Nonlinear viscous fluid patterns in a thin rotating spherical domain
by user
Comments
Transcript
Nonlinear viscous fluid patterns in a thin rotating spherical domain
Nonlinear viscous fluid patterns in a thin rotating spherical domain and applications Ranis N. Ibragimov Citation: Phys. Fluids 23, 123102 (2011); doi: 10.1063/1.3665132 View online: http://dx.doi.org/10.1063/1.3665132 View Table of Contents: http://pof.aip.org/resource/1/PHFLE6/v23/i12 Published by the American Institute of Physics. Related Articles Vortex motion around a circular cylinder Phys. Fluids 23, 123601 (2011) Inertial, barotropic, and baroclinic instabilities of the Bickley jet in two-layer rotating shallow water model Phys. Fluids 23, 126601 (2011) The instability of the boundary layer over a disk rotating in an enforced axial flow Phys. Fluids 23, 114108 (2011) A study of similarity solutions for laminar swirling axisymmetric flows with both buoyancy and initial momentum flux Phys. Fluids 23, 113601 (2011) On particle spin in two-way coupled turbulent channel flow simulations Phys. Fluids 23, 093302 (2011) Additional information on Phys. Fluids Journal Homepage: http://pof.aip.org/ Journal Information: http://pof.aip.org/about/about_the_journal Top downloads: http://pof.aip.org/features/most_downloaded Information for Authors: http://pof.aip.org/authors PHYSICS OF FLUIDS 23, 123102 (2011) Nonlinear viscous fluid patterns in a thin rotating spherical domain and applications Ranis N. Ibragimov Department of Mathematics, University of Texas at Brownsville, Brownsville, Texas 78520, USA (Received 6 September 2011; accepted 7 November 2011; published online 9 December 2011) We study the nonlinear incompressible fluid flows within a thin rotating spherical shell. The model uses the two-dimensional Navier-Stokes equations on a rotating three-dimensional spherical surface and serves as a simple mathematical descriptor of a general atmospheric circulation caused by the difference in temperature between the equator and the poles. Coriolis effects are generated by pseudoforces, which support the stable west-to-east flows providing the achievable meteorological flows rotating around the poles. This work addresses exact stationary and non-stationary solutions associated with the nonlinear Navier-Stokes. The exact solutions in terms of elementary functions for the associated Euler equations (zero viscosity) found in our earlier work are extended to the exact solutions of the Navier-Stokes equations (non-zero viscosity). The obtained solutions are expressed C 2011 American Institute of Physics. in terms of elementary functions, analyzed, and visualized. V [doi:10.1063/1.3665132] I. INTRODUCTION The large-scale atmospheric dynamics is usually described by latitude-dependent viscous or nonviscous three-dimensional flows within the theory of shallow water approximation,1–5,7,11 where the surface is considered to be free and the depth of the atmosphere layer obeys an additional equation of motion. The modelling of such moving air masses plays an important role in understanding the impacts of climate change.3,8,9 We address the nonlinear three-dimensional Euler (zero viscosity) and Navier-Stokes (nonzero viscosity) equations for an incompressible fluid in the Lamb’s form,10 ~ @~ u u~ u p ~ þ D~ ~ u rotu ~ u ¼ r þ þ W þ 2~ uX u; @t q0 2 (1) r~ u¼0 in a thin spherical shell P ¼ x 2 R3 : r0 < jxj < r0 þ ; (2) (3) where ! 0 is a small parameter for the thickness of the spherical shell, ~ u is the velocity vector, p is the pressure pressure, q0 is the constant density, is the kinematic viscosity, and W is the potential of the gravitational force which ~ is includes the Newtonian attraction. The term 2~ uX referred to as the Coriolis acceleration. Additionally, r0 is a fixed altitude above the earth. Since the velocity vector ~ u and the pressure p are coupled together by the incompressibility constraint r ~ u ¼ 0; it is difficult to analyze the full set of three-dimensional Eqs. (1) and (2). A common approach to simplify the nonlinear model in the question is to use the artificial methods such as the pressure stabilization and projection.11 As has been shown in Ref. 11, the error estimate of the pressure stabilization and projection methods is not however mathematically precise. Instead, we employ the usual spherical coordinates (r,h,/) and use the 1070-6631/2011/23(12)/123102/8/$30.00 reduction of the Eqs. (1) and (2) to the two-dimensional equations on the sphere S ¼ fðh; /Þ; 0 < h < p; 0 / < 2pg; (4) which is obtained in the limit ! 