1: \documentclass{jfm}

2: %\documentclass[referee]{jfm}

3: \usepackage{graphicx,bm}

4: \def\dt{{\rm d}t}

5: \def\d{{\rm d}}

6: \def\dk{{\rm d}k}

7: \def\dx{{\rm d}{\bm x}}

8: \def\x{{\bm x}}

9: \def\k{{\bm k}}

10: \def\etal{{\it et al. }}

11: \def\norm#1{\left|\mkern-2mu\left|#1\right|\mkern-2mu\right|}

12: 

13: \title[Impeded inverse transfer in the CHM model of quasi-geostrophic 

14: flows]{Impeded inverse energy transfer in the Charney--Hasegawa--Mima 

15: model of quasi-geostrophic flows}

16: \author[C.V. Tran and D. G. Dritschel]

17: {C\ls H\ls U\ls O\ls N\ls G\ns V.\ns T\ls R\ls A\ls N \ns \and

18: D\ls A\ls V\ls I\ls D\ns G.\ns D\ls R\ls I\ls T\ls S\ls C\ls H\ls E\ls L}

19: \affiliation{School of Mathematics and Statistics, University of St Andrews, 

20: St Andrews KY16 9SS, UK}

21: 

22: %\pubyear{2005}

23: %\volume{538}

24: %\pagerange{119--126}

25: %\date{12 September 2005}

26: %\setcounter{page}{119}

27: 

28: \begin{document}

29: 

30: \maketitle

31: 

32: \begin{abstract}

33: 

34: The behaviour of turbulent flows within the single-layer quasi-geostrophic

35: (Charney--Hasegawa--Mima) model is shown to be strongly dependent on the

36: Rossby deformation wavenumber $\lambda$ (or free-surface elasticity). 

37: Herein, we derive a bound on the inverse energy transfer, specifically on 

38: the growth rate $\d\ell/\dt$ of the characteristic length scale $\ell$ 

39: representing the energy centroid. It is found that 

40: $\d\ell/\dt\le2\norm q_\infty/(\ell_s\lambda^2)$, where $\norm q_\infty$ 

41: is the supremum of the potential vorticity and $\ell_s$ represents the 

42: potential enstrophy centroid of the reservoir, both invariant. This result 

43: implies that in the potential energy dominated regime 

44: ($\ell\ge\ell_s\gg\lambda^{-1}$), the inverse energy transfer is strongly 

45: impeded, in the sense that under the usual time scale no significant 

46: transfer of energy to larger scales occurs. The physical implication is 

47: that the elasticity of the free surface impedes turbulent energy transfer 

48: in wavenumber space, effectively rendering large-scale vortices long-lived 

49: and inactive. Results from numerical simulations of forced-dissipative 

50: turbulence confirm this prediction.

51: 

52: \end{abstract}

53: 

54: \section{Introduction}

55: 

56: The atmosphere and oceans are hugely influenced by the background planetary 

57: rotation and stable density stratification. These features virtually suppress 

58: motion in the vertical direction, rendering geophysical fluid systems, 

59: intrinsically vast in their horizontal extent compared with their limited 

60: depth, approximately two-dimensional (2D) at large scales. Many models 

61: derived from the theory of quasi-geostrophy exploit this simplification.

62: One such model, governed by the Charney--Hasegawa--Mima (CHM) equation, 

63: is the subject of the present study.

64: 

65: The CHM equation, which describes the evolution of potential vorticity in

66: a shallow fluid layer with a free surface, takes the form (Pedlosky 1987)

67: \begin{eqnarray}

68: \label{CHM}

69: \frac{\partial q}{\partial t}+J(\psi,q) &=& 0.

70: \end{eqnarray}

71: Here $\psi(x,y,t)$, the streamfunction, is assumed to be periodic in both

72: $x$ and $y$. Note that $\psi$ is proportional to the free surface height

73: anomaly. The quantity $q=(\Delta-\lambda^2)\psi$ is the potential 

74: vorticity,\footnote{This form of potential vorticity ignores the 

75: so-called $\beta$-term. For some studies concerning the effects of 

76: this term, see Kukhakin \& Orszag (1996), Okunu \& Masuda (2003) and 

77: Smith (2004).} where $\lambda$ is the Rossby deformation wavenumber 

78: ($\lambda=f/c$, where $f$ is the Coriolis frequency and $c$ is the 

79: small-scale gravity waves speed, both assumed constant). The differential 

80: operators $J(\cdot,\cdot)$ and $\Delta$ are, respectively, the Jacobian 

81: and two-dimensional Laplacian. Eq.~(\ref{CHM}) expresses material 

82: conservation of potential vorticity. It also governs the evolution of 

83: quasi-2D fluctuations of the electrostatic potential in a plane 

84: perpendicular to a strong magnetic field applied uniformly to a plasma, 

85: in which case $\psi$ is the electrostatic potential and $\lambda^{-1}$ 

86: is the ion Larmor radius (Hasegawa \& Mima 1978). 

