Variations in phosphorus (P) speciation were compared for two types of pilot scale wetlands: a blast furnace slag-based constructed wetland (SCW), and a gravel-based constructed wetland (GCW). Synthetic secondary effluent was used as the influent of the wetlands, which contained 1.1 mg/L P with more than 95% present as soluble reactive P (SRP). However, dissolved organic phosphorus (DOP) and particulate phosphorus (PP) emerged in the water along both wetlands. The levels of these three P species varied between the two wetlands. The GCW was more efficient than SCW at removing SRP, but showed a different trend. SRP decreased continually in the SCW, while it increased at the end of the GCW due to biological release. DOP was constant in SCW and GCW, and the mean value was 0.023 and 0.020 mg/L, respectively. The mean values of PP in the GCW ranged between 0.093 and 0.216 mg/L; much higher than the 0.05 l 0.01 mg/L measured in the SCW. Sequential extractions showed that iron bound PP (Fe-PP), aluminum bound PP (Al-PP), organic PP (Org-PP) and occluded PP (Oc-PP) were the major components of PP at most locations of the GCW. Fe-PP decreased from 0.53 to 0.14 mg/L in the upper layer, with DO steady at about 0.15 mg/L at the bottom. Oc-PP increased at each layer. Al-PP and Org-PP were steady in the first 140 cm of the GCW, but decreased sharply at the end. Considering the variation of SRP, DOP and PP fractionations, it can be seen that PP exchanged intensively with SRP in the GCW, and might act as an intermediate in the P removal process. Part of the SRP was first transformed into PP, and then absorbed by substrates or deposited in the wetlands. P removal was mainly via Ca precipitation in the SCW, but involved multiple mechanisms in the GCW, such as precipitation, adsorption and biological interactions. The multiple P removal mechanisms might be the reason for the low proportion of SRP, and the better P removal efficiency observed in the GCW.