Sky View Factor footprints for urban climate modeling

Abstract

Urban morphology is an important multidimensional variable to consider in climate modeling and observations, because it significantly drives the local and micro-scale climatic variability in cities. Urban form can be described through urban canopy parameters (UCPs) that resolve the spatial heterogeneity of cities by specifying the 3-dimensional geometry, arrangement, and materials of urban features. The sky view factor (SVF) is a dimension-reduced UCP capturing 3-dimensional form through horizon limitation fractions. SVF has become a popular metric to parameterize urban morphology, but current approaches are difficult to scale up to global coverage. This study introduces a Big-Data approach to calculate SVFs for urban areas from Google Street View (GSV). 90-degree field-of-view GSV photos are retrieved and converted into hemispherical views through equiangular projection. The fisheyes are segmented into sky and non-sky pixels using image processing, and the SVF is calculated using an annulus method. Results are compared to SVFs retrieved from GSV images segmented using deep learning. SVF footprints are presented for urban areas around the world tallying 15,938,172 GSV locations. Two use cases are introduced: (1) an evaluation of a Google Earth Engine classified Local Climate Zone map for Singapore; (2) hourly sun duration maps for New York and San Francisco.