AeroSim Docs
Street-Canyon Pollutant Dispersion (CODASC)¶
Why this case matters¶
The street canyon is the defining urban-dispersion geometry: a road flanked by two buildings, with traffic emissions trapped in a recirculating cell between the facades. Pedestrian exposure depends on where that cell deposits the pollutant, which makes the canyon the flagship test of a CWE dispersion solver. It exercises the entire scalar stack at once: impermeable building facades, a continuous ground-level emission, a turbulent atmospheric inlet and the subgrid mixing that distributes the tracer between the leeward and windward walls.
The CODASC database (Concentration Data of Street Canyons, KIT Karlsruhe) provides wind-tunnel facade-concentration measurements for exactly this configuration. The underlying experiments [1][2] are widely used as a dispersion-model benchmark, evaluated with the COST Action 732 framework [3]. This case targets the tree-free reference canyon, validating the zero-flux facades, the ground line source, the ABL inlet and the turbulent-Schmidt coupling together against a documented dataset. It is the most complete pollutant-dispersion validation in the suite.
Physical description¶
Two parallel buildings of equal height form a street canyon of unit aspect ratio (width equal to height). A neutral atmospheric boundary layer approaches perpendicular to the canyon axis, separating over the upstream (leeward) building and driving a recirculating vortex inside the canyon. A continuous line source on the canyon floor releases a passive tracer; the canyon vortex transports it preferentially toward the leeward facade (wall A), with lower concentrations on the windward facade (wall B).
In nassu both buildings are voxelized, with their facades carrying the no-slip velocity condition and a zero-flux scalar condition (impermeable to the tracer). The line source is modelled as a localized volumetric source term on the canyon floor, a thin band centred on the canyon mid-line and spanning the source length. The atmospheric boundary layer is fed at the inlet by the Synthetic Eddy Method (SEM), reused from the atmospheric case, and the subgrid mixing of the tracer is controlled by the turbulent Schmidt number \(\mathrm{Sc}_t\) (see LES Subgrid Coupling).
Governing equations¶
The tracer is a passive scalar transported by the resolved flow with a floor source term (see Scalar Transport). The validated quantity is the dimensionless wall concentration along each facade:
where \(c\) is the measured facade concentration, \(U_H\) the reference velocity at roof height, \(H\) the building height, \(l_s\) the line source length and \(Q\) the total volumetric source (tracer) emission rate. This matches the CODASC database normalisation \(c^+ = c\, U_H\, H / (Q / l_s)\). \(K\) is measured along the leeward facade (wall A) and the windward facade (wall B) and compared profile by profile against the wind-tunnel data.
Simulation setup¶
The case follows the CODASC empty (tree-free) canyon at model scale 1:150, with a unit-aspect-ratio canyon and a perpendicular neutral approach flow.
Parameter |
Value |
|---|---|
Building height \(H\) |
0.12 m (model), 24 lattice cells at level 0 |
Canyon aspect ratio \(W / H\) |
1 |
Wind angle |
\(90^\circ\) (perpendicular approach) |
Reference velocity \(U_H\) (roof height) |
4.65 m/s (model), 0.05 lattice units (\(\mathrm{Ma} = 0.087\)) |
Reynolds number on \(H\) |
37000 |
ABL mean profile |
power law \(u(z) = U_H (z / H)^{0.3}\) (exponent \(\alpha = 0.3\)) |
Tracer |
\(\mathrm{SF}_6\) (molecular \(\mathrm{Sc} = 1\)) |
Turbulent Schmidt number \(\mathrm{Sc}_t\) |
0.8 |
Fluid velocity set / operator |
D3Q27 / RRBGK |
Fluid relaxation time \(\tau\) |
0.5000973 (\(\nu = 3.243 \times 10^{-5}\); \(\omega \approx 1.9996\)) |
Scalar velocity set / operator |
D3Q7 / RRBGK |
LES model |
Smagorinsky (\(C_S = 0.17\)) |
Inlet turbulence |
SEM, neutral ABL profile (reused from the atmospheric case) |
Facade wall treatment |
voxelized band: |
Source treatment |
continuous ground-level line source on the canyon floor (localized source term) |
Ground |
IBM terrain with the equilibrium log-law wall model |
Refinement |
static level-1 shell and level-2 slab over the canyon (effective \(H \approx 96\) cells) |
The canyon is refined to level 2, giving an effective canyon resolution of about 96 cells per building height, well above the COST 732 guidance of at least 10 cells per building dimension in the region of interest. The RRBGK operator stays stable as \(\omega \to 2\), which is required to reach this Reynolds number on the available grid.
Reference and acceptance¶
The reference is the CODASC database (www.codasc.de, KIT Karlsruhe), from the wind-tunnel experiments of Gromke[1] and Gromke and Ruck[2], evaluated with the COST Action 732 metrics [3]. The acceptance targets are the standard quality bands:
factor-of-two of observations \(\mathrm{FAC2} \geq 0.5\),
normalised mean-square error \(\mathrm{NMSE} < 4\),
hit rate \(q \geq 0.66\).
Note
The exact codasc.de case/file label for the empty \(W/H = 1\), \(\alpha = 90^\circ\) configuration is to be confirmed on the website before the facade-concentration comparison data is populated.
Results¶
Note
The quantitative comparison notebook (the leeward / windward wall concentration \(K\) against CODASC and the COST 732 metrics) will be added once the GPU validation runs are committed, mirroring the workflow of the other v2.0 scalar cases.