ea SciENCE Global Observations of Oceanic Rossby Waves Dudley 8. Chelton* and Michael G. Schlax Copyright © 1996 by the American Association for the Advancement of ScienceGlobal Observations of Oceanic Rossby Waves Dudley B. Chelton* and Michael G. Schlax Rossby waves play a critical role in the transient adjustment of ocean circulation to changes in large-scale atmospheric forcing. The TOPEX/POSEIDON satelite altimeter has detected Rossby waves throughout much of the world ocean from sea level signals with =10-centimeter amplitude and =500-kiometer wavelength. Outside of the tropics, Rossby waves are abruptly amplified by major topographic features. Analysis of 3 years, Of data reveals discrepancies between observed and theoretical Rossby wave phase speeds that indicate that the standard theory for free, linear Rossby waves is an incom plete description of the observed waves, Ohne of the major breakthroughs in the development ofa theoretical understanding of the large-scale circulations of the ocea that owe theit existence to the These plane- tary Rossby waves (2) are easily observed in the atmosphere asthe large meanders of the mid-latitude jt stream that are responsible for the prevailing seasonal weather patterns and their day-to-day. variations. Rossby waves have been much more difficule t0 1 naan bos of he came Ser em or mar isnt ems unis wo tho beer We pest hve a summary global observations of oceanic Rosshy waves from the joint United States-French TOPEX/POSEIDON | satellite altimeter mission and a comparison of the observa tions with predictions based on the stan- dard theory for freely propagating, line Rossby waves. Rosby waves are central to all modern theories of large-scale ocean circulation They are responsible for establishing the most fundamental featute of the large-scale circulation: the westward intensification of ‘circulation gyres (3). In the North Atlantic, ‘Tie auhors ae ao Oatege 0 Oceans wna imo here Sesenoes, Oregon Soe Uns, Cone. OF ‘rst-5500, USS SCIENCE + VOL.272 this is manifest as the intense Gulf Stream western boundary current. There are coun- ip the other major ocean basins. fn Targe-scafe atmospheric forcing Tn-concert with coastal-trapped waves along the easter boundary of an ‘waves are a mechanism that Rossby waves generated by EL Nisio events (6) may account for ‘ocean circulation anomalies a decade later in the mid-latitude North Pacific. Such ‘ocean changes might significantly influence ‘weather patterns over North The are the the ocean to (heating and cooling) att woundaries.and_over_the They can al by perturbations along the easter boundaries associated with coasta-trapped waves originating at lower latitudes. Although it is posible for Rossby wave characteristics (amplitude and propagation speed) to be altered by wind or buoyancy forcing that is coherent with the wave at precise wave numbers and frequen= cies (7), there is no evidence at present to indicate thae such resonance exists over the broaal ranges of wave numbers and frequen: cies and the global geographical domain ‘over which sea level signals with Rossby wave-like characteristics are observed in +12 APRIL 1996the “TOPEXIPOSEIDON data analysed here, The premise of this analysis is that Roby waves generated by forcing at the easter boundary oF over the oe tie ssn pone aay fam the “The standard theory for fcely prop ing: incor Rowdy waves can be dette aitation of the unforced eat "sale, low-frequency motion shoe a state of rest. This yields a wave Caqation forthe verical normal moses (8). The restoring force is the so-called “B ef- fect” of latina variation of the foal vercal component of the Earth's angular oration vector. The curvature of the anh’ surface is ths esertial co the exis tence of Rossby waves With appropriate surface ned bottom boundary conditions, one can