Abstract: Turbulent fluctuations dominate the flow within urban canopies and impact the dispersion of passive scalars. This study aims to better understand the stochastic nature of the turbulent flow and its influence on puff dispersion in idealized street networks using direct numerical simulations. A fixed wind direction of 45 degrees was simulated and passive scalars were released from a point source at the ground at three locations, corresponding to a short street, an intersection and a long street. First, the probability density functions (PDFs) characterizing the 3D velocity components were analyzed at different horizontal locations and heights. Secondly, time series of puff concentration from different source locations, and parameters such as dosage (integral of concentration over time) and arrival time (time when 5% of the total dosage is reached) were calculated for each street. The differences in the shape of the PDFs for the wind velocity components reflect the three- dimensionality of the stochastic flow. The distributions generally depart from Gaussian, especially within the canopy. However, the shape of the distribution also depends on the location within the street network. Above a transition zone above the canopy approximately Gaussian behavior is observed. The flow regime around the source is fundamental in determining how the puff will propagate. The flow is faster in long streets and, consequently, the highest puff dosages were also observed along the long streets for all source locations. There is large vertical scalar transport close to the sources, especially for releases in the short and long streets. The dosage is split more equally along perpendicular streets for the source at the intersection. Analyzing arrival times, we observe a faster lateral spread above the canopy, with the plume first reaching more distant regions from the source above the canopy.

Keywords: Urban areas, DNS, stochastic flow, puff propagation.