Evaluating the impact of urban climate on building energy: discrepancies during the hottest period in a subtropical city

Urban building energy simulations often rely on suburban weather files that overlook intra-urban microclimates, leading to biased cooling estimates. This study quantified that bias for Guangzhou, China, in July 2023. WRF with BEP/BEM, driven by local climate zone (LCZ) and GloUCP data, generated 1 km hourly urban meteorology over 1,501 grids, which was then used to run EnergyPlus for a representative high-rise residential building. Urban air temperature was up to 4 °C warmer than the suburban station, and using the station file introduced systematic cooling energy errors: most urban grids were underestimated, while a few cooler peripheral cells were slightly overestimated. Citywide, mean hourly cooling was biased by about 11.5 % and peak cooling by about 16.5 %, with the largest downward errors in compact LCZs (often 20–35 %). A minimal-input XGBoost model using only temperature and humidity from sparse urban sites substantially reduced both mean and peak biases. These results indicate that UHI- and LCZ-dependent boundary conditions must be represented to obtain realistic urban cooling assessments. The results show that UHI-related, LCZ-dependent boundary conditions must be made explicit in urban building energy assessments and that city-to-building coupling can support more resilient, climate-aligned planning.