Held–Larichev Baroclinic Eddy Diffusivity Scaling

Held–Larichev baroclinic eddy diffusivity scaling is an analytical closure and scaling theory in geophysical fluid dynamics, formulated by atmospheric physicist Isaac M. Held and physical oceanographer Vitaly D. Larichev in 1996 in the Journal of the Atmospheric Sciences, that relates the macroturbulent eddy diffusivity (D), root-mean-square eddy velocity (V), and energy-containing mixing length (L_mix) of horizontally homogeneous, baroclinically unstable flow on a beta plane to the vertical wind shear (U), the Rossby deformation radius (L_D), and the planetary vorticity gradient (β). By unifying Rick Salmon's (1980) two-layer triad phenomenology of geostrophic baroclinic turbulence with Peter B. Rhines's (1975) beta-plane arrest of the two-dimensional inverse energy cascade, Held and Larichev showed that in a strongly supercritical baroclinic flow, the potential vorticity mixing length is not fixed at the internal Rossby radius of deformation (L_D), but expands to the Rhines arrest scale (k_β⁻¹ ∼ L_D ξ), where ξ = f₀ U / (β L_D²) is the dimensionless baroclinic supercriticality parameter. Consequently, the macroturbulent eddy diffusivity scales as the square of the supercriticality times U L_D—or as the third power of the thermal-wind shear (D ∼ f₀² U³ / [β² L_D³])—and the poleward baroclinic eddy heat flux scales as the fourth power of the meridional temperature gradient (V′T′ ∝ [∂T̄/∂y]⁴).

Across general circulation theory and climate modeling, the Held–Larichev scaling bridged linear baroclinic instability closures—pioneered by John S. A. Green (1970) and Peter H. Stone (1972)—and nonlinear geostrophic turbulence. In physical oceanography, it provides the dynamical basis for flow-dependent thickness diffusivity in the Gent–McWilliams (GM) mesoscale eddy parameterization (Visbeck et al., 1997; Eden & Greatbatch, 2008), governing how coarse-resolution climate models represent poleward heat transport in midlatitude storm tracks, the Antarctic Circumpolar Current, and western boundary current extensions.

From Charney–Eady Linear Instability to Macroturbulent Fluxes

Following the discovery of linear baroclinic instability by Jule G. Charney (1947) and Eric T. Eady (1949), meteorologists recognized that synoptic-scale cyclones and anticyclones release available potential energy (APE) stored in the equator-to-pole temperature gradient. Through the thermal wind relation, a meridional temperature gradient (∂T̄/∂y) across a rotating fluid of Coriolis parameter f₀ maintains a vertical shear of the zonal mean wind (∂ū/∂z = −[g / (f₀ T₀)] ∂T̄/∂y). To parameterize unresolved eddies in climate and ocean models, theorists sought a closure expressing the meridional eddy heat flux (⟨v′T′⟩) and potential vorticity flux (⟨v′q′⟩) via a macroturbulent eddy diffusivity (D ∼ V L_mix):

⟨v′T′⟩ = −D (∂T̄/∂y), where V is the RMS meridional eddy velocity and L_mix is the mixing length.

Limitations of Linear Mixing-Length Closures: Green (1970) and Stone (1972)

Prior to 1996, two classical parameterizations dominated atmospheric and oceanic theory:

1. Stone's (1972) Eady-Scale Closure: Building on Eady (1949) on an f-plane (β = 0), Peter H. Stone assumed that L_mix is fixed at the internal Rossby radius of deformation (L_D = N H / f₀) and V saturates at the background shear U, yielding D_Stone ∼ U L_D and ⟨v′T′⟩ ∝ (∂T̄/∂y)² independent of β.
2. Green's (1970) Domain-Scale Closure: John S. A. Green argued that baroclinic eddies grow until bounded by the width of the baroclinic zone (L_zone), giving D_Green ∼ U L_zone.

Numerical simulations of quasigeostrophic turbulence in the 1980s and early 1990s (Rick Salmon, Geoffrey K. Vallis, Isaac M. Held, and Vitaly D. Larichev) showed that when L_D is much smaller than the domain size (as in the ocean, where L_D ≈ 15–50 km), nonlinear triad interactions transfer energy to larger horizontal scales until halted by Rossby wave radiation on the beta plane—making D dependent on both L_D and β.

