WorksPhysics and Astronomy
Critical-level wave absorption controls chaos in a wave–mean flow system
Abstract
Wave-mean flow interactions in stratified fluids can drive spontaneous oscillations, yet the mechanisms controlling transitions between periodic and chaotic regimes remain poorly understood. Using a minimal one-dimensional model of wave-driven mean-flow reversals motivated by the quasi-biennial oscillation (QBO), we identify two distinct mechanisms that suppress the formation of multiple extrema in the velocity profile: critical-level momentum deposition by a broad spectrum of undamped small-scale gravity waves, and wave-amplitude growth due to exponential density stratification. Although physically distinct, both mechanisms inhibit the route to chaos and reduce sensitivity to small-scale viscosity, substantially widening the region of periodic behavior in parameter space. These results identify a general internal mechanism for the stabilization of wave-driven oscillations, with direct implications for QBO-like phenomena in geophysical and astrophysical flows.
Wave–mean flow interactions in stratified fluids can drive spontaneous oscillations across a wide range of physical systems, from laboratory experiments [1, 2, 3] to planetary atmospheres and models of stellar interiors [4, 5]. A prominent geophysical example is the quasi-biennial oscillation (QBO) of equatorial stratospheric winds, which influences interannual variability at the planet scale [6]. A wide range of QBO-like regimes, including irregular and chaotic oscillations, has been reported across modeling hierarchies, from one-dimensional reduced models [7, 8] to idealized direct numerical simulations [9, 10, 11] and general circulation models of planetary atmospheres and stratified layers of stellar interiors [5, 12], indicating that the mechanisms controlling the robustness of nearly-periodic mean-flow reversals remains incompletely understood.
These dynamics hinge on wave–mean flow interaction, driven by a spectrum of internal gravity waves whose parameterization in general circulation models remains challenging [13, 14, 15, 16]. We focus on the minimal configuration leading to low-frequency oscillations: vertically propagating waves generated below a vertically sheared zonal flow and absorbed within it. Horizontally propagating waves in laterally sheared flows are not required for the oscillation itself, although they can perturb these oscillations, as in the case of recent QBO disruptions on Earth [17].
Early reduced models by Holton, Lindzen and Plumb explained the spontaneous emergence of oscillations by considering a standing planetary wave source [18, 19]. Subsequent work showed that modifying the wave spectrum or its parameterization can qualitatively change the dynamics [14, 20], including the emergence of oscillations [21, 22] or, surprisingly, the stabilization of oscillations against chaos under broadband wave forcing [23], though the mechanisms behind these effects remain unclear.
In this Letter, we isolate the mechanisms responsible for the stabilization of periodic oscillations in the presence of broadband wave forcing, building on the finding that transitions to chaos in wave–mean flow systems are associated with the amplification of vertical modes containing multiple velocity extrema [24]. Using a minimal extension of the Holton–Lindzen–Plumb framework, we show that two distinct processes suppress these multiple extrema: (i) state-dependent momentum deposition through critical-level absorption of undamped small-scale gravity waves, captured by the seminal Lindzen-Holton 1968 parameterization [13]; and (ii) density-driven amplification of wave forcing due to exponential stratification. Although physically distinct, both mechanisms suppress secondary shear layers and thereby inhibit the route to chaos, enlarging the periodic regime.
Model
We consider a stratified fluid layer of depth , with an exponential background density profile characterized by a scale height . The flow is forced from below by two counter-propagating planetary waves of phase speeds and momentum fluxes , subject to radiative attenuation over a vertical length scale , and supplemented by a broad spectrum of undamped small-scale gravity waves (SGW) carrying zero net momentum flux, with eastward and westward contributions and , respectively. The system is governed by four nondimensional parameters,
which control, respectively, the relative importance of wave forcing and viscosity, the relative amplitude of SGW forcing, the degree of planetary-wave attenuation, and the degree of exponential stratification.
In the quasilinear approximation, the horizontally averaged zonal flow evolves under the divergence of wave-driven momentum fluxes and viscous diffusion. In an anelastic medium, density stratification amplifies wave amplitudes with altitude, introducing an exponential factor that weights the wave-flux divergence. Introducing nondimensional variables , , and , we obtain
with , no-slip at the lower boundary , and a stress-free condition at the top. Here and denote the vertical momentum fluxes associated with planetary and small-scale gravity waves, respectively. Both wave populations are forced exclusively at the lower boundary, representing wave generation in the underlying troposphere. Changing the lower boundary condition for the mean-flow to free-slip does not qualitatively affect periodicity recovery (Supplementary Material), consistent with structural stability observed in previous studies on this model [25, 26].
