Monin–Obukhov Similarity Theory

• Monin–Obukhov Similarity Theory is a framework that scales near-surface turbulence using stability parameters such as z/L to relate mean gradients and fluxes.
• It underpins bulk flux algorithms through canonical gradient and profile relations, enabling predictions of wind and temperature structures across stable and unstable regimes.
• Recent advances incorporate turbulence anisotropy and regime-dependent logic to extend MOST’s applicability beyond the classical constant-flux, homogeneous conditions.

Monin–Obukhov Similarity Theory (MOST) is a similarity framework for turbulence in the atmospheric surface layer, the near-surface region in which properly scaled mean gradients, variances, and fluxes are expressed as functions of a stability parameter built from height and the Obukhov length. In its classical form, MOST assumes stationary, horizontally homogeneous flow, negligible mean subsidence, and a constant-flux layer in which turbulent momentum and heat fluxes are approximately invariant with height. It underpins nearly all marine bulk aerodynamic algorithms and is used in virtually every Earth System Model to parameterize near-surface turbulent exchanges, yet a substantial modern literature shows that its accuracy depends strongly on regime, measurement height, terrain complexity, and the degree to which its foundational assumptions hold (Foxabbott et al., 8 Jul 2025, Mosso et al., 2023, Stiperski et al., 2022).

1. Foundations and scaling variables

MOST rests on the hypothesis that near-surface turbulence is controlled by a small set of wall and buoyancy scales. In common notation, the stability parameter is $\zeta = z/L$, where $z$ is height above the surface and $L$ is the Obukhov length. A frequently used formulation based on virtual potential temperature is

$$u_* = \left(\overline{u'w'}^2 + \overline{v'w'}^2\right)^{1/4}, \qquad \theta_* = -\frac{\overline{w'\theta_v'}}{u_*},$$

$$L = -\,\frac{u_*^3\,\overline{\theta_v}}{\kappa g\,\overline{w'\theta_v'}}, \qquad \zeta = \frac{z}{L},$$

with $\kappa \approx 0.4$ the von Kármán constant. Dry formulations using potential temperature rather than virtual potential temperature are also used, especially in idealized surface-layer treatments (Mosso et al., 2023, Basu, 2017).

The classical assumptions are repeatedly stated in the recent literature: stationary, horizontally homogeneous surface-layer flow; negligible subsidence and large-scale advection; monotonic wind and temperature profiles; and vertically constant turbulent fluxes of momentum and heat over the measurement range. These assumptions define the constant-flux layer that MOST was originally intended to describe. In stable boundary layers, a local-scaling formulation replaces surface fluxes by local fluxes at height $z$, so that the Obukhov length and similarity functions become explicitly height dependent (Foxabbott et al., 8 Jul 2025, Grachev et al., 2012).

The physical role of $L$ is to measure the relative importance of shear and buoyancy. Neutral conditions correspond to $L \to \infty$, so that buoyancy corrections vanish and logarithmic wall-layer behavior is recovered. Unstable conditions have $L<0$, stable conditions $z$0, and the near-neutral, weakly stratified, horizontally homogeneous surface layer remains the regime in which MOST is most robust (Basu, 2017, Foxabbott et al., 8 Jul 2025).

2. Canonical flux–gradient and flux–profile relations

MOST is most commonly expressed through dimensionless gradient functions,

$z$1

or through integrated profile relations for mean wind and temperature. A standard profile form is

$z$2

$z$3

with $z$4 and $z$5 the roughness lengths for momentum and heat, and $z$6 the integrated stability corrections (Foxabbott et al., 8 Jul 2025, Basu, 2017).

The most widely cited empirical closures are Businger–Dyer type relations. In unstable stratification, one commonly uses

$z$7

while in stable stratification a common linear form is

$z$8

Integrated corrections for the unstable Businger–Dyer forms are available in closed form; in neutral conditions $z$9, and the wind and temperature profiles reduce to logarithmic laws (Basu, 2017, Heisel et al., 2023).

These canonical functions are the basis of many practical surface-layer algorithms. Marine bulk schemes typically use Businger–Dyer-type functions or the related formulations embedded in COARE, and recent analytical models of the atmospheric boundary layer still couple outer-layer dynamics to a MOST-consistent inner surface layer through these profile relations (Foxabbott et al., 8 Jul 2025, Narasimhan et al., 2023).

