Skip to content

Barotropic, Baroclinic and Primitive-Equation Models

CSIR NET meteorology: barotropic model, Rossby–Haurwitz waves, quasi-geostrophic baroclinic models, primitive equations, σ coordinates, spectral models.

  • 20 min read
  • Updated
On this page · 14 sections
  1. Prerequisites
  2. The Hook: The First Successful Computer Forecast
  3. Barotropic and Baroclinic Atmospheres
  4. The Barotropic Vorticity Model
  5. Baroclinic Models: Quasi-Geostrophic Theory
  6. Primitive-Equation Models
  7. The Model Hierarchy at a Glance
  8. Memory Tricks
  9. Don’t Confuse
  10. Common Mistakes
  11. Practice
  12. Quick Revision Table
  13. Key Takeaways
  14. What’s Next

Prerequisites

The Hook: The First Successful Computer Forecast

Barotropic and Baroclinic Atmospheres

BarotropicBaroclinic
DefinitionDensity depends on pressure only: \( \rho = \rho(p) \)Density depends on pressure and temperature
Isobars and isotherms (isopycnals)ParallelIntersect
Temperature gradient on a pressure surfaceZeroNon-zero
Thermal windZero: geostrophic wind does not change with heightNon-zero: vertical wind shear
Energy source for disturbancesHorizontal shear of the flow (barotropic instability)Available potential energy of the horizontal temperature gradient (baroclinic instability)
Typical exampleTropical easterlies (nearly)Mid-latitude westerlies with fronts

The real mid-latitude atmosphere is baroclinic; that is why its weather systems grow. A barotropic model can only move existing systems about; it cannot create a new cyclone from a temperature contrast.

The Barotropic Vorticity Model

The equation

For horizontal, non-divergent flow on a single level, the wind is given by a streamfunction \( \psi \): \( u = -\partial\psi/\partial y \), \( v = \partial\psi/\partial x \), and the relative vorticity is \( \zeta = \nabla^2\psi \). The vorticity equation reduces to

$$\boxed{\frac{\partial \zeta}{\partial t} + \mathbf{V}_\psi \cdot \nabla(\zeta + f) = 0 \quad\Longleftrightarrow\quad \frac{D(\zeta + f)}{Dt} = 0}$$

or, with the Jacobian \( J(a, b) = a_x b_y - a_y b_x \),

$$\frac{\partial}{\partial t}\nabla^2\psi = -J(\psi, \nabla^2\psi + f).$$

Absolute vorticity is conserved following the motion. A parcel moving poleward (\( f \) increasing) gains anticyclonic relative vorticity; one moving equatorward gains cyclonic relative vorticity. That is the mechanism of Rossby waves.

The forecast procedure

graph TD
    A["Analysed 500 hPa heights"] --> B["Streamfunction ψ
(geostrophic: ψ = Φ/f₀)"] B --> C["ζ = ∇²ψ"] C --> D["Vorticity tendency
∂ζ/∂t = −J(ψ, ζ + f)"] D --> E["Solve Poisson equation
∇²(∂ψ/∂t) = ∂ζ/∂t"] E --> F["Step ψ forward in time
(leapfrog)"] F --> C

Each step needs an elliptic (Poisson) solve for the streamfunction tendency, with boundary conditions on the edge of a limited domain (or none on the globe). The time step is limited only by advection (tens of m/s), not by gravity waves, because the model has none.

Rossby–Haurwitz waves

Linearise about a uniform westerly \( U \) on a \( \beta \)-plane (\( \beta = df/dy = 2\Omega\cos\phi/a \)). A wave \( \psi' \propto e^{i(kx + ly - \nu t)} \) satisfies the dispersion relation

$$\boxed{c = \frac{\nu}{k} = U - \frac{\beta}{k^2 + l^2}}$$
  • Rossby waves move westward relative to the flow. Long waves (small \( k \)) retrogress fastest.
  • A wave is stationary when \( U = \beta/K^2 \), with \( K^2 = k^2 + l^2 \). For a purely zonal wave (\( l = 0 \)) the stationary wavelength is
$$L_s = 2\pi\sqrt{\frac{U}{\beta}}.$$

For \( U = 20 \) m/s at 45° (\( \beta \approx 1.62 \times 10^{-11} \) m⁻¹ s⁻¹), \( L_s \approx 6\,980 \) km. Waves shorter than \( L_s \) move east; longer ones move west.

