The aim of this study was to explore the efficiency of three horizontal subsurface flow (HSF) constructed wetlands (CWs) to remove S-metolachlor from contaminated water. The three CWs, with code names MG-R, MG-C and MG-Z, all contained medium gravel as porous media. MG-R and MG-C were planted with Phragmites australis and Typha latifolia, respectively, and MG-Z was left unplanted and used as control unit to study the effect of vegetation. The CWs units in this experiment operated continuously for one year but they were mature systems operating in various experiments for more than fourteen years. The results showed that the mean percent S-metolachlor removal values were 68.9%, 47.8% and 40.8% for MG-R, MG-C and MG-Z units, respectively. The vegetation (Phragmites australis) and the hydraulic residence time were found as the most important parameters affecting the CW capacity in herbicide removal. The current data suggest that HSF CWs are an efficient and low-cost technology in treating S-metolachlor-contaminated surface runoff in agricultural areas.