betaFeedback

WorksPhysics and Astronomy

Dynamics of equatorial jets generated by temperature fronts

Abstract

The theory of temperature jets gets extended to account for the influence of the beta effect on their dynamics. Including this effect noticeably changes symmetry properties and laws of conservation inherent to models without the beta effect. Especially nontrivial, the dynamics of jets become near the equator, where the model admits multi-valued solutions that look like jets crossing the equator. For example, the so-called loop solitons moving along the equator with velocities inversely proportional to the cube of their amplitude turn out among them. Estimates based on simple qualitative considerations show that, owing to the beta effect, the temperature jets can form a specific equatorial turbulence, in which they play the role of structural elements. Notably, the spectral slope of the energy density at large wave numbers in such turbulence becomes equal to unity.

Introduction

Frontal jet currents in the ocean and the atmosphere are caused by horizontal pressure gradients that in rotating nonhomogeneous fluids occur due to having narrow transition zones between regions of different potential vorticity (PV) and/or temperature under the action of the Coriolis and gravity forces[1]. With varying degrees of detail, this effect is reproduced by many models of geophysical hydrodynamics. But, most simply, it can be mathematically formalized and studied in the context of vertically averaged quasi-geostrophic models.

One such approach can be developed within the thermal rotating shallow-water (TRSW)[2, 3, 4] model, which describes a two-dimensional (columnar) motion of incompressible horizontally-nonuniform fluid in a sufficiently thin layer under the shallow-water approximation and in neglecting any dissipation. In the TRSW model, two scenarios of jet generation are possible, depending on which field (potential vorticity or temperature) undergoes the jump at the frontal interface.

The theory of thin jets initiated by cross-frontal PV jumps can be developed only assuming the horizontal homogeneity of the temperature field. Otherwise, the potential vorticity has no Lagrangianity property, and it is this property that allows us to trace the jet’s trajectory during its evolution. A brief description of some models and theoretical approaches[5, 6, 7, 8, 9, 10, 11, 12] that have been used at different times to study the trajectory dynamics of jets generated at PV fronts can be found in Ref.[13].

An alternative scenario of thin jets resulting from temperature jumps has been recently analyzed in Ref.[15]. Like PV jets, temperature jets also have a Lagrangian character, but their trajectories are localized not at lines of PV jumps but at temperature jump lines. Although, in deriving, both theories assume the same relation L∼T1/3L \sim T^{1/3} between the characteristic scales of length LL and time TT, their equations of motion, symmetries, and hence solutions turn out to be fundamentally different.

Another essential difference between the theories exhibits when accounting for the beta effect. If, for the PV jets, the inclusion of the beta effect leads to a Doppler shift without any violation of the translational invariance of the theory, for the temperature jets, it is not so at all. Moreover, in this case, their path equations lose, as well, some other symmetries to be inherent for the planar version of the theory when β=0\beta= 0. For instance, path equations become no longer invariant under rotations and inversions. Instead of them, the theory gains the properties that open possibilities for existing solutions of a new type. These solutions, known as loop solitons[8, 11, 13], destroy the connectedness of the flow by breaking it into closed domains and can enhance large-scale zonal transport.

One more application area for the concept of temperature jets is large-scale turbulence. That means the collective contribution of temperature jets to atmospheric dynamics can be considered a statistical process in which they play the role of structural elements. In this respect, studying the spectral properties of equatorial turbulence and investigating the beta-effect influence on temperature jets are two sides of the same coin.

The paper is structured as follows. In § 0.2, we outline a general statement of the problem and formulate the variational principle for a depth-averaged, horizontally inhomogeneous layer of a rotating fluid. In Sec. III, by using the scaling of variables, we develop a self-consistent perturbation theory to find the thin jets Lagrangian in the equatorial beta-plane approximation. Sec. IV analyzes the motion equations describing the dynamics of equatorial jets and includes some solutions, among which loop solitons. In § 0.5, we discuss obtained results. The validity of the equatorial parameterization for Coriolis is argued in Appendix § 0.4.1 in the context of its applicability to a layer on a rotating sphere. Appendix § 0.4.2 addresses the spectral properties of turbulence for which straight-line temperature jets are its structural elements.

General statement

The shortest path to formulate the equations of motion for frontal temperature jets, with the account of the beta effect, is to use the variational least action principle

δS=0,S=∫L dt,(1)\delta S = 0,\qquad S = \int\mathcal{L}\,dt, \tag*{(1)}

where SS is the action and L\mathcal{L} is the Lagrangian, which, based on physically reasonable approximations, needs to be successively simplified up to its final form.

As a first step towards this goal, we consider the model of a two-layer inviscid incompressible fluid that moves under the action of gravity and Coriolis forces and occupies the half-space z>0z > 0, where zz is the vertical coordinate. Let the lower layer of a sufficiently small height hh rest on the impermeable bottom z=0z = 0 and be horizontally inhomogeneous. with the density ϱ0+Δϱ\varrho_0 + \Delta\varrho. But in doing so, the upper layer, filled with the homogeneous fluid of the density ϱ0\varrho_0, spans to infinity. Moreover, if all kinetic energy is assumed to be determined mainly by horizontal motions and concentrated in the lower layer above which there is a weakly perturbed, almost homogeneous fluid, then, after integration over zz, we obtain the Lagrangian

L=12∫h(u2+v2−2uR−hb) dxdy.(2)\mathcal{L} = \frac{1}{2} \int h\left(u^2 + v^2 - 2uR - hb\right)\,\mathrm{d}x\mathrm{d}y. \tag*{(2)}

In this expression, besides those already specified, we use the following notations: xx, yy are the horizontal Cartesian coordinates, tt is the time, subscripts denote the partial derivatives, uu, vv are components of the horizontal velocity. Responsible for rotation, the function RR depends only on yy, and its derivative gives the Coriolis parameter ff:

