Fast traffic noise mapping of cities using the Graphics Processing Unit of a personal computer

Traffic noise mapping of cities requires large computer calculation times. This originates from the large number of point-to-point sound propagation calculations that must be performed. In this article it is demonstrated that noise mapping calculation times can be reduced considerably by the use of parallel computation on the Graphics Processing Unit (GPU) of a personal computer. Comparisons are presented between a GPU implementation and a conventional CPU implementation for various urban areas, from which a GPU speedup factor of 720 is obtained (compared with a single CPU). The complete traffic noise map of Amsterdam is calculated with the GPU implementation in 2 h. Local modifications of the noise map, for example to investigate the effect of a new building or a road, are calculated in a time of the order of a minute.