Constructed wetland is an important approach of removing arsenic in water, and substrates are key factors determining the removal efficiency. We researched the efficiency and impact factors of arsenic removal through adsorption experiments (adsorption kinetics, adsorption isotherm, and influence factors) on four substrates (gravel, manganese sand, zeolite, and ceramsite). Results reveal that the arsenic adsorption is time-dependent and could reach equilibrium in 24 hours for all the four substrates. The adsorption process consists with the pseudo-first-order kinetic model and pseudo-second-order kinetic model. The maximum arsenic adsorption capacity are as follows: manganese sand (36.62 mg/kg), ceramsite(25.39 mg/kg), zeolite (11.96 mg/kg), and gravel (7.04 mg/kg). The arsenic adsorption capacity of all tested substrates could best explained by the Freundlich adsorption isotherm. Given the range from 0.25mm to 0.50mm, the grain size of substrate has insignificant impact on the adsorption capacity of manganese sand and zeolite. The presence of ammonia-nitrogen (concentration 0.50-2.50 mg/L) shows negligible effect on arsenic adsorption capability. When the arsenic concentration is lower than 0.4 mg/L, the effect of phosphate (concentration 0.25-0.50 mg/L) is insignificant; when arsenic concentration is higher than 0.4 mg/L, the maximum arsenic adsorption capacity decreased significantly (ceramsite decreased 2.57 mg/kg and manganese sand decreased 1.85 mg/kg) when phosphate concentration rose from 0.25 mg/L to 0.50 mg/L.