Ry=f.(3)R_y = f. \tag*{(3)}

To account for the influence of Earth’s sphericity, we will use the so-called equatorial beta-plane approximation

f=2Ωay,(4)f = \frac{2\Omega}{a}y, \tag*{(4)}

where Ω\Omega is the angular rotation rate, and aa is the planetary radius. Integrating this parameterization over yy leads to the expression

R=Ωay2.(5)R = \frac{\Omega}{a}y^2. \tag*{(5)}

As well known, the beta effect is one of the keystones in geophysical fluid dynamics. For the model under study, that is quite explainable since the presence of the equator removes some symmetries from the dynamics of the jets, which are inherent to them in the planar case. Notably, if we consider the above model on a rotating sphere and use Lambert’s cylindrical projection (see Appendix § 0.4.1 for more details)

y=asin⁡φ(6)y = a\sin\varphi \tag*{(6)}

to introduce the variable yy instead of the latitude φ\varphi, then, after such the transformation, the parameterization (5) becomes exact.

At last, the latter term in Lagrangian (2) is responsible for potential energy and includes the variable bb denoting the relative buoyancy. When the deviations of density and temperature are related linearly, it is possible to determine this quantity alternatively as

b=gΔϱϱ0=−gΔθθ0,(7)b = \frac{g\Delta\varrho}{\varrho_0} = -\frac{g\Delta\theta}{\theta_0}, \tag*{(7)}

where gg is gravity, Δϱ(x,y,t)\Delta\varrho(x,y,t) and Δθ(x,y,t)\Delta\theta(x,y,t) are the density and temperature deviations from the respective reference values ϱ0\varrho_0 and θ0\theta_0.

Depending on which is the description of the fluid continuum, there are mainly two ways to apply the principle of least action to similar types of models. One of these ways relies on the Eulerian conception of fluid motion and uses Lin’s constraints, while the other, called Lagrangian, requires tracking individual fluid particles.

In the context of depth-averaged fluid models, a pivotal point of such an approach is to interpret flow as the motion of infinitely narrow liquid columns whose Cartesian positions

x=x^(s,n,t),y=x^(s,n,t)(8)x=\hat{x}(s,n,t),\qquad y=\hat{x}(s,n,t) \tag*{(8)}

are functions of time tt and labeling coordinates (s,n)(s,n) of the special kind. Being assigned so that equal infinitesimal areas in (s,n)(s,n)-space correspond to the same volumes, these coordinates, sometimes termed straightened14, obey the condition

H ds dn=h dx dy,(9)H\,ds\,dn=h\,dx\,dy, \tag*{(9)}

in which HH is a constant and means the average layer depth.

Since, apart from (9), the transformation (8) entails the rules both for velocities

u→x^t,v→y^t,(10)u\to\hat{x}_t,\qquad v\to\hat{y}_t, \tag*{(10)}

and also for buoyancy and layer height

b→B(s,n),h→HJ(x^,y^),(11)b\to B(s,n),\qquad h\to\frac{H}{J(\hat{x},\hat{y})}, \tag*{(11)}

where JJ is the transform Jacobian

J(x^,y^)=x^sy^n−x^ny^s,(12)J(\hat{x},\hat{y})=\hat{x}_s\hat{y}_n-\hat{x}_n\hat{y}_s, \tag*{(12)}

we find that the Lagrangian (2) written in terms of straightened variables looks as

L=H2∫(x^t2+y^t2−2Ωax^y^2−HBJ(x^,y^)) ds dn.(13)\mathcal{L}=\frac{H}{2}\int\left(\hat{x}_t^2+\hat{y}_t^2-\frac{2\Omega}{a}\hat{x}\hat{y}^2-\frac{HB}{J(\hat{x},\hat{y})}\right)\,ds\,dn. \tag*{(13)}

Scaling and perturbation theory on equatorial beta-plain

In any geophysical context, the theory of thin jets has an approximate character and, as a rule, for one’s construction, requires the use of the multiscale expansions method. Because of this, to more effectively apply perturbation theory and to control all small parameters, the best way is to convert the Lagrangian (13) to a dimensionless form.

For frontal jets caused by temperature gradients and formed by the influence of gravity and rotation, it is natural to take the following quantities

T=12Ω,L=T(gHΔθ∗θ0)1/2,(14)T = \frac{1}{2\Omega}, \qquad L = T\left(gH\frac{\Delta\theta^{*}}{\theta_0}\right)^{1/2}, \tag*{(14)}

as the typical scales of time and horizontal length, respectively. Here HH denotes the layer depth reached far from the jet in a quiescent region, and Δθ∗\Delta\theta^{*} is the peak value of temperature deviation. In particular, in conditions typical of the Earth’s atmosphere, where

H≈5 km,g≈10 m⋅s−2,Ω≈7.3⋅10−5 s−1,Δθ∗/θ0≈5⋅10−3,(15)H \approx5\ \mathrm{km}, \qquad g \approx10\ \mathrm{m}\cdot\mathrm{s}^{-2}, \qquad\Omega\approx7.3\cdot10^{-5}\ \mathrm{s}^{-1}, \qquad\Delta\theta^{*}/\theta_0 \approx5\cdot10^{-3}, \tag*{(15)}

the definition (14) allows us to estimate the Rossby length scale as L≈100 kmL \approx100\ \mathrm{km}.

Now, upon transforming by the rules

(x^,y^,s,n)→L(x^,y^,s,n),t→Tt,(16)(\hat{x},\hat{y},s,n)\to L(\hat{x},\hat{y},s,n), \qquad t\to Tt, \tag*{(16)}
B→BL2T2H,L→LL4T2,(17)B\to B\frac{L^2}{T^2H}, \qquad\mathcal{L}\to\mathcal{L}\frac{L^4}{T^2}, \tag*{(17)}

we can reduce the Lagrangian (13) to the dimensionless form

L=12∫(x^t2+y^t2−δx^y^t2−BJ(x^,y^)) ds dn.(18)\mathcal{L}=\frac{1}{2}\int\left(\hat{x}_t^2+\hat{y}_t^2-\delta\hat{x}\hat{y}_t^2-\frac{B}{J(\hat{x},\hat{y})}\right)\,ds\,dn. \tag*{(18)}

Here δ=L/a<1\delta=L/a<1 is the parameter responsible for the beta effect.