87: 

88: The CHM equation has been a subject of active research for decades (see 

89: for example, Hasegawa \& Mima 1978; Fyfe \& Montgomery 1979; Yanase \& 

90: Yamada 1984; Larichev \& McWilliams 1991; Ottaviani \& Krommes 1992; 

91: Kukharkin, Orszag \& Yakhot 1995; Iwayama, Shepherd \& Watanabe 2002; 

92: Arbic \& Flierl 2003; Tran \& Bowman 2003). A majority of these studies 

93: have been concerned with the dual transfer of the total energy 

94: $\langle-\psi q\rangle/2=\langle|\nabla\psi|^2\rangle/2+\lambda^2\langle

95: \psi^2\rangle/2$ and potential enstrophy $\langle\Delta\psi q\rangle/2=

96: \langle|\Delta\psi|^2\rangle/2+\lambda^2\langle|\nabla\psi|^2\rangle/2$, 

97: which are conserved by the CHM nonlinearities. Here $\langle\cdot\rangle$ 

98: denotes a spatial average. The kinetic energy and the usual enstrophy are 

99: $\langle|\nabla\psi|^2\rangle/2$ and $\langle|\Delta\psi|^2\rangle/2$, 

100: respectively (these are not conserved). The quantity 

101: $\lambda^2\langle\psi^2\rangle/2$ is the potential energy. In addition

102: to the total energy and potential enstrophy, the CHM dynamics also 

103: conserve an infinite class of potential vorticity norms, including

104: the $L^\infty$ norm $\norm q_\infty$. Similar to 2D Navier--Stokes (NS) 

105: turbulence, the dual conservation law implies that the total energy 

106: (potential enstrophy) is preferentially transferred to lower (higher) 

107: wavenumbers. Unlike 2D NS turbulence, each of the invariants in CHM 

108: turbulence consists of two components. The presence of the second 

109: component is due to the elasticity of the free surface, which is known 

110: to have a fundamental influence on the dynamics. A remarkable effect is 

111: that the inverse energy transfer for wavenumbers $k\ll\lambda$ (the 

112: so-called potential energy regime)\footnote{For scales smaller than 

113: $\lambda^{-1}=c/f$, the motion is inherently three-dimensional (cf. 

114: Dritschel \& de la Torre Ju\'arez 1996; Dritschel, de la Torre Ju\'arez 

115: \& Ambaum 1999) and therefore poorly described by (\ref{CHM}).} is 

116: predominantly a transfer of the potential energy (Tran \& Bowman 2003), 

117: i.e. of the quadratic quantity $\langle\psi^2\rangle$. 

118: 

119: Numerical studies of CHM turbulence appear to suggest that in the 

120: potential energy regime, the inverse cascade is strongly impeded. For 

121: turbulence decaying from an initial energy reservoir, \cite{Larichev91} 

122: observe a ``slowness'' in the turbulent evolution. Iwayama \etal (2002) 

123: notice the formation of a sharp spectral peak, a manifestation of an 

124: inverse energy transfer with limited extent in wavenumber space. 

125: \cite{Arbic03} find that the inverse cascade slows down and that vortex 

126: merging takes a very long time. For forced turbulence, Kukharkin \etal 

127: (1995) observe the formation of vortical ``quasicrystals'', states of 

128: long-lived inactive vortices. While these observations are consistent 

129: with the finding of Tran \& Bowman (2003) that the inverse energy transfer 

130: involves virtually no kinetic energy when $k\ll\lambda$, they are not 

131: fully explained by this fact. A more satisfactory explanation would be 

132: that the inverse transfer of potential energy is also impeded in this regime. 

133: 

134: In this Note we derive an upper bound for the growth rate of a 

135: characteristic length scale associated with the inverse energy transfer.

136: We infer from this upper bound that in the potential energy regime,

137: the inverse energy transfer is strongly impeded by the elasticity of

138: the free surface. An interesting physical interpretation of this effect 

139: is that this elasticity prohibits the merger of large-scale vortices to 

140: form larger energy-carrying scales, effectively rendering such vortices 

141: long-lived and inactive. Although this effect is deduced in the framework 

142: of unforced and inviscid dynamics, it also manifests itself in a 

143: forced-dissipative version of (\ref{CHM}), as is illustrated by some 

144: numerical results. 

145: 

146: \section{Inverse energy transfer}

147: 

148: The nonlinear transfer of energy to lower wavenumbers of an energy reservoir

149: (when it occurs) necessarily entails an increase of a suitably defined length 

150: scale representing the energy centroid of the reservoir. Here we consider the 

151: length scale $\ell$ defined by  

152: \begin{eqnarray}

153: \label{ell}

154: \ell=\frac{\langle(-\Delta)^{-1/2}\psi q\rangle}{\langle\psi q\rangle}

155: =\frac{\langle|(-\Delta)^{1/4}\psi|^2\rangle+\lambda^2\langle|(-\Delta)^

156: {-1/4}\psi|^2\rangle}{\langle|\nabla\psi|^2\rangle+\lambda^2\langle

157: \psi^2\rangle}. 

