Concentration averaging and long-term exposure assessment methodologies in urban areas

Assessing long-term exposure to urban air pollution through modelling often requires significant computational power, particularly when using large-scale or complex models. This article explores methodologies designed to reduce the computing time needed to assess average pollutant concentrations over extended periods leveraging wind statistics, illustrated with both Computational Fluid Dynamics (CFD) and Artificial Intelligence (AI) models. These methodologies are evaluated against in-situ measurement data and an exhaustive modelling approach that computes concentration maps for every hour of the year. The approach includes two variants: one utilizing continuous wind data and another using discrete wind data. Both variants achieve errors below 10 % compared to the exhaustive hourly computation, given appropriate parameter choices. The statistical method performs the computation in under 2 min compared to 5 h for the reference method to obtain an average concentration map. Additionally, the article uses a secondary methodology that estimates the error from using 9 wind directions to discretize the wind rose versus 18 directions, without requiring additional computations for all 18 directions.