Visualising the Heat: A Street-Level Approach to Urban Temperature Prediction

Urban heat islands (UHIs) pose challenges to urban sustainability and citizen well-being. This study explores the interaction between Land Surface Temperature (LST) and the Green View Index (GVI) in Milan. Using advanced techniques such as image segmentation and machine learning, this analysis scrutinises the nuanced relationship between urban temperatures and visual elements observed across the four cardinal directions. The processing of street-level images through image segmentation not only reveals substantial correlations, but also discerns varied effects based on image heading. Using spatial regression, the results reveal the substantial impact of different image features, highlighting the predominant influence of greenery on temperatures. Adopting Gradient Boosting for LST prediction, this approach highlights areas where temperature underestimation may occur. This method highlights potential biases in citizens’ heat estimation of heat islands who evaluate surrounding conditions just by observing the environment, underscoring the critical need for real-time georeferenced information to fortify public awareness and mitigate risks associated with urban heat islands.