In order to explore the approaches of ecological engineering for a non-point source pollution caused by paddy field runoff in the Sanjiang Plain, and to seek suitable wetland plants which are appropriate to be used under the local natural conditions in constructed wetlands two subsurface flow constructed wetlands (SSFCWs), 2.0 m in length, 0.5 m in width and 0.65 m in depth, were set up by selecting the local soil and slag (3 - 5 mm in diameter) as the substrates and transplanting two typical natural marsh wetland plants, Calamagrostis angustifolia and Phragmites australis. The wetlands were dosed in batch operations using synthetic wastewater without the addition of any organic carbon but with the addition of ammonium-N (30.32 mg/L), nitrate-N (28.74 mg/L) and phosphate-P (10.15 mg/L) as major pollutants. The results suggested that both of the two plant species maintained normal growth in the wastewater, which demonstrates their good pollutant resistance. During the experimental period, the relative growth rate of Phragmites australis was 0.042 d-1, which was 2.82 times higher than that of Calamagrostis angustifolia. During a 7-day wastewater retention, an average TN removal rate for Calamagrostis angustifolia and Phragmites australis of 65.1% and 99.6%, respectively, and a TN removal load of 1.66 g?m-2?d-1 and 2.53 g m-2 d-1, respectively, were obtained. The wetland with Phragmites australis demonstrated a higher nitrogen removal efficiency. In the course of direct nutrient uptake, the contribution ratios to TN removal achieved by Calamagrostis angustifolia and Phragmites australis were 14.7% and 61.7%, respectively; whereas for TP, the ratios were 11.7% and 12.9%, respectively. Apart from its uptake rate, the accumulation and concentration of nitrogen in the plants were higher than those of phosphorus, whereas a significant difference of nutrient assimilation between different plants was observed. The nitrogen uptake rate of Phragmites australis was 0.519 g m-2?d-1, i.e. 3.94 times higher than that of Calamagrostis angustifolia. The phosphorus uptake rate of Phragmites australis was nearly the same as that of Calamagrostis angustifolia (0.012 g?m-2?d-1 and 0.011 g?m-2?d-1, respectively). The concentrations of TN and TP in the biomass of Phragmites australis were measured as 29.2 mg?g-1 and 3.41 mg?g-1, respectively, which were 1.83 and 1.49 times as those of Calamagrostis angustifolia. The N and P content in the leaves was found to be higher than that of the stems, for both Calamagrostis angustifolia and Phragmites australis. For Calamagrostis angustifolia, the N and P concentrations in the leaves were 3.66 and 1.03 times higher than those of the stems, and for Phragmites australis, the N and P concentrations in the leaves were 3.17 and 0.68 times higher than those of stems, respectively. The N and P assimilation capacity of Calamagrostis angustifolia was lower than that of Phragmites australis. The removal effectiveness of NH4 -N exceeded that of NO3--N in each of the SSFCWs. A visible segregation effect of purification for nitrogen and phosphate pollution was found in SSFCWs, and the purification capacity of the top layer was higher than that of the bottom layer in each wetland. The pH values of the top layer of the SSFCWs were lower than those of the bottom layer during the operation process of the SSFCWs. The variation curves of the pH values decreased firstly and then increased. The NH4 -N concentration affected the pH values markedly. The break-point of the pH curve was suggested to be an indicator of the end of the NH4 -N reaction. The results clearly demonstrated that constructed wetlands could be a potential technology for nutrient removal from paddy field runoff. The purification capacity of the wetland planted with Phragmites australis exceeded that of the wetland with Calamagrostis angustifolia.