Let the jet be sufficiently thin and localized at the zero-buoyancy line described parametrically as

x=X(s,t),y=Y(s,t).(19)x=X(s,t), \qquad y=Y(s,t). \tag*{(19)}

The thin-jets approximation implies an expansion of the field variables in the vicinity of the jet trajectory, and it assumes that the dependent and independent variables appearing in the Lagrangian (18) have the following orders:

s∼x^∼y^∼1/ϵ,n∼B∼1,t∼1/λ,(20)s\sim\hat{x}\sim\hat{y}\sim1/\epsilon,\qquad n\sim B\sim1,\qquad t\sim1/\lambda, \tag*{(20)}

where ϵ\epsilon and λ\lambda are two small perturbation parameters. How they relate to each other and the parameter δ\delta will be found later. Note that, although the smallness of ϵ\epsilon is arbitrary, it will be measured as the ratio ϵ∼l/L\epsilon\sim l/L, where ll is the width of the jet.

Then, expanding the functions x^\hat{x} and y^\hat{y} by powers of nn and ϵ\epsilon, one can find that

x^=1ϵ(X+ϵnX1+ϵ2n2X2+O(ϵ3)),(21)\hat{x} = \frac{1}{\epsilon}\left(X+\epsilon nX_1+\epsilon^2n^2X_2+O(\epsilon^3)\right), \tag*{(21)}
y^=1ϵ(Y+ϵnY1+ϵ2n2Y2+O(ϵ3)),(22)\hat{y} = \frac{1}{\epsilon}\left(Y+\epsilon nY_1+\epsilon^2n^2Y_2+O(\epsilon^3)\right), \tag*{(22)}

where X1(s,t)X_1(s,t), Y1(s,t)Y_1(s,t) are the displacements in the first order of perturbation theory, and X2(s,t)X_2(s,t), Y2(s,t)Y_2(s,t) are the displacements in the second order.

The expansion for buoyancy we can find analogously. The fact that along-jet velocities localize near zero-buoyancy lines and reach their peak values therein allows us to conclude that the function B(s,n)B(s,n) must be odd relative to nn, so its expansion must begin with a linear term

B=nγ+O(ϵ2).(23)B=n\gamma+O(\epsilon^2). \tag*{(23)}

Here the coefficient γ(s)\gamma(s) characterizing the across-jet temperature gradient (or the along-jet velocity) can depend only on the longitudinal coordinate ss.

The Jacobian JJ is one more quantity that we should expand into a power series in ϵ\epsilon. Substituting expansions (21) and (22) into the Jacobian (12), it is easy to find

J=J0+nϵJ1+O(ϵ2),(24)J=J_0+n\epsilon J_1+O(\epsilon^2), \tag*{(24)}
J0=XsY1−YsX1,(25)J_0=X_sY_1-Y_sX_1, \tag*{(25)}
J1=X1sY1−Y1sX1+2(XsY2−YsX2).(26)J_1=X_{1s}Y_1-Y_{1s}X_1+2\left(X_sY_2-Y_sX_2\right). \tag*{(26)}

Now, in analogy with [15], we can find the thin-jet Lagrangian in the equatorial beta-plane approximation. By inserting the expansions (21)–(26) directly into (18), integrating it over −1≤n≤1-1\le n\le1, and keeping the expansion terms up to the first order of ϵ\epsilon, λ\lambda, and δ\delta we arrive at the expression

L=∫(−λδϵ3XtY2+ϵ3γJ1J02)ds+O(ϵ,λ).(27)\mathcal{L}=\int\left(-\frac{\lambda\delta}{\epsilon^3}X_tY^2+\frac{\epsilon}{3}\gamma\frac{J_1}{J_0^2}\right)ds+O(\epsilon,\lambda). \tag*{(27)}

To obtain a closed description in the leading order of perturbation theory, we need to require that both terms in the integral (27) have the same order. Whence, by comparing them, we get

λ=ϵ43δ.(28)\lambda=\frac{\epsilon^4}{3\delta}. \tag*{(28)}

In addition, following [15], we have to impose the conditions

XsX1+YsY1=0,XsY2−YsX2=0,(29)X_sX_1+Y_sY_1=0,\qquad X_sY_2-Y_sX_2=0, \tag*{(29)}

which say that first-order displacements are orthogonal to the jet trajectory while second-order ones are parallel. Note that these conditions provide self-consistency of the perturbation theory so that it would not contain any second-order displacements in the first-order perturbations for the Jacobian JJ.

With the use of the relation (25) and the orthogonality condition from (29), one can find

X1=−J0YsXs2+Ys2,Y1=J0XsXs2+Ys2,J1=J02YsXss−XsYss(Xs2+Ys2)2.(30)X_1=-\frac{J_0Y_s}{X_s^2+Y_s^2},\qquad Y_1=\frac{J_0X_s}{X_s^2+Y_s^2},\qquad J_1=J_0^2\frac{Y_sX_{ss}-X_sY_{ss}}{(X_s^2+Y_s^2)^2}. \tag*{(30)}

As a result, after excluding J1J_1 from the Lagrangian (27), it reduces to a more simple form

L=∫(−XtY2+γYsXss−XsYss(Xs2+Ys2)2) ds.(31)\mathcal{L}=\int\left(-X_tY^2+\gamma\frac{Y_sX_{ss}-X_sY_{ss}}{(X_s^2+Y_s^2)^2}\right)\,ds. \tag*{(31)}

