Reconstruction of scalar dispersion using a coarsely sampled Green's function

Fast and reliable urban dispersion models are essential for air quality assessment, city planning, and emergency response. This study uses high-fidelity Large-Eddy Simulations (LES) to examine scalar dispersion around an idealized cubical building under both laminar and turbulent inflow conditions. For the laminar inflow, two Blasius profiles are used to represent thin and thick boundary layers, with thicknesses of 0.25 and 2 times the cube height H, respectively. For the turbulent inflow, realistic upstream turbulence is generated by coupling the domain with a concurrent precursor simulation of a turbulent half-channel, mimicking an incoming atmospheric boundary layer with thickness 3H and Reynolds number based on the inlet velocity at cube height about 4,000. Nine passive scalars are released from distinct upstream locations, selected based on the observed mean flow field structures, to span the main plume behaviors observed across the domain. Their downstream responses are used as a coarse sampling of the Green's function. A reduced-order modeling approach is developed, reconstructing arbitrary scalar release scenarios through weighted superposition of these precomputed fields. By comparing reconstructions under different inflow conditions, we show how incoming turbulence and boundary layer thickness influence plume evolution and impact the robustness and accuracy of fast-response superposition-based predictions.