Two-Layer Quasi-Geostrophic Formulation on a Beta Plane

Held and Larichev (1996) formulated their scaling within the Phillips (1954) two-layer quasi-geostrophic (QG) model on a doubly periodic beta plane ("homogeneous baroclinic turbulence") with uniform background shear +U in the upper layer and −U in the lower layer. Writing the barotropic streamfunction ψ = (ψ₁ + ψ₂) / 2 and baroclinic (thermal) streamfunction τ = (ψ₁ − ψ₂) / 2, the prognostic equations for barotropic (q_BT = ∇²ψ) and baroclinic (q_BC = ∇²τ − k_R² τ) potential vorticity are:

• Barotropic Mode: ∂(∇²ψ)/∂t + J(ψ, ∇²ψ) + J(τ, ∇²τ) + β (∂ψ/∂x) + U ∂(∇²τ)/∂x = −D_bottom(ψ)
• Baroclinic Mode: ∂(∇²τ − k_R² τ)/∂t + J(ψ, ∇²τ − k_R² τ) + J(τ, ∇²ψ) + β (∂τ/∂x) + U ∂(∇²ψ + k_R² ψ)/∂x = 0

Here J(A, B) is the horizontal Jacobian operator, D_bottom is bottom Ekman drag, k_R = L_D⁻¹ is the internal Rossby deformation wavenumber, and the mean layer potential vorticity gradients are ∂Q̄₁/∂y = β + k_R² U and ∂Q̄₂/∂y = β − k_R² U.

The Supercriticality Parameter (ξ)

By the Charney–Stern condition, the flow is baroclinically unstable when ∂Q̄₂/∂y < 0 (k_R² U > β). Held and Larichev defined the dimensionless supercriticality parameter ξ as:

ξ = (k_R² U) / β ∼ (f₀ U) / (β L_D²)

When ξ < 1, β stabilizes the flow at all wavenumbers. When ξ ≫ 1 (strongly supercritical regime), baroclinic shear drives geostrophic macroturbulence across a wide spectral range between k_R and a smaller arrest wavenumber k_β.

Salmon's (1980) Dual-Cascade Picture of Geostrophic Turbulence

To connect k_R to larger scales when ξ ≫ 1, Held and Larichev built upon the triad cascade architecture of Rick Salmon (1978, 1980):

1. Forward Baroclinic Cascade (k < k_R): At scales larger than the deformation radius (k ≪ k_R), the Jacobian J(ψ, −k_R² τ) shows that τ is advected like a passive scalar by the barotropic velocity v_BT = ẑ × ∇ψ, so available potential energy (APE) extracted at large scales (k ∼ k_β) cascades downscale from k_β toward k_R.
2. Deformation-Scale Conversion (k ≈ k_R): Near k ≈ k_R, baroclinic instability converts baroclinic potential energy into barotropic eddy energy (barotropic velocity variance, ½ |∇ψ|²), while sending potential enstrophy downscale (k > k_R).
3. Inverse Barotropic Energy Cascade (k_R → k_β): Conserved two-dimensional barotropic vorticity drives an inverse energy cascade toward larger horizontal scales (k < k_R) with a Kraichnan k^−5/3 spectrum, until halted at the Rhines arrest wavenumber (k_β) by Rossby wave dispersion and bottom friction.

Held and Larichev (1996) closed Salmon's dual-cascade loop by balancing the production of available potential energy at the arrest scale k_β against the inverse cascade dynamics. Let V denote the root-mean-square barotropic eddy velocity at wavenumber k_β, and let τ denote the root-mean-square baroclinic thermal streamfunction at k_β.

Step 1: Mixing Length Set by the Rhines Arrest Scale (L_mix ∼ k_β⁻¹)

In the inverse barotropic cascade, turbulent eddy turnover at wavenumber k has frequency ω_turb ∼ k V, whereas barotropic Rossby waves have frequency ω_Rossby ∼ β / k. The inverse cascade halts where ω_turb ≈ ω_Rossby, defining the Rhines wavenumber (k_β ∼ √[β / V]) and mixing length (L_mix ∼ k_β⁻¹ ∼ √[V / β]). Because the barotropic eddies at scale k_β⁻¹ stir the temperature field, the effective eddy diffusivity for baroclinic potential vorticity and heat is D ∼ V L_mix ∼ V k_β⁻¹.

Step 2: Eddy Energy Production Rate (ε) and Thermal Variance Balance

In statistical equilibrium, the rate of available potential energy generation (per unit mass) by baroclinic eddy heat fluxes extracting energy from the background vertical shear U is ε = k_R² U ⟨v_BT τ⟩ ∼ k_R² U² V k_β⁻¹. At the arrest scale k_β, the available potential energy per unit mass stored in the large-scale temperature fluctuations (E_BC ∼ k_R² τ²) cascades downscale toward k_R on the eddy turnover timescale T_eddy ∼ (V k_β)⁻¹, giving ε ∼ (k_R² τ²) (V k_β). Equating these two expressions for ε cancels k_R² V and yields τ ∼ U k_β⁻¹, meaning that the root-mean-square baroclinic thermal velocity at the arrest scale (k_β τ) equals the background thermal-wind shear (U).

