Several small-scale surface flow constructed wetlands unplanted and planted (monoculture) with various macrophytes (Phragmites australis, Typha orientalis, Pennisetum purpureum, Ipomoea aquatica, and Pistia stratiotes) were established to continuously receive nitrate-contaminated groundwater. Soil characteristics and their effects on nitrate removal and soil denitrification were investigated. The results showed that planted wetland cells exhibited significantly higher (P < 0.05) nitrate removal efficiencies (70-99%) and soil denitrification rates (3.78-15.02 μ g N2O-N/g dry soil/h) than an unplanted covered wetland cell (1%, 0.11 μ g N2O-N/g/h). However, the unplanted uncovered wetland cell showed a nitrate removal efficiency (55%) lower than but a soil denitrification rate (9.12 μg N2O-N/g/h) comparable to the planted cells. The nitrate removal rate correlated closely and positively with the soil denitrification rate for the planted cells, indicating that soil denitrification is an important process for removing nitrate in constructed wetlands. The results of nitrogen budget revealed that around 68.9-90.7% of the overall nitrogen removal could be attributed to the total denitrification. The soil denitrification rate was found to correlate significantly (P < 0.01) with the extractable organic carbon, organic matter, and in situ-measured redox potential of wetland soil, which accordingly were concluded as suitable indicators of soil denitrification rate and nitrate removal rate in nitrate treatment wetlands.