Part One GeneralOceanCirculation

1.1 IntroductionIDeep Circulationof theWorld Ocean

Bruce A. Warren

Historically, the deep circulation of the ocean has beenviewed from the perspective of property fields, mainlythe distributions of temperature, salinity, density, anddissolved-oxgyen concentration. The practical reasonfor not considering velocity measurements as well, ofcourse, was a technical incapacity for making themuntil very recently. On the whole, this was probablynot a bad thing: not merely because the property dis-tributions are as interesting in themselves as the mo-tion field, but also because the scalar fields are so muchmore stable than the velocity vectors-allowing spotmeasurements from different areas even years apart tobe combined into coherent pictures that tell a gooddeal about general patterns of deep flow, albeit indi-rectly. The slight differences between correspondinghydrographic sections in the atlases by Fuglister (1960)and by Wust and Defant (1936), when compared withthe fluctuations much larger than the means of deepvelocities observed by the MODE Group (1978), forexample, demonstrate how much easier it is to obtainstatistically significant information pertinent to theoverall global deep circulation from water-propertydata than from current measurements.

On the other hand, the information gained from theproperty fields allows only a limited view of the deepmotions, at very best some kind of long-term average.Although oceanographers have usually been mindful ofvariability in the deep flow, even if only to accomplisheddy mixing, it seems extremely unlikely that anyoneimagined the highly energetic low-frequency meso-scale motions that current records have revealed. In-stead, because of the stability of the property fields, itwas stationarity rather than variability that was em-phasized, however implicitly, in the circulation pic-tures derived from them. That stability also was surelythe basis for the conceptual structure of water typesand masses that has been so enormously useful in sum-marizing and comprehending the temperature-salinitystructure of the ocean and in identifying features inthe property fields that can be exploited as tracers forthe flow. Without velocity information, though, suchdescriptions of oceans have sometimes degeneratedinto taxonomic sterility (naming something doesn'texplain it), and perhaps sometimes there has been tooelemental a character ascribed to water masses (as ifthey were truly building blocks rather than names forfeatures), leading to pictures of the ocean more sugges-tive of rigid geological strata than of the real motionfield that forms the distributions.

Plainly there cannot be a satisfying description ofthe deep ocean circulation that does not meld stationdata with current records. It does not seem to me,though, that such a description is yet possible. Far too

6Bruce A. Warren

few current records have been obtained to describe thedeep, low-frequency motions in a global sense; it isonly in the western North Atlantic that one can evencontemplate making a basin-wide description. More-over, we simply have not learned how to combine thestable station data with the fluctuating-velocity rec-ords to tell a story that is both consistent and inform-ative. For example, Reid, Nowlin, and Patzert (1977)reported a record (Cato 2) from a current meter mooredon the South American continental slope in the coreof the North Atlantic Deep Water: for 2 weeks thedaily-averaged velocity vectors were directed south-westward, parallel to the isobaths, as one would haveexpected in this particular deep western boundary cur-rent; but then the flow abruptly changed direction andwent eastward for nearly 2 weeks. Nevertheless, withdue regard for different density and accuracy of obser-vations, the high-salinity core of the current lookedvery much as had been depicted by Fuglister (1960) andWuiist and Defant (1936). How are we to approach thesetwo different sets of data, to reconcile the variabilityof the one to the steadiness of the other, and to learnsomething significant from their combination aboutthat boundary current?

Finally, it is not at all clear what effect the low-frequency velocity fluctuations have on the long-termmean flow. There is enough theoretical reason (e.g.,Rhines, 1977) to suspect that their role in its dynamicsmay be substantial, but measurements of deep Rey-nolds stresses are meager. In fact, values reported bySchmitz (1977) from the Sargasso Sea well south of theGulf Stream actually favor a negligible contribution tothe vorticity balance there, but those measurementsare far too few to give a general characterization of thedeep open ocean.

