Hideki Kikumoto

Urban environments exhibit complex physical phenomena involving multiple interacting and imbalanced variables, making high-resolution measurements challenging and expensive. This study aimed to reduce the measurement demand while accurately estimating the mean flow and pollutant concentration fields using physics-informed neural networks (PINNs). To improve the robustness of and mitigate the difficulties faced by PINNs when handling imbalanced outputs, this article proposes a two-step framework in which the airflow and pollutant dispersion within a two-dimensional street canyon are trained separately. In the first step, a PINN reconstructs the mean flow field from sparse velocity data derived from a large eddy simulation and simultaneously estimates the eddy viscosity. In the second step, these results are imposed as fixed fields in the advection–diffusion equation, enabling estimation of the pollutant dispersion from a ground-level line source. By leveraging the accurately reconstructed flow field, the second-step PINN achieves reliable reconstruction of the mean concentration field and turbulent diffusivity. Incorporating the eddy viscosity as additional physical information not only reduces the training cost and improves the accuracy, but also substantially reduces the data requirements. The eddy viscosity data ensure system closure, enabling physics-driven concentration field estimation even in the absence of concentration measurements. Furthermore, a comparison of the optimization strategies shows that the L-BFGS is more effective than Adam because of its second-order characteristics, facilitating convergence in ill-conditioned residual loss landscapes. Overall, the proposed method demonstrates the strong capability of PINNs to handle multiple outputs with imbalanced scaling in urban environments, highlighting their potential for broader applications.

This study quantifies uncertainty in CFD‐based urban flow simulations over central Paris using three modeling approaches: steady RANS with OpenFOAM, unsteady RANS with code_saturne, and LES with PALM. All simulations are validated against field measurements and show overall good agreement. At pedestrian height, the three approaches reproduce similar mean wind distributions, while LES predicts stronger wind components, higher turbulent kinetic energy, a more developed urban canopy layer, and higher area‐averaged temperatures, partly due to grid‐scale turbulence resolution and online radiation coupling. Vertical profiles further indicate that LES produces stronger near‐surface winds and a lower but more intense turbulence peak compared with RANS models. Upwind tall buildings affect turbulence patterns similarly across the cases. The influence of trees is also examined: RANS models consistently predict tree‐induced reductions in mean kinetic energy and friction velocity, whereas LES captures greater spatial variability. Tree shading reduces pedestrian‐level air temperature. These findings clarify model‐dependent uncertainties and provide guidance for selecting suitable CFD approaches and settings for realistic urban microclimate simulations.

This study applies the rainbow particle tracking velocimetry (rainbow PTV) technique—a single‐camera method capable of three‐dimensional, three‐component (3D3C) velocity measurements—to airflow around a cubic block in a wind‐tunnel environment. Helium‐filled soap bubbles were used as tracer particles, and volumetric 3D3C velocity information was obtained over a measurement depth of 140 mm, with the cubic building model having a height of 100 mm. A dedicated processing pipeline, including color‐to‐depth mapping, particle tracking, and depth smoothing, was used to obtain particle trajectories and instantaneous vectors within the measurement volume. The technique successfully captured impingement, lateral divergence, upward motion over the block, and vortex shedding in the wake. Mean velocity fields were compared with reference hot‐wire anemometry measurements and showed good agreement in both upstream and wake regions. These results demonstrate that rainbow PTV can provide reliable volumetric velocity measurements in air using a single camera, offering substantially higher spatial coverage than pointwise methods and enabling analysis of unsteady flow phenomena.