determine al modes unericaly by solving mivaluc problem thar depends only Gin the local statfication (8-10). There tit an infinite number of Raby wave fromnal odes, ordered by decreing phase speed. The phase speed are westward forall Of the motes, and low-frequency, lone rath solaions ae zonally nondspe five (tha & the westwardcomponent of phase speed is independent of wavelength). “The lowest onder (barotropic) mae is uniform vertically sndependent of stat cation, and. propagates actos an entire scean basin int peri of about a week This iso fat torbe resolved in mean sents from an orbiting siete, The next Igher onder (Hint birclinie) mode ss fice intensified, depens strongly on the stratification, has a velociry. profile that changes sgn at the depth of the thermo sline (11), and propagates much more sw Ty rearing months to decades to cross an vein havin (Fig. 1). The basin etesing time increases with increasing latitude ane of the larituinal variation of phase speed owing to the B effect. Asa result, the crests and troughs of Rosby waves inated Simultaneously along the entre eastern boundary of an ocean basin arrive at the westem boundary much sooner at Tow lt tes than at igh latte, a effect often refer to a B refraction, ‘An important feature of the first ba roclnie mode is that variationa ofthe sea terface height ave enivored as thermocline depch versions of the opposite sign with out thee oer of magic great pli (8) A sea level variation of 5 cm thus coresponds toa thermocline dspace tment of abet 50 m. Such lage Sf upper ocean thermal structure have im portant implications about the role of the ‘ocean in shortcerm climate variations. in most reyions of the ocean it fs gen erly believed sha the lower, higher omer baroclinic Rossy wave modes play a ser role in large-scale ocean circulation dynam jes. The focus here is therefore on first ‘mode baroclinic Rasby waves. Observations of westward propagation Evidence of firstmode haroclinic Ressby waves has been presented from ship-hased ‘measurements of the upper ocean thermal 7, 12). Because most ofthe ocean fs severely undersimpled, unambiguous in- tempretation of Rosby wave signals in the thermal measurements is difficult at all hur a small number of locations. The advent of satelite altimetry inthe lace 1970s established a foundation for glob al observations of baroclinic Rosshy waves (13). The measurement accuracy and mis. sion duration of the early GEOS-3 and SEA. SAT altimeters (74) were inadequate for detecting Rossby waves. The GEOSAT al timeter (15) measured sea level from 1985 through 1989 with sufficient accursey (16. 17), but the orbit configuration wasill suited to Rosshy wave studies because tidal errors aliased into wave numbers and frequencies that are difficule to distinguish from annual- period baroclinic Rossby waves (18, 19) ‘The TOPEX/POSEIDON launched in August 1992 exactly repeating orbit eycle, is the flown in an orbit specifically designed co avoid tidal aliasing (19, 20). An esentially continuous record of sea level measure- iments with an accuracy of ~4 cm his been ‘obained since the instruments became fully ‘operational in October 1992 (21). Large- scale sea level signals with amplitudes as small as ~T om can he detected after ap- propriate filtering of the data to reduce the ‘quasi-random component of measurement terrors. The >3-year data record thus pro- vides the ability to detect Rosby waves unambiguously over the entire world ocean. For the analysis presented here, the TOPEX/POSEIDON data were filtered (22) specifically to highlight the long Rossby waves predicted by the theoretical altimete nto a 10s Laude (degrees noth) Fg. 1. Latusinal variation ofthe tine requred for baodlnic Rossby waves fo cross an oceen basin ‘wth the gpometry ofthe Noth Paci. Tose ra 'Sitimes are based onthe pase speeds