Consequently, although I recognize its incomplete-ness, the following account of the deep circulation ofthe world ocean is undertaken mainly from the tradi-tional perspective of hydrographic station data, withreference to current measurements only where theyseem helpful in estimating velocities and transports ofthe prominent currents. The emphasis is on mean ther-mohaline circulation. What has been learned about thelow-frequency motions is. described in detail byWunsch in chapter 11 of this volume.

In section 1.2, I have attempted a historical reviewof what seem to me to be the important events anddates in the development of ideas about the deep cir-culation, from the first deep temperature measure-ments through Sverdrup's comprehensive synthesis inchapter XV of The Oceans (Sverdrup, Johnson, andFleming, 1942). In section 1.3, I have discussed thedynamical ideas of Stommel and his colleagues thatled to the overthrow of a substantial part of Sverdrup'spicture, and its revision in contemporary thinking withdynamically consistent models of circulation. Section

1.4 is an account of the sinking processes that supplywater to the deep ocean from the surface layer. Section1.5 is a consideration of how well the kinds of deep-circulation patterns envisioned in dynamical theorystand up to observation; it is necessarily mainly a di-gest of the evidence for deep western boundary currentsin the world ocean. Finally, in section 1.6, I have spec-ulated about some fundamental aspects of the deepcirculation that seem to me to be not very well under-stood at this time.

The focus throughout is more on the circulation ofdeep water than on the complementary problem of itsproperties, because the subject of deep-water charac-teristics has been treated in detail by Worthington inchapter 2 of this volume. There is, of course, someoverlap with that chapter, as well as with Reid's gen-eral discussion of the mid-depth circulation (chapter3). To the extent that our opinions are in conflict, wehope that readers will recognize subjects for furtherobservation and thought.

1.2 Historical Development of Ideas about the DeepCirculation

In 1751 Henry Ellis, captain of the British slavetraderEarl of Halifax, wrote to the Reverend Stephen Halesto describe some deep temperature measurements thathe had made at lat. 25°13'N, long. 25°12'W, with a"bucket sea-gage" devised and provided for him byHales. This instrument, to be attached to a soundingline, was a "common houshold pail" covered at topand bottom by valves that would be forced open duringdescent and pushed shut by drawing the bucket backto the surface; it was furnished with a thermometer sothat the temperature of the water sample thus trappedcould be read when the bucket was returned to theship. In his letter, which Hales transmitted to theRoyal Society of London, Ellis (1751) reported:

Upon the passage, I made several trials with the bucketsea-gage, in latitude 25'-13" north; longitude 25'-12"west. I charged it and let it down to different depths,from 360 feet to 5346 feet; when I discovered, by asmall thermometer of Fahrenheit's, made by Mr. Bird,which went down in it, that the cold increased regu-larly, in proportion to the depths, till it descended to3900 feet: from whence the mercury in the thermom-eter came up at 53 degrees; and tho' I afterwards sunkit to the depth of 5346 feet, that is a mile and 66 feet,it came up no lower. The warmth of the water uponthe surface, and that of the air, was at that time by thethermometer 84 degrees. I doubt not but that the waterwas a degree or two colder, when it enter'd the bucket,at the greatest depth, but in coming up had acquiredsome warmth.

Ellis's guess of one or two degrees warming duringthe ascent was based on changes observed in the tem-perature of the sample while on deck. Modern data,however, indicate that Ellis's values were some 10-

7Deep Circulation of the World Ocean

13°F too high. Perhaps he underestimated the conduc-tion through the walls of the sampler when it was inthe ocean, or perhaps the valves were not tight enoughto prevent exchange of water while it was being raised(especially if raised unevenly).

Nevertheless, these are the earliest recorded subsur-face temperature measurements in the open ocean-indeed, the first anywhere to a substantial depth-andthey pointed to what must be the most fundamentaland striking physical feature of the ocean: that deepwater is all cold, and warm water is confined to arelatively thin layer near the surface in the tropics andsubtropics.

Ellis himself did not seem to realize the far-reachingsignificance of his data, for he remarked later in hisletter to Hales:

This experiment, which seem'd at first but mere foodfor curiosity, became in the interim very useful to us.By its means we supplied our cold bath, and cooled ourwines or water at pleasure; which is vastly agreeableto us in this burning climate.