Planetary-wave forcing follows Plumb [19]. The two waves undergo Doppler-shifted radiative attenuation, yielding the net momentum flux
where corresponds to critical levels at which absorption becomes complete. For simplicity, we assume an altitude-independent attenuation length , whereas it varies in real atmospheres (see Supplementary Material for the consequences).
Small-scale gravity-wave forcing follows Lindzen and Holton [13] (LH68 hereafter), assuming that their attenuation length scale is much larger than the domain height. The key mechanism is critical-level absorption: a wave of phase speed (in units of ) deposits all its momentum at the lowest altitude where , and carries none higher. Assuming a uniform distribution of phase speeds at the lower boundary and integrating over the spectrum yields the effective closure
where the condition encodes critical-level absorption: the flux carried by a given wave with phase speed vanishes above its lowest critical level, making a nonlinear, state-dependent functional of . We emphasize that $\mathcal{F}_{sgw} is not an arbitrary additional term: it is the weak-attenuation limit of a continuum of waves with different phase speeds, each obeying the P77 absorption law [20]. LH68 thus captures, more simply and cheaply than P77 (see Supplemental Material), what we expect to be the generic effect of nearly undamped gravity waves. For analytical simplicity, we use a waterbag momentum-forcing distribution, sufficient to reproduce the generic behavior found for broad wave-forcing distributions in [23].
In the limit and , we recover the textbook model as formulated by Plumb [19] (P77 hereafter), who also discussed the case of finite (P77 hereafter). Representative terrestrial values are , and [27], and yields QBO-like oscillations. However, and especially are poorly constrained in practice: owing to the uncertainty in turbulent viscosity, spans 2–400 [8], which in the P77 model encompasses the full range of dynamical regimes, from rest to chaos. Why, then, are nearly periodic states so prevalent in both more realistic models and observations?
Results. Figure 1 illustrates the central result of this Letter. In the absence of small-scale gravity-wave forcing (), the P77 model produces a chaotic mean-flow reversal pattern characterized by the irregular development of high-altitude secondary jets (panel a). Adding the LH68 parameterization () at otherwise identical parameter values restores a clean periodic oscillation (panel b). This dramatic transition from irregular, broadband reversals to a well-defined QBO-like cycle motivates the mechanistic analysis that follows.
Figure 1. Simulations of Holton-Lindzen-Plumb model for horizontal mean flow reversals a) forced by a single standing planetary-scale wave ( in the text, with ) b) forced with an additional contribution modelled by the Lindzen-Holton 1968 parameterization (LH68 with ). In both cases, , and .
Figure 2 compares the early evolution of a wind profile with multiple extrema under three configurations: P77 alone (, ), LH68 alone (no planetary waves, with ), and P77 (, ). Under P77, local extrema are amplified as fronts descend, with velocities converging toward [19]; at large , secondary shear layers can grow unchecked [24]. Under LH68, fronts also form and propagate downward, but the velocity field cannot develop amplitudes beyond its initial range, and existing extrema are progressively erased. Although multiple jets initially co-exist, the system eventually reaches a steady state with a homogeneous velocity profile above a lower boundary layer [13]. Under P77, the exponential amplification of wave forcing at upper levels () dominates over radiative attenuation when (i.e. when the density scale height is smaller than the wave attenuation length scale ), accelerating convergence toward at altitude and suppressing secondary extrema before they can develop. Since chaotic dynamics in the HLP model arise through the amplification of modes with multiple velocity extrema [24], both mechanisms are expected to inhibit this route to chaos.
Figure 3 shows bifurcation diagrams as a function of with and without LH68, at fixed and . Including LH68 with slightly lowers the oscillation threshold (from to 3), consistent with the additional momentum flux from the small-scale wave spectrum. More strikingly, it strongly expands the range of periodic and quasi-periodic states: chaotic dynamics present above in P77 are suppressed up to for . Thus, adding small scale gravity wave forcing does not destabilize the system but instead restores regularity, extending the periodicity-resurgence results of Léard et al. [23] and Chartrand et al. [24] to the minimal LH68 parameterization.

Figure 3. Effect of adding LH68 parameterization to P77 model, with , . a) Bifurcation diagram comparing a situation without LH68 () and a case with LH68 parameterization () b) Hovmöller diagram showing the structure of the QBO state at , with and without additional LH68 parameterization.
At , we observe in Figure 3 a transition from a chaotic state () to a quasi-periodic state (), with fast oscillations confined to the lower domain while the upper part approaches a period-one structure (panel b). We interpret this as a competition between the two forcings: P77 dominates near the bottom where radiative attenuation is strongest, while LH68 — whose amplitude is altitude-independent — controls the upper domain. If is further increased, LH68 progressively wins this competition, driving the full profile toward periodic behavior.
