Sensitivity analysis in pollution dispersion within street canyons

The trend of urbanization is increasing due to socioeconomic factors, leading to a decline in air quality and heightened exposure for individuals. Modeling dispersion of emissions within street canyons is crucial for understanding local-scale air pollution exposure indicators and devising effective strategies to enhance urban air quality. However, microscale urban air quality simulations are complex and involve a large number of interacting physical processes. Phenomena taken into consideration in simulations of urban pollution dispersion largely vary from one study to another. In this context, we aim at identifying the most influential parameters governing this phenomenon.First, we present a comprehensive database of pollution dispersion in street canyon by perturbing the most important parameters. In this study, these parameters include wind speed, background concentration, emission rates, background turbulence intensity, temperature disparity between building surfaces and the atmosphere, as well as model-specific parameters. We employ COMSOL Multiphysics for the computational fluid dynamic simulations. To keep feasible computational times and create a large database, turbulence is modeled by means of the Reynolds Averaged Navier-Stokes (RANS) k- technique and the domain is considered two-dimensional.This simulation bank is used to develop a surrogate model based on Gradient Boosting Machine and the Principal Component Analysis to reduce system complexity. This surrogate model offers a solution to carry out a multivariate global sensitivity analysis. To this end, we compute the Sobol Indexes, providing insights into the global contributions of input factors to the resulting pollution concentrations at each point of the physical domain.Results show that, as expected, background concentration dominates the pollution field far from the emission point. The temperature of building surfaces plays a secondary role, acting mainly on the interface between the street canyon cavity and the flow above the buildings. Emission plays a fundamental role, especially in the vertical concentration profile. Other model-specific parameters have a minor role when compared to the other physical variables. Wind speed plays a key role, along with temperature, in determining how pollution exits the cavity.By integrating advanced simulation techniques with rigorous sensitivity analyses, this study aims to provide valuable insights into the factors influencing air quality in urban street canyons. Such insights are crucial to perform model calibration, and to understand the implications of typical approximations such as non-buoyancy flows.