0. In our modeling, the depth is sent to zero in the limiting procedure for the non-free surface model. The latter reduction relies on the theorem from the work Temam and Ziane.12 Namely, as follows directly from our previous analysis in Refs. 13 and 16, the latter reduction is justified by theorem B in Ref. 12, which states that the strong global solution of the three-dimensional Navier-Stokes equations (1) and (2) converges as ! 0 to the strong global solution ~ uðh; /; tÞ of the two-dimensional Navier-Stokes or Euler equations on a spherical surface (4), where ~ vðh; /; tÞ ¼ lim 1 !0 r0 r0 þr0 r0 r~ uðr; h; /; tÞdr ¼ 0; vh ; v/ ; (5) in which r0 is constant and is a nondimensional small parameter. The vector ~ vðh; /; tÞ can thus be interpreted as the average velocity with respect to r. In terms of physical interpretation, the above reduction can be interpreted as follows: let the earth be a sphere of radius r0. Then, for r ¼ r0, we must set ur ¼ 0. In addition, for r ! 1, we must assume that (qur) ! 0 (the atmosphere neither loses nor gains mass from outside). However, there exists a height for which qur attains a maximum (the other possibility corresponds to the case when the vertical velocity ur ¼ 0), i.e., where @qur ¼ 0: @r r¼r0 It is thus postulated in the present work that at least one such level exists for the entire atmosphere of the earth. We call it the "mean level." In practice (see e.g., Refs. 2 and 14), we can speak about it as a level of 3–5 km. Thus, in the limit 23, 123102-1 C 2011 American Institute of Physics V 123102-2 Ranis N. Ibragimov Phys. Fluids 23, 123102 (2011) ! 0, the non-stationary three-dimensional viscous rotating fluid flow is confined on a sphere (4) of the radius r0 parametrized by the polar (latitude) angle h and azimuthal (longitude) angle /. Recently, the flow of a thin layer of incompressible fluid on a rotating sphere has been studied in Ref. 9 in order to investigate a linearized theory for small amplitude perturbations about a base westerly flow field, allowing calculation of the linearized progressive wavespeed. The authors of Ref. 9 have extended their result to the numerical solution of the full model in the absence of viscosity, to obtain highly non-linear large-amplitude progressive-wave solutions in the form of Fourier series. It is shown in Ref. 9 that the formation of localized low pressure systems cut off from the main flow field is an inherent feature of the non-linear dynamics, once the amplitude forcing reaches a certain critical level. Our previous analysis in Ref. 15 extends the earlier results in Ibragimov and Pelinovsky13,16 to the case of nontrivial and non-steady nonlinear solutions in the particular case of zero viscosity. This work is an extension of the generalized results in Ref. 15 to the case of nonzero viscosity. Particularly, here, we focus on integration of the nonlinear non-steady latitude-dependent Navier-Stokes equations in a thin rotating spherical shell. The inquiry is motivated by dynamically significant Coriolis forces in meteorology and oceanographic applications such as climate variability models, the general atmospheric circulation model, weather prediction (see e.g., Refs. 17–21), as well as a variety of applications to large-scale dispersant operations e.g., Deepwater Horizon incident studied recently in Ref. 22. II. PRELIMINARIES As follows directly from our previous work in Refs. 13 and 16, in the particular case of no rotation, there exists an exact stationary solution to the three dimensional NavierStokes equations in spherical coordinates given by ur ¼ 0; uh ¼ a ; r sin h u/ ¼ 0; p¼b a2 ; (6) 2r 2 sin2 h where (a, b) are arbitrary parameters. The stationary solution (6) describes the flow tangential to a sphere of any given radius r. The stationary flow has two pole singularities at h ¼ 0 and h ¼ p. These singularities correspond to the source and sink of the velocity vector at the North and South poles of the spherical shell P: the fluid is injected at the North pole from an external source and it leaks out at the South pole to an external sink. It has been shown in Ref. 16 that the steady-state solution (6) is asymptotically stable with respect to symmetry-preserving perturbations for all possible Reynolds numbers. If the effects of rotation are ignored, the exact solution (6) corresponds to the exact stationary solution