Let us now turn to the the effect of exponential density variations. Figure 4 shows that decreasing drives the system toward periodic regimes regardless of . In the low-Reynolds case (panel a), exponential amplification of the forcing at small confines oscillations to the upper part of the domain. Panel b shows an abrupt transition from chaotic to periodic dynamics below a critical value . No chaotic states are observed for up to , indicating a stronger stabilization than that achieved by LH68 alone. This results from the exponential amplification of wave forcing at upper levels, which suppresses secondary extrema and accelerates convergence toward , preventing interior shear layers from developing before they can seed chaos.
Figure 4. Effect of exponential density decay (with ). a) . b) .
Conclusion. We have identified two distinct ways to restore periodic oscillations in the Holton–Lindzen–Plumb system at large : (i) adding the critical-level absorption of nearly undamped small-scale gravity waves, as described by the LH68 parameterization; or (ii) accounting for the exponential decay of the background density (P77). With either effect, QBO-like periodicity becomes much more robust and less dependent on small-scale viscosity than in the P77 case. Result (i) highlights the importance of the vertical attenuation length of the waves: weakly damped waves (with attenuation length larger than the domain height), modeled either by spectra of P77-type waves, as in [23, 24], or by LH68, favor periodicity. We conjecture that this holds for other parameterizations of weakly damped waves, including wave breaking below critical levels and saturation [22]. This remains to be established, as does the extension to a turbulent viscosity enhanced where waves break [28].
The competition between chaotic and periodic states can be summarized as follows. On one side, stronger standing-wave forcing (high ) destabilizes an increasing number of high vertical modes, that is, vertical profiles of the zonal wind with multiple extrema, whose amplification marks the transition to chaos [24]. On the other side, undamped-wave forcing (LH68) and density stratification each counteract this growth. LH68 sweeps secondary extrema toward the lower boundary through localized critical-level absorption before they can grow. Exponential amplification of the forcing by density variations instead favors the formation of a front aloft, as , and accelerates its descent before secondary extrema have time to develop. In both cases, extrema are removed faster than they can grow, preventing chaos.
These mechanisms differ from standard strategies for taming chaos, which rely on feedback or weak periodic forcing applied from outside the system [29, 30]. Here, all forcings are steady and periodicity emerges from wave–mean flow interactions only, which provides a route complementary to synchronization with an imposed annual cycle [25]. Considering such external forcing and generalizing the present framework to account for meridional circulation, either in a two-dimensional model [31] or through parameterized upwelling in one dimension [14, 20, 25, 27], will be necessary to identify the dominant mechanisms governing the flow-reversal periodicity.
The question of what sets regularity versus chaos in a wave-driven mean flow extends beyond QBO-like phenomena. On the one hand, momentum deposition by critical-layer absorption of a broadband wave spectrum, as in LH68, could play a role in equatorial deep oceanic jets [32], in the dynamics of multiple zonal jets in fluids and plasmas [33, 34], and in wave–vortex interaction models [35]. On the other hand, the momentum forcing in LH72 and P77 has inspired similar models for strato-rotational instabilities [2, 36], and is a streaming effect closely related to Faraday-wave-driven flows [37] acoustic streaming [38]. Whether the mechanisms favoring periodic states identified here have counterparts in these systems remains an open question.
Acknowledgement
We acknowledge Cerasela Calugaru and the whole team from PSMN/CBP at ENS de Lyon for computing facilities. We warmly thank Ugo Martinez for his preliminary work on the numerical model, Louis Couston and Jason Reneuve for useful discussions on wind reversals with multiple wave forcing. This work was funded by French Research Agency (ANR) under BOURGEONS project, grant ANR-23-CE40-0014-01, SEEDLING project, grant ANR-24-EXMA-0001, and CliMath project, grant ANR-23-EXMA-0003, PEPR Math-Vives.
Supplementary Material
These supplementary materials contain
Appendix A: Details on model derivation and adimensionalization.
Appendix B: LH68 recovered from P77 with a continuum of weakly damped internal waves.
Appendix C: Numerical details.
Appendix D: Hovmöller diagrams of QBO-like states with exponential density variations.
Appendix E: Adding back LH68 parameterization to the original Holton-Lindzen 72 model, including altitude-dependent wave attenuation length scale
Details on model derivation and adimensionalization
The aim of this appendix is to show how the nondimensional P77+LH68 model presented in the letter is related to the classical dimensional Holton-Lindzen-Plumb model augmented with Lindzen-Holton 68 parameterization.
