Previous studies examining the impact of large-scale photovoltaic (PV) roofs on urban heat islands (UHI) have reported inconsistencies, primarily due to reliance on simulations without robust experimental validation. This study addresses this gap through a six-month experimental investigation of four 200m2 rooftop sites in subtropical Hong Kong. We compared a conventional bare roof, a PV roof, and two PV integrated green roofs (PVIGRs), providing the first real-world comparison of these configurations. Results reveal that hourly air temperatures above PV rooftops exceeded those above bare roof by over 4°C on sunny days, with a monthly peak PV heat island (PVHI) intensity of 1.18°C at noon in July. The PVHI was primarily driven by PV surface temperatures, solar irradiance, and ambient air temperatures. Additionally, a notable PV-canopy heating effect was observed under PV panels. While PVIGRs did not exhibit cooling above panels, they mitigated the heating effect underneath by up to 1.26°C in July, lowering PV surface temperatures and building heat conduction. This dual benefit enhances PV efficiency and reduces buildings cooling loads. These findings suggest refining urban land surface models to better estimate the climatic consequences of widespread PV installations. The proposed PV parameterization scheme should consider the heating effects beneath PV canopies and surface roughness length of PV configurations. Additionally, integrated building energy models with urban canopy models could help simulate waste heat from air conditioning influenced by PV rooftops. These insights can inform urban planning and efficient PV deployment strategies.