of the twodimensional Navier-Stokes equations with nonzero viscosity on the unit sphere S vh ¼ a ; sin h v/ ¼ 0; where (a, b) are arbitrary parameters. p ¼ b; (7) However, the exact solution (7) is destroyed by the effects of rotation, whereas another stationary solution ð F 1 vh ¼ 0; v/ ¼ FðhÞ; p ¼ p ¼ F cos h dh þ p0 ; þ sin h R0 p0 ¼ const: (8) persists with respect to the rotational terms provided that F¼ 1 sin h (9) if the effects of viscosity are included. We remark that the both stationary flow solutions have singularities in the pressure, which cannot be removed by a coordinate transformation. Particularly, in the spherical geometry, the curvature of the shape induces singularities in the velocity vector and in the pressure terms. If the spherical layer is transformed to a planar layer by a homotopy transformation, the stationary flows with singularities transform to a regular solution with a constant velocity field. III. NONLINEAR NON-VISCOUS FLOWS IN NON-ROTATING REFERENCE FRAME The presence of Coriolis effects originates the general west-to-east flows caused by the Earth’s rotation which are related to jet streams, i.e., zones of fastmoving west-to east winds in the upper atmosphere between the Ferrell and Polar cells. Such flows exist due to the presence of the cold North and South poles and warm equator so that the pressure is low at high latitudes above the poles and is high in the temperature zones and even higher in the equatorial zones. Thus, as has been discussed in Ref. 23, under the assumption of no friction, and a distribution of temperature dependent only upon latitude and altitude, the stationary solution (8) can be associated with a zonal flow directed from the west to east. In view of the importance of zonal flows in terms of meteorological applications,14,23,24 we first focus on the system (1) and (2) on the sphere S in the limiting case Re ! 1 (nonviscous flows) and Ro ! 1 (no rotation), and with the superimposed stationary flows of the form (8) with F being an arbitrary function of h. In particular, as it has been remarked in Ref. 23, under the assumption of no friction and a distribution of temperature dependent only upon latitude, starting from some height, the west-to-east flows always exist in the earth’s atmosphere if the atmosphere overtakes the earth in the west-to-east motion. The possible modeling scenario when a steady condition of motion can occur in the atmosphere with an unvarying distribution of temperature (i.e., atmospheric movement that remains constant at each individual point in the atmosphere) has also been discussed in Ref. 23. The experimental results in Ref. 24 justify many of the conclusions discussed in Ref. 23. We thus look for solution in the form vh ; vh ¼ l^ v/ ¼ FðhÞ þ l^ v/ ; p ¼ p þ l^ p; (10) where l > 0 is a parameter, v^h ; v^/ , and p^ are spatial and time dependent disturbances of the basic flow (8). Since we are 123102-3 Nonlinear viscous fluid patterns Phys. Fluids 23, 123102 (2011) interested in exact solution of the nonlinear model, the case l 1 is of special interest. The particular case l 1 was studied in much details in our previous work in Refs. 13 and 16. Thus, after the above averaging procedure, in the double limiting case when Re ! 1 and Ro ! 1, the basic model is given by means of the perturbed Euler equations of motion of inviscid fluid in non-rotating reference frame as (hereafterm, the symbol hat is omitted and, without loss of generality, the parameter l is set to be one) @vh F @vh @p @vh þ 2Fv/ cot h þ þ vh @t sin h @/ @h @h v/ @vh þ v2/ cot h ¼ 0; sin h @/ (11) (13) where prime means differentiation. The pressure terms can be eliminated in order to write the Euler equations (11)–(13) as a single equation in terms of the stream function w(t,h,/) as @DS w F @DS w w/ L1 F 1 þ J ðw; DS wÞ ¼ 0; (14) þ @t sin h @/ sin h sin h where 1 @w @w ; v/ ¼ ; sin h @/ @h 1 @ @ 1 @2 DS ¼ sin h þ 2 sin h @h @h sin h @/2 vh ¼ (15) is the Laplace-Beltrami operator in spherical angles and 1 d d 1 (16) sin h 2 L1 ¼ sin h dh dh sin h is the Stourm-Liouville operator for the associated Legendre functions. Additionally, J ða; bÞ ¼ ah b/ a/ bh in Eq. (14) stands for the nonlinear Jacobian operator with the subscripts meaning the partial differentiation. The model (14) corresponds to the Euler model perturbed by zonal flows (8) in the non-rotating reference frame with an arbitrary choice of F(h). It has been shown in our recent work in Ref. 15 that