predicted bythe standara theory for recy propagating, non Capers. incor fst mods arocinc Rossby waves (0, SCIENCE + VOL.272 + 12 APRIL 196 ‘considerations outlined above. The expect ‘ed westward propagation of alternating pos tive and negative sea level signals isa ver ‘common feature in the resulting filtered se Tevel fields (Figs. 2 and 3). The petios of these signals generally full between 0.5 and Deyeles per year, and the wavelengths typ- ‘cally range ftom 10,000 km or longer in the tropics to about 500 km at 50° latitude ‘These propagating signals account for more than 30% of the toral sea level variance at latitudes lower than 15° but only about 10% ofthe variance at lavitudes higher than 30°. ‘The mechanisms that generate this widespread propagation have not yet been determined, In many cases, the waves orig inate at the eastern houndaties of ocean ther from local wind ot buoyancy foci with coastal trapped waves originating at lower lat In some cases, the waves appear to originate in the ocean interior. The propagating signals in the extr tropical examples (Fig. 2) are larger in am- plitude west of major topoxr (the Emperor Seamounts at 170°E, the southeast flank of Hess Rise at 1 Ne and the Hawaiian Ridge at 1 °N}. This relation berween otto pography and wave amplitude has previous- Ty been reported in the North Atlantic (17 19). Iris observed in this analysis in. many asin, 00 geo ol Yar SOE 100 100 00 Yo 150° 180" 150W_ 120 1000 00 400 Yeo BOE 10 SOW 120 — oo a a Sea love (em) Fig. 2. Time longiude socions offered sealove (22) in the Pactio Ocean along 39", 22", and 2I'N. These examples are representatve of fextratropical latitudes throughout the word 235Fig. 3. _Time-longiude sections of fitred sea evel (22) In the Pactic Ocean ‘song Nand the equator. AA section along 4°5 i a ‘mest identical to tho 4°N section, The te ais Ssteiched compared with Fag. 2 to ad nthe ent ccalon of the rapid east Ware and west pop gating tropical sea lev! ssonas. The symbols co respond to the mes ana locations of the matching symbols nig. 4. cof the time-longitule sections of TOPEX/ POSEIDON data throughout the world ‘ocean. In the 21°N and 39°N examples, the westward-propayating signals appear to originate atthe houndary and abruptly amplify over the mic-ocean topos raphy. In the 32°N example, there is: no detectable westward propagation east of Hess Rise, suggesting that the topography may be a source of the waves. Previous theoretical stlies have suggested at least three possible says that topography could affect aroclinie Rosby waves (23), Additional analysis is required to identify the mechanisin responsible for the strong top sriphic interaction observed by TOPEX/PO- SEIDON. Also in nced of further study is the suprisingly weak westward penetration of the ‘Sea level em) Fig. 4 Gobsl maps o ited sea evel (22) on) 13 Ape 1983 (eye 21) and 3.5 months later on (B) 81 Jdy 199G (eycle 92}. White ines identity a west \warG-propagating, B-efracted Fossby wave tou. the equatonal Klin wae rough Othe Rossby wave crests (open tangle 236 ‘Sea level (em) observed energetic sea level variability: near the easter boundary (Fig. 2), presumably a sociated with seasonal wind forcing Another common feature in the data is n increase of the phase speed of the ob- served waves inthe western basins. This phenomenon is evident in the 21°N exam- ple, where the phase speed in the far west- fem basin exceeds that in_ che far eastern basin by more dl :. This longitudinal variation of phase speed i a consequence of the deepening of the thermocline in the west (10) ‘Waves within about 2° of the equator are distinetly different from extratropical waves. Most of the observed sea level var ability along the equator consists of east ward-propayating signals with phase speeds thetimes ana The time evolutons of frames rom