Step 3: Equipartition at the Arrest Scale and Scaling of Eddy Velocity (V)

In Kraichnan's k^−5/3 inverse barotropic energy cascade from k_R down to k_β, the barotropic velocity variance at k_β obeys Kolmogorov–Kraichnan scaling (V² ∼ ε^2/3 k_β^−2/3, or ε ∼ V³ k_β). Comparing ε ∼ V³ k_β with ε ∼ k_R² τ² V k_β from Step 2 yields energy equipartition between barotropic velocity variance (V²) and available potential energy (k_R² τ²) at the arrest scale k_β: V ∼ k_R τ. Substituting τ ∼ U k_β⁻¹ into V ∼ k_R τ gives V ∼ U (k_R / k_β). Since k_β² = β / V, eliminating V against k_β yields the four scaling laws of Held and Larichev (1996):

1. Wavenumber Ratio and Arrest Scale: k_R / k_β ∼ ξ ⇒ L_mix = k_β⁻¹ ∼ L_D ξ = (f₀ U) / (β L_D)
2. RMS Barotropic Eddy Velocity: V ∼ U ξ = (k_R² U²) / β = (f₀² U²) / (β N² H²)
3. Baroclinic Eddy Diffusivity: D ∼ V k_β⁻¹ ∼ (U L_D) ξ² = (k_R³ U³) / β² ∼ (f₀² U³) / (β² L_D³)
4. Energy Generation Rate and Poleward Eddy Heat Flux: ε ∼ (k_R⁵ U⁵) / β², and ⟨v′T′⟩ ∼ −D (∂T̄/∂y) ∝ U⁴ ∝ (∂T̄/∂y)⁴

Sensitivity of Eddy Transport to Temperature Gradients and Stratification

The cubic dependence of the eddy diffusivity on vertical wind shear (D ∝ U³) and the quartic dependence of poleward eddy heat flux on the meridional temperature gradient (⟨v′T′⟩ ∝ [∂T̄/∂y]⁴) have two primary consequences for planetary climate dynamics:

• Nonlinear Baroclinic Adjustment: Because the eddy heat flux increases as the fourth power of ∂T̄/∂y, a 10% increase in radiative differential heating between the tropics and poles requires only a ~2.4% increase in the equilibrium meridional temperature gradient to boost poleward eddy heat transport by 10%. This power-law exponent provides a turbulent mechanism for Stone's (1978) "baroclinic adjustment", whereby midlatitude storm tracks act as a dynamical thermostat.
• Inverse Cube Sensitivity to Rossby Deformation Radius (D ∝ L_D⁻³): Unlike Stone's closure (where D ∝ L_D increases with stratification), the Held–Larichev diffusivity decreases as the deformation radius L_D (and static stability N) increases: D ∝ L_D⁻³ ∝ N⁻³. Stronger vertical stratification increases L_D, reduces ξ, and suppresses baroclinic eddy diffusivity.

Lower-Layer Potential Vorticity Mixing vs. Upper-Layer Jet Formation

In the supercritical limit (ξ ≫ 1), both layers have mean potential vorticity gradients of magnitude ~k_R² U, which is ξ times larger than β. At the energy-containing scale k_β, the planetary wave term β is small compared to k_R² U in the thermal mode, while β enters the problem by arresting the barotropic mode (where k_R² U cancels out of ∂Q̄_BT/∂y = β). In numerical realizations, this causes the eddies to organize into alternating zonal barotropic jets (Rhines jets) of meridional spacing ~k_β⁻¹ while homogenizing potential vorticity in the lower layer.

Bottom Friction Arrest and Zonal Jet Anisotropy

As shown by Larichev and Held (1995), K. Shafer Smith and Geoffrey K. Vallis (2002), and Brian K. Arbic and Glenn R. Flierl (2004), if bottom Ekman drag (r ∇²ψ) is moderate or strong, the inverse cascade is arrested at the frictional arrest wavenumber (k_r ∼ r / V) before reaching k_β, shifting the diffusivity from cubic scaling (D ∝ U³) toward D ∝ U² or U L_D. Furthermore, because Rossby wave dispersion vanishes for zonal flows (k_x = 0), energy funnels into zonal jets; only the meridional barotropic eddy velocity (v_BT) contributes to poleward diffusivity. Isaac M. Held (1999) and Tapio Schneider (2004) showed that in Earth's midlatitude troposphere, supercriticality is moderate (ξ ≈ 1.2 to 2.0), whereas on an f-plane Andrew F. Thompson and William R. Young (2006) showed that coherent barotropic vortices confirm that V and L_mix outgrow the linear Eady scales U and L_D.