Quasi-linear framework. The model assumes the dynamics is confined in a vertical plane above the equator, in a stratified fluid described by anelastic equations (sound waves are filtered out but strong variation of atmospheric density are accounted for). The starting point of the model is a decomposition between the zonally () averaged equatorial wind, and its fluctuations (), with no net vertical flow (). Owing to periodicity in direction, the zonally averaged horizontal momentum equation reads
where is a constant viscosity, and is the background atmospheric density with scale height . The total wave momentum flux has been separated into a planetary wave () component and a small-scale gravity wave () component , neglecting (or time-filtering) cross-correlation terms. Importantly, the sgw component describes the effect of a continuum of small scale gravity waves, not just a single standing wave. Those quadratic terms are computed using a quasilinear approach, assuming a time scale separation between mean flow evolution and the adjustement of the wave-field around this mean flow, and a scale separation between mean flow variations in direction and vertical wavelength, allowing for a WKB computation of the waves. More details on the derivation of this model and underlying hypotheses are given in [26]. In particular, the derivation relies on a distinguished limit of infinitely large Richardson number (large buoyancy frequency compared to the vertical shear of the horizontal mean flow). In this limit, wave momentum is fully absorbed at critical layers, and no shear instability occurs. In the real stratosphere, the Richardson number is generally large but can be locally small, potentially leading to wave over-reflection and instabilities. These effects are not accounted for in the present model; incorporating them is an interesting direction for future work.
Holton-Lindzen 1972 (HL72) model for planetary waves. In the original formulation of the model [18], the planetary wave flux is modeled as a combination of equatorial Kelvin and Rossby waves:
where and are the horizontal phase speeds of the Kelvin and Rossby waves. Their vertical attenuation is controlled by radiative damping, with characteristic length scales
depending on the buoyancy frequency , damping rate , and horizontal wavenumbers and . Rotation of the planet is controlled by with the gradient of Coriolis parameter at the equator.
Plumb 1977 (P77) simplification. The special case , , and corresponds to Plumb [19], and involves a single standing internal gravity wave forcing. The model can also be generalized to multiple wave forcings, where each wave is associated with a pair , or equivalently with a given frequency and wavenumber. Each wave then contributes independently of the others [20].
LH68 parameterization. The momentum flux induced by a continuum of small scale gravity waves is modelled using LH68 formulation. [13] built on [39] to describe how undamped small-scale gravity waves deposit their momentum at the lowest critical level, where their phase speed coincides with the mean flow velocity:
Here, is the momentum flux density at the bottom of the stratosphere, integrated over all phase speeds:
with the sign function. The LH68 parameterization transfers momentum to the mean flow once a wave encounters their lowest critical level.
In presence of multiple critical levels in a given profile, only the lowest one is subject to momentum forcing. Some part of the wind profile are therefore not subject to any forcing. This concerns also altitudes where the wind velocity exceeds the range of horizontal phase speed in the forcing.
Adimensionalization. We now consider the P77 setting of HLP, with , , in Eq. (6).
We assume that small-scale gravity wave momentum flux at the bottom is described by a water-bag density distribution with constant amplitude, i.e.,
and elsewhere. This simple waterbag momentum-forcing distribution is sufficient to reproduce the generic behavior obtained for a broader range of wave-forcing distributions in [23]. Note also that results are unchanged for a waterbag distribution where for for , when .
Let us recall for completeness that the system admits four non-dimensional parameters
with the bottom amplitude of the momentum flux per unit mass for each of the planetary waves. Following [40], we adimensionalize the model using
which leads to the rescaled equation and boundary conditions:
with whenever there exists a depth such that , and otherwise, consistently with the LH68 critical-level formulation with waterbag forcing. The system is complemented with boundary conditions
These equations provide a minimal model capturing the combined effects of planetary waves (controlled by ) and small-scale gravity waves (controlled by ).
The textbook model for the QBO explanation is recovered in the limits (Boussinesq atmosphere), (semi-infinite stratosphere), and (planetary waves only). In that case, the first bifurcation from a rest state to a limit cycle as is increased was described by [19] and studied in depth by [3]. The emergence of other oscillatory regimes as is further increased has been explored by [8], while the impact of changes in the effective stratospheric depth on both the first bifurcation and chaotic states was addressed by [24]. The influence of the density scale height on the periodic regime was described first by [19]. The role of the small-scale gravity wave parameter in promoting transitions toward periodic states was discussed by [14]. This effect has also been addressed by [21, 22] for more general small-scale gravity-wave parameterizations. Finally, [27] examined how shapes the QBO buffer layer. In the present study, we show that increasing and decreasing suppress chaotic states, thereby enhancing the robustness of regular QBO-like oscillations.