He does not seem to have made any further measure-ments, nor does his work appear to have stimulatedimmediate exploration of the deep water by others, for,according to Prestwich's (1875) tabulation, it was morethan 60 years before any additional temperature meas-urements were made to the depth that Ellis reached(though others were made during this time at lesserdepths).

What Ellis's observations meant, of course, was thatdeep water in the tropics must derive from polar re-gions, and that, accordingly, there must be a meridionalcirculation system in the ocean to carry deep waterequatorward. Evidently (M. Deacon, 1971), the firstrecorded recognition of this profound implication wasby Count Rumford in his essay, "The Propagation ofHeat in Fluids," first published in 1797.1 Rumford hadmade the experimental discovery of convection cur-rents in liquids (the basic subject of this essay), and, inconsidering their possible role in nature, he reasoned(Rumford, 1800):

But if the water of the ocean, which, on being deprivedof a great part of its Heat by cold winds, descends tothe bottom of the sea, cannot be warmed where itdescends, as its specific gravity is greater than that ofwater at the same depth in warmer latitudes, it willimmediately begin to spread on the bottom of the sea,and to flow towards the equator, and this must nec-essarily produce a current at the surface in an oppositedirection.

In advancing evidence for such a deep flow, he drewthe correct inference from Ellis's measurements, in asclear and straightforward a fashion as one could want:

But a still more striking, and I might, I believe, say, anincontrovertible proof of the existence of currents ofcold water at the bottom of the sea, setting from the

poles towards the equator, is the very remarkable dif-ference that has been found to subsist between thetemperature of the sea at the surface and at great depth,at the tropic-though the temperature of the atmos-phere there is so constant that the greatest changesproduced in it by the seasons seldom amounts to morethan five or six degrees; yet the difference between theHeat of the water at the surface of the sea, and that atthe depth of 3600 [sic] feet, has been found to amountto no less than 31 degrees; the temperature above or atthe surface being 84°, and at the given depth below nomore than 53°.

It appears to me to be extremely difficult, if not quiteimpossible, to account for this degree of cold at thebottom of the sea in the torrid zone, on any othersupposition than that of cold currents from the poles;and the utility of these currents in tempering the ex-cessive heats of these climates is too evident to requireany illustration.2

During the early nineteenth century, Humboldt(1814, 1831, 1845) popularized the notion of deep cur-rents flowing from polar regions toward the equator.He mentioned no sources for the idea, but he did citeEllis's measurements and later ones of better qualityas supporting data.3

The physicist Lenz (1845) had interested himself inhow the subsurface vertical temperature gradient var-ied with latitude. In the course of his investigation hediscovered the shoaling of the thermocline at the equa-tor in the Atlantic; and the much sparser data availablefrom the Pacific suggested the same phenomenon tohim there. He regarded that shoaling as evidence forupwelling of deep water to the sea surface, and, ac-quainted with Humboldt's work, he proposed a morespecific scheme of closed meridional circulations thanhad been given hitherto. This scheme involved greatconvection cells symmetric about the equator, withwater sinking in high latitudes, flowing equatorwardat depth, and rising in the tropics to return polewardnear the surface. He considered that this upwelling alsocontributed to the reduction in surface salinity that hehad observed at the equator in the Atlantic. His un-derstanding of the driving mechanism was not moreprofound than Rumford's, but he did recognize thatwind stress and the earth's rotation would distort themotions substantially from the simple cellular form.He made no explicit reference to the Indian Ocean, buthe discussed the symmetric cells in terms of generalvalidity. Variants on Lenz's construction, some perhapsconceived independently, were to influence interpre-tations of the deep circulation into the first decade ofthe twentieth century (e.g., Schott, 1902; Brennecke,1909).

