The concept of hub clouds and eye moats comes from aircraft observations made by Simpson and Starrett (1955). Figure 1 is adapted from their schematic diagram of Hurricane Edna (9–10 September 1954). Of particular interest is the hub cloud near the circulation center and the clear moat at the edge of the eye. In later years, intense storms like Edna have been found that also possess a warm-ring thermal structure in the lower troposphere—where the temperature surrounding the center of a tropical cyclone is greater than the center, which is in contrast to a warm-core thermal structure where temperature decreases radially outward from the center. A good example is Hurricane Isabel on 13 September 2003, when it had tangential winds in excess of 70 m s⁻¹. Figure 2 shows NOAA WP-3D aircraft data for this storm. The two panels show tangential wind (black curves), temperature (red curves), and dewpoint temperature (blue curves) for a 2.1 km altitude radial leg (lower panel) and a 3.7 km altitude radial leg (upper panel). A warm-ring thermal structure occurs at both levels, with the warmest and driest air at 25 km radius, which is at the outer edge of the eye. In temperature, the warm ring is approximately 4.2°C warmer than the vortex center at 3.7 km and 5.0°C warmer than the vortex center at 2.1 km. As can be seen from the dewpoint depressions, the warm-ring region near 25 km radius is associated with dry, subsiding air at the outer edge of the eye. Enhanced subsidence at the outer edge of the eye tends to produce an eye-moat. Note that the center of the eye is saturated at z = 2.1 km but is not saturated at z = 3.7 km, which is consistent with the top of the hub cloud being located at z ≈ 3 km. The tangential wind profiles reveal a vorticity structure more complicated than the simple structure that will be assumed in Section 3. This is evident from the kinks that occur near r ≈ 20 km for z = 2.1 km and near r ≈ 27 km for z = 3.7 km. In other words, Isabel had a somewhat “hollow” vorticity structure compared to the radially uniform structure that is assumed in Section 3.

The photograph shown here as Figure 3 was taken from the WP-3D aircraft near the edge of the eye, looking towards the hub cloud at the center of the eye. The top of the hub cloud is near 3 km altitude, so the radial leg in the upper panel of Figure 2 is just above the top of the hub cloud, while the radial leg in the lower panel is just below the top of the hub cloud, as is evident in the dewpoint depressions. Since the first internal mode Rossby length in the eye of Isabel at this time is on the order of 10–15 km, the eye diameter ( ∼ 60 km) is approximately 5 Rossby lengths, allowing for the rare opportunity to view balanced dynamical structure over several Rossby lengths in a single photograph. For comprehensive discussions of Hurricane Isabel, see Montgomery et al. (2006), Aberson et al. (2006), Bell and Montgomery (2008), Nolan et al. (2009b,a), and Stern et al. (2016). The eye-moat and mesovortex structure of Hurricane Isabel on 12 September 2003 is discussed by Kossin and Schubert (2004) and Rozoff et al. (2006).

Another common feature of intense tropical cyclones is the “stadium effect,” caused by the outward slope of the eyewall. An example of this effect, taken from the CloudSat data archive, is shown in Figure 4. This vertical cross-section was obtained when CloudSat’s 94 GHz Cloud Profiling Radar made a fortuitous pass directly over the eye of Typhoon Choi-Wan on 15 September 2009. In the top panel of Figure 4, the vertical scale is stretched to highlight the vertical structure of the radar reflectivity. In the bottom panel, only the region inside a radius of 50 km is shown, so the aspect ratio is one-to-one, which clearly reveals the approximate 45° baroclinic tilt of the eyewall updraft. The incorporation of such baroclinic tilt is an important aspect of the theoretical analysis presented here in Sections 2 and 3.

An interesting feature of nonhydrostatic, full-physics, tropical cyclone models is that they can produce intense storms with a temperature field that has a warm-core structure at upper levels, but a warm-ring structure at lower levels. The first example of this was presented by Yamasaki (1983), whose Figure 10A is reproduced here as Figure 5. The upper tropospheric warm-core anomaly is 17°C and is centered at a height of 14 km. A warm-ring thermal structure is found between heights of 2 and 8 km, and at a radius of approximately 7–8 km. Much of the present paper is devoted to a balanced dynamical interpretation of the production of such an overall thermal structure.