solution of determining equations (see e.g., the Eq. (6.16) in Ibragimov and Ibragimov25) provides the following l–independent two invariant solutions c1 c2 h 1 ½1 (17) w ¼ þ lntan ; w½2 ¼ Uðk0 Þ; t t 2 t where and U is given in terms of elementary functions of its argument as 1n ; (18) UðnÞ ¼ c1 þ c2 ln 1þn in which jk0j = 1 and c1 and c2 are arbitrary constants. IV. NONLINEAR VISCOUS FLOWS IN ROTATING REFERENCE FRAME @v/ F @v/ 1 @p @v/ þ ðF0 þ cot hÞvh þ þ þ vh @t @h sin h @/ sin h @/ v/ @v/ þ þ vh v/ cot h ¼ 0; (12) sin h @/ @ @v/ ¼ 0; ðvh sin hÞ þ @/ @h k0 ¼ sin h cos / The inclusion of the Coriolis force creates a cyclonic rotation around the poles, i.e., west-to-east winds.26 Namely, as has been indicated in Ref. 23, the temperature difference between the equator and the poles of a sphere gives rise to waves of two kinds. The first kind consists of waves that advance in the direction of the meridian; the second kind includes waves that advance in the direction of the circles of latitude. The atmospheric pressures and motions resulting from the combination of these two groups of intersecting waves give rise to the cyclonic and anticyclonic phenomena which are nowadays a paramount topic of research in atmospheric modeling. For example, the mechanisms of cyclone formation in the Earth’s atmosphere have been recently studied in Ref. 14 with the help of numerical modeling using the complete system of gas-dynamic equations. The authors of Ref. 14 have shown that cyclones can appear in horizontal stratified shear flows of warm and wet air masses with a horizontal direction of gradients of the wind velocity components as a result of small disturbances of pressure which can be produced by Rossby waves. Recent laboratory experiments on cyclone and anticyclone formation in a rotating stratified fluid have been studied in Ref. 24. The importance of understanding the formation of cyclones and their time evolution in the Earth’s atmosphere for the creation and distribution of weather systems throughout the world is summarized in Ref. 8. In terms of applications to atmospheric sciences, it is useful to note atmospheric patterns at the North and South Poles spin around itself in the anticlockwise sense at a rate X ¼ 2p rad/day, whereas atmospheric patterns in the domain h [ [h0,p h0] do not spin around themselves but simply translate provided h0 2 0; p2 . Owing to the Coriolis effects, the achievable meteorological flows rotating around the poles correspond to the flows that are being translated along the equatorial plane. This translational motion is well captured by the zonal flows considered in our work. We consider now the model Eq. (14) superimposed by nonzero viscosity (Re = 1) and nonzero rotation (Ro = 1). The corresponding two-dimensional Euler equations (11)–(13) are then generalized to the Navier-Stokes equations in rotating reference frame as follows: @vh F @vh v/ cos h þ 2Fv/ cot h þ @t sin h @/ Ro @p @vh v/ @vh þ v2/ cot h þ vh þ @h sin h @/ @h 1 vh 2 cos h @v/ þ ¼ 0; DS vh 2 Re sin h sin2 h @/ (19) 123102-4 Ranis N. Ibragimov Phys. Fluids 23, 123102 (2011) @v/ dF F @v/ vh cos h þ þ Fvh cot h þ vh þ @t dh sin h @/ R0 1 @p @v/ v/ @v/ þ vh þ þ vh v/ cot h þ sin h @/ @h sin h @/ 1 v/ 2 cos h @vh ; þ DS v / 2 þ Re sin h sin2 h @/ @ @v/ ¼ 0: ðvh sin hÞ þ @/ @h (20) (21) V. DISCUSSION AND VISUALIZATION OF INVARIANT SOLUTIONS We next introduce t k ¼ sin h cos / þ 2Ro eral case of viscous flows on a rotating spherical surface, we apply the method of approximate equivalence transformations (the general algorithm is assembled in Ibragimov and Ibragimov25 and its application to a certain class of nonlinear wave equation has been demonstrated in Ibragimov27). Our main results on the extension of invariant solutions (17) are summarized in Table I. The remainder of this article is devoted to discussion and visualization of the tabulated results. ½i (22) and denote H(h) for an integral of F, i.e., H0 (h) ¼ F(h). In order to extend the invariant solutions (17) to the most gen- ½i The non-trivial non-steady exact solutions vh and v/ for the Euler equations (Re ¼ 1) were obtained in our previous analysis in Ref. 15. Although the particular result for the Navier-Stokes model (Re = 1) in a non-rotating reference frame (Ro ¼ 1) was not reported in Ref. 15, one can check TABLE I. Summary of invariant solutions and limiting cases. Since we are interested in the exact solutions of strongly nonlinear equations, we set