Interactions between urban features and local climate shape the spatial distribution of temperature in cities, which affects outdoor thermal comfort and public health. Modeling these interactions through physics-based climate models, such as the WRF, can provide high-resolution and physically consistent data but are computationally intensive, making high-resolution city-scale simulations challenging. This proposal combines the strengths of physics-based modeling and machine learning (ML) to develop an emulator that efficiently predicts urban temperatures based on urban characteristics. High-resolution meteorological data are generated using WRF simulations for representative weather patterns over the study region. Correlation analysis is then used to select the most important climatic attributes, including different lead times, for predicting hourly ambient temperatures. Static datasets describing landscape characteristics and urban morphology are then incorporated along with the climatic variables as inputs for a ML model. The ML model—trained with XGBoost and Random Forest algorithms— is developed to predict spatial temperature patterns while identifying key urban characteristics that influence temperatures through feature importance analysis. This urban climate emulator will enable efficient scenario testing and can be a scalable, data-driven tool for modeling heat stress in cities.