1. Introduction
Inertia–gravity waves (GWs) play an important role in the global circulation and transport by redistributing momentum and energy in the atmosphere (Alexander et al. 2010). However, because of their small scale, the actions of GWs are usually incorporated in climate models using parameterization schemes. The parameterization schemes include various assumptions because the mechanism of GW generation and the GW propagation properties are not very clear, and computational resources are limited. In particular, the dynamics of nonorographic GW generation is nonlinear and quite complicated, unlike that of orographic GW generation, and this leads to a lot of uncertainties in the parameterization schemes (e.g., Hines 1997; Alexander and Dunkerton 1999; Warner and McIntyre 2001). Thus, gaining a better understanding of the nonorographic GW dynamics is important not only for the geophysical fluid dynamics but also for the improvement of parameterization schemes.
Recent studies have clarified that GWs are spontaneously radiated from an approximately balanced flow as the flow evolves (e.g., O’Sullivan and Dunkerton 1995; Vanneste 2013; Plougonven and Zhang 2014). The amount of spontaneously radiated GWs is considered to be comparable to that of orographic GWs (e.g., Plougonven et al. 2013). Many theoretical studies on spontaneous radiation, including the present study, are partly based on balanced-flow dynamics. Thus, we first briefly review balanced-flow dynamics.
The time evolution of a balanced flow is governed by a balanced model (BM). The BM is a system of equations that consists of a time-evolution equation for a slow variable and diagnostic formulas for fast variables (e.g., Leith 1980; Lorenz 1980; Warn et al. 1995; McIntyre and Norton 2000; Wirosoetisno et al. 2002; Saujani and Shepherd 2006; Mohebalhojeh and McIntyre 2007a,b). Here, a slow variable (e.g., potential vorticity) is a dependent variable that is zero for the GW solution but nonzero for the geostrophic solution in the linear system. In contrast, fast variables (e.g., horizontal divergence and ageostrophic vorticity) are dependent variables that are zero for the geostrophic solution but nonzero for the GW solution in the linear system. In BMs, fast variables are diagnostically obtained from the slow variable, which are called slaved components. BMs are generally obtained by imposing diagnostic relations between physical quantities such as geostrophic and hydrostatic balances on the primitive equations. The most familiar BM may be the quasigeostrophic (QG) model, in which the slow variable is QG potential vorticity (QGPV), while the slaved component is ageostrophic velocity. The QGPV is obtained from the time-evolution equation, while the ageostrophic velocity is diagnostically determined by the omega and continuity equations with a given QGPV field. GWs are completely filtered out in a BM; hence, the BM has a lower degree of freedom than the primitive equations. In other words, a BM describes the evolution on the slow manifold, which is a hypersurface in the phase space defined by the balanced relations (e.g., Vanneste 2013). In balanced-flow dynamics, the balanced flow should be distinguished from the vortical flow, because the vortical flow is a component composing the PV field, while the balanced flow (i.e., the solution of the BM) is the sum of the vortical flow and slaved components.
Theoretical studies on spontaneous GW radiation can be classified into two groups:
The initial state does not contain any GWs, and then GWs are radiated as the flow evolves.
The initial state contains GWs with quite small amplitudes, and then the GWs are amplified.
Theories classified into the second group include instabilities that occur as an interaction between balanced and imbalanced modes (e.g., Plougonven et al. 2005) and transient amplifications by a superposition of neutral modes composed of balanced and imbalanced parts (Lott et al. 2010, 2012). The classification of Lott et al. (2010, 2012) may be ambiguous, because they demonstrated that it was impossible to rigorously separate a balanced part from an imbalanced part in a constant vertical shear field. This issue will be further discussed in section 4b. In the second-group theories, GWs are not radiated if the initial state is exactly balanced.
It should be noted here that another classification is possible from a different viewpoint, as made in the review paper (Plougonven and Zhang 2014): the theories by Lott et al. (2010) and Vanneste and Yavneh (2004) are classified into the same group named, “transient generation by sheared disturbances,” while they are classified into different groups in the present paper.
The present study is classified into the first group (group 1). Roughly speaking, theories in the first group are further classified into three subgroups: Lighthill radiation, transient generation, and balance adjustment.
The Lighthill radiation theory is based on the theory of vortex sound and describes GW generation in shallow-water systems quite well (e.g., Ford et al. 2000; Sugimoto et al. 2007, 2008). The transient generation describes the generation of GWs having high ground-based frequencies in a three-dimensional system for the small Rossby number limit (Vanneste and Yavneh 2004; Ólafsdóttir et al. 2008). In contrast, the present study treats significantly Doppler-shifted GWs, whose time scales are comparable to that of the balanced flow, in a three-dimensional system for a finite but sufficiently small Rossby number. Thus, our study should be distinguished from the first and second subgroups of the first-group studies.
Balance adjustment, the third subgroup of the first group, is the mechanism for GW radiation from imbalanced sources that are produced continuously as an approximately balanced large-scale flow evolves in time (Zhang 2004). This imbalance is defined as the difference between the real flow and the balanced one. Note that the balance adjustment is not an initial-value problem, such as the classical geostrophic adjustment (or Rossby adjustment) problem (e.g., Holton 2004). Plougonven and Zhang (2007) derived a system of linear equations describing balance adjustment on the basis of a scale analysis. In their theory, residuals of the balanced-model solution are called imbalances, and the Doppler-shifted GWs, having low ground-based frequencies, are radiated from these residuals. There have been other linear theories (e.g., Snyder et al. 2009) quite similar to that of Plougonven and Zhang (2007). In the present paper, linear equations that describe the radiation of Doppler-shifted GWs from residuals of the balanced-model solution are referred to as balance adjustment equations (BAEs). The BAEs have successfully reproduced the distribution of spontaneously radiated GWs in the vortex dipole with high accuracy (Snyder et al. 2009; Wang and Zhang 2010).
However, there are a few important issues that are not elucidated quite well by the BAEs. The residuals of the balanced-model solution, which are called imbalances in the BAEs, as mentioned above, depend on which BM is used to construct the BAEs. Thus, radiated GWs may significantly depend on the choice of BMs. In addition, the physical mechanism for the continuous production of the imbalance is not clear. In the BAEs, the imbalances always exist when the balanced-model solution is not an exact solution for the primitive equations. In other words, it is physically unclear why the balanced-model solution is not an exact solution. Last but not least, the large-scale flow should be continuously modified by the continuous GW radiation, as discussed by McIntyre (2009). This “adjustment” process is, however, not described in the BAEs, because they are linear equations describing only the GW generation.
