOM) model was applied to elucidate heavy metal removal mechanisms and to predict heavy metal concentrations in experimental constructed wetlands treating urban runoff. A newly developed SOM map showed that nickel in constructed wetland filters is likely to leach under high conductivity in combination with low pH in winter. In contrast, influent pH and conductivity were not shown to have clear relationships with copper (Cu) concentrations in the effluent, suggesting that the mobility of Cu was not considerably affected by salt increase during winter. The accuracy of prediction with SOM was highly satisfactory, suggesting heavy metals can be efficiently estimated by applying the SOM model with input variables such as conductivity, pH, temperature and redox potential, which can be monitored in real time. Moreover, domain understanding was not required to implement the SOM model for prediction of heavy metal removal efficiencies.