MULTIPLE EQUILIBRIA AND THEIR STABILITY IN A BAROTROPIC AND BAROCLINIC ATMOSPHERE by SANDRO RAMBALDI Laurea in Fisica, University of Bologna (1974) SUBMITTED TO THE DEPARTMENT-OF METEOROLOGY AND PHYSICAL OCEANOGRAPHY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY August 1982 Massachusetts Institute of Technology 1982 Signature of Author Department of Meteorology and Physical Oceanography August 1982 Certified by Glenn R. Flier1 Thesis Supervisor Accepted bp: Ronald G. Prinn Chairman, Departmental Graduate Committee MIT PAGE 2 MULTIPLE EQUILIBRIA AND THEIR STABILITY IN A BAROTROPIC AND BAROCLINIC ATMOSPHERE by SANDRO RAMBALDI Submitted to the Department of Meteorology in August 1982 Physical Oceanography and in partial fulfillment of the requirements Philosophy Degree of Doctor of for the ABSTRACT This work is divided into three independent sections; the of in the forcing topography common subject is the role of stationary waves. first section we study the multiple equilibria In the problem in a barotropic atmosphere and we examine the considering in in a simple model, form-drag instability how the study We processes involved. some detail the or departs equilibria (stable case) approaches the system Then we consider the non-linear them (unstable case). from of the solution using the properties discussing problem and of energy and potential enstrophy the conservation The zonal flow. function for the introducing a potential dissipation and forcing is then discussed small limit for and compared with numerical solutions. highly examine a two-layer section we second In the multiple study to with topography channel model truncated that We find stability. their equilibria and to discuss require meridional all the Charney-Straus multiple equilibria temperature gradients which are highly baroclinically unstable. After a detailed study we demonstrate that the flux of angular produce one is able to latitudes tropical from momentum (in the two layer baroclinically stable equilibrium which is model). For low values of the external radiation heating the stationary topographically forced wave alone satisfies all PAGE 3 the balances, while for larger values of the external topographically forced forcing the stationary radiation wave collaborates with shorter, more baroclinically unstable, moving, synoptic scale waves. last section we have presented some In the third and transport of upward theorems on the simple but general stationary quasi-geostrophic energy by momentum and zonal induced by circulation meridional the mean and by waves these waves. Thesis Supervisor Title : Dr. Glenn R. Flierl -Associate Professor of Oceanography PAGE 4 Acknowledgments This work began and was almost completed under the guidance of the late Prof. Jule G. Charney. His constructive criticisms and helpful comments deepened my understanding of my particular problem and its relation to more general ones. Throughout all my studies and Ray Pierrehumbert were Profs. Glenn Flierl, Peter Stone ready sources for advice and consultation. During 1982 Prof. Glenn Flierl became my advisor and available for supported my efforts. He was always discussion, providing useful alternative viewpoints. The manuscript was carefully corrected by Joel Sloman and the figures were skillfully and promptly Kole. The research was supported by Foundation under grant NSF 76-20070ATM. drawn by Ms. Isabelle the National Science PAGE 5 Table of contents Page Abstract ........................................... Acknowledgments .................................... 4 Table of contents.................................... 5 SECTION 1 FORM-DRAG INSTABILITY AND MULTIPLE EQUILIBRIA IN THE BAROTROPIC CASE 1. Introduction ................................... 7 2. Equations of motion ............................ 10 3. Form-drag instability .......................... 14 4. Non-linear behaviour for the inviscid case ..... 21 5. Approach to the equilibria in the viscous case 27 6. Stability of the equilibria .................... 33 7. Numerical integration . .......................... 33 References . ......................................... 35 Figure captions . .................................... 36 F igur es............................................... 37 SECTION 2 MULTIPLE EQUILIBRIA IN A BAROCLINIC MODEL 1. Introduction . ................................... 48 PAGE 6 Page 2. Equations of motion 3. Low-order model .... 55 4. Multiple equilibria 58 5. Search for more stab 6. Conclusion ......... 76 References ............. 78 Figure captions ........ 79 Table captions ......... 82 Figures ................ 81 Tables ................. 102 71 equilibr SECTION 3 STATIONARY ADIABATIC FRICTIONLESS FLOW OVER TOPOGRAPHY ....... 1. Introduction 2. Equations of motion . . .. ..... .. 3. Simple 4. A simple non-linear analytical theorems .... 103 ........... 104 107 . .......... References ............ . -.- -.- -.- -.- -.- solution 115 123 PAGE 7 SECTION 1 FORM-DRAG INSTABILITY AND MULTIPLE EQUILIBRIA IN THE BAROTROPIC CASE 1. Introduction In 1979, Charney and his coauthor-s wrote a series of which papers developed the idea that the atmosphere first may have several possible equilibria forcing. external in a beta In the simplest problem, barotropic flow channel plane same the for states with topography, forced by a momentum source, Charney and Devore (1979) demonstrated that truncation forcing, did have within multiple certain Given the states. stable parameter spectral order the non linear equations arising from a low ranges, possible two equilibria were found, one with a weak zonal flow and strong waves and one with strong higher resolution zonal numerical flow code and was weak waves. A used to assess the degree to which their truncated model rapresented solutions of the original problem. In further work, Charney and Straus (1980), examined the effects of baroclinicity, Charney, Shukla and Mo (1981), PAGE B of suggests possible directions in which (1981), Reinhold work the finally applied the model to earth's topography; the principles of multiple equilibria might be modified when scale synoptic occur instabilities in addition is to to weak topographic forcing of the long waves. model simplest with form-drag to a model similar equilibria: and Flierl Charney however, here, purpose Our (1980). (1979) Hart of multiple and instability that the examine or We shall consider the processes involved in form drag instability in some detail in order to the explicate in mechanism the straightforward most barotropic case. how As new physical contributions we point out form drag is felt by the zonal flow through the ageostrophic component of the flow and in paragraph 3 and 5 the properties the and mechanism of describe we the form-drag emphasizing the study of the quantities that can instability be the paragraph In measured. 6 we instability as a simple consequence of energy and conservation discuss of potential the the form-drag simultaneous In the enstrophy. viscous case some numerical calculations have been done to show how the system approaches the stable equilibria. As technical contribution we have introduced potential function; in this a new way we have three degree of freedom and three conserved quantities and we are able to PAGE 9 give a simple but complete qualitative discussion of the properties of our system in the inviscid case. PAGE 10 2. Equations of motion The simplest set of equations which form drag Hart (1979) illustrate the instability and multiple equilibria are those of describing the coupled evolution of the x-independent flow and the topographically forced waves. derive similar expansion. equations This new without derivation using any clearly asymptotic points importance of the ageostrophic components in the between the waves and the vorticity equation for wave topography is +~ zonal u flow. flows The over out the interaction potential y-independent ,+ ~ We C~±2.1I) is the geostrophic streamfunction where U is the eastward component of the geostrophic component of the geostrophic wind V is the northward wind f, is the Coriolis parameter at mid-latitude is derivative (df /dy) of the Coriolis parameter = PAGE 11 h is the height of the topography H is the depht of the fluid t' and upper is the dissipation time scale (due and lower Ekman layers ). zonal wind U cames from the to friction in The evolution of the mean x-averaged momentum equation where periodicity has been assumed T~--(V/ (2.2) - When the topography is y-independent the Reynolds' disappear and the changes in dissipation, external forcing, U part of the y-component of the zonal flow are due to and to the stresses the velocity ageostrophic V' . This ageostrophic part of the velocity can be directly related to the form drag as follows: we calculate the mean meridional flux of mass here, denoted by M M (H- The form drag LX H - V- (2.3) is given by L2 LaL rDx X rax - PAGE 12 where P is the atmospheric ground angle from pressure and G the is horizontal to the inward normal to the surface. To lower order the meridional mass flux simplifies to M + +(2.5) where the first term represent the geostrophic contribution to the form drag. In a y-independent problem, the mean meridional mass flux could still on time but th'e simplest case is depend that of no net flux M= o (which can be confined 2h .=-_. in justified if the a very wide channel). flow is (2.6) assumed to be With these assumptions, our system of equations becames X'D -j (2.8)D (2.8) The solution of the system (2.7) and a with found Fourier decomposition PAGE 13 can be topography of and streamfunction t4)= Equations sin( h h = .4 - Uy + A Al (2.7) and H k x+ T) AA sin( .M~%ILJ4J~A k x+ + B cos( 14 x+ -4 f) (2.8) became ~A4A4 (QU)- k t)i A4 (2.9) U*) (2.10) t AAN U- _ -BM (2.11) PAGE 14 3. Form-drag The instability simplest system containing form the drag interaction can be obtained with a single mode topography naL )( (3.1) Under this assumption our system of equations can be divided into a set of sub-systems for k #h k we get (3.2) - A - Even if U is time dependent (3.3) we can see from the energy equation -- (. (3. 4) that these modes simply dissipate. (3.3) can be solved exactly; the solution is given by: The system (3.2) and introducing a complex notation PAGE 15 (3.5) ' A This is just a decaying travelling Rossby wave. To determine the time dependence of the zonal flow and to study the role of the topography we must consider the case k = k we then have the following system (jil~U*) -~ (U= A)(U- =A.cU-/3Id)A a To clarify involved let the us role introduce (3. 6) -- (3.7) +AiU--LB '-' H of the (3.8) different non-dimensional parameters variables as follows: (~IA) LI Z)-U (/3J V (3. 9) r6 H "i') PAGE 16 With these substitutions we get two non-dimensional parameters -^f- (3.10) that describes the importance of the mountain and - _ which describes the dissipative zonal number wave the earth radius and to h A/ I.-4 f .2, Y k = n*( 27V/2 rr a cos , f is the Introducing effects. a middle latitude, ),where a is chosen here and the following typical atmospheric numbers, 45*, be n 11) __(3. I -- sec 10 t 4. 10 days 4S .6*10 5 sec, we get the following values: . - ' . 0 With these non-dimensional parameters our system of (3.6)-(3.7)-(3.8),dropping the primes, becomes equations, - ~~* -- (3. 12) (3. 13) PAGE 17 The mechanism of form-drag instability can be seen quite clearly from these equations and figures 1 and 2. Let us neglect forcing and dissipation and the situation consider first UJ4C when U is less than the resonant speed U ( , or in dimensional quantities. The wave generated by the flow has vorticity which is in phase with the topography since the flow is dominantly constrained to follow the geostrophic south contours, on and north on downslopes, corresponding to cyclonic upslopes curvature over the peaks. situation there wave-mean flow is non no In form interaction this inviscid drag, in an the steady-state illustration absence of of time dependence or friction (Pedlosky, private comunication). If we perturb this which by introducing vorticity is 9e out of phase with the topography, the flow will evolve in time. flow: flow There are two different tendencies for the an acceleration or deceleration of the mean flow and a change in the vorticity distribution. If topography the initial vorticity by 90 (figure 1.b), corresponding to B This discussion of form-drag instability was by Prof. Glenn R. Flierl lags perturbation the O, the 0? contributed PAGE 18 pressure disturbance will form drag lead the topography decelerate will the Changes in vorticity are induced propagation and subresonant mean tendency eastward flow, mean by so flow that the (figure 1.c). westward Rossby wave advection by the mean flow. the westward For propagation wave larger than the eastward mean flow so that the is vorticity begins to increase over the crest (figure 1.d). From the first derivative patterns, we the second whether generating field, to see There three are different mechanisms for wave propagation and mean advection of the ,: by generation topographic the (4 additional advection of the basic state vorticity the L4 continues to lie propagation wave The flow. opposite or reinforced is perturbation initial counteracted. explore derivative of the vorticity, time the can the toward field by plus mean advection leading west, and field to ) a of from the original perturbation (figure 1.e). sign The decrease in the flow over topography leads to a decrease in in vorticity Finally, topography by 90*. the a which advection (the term which would correspond to vorticity differentiated since troughs, the correlated negative, with contributions to h. W For are and U < I the - LQ form of Eq. (3.14)) also leads the is U of the leads basic in the topography the basic vorticity , then, all of is the of opposite sign to the initial PAGE 19 that perturbation so is wave topographic the neutrally stable. U > When different { the however, (figure 2). The basic equilibrium distribution of since vorticity is anti-correlated with the topography in vorticity tend to balance the we still get a form drag retarding the propagation tendency now has but the a reversed sign. The we As westward. a than propagates it cha nges in the term, flow leading the topogra phy (figure 2.c). The over the the decrease contribution now in flow mean the But the has the since sign opposite (figure 2.g). possibility Thus there is the when the equilibrium state vorticity is sufficiently which occurs near resonance) that this other two and I will term basic mechanism for form drag as follows: large ( will dominate the original reinforce perturbation, leading to an instability of summarized the equilibrium vorticity is now of opposite sign from the topography the still propagation wave is also the same as before (figure 2.f). topography original the doppler-shifted the tendency from third at look flow mean perturbation advects downstream faster generates the If we introduce the same form of perturbation, topography. however, over stretching vortex the planetary than rather vorticity relative changes wave quite is situation the flow. The instability, therefore, can be when the zonal flow is PAGE 20 supercritical, topography. the vorticity Decreases in the anti-correlated vorticity is anticorrelated zonal shifting flow result westward with the in leading pressure patterns which have highs on the upslopes and on the downslopes. this to lows This in turn leads to a form drag which decelerates the zonal flow further. PAGE 21 4. Non-linear behaviour for the inviscid case _I As we have seen can our solve parameter and we by developing all our unknowns in a system : power series of (F-B) is a small ) , = (iA + + ( We shall study the non-linear behaviour only of the zero-order equations which are: (4. 1) 0 .: -. (4.2) .( 0 - B B - (U,0-- ) A , U0 - This inviscid problem has two independent (4.3) conserved quantities 1 .4-I 40 0 2- (0- -) At~ ~ +-BI- (4. 4) (%U+ .4 xno (4.5) (.5 PAGE 22 pointed DeVore and Charney As first the out, quantity represents the total energy and the second quantity a allows solution analytical complete a potential the with energy total existence of these two integrals of motion The enstrophy. the of combination system the of (4.1)-(4.2)-(4.3) in terms of elliptic functions. As we shall see a full discussion of the of properties the solutions of this system can be done without solving in only equation order As first step we derive a simple second explicitly. we substitute the definition of Q in U; Eq. (4.3) getting ---- 4Q +o( o~ do--- After of derivative time a 4.6 ) and (4.1) the substitution of B,,from (4.6) we get: ---(L)--iMU-')-- q --- It is useful at expression explicit for this Q point as to U (4.7) introduce a more function of the stationary solutions. If U has any chosen value U1 the system PAGE 23 (4.i)-(4.2)-(4.3) gives a stationary forced wave with amplitude 0~o (4.8) (4.9) U-i On figure 3, the constant 3,, Eq. (4.5), light lines while the heavy stationary solutions, from here curve, Eq. (4.9). on lines of line represents the called the equilibria At least one intersection on the branch number one always exists; 0 can represent therefore the conserved quantity always be written as a function of the coordinates of this intersection, here denoted with ( U , A , B ),as follows Q -. (0--) -+. X (4.10) Using this definition and defining u = U, - A = - U "distance" from the U "distance of the equilibrium from the resonance" equilibrium PAGE 24 (4.7) can be rewritten as Eq. LL--~AU + QL (4. 11) L or multiplying both sides by u. (4. 12) e + LL) where LL + We have a evolution of defined is a potential function the L. - analogy mathematical 4- by LL (4.13) the between time wind and the motion of a particle zonal moving in a one-dimensional space with potential a energy given by With this behaviour the of potential our solutions: we can discuss the general in particular we can see that linear stability properties of the points of equilibrium are governed by the sign of I+i , in agreement with the linear analysis of Charney and DeVore; behaviour is dominated the non linear by the positive sign of the fourth PAGE 25 order term and we have a non linear stability for all the points of equilibrium. Depending on the number of points of equilibrium our system A behaves ,the in points =0 , different of ways; equilibrium for are a given given Q,,or u = 0 by or i. e. d tA 6ALt the u u = 0 root A, B), - S represents previously the discussed .3=0 <4.14) equilibrium point, while the other two roots AZ,3 represent the other intersections of with the equilibria curve. the (4.15) constant The point P =U ,A B) G. line will be the only equilibrium if <(4.16) In this case the potential function, Eq. (4. 