l ¼ 1. Rotation Viscosity Zonal flows Invariant solutions ½1 vh ¼ 0; [1] Ro ¼ 1 Re ¼ 1 F¼0 [2] Ro ¼ 1 Re = 1 F¼0 [3] Ro = 1 Re ¼ 1 F¼0 ½2 ½1 0 c2 v/ ¼ t sin h vh ¼ U ðtk0 Þ sin / 0 v/ ½2 ¼ U ðtk0 Þ cos h cos / c2 ½1 ½1 vh ¼ 0; v/ ¼ t sin h U0 ðk0 Þ ½2 sin / vh ¼ t 0 U ðk0 Þ ½2 cos h cos / v/ ¼ t½1 c2 sin h ½1 vh ¼ 0; v/ ¼ t sin h 2Ro U0 ðkÞ t ½2 sin / þ vh ¼ t 2Ro U0 ðkÞ t sin h ½2 cos h cos / þ v/ ¼ t 2Ro 2Ro ½1 c2 sin h ½1 v/ ¼ FðhÞ t sin h 2Ro 0 U ðkÞ t sin / þ ¼ t 2Ro U0 ðkÞ t sin h cos h cos / þ ¼ FðhÞ t 2Ro 2Ro c2 ½1 ¼ 0; v/ ¼ FðhÞ t sin h U0 ðk0 Þ sin / ¼ t U0 ðk0 Þ cos h cos / FðhÞ ¼ t vh ¼ 0; [4] Ro = 1 Re ¼ 1 F ¼ F(h) ½2 vh ½2 v/ ½1 vh [5] Ro ¼ 1 Re ¼ 1 F ¼ F(h) ½2 vh ½2 v/ ½1 vh ¼ 0; [6] Ro ¼ 1 Re = 1 F ¼ F(h) ½2 ½2 U0 ðk0 Þ sin / t 0 U ðk0 Þ cos h cos / FðhÞ ¼ t ½1 c2 ½1 v/ ¼ t sin h U0 ðkÞ t sin / þ ¼ t 2Ro U0 ðkÞ t cos h cos / þ ¼ t 2Ro vh ¼ 0; Ro = 1 Re = 1 F¼ sin h 2Ro c2 FðhÞ t sin h vh ¼ v/ [7] ½1 v/ ¼ ½2 vh ½2 v/ 123102-5 Nonlinear viscous fluid patterns Phys. Fluids 23, 123102 (2011) by direct differentiation that the invariant solutions of the corresponding Navier-Stokes model are the same as for the Euler model (Re ¼ 1) in non-rotating reference frame for arbitrary choice of F(h) (see also entry [6] in Table I). The crucial difference between the results obtained in Ref. 15 and the present studies is that application of approximate equivalence transformations25,27 extends the invariant solution (17) to the Navier-Stokes model (Re = 1) in a rotating reference frame (Re = 1) as shown in the entry [7] of the Table I. Namely, in terms of the stream function, the extended invariant solution for arbitrary l > 0 can be written as cos h H ðhÞ c1 c2 h ½1 þ þ lntan ; w ¼ (23) 2lRo l 2 t t w½2 ¼ cos h H ðhÞ 1 þ UðkÞ; 2lRo l t (24) where k is given by Eq. (22). According to the entry [7], we remark that the extended invariant solutions (23) and (24) are valid provided that F(h) is constrained by the choice F¼ sin h : 2Ro (25) One can check by direct differentiation that the both w[1] and w[2] given by Eqs. (23) and (24) solve the “viscous corrections” of the Euler equations (see also Eqs. (19) and (20), i.e., DS vh vh 2 cos h @v/ ¼ 0; sin2 h sin2 h @/ (26) DS v/ v/ 2 cos h @vh þ ¼ 0; sin2 h sin2 h @/ (27) if F(h) is the solution of the non-homogeneous equation L1 F ¼ sin h : Ro It is also a matter of direct differentiation to check that F given by Eq. (25) is a solution of the linear ordinary differential equation (ODE) (28). We next visualize the obtained invariant solutions in terms of the velocity and pressure fields associated with the stream functions w[1] and w[2] given by Eqs. (23) and (24) (see also the entry [7] in Table I). For example, the exact solutions corresponding to the invariant solution w[1] for the velocity field and the pressure are written in terms of elementary functions as follows: 1 @w½2 2c2 t ½ 2 ; ¼ sin / þ vh ¼ sin h @/ 2Ro t 1 k2 @w½2 2c2 t ½ 2 ¼ cos h cos / þ v/ ¼ @h 2Ro t 1 k2 p½2 ¼ c2 4c22 ; R ð t; h; / Þ t2 2t2 ð1 k2 Þ where 2 3 sin h tan 6 t 2 7 7 Rðt; h; /Þ ¼ arctan6 4cot / þ 2R0 þ t 5 sin / þ 2R0 2 3 sin h tan t 6 7 4 2 5: arctan4cot / þ t 2R0 sin / þ 2R0 The exact solutions for the velocity field and the pressure corresponding to the second invariant solution w[2] can be written in terms of elementary functions likewise. For the purpose of visualization, the solutions are plotted on a spherical surface using the symmetric interval [z0, z0], where z0 ¼ cos h0, which corresponds to the truncated annular domain S0 ¼ fðh; /Þ : h0 h p h0 ; (28) 0 / 2pg; (29) where 0 < h0 < p/2. Without loss of generality, the spherical layer S0 is truncated symmetrically at the two rings located FIG. 1. (Color online) Comparison of exact solutions from entry [7] in Table I written in terms of the azimuthal velocity v/ component plotted on a spherical surface at different values of time t. The upper panel shows the invariant ½1 solution v/ at the initial time t ¼ 1 and latter times of t ¼ 2, 6, 12, 24, and 40. The lowermost panel is used to compare the invariant solution ½2 v/ at the same values if time. The constant Ro ¼ 68.493. 