158: \end{eqnarray}

159: In (\ref{ell}) the operator $(-\Delta)^\theta$ is defined by

160: $(-\Delta)^\theta\widehat\psi(\k)=k^{2\theta}\widehat\psi(\k)$, where

161: $k=|\k|$ is the wavenumber and $\widehat\psi(\k)$ the Fourier transform

162: of $\psi(\x)$. The growth rate $\d\ell/\dt$ provides a quantitative 

163: measure of the inverse transfer. In particular, a smaller rate 

164: corresponds to an inverse transfer with a more limited extent in 

165: wavenumber space than one with a greater rate. In this respect, 

166: $\d\ell/\dt$ may be considered as a measure of locality of the 

167: inverse transfer, in the sense that for a given energy distribution 

168: a smaller rate implies a more local transfer than a greater one.

169: 

170: In the case where most of the kinetic energy resides in wavenumbers 

171: $k\ll\lambda$ (the potential energy regime),  

172: $\langle|(-\Delta)^{1/4}\psi|^2\rangle$ is negligible compared with 

173: $\lambda^2\langle|(-\Delta)^{-1/4}\psi|^2\rangle$, and similarly 

174: $\langle|\nabla\psi|^2\rangle$ is negligible compared with

175: $\lambda^2\langle\psi^2\rangle$. More precisely, in this regime we 

176: have $$\frac{\langle|(-\Delta)^{1/4}\psi|^2\rangle}{\lambda^2\langle

177: |(-\Delta)^{-1/4}\psi|^2\rangle}\le\frac{\langle|\nabla\psi|^2\rangle}

178: {\lambda^2\langle\psi^2\rangle}\ll1,$$ where the first inequality is due

179: to H\"older's inequality (see also Tran \& Shepherd 2002 and Tran 2004), 

180: implying 

181: $\ell\approx\langle|(-\Delta)^{-1/4}\psi|^2\rangle/\langle\psi^2\rangle$. 

182: For an energy reservoir in the potential energy regime, the length scale 

183: $\ell$ specifies where, in wavenumber space, most of $\langle\psi^2\rangle$ 

184: (or equivalently most of the potential energy) is distributed. Its growth 

185: rate is thus the rate at which the spectral profile of the potential 

186: energy proceeds toward lower wavenumbers.

187: 

188: The evolution equation for the characteristic length scale $\ell$, defined 

189: in the preceding paragraphs, is obtained by multiplying (\ref{CHM}) by 

190: $(-\Delta)^{-1/2}\psi/\langle\psi q\rangle$ and taking the spatial average 

191: of the resulting equation:

192: \begin{eqnarray}

193: \label{evol1}

194: \frac{1}{2}\frac{\d\ell}{\dt} &=&

195: \frac{\langle(-\Delta)^{-1/2}\psi J(\psi,q)\rangle}

196: {\langle|\nabla\psi|^2\rangle+\lambda^2\langle\psi^2\rangle}

197: =-\frac{\langle q J(\psi,(-\Delta)^{-1/2}\psi)\rangle}

198: {\langle|\nabla\psi|^2\rangle+\lambda^2\langle\psi^2\rangle}\nonumber\\

199: &\le&

200: \frac{\langle|q||\nabla\psi||\nabla(-\Delta)^{-1/2}\psi|\rangle}

201: {\langle|\nabla\psi|^2\rangle+\lambda^2\langle\psi^2\rangle}

202: \le \frac{\langle|\nabla\psi|^2\rangle^{1/2}\langle\psi^2\rangle^{1/2}}

203: {\langle|\nabla\psi|^2\rangle+\lambda^2\langle\psi^2\rangle}

204: \norm q_\infty,

205: \end{eqnarray} 

206: where the various manipulations of the nonlinear term are straightforward.

207: In particular, the last step can be readily seen by expressing the 

208: streamfunction in terms of a Fourier series.

209: 

210: The ratio in front of $\norm q_\infty$ on the right-hand side of 

211: (\ref{evol1}) can be closely estimated in terms of conserved quantities. 

212: For that purpose, let us denote by $\ell_s$ the conserved length scale 

213: representing the enstrophy centroid of the reservoir:

214: \begin{eqnarray}

215: \label{enscentroid}

216: \ell_s &=& 

217: \frac{\langle-\psi q\rangle^{1/2}}{\langle\Delta\psi q\rangle^{1/2}}=\frac

218: {\left(\langle|\nabla\psi|^2\rangle+\lambda^2\langle\psi^2\rangle\right)^{1/2}}

219: {\left(\langle|\Delta\psi|^2\rangle+\lambda^2\langle|\nabla\psi|^2\rangle

220: \right)^{1/2}}.