In the present study, we derive a new theory that is similar to BAEs, but we propose a new mechanism. The most important point in our theory is that Doppler-shifted GWs are radiated through a quasi resonance with a part of the balanced flow. More specifically, GW sources are the slaved components of the balanced flow, and GWs are radiated when space–time scales of the GWs are comparable with those of the sources. Our theory treats balance adjustment as interactions between GWs, slaved components, and vortical flow in a coexistence system, as illustrated in Fig. 1. As discussed in section 4a, our theoretical equation can be formally reduced to a BAE and will clarify that BAEs actually describe the GW radiation from the slaved components. This point is discussed briefly here because it is the motivation for the present study.
Conceptual diagram of RGE theory. The vortical flow is a component composing PV, the slaved components are part of the fast variables, which are diagnostically determined by the PV distribution, and the balanced flow is the sum of the vortical flow and slaved components. GWs that are significantly Doppler shifted as a result of the vortical flow are radiated from the slaved components through the quasi resonance, while the vortical flow and slaved components (GW sources) are affected by the GW radiation reactions. In RGE theory, the large-scale-flow system is considered to be a coexistence system composed of GWs, slaved components, and vortical flow that interact with each other. These components should not be considered separately.
Citation: Journal of the Atmospheric Sciences 72, 3; 10.1175/JAS-D-13-0370.1
We have developed a new theory by focusing on two important facts elucidated in previous studies. The first fact is the existence of a descent–ascent couplet on a local jet axis in a vortex dipole whose characteristic spatial scale is larger than those of GWs (see Fig. 2). Viúdez (2007) suggested that this couplet was the initial GW packet originating from the material rate of change of the ageostrophic differential vorticity. However, this suggestion does not seem consistent with the fact that this couplet exists even in the dipole in which Ro is so low that GW radiation is hardly recognized (Snyder et al. 2007). This fact means that this couplet may be attributable to high-order slaved components. We inferred from these two studies that the origin of GW sources is high-order slaved components that a local jet accompanies.
Schematic illustration of spontaneous GW radiation from a localized jet stream caused by the mountain-wave-like mechanism proposed by McIntyre (2009). The blue arrows show the vertical flow composed of the slaved components. Doppler-shifted GWs are radiated because of the quasi resonance with the slaved components (i.e., the vertical flow couplets) when the space–time scales of the GWs are partially comparable to those of the slaved components. The wavelengths of the radiated GWs decrease through the wave-capture process. As described in Part II, these vertical flow couplets are a compensation flow for the strong localized jet and are produced by the vortical flow over the deformed potential temperature (θ) surfaces, which are also the slaved components. These potential temperature surfaces act as mountains, such as ones that generate orographic GWs. The RGE theory can describe two different mechanisms: the mountain-wave-like mechanism (this figure) and the velocity-variation mechanism (Fig. 3).
Citation: Journal of the Atmospheric Sciences 72, 3; 10.1175/JAS-D-13-0370.1
The second fact is a significant Doppler shift by the vortical flow (i.e., the main part of the balanced flow). The intrinsic frequencies of GWs are usually much higher than those of the balanced flow. However, when GWs are significantly Doppler shifted by the strong vortical flow, their ground-based frequencies may be comparable to those of the balanced flow. Note that the Doppler shift is also essential for the BAEs (e.g., Plougonven and Zhang 2007).
Taking the above two facts into account, we have reached a new mechanism: GWs are radiated from slaved components (i.e., part of the balanced flow) through a quasi resonance when the space–time scales of the GWs are partially comparable to those of the slaved components. Note that the term “quasi resonance” is used here because the frequencies of radiated GWs are not always identical to those of the slaved components. This new mechanism explains the continuous radiation of GWs because the GW sources are the slaved components, which are part of the balanced flow; hence, they are always present in the large-scale flow field. More importantly, this theoretical framework can address more than one mechanism regarding the spontaneous radiation, because there may be various physical mechanisms that produce slaved components. In other words, a physical mechanism of the GW radiation depends on how the GW source (slaved component) is produced. In fact, as discussed in Yasuda et al. (2015, hereafter Part II), the GW radiation mechanisms in vortex dipoles are classified into two types: one is the mountain-wave-like mechanism (Fig. 2) discussed by McIntyre (2009), and the other is the velocity-variation mechanism (Fig. 3) proposed by Viúdez (2007). These two mechanisms are described in the theoretical framework proposed in this study, even though they look very different on first sight. A new system of theoretical equations describing these mechanisms is constructed by using the renormalization group (RG) method (Chen et al. 1994, 1996). The outline for the derivation is described in the following.
Schematic illustration of spontaneous GW radiation from a localized jet stream caused by the velocity-variation mechanism proposed by Viúdez (2007). Two short red arrows show the vortical flow (localized jet shown as a long red arrow) acceleration and deceleration. The blue ellipse shows the region where the vortical flow acceleration has horizontal convergence (i.e., negative divergence). This horizontal convergence is obtained as the slaved components. Doppler-shifted GWs are radiated as a result of the quasi resonance with the slaved components (i.e., the acceleration convergence), when the space–time scales of the GWs are partially comparable to those of the slaved components. The wavelengths of the radiated GWs decrease through the wave-capture process. The RGE theory can describe two different mechanisms: the mountain-wave-like mechanism (Fig. 2), and the velocity-variation mechanism (this figure).
Citation: Journal of the Atmospheric Sciences 72, 3; 10.1175/JAS-D-13-0370.1
The RG method (Chen et al. 1994, 1996), which is a singular perturbation method, makes it possible to systematically and naturally extract time-evolution equations on slow time scales from a system containing multiple time-scale motions. Thus, this method is relevant to obtain the theoretical equations describing the interaction including quasi resonance between the vortical flow and Doppler-shifted GWs, which both have slow time scales. In the present paper, the RG method is applied to the hydrostatic Boussinesq equations on the f plane (herein referred to as primitive equations).
We use linear potential vorticity q as a slow variable and horizontal divergence δ and ageostrophic vorticity γ as fast ones. The linear theory indicates that δ and γ have high-frequency components, including (non-Doppler-shifted) GWs, but some of their nonlinear terms vary slowly in time because they consist only of q. These slowly varying nonlinear terms are mainly due to slaved components of balanced flow and should be distinguished from radiated GWs but should be included in the theory for the slow-time-scale motions. Thus, five dependent variables are introduced separately: Doppler-shifted fast variables (δGW and γGW), diagnostic fast variables (δdiag and γdiag), and a slow variable (q). Doppler-shifted fast variables correspond to the spontaneously radiated GWs, while the diagnostic ones correspond to the slowly varying nonlinear terms. Note again that all five variables should have slow time scales. The GW components (δGW and γGW) that include the Doppler effect by the vortical flow are obtained as the eigenmodes of the fast variables in a given vortical flow field after the renormalization is made. On the contrary, δdiag and γdiag are directly and diagnostically obtained from renormalized q, δGW, and γGW, and they correspond to the sum of the slaved components and the GW radiation reactions.
