Traffic at street level emits pollutant particles that get caught in the canyon's own recirculating wind cell. Wind blowing over the rooftops drives a single rotating vortex inside the street canyon — down the windward wall, along the street, up the leeward wall, out over the roof. The top canvas is that cross-section (looking along the street):
v(r) = U·(r/Rc) for r ≤ Rc (solid-body core)
v(r) = U·(Rc/r) for r > Rc (irrotational outer flow)
Rc = 0.5·min(W, H) — Rankine-vortex model of canyon circulation
Every time a particle's path carries it through a tree canopy, some of it is captured on the leaf surfaces — dry deposition, the same process i-Tree and similar urban-forestry models use:
Flux F = Vd · C (deposition flux = velocity × concentration)
Vd_eff = Vd_species · (LAI / 4)
P(capture per pass) = 1 − exp(−Vd_eff · dt / L), L ≈ leaf path length
Needle-leaved conifers hold a higher deposition velocity per unit LAI than broadleaf trees (rougher, denser needle surfaces), so the Species control changes Vd, while LAI scales how much leaf area a particle actually has to pass through.
But denser canopy is not free of trade-offs: once neighbouring canopies begin to close over the street, they block part of the "sky view" at rooftop level and slow the exchange of air with the atmosphere above — real studies (e.g. Vos et al. 2013; Nowak et al., i-Tree) find this can locally raise street-level concentration even as total pollutant mass captured goes up. That trade-off is tracked live as Canopy closure, which throttles the vortex's escape probability at roof height and feeds into the PM2.5 index alongside the measured deposition rate. The bottom-left canvas is a top-down view along the street — drag it to pan and see where particles sit relative to the tree rows; the bottom-right canvas is a rolling history of the deposition/escape outcome of the last 150 particles.