Unravelling the street thermal features network and spatial visual thermal inertia : a study in Hong Kong

Abstract

Urban streets are key areas of heat exchange, yet their fine-scale thermal dynamics remain poorly quantified. This study develops an image-driven framework to identify street-element networks and measure spatial visual thermal inertia in Kowloon, Hong Kong, one of the world’s densest urban fabrics. We synchronised vehicle-mounted measurements of air temperature and helicopter-mounted land surface temperature (LST) with Google Street View images. Streetscape visual features were extracted using semantic segmentation, then linked to air temperature through three complementary lag-sensitive models. All models were converged on a dominant impervious triad LST-roads-building, whose combined warming influence extends laterally up to 40 meters during early evening. Results reveal that vegetation offers strong but highly localized cooling effects within 15 meters, while sky remains nearly neutral inside narrow canyons. These relationships were synthesized into a visual thermal inertia index (VTII), converting lag coefficients into intuitive distance-decay curves, and into a graph representation exposing thermal “hubs,” “bridges,” and communities. The findings suggest that marginal greening or isolated cool-material retrofits are insufficient to offset the spatial influence of impervious surfaces. Instead, 40–50 m continuous high-albedo road would attenuate the VTII by 40–60 %, and connected 30–40 m green corridors are yielding additional cooling ΔTa ≈ 0.45–0.68 °C at 10–20 m. By integrating synchronously collected field data with image analytics, this study offers perspective on urban microclimate processes. The VTII and network metrics provide actionable benchmarks for street-by-street heat mitigation strategies and open new avenues for integrating microclimate evidence into urban design.