3. Stable, convective, and local-scaling regimes

In stable boundary layers, a central question is the range over which similarity theory remains applicable. Spectral analysis over Arctic pack ice showed that when both the gradient Richardson number $L$0 and the flux Richardson number $L$1 exceed about $L$2–$L$3, the Richardson–Kolmogorov inertial cascade collapses, high-frequency flux-carrying eddies vanish, and local MOST ceases to be appropriate. When supercritical cases are filtered out, the data follow classical local z-less predictions, with

$L$4

and $L$5–$L$6, while $L$7 is nearly constant in the subcritical regime (Grachev et al., 2012).

In convective boundary layers, the central difficulty is different. Large-eddy simulations showed that the nondimensional gradients $L$8 and $L$9 broadly align with Monin–Obukhov scaling across cases, but within each profile their decay with increasing height is steeper than classical Businger–Dyer theory predicts. Departures become substantial well below the conventional surface-layer height of $$u_* = \left(\overline{u'w'}^2 + \overline{v'w'}^2\right)^{1/4}, \qquad \theta_* = -\frac{\overline{w'\theta_v'}}{u_*},$$0, and an exponential cutoff in $$u_* = \left(\overline{u'w'}^2 + \overline{v'w'}^2\right)^{1/4}, \qquad \theta_* = -\frac{\overline{w'\theta_v'}}{u_*},$$1,

$$u_* = \left(\overline{u'w'}^2 + \overline{v'w'}^2\right)^{1/4}, \qquad \theta_* = -\frac{\overline{w'\theta_v'}}{u_*},$$2

$$u_* = \left(\overline{u'w'}^2 + \overline{v'w'}^2\right)^{1/4}, \qquad \theta_* = -\frac{\overline{w'\theta_v'}}{u_*},$$3

improves similarity from approximately $$u_* = \left(\overline{u'w'}^2 + \overline{v'w'}^2\right)^{1/4}, \qquad \theta_* = -\frac{\overline{w'\theta_v'}}{u_*},$$4 to above $$u_* = \left(\overline{u'w'}^2 + \overline{v'w'}^2\right)^{1/4}, \qquad \theta_* = -\frac{\overline{w'\theta_v'}}{u_*},$$5 (Heisel et al., 2023).

The literature therefore distinguishes clearly between the surface layer, where MOST remains an organizing framework, and the lower convective or stable boundary layer, where additional parameters such as $$u_* = \left(\overline{u'w'}^2 + \overline{v'w'}^2\right)^{1/4}, \qquad \theta_* = -\frac{\overline{w'\theta_v'}}{u_*},$$6, local fluxes, or alternative similarity lengths become necessary. This suggests that the phrase “validity of MOST” is regime dependent rather than absolute: in subcritical stable surface layers and near-neutral homogeneous flows it remains effective, whereas transition regions to the mixed layer or strongly stratified intermittent regimes require extensions or filters (Grachev et al., 2012, Heisel et al., 2023).

4. Documented departures from classical assumptions

Recent marine observations show especially direct failures of classical MOST assumptions. Using CLASI ASIS buoy data with measurements at typically $$u_* = \left(\overline{u'w'}^2 + \overline{v'w'}^2\right)^{1/4}, \qquad \theta_* = -\frac{\overline{w'\theta_v'}}{u_*},$$7 m and $$u_* = \left(\overline{u'w'}^2 + \overline{v'w'}^2\right)^{1/4}, \qquad \theta_* = -\frac{\overline{w'\theta_v'}}{u_*},$$8 m above the sea surface, wind speed decreased with height in $$u_* = \left(\overline{u'w'}^2 + \overline{v'w'}^2\right)^{1/4}, \qquad \theta_* = -\frac{\overline{w'\theta_v'}}{u_*},$$9 of all observations, contradicting the assumption of a monotonic wind profile. Large vertical gradients in sensible heat flux also occurred over only $$L = -\,\frac{u_*^3\,\overline{\theta_v}}{\kappa g\,\overline{w'\theta_v'}}, \qquad \zeta = \frac{z}{L},$$0 m, with “extreme” gradients defined as $$L = -\,\frac{u_*^3\,\overline{\theta_v}}{\kappa g\,\overline{w'\theta_v'}}, \qquad \zeta = \frac{z}{L},$$1, contrary to the constant-flux expectation. These anomalies were strongly modulated by coastal proximity and wind direction, with the highest occurrence rates near shore under offshore winds, and they co-occurred far more often than chance would predict, with odds ratio $$L = -\,\frac{u_*^3\,\overline{\theta_v}}{\kappa g\,\overline{w'\theta_v'}}, \qquad \zeta = \frac{z}{L},$$2 and $$L = -\,\frac{u_*^3\,\overline{\theta_v}}{\kappa g\,\overline{w'\theta_v'}}, \qquad \zeta = \frac{z}{L},$$3 confidence interval $$L = -\,\frac{u_*^3\,\overline{\theta_v}}{\kappa g\,\overline{w'\theta_v'}}, \qquad \zeta = \frac{z}{L},$$4 (Foxabbott et al., 8 Jul 2025).

