The performance of 14 large pilot-scale horizontal flow wetlands (similar to 60 m(2)each) designed to treat polluted river water was tested in a 3-year study at the Chenshan Botanical Garden in Shanghai, China. Five local species,Arundo donax,Cyperus alternifolius,Phragmites australis,Thalia dealbataandTypha orientalis, were planted in monocultures and in a polyculture ofPhragmites,ThaliaandTypha.Total nitrogen (TN) removal was compared among species and to an unplanted bed.Each bed was replicated once, and the parallel setup received a TN inflow concentration of 16 mg L(-1)on average over each summer season, a level twice as high as for the beds receiving unspiked inflow. During the first 2 years of operation, the only significant differences in pollutant removal were between planted and unplanted systems. In the third year, significant differences appeared among planted beds in the high nutrient systems, withPhragmitesbeing the most efficient species in TN removal andArundothe least, suggesting that greater inflow and more maturity may be needed before differences can be detected. The polyculture was never significantly more efficient than the other planted systems. However, it ranked among the best systems in 2018. Yet, while a greater removal of the polyculture over the average TN removal of all monocultures taken together may develop over time, we found no evidence that the polyculture may get more efficient than the best monoculture. Observation over all phases of plant establishment made it possible to draw conclusions for future design and operation.