We show in Figure S1 that changing the lower boundary condition from no-slip to free-slip does not modify qualitatively the periodicity recovery described in Figure 1 of the main text. The fact that free-slip and no-slip boundary conditions does not change the system behavior at a qualitative level is consistent with the results of [25] when addressing QBO synchronisation with seasonal forcing and [26] on the first bifurcation and subsequent quasi-periodic route to chaos.

Figure S1. Simulations of Holton-Lindzen-Plumb model for horizontal mean flow reversals identical to Figure 1 of the main text except for the lower boundary condition .
LH68 recovered from P77 with a continuum of weakly damped internal waves.
We show in this appendix that LH68 can be interpreted as a mean-field model for a weakly attenuated continuum of waves described by Plumb 77 parameterization. Let us consider a continuum of internal waves spanning a range of phase speeds and fixed wavenumber within the HLP framework. Using notation from the main text, the small-scale gravity wave momentum flux can be expressed as
where is the attenuation length of each wave and is the momentum flux density at the base of the stratosphere.
Assuming that the wavenumber of the small-scale waves is much larger than that of planetary waves, consistent with generation by localized convection, we consider the limit . Physically, this corresponds to weakly damped waves that propagate nearly freely until they reach their first critical level, with (stratosphere height). In this limit, the exponential term approaches a Heaviside function:
where is the lowest altitude where the critical level condition is satisfied. Using this, the total momentum flux becomes
where is the Heaviside step function. Taking the vertical derivative and applying the identity for such that , we obtain:
This expression applies to the first critical level encountered by each wave; at higher critical levels, all momentum has already been absorbed, so the flux divergence is zero. With the no-slip boundary condition , we have at altitudes where LH68 is active, recovering the original parameterization (Eq. (8)). This allows us to safely use LH68 to interpret the role of weakly damped waves described by [23, 24] within the HLP framework.
To validate our implementation of the LH68 parameterization, we show in Figure S2 a configuration similar to Section 4 of the original Lindzen and Holton [13] paper [13]. Panel (a) presents a case without viscosity, as in Figure 2 of [13], while panel (b) presents a case with the same viscosity as in the main text (our Figure 2).

Figure S2. Temporal evolution of an initially linear wind profile (plain black line), forced by the LH68 parameterization with a waterbag distribution of undamped waves at the bottom boundary ( constant for , zero otherwise). a) Without explicit viscosity (a configuration similar to that of Sec. 4 of LH68) b) With a weak viscosity, set by (and , since there is no standing-wave forcing). In both panels, .
Numerical details
We solve (S9) using a finite difference semi-implicit scheme: the viscous term is treated implicitly, while the nonlinear advection term is advanced using a first-order upwind scheme. Higher-order schemes were tested but found to produce spurious oscillations near the front where , which destabilize the simulation since any overshoot violates the physical bound .
Figure S3. Hovmöller diagrams illustrating the effect of resolution the periodicity recovery. Parameters are the same as in figure 1 of the main text, only the number of vertical levels varies. Figure 1 was done using .
Instabilities also arise when the vertical resolution is too coarse. The WKB integral in the wave forcing term becomes poorly resolved near the critical level, leading to spurious overestimation of the forcing just above it and to inaccurate localization of the critical-level height. Both effects produce strong fluctuations in the forcing near the critical level.
To determine empirically compatible values of and ensuring both CFL condition and convergence of the WKB wave forcing estimate, we defined a critical time step as follows: for given parameters and given , starting from an initial profile that reaches at a known altitude, we advance the mean flow by one explicit time step and identify the largest for which the bound is not violated at or below the front. Because of the previous effects, this critical time step becomes strongly dependent on the vertical resolution when the grid is too coarse, and converges as goes to 0. We consider that the integral is well resolved when this critical time step becomes constant with respect to the grid size
This provides a physically grounded criterion that is more stringent than a standard CFL condition. With this criterion, it was not necessary to use the ad hoc threshold on the maximum allowed velocity that was needed in [8, 23, 24].
Based on this procedure, we adopted as a conservative choice: at this resolution, the critical time step varies smoothly with and , indicating that the WKB integral is well-resolved across the parameter range explored in this study. The time step is then set just below the critical value for each parameter combination.
The sensitivity of the HLP model to numerical resolution was studied by [8], which demonstrated the robustness of the quasi-periodic route to chaos. Figure S3 further illustrates the robustness of periodicity recovery through exponential density stratification and small-scale gravity waves parameterized with LH68. Note that in the chaotic regime, the solution behavior is robust to changes in vertical resolution, though the solution itself is not, as expected due to sensitivity to initial conditions.