Our theory clarifies that the main GW source is the slaved component that is a part of δdiag and γdiag. It is worth noting that the derived equations describe the tendencies of q and GW sources, including the GW radiation reactions. In addition, this new theoretical system proposes a new adjustment process (i.e., resonance adjustment) that is different from the balance adjustment process (Zhang 2004), as discussed in section 4b. In the new process, GWs are radiated continuously because the GW sources are always present, and the large-scale flow becomes weak as GWs are radiated. When the flow is too weak to cause the time-scale matching between the Doppler-shifted GWs and slaved components, the GW radiation is ceased, and the weak large-scale flow without GWs is left. In this paper (Part I), we describe the formulation of the theoretical equations.
The structure of this paper is as follows. In section 2, the primitive equations are transformed before applying the RG method. In section 3, by applying RG method to the transformed primitive equations, the time-evolution equations on slow time scales are derived. These equations are referred to as renormalization group equations (RGEs). In section 4, the relation between the RGEs and BAEs is discussed. Advantages and disadvantages of the RGE system are described in comparison with the BAEs. A discussion is also made on how the wave capture is described in the RGE system. A summary and the concluding remarks are given in section 5. In appendix A, the details of the RG method are presented by using an example of an ordinary differential equation.
One possible way to confirm the validity of this new theory is to compare the predictions from this theory with the results from numerical simulations of a compressible nonhydrostatic equation model. For this purpose, quasi-steady spontaneous GW radiation in a vortex dipole is addressed in Part II. Details of the physics on the GW sources and processes of the spontaneous radiation, as in Figs. 2 and 3, are also examined in Part II.
2. Transformation of dependent variables and wavenumber representation of primitive equations
The representations of the primitive equations [i.e., hydrostatic Boussinesq equations on the f plane (e.g., Vallis 2006)] are transformed before applying the RG method. After dependent variables, such as velocity, temperature, and pressure, are converted to fast and slow variables, all equations are represented in the wavenumber space. No approximations, such as geostrophic balance approximation, are needed for the manipulations in this section.
When Ro is so small as to enable all nonlinear terms on the RHSs of (24)–(26) to be ignored, qk is constant, while γk and δk oscillate with the frequencies of ±ωk. This oscillation corresponds to a plane GW in the motionless atmosphere. These GWs are not Doppler shifted, because all nonlinear terms, including the Doppler shift for GWs, are neglected. The above discussion suggests that qk varies on the slow time scale characterized by s(≡Rot), while γk and δk vary on the fast time scale characterized by t, when Ro is small but the nonlinear terms in (24)–(26) cannot be ignored. The slow time scale is separated to a greater degree from the fast time scale for smaller Ro, indicating that high-frequency GWs are hardly excited by the vortical flow. Another important point is that the first terms on the RHSs of (25) and (26) are nonlinear terms composed only of qk. Thus, the fast variables contain not only high-frequency components but also low-frequency ones as a result of the nonlinear interactions. Roughly speaking, these slowly varying nonlinear terms are slaved components to the vortical flow and should be distinguished from GWs.
3. Application of RG method to the primitive equations
In this section, by applying the RG method to the transformed primitive equations [(24)–(26)], the time-evolution equations (RGEs) are derived. In section 3a, the assumptions used in this paper are shown and discussed. The RGEs are derived in section 3b, and their characteristics are discussed in section 3c. Separability between the diagnostic components and GW components in this RGE theory is discussed in section 3d.
The RGE system derived in this section describes the new mechanism: GWs are radiated from slaved components through a quasi resonance when the space–time scales of the GWs are partially comparable to those of the slaved components. Spontaneous GW radiation tends to occur at a later time, even when the initial state is well balanced, because GW sources (i.e., slaved components) are always present. Thus, the RGE theory indicates that the primitive equations do not have an invariant slow manifold. In the RGE theory, a flow system is considered to be a coexistence system composed of the GWs, slaved components, and vortical flow that interact with each other (Fig. 1).
a. Assumptions and their validity
The following six assumptions were made when applying the RG method:
The initial state is balanced. The GWs excited are so significantly Doppler shifted by the vortical flow as to have low ground-based frequencies (i.e., slow time scales). The excitation of GWs having high ground-based frequencies that are not significantly Doppler shifted is negligible.
To take the significant Doppler shift into account, an eigenmode expansion of the fast variables in a given vortical flow is carried out after applying the RG method. Eigenmodes with eigenvalues (i.e., ground-based frequencies) that are O(Ro) are referred to as GW modes. A superposition of the GW modes is considered as the spontaneously radiated GWs. When the GW modes are represented by the plane-wave bases, they are referred to as GW components (
The amplitude of the GW component is nondimensionalized on the basis of the characteristic space–time scales of the large-scale flow, which is characterized by the balanced flow.
The characteristic length of the large-scale flow is O(1), while that of the GWs is O(Ro). This assumption is represented in terms of wavenumbers;
The interaction coefficients between the vortical flow and GW component and those between the diagnostic component and GW component are expanded with respect to Ro.
The interaction coefficients of O(Ro−1) between the diagnostic component and GW component are considered only in the time-evolution equations for the GW components (or modes).
The validity of the above six assumptions is discussed in the following. Wirosoetisno et al. (2002) applied the third-order RG method to the primitive equations, but they did not include the Doppler effect on GWs. They showed that the flow field governed by the RGEs was always balanced when the initial state was balanced. Note that the primitive equations addressed in their paper were abstract equations having some properties that the hydrostatic Boussinesq equations have. The time scales of the vortical flow and non-Doppler-shifted GWs are completely separated in the regular perturbation solution. This fact means that the quasi resonance does not occur in their RGEs, and (non-Doppler-shifted) GWs are not excited in the balanced-flow field. Thus, assumption (i), which eliminates high-frequency GWs, is plausible.3
In general, the fast variables have slowly varying nonlinear terms that are distinguished from GWs such as the first terms on the RHSs of (25) and (26). Roughly speaking, these terms are regarded as slaved components. To include the slowly varying nonlinear terms separately from GWs in the theory, the RG method is applied to δ and γ in the same way as Wirosoetisno et al. (2002). Under assumptions (i)–(vi), the distributions of these components are determined diagnostically from qk, 
On the other hand, the GW modes are obtained by the eigenmode expansion in a given vortical flow after applying the RG method. The characteristic frequency scale of the GW modes should be O(Ro) so that the derived RGEs describe the excitation of GWs through the quasi resonance with the balanced flow with a characteristic frequency scale of O(Ro). Thus, the GW modes are defined as in assumption (ii). This definition has uncertainty, because the range of eigenvalues (i.e., ground-based frequencies of GW modes) is not specified. This uncertainty is discussed in Part II.