221: \end{eqnarray} 

222: Solving (\ref{enscentroid}) for $\langle\psi^2\rangle$ and 

223: substituting the result into the said ratio we obtain

224: \begin{eqnarray}

225: \label{ratio}

226: \frac{\langle|\nabla\psi|^2\rangle^{1/2}\langle\psi^2\rangle^{1/2}}

227: {\langle|\nabla\psi|^2\rangle+\lambda^2\langle\psi^2\rangle} 

228: &=& \frac{\langle|\nabla\psi|^2\rangle^{1/2}

229: \left[\ell_s^2\langle|\Delta\psi|^2\rangle+(\ell_s^2\lambda^2-1)\langle

230: |\nabla\psi|^2\rangle\right]^{1/2}}{\lambda\left[\langle|\nabla\psi|^2

231: \rangle+\ell_s^2\langle|\Delta\psi|^2\rangle+(\ell_s^2\lambda^2-1)\langle

232: |\nabla\psi|^2\rangle\right]}\nonumber\\ 

233: &\le& \frac{\langle|\nabla\psi|^2\rangle^{1/2}

234: \left(\langle|\Delta\psi|^2\rangle+\lambda^2\langle

235: |\nabla\psi|^2\rangle\right)^{1/2}}{\ell_s\lambda\left(\langle|

236: \Delta\psi|^2\rangle+\lambda^2\langle

237: |\nabla\psi|^2\rangle\right)}\nonumber\\ 

238: &=& \frac{\langle|\nabla\psi|^2\rangle^{1/2}}{\ell_s\lambda

239: \left(\langle|\Delta\psi|^2\rangle+\lambda^2\langle

240: |\nabla\psi|^2\rangle\right)^{1/2}}

241: \le\frac{1}{\ell_s\lambda^2}.

242: \end{eqnarray} 

243: This is a very sharp estimate, especially for a reservoir in the 

244: potential energy regime. We note that in general the inequality 

245: $\ell\ge\ell_s$ holds. This is a consequence of the H\"older inequality

246: and can be shown as follows. Let us denote by $U(k)$ the power spectrum 

247: of $\langle-(-\Delta)^{-1/2}\psi q\rangle$. The length sales $\ell$

248: and $\ell_s$ can then be expressed in terms of $U(k)$ as 

249: $$\ell=\frac{\int U(k)\,\dk}{\int kU(k)\,\dk}$$ and 

250: $$\ell_s=\left(\frac{\int kU(k)\,\dk}{\int k^3U(k)\,\dk}\right)^{1/2}.$$ 

251: Now by H\"older's inequality we have 

252: $$\int kU(k)\,\dk\le\left(\int U(k)\,\dk\right)^{2/3}

253: \left(\int k^3U(k)\,\dk\right)^{1/3},$$ from which the inequality

254: $\ell\ge\ell_s$ indeed follows immediately.

255: 

256: Putting these results together we have

257: \begin{eqnarray}

258: \label{evol2}

259: \frac{\d\ell}{\dt} &\le& 2\frac{\norm q_\infty}{\ell_s\lambda^2}

260: \le 2\frac{\norm q_\infty}{\ell_s^2\lambda^2}\ell

261: \end{eqnarray}

262: using $\ell\ge\ell_s$. Hence the exponential growth rate of $\ell$ is 

263: bounded from above by $2\norm q_\infty/(\ell_s^2\lambda^2)$. Given a 

264: finite $\norm q_\infty$, the upper bound $2\norm q_\infty/(\ell_s^2\lambda^2)$ 

265: can be made arbitrarily small provided $\ell_s^2\lambda^2$ is sufficiently 

266: large. This condition is satisfied for an initial energy reservoir in a 

267: sufficiently remote wavenumber region of the potential energy regime,

268: where both $\ell_s\lambda\gg1$ and $\ell\lambda\gg1$. Under these 

269: circumstances, no significant growth of $\ell$ occurs when the initial 

270: energy reservoir freely spreads out, and hence no significant inverse 

271: energy transfer is possible. In passing it is worth mentioning that the 

272: upper bound for the exponential growth rate of $\ell$ in (\ref{evol2}) 

273: consists of the constant $\lambda$ and the two invariants $\norm q_\infty$ 

274: and $\ell_s$, and therefore can be completely determined by initial data.

275: 

276: The result in the preceding paragraph is an analytic version of the 

277: qualitative statement in the literature that the turbulence evolves 

278: on a rescaled (slower) time (cf. Larichev \& McWilliams 1991; Kukharkin, 

279: Orszag \& Yakhot 1995). To clarify this point let us rewrite 

280: (\ref{evol2}) as

281: \begin{eqnarray}

282: \label{evol3}

283: \ell_s^2\lambda^2\frac{\d\ell}{\dt} &\le& 2\norm q_\infty\ell.

284: \end{eqnarray}

285: The left-hand side of (\ref{evol3}) is the (instantaneous) growth rate 

286: of $\ell$ on the slow time $\tau=t/(\ell_s^2\lambda^2)$. For an initial

287: reservoir with a finite $\norm q_\infty$, the inverse transfer of 

288: potential energy can be considerable only on this slow time scale. In 

289: the limit $\ell_s\lambda\rightarrow\infty$, the invariants become 

290: $\langle\psi^2\rangle/2$ and $\langle|\nabla\psi|^2\rangle/2$ and the 

291: length scales $\ell$ and $\ell_s$ become

292: $\langle|(-\Delta)^{-1/4}\psi|^2\rangle/\langle\psi^2\rangle$ and 

293: $\langle\psi^2\rangle^{1/2}/\langle|\nabla\psi|^2\rangle^{1/2}$, 

294: respectively. The rescaled time $\tau$ becomes infinitely long. 