We also analyze the sensitivity of the LH68 parameterization to vertical resolution by comparing the temporal evolution of the profile shown in Figure 2 of the main text for different grid spacings. The comparison between panel (b) of Figure 2 in the main text and Figure 3 of this supplementary material demonstrates that our vertical resolution is sufficient to ensure that the LH68 dynamics are well-resolved.
Figure 2. Temporal evolution of a wind profile (initial state: plain black line) a) b) idem with and c) same as panel a with . In dimensional units, cases (a) and (c) correspond to , , while case (b) corresponds to , . All cases share the same viscosity value. The corresponding imposed momentum flux density is plotted in each case. Time between initial and final profiles: for cases a, b, c respectively.
Hovmöller diagrams of QBO-like states with exponential density variations
Figure S5 shows typical QBO-like states obtained with the P77 configuration, i.e. planetary-wave forcing with exponential background density variations. In panel a, weak forcing combined with exponential density decay leads to intensified dynamics near the top of the domain, so that the oscillation threshold is exceeded only at high altitude. Note the qualitative similarity with the observed pattern of the Semi-Annual Oscillation in the upper stratosphere. Panel b illustrates how a state that is chaotic in the uniform-density limit () recovers periodicity when is sufficiently small.

Figure S5. Hovmöller diagrams a) b)
Figure S4. Temporal evolution of a wind profile (initial state : plain black line) for different vertical resolutions. Parameters are the same as in panel b) of figure 2 from the main text, only the number of vertical levels varies. Figure 2 was done using .
adding LH68 back to HL72, including altitude-dependent wave attenuation length scale
We now show that the results obtained with P77 model holds for the more general HL72 model, where the two counter-propagating waves driving the QBO-like oscillation have distinct propagation characteristics. This asymmetry is a direct consequence of Earth’s rotation on the equatorial waveguide: the two waves correspond to an equatorially trapped Kelvin wave and a Rossby wave, which differ in both their dispersion relations and attenuation length scales.
The HL72 framework is of particular interest here for two reasons. First, it is historically the reference model for the QBO mechanism. Second, and more relevantly for the present study, it provides the natural setting in which to examine the combined effect of planetary and small-scale internal waves.
Concretely, the asymmetry is captured by the parameter appearing in the Rossby-wave momentum flux term in Eq. (6), and by the difference between the attenuation length scales and of the Kelvin and Rossby waves respectively. Following HL72 we consider , and . As a result of this symmetry breaking, the zero-wind state is no longer a steady solution of the system, and the transition to a QBO-like oscillation is more complex than in the P77 case. Nevertheless, a quasi-periodic route to chaos persists [8], and the insights gained from the symmetric case carry over, as we discuss below.
HL72 also emphasized the need for an increasing damping rate with altitude to reproduce a realistic QBO structure. We adopt the same formulation here, defining:
with , and . In that case the attenuation length scale is a function of altitude. The importance of this altitude-dependent parameter in shaping the vertical structure of the QBO on Earth has been emphasized by [41]. Based on the analysis performed in the main text, we expect that this increase of wave attenuation would counteract the effect of exponential density variations and favor chaotic over periodic regimes at large , as we verify in the following.
Figure S6 illustrates different dynamical regimes in this setup. Non-dimensional parameters , and are obtained by considering the value of at . Panel a reproduces the original HL72 parameters, yielding periodic oscillations. This periodic behavior persists both when exponential density stratification is removed (panel b) and when increased wave attenuation is removed (panel c, with held fixed), highlighting the key role of enhanced wave attenuation at high altitude in producing QBO-like oscillations. Indeed, this enhanced attenuation compensates for the exponential growth of wave forcing with altitude.
Figure S6. Hovmöller diagrams illustrating the dynamics of the modified HL72 model a) Reference HL72 simulation with varying attenuation coefficient (Eq. (20)) b) Reference HL72 simulation with c) Reference HL72 simulation with constant attenuation lengthscale d) e) f)
So far, we have assumed a depth-independent radiative attenuation. More realistic models of the stratosphere instead let the attenuation increase with altitude [18], a feature that plays a key role in shaping the vertical structure of the observed QBO on Earth [41]. Based on our analysis, we expect that this increase of wave attenuation would counteract the effect of exponential density variations and favor chaotic over periodic regimes at large , as we verified in the original Holton and Lindzen [18] setting (see Supplementary Material).
Panels d and e show quasi-periodic and chaotic states obtained by increasing the Reynolds number relative to the original HL72 study, with and without exponential density stratification respectively. Panel f demonstrates that adding the LH68 parameterization restores order: the chaotic state of panel e transitions to a quasi-periodic regime.