123102-6 Ranis N. Ibragimov Phys. Fluids 23, 123102 (2011) FIG. 2. (Color online) Comparison of exact solutions from entry [7] in Table I written in terms of the azimuthal velocity v/ component plotted on a spherical surface at different values of time t. The circles along the centerline (red online) correspond to the stationary ½IP trivial solution v/ ¼ 1= sin h whose asymptotic stability was investigated in our previous studies in Ref. 16. The dotted and dashed lines correspond to the invariant ½1 ½2 solutions v/ and v/ , respectively, at / ¼ 50. The constant R0 is the same as in Figure 1 and values of t [ [0.5, 10]. in the northern and southern semispheres so that the invariant solutions are free of pole singularities in S0. The values of the constants were derived from the boundary conditions given by Eq. (42) in Ref. 16. Namely, the solution for the truncated spherical layer S0 given by Eq. (29) is defined on the closed interval h0 < h p h0 for h < h0 < p/2 is subject to the boundary conditions w(h0) ¼ w0 (h0) ¼ w (p h0) ¼ w0 (p h0) ¼ 0 meaning that components of the velocity vector vanish at the regular end points of the domain. Since h ¼ 0 and h ¼ p are singular points of the interval 0 < h p, the solution w(h) for the complete sphere S given by Eq. (4) with the singular end points is defined on an open interval 0 < h < p satisfying the boundary conditions lim wðhÞ ¼ lim wðhÞ ¼ lim sin hw0 ðhÞ ¼ lim sin w0 ðhÞ ¼ 0 h!0 h!p h!0 h!p meaning that the components of the vorticity @ sin hv/ @vh @h @/ vanish at the singular end points of the domain. The treatment of the geometric singularity in spherical coordinates has for many years been a difficulty in the development of analytic and numerical simulations for oceanic and atmospheric flows around the Earth. In particular, vorticity equations were considered by Ben-Yu28 using the spectral method and pressure stabilization method has been developed in Shen.11 In terms of numerical simulations, most of studies use a powerful Jacobi-Davidson “QZ” method (Sleijpen and Van der Vorst29) to solve the resulting linearized eigenvalue problem (see also Weijer et al.7 for implementation of this method in studying the barotropic Rossby basin modes of the Argentine Basin). The common results on computational experiments seem to provide credible evidence to support the assertion that singular solutions to the shallow water equations may exist on a stationary sphere. They also suggest that singular solutions are less likely on a rotating sphere. However, the experiments conduced up to the date are not sufficiently extensive to support any credible assertions about the existence or nonexistence of singular solutions to the shallow water equations on a rotating sphere. Even less is known about singular terms used on a spherical surface. In terms of analytic treatment, the stability analysis of stationary flows in the limiting case z0 ! 1 has been considered in our earlier works in Refs. 13 and 16. The method is based on calculating eigenvalues that are closest to a prespecified target value. Figure 1 is used to show the exact solutions from entry ½1 [7] in Table I written in terms of the azimuthal velocities v/ ½2 (the upper panel) and v/ (the lower panel) plotted on a spherical surface at different values of time t. The value R0 ¼ 68.4932 has been chosen for the following reason; for a typical atmosphere, we can choose the characteristic velocity and length scale to be c0 10 m/s, and r0 103 (see also Table I in Ref. 7) so that, according to the definition for the Rossby number R0, we have 1/Ro 0.0146. A small Rossby number R0 signifies a system which is strongly affected by Coriolis force, and a large Rossby number signifies a system in which inertial and centrifugal forces dominate. As is seen FIG. 3. (Color online) Comparison of the exact solutions from entry [7] in ½2 Table I for the velocity the velocity vh component plotted on a spherical surface at the initial time t ¼ 1 and latter times of 5, 10, 15, 20, and 40. 123102-7 Nonlinear viscous fluid patterns Phys. Fluids 23, 123102 (2011) FIG. 4. (Color online) Comparison of the exact solutions corresponding the entry [7] in Table I for the pressure distribution on a spherical surface at the different values of time. The upper panel shows the exact solution p[1] at initial time t ¼ 1 and latter times of 5, 10, 20, and 40. The lower panel shows the exact solution p[2] at the same values of time. from the analysis in this work, our model is related to the second situation of the large Rossby numbers. For the purpose of visualization, the maximum and minimum values of solutions in Figure 1 and the forthcoming plots have been normalized to þ1 and 1, respectively. Since we are only interested in qualitative analysis, the quantitative in-depth analysis including the