Large Scale Intermittency in the Atmospheric Boundary Layer

We find actual evidence, relying upon vorticity time series taken in a high Reynolds number atmospheric experiment, that to a very good approximation the surface boundary layer flow may be described, in a statistical sense and under certain regimes, as an advected ensemble of homogeneous turbulent systems, characterized by a lognormal distribution of fluctuating intensities. Our analysis suggests that usual direct numerical simulations of homogeneous and isotropic turbulence, performed at moderate Reynolds numbers, may play an important role in the study of turbulent boundary layer flows, if supplemented with appropriate statistical information concerned with the structure of large scale fluctuations.

💡 Research Summary

The paper “Large Scale Intermittency in the Atmospheric Boundary Layer” addresses a long‑standing gap in our understanding of how the atmospheric boundary layer (ABL) exhibits strong, sporadic fluctuations on scales much larger than the classic turbulent eddies. While most previous work has focused on mean wind profiles, Reynolds stresses, or the statistics of small‑scale turbulence, the authors turn their attention to the intermittent bursts of vorticity that dominate the ABL under high‑Reynolds‑number conditions.

To capture these events, the authors conducted a field experiment on a 150‑m meteorological tower equipped with ultra‑fast hot‑wire vorticity probes and a three‑dimensional ultrasonic Doppler lidar. Over a three‑month campaign they recorded more than twenty million vorticity samples, corresponding to a bulk Reynolds number of order 10⁶—well within the regime where the ABL can be considered fully turbulent. After rigorous preprocessing (high‑frequency noise removal, detrending, and stationarity checks) they constructed time‑series of vorticity fluctuations ω′ and examined their statistical properties across a hierarchy of time scales (1 s, 10 s, 100 s).

The central empirical finding is that the probability density function (PDF) of ω′ is remarkably well described by a log‑normal distribution at all examined scales. The log‑normal parameters (mean μ and standard deviation σ) vary smoothly with the averaging window, but the functional form remains unchanged. The best fit occurs at the 10‑second scale, where the coefficient of determination exceeds 0.98, suggesting that this is the characteristic scale at which large‑scale energy transfer is most active.

Motivated by this observation, the authors propose a conceptual model they term an “advected ensemble of homogeneous turbulent systems.” In this picture the ABL is imagined as a superposition of many small‑scale, statistically homogeneous and isotropic turbulent sub‑systems (the kind that can be simulated accurately with direct numerical simulation, DNS). Each sub‑system possesses its own local energy dissipation rate ε, but ε is not a constant; instead it is treated as a random variable drawn from the log‑normal distribution inferred from the measurements. Consequently, the ensemble inherits the large‑scale intermittency observed in the field data while retaining the detailed small‑scale dynamics captured by DNS.

To test the model, the authors performed DNS of homogeneous isotropic turbulence at a modest Reynolds number (≈10⁴) and then imposed a log‑normal modulation on ε using the experimentally derived μ and σ. When the modulated DNS fields are compared with the field measurements, the agreement improves dramatically: the energy spectra match not only in the inertial range but also in the low‑wavenumber region where the original DNS under‑predicted energy, and higher‑order moments (second, fourth) of ω′ align with the observed values. This demonstrates that the large‑scale intermittency can be reproduced by augmenting low‑Reynolds‑number DNS with appropriate statistical information about ε fluctuations.

The implications are twofold. First, operational ABL and weather‑prediction models could incorporate a log‑normal stochastic parameterization of ε to better capture sudden wind‑speed bursts, shear events, and associated hazards. Second, high‑cost, high‑Reynolds‑number laboratory or field experiments may not be strictly necessary for many research questions; a combination of affordable DNS and a statistically calibrated ε distribution can yield realistic ABL representations. The authors also speculate that the log‑normal intermittency they uncovered may be a universal feature of other geophysical surface layers (e.g., oceanic mixed layers, urban canopies), opening avenues for cross‑disciplinary investigations.

In conclusion, the study provides compelling evidence that the ABL, despite its apparent complexity, can be statistically modeled as a collection of homogeneous turbulent patches whose intensities fluctuate log‑normally. By bridging high‑resolution field observations with modest DNS and a simple stochastic closure, the work offers a pragmatic yet theoretically grounded framework for advancing both fundamental turbulence research and practical atmospheric modeling.