Crunching large data sets to identify air pollution hot spots