13), has only one PAGE 26 minimum ( U, A_, oscillates the for potential functions evident that the second equilibrium one on the branch 2. if periodic; we I is The motion equilibria the basin to the other. system is start close to the equilibria it consists of a simple oscillation while if we start far the always In figure 4 we have drawn the potential function for this case. always unstable an Some straight-forward analysis shows that the minimum on the branch the (4.13), it is is point point while the other two are stable. than from three intersections are present; > expression lower around periodically B ). For the system the and oscillates from from one attractor PAGE 27 5. Approach case; to the equilibria in the viscous case Until now we have in of a small dissipation and forcing the presence considered the non-dissipative evolution of our system is essentially similar now we except that no longer have an equilibrium curve but only one or three equilibria and from any points of our phase-space system goes conserving better towards the stable points while approximately To get a the potential enstrophy and the energy. understanding of the way in which the approaches the stable equilibria we can study the system behaviour of small perturbations around these stable equilibria. Eqs. the From (3.12)-(3.13)-(3.14) we find that these equilibria are defined by: 8) -U - -- U(5.1) B (5.2) . C (5.3) U-A It coordinates for 'Xft *. - is of convenient the 0 to get equilibrium we have: an expansion of point in powers of the 2? PAGE 28 (5.4) Pt; U +(5.5) + )(5.6) where A and U are related by: (5.7) This is a zero order relation and equilibria implies that the point must be on the inviscid equilibria curve. Secondly, P ~ we have ~ - (5.8) which comes from the first order equations and implies the balance between the generation and the dissipation. A graphical solution of this system is shown in figure I for different values of the external forcing U* The heavy line represents the equilibria curve (5.7), while the light curve rapresents points where the generation is in balance with the (U,A). we As dissipation, can Eq. (5.8), in the plane see for small forcing we have only one PAGE 29 stable equilibrium and for larger values of the forcing can the get three equilibria with the one in we middle unstable. Denoting, as before, with (u,a, b) the from (U ,A ,B perturbations ) we get the following equations: (5.9) S- S= -- (5.10) i - (5.11) U) Having a system with constant coefficients, look for a the form (u,ab)*e solution in we can and we get the following eigenvalue problem: (t+ + Ci'+2) - )- - (5.12) -o where expanding the eigenvalues -. in power of )(5. 13) PAGE 30 - C3 + > (I 0 +. -3 (. ) = we get Q - -o Z. . +-_-_ -i (5. 15) one real We have three solutions: g = - (5. 14) (I- ) + ( (5. 16) 0+ o (5. 17) and two complex coniugates The first eigenvalue has for its eigenvector (5. 18) curve in the plane B = 0 of figure eigenvalues have an oscillatory part time scale and therefore, on the the to tangent corresponding to the direction with equilibria The other two 3. a much shorter average, the solution PAGE 31 approaches the point of equilibrium moving the along "equilibria curve" with a damped motion. These two complex eigenvalues with their eigenvectors b C LL (5. give rise to a damped oscillation with damped standing and 19) in the zonal flow together propagating waves. In fact the stream-function for the wavy part of these perturbations can be written as: FWZCx,~) e~~L/ 0 (5. 20) where Z(~,t) = 0- z. 1~ (~) - We have two cases: L> d. Z (xQ) , we have = .4. c { -IZ. for -l .- c and ) +-x z-. the super-resonant case, PAGE 32 an where the first term rapresents eastward moving "slow Rossbu wave" and the second term a smaller standing wave out of phase with the topography; U we have where the first term Rossby for the sub resonant-case, and represents a westward moving "fast wave" and the second term a smaller standing wave in phase with the topography. PAGE 33 6. Stability of the equilibria Using figures 6. (a) and 6. (b) properties of the equilibria; we for a parabola in figure 6.(a), we have three the equilibria curve, heavy line; can discuss chosen 0 we can Eq. For the ellipsoid intersects with fields can (4.4). see, for the two stable equilibria C and E, any reached. points with for each intersection we nearby point, on the QGcurve, has larger energy be dotted , intersections have drawn the energy ellipsoid, dashed lines, As the unstable the 0 and cannot equilibrium D, the energy surface so that neighboring finite B's and smaller energies in the U and A be enstrophy; this reached conserving trajectory is energy the and potential heavy dashed line of figure 6. (b). Numerical integration 7. To illustrate part of dynamics with a of fourth = .8 considered, linear and non linear the preceeding discussions we have integrated, order (3.12)-(3.13)-(3.14). x the and for Runge method, the system We have chosen the following values .05 . different with three equilibria Kutta In figures 7a,7b,7c initial (CD,E); as we have conditions, the system we can see, the two PAGE 34 stable equilibria almost on the displacement (C,E) isolines are of approached with oscillations G, together with a net along the inviscid equilibria curve toward the stable equilibrium. In figure 8 we have considered parameters such the system has only one equilibrium. that We no longer have any forced and damped equilibria on the branch number two of the equilibrium curve, but the equilibrium curve is still a zero order equilibrium oscillations, solution so that with 0 nearly conserved, we can see together with a mean motion along the equilibrium curve toward the saddle After crossing damped point. this point, the oscillations grow again and the system evolves toward the only equilibrium left. PAGE 35 REFERENCES J.G. Charney, J.G., DeVore, 1979: Multiple flow equilibria in the atmosphere and blocking. J. Atmos. Sci., 36, 1205-1216. Charney, J.G. and D.M. Strauss, 1980: multiple in equilibria and propagating baroclinic, orographically systems. J. Atmos. Sci., Charney, J.G., Form-drag instability, forced, planetary 37, j. Shukla and K.C. Mo, barotropic blocking planetary waves wave 1157-1176 1981: Comparison of a theory with observation. J. Atmos. Sci. 38, Charney, J.G. and G.R. Flier1, 1980: large-scale atmospheric motions. oceanography. Oceanic analogues of Evolution of physical Edited by B.A. Warren and C.Wunsch M.I.T. Press. 504-548 Hart, J.E., 1979: anisotropic Reinhold, B.B., stationary Barotropic mountains. 1981: quasi-geostrophic J. Atmos. Sci., Dynamics of weather waves and Cambridge, Mass. 02139. blocking. flow over 36, 1736-1746 regimes: quasi Ph.D. Thesis, M.I.T., PAGE 36 FIGURE CAPTIONS Figure 1 : see text Figure 2 : see text Figure 3 : The heavy line represents the equilibrium curve, and the light lines are contours of constant Q in the plane (UA). Figure 4 : Shape of the potential function for different value of Figure 5 : Graphical solution of Eq. (1.48), for different values of the forcings U Figure 6 : The equilibria the which , line, Eq. (1.49). points (C,DE) are points in the and energy heavy potential enstrophy surfaces-are tangent and the stability is determined by whether nearby points can be reached. Figure 7 : Examples evolutions, starting of time from different regions of the phase space in presence of three equilibria (CD,E). The heavy line represents the trajectory of the system, the medium heavy line represents the equilibria curve and the light lines are the intersections of the with the (UA) plane. For this As Figure 8 one 9= .05 and U = 4.8, ch osen: in figure equilibrium, C. chosen: U = 4.8, constant G :? 7 but in For .05 and this evolution surfaces we have .8 presence of only evolution we have .8. PAGE 37 (U <#8/k' CASE) (0) + basic state /~ Th + perturbation -1 - (b) 7%> / (c) Ut / / (d) + + / -I- (e) + - - from wave propagation (f) + from ftt Ut * Vh ttt from Ut 'V FIG. 1 - / (9) PAGE 38 (U> CASE) / /k 2 (a) + basic state / / -I- (b) + - perturbation / (c) Ut // / (d) St + 00, /0 (e) tt from wave propagation - Ctt (f) from Ut. Vh (g) Ctt f rom Ut 0 Vt FIC. 2 PAGE 39 branch *2 bran PAGE 40 A = 3.0 v (U) A = 2.6 A = 1.6 30- 20- -5 -4 PAGE 41 -4 -3 -2 -I 10 -5 T r i U'= 6 =2 PAGE 42 N. V' S-I / - PAGE 43 PAGE 44 PAGE 45 PAGE 46 PAGE 47 CTfl 0 PAGE 48 SECTION 2 MULTIPLE EQUILIBRIA IN A BAROCLINIC MODEL 1. Introduction In equilibria the barotropic is the multiplicity Because the form-drag oresents a maximum the speed of the zonal flow goes close to a resonance, the mean for a constant zonal flow external has forcing, on the resonance. of view, From this point the of statistical equilibrium values airplane possibility, reaching different two sides we analogy with the two different equilibrium an of due essentially to the resonant structure of the form-drag. when case of have a strong velocities a supersonic considerations continue to case, but here we speed. apply also to the baroclinic to (Reinhold, 1981). We have organized our study in a to the have some evidence that the equilibria synoptic scale baroclinic instability us a These general qualitative found could be unreachable if the system is too unstable allows that can reach, for a constant propelling force: subsonic speed or, after having overcome the maximum in form-drag, the compact way that discuss the multiple equilibria problem in a PAGE 49 more general and simple way We forcing. thermal function as of the external demonstrated that the multiple have Straus and equilibria suggested by Charney for realistic parameter values are baroclinically unstable to short waves. We have found that the balance implied by one of equations, the barotropic zonal flow model equation, is quite different from the real atmospheric balance. shown the We have then that a simple correction to this equation, simulating a constant flux of zonal angular momentum from the leads to baroclinically more stable tropics, equilibria. New equilibria are then hypothesized in which the solutions have a large scale stationary dependent part. stability boundary this but part large the and a synoptic scale time scale state is near the synoptic motions also play an important role in the heat and momentum balances. PAGE 50 Equations of motion 2. In the barotropic case it was decided to that arguing directly, flow zonal forcing, this baroclinic atmosphere, would correspond to the thermal the driven by direct thermal field. radiation forcing and in a wind To take into account the effects the study to the force of baroclinicity of the flow, Charney and Straus (1980) decided for to look two-layer equilibria multiple in a highly truncated Following their notation, we consider the model. potential vorticity equations for the two layers Lb (4. C, (2. 