THANS TION FROM ONE WAVE REGIME TO ANOTHER' I-1 A TWO-LAYER GEOSTROPHIC MODEL U by MAN-KIN MA B.A.8$,University of Toronto (1903) SUBMITTSD IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE at the MASSACRUSETTS INSTITUTE OF TECHNOLOGY January 1986 Signature of Author. . . . . . Department of Meteorology, January 1966 Certified by. Thesis Suverzvisor ( Accepted by . Na Cirm,, bepartmental dwittee on *Graduate Students Libraries MIT Document Services Room 14-0551 77 Massachusetts Avenue Cambridge, MA 02139 Ph: 617.253.5668 Fax: 617.253.1690 Email: docs@mit.edu http://libraries.mit.edu/docs DISCLAIMER OF QUALITY Due to the condition of the original material, there are unavoidable flaws in this reproduction. We have made every effort possible to provide you with the best copy available. If you are dissatisfied with this product and find it unusable, please contact Document Services as soon as possible. Thank you. Poor text quality. This is the best copy available. 1' ' -- - I - - Iq TRANSITOA FRhM ONE WAVE REGIME TO PNTHER IN A TWO-LAYER GEOSTROPHIC MODEL by Man-Kin Mak Submitted to the Department of Meteorology on December 24, 1965 in partial fulfillment of the requirements for the degree of Master of Science ABSTRACT A truncated spectral representation of a to-layer geostrophic model with 'parameterised effects of friction and heating is used for investigating the steady states as well an the transient behaviour of an idealized rotating fluid when it undergoes a transition from one wave number to another in the Rossby regime. Wave number 2 and 3 are studied in this paper. For the equilibrium states, it i found that the total energy of the system is independent of the thermal forcing, &A , as is within the range where both waves coexist stably. long as o A given increase of $* leads to a certain increase of energy associated with wave 2 and a decrease In equal amount of energy associated with wave 3. but the zonal component remains unchanged. appears The transient behaviour following a change of / to have throe distinct stages of development. In the case of increasing 8,f from the lower to the upper bound of the transition range, the initial development is a growth of all components; for the smaller rate of increase of I a the initial developmat lasts longer but has a weaker intensity. The second stage of development is characterised by an exponential growth of one wave and decay of the other wave, the growth rate being relatively independent of how &A is actually varied. The third stage is a gradual readjustment to a now state of equilibrium. No transition can take place if we hold the static stability fixed even though its initial and final values are known to be equal. A dynamical account for the outlined major characteristics is proposed. The transition phenomenon is viewed as a chain of processes with baroclinic instability of the waves with respect to the zonal flow as the basic mechanism. Variability of the static stability serves the role of intensifying the sonal flow until instability begins0 and thus is an indispensable factor. It is hoped that this study sheds some light on an analogous, but much more complicated phenomenon of wave transition observed in the dishpan eperiment. Thesis Supervisor: Title: Edward N. Lorenz Professor of Meteorology TABLE OF CONTEfT6 INTRODUCTION TWO-LAYER GEOSTROPHIC MDDEL 3 EQUILIMIUM STEADY STATES 7 TRANSIENT BEHAVIOUR DURING TRANSITION FROM ONE WAVE REGIE T ANOTHER 17 EFFECT OF VARIABILITY OF TE STATIC STABILIT 27 CONCLUSION 34 APPENDIX A 37 APPENDIX B 42 ACKNOWLEDGEMENTS I wish to express my thanks to my thesis advisor, Professor Edward N. Lorenz, for taking active interest in. Inspiring much optimism of and giving some valuable advice throughout this research. I also greatly appreciate the enthusiastic spirit of my follow student, Mr. John Young, in many of our unending discussions. My thanks are also due to Mles Pauline Look Bong for her able assistance in helping me to prepare the manuscript. NRODUCTION The dish-pan apparatus was first found indepeandenty by Fultz a'd Hide (see Fultz (1953) Hide (1953)) to be a powerful tool for stdying atmospheric motions, particularly those of planetary scale.o Such experimental flows and the observed large-scale atmospheric flows have been shown in many studio but als to be not only visually similar dynamically and energetically similar. Specifically, Fults and Hide demonstrated that axially symmetric flow, asymmetric flows and essentially turbulent flow can be reproduced by controlling the combination of the heating rate and the rotation rate of the apparatue. The symmetric and asymmetric flows have since then been named the Hadley regime and the Rossby regime respectively as analogs of the atmospheric flows observed in the tropics and the mid-latitudes. Fults and Hide also obtained quantitative criteria for the transition between the Hadley and Rossby regimes and the transition between different wave numbers within the Rosuby regime. Since then several theoretical hydrodynaists have attempted to explain such observations withmathematical models. (1956) and Lorenz (1962). that of Lorenz. Notable are those of Kuo (1954)0, Davies Of particular relevance to this study is With the use of a two-layer geostrophic "numerical- predicton*' model 0 and with the technique of spectral representation of each dynamic and thermodynamic quantity of the system, Lorenz was able to obtain stability criteria for the symetric and asymmetric flows. His theoretical stability criteria were found to agree qualitatively with the experimental resulti. However, his analysis is only applicable toi the steady equili- brium situation. It Is incapable of describing how the transition from one wave number to another actually occura. Thus we cannot deduce what the mechanisms governing the rate of evolution are. This study is primarily aimed at answering these unsolved questions. Before we investigate the transient behaviour, we will first verify the theoretical predictions of the steady states, with particular interest in the variation with A of the distribution f the energy among the various components of the system. In this study we adopt Lorens's basic approach,) but usenumerh-.a IntegratAon' to investigate the following aspectn of the problem of wave transition: (1) Structure and energy of the steady states as functions of the externally imposed heatingr (2) and 9A Transient behaviour of the eyatem under four different types of time varying thermal forcing 1; , namely: (1) sudden increase of 6 from an initial value to a final value (1i) e is increased hyperbolically towards a final value (111) exponential increase of 6A* (iv) sinusoidal fluctuation of Q* (3) The effect of the variation of static stability on the wave transition within the range of in finite amplitude. A where both waves coexist i-.- c~m~ ~. Th~ 2'w~ y~ The physical Bystem that we wish to investigate is a rotating cylinder with a controllable temperature distribution at the bottom. The numerical model that is used for representing this system can best be described pietorially as follows: c~ o =o - * 4+ -------> q e 3 - * W=O 4 .7 -,ez rr -P ------------- 0 ~sII, jill t-T -aC -2X cz k The mathematical formulation of this system wa first done by Lorenz (1962). For the sake of completeness the detailed derivation is given in Appendix A. Only the subsequent set of equations that govern the dynamical and thermodynamical quantities are presented in this chapter. Lorenz*@ approach is based on the idea that each physical quantity be adequately approximated by a truncated series of orthogonal functions. II- 4' The Choice V9 the orthognalu the system to be studied. dicatd by the geonetry of In the cae of a cylindrical domainu Besel expansion for / o T 0 and a V series after the five terms. Is appropriate. We truncate the Such a truncation correeponds to suppressing all components except the sonal . the wave number 2 and the wave number 3 components. The expansions are as follows: 0- t '433 2 2 21 -1 Where C-= radius of outer cylinder = depth of the liquid = coefficient of thermal expansion = gravity f Coriolis parameter As shown in the Appendix A 9A , ,- 'r 9 1 for 71 and also the quantities = 2.3 are governed by the following set of 12 ordinary differential equations. For the special choice: (2) 0 (i) 6,=-ff Udi -act-19 ,') 4 £>4,- 9 40 =Z - L Lv) &3d- , ,174 A') 2,1] 00 2 N-414 1') 0-44i7+2-rX ) tJn: r2~~o~ (] X) v - ,-c,.F - 4- ./~IAhJ~~PAl -Q] (X;) (Y;;) ~ 4 = r.<.- A = &( 0 -4i * ' 0) + +- ,-p 4 C, (te ,0 o F,by -4] 40 +&[ iA3 9,,z r.3. 7 ,'4 (4 -j +2 ) ,,)+ +0 2' 0 '# -i*) ,,j Ilipiil bU il lloilillillkil, , 11, 6. It is very difficult to study this set o2 equations with an arbitrary forcing $ . We therefore resort to the technique of numeri.4 cal integration. Although this is aesthetically less appealing and loss conclusive than analytic analysis, it nevertheless is a powerful tool for bringing to us some insight about the behaviour of this system when two finite amplitude Rossby waves simultaneously interact with the thermally forced sonal component. It is a relatively staple matter to numerically integrate such a set of ordinary differential equations. However, in order to avoid computational instability and excessive truncation error, care should be given to the choice of the integration scheme. procedure is used in this study. A double-approximation Such a scheme has been showin (Lorenz, 1963) to be appropriate for both periodic and non-periodic flows. The validity of this computational procedure is checked by applying Lorenz s analytic expressions to the quantities of the steady state obtained with numerical integration. If those expressions agree with the com- putations we can then say with some confidence that our computational procedure is reliable. CHAIPTER Equ1A-Ai liM S§ta& tesa hbat of the features related to the various equilibrium ster icy states of such a system have been investigated analytically by Lorenz (1962), Of particular relevance to this study are those that are concerned with the coexistence of two Rossby waves. these two waves by wave numbers/4 and )) * We designate For mathematical cot- venience we only permit interaction between each wave and the son1 component and also we choose the frictional heating coefficient to Z 2 Lorenz found that for a system in which all wave allowed to exist e numbers are if two waves are to coexist stably the wave numbe s must be consecutive integers, i.e. and P =7 f / . Two stabIl4 ty criteria have to be satisfied simultaneously. (3) 7 A (a) (b) & A P Furthermore, the other quantities of the system are found to satisfy the following relationship: (c) (e) + A (2) ,k; 4 I 44 (k) (J) *- A + AJ p =os / (n) (0) (n) (p) (ci) &<;4 These equations would degenerate to those that describe the system consisting of one wave, say wave number/ . if we put A and remve all quantities with subscript P An shown by Lorens, the stability criteria when represented graphically, do have a strong resemblance to the corresponding experimental curves. But in view of the omlicated algebraic relation, it opet 4S difficult to estimate the relatIve Magnitude-v oft02rou ln facto the thaoretical of the flow for dlfferent thermal forcings. analyals does not yield a functional dependence of on and and therefo / With the aid of a computer, we cano however, read1.ly * obtain that information from the various steady states corresponding This is done by numerically integratIng to different thermal forcing. the original set of equations with an arbitrary initial condition. This is indeed the starting point of this study. Our objective is t find the quantitative dependence of the steady state on 9 within The theoretical results described above can the transition range. thereby be checked directly. The friction and Wave numbers 2 and 3 are used in this study. heating parameters have the following values; =0.1 # = Ge =1. I 0. is varied over the range from 0.8 to 1.8. An arbitrary set of initial conditions is used. It is found that the two waves coxist stably for the range of from 1.00 to 1.28. Lorenx's analysiso range of 0' temperature , A As predicted by all do not vary within that and a whereas the wave amplitudes of the stream function and J dependence on ., A and . 70 =, 3) do have a dofinite 7 These results are shown in Figure 1. Before we discuss their physical implications, we should first ascertain the reliability of these results. (3c) To do that we compute with equation based on the quantities of the steady state obtained by numerical integrations and then find out by how much this value differs from the actual value of 0 used. A This was done and it was found that iMiUi*iinin.~ W1111iiiia~m I to. ~ "4 .5A Fig la Streamfunction amplitude of the waves, vs thermal forcing, g , A Pig lb Static stabilit', zonal temperature, O in equilibrium states. zonal streanfunction,4 , , A vs O A , and in equilibrium states. *9987 While P= 1.00. Thum we have a Uab- stantial evidence concerning the validity of the double appromination procedure. We now examine the physical implications of the results displayed in Figure 1. That ; ,, 7 remain constant implies that the strength of the zonal flow, the zonal temperature field, the static stability and the overall average of the temperature field are uneffected after a partial or complete transition. It is difficult to tell how the waves* amplitude varies with &A. However, if we plot the square of the wave amplitude from Fig. la instead of the amplit;ude itself, against , we would get a simple linear relationship. This & is shown in Figure 2. This finding immediately suggests that there might be some simple energetic relationship between the two waves in the transition region. With this expectation we will first examine the analytic expression for the energy of each wave and then seek a simple link between them, which hopefully will be consistent with the finding mentioned above. The derivation of the energy components of the system is quite straightforward. It is presented in Appendix B. According to equation (B-4b) and (B-40), total kinetic energy of the waves (EAT) - 2 Substituting equation (4g), (4h) we get Fig 2 Square of the streamfunction of the maves, ('+ W,(,) vs and stateso '30- 2') '2-0 ( NK ) '1 .,.I /.' i equilibrium 13. k3t A Since according to (3-p) and (3-q)o are constant for the range of , slop* is equal to ... IA)(/f* ) thersal forcing. Figure 3 Is such a Indeed we find a straight line whose slope is -. 73. (3-p), (3-q) we find that ) the we can then say that plot. (/ 2(/--4 is a linear plot and if A+z - e u remains constant for ( where both waves coexist stably. 42 ( Thus If are functions and alone, the two coefficients of i EKE A Using 0/7 This is the physical implication, implicit in Fig. 20 that we set out to seek. Now we are in a position to describe how the various energy components vary with 0 in steady states. For sufficiently low value of 19* i.e. weak heating, we should expect only wave number 3. The sonal ft. (A)-D the .E of wave 3 ( K ) and the total potential energy (TP ) are expected to increase with tion region i.e. - 6 until 1.0. # reaches the lower limit of the transi- Beyond this point E , 7P( and Cr remain constant, and wave 2 progressively intensifies at the expense of wave 3. When reaches the upper limit of the transition zone, 3 virtually vanishes. From thereon the system absorbs energy again through the bottom; A ,Pe bigger. ~ 1.25, wave as well as K2 all become progressively In view of other studies, we should expect that at a certain Fig ( (~ &+ 3 W3 FWg ). . vs S.W * ot )j,)within the range of C4 'IiIi Ii,1 4 1O-1.25 1o. otageo wava 2 also otops growing and avenitually regime will dominate. i e da Hadly Since our study is merely one coneemring the phenomenon of transition between two Roasby waves, our investigation stops at the point where wave 3 clearly vanishes and wave 2 Is clearly well-established. Knowing the various steady states of the system for a large range of f (see Table 1)0 we can readily verify the description of the variations of the various energy coWonents. Figure 4. The previous description is thus proved to be accurate. TABLE 1. (kA 4 klPg 42, A ex Steady States and Energy Variations with .103 .0000 .2004 .0 .0000 .0 .2018 -. 2474 .183 -. 0000 -. 1513 -. 3400 .2003 .0 .2158 .096 .2158 .2683 .2203 .2405 -. 0220 -. 4434 -. 1883 -. 0089 .2763 .2818 -. 0622 .4601 *2167 -. 2093 .2157 .1160 .1053 .1411 .2157 .2092 .0 .0 .2203 -. 2569 .0 .0 .2404 .1034 -. 0492 .1535 .0225 .1696 .3109 .0844 .0 .0 -. 0000 4 , 04 6. -. 0224 -. 1741 -. 1781 -. 0743 1.3379 .3379 1.3830 .3830 A' .0435 .0 .0507 .0276 .0 .188 *1858 .0517 .0 .0409 .0400 .0 .2414 .2414 Ag K3 1.2703 .2703 .0 .2158 -. 3578 -. 1299 .2158 .0460 1'1 A, They are shown in -. 1737 .0119 .0881 -. 1472 1.3830 .3630 1.3630 .3630 1.3781 .3781 1.4419 .4419 .05061 .0055 .0420 .0466 .0125 .2489 .2614 .0547 .0435 .0256 .0465 .1022 .1560 .2562 .0534 .0837 .0085 .0465 .2014 .0832 .2546 .0531 .1037 .0 .0485 .2650 .0 .2680 .0554 .1030 .0 .0578 .3175 .0 .3175 .0178 .0 .0 4' Fig 4 Kinetic energy of variouB components equilibrium states* .25- 2K -2I' 1.4 vs Ain C.APTER I I Transient BehaAour of the System durinM Transition frM Oe..PredJminatt ae_Numb2r-to A thLr The reason that we want to examine the transient behaviour of the system is twofold. First, the transient behaviour is interesting and important in its own right. Second, it is hoped that studies of the major features during the transition evolution of the system might lead to a deeper understanding of the transition phenomenon. To fulfill the first objective Is essentially to give a synoptic description of the evolution of the system as revealed in the numerically integration. But the second objective is a mauch more difficult one to be fulfilled. The synoptic picture is most probably the net effects of several unknown interwoven processes. We should therefore follow the approach of a chemist who tries to aalyse the structure of a complicated molecule. In other words, we should change the controllable factors one at a time and examine the corresponding effects in the hope of being able to isolate the crucial factors, In this problem, there are two factors that we can control and hope for interesting results. thermal forcingo Of One is the wave that the , approaches its final value. or not the static stability is allowed to vary. The other is whether Although a lot of other parameters such an the frictional and heating coefficients can also be changed arbitrarily, doing so is unlikely to yield qualitatively difforent results. Figure 5 shows the evolution of the amplitude of the wave components of the vertical mean streamfunction for three different types of variation of 9A from 1.01 V;to 1.20. Curve (1) corresponds to PA being increased stepwisely; curve (1I) l.-IIly. hyperbo'lically; curve (MI) egonn Figure 6 shows the evolution of the static stabilit1 and the zonal component as gall as the variation. The main features of the evolution of the three components can be summarized as follows: 1. The faster 0 is increased to the final value, the will the transition be completed. 2. A is Except for case (I) where the rate of increase of infinite, the initial growth rate of wave 2 is very much smaller than the growth rate during the major portion of the transition. Curves (I), (II) and (III) of wave 2 are to a first approximation parallel to one another. The slope of the main portion gives the major growth rate of an e-folding time equal to 200 non-dimensional units, or equivalent to 16 revolutions (days). ooner 3. Wave 3 also has an initial growth although it subsequently decays exponentially to a smaller steady value. The portions of the curves that correspond to this exponential decay for the three cases are also parallel to one another. 4. The zonal component and the static stability also increase initially for about the same period as wave 3 does before they return to their original value. 5. The initial growth of all components have the following comn characteristics: the faster 6A is increased, the smaller is the period of the initial growth, and also the larger is the initial growth rate. Furthermore, the period of the initial growth of all components is roughly the same. 6. The phase speed of both waves changes in a similar manner as the zonal flow does. Feature 1 is what we would intuitively expect, but it yields little information about the underlying processes. Feature 6 indicates that the waves are advected along by the monal flow, thus their phase speed is proportional to the strength of the zonal flow. -ir- "4 4oo a 0 4e'o Fig 5 Evelution of 4 (1) 6 goo 40/000 = 1.01 = 1.20 +)f and e t=0 ot>0 /00 zoo 4eo/zoo with time for three cases (II) -= i.01 + (.19)tanh(.02t) (III)tfA*;- 1.01 exni(.oo14t) 4/-ZO therea 1.20 Fig :A Varition of the t- therrnal forcing with time 3o Fig it Evolution of the 6Ob static stabilit/ T0 '3o' 4 .25- /. ,, ;00 Pig & i; olution of the ,e. zonal streamfunction i 2R0 Featurer 2 to v tholwn Umplications. Ao -a002ew waves are concern&d. the whole transitaon can be divided into threO stages: an initial development, a major development and a final adjustment to a new equilibrium. The initial development i qua7- tatively different from the major development. For wave 2 the Itajor development is much more intense and lasts much longer than tk e ' initial development. For wave 3 apart from the large difference In magnitude, the initial development is even in opposite senso of the major development. Above all, the initial development has a distinct dependence on how the thermal forcingo I e is varied, Wile the major development is relatively independent of it. Such difference suggests that the initial development is of a forced neture and the major development is of an inertial nature. Since the major develop- ment of the waves begins at about the time when the rxrnal flow and its vertical shear reach their maximum values, it it suggestive to associate this development with a baroclinic instability process of the waves with respect to the sufficiently intensified xonal flow. This suggestion is consistent with the fact that there is not a counterpart major development in the evolution of the zanal flow because during the baroclinic instability process the growing wave gets its energy mainly from the decaying wave. So it appears that the transition is a consequence of baroclinic instability of the waves with respect to an intensified zonal flow which is caused by the initial development. In order to complete the reasoning, we need to know the crucial factor responsible for the intensification of the zonal flow. That piece of Ii ___________________________________ ______________________________________ missing inte~rnation0 Is founMd by the following attempt Two sets of runs were made with the objective of exploring the effect of the variation of static stability. e Is In each run, A stepwisely increased from 1.01 to a final value with the equilibrium A= 1.01 as the initial state. state for free to vary while in stability is The result is In one set (A) the static the other set (B) it is held fixed. very remarkable and indeed is against what one would Figure 7 shows the evolution of the kinetic intuitively expect. energy of each wave in differenO cases. where is that for set (B). The most surprising feature = 0s no transition from wave 3 to wave 2 can occur. Even though the initial and final values of are the same for 0A within the range from 1.0 to 1.25, the whole transition process is suppressed merely by suppressing the variability of the static stability. If we examine closely the effects displayed in Fig. 100 we find that the initial growth of the zonal flow Is also suppressed although that of wave 2 and 3 still exists. So we now come to preliminary conclusions - variability of the static stability is a crucial factor responsible for the intensification of the zonal flow that subsequently leads to the baroolinic instability of the waves. As to how this factor works from the energy point of view is yet to be explored. This aspect is the subject matter of Chapter IV. It is of interest to check whether anything qualitatively different might occur if 1.20 to 1.01. PA is decreased in a similar manner as above from The results are given in Fig. 8. The basic features are again apparent. So the proposed dynamical account should also be ------------- I- h' , - c/A - I -. 1 / l.~o / I Z%0~ tA =/-&o /0o /0o /90 and / , due to a stepwise increase Fig 7 variation with time of the waves kinetic energy, of $4* from 1.01 to 1.20 in two caseso <r is free to vary (broken line) ( ) (ii ) Cr- is held fixed at the equilibrium value (solid line) C4-i 2-2 A Ar 2'ooo Pig 8 1:volution of' (IV) 6,*= 1. __ -Pl- j/++1 20 = 1.01 (VI) and with time for three cases. ^ t =0 R t > 0 (V)A = 1.20 - (.19)tanh(.02t) 1.2 exp(-.oo0114t) t*= L01 > 1.01 0, 03o 4oo Fig 9 Evolution of due to thermal forcing .+ (g t i Lastlyo w e xmine the behwAvour of the system if 9A is varied sinusoidally between 1.01 and 1.20 over a wide range of frequency. in the light of the above findings we should expect a simple forced oscillation of the system for high frequency variation of because the system would be always in the stage of initial dvelopmanto i.e. before the major development,, namely baroolinic instability of the waves 0 can set in A would be decreased and thus a reverse "initial" development would begin. period of the 9 shown in Fig. 9. Indeed, this is what we observe when the A oscillation it 112 non-dimensional units. This is Presumably if a nwh lower frequency is used0 part of the expected major development may set in and thus a distorted sinusoidal response should be found. Perhaps it is worth pointing out two other features in Fig. 9. (I) that of The amplitude of 4 A fluctuation is larger than that of Is larger than that of (O* -I. ) , although they fluctuate about their steady state value as empeoted. (11) is in phase with whereas all other quantities have definite but different phase shift with respect to , I cannot offer convincing reasons to account for these features. I. ___________________ CHAPTER IV Eff- o:1Xf at . t t _ A- One fruitful approach to examining the dynamical effects of the variation of the static stability is to study the energy transformaGates (1961) took this approach. tion processes. Hie analysis reveals two additional terms, which. would aot exist if the static stability Is held constant, relating the eddy available potential energy and the But without direct verification he zonal available potential energy. had to contend with just a qualitative discussion of the effects of the variation of static stability. In this study0 with the benefit of using spectral representation, we can examine its effects on each wave through numerical integration. As was shown In Chapter III the transition is actually suppressed when * we use. of what value of =e=mination, of Cr is held constant regardless Further insight can be gained only from the analytic expressions of the energy transformation. Those expressions can be obtained by simply taking the time derivative of equations (B-da) : (B-4f). (Note: Begause of the difference in the definitions of the temperature field, our expressions are different from Gates*.) m.. (7) {4.K -2 284 wh{re 14 = -- - CA, 3' ) 3K 3 g %3 t 6 43 C 1[4 and .jare frictional dissipation. 4+c Since -- ej is positive for warm air rising and cold air sinking, the other terms on the right sides of equations (6),, (7) and (8) are then energy transformation terus from XPr to kinetic energy [.. (9) where ....- . 2 Oet&A Similarly (10) vdiere expressions for 3'A 3 f 3 3 fA 3 ]3 are similar to the corresponding ones in (9) except interchanging subscripts 2 and 3. 203 Equations (6) to (11) can be sensarised schematically In the folloudag energy diagram. 2U E S/N/K / - To F r f/NAL ,2?,DISS1 P4 7/ OIV We should notice that f ' Vs allow %o to vary with time. and arise from the fact that Wenever they differ from zero it would mean that these are two additional channels for energy flow which would 30. have been wuppreosd hac baen held constant, Their signS have explicit phyeical interpretation, ~L%~Lb ~- (ocq~ e~j (&~+ L~tA~ A- A3 Each side of the inequality stands for the rate of conversion of per unit amount of zonal combined. - where ; stands for either 2 or 3 and It means that static stability would vary in such a way as to channel further amount A to Each side can be called the "specific" conversion rate from A.P.E. to K.E. loss rate of . and A3per unit ( of 4 2 to 4A 3 ) A when the combined exceeds that associated with wave 2. Similarly jA3 A3 A . The not effect of the variation of the static stability in then to replenish the A.P.E. of a component - iben its "specific" conversion rate is larger than the specific conversion rate of the other components combined. This mechanism should therefore be very effective soon after one component begins to grow and acts as a boost for further growth. Hence it is highly suggestive to infer that this mechanism is a major process to precipitate the exponential growth of wave 2 soon after its growth rate becomes perceptible. In order to see the detailed effects of the variation of , we examine 1. 3A3 04. / 3/ / /A 9 ' K a- 4o e Fig 10 - too &,o -- V) Ks / /000"o ;j ee$ 0( Variation of the kinetic energy and the available potential energy of various components due to a stepuise increase of 9* from 101 'to 1.20 in two cases, ( (i ) (broken lines) , is free to vary 1 is held fixed at the equilibrium value (solid lines) 32. Fig. 10 which is a mmary of the evolution of each energy component is stepwisely increased from 1.01 to 1.2 when for two different cases: (1) C, Is held constant and (2) <q is free to adjust itself. Besides the obvious fact that no transition takes place in case (1) we also notice two distinct features. First, grows noticeably slower in Secondly, when case (2) than case (1). , = 0 K is esentially suppressed, is free to vary. whereas it grows significantly when TO Now we are in a position to recapitulate and complete our reasoning behind the dynamical account for the transition phenomenon. Z is When first increased, there is a period of initial growth of all components. This arises from the fact that there is a not Inflow of energy into the fluid system through the bottom interface. Of the four processes that convert potential available energy between the wave and the zonal components0 two arise from the variability of the static stability. is positive. tA the initial condition and the initial growth, Although part of the increase of A is simultaneously transferred to the transfer can be so large that there is a net increase of fore substantial growth of K is possible. A A 2 and there- Now as soon as K is large enough, instability for wave 2 is triggered on. time For We notice that by this CQ increases at a such lower rate and soon begins to decrease. As long as A- remains sufficiently strong wave 2 continues to be unstable. While K and A are continually drained by the growing wave 2. they are simultaneously replenished by decaying wave 3. stained .growth of wave 2 is possible. Thus a sub- The montonic decrease of CJ, has 334 two effects. At firato the baroolinic growth of wave 2 io zurtr boosted. This Is largely due to the existence of on as k Is reduced to a sufficiently low value, - . ter a. decreases at a very low rate, wave 3 Is stabilized and the growth of wave 2 olows down. Eventually a new equilibrium state Is reached. However, there is a serious weakness in our approach. some external processes Even if by C( is maintained approximately the same, our previous results does not necessarily predict correctly that transition would not occur. The reason lies In the fact that we use one rather restrictive assumption in our model. That is that each Rossby wave is assumed only to Interact with the sonal component and is not able to interact with one another directly. When one deals with finite amplitude waves like those encountered in this study, such an assumption is a very drastic one. Therefore, even for non-linear interaction between the waves could conceivably bring about transition through direct transfer of energy without relying on mechanIsms represented fA,- and -A, . 4. CHAPTER V It is found in this study that with the technique of numerical integration a truncated spectral representation of each dynamical and thermodynamical quantity offers an acceptable approach for investigating the transient behaviour of a fluid system like the dish pan experiment. We have only considered the simplest type of transition between Rossby waves, namely between 2 waves of consecutive wave nupbers. In addition to verifying Lorenz's theoretical results concerning the relationships among the marious quantities in steady state, we also find that the total energy of the system remains constant within the range of ever, as 0A where two Rossby waves coexist in finite amplitude. How- is increased, wave number 2 in the new state has more energy than initially by an amount exactly equal to that lost by wave number 3. The reverse is true when A is decreased. This finding was shown to be implicit in Lorenas theoretical results. In examining the transient behaviour of the system during the transition, we find that the system evolves in a rather surprising manner for a short period initially before the actual transition takes place. It is found that both the zonal component and the wave component that subsequently decays actually intensity for a while. that the other growing wave dominates the developments flow gradually returns to the original value. Soon after while the zonal Although the rate of transition is found to be proportional to the rate of increase of the difference is reflected more by the delay of the growth of wave 35. number 2 than by the dependence of the growth rates on the rate of increase of 0 4. The variation of the static stability is found to play a central role in the transition phenomenon. By holding the static stability constant, we find that transition cannot occur. Instead the system quickly adjusts itself to a new equilibrium state very different from the laboratory and Lorenz"e theoretical results. It is proposed that transition from one wave number to another in primarily a consequence of baroolinic instability of the waves with respect to the sufficiently intensified zonal flow when A is changed. Energy is transferred from one wave to another via the sonal component. Thus the whole phenomenon is hinged upon vhether or not the zonal flow can be sufficiently intensified before it returns to its steady value. This crucial factor is found to be determined by the variability of the static stability. the rate with which bility. The rate of increase of K only determines is increased to the critical value for insta- Once instability begins the growing wave derives its energy from that stored in the other wave and therefore should be relatively unaffected by the actual manner whereby is increased. As more and more energy is transferred from one wave to another by way of the zonal flow, the original unstable configuration is eased up and the variability of 03 in turn acts to oppose further development. Thus the system eventually approaches a new steady state. The objectives of this study as outlined in the introduction are essentially fulfilled. However, one aspect of the transient behaviour 36. of the system is still development. not fully understood, and that Ia the Initial Although the initial intensification of the mona1 flows is found to depend critically on the variability of the static stability we still do not know about the processes responilble for the initial growth of the waves. Those unexplored processes however do have interesting properties. In particular, we find some definite dependence of the amplitude and phase of each quantity upon the sinusoidal variation of . It seems quite difficult to identify the crucial factors behind those processes and I cannot offer suggestions to do so. Fortunately the transition phenomenon does not critically depend upon those processes. 37. APPNDIX A Tmrncated Functional Representation of a TwoLayer Quai-Gostrophic Model A frictionless, hydrostatic and adiabatic fluid system in (XzyapOt) coordinate is governed by the following met of equations. Vortieity equation: + lk-cu XI (A-2) v~ 0 -0V x -C) Divergence equation: = v~ .vv~ First Law of Themodynaics: Equation of state: C< k(= T ) + Q.) (per fect gas) (nc ompressible liquid) The hydrostatic equation and the continuity equation are embodied in these equations. The next step is to introduce mmxiinm simplification which are based on the property of quasi-geostrophic flows, but they must not violate two fundamental invariants of the system, namely (1) total vorticity is conserved; (2) total energy is conserved. those problems of geophysical interests, are both small. Wx and l' For W If both of then are neglected, the two mentioned con- straints will still not be violated. In addition, quasi-geostrophic 38, and (ii) (1) Elows have two distinctive properties: V' * a It follows that advoction Is essentially duo to the non-divergent part of the horizontal wilnd in (2). Thus if w write V =. / ' V, . ( ) can be safely ignored 4- V,7( (A-(2)0 (A-(3) for constant forms of (A-() 0 (A-4) and that ( then the simplified f becomes )_- (5a) can further be simplified by using the equation of state and hydrostatic equation. (perfect gas) 2 0(liquid) 1 = gas constant where coefficient of thermal expansion 2!) depth of fluid Now it we substitute the quantities of the nodel described an page 3 into (A-4), (A-5b), (A-6) ve would obtain the consistent and maximum simplified set of equations as follows: 39. (A-7) (4)4 (c) .D gCt) D 4'e) + + (/cr) ... D V )- - + = =0 VX + Jc(-(,)-6- jV X) -ZO 0 Frictional and heating effects are accomdated into (A-7) as follows. They are parameterised with the use of the following physi- cal approximations. We introduce a coefficient of friction, at the underlying surface and at the separating surface. 2 The drag then is assumed to be proportional to the relative velocity at each surface. Similarly we introduce a coefficient of heatng, 2 at the underlying surface, and at the separating surface. The heat flux is assumed to be thettemperature difference across each surface. With such assumptions, the following terms U .?0 + /T ( -~~ - ~ 4-2 L2~ If~ 19 i 40# ahould be added to (7a)O (7b), (70,) (7d) and (7a) respectiveiy. In this study we want to study the simplest transition phenomenonthe one between two Rossby waves, wave number 2 and 3. Thus a trun- cated functional representation for each quantity is appropriate. For mathematical convenience we only allow interaction between each wave and the zonal component. But we permit the static stability -9. vary with time, although not with space. Because of the cylindrisal geometry of the systemu, a logical choice of the orthogonal funct c-n is Fourier-Bessel function. The actual expansion of each quant'ty is given on page 4 and is therefore not repeated here. them in equation (A-7) It we subtitute and make use of the following properti i of the Fourier-Bessel functions of are conwitants we can equate the coefficient of each orthogonal function. Tat end result then is the following set of ordinary differential eqations: (A-8) Complete set of the governing equations: (IV 13 3 3A L A3 41. (v) '4' -- (V) (0 (var) , fd -0( 4 A e) - A = -d iK + - 4" -( ft+4 -Q %ki-de24- +A,0+46 . 0- c;. it. --- pu 19 - Ct+2E3 -+ (va) -) #A - %+Ry- 0ff3( %* O9'PN3) -0- by (Ex) ) -0 649A4 C P 619 + bd(-k{t k+*)3 3 -. (1i) 60 l'iii) 9, (xiii) -C7 - -,(%, PI A (xv) % (yv;) b (xv)i i 'G) K- 0,( =- k3 --'fee's Ak t T ~-G~#4) 01 ' 9 - -AW S-4, kJ440 LL. Ke T. 6UL3 ) -+ OALA kCu-'L *A -ion+ + T 0+ ,) 9K ...3 Since thetas *s in (A-8v) - 0 K(k 4~.L3 Q9 'k ' r;- =~~ -* + ' 3 k3 e~ " c A 3L '-A- - (A-Sxvii) Can easily be eliminatedo we will eventually obtain 12 ordinary differential equations (21) - (2xii) for 12 variables depending on Z alone. They are given on page 8. L 42. APPENDIX B Derivation of the Various Energy Components of the System Total kinetic energy (TR) of the horizontal flow. (B-1) TCE - >l OJ40 // 4?y v3v + t7 , V'k-0- 7t -7 -) ar 4veraoe Are4/ at'Pra11e ~1 vo - vib .01 F 4 Vekit -711 4F 4- Lh 211 A 0/ 12. 4;tvlFit/ 11+ C4[o F~i- -(~F~ I LA.?~S k,~ 4 F 24 V1A: c-- I4ZL L7?V2 z 4- k q-1 4'~.. / c (k-A (f 1t2~ C 2 4.. ~1 9 r L3 430 Total poiential and itrne.1 energy (E) WIVERC 4 f 0 -I 0 )-2) 7r 4 z - ( c- Available potential energy [6w3' , + A4 -kdo) .APE) A.,Jlad 0 '7- E .4- 2 , 8, + 01:n"F '41 c' +. 2. L Writing g-. =-L, + 9 t0 I. 7- 9 + + T-, -+- & T 3 lj; We can associate a part of each type with one of the three components of the system, namely the zonal part, wave : imber 2, and wave number 3. "4 e) (B-4) 14CT Kj f- 14A1z 2) Bibi omaAh Davis, T. V.0 1956: The forced flow due to heating of a rotating liquid. Royal Trans. Roy. Soc. London (A) 249g p. 27-64. Fults, D., 1953: A survey of certain thermally and mechanically driven fluid systems of meteorological interests. Fluid models in geophysics Proc. lot Symp. on the use of models in geophysical fluid dynamics, Baltimore, p. 27-63. Gates, W. L. 1961: The stability properties and energy transformations of the two-layer model on variable static stability. Dynamical Weather Prediction Project sei. Rqport No. 6 UCLA. Hide, R., 1983: Some experiments on thermal convection in a rotating liquid. Quart. g. R. Meteor. 8oo. 79, p. 161. Kuo, H. L., 19531 On convective instability of a rotating liquid. Fluid Models in Geonhsi.. Proc. 1st symposium on the use of models in geophysical fluid dynamics, Baltimore, p. 101-116. Lorens, g. N., 1962: Simplified dynamic equations applied to the rotating basin experiments. J. Atm. Sci., Vol* 19, No. l p. 39-51. Lorens, B. N.0 1963: Deterministic Nonperiodie Flow. Vol. 20, No. 20 p. 130-141. i. Atm. Sci.