Street-View Derived Facade Albedo and Urban Heat in Five U.S. Cities

This dataset contains the data generated for the paper "Scalable quantification of three-dimensional material albedo reveals consistent cooling mechanisms and unequal heat mitigation potential". Relevant code can be found in: https://github.com/carolgxy/FacadeonHeat/tree/main. Our dataset was constructed using the foundational data and data processing pipeline described in the paper "Enhancing urban building energy models with Vision Transformers: A Case study in material classification from Google street view". Readers seeking detailed methodologies for facade albedo detection are referred to: https://doi.org/10.1016/j.enbuild.2025.115457. Paper title: Scalable quantification of three-dimensional material albedo reveals consistent cooling mechanisms and unequal heat mitigation potential Abstract: As global warming worsens urban heat, accurate urban heat modelling is key to cooling strategies, which traditionally overemphasises coarse land cover while overlooking detailed 3D materials like building façades and roads. This study operationalised a high-throughput pipeline to quantify 3D material envelope information across five diverse U.S. cities: Boston, Chicago, Los Angeles, New York, and Seattle. We provided the first city-wide, street-level 3D material albedo dataset and revealed a consistent cross-city pattern: façade albedo shows a robust cooling association across diverse urban morphologies, supported by SHAP interpretation and counterfactual scenarios, whereas road albedo interacts nonlinearly with canyon geometry. Our findings suggest a previously unquantified material–geometry regime and indicate effective mitigation measures to prioritise the hottest low-density suburban areas where low-income households dominate; high albedo renewal can yield a cooling potential of up to 1.14℃ in Boston. This study highlights the critical role of 3D material albedo in urban heat, offering actionable insights for equitable climate adaptation. Researchers should cite this dataset and the relevant paper when using it in publications or presentations.