Next, we discuss the characteristic amplitude and spatial scales of the GWs to show that assumptions (iii) and (iv) are reasonable. As mentioned earlier, we consider that the GWs are radiated through the quasi resonance and the main GW sources are the slaved components composed only of q. This means that the characteristic magnitudes of the slaved and GW components should be of the same order. In other words, δGW and γGW should be nondimensionalized on the basis of the characteristic scales of the large-scale flow, like the slaved components are.
Previous studies using numerical model simulations have indicated that the characteristic spatial scales of the GWs generated from a jet are much smaller than that of the local jet itself (O’Sullivan and Dunkerton 1995; Zhang 2004; Plougonven and Snyder 2007). This result is consistent with the scale analysis on the dispersion relation of spontaneously radiated GWs (Snyder et al. 2009). They showed that the characteristic spatial scale of the GWs should be O(Ro) in the large-scale flow with the characteristic spatial scale of O(1) so that their time scales were comparable by the strong Doppler effects. Note again that this time-scale matching between the large-scale flow and GWs is necessary for the quasi resonance. These are the reasons for assumptions (iii) and (iv).
Assumptions (v) and (vi) are related to (iv). All interaction coefficients in the nondimensional primitive equations [(24)–(26)] are zeroth-order homogeneous with respect to wavenumbers (see appendix B for details). Thus, the magnitude of an interaction coefficient describing the local interaction in the wavenumber space is O(1), because all wavenumbers in the coefficient have the same order. On the other hand, for the nonlocal interactions, the magnitude of an interaction coefficient can be greater than O(1), because the magnitude of the numerator of the coefficient may be much larger than that of the denominator. In particular, the interaction between the vortical flow and GWs is considered to be nonlocal, because their characteristic wavelengths are largely different. Actually, it is evident from appendix B that these interaction coefficients are O(Ro−1), meaning that these coefficients must be expanded with respect to Ro in order to correctly obtain the regular perturbation solution (procedure A of the RG method in appendix A). In addition, the diagnostic components include the slaved components; hence, the low-wavenumber diagnostic components are not negligible. This fact suggests that the expansion is also necessary for the interaction coefficients between the diagnostic component and GW component.
Similar to (28), coefficients (30) and (31) represent the advection of the GW component by the velocity associated with δdiag and γdiag. Higher-order interaction coefficients between the diagnostic component and GW component do not include such an advection effect but represent shear effects of the velocity associated with the diagnostic components.
It should be emphasized again that all expanded interaction coefficients of O(Ro−1) represent the advection of the GW component. The interaction coefficients of O(Ro−1) with “superscripts” qk are zero in (27) and (29), because the advection of GWs is not included in the time-evolution equation for qk. As this case, it is natural that the advection of GWs is included only in the time-evolution equation for the GW component. This also means that the interaction coefficients of O(Ro−1) should not be included in the equations for the diagnostic component, as in assumption (vi).
b. Derivation of the renormalization group equations
In this subsection, by taking Ro as a small number, the RG method is applied to the transformed primitive equations [(24)–(26)]. First, all dependent variables (i.e., 
First, the zeroth-order regular perturbation solutions are examined. For the vortical flow, 
One may think that the eigenmode representation should be introduced here because the matrices 
For the diagnostic components, the time derivatives in (25) and (26) are ignored because of assumption (i), and the terms whose interaction coefficients are O(Ro−1) are also ignored as a result of assumption (vi). Thus, 
Similar to 
Next, the secular terms are renormalized using the RG method (Chen et al. 1994, 1996). All t in the secular terms are transformed into t = (t − τ) + τ (procedure B of the RG method). To cancel the divergence due to τ at each order of Ro, renormalization constants and renormalized variables are introduced (procedure C of the RG method). For example, by introducing the renormalization constant 
In this way, we have finally obtained the system of first-order RGEs that is composed of the time-evolution equations for the linear potential vorticity [(46)] and the GW mode [(50)], and the formulas giving the diagnostic components [(52)]. Note that all variables vary in slow time s (≡Rot). See appendixes C and D for the derivations of the second-order regular perturbation solutions and RGEs [(D1), (D3), and (D5)], respectively. All slaved components in the second-order balanced model are shown in appendix E.
c. Characteristics of the RGE system
In this theory, it is considered that the quasi resonance could occur and generate the GWs from the slaved components, because the GW modes have the same-order ground-based frequencies as the balanced flow. The exact frequency matching between the GW modes and their sources is not necessary, which is different from a pure resonance. This is because 
It is also important that the nonlinear terms containing the GW components can be regarded as the GW radiation reactions in (35), (38), (39), (42), (43), (46), and (52). The existence of these terms means that the variations in the vortical flow, the diagnostic components, and the GW sources are affected by the spontaneously radiated GWs (i.e., the reaction of the GW radiation). The GW sources also include an implicit reaction through the slaved components, which are varied by the reaction on q. In particular, the reactions on q [i.e., 
It is necessary to confirm the validity of the RGE theory, because we made several assumptions to construct the theory. Here, it is verified on the basis of the forms of 
In the linear theory, GWs do not have potential vorticity (PV), which indicates that a PV field is generally hardly affected by GWs in the real atmosphere. However, a finescale spatial structure of PV, which is likely due to the GWs, has been recognized slightly above the tropopause at midlatitudes in the general circulation model in which GWs are explicitly resolved (Miyazaki et al. 2010). The RGE theory likely describes such fine PV structures as the reactions of GW radiation. This also suggests that it may sometimes be hard to extract and/or remove GWs from observation and/or numerical model data.
Finally, two remarks need to be made regarding the application of the RGEs to numerical calculations. First, it is necessary to change the basis functions from the eigenmodes to the plane waves at each step when integrating the RGEs. This is because most processes, such as the GW radiation reactions on 
d. Separability between diagnostic and GW components on RGE theory
A major difference in our treatment from previous studies is that the Doppler shift has been taken into account for the GW components (or modes). The GW modes having low ground-based frequencies are mainly composed of high wavenumbers so as to be affected by a significant Doppler shift. In contrast, the diagnostic components are mainly composed of the slaved components having low wavenumbers. However, the spatial scales of GWs need to be partially comparable to those of diagnostic (or slaved) components in order that a quasi resonance could occur, as discussed in the previous subsection. In other words, projections of the GW sources on some GW eigenmodes need to be present for the quasi resonance [i.e., 
The diagnostic components are mainly composed of the slaved components that are proportional to the products of 
Next, we discuss changes in the RGE theory due to the inseparability. The separability may not be certain when “characteristic” spatial scales of diagnostic components (determined by the wavenumbers at the peaks of energy density spectra) are comparable to those of GW components. It is considered that the first-order diagnostic components are separable from the GW components more clearly than the second-order diagnostic ones, because the first-order ones consist of lower-order nonlinear interactions (or terms).
In the first-order RGE system, only the first-order diagnostic components are present. This fact indicates that the separability between diagnostic and GW components is generally valid, compared with the separability in the second-order RGE system, in which the second-order diagnostic components as well as first-order ones are present. When the first-order diagnostic components are not separable from the GW ones in the first-order RGE system, the first-order diagnostic ones and slaved ones cannot be defined. This means that all terms including the first-order diagnostic components are taken to be zero. In addition, the main GW sources in (38) and (39) cannot be replaced with the first-order slaved components; for example, the first term in (38) remains 
In the second-order RGE system, the first- and second-order diagnostic components are present. The separability between the second-order diagnostic and GW components may not be certain, compared with that between the first-order diagnostic and GW components. Thus, the former is discussed further. When the second-order diagnostic components are not separable from the GW ones in the second-order RGE system, the second-order diagnostic ones, as well as the second-order slaved ones, cannot be defined. This means that all terms including the second-order diagnostic components are taken to be zero. In addition, some main GW sources in (D5) cannot be replaced with the second-order slaved components: for example, the first term in the first row of (D8), which is the main part of the GW source 
In this way, only the interpretation on the GW radiation may be changed depending on the separability, while it is considered that the first- and second-order RGE systems can be derived regardless of the separability. In the present paper and in Part II, we consider that the first- and second-order diagnostic components are separable from the GW components, as discussed above. One of our future tasks is to examine in detail the degree of separability between diagnostic (or slaved) components and GW components.
The higher-order diagnostic components (higher than second order) may include significant high-wavenumber components because of multiple nonlinear interactions (or high-order nonlinear terms). This fact suggests a possible inseparability between diagnostic and GW components; hence, higher-order RGEs (higher than second order) may not be able to be derived in the same way as used for the first- and second-order RGEs.
4. Discussion
In this section, the relations between our new theory and the theories proposed in previous studies are discussed. In section 4a, it is shown that the RGE can be formally reduced to a BAE (e.g., Plougonven and Zhang 2007). The advantages and disadvantages of the RGE system are compared with those of BAEs. In section 4b, the adjustment process described by the RGE system is compared with the balance adjustment process (Zhang 2004). Section 4c features a discussion on how the wave capture (Bühler and McIntyre 2005) is described in the RGE system.
a. Relation to balance adjustment equations
BAEs are linear equations that describe the GW radiation from the residuals of the balanced-model solution. In contrast, RGEs are nonlinear equations that describe time evolutions of both the large-scale flow (q, γdiag, and δdiag) and the GWs (γGW and δGW). The nonlinear terms in the RGE for the GWs are regarded as the GW radiation reactions on the GW sources, as discussed in section 3c. The RGE is formally reduced to a BAE if these nonlinear terms in the GW sources are neglected. To demonstrate this, the second-order RGE is compared with the BAE derived from the first-order BM. The first-order BM is obtained by eliminating all O(Ro2) terms in the second-order BM (see appendix E). The first-order BM derived using the RG method corresponds to the QG model (Wirosoetisno et al. 2002).
It should be emphasized here that all second-order slaved components in (58) are seemingly identical to those in the second-order BM, but they are not exactly identical. The symbol 
In summary, the linearized RGE (54) and the BAE (58) are linear equations that describe the GW radiation from the sources mainly composed of the slaved components. In particular, as shown in Part II, two main GW sources in the linearized RGE (i.e., 
However, this fact does not mean that the only difference between BAEs and our theory is in the interpretation on GW sources. The most essential difference is in the proposed mechanism of GW radiation. In the BAE theory, it is considered that GWs are radiated from the residuals of the balanced-model solution that are called “imbalances.” In contrast, in our theory, GWs are considered to be radiated through the quasi resonance with the slaved components of the balanced flow. Note again that the slaved components are a part of the balanced flow. In other words, the RGE theory physically clarifies that imbalances (i.e., GWs) can be produced in a balanced-flow field because of the time-scale matching (or, equivalently, the quasi resonance). To treat a quasi resonance correctly, the RGE theory requires the eigendecomposition of 
In addition, the original RGE system can describe the GW radiation reaction on the large-scale flow, unlike the BAEs. However, there are a few disadvantages in the original and linearized RGEs. The most critical one is that it may be difficult to apply the RGEs to high-resolution observation and/or numerical model data, unlike with the BAEs. There are two reasons for this. First, the Fourier expansion is used to construct the RGE theory, which suggests that it may be impossible for the RGEs to describe GW radiation near the ground, such as in a surface dipole (Snyder et al. 2007). Second, the eigendecomposition of 
Another disadvantage is that the RGEs cannot describe the GW radiation from the true imbalances, because imbalances are not originally included in the RGE system. In addition, the RGEs cannot describe the radiation of GWs having high ground-based frequencies. Some studies have shown that significant GWs including high-frequency ones are radiated during the initial adjustment (e.g., Snyder et al. 2007, 2009). This initial adjustment (i.e., three-dimensional Rossby adjustment) cannot be addressed using the RGE theory. The BAEs may have the potential to describe the GW radiation from the true imbalances and to address such initial adjustments, although a detailed investigation is needed.
b. Adjustment process
Our RGE theory proposes a new adjustment process associated with spontaneous GW radiation, which is very different from the balance adjustment theory (Zhang 2004). We call this new adjustment process “resonance adjustment.” In the balance adjustment process, it is considered that imbalances are continuously produced by the time evolution of an approximately balanced large-scale flow. GWs are radiated from these imbalances. This cycle is continuously repeated until the large-scale flow becomes so weak that the GW radiation is hardly recognized.
To explain the new adjustment process using our theory (i.e., resonance adjustment), let us consider that an approximately balanced large-scale flow is sufficiently strong at first. GW sources are always present in the large-scale flow field, because the GW sources are slaved components to the vortical flow; namely, they are part of the balanced flow. Note that the slaved components are diagnostically obtained by the PV distribution. The strong large-scale flow makes GWs significantly Doppler shifted, leading to the time-scale matching between slaved components and Doppler-shifted GWs that is needed for the quasi resonance. This quasi resonance is the excitation mechanism of GWs. The large-scale flow becomes weak by the GW radiation, because the GWs take the energy away from the balanced flow. Subsequently, the Doppler shift becomes insignificant, and the quasi resonance rarely occurs. On the whole, the large-scale flow weakens and the GW radiation is ceased. This scenario can answer the question as to why the spontaneous GW radiation is sometimes continuous in time, as in a vortex dipole (Snyder et al. 2007; Viúdez 2007), and it also explains why the balanced-model solution is usually not an exact solution for the primitive equations.
The time-scale matching was also addressed by Lott et al. (2010, 2012). They proposed that spontaneous radiation can be considered as transient amplifications by a superposition of neutral modes composed of balanced and imbalanced parts by analyzing the linear primitive equations with a constant vertical shear flow in an infinite domain. In their papers, the Doppler effect is infinitely stronger as the altitude gets higher; hence, the balanced wave solution is considered to be smoothly connected with the imbalanced wave (i.e., GW) solution at higher altitudes. However, in the real atmosphere, a region with a strong basic flow is confined rather than extended infinitely in space. Thus, time-scale matching should be realized locally, not globally, in space. In the present paper, this local time-scale matching has been achieved using the eigenmodes of the fast variables in a given vortical flow, which are composed of some plane-wave components, such as wave packets, but are not identical to the one plane-wave solution.
c. Expression of wave capture
Spontaneously radiated GWs may be captured in a large-scale flow through the process of “wave capture” (Bühler and McIntyre 2005; Plougonven and Snyder 2005; Wang et al. 2010). Wave capture is important for wave–mean flow interactions and wave propagation into the middle atmosphere. GWs in a large-scale flow having significant deformation are contracted by the advection, meaning that the GW group velocities asymptotically approach the large-scale flow velocity, and finally the GWs are captured in the large-scale flow. The BAEs can describe the wave-capture process (Snyder et al. 2009; Wang and Zhang 2010; Wang et al. 2010). As described earlier, the RGE can be formally reduced to the BAE, indicating that the RGE can also describe the wave-capture process, although the process is not explicitly described in the RGE system.
In the RGE system, a GW packet is regarded as a superposition of eigenmodes determined by the large-scale vortical flow. This means that the advection effect on the GW packet is contained in the eigenmodes and eigenvalues. The initial amplitudes of the eigenmodes composing the GW packet are determined by the GW sources, and each eigenmode evolves in time according to its eigenvalue (i.e., ground-based frequency). The above description makes it clear that the structure and evolution of GW packets depend not only on large-scale flow but also on the initial GW sources, although the wave capture is implicitly described in the RGE system. Bühler and McIntyre (2005) showed that GWs generally break near the edge of a capture region as a result of eddy (or turbulent) diffusion. In Part II, by integrating the RGE with diffusion, it will be shown that the RGE can describe the wave capture and that the structure of GW packets in the dipole is affected by the GW sources as well as the background flow.
5. Summary and concluding remarks
We have developed a new theory on spontaneous GW radiation that deals with the coexistence system composed of GWs, slaved components, and vortical flow that interact with each other (Fig. 1). The essence of this theory is that GW radiation is due to a quasi resonance between the largely Doppler-shifted GWs and the slaved components (i.e., part of the balanced flow) that occurs when the space–time scales of the GWs are partially comparable to those of the slaved components.
The theoretical equations were derived from the hydrostatic Boussinesq equations on the f plane using the renormalization group (RG) method (Chen et al. 1994, 1996), which is a singular perturbation method. In general, the RG method makes it possible to extract the slowly varying components systematically and naturally from a system containing multiple time-scale motions. The derived equations, called renormalization group equations (RGEs), describe not only the spontaneous GW radiation from the slaved components through the quasi resonance but also the GW radiation reactions on the balanced flow and on the GW sources.
It was shown that the RGE for the time evolution of GWs can be formally reduced to a balance adjustment equation (BAE) (e.g., Plougonven and Zhang 2007) if the GW radiation reactions on the GW sources are neglected. This fact suggests that the RGE can describe the wave-capture process like the BAE can, although the process is not explicitly described. The RGE theory clarified two important facts on the physical nature inherent in the BAEs. First, the GW sources, which were called imbalances in the BAEs, are composed mainly of the slaved components (i.e., a part of the balanced flow). Second, the GW radiation mechanism is the quasi resonance with the slaved components. Thus, it is not generally appropriate to call all GW sources in the BAEs (i.e., residuals of the balanced-model solution) imbalances, because imbalances are generally defined as the difference between the real flow and the balanced one. On the other hand, there are a few disadvantages in the RGE theory. The most critical one is that the RGEs are hard to use for analyses of observation and/or numerical model data.
The RGE theory provides a new scenario of the adjustment process (i.e., resonance adjustment). GWs are radiated continuously through the quasi resonance, because the GW sources (i.e., slaved components) are always present, and then the large-scale flow weakens because of the GW radiation. When the flow becomes too weak to make GWs significantly Doppler shifted, the time scales of GWs cannot be matched with those of slaved components. Eventually, quasi resonance rarely occurs, and the GW radiation is ceased.
In Part II, the validity of the RGE theory derived in this paper will be confirmed using numerical simulations. Moreover, the physical interpretations on the GW sources, as in Figs. 2 and 3, are discussed in detail.
There is an interesting issue regarding the GW radiation reaction. As discussed in section 3c, possibly the most important reactions are represented by 
Our theory is in contrast to the theories proposed by Vanneste and Yavneh (2004) and Ólafsdóttir et al. (2008), in which the exponential asymptotics has been used for the small Ro (Rossby number) limit. According to their theories, the amplitude of high-ground-based-frequency GWs excited from a balanced flow is proportional to exp(−α/Ro), where α is a constant real number. On the contrary, we showed that low-ground-based-frequency GWs could be excited whose amplitudes are proportional to the power of Ro. The difference in the GW frequencies is attributed to the degree of Doppler shift. Note that although Vanneste and Yavneh (2004) and Ólafsdóttir et al. (2008) considered the Doppler shift, the Doppler shift did not affect the GW radiation, because both GWs and balanced flow were Doppler shifted in the same way. These results suggest that the GW amplitude with respect to Ro depends on the degree of Doppler shift. In Part II, by using the numerical model data, it will be shown that the Doppler-shifted GW amplitude in the vortex dipole is proportional to the power of Ro in the range of 0.15 ≤ Ro ≤ 0.4.
Acknowledgments
This study was supported partly by Grant-in-Aid for Research Fellow (25 8466), by Grant-in-Aid for Scientific Research (A) (25247075), and by Grant-in-Aid for the Young Scientists (B) (25800265) of the Ministry of Education, Culture, Sports and Technology (MEXT), Japan, and by the Leading Graduate Course for Frontiers of Mathematical Sciences and Physics (FMSP) program at the University of Tokyo.
APPENDIX A
Example Application of the RG Method
In this section, an application of the RG method (Chen et al. 1994, 1996) is explained using an ordinary differential equation. The RG method is a singular perturbation method that has several features, as follow: (i) The derived differential equations referred to as renormalization group equations (RGEs) describe the slow variation of the system with respect to time and/or space. (ii) Many singular perturbation methods can be regarded as RG methods [i.e., the boundary layer method, WKB method, multiple-scale method, and reductive perturbation method; see Bender and Orszag (1999) for a detailed discussion on these methods]. (iii) No particular preparations, such as the proper introduction of multiple time scales in the multiple-scale method, are required. (iv) The solution is sometimes more accurate than that obtained by using another method, such as the boundary layer method.