Path equations, some symmetries and solutions

Formulated based on the variational principle (1) with the Lagrangian (31), the equations of equatorial dynamics for temperature jets already externally look different than those derived for a planar version of this theory[15]. Formally, the equatorial version follows directly from the planar one, at once after substituting Xt→XtYX_t\to X_tY and Yt→YtYY_t\to Y_tY into the left-hand sides of planar version equations. In particular, when the temperature gradient γ\gamma across the jet is assumed to be constant, this trick leads us to the path equations

XtY=∂∂s(Xss(Xs2+Ys2)2),YtY=∂∂s(Yss(Xs2+Ys2)2),(32)X_tY=\frac{\partial}{\partial s}\left(\frac{X_{ss}}{(X_s^2+Y_s^2)^2}\right),\qquad Y_tY=\frac{\partial}{\partial s}\left(\frac{Y_{ss}}{(X_s^2+Y_s^2)^2}\right), \tag*{(32)}

whose symmetries are essentially different from those inherent to the planar version.

One can easily check that Eqs. (32) lose some of those symmetries, due to which they remained invariant in the planar version. For example, they become no invariant under translations in YY, rotations, and inversions. Instead, Eqs. (32) gain new symmetry properties

X→X+X0,(33)X\to X+X_0, \tag*{(33)}
X(s,t)→X(−s,t),Y(s,t)→−Y(−s,t).(34)X(s,t)\to X(-s,t),\qquad Y(s,t)\to-Y(-s,t). \tag*{(34)}

From all the integrals of motion found for the planar version[15], the equatorial version (32) leaves invariant only the energy integral

E=12∫YssXs−XssYs(Xs2+Ys2)2 ds.(35)E=\frac{1}{2}\int\frac{Y_{ss}X_s-X_{ss}Y_s}{(X_s^2+Y_s^2)^2}\,ds. \tag*{(35)}

But additionally, it provides time invariance of two more integrals:

I1=∫Y2 ds,I2=∫XsY2 ds.(36)I_1 = \int Y^2\,ds,\qquad I_2 = \int X_sY^2\,ds. \tag*{(36)}

There is a fundamental difference between how the beta effect impacts the dynamics of temperature jets and how it influences the motion of PV jets generated by potential vorticity jumps. While, for the PV jets[13], the beta-effect leads to Doppler’s shift of the phase velocity, its influence on temperature jets turns out to be more nontrivial.

Unfortunately, it is currently unknown whether the Eqs. (32) are exactly integrable. But despite this, they admit a wide variety of steadily propagating solutions, including spatially periodic waves. Herewith, the beta-effect influence manifests itself most brightly in the soliton-type solutions. In particular, due to symmetry (34), the soliton solutions for equatorial jets are not unipolar. The sign of their amplitude depends on their motion direction. Another remarkable feature of temperature jets is, owing to the beta effect, they can evolve in the shape of loop solitons (both with- and non-intersection). As known in geophysical fluid dynamics[8, 11], the solutions of such type can be responsible for the fluid detachment from jet flows and are used to explain the formation of warm/cold rings in the Gulf Stream.

Consider the solutions of Eqs. (32) within a specific class. Quite obviously, in the case X=sX=s, i.e., when ss coincides with the axis XX, these equations reduce to one single

YtY=∂∂XYXX(1+YX2)2,(37)Y_tY = \frac{\partial}{\partial X}\frac{Y_{XX}}{(1+Y_X^2)^2}, \tag*{(37)}

with the variable YY to be considered as a function of XX and tt.

Note that being scale-invariant under the transformation

(X,Y,t)→(αX,αY,α4t),(38)(X,Y,t) \to(\alpha X,\alpha Y,\alpha^4t), \tag*{(38)}

with rescaling factor α\alpha, Eq. (37), in particular, conserves the integrals

Q1=∫Y2 dX,Q2=∫XY2 dX,Q3=∫Y3 dX.(39)Q_1 = \int Y^2\,dX,\qquad Q_2 = \int XY^2\,dX,\qquad Q_3 = \int Y^3\,dX. \tag*{(39)}

Let us look for the steadily translating solutions Y(σ)Y(\sigma), (σ=X−ct\sigma= X-ct), propagating with a constant velocity cc without profile deformation along the XX axis. Then, after twofold integration, scaling, and the introduction of the calibration parameter κ\kappa

y=YY0,x=σY0,κ=3cY03,(40)y = \frac{Y}{Y_0},\qquad x = \frac{\sigma}{Y_0},\qquad\kappa= \frac{3}{cY_0^3}, \tag*{(40)}

which links the amplitude Y0Y_0 and phase velocity cc, Eq. (37) can be reduced to the form

yx2=−Pκ+P,P=y3+c1y+c2.(41)y_x^2=\frac{-P}{\kappa+P},\qquad P=y^3+c_1y+c_2. \tag*{(41)}

Here and below, PP is an incomplete cubic polynomial, and c1c_1 and c2c_2 are integration constants chosen to satisfy the appropriate boundary conditions at infinity.

Cusped jets

Limiting ourselves to the so-called solitary solutions, we first consider those to have a power decay at infinity. One can show that such solutions become possible if c1=c2=0c_1=c_2=0. In this case, assuming without any loss of generality κ=−1\kappa=-1, Eq. (41) takes the form

yx2=y31−y3.(42)y_x^2=\frac{y^3}{1-y^3}. \tag*{(42)}

Solutions of this equation vary in the range 0<y≤10<y\leq1 and, being translational invariant, depend on xx with the accuracy of an additive constant omitted next for brevity. In implicit form, the general solution of (42) can be written as

x=±B(56,12)(1−y−1/2P1/2(−16,−32)(1−2y3)),(43)x=\pm B\left(\frac{5}{6},\frac{1}{2}\right)\left(1-y^{-1/2}P_{1/2}^{\left(-\frac{1}{6},-\frac{3}{2}\right)}\left(1-2y^3\right)\right), \tag*{(43)}

through the fractional Jacobi function[16]. To find the phase velocity cc of such solutions, one has to use the third relation from (40), whence, at κ=−1\kappa=-1, it follows the dispersion relation

c=−3/Y03,(44)c=-3/Y_0^3, \tag*{(44)}

for their phase velocity.

