Supplementing street view imagery to local climate zones for modeling land surface temperature: a case study of Guangzhou

Abstract

Urban Heat Island (UHI) effect has driven increasing concern due to climate change, with extreme heatwaves posing significant health risks and socioeconomic challenges. Surface Urban Heat Island (SUHI), described by land surface temperature (LST), has been the focus of extensive research. However, most studies rely on geographic information system (GIS) and remote sensing (RS) data, which have limitations such as coarse resolution, slow data updates, and a bird's-eye perspective that overlooks the human-centered, micro-scale view of the built environment. Consequently, the relationship between the built environment and UHI remains insufficiently explored. Street view imagery (SVI), as a widely available resource, could address these gaps by capturing ground-level, three-dimensional streetscapes (e.g., trees) and land cover features (e.g., pavement materials). This study investigates the UHI phenomenon in Guangzhou's central districts, extracting various streetscape elements from SVI. The correlation between SVI factors and LST was assessed under local climate zones (LCZ). By integrating streetscape indices into a GIS- and RS-based LCZ LST evaluation framework, the study explores SVI's potential to enhance UHI modeling. The findings reveal that while GIS- and RS-based models effectively explain LST-related heat risks, models incorporating SVI data (R² = 0.557, 0.644) exhibit stronger explanatory power compared to those using only GIS and RS data (R² = 0.524, 0.632). Considering the presence of spatial autocorrelation, we employed spatial regression methods, which revealed significant positive spatial spillover effects of UHI on neighboring areas. Streetscape elements such as walls, trees, and sky were found to have a substantial impact on UHI. These findings are valuable for optimizing streetscape design in climate responsive urban design strategies.