The equivalent barotropic model and its limits

The real atmosphere is not barotropic, but in the mid-troposphere, near the level of non-divergence (roughly 500–600 hPa), divergence is small and the barotropic vorticity equation describes the movement of large-scale troughs and ridges reasonably well for a day or so. A model applied there, treating the flow as if barotropic, is called an equivalent barotropic model.

Its limitations follow from what it leaves out:

Missing ingredientConsequence
Temperature gradients (baroclinicity)No development: cyclones cannot form or intensify by baroclinic instability
Vertical motion and divergenceNo weather (cloud, rain); no stretching of vortex tubes
Divergence of the very long wavesThe ultralong waves retrogress far too fast; early operational models added a divergence term (the divergent barotropic model) to slow them
Diabatic heating, frictionNo energy input or loss

The divergent barotropic or shallow-water model keeps a free surface (height \( h \)), so it has divergence and external gravity waves with speed \( \sqrt{gH} \). It is the simplest model that shows geostrophic adjustment, and the standard test-bed for numerical schemes for PE models.

Baroclinic Models: Quasi-Geostrophic Theory

The QG equations

Quasi-geostrophic (QG) theory applies when the Rossby number \( Ro = U/(fL) \) is small: the flow is nearly geostrophic, horizontal scales are synoptic, and \( f \) is replaced by a constant \( f_0 \) except where it is differentiated (\( \beta \)-plane). In pressure coordinates with \( \omega = Dp/Dt \):

$$\text{Vorticity:}\quad \frac{\partial \zeta_g}{\partial t} = -\mathbf{V}_g \cdot \nabla(\zeta_g + f) + f_0\frac{\partial \omega}{\partial p}$$$$\text{Thermodynamic:}\quad \frac{\partial}{\partial t}\left(-\frac{\partial \Phi}{\partial p}\right) = -\mathbf{V}_g \cdot \nabla\left(-\frac{\partial \Phi}{\partial p}\right) + \sigma\omega$$

with \( \zeta_g = \nabla^2\Phi/f_0 \) and \( \sigma \) the static stability parameter. The divergent wind appears only through \( \omega \) in the stretching term; the advecting wind is geostrophic. Eliminating \( \omega \) gives a single prognostic equation for the quasi-geostrophic potential vorticity,

$$q = \frac{1}{f_0}\nabla^2\Phi + f + \frac{\partial}{\partial p}\left(\frac{f_0}{\sigma}\frac{\partial \Phi}{\partial p}\right), \qquad \frac{D_g q}{Dt} = 0,$$

and eliminating the time derivative gives the diagnostic omega equation for vertical motion. The model therefore predicts only the slow, balanced flow; gravity waves are filtered.

The two-level model

The simplest baroclinic model represents the atmosphere by two layers. A standard textbook layout (as in Holton’s An Introduction to Dynamic Meteorology):

LevelWhat is carried there
0 hPatop boundary: \( \omega = 0 \)
250 hPa (level 1)vorticity, streamfunction \( \psi_1 \)
500 hPa (level 2)\( \omega_2 \), temperature (thickness \( \propto \psi_1 - \psi_3 \))
750 hPa (level 3)vorticity, streamfunction \( \psi_3 \)
1000 hPalower boundary: \( \omega = 0 \) (flat ground, no friction)

With two levels the model has a vertical shear (a thermal wind, \( \propto \psi_1 - \psi_3 \)) and a temperature field, so it can represent:

  • Baroclinic instability: waves longer than a short-wave cutoff grow when the vertical shear exceeds a critical value, drawing on available potential energy. Very short waves are stabilised by static stability; very long waves by \( \beta \).
  • Vertical motion: rising air ahead of upper troughs (warm advection, differential vorticity advection), sinking behind.
  • Energy conversion: from available potential energy to eddy kinetic energy, the energy cycle of mid-latitude weather.