295: 

296: An estimate which may be significantly sharper than (\ref{evol2}) can be 

297: derived by replacing the potential vorticity $q$ in (\ref{evol1}) by the 

298: relative vorticity $\Delta\psi$. By doing that, instead of (\ref{evol2}), 

299: we obtain

300: \begin{eqnarray}

301: \label{evol4}

302: \frac{\d\ell}{\dt} &\le& 2\frac{\norm{\Delta\psi}_\infty}

303: {\ell_s^2\lambda^2}\ell.

304: \end{eqnarray}

305: In remote regions of the potential energy regime, where 

306: $\ell\lambda\ge\ell_s\lambda\gg1$, we expect $\lambda^2\psi$ to 

307: dominate $\Delta\psi$. Therefore in these regions it is plausible that

308: $\norm q_\infty\gg\norm{\Delta\psi}_\infty$, allowing (\ref{evol4}) to 

309: become significantly sharper than (\ref{evol2}). The price for this 

310: improvement is that $\norm{\Delta\psi}_\infty$ is not conserved. Note

311: that in such regions $\ell$ and $\ell_s$ are the length scales of 

312: potential and kinetic energy, respectively.

313: 

314: A straightforward physical interpretation of the above result is that 

315: the elasticity of the free surface suppresses energy transfer for scales 

316: larger than the Rossby deformation radius. This effect has a profound 

317: influence on the vortex dynamics. Let us consider a hypothetical flow 

318: consisting of vortices (having smooth potential vorticity---no small 

319: length scales) of comparable scales, which are much larger than the 

320: deformation radius $\lambda^{-1}$. For these scales the inverse energy 

321: transfer is strongly impeded, allowing for virtually no formation of 

322: larger energy-carrying scales. This keeps the vortices from merging, 

323: effectively establishing a state of long-lived inactive vortices. This 

324: may explain the crystallisation of vortices observed by Kukharkin \etal 

325: (1995). The impeded inverse energy transfer implies a similar effect on 

326: the direct transfer of potential enstrophy, a consequence of the dual 

327: conservation law. One can see from the ``symmetry'' of the dual 

328: conservation law that the available potential enstrophy for direct 

329: transfer depends on the inversely-transferred energy. Similarly, the 

330: available energy for inverse transfer depends on the directly-transferred 

331: potential enstrophy. Hence impeded inverse energy transfer means impeded 

332: direct potential enstrophy transfer (and vice versa). This implies that 

333: smooth potential vorticity with scales $\gg\lambda^{-1}$ remains smooth.  

334: 

335: \section{Numerical results}

336: 

337: We now consider results obtained from numerical simulations of a 

338: forced-dissipative version of (\ref{CHM}). These convincingly illustrate 

339: the strong reduction of the inverse transfer in the potential energy 

340: regime, and may help explain an observation by Iwayama \etal (2002) in 

341: their numerical simulations of freely decaying CHM turbulence wherein an 

342: initial energy reservoir evolves into a single sharp peak. A plausible 

343: explanation for this observation is that the sharp peak is due to both 

344: the viscous dissipation (which suppresses the formation of small scales) 

345: and the impeded inverse transfer deduced in the preceding section: their 

346: collective effect ``squeezes'' energy into a narrow band of wavenumbers, 

347: resulting in the observed peak. In our forced-dissipative case we observe 

348: that in the potential energy regime the inverse energy flux is arrested 

349: and deposits most of the potential energy at the lower-wavenumber end of 

350: the quasi-steady energy range. We also observe a relaxation of the kinetic 

351: energy spectrum of the inertial range (from $k^{-5/3}$ toward $k^{-1}$) 

352: as the inverse cascade proceeds from $k>\lambda$ to $k<\lambda$. 

353:  

354: We simulate Eq. (\ref{CHM}), with the addition of molecular viscosity 

355: and a body force, in a doubly periodic square $2\pi\times2\pi$ with 

356: a dealiased $1024^2$ pseudospectral method. No large-scale dissipation

357: mechanisms are employed and no steady dynamics are sought since we are

358: primarily interested in the near arrest of the inverse energy cascade in

359: a quasi-steady state. (For an analysis of steady CHM turbulence in the

360: presence of large-scale damping, see Smith \etal 2002.) Three different 

361: values of $\lambda$ are used: $\lambda=20,40$ and $60$. The forcing is 

362: spectrally localised in the wavenumber interval $K=[49.5,50.5]$, in the 

363: sense that its Fourier components $\widehat f(\k)$ are nonzero only for 

364: those modes $\k$ having magnitudes $k$ lying in the interval $K$: 

365: \begin{eqnarray}

366: \label{forcing}

367: \widehat{f}(\bm k)&=&\frac{\epsilon}{N}\frac{\widehat{\psi}(\bm{k})}

368: {\displaystyle\sum_{|\bm{p}|=k}|\widehat{\psi}(\bm{p})|^2}.

