Abstract Summary
According to the latest IPCC reports heat waves will increase in both intensity and frequency. Due to the physical nature of cities, they are particularly vulnerable to this increase in heat. Urban heat island (UHI) effect analysis often relies on satellite imagery, which gives a planar representation of often three-dimensional features. In this study, we propose to integrate street view data to create a large-scale approximation of local street-level micro-climates in urban environments, using Amsterdam as a case study. We present a method that incorporates street view images with a semantic segmentation model to capture finer urban elements from the panoramic images, such as sky, buildings, trees, pervious and impervious surfaces. Furthermore, our approach also involves the calculation of view factors derived from panoramic street view images, employing a hemispherical azimuthal projection technique to accurately capture the 3D element of the urban environment. This allows us to assess the impact of various environmental features on LST, considering elements such as tree view factor (TVF), sky view factor (SVF) and building view factor (BVF). We then use the extracted features to model the relationship between these features and LST using machine learning algorithms such as Support Vector Regression, Gradient Boosting Tree, and a Random Forest.