References
- [1]R. A. Plumb and A. D. McEwan, The instability of a forced standing wave in a viscous stratified fluid: A laboratory analogue of the quasi-biennial oscillation, Journal of the Atmospheric Sciences 10.1175/1520-0469(1978)035<1227:TIOAFS>2.0.CO;2 (1978).doi.org/10.1175/1520-0469(1978)035<1827:tioafs>2.0.co;2
- [2]T. Seelig and U. Harlander, Can zonally symmetric inertial waves drive an oscillating zonal mean flow?, Geophysical & Astrophysical Fluid Dynamics 109, 541 (2015).
- [3]B. Semin, N. Garroum, F. Pétrélis, and S. Fauve, Nonlinear saturation of the large scale flow in a laboratory model of the quasi-biennial oscillation, Physical Review Letters 121, 134502 (2018).
- [4]M. P. Baldwin et al., The quasi-biennial oscillation, Reviews of Geophysics (2001).DOI
- [5]A. P. Showman, X. Tan, and X. Zhang, Atmospheric circulation of brown dwarfs and jupiter and saturn-like planets: Zonal jets, long-term variability, and qbo-type oscillations, The Astrophysical Journal 883, 4 (2019).
- [6]J. A. Anstey, S. M. Osprey, J. Alexander, M. P. Baldwin, N. Butchart, L. Gray, Y. Kawatani, P. A. Newman, and J. H. Richter, Impact processes and projections of the quasi-biennial oscillation, Nature Reviews Earth & Environment 3, 588 (2022).DOI
- [7]E.-j. Kim and K. B. MacGregor, Gravity wave–driven flows in the solar tachocline, The Astrophysical Journal 556, L117 (2001).DOI
- [8]A. Renaud, L.-P. Nadeau, and A. Venaille, Periodicity disruption of a model quasi-biennial oscillation of equatorial winds, Physical Review Letters 122, 10.1103/PhysRevLett.122.214504 (2019).
- [9]L.-A. Couston, D. Lecoanet, B. Favier, and M. Le Bars, Order out of chaos: Slowly reversing mean flows emerge from turbulently generated internal waves, Physical Review Letters 120, 244505 (2018).DOI
- [10]F. Daniel and D. Lecoanet, A self-consistent numerical model of internal wave-induced mean flow oscillations in polar geometry, Journal of Fluid Mechanics 1028, A38 (2026).arxiv.org/abs/2509.19482
- [11]V. Brémaud, A. Podglajen, A. Hertzog, and R. Plougonven, An idealized two-dimensional atmospheric model of the quasi-biennial oscillation and wave-mean flow interactions in the equatorial channel, Authorea Preprints (2025).DOI
- [12]D. Bardet, A. Spiga, and S. Guerlet, Joint evolution of equatorial oscillation and interhemispheric circulation in saturn’s stratosphere, Nature Astronomy 6, 804 (2022).DOI
- [13]R. S. Lindzen and J. R. Holton, A theory of the quasi-biennial oscillation, Journal of the Atmospheric Sciences (1968).doi.org/10.1175/1520-0469(1968)025<1095:atotqb>2.0.co;2
- [14]T. J. Dunkerton, The role of gravity waves in the quasi-biennial oscillation, Journal of Geophysical Research (1997).DOI
- [15]B. R. Sutherland, U. Achatz, C.-c. P. Caulfield, and J. M. Klymak, Recent progress in modeling imbalance in the atmosphere and ocean, Physical Review Fluids 4, 010501 (2019).
- [16]L. A. Holt, F. Lott, R. R. Garcia, G. N. Kiladis, Y.-M. Cheng, J. A. Anstey, P. Braesicke, A. C. Bushell, N. Butchart, C. Cagnazzo, et al., An evaluation of tropical waves and wave forcing of the qbo in the qboi models, Quarterly Journal of the Royal Meteorological Society 148, 1541 (2022).
