Abstract: Six field experimental devices have been used to study the performance of bioretention tanks with different fillers in purification of urban road runoff. Through field tests, mathematical statistics and numerical simulation, this study focused on the control effect of bioretention tanks under the influences of five major factors: filler types, filler combinations, filler thickness, pollutant characteristics, and vegetation conditions. Multi-level orthogonal tests on these factors show that the best way to remove pollutants is to select fly ash for the filler and a filter thickness of 45cm, along with the vegetation of boxwood combined with Ophiopogon japonicus. We developed multiple regression models of removal rate of the pollutants (total nitrogen and total phosphorus) versus the influencing factors. Our simulation of the migration of TP in the tanks in three inflow concentration scenarios using Hydrus-1D achieved satisfactory results of mean relative errors 16.51%, 13.03%, and 23.49% relative to the measured values.