1) +1 ( qz.) .)19 _j 7- = _F V1 V 2- q7 t where the subindex 1 indicates the upper layer the subindex 2 indicates the lower layer 1represents internal dissipation V< represents bottom Ekman dissipation represents the Rossby deformation radius The starred quantities represent the external PAGE 51 forcing of the temperature field, here parameterized as a Newtonian relaxation with an inverse time scale of The potential vorticities in the two layers are defined by + [ - where h is the height of the topography and H is the mean depth of each layer. We consider this flow in a channel walls at y = 0 and y = by , u and v by L. T with meridional Let x and y be scaled by L, t L by /9 L by , h by H and define the non dimensional parameters XL respectively. variables, functions, that Vol y~K~'~V To get a spectral model we now here F- on dimensionless, and complete F by denoted F F -T. U a develop ; LV we let all set our of such PAGE 52 f3 = (cZ.o) = F' F and we define horizontal crii average (c:- CI on the cj-)f where the channel. bar implies Substituting a these definitions in (2. 1) we get L- L AL (2. 2) LL P j9All L9Q ~ ~~ ~(~ WL P I [ i ±~\ 97IA"4kzVIL ~ Y J &:j WJ i wh ere Qi =L4 1L 1-- . (W and An introd uced their alternative set to the discuss interactions: the of variables is usually dynamics of the two layers and barotropic compon ents of the stream-function and baroclinic PAGE 53 eat) \~'iL' 8*-. L ± - 4 ~ ( .Q) and the barotropic and baroclinic component of the potential vorticity (9)= (9 93) = -( the With these new variables ) + tK - of equations motion (2. 1) bec ome kwz p) +i (91~ fX zj '~Q (2.3) = &. ~A()~- ~ (v~~± (Q~O9 -.-).A - aCX~Q In our equation obtained multiplied by '(-and we discussion 1I summing from to shall consider the energy adding the first equation of (2.2) second the over \('L i. After equation changing multiplied to the by new PAGE 54 variables, the energy equation takes the form ole. -.- 2: We equations shall for . o9+ A1 4.// also both consider layers -v.Y 2. - the A~3~q dCL 9. potential (2.4) enstrophy obtained by summing over i the . first equation of (2.2) multiplied by equation multiplied by (e)9L9 ; for the and the second upper layer we get zU -C1 (2 LiL %L 5 and for the lower layer a4 L -L. 9. .IL L 9 .. R( J- (2.6) PAGE 55 Low-order model 3. Following the to model procedure for the barotropic equilibria we consider only three multiple find used now on system to from modes, the same as Charney and Straus (1980; referred as C-B): a zonal mode Fa tw = and = c ose( y) I two wavy modes F = 2 sin( y )*cos( m*x F = .3 2 sin( x= 1+ .A7 = y )*sin( m*x .3 This choice, in particular, enables I -M our have baroclinic instability and to balance the generation of heat with the divergence of The heat. interaction the eddy fluxes of sensible coefficient between these modes is given by The and we choose term has projection only on F , / to match the C-S value; PAGE 56 we consider a single mode topography with projection only on F, and The evolution equations for the zonal flow become 3 A. + _j+ - In a long time (3. -( .2K mean we can + (- see (3.2) from equation that, the zonal flow to balance the dissipation due to bottom the can boundary see first in the lower layer, the form drag must push layer. the that the Ekman In this simple model the Reynolds' stresses are identically zero, and from the second we 1) eddy fluxes equation of sensible heat, the internal dissipation and the Newtonian cooling must balance the external generation of heat. If complex variables are part of the flow: /141+ L introduced for the wavy PAGE 57 Eqs. (2. 3) can be rewritten as: OX1 + IZ. - 6 (3.3) C 2.1 +- 6 (3.4) +09 where aI aI a# and b a and zonal quantit ies, -L Z. 19i = G are + functions only defined by x x , LC N - (-X?-- 1) ? I ) -- QVZ Q aZ G '7 of the PAGE 58 4. Multiple equilibria Without the topography behavior. For a this small forcing system has a it has a stable stationary solution, with no waves and with the heat balance by the mean increased meridional this instability circulation; solution and the becomes system if unstable reaches a the satisfied the forcing is to baroclinic different equilibration with the meridional temperature to simple stable gradient set two-layer critical value for baroclinic instability and with a translating baroclinic wave large enough to transfer all forcing. This second state corresponds to a stable cycle in the heat necessary to balance the external the phase space. The presence of topography enriches the system several new behaviors and, as C-S discussed, it can show stationary, periodic or aperiodic the range of parameters with solutions, chosen and depending on the on initial conditions. In this paper we shall mainly consider the stationary solutions and their stability. To compute Eqs. (3.1)-(3.2)-(3.3)-(3.4) dependence. is each are solved A graphical solution in the presented point stationary ( in kV , Fig.1. GI solutions without plane the ( time ) This is obtained by computing at ) the wavy part, from Eqs. (3.3)-(3.4) PAGE 59 looking of curves represent 9 forcing computing the external lines zonal flow is specified, then the once linear are which balance in are straight heavy line at forty solutions Hadley no with and form-drag finally Ekman the (heavy lines), Eq. (3.1). The represents the five degrees wave present and with zero wind the closed heavy speed in the lower layer; and ) constant for the points where the dissipation from Eq. (3.2) (the light line represents non trivial balances between form drag and dissipation. The intersections between line light selected the heavy lines and a define the stationary solutions for a selected strengh of forcing In the case here studied the external parameters have been set to the following values: channel width 7TL = 5000 Km H = 5 mean depth of each layer Km Rossby deformation radius = 1000 Km 2L A = 5. The internal and Ekman dissipation well as the time scales as Newtonian heating time scale have been set to ten days corresponding = h to: =.0114 PAGE 60 For the topographically forced wave wave number three, n = 3 and we m = have chosen ( n/2.83 ), zonal and the amplitude of the sinusoidal topography has been set to 1 Km. With these choices we get the same equations with the same parameter values as C-S. Before discussing the stationary full system it -solutions of the is useful to consider the stationary waves that are forced by the topography for a fixed ( In Figs.2 and 3 the amplitude and the phase, relative to the topography, of the barotropic and are drawn. baroclinic forced waves Because the topography interacts only with the lower-layer flow we also present the amplitude and the phase of the forced wave on this layer divides the plane in two part it the has zonal form The straight line regions; in the upper flow has easterlies while in the lower part westerlies positive (Fig.4). in drag, flow eastward (i.e., have been shaded the lower-layer. The region of where the topography pushes the zonal the phase is between - Tr and 0 ), (Fig.4b). As we can see the amplitude of the forced solutions is dominated by the intensity of the zonal flow in the lower layer, In the ( 'j-91), inviscid and by case the presence of this a resonant curve. resonant curve (heavy line of Fig.5) is the two-layer baroclinic extension of the resonant barotropic point, 0 and 4- , and in particular PAGE 61 interior it can be shown that, for westerly flows in the (shaded semicircle this baroclinic forced waves are in phase cyclonic with curvature topography, the with peaks the on barotropic and the the region), anticyjclonic and subresonant barotropic curvature in the valleys) as in the of case. It is interesting point to system the that out (3.3)-(3.4) has a true resonance, infinite response, even in P that lies fact In the presence of dissipation. on the closed heavy point a is there line of Fig.1, where the dissipation is balanced by the generation and it is possible to have a stationary wave also without the topography; course, this stationary finite value of the solution cannot external resonance when the determinant of , Eqs. (3.3)-(3. 4), QO -Q be coefficients for a We find a . forcing 4 of and vanishes. Q = The real part vanishes on the heavy while reached of line of the immaginary part vanishes on the dashed line; Fig.5 the intersection between this two curves defines the position of the infinite resonance. Thus unlike the barotropic case where the amplitude of the wave is limited by dissipation as PAGE 62 the forcing is increased, here grow as the stationary wave amplitudes when the forcing is increased. Due to the presence of this true anticipate that in resonance we can its neighborhood the forced stationary wave are large enough to tranfer meridionally all the heat necessary to balance the external heat forcing and therefore we can expect multiple equilibria. having solved forcing ( As we can see, after the problem, (Fig.1) for small values of the K. 140) we have only one intersection corresponding to a stable direct Hadley circulation, without any forced wave. For values of the forcing between .140 and .175 two new equilibria are present; wind ( > PI) one has easterly zonal in the lower layer and the forced waves exert a negative form drag while the other has westerly zonal wind ).in the positive lower layer and the forced form drag. westerlies ground. exert a For values of the forcing between .175 and .200 two other equilibria have waves and for When the forcing appear; for C.195 both ,>.195 one has easterlies at the is increased further, the two solutions with equilibria. One is the Hadley solution that has only a mean meridional efficient gives easterlies circulation in transfering meridional forced ones & . and, heat temperature disappear as we can leaving see, is not very meridionally gradients The other two equilibria three and 9i therefore close are to the locked to PAGE 63 the true resonance and, in analogy with the barotropic case, other one can be called subresonant and the whether on depending superresonant, it is inside or outside the resonant curve of Fig.5. Before continuing our study of these equilibria, let us try to state why these equilibria are important and from which characteristics they get their importance. more or, points equilibria of presence The generally, the existence of regions in the phase space where some are components predictability In our case these are these components. of the increases varying, slowly the mean zonal flow and those modes directly the by forced Following Speranza (private comunication) these topography. regions can be seen as examples of islands of predictability in a stochastic, unpredictable space. we are stationary considering predictability In our case, because high the solutions. is a consequence of the high persistence. islands There are four classes of persistence that we can find: * a) simple and fortunate stable solution, inside its b) solution * stationary Stable for most case we could find a full stationary which attractor basin, * Unstable the In system the once it stationary confined remains gets there. solution * In a s t ill PAGE 64 case simple full non-linear stationary a find could we is close enough to the equilibrium system the case, once point, it continues to go closer in some directions, on linear the of system their of values the rates. growth system Lorenz stationary points of the such point becomes directly this eigenvectors related to the amplitudes of the most unstable and of the amount of time that i.e., to close spends around theory stability equilibrium, its persistence , the unstable the on the Once the system enters the region manifold, it departs. validity others, along while manifold, stable this In solution with some linearly unstable eigenvectors. in unstable The the aperiodic reqime are examples of this case. c) * corresponds to stationary solution * quasi-stationary Stable This a stable solution with only some components nearly or this where stationary, quasi-stationarity is due to the non-linear interaction with the full non-stationary some constant components d) * Unstable A components. stable cycle with is a simple example of this case. quasi-stationary solution* In the more general case we can find an unstable solution with only some components nearly stationary. time behavior of the system point, we have to consider the Here, to understand the in the neighborhood of this instability of this state PAGE 65 role the account into taking the of non-stationary components. full that evidence We do not have any stationary solutions do exist, but we have some experimental evidence blocking situations - and until found highly truncated models spectral seen been have have stationary these of now therefore, only are solutions must be to real persistent question open on the truncated solutions, but we realize that these solutions can be very they - have been obtained with and, an The part. approximations Here we still solutions. validity possible as only suggestions of the existence of solutions quasi-stationary with a non trivial that theory - equilibria multiple theoretical some relevant even if approximations of unstable quasi-stationary solutions. Once the system is supposed to be in the presence of one these islands we still have to determine its impact of on the time behavior of our importance the from system. island An gets its amount of time spent by the system in its neighborhood and this is a function of the extent of the attractor its in basin stable manifold and on the growth rates present in the unstable manifold. Here we believing conclude our that approximations of our truncated real general considerations and solutions are solutions we stationary quasi-stationary PAGE 66 begin To study. our continue unstable manifold around the in the instabilities subspace including only the forced waves and the zonal and we solutions stationary "internal" linear the study shall C-S understanding of the our flow then we shall consider some "external" instabilities in the subspace containing the most unstable baroclinic wave. Let us first consider the of zonal flow and meridional perturbations in the solutions, i.e., stationary these instabilities "internal" in the wave field with the and zonal same number of the topography, described by the wave the system of Eqs. (3.1)-(3.2)-(3.3)-(3.4) linearized around chosen As solution. stationary we as see, shall a consequence of the introduction of another external forcing, all the states; unstable ( points therefore modes on I, 91 ) can the compute we correspond the whole plane. growth roots is plotted unstable non propagating in rates of the We have a sixth-order Fig.6a number The eigenvalue problem with real coefficients. unstable stationary to of the number of and (real) roots is plotted in Fig.6b. Without the topography we do not find any finite area region of growing stationary perturbations; (dashed line in Fig.6b) that there is only the divides propagating from the westward propagating ones. a line eastward As we see, while this line now has been broadened to a region of finite area new regions of orographic instabilities (here defined PAGE 67 as instabilities that grow in place) have appeared. In Fig.7 we have plotted isolines of the largest growth rate in presence of the of topography. largest instability growth problem In Fig.8 we have plotted isolines rate without obtained, solving topography; basic state that we perturb has no wave baroclinic instability. the same in this case the and we have pure As we can see the barotropic part of the flow influences the growth rate as well as the speed of the perturbation; this is because the describing the zonal flow always vanishes at the and a change in the barotropic Galileian transformation. the presence of a phase mode two walls part is not equivalent to a Comparing Fig.7 with Fig.B we see strong orographic that instability dominates the picture for low values of the baroclinic zonal flow near true the resonance. The indentation isolines in the upper part of the figure shows the influence of the of orography of instabilities: pure baroclinic stronger topography on those unstable modes that, also without topography, were growing almost in presence on the place. In (Fig.7) we have essentially two kinds for high values of and instability 61 the we have source almost of the perturbation energy is the available potential energy of the zonal flow; for low value of & and close to the true resonance we have almost a pure orographic the instability and forced stationary wave is the main source of energy for PAGE 68 super-resonant the instability as in the The barotropic case. line of Fig.7 indicates the stationary solutions heavy of the C--S model for different values of the forcing and see that some of these unstable, to "internal" some instabilities, are stable, or weakly solutions perturbations suggesting topography), of we wavenumber (same as are the dominant that, if these these equilibria can play an important role in the time behavior of the system. of At this first level find a stable stationary truncation we could Another subspace that solutions. has been studied by C-S is the one including modes with same theu number; wave stationary solutions The effects of this found the but with a higher zonal wave number of the topography, meridional still all that their were unstable to these perturbations. y-mode higher instability have been collaborators (private comunication) who ran