Phosphorus (P) removal efficiency was compared between two types of wetlands: a blast furnace slag (BFS)-based constructed wetland (SCW) and a gravel-based constructed wetland (GCW). Both wetlands were 2.4m long, 0.4m wide, and 0.65m high. Synthetic water to be reclaimed contained 1.1mg . L-1 P. Hydraulic retention time (HRT) was set at 1 day in the first four phases and then increased to 2 days in the last two phases. Results showed that P removal with 1 day HRT application decreased from 80.4 to 56.4% in SCW and from 73.0 to 61.6% in GCW. With the increase of HRT, P removal decreased continually to 54.6% in SCW while it increased to 88.6% in GCW. The lower amount of P removed by SCW in the latter phases was mainly the result of substrates. P adsorption capacities did not show much difference between BFS and gravel in the batch experiment with a low initial P concentration less than 1.4mg . L-1. However, chemical properties of substrates presented an obvious impact on P removal in the pilot experiment. Scanning electron microscopy (SEM) and electron dispersive spectrometry (EDS) analyses revealed that slag surfaces contained 1.07% P, which precipitated with CaCO3, whereas gravel surfaces contained 2.12% P, which was bound to Fe and Al oxides. P removal in SCW decreased synchronously with the pH decrease of 9.82 to 9.04. Decrease of pH and low P concentration was the basic reason for the formation of CaCO3, which led to the poor P removal. P accumulated by plants also influenced the P removal in SCW and GCW, which accounted for 9.63 and 16.17%, respectively. Besides, phosphatase activity in the SCW increased in the last two phases, which brought a large P release. This article shows that P removal by substrate varied with the initial P concentration and was influenced largely by pH in the wetlands. BFS used in this experiment is not a good substrate candidate for wetlands to treat reclaimed water.