🔬 physics

Snowflake-tracing large-scale PIV enables full-scale measurements of atmospheric flow around a wall-mounted cube

This study introduces a novel snowflake-tracing large-scale particle image velocimetry (LPIV-s) technique to perform the first full-scale, high-resolution measurements of atmospheric flow around a wall-mounted cube, revealing unique unsteady wake behaviors like "wake breathing" and distinct flow states driven by high-Reynolds-number turbulence.

Original authors: Biao Li, Kun Zhang, Zhuohang Ji, Qingwei Lyu, Xinhua Su, Shengli Mao, Wenchao Su, Yong Shuai, Cunyan Jiang, Yingli Xuan, Akashi Mochida

Published 2026-08-24

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Biao Li, Kun Zhang, Zhuohang Ji, Qingwei Lyu, Xinhua Su, Shengli Mao, Wenchao Su, Yong Shuai, Cunyan Jiang, Yingli Xuan, Akashi Mochida

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Wind is not a steady, uniform stream; it is a chaotic, churning fluid that changes speed and direction in ways that are difficult to predict. When this moving air encounters a solid object, such as a building or a bridge, it separates from the surface, creating a turbulent wake behind the obstacle. For decades, engineers have relied on small-scale models tested in wind tunnels or computer simulations to understand how these forces behave, hoping to design safer structures. However, a persistent gap exists between these controlled environments and the real world. In a wind tunnel, the air is confined, the turbulence is limited, and the scale is small, which often fails to capture the complex, large-scale interactions that occur when wind hits a full-sized object in the open atmosphere. To truly understand how wind loads act on real structures, scientists need to measure the air itself as it moves around a full-sized object in nature, a task that has proven incredibly difficult because the air is invisible and the scales are vast.

A team of researchers from Harbin Institute of Technology and international partners has now bridged this gap by turning a winter day in Harbin, China, into a massive, natural laboratory. Instead of trying to generate artificial wind or inject tracers into the air, they used the weather itself. In December 2024, on a frozen stretch of the Songhua River, they placed a two-meter-high concrete cube in the open field. As natural snow fell, they illuminated the falling flakes with a powerful laser sheet, turning the snow into millions of tiny, moving markers. By filming these snowflakes with high-speed cameras, the team was able to track the movement of the air around the cube with a level of detail never before achieved in a full-scale, outdoor setting. This technique, which they call snowflake-tracing large-scale particle image velocimetry, allowed them to see the invisible flow of the atmosphere as it interacted with the cube, capturing the true behavior of wind at a Reynolds number of approximately 345,000, a scale far beyond what wind tunnels can typically replicate.

The measurements revealed that the wind flowing around the real cube behaved differently than the wind around smaller models. The wake, or the region of swirling air behind the cube, was shorter and more compact than what engineers usually expect from laboratory experiments. The air recovered its speed more quickly after passing the obstacle, and the center of the swirling vortex sat lower than predicted. These differences suggest that the high turbulence and large, churning eddies present in the natural atmosphere mix the air more vigorously than the smoother, more controlled air in a wind tunnel. This enhanced mixing pulls faster-moving air from above down into the wake, helping the flow to reattach and stabilize sooner. The study confirms that relying solely on small-scale models might lead to an incomplete picture of how wind actually behaves around full-sized structures.

Perhaps the most striking discovery was that the wake was not a static shape but a living, breathing entity that expanded and contracted over time. The researchers observed that the length of the recirculation zone behind the cube varied significantly, stretching from about 1.02 times the height of the cube to 1.53 times its height. This "wake breathing" was driven by low-frequency fluctuations in the incoming wind, large-scale gusts that meander through the atmosphere over periods much longer than the rapid spinning of individual vortices. When the wind speed increased, the wake contracted; when it slowed, the wake expanded. This dynamic behavior showed that the flow does not simply respond to the average wind speed but is constantly modulated by the unsteady, natural rhythm of the atmosphere. The flow also switched between two distinct states: one where the air separated from the top of the cube and rolled directly into the wake, and another where the air bent down, briefly stuck to the top surface, and then separated again. The researchers found that the air reattached to the top of the cube about 77 percent of the time, a detail that significantly alters the forces acting on the structure.

By analyzing the energy within the flow, the team found that the leading edge of the cube was the primary site where wind energy was converted into turbulence. The shear layer, the thin region where fast-moving air slides over slower-moving air, grew rapidly as it moved downstream, creating intense mixing and energy dissipation. This rapid growth and the subsequent transport of energy helped explain why the wake was so compact and why the flow recovered so quickly. The study provides the first high-resolution, two-dimensional view of these full-scale atmospheric interactions, offering a new benchmark for validating computer models and wind-tunnel tests. It demonstrates that the real atmosphere is far more complex and dynamic than previously captured in standard engineering assessments, and that understanding the full-scale, unsteady nature of wind is essential for designing structures that can withstand the true forces of nature. The success of using natural snow as a tracer opens a new path for observing the invisible world of fluid dynamics in its most authentic environment, proving that sometimes the best tools for science are already falling from the sky.