Gent–McWilliams Ocean Eddy Parameterization and Southern Ocean Eddy Saturation

Because the oceanic first baroclinic Rossby radius of deformation is 10 to 50 km at mid- and high latitudes, coarse-grid climate models (50–100 km spacing) parameterize mesoscale eddies via the Gent–McWilliams (GM, 1990) thickness diffusivity κ_GM. Drawing on Held–Larichev (1996) and Green (1970)/Stone (1972), Martin Visbeck et al. (1997) and Carsten Eden and Richard J. Greatbatch (2008) introduced flow-dependent mesoscale eddy diffusivity closures combining Eady growth rates (σ_Eady = |S_iso| N) and Rhines mixing lengths, yielding κ_GM values from 100 m²/s in subtropical gyre interiors up to 3,000 to 10,000 m²/s in western boundary currents and the Antarctic Circumpolar Current (ACC). In the ACC, the Held–Larichev scaling (D ∝ U³ ∝ S_iso³) explains eddy saturation, whereby strengthening Southern Hemisphere westerly winds drive a large increase in transient mesoscale eddy energy and poleward buoyancy transport while leaving the time-mean isopycnal slope and circumpolar volume transport nearly invariant.

What is the baroclinic supercriticality parameter (ξ) in Held–Larichev theory?

The supercriticality parameter ξ = (k_R² U) / β = (f₀ U) / (β L_D²) is the dimensionless ratio of the destabilizing vertical shear (U / L_D²) to the stabilizing planetary vorticity gradient (β = ∂f/∂y). When ξ < 1, planetary beta suppresses baroclinic instability; when ξ ≫ 1, baroclinic eddies drive an inverse energy cascade to horizontal scales ξ times larger than the deformation radius L_D.

Why does Held–Larichev eddy diffusivity scale as the cube of the wind shear (U³) instead of linearly (U)?

In linear closures like Stone (1972), the eddy velocity V equals the thermal-wind shear U and the mixing length L_mix remains fixed at L_D (D ∼ U L_D). Held and Larichev showed that in nonlinear geostrophic turbulence, barotropic energy cascades upscale from L_D to the Rhines arrest scale L_mix ∼ L_D ξ while the RMS barotropic eddy velocity amplifies to V ∼ U ξ, making D ∼ V L_mix ∼ (U L_D) ξ² ∝ U³.

How does Held–Larichev scaling explain why the poleward heat flux scales as the fourth power of the temperature gradient?

Because the poleward baroclinic eddy heat flux is the product of the eddy diffusivity D and the background meridional temperature gradient ∂T̄/∂y (proportional to U by thermal wind balance), a diffusivity scaling as D ∝ U³ ∝ (∂T̄/∂y)³ produces a heat flux ⟨v′T′⟩ = −D (∂T̄/∂y) ∝ (∂T̄/∂y)⁴.

Why doesn't Earth's midlatitude atmosphere exhibit a pure U³ diffusivity scaling everywhere?

In Earth's midlatitude troposphere, the Rossby deformation radius (L_D ≈ 1,000 km) is comparable to the width of the baroclinic storm-track zone, so the atmospheric supercriticality is moderate (ξ ≈ 1.2 to 2.0) and surface Ekman friction arrests the inverse cascade early. In the ocean (L_D ≈ 15–50 km) and in gas-giant atmospheres, ξ can be much larger than 1, allowing a broader inverse cascade between L_D and the Rhines scale.

1. Held, I. M., & Larichev, V. D. (1996). "A scaling theory for horizontally homogeneous, baroclinically unstable flow on a beta plane." Journal of the Atmospheric Sciences, 53(7), 946–952. https://doi.org/10.1175/1520-0469(1996)053<0946:ASTFHH>2.0.CO;2053<0946:ASTFHH>2.0.CO;2)
2. Larichev, V. D., & Held, I. M. (1995). "Eddy amplitudes and fluxes in a homogeneous model of fully developed baroclinic instability." Journal of Physical Oceanography, 25(10), 2285–2297. https://doi.org/10.1175/1520-0485(1995)025<2285:EAAFIA>2.0.CO;2025<2285:EAAFIA>2.0.CO;2)
3. Salmon, R. (1980). "Baroclinic instability and geostrophic turbulence." Geophysical & Astrophysical Fluid Dynamics, 15(1), 167–211. https://doi.org/10.1080/03091928008241178
4. Visbeck, M., Marshall, J., Haine, T., & Spall, M. (1997). "Specification of eddy transfer coefficients in coarse-resolution ocean circulation models." Journal of Physical Oceanography, 27(3), 381–402. https://doi.org/10.1175/1520-0485(1997)027<0381:SOETCI>2.0.CO;2027<0381:SOETCI>2.0.CO;2)
5. Thompson, A. F., & Young, W. R. (2006). "Scaling baroclinic eddy fluxes: Vortices and energy balance." Journal of Physical Oceanography, 36(4), 720–738. https://doi.org/10.1175/JPO2874.1