The procedures for applying the RG method are summarized as follows:
A) The regular perturbation solution for x(t) is derived.
B) The independent variables t, which cause the divergence of the secular terms, are transformed into tn = (tn − τn) + τn, where n is a natural number and τ is an arbitrary real number.
C) Renormalized variables and renormalization constants are introduced. The constants of integration in the zeroth-order solution are represented by these quantities.
D) The renormalization constants are determined so that the divergence due to τ is removed at each order, and the renormalized perturbation solution x(t, τ) is obtained.
E) The RGEs are derived by ∂x(t, τ)/∂τ = 0 for any t. The approximate solution for the original equation is obtained by using the solutions for the RGEs and substituting t into τ.
APPENDIX B
Specific Forms of Interaction Coefficients
APPENDIX C
Derivation of the Second-Order Regular Perturbation Solutions
APPENDIX D
Derivation of the Second-Order RGEs
In this section, the second-order RGEs are derived by using the second-order regular perturbation solutions in appendix C.
APPENDIX E
Balanced Model Derived by the Second-Order RG Method
REFERENCES
Alexander, M. J., and T. J. Dunkerton, 1999: A spectral parameterization of mean-flow forcing due to breaking gravity waves. J. Atmos. Sci., 56, 4167–4182, doi:10.1175/1520-0469(1999)056<4167:ASPOMF>2.0.CO;2.
Alexander, M. J., and Coauthors, 2010: Recent developments in gravity-wave effects in climate models and the global distribution of gravity-wave momentum flux from observations and models. Quart. J. Roy. Meteor. Soc., 136, 1103–1124, doi:10.1002/qj.637.
Bender, C. M., and S. A. Orszag, 1999: Advanced Mathematical Methods for Scientists and Engineers: Asymptotic Methods and Perturbation Theory. Springer, 593 pp.
Bühler, O., and M. E. McIntyre, 2005: Wave capture and wave–vortex duality. J. Fluid Mech., 534, 67–95, doi:10.1017/S0022112005004374.
Chen, L.-Y., N. Goldenfeld, and Y. Oono, 1994: Renormalization group theory for global asymptotic analysis. Phys. Rev. Lett., 73, 1311–1315, doi:10.1103/PhysRevLett.73.1311.
Chen, L.-Y., N. Goldenfeld, and Y. Oono, 1996: Renormalization group and singular perturbations: Multiple scales, boundary layers, and reductive perturbation theory. Phys. Rev., 54E, 376–394, doi:10.1103/PhysRevE.54.376.
Dritschel, D. G., and Á. Viúdez, 2003: A balanced approach to modelling rotating stably stratified geophysical flows. J. Fluid Mech., 488, 123–150, doi:10.1017/S0022112003004920.
Ford, R., M. E. McIntyre, and W. A. Norton, 2000: Balance and the slow quasimanifold: Some explicit results. J. Atmos. Sci., 57, 1236–1254, doi:10.1175/1520-0469(2000)057<1236:BATSQS>2.0.CO;2.
Hines, C. O., 1997: Doppler-spread parameterization of gravity-wave momentum deposition in the middle atmosphere. Part 2: Broad and quasi monochromatic spectra, and implementation. J. Atmos. Sol.-Terr. Phys., 59, 387–400, doi:10.1016/S1364-6826(96)00080-6.
Holton, J. R., 2004: An Introduction to Dynamic Meteorology. International Geophysics Series, Vol. 88, Elsevier Academic Press, 535 pp.
Kunihiro, T., 1995: A geometrical formulation of the renormalization group method for global analysis. Prog. Theor. Phys., 94, 503–514, doi:10.1143/PTP.94.503.
Kunihiro, T., 1997: The renormalization-group method applied to asymptotic analysis of vector fields. Prog. Theor. Phys., 97, 179–200, doi:10.1143/PTP.97.179.
Leith, C. E., 1980: Nonlinear normal mode initialization and quasigeostrophic theory. J. Atmos. Sci., 37, 958–968, doi:10.1175/1520-0469(1980)037<0958:NNMIAQ>2.0.CO;2.
Lorenz, E. N., 1980: Attractor sets and quasigeostrophic equilibrium. J. Atmos. Sci., 37, 1685–1699, doi:10.1175/1520-0469(1980)037<1685:ASAQGE>2.0.CO;2.
Lott, F., R. Plougonven, and J. Vanneste, 2010: Gravity waves generated by sheared potential vorticity anomalies. J. Atmos. Sci., 67, 157–170, doi:10.1175/2009JAS3134.1.
Lott, F., R. Plougonven, and J. Vanneste, 2012: Gravity waves generated by sheared three-dimensional potential vorticity anomalies. J. Atmos. Sci., 69, 2134–2151, doi:10.1175/JAS-D-11-0296.1.
McIntyre, M. E., 2009: Spontaneous imbalance and hybrid vortex–gravity structures. J. Atmos. Sci., 66, 1315–1326, doi:10.1175/2008JAS2538.1.
McIntyre, M. E., and W. A. Norton, 2000: Potential vorticity inversion on a hemisphere. J. Atmos. Sci., 57, 1214–1235, doi:10.1175/1520-0469(2000)057<1214:PVIOAH>2.0.CO;2.
Miyazaki, K., K. Sato, S. Watanabe, Y. Tomikawa, Y. Kawatani, and M. Takahashi, 2010: Transport and mixing in the extratropical tropopause region in a high-vertical-resolution GCM. Part II: Relative importance of large-scale and small-scale dynamics. J. Atmos. Sci., 67, 1315–1336, doi:10.1175/2009JAS3334.1.
Mohebalhojeh, A. R., and D. G. Dritschel, 2001: Hierarchies of balance conditions for the f-plane shallow-water equations. J. Atmos. Sci., 58, 2411–2426, doi:10.1175/1520-0469(2001)058<2411:HOBCFT>2.0.CO;2.
Mohebalhojeh, A. R., and M. E. McIntyre, 2007a: Local mass conservation and velocity splitting in PV-based balanced models. Part I: The hyperbalance equations. J. Atmos. Sci., 64, 1782–1793, doi:10.1175/JAS3933.1.
Mohebalhojeh, A. R., and M. E. McIntyre, 2007b: Local mass conservation and velocity splitting in PV-based balanced models. Part II: Numerical results. J. Atmos. Sci., 64, 1794–1810, doi:10.1175/JAS3934.1.
Muraki, D. J., C. Snyder, and R. Rotunno, 1999: The next-order corrections to quasigeostrophic theory. J. Atmos. Sci., 56, 1547–1560, doi:10.1175/1520-0469(1999)056<1547:TNOCTQ>2.0.CO;2.
