Predicting aerosol dispersal in indoor spaces and the associated uncertainties due to turbulence

Turbulent dispersal of aerosol pollutants in a ventilated indoor space is often characterized by the mean concentration using theoretical frameworks such as the well-mixed, near-field/far-field, and eddy diffusivity models. Although their accuracy and efficiency in predicting this mean behavior have been well-established in the literature, this problem demands a deeper analysis, involving the uncertainties associated with the fluid mechanics of the problem. Quantifying these uncertainties requires a rigorous statistical description of the inherent stochastic turbulent dispersal process and the inhomogeneous nature of the indoor flow. This work addresses the importance of going beyond accurate prediction of the mean ensemble-averaged exposure, by evaluating the expected level of variability in individual realizations. We leverage large datasets from turbulence-resolving Euler-Lagrange simulations of aerosol dispersal in varying indoor geometries. The statistical information is used to develop a data-driven, stochastic model that can accurately predict the mean behavior as well as the variances as a function of the separation distance between the pollutant source and the receiver. The validity of the model's predictions is further demonstrated with comparisons to experiments and existing theoretical models.