Researchers develop a new way to measure the turbulent air wakes left behind wind turbines
Every spinning wind turbine carves an invisible corridor of disturbed air in its wake — a churning zone of reduced wind speed and elevated turbulence that quietly saps power from neighboring turbines and accelerates mechanical wear on rotors downstream. Capturing the detailed structure of that wake in real time has long frustrated engineers, because conventional remote-sensing tools tend to assume the air is behaving more uniformly than it actually is.
Acoustic tomography takes a different approach: it listens to how sound pulses travel between sensors to reconstruct what the wind is doing across a patch of atmosphere. A new study suggests that feeding turbulence data from high-fidelity computer simulations into that listening process may give the technique enough precision to matter on a real wind farm.
The problem with wind turbine wakes
When a turbine extracts energy from the wind, it leaves behind a momentum deficit — a corridor of slower, choppier air that can reduce a downstream turbine’s power output and accelerate rotor fatigue. Managing that deficit is central to wind farm design, yet measuring it accurately remains surprisingly hard.
Those LES-derived covariances were plugged into the TDSI operator, allowing the inversion to account for the wake’s lopsided turbulence structure.
Wake turbulence is strongly inhomogeneous and anisotropic: its statistical properties shift sharply depending on where you look — near the rotor hub, at the blade tips, or several diameters downstream. Doppler lidars have advanced wake science considerably, but they rely on flow-homogeneity assumptions that simply don’t hold inside a wake. Without accurate wake characterization, engineers are working with an incomplete picture of what the air is actually doing.
How acoustic tomography listens to the wind
Acoustic tomography measures how long sound pulses take to travel between pairs of speakers and microphones arranged around a patch of atmosphere. Wind and temperature along each path alter those travel times in predictable ways. The time-dependent stochastic inversion (TDSI) technique then mathematically inverts those travel-time differences to reconstruct fluctuating velocity and temperature fields across a two-dimensional plane — a bit like using sound distortion to X-ray the air.
Classical TDSI relies on Gaussian covariance functions that assume homogeneous, isotropic flow. Inside a wake, that assumption breaks down badly. The virtual array modeled in the study mirrors the physical setup at the National Laboratory of the Rockies’ Flatirons Campus in Colorado.
Feeding simulations into the inversion
The core innovation is straightforward in concept: replace smooth Gaussian covariance functions with ones extracted directly from high-fidelity simulations of an actual turbine wake. Researchers ran large-eddy simulations (LES) of a Vestas V27 turbine — a 88.5-foot rotor, 103-foot hub height — under neutral and convective atmospheric boundary layer conditions, using the AMR-Wind solver.
Those LES-derived covariances were plugged into the TDSI operator, allowing the inversion to account for the wake’s lopsided turbulence structure. Three approaches were compared: heterogeneous LES-derived covariances, homogeneous covariances from undisturbed precursor simulations, and classical analytical Gaussian models.
What the numbers revealed
Heterogeneous covariances produced the lowest retrieval errors for streamwise velocity and turbulent kinetic energy, particularly at moderate numbers of observation frames. They also captured fine-scale wake structures near the rotor that analytical models smoothed over entirely. LES-informed retrievals matched simulated velocity spectra up to roughly 2 Hz — outperforming both precursor-based and analytical approaches in the wake region, a frequency range critical for resolving fatigue-driving fluctuations.
One exception is worth noting. At high frame counts in the neutral boundary layer, the classical Gaussian approach outperformed LES-derived covariances. Analytical models remain competitive when their simplifying assumptions actually fit the flow.
Limits, caveats, and the road to real turbines
The study is explicitly idealized. Travel times were computed analytically rather than extracted from simulated acoustic waveforms, and acoustic refraction, instrumental noise, and multipath propagation were not modeled. Temperature retrieval remains the weakest link — errors reached roughly 0.4 K in the neutral case, likely near the practical resolution ceiling.
Scaling to utility-scale turbines introduces further challenges. Modern rotors exceed 393 feet in diameter, requiring proportionally larger arrays. Aeroacoustic noise from rotor blades could overlap tomography signal frequencies, demanding careful signal design — coded pulse sequences in the 100–500 Hz range, for instance — to separate measurement from noise.
What to watch for next
The Flatirons Campus in Colorado is already operating a physical acoustic tomography array near working turbines, and the authors identify field validation there as a necessary next step. Simulation results establish a performance ceiling; real-world deployments will introduce additional uncertainty that only field data can quantify.
Longer term, the study points toward machine-learning inversion strategies, adaptive covariance models, and hybrid approaches blending analytical simplicity with numerically informed anisotropy. If LES-informed covariances can be made practical — through faster simulations or observational surrogates — acoustic tomography could become a genuinely operational tool for reading the invisible scars that turbines leave in the air.
Reference: Hamilton, N. and Bidadi, S.: Wake-resolving acoustic tomography: advances through numerical covariance methods, Wind Energ. Sci., 11, 2621–2645, https://doi.org/10.5194/wes-11-2621-2026, 2026.
Daniel García is an Editor-in-Chief with strong expertise in structural work and engineering principles. He combines this technical foundation with deep knowledge of energy, spatial design, and emerging technologies, bringing a forward-thinking and analytical approach to editorial leadership.