Dr. Carpenter, for one (Carpenter, Jeffreys, andThomson, 1869), invoked the general idea to accountfor differences in deep temperatures observed in thenorthern North Atlantic, and his resulting prolongedand contentious controversy with Croll attracted agreat deal of attention to it (M. Deacon, 1971).4 Car-

8Bruce A. Warren

penter had a more elastic conception of the convectioncells than did Lenz, though, believing that their warm,poleward-directed limbs could touch bottom in places;nor was he rigid about the equatorial symmetry, con-sidering that the northern cells would be weaker andless extended laterally than the southern ones, in con-formity with the oceans being much more open topolar water in the south than in the north-wherebydeep water from the Antarctic could be expected topenetrate into the northern hemisphere. He referredexplicitly to recent subsurface temperature measure-ments in the Arabian Sea to argue that all deep waterin the Indian Ocean derived from the Antarctic, and heimplied a belief that most of the deep water in thePacific would also be found to come from the south.

Nevertheless, Prestwich (1875) seems to have beenthe first to demonstrate from observations the nonex-istence in the Pacific of relatively strong deep upwell-ing near the equator and the lack of a northern sourcefor the deep water there. These facts showed that deepwater in the North Pacific, as well as that in the SouthPacific, was supplied from the Antarctic, and thus in-validated Lenz's picture of symmetric convection cells.Prestwich believed, however, that it was still a correctdescription of the meridional circulation in the Atlan-tic.

Following the Challenger expedition, Buchan (1895)produced world-ocean maps of the distribution of tem-perature at various levels, based mainly on the Chal-lenger data. His map for 2200 fathoms illustrated thespreading of deep water from the Antarctic into thethree oceans, including details of how its course isshaped by submarine ridges:

It is also to be noted that the lowest deep-sea temper-atures are found in those parts of the ocean which liein the Southern hemisphere, and that, on the whole,higher temperatures are encountered as we recede fromthe Antarctic region. It may also be pointed out thatthe lower deep-sea temperatures extend farther to thenorth from the Southern Ocean, just over those depthsof the sea which appear to have, and probably do have,a direct communication with the south; that is, are notcut off by any intervening submarine ridge separatingthem from the cold waters of Antarctica.

... There can be no doubt that these very low deep-sea temperatures have their origin in the Southern orAntarctic Ocean, the icy cold waters of which are prop-agated northward, the rate of propagation being soslight as to be regarded rather as a slow creep than asa distinctly recognizable movement of the water.

Buchan (1895) also inferred a southward movementof deep water from the North Atlantic into the SouthAtlantic. As Merz and Wiist (1922) pointed out later,this perception, together with Buchanan's (1884) spe-cific-gravity profile that showed the intermediate-water salinity minimum extending well north of theequator from high southern latitudes, would have beensufficient to disprove Lenz's equatorial upwelling from

great depth to the sea surface-and thereby his sym-metric convection cells for the Atlantic; but neitherBuchan nor Buchanan attempted a comprehensive dis-cussion of the meridional circulation.

Buchan formed his idea of southward flow after no-ticing in the Challenger data that the vertical temper-ature gradients at depths from 800 fathoms to 1500fathoms and deeper (how much deeper not stated) weremuch smaller in the South Atlantic than in the NorthAtlantic. He linked this difference to the distributionof salinity (specific gravity):

The specific gravities at the bottom of the ocean afforda ready explanation for this remarkable distribution oftemperature. Owing to the higher specific gravities ofthe North Atlantic, an extensive deep-sea current fromthe North to the South Atlantic, carrying a higher tem-perature with it, sets in at depths at which the influ-ence of the surface currents is no longer felt, and be-comes more pronounced as the depth below 1000fathoms is increased. Hence the North Atlantic re-ceives large accessions to its salinity from the surfacecurrents, which the deep-sea currents again return tothe South Atlantic.

This was probably the first recognition of deep south-ward flow across the equator, but unfortunately it isall that Buchan (1895) said on the subject. One cannottell whether he also realized that the deep flow waslayered, with North Atlantic water overriding the bot-tom water from the Antarctic, or whether he viewed"deep water" in the terms of his day as a single unitmoving uniformly northward or southward, wherebythe oppositely directed North Atlantic and Antarcticflows would simply collide and upwell at some middlesouthern latitude.