Technical Summary: Snowflake-Tracing Large-Scale PIV for Full-Scale Atmospheric Flow Measurements

Problem Statement
Full-scale, high-resolution measurements of bluff-body flows in atmospheric turbulence are critical for bridging the gap between laboratory models and engineering prototypes. However, extrapolating wind-tunnel results to the atmospheric boundary layer (ABL) is hindered by scale gaps, Reynolds-number mismatches, truncated turbulence integral scales, and the inability to reproduce unsteady atmospheric gusts. While Direct Numerical Simulation (DNS) is limited by exponential grid requirements at high Reynolds numbers, existing full-scale field experiments are largely restricted to surface pressure sensors or intrusive point-velocity measurements. These methods fail to resolve two-dimensional flow fields, separation, shear-layer entrainment, and wake-vortex evolution. Furthermore, extending Particle Image Velocimetry (PIV) to large outdoor domains is challenging due to tracer visibility and dynamic response limitations; existing tracers like helium-filled soap bubbles (HFSBs) are unsuitable for spatial scales exceeding 10 meters.

Methodology
The authors propose and validate Snowflake-Tracing Large-Scale Particle Image Velocimetry (LPIV-s), a non-intrusive technique utilizing natural snowfall as flow tracers.

Experimental Setup: The study was conducted in December 2024 on a frozen, open reach of the Songhua River in Harbin, China, serving as a natural open-air wind tunnel. A 2-meter full-scale cube (HIT-Cube) was deployed in a near-neutral atmospheric boundary layer with a Reynolds number based on height ( $Re_H$ ) of approximately $3.45 \times 10^5$ .
Imaging System: A 30 W continuous-wave laser (520 nm) illuminated a 12 m $\times$ 6.75 m field of view. A 4K camera recorded naturally falling snowflakes at 120 Hz.
Tracer Physics: Snowflakes were characterized by an equivalent diameter of ~0.4 mm and a density of ~91 kg/m³. While possessing finite inertia (Stokes number $\approx$ 0.136 based on Kolmogorov timescales), their characteristic response scales are sufficiently small relative to the large-scale flow structures to allow tracking. Their complex crystal structures provide strong geometric scattering, offering signal intensities ~3,000 times higher than standard 3 $\mu$ m olive-oil droplets, enabling long-range imaging.
Data Processing: A hybrid algorithm combined ensemble correlation for mean flow reconstruction and cross-correlation for instantaneous fluctuations. Crucially, a dynamic correction based on the simplified Basset-Boussinesq-Oseen (BBO) equation was applied to compensate for snowflake inertia, recovering the true fluid velocity from the measured particle velocity.
Uncertainty: The combined relative uncertainty for instantaneous velocities was estimated at approximately 6.6%.

Key Results

Mean Flow Topology: The full-scale measurements revealed a wake topology similar to wind-tunnel and DNS results but with distinct differences driven by high Reynolds numbers and atmospheric turbulence. The full-scale wake was more compact, with the vortex core descending to $z/H \approx 0.8\text{--}1.0$ (compared to $1.0\text{--}1.2$ in lower-Re studies) and a mean recirculation length of approximately 1.2H, shorter than the typical 1.3H–1.6H observed in wind tunnels. Momentum recovery was significantly faster in the full-scale experiment.
Wake Breathing: Low-frequency, large-scale atmospheric fluctuations (Strouhal number $St \approx 0.02$ ) were found to modulate the wake structure, producing a pronounced "wake breathing" phenomenon. The recirculation length varied dynamically between 1.02H (during high-speed inflow phases) and 1.53H (during low-speed phases). This indicates that the wake response is not a simple quasi-steady function of instantaneous wind speed but is jointly modulated by multiscale unsteadiness.
Instantaneous Flow States: The flow alternates between two distinct instantaneous states:
- Unattached State (23% of samples): The leading-edge shear layer passes over the cube without reattaching, shedding diffuse, weakly coherent vortices.
- Reattached State (77% of samples): The shear layer bends downward, reattaches to the top surface, and separates again at the trailing edge. This state is associated with enhanced shear-layer instability, stronger vortex shedding, and more coherent wake structures.
Turbulence Statistics: The separated shear layer exhibited rapid spatial growth with a thickness growth rate of $d\delta/dx \approx 0.404$ , exceeding canonical planar mixing layer values. High Reynolds stresses and Turbulent Kinetic Energy (TKE) production were concentrated in the leading-edge separated shear layer, which acts as the primary site for energy conversion and transport. The shear layer evolved rapidly toward isotropy downstream.

Significance and Claims
The paper claims to present the first full-scale, two-dimensional flow-field measurements of a bluff-body wake using natural snowfall tracers. Its primary contributions are:

Methodological: Establishing LPIV-s as a feasible route for spatially resolved, large-scale flow measurements in the real atmosphere, overcoming the tracer limitations of previous methods.
Physical Insight: Providing experimental evidence that high-Reynolds-number atmospheric turbulence and low-frequency inflow modulation significantly alter wake structure, leading to more compact mean wakes and complex, non-stationary dynamics (wake breathing and state transitions) that are not captured by steady or statistically stationary assumptions.
Validation: Offering a new benchmark for assessing the extrapolation of reduced-scale wind-tunnel experiments and numerical models to full-scale atmospheric conditions, particularly regarding the interaction between high-Re turbulence and bluff-body wake structures.

The authors conclude that full-scale bluff-body flows should not be evaluated using a single mean wake scale; instead, the statistical range of recirculation lengths, low-frequency modulation, and transitions between instantaneous flow states must be considered for accurate wind-load assessments.

More like this