Note that found solutions are cusped (see Figure 1). At infinity, they vanishes by the power law y∝x−2y\propto x^{-2}, but at peak point, x=0x=0, they take the finite value y=1y=1 and near it behave like 1−y∝x2/31-y\propto x^{2/3}. Such behavior of cusped solutions cause their derivatives at these points to have power asymptotics

Figure 1. The cusped-jet solution.

yx∣x→0∝x−1/3,yx∣x→∞∝x−3.(45)\left.y_x\right|_{x\to0}\propto x^{-1/3},\qquad\left.y_x\right|_{x\to\infty}\propto x^{-3}. \tag*{(45)}

Loop jets

Another class of steady-state solutions admissible for Eq. (41) are the so-called solitons. Given their specific behavior, we will assume that they reach a maximum of y=1y=1 at x=0x=0, while, at infinity, exponentially decrease as e−∣x∣/λe^{-|x|/\lambda}, approaching there a constant value. A simple analysis shows that such solutions are possible only if c1=−3/4c_1=-3/4, c2=−1/4c_2=-1/4, and imply the fulfillment of the following conditions:

P=(y−1)(y+12)2,−12≤y≤1,κ=32λ2≥12.(46)P=(y-1)\left(y+\frac{1}{2}\right)^2,\quad-\frac{1}{2}\le y\le1,\quad\kappa=\frac{3}{2}\lambda^2\ge\frac{1}{2}. \tag*{(46)}

Here the last equality can be viewed as a relation between the parameter κ\kappa and the length scale λ\lambda that defines how strongly localized the solution is. If we take advantage of the proper equalities in Eqs. (40) and (46), then excluding κ\kappa from them leads us to the dispersion relation

c=2λ−2Y0−3.(47)c=2\lambda^{-2}Y_0^{-3}. \tag*{(47)}

It demonstrates that the phase velocity for solitons depends not only on their amplitude, as in the case of cuspons, but also on their localization length.

Thus, with the account of (46) and after replacing

y=cos⁡ψ,x=x(ψ),∣ψ∣≤2π/3,(48)y=\cos\psi,\quad x=x(\psi),\quad|\psi|\le2\pi/3, \tag*{(48)}

Eq. (41) looks like

xψ2=sin⁡2ψ[2κcsc⁡2(3ψ2)−1].(49)x_\psi^2=\sin^2\psi\left[2\kappa\csc^2\left(\frac{3\psi}{2}\right)-1\right]. \tag*{(49)}

As shown in Figure 2, this equation has a family of one-valued solitary solutions only under the condition κ>1/2\kappa>1/2. The multi-valued solitary solutions arise in the limiting case $\kappa=1/2` when Eq. (49) takes the form

Figure 2. Single-valued solitons for the sequentially increasing parameter κ\kappa, running values 0.5, 0.65, 1, 2, 8. As κ\kappa increases, the solitons decrease with distance more and more slowly. The limiting value κ=0.5\kappa= 0.5 corresponds to the red curve. The dashed curve corresponds to the loop soliton with the intersection.

xψ2=sin⁡2ψcot⁡2(3ψ2).(50)x_{\psi}^{2} = \sin^{2}\psi\cot^{2}\left(\frac{3\psi}{2}\right). \tag*{(50)}

In this instance, as we shall see below, it is integrated analytically. However, the final result will depend on choosing the so-called signum function that arises in extracting the square root.

To begin with, let us consider the simplest multi-valued soliton solution constructed without using any signum functions:

x=F(ψ),y=cos⁡ψ,(51)x = F(\psi), \quad y = \cos\psi, \tag*{(51)}
F(ψ)=23arcth⁡(13tan⁡(ψ2))−sin⁡ψ.(52)F(\psi) = \frac{2}{\sqrt{3}}\operatorname{arcth}\left(\frac{1}{\sqrt{3}}\tan\left(\frac{\psi}{2}\right)\right) - \sin\psi. \tag*{(52)}

Here F(ψ)F(\psi) is the particular solution of Eq. (50) which, as shown in Figure 2, corresponds to the so-called loop soliton with self-intersection.

If we want to find loop solitons free of self-intersections, we must construct the relevant signum functions

S+=sign⁡(ψ+π3),S−=sign⁡(ψ−π3),(53)S_{+} = \operatorname{sign}\left(\psi+ \frac{\pi}{3}\right), \quad S_{-} = \operatorname{sign}\left(\psi- \frac{\pi}{3}\right), \tag*{(53)}

whose zeros ψ=±π/2\psi= \pm\pi/2 coincide with the zeros of the derivative dF/dψdF/d\psi. As a result, we get three possible solutions

x=x±=S±(F(π3)±F(ψ)),x=F(ψ)−x+−x−.(54)x = x_{\pm} = S_{\pm}\left(F\left(\frac{\pi}{3}\right) \pm F(\psi)\right), \qquad x = F(\psi) - x_{+} - x_{-}. \tag*{(54)}

Since differentiating signum functions over ψ\psi leads to delta-functions, it is easy to check that all solutions listed in (54) and illustrated in Figure 3 satisfy Eq. (50). Note also that, unlike the first two solutions, the last solution is one-valued and coincides with (52). Figure 2 and Figure 3 show it in red.

Figure 3. The loop solitons without self-intersection. The blue curve corresponds to the x+x_{+}-solution, the yellow curve to the x−x_{-}-solution, and the red curve to the xx-solution of Eq. (50).