discussion of magnitudes of the flow values for the given parameters has been omitted from the present studies. ½1 ½2 We observe the convergence of v/ and v/ as time increases to a steady-state regime. To understand better the nature of such convergence, we compare the latter exact solutions at certain fixed angle / versus latitude as shown in Figure 2 at / ¼ 50 as the particular example. Namely, as we observe from Figure 2, the convergence occurs at the vicinity of the value v/ ¼ 100/sinh. This convergence can be explained as follows: we remark that the general solution of the homogeneous equation L1F ¼ 0 is given by FðhÞ ¼ a þ b; sin h (30) where a and b are arbitrary parameters. The stationary solution (30) is the exact solution of the Navier-Stokes model (19)–(21), a non-rotating reference frame (Ro ¼ 1) whose asymptotic stability has been studied earlier in our work in Ref. 16. Thus, the steady state regime shown in Figure 2 corresponds to the asymptotically stable exact solution (8) in which we arbitrarily set a ¼ 100 and b ¼ 0. ½2 Likewise, the coalescence of the invariant solution vh and the corresponding steady state solution can also be observed on Figure 3 which is used to visualize the latter invariant solution on a spherical surface. Figure 4 displays a similar coalescence phenomena for the invariant solutions p[1] and p[2]. Figure 5 demonstrates the spinning phenomena for the ½2 invariant solution v/ . Namely, Figure 5 shows the time se½2 ries for v/ from which we observe the rotation of atmospheric patterns around the polar axis, in an anticlockwise sense looking above the North pole. VI. CONCLUDING REMARKS The exact solutions to the Navier-Stokes equations in a thin rotating spherical shell has been found using Lie group ½2 FIG. 5. (Color online) Spinning phenomena for the invariant solution v/ on a spherical surface (top view) at different values of time t. The atmospheric patterns look rotating around the polar axis, in an anticlockwise sense looking above the North Pole. FIG. 6. (Color online) Discussion of the invariant solutions w[1] and w[2] corresponding to the entries in Table I depending on the limiting case. 123102-8 Ranis N. Ibragimov methods. Particularly, it has been shown that the exact solutions were found as the extension of the invariant solutions for the corresponding Euler equations. Namely, as one can check by direct substitution, the governing Eqs. (19)–(21) are invariant with respect to the obvious translation / ¼ / þ a of the angle /, where the constant a is the group parameter. Apparently, it is so happens that application of the approximate equivalence transformations shows that the Navier-Stokes equations are also invariant with respect to the one-parameter transformation group of a more complex form, namely: t / ¼ / þ 1 e2lR0 a ; 2R0 cos h 2R H ð h Þ 0 w ¼ we2lR0 a þ 1 e2lR0 a : 2lR0 t ¼ te2lR0 a ; h ¼ h; The observation tailored to construct the exact solution presented in the entry [7] of the Table I. We also remark that the above transformations are valid for the Euler equations (11)–(13) which agrees with the general conclusions on the vanishing viscosity in Cauchy’s problem for hydrodynamics equations (Golovkin30). The summary of the exact solutions corresponding to the limiting cases of zero/non-zero viscosity and zero/nonzero rotation is also shown schematically in Figure 6 for different choices of F(h). The exact solutions discussed here are used to describe physically relevant zonal flows. For example, the presence of the Coriolis force creates a cyclonic rotation around the poles, i.e., west-to-east winds. From a mathematical standpoint, the velocity and the pressure terms are unbounded in the neighborhood of the pole. Unbounded terms are common in differential equations posed in spherical coordinates and it introduces a host of computational problems that are collectively known as the "pole problem" (see e.g., Swarztrauber19). It has been found most recently that the nonlinear system (19)–(21) describing the dynamics of atmospheric motion in a thin rotating shell has a remarkable property to be self-adjoint. This property is crucial for constructing conservation laws. The more detailed group classification along with the resulting analysis for new conservation laws for atmospheric flows is the topic of current work and will appear elsewhere. Additionally, an interesting topic for further studies could include an investigation of the "pole problem" in modeling of west-to-east winds from group theoretical point of view. 1 E. N. Blinova, “A hydrodynamical theory of pressure and temperature waves and of centres of atmospheric action,” C. R. (Dokl.) Acad. Sci USSR 39, 257 (1943). 