The thermal performance of green roofs is usually site-specific and changes temporally. Hence, thermal performance evaluation is necessary to optimize green roof design and its cooling effect. In this paper, we evaluated the outdoor spatio-temporal performance of a full-scale extensive green roof (EGR) in Nanjing, China throughout a summer at three heights (30, 60and 120cm). We found the EGR exhibited an overall slight diurnal cooling effect at all three heights (-0.09, -0.23, and - 0.09 °C, respectively), but there was an obvious warming effect at a couple of specific hours during daytime. Especially on sunny days, the maximum warming effect at all three heights was 1.59, 0.59, and 0.38 °C, respectively. During the night, the EGR had a pronounced cooling effect of -0.63, -0.40, and - 0.15 °C, respectively. Among the weather scenarios, sunny days had the highest impact on the EGR's thermal performance, while effects were less pronounced on cloudy and rainy days. The average range of hourly air temperature difference at 30 cm between EGR and a bare roof on selected days was 4.02 (sunny), 2.67 (cloudy), and 0.74 °C (rainy). The results of multiple-regression analyses showed strong and significant correlations of air temperature difference between the EGR and a bare roof with differences in relative humidity, net radiation, several measures of soil and surface temperature, and soil moisture as well as average solar radiation, air temperature and wind speed. The results implied that both the components of the EGR, such as green vegetation and the soil substrate layer, and the microclimate created by the EGR can feed back and contribute to the thermal performance of an EGR. Through this full-scale EGR research in a subtropical monsoon climate, we provide the scientific basis and actionable practices for green roof planning and design to alleviate the urban heat island effect towards designing climate-resilient cities.