369: \end{eqnarray}

370: Here $\epsilon=1$ is the constant energy injection rate and~$N$ is the 

371: number of distinct wavenumbers in~$K$. The viscosity coefficient is 

372: $\nu=5\times10^{-4}$. The simulations were initialised with the kinetic 

373: energy spectrum $E(k)=10^{-5}\pi k/(50^2+k^2)$.

374: 

375: \begin{figure}

376: \centerline{\includegraphics{atmosp234}}

377: \caption{The kinetic energy spectrum $E(k)$ {\it vs.}\ $k$ averaged over 

378: periods $[139,140]$, $[155,156]$ and $[160,161]$ for $\lambda=20$, $40$ 

379: and $60$, respectively.} 

380: \label{atmosp234}

381: \end{figure} 

382: 

383: Figure \ref{atmosp234} shows the quasi-steady kinetic energy spectrum 

384: $E(k)$ vs. $k$ 

385: averaged over periods $[139,140]$, $[155,156]$ and $[160,161]$ for 

386: $\lambda=20$, $40$ and $60$, respectively. In all three cases, the 

387: accumulation of energy at low wavenumbers is clearly visible. This is 

388: attributed to the impeding effect discussed above, and not finite-size 

389: effects since the energy levels at the lowest wavenumbers are still 

390: small compared with the energy peaks. This is true for both kinetic 

391: and potential components. For $\lambda=20$ we have a limited kinetic 

392: energy regime in the wavenumber interval $[20,50]$. Before the energy 

393: peak reaches $\lambda=20$, the dynamics of this kinetic energy regime 

394: are expected to resemble those of 2D NS turbulence: the energy cascades 

395: to low wavenumbers via a $k^{-5/3}$ kinetic energy spectrum. We indeed 

396: observe a discernible $k^{-5/3}$ spectrum (not shown) in the wavenumber 

397: range $[20,50]$ before the energy peak reaches $\lambda=20$. The dynamics 

398: after the energy peak passes $\lambda$ undergo profound changes. Most 

399: notably, the growth of potential energy occurs mainly around the energy 

400: peak, consistent with the reduction of the inverse energy transfer in 

401: the conservative case. Another remarkable change is in the spectral 

402: slope of the energy spectrum from $k=\lambda=20$ up to the forcing 

403: wavenumber $k=50$. The slope of the kinetic energy spectrum in this 

404: region appears to relax from $-5/3$ to $-1$. The quasi-steady\footnote

405: {Note that after the energy peak passes $\lambda$, growth of energy 

406: is due primarily to that of the potential energy and the kinetic 

407: energy becomes quasi-steady.} spectrum becomes $k^{-1}$ with an 

408: excessive accumulation of energy at the lower-wavenumber end. This 

409: unexpected feature is common to the other two cases, for which hardly 

410: any inverse transfer region exists between the forcing wavenumber and 

411: $\lambda$ and for which no clear $k^{-5/3}$ spectrum is formed initially.

412: 

413: \begin{figure}

414:     \begin{center}

415:     \includegraphics[scale=0.625]{ell-ori.eps} \,\,

416:     \includegraphics[scale=0.625]{ell-sca.eps} 

417:     \end{center}

418: \caption{The energy centroid $\ell$ versus time $t$ (left) and scaled

419: time $t(20/\lambda)^2$ (right) for $\lambda=20$, $40$ 

420: and $60$, as labelled.} 

421: \label{ell_vs_time}

422: \end{figure} 

423: 

424: The evolution of the energy centroid $\ell$ for these three cases is

425: shown in Figure \ref{ell_vs_time}, on the left versus unscaled time $t$

426: and on the right versus scaled time $t(20/\lambda)^2$.  Both the 

427: reduction in ${\d\ell}/{\dt}$ with increasing $\lambda$ {\it and} 

428: the collapse of the curves when plotted versus scaled time are 

429: consistent with the growth rate bound (\ref{evol2}).\footnote

430: {We cannot directly compare the numerical growth rates with 

431: the theoretical bound because the imposed forcing violates 

432: conservation of $\norm{q}_\infty$ and $\ell_s$ used 

433: in (\ref{evol2}).}

434: Only the $\lambda = 20$ case stands out slightly, due to 

435: an early kinetic energy transfer which is virtually absent in 

436: the other two cases.  

437: 

438: \section{Summary and discussion}

439: 

440: The main result of this work is an upper bound for the growth rate of

441: the characteristic length scale associated with the inverse energy 

442: transfer in CHM turbulence. This upper bound is expressible in terms of 

443: the supremum of the potential vorticity, the characteristic length scale

444: representing the potential enstrophy centroid of the reservoir and the 

445: Rossby deformation radius. These are constants of motion, allowing the 

446: bound to be completely described in terms of initial data. It is found 

447: that in remote regions of the potential energy regime, the turbulent 

448: transfer from an initial energy reservoir becomes strongly impeded, in 

449: the sense that hardly any potential energy is transferred to lower 

450: wavenumbers. This effect has been illustrated by numerical simulations 

451: of a forced-dissipative version of the CHM equation. The physical 

452: implication of the present finding is that the elasticity of the free

453: surface hinders the nonlinear transfer of potential energy to larger scales. 