2 E. N. Blinova, “A method of solution of the nonlinear problem of atmospheric motions on a planetary scale,” Dokl. Bolg. Acad. Nauk 110, 975 (1956). 3 J. L. Lions, R. Teman, and S. Wang, “On the equations of the large-scale ocean,” Nonlinearity 5, 1007 (1992). 4 D. Iftimie and G. Raugel, “Some results on the Navier-Stokes equations in thin 3D domains,” J. Differ. Equations 169, 281 (2001). Phys. Fluids 23, 123102 (2011) 5 J. L. Lions, R. Teman, and S. Wang, “New formulations of the primitive equations of atmosphere and applications,” Nonlinearity 5, 237 (1992). 6 R. N. Ibragimov, “Shallow water theory and solutions of the free boundary problem on the atmospheric motion around the Earth,” Phys. Scr. 61, 391 (2000). 7 W. Weijer, F. Vivier, S. T. Gille, and H. Dijkstra, “Multiple oscillatory modes of the Argentine Basin. Part II: The spectral origin of basin modes,” J. Phys. Oceanogr. 37, 2869 (2007). 8 C. P. Summerhayes and S. A. Thorpe, Oceanography, An Illustrative Guide (Wiley, New York, 1996). 9 H. H. Chang, J. Zhou, and M. Fuentes, “Impact of climate change on ambient ozone level and mortality in southeastern United States,” Int. J. Environ. Health Resour 7, 2866 (2010). 10 H. Lamb, Hydrodynamics, 5th ed. (Cambridge University Press, Dover Publications, Inc., 1924). 11 J. Shen, “On pressure stabilization method and projection method for unsteady Navier-Stokes equations,” in Advances in Computer Methods for Partial Differential Equations (IMACS, New Brunswick, NJ, 1992), pp. 658–662. 12 R. Temam and M. Ziane, “Navier-Stokes equations in thin spherical domains,” Contemp. Math. 209, 281 (1997). 13 R. N. Ibragimov and D. E. Pelinovsky, “Incompressible viscous fluid flows in a thin spherical shell,” J. Math. Fluid Mech. 11, 60 (2009). 14 O. M. Belotserkovskii, I. V. Mingalev, and O. V. Mingalev, “Formation of large-scale vortices in shear flows of the lower atmosphere of the earth in the region of tropical latitudes,” Cosmic Res. 47(6), 466 (2009). 15 N. H. Ibragimov and R. N. Ibragimov, “Integration by quadratures of the nonlinear Euler equations modeling atmospheric flows in a thin rotating spherical shell,” Phys. Lett. A 375, 3858 (2011). 16 R. N. Ibragimov and D. E. Pelinovsky, “Effects of rotation on stability of viscous stationary flows on a spherical surface,” Phys. Fluids 22, 126602 (2010). 17 S. Balasuriya, “Vanishing viscosity in the barotropic b-plane,” J. Math. Anal. Appl. 214, 128 (1997). 18 D. T. Shindell and G. A. Schmidt, “Southern Hemisphere climate response to ozone changes and greenhouse gas increases,” Res. Lett. 31, L18209 (2004). 19 P. N. Swarztrauber, “Shallow water flow on the sphere,” Mon. Weather Rev. 132, 3010 (2004). 20 T. G. Callaghan and L. K. Forbes, “Computing large-scale progressive Rossby waves on a sphere,” J. Comput. Phys. 217, 845 (2006). 21 D. Williamson, “A standard test for numerical approximation to the shallow water equations in spherical geometry,” J. Comput. Phys. 102, 211 (1992). 22 P. A. Hsieh, “Application of modflow for oil reservoir simulation during the Deepwater Horizon crisis,” Ground Water 49(3), 319 (2011). 23 E. Herrmann, “The motions of the atmosphere and especially its waves,” Bull. Am. Math. Soc. 2(9), 285 (1896). 24 C. Cenedese and P. F. Linden, “Cyclone and anticyclone formation in a rotating stratified fluid over a sloping bottom,” J. Fluid Mech. 381, 199 (1999). 25 N. H. Ibragimov and R. N. Ibragimov, “Applications of Lie Group Analysis in Geophysical Fluid Dynamics,” in Series of Complexity, Nonlinearity and Chaos (World Scientific Publishers, 2011), Vol. 2. 26 G. K. Bachelor, An Introduction to Fluid Dynamics (Cambridge University Press, Cambridge, 1967). 27 R. N. Ibragimov, “Approximate equivalence transformations and invariant solutions of a perturbed nonlinear wave equation,” Electron. J. Differ. Equations 23, 1 (2001). 28 G. Ben-Yu, “Spectral method for vorticity equations on spherical surface,” Math. Comput. 64, 1067 (1995). 29 G. L. G. Sleijpen and H. A. Van der Vorst, “A Jacobi-Davidson iteration method for linear eigenvalue problems,” SIAM J. Matrix Anal. Appl. 17, 410 (1996). 30 H. Golovkin, “Vanishing viscosity in Cauchy’s problem for hydromechanics equation,” Proc. Steklov Inst. Math. 92, 33 (1966).