The same study identified three distinct mechanisms for breakdowns in MOST assumptions: internal boundary layers formed by offshore continental flow, wave-driven wind jets associated with high wave age, and thermally stable boundary layers over cold sea surfaces where warm, moist air overlies cooler water far from shore. Quantitative thresholds were extracted for each regime, including nearshore offshore conditions with relative humidity $$L = -\,\frac{u_*^3\,\overline{\theta_v}}{\kappa g\,\overline{w'\theta_v'}}, \qquad \zeta = \frac{z}{L},$$5–$$L = -\,\frac{u_*^3\,\overline{\theta_v}}{\kappa g\,\overline{w'\theta_v'}}, \qquad \zeta = \frac{z}{L},$$6, air temperature $$L = -\,\frac{u_*^3\,\overline{\theta_v}}{\kappa g\,\overline{w'\theta_v'}}, \qquad \zeta = \frac{z}{L},$$7–$$L = -\,\frac{u_*^3\,\overline{\theta_v}}{\kappa g\,\overline{w'\theta_v'}}, \qquad \zeta = \frac{z}{L},$$8, and pressure $$L = -\,\frac{u_*^3\,\overline{\theta_v}}{\kappa g\,\overline{w'\theta_v'}}, \qquad \zeta = \frac{z}{L},$$9–$\kappa \approx 0.4$0; swell regimes with wave age $\kappa \approx 0.4$1–$\kappa \approx 0.4$2 and wind speed $\kappa \approx 0.4$3–$\kappa \approx 0.4$4; and stable offshore regimes with $\kappa \approx 0.4$5–$\kappa \approx 0.4$6, relative humidity $\kappa \approx 0.4$7–$\kappa \approx 0.4$8, and sensible heat flux from $\kappa \approx 0.4$9 to $z$0 (Foxabbott et al., 8 Jul 2025).

Stable-boundary-layer flux estimation provides another example. In winter observations from Utqiagvik and Wendell, conventional MOST underperformed observed eddy-covariance fluxes under stable conditions, while REA and the A22 mixing-length parameterization outperformed MOST. The paper attributes this to the frequent absence of MOST’s ideal conditions in stable flows, including intermittent turbulence, anisotropy, and departures from a constant-flux, stationary surface layer (Allouche et al., 2024).

These cases do not imply universal failure. The marine study explicitly states that MOST remains more robust in neutral to weakly stratified, horizontally homogeneous open-ocean conditions with moderate winds and lower relative humidity, and the Arctic pack-ice study shows that once supercritical stable cases are removed, classical local similarity re-emerges cleanly (Foxabbott et al., 8 Jul 2025, Grachev et al., 2012).

5. Generalizations beyond a single-parameter theory

A major contemporary direction is to generalize MOST by adding turbulence anisotropy as a second non-dimensional variable. In this framework, the Reynolds-stress anisotropy tensor is

$z$1

and a barycentric invariant $z$2 measures the degree of anisotropy, with $z$3 for highly anisotropic states and $z$4 for isotropy. Flux–gradient and variance relations then become functions of both $z$5 and $z$6, such as $z$7 and $z$8, rather than of $z$9 alone (Mosso et al., 2023, Stiperski et al., 2022).