454: 

455: An interesting implication of the present result is that vortices having

456: smooth potential vorticity and scales much larger than the Rossby 

457: deformation radius are kept from merging to form larger energy-carrying 

458: scales and remain smooth. Such vortices are inactive and long-lived 

459: compared to their counterparts in 2D NS turbulence.

460: 

461: The present theory is fully consistent with other results for both forced

462: (Kukharkin \etal 1995) and freely decaying (Larichev \& McWilliams 1991; 

463: Iwayama \etal 2002; Arbic \& Flierl 2003) CHM turbulence, as discussed above.

464: These, we argue, are explained by the growth rate bound (\ref{evol3}) and 

465: its physical interpretations given at the end of \S\,2. 

466: 

467: Analyses of atmospheric and oceanic data yield results that could plausibly 

468: be explained by the present work. \cite{Boer83} find that the spectral flux 

469: in atmospheric wind data is arrested with essentially no inertial range. 

470: \cite{Scott05} find the inverse energy transfer in the oceans to be 

471: qualitatively similar to the result of \cite{Boer83}. These are in 

472: remarkable agreement with the idealised numerical results noted above and 

473: with the present findings.

474: 

475: Iwayama \etal (2002) observe ``well-behaved'' tails of the potential 

476: vorticity probability density function (PDF)---much steeper than observed

477: in NS turbulence, see their figure 5---and raise the question why it 

478: is that CHM turbulence in the potential energy regime exhibits no 

479: intermittency. A speculative answer provided by the authors is based 

480: on an asymptotic version of the CHM equation wherein the relative 

481: vorticity is dropped from the time-derivative term: 

482: $$\lambda^2\,\frac{\partial\psi}{\partial t}+J(\Delta\psi,\psi) = 0.$$ 

483: This equation, as noted by Iwayama \etal (2002), describes advection of

484: $\psi$ by $\Delta\psi$ (i.e. by the smaller scales of motion). These 

485: authors argue that this may lead to more ``random walk'' behaviour, 

486: effectively suppressing intermittency. The present study provides an 

487: alternative answer: the reduction of turbulent transfer both renders 

488: large-scale vortices inactive and discourages erratic small-scale 

489: motion. This also can account for the observed steep PDF. In some sense 

490: CHM turbulence in the potential energy regime is far less turbulent 

491: than its 2D NS counterpart. 

492: 

493: \begin{acknowledgments}

494: The numerical simulations in \S\,3 were carried out, using a computer 

495: code developed by John Bowman, when CVT was a Pacific Institute for the 

496: Mathematical Sciences postdoctoral fellow at the University of Alberta. 

497: Constructive comments from two anonymous referees were very much appreciated.

498: \end{acknowledgments} 

499: 

500: \begin{thebibliography}{}

501: 

502: \bibitem[Arbic \& Flierl (2003)]{Arbic03} 

503:      \textsc{Arbic, B. K. \& Flierl, G. R.} 2003

504:      {Coherent vortices and kinetic energy ribbons in asymptotic, quasi 

505:       two-dimensional f-plane turbulence.} 

506:      \textit{Phys. Fluids} {\bf 15}, 2177--2189.

507: 

508: \bibitem[Boer \& Shepherd (1983)]{Boer83} 

509:      \textsc{Boer, G. J. \& Shepherd, T. G.} 1983

510:      {Large-scale two-dimensional turbulence in the atmosphere.} 

511:      \textit{J. Atmos. Sci.} {\bf 40}, 164--184.

512: 

513: \bibitem[Dritschel, de la Torre Ju\'arez \& Ambaum (1999)]{Dritschel99}

514:      \textsc{Dritschel, D. G., de la Torre Ju\'arez, M. \& Ambaum, M. H. P.} 

515:       1999

516:      {The three dimensional vortical nature of atmospheric and oceanic

517:       turbulent flows.} 

518:      \textit{Phys. Fluids} \textbf{11}, 1512--1520.

519: 

520: \bibitem[Dritschel \& de la Torre Ju\'arez (1996)]{Dritschel96}

521:      \textsc{Dritschel, D. G. \& de la Torre Ju\'arez, M.} 1996

522:      {The instability and breakdown of tall columnar vortices in a

523:       quasi-geostrophic fluid.} 

524:      \textit{J. Fluid Mech.} \textbf{328}, 129--160.

525: 

526: \bibitem[Fyfe \& Montgomery (1979)]{Fyfe79}

527:      \textsc{Fyfe, D. \& Montgomery, D.} 1979

528:      {Possible inverse cascade behaviour for drift-wave turbulence.} 

529:      \textit{Phys. Fluids} \textbf{22}, 246--248.

530: 

531: \bibitem[Hasegawa (1978)]{Hasegawa78} 

532:      \textsc{Hasegawa, A. \& Mima, K.} 1978

533:      {Pseudo-three-dimensional turbulence in magnetized nonuniform plasma.} 

534:      \textit{Phys. Fluids} {\bf 21}, 87--92.

