The share of the world’s population living in cities is rapidly increasing, and it is expected to rise to 80% by 2050. It becomes therefore crucial to have efficient and reliable methods to model the urban microclimate; in fact, these models can support urban planners and policymakers to create more comfortable and sustainable cities. Since at the urban level, pollutant dispersion depends on daily weather conditions, computational fluid dynamics models with low time scales, repeated evaluation, and fine mesh discretization must be used. The former requirements translate into huge memory requirements, making it essential to use HPC facilities to get results in reasonable time frames. However, the problem is suitable for the employment of Reduced Order Models (ROMs) to achieve fast converged solutions with limited loss of accuracy. The developed methodology is used to achieve real-time prediction of urban air pollution for different test cases. We show a first example on a small portion of the city of Bologna and some preliminary results on a larger test case. The problem is paremetrized by the direction and intensity of the velocity field at the boundary of the computational domain.