Spatiotemporal mapping of urban air temperature and UHI under TMY condition: A reference station based machine learning approach

Urban heat island (UHI) has been one of the most prominent results of anthropogenic related land use change. To achieve accurate and computationally efficient spatiotemporal mapping of air temperature and UHI under typical climate conditions, in this study, a reference weather station-based framework is presented for high-resolution and representative urban temperature mapping in a cost-effective and easy-to-implement way using Shenzhen as a case study. The method employs multi-source data including Local Climate Zone (LCZ) classification, remote sensing data, and machine learning techniques to produce spatially and temporally continuous air temperature fields, rather than land surface (LST) temperatures typically used in previous studies. The XGBoost-based framework achieves good predictability (MAE: 0.56 °C, R²: 0.980) while requiring the weather data from only one single reference station during spatiotemporal mapping. Then, integrated with Typical Meteorological Year (TMY) data of the reference station, it is found that the annual mean UHI intensity (UHII) across all time periods and urban typologies in Shenzhen varies from −0.93 °C to 1.11 °C, with peak instantaneous UHII exceeding 1.2 °C during early afternoon hours (13:00–15:00) in high-rise urban areas. The research shows that high-rise urban areas in Shenzhen experience maximum temperature rises during early afternoons while vegetated areas remain cooler throughout the day. It is also found that urban morphology can significantly influence local temperature patterns, with buildings and vegetation density playing an important role in shaping how temperatures vary across urban areas. The proposed framework enables the integration of TMY data to develop applicable microclimates that serve as foundation for building energy simulations and urban planning related studies. It also provides practical value through its capability to create high-resolution air temperature mapping while requiring low infrastructure, making it accessible for cities worldwide facing urban heating challenges.