Background The rapid development of urbanization has increased the impervious surfaces of cities, which has caused serious problems such as flooding. Green roofs, as one of the common practices for the Sponge City construction, have been increasingly gaining popularity in urban stormwater management and ecosystem restoration.

Methods Based on observations of rainfall-runoff processes of 22 rainfall events on 6 Sedum lineare green roofs located in Beijing with 3 types (i.e., local planting soil, engineered soil, and light growing medium) and 2 depths (i.e., 10 cm and 15 cm) of substrates in 2017, the influence of types and depths of substrates on the runoff management performances of green roofs were analyzed using runoff reduction rate, peak discharge reduction rate, and delayed times in runoff generation and peak discharge as the stormwater management performance indices.

Results The average runoff and peak discharge reduction rates, and average delayed timein runoff generation on the local planting soil and engineered soil green roofs were all lower than that of light growing medium green roofs. The average runoff reduction rates of light growing medium, local planting soil, engineered soil and green roofs were 82.2%、88.4%, and 89.4%, respectively, the average peak discharge reduction rates were 55.9%, 65.6%, and 63.9%, respectively, the average delayed time in runoff generation were 48.8, 106.2, and 163.9 min, respectively. However, the local planting soil was not recommended to be used as green roof substrate because of its high bulk density and poor hydraulic conductivity. The hydrologic performances of green roofs with 15 cm depth of all 3 types of substrates were higher than that of the green roofs with 10 cm depth of corresponding substrates. The impacts of types and depths of substrates on the hydrological performance of green roofs changed with rainfall conditions. For the 11 rainfall events < 10 mm, the 6 green roofs did not show obvious difference in hydrological performances, with all the runoff reduction rates of them greater than 95%. For the rainfall events larger than 10 mm, however, the hydrological performances of green roofs with local planting soil and engineered soil as substrates were better than that of the light growing medium green roofs. For rainfall events >10 mm, increasing the depth of substrate layers from 10 cm to 15 cm for all the three types of substrates could significantly improve the runoff mangement performances of green roofs (P < 0.05).

Conclusions Engineered soil and local planting soil were more excellent than light growing medium in runoff reduction rates, peak discharge reduction rates and delayed time in runoff generation. The hydrologic performances on green roofs with 15 cm depth were higher than that on the green roofs with 10 cm depth. The results of this experiment provide scientific references for the hydrologic design and performance evaluation of green roofs in cities of northern China.