an SCIENCE + VOL 272 + 12 APRIL 1996 1998, 1999 150° 180) of about 270 em s~ (Fig. 3). These signals can be interpreted as equatorially trapped Kelvin waves, often originating in the cen: tral Pacific where they are generated by shottsperiod equatorial wind events at about 170°E (24-26). Ie has recently been suggested (24) that these intraseasonal Kelvin waves may play an important role in the onset of interannual El Nifo events iperimposed on the rapidly propagat= shortperied Kelvin waves are two long-period, relatively slow, eastwand-prop- agating, postive sea level signals. The first friginated at the western boundary in Oc- tober 1992 and reached the eastern bout any in June 1993, and the second originated in June 1994 and reached the eastern Foundry in January 1995. These sig 10 20's ‘Sea level (em) andopen cite}, andthe Rossby wave trough sai irl) can be traced trom ions ofthe matching symbols n Fig 3, These womans are imation of TOPEX/POSEIDON data hats avalablo onthe ‘Worle wide Web at ftp: /topex- nw jl asa.gos/contichellor/rossby/coincident with the second and third pales of the recent, long-lasting El Nito event (27). Because their propagation speed is ‘only about one-thied that of Kelvin waves, these slow eastward-propagating signals sre likely coupled atmosphere-ocean phenome za (6) rather than freely propagating waves. In addition to the fist and slow east ward-propagating equatorial signals, there smaller amplituie, long-period west- ‘ward-propagating sea level signals that can be interpreted as opical Rossby waves with, westward phase speeds of abour 100 cm s~! (12, 24-26). The latitudinal structure of sea level associated with che dominant op eal Rossby waves has a local minimum at the equator with symmetric maxima at about 4°N and 4°S (25, 26). The observed westward propagation “is therefore much ‘more apparent along 4°N than along the equator (Fig. 3). A noteworthy feature of these tropical Rossby waves is a monotoni= cally increasing westward phase speed be: the South American coast and about 140°W and relatively constant phase speed farther west. As with the geographical vari- ation of the phase speeds of waves noted along 21°N, this shift is a consequence of the deepening of the thermocline in. the ‘western tropical Pacific (10), Spatial structure of the observed Rossby waves. Two frames from an anima- tion of the filtered TOPEX/POSEIDON lata are shown in Fig. 4. In the 13. April 1993 frame, a negative sea level sig propagated along the equator from South An America to the far western, Paci ‘elongated negative sea level pattern 1 behind the equatorial signal at higher tuxles can he traced to abou 40° latitude in both hemispheres, This feature is character- istic of a refracted Rossby wave trom The remnants of an earlier refracted Rossby wave crest are evident westward and pole- ward of the trovgh, and a new erest is seen to be forming in the far eastern tropical Pacific, Similar refracted Rossby wave crests ang troughs can be seen in hoth the Atlan- tie and Indian oceans, In the 31 July 1993 frame, the Rosby ‘wave trough has impinged on the western oundary of the Pacific, and an equatorial Kelvin wave tough centered at about 140°W has propagated rapidly eastward mote than halfway across the Pacific, split- ting the newly formed Rossby wave crest that has propagated westward from South America, An important and controversial {question presently under investigation (26) is whether equatorally rapped Kelvin waves such as this can he generated by the reflee- tion of Rosby waves from the western boundary (a, purely oceanic process) or whether the Kelvin waves are forced by trade wind variations in response to the changes of| sea surface temperature associated with the Rosy aves (an air-sea interaction pro- ees). The dense coverage and continuing data record