Whatever Buchan may have believed, the first un-ambiguous, well-documented statement of deep south-ward flow from the North Atlantic between the bottomand intermediate waters from the Antarctic was madeby Brennecke (1911) on the basis of stations occupiedby the Deutschland along the western rim of the SouthAtlantic:

Das Hauptergebnis unserer Reihenmessungen ist dieFeststellung eines Tiefenstromes in etwa 1500 bis 3000m Tiefe der vom Nordatlantischen Ozean nach Sildenvordringt und durch hohe Temperatur und hohen Salz-gehalt sich von der uber- und unterlagernder Schichtabhebt. Dieser Tiefenstrom konnte bei allen Rehen-messungen von 5°S-Br. bis 40°S-Br. klar erkannt wer-den. Wenn bislang noch nicht erkannt worden ist, soliegt dies einerseits an der geringen Zahl der Messun-gen, die im diesen Schichten tatsachlich ausgefiihrtworden sind, andererseits an den friihen vielfach ben-utzten Maxima-Mimima-Thermometem, die eineTemperaturumkehr in der Tiefe nicht anzeigen. Soweitunserer Messungen Aufschluss geben, wird dieser Tie-fenstrom bei seinem Vordringen nach SUiden in gr6ssereTiefe gedrangt, bzw. in seiner Oberschicht mehr undmehr durch den schon von fruiheren Forschungen herbekannten, entgegengesetzt gerichteten, d.h. nordwirts

9Deep Circulation of the World Ocean

vordringenden Tiefenstrom in 1000 m (ausgezeichnetdurch das Minimum des Salzgehalts) gemischt.

The Challenger thermometers having been of the min-imum-type, Buchan could not have detected the tem-perature inversion. Another factor that contributed toBrennecke's success was that, although he could nothave known it, he actually made his stations in theswiftest part of the southward flow, where the inver-sion is much more markedly developed than to theeast, and where, happily, the strongest evidence for theflow was available.

In the early 1920s, Merz amassed a file of all availabledeep observations of temperature and salinity, and be-gan a program of systematic reexamination of thesedata to delineate the deep circulation of the wholeworld ocean. In the course of this study Merz and Wiist(1922) and Merz (1925) explicitly refuted the existenceof Lenz's symmetric convection cells in the Atlantic,on the evidence already cited, and they proposed acomprehensive new picture of the meridional circula-tion there, whose essential feature was hemisphericexchange of water: ocean-wide northward flow acrossthe equator in the upper kilometer, compensated bysouthward flow below to about 4000 m, with basin-wide northward and southward flows of bottom waterin the western and eastern basins, respectively.5 Theirworking materials were north-south sections of tem-perature and salinity, and they regarded the depressionof isotherms and isohalines in the upper 2 km in thenorthern subtropics as indicating sinking from the sur-face in those latitudes, and thus formation of the NorthAtlantic deep water there. In making this misinterpre-tation of that feature, they evidently did not fully con-sider either the baroclinic character of the Gulf Streamgyre or the development of anomalously high temper-atures and salinities at mid-depth from mixing of theMediterranean outflow into the gyre.6

Following Merz's death in 1925 on the Meteor Ex-pedition, M11ller and Wust undertook to complete hisprogram for the Indian and Pacific Oceans, respec-tively. In her study of the Indian Ocean data set Miller(1929) joined with Schott (1926) to advance the Merzscheme of meridional circulation, emphasizing he-mispheric exchange, for that ocean too-northwardflow of intermediate and bottom water, southward flowof deep water in between-with the deep water sinkingfrom the surface in the Arabian Sea, and receiving con-tributions from the Red Sea and Persian Gulf outflows.Thomsen (1933), however, found systematic error insome of these salinity values, and new data from theDana showed much less extensive southward penetra-tion of high-salinity deep water from the North IndianOcean than Miller had described, implying a differentcharacter of deep-layer flow from that in the Atlantic:certainly weaker, and if directed across the equator atall, only in response to the northward movement of

intermediate and bottom water. Implicit in this con-clusion was rejection of the idea of deep water sinkingfrom the surface of the Arabian Sea, although the influxof Red Sea water was recognized. From the distributionof bottom potential temperature, Wust (1934) clarifiedthe role of the Central Indian Ridge in dividing thebottom flow into at least two separate regimes (theNinetyeast Ridge being then unknown), with bottomwater entering from the south into both the westernand the eastern basins of the Indian Ocean, as Schott(1902) had suggested in his report on the Valdivia ob-servations.

