GPU Accelerated Parallel Computing for Estimating Continuous Sky view Factor Map

Abstract

The sky view factor (SVF) that represents the fraction of visible sky on a hemisphere or the percentage of radiation reaching the planar ground in the entire hemisphere’s input radiation is an important parameter for urban climate studies. However, the estimation of a continuous SVF map is very time-consuming, which limits the applications of SVF to small geographical areas. This study proposed to use graphics processing unit (GPU) parallel computing to accelerate the computing of SVF in the city of Philadelphia, Pennsylvania, USA. This study implemented and compared both the GPU-accelerated version and regular CPU version of two major methods for estimating continuous SVF maps, ray tracing-based algorithm and shadow casting-based algorithm based on the high-resolution building height model. Results show that the GPU-accelerated algorithms can reduce the time consumption dramatically and estimate the SVF map for the city of Philadelphia in less than 20 minutes on a personal computer with one NVIDIA GPU. The ray tracing-based algorithm has a much more efficiency increase than the shadow casting-based algorithm on GPU. The proposed method makes it possible to generate large-scale continuous SVF maps using regular personal computers with GPU. The proposed GPU-accelerated SVF estimation methods would benefit urban climate studies.