With continued urbanization pressure, regulators and developers alike are increasingly looking to new forms of green infrastructure and low-impact development technologies as a means of appropriately integrating built infrastructure into the landscape. This paper describes the results of a series of experiments designed to simulate the hydrologic performance of green roofs under variable precipitation conditions. The experiments were designed in order to test performance under both steady, low-intensity rainfall, as well as under short duration, high-intensity rainfall conditions. A control membrane roof and prototype green roofs of 2.5-, 6.3-, 10.1-cm depths were subjected to simulated precipitation in a laboratory setting. The green roofs delayed, prolonged, and reduced the peak rates of green roof discharge to 22-70% that of a standard roof surface, with greater percent reductions associated with deeper roofs. Negligible discharge was observed from all of the prototypes during the first 10 min of simulated precipitation. Although the fate of the 0.35 cm of precipitation that were applied over this time period can only be determined through additional controlled testing of the prototypes with shorter duration rain events, the potential significance of green roofs that retain this quantity of water is discussed in the context of the historical New York City precipitation record. The results also indicated that nearly all of the precipitation applied was discharged as drainage over the 24 h period immediately following the experiment, suggesting that the percentage of large storms that are retained in green roofs may be insignificant. Green roof runoff coefficients computed from an analysis of the discharge hydrographs ranged from 0.2-0.7, consistent with other studies. Two approaches to predicting the observed discharge using the U.S. EPA's Stormwater Management Model (SWMM) are also presented. The "storage node" approach achieves better overall predictions than the "curve number" approach, which itself tends to significantly underpredict discharge from these systems. Although reasonable sets of predictions were eventually obtained, the selection of appropriate model parameters would not have been possible without the availability of experimental data with which to calibrate the models. The experimental results support the argument that the storm water benefits of green roofs could be significant. However, the writers urge caution in interpreting the results of green roof drainage discharge calculations made using SWMM until additional calibration and validation attempts have been performed.