a two-layer highly truncated spectral by studied Arakawa model including ones; they found and his one zonal wave number and three meridional that, instead of multiple stationary solutions,1 multiple statistical steady states can be reached resemblina the C-S multiple equilibria. In this way they have shown that these instabilities are not able to take and to keep the system completely away from the they weakly modify the equilibration. equilibria but In particular, these unstable modes participate in the meridional heat transfer PAGE 69 meridional value equilibrium temperature lowering the gradient. At this second level of truncation they show that there may be stable quasi-stationary solutions. baroclinic unstable containing the most the C-S consider the subspace we when arise equilibria multiple of importance Serious concerns about the the With wave. parameters chosen in this model the most unstable baroclinic wave has the largest meridional wave lenght and direct topographic forcing on the most wave baroclinic unstable zonal number five and we study a new instability with probleem, wave we neglect the influence its consider To five. number zonal to adding in problem can be factored the describing of number five. This the first sub-problems: two zonal flow and the topographic forced waves and that we have just studied instability wave zonal for equations analogous the (3.1)-(3.2)-(3.3)-(3.4) Eq.s wave the second the describing number five in a constant zonal flow. In Fig. 9 we have plotted isolines of the largest growth rate obtained in this wave number five). stationary states Fig.9 (heavy lines). problem (no topography and To understand the stability of zonal the C-S for the most unstable baroclinic wave solutions we have superimposed stationary second the for curve different that delineates the values of the forcing C-S in As we can see only the Hadley solution for small value of the forcing is stable. PAGE 70 Stone real (1978) has shown that at middle latitudes atmosphere baroclinic result is very close to the neutral value for instability in the two-layer model) and the solutions require The fact that the C-S indeed does equilibria. it not To has been stay prove shown close two zonal wave importance, (Reinhold, 1981) to any of these that the multiple the preceding point Reinhold used a two-layer highly truncated model and stationary a meridional temperature gradient almost twice as big raises some suspicions about their system same has been confirmed with a numerical time integration of a two-layer model. and the numbers, including one the zonal directly forced by the topography and another more baroclinicly unstable, with meridional interactions. wave numbers, to flow allow some two non-linear PAGE 71 5. Search for more stable equilibria From Fig.9 equilibria we are have seen strongly how all the baroclinically C-S wavy unstable. To understand if this negative result is a consequence of neglected important physical assumptions of the C-S model. parameters involved and, verified that this model in phenomena, we some studied the We started by considering the as pointed out by Reinhold, we is sensitive to reasonable changes the parameters, but is also very consistent in producing multiple equilibria which are strongly baroclinically unstable. To understand the reason for must understand why their Because . we the non-trivial balances between form drag and dissipation are obtained only for 9B instability high values of this is a complex balance we did not find a simple explanation and we decided to compare this balance in our model with the one in the real atmosphere. The model or equation (3.1) states that, in equilibrium in a long time average, the mountain acts to balance the dissipation of zonal boundary layer. This observational studies torque and surface Bowman (1974) found torque and the momentum due statement on the friction. that, on to does relative Newton the the Ekman not agree with the role (1971) average, surface friction work surface of mountain and Oort and the mountain together against the PAGE 72 convergence of angular momentum flux, neglected in the Reynolds' stresses our model. It is interesting to point out that this problem not is in the barotropic studies where the zonal flow present directly forced, and that also if we compute used by Charney and DeVore (1979), we get a somewhat higher still realistic value of the atmospheric convergence of momentum due to Reynolds' the computed In stresses. Table 1 we the 60* N for 30* to we calculation this for the data of Oort and using seasons extreme Rasmusson (1971); divided the in two layers, one from the ground to 500 mb and atmosphere As we can see the other from 500 mb to 50 mb. source a have barotropic and baroclinic convergence of eddy flux of angular momentum in the latitudinal band from are the of momentum necessary to give the zonal forcing convergence but is of both barotropic the tropics and baroclinic angular momentum for the mid-latitudes, and these momentum exchanges are (that neglected condition, boundary in channel models when the meridional v = 0, is applied) can play an important role on our stationary solutions. To take this momentum source for the zonal flow into we account ( , F equation: decided to include a constant forcing, ), in the barotropic and baroclinic zonal momentum PAGE 73 C- 3 k (Z ±y The balance realistic - 9 for and 3t the -+(5.1) (&''(ti~)± + 3 barotropic part flux convergence. balance P The balance for the baroclinic drastically influenced; in fact momentum the the 91 , ( zonal flow is not so flux of baroclinic vertical can simply continue our discussion by we discussion of function . I" however , the form the drag data in the ( have to as becomes less positive; become that this is not the parameter interest, since the momentum forcing is lower layer. between a selected heavy line light line parameterized Presumably as will suggests range of atmospheric positive However the baroclinic flux of momentum will no of the wave field. Nw shear introducing a given by longer be so simple once this quantity is than the in in trying to set a larger new effective forcing the more momentum simply adds its contribution to the external radiation forcing and now To discuss the effects of this forcing we looked for the equilibria solving for zonal becames we can have stationary states where mountain torque and friction work together to plane. (5.2) ( In Fig. 10 the intersections const) and a selected const) define the stationary solutions for PAGE 74 e. and the particular values of the forcings As Irt. we can see the equilibria move in the right direction, becoming values for realistic more baroclinicly stable and, the of momentum forcing, we start to have, in the lower left zonal which waves long stationary with equilibria Fig.10, side of energy considerable are in baroclinicly stable for shorter, synoptic-scale waves. into taken been have Along with the other factors that account, we want to mention that changes in the ground ground the to (extrapolation parameterization friction instead of lower layer value being used to compute the Ekman dissipation, the Ekman term i.e., V V4z.. Eq. (2.1) has but not to the point improve, been replaced b y34) in of stability, the stability of the wavy equilibria (Fig.11). neglected, being until term physical interesting Another to parameterize, but which can difficult play a role in the stabilization of our equilibria, latitudinally varying moist convection. equilibria to simple Newtonian term * equilibria forcing. forcing heat due is heat non-symmetric perturbation for a our forcing we considered a the zonal same wave we introduced a wavy forcing Fig.12a, b, c, d In in Eq. (2.3). obtained i.e. with the to radiation and To get an idea of the sensitivity of number as the topography, now strong present we zonally varying The amplitude of the forced wave is 40 K and the heat the PAGE 75 time Newtonian scale is ten days, with different phase relations between the topography and the heat forcing. Fig.12b we From have the suggestion that when the heating is in phase with the topography multiple baroclinicly stable can be obtained. equilibria which are PAGE 76 6. Conclusion We have seen how, taking into account way the fluxes of angular regions, our highly stationary This truncated forced the baroclinic momentum in from two-layer a the simple tropical system can large scale waves stable with respect to instability of shorter synoptic scale state requires a barotropic baroclinic zonal have waves. zonal flow about twice the flow in reasonable agreement with what we measure in the real atmosphere. To discuss the properties of this equilibrium state, varying V ~v, and we consider a low order model with only one meridional wave number numbers: a planetary wave number three, wave, and with topographically the most two zonal wave forced wave, zonal baroclinically unstable zonal wave number five in our model. For small value equilibria with and no the tropical forcing 1g the obtained with a stationary wave number three and wave number > I unstable, 9 baroclinic instability, therefore the close reaching of to its five are strongly 9 (where dashed is the neutral value for line of shorter wave develops and 9 preventing baroclinically the Fig.10), and keeps the value of forced long wave from equilibria. When we increase becomes tangent to 0 = and the isoline tp constant (dashed line of Fig.10), we have PAGE 77 external the two situations depending upon the strength of fC radiation forcing. 0, value of For values of less than 91 = on the intersection between (the and the ) we can have a stationary solution in which isoline of the long stationary wave is the only- wave and forcing the external radiation and it the heat necessary to balance all meridionally transports present meridional the keeps temperature gradient below the critical value for baroclinic For values of instability. G while the isoline the between intersection 9, than greater F difference between constant and 9 to balance zonal and states the atmospheric in which long stationary waves are present and we have found that a determinant factor for realization the . We have explored the possibility of the reaching at long shorter baroclinic wave reaches an amplitude large enough to transfer the heat required system the state equilibrium the reaches wave forced 9 e. the of such a state is the presence of a barotropic momentum mid-latitudes. flux coming This angular from the momentum tropics the flux is especially strong in winter when examples of atmospheric more frequent. into blocking are PAGE 78 REFERENCES Multiple flow equilibria DeVore, 1979: J.G. Charney, J.G., J. Atmos. Sci., in the atmosphere and blocking. 36, 1205-1216. Charney, J.G. equilibria multiple in Form-drag and D.M. Straus, 1980: and propagating momentum balance. Qort, A.H., and its 28, 1974: interannual hemisphere. J. Atmos. Sci. Oort, A.H., wave 1157-1176. Sci. Atmos. J. and Bowman II H. D., torque waves Mountain torques in the global angular 1971: Newton, C. W., 37, J. Atmos. Sci., planetary forced, planetary baroclinic, orographically systems. instability, 31, 623-628. A study of the mountain variations in the northern 1974-1982. and E.M. Rasmusson, 1971: Atmospheric circulation NOAA Prof. Paper 5, Govt. Printing Office, statistics. 323 pp. Reinhold, B. B., stationary 1981: Dynamics of weather waves and blocking. regimes: quasi Ph. D. Thesis, M. I.T., Cambridge, Mass. 02139. Stone, P. H., 561-571. 1978: Baroclinic adjustment. J. Atmos. Sci. 35, PAGE 79 FIGURE CAPTIONS Figure 1 : Graphical ( . & 91 , ), solution of the equilibria in the plane represent lines of constant the light lines indicates balances between the the heavy line form drag and the Ekman dissipation. Figure 2a: Amplitude dimensional Figure 2b: of barotropic forced wave in non- the units. Phase, relative to topography, of the barotropic forced wave. Figure 3a: Amplitude baroclinic forced wave in non- of the dimensional units. Figure 3b: Phase, relative to topography, of the baroclinic forced wave. Figure 4a: Amplitude in non-dimensional Figure 4b: forced wave in the lower layer of the units. Phase, relative to topography, of the forced wave lower layer. in the Figure 5 : Resonant without (heavy dissipation dissipation there is channel model our two layer curve for still line). a line of presence In on the which dissipative effects balance and in P we therefore have a true infinite response. Figure 6a: areas Number of "Internal we have instabilities". In the white no unstable roots, while in the other regions we find 1,2 or 3 unstable roots as written. PAGE 80 Figure 6b: Number of "Internal instabilities" with zero phase velocity. In the white areas we have no stationary unstable roots, while in the other regions we find 1,2 or 3 unstable stationary roots as written . The dashed line represents the only unstable stationary root in absence of topography. Figure 7 : Largest growth rate for the "internal The heavy instability". line, representing balances between form drag and Ekman dissipation, indicates possible equilibria. Figure 8 : Largest growth rate for the "internal instability" without topography; i.e. growth rate for the baroclinic instability of wave number three. Figure 9 :Largest baroclinic growth mode rate for present in our the most unstable channel model: zonal. wave number five. Figure 10: ( Graphical solution of the equilibria in the plane ,9) in presence of a zonal momentum the light heavy Figure 11: forcing lines represent lines of constant line represent lines of constant ,the , where As in Fig.10 but with ground extrapolation to compute the Ekman dissipation. Figure 12a: ( ,01:) Graphical solution of the equilibria in the plane in presence of a zonal momentum and a latitudinal Newtonian heat forcing of 400 and a heating Newtonian time scale of is in phase with respect to the forcing (with amplitude ten days). The topography. The PAGE 81 light lines represent lines of constant lines represent lines of constant Figure 12b: is out of phase topography. As in Fig.12a but here the heating is leading the topography Figure 12d: the heavy where As in Fig.12a but here the heating with respect to the Figure 12c: e by 904 . As in Fig.12a but here the topography by 90". heating is behind the PAGE 82 TABLE CAPTION Table 1 :Barotropic and baroclinic eddy flux of angular momentum and its convergence between 300 in m/sec (or in and 60* north. our non--dimensional units). PAGE 83 20- .15- .10 5- 5 0 PT(0 1 ,10 .15 .20 f1 .25 PAGE 84 .20 .15 .10 5 CrrT 7n T .10 .15 .20 .25 PAGE 85 .2 5 al .20- .15- 10 .5 0.5 ,10 .15 - .20 .,-25 PAGE 86- .20 .04 .02 .00 .02 5 FIG, 3a 10 .15 -20 #, .25 PAGE 87 .25 0i .20- .15 - ,10- 5.- O0 CT 5 ''Th .10 .15 .20 *4s .25 PAGE 88 .25 0, .20- .15- .10- .5- .5 r-Tr / A/ .10 .15 .20 qll .25 PAGE 89 .251 .20. .15- 10 5 .10 .15 .20 *, .25 PAGE 90 .25 20 0 5 P T( Ci lo 5 5 e 20 2 .25 PAGE 91 .20 .15.\0- .10- O5 5 yr- r r% 1- .10 .15 .20 .25 PAGE 92 e, .20 o 15 .10 .15 .20 PAGE 93 5 PT0 7 .10 .15 .20 .25 PAGE 94 25 e, 20 15 10 .5 0.5 .10 .15 .20 PAGE 95 25 .18 .16 )0 .14 .12 15 .to 10 .06 00 .02 5 0 C-T f\ .. O 10 .15 .20 'I, .25 PAGE 96 .20 .15 .10 .5- 0 .10 .15 20 *pg .25 PAGE 97 .25 al 20. 15- 10- 5 .10 .15 .20 -25 .p PAGE 98 25 ei 20- I5 .10 .5 5 r- Ir flX -d 111 - .10 .15 .20 #i .25 PAGE 99 .20 #1I ,25 PAGE 100 25 200 .1=.28 .24 .22 .0 .10 - . .N -14 .12 9 10 .6-- .5IN .l O .15 .20 .25 PAGE 101 .25 e, 20 26 .15- .22 .20 .18 .10- .16 .12 .10o 0 '1 .5 >vas .10 .1 PAGE 102 Summer Winter Eddy fluxes of barotropic angular momentum (m /s ) : BR Eddy convergence of barotropic angular momentum between 30* (m/s) and 60 N. at 300 16.4 7.4 at 60" -3. 6 -1. 2 8. 5*10 3. 6*10 o * -3 .53*10 (non-dimensional units) at 30* Eddy fluxes of baroclinic angular momentum (m'/s ) at 600 - -1.9 (non-dimensional units) TABLE 1 - - -3 -0.7 - -. angular momentum between 300 (m/sE 23*10 4.2 7.0 Eddy convergence of baroclinic and 60* N. . 3. 8*10 2. 1*10 .24*10 .13*10 a C PAGE 103 SECTION 3 STATIONARY ADIABATIC FRICTIONLESS FLOW OVER TOPOGRAPHY 1. Introduction In this last section we shall examine quasi-geostrophic vertical analyze inviscid over stationary topography propagation of zonal momentum and energy; and the we then in some detail how these transports are accomplished in a simple fully non-linear shall flow a consider solution. In particular we the implication of a top radiation boundary condition on the mean meridional circulation. Our study clarifies the role of the stationary waves in forcing a mean meridional circulation and in particular suggests that there will be a strong meridional mass flux in the layer containing the topography. It also points out the necessity of studying the time dependent problem to increase our understanding of the mean meridional circulation in the upper atmosphere. PAGE 104 2. Equations of motion We study a quasi-geostrophic adiabatic flow beta channel. periodic zonally a in frictionless We assume the validity of the hydrostatic balance and we choose to work in pressure coordinates. the consider We following quasi-geostrophic equations: zonal momentum equation - 4 L- - - _( - (2. 1) +o (2.2) + (2.3) vorticity equation avk' thermodunamic -- - .J ( q+7 equation + J - -- mass conservation equation PAGE 105 I - -- 0 (2. 4) and the following energy equation I ( - es ,I±f.I 4- (( S -) - ( .W adding > where T / (2. 5) (2.2) by , (2.3) by IVq1Z J3t- which can be obtained by multiplying S ')P TT"D 41 tP them together and averaging over a p-surface. LAL eastward V northward component of the geostrophic wind component of the geostrophic wind eastward component of the ageostrophic wind V northward component of the ageostrophic wind Coriolis parameter = t 2.bvMMf Coriolis parameter evaluated .j at mid-latitude derivative of the Coriolis parameter geopotential y V + stream-function for the geostrophic wind horizontal divergence or gradient d4- PAGE 106 C2D individual pressure change d1 dry static stability = - zonal average at constant p and y C horizontal average at constant pressure PAGE 107 3. Simple theorems We now proceed to prove three statements. Theorem 1. For a stationary channel flow the flux of energy, Proof. This thermodynamic kV total upward must vanish. is a straightforwar-d equation. In fact, consequence multiplying of the (2.3) by and horizontally averaging, we have: CV -- - A4I~y /?3(V =- ' I Dq (3. 1) where we have used the zonal periodicity and the presence of lateral meridional walls. Lemma 1. energy If we upward have we a must stationary wave that propagates have a mean meridional circulation that transports back the same amount of energy. Lemma 2. No level Theorem 2. to can be a net source of energy. For a stationary channel flow the form drag due the topography must be balanced by the Coriolis force in the layer containing the topography. Proof. We start from the zonal momentum equation (2. 1), PAGE 108 and, to avoid the problem arising where p-surfaces intersect the ground, (quantities we integrate evaluated at vertically from ground the ground are hereafter denoted p = P (xy), to a with the sub-index g), the generic level p, then we take the spatial derivatives outside the integral to get P PF P IDX /ZX p P where the term in the (u L-) ) (3.2) PCP has been neglected been Rossby number. Taking a p-level, order second p = P just above the topography we can perform a horizontal average. All the terms having in front a horizzontal derivative vanish due to our zonal periodicity and our meridional boundary conditions and we are left with - V+/cP P Lemma 3. = (3.3) If the net meridional flux of mass is zero at latitudes valid e-3 once particular the p form Is drag must van i sh. lower we can choose Because we a Te than any all In fact Eq. (3.2) is ground pressure; in p = 0. working with quasi geostrophic PAGE 109 equation we conclude that the form-drag must be a second order term in the Rossby number. this result To point out more clearly we can do the quasi-geostrophic scaling on the zonal momentum equation D X /-k /~, Dp /D where now (U,V,cO) are the full components of field. x the velocity We scale all our variables in the quasi geostrophic framework L/ ((L )) / L. where Ro is the typical scale value the number Rossby of vertical and P the the ground pressure, has been choosen to variations of the pressure field. Substituting we get -, 4 V 4- -, r- PAGE 110 Integrating from P to P=0 and horizontally averaging we find: D 4) Q~J' P and assuming no meridional mass flux Z 1 _x >. ;IAw~ Transforming the x-derivative following to the ground x-derivative at constant height, using the definition of geostrophic wind in z-coordinates and neglecting horizontal variations of the ground density, 1 4) the we can write: "R7('' where now H =(g h/R h/H /P0; 2 Ro, /10J VyI'J7 A/- H is the Assuming we-find that internal a -the height realistic part of the 1 as scale defined small topography, meridional wind PAGE 111 correlated number. waves with the In the topography linear theory must of be of order Rossby topographically forced this does not hold when we .go close to a resonance so that the linear theory must break down. Our third drag and the statement is a relation between vertical the form component of the Palm and Eliassen flux. We start by considering the zonally averaged zonal momentum equation .- x& f/'X 4- We write the ageostrophic wind (3. 4) using the zonally averaged continuity and thermodynamic equations V= We boundary CDPl'f77 --- integrate condition V once in V 0 y, using the meridional on the lateral boundary, to get jX .-- I --a P 5 (3. 5) / PAGE 112 X We then substitute (3.4) getting .-0/'O 7-VX VX where now on the right hand side we have the the Palm and Eliassen flux divergence (Andrew and McIntyre, 1976). second term on the R. H. S can be interpreted as convergence flux of (3. 6) "geostrophic vertical a of The vertical of zonal momentum." Theorem 3. vertical and it For a stationary flux flow the total "geostrophic of zonal momentum" is constant with pressure is related to the topography by: - LA Proo f. We consider -p e-D 10 X the zonal X k4- averaged (3.7) thermodynamic equation on a pressure vertically surface integrate P above the topography, then we the conservation of mass from P (xy) PAGE 113 to P0 and we zonally average to get: = S Equating these two - V+ V') -- expressions for -Z3 X and integrating once in the meridional direction we get: -D (3. 8) P Using neglecting rewrite Eq. (3. 3), second--order substituting geostrophic __ -~ proving the theorem. streamfunction, x-derivative outside the integral sign integrating once by part we find: -V and ~4~ substituting the definition of partial -Py X14P the f" as x eA taking by terms in the Rossby number, we can the R.H.S of Eq. (3.8) 1 and f PAGE 114 Lemma 4. For small topography _____x ~L v~V .5 P R[D (3. 9) comes from (3.8) This result, without any y-integration we neglect the the topography; if coupling between the ageostrophic wind and i. e. we neglect p PAGE 115 4. A simple non-linear analytical solution We now solve a particular case and we study how these requirements are satisfied. We use equation, the quasi-geostrophic obtained by eliminating potential e-) vorticity between (2.2) and (2. 3), - (4.1) To define the bottom boundary condition we consider that the ground is a material surface: Where in the stationary case the ground pressure only, a function of x and y. Transforming the horizontal derivatives following the ground to at constant height, is horizontal derivatives and using the hydrostatic balance, we can write: (V~v) -~ ~~S 4 3~ O~.Vh (4.2) PAGE 116 In the last expression we have neglected the ageostrophic pressure, which topography and ground density level P0 , instead that of we assuming assume a a constant (4.2) on a constant pressure. on p = P 1 Eliminating (2.3) and ch between the thermodynamic equation (4.2) we get J~n4} 9) the f(4.3) 3O An analytical solution of (4. 1) and assuming condition consistently and we apply advection is a second order term. We approximate this boundary small the (4.3) can be obtained quasi-geostrophic potential vorticity to be a linear function of the geostrophic stream-function 4 - -- ;a (4.4) and assuminQ -P Putting -~ in explicity the barotropic zonal flow (4.5) PAGE 117 . and setting C - --A/U = - 4 (4.6) we can rewrite (4. 5) and (4.6) as: I ~ -s ~~Z:~p U (4. 7) These equations are linear and can be easily solved for any assigned topography. To solve we approximate, the dry static stability as: Wiin-Nielsen (1959), _S where H problem the following V V4 z VO 4-, at mid-latitude is 1-4 . approximatly AAA 6 1.07 * We define the following non-dimensional quantities: T0P L 4z L >' H h ' (4.8) PAGE 118 where 7i-L reference is the meridional width of the channel, pressure, P = ground and H is an internal With these scale substitutions coordinate 1000 mb, height, and H k is a To Po J introducing . 8 km. a new vertical Eq. (4.7) becomes: (4.9) Assuming no topography on the meridional walls of Because of the channel we can use a Fourier expansion m' 2-M I Aim X4 ) rL =A. 1- Where the and L is the channel lenght. linearity of the problem we can consider each component independenly. -/3 L( We must consider two cases, for L (pp)Lz~ the forced wave decays exponentially given by: L7 and the solution is PAGE 119 X+ + C (4. 10) and applying the bottom boundary condition we get 'Y;o MI (4. 11) topography We see that the forced wave is in phase with the and theref~ore the form drag vanishes. Uii L- the second case, In L - ZC-Z the solution has upward, A and A' and downward, B and B', energy propagating waves and it is given by: Ada X +Am 4- 4-? Cr+ cl (4. 12) and applying the bottom boundary condition we get A(A-A = ( B +B) ± (4--B) + PAGE 120 Applying a top radiation boundary condition to determine the solution we have: B 0 (4. 13) I A 4 -L4 =7, - -3'LU R Because the solution is not in phase with now have a topographic form drag. the topography we To get no form drag and to satisfy the bottom boundary condition we must have A i.e., we FI= B= B'= O 0= have two waves with the same -- Al amplitude, one propagating upward and the other downward. The most interesting solution is that obtained the top radiation boundary condition because it does not require a source of wave eneTgy at infinity; a single-mode topography with in presence of this is given by L(4.14 (4. 14) S-Z 'U4 c.'O" ( J4X, -- AU) PAGE 121 As we have said, hence, for this solution transfers wave energy upward; Theorem 1, we circulation that transfer energy must energy have a mean meridional downward. The vertical flux due to the mean meridional circulation is given TL by The ---x vertical motion can be computed from the thermodynamic equation (2.3) (4. 15) and the zonal average --X c. reduces to (4.16) independent meridional of the motion pressure. we have a Hence, downward due flux to the of mean energy, independent of height, that balances exactly the upward flux of wave energy. To understand the source of this energy at infinity we should consider the time dependent problem and PAGE 122 study the upward propagation of the wave front; meridional circulation should be front set up behind the mean the wave that should be also the source of the energy which is advected downward. From Eg. (4.16) we also see that at some latitudes we get a constant upward mass flux starting from the surface p =P and leaving the atmosphere while at other latitudes get a down mass to flux the we entering the upper atmosphere and coming p =P surface . From Theorem 2 we can anticipate that in the layer containing the topography there will be a net meridional lower part of the transfer of mass able to close meridional upper part of the mean meridional we should, as the wave front. circulation. the To close the circulation we think that we said before, study the time evolution of PAGE 123 REFERENCES Andrews, D.G., and M.E. McIntyre, 1976: horizzontal and vertical shear: Planetary the waves in generalized Eliassen-Palm relation and the mean zonal acceleration. J. Atmos. Sci., Wiin-Nielsen, A., 33, 1959: 2031-2048 On barotropic and baroclinic models, with special emphasis on ultra-long waves. Mon. Wea. Rev. 87, 171-183.