Ólafsdóttir, E. I., A. B. O. Daalhuis, and J. Vanneste, 2008: Inertia–gravity-wave radiation by a sheared vortex. J. Fluid Mech., 596, 169–189, doi:10.1017/S0022112007009408.
O’Sullivan, D., and T. J. Dunkerton, 1995: Generation of inertia–gravity waves in a simulated life-cycle of baroclinic instability. J. Atmos. Sci., 52, 3695–3716, doi:10.1175/1520-0469(1995)052<3695:GOIWIA>2.0.CO;2.
Plougonven, R., and C. Snyder, 2005: Gravity waves excited by jets: Propagation versus generation. Geophys. Res. Lett., 32, L18802, doi:10.1029/2005GL023730.
Plougonven, R., and C. Snyder, 2007: Inertia–gravity waves spontaneously generated by jets and fronts. Part I: Different baroclinic life cycles. J. Atmos. Sci., 64, 2502–2520, doi:10.1175/JAS3953.1.
Plougonven, R., and F. Zhang, 2007: On the forcing of inertia–gravity waves by synoptic-scale flows. J. Atmos. Sci., 64, 1737–1742, doi:10.1175/JAS3901.1.
Plougonven, R., and F. Zhang, 2014: Internal gravity waves from atmospheric jets and fronts. Rev. Geophys., 52, 33–76, doi:10.1002/2012RG000419.
Plougonven, R., D. J. Muraki, and C. Snyder, 2005: A baroclinic instability that couples balanced motions and gravity waves. J. Atmos. Sci., 62, 1545–1559, doi:10.1175/JAS3426.1.
Plougonven, R., A. Hertzog, and L. Guez, 2013: Gravity waves over Antarctica and the Southern Ocean: Consistent momentum fluxes in mesoscale simulations and stratospheric balloon observations. Quart. J. Roy. Meteor. Soc., 139, 101–118, doi:10.1002/qj.1965.
Saujani, S., and T. G. Shepherd, 2006: A unified theory of balance in the extratropics. J. Fluid Mech., 569, 447–464, doi:10.1017/S0022112006002783.
Snyder, C., D. J. Muraki, R. Plougonven, and F. Zhang, 2007: Inertia–gravity waves generated within a dipole vortex. J. Atmos. Sci., 64, 4417–4431, doi:10.1175/2007JAS2351.1.
Snyder, C., R. Plougonven, and D. J. Muraki, 2009: Mechanisms for spontaneous gravity wave generation within a dipole vortex. J. Atmos. Sci., 66, 3464–3478, doi:10.1175/2009JAS3147.1.
Sugimoto, N., K. Ishioka, and S. Yoden, 2007: Gravity wave radiation from unsteady rotational flow in an f-plane shallow water system. Fluid Dyn. Res., 39, 731–754, doi:10.1016/j.fluiddyn.2007.07.001.
Sugimoto, N., K. Ishioka, and K. Ishii, 2008: Parameter sweep experiments on spontaneous gravity wave radiation from unsteady rotational flow in an f-plane shallow water system. J. Atmos. Sci., 65, 235–249, doi:10.1175/2007JAS2404.1.
Vallis, G. K., 2006: Atmospheric and Oceanic Fluid Dynamics: Fundamentals and Large-Scale Circulation. Cambridge University Press, 745 pp.
Vanneste, J., 2013: Balance and spontaneous wave generation in geophysical flows. Annu. Rev. Fluid Mech., 45, 147–172, doi:10.1146/annurev-fluid-011212-140730.
Vanneste, J., and I. Yavneh, 2004: Exponentially small inertia–gravity waves and the breakdown of quasigeostrophic balance. J. Atmos. Sci., 61, 211–223, doi:10.1175/1520-0469(2004)061<0211:ESIWAT>2.0.CO;2.
Viúdez, Á., 2007: The origin of the stationary frontal wave packet spontaneously generated in rotating stratified vortex dipoles. J. Fluid Mech., 593, 359–383, doi:10.1017/S0022112007008816.
Wang, S., and F. Zhang, 2010: Source of gravity waves within a vortex-dipole jet revealed by a linear model. J. Atmos. Sci., 67, 1438–1455, doi:10.1175/2010JAS3327.1.
Wang, S., F. Zhang, and C. C. Epifanio, 2010: Forced gravity wave response near the jet exit region in a linear model. Quart. J. Roy. Meteor. Soc., 136, 1773–1787, doi:10.1002/qj.676.
Warn, T., O. Bokhove, T. G. Shepherd, and G. K. Vallis, 1995: Rossby number expansions, slaving principles, and balance dynamics. Quart. J. Roy. Meteor. Soc., 121, 723–739, doi:10.1002/qj.49712152313.
Warner, C. D., and M. E. McIntyre, 2001: An ultrasimple spectral parameterization for nonorographic gravity waves. J. Atmos. Sci., 58, 1837–1857, doi:10.1175/1520-0469(2001)058<1837:AUSPFN>2.0.CO;2.
Wirosoetisno, D., T. G. Shepherd, and R. M. Temam, 2002: Free gravity waves and balanced dynamics. J. Atmos. Sci., 59, 3382–3398, doi:10.1175/1520-0469(2002)059<3382:FGWABD>2.0.CO;2.
Yasuda, Y., K. Sato, and N. Sugimoto, 2015: A theoretical study on the spontaneous radiation of inertia–gravity waves using the renormalization group method. Part II: Verification of the theoretical equations by numerical simulation. J. Atmos. Sci., 72, 984–1009, doi:10.1175/JAS-D-13-0371.1.
Zhang, F., 2004: Generation of mesoscale gravity waves in upper-tropospheric jet–front systems. J. Atmos. Sci., 61, 440–457, doi:10.1175/1520-0469(2004)061<0440:GOMGWI>2.0.CO;2.
As discussed by Warn et al. (1995), this expansion may not be mathematically valid for a long time, because both slow and fast variables may be expanded. More accurate discussion will be made in section 4a.
When linear potential vorticity is used, it is possible to construct much more accurate balanced models than the QG model. See Mohebalhojeh and Dritschel (2001) for details.
Our theory is constructed by eliminating little-Doppler-shifted GWs at the first step. A similar theory can be constructed without using this elimination but based on the assumptions (ii)–(vi). In this case, however, it is necessary to deal with twice as many terms as for our method.
The second-order solutions for the diagnostic components have secular terms. These terms are automatically renormalized when renormalizing 
In addition, note that the assumption (iv) with respect to the GW characteristic scale has not been used to obtain the GW modes. As shown in Part II, the GW modes, which are obtained as eigenmodes, contain motions with intermediate spatial scales as well as short ones [=O(Ro)]. Here, the intermediate scale means the spatial scale between the characteristic scale of the balanced flow and the characteristic scale of GWs.