Abstract
This study investigates the stochastic nature of turbulent flow switching at street intersections in urban canopies. Direct numerical simulations of turbulent flow over a uniform array of rectangular buildings of height h with a 45° inflow direction reveal bistable switching of horizontal wind direction at heights z=0.125h and z=0.25h, where the flow alternates between alignment with the short-street (0°) and long-street (90°) axes. The characteristic switching time scale is approximately 10.5 T (where T=h/uτ is the eddy turnover time), with a standard deviation of 3.7 T and a longer residence time in the long-street-aligned state. Wind-tunnel experiments for different idealized urban configurations confirm that this bistable switching is a generic process, with time scales that depend on building geometry and incident wind direction. Analysis of field measurements at a real intersection in London (UK) also shows intermittent bistable switching, indicating that the phenomenon is robust but sensitive to external forcing and building heterogeneity. Flow switching at intersections influences scalar transport, particularly in the near field, where the strongest correlations between wind direction and concentration are found in the adjacent short street. In the far field, these correlations are weakened by increased mixing and street-network-scale interactions. Overall, the findings show that stochastic flow switching is an intrinsic mechanism that shapes urban ventilation pathways and modifies pollutant dispersion patterns, highlighting its relevance for both dispersion modelling and real-world air-quality assessments. With these applications in mind, simple approaches for including this process in street network dispersion models are outlined.
•Bistable wind direction switching occurs at street intersections for oblique flows.•The switching and residence time scales are computed from DNS data.•Bistable switching also occurs in wind tunnel setups and real irregular intersections.•Wind switching strongly correlates with near-field scalar dispersion.•This stochastic switching can be represented in street network models.
[Display omitted]
