In this study, the performance of an irrigation-integrated green roof (GR) system is analyzed through numerical simulations using the vertically-resolved Princeton ROof Model (PROM). The simulations are driven by a 63-day series of summertime meteorological forcing measured at a GR site in Beijing, China. Due to the importance of the medium layer depth of GR systems, its effect on the dynamics of heat and moisture transports and consequently on the thermal performance is first examined under a no-irrigation scenario. The results confirm that the medium layer depth affects heat and moisture transports significantly, but non-monotonic trends emerge. A deeper layer is found to redistribute more water into the bottom section, thus limiting surface evaporation, while a thin layer does not store enough water, dries up fast, and decreases performance too. This indicates that an optimal layer thickness exists somewhere in the middle. Given a fixed medium layer depth, different irrigation scenarios are then investigated. Higher irrigation control limits (i.e. the soil moisture at which irrigation is initiated) enhance the thermal performance of the GR, but this enhancement plateaus at high limits. Based on these findings, a GR (constructed over an un-insulated concrete slab roof in Beijing) is simulated and a simplified operational cost-benefit analysis is performed by comparing the cost of irrigation to that of the energy saved due to lower air-conditioning (AC) requirement when the GR is wetted. The analysis indicates that the irrigated GR system costs are lower than the AC costs for an unirrigated GR. For instance, an irrigated GR system with an area of 100 m2 under an irrigation control limit of 0.3 m3 m−3 will save ￥110 (18 USD) for the chosen 63-day simulation period in Beijing. Therefore, irrigation-integrated GR systems emerge as potentially viable solutions for improving building energy efficiency in a temperate climate.