## 1. Introduction

The Met Office’s Unified Model (MetUM; Martin et al. 2006), used for climate and weather prediction, is one of the few general circulation models (GCMs) based on the general, nonhydrostatic primitive equations. Another is the nonhydrostatic icosahedral atmospheric model (NICAM) described by Tomita and Satoh (2004). Before suitable semi-implicit numerical methods were developed, use of the general equations in GCMs was not practicable and the hydrostatic primitive equations (HPEs) were the preferred option. The HPEs, given in log-pressure coordinates by Andrews et al. (1987, hereafter AHL), for example, can be derived from the general equations by making the approximations described in White et al. (2005) and given later in this paper. They are still used today by most GCMs.

The general nonhydrostatic primitive equations, unlike the HPEs, are formulated in a Euclidean space, which is advantageous since this better represents the actual atmosphere. Further, the greater accuracy of the general primitive equations allows the modeling of subtle physical processes that are not captured when using the HPEs. An example (White and Bromley 1995) is the zonal velocity arising from conservation of absolute axial angular momentum of a fluid parcel moving in the vertical. We therefore expect that as computing power continues to increase, more GCMs will switch to using the general equations rather than the HPEs.

Conceptually, the meridional circulation from equator to pole can be split into two parts: a balanced part due to differential heating and a wave-driven part.^{1} The primitive equations can be rearranged to form the so-called transformed Eulerian mean (TEM) equations in which the wave-driven (hereafter residual) circulation is made explicit. The wave driving (at least, that by resolved waves) is studied using the Eliassen–Palm (EP) diagnostics. The TEM equations and EP diagnostics are given by AHL for the HPEs in log-pressure coordinates, following the pressure-coordinate versions derived by Andrews and McIntyre (1978a, hereafter AM). The formalism given in AHL is that commonly used by the stratospheric community.

**∇**·

**F**is the EP flux divergence,

*A*is the wave activity density,

*D*denotes dissipation due to nonconservative (i.e., frictional and diabatic) effects, and

*α*is a measure of wave amplitude (AHL, p. 131). This relation, when used with the TEM equations, leads to the

*nonacceleration theorem*:

**∇**·

**F**= 0 in a steady, conservative flow, and under such conditionsis a solution of the TEM equations under appropriate boundary conditions (AHL, p. 132). Here,

*u*

*θ*

*υ*

*w*

*A*and

*D*in the HPE case are given in AHL, and the general, nonhydrostatic, version of the nonacceleration theorem is given in AM [see their Eq. (6.23a)].

AM also derive the TEM equations and EP diagnostics in the general, nonhydrostatic case (appropriate for the MetUM) and in a coordinate-independent form. The present paper shows that a naive application of AHL’s HPE diagnostics, in log-pressure coordinates, to a GCM based on the general, nonhydrostatic equations, in geometric coordinates, gives results that are significantly different from those obtained with the exact diagnostics of AM. Since there is a significant difference, the residual mean circulation calculated from the AHL equations will not satisfy the nonacceleration theorem for steady, conservative waves and thus could give misleading results in such GCMs. Further, we show physically why this difference occurs. The aim of this paper, then, is to provide advice to modeling groups as to which formulation of the TEM equations can be meaningfully applied to their model output. The paper, therefore, includes the equations of AM presented in a way that is directly applicable to general, nonhydrostatic GCM output.

One common use of the EP diagnostics in the modeling community is to compute the so-called downward control streamfunction (Haynes et al. 1991). In this paper, we therefore extend the downward control formulation to the general, nonhydrostatic case and consider differences between this general streamfunction and its more familiar approximate counterpart.

## 2. The TEM equations and EP diagnostics

### a. The general, nonhydrostatic equations

We derive EP diagnostics for the fully compressible, nonhydrostatic equations without making any “shallow atmosphere” assumption. We follow the procedure outlined in section 6 of AM but work in standard meteorological spherical coordinates *λ*, *ϕ*, and *r*, with corresponding velocity components *u*, *υ*, and *w*, rather than in the cylindrical coordinates used by AM.

*ρ*:where Ω is the earth’s rotation rate,

*p*is pressure,

*G*represents any external zonal force, andUsing mass conservation,where

_{λ}**u**is the velocity vector; thus, the term on the lhs of Eq. (3) can be writtenwhere subscripts preceded by a comma denote derivatives. This expression can be substituted in Eq. (3) to give the “flux form” of the zonal momentum equation; after collecting the terms in

*uυ*and

*uw*in the latter, and taking the zonal average at fixed

*ϕ*,

*r*, and

*t*(denoted by an overbar), we obtainFollowing AM, we put

*ρu*

*ρ u*

*ρ*′

*u*′

*u*′ =

*u*−

*u*

*ϕ*,

*r*, and

*t*). We separate the zonal means of triple products into “mean” and “eddy” parts as follows:Equation (5) can therefore be writtenwhere the eddy terms have been put on the rhs together with the forcing term. It is straightforward to show that the zonal mean of the mass conservation equation [Eq. (4)] can be written as follows:Equation (7) can be multiplied by

*u*

*ψ*defined in terms of the component of the mean eddy potential temperature flux along the mean isentropes:where

**e**

*is a unit vector in the*

_{λ}*λ*direction, and

*θ*is the potential temperature. Expanding the scalar triple product, we obtainAs noted by AM in their section 6c, there is flexibility in the choice of

*ψ*; this particular definition is independent of the (

*ϕ*,

*r*) coordinates. An alternative definition in terms of the northward potential temperature flux, that is, of the formhas been used extensively, especially for studies based on the hydrostatic primitive equations.

^{2}Figure 1 shows that for the month of January in the MetUM,

*ψ*and

*ψ*

_{1}are virtually identical. The difference between them is six orders of magnitude smaller than their absolute value (not shown). Difference streamfunctions are shown for January, which is the month that we simulate in this study (see section 3).

**u**

*υ*

*w*

**F**= (0,

*F*

^{(ϕ)},

*F*

^{(r)}) byThen, after some manipulation, we obtain the TEM zonal momentum equationwhereNote that the divergence of the EP flux

**∇**·

**F**appears as part of the eddy forcing in the TEM equation [Eq. (15)]. As shown by AM, the nonacceleration theorem, mentioned in section 1 of the present paper, follows from this TEM equation.

### b. Approximations to the general equations

#### 1) Hydrostatic primitive equation version, in geometric coordinates

- use geometric height
*z*=*r*−*a*as the vertical coordinate, where*a*is the mean radius of the earth, - omit all metric terms not involving tan
*ϕ*, - omit Coriolis terms that vary as cos
*ϕ*, and - replace
*r*with*a*in all terms that remain.

The corresponding versions of the EP diagnostics for the HPEs can then be derived in the same manner as in section 2a. We just quote the results here.

*ψ*is defined byThe EP flux divergence is given byThe TEM zonal momentum equation [Eq. (15)] is replaced byBelow, this formulation is referred to as HPE(

*z*).

#### 2) Hydrostatic primitive equation version, in log-pressure coordinates

*z*) approximation, a transformation from geometric height coordinates to log-pressure coordinates is required to obtain the familiar TEM equations and EP diagnostics of section 3.5 of AHL. To avoid confusion with geometric height

*z*, we denote log-pressure height by the symbol

*Z*, defined bywhere

*H*=

*RT*

_{r}g^{−1}is a suitable pressure scale height for the stratosphere (

*H*= 6800 m is used in this paper),

*p*is a reference pressure (1000 hPa here),

_{r}*T*is a reference temperature (232.4 K corresponds to the value of

_{r}*H*used here),

*R*= 287 J K

^{−1}kg

^{−1}is the specific gas constant for air, and

*g*is acceleration due to gravity. It is convenient to introduce a quantity

*ρ*

_{0}(

*Z*) defined bywhere

*ρ*=

_{r}*p*(

_{r}*RT*)

_{r}^{−1}(≈1.5 kg m

^{−3}here), chosen to ensure that the mass element

*ρdxdydz*in geometric coordinates equals

*ρ*

_{0}

*dxdydZ*in log-pressure coordinates, under hydrostatic balance. Note that

*ρ*

_{0}is proportional to

*p*; although its contribution in log-pressure coordinates is analogous to the density

*ρ*in geometric coordinates,

*ρ*

_{0}is not in general an approximation to

*ρ*. Indeed, it follows from the ideal-gas law

*p*=

*RTρ*that

*ρρ*

_{0}

^{−1}=

*T*

_{r}T^{−1}; thus,

*ρ*

_{0}differs significantly from

*ρ*wherever the temperature

*T*differs significantly from

*T*(e.g., at the ground).

