## 1. Introduction

The motivation for this work originated in the observation of weak but distinct minima in total ozone over monsoonal systems (Zhou and Chao 1994; Zou 1996). Viewing a monsoonal system as being forced by diabatic heating in the troposphere (Ting 1994), one may expect a secondary circulation with upward air motion in the center of the heating. To the degree that the secondary circulation extends upward into the stratosphere, it would dynamically cause a local decrease of total ozone.

We shall completely refrain from dealing with the complexities of monsoonal systems. Instead, a highly idealized problem is considered as a step toward an understanding of the underlying basic dynamics. We investigate balanced flow of a stably stratified dry non-Boussinesq atmosphere on the *f* plane. The flow is assumed to be rotationally symmetric about an axis perpendicular to the *f* plane. The circulation is forced in the troposphere through thermal relaxation toward a specified equilibrium temperature and is damped through Rayleigh friction in the interior of the domain. In addition, surface friction is sufficiently strong to ensure weak surface winds.

Plumb and Hou (1992, hereafter referred to as PH92) have shown that there exists threshold behavior in a related problem. They studied the response to subtropical thermal forcing of a zonally symmetric atmosphere on the sphere. In the inviscid limit they obtained a thermal equilibrium (TE) solution for subcritical forcing and a so-called angular momentum conserving (AMC) solution for supercritical forcing. The latter is highly nonlinear and differs from the former in that there is a nonzero secondary circulation even in the inviscid limit. For a clear distinction we refer to PH92’s problem in the following as the “zonally symmetric problem,” while our current problem is called the “*f*-plane axisymmetric problem.” One major difference between the two idealizing geometries is that only the present one is able to represent truly local thermal forcing, which appears desirable in connection with monsoonal systems. Although the *f*-plane axisymmetric problem and the zonally symmetric problem share a number of features, noticeable differences arise, mostly because the thermal wind equation has a stronger nonlinearity and there is nonzero thermal forcing right on the axis of symmetry in the present case. Of course, real monsoonal systems are neither zonally symmetric nor *f* plane axisymmetric. Nevertheless, it is considered worthwhile to shed some light on the *f*-plane axisymmetric problem, too, since it represents another possible symmetry that allows a similar quasi-analytical approach to that of the zonally symmetric problem.

We will investigate the dependence of the flow on the Rayleigh friction parameter and consider, in particular, the frictionless limit. The motivation for this approach is similar to previous work (Schneider 1977; Held and Hou 1980, hereafter referred to as HH80; Lindzen and Hou 1988; Hou and Lindzen 1992; PH92): The complex circulation of the real atmosphere, being highly variable in both time and space, is idealized by constraining it to be stationary and symmetric. Comparing the frictionless limit of this idealized flow with the suitably averaged real flow then yields a particular view on the role of “eddies,” that is, deviations from stationarity and symmetry, which exist in real flows (cf. Schneider 1987; Becker et al. 1997). Unlike the above-quoted studies we assume a non-Boussinesq atmosphere extending far beyond the tropopause. This is similar to Dunkerton (1989), although the latter study was mainly concerned with stratospheric forcing, while we will focus on tropospheric forcing. The extension of the model domain beyond the tropopause allows us to study the penetration of the secondary circulation into the stratosphere, which may have implications for the dynamical modification of lower-stratospheric ozone or the transport of water vapor from the troposphere to the stratosphere. Similar to Dunkerton (1989), but in contrast to the other quoted studies, we will use an Eliassen balanced vortex model for the numerical calculations.

The paper is organized as follows. In section 2 we present the basic material including model equations, boundary conditions, and numerical methods of solution. Section 3 deals with threshold behavior in the frictionless limit and the subcritical regime (TE solution). Additional insight will be obtained in section 4 through an analysis of the linearized system of equations. The supercritical regime (AMC solution) is investigated in section 5. Section 6 provides a summary and discussion. Some of the more technical details are deferred to the appendixes.

## 2. The model

### a. Model equations

*f*plane for a dry fluid. The flow is assumed to be rotationally symmetric about the vertical axis. Moreover, it is assumed to be balanced in the sense that the equation for radial momentum can be replaced by the gradient wind equation [see (4) below]. As coordinates we use radius

*r*and log-

*p*altitude

*z*= −

*H*ln(

*p*/

*p*

_{0}), where

*p*is pressure,

*H*= 7000 m is a constant scale height, and

*p*

_{0}= 1000 hPa is a constant reference pressure. The domain considered is the whole volume above the surface. For the numerical calculations, the domain is restricted to

*r*⩽

*r*

_{max}= 2000 km and

*z*⩽

*z*

_{max}= 35 km, which was chosen to be considerably larger than the region of the thermal forcing (to be described below). Diabatic heating is modeled as Newtonian relaxation toward a specified temperature

*T*

_{e}(

*r, z*) with a constant relaxation coefficient

*α*

_{n}. Damping of tangential momentum is modeled as Rayleigh friction with the coefficient

*α*

_{r}. The equations for steady flow are

*u,*

*υ*, and

*w*denote the tangential, radial and vertical wind, respectively; Φ is the geopotential;

*T*is the temperature;

*ρ*

_{0}(

*z*) =

*p*(

*z*)/(

*gH*);

*f*is the constant Coriolis parameter;

*g*is the gravity acceleration;

*T*

_{s}=

*gH*/

*R*;

*R*is the gas constant for dry air;

*κ*=

*R*/

*c*

_{p}= 2/7; and

*c*

_{p}is the specific heat at constant pressure. The symbols

*X*and

*Q*denote the nonconservative terms in the equation for tangential momentum and temperature, respectively. The fields

*u*and

*T*characterize the (primary) vortex and its circulation. Combining (4) and (5) yields the thermal wind equation

*T*

_{e}(

*r, z*) with ∂

*T*

_{e}/∂

*r*≠ 0 in a finite subdomain (restricted to the troposphere) will be called thermal forcing in the following (see section 2b). The term “forcing” does not imply a cause-and-effect relationship. The resulting diabatic heating or cooling

*Q*(

*r, z*) is part of the solution and not given a priori. In our whole study we restrict the attention to forcing that is weak enough so that the atmosphere remains stably stratified, that is,

*N*is the Brunt–Väisälä frequency.

The timescale for thermal relaxation is assumed to be considerably longer than a day but considerably shorter than a month. We choose *α*_{n} = 0.1 day^{−1} and keep this value fixed throughout our calculations. The coefficient *α*_{r}, on the other hand, is varied within the range 0 ⩽ *α*_{r} ⩽ 0.1 day^{−1}. In our attempt to study the nearly frictionless behavior, we pushed the value of *α*_{r} in the numerical calculations as close as possible toward zero without violating a certain constraint guaranteeing the existence of the numerical solution (see below). The remaining parameters were chosen as *g* = 9.81 m s^{−2}, *R* = 287 J kg^{−1} K^{−1}, and *f* = 7.292 × 10^{−5} s^{−1} (corresponding to a latitude of 30°N).

*θ*

*Te*

^{κz/H}

*ω*=

*u*/

*r*denotes the angular velocity of the flow. Whenever we talk about “angular momentum” in the context of the

*f*-plane axisymmetric model, we mean absolute angular momentum

*m*as defined above.

*τ*

_{s}is modeled through a linear drag law according to

*τ*

_{s}

*c*

_{d}

*u*

_{s}

*u*

_{s}is the tangential wind at

*z*= 0 and

*c*

_{d}is a drag coefficient. We used the value

*c*

_{d}= 5 × 10

^{−3}m s

^{−1}throughout our numerical calculations. Whenever we talk about the frictionless limit, we mean

*α*

_{r}→ 0 with

*c*

_{d}remaining constant. The other boundaries are assumed to be stress free.

*υ*and

*w*) can be described in terms of a cross-vortex streamfunction

*ψ*(

*r, z*), which is defined through

*Q*and

*X*as given, one can—by forming

*gT*

^{−1}

_{s}

*r*− ∂[(

*f*+ 2

*u*/

*r*)(1)]/∂

*z*—derive the following linear partial differential equation for

*ψ*:

*AC*−

*B*

^{2}) > 0, which had to be satisfied in all numerically determined solutions and, thus, restricted the range of values for

*α*

_{r}that could be used. Equation (13) allows one to diagnose the secondary circulation (

*ψ*) from knowledge about the primary vortex circulation (

*u*and

*T*) and the nonconservative terms (

*Q*and

*X*). As boundary conditions we specify

*ψ*= 0 at

*r*= 0 and at

*z*=

*z*

_{max}, and

*w*= 0 at

*r*=

*r*

_{max}. Since our domain is considerably larger than the size of the forcing region, our results are not sensitively dependent on the choice of the upper and outer boundary condition. At the lower boundary, the drag law (11) can approximately be translated to the following condition for

*ψ*(cf. Schubert and Hack 1983; Wirth 1995):

*z*= 0 is interpreted as the top of the boundary layer. Assuming that the boundary layer is infinitely shallow, the actual surface is at

*z*= 0

_{−}.

### b. Basic state and thermal forcing

If the equilibrium temperature *T*_{e} is a function of altitude only, that is, *T*_{e} = *T*_{o}(*z*), a possible solution of the problem (1)–(5) plus boundary conditions is given by (*u,* *υ*, *w*) = (0, 0, 0) and *T* = *T*_{o}. One can view *T*_{o}(*z*) as a reference temperature profile in the absence of thermal forcing. It is specified through the corresponding potential temperature profile *θ*_{o}(*z*) as piecewise linear in *z* with *θ*_{o} = 300 K at *z* = 0, ∂*θ*_{o}/∂*z* = 4.375 K km^{−1} for 0 ⩽ *z* ⩽ *z*_{tp}, and ∂*θ*_{o}/∂*z* = 38 K km^{−1} for *z*_{tp} ⩽ *z* ⩽ *z*_{max}. The symbol *z*_{tp} denotes the height of the tropopause, which is taken to be at *z*_{tp} = 16 km.

*T*

_{e}(

*r, z*) has to deviate from the reference temperature

*T*

_{o}(

*z*). The deviation equilibrium temperature

*T*

^{′}

_{e}

*r, z*) =

*T*

_{e}(

*r, z*) −

*T*

_{o}(

*z*) is specified as

*r*

_{o}= 1000 km and

*z*

_{o}=

*z*

_{tp}= 16 km. The quantity

*T*

_{eo}is called amplitude of the thermal forcing. We restrict our attention to

*T*

_{eo}≥ 0, which means that we impose heating rather than cooling in the vortex center. Because of the strong nonlinearity of the model equations, some of our results do not apply to

*T*

_{eo}< 0.

Figure 1 shows the forcing *T*^{′}_{e}*r, z*) for *T*_{eo} = 0.5 K (solid contours) and the reference potential temperature *θ*_{o}(*z*) (dashed contours). Note that in this and all the following altitude–radius sections only the lowermost 25 km of the computational domain are displayed. The gray shading denotes the stratosphere, which is defined via potential vorticity *P* as the region where *P* > 4 PVU (with 1 PVU = 10^{−6} K m^{2} kg^{−1} s^{−1}).

### c. Numerical solution

It is possible to use (13) even in a nonstationary initial value problem, as long as the corresponding time evolution is slow (Eliassen 1952). This was exploited for solving the stationary problem (1)–(5) numerically by starting from a state at rest and integrating the time-dependent version of (1)–(5) until a steady state was reached. At each time step the cross-vortex streamfunction *ψ* was diagnosed by solving (13). Then the equation for tangential momentum was integrated forward by one time step using a second-order Adams–Bashforsh scheme. In addition, the temperature at the outer boundary was integrated forward in time. Finally, the temperature in the interior of the domain was calculated from the thermal wind equation (6).

One major advantage of a balanced model is that it allows a large time step. Using a grid spacing of Δ*r* = 59 km and Δ*z* = 530 m, we could typically take Δ*t* = 24 h without incurring numerical instability. A more complete description of the numerical details can be found in Wirth (1995).

## 3. Threshold behavior in the frictionless limit

*T*=

*T*

_{e}and (

*υ*,

*w*) = (0, 0) the TE solution. This solution trivially satisfies (1), (2), and (3) for any

*T*

_{e}(

*r, z*). Its existence hinges on whether the thermal wind equation (6) can be satisfied with

*T*=

*T*

_{e}. Assuming that

*r*

^{−1}∂

*T*/∂

*r*is finite for

*r*→ 0, (6) can be rewritten as

*ω*= 0 at

*z*= 0, (20) is integrated to give

*r, z*

*f*

^{2}

*ω*becomes imaginary somewhere. Relation (23) yields a criterion for the existence of the TE solution through simply replacing

*T*by

*T*

_{e}in the definition of Λ. With our forcing (19) one obtains

*e*denotes the TE solution and where

*r*

_{o}] × [0,

*z*

_{o}] whenever

*T*

_{eo}>

*T*

_{c}. Any forcing that satisfies condition (23) will be called subcritical, otherwise it will be called supercritical.

The dependence of the threshold temperature on the Coriolis parameter *f* and the parameters defining the geometry of the forcing (*r*_{o} and *z*_{o}) in (25) is essentially the same as in the zonally symmetric problem of PH92. For the values used in this study one obtains *T*_{c} = 0.645 K. This can be considered as a very low threshold, and typical synoptic-scale thermal forcing in subtropical monsoonal systems is most likely supercritical. It follows that the TE solution is probably of little relevance for such real atmospheric conditions, as was already pointed out by PH92.

For supercritical forcing the TE solution does not exist. Instead, in the frictionless limit one obtains the so-called AMC solution, which will be discussed in more detail in section 5. One can also show that the AMC solution does not exist as long as the TE solution exists. Roughly speaking, in that case the term ∂*T*/∂*r* is too small for the thermal wind equation (20) to be consistent with *ω* = 0 at the bottom and *ω* = −*f*/2 at the top, which both have to be satisfied for the AMC solution. A formal proof is provided in appendix A. In short, the TE and the AMC solution are mutually exclusive. Incidentally, the “breakdown of the gradient wind balance,” that is, the nonexistence of the TE solution, was pointed out by Shine (1987) in the context of middle atmosphere thermal equilibrium temperatures near the summer pole.

There is marked difference concerning the threshold behavior in our case and in the zonally symmetric case of PH92. It is related to the different geometry implied by our symmetry. Both here and in PH92, the thermal wind equation is nonlinear, but the nonlinearity in PH92 is much weaker, since it involves the radius of the earth rather than the radius of the vortex in the denominator of one of its terms [cf. Eq. (5) in PH92 with our Eq. (6)]. Supercritical forcing in a zonally symmetric atmosophere means that, while the TE solution may well exist, the angular momentum about the earth’s axis of rotation has a local extremum somewhere in the interior of the atmosphere. It follows that for diffusive friction the TE solution is not a regular solution of the equations; that is, it does not represent the inviscid limit of a solution with small but nonzero viscosity. On the other hand, supercritical forcing in our case means that the TE solution simply ceases to exist. In fact, as long as it exists, it is regular everywhere in the above sense (the proof is given in appendix B). In a certain sense, the strong nonlinearity of our thermal wind equation makes the problem easier.

The vortex circulation *u*_{e} = *r**ω*_{e} of the TE solution (24) is plotted in Fig. 2 together with the corresponding balanced potential temperature *θ*_{e} for *T*_{eo} = 0.5 K. The balanced wind *u*_{e} reflects the forcing *T*^{′}_{e}*α*_{r} = 2 × 10^{−3} day^{−1}. Apparently, the surface friction is strong enough to keep the surface wind close to zero. Comparison of Fig. 3a with Fig. 2 shows how close the “nearly frictionless” numerical solution is to the truly frictionless limit. This can be seen more systematically in Fig. 4a, where the solid line is *u*_{e}(*r, z*_{o}) and the dashed lines represent numerical calculations of *u*(*r, z*_{o}) for different values for *α*_{r}. Apparently, as the value of *α*_{r} decreases (increasing dash length), the numerical solution approaches the TE solution.

*N*

^{2}across the tropopause is not visible in the plot of

*ψ*, but it would become noticeable in a plot of

*υ*∝ ∂

*ψ*/∂

*z*(not shown); we will come back to this point in the following section. The vertically integrated radial heat flux due to the secondary circulation is defined as

*h*(

*r*) = 0 for all values of

*r,*since

*ψ*= 0 everywhere. Figure 4b demonstrates, again, that the numerical solution (dashed lines) approaches the frictionless limit (solid line) as

*α*

_{r}→ 0.

## 4. The linear regime

*ζ*

_{a}. As a coefficient multiplying

*υ*in (1),

*ζ*

_{a}determines how strongly the radial flow interacts with the primary vortex circulation and, as a consequence, how much friction

*X*is needed to keep the system stationary in the presence of a secondary circulation. Similarly, in the case of solid body rotation the first factor on the left-hand side of the thermal wind equation (6) is the absolute vorticity. The ratio

*ζ*

_{a}/

*f*can, therefore, be taken as a measure of linearity that is relevant for both the (primary) tangential flow and its interaction with the secondary circulation. It ranges from

*ζ*

_{a}/

*f*≈ 1 (approximately linear) to

*ζ*

_{a}/

*f*≈ 0 (fully nonlinear). Figure 5 demonstrates how

*ζ*

^{min}

_{a}

*f*depends on the forcing amplitude

*T*

_{eo}and the Rayleigh friction coefficient, where

*ζ*

^{min}

_{a}

*α*

_{r}. For

*α*

_{r}= 10

^{−1}day

^{−1}(short dashes),

*ζ*

^{min}

_{a}

*f*within the considered range of forcing amplitudes, which means that the solution is in a linear regime to a good approximation. On the other hand, for

*α*

_{r}= 2 × 10

^{−3}day

^{−1}(long dashes) one obtains

*ζ*

^{min}

_{a}

*T*

_{eo}= 5 K, indicating that this combination of forcing and friction leads to a highly nonlinear regime. Note also that the TE solution can be highly nonlinear; that is, the two terms in the parentheses on the left-hand side of (6) can be similar in magnitude. This happens whenever

*T*

_{eo}is close to

*T*

_{c}. It is anticipated that linear theory gives a good approximation either when

*T*

_{eo}≪

*T*

_{c}(for any friction coefficient) or when the Rayleigh friction is large enough (for any forcing amplitude).

*S*

_{o}(

*z*) =

*g*

^{−1}

*T*

_{s}

*N*

^{2}

_{o}

*z*) = ∂

*T*

_{o}/∂

*z*+

*κ*

*T*

_{o}/

*H.*Inspection of (29)–(32) shows that for any nontrivial

*T*

_{e}(i.e., for ∂

*T*

_{e}/∂

*r*≠ 0 somewhere in the domain) the TE solution exists if and only if

*α*

_{r}= 0. In this sense, the secondary circulation in the linear regime can be viewed as being forced by friction, although it is by no means linear in

*α*

_{r}(see below).

Two measures are introduced in order to quantify two different aspects of the cross-vortex circulation with a single number each. First, the “total strength” is diagnosed as the maximum absolute value *ψ*_{max} of the streamfunction *ψ*. This choice is physically motivated by the fact that 2*π**ψ*_{max} is the total mass per unit time (in kg s^{−1}) crossing the horizontal circular area of radius *r̂* at altitude *ẑ,* where (*r̂, ẑ*) is the location at which the streamfunction assumes its maximum value [this interpretation follows from the definition of *ψ* in (12)]. The dependence of *ψ*_{max} on *α*_{r} is shown in Fig. 6a. Apparently, the slope of the curve decreases with increasing *α*_{r}; in other words, the depencence on *α*_{r} is particularly sensitive as *α*_{r} → 0. Second, in order to quantify the stratospheric penetration of the secondary circulation, we consider the vertical wind at the vortex center at *z* = 22 km, denoted by *w*_{strat}. This can be taken as an approximate measure for the rate of change of total ozone due to the cross-vortex circulation. The latter is proportional to ∫ *w**χ*_{z}*ρ*_{0} *dz,* where *χ*_{z} is the vertical derivative of the ozone mixing ratio. With the tropical ozone profile from the McClatchey reference atmosphere (McClatchey et al. 1972), the function *χ*_{z}(*z*)*ρ*_{0}(*z*) has a pronounced peak at *z* = 22 km, which is why we consider the vertical wind at this particular altitude. Typical values of *w*_{strat} are of the order of *w*_{strat} = 10^{−5} m s^{−1}, which for the McClatchey tropical profile corresponds to a rate of decrease of total ozone by 0.05 Dobson units per day. Figure 6b shows the dependence of *w*_{strat} on *α*_{r}. There is a strong increase for very small values of *α*_{r}, followed by a more gentle decrease for larger values of *α*_{r}. Thus, the stratospheric part of the secondary circulation differs qualitatively from the tropospheric part as far as its dependence on *α*_{r} is concerned.

*α*

_{r}. For

*α*

_{r}≠ 0, the linearized version of (13) is

*ψ*is linear in

*T*

_{e}. Note that the second term of the differential operator on the left-hand side of (33) differs from the respective term in the elliptic operator connecting potential vorticity and streamfunction in standard quasigeostrophic theory. The difference is such that a discontinuity of

*N*

_{o}at the tropopause in the latter theory renders ∂

*ψ*/∂

*z*discontinuous (Bishop and Thorpe 1994), while in our case ∂

*ψ*/∂

*z*remains continuous. With

*T*

_{eo}from (19), the right-hand side becomes

*ρ*

_{0}= const. and

*N*

^{2}

_{o}

*F̃*

_{δ}= −

*r*

*δ*(

*r*−

*r*′)

*δ*(

*z*−

*z*′) with (

*r*′,

*z*′) = (

*r*

_{o}/2,

*z*

_{o}/2). An analytical solution to this problem is derived in appendix C. It is found that the vertical wind on the axis of symmetry,

*w*(0,

*z*), goes asymptotically like

*D*is independent of

*α*

_{r}, while for large values of

*α*

_{r}one obtains

*w*(0,

*z*) exhibits a maximum at some intermediate value of

*α*

_{r}. In fact, there is good qualitative agreement between the analytical solution of this simplified problem (Fig. 7) and the numerical solution shown in Fig. 6b. Similarly, it is shown in appendix C that

*w*

*z*

*D*

*α*

_{r}

*α*

_{r}

*z*

*z*

*α*

_{r}

*α*

_{r}→ 0 and

*α*

_{r}→ ∞. This is qualitatively consistent with the curve shown in Fig. 6a.

*w*

_{strat}on

*α*

_{r}, we note that the principal part

*G*

_{1}of the Green’s function

*G*for (33) in the neighborhood of (

*r*′,

*z*′) is given by

*r*′,

*z*′) with half axes in the radial and vertical direction proportional to

*α*

_{r}

*α*

_{n}

*f*/

*N*

_{o}, respectively. Consider a point, (

*r, z*) ≠ (

*r*′,

*z*′). For large

*α*

_{r}, the value of

*G*

_{1}may be large, but the the ellipses are elongated in the radial direction and the vertical wind, which involves the radial derivative, can be rather small. For decreasing

*α*

_{r}, the value of

*G*

_{1}decreases, but the ellipses become more and more elongated in the vertical direction such that the radial derivative and, hence, the vertical wind can increase. It is the systematic change in the aspect ratio of the ellipses that makes the streamfunction and the vertical wind behave qualitatively differently in their dependence on

*α*

_{r}.

## 5. Frictionless limit for supercritical forcing: AMC solution

Following PH92, we call the frictionless limit of the solution for supercritical forcing the angular momentum conserving (AMC) solution. As will become apparent below, the AMC solution differs qualitatively from the TE solution. It is characterized by a nonvanishing secondary circulation with upward flow in the vortex center, outward flow in the upper troposphere, downward flow at larger radii, and a return flow in the lowest part of the atmosphere. The secondary circulation modifies the temperature in such a way as to make the integration of the nonlinear thermal wind equation possible throughout the domain.

### a. Thermal forcing right on the axis of symmetry

In a problem with rotational symmetry, the axis of symmetry is a singularity. In the zonally symmetric problem this singularity prevents the secondary circulation of the AMC solution from reaching the pole, since otherwise one would obtain infinite zonal wind speeds. On the other hand, the current problem involves nonzero thermal forcing right on the axis of symmetry, and the secondary circulation is nonzero right on the axis of symmetry. Fortunately this does not necessarily lead to a problem in connection with the singularity. The important difference is the sense of the secondary circulation: In the zonally symmetric cases with tropical or subtropical thermal forcing the flow toward the axis of symmetry takes place in the upper troposphere, while in the present case it takes place in the lower troposphere and, in particular, in the boundary layer. Thus, for our AMC solution to be physically meaningful it is crucial that the return flow does not conserve angular momentum. This is consistent with one of our basic assumptions, namely, nonzero boundary layer drag.

### b. Numerical solution

*α*

_{r}< 2.5 × 10

^{−2}day

^{−1}, we used

*α*

_{r}, where

*α*

_{bl}= 2.5 × 10

^{−2}day

^{−1}and

*z*

_{bl}= 5 km, giving a smooth transition from a (moderate) value

*α*

_{bl}at

*z*= 0 to the smaller constant value

*α*

_{r}above

*z*

_{bl}. Varying Rayleigh friction means varying

*α*

_{r}while keeping

*α*

_{bl}constant. The use of such a modified

*α̂*

_{r}(

*z*)

*ζ*

_{a}

*υ*in the time-dependent version of (1) produced strong cyclonic spinup in the lowest few kilometers close to the axis of symmetry. Interaction of this cyclonic circulation with our parameterization of surface friction and Newtonian relaxation then precluded a steady state to be reached. It is conjectured that this problem would be less severe (or nonexistent) in the case of a zonally symmetric atmosphere in the subtropics and even less so on a zonally symmetric equatorial

*β*plane. Only in our axisymmetric

*f*-plane geometry does a parcel with finite absolute angular momentum at

*r*≠ 0 acquire infinite cyclonic wind speed if brought toward the center of the thermal forcing in the absence of friction.

A numerical solution for supercritical forcing and small but nonzero Rayleigh friction is shown in Fig. 8. The tangential wind *u* (Fig. 8a) is much stronger in comparison with the previous example of subcritical forcing (Fig. 3); furthermore, it does not reflect the shape of the deviation equilibrium temperature *T*^{′}_{e}*r, z*) as was the case before. Instead, the upper-tropospheric *u* increases approximately linearly with radius out to about *r* = 950 km. At larger radii it decays to zero rather abruptly. The tropopause is roughly 1 km above its reference position in the vortex center. There is weak cyclonic circulation in the neighborhood of (*r, z*) = (500 km, 4 km), which would be stronger if we had used *α*_{r} instead of the modified *α̂*_{r}*ψ* at the tropopause: it appears to avoid the stratosphere and to develop a kink close to the tropopause. We will come back to this point in the following section. As a measure of absolute angular momentum about the axis of rotation, Fig. 8c shows the distribution of potential radius *R*; the latter is defined via *m* ≡ *fr*^{2}/2 + *ur* = *fR*^{2}/2, so that lines of constant *R* are lines of constant *m.* Apparently, the circulation is approximately angular momentum conserving: the contours of *R* are approximately parallel to the contours of *ψ* except where friction is large, that is, except close to the lower boundary (where *α̂*_{r}*u* is large and, hence, *X* is nonnegligible). Figure 8d gives the diabatic heating *Q,* illustrating that there is heating in regions of upward motion and cooling in regions of downward motion; in addition, the change in static stability at the tropopause is clearly visible.

### c. Vertical extent of the secondary circulation

The secondary circulation of the AMC solution cannot extend beyond the maximum altitude *z*_{o} of the thermal forcing. We are going to prove this statement by assuming the opposite and deriving a contradiction to this assumption. Figure 9 serves for illustration.

*z*

_{l}>

*z*

_{o}, but that the thermal forcing does not reach beyond

*z*

_{o}, that is,

*T*

^{′}

_{e}

*z*>

*z*

_{o}. Further, let

*r*

_{a}be the maximum radius out to which the secondary circulation extends. Owing to angular momentum conservation, a parcel in stationary flow moving radially outward at altitude

*z*=

*z*

_{1}starting at

*r*= 0 has

*ω*

*r, z*

_{1}

*f*

*r*

*r*

_{a}

*z*

_{o}to

*z*

_{1}together with (40) gives

*z*

_{o}and

*z*

_{1}. Consider two radii

*r*

_{u}and

*r*

_{d}such that between

*z*

_{o}and

*z*

_{1}there is upwelling for 0 ⩽

*r*⩽

*r*

_{u}and downwelling for

*r*

_{d}⩽

*r*⩽

*r*

_{a}. Since in our whole study we assume a statically stable atmosphere, this renders

*T̃*<

*T̃*

_{e}for 0 ⩽

*r*⩽

*r*

_{u}and

*T̃*>

*T̃*

_{e}for

*r*

_{d}⩽

*r*⩽

*r*

_{a}. By assumption we have

*T*

^{′}

_{e}

*T*

_{e}=

*T*

_{o}(

*z*) for

*z*≥

*z*

_{o}. It follows that

*T̃*

_{e}= const. and, hence, ∂

*T̃*/∂

*r*> 0 somewhere in 0 ⩽

*r*⩽

*r*

_{a}, resulting in a negative argument of the square root in (41). The initial assumption thus leads to an unphysical solution in connection with the nonlinear thermal wind equation. It follows that the secondary circulation cannot extend to altitudes higher than

*z*

_{o}.

We come back to the kink in the streamfunction close to the tropopause mentioned in connection with Fig. 8b. This near discontinuity of ∂*ψ*/∂*z* arises because Eq. (13) ceases to be elliptic in the limit *α*_{r} → 0. Of course, in the numerical calculation the condition *AC* − *B*^{2} > 0 is satisfied everywhere, but in the neighborhood of the tropopause for 0 ⩽ *r* ⩽ *r*_{a} the coefficient *C* is only a small fraction compared with its value in the linear case. This means that the penalty of high curvature ∂^{2}*ψ*/∂*z*^{2} becomes low, allowing the solution to develop high values of ∂^{2}*ψ*/∂*z*^{2} at *z* = *z*_{o}. In this sense, the kink is a property of the elliptic equation (13) in the limit *α*_{r} → 0. On the other hand, a kink in *ψ* across the tropopause is equivalent to a discontinuity of the velocity component parallel to the tropopause, allowing nonzero tropopause-parallel flow in the troposphere right underneath the tropopause but, at the same time, zero flow right above the tropopause. Thus, the kink of *ψ* in the almost frictionless numerical solution reflects the fact that in the limit *α*_{r} → 0 the cross-vortex circulation cannot penetrate above *z*_{o}.

### d. Approximate analytical theory

*ψ*

_{a}= 0 for

*z*>

*z*

_{o}, where the subscript

*a*denotes the AMC solution. As a consequence,

*T*=

*T*

_{e}=

*T*

_{o}(

*z*) and ∂

*u*/∂

*z*= 0 for

*z*>

*z*

_{o}. One can restrict attention to

*z*⩽

*z*

_{o}and proceed in a similar way as HH80 or PH92. Consider a parcel in the vortex center that is lifted up to altitude

*z*

_{o}and subsequently advected radially outward to some radius

*r*

_{a}. Since

*m*= 0 in the vortex center, it follows from (10) and the conservation of angular momentum that along the parcel’s streamline

*ω*

_{a}

*r, z*

_{o}

*f*

*r*

*r*

_{a}

*u*

_{a}

*r, z*

_{o}

*f*

*r*

*r*

*r*

_{a}

*r*>

*r*

_{a}the secondary circulation is assumed to be zero and the tangential wind to be in thermal wind balance with

*T*

_{e}. Note that (43) renders the factor multiplying

*υ*in (1) equal to zero for 0 ⩽

*r*⩽

*r*

_{a}at

*z*=

*z*

_{o}. In contrast to the situation with the linearized equation (29), the nonlinearity allows nonzero

*υ*even for

*α*

_{r}= 0.

*z*= 0

_{−}denotes the earth’s surface, as opposed to

*z*= 0, which is the top of the boundary layer at which

*ψ*is not necessarily zero [owing to (18)]. The vertical average of (20) gives

*ω*= 0 at the earth’s surface. Multipliying by

*r,*using (42), and integrating radially leads to

*T*

_{ao}and

*r*

_{a}are obtained by invoking the conservation of energy and the continuity of the temperature field. The heat equation (2) together with (3) is rewritten in terms of potential temperature as

*α*

_{n}= const., (48) with (49) becomes

*f*-plane axisymmetric analogue of the “equal-area” construction in HH80, fixes

*T*

_{ao}and

*r*

_{a}in (46). Some more details are given in appendix D.

The dependence of the radius *r*_{a} on the forcing amplitude *T*_{eo} is shown in Fig. 10. For marginally supercritical forcing the dependence is very sensitive. As shown in appendix D, *r*_{a} = *r*_{o} for *T*_{eo}/*T*_{c} = (*π*/2)^{2}(1 − 4/*π*^{2})^{−1} ≈ 4.15, that is, for *T*_{eo} = 2.68 K. At higher forcing amplitudes, the dependence becomes rather weak, with *r*_{a} increasing from 1200 to 1400 km between *T*_{eo} = 5 and 10 K.

We tested the prediction (43) of the approximate analytical theory for *u*(*r, z*_{o}) by running the numerical model with *T*_{eo} = 5 K and different values of *α*_{r}. Figure 11a shows the result. As the value of *α*_{r} decreases, the numerical solution (dashed lines, with increasing dash length for decreasing *α*_{r}) broadly approaches the theoretical prediction (solid line); in particular, the predicted radius *r*_{a} appears to be relevant for the numerical solution for small values of *α*_{r}. For small radii, the slope ∂*u*/∂*r* of the numerical solutions is somewhat smaller than predicted. This is similar to a feature found by HH80 and is believed to be (partly) due to the width of the rising branch of the secondary circulation.

### e. Secondary circulation

_{−}to ∞ gives

*T*

_{e}(

*r*) and

*T*

_{a}(

*r*) are known, this equation can readily be integrated with

*h*= 0 at

*r*= 0 to yield

*h*(

*r*). The result is displayed as a solid curve in Fig. 11b. Comparison with the numerical solution (dashed lines) suggests, again, a fairly good agreement in the limit

*α*

_{r}→ 0.

Figure 12 shows the dependence of *ψ*_{max} and *w*_{strat} on *α*_{r} for *T*_{eo} = 5 K. For the larger values of *α*_{r} displayed in the figure, both the total strengh and the stratospheric portion of the secondary circulation are a factor of 10 times larger than in Fig. 6, where we had *T*_{eo} = 0.5 K. This is consistent with the analysis of linearity from Fig. 5 and the result that *ψ* ∝ *T*_{eo} in the linear regime [see (33)]. Significant nonlinearity appears in Fig. 12 at small values of *α*_{r}, for which the actual curve (solid line) deviates from the curve that one would obtain from the linearized equations (dashed curve). Linear theory underpredicts *ψ*_{max} while it overpredicts *w*_{strat}. The qualitative dependence of *w*_{strat} on *α*_{r} is similar in the subcritical and the supercritical regime, with *w*_{strat} → 0 for both *α*_{r} → 0 and *α*_{r} → ∞. On the other hand, there is clearly a qualitative difference between subcritical and supercritical forcing for *ψ*_{max} as *α*_{r} → 0: while *ψ*_{max} → 0 for subcritical forcing, *ψ*_{max} → const. ≠ 0 for supercritical forcing. This reflects the distinguishing property of the AMC solution to have a nonzero secondary circulation.

## 6. Summary and discussion

We investigated stationary axisymmetric balanced flow of a stably stratified dry non-Boussinesq atmosphere on the *f* plane. The circulation is forced in the troposphere through thermal relaxation toward a specified equilibrium temperature and is damped through Rayleigh friction in the interior of the domain. Surface friction is sufficiently strong to ensure weak surface winds. In contrast to the zonally symmetric problem of PH92, the geometry of the current problem is able to represent truly local thermal forcing. We studied the dependence of the flow on the strength of friction with particular focus on the frictionless limit, and we investigated the upward penetration of the secondary circulation into the stratosphere. Both approximate analytical and numerical calculations were performed; for the latter an Eliassen balanced vortex model was used. Our main results are the following.

There is threshold behavior in the frictionless limit with a thermal equilibrium (TE) solution for subcritical forcing and a so-called angular momentum conserving (AMC) solution for supercritical forcing. The existence of a threshold separating two qualitatively different regimes is quite analogous to the corresponding result of PH92 for the zonally symmetric case.

By definition, the TE solution has a primary vortex circulation that reflects the equilibrium temperature, while its secondary (cross vortex) circulation is zero. When the forcing is subcritical, but close to the critical threshold, the TE solution is strongly nonlinear in the sense that the nonlinear term in the thermal wind equation is about as important as the linear term. In contrast to the zonally symmetric problem, the TE solution does not exist for supercritical forcing and it is regular throughout its range of existence (i.e., for subcritical forcing). The difference arises from the stronger nonlinearity of the thermal wind equation in the current problem compared with the zonally symmetric problem. Here, a solution is called regular if it can be the inviscid limit of a solution with nonzero viscosity.

The AMC solution is characterized by a sharp outward edge of the primary vortex circulation and a nonvanishing secondary circulation with upward flow in the vortex center. In contrast to the zonally symmetric problem, the singularity at the axis of symmetry poses no constraint on the the secondary circulation, since its sense is such that the flow toward the axis of symmetry takes place in the boundary layer, which is not frictionless by assumption. For forcing that is well into the supercritical regime, the location of the outward edge depends only weakly on the amount of the forcing. Similar to the zonally symmetric problem one can find an approximate analytical theory for the AMC solution; it predicts well the salient features of the numerical solution for small (but nonzero) friction.

For small but nonzero friction our numerical Eliassen balanced vortex model approximately represents the (strictly speaking hypothetical) AMC solution. This success suggests that one can view the nonzero secondary circulation in the AMC regime as a thermally and frictionally forced Eliassen cross-circulation in the limit of small friction.

The cross-vortex circulation of the AMC solution cannot extend above the maximum altitude of the thermal forcing. For thermal forcing that is confined to the troposphere it means that the cross-vortex circulation cannot extend into the stratosphere. This property of the AMC solution is consistent with the behavior of the elliptic equation for the Eliassen secondary circulation when friction becomes small.

The overall strength of the secondary circulation in the troposphere increases monotonically, albeit nonlinearly with the friction coefficient. On the other hand, the impact of the secondary circulation on the stratosphere (as quantified by the vertical wind in the vortex center) has a maximum at an intermediate value of the friction coefficient and decreases for larger values. These dependencies are a property of the elliptic equation for the Eliassen secondary circulation in the linearized problem and are reproduced by the asymptotic behavior of the corresponding Green’s function solution.

It is not exactly straightforward to put these results into the context of the original motivation, namely, the dynamics of monsoonal systems. Although a monsoonal system may persist for several months, it is by definition a transient phenomenon, and one may ask to what degree a stationary flow regime is relevant at all. Our numerical method of solution gives us a clue, since it involves a switch-on time integration that, in a certain sense, simulates the onset of a monsoon. In the experiment shown in Fig. 8, both the primary and the secondary circulation in the troposphere had practically reached their steady-state limit after about 150 days, with their general characteristics being established even earlier. In view of the rather long timescale *τ*_{n} we used for the thermal forcing (*τ*_{n} = *α*^{−1}_{n}

*except*for the factor

*α*

_{n}/

*α*

_{r}missing in front of the second term on the left-hand side. Since (52) is a truly elliptic equation independent of the values of

*α*

_{r}and

*α*

_{n}, its solution

*ψ*extends beyond the forcing region. This (transient) secondary circulation will dynamically affect ozone and may transport water vapor from the troposphere into the stratosphere. For moderate and higher values of the friction coefficient (

*α*

_{r}≥ 10

^{−2}day

^{−1}), both the tropospheric and the stratospheric circulation reached a steady state within about 150 days. A full discussion of the various timescales involved is beyond the scope of this paper.

Our parameterization for friction was deliberatly chosen to be the simplest possible, since it facilitated an approximate analytical investigation of the dependence on friction in the linearized regime. As far as the frictionless limit is concerned, the exact parameterization of friction shoud be irrelevant. In fact, we did some numerical experiments with a viscous parameterization instead of Rayleigh friction and found broadly similar results.

Many of our results depend on the assumed relaxational character of the thermal forcing. Such an approach is often accepted for global studies, since the differential solar radiation externally forces horizontal temperature differences toward which the actual atmosphere relaxes through a combination of dynamics and various diabatic processes. In a monsoonal system, on the other hand, the diabatic heating is dominated by moist convection, and a simple parameterization in terms of Newtonian relaxation is more questionable. It is true that moist convection is sometimes parameterized as relaxation toward an equilibrium atmosphere, which in turn is determined by surface and boundary layer conditions (e.g., Raymond 1994). Nevertheless, the relaxation toward a specified equilibrium profile can be accepted as a first approximation only to the degree that these surface conditions are not affected by the flow itself. The latter may be a poor assumption in the context of a monsoonal system and the differences between a dry and a moist model may be significant (e.g., Hunt 1973).

The symmetry exploited in the present work crucially depends on the *f*-plane approximation. Breaking this symmetry, for example, by going over to the *β*-plane approximation, is likely to result in important changes in the character of the solution. There is work in progress that investigates the consequences of breaking the symmetry (Hsu and Plumb 1997).

In conclusion, the dynamics of real monsoonal systems certainly depend on flow asymmetries, on transient effects, on the more complicated interactions between dynamics and diabatic effects, and on the precise way of momentum damping. Nevertheless, the basic features of the idealized flow investigated in this paper and the differences with respect to the related zonally symmetric problem may prove a helpful step toward an understanding of the complex interactions in fully three-dimensional time-dependent models with realistic physical parameterizations and, eventually, in the real atmosphere.

## Acknowledgments

The author wants to thank his colleagues, especially M. Juckes, for numerous stimulating discussions. The instructive comments of three anonymous reviewers are gratefully acknowledged.

## REFERENCES

Becker, E., G. Schmitz, and R. Geprägs, 1997: The feedback of midlatitude waves onto the Hadley cell in a simple general circulation model.

*Tellus,***49A,**182–199.Bender, C. M., and S. A. Orszag, 1978:

*Advanced Mathematical Methods for Scientists and Engineers.*McGraw-Hill, 593 pp.Bishop, C. H., and A. J. Thorpe, 1994: Potential vorticity and the electrostatics analogy: Quasi-geostrophic theory.

*Quart. J. Roy. Meteor. Soc.,***120,**713–731.Dunkerton, T. J., 1989: Nonlinear Hadley circulation driven by asymmetric differential heating.

*J. Atmos. Sci.,***46,**956–974.Eliassen, A., 1952: Slow thermally or frictionally controlled meridional circulation in a circular vortex.

*Astrophys. Norv.,***5**(2), 19–60.Held, I. M., and A. Y. Hou, 1980: Nonlinear axially symmetric circulations in a nearly inviscid atmosphere.

*J. Atmos. Sci.,***37,**515–533.Hou, A. Y., and R. S. Lindzen, 1992: The influence of concentrated heating on the Hadley circulation.

*J. Atmos. Sci.,***49,**1233–1241.Hsu, C. J., and R. A. Plumb, 1997: Non-axisymmetric thermally-driven circulations. Preprints,

*11th Conf. on Atmospheric and Oceanic Fluid Dynamics,*Tacoma, WA, Amer. Meteor. Soc., 59–61.Hunt, B. G., 1973: Zonally symmetric global general circulation models with and without the hydrologic cycle.

*Tellus,***25,**337–354.Lindzen, R. S., and A. Y. Hou, 1988: Hadley circulations for zonally averaged heating centered off the equator.

*J. Atmos. Sci.,***45,**2416–2427.McClatchey, R. A., R. W. Fenn, J. E. A. Selby, F. E. Volz, and J. S. Garin, 1972: Optical properties of the atmosphere (3d ed.). AFCRL-047297, 108 pp. [NTIS AD753075.].

Plumb, R. A., and A. Y. Hou, 1992: The response of a zonally symmetric atmosphere to subtropical thermal forcing: Threshold behavior.

*J. Atmos. Sci.,***49,**1790–1799.Raymond, D. J., 1994: Convective processes and tropical atmospheric circulations.

*Quart. J. Roy. Meteor. Soc.,***120,**1431–1455.Schneider, E. K., 1977: Axially symmetric steady-state models of the basic state for instability and climate studies. Part II: Nonlinear calculations.

*J. Atmos. Sci.,***34,**280–296.——, 1987: A simplified model of the modified Hadley circulation.

*J. Atmos. Sci.,***44,**3311–3328.Schubert, W. H., and J. J. Hack, 1983: Transformed Eliassen balanced vortex model.

*J. Atmos. Sci.,***40,**1571–1583.Shine, K., 1987: The middle atmosphere in the absence of dynamical heat fluxes.

*Quart. J. Roy. Meteor. Soc.,***113,**603–633.Ting, M., 1994: Maintenance of northern summer stationary waves in a GCM.

*J. Atmos. Sci.,***51,**3286–3308.Wirth, V., 1995: Diabatic heating in an axisymmetric cut-off cyclone and related stratosphere-troposphere exchange.

*Quart. J. Roy. Meteor. Soc.,***121,**127–147.Zhou, X., and L. Chao, 1994: Ozone valley over Tibetan plateau.

*Acta Meteor. Sinica,***8,**505–506.Zou, H., 1996: Seasonal variation and trends of TOMS ozone over Tibet.

*Geophys. Res. Lett.,***23,**1029–1032.

## APPENDIX A

### Nonexistence of the AMC Solution for Subcritical Forcing

We want to prove that the AMC solution does not exist for subcritical forcing. This is done by assuming that both the TE and the AMC solution exist and showing that this assumption leads to a contradiction. The argument is similar to that in PH92.

*r*

^{4}, together with (A1), yields

*m*

^{2}

_{a}

*z*

_{o}) = 0 in 0 ⩽

*r*⩽

*r*

_{a}for the AMC solution. Furthermore, the assumed existence of the TE solution guarantees that

*ω*

_{e}+

*f*/2 is real and nonnegative everywhere [see (21)]. It follows that

*m*

_{e}(

*r, z*

_{o}) ≥ 0 [see (10)], which explains the ⩽ in (A3).

*T*

_{a}below the equilibrium temperature

*T*

_{e}in the vortex center and above

*T*

_{e}in the environment, but at the same time it is weak enough so that the actual temperature

*T*

_{a}is still warmer in the vortex center than in the environement. It follows that

## APPENDIX B

### Regularity of the TE Solution

*ν*), (1) can be rewritten in terms of absolute angular momentum as

*m*in the interior for any solution with small but nonzero viscosity; the argument has been given several times before (e.g., HH80; PH92). In our case, the axis of symmetry deserves special attention, because in a certain sense it is “in the interior” of the flow and there is a global minimum of the angular momentum (

*m*= 0 at

*r*= 0 and ∂

*m*/∂

*r*> 0 for

*r*> 0; see below). Yet this poses no problem in terms of regularity, since the effect of vertical diffusion is zero at

*r*= 0 (because ∂

*m*/∂

*z*= 0 at

*r*= 0). Correspondingly, the condition for regularity of a frictionless solution in our

*f*-plane axisymmetric case is is that there can be no extrema of absolute angular momentum away from the axis of symmetry except on the lower boundary.

*ω*

_{e}of the TE solution satisfies

*ω*

_{e}

*f*

*r*

^{−1}∂[

*r*

^{2}(B3)]/∂

*r*leads to

*u*

_{e}=

*r*

*ω*

_{e}, which proves that

## APPENDIX C

### Green’s Function for the Linear Problem

*F̃*

_{δ}(

*r, z*) = −

*r*

*δ*(

*r*−

*r*′)

*δ*(

*z*−

*z*′), where (

*r*′,

*z*′) = (

*r*

_{o}/2,

*z*

_{o}/2). Assuming that

*ρ*

_{0}= const. and

*N*

_{o}= const. and rescaling the vertical coordinate according to

*z*→

*N*

_{o}

*z*/

*f*leads to the following problem:

*ψ*→ 0 as

*z, r*→ ∞, 0 <

*w*= (

*r*

*ρ*

_{0})

^{−1}∂

*ψ*/∂

*r*< ∞ at

*r*= 0, and ∂

*ψ*/∂

*z*= 0 at

*z*= 0. The last condition is motivated by our numerical results (Fig. 3). In this appendix we will derive a solution to (C1) plus boundary conditions and investigate its asymptotic dependence on

*α*

_{r}.

*ψ*/

*r*gives

^{m}denote the Fourier coefficients, and

*r*

*α*

_{n}/

*α*

_{r}

*m*|

*r,*one obtains

*r*= 0 is

*K*

_{i}and

*I*

_{i}are the modified Bessel functions of

*i*th order, and

*r*

_{>}and

*r*

_{<}denote the larger and smaller of

*r*

*r*

*ψ*at the lower boundary is equivalent to ∂Φ/∂

*z*= 0 at

*z*= 0. It is implemented by adding a mirror charge at −

*z*′, yielding

^{m}= Φ

^{−m}was used.

*r*= 0) this gives

*x*

^{−1}

*d*(

*xI*

_{1})/

*dx*=

*I*

_{o}(

*x*) = 1 at

*x*= 0 was used. It follows that

*α*

_{n}= 0.1 day

^{−1},

*r*′ = 500 km,

*z*′ = 8 km ×

*N*

_{o}/

*f, z*= 22 km ×

*N*

_{o}/

*f,*and

*N*

_{o}/

*f*= 120, that is, for values of the relevant parameters that approximately correspond with the numerical solution shown in Fig. 6b. Apparently, the functional dependence on

*α*

_{r}in the numerical solution is borne out rather well by the Green’s function solution. In particular the maximum appears at about the right value of

*α*

_{r}.

*w*(0,

*z; r*′,

*z*′) in

*α*

_{r}. The known asymptotic behavior of

*K*

_{i}and

*I*

_{i}gives

*x*) at large

*x*ensures that the integral in (C11) is finite. For small values of

*α*

_{r}both cosines in (C10) can be approximated by unity, thus yielding

*z*=

*z*′, the leading asymptotic behavior for

*α*

_{r}→ ∞ is similar, because cos

^{2}

*φ*= [1 + cos(2

*φ*)]/2 and the rapid oscillations of cos(2

*φ*) do not contribute to leading order:

*z*≠

*z*′ one obtains

*z̃*

_{j}=

*z*−

*z*′ for

*j*= 1,

*z̃*

_{j}=

*z*+

*z*′ for

*j*= 2, and

*y*=

*α*

_{r}/

*α*

_{n}

*z̃*

_{j}

*x*/

*r*′. The integral on the right-hand side is evaluated by dividing the interval [0, ∞) into an infinite number of subintervals of length 2

*π*. For large

*α*

_{r}the function Θ can be approximated in each subinterval

*i*by a Taylor series according to

*x*) =

*d*Θ/

*dx*and

*x*

_{oi}denotes the midpoint of subinterval

*i.*The first two terms of the expansion, multiplied by the cosine and integrated over each subinterval, give zero, leaving

*α*

_{r}. At the upper limit, Θ′(

*x*) → 0 as

*x*→ ∞ was used. Since Θ′(0) = 0, it follows that

*α*

_{r}, that is,

*α*

_{r}

*w*→ 0 for

*α*

_{r}→ ∞ (cf. Bender and Orszag 1978).

## APPENDIX D

### Approximate Analytical Theory for the AMC Solution

*x*=

*r*/

*r*

_{o}, Eq. (46) becomes

*x*

_{a}, that is,

*T*

_{a}(

*x*

_{a}) =

*T*

_{e}(

*x*

_{a}), gives

*x*∗ = min(

*x*

_{a}, 1). By evaluating the integrals in (D4) and using (D3) to eliminate

*T*

_{ao}one obtains

*x*

_{a}when

*T*

_{c}and

*T*

_{eo}are known. Inserting the value of

*x*

_{a}thus calculated into (D3) yields

*T*

_{ao}.

*T*

_{eo}for which

*r*

_{a}=

*r*

_{o}is obtained by setting

*x*

_{a}= 1 in (D5), which gives

The TE solution for subcritical forcing (*T*_{eo} = 0.5 K): tangential wind *u*_{e} (in m s^{−1}, solid, contours every 0.5 m s^{−1}, zero contour omitted) and potential temperature *θ*_{e} (in K, dashed, contours every 10 K). The tangential wind satisfies *u*_{e} = 0 at *z* = 0. The gray shading denotes the stratosphere.

Citation: Journal of the Atmospheric Sciences 55, 19; 10.1175/1520-0469(1998)055<3024:TFSAFO>2.0.CO;2

The TE solution for subcritical forcing (*T*_{eo} = 0.5 K): tangential wind *u*_{e} (in m s^{−1}, solid, contours every 0.5 m s^{−1}, zero contour omitted) and potential temperature *θ*_{e} (in K, dashed, contours every 10 K). The tangential wind satisfies *u*_{e} = 0 at *z* = 0. The gray shading denotes the stratosphere.

Citation: Journal of the Atmospheric Sciences 55, 19; 10.1175/1520-0469(1998)055<3024:TFSAFO>2.0.CO;2

The TE solution for subcritical forcing (*T*_{eo} = 0.5 K): tangential wind *u*_{e} (in m s^{−1}, solid, contours every 0.5 m s^{−1}, zero contour omitted) and potential temperature *θ*_{e} (in K, dashed, contours every 10 K). The tangential wind satisfies *u*_{e} = 0 at *z* = 0. The gray shading denotes the stratosphere.

Citation: Journal of the Atmospheric Sciences 55, 19; 10.1175/1520-0469(1998)055<3024:TFSAFO>2.0.CO;2

Numerical solution for subcritical forcing (*T*_{eo} = 0.5 K) with small but nonzero friction *α*_{r} = 2 × 10^{−3} day^{−1}. (a) Tangential wind *u* (in m s^{−1}, solid, contours every 0.5 m s^{−1}, zero contour omitted) and potential temperature *θ* (dashed contours). (b) Cross-vortex streamfunction *ψ*. The solid contours in (b) are drawn with a contour interval of 0.1*ψ*_{max}, and the dashed contours are at 0.01*ψ*_{max} and 0.03*ψ*_{max}, respectively, where *ψ*_{max} = 1.95 × 10^{6} kg s^{−1}; the arrows indicate the direction of the secondary circulation. In both panels the gray shading denotes the stratosphere.

Numerical solution for subcritical forcing (*T*_{eo} = 0.5 K) with small but nonzero friction *α*_{r} = 2 × 10^{−3} day^{−1}. (a) Tangential wind *u* (in m s^{−1}, solid, contours every 0.5 m s^{−1}, zero contour omitted) and potential temperature *θ* (dashed contours). (b) Cross-vortex streamfunction *ψ*. The solid contours in (b) are drawn with a contour interval of 0.1*ψ*_{max}, and the dashed contours are at 0.01*ψ*_{max} and 0.03*ψ*_{max}, respectively, where *ψ*_{max} = 1.95 × 10^{6} kg s^{−1}; the arrows indicate the direction of the secondary circulation. In both panels the gray shading denotes the stratosphere.

Numerical solution for subcritical forcing (*T*_{eo} = 0.5 K) with small but nonzero friction *α*_{r} = 2 × 10^{−3} day^{−1}. (a) Tangential wind *u* (in m s^{−1}, solid, contours every 0.5 m s^{−1}, zero contour omitted) and potential temperature *θ* (dashed contours). (b) Cross-vortex streamfunction *ψ*. The solid contours in (b) are drawn with a contour interval of 0.1*ψ*_{max}, and the dashed contours are at 0.01*ψ*_{max} and 0.03*ψ*_{max}, respectively, where *ψ*_{max} = 1.95 × 10^{6} kg s^{−1}; the arrows indicate the direction of the secondary circulation. In both panels the gray shading denotes the stratosphere.

Approach of the frictionless limit (TE solution) for subcritical forcing (*T*_{eo} = 0.5 K). (a) Tangential wind *u*(*r, z*_{o}) at the top of the thermal forcing region (i.e., at *z*_{o}) as a function of radius *r.* (b) Vertically integrated radial heat flux *h*(*r*) as a function of radius *r.* In both panels the solid line represents the (frictionless) TE solution. The dashed lines represent numerical calculations with *α*_{r} = 6 × 10^{−3} day^{−1}, 2 × 10^{−3} day^{−1}, 6 × 10^{−4} day^{−1}, and 2 × 10^{−4} day^{−1}, respectively, where an increase in the length of the dashes corresponds to a decrease in *α*_{r}.

Approach of the frictionless limit (TE solution) for subcritical forcing (*T*_{eo} = 0.5 K). (a) Tangential wind *u*(*r, z*_{o}) at the top of the thermal forcing region (i.e., at *z*_{o}) as a function of radius *r.* (b) Vertically integrated radial heat flux *h*(*r*) as a function of radius *r.* In both panels the solid line represents the (frictionless) TE solution. The dashed lines represent numerical calculations with *α*_{r} = 6 × 10^{−3} day^{−1}, 2 × 10^{−3} day^{−1}, 6 × 10^{−4} day^{−1}, and 2 × 10^{−4} day^{−1}, respectively, where an increase in the length of the dashes corresponds to a decrease in *α*_{r}.

Approach of the frictionless limit (TE solution) for subcritical forcing (*T*_{eo} = 0.5 K). (a) Tangential wind *u*(*r, z*_{o}) at the top of the thermal forcing region (i.e., at *z*_{o}) as a function of radius *r.* (b) Vertically integrated radial heat flux *h*(*r*) as a function of radius *r.* In both panels the solid line represents the (frictionless) TE solution. The dashed lines represent numerical calculations with *α*_{r} = 6 × 10^{−3} day^{−1}, 2 × 10^{−3} day^{−1}, 6 × 10^{−4} day^{−1}, and 2 × 10^{−4} day^{−1}, respectively, where an increase in the length of the dashes corresponds to a decrease in *α*_{r}.

Minimum absolute vorticity *ζ*^{min}_{a}*f*) as function of the forcing amplitude *T*_{eo} (in K) for *α*_{r} = 10^{−1} day^{−1} (short dashes), *α*_{r} = 2 × 10^{−2} day^{−1} (medium dashes), *α*_{r} = 2 × 10^{−3} day^{−1} (long dashes), and in the frictionless limit (solid line). The frictionless limit represents the TE solution for *T*_{eo} ⩽ *T*_{c} and the AMC solution for *T*_{eo} > *T*_{c} (*T*_{c} = 0.645 K). Each dot represents a numerical solution of (1)–(5); the lines are drawn to guide the eye by connecting the respective points and have no other significance.

Minimum absolute vorticity *ζ*^{min}_{a}*f*) as function of the forcing amplitude *T*_{eo} (in K) for *α*_{r} = 10^{−1} day^{−1} (short dashes), *α*_{r} = 2 × 10^{−2} day^{−1} (medium dashes), *α*_{r} = 2 × 10^{−3} day^{−1} (long dashes), and in the frictionless limit (solid line). The frictionless limit represents the TE solution for *T*_{eo} ⩽ *T*_{c} and the AMC solution for *T*_{eo} > *T*_{c} (*T*_{c} = 0.645 K). Each dot represents a numerical solution of (1)–(5); the lines are drawn to guide the eye by connecting the respective points and have no other significance.

Minimum absolute vorticity *ζ*^{min}_{a}*f*) as function of the forcing amplitude *T*_{eo} (in K) for *α*_{r} = 10^{−1} day^{−1} (short dashes), *α*_{r} = 2 × 10^{−2} day^{−1} (medium dashes), *α*_{r} = 2 × 10^{−3} day^{−1} (long dashes), and in the frictionless limit (solid line). The frictionless limit represents the TE solution for *T*_{eo} ⩽ *T*_{c} and the AMC solution for *T*_{eo} > *T*_{c} (*T*_{c} = 0.645 K). Each dot represents a numerical solution of (1)–(5); the lines are drawn to guide the eye by connecting the respective points and have no other significance.

Dependence of the cross-vortex circulation on the Rayleigh friction coefficient *α*_{r} for subcritical forcing (*T*_{eo} = 0.5 K). (a) Total strength as quantified by *ψ*_{max}, and (b) the stratospheric part of the secondary circulation as quantified by *w*_{strat}. Each dot represents a numerical solution of (1)–(5). The lines connecting the dots are drawn to guide the eye and have no other significance.

Dependence of the cross-vortex circulation on the Rayleigh friction coefficient *α*_{r} for subcritical forcing (*T*_{eo} = 0.5 K). (a) Total strength as quantified by *ψ*_{max}, and (b) the stratospheric part of the secondary circulation as quantified by *w*_{strat}. Each dot represents a numerical solution of (1)–(5). The lines connecting the dots are drawn to guide the eye and have no other significance.

Dependence of the cross-vortex circulation on the Rayleigh friction coefficient *α*_{r} for subcritical forcing (*T*_{eo} = 0.5 K). (a) Total strength as quantified by *ψ*_{max}, and (b) the stratospheric part of the secondary circulation as quantified by *w*_{strat}. Each dot represents a numerical solution of (1)–(5). The lines connecting the dots are drawn to guide the eye and have no other significance.

Approximate analytical solution of *w*_{strat} according to (C10) as a function of *α*_{r} for point like forcing at (*r*′, *z*′) = (*r*_{o}/2, *z*_{o}/2) and *ρ*_{0} = const., *N*_{o} = const. The function is normalized so that its maximum is 1. The ratio *N*_{o}/*f* = 120 is chosen such that this calculation approximately corresponds with the numerical calculation shown in Fig. 6b.

Approximate analytical solution of *w*_{strat} according to (C10) as a function of *α*_{r} for point like forcing at (*r*′, *z*′) = (*r*_{o}/2, *z*_{o}/2) and *ρ*_{0} = const., *N*_{o} = const. The function is normalized so that its maximum is 1. The ratio *N*_{o}/*f* = 120 is chosen such that this calculation approximately corresponds with the numerical calculation shown in Fig. 6b.

Approximate analytical solution of *w*_{strat} according to (C10) as a function of *α*_{r} for point like forcing at (*r*′, *z*′) = (*r*_{o}/2, *z*_{o}/2) and *ρ*_{0} = const., *N*_{o} = const. The function is normalized so that its maximum is 1. The ratio *N*_{o}/*f* = 120 is chosen such that this calculation approximately corresponds with the numerical calculation shown in Fig. 6b.

Numerical solution for supercritical forcing (*T*_{eo} = 5 K) with small but nonzero friction *α*_{r} = 2 × 10^{−3} day^{−1}. (a) Tangential wind *u* (in m s^{−1}, solid, contours every 2.5 m s^{−1}, zero contour omitted) and potential temperature *θ* (dashed contours). (b) Cross-vortex streamfunction *ψ*, with plotting conventions as in Fig. 3b; the maximum value of the streamfunction is *ψ*_{max} = 3.8 × 10^{7} kg s^{−1}. (c) Potential radius *R* (in km, solid contours) and potential temperature *θ* (dashed contours). (d) Diabatic heating *Q* (in K day^{−1}, contours every 0.02 K day^{−1}, solid contours and dark shading for positive values, dashed contours and white shading for negative values).

Numerical solution for supercritical forcing (*T*_{eo} = 5 K) with small but nonzero friction *α*_{r} = 2 × 10^{−3} day^{−1}. (a) Tangential wind *u* (in m s^{−1}, solid, contours every 2.5 m s^{−1}, zero contour omitted) and potential temperature *θ* (dashed contours). (b) Cross-vortex streamfunction *ψ*, with plotting conventions as in Fig. 3b; the maximum value of the streamfunction is *ψ*_{max} = 3.8 × 10^{7} kg s^{−1}. (c) Potential radius *R* (in km, solid contours) and potential temperature *θ* (dashed contours). (d) Diabatic heating *Q* (in K day^{−1}, contours every 0.02 K day^{−1}, solid contours and dark shading for positive values, dashed contours and white shading for negative values).

Numerical solution for supercritical forcing (*T*_{eo} = 5 K) with small but nonzero friction *α*_{r} = 2 × 10^{−3} day^{−1}. (a) Tangential wind *u* (in m s^{−1}, solid, contours every 2.5 m s^{−1}, zero contour omitted) and potential temperature *θ* (dashed contours). (b) Cross-vortex streamfunction *ψ*, with plotting conventions as in Fig. 3b; the maximum value of the streamfunction is *ψ*_{max} = 3.8 × 10^{7} kg s^{−1}. (c) Potential radius *R* (in km, solid contours) and potential temperature *θ* (dashed contours). (d) Diabatic heating *Q* (in K day^{−1}, contours every 0.02 K day^{−1}, solid contours and dark shading for positive values, dashed contours and white shading for negative values).

Schematic illustration of a hypothetical situation in which the AMC solution extends to an altitude, *z*_{1}, that is above the top *z*_{o} of the thermal forcing region. The thermal forcing region is marked by gray shading. In the radial direction the secondary circulation extends out to radius *r*_{a}. Furthermore, *r*_{u} and *r*_{d} are defined such that there is upwelling for 0 ⩽ *r* ⩽ *r*_{u} and downwelling for *r*_{d} ⩽ *r* ⩽ *r*_{a}.

Schematic illustration of a hypothetical situation in which the AMC solution extends to an altitude, *z*_{1}, that is above the top *z*_{o} of the thermal forcing region. The thermal forcing region is marked by gray shading. In the radial direction the secondary circulation extends out to radius *r*_{a}. Furthermore, *r*_{u} and *r*_{d} are defined such that there is upwelling for 0 ⩽ *r* ⩽ *r*_{u} and downwelling for *r*_{d} ⩽ *r* ⩽ *r*_{a}.

Schematic illustration of a hypothetical situation in which the AMC solution extends to an altitude, *z*_{1}, that is above the top *z*_{o} of the thermal forcing region. The thermal forcing region is marked by gray shading. In the radial direction the secondary circulation extends out to radius *r*_{a}. Furthermore, *r*_{u} and *r*_{d} are defined such that there is upwelling for 0 ⩽ *r* ⩽ *r*_{u} and downwelling for *r*_{d} ⩽ *r* ⩽ *r*_{a}.

Dependence of the radius *r*_{a} on the forcing amplitude *T*_{eo}. The parameter *r*_{a} is the maximum radius out to which the secondary circulation extends in the AMC regime according to the approximate analytical theory.

Dependence of the radius *r*_{a} on the forcing amplitude *T*_{eo}. The parameter *r*_{a} is the maximum radius out to which the secondary circulation extends in the AMC regime according to the approximate analytical theory.

Dependence of the radius *r*_{a} on the forcing amplitude *T*_{eo}. The parameter *r*_{a} is the maximum radius out to which the secondary circulation extends in the AMC regime according to the approximate analytical theory.

Approach of the frictionless limit (AMC solution) for supercritical forcing (*T*_{eo} = 5 K). (a) Tangential wind *u*(*r, z*_{o}) at the top of the thermal forcing region (i.e., at *z*_{o}) as a function of radius *r.* (b) Vertically integrated radial heat flux *h*(*r*) as a function of radius *r.* In both panels the solid line represents the (frictionless) AMC solution according to the approximate analytical theory. The dashed lines represent numerical calculations with *α*_{r} = 10^{−2} day^{−1}, 5 × 10^{−3} day^{−1}, 2 × 10^{−3} day^{−1}, and 10^{−3} day^{−1}, respectively, where an increase in the length of the dashes corresponds to a decrease in *α*_{r}.

Approach of the frictionless limit (AMC solution) for supercritical forcing (*T*_{eo} = 5 K). (a) Tangential wind *u*(*r, z*_{o}) at the top of the thermal forcing region (i.e., at *z*_{o}) as a function of radius *r.* (b) Vertically integrated radial heat flux *h*(*r*) as a function of radius *r.* In both panels the solid line represents the (frictionless) AMC solution according to the approximate analytical theory. The dashed lines represent numerical calculations with *α*_{r} = 10^{−2} day^{−1}, 5 × 10^{−3} day^{−1}, 2 × 10^{−3} day^{−1}, and 10^{−3} day^{−1}, respectively, where an increase in the length of the dashes corresponds to a decrease in *α*_{r}.

Approach of the frictionless limit (AMC solution) for supercritical forcing (*T*_{eo} = 5 K). (a) Tangential wind *u*(*r, z*_{o}) at the top of the thermal forcing region (i.e., at *z*_{o}) as a function of radius *r.* (b) Vertically integrated radial heat flux *h*(*r*) as a function of radius *r.* In both panels the solid line represents the (frictionless) AMC solution according to the approximate analytical theory. The dashed lines represent numerical calculations with *α*_{r} = 10^{−2} day^{−1}, 5 × 10^{−3} day^{−1}, 2 × 10^{−3} day^{−1}, and 10^{−3} day^{−1}, respectively, where an increase in the length of the dashes corresponds to a decrease in *α*_{r}.

Dependence of the cross-vortex circulation on the Rayleigh friction coefficient *α*_{r} for supercritical forcing (*T*_{eo} = 5 K). (a) Total strength as quantified by *ψ*_{max} and (b) the stratospheric part of the secondary circulation as quantified by *w*_{strat}. Each dot represents a numerical solution of (1)–(5). The solid lines connecting the dots are drawn to guide the eye and have no other significance. The dashed lines represent the respective solid lines from Fig. 6 after multiplication by 10; that is, they indicate the hypothetical functional dependence on *α*_{r} if the dependence on *T*_{eo} were linear.

Dependence of the cross-vortex circulation on the Rayleigh friction coefficient *α*_{r} for supercritical forcing (*T*_{eo} = 5 K). (a) Total strength as quantified by *ψ*_{max} and (b) the stratospheric part of the secondary circulation as quantified by *w*_{strat}. Each dot represents a numerical solution of (1)–(5). The solid lines connecting the dots are drawn to guide the eye and have no other significance. The dashed lines represent the respective solid lines from Fig. 6 after multiplication by 10; that is, they indicate the hypothetical functional dependence on *α*_{r} if the dependence on *T*_{eo} were linear.

Dependence of the cross-vortex circulation on the Rayleigh friction coefficient *α*_{r} for supercritical forcing (*T*_{eo} = 5 K). (a) Total strength as quantified by *ψ*_{max} and (b) the stratospheric part of the secondary circulation as quantified by *w*_{strat}. Each dot represents a numerical solution of (1)–(5). The solid lines connecting the dots are drawn to guide the eye and have no other significance. The dashed lines represent the respective solid lines from Fig. 6 after multiplication by 10; that is, they indicate the hypothetical functional dependence on *α*_{r} if the dependence on *T*_{eo} were linear.