Discussion and conclusions

In this paper, based on the variational principle formulated for the theory of temperature-frontal jets, the planar version[15] of this theory gets extended to include the beta effect. In contrast to jets on potential vorticity jumps[13], where the beta effect results only in a Doppler shift, for jets on temperature jumps, we found that the beta effect leads to the fundamental changes of dynamical symmetries in the model. As a result, the corresponding path equations admit a new class of solutions. In particular, as our analysis has shown, among them turn out equatorial jets in the form of cuspons and loop solitons.

As follows from (44) and (47), the phase velocities of both solutions have the same dependence on the amplitude and, being opposite in sign, are in the relation 2∣csol∣≤∣ccusp∣2|c_{\mathrm{sol}}| \le|c_{\mathrm{cusp}}|, where equality is achieved on loop solitons. Based on conditions (15) the velocity estimations at Y0≈L≈100 kmY_0 \approx L \approx100\ \mathrm{km} give us the values ∣ccusp∣≈6 m⋅s−1|c_{\mathrm{cusp}}| \approx6\ \mathrm{m}\cdot\mathrm{s}^{-1} and ∣cloop∣≈12 m⋅s−1|c_{\mathrm{loop}}| \approx12\ \mathrm{m}\cdot\mathrm{s}^{-1}, respectively for cusped and loop solutions.

Unfortunately, due to the intense mixing and turbulence in the real atmosphere, these solutions are almost impossible to identify with sufficient reliability in experimental in-situ data or direct satellite observations. However, if we assume that some of these solutions can play the role of structural elements of large-scale turbulence, this hypothesis can be tested by indirect methods, both analytically and numerically.

One way to implement this approach is to consider large-scale turbulence as a collective effect provided by a statistical ensemble constructed from regular solutions found analytically. In so doing, the free parameters of such solutions should be treated as random variables. It is worth noting that similar approaches have been applied in Refs.[17, 18, 19, 20, 21, 22] but in other contexts.

In the spirit of these ideas, it can be demonstrated that, at the expense of equatorial temperature jets, the beta effect may be responsible for modifying the spectral properties of turbulence in the near equator. In particular, as shown in Appendix § 0.4.2 (see also Figure 4), even neglecting trajectory dynamics of temperature jets, their impact on the spectral slope in the short-wave range can be sufficiently strong.

Figure 4. The energy spectrum of turbulence created by equatorial temperature jets. The dashed lines on the log-log plot show power asymptotes. In the lower plot, the equatorial beta effect is off.

Acknowledgments

This work was supported by the Russian Science Foundation, 23-17-00273.

Temperature jets on rotational sphere

Because inertial effects play a secondary role in the dynamics of thin temperature jets, to extend the theory to a rotational sphere, let us start at once with the truncated Lagrangian

L=a22∫h(2a2Ωdϑdtcos⁡2φ−hb)cos⁡φ dφ dϑ,(55)\mathcal{L} = \frac{a^{2}}{2} \int h \left(2a^{2}\Omega\frac{\mathrm{d}\vartheta}{\mathrm{d}t}\cos^{2}\varphi- hb\right)\cos\varphi\,\mathrm{d}\varphi\,\mathrm{d}\vartheta, \tag*{(55)}

ignoring all other effects except Coriolis and buoyancy. Here, aa is the radius of the sphere, hh is the thickness of an atmospheric layer on it, φ\varphi and ϑ\vartheta are the geographic coordinates (latitude ∣φ∣≤π/2|\varphi| \le\pi/2 and longitude ∣ϑ∣≤π|\vartheta| \le\pi), Ω\Omega is the rotational angular speed, bb is the relative buoyancy. In the Eulerian description, the dynamic variables hh, bb, and the azimuthal velocity dϑ/dt\mathrm{d}\vartheta/\mathrm{d}t are treated as functions of the spherical coordinates, φ\varphi and ϑ\vartheta, and time tt.

The Lagrangian (A1) describes depth-averaged (columnar) motions of an incompressible fluid with variable buoyancy that covers the rotating sphere by a thin layer (h≪ah \ll a) and is under the action of gravity and Coriolis forces. To formulate the variational principle similar to (1), we must free ourselves from constraints imposed implicitly in the Eulerian formulation. To do this, let us pass from the Eulerian description to the Lagrangian one, considering the geographic coordinates of fluid columns

ϑ=ϑ^(s,n,t),φ=φ^(s,n,t),(56)\vartheta= \widehat{\vartheta}(s,n,t), \qquad\varphi= \widehat{\varphi}(s,n,t), \tag*{(56)}

as functions of curvilinear labeling coordinates, ss and nn, and time tt.

Since the motion of such a fluid is assumed incompressible and without dissipation, each infinitesimal fluid column must preserve its volume and buoyancy. So, the Lagrangian (A1) must be consistent with the following constraints

∂t(h^J(ϑ^,φ^)cos⁡φ^)=0,J(ϑ^,φ^)=ϑ^sφ^n−ϑ^nφ^s,b^t=0,(57)\partial_{t}\left(\widehat{h}J(\widehat{\vartheta},\widehat{\varphi})\cos\widehat{\varphi}\right)=0,\qquad J(\widehat{\vartheta},\widehat{\varphi})=\widehat{\vartheta}_{s}\widehat{\varphi}_{n}-\widehat{\vartheta}_{n}\widehat{\varphi}_{s},\qquad\widehat{b}_{t}=0, \tag*{(57)}

where J(ϑ^,φ^)J(\hat{\vartheta},\hat{\varphi}) denotes the Jacobian for the transformation (A2), and the hat symbol indicates that arguments of hatted functions are taken at ϑ=ϑ^\vartheta=\hat{\vartheta}, φ=φ^\varphi=\hat{\varphi}.