_{r}*z*replaced by

*Z*, gives the following definitions. The residual mean meridional circulation is defined byNote that the overbar here denotes a zonal average at fixed

*ϕ*,

*Z*, and

*t*; since

*Z*surfaces are generally different from

*z*or

*r*surfaces, this average is not the same as that defined above, in sections 2a and 2b(1). In the same way, the deviations (…)′ differ from those defined above, which will introduce differences of

*O*(

*α*

^{2}) (see the appendix).

*ψ*. Figure 1 showed, for the middle atmosphere, that the equivalent of

*ψ*

_{1}used here is virtually identical to the form of

*ψ*used in the general case.)

Below, this formulation is referred to as HPE[ln(*p*)].

## 3. Numerical results

### a. Model description and setup

The version of the MetUM used in this paper is HadGAM1, the atmospheric component of the Hadley Centre Global Environmental Model, version 1, (HadGEM1); see Davies et al. (2005). As noted above, the model is a nonhydrostatic, fully compressible, deep atmosphere formulation. It uses a terrain-following, height-based vertical coordinate, semi-Lagrangian advection of all prognostic variables except density, an Arakawa-C horizontal grid (in which zonal and meridional velocity components and thermodynamic variables are staggered), and a Charney–Phillips vertical grid (in which vertical velocity and potential temperature are held at the same levels, staggered with respect to the levels at which the horizontal velocity components are held). The model resolution (as used here) is 1.25° latitude by 1.875° longitude (N96), with 60 levels ranging from the ground to 84.1 km. A time step of 20 min is used. Other details, such as radiation and gravity wave drag schemes, can be found in Martin et al. (2006).

An ensemble of five 1-month simulations is performed for January conditions. The initial conditions are taken from 1 January for five Januaries of a long multiyear control model run and are assumed to be independent. Daily mean dynamical fields are used to compute the EP diagnostics, which are then averaged over the five months to form the means presented in the following sections.

### b. Model results—EP diagnostics

*w*

*υ*

*F*

^{(ϕ)}and

*F*

^{(r)}from Eqs. (13) and (14), and

**∇**·

**F**from Eq. (16). We define SDIVF =

**∇**·

**F**/

*ρ*

*r*cos

*ϕ*as the scaled EP flux divergence (or force per unit mass) as it appears on the rhs of the momentum equation [Eq. (15)]. Here,

*z*is the vertical coordinate. The EP diagnostics corresponding to the hydrostatic primitive equations, as given in AHL, are shown in Figs. 3a,c,d,e,f and denoted HPE[ln(

*p*)];

*υ*

*w*

*F*

^{(ϕ)},

*F*

^{(Z)}, and

**∇**·

**F**are as defined by Eqs. (22), (23), (25), (26), and (27), respectively. In this HPE[ln(

*p*)] case we define SDIVF =

**∇**·

**F**/

*ρ*

_{0}

*a*cos

*ϕ*as on the rhs of the momentum equation [Eq. (24)]. Here,

*Z*(

*p*) is the vertical coordinate:The diagnostics, shown in the stratosphere and lower mesosphere, look qualitatively similar between the two cases; however, the residual circulation is larger in the general case.

Figure 4a shows the difference in SDIVF between the general and HPE[ln(*p*)] cases. A greater negative SDIVF in the general case is consistent with a larger residual circulation in that case. This difference, although subject to interannual variability, is remarkably similar when calculated for any of the five months simulated here (not shown) and therefore can be considered generic across the five simulations in this paper. The important point to note here is that there is a significant difference in SDIVF seen throughout the winter stratosphere. In fact, there is a difference in all the EP diagnostics between the general and HPE[ln(*p*)] cases throughout the winter lower stratosphere that is of the same magnitude as their absolute values (not shown).

Figures 4b and 4c show that these differences are due almost entirely to the differences in SDIVF between the HPE(*z*) and HPE[ln(*p*)] cases. The HPE(*z*) and HPE[ln(*p*)] cases are compared (Fig. 4b) by interpolating the HPE(*z*) diagnostics to pressure levels using Eq. (28). The interpolation could also have been done using the zonal mean of the modeled pressure field, *p*(*ϕ*, *z*, *t*); however, *p*(*ϕ*, *z*, *t*) is found to be essentially time and latitude independent, apart from a weak poleward downslope of pressure surfaces in the winter hemisphere extratropical stratosphere.^{3} Using *p*(*ϕ*, *z*, *t*) is found to make almost no difference to using *Z*(*p*) (not shown), so *Z*(*p*) is used for simplicity.

The difference between the general and HPE(*z*) cases is negligible (Fig. 4c), and this is true for all the EP diagnostics (not shown). In other words, the shallow atmosphere approximation and the hydrostatic approximation make essentially no difference to the EP diagnostics. *The major differences between the general and HPE[ln(p)] cases are thus due to the vertical coordinate used*. Both the general and HPE[ln(*p*)] formulations are self-consistent and exact in their respective coordinate systems (geometric and log-pressure height). However, recall that taking the zonal mean of quantities will lead to *O*(*α*^{2}) differences when transforming between coordinate systems (see the appendix). A difference between the general and HPE[ln(*p*)] cases is thus expected.

Figure 5 shows the difference in the EP diagnostics between the HPE(*z*) case (essentially identical to the general case, as noted in the previous paragraph), and the HPE[ln(*p*)] case. The much larger negative SDIVF in the HPE(*z*) case (Fig. 4b) is found to be due largely to a decreased ∂*F*^{(Z)}/∂*Z* (Fig. 5b) in the lower stratosphere. This causes a poleward and downward circulation in the extratropics in the difference fields (Figs. 5c,d), leading to a stronger residual circulation in the HPE(*z*) case.

The work of this section shows that the EP diagnostics appropriate to the general and HPE[ln(*p*)] formalisms of the primitive equations differ by the same order of magnitude as their absolute values throughout the winter stratosphere. For a given GCM, to compute the residual circulation that satisfies the nonacceleration theorem in that GCM, it is essential to compute EP diagnostics formulated within the same vertical coordinate system as that of the particular equations on which the GCM is based.

Note that this conclusion has implications for the validation of GCM output with reanalysis data, since reanalysis data is often only available on pressure levels. As discussed, the general or HPE(*z*) formulation of the TEM equations should be used for model level output in the case that a model is based on a geometric height coordinate, and this cannot be directly compared to the HPE[ln(*p*)] formulation, which should be used for pressure level-based reanalysis data. This is because of the differences that arise from the interpolation of TEM diagnostics from height coordinates to pressure coordinates and from the taking of zonal means in the two different coordinate systems. A solution is to interpolate the model dynamical fields from model to pressure levels at each time step during the model run and using the model pressure field *p*(*λ*, *ϕ*, *r*). This will give the most accurate dynamical output on pressure surfaces, which can then be meaningfully used in the HPE[ln(*p*)] formulation of the TEM equations and thus compared directly to the reanalysis data for use in model validation.

## 4. Effect on the wave-driven circulation

One of the most common ways in which the EP diagnostics are used is to compute the so-called downward control streamfunction (Haynes et al. 1991) and thus obtain a measure of the strength of the wave-driven meridional circulation. We now consider the extension of the downward control principle (Haynes et al. 1991) to the general case and the differences between the streamfunctions as computed in both general and HPE[ln(*p*)] cases.

### a. The downward control equations

#### 1) General, nonhydrostatic version

*ρ*

_{,t}= 0 and may therefore define a streamfunction Ψ(

*ϕ*,

*r*) for the residual circulation byThese expressions can be used to substitute for

*υ*

*w*

*u*

_{,t}= 0, we obtainwhere

*M*. It follows thatwhere the integral is taken along a contour

*ϕ*(

*r*), say, of constant

*M*. The residual velocity components can then be found by substituting for Ψ in Eq. (30), for example,

#### 2) Hydrostatic version in geometric coordinates

#### 3) Hydrostatic version in log-pressure coordinates

*ψ*multiplied by

*a*to maintain the same dimensions as in the previous cases. Now,where the integral is again taken along a contour

*ϕ*(

*Z*), say, of constant

*M*,

### b. Downward control results

Figure 6a shows Ψ as calculated from *w**p*)] case [Eqs. (35) and (36)]. Notice that the streamfunction calculated from *w**p*)] cases, demonstrates that the difference in streamfunction as calculated from *w**ρ**ρ*_{0} in Eq. (35). As already mentioned, *ρ**not* well approximated by *ρ*_{0}(*Z*) in general. The differences between the general and HPE(*z*) cases are again negligible (Fig. 6d).

## 5. Discussion and conclusions

In this paper it has been demonstrated that in a general circulation model (GCM) based on the general, nonhydrostatic primitive equations and based on a geometric height vertical coordinate, naive use of the approximated form of the transformed Eulerian mean (TEM) equations and Eliassen–Palm (EP) diagnostics given in AHL, denoted by the HPE[ln(*p*)] case, leads to misleading features in the EP diagnostics in the stratosphere that are of the same magnitude as the absolute values of those diagnostics. These features are largely due to the vertical coordinate system of the model output and TEM formalism used: geometric height in the general case and log-pressure height in the HPE[ln(*p*)] case. Differences between zonal mean quantities calculated in geometric and in log-pressure coordinates are on the order of wave amplitude squared. In the example shown in this paper, these differences lead to a slightly larger wave-driven circulation in the general case than is seen in the HPE[ln(*p*)] case. The point is made that to validate such models against reanalysis data, which is often available only on pressure surfaces, the model dynamical fields should be interpolated to pressure levels at each time step during the model run and using the modeled pressure field. The HPE[ln(*p*)] case could then be used with both model and reanalysis pressure level data. Reanalysis data is sometimes available only at coarse resolution, but little difference is found between the EP flux diagnostics as calculated from 1.25° latitude by 1.875° longitude resolution dynamical data and those calculated from 2.5° latitude by 3.75° longitude resolution dynamical data (not shown).

The downward control principle of Haynes et al. (1991) is found to extend easily to the general case. Differences between the general case and the HPE[ln(*p*)] case in the mass streamfunction, computed both from the residual vertical velocity *w**w**p*)] case in the Met Office GCM (the MetUM).

Modeling groups with GCMs based on the general primitive equations will, strictly speaking, only obtain EP diagnostics that obey mass conservation and the nonacceleration theorem if they use the general TEM equations of AM—Eqs. (15) and (16) here. However, the work of this paper demonstrates that using the so-called HPE(*z*) approximation [Eqs. (20) and (21)] gives virtually the same results as using the general TEM equations.

The implication for the modeling community is that to produce accurate EP diagnostics from geometric height-based model output, existing EP diagnostics written using the HPE equations of AHL [their Eqs. (3.5.2) and (3.5.3)] can be modified in a straightforward way to use the form of the HPEs appropriate for geometric height coordinates [Eqs. (20) and (21)]. Further generalization toward the AM equations gives little extra gain.

This work was supported by the Joint DECC and Defra Integrated Climate Programme–DECC/Defra (GA01101).

## REFERENCES

Andrews, D. G., , and M. E. McIntyre, 1978a: Generalized Eliassen–Palm and Charney–Drazin theorems for waves on axisymmetric mean flows in compressible atmospheres.

,*J. Atmos. Sci.***35****,**175–185.Andrews, D. G., , and M. E. McIntyre, 1978b: An exact theory of nonlinear waves on a Lagrangian-mean flow.

,*J. Fluid Mech.***89****,**609–646.Andrews, D. G., , J. R. Holton, , and C. B. Leovy, 1987:

*Middle Atmosphere Dynamics*. Academic Press, 489 pp.Davies, T., , M. J. P. Cullen, , A. J. Malcolm, , M. H. Mawson, , A. Staniforth, , A. A. White, , and N. Wood, 2005: A new dynamical core for the Met Office’s global and regional modelling of the atmosphere.

,*Quart. J. Roy. Meteor. Soc.***131****,**1759–1782.Glickman, T., Ed. 2000:

*Glossary of Meteorology*. 2nd ed. Amer. Meteor. Soc., 855 pp.Haynes, P. H., , C. J. Marks, , M. E. McIntyre, , T. G. Shepherd, , and K. P. Shine, 1991: On the “downward control” of extratropical diabatic circulations by eddy-induced mean zonal forces.

,*J. Atmos. Sci.***48****,**651–678.Held, I. M., , and T. Schneider, 1999: The surface branch of the zonally averaged mass transport circulation in the troposphere.

,*J. Atmos. Sci.***56****,**1688–1697.Martin, G. M., , M. A. Ringer, , V. D. Pope, , A. Jones, , C. Dearden, , and T. J. Hinton, 2006: The physical properties of the atmosphere in the new Hadley Centre Global Environmental Model (HadGEM1). Part I: Model description and global climatology.

,*J. Climate***19****,**1274–1301.Tomita, H., , and M. Satoh, 2004: A new dynamical framework of nonhydrostatic global model using the icosahedral grid.

,*Fluid Dyn. Res.***34****,**357–400.White, A. A., , and R. A. Bromley, 1995: Dynamically consistent, quasi-hydrostatic equations for global models with a complete representation of the Coriolis force.

,*Quart. J. Roy. Meteor. Soc.***121****,**399–418.White, A. A., , B. J. Hoskins, , I. Roulstone, , and A. Staniforth, 2005: Consistent approximate models of the global atmosphere: Shallow, deep, hydrostatic, quasi-hydrostatic, and non-hydrostatic.

,*Quart. J. Roy. Meteor. Soc.***131****,**2081–2107. doi:10.1256/qj.04.49.

# APPENDIX

## Comparing Zonal Averages on z Surfaces and p Surfaces

*χ*(

*λ*,

*ϕ*,

*z*,

*t*) expressed in geometric coordinates; for brevity we shall take the

*ϕ*and

*t*dependence as understood and write this quantity as

*χ*(

*λ*,

*z*). It can be separated into its zonal average at constant geometric height

*z*(this has been called an isohypsic average; see Glickman 2000, 421–422),and the deviation from this zonal average,Next, consider the zonal average of the same quantity

*χ*(

*λ*,

*z*) at constant pressure

*p*; we denote this isobaric average by

*χ̃*(

*p*). It is given bywhere

*z*(

*λ*,

*p*) is the geometric height of the isobar at

*p*. We can also define the isobaric deviation:

*χ*=

*z*(

*λ*,

*p*) we havewhereis the zonal-mean height of the isobar, and

*z`*is the deviation of the height from its zonal mean (see Fig. A1). We can therefore writeAssuming that

*z`*is small, say on the order of

*α*where

*α*≪ 1, we can Taylor-expand the integrand in (A1) to obtainwhere the subscript

*z*denotes a derivative. Then, putting

*χ*=

_{z}*χ*

_{z}+

*χ*′

_{z}and

*χ*=

_{zz}*χ*

_{zz}+

*χ*′

_{zz}, where the deviations

*χ*′

_{z}and

*χ*′

_{zz}are also assumed to be on the order of

*α*, we obtainThe

*ϕ*and

*t*dependence can be reinstated throughout if required.

Equation (A2) shows that for small-amplitude waves, there is an *O*(*α*^{2}) difference between *χ̃*(*p*) on the isobar at *p* and *χ**z̃*), taken at the mean height *z̃* of that isobar. This difference is similar (but not identical) in form to the difference between the Lagrangian mean and the Eulerian mean [the “Stokes correction”; see Andrews and McIntyre 1978b, their Eq. (2.27)]; however, its physical interpretation is not the same, since *z`* is not a vertical Lagrangian displacement.

This difference between zonal-mean quantities using different vertical coordinates must be borne in mind when comparing zonal-mean diagnostics, such as the geometric-coordinate expressions in sections 2a and 2b(1) and the log-pressure expressions in section 2b(2).^{A1} In practice, wave amplitudes may not be small, so there may be large differences between the two types of average.

^{1}

This split is not so clear-cut in practice. The differential heating itself can be partly forced by wave driving. Furthermore, alternative versions of the “difference streamfunction” given later in the paper show that there is no unique definition of the wave-driven (residual) circulation.

^{2}

By contrast, a difference streamfunction defined in terms of the vertical potential temperature flux is used by Held and Schneider (1999).

^{3}

This poleward downslope is expected insofar as the zonal wind and pressure fields are in geostrophic balance.

^{}

Note that the overbars in section 2b(2) refer to isobaric averages and the primes to isobaric deviations, equivalent to