535: 

536: \bibitem[Iwayama, Shepherd \& Watanabe (2002)]{Iwayama02}

537:      \textsc{Iwayama, T., Shepherd, T. G. \& Watanabe, T.} 2002

538:      {An `ideal' form of decaying two-dimensional turbulence.}

539:      \textit{J. Fluid Mech.} \textbf{456}, 183--198.

540: 

541: \bibitem[Kukharkin \& Orszag (1996)]{Kukharkin96}

542:      \textsc{Kukharkin, N. \& Orszag, S. A.} 1996

543:      {Generation and structure of Rossby vortices in rotating fluids.}

544:      \textit{Phys. Rev.} E \textbf{54}, 4524--4527.

545: 

546: \bibitem[Kukharkin, Orszag \& Yakhot (1995)]{Kukharkin95}

547:      \textsc{Kukharkin, N., Orszag, S. A. \& Yakhot, V.} 1995

548:      {Quasicrystallization of vortices in drift-wave turbulence.}

549:      \textit{Phys. Rev. Lett.} \textbf{75}, 2486--2489.

550: 

551: \bibitem[Larichev \& McWilliams (1991)]{Larichev91}

552:      \textsc{Larichev, V. D. \& McWilliams, J. C.} 1991

553:      {Weakly decaying turbulence in an equivalent-barotropic fluid.}

554:      \textit{Phys. Fluids} A \textbf{3}, 938--950. 

555: 

556: \bibitem[Okuno \& Masuda (2003)]{Okuno03}

557:      \textsc{Okuno, A. \& Masuda, A.} 2003

558:      {Effects of horizontal divergence on the geostrophic turbulence on a

559:       beta-plane: suppression of the Rhines effect.}

560:      \textit{Phys. Fluids} \textbf{15}, 56--65. 

561: 

562: \bibitem[Ottaviani \& Krommes (1992)]{Ottaviani92}

563:      \textsc{Ottaviani, M. \& Krommes, J. A.} 1992

564:      {Weak- and strong-turbulence regimes of the forced Hasegawa--Mima

565:       equation.}

566:      \textit{Phys. Rev. Lett.} \textbf{69}, 2923--2026.

567: 

568: \bibitem[Pedlosky (1987)]{Pedlosky87}

569:      \textsc{Pedlosky, J.} 1987

570:      \textit{Geophysical Fluid Dynamics,} 2nd Edn. Springer. 

571: 

572: \bibitem[Scott \& Wang (2005)]{Scott05} 

573:      \textsc{Scott, R. B. \& Wang, F.} 2005

574:      {Direct evidence of an oceanic inverse kinetic energy cascade from 

575:       satellite altimetry.}

576:      \textit {J. Phys. Ocean.} \textbf{35}, 1650--1666.

577: 

578: \bibitem[Smith (2004)]{Smith04}

579:      \textsc{Smith, K. S.} 2004

580:      {A local model for planetary atmospheres forced by small-scale 

581:       convection.}

582:      \textit{J. Atmos. Sci.} \textbf{61}, 1420--1433.

583: 

584: \bibitem[Smith \etal (2002)]{Smith02}

585:      \textsc{Smith, K. S., Boccaletti, G., Henning, C. C., Marinov, I. N.,

586:       Tam, C. Y., Held, I. M. \& Vallis, G. K.} 2002

587:      {Turbulence diffusion in the geostrophic inverse cascade.}

588:      \textit{J. Fluid Mech.} \textbf{469}, 13--48.

589: 

590: \bibitem[Tran (2004)]{T04} 

591:      \textsc{Tran, C. V.} 2004

592:      {Nonlinear transfer and spectral distribution of energy in 

593:      $\alpha$ turbulence.}

594:      \textit {Physica} D \textbf{191}, 137--155.

595: 

596: \bibitem[Tran \& Bowman (2003)]{TB03} 

597:      \textsc{Tran, C. V. \& Bowman, J. C.} 2003

598:      {Energy budgets in Charney--Hasegawa--Mima and surface 

599:      quasigeostrophic turbulence.}

600:      \textit{Phys. Rev.} E \textbf{68}, 036304.

601: 

602: \bibitem[Tran \& Shepherd (2002)]{TS02}

603:      \textsc{Tran, C. V. \& Shepherd, T. G.} 2002 

604:      {Constraints on the spectral distribution of energy and enstrophy 

605:      dissipation in forced two-dimensional turbulence.} 

606:      \textit{Physica} D \textbf{165}, 199--212. 

607: 

608: \bibitem[Yanase \& Yamada (1984)]{Yanase84}

609:      \textsc{Yanase, S. \& Yamada, M.} 1984

610:      {The effect of the finite Rossby radius on two-dimensional isotropic

611:      turbulence.}

612:      \textit{J. Phys. Soc. Japan} \textbf{53}, 2513--2520.

613: 

614: \end{thebibliography}

615: 

616: \end{document}

617: