Search arXiv⌕ Search
Home/Archive/arXiv

arXiv · 2610.10430

Conditional Flow Matching for Generation of 3D Multi-variable Instantaneous Urban Microclimate Fields

Peng Liu·Shaoxiang Qin·Theodore Potsis·Lili Ji·Dingyang Geng·Liangzhu Leon Wang
View original source ↗Read source document ↗Export metadata ↓

Abstract

Rapid and accurate prediction of urban wind and temperature fields is important for urban microclimate design and climate adaptation. Large-eddy simulation (LES) effectively resolves these instantaneous fields, but its application is limited in iterative design of urban microclimate applications due to high computational cost. Existing regressive data-driven models offers quick outputs, but they produce only deterministic point predictions that inherently fail to represent turbulent stochasticity. This paper adopts a novel generative framework of Conditional Flow Matching (CFM) that uses building geometry and mean flow as guidance to generate plausible three-dimensional instantaneous velocity and temperature fields for urban microclimate in seconds. To overcome the GPU memory bottleneck of pixel space 3D generation, the model operates in parallel on overlapping pixel space through a shared-noise initialization that preserves high spatial continuity of flow structure across the entire domain. Against reference LES data, the CFM surrogate can rapidly and accurately restore the first-order statistics with Normalized Root Mean Square Error (NRMSE) of 2.99% for wind and 1.77% for temperature, second-order turbulence metrics with NRMSE of 7.17% for wind and 8.84% for temperature, turbulent kinetic energy with NRMSE of 7%, probability density function and vertical profiles in representative locations. Wind engineering application of local gust prediction demonstrate that the speed and accuracy of CFM, supporting the use of generative AI for making turbulence-aware resilient urban design and climate adaptation more computationally feasible.

Explore related subjects

physics.flu-dyncs.CEcs.LGphysics.ao-ph

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Peng Liu, Shaoxiang Qin, Theodore Potsis, Lili Ji, Dingyang Geng, Liangzhu Leon Wang. 2026-10-07. Conditional Flow Matching for Generation of 3D Multi-variable Instantaneous Urban Microclimate Fields. https://arxiv.org/abs/2610.10430

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Unifying Viscocapillary and Inertial Regimes in Selective Withdrawal

Selective withdrawal extracts only a single phase from a stratified multi-layer system. Entrainment occurs when a critical condition draws up the static layer which is not being withdrawn. Existing studies provide robust scalings within distinct limiting regimes. These include viscocapillary-dominated entrainment at low Reynolds number. They also include inertia-dominated entrainment at high Reynolds number. However, a single unifying representation remains to be explored in the literature. This limitation is most evident in transitional conditions between classical limits. It is also pronounced when the lower layer is non-Newtonian. Here we report selective-withdrawal experiments spanning these conditions. The upper layer is Newtonian, using PDMS or soybean oil. The lower layer is either Newtonian water or shear-thinning xanthan-gum solutions. We propose a unified framework that connects these previously separated regimes. The framework adopts a ``Moody diagram'' type representation for selective withdrawal. We collapse normalized critical submergence height using a Reynolds-like control parameter. Surface-tension effects enter subdominantly through the capillary length. The resulting master curve captures the transition between dominant balances. It connects viscous and shear-controlled entrainment to inertial entrainment. The collapse also clarifies how shear thinning enters the organization. Shear thinning primarily renormalizes the viscous correction through an effective viscosity. It does not alter the inertial baseline scale that anchors the normalization. This regime-spanning representation avoids regime-by-regime correlation switching. It provides a compact diagnostic for entrainment thresholds across fluid types. The diagnostic applies to Newtonian and generalized-Newtonian two-layer systems.

physics.flu-dyn↗
arXiv2026

Fluids You Can Trust: Property-Preserving Operator Learning for Incompressible Flows

We present a novel property-preserving kernel-based operator learning method for incompressible flows governed by the incompressible Navier--Stokes equations. Traditional numerical solvers incur significant computational costs to respect incompressibility. Operator learning offers efficient surrogate models, but current neural operators fail to exactly enforce physical properties such as incompressibility, periodicity, and turbulence. Our kernel method maps input functions to expansion coefficients of output functions in a property-preserving kernel basis, ensuring that predicted velocity fields \emph{analytically} and \emph{simultaneously} preserve the aforementioned physical properties. We present universal approximation results and worst-case a priori convergence rates for our framework; empirically, the observed convergence rates exceed the pessimistic predictions across most benchmarks, motivating a formulation of more optimistic convergence rates. Another central contribution of this work is a novel computational framework revolving around streaming construction of kernel Gramians and recursive Schur-complement algorithms to solve the large block linear systems arising from property-preserving kernel methods. This kernel-based framework makes operator learning practical at scales up to $10{,}000$ training functions sampled at up to $10{,}000$ spatial locations. We evaluate the method on challenging 2D and 3D, laminar and turbulent, incompressible flow problems. Our method achieves up to six orders of magnitude lower relative $\ell_2$ errors upon generalization and trains up to five orders of magnitude faster compared to neural operators. Moreover, while our method enforces incompressibility analytically, neural operators exhibit large divergence errors. Our results show that our method provides an accurate and efficient surrogate for incompressible flows.

physics.flu-dyn↗
arXiv2026

Implicit Velocity Correction Schemes for Scale-Resolving Simulations of Incompressible Flow: Stability, Accuracy, and Performance

Scale-resolving simulations of high-Reynolds-number incompressible flow are often limited by the Courant--Friedrichs--Lewy (CFL) restriction of explicit time-stepping schemes, leading to small steps and long times-to-solution. We systematically compare two implicit formulations of the velocity correction scheme -- a linear-implicit approach and a sub-stepping (semi-Lagrangian) method -- with a standard semi-implicit formulation in a high-order spectral/hp element framework. The schemes are assessed in terms of stability limits, temporal accuracy, and computational performance for implicit large-eddy simulation of the Imperial Front Wing benchmark, a complex high Reynolds number geometry with curved surfaces that imposes strict CFL constraints. The sub-stepping scheme extends the stable time step size by a factor of 20, and the linear-implicit scheme by up to a factor of 100 when equal-order velocity and pressure spaces are used. While increasing the cost per time step, the linear-implicit scheme reduces the time-to-solution by up to a factor of 8.7 at this largest step, whereas the sub-stepping scheme reduces it by only 12%. With application targets of 1% for mean lift and 5% for mean drag, the fastest accepted linear-implicit case reduces time-to-solution by a factor of 1.53 at ten times the CFL-limited step. The main-plane separation bubble is already substantially shortened at this step, showing that force agreement can mask changes in transition-sensitive flow structure. The results quantify the trade-off between stability, accuracy, and computational cost for implicit velocity correction schemes on complex geometries and provide guidance for selecting time integration strategies in large-scale scale-resolving simulations.

physics.flu-dyn↗