Two levels is the minimum for baroclinic development: one level cannot carry vertical shear or temperature. Multi-level QG models add vertical resolution but no new physics.

N. A. Phillips’s two-level QG integration in the mid-1950s, driven by heating and friction, produced a realistic jet and growing eddies and is usually cited as the first general-circulation experiment. The general-circulation side of this story belongs to climate modelling.

Limitations of QG models

LimitationWhy
Poor in the tropics\( Ro \) is not small and \( f \to 0 \): geostrophy fails
Poor at fronts and jetsAgeostrophic advection matters; QG fronts are too slow to sharpen and form at the wrong level
Constant static stability \( \sigma \), \( f_0 \)Simplifies the elliptic equations but misses real variations
No gravity wavesGood for stability, but no geostrophic adjustment or mountain waves
Diagnostic \( \omega \) onlyMoist processes and heating handled crudely

The balance equations (non-linear balance between the wind and mass fields) sit between QG and PE models; they were used briefly before operational centres moved to PE models.

Primitive-Equation Models

The equation set

“Primitive equations” are the hydrostatic equations of motion with little further approximation. In pressure coordinates:

EquationForm (pressure coordinates)Role
Horizontal momentum\( \dfrac{D\mathbf{V}}{Dt} = -\nabla\Phi - f\mathbf{k}\times\mathbf{V} + \mathbf{F} \)prognostic \( u, v \)
Hydrostatic\( \dfrac{\partial \Phi}{\partial p} = -\dfrac{RT}{p} \)diagnostic \( \Phi \) (hypsometric)
Continuity\( \nabla\cdot\mathbf{V} + \dfrac{\partial \omega}{\partial p} = 0 \)diagnostic \( \omega \)
Thermodynamic\( \dfrac{DT}{Dt} - \dfrac{RT}{c_p p}\omega = \dfrac{Q}{c_p} \)prognostic \( T \)
Moisture\( \dfrac{Dq}{Dt} = S_q \)prognostic \( q \) (and cloud species)
Surface pressure tendency\( \dfrac{\partial p_s}{\partial t} = -\displaystyle\int_0^{p_s}\nabla\cdot\mathbf{V}\,dp \) (in σ-type coordinates)prognostic \( p_s \)

Prognostic variables (with a time-derivative equation): \( u, v, T, q \) and surface pressure. Diagnostic variables (computed from the others at each step): \( \omega \) (continuity), \( \Phi \) (hydrostatic), and the derived fields.

Compared with QG models, PE models:

  • use the actual wind, not the geostrophic wind, so they are valid in the tropics and at fronts;
  • include divergence and gravity waves explicitly, so they need semi-implicit time stepping and balanced initial data (initialization);
  • allow variable \( f \), static stability and full topography;
  • couple naturally to physical parameterizations.

All operational global NWP models are PE models, hydrostatic at grid lengths above roughly 10 km, and non-hydrostatic (full vertical momentum equation, with sound waves filtered or treated implicitly) at the kilometre scales of convection-permitting regional models.

Vertical coordinates

CoordinateDefinitionStrengthWeakness
Height \( z \)geometricnatural for non-hydrostatic modelscoordinate surfaces intersect topography
Pressure \( p \)isobaricsimple equations (continuity has no density)surfaces intersect mountains; lower boundary awkward
Sigma \( \sigma = p/p_s \)1 at the ground, 0 at the top (introduced by Phillips)lower boundary is a coordinate surface: \( \dot\sigma = 0 \) at \( \sigma = 1 \)pressure-gradient force is a small difference of two large terms over steep slopes
Hybrid σ–p (η)terrain-following near the ground, relaxing to pressure aloftcombines bothslightly more complex
Potential temperature \( \theta \)isentropicadiabatic flow stays on surfaces; good for tracers and PVsurfaces intersect the ground; poor in neutral layers

In σ coordinates the horizontal pressure-gradient force becomes

$$-\nabla_p\Phi = -\nabla_\sigma\Phi - RT\,\nabla_\sigma \ln p_s,$$

two terms that nearly cancel over a mountain slope. Small errors in either produce a spurious force; this pressure-gradient error is the classic weakness of terrain-following coordinates and one reason for hybrid coordinates.

Horizontal discretisation: grid point versus spectral

Grid-point (finite difference / finite volume)Spectral (transform method)
Representationvalues at grid points (Arakawa grids)coefficients of spherical harmonics \( Y_n^m \)
Horizontal derivativesfinite differencesexact for the resolved waves
Nonlinear termscomputed at grid pointstransform to a Gaussian grid, multiply, transform back
Pole problemneeds filtering or a special gridnone: uniform resolution on the sphere
Aliasingneeds conservative differencing or filtersavoided with a quadratic (alias-free) transform grid
Semi-implicit Helmholtz solveiterative elliptic solvertrivial: \( \nabla^2 \) is diagonal in spectral space
Physicsat grid pointsat transform-grid points
Weaknessesphase errors for short wavesGibbs ripples near sharp features (rain, mountains); transform cost grows quickly with resolution

Truncation. Spectral models are labelled by their truncation, for example T106 or T639: “T” for triangular truncation (all \( n \le N \), \( \lvert m\rvert \le n \)), which gives the same resolution in every direction on the sphere. The shortest resolved wavelength around a great circle is about \( 2\pi a/N \) (about 62.6 km for \( N = 639 \)); the grid spacing of the transform grid is smaller than this. “TL” denotes a linear grid, used with semi-Lagrangian advection. Rhomboidal truncation (R) was used in some early models.

Physical parameterizations

Processes smaller than the grid, or not described by the resolved equations, enter as source terms \( \mathbf{F} \), \( Q \) and \( S_q \):

ProcessWhat the scheme supplies
Radiation (short-wave and long-wave)heating rates; depends on clouds, water vapour, CO₂, O₃, aerosols
Cumulus convectionvertical redistribution of heat and moisture and convective rain (mass-flux or adjustment schemes)
Cloud microphysicscondensation, cloud water and ice, precipitation
Planetary boundary layerturbulent fluxes of momentum, heat and moisture
Land surfacesoil temperature and moisture, evapotranspiration, snow
Gravity-wave drag (orographic and non-orographic)momentum deposition by sub-grid waves

Convection parameterization is the main source of error in tropical and monsoon forecasts. At grid lengths of a few km, deep convection begins to be resolved explicitly (“convection-permitting”), and models must be non-hydrostatic.

Global and limited-area models

A global model has no lateral boundaries. A limited-area (regional) model covers a region at higher resolution and needs lateral boundary conditions from a global model. Fields near the boundary are relaxed towards the driving model over a few grid rows (a buffer zone, often following Davies’ relaxation method), to stop waves reflecting. Nesting can be one-way (driving model unaffected) or two-way.

The Model Hierarchy at a Glance

Barotropic (non-divergent)Shallow waterQG baroclinicPrimitive equation
Levels11 (free surface)≥ 2many
Conserved / predictedabsolute vorticityPV of a layerQG potential vorticity\( u, v, T, q, p_s \)
WavesRossbyRossby + gravityRossbyRossby + gravity (+ Lamb)
Developmentnonoyes (baroclinic instability)yes
Vertical motionnonenonediagnostic (omega equation)diagnostic (continuity)
Tropicsweakpossibleinvalidvalid
Gravity waves need controlnoyesnoyes (semi-implicit, initialization)

Memory Tricks

Don’t Confuse

ThisvsThat
Barotropic atmosphere: \( \rho = \rho(p) \), no thermal windEquivalent barotropic model: a real (baroclinic) atmosphere treated as barotropic near the level of non-divergence
Barotropic instability: energy from horizontal shearBaroclinic instability: energy from horizontal temperature gradient (APE)
QG: geostrophic wind advects, \( \omega \) diagnostic, Rossby waves onlyPE: actual wind advects, gravity waves present
Prognostic variable (has a tendency equation)Diagnostic variable (computed each step)
Stationary Rossby wavelength \( 2\pi\sqrt{U/\beta} \)Rossby radius of deformation \( NH/f \) (or \( \sqrt{gH}/f \))
Spectral “T639”: truncation wavenumberGrid spacing of the transform grid (finer than \( 2\pi a/N \))
Hydrostatic PE modelNon-hydrostatic model (needed at a few km resolution)

Common Mistakes

Practice

All questions below are original practice questions written for this page. They are not from any CSIR NET paper. Use \( \Omega = 7.292 \times 10^{-5} \) s⁻¹ and \( a = 6371 \) km.

Level 1: Foundation

  1. Which quantity does the non-divergent barotropic vorticity equation conserve following the motion?
  2. What is the minimum number of levels a QG model needs to represent baroclinic instability?
  3. Name two diagnostic variables of a hydrostatic primitive-equation model in pressure coordinates.
Answers
  1. Absolute vorticity \( \zeta + f \).
  2. Two. One level cannot carry vertical shear or a temperature field.
  3. Vertical velocity \( \omega \) (from continuity) and geopotential \( \Phi \) (from the hydrostatic equation).

Level 2: Part B Style

Practice Q1. A purely zonal Rossby wave of wavelength 6000 km sits in a uniform westerly of 15 m/s at 45° latitude. Its phase speed relative to the ground is closest to:

(a) −15 m/s (b) 0.2 m/s (c) 15 m/s (d) 29.8 m/s

Answer: (b) about 0.2 m/s

\( \beta = 2\Omega\cos 45°/a = 2(7.292 \times 10^{-5})(0.7071)/(6.371 \times 10^{6}) = 1.62 \times 10^{-11} \) m⁻¹ s⁻¹.

\( k = 2\pi/(6 \times 10^{6}) = 1.047 \times 10^{-6} \) m⁻¹; \( k^2 = 1.097 \times 10^{-12} \) m⁻².

\( \beta/k^2 = 14.76 \) m/s, so \( c = 15 - 14.76 \approx 0.24 \) m/s. The wave is almost stationary. Option (d) adds instead of subtracting.

Practice Q2. In a primitive-equation model on σ levels, which set is entirely prognostic?

(a) \( u, v, \omega, \Phi \) (b) \( u, v, T, q, p_s \) (c) \( \zeta, \psi, \omega \) (d) \( T, \Phi, \omega, q \)

Answer: (b)

Momentum, thermodynamic, moisture and surface-pressure tendency equations give \( u, v, T, q, p_s \). Vertical velocity and geopotential are diagnosed.

Practice Q3. A spectral model uses triangular truncation T639. The shortest wavelength it represents along a great circle is about:

(a) 16 km (b) 31 km (c) 63 km (d) 125 km

Answer: (c) about 63 km

\( 2\pi a/N = 2\pi(6371)/639 \approx 62.6 \) km. The transform grid spacing is finer than this, which is why resolution quoted “in km” for a spectral model depends on the convention used.

Practice Q4. A station at the ground under a plateau has surface pressure 600 hPa. The model level \( \sigma = 0.5 \) above it lies at:

(a) 300 hPa (b) 500 hPa (c) 550 hPa (d) 600 hPa

Answer: (a) 300 hPa

\( p = \sigma p_s = 0.5 \times 600 = 300 \) hPa. Over sea level (\( p_s \approx 1000 \) hPa) the same level would be near 500 hPa: σ surfaces follow the terrain.

Level 3: Part C Style

Practice Q5. In a barotropic non-divergent flow, an air column with zero relative vorticity at 30° N drifts to 50° N. Consider:

  1. Its relative vorticity at 50° N is about \( -3.9 \times 10^{-5} \) s⁻¹.
  2. It has become anticyclonic.
  3. A column drifting from 50° N to 30° N with zero initial relative vorticity would become cyclonic.
  4. The result would be the same in a QG two-level model with strong ascent at the column’s level.

Which are correct?

(a) 1 and 2 only (b) 1, 2 and 3 only (c) 2 and 4 only (d) 1, 2, 3 and 4

Answer: (b) 1, 2 and 3 only

\( f_{30} = 2\Omega\sin 30° = 7.29 \times 10^{-5} \) s⁻¹; \( f_{50} = 2\Omega\sin 50° = 1.117 \times 10^{-4} \) s⁻¹.

Conservation: \( 0 + f_{30} = \zeta + f_{50} \Rightarrow \zeta = -3.88 \times 10^{-5} \) s⁻¹: negative, so anticyclonic in the northern hemisphere (1, 2 true). The reverse trip gives \( +3.88 \times 10^{-5} \) s⁻¹, cyclonic (3 true).

4 is false: with vertical motion the stretching term \( f_0\,\partial\omega/\partial p \) changes vorticity too, so absolute vorticity is no longer conserved.

Practice Q6. For a zonal flow \( U = 20 \) m/s at 45° N, which statements about barotropic Rossby waves (\( l = 0 \)) are correct?

  1. The stationary wavelength is about 7000 km.
  2. A 4000 km wave moves eastward relative to the ground.
  3. A 10 000 km wave moves westward relative to the ground.
  4. All Rossby waves move westward relative to the mean flow.

(a) 1 and 4 only (b) 1, 2 and 4 only (c) 1, 2, 3 and 4 (d) 2 and 3 only

Answer: (c) 1, 2, 3 and 4

\( L_s = 2\pi\sqrt{U/\beta} = 2\pi\sqrt{20/(1.62 \times 10^{-11})} \approx 6.98 \times 10^{6} \) m ≈ 7000 km (1 true).

Shorter than \( L_s \): \( \beta/k^2 < U \), so \( c > 0 \), eastward (2 true). Longer: \( \beta/k^2 > U \), \( c < 0 \), westward (3 true; this is the retrogression of ultralong waves). Relative to the flow, \( c - U = -\beta/k^2 < 0 \) always (4 true).

Practice Q7. A forecaster compares three models over a developing mid-latitude cyclone, a monsoon depression and a mountain-wave event. Which matching of model to failure is most defensible?

(a) Barotropic: fails to deepen the cyclone; QG two-level: misrepresents the monsoon depression; hydrostatic PE at 50 km: cannot resolve the non-hydrostatic mountain waves (b) Barotropic: fails on the monsoon depression only; QG: fails to deepen the cyclone; PE: no failure (c) QG: generates spurious gravity waves; PE: cannot represent baroclinic instability (d) All three handle the cyclone’s development equally well

Answer: (a)

The barotropic model has no baroclinic energy source, so no deepening. QG theory needs small Rossby number and \( f \) away from zero, so it is poor in the tropics. A hydrostatic model at 50 km resolves neither the short horizontal scale nor the non-hydrostatic dynamics of trapped mountain lee waves. (c) inverts the facts: QG filters gravity waves, and PE models capture baroclinic instability.

Quick Revision Table

IdeaFormula or fact
Barotropic\( \rho = \rho(p) \); no thermal wind; no development
Barotropic vorticity equation\( \partial_t\nabla^2\psi = -J(\psi, \nabla^2\psi + f) \); \( D(\zeta + f)/Dt = 0 \)
Rossby–Haurwitz\( c = U - \beta/(k^2 + l^2) \); \( \beta = 2\Omega\cos\phi/a \)
Stationary wavelength\( L_s = 2\pi\sqrt{U/\beta} \) (about 7000 km for 20 m/s at 45°)
Equivalent barotropic levelnear the level of non-divergence, about 500–600 hPa
Shallow wateradds divergence and gravity waves \( \sqrt{gH} \)
QGgeostrophic advection; \( \omega \) from the omega equation; conserves QGPV; gravity waves filtered
Two-level model250 and 750 hPa vorticity, \( \omega \) at 500 hPa; minimum for baroclinic instability
PE prognostic\( u, v, T, q, p_s \)
PE diagnostic\( \omega \) (continuity), \( \Phi \) (hydrostatic)
σ coordinate\( \sigma = p/p_s \); \( \dot\sigma = 0 \) at the ground; pressure-gradient error over slopes
Spectralspherical harmonics; T = triangular; shortest wavelength \( \approx 2\pi a/N \)
Limited-area modelneeds lateral boundary conditions; relaxation zone
Non-hydrostaticneeded at grid lengths of a few km

Key Takeaways

What’s Next