Assessing social equity and urban heat risks with machine learning of remote sensing imagery: A Pittsburgh case study

Rapid urban development has intensified soil sealing by impervious surfaces, contributing to extreme urban heat, flooding, and health risks. Although research on the impact of impervious surfaces on urban heat has grown in recent years, most studies overlook variations in surface characteristics, such as surface color, which influence urban heat. Consequently, city-wide data on sub-categorized impervious surfaces remain limited. This research fills this gap by using high-resolution remote-sensing imagery classification in ArcGIS Pro to map surface characteristics in Pittsburgh, Pennsylvania, and examine their relationship with land surface temperatures (LST) and social vulnerability. Results show that impervious surfaces cover 55% of the city, including 22% roofs, 30.7% roads and 2.3% parking lots, with 52% of these surfaces classified as dark. On average, historically redlined neighborhoods are 2.6 °C (4.7 °F) hotter and contain a higher proportion of dark surfaces. These results underscore the role of surface color and composition in shaping urban thermal inequities and emphasize the need for evidence-based decision-making in surface material selection to build more equitable and sustainable cities. • Pittsburgh’s overall roofs, streets, and parking lots are 52% dark. • Summer surface temperatures can reach 68.6 °C (155.4 °F). • Historically redlined neighborhoods have 12.6% more dark surfaces and are 2.6 °C (4.7°F) hotter. • Each 1% more dark surface adds 0.18 °C (0.32 °F) to summer LST. • Dark and light surfaces show modest correlations with SVI across Pittsburgh.