Abstract Green roofs, integrated into conventional buildings, offer valuable thermal benefits, with extensive research and validated models focused on specific climates and summer conditions. However, their applicability requires further refinement, as conflicting year-round performance results have been observed. This study investigates the thermal performance of unvegetated green roofs with various depths in Melbourne, Australia, over almost a year. Utilizing Sailor’s green roof model, the research reveals reasonable accuracy for winter and deeper roofs, but notable discrepancies for 100 mm thick green roofs during the day. Additionally, the study provides experimental data on non-vegetated green roof top temperatures of a typical building in South-East Australia. Night and day comparisons are emphasized to identify model limitations due to different energy balance mechanisms of green roofs. Despite recognizing the green roof substrate’s role in regulating heat flux, few studies independently examine its thermal benefits. Implemented in EnergyPlus, the green roof model is modified for input ranges, including thermal conductivity and leaf area index (LAI), enabling simulations of the three experimental unvegetated green roofs. A refined green roof model is suggested, considering substrate moisture content variations, associated thermal conductivity, and distinct thermal mechanisms during day and night.