Building energy modeling (BEM) is essential for predicting energy use and improving thermal performance in buildings. Traditionally, weather data for BEM comes from built-in tool datasets. Additionally, global atmospheric reanalysis datasets like ERA5, have been used in recent years for BEM. However, the spatial resolution of global atmospheric reanalysis datasets is generally coarse relative to cities, limiting their accuracy in capturing local urban climate effects. Adopting ERA5 as the forcing data, this study examines the use of two urban land surface models, Urban Tethys-Chloris (UT&C) and Urban Weather Generator (UWG), to generate localized weather data for Singapore. The generated local weather data are compared with the data from an on-campus weather station and other weather datasets. Subsequently, these weather datasets are employed as input for an educational building’s energy model that has been validated with energy meter data. The results demonstrate a better agreement between the generated local weather data and locally measured data, compared to the original ERA5 data and typical meteorological year weather data. This leads to an improved accuracy in building energy prediction. By leveraging the global availability of atmospheric reanalysis datasets, this framework for generating local weather data can serve as a universally applicable approach to support building energy design in tropical cities.@en