For the Pacific, Wust (1929) found that the generalquality of data available was inferior to that for theother two oceans, and, with deep property differencesbeing much smaller there anyway, he did not reachsuch definite conclusions as for the Atlantic. Never-theless, his analysis suggested to him that the Pacificwas different from the Atlantic, in that hemisphericexchange was insignificant, and that, in both the Northand South Pacific, intermediate and bottom waterspread equatorward and deep water poleward (i.e., thatdeep water sank from the surface layer near the equa-tor, rather than in middle northern latitudes, as hethought it did in the Atlantic). In one respect, Wiist's(1929, 1938) work on the deep Pacific in this decadewas actually retrogressive, because some temperaturemeasurements made by the Tuscarora in 1874 that areincompatible with modem data led him to infer anorthern source for bottom water in the Sea of Okh-otsk, contrary to Prestwich's (1875) correct finding thatsinking to great depth does not occur anywhere in theNorth Pacific.

From his study of the Carnegie observations, Sver-drup (1931) realized that there was not, in fact, anyequatorial downwelling in the Pacific. He also recog-nized that Pacific bottom water did not derive fromsinking directly to the south near Antarctica, as is thecase for the Atlantic, but that all water below the low-salinity intermediate layers in the Pacific was, in effect,carried to it by the Circumpolar Current. He thoughtit likely that at least in the South Pacific there wassome southward return flow of deep water, as in theother two oceans, but the very low oxygen concentra-tion in the deep North Pacific indicated only smallhemispheric exchange.

The idea of a deep-water connection among the At-lantic, Indian, and Pacific Oceans via the CircumpolarCurrent was reinforced by the systematic cruises of theDiscovery in the Antarctic, which obtained evidencefor the extension of high-salinity deep water all theway from the Atlantic to the Pacific (Deacon, 1937).

By the time Wust began his comprehensive treat-ment of the Meteor results, Helland-Hansen and Nan-sen (1926) had shown clearly-and Defant (1931) hadreaffirmed-that the high salinities at mid-depths in

IOBruce A. Warren

the subtropical North Atlantic were due to the Medi-terranean outflow; and Wattenberg (1929) had tracedthe maximum in oxygen concentration below to highnorthern latitudes. Wiist (1935) therefore abandonedthe broad mid-latitude source for the deep water thatMerz and he (1922) had proposed, and that he (1928)had later modified somewhat. Instead, he distinguishedthree layers within the deep water, all spreading south-ward, and each with its own "formation" site: UpperNorth Atlantic Deep Water, characterized by the deepsalinity maximum attributed to the Mediterraneanoutflow; Middle North Atlantic Deep Water, identifiedby a maximum in dissolved-oxygen concentration, andtraced to the Labrador Sea; and Lower North AtlanticDeep Water, identical to Wiist's (1928) earlier "NorthAtlantic Bottom Water," defined by a second anddeeper oxygen maximum, and thought to be formedsomewhere in the waters south or southeast of Green-land by deep convection in winter.

The picture of deep circulation for the world oceanthat Sverdrup (Sverdrup, Johnson, and Fleming, 1942,chapter XV) then constructed was broadly Merzian:ocean-wide or basin-wide flows in deep and bottomlayers, the bottom water moving generally northward,the deep water southward-with the greatest hemi-spheric exchange in the Atlantic and the least in thePacific. Sverdrup departed qualitatively from the Mer-zian schematic only in inferring that the principalmovement of deep (and bottom) water in the NorthPacific was a slow, nearly closed, clockwise circula-tion. By this time, though, the areal extent of regionswhere deep sinking was thought to occur had beenmuch reduced from what Merz had supposed: sinkingin the southern hemisphere limited mainly to the Ant-arctic continental slope in the Weddell Sea (Brennecke,1915, 1921; Mosby, 1934), and sinking in the northernhemisphere confined to high latitudes in the westernbasin of the Atlantic, with probably smaller, subsurfaceinfluxes of relatively dense water from such marginalseas as the Mediterranean and Red Seas.

Even then, however, Wiist had already realized thatthe idea of deep and bottom flows in the Atlantic thatwere uniform in a basin-wide sense was oversimplifiedto the point of being misleading. His maps of tracerdistributions from the Meteor data (Wiist, 1935) dem-onstrated more intense southward propagation ofNorth Atlantic Deep Water in the western than in theeastern South Atlantic, and geostrophic velocity cal-culations (Wiist, 1938) showed relatively strong north-ward flow of Antarctic Bottom Water (at speeds ofseveral centimeters per second) only close to the SouthAmerican continental slope. In discussing property dis-tributions, Wiist (1935) was careful to speak of the"spreading" ("Ausbreitung") of water types, statingthat such spreading had a current-like character onlyon the western side of the Atlantic, and that elsewhere

it was accomplished, in effect, by eddy fluxes. He re-garded this distribution of currents as an extremelystrange phenomenon, for which he was unable to givea satisfactory explanation; and, indeed, the real signif-icance of these "deep western boundary currents" wasnot to be appreciated until dynamically consistentmodels of the deep circulation were developed.

It should be acknowledged, however, that, as pene-trating as Wuist's insights were, it was not, in fact, hewho first discovered the bottom-water boundary cur-rent, but Buchanan, who was presumably the chiefcontributor to chapter XXII of the Challenger ReportNarrative. The Challenger stations were the first setthat by virtue of their spacing could possibly have re-vealed such a feature, and Buchanan (Tizard, Mosely,Buchanan, and Murray, 1885a) did not fail to see it:

Another equally remarkable current is that whichbrings the cold water from the south polar regions upalong the South American coast at the bottom as far asthe Equator.... That this does take place is shown bythe low bottom temperatures observed at every Stationalong the western side of the South Atlantic almost tothe Equator, and that there is here a sensible currentof cold water along the bottom is shown by the suddenchange in the rate of decrease of temperature withincreasing depth which is observed at depths of from2000 to 2200 fathoms.

... The temperature of the water from 2400 fathomsto the bottom was uniform, the mean result of sixobservations being 32.43° . This water underruns thebody of the Atlantic water, which at 1500 to 2000fathoms has here a temperature of 37°, producing atemperature gradient of about 1.3° per hundred fathomsat the steepest. For the preservation of this gradient, aconsiderable supply of cold water is requisite, and itmust be drawn from higher latitudes. But any motionof the water towards the Equator will be accompaniedby a strong deflection to the westward (proportional tothe change of the cosine of the latitude). A measure ofthis deflecting force is furnished by the rise of this coldwater at the more inshore Station on the 2 8TH, wherethe maximum gradient is at about 1750 fathoms, whileon the 2 9TH, at a distance of 120 miles,7 it is at about2100 fathoms.

In the last two sentences Buchanan seems to havebeen groping toward some notion of the thermal-windrelation, and to have correctly associated the lateralscale of the slope in the maximum-gradient surfacewith that of the northward flow beneath. He had badluck with his discoveries, though. Not only was thisone ignored [except for an allusion by Wiist (1933)],but, as will be recalled, his almost simultaneous dis-covery of the Atlantic Equatorial Undercurrent wascompletely forgotten after three decades.

1.3 A Dynamical Framework

Paralleling the intuitively appealing concept of ocean-wide or basin-wide meridional flows were two otherideas that seemed equally reasonable. One is that, in

IIDeep Circulation of the World Ocean

compensation for the known sinking of near-surfacewater in small regions, there is a general slow upwardmovement of deep water over most