For the study of eye subsidence, one can envision using (at least) four different sets of independent variables in space: radius and log-pressure, (r, z); potential radius and log-pressure, (R, Z); radius and potential temperature, (r, θ); or potential radius and potential temperature, (R, Θ). Note that Z = z and Θ = θ, but the upper case symbols Z and Θ are used because (∂/∂Z) ≠ (∂/∂z) and (∂/∂Θ) ≠ (∂/∂θ). Concerning the use of (R, Θ)-coordinates, since the flow in the eye is inviscid and adiabatic, both R and Θ are Lagrangian coordinates, which means that a given parcel in the eye stays on its original R-surface and its original Θ-surface. In other words, when the mathematical analysis is performed in (R, Θ)-space, there is no need for a transverse circulation equation, and the dynamics is more easily understood in the framework of PV and its invertibility principle, as discussed in a theoretical context by Schubert and Alworth (1987), Möller and Smith (1994), and Schubert (2018), and in an observational context by Martinez et al. (2019). This is in sharp contrast to the use of (r, z)-coordinates, where neither r nor z is a Lagrangian coordinate. In this paper, we have analyzed the transverse circulation problem in (R, Z)-coordinates, a setting in which one coordinate is Lagrangian and the other is not. There is a duality between the use of (R, Z)-coordinates and the use of (r, θ)-coordinates, since in the (r, θ)-formulation, one coordinate is Lagrangian (recall that θ ̇ = 0 in the eye) and the other is not. Thus, the essential results obtained in Section 3 using the (R, Z)-formulation could also be obtained using the (r, θ)-formulation. However, some details involved in the two formulations are different. For example, in the (r, θ)-formulation, a convenient vertically sheared Rankine core has solid-body rotation on each isentropic surface rather than on each isobaric surface. Also, in the elliptic problem (Eq. 16), the outer boundary is the coordinate surface R = R _ew, while in the (r, θ)-formulation a sloping eyewall outer boundary would not be an r-surface, thereby producing an additional difficulty in the solution of the elliptic problem. Thus, while the difficulties of including baroclinic effects in the (r, z)-formulation can be overcome by using either the (R, Z)-formulation or the (r, θ)-formulation, the analysis presented in this paper is indicative of our slight preference for the (R, Z)-formulation.

The paper is organized as follows. Section 2 first presents the balanced vortex equations in the (r, z)-formulation, and then transforms to the (R, Z)-formulation. For certain baroclinic vortices, the (R, Z) formulation of the transverse circulation equation can be solved semi-analytically using a vertical transform approach (Section 3). These semi-analytical solutions are used to generalize the barotropic vortex results of Schubert et al. (2007) and to better understand the role of baroclinicity in the distribution of subsidence in the eye of an intense tropical cyclone.