- [17]S. M. Osprey, N. Butchart, J. R. Knight, A. A. Scaife, K. F. Hamilton, J. A. Anstey, V. Schenzinger, and C. Zhang, An unexpected disruption of the atmospheric quasi-biennial oscillation, Science 353, 1424 (2016).pubmed.ncbi.nlm.nih.gov/27608666
- [18]J. R. Holton and R. S. Lindzen, An updated theory for the quasi-biennial cycle of the tropical stratosphere, Journal of the Atmospheric Sciences (1972).doi.org/10.1175/1520-0469(1972)029<1076:autftq>2.0.co;2
- [19]R. A. Plumb, The interaction of two internal waves with the mean flow: implications for the theory of the quasi-biennial oscillation, Journal of the Atmospheric Sciences (1977).doi.org/10.1175/1520-0469(1977)034<1847:tiotiw>2.0.co;2
- [20]R. Saravanan, A multiwave model of the quasi-biennial oscillation, Journal of the Atmospheric Sciences (1990).doi.org/10.1175/1520-0469(1990)047<2465:ammotq>2.0.co;2
- [21]L. J. Campbell and T. G. Shepherd, Constraints on wave drag parameterization schemes for simulating the quasi-biennial oscillation. part i: Gravity wave forcing, Journal of the atmospheric sciences 62, 4178 (2005).DOI
- [22]L. J. Campbell and T. G. Shepherd, Constraints on wave drag parameterization schemes for simulating the quasi-biennial oscillation. part ii: Combined effects of gravity waves and equatorial planetary waves, Journal of the atmospheric sciences 62, 4196 (2005).DOI
- [23]P. Léard, D. Lecoanet, and M. Le Bars, Multimodal excitation to model the quasi-biennial oscillation, Physical Review Letters 125, 10103/PhysRevLett.125.234501 (2020).
- [24]X. Chartrand, L.-P. Nadeau, and A. Venaille, Recovering quasi-biennial oscillations from chaos, Journal of the Atmospheric Sciences 81, 1213 (2024).arxiv.org/abs/2311.10907
- [25]K. Rajendran, I. M. Moroz, P. L. Read, and S. M. Osprey, Synchronisation of the equatorial qbo by the annual cycle in tropical upwelling in a warming climate, Quarterly Journal of the Royal Meteorological Society 142, 1111 (2016).DOI
- [26]A. Renaud and A. Venaille, On the holton–lindzen–plumb model for mean flow reversals in stratified fluids, Quarterly Journal of the Royal Meteorological Society 146, 2981 (2020).DOI
- [27]A. Match and S. Fueglistaler, Mean-flow damping forms the buffer zone of the quasi-biennial oscillation: 1d theory, Journal of the Atmospheric Sciences 77, 1955 (2020).DOI
- [28]V. Brémaud, A. Podglajen, A. van Niekerk, A. Hertzog, and R. Plougonven, Sensitivity of the qbo to the turbulent diffusion scheme in an idealized model, ESS Open Archive 2026 (2026).
- [29]E. Ott, C. Grebogi, and J. A. Yorke, Controlling chaos, Physical review letters 64, 1196 (1990).
- [30]Y. Braiman and I. Goldhirsch, Taming chaotic dynamics with weak periodic perturbations, Physical Review Letters 66, 2545 (1991).pubmed.ncbi.nlm.nih.gov/10043550
- [31]R. A. Plumb and R. C. Bell, A model of the quasi-biennial oscillation on an equatorial beta-plane, Quarterly Journal of the Royal Meteorological Society 108, 335 (1982).
- [32]C. Ménesguen, A. Delpech, F. Marin, S. Cravatte, R. Schopp, and Y. Morel, Observations and mechanisms for the formation of deep equatorial and tropical circulation, Earth and Space Science 6, 370 (2019).DOI
- [33]B. Galperin and P. L. Read, Zonal Jets (Cambridge university press, 2019).DOI
- [34]D. Lemasquerier, Emergence and coalescence of zonal jets: A quasilinear rossby wave-mean flow interaction model, Physical Review Fluids (2025).DOI
- [35]C. Nolan and E. Benilov, Hyper-reflection of inertia-gravity waves by vortices, Physics of Fluids 38 (2026).DOI
- [36]G. Meletti, S. Abide, S. Viazzo, J. Curbelo, and U. Harlander, Wave-like spirals and spontaneous oscillations in strato-rotational flows, Journal of Fluid Mechanics 1035, A38 (2026).DOI
- [37]M. Higuera, J. Porter, and E. Knobloch, Faraday waves, streaming flow, and relaxation oscillations in nearly circular containers, Chaos: An Interdisciplinary Journal of Nonlinear Science 18 (2008).DOI
- [38]G. Launay, T. Cambonie, D. Henry, A. Pothérat, and V. Botton, Transition to chaos in an acoustically driven cavity flow, Physical Review Fluids 4, 044401 (2019).DOI
- [39]J. R. Booker and F. P. Bretherton, The critical layer for internal gravity waves in a shear flow, Journal of Fluid Mechanics 10.1017/S0022112067000515 (1967).DOI
- [40]G. H. Wallis, Atmospheric and oceanic fluid dynamics (Cambridge University Press, 2017).
- [41]K. Hamilton, The vertical structure of the quasi-biennial oscillation: Observations and theory, Atmosphere-Ocean 19, 236 (1981).DOI