Due to the arbitrariness of choosing the labeling coordinates ss and nn, it is convenient to assign them so that equal areas on the sphere contain equal volumes:

H ds dn=hcos⁡φ dφ dϑ,(58)H\,ds\,dn=h\cos\varphi\,d\varphi\,d\vartheta, \tag*{(58)}

where HH is some constant. Then, after integrating Eqs. (57) over time, we obtain that

h^=HJ(ϑ^,sin⁡φ^),b^=B(s,n).(59)\hat{h}=\frac{H}{J(\hat{\vartheta},\sin\hat{\varphi})},\qquad\hat{b}=B(s,n). \tag*{(59)}

Here, as before, in § 0.2, the constant HH means the average layer depth, while the time-independent function BB defines the initial buoyancy distribution.

Now, it is possible to apply the transformation (A2) directly to the Lagrangian (A1) and make it free from constraints. Using the relations (A5), and dϑ/dt=ϑ^td\vartheta/dt=\hat{\vartheta}_t, as well as given that the Lagrangian is defined only with accuracy to total time derivative of an arbitrary function, we find

L=−Ha22∫(2a2Ωϑ^tsin⁡2φ^+HBJ(ϑ^,sin⁡φ^)) ds dn.(60)\mathcal{L}=-\frac{Ha^2}{2}\int\left(2a^2\Omega\hat{\vartheta}_t\sin^2\hat{\varphi}+\frac{HB}{J(\hat{\vartheta},\sin\hat{\varphi})}\right)\,ds\,dn. \tag*{(60)}

It is easy to see that if we introduce notations

x^=aϑ^,y^=asin⁡φ^,(61)\hat{x}=a\hat{\vartheta},\qquad\hat{y}=a\sin\hat{\varphi}, \tag*{(61)}

and renormalize the labeling coordinates s→ass\to as, n→ann\to an, then the Lagrangian (A6) takes the form

L=−H2∫(2Ωax^ty^2+HBJ(x^,y^)) ds dn,(62)\mathcal{L}=-\frac{H}{2}\int\left(\frac{2\Omega}{a}\hat{x}_t\hat{y}^2+\frac{HB}{J(\hat{x},\hat{y})}\right)\,ds\,dn, \tag*{(62)}

and coincides with the Lagrangian (13) in the inertial-free approximation.

This fact means that the requirement ∣y/a∣≪1|y/a|\ll1, traditional for equatorial models when choosing the Coriolis parameterization (4), is redundant, in our case, and can be replaced by a much softer condition ∣y/a∣≤1|y/a|\le1.

Turbulence of equatorial straight-line jets

Let us consider turbulence occurring in the narrow equatorial belt as a collective effect to be created by a statistical ensemble of thin temperature jets. To exclude the influence of all other factors except the beta effect, we restrict ourselves to the simplest model that ignores the bending of the jets and treats them as straight lines parallel to the equator.

The random field of velocity produced by such jets can be found from the condition of equatorial geostrophic balance, putting, for simplicity, the layer depth hh is constant. Then, being written in a dimensionless form, this condition looks like the relation

εyu=−12by,(63)\varepsilon y u = -\frac{1}{2}b_y, \tag*{(63)}

between the zonal velocity component uu and the derivative of buoyancy bb over the meridional coordinate yy. The small parameter ε=L/a\varepsilon= L/a results from scaling to be expressed through the characteristic scale LL and the Earth’s radius aa.

Since the Fourier transform is traditional for studying spectra, it is reasonable to use it for reformulating (B1) as the boundary value problem

uk=−14πε∫−∞∞bye−iky dy,u∣k=0=∫−∞∞u(y) dy=0.(64)u_k = -\frac{1}{4\pi\varepsilon}\int_{-\infty}^{\infty} b_y e^{-iky}\,\mathrm{d}y,\qquad\left.u\right|_{k=0} = \int_{-\infty}^{\infty}u(y)\,\mathrm{d}y = 0. \tag*{(64)}

Thus the Fourier amplitude of velocity uu treated as a function of the wave number kk is subject to the first-order linear differential equation and the boundary condition. It is worth noting that this condition follows directly from the ergodicity hypothesis (equivalence of spatial and statistical averaging) and the assumption of the turbulence in question being a zero-mean random process.

By considering the sum of step functions as the buoyancy distribution

b=∑iγiθ(y−yi),(65)b = \sum_i \gamma_i\theta\left(y-y_i\right), \tag*{(65)}

and putting it into (B2), after some algebra, we get

u=12πε∑iγiyisin⁡(kyi2)e−ikyi/2.(66)u = \frac{1}{2\pi\varepsilon}\sum_i\frac{\gamma_i}{y_i}\sin\left(\frac{ky_i}{2}\right)e^{-iky_i/2}. \tag*{(66)}

This expression describes the Fourier amplitude of the velocity field produced by an equatorial ensemble of thin temperature jets. Free model parameters, which include buoyancy jumps γi\gamma_i and their distances yiy_i from the equator, are listed by the index ii and, below, play the role of independent random variables.

Now, to calculate the energy density spectrum E(k)E(k), it is necessary to use the standard procedure

E=12⟨∣u(k)∣2⟩,(67)E = \frac{1}{2}\left\langle\lvert u(k)\rvert^2\right\rangle, \tag*{(67)}

where angle brackets imply averaging.

In particular, when yiy_i has a uniform distribution in the narrow belt ∣yi∣≤1|y_i| \le1, the calculation gives the result

E=N⟨γ2⟩2ε2∫011y2sin⁡2(12ky) dy∝cos⁡k−1+kSi⁡(k),(68)E = \frac{N\langle\gamma^2\rangle}{2\varepsilon^2}\int_0^1 \frac{1}{y^2}\sin^2\left(\frac{1}{2}ky\right)\,dy \propto\cos k - 1 + k\operatorname{Si}(k), \tag*{(68)}

where NN is the mean number of jets per unit width of the equatorial belt, ⟨γ2⟩\langle\gamma^2\rangle is the mean square magnitude of buoyancy jumps, and Si⁡(k)\operatorname{Si}(k) is the integral sinus.

Thus, in the case of the equatorial beta effect, the energy density spectrum has two parts with different power laws. At small wave numbers, if k≪1k \ll1, the spectral density behaves as E∝k2E \propto k^2, and at large ones, if k≫1k \gg1, as E∝kE \propto k.

In conclusion, we note that turning off the equatorial beta effect, in other conditions being equal, leads to a rearrangement of the spectrum (B6), so after that, it converts in

E∝1−sin⁡kk.(69)E \propto1 - \frac{\sin k}{k}. \tag*{(69)}

Because of this, as shown in Figure 4, the short-wave part of the spectrum loses its slope.

References

  1. [1]¹B. Cushman-Roisin and J. M. Beckers, Introduction to Geophysical Fluid Dynamics: Physical and Numerical Aspects, Series in International Geophysics. Academic Press, 2nd edition, 2011. https://books.google.ru/books?id=3KS8sD5ky9kC
  2. [2]²P. Ripa, “Conservation laws for primitive equations models with inhomogeneous layers,” Geophys. Astrophys. Fluid Dynam. 70, 85–111 (1993). https://doi.org/10.1080/03091929308203588
  3. [3]³P. Ripa, “On improving a one-layer ocean model with thermodynamics,” J. Fluid Mech. 303, 169–201 (1995). https://doi.org/10.1017/S0022112095004228
  4. [4]⁴E. S. Warneford and P. J. Dellar, “The quasi-geostrophic theory of the thermal shallow water equations,” J. Fluid Mech. 723, 374–403 (2013). https://doi.org/10.1017/jfm.2013.101
  5. [5]⁵A. R. Robinson and P. P. Niler, “The theory of free inertial jets: I. Path and structure,” Tellus 19, 269-291 (1967). https://doi.org/10.3402/tellusa.v19i2.9788
  6. [6]⁶J.R. Luyten and G. Flierl, “On the theory of thin rotating jets: A quasi-geostrophic time-dependent mode,” Geophys. Fluid Dyn. 6, 211-244 (1975). https://doi.org/10.1080/03091927509365796
  7. [7]⁷G.R. Flierl and A.R. Robinson, “On the time dependent meandering of a thin jet,” J. Phys. Oceangr. 14, 412-443, 1984. https://doi.org/10.1175/1520-0485(1984)014<0412:OTTDMO>2.0.CO;2
  8. [8]⁸L.J. Pratt and M.E. Stern, “Dynamics of potential vorticity fronts and eddy detachment,” J. Phys. Oceangr. 16, 1101–1120 (1986). https://doi.org/10.1175/1520-0485(1986)016<1101:DOPVFA>2.0.CO;2
  9. [9]⁹L.J. Pratt, “Meandering and eddy detachment according to a simple (looking) path equation,” J. Phys. Oceangr. 18, 1627-1640 (1988). https://doi.org/10.1175/1520-0485(1988)018<1627:MAEDAT>2.0.CO;2
  10. [10]¹⁰J. Nycander, D.G. Dritschell and G.G. Sutyrin, “The dynamics of long frontal waves in the shallow water equations,” Phys. Fluids A, 5, 1089-1091 (1993). https://doi.org/10.1063/1.858592
  11. [11]¹¹E.A. Ralph and L.J. Pratt, “Predicting eddy detachment for an equivalent barotropic thin jet,” J. Nonlinear Sci. 4, 355–374 (1994). https://doi.org/10.1007/BF02430638
  12. [12]¹²B. Cushman-Roisin, J.A. Proehl and D.T. Morgan, ”Barotropic thin jets over arbitrary topography,” Dynam. Atmos. Oceans 26, 73-93 (1997). https://doi.org/10.1016/S0377-0265(96)00484-8
  13. [13]¹³B. Cushman-Roisin, L.J. Pratt and E.A. Ralph, “A general theory for equivalent barotropic thin jets,” J. Phys. Oceangr. 23, 91–103 (1993). https://doi.org/10.1175/1520-0485(1993)023<0091:AGTFEB>2.0.CO;2
  14. [15]¹⁵V.P. Goncharov, “Dynamics of thin jets generated by temperature fronts,” Phys. Rev. Fluids 6, 103801(14) (2021). https://link.aps.org/doi/10.1103/PhysRevFluids.6.103801DOI
  15. [16]¹⁶E. Gogovcheva and L. Boyadjiev, “Fractional Extensions of Jacobi Polynomials and Gauss Hypergeometric Function,” Fract Calc Appl Anal. 8, 431-438 (2005). https://www.researchgate.net/publication/228544128
  16. [17]O. M. Phillips, “The equilibrium range in the spectrum of wind-generated waves,” J. Fluid Mech. 4, 426–434 (1958). https://doi:10.1017/S0022112058000550DOI
  17. [18]P. G. Saffman, “On the spectrum and decay of random 2D vorticity distributions at large Reynolds number,” Stud. Appl. Math. 50, 377 (1971). https://doi.org/10.1002/sapm1971504377
  18. [19]B. B. Kadomtsev and V. I. Petviashvili, “On acoustic turbulence,” Sov. Phys. Dokl. 18, 115 (1973). Dokl. Akad. Nauk SSSR, 208, 794–796 (1973) https://www.mathnet.ru/eng/dan37430
  19. [20]E. A. Kuznetsov, “Turbulence spectra generated by singularities,” JETP Lett. 80, 83–89 (2004). https://doi.org/10.1134/1.1804214
  20. [21]E. A. Kuznetsov, V. Naulin, A. H. Nielsen, and J. J. Rasmussen, “Effects of sharp vorticity gradients in two-dimensional hydrodynamic turbulence,” Phys Fluids 19, 105110 (2007). https://doi.org/10.1063/1.2793150
  21. [22]V. P. Goncharov, V. I. Pavlov. “Whether the Sun’s supergranulation possesses a scaling?” Jetp Lett. 99, 317-322 (2014). https://doi.org/10.1134/S002136401406006X

Paper details

vlnc:W_zudgdbrh
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents