Constructed wetlands (CWs) are a promising technology to treat pesticide contaminated water, but its implementation is impeded by lack of data to optimize designs and operating factors. Unsaturated and saturated CW designs were used to compare the removal of triazole pesticide, tebuconazole, in unplanted mesocosms and mesocosms planted with five different plant species: Typha latifolia, Phragmites australis, Iris pseudacorus, Juncus effusus and Berula erecta. Tebuconazole removal efficiencies were significantly higher in unsaturated CWs than saturated CWs, showing for the first time the potential of unsaturated CWs to treat tebuconazole contaminated water. An artificial neural network model was demonstrated to provide more accurate predictions of tebuconazole removal than the traditional linear regression model. Also, tebuconazole removal could be fitted an area-based first order kinetics model in both CW designs. The removal rate constants were consistently higher in unsaturated CWs (range of 2.6-10.9 cm d-1) than in saturated CWs (range of 1.7-7.9 cm d-1) and higher in planted CWs (range of 3.1-10.9 cm d-1) than in unplanted CWs (range of 1.7-2.6 cm d-1) for both designs. The low levels of sorption of tebuconazole to the substrate (0.7-2.1%) and plant phytoaccumulation (2.5-12.1%) indicate that the major removal pathways were biodegradation and metabolization inside the plants after plant uptake. The main factors influencing tebuconazole removal in the studied systems were system design, hydraulic loading rate and plant presence. Moreover, tebuconazole removal was positively correlated to dissolved oxygen and all nutrients removal.
Keywords: Artificial neural network; Biofilm reactors; Contaminants of emerging concern; Fungicides; Phytoremediation.
Copyright © 2017 Elsevier Ltd. All rights reserved.