arXiv subjects

Jingzi Huang

Publications and source records attributed to Jingzi Huang.

5 recordsLinked to original sources

arXiv2026

Inpainting U-Net for seamless pedestrian-level wind prediction across urban morphologies

Pedestrian-level wind prediction is essential for urban design and wind-comfort assessment, but high-fidelity simulations such as LES remain computationally expensive for rapid evaluation. This study develops a two-stage U-Net framework for efficient prediction of time-averaged pedestrian-level wind speed over realistic urban morphologies. The model is trained and evaluated using the UrbanTALES dataset, which contains realistic city configurations under different approaching wind directions. In the first stage, a baseline U-Net model (M1) predicts wind fields patch-by-patch from normalised building height and fetch information. This formulation allows application to urban domains of arbitrary size, but independent patch inference can introduce discontinuities at patch boundaries. To address this, a second U-Net model (M2) is introduced as an inpainting-based refinement model. M2 uses a larger contextual window containing the initial M1 prediction and local morphology to reduce discontinuities using neighbouring flow information. During full-field inference, M2 is applied iteratively using a Gauss-Seidel scheme until convergence. Results show that M1 captures the main spatial distribution of pedestrian-level wind speed and performs well in low- and moderate-velocity regions, although high-velocity peaks are less accurate. M2 substantially reduces patch-boundary artefacts and improves spatial coherence. Across unseen urban cases, the framework reproduces mean velocity and spatial variability reasonably well, while maximum velocities remain underestimated. Overall, the proposed framework provides an efficient and flexible surrogate model for high-resolution pedestrian-level wind prediction across realistic urban morphologies.

cs.CV↗

arXiv2026

Building drag and shielding in a realistic urban environment

Shielding by upstream buildings is a fundamental control on urban drag, yet its influence remains poorly quantified in realistic urban environments. Here, we investigate shielding effects using building-resolved large-eddy simulations of the University of Bristol campus, comprising 110 buildings of varying height, shape and orientation. Twenty-four wind directions are considered, allowing each building to experience a wide range of upstream shielding conditions. While the total drag of the campus exhibits only moderate directional variability, the drag acting on individual buildings varies substantially. In the present case, approximately $20\%$ of buildings account for $80\%$ of the total drag, which is primarily attributed to a small number of large buildings that contribute disproportionately high drag forces. To quantify shielding, we introduce two dimensionless parameters: the upstream fetch ratio, $L_s/H_s$, and the relative height ratio, $H_s/H$, where $L_s$ is the distance to the nearest upstream obstacle, $H_s$ is the height of the upstream obstacle, and $H$ is the height of the target building. These parameters distinguish between near- and far-wake conditions and between sheltered and exposed buildings, providing a simple method to characterise shielding effects in realistic urban environments. The study provides valuable quantitative insight into drag and shielding in the Bristol campus morphology; more importantly, it establishes a general framework for analysing drag and shielding that can be applied in other complex urban environments. The results identify shielding as a primary control on building drag and motivate shielding-aware measures of effective frontal area and drag coefficient

physics.ao-ph↗

arXiv2026

Multi-scale flow analysis for scale-aware urban-canopy models

As Numerical Weather Prediction (NWP) models approach hectometric resolution, they increasingly enter a regime where urban heterogeneity is only partially resolved and the assumptions underlying conventional urban canopy models (UCMs) become questionable. To address this scale gap, we apply a multi-scale coarse-graining framework (van Reeuwijk and Huang 2025, Boundary-Layer Meteorology) to building-resolving Large-Eddy Simulations (LES) of the University of Bristol campus. Two related morphologies are considered: an original layout with large open-space contrasts and a modified configuration with these regions infilled. By systematically filtering the LES fields, we quantify how flow heterogeneity evolves with resolution and identify a characteristic urban length scale at which resolved and unresolved variability are comparable. This scale is strongly morphology-dependent, with values of about 256 m for the original layout and 64 m for the modified case, showing that neighbourhood-scale organisation can remain important at resolutions relevant to next-generation NWP. We then perform an a priori assessment of distributed drag and turbulent-stress parameterisations. Parameterisations derived from idealised geometries perform reasonably well only at sufficiently coarse resolutions, where horizontal transport is negligible and the flow appears approximately homogeneous. At finer resolutions, their fidelity degrades rapidly because of increasing heterogeneity and filter-to-filter variability in morphology, with stronger limitations in realistic layouts than in idealised cuboid arrays. Overall, the results show that the applicability of urban parameterisations depends critically on the relationship between model resolution and a morphology-dependent heterogeneity scale, providing a systematic route for developing scale-aware UCMs for high-resolution NWP.

physics.flu-dyn↗

arXiv2025

Multi-scale analysis of flow over heterogeneous urban environments

A computationally efficient multi-scale planar-averaging framework for urban areas is developed, which enables efficient computation of coarse-grained velocity and scalar fields. We apply the multi-scale framework to a large-eddy simulation of an idealised heterogeneous urban environment of 512 buildings based on a typical London height distribution. We observe that for this geometry, the characteristic urban lengthscale $\ell \approx$ 50 m, which is the averaging lengthscale L at which as much variance in the mean flow is resolved as is unresolved. For $L>400$ m, the statistics become approximately homogeneous, suggesting that non-building-resolving numerical weather prediction (NWP) models can be applied without modification at resolutions of 400 m and above for the case under consideration. We derive the multi-scale plane- and Reynolds-averaged momentum equation and show that for neutral cases, NWP models require parameterisation of the distributed drag and the unresolved turbulence and dispersive stress. An a priori analysis reveals that the drag parameterisation from Sutzl et al. 2020, Bound-Layer Meteorol., 178:225-248 holds reasonably well for resolutions $L$ above 200 m. Below this value, the problem becomes inhomogeneous and the parameterisation works less well. The unresolved stresses are well represented by a $k-ω$ closure with a value of $ω=0.4/s$. However, an even more accurate closure can be derived from the Sutzl drag parameterisation that does not require further turbulence information.

physics.flu-dyn↗

arXiv2023

The internal structure of forced fountains

We study the mixing processes inside a forced fountain using data from direct numerical simulation. The outer boundary of the fountain with the ambient is a turbulent/non-turbulent interface. Inside the fountain, two internal boundaries, both turbulent/turbulent interfaces, are identified: 1) the classical boundary between upflow and downflow which is composed of the loci of points of zero mean vertical velocity; and 2) the streamline that separates the mean flow emitted by the source from the entrained fluid from the ambient (the separatrix). We show that entrainment due to turbulent fluxes across the internal boundary is at least as important as that by the mean flow. However, entrainment by the turbulence behaves substantively differently from that by the mean flow and cannot be modelled using the same assumptions. This presents a challenge for existing models of turbulent fountains and other environmental flows that evolve inside turbulent environments.

physics.flu-dyn↗