Across five well-known datasets, anisotropy-augmented flux–gradient relations reduced scatter substantially. Relative to Högström-type reference functions, skill scores were reported as $L$0 for unstable $L$1, $L$2 for stable $L$3, $L$4 for unstable $L$5, and $L$6 for stable $L$7. The same framework also resolved the long-debated free-convection behavior of $L$8, yielding

$L$9

once anisotropy is accounted for, and implied that the turbulent Prandtl number tends to zero in the free-convection limit (Mosso et al., 2023).

The anisotropy program has expanded in two directions. First, a 47-site NEON analysis found that anisotropy-generalized MOST extends across vegetated canopies and complex terrain, with robust performance over a wide range of canopy and terrain configurations; the strongest systematic gains were for $L \to \infty$0 in stable regimes and for $L \to \infty$1 in unstable regimes (Waterman et al., 3 Feb 2025). Second, interpretable machine-learning work sought predictors of anisotropy itself and found that non-dimensional groups outperformed dimensional terrain descriptors. The dominant daytime predictor was the refined stability parameter

$L \to \infty$2

while the ratio $L \to \infty$3 or $L \to \infty$4 and rapid-distortion parameters dominated at night. Contrary to expectation, terrain variables were not found to significantly impact turbulence anisotropy directly (Mosso et al., 19 Mar 2025).

Other generalizations modify the similarity variable rather than the closure coefficients. Evidence from LES and CASES-99 supports a mixed stable parameter

$L \to \infty$5

which improved mean-profile similarity for wind speed and temperature relative to $L \to \infty$6 and yielded linear relations $L \to \infty$7 and $L \to \infty$8 over $L \to \infty$9 (Heisel et al., 2022). A different alternative uses the Dougherty–Ozmidov length

$L<0$0

leading to a DO-based stability parameter $L<0$1 and, under local equilibrium, the compact results $L<0$2 and $L<0$3 in the stable boundary layer (Grachev et al., 2014).

6. Applications, reinterpretations, and ongoing debates

MOST remains the standard entry point for bulk flux estimation, wall models, and boundary-layer parameterization, but its implementation has become increasingly regime aware. One example is the hybrid profile–gradient method, which exploits three-level wind or temperature measurements to form stability indices

$L<0$4

thereby allowing inversion for $L<0$5 from wind-only or temperature-only data. In noise-free Monte Carlo experiments, these hybrid methods were nearly equivalent to the full profile method and better than the gradient method, though less competitive in the presence of random errors (Basu, 2017).

Analytical models also continue to embed MOST within broader boundary-layer structure. In convective boundary layers, composite theories couple a MOST surface layer to perturbation-based flux profiles and a convective logarithmic friction law, yielding closed-form wind and potential-temperature-flux profiles across the entire CBL and agreement with LES over $L<0$6 (Liu et al., 2023). In stably stratified channels, MOST-like ideas have been reformulated with confinement-aware scalings, where the ratio $L<0$7 governs the layered structure and reconstructed velocity profiles predict the skin-friction coefficient within approximately $L<0$8 of DNS across most cases (Kotturshettar et al., 5 Aug 2025).

A notable current controversy concerns logarithmic laws under buoyancy. One study of stable atmospheric boundary layers argues that buoyancy does not destroy the logarithmic nature of the near-wall velocity profile but modifies its slope, with an effective $L<0$9 depending on $z$00 and $z$01 rather than on local $z$02 inside the log region (Cheng et al., 2022). A companion convective study argues similarly for temperature: the mean potential-temperature profile in the constant-flux layer remains logarithmic, while buoyancy modulates the slope through $z$03 rather than through a conventional $z$04 correction (Cheng et al., 2020). These results do not abolish MOST; they indicate that its canonical integrated correction functions may not always be the most effective description of buoyancy effects.

The modern interpretation of MOST is therefore dual. It remains the central framework for organizing surface-layer turbulence and for constructing operational flux algorithms, but it is no longer treated as a universally sufficient one-parameter closure. Coastal marine observations, stable intermittency, convective mixed-layer transitions, canopies, complex terrain, and anisotropy all demonstrate that additional state variables, filters, or regime logic are often required. A plausible implication is that future parameterizations will preserve MOST as the backbone of surface-layer scaling while conditioning its application on anisotropy, boundary-layer depth, wave state, or local flux divergence rather than on $z$05 alone (Foxabbott et al., 8 Jul 2025, Mosso et al., 2023, Mosso et al., 19 Mar 2025).