Numerical studies of urban pollutants have focused on chemical species. This represents a simplification because aerosols are important components of urban air pollutants. The concentration of aerosols is sensitive to many complex factors, e.g. source locations, emission spectra, and aerosol dynamical processes. In this thesis, the dispersion of cooking-generated aerosols is investigated using numerical simulations. Box models with improved accuracy and computational efficiency are developed.

First, the dispersion of cooking-generated aerosols from an urban street canyon is examined with computational fluid dynamics (CFD). With a representative choice of the source flux, the inclusion of aerosol dynamic processes decreases the mean canyon-averaged number concentration by 15%-40%, whereas the effect is weaker for traffic-generated aerosols. Coagulation dominates for deep-frying, while the effects of deposition and coagulation are comparable for boiling. It is argued that, for a specific emission spectrum, the qualitative nature of the aerosol dynamics within urban canopies is determined by the ratio of the aerosol timescales to the relevant dynamical timescale (e.g. the mean age of air).

Second, box models describing the evolution of urban aerosols emitted inside an urban street canyon are developed. By analogy with box models for gas phase species, which describe the evolution of different reactants, the evolution of the discrete size spectrum is analysed. By contrast with previous box models of urban pollutants, multiple boxes are introduced by partitioning the canyon space into dynamically distinct boxes. The results reveal that when coagulation and deposition processes occur simultaneously, the discrepancy in mean concentrations between the box model and large-eddy simulations (LES) decreases from 27% to 5% using the multi-box model. The multi-box model also shows the spatial structure of the particle number concentration. The results illustrate that the box modelling methodology can be generalised to urban aerosols. The inclusion of extra boxes improves accuracy and eliminates the need for segregation corrections.

Third, standard box models usually describe the ventilation using the mean values, e.g. mean exchange velocity; therefore only mean concentrations are predicted. The influence of turbulent fluctuation is incorporated by developing an ensemble box model in which the mean residence time is replaced by the probability distribution of residence times. Application to the standard NOx photochemistry shows that the introduction of the ensemble allows for the mean concentration to be predicted more accurately and a useful estimate of the standard deviation to be made. The amplitude of the temporal fluctuations is captured well by the ensemble (e.g. the standard deviation is underestimated by around 5% compared to LES). The computational cost of the ensemble box models is significantly lower (i.e. by ~90%) than that of the CFD model.

The findings of this thesis improve understanding of the physical processes governing urban aerosol evolution and offer a new perspective on box modelling of urban aerosols. The inclusion of aerosol processes leads to deviations from the evolution of a passive scalar, coagulation requires the aging of pollutants and deposition is promoted when particles contact solid surfaces. Computationally efficient alternatives to CFD for aerosols and gas-phase pollutants can be obtained by implementing aerosol dynamical processes using multi-box or ensemble models. Long-term implications of this work include more accurate numerical models of urban pollutants and improved operational air quality predictions.