For simplicity, the analysis presented here considers an axisymmetric, balanced flow in the inviscid fluid that lies above the frictional boundary layer. To simplify the primitive equation model to a balanced vortex model, we assume that the azimuthal flow remains in a gradient balanced state, i.e., we discard the exact radial equation of motion and replace it with the gradient balance condition given below as the first entry in Eq. 1. A sufficient condition for the validity of this assumption is that the diabatic forcing effects have slow enough time scales that significant, azimuthal mean inertia-gravity waves are not excited. We shall describe this inviscid flow using the log-pressure vertical coordinate z = H ln(p ₀/p), where H = R _d T ₀/g is the constant scale height, p ₀ and T ₀ are constant reference values of pressure and temperature, R _d is the gas constant for dry air, and g is the acceleration of gravity. Under the balance condition, the governing equations are f + v r v = ∂ ϕ ∂ r , ∂ v ∂ t + w ∂ v ∂ z + f + ∂ r v r ∂ r u = 0 , ∂ ϕ ∂ z = g T 0 T , ∂ r u r ∂ r + ∂ ρ w ρ ∂ z = 0 , ∂ T ∂ t + u ∂ T ∂ r + w ∂ T ∂ z + κ T H = Q c p , (1) where κ = R _d/c _p, c _p is the specific heat at constant pressure, f the constant Coriolis parameter, ρ(z) = ρ ₀ e ^−z/H the pseudo-density in the log-pressure coordinate, ρ ₀ = p ₀/(R _d T ₀) the constant reference density, ϕ the geopotential, u the radial velocity component, v the azimuthal velocity component, w the log-pressure vertical velocity, and Q the diabatic heating. The potential vorticity (PV) equation, derived from Eq. 1, is D P D t = 1 ρ − ∂ v ∂ z ∂ θ ̇ ∂ r + f + ∂ r v r ∂ r ∂ θ ̇ ∂ z , (2) where D/Dt = (∂/∂t) + u(∂/∂r) + w(∂/∂z) is the material derivative, P = 1 ρ − ∂ v ∂ z ∂ θ ∂ r + f + ∂ r v r ∂ r ∂ θ ∂ z = f R ρ r ∂ R , θ ∂ r , z (3) is the potential vorticity, θ = T ( p 0 / p ) κ = T e κ z / H is the potential temperature, c p θ ̇ / θ = Q / T , and the potential radius R is defined in terms of the absolute angular momentum by 1 2 f R 2 = r v + 1 2 f r 2 .

Using the mass conservation principle, we define a streamfunction ψ such that ρ u = − ∂ ψ ∂ z and ρ w = ∂ r ψ r ∂ r . (4) For convenience, we shall refer to ψ as the “streamfunction,” although it is worth noting that it is actually rψ, rather than ψ, that is the “streamfunction” for the transverse mass flux. This flexibility with the factor r proves convenient for the analytical solutions presented in Section 3. Using the gradient balance relation in the tangential wind equation, and using the hydrostatic relation in the thermodynamic equation, we can write ∂ ϕ t ∂ r − ρ B w + ρ C u = 0 , ∂ ϕ t ∂ z + ρ A w − ρ B u = g Q c p T 0 , (5) where ϕ _t = (∂ϕ/∂t) is the geopotential tendency, and where the static stability A, the baroclinicity B, and the inertial stability C are given by ρ A = g T 0 ∂ T ∂ z + κ T H , ρ B = − f + 2 v r ∂ v ∂ z = − g T 0 ∂ T ∂ r , ρ C = f + 2 v r f + ∂ r v r ∂ r . (6)

Eliminating ϕ _t between the two equations in Eq. 5, then expressing u and w in terms of ψ via Eq. 4, and requiring that w = 0 at the top and bottom boundaries, we obtain the following transverse circulation problem (Eliassen, 1951). Secondary Circulation Problem on the Full r , z - Domain: ∂ ∂ r A ∂ r ψ r ∂ r + B ∂ ψ ∂ z + ∂ ∂ z B ∂ r ψ r ∂ r + C ∂ ψ ∂ z = g c p T 0 ∂ Q ∂ r for 0 ≤ r < ∞ and 0 ≤ z ≤ z T with ψ 0 , z = ψ r , 0 = ψ r , z T = 0 , r ψ r , z → 0 as r → ∞ . (7)

Note that A C − B 2 = g T 0 − 1 ρ 0 − 1 e ( 1 − κ ) z / H f + 2 v / r P , so that this problem is elliptic if (f + 2v/r)P > 0, which is typically the case for the tropical cyclone core region studied in this paper.

According to Eq. 7, subsidence in the hurricane eye at a particular time is forced by the (∂Q/∂r) term and is shaped by the three spatially varying coefficients A, B, C at that time. Analytical progress in understanding eye subsidence can more easily be made if we obtain a transformed version of Eq. 7 that contains only two spatially varying coefficients and does not contain any second order mixed derivative terms. The balanced vortex model and the associated transverse circulation equation take simple forms when the original independent variables (r, z, t) are replaced by the new independent variables (R, Z, τ), where Z = z and τ = t but ∂/∂Z and ∂/∂τ imply fixed potential radius R. This transformation (Schubert and Hack, 1983) makes use of ∂ ∂ r , ∂ ∂ z , ∂ ∂ t = ∂ R ∂ r ∂ ∂ R , ∂ R ∂ z ∂ ∂ R + ∂ ∂ Z , ∂ R ∂ t ∂ ∂ R + ∂ ∂ τ , (8) from which it follows that D D t = ∂ ∂ t + u ∂ ∂ r + w ∂ ∂ z = ∂ ∂ τ + w ∂ ∂ Z . (9) Defining U, V, W, and Φ by U = r R u + w f ∂ V ∂ Z , V = r R v , W = f f + ζ w , Φ = ϕ + 1 2 v 2 , (10) where ζ = r ⁻¹ ∂(rv)/∂r is the relative vorticity, the governing equations (Eq. 1) transform to (see Appendix A for further details) f ̂ V = ∂ Φ ∂ R , ∂ V ∂ τ + f U = 0 , ∂ Φ ∂ Z = g T 0 T , ∂ R U R ∂ R + ∂ ρ W ρ ∂ Z = 0 , ∂ T ∂ τ + T 0 g N 2 W = Q c p , (11) where the effective inertial frequency f ̂ and the effective buoyancy frequency N are defined by f ̂ 2 = f + 2 v r 2 = f 2 R 4 r 4 = f 4 f − 2 V R − 2 , N 2 = f + ζ f g T 0 ∂ T ∂ Z + κ T H = e − κ Z / H f + ζ f g T 0 ∂ θ ∂ Z = e − κ Z / H g ρ P T 0 f . (12) It follows that r and R are related by r = ( f / f ̂ ) 1 / 2 R and that rv = RV. Note that the system in Eq. 11 is formally simpler than the system in Eq. 1, even though no additional approximations have been introduced. Also note that the effective buoyancy frequency N can differ greatly from the usual buoyancy frequency in the cyclone core, but they are essentially the same in far-field regions where |ζ| ≪ f and the angular momentum surfaces are nearly vertical, so that (∂T/∂Z) ≈ (∂T/∂z) and (∂θ/∂Z) ≈ (∂θ/∂z).

Using the fourth entry in Eq. 11, we define the streamfunction Ψ such that ρ U = − ∂ Ψ ∂ Z and ρ W = ∂ R Ψ R ∂ R . (13) From Eqs. 4 and 13, and the transformation relations in Eq. 8, it can be shown that RΨ and rψ differ only by a constant, which, without loss of generality, we can take to be zero, so that RΨ = rψ. Using the gradient balance relation in the tangential wind equation, and using the hydrostatic relation in the thermodynamic equation, we can write ∂ Φ τ ∂ R + f ̂ 2 U = 0 and ∂ Φ τ ∂ Z + N 2 W = g Q c p T 0 , (14) where Φ _τ = (∂Φ/∂τ) = (∂ϕ/∂t) is the “geopotential tendency.” Eliminating Φ _τ between these two equations, and then expressing U and W in terms of Ψ via Eq. 13, we obtain the following transverse circulation problem. Secondary Circulation Problem on the Full R , Z - Domain: ∂ ∂ R N 2 e Z / H ∂ R Ψ R ∂ R + ∂ ∂ Z f ̂ 2 e Z / H ∂ Ψ ∂ Z = g ρ 0 c p T 0 ∂ Q ∂ R for 0 ≤ R < ∞ and 0 ≤ Z ≤ Z T with Ψ 0 , Z = Ψ R , 0 = Ψ R , Z T = 0 , R Ψ R , Z → 0 as R → ∞ . (15) The boundary conditions on Ψ come from the requirement that the log-pressure vertical velocity vanishes at the bottom and top boundaries, that the radial component of the secondary circulation vanishes at R = 0, and that the secondary circulation goes to zero as R → ∞. To summarize, the secondary circulation in the entire region is obtained by solving the elliptic problem in Eq. 15 for specified N(R, Z), f ̂ ( R , Z ) , and Q(R, Z). Note that Eq. 15 is formally simpler than Eq. 7 since it does not contain baroclinic, second order cross derivative terms. Baroclinic effects are implicit in Eq. 15 and are recovered when the solution Ψ(R, Z) is transformed back to the physical space form ψ(r, z) using ψ = (R/r)Ψ.

We can understand several aspects of eye subsidence by solution of a restricted version of the full elliptic problem in Eq. 15. In this simplified problem, we restrict our attention to the eye region 0 ≤ R ≤ R _ew, where the eyewall potential radius R _ew (i.e., the inner edge of the eyewall) is assumed to be a constant. In the eye region, we assume that Q = 0, so that the elliptic problem in Eq. 15 simplifies as follows. Secondary Circulation Problem on the Restricted R , Z - Domain: ∂ ∂ R N 2 e Z / H ∂ R Ψ R ∂ R + ∂ ∂ Z f ̂ 2 e Z / H ∂ Ψ ∂ Z = 0 for 0 ≤ R ≤ R ew and 0 ≤ Z ≤ Z T with Ψ 0 , Z = Ψ R , 0 = Ψ R , Z T = 0 , Ψ R ew , Z = Ψ ew Z . (16)

To summarize, the secondary circulation in the eye is obtained by solving the homogeneous, elliptic problem in Eq. 16 with specified N(R, Z), f ̂ ( R , Z ) , and Ψ_ew(Z). Note that all the dynamics in the region R _ew < R < ∞, described explicitly in Eq. 15, has been replaced in Eq. 16 by the specification of Ψ(R, Z) at R = R _ew. Because the elliptic problem in Eq. 16 is homogeneous, the solution Ψ(R, Z) has no local maxima or minima in the interior of the restricted domain (i.e., in the eye). For the problems considered here, we have chosen Ψ_ew(Z) ≤ 0, so that Ψ(R, Z) attains its minimum value on the outer boundary R = R _ew and its maximum value of zero on the other three boundaries, R = 0 and Z = 0, Z _T, where Z _T = z _T represents the tropopause. By concentrating attention on the restricted problem (Eq. 16), the goal is to understand how the spatial structure of eye subsidence is shaped by N(R, Z), f ̂ ( R , Z ) , and Ψ_ew(Z). A semi-analytical solution of a simplified version of Eq. 16 is given in Section 3, using vertical transform methods. An alternative approach, not explored here, is to study solutions of the geopotential tendency equation (see Appendix B).

Two problems for the secondary circulation in R-space have been formulated. The first is the full domain elliptic problem (Eq. 15), which requires knowledge of the coefficients f ̂ ( R , Z ) and N(R, Z), and the forcing Q(R, Z) over 0 ≤ R < ∞. The second is the restricted domain elliptic problem (Eq. 16), which requires knowledge of the coefficients f ̂ ( R , Z ) ] and N(R, Z) over 0 ≤ R ≤ R _ew and specification of the lateral boundary values Ψ_ew(Z) at R = R _ew. For the barotropic vortex core and the vertically sheared Rankine core, the restricted domain elliptic problem has been solved semi-analytically in Section 3. The results for the middle-tropospheric forcing barotropic case shown in the upper left panels of Figure 10 and Figure 11 are closest to the results shown in the previous study (Schubert et al., 2007), which were based on full domain, physical space solutions of the transverse circulation equation with forcing of the first internal mode only. In the results for the lower-tropospheric forcing barotropic case shown in the lower left panels of Figure 10 and Figure 11, the minimum value of Ψ_ew(Z) has been shifted to a lower level, which has the effect of producing an enhancement of ∂θ/∂t in the lower troposphere at the edge of the eye. The results for the vertically sheared (baroclinic) Rankine core, shown in the right panels of Figure 10 and Figure 11, more realistically capture the outward tilt of the eyewall and illustrate to a greater degree the preference for the potential temperature tendency to be largest in the lower troposphere at the edge of the eye in baroclinic vortices for which the minimum value of the eyewall streamfunction Ψ_ew(Z) occurs below middle-tropospheric levels. The right two panels of Figure 11 are consistent with the concept of a potential temperature tendency field that is nearly uniform in radius at upper tropospheric levels but highly biased toward the edge of the eye at lower tropospheric levels, i.e., consistent with the development of a warm core at upper levels and a warm ring at lower levels, as seen in the full-physics simulations of Figure 5.

It is important to note that the model of eye subsidence used here is highly idealized, i.e., it is axisymmetric, gradient balanced, inviscid, adiabatic, and for a restricted domain. The adiabatic idealization results because, in the formulation of the restricted problem (Eq. 16), it has been assumed that Q = 0 for R ≤ R _ew. Careful inspection of the lower panel of Figure 4 indicates that this assumption might be violated in the upper troposphere near the edge of the eye, where frozen condensate can be advected inward and subsequently sublimated, producing a region where Q < 0. Malkus (1958), Willoughby (1998), and Zhang et al. (2002) have considered the role that such cooling might play in producing deep, narrow downdrafts at the edge of the eye. The relative roles of such diabatic dynamics and the adiabatic dynamics studied here deserve further study. The idealization of gradient balance filters inertia-gravity waves, which results in a “slow manifold” view of eye dynamics. With their mesoscale scanning techniques, the GOES-R series of geostationary satellites can image a 1000 km × 1000 km hurricane area with 30 s time resolution. When viewing such rapid-scan loops of a major hurricane, one is struck by the highly dynamic nature of the inner core. Some of the high frequency variability of the inner core is probably due to inertia-gravity wave oscillations that are not captured by the simplified dynamics of the balanced vortex model. However, the balanced vortex model does capture the slow-manifold dynamics upon which we can crudely imagine the higher frequency inertia-gravity waves are superposed. To improve this crude view, much work remains to understand how the inner core PV dynamics can become frequency matched with the inertia-gravity wave oscillations, so the two types of dynamics can evolve in a strongly coupled fashion.

In closing, we comment on the possible effects of lower-tropospheric warm-ring structure on the stadium effect. General experience with solutions of the Eliassen transverse circulation equation supports the notion that baroclinic effects play an important role in determining the outward tilt of the eyewall, as seen for example in the CloudSat observations of Figure 4. We have studied the secondary circulation in the eye when the vortex has a warm-core structure at all levels, as shown in the right panel of Figure 7. The results indicate that a lower-tropospheric warm-core structure can be modified to a lower-tropospheric warm-ring structure when the subsidence is enhanced in the lower troposphere at the edge of the eye, as shown in panels D of Figure 10 and Figure 11. When a lower-tropospheric warm ring develops in a tropical cyclone, the baroclinic terms acquire a more complicated spatial structure, which means the absolute angular momentum surfaces also acquire a correspondingly more complicated spatial structure. To see this, consider the thermal wind equation written in the form f ²(R/r)³(∂R/∂z) = (g/T ₀)(∂T/∂r). From this form of the thermal wind equation, we can easily deduce the following general rules. Radially inward of a warm ring: ∂ T ∂ r > 0 ⇒ ∂ R ∂ z > 0 ⇒ R - surfaces tilt inward with height; Radially outward of a warm ring: ∂ T ∂ r < 0 ⇒ ∂ R ∂ z < 0 ⇒ R - surfaces tilt outward with height. (35) Thus, if a tropical cyclone has a warm core at all pressure levels, the R-surfaces tilt outward with height everywhere. Since the R-surfaces help shape the secondary circulation, the eyewall updraft would generally be expected to tilt outward at all levels. However, if a tropical cyclone has a warm-core structure in the upper troposphere but a warm-ring structure in the lower troposphere, the R-surfaces tilt inward with height in the lower troposphere just inside the radius with maximum temperature anomaly. The effect is to make the secondary circulation outward-tilted at upper levels but more vertical at lower levels—an interesting refinement of the stadium effect. To better understand such refinements, additional solutions of Eq. 16 for baroclinic vortices with warm-ring structure would be helpful and are a topic for future work since such solutions would require a numerical approach to Eq. 16. The semi-analytical approach used here is restricted to the vertically sheared, Rankine vortex and simply provides a snapshot in time (rather than a complete time evolution) that illustrates the tendency for a vortex with a warm-core structure to transition to a vortex with a lower-tropospheric warm-ring structure.