provided by TOPEX/POSEI- DON are likely to provide important insight into this dynamical process Outside of the topics, the positive and negative sea level signals lose thei latitudi- nally coherent refractive structure soon af- ter generation. The resulting discontinuous sea level features fluctuate in amplitude but continue propagating westward across the basin. The amplitude fluctuations are evic dentin the time-longitude sections in Fig.2 as ripples along the propagating positive and negative sea level signals. The breakup, of the coherent structures may be indicative of the effects of distortions of freely propa gating waves by local wind forcing or bot tom topography ‘Comparison. between observations and theory. Time-longitude sections like those in Figs 2 and 3 were constructed throughout the world ocean. The phase speeds of westward- propagating sigaals estimated from these see- ‘ions (28) are shown in Fig. 5. Although the srtong. lrinuinal variation ofthese observed a0 1 a 20 Latitude Fig. 5. (A) Gobaly distibuted estimates of the phase spoeds of wostward-propagating ses level Signals estimated from 3 years of TOPEXPOSE: DON altimeter observatons. The sold eres coe respond to Pactic estates, and the open cries Comespond to Alantic aid Indian Ocean est ‘mates. The gobal average litdinal vation of the pase speed predicted bythe standart theory for exvatroical Weely propagating, rondsper sve, near, frstrode baraclnic Rassby waves (0)'s suparimposod as tho continous Ine. (B) ato of the obso10 pase speeds to the phase ‘sp00ds predicted by the standard theory at the same geographical bcatons as the observations SCIENCE © VOL.Z72 + 2 APRIL 1996 ‘Phase speeds is consistene with the expected B refraction of Rashy svaves, the observed phase speeds outside ofthe topical hand! fom TO°N eo 10°S are systematically greater than those predicted by the standard theory for fively propagating linear, first-moe has rwoclinic Rossby waves (29). ‘The discrepancies herween observed and predicted phase speeds are too large to be artifacts, The possibility thae they may have ‘heen spuriously introduced by sampling er- ror of data processing has also been ruled ‘out on the basis of simulations of each of the processing steps applied to the TOPEX/ POSEIDON data, We thus conclude that the standard theory for freely propagating, linear, baroclinic Ressby waves is deficient in predicting the observed phase speeds. ‘The inadequacy of the standard theory for prediction of Rosby wave phase speeds hhas heen noted previously from upper ocean thermal data (7, 12, 30). Because of the paucity of in. sity thermal measurements available for Rosshy wave detection, these studies have been restricted to the region between 8° and 22°N in the Pacific. In addition to coreoborating these widely urap- preciated result, this analysis of TOPEX/ POSEIDON dats underscores the global sig nificance of the deficiency of the standard theory for fee, linear Rossby waves, Ac least two mechanisms ave been suggested £0 ex- plain the high observed phase speeds (7, 31), but the iswe is not yet resolved “The implication of these result is that the ocean reacts more rapidly than is gen- ‘erally believed; the transient baroclinic ad- justment time of the ocean at 35° latitude, for example, is only about half as long as that predicted by the standard theory. The transoceanic transit times shown in Fig. 1 for the North Pacific must be modified t account for the observed propagation speeds of extratropical baroclinic Rossby waves. The TOPEX/POSEIDON observa- tions of high phase speeds, as well as the apparent effects of bottom topography on extratropical Rossby waves (Fig. 2), provide ‘important consistency checks for evaluating the performance of ocean general cicula- tion models (32) for studies of the tole of the ocean in interannual and decadal clt- ‘mate variability REFERENCES AND NOTES 1, © 6, Pasi eat, Mar F238 9099 6. Fostty, ft Mao Soe 6, 667040, 2. GW Bataman, JF, Mote Soo 94, 25 (as 8, DLL T,Adrson and A. Git, Deop-So Rs. 22, Ss (1978H0 LT Aegon Bjan AEF G Pacanonh J. Geophys Be, 8447081978, 4, PeieCreay) Pye Oosanogr 665211379: i humour 6. kd dd OB, p62 GA downbeat 370, 320908) ‘paret Latfr : Rooc of Clr 1 taot 1B Nes La Fr vas ae ‘Aad Mech 96, 617 (1094; D-S. Barak wd ES 272 10. 2 13 ¥, 1. Saraonk, Av. Geooty= $8 (sup), 1967 (1996. WEB. White rye. Goaarogy 7.80 (1977 Dd. 1, 49 (1086) a releencoe ren, lit Lebena ana LA Mak Waves Be Ooaan IBsover Now York 107s) AE, Gl Amore (Cesar Eyam Acar ross, Cambie, UK, ‘a Wal Emer, WG. Loo L. Mogae J Prys. Ocean- fp1 14 28 1084) § fay, €- Bombrowsey, (eioy oF Minster, 17,7619 1967, 1.8: Choton MG. Sela A A caSscoke, KEL Nagpar NSor prepa, {he termosine's a rogen of 2009 verkcal ‘Guts of operat and erst at operates ‘dene, and ney Ramogengous dees Waar fram the wa and weakly sateod une Wate cores WS Keser Goons. Aes 8,518 (1900, CC Rivand RE” Chany. Alt Gooey. 88 {au 29 1058, ‘Apatoance cmparaon betcen the GEDS.3 SeASAT. and GEOSATatmatreand rac ox pcatens of TOPEUPOSECON e sumrareed Oy Ci 0..chaten and. ori foeoaroe ay. (rea 0. Daugis ana RE Crane, J Goons. Fs 96, 26581000, Exams of apparent Rosey waves nred rm ‘guna anaes oTGEOSAT ta nekce PE Mat {ewe M.A otrson 0 Or. 97,1789 {ond} ©. Pargaad and P. Daouse, Sot6d: MC Van Wort a ot MPs, ba 8, {aerg t0a3:P. Le aon ane-F Mist 12918; 5 Ao S Imowat K Ichkawe, 00,29 (1985, Fort Tokmakan 2 PG. raloee ot. Gaantye Fes 99, <7 (989. GrA.vocobe, SH: om, ME Parks, PC. Alen, ‘Be 87179193982} MG Sohgs and D8. Cha 2. 2. 28, ton, bi, 9, 12000 (1994) MG. Sonu are D.B. Craton, bap. 2476% NE Pare, AH Stowat OL Fares, 0.€ Car wi ‘Sreoedase lve ils angi @tortan ery ‘Sale sponertan 4 Joon fore © 5" andshoter onde idea umare 5" BUR Fines J Foe Moh 37, t61 (1080: 8. art Pj. Oosaroge 18,247 (188 FG ‘ee gr G Weber, 21, 1900(1381 AWC Reser, Ned MePhaden, Me Wecimam, 1 czoonye Fee 10, 0643 15 W . Kosser Sno ied Pads J Co 817871985. ‘Oakrig Peans .En, Coops. Ae 96, 4240 00) T Doron -P Boulanger, F Mos. Meno. 08, 25035 (104) S1-B Bourqor and C. Menkes, ec. 100, 25081 (todo -P Bouangar ana. Fu,n precaraton (Ce teenth ans Hoar, Geoony A Lat 23,57 (1950, ‘The timelongtuce pets wore sxamines vy {0 det tagone of unambiguous estar ‘repagaton The phase Spex vas thn estnatd hyena for sath aecton for wtien the vb fropagaton spanned at ast 0" longa The Sojete etna nae obtsed tom a engi SCIENCE + VOL,272 + 12 APRIL 1996 eon of na Race wanton common fered qoasa slot tage" he sourelogeal Kate [Ak Jn Funan of gal rage Process InatPremen ah Enlowcoa Cif, Ma 19880. ‘Glnbout, mageg the Ears ore (cel, ‘Sxl, Ui, 1085, Sea vel was armed long Is of constant phase seed none gts flr octane of grase pra, The be Song 100'samplelines for sah pase seed were ‘hen squared ana sumed Te phase speed cor ‘Bepanting oth madam sum ot suse con Sues te enpicct extra of pase sped Ech cbwcive cata was vay checks for onstercy wen the trelongtae sacton om vacant “he tana hoc) ares ete breaks dow she about 10" ot ho ety, whee the erica ‘Component cf the Eas agua otton vet ene venshngy sma An tos or Tae ten debe fr fe uate band [SP lander nie, Lat, andthe Southam Osanon (cadre Pose, San Diogo, 100), ae ob ‘Sac phase spaods betwen 10 and TOW are Corea wh ee an 25. 20, (G eojers Pre. Ooonog. 8,068 (579 BL Sram anh Ker Daop Sears 426, i033 1979, 2 akon as. Sth, J, Gots. Res ‘ba joo 4)-A Serna, Sonne 29.1879 11805. 7 “ne retarch was super by coat a8 ‘rom Jo Propason ateratory, by art NACW. $10 tom Pe Ratonal Aeronautics an Space A Irneraton say gat OOe 3202896 fom NSF ‘Moar axSatake M Frei 5. aud, onsior Mato, Re ier. Siowat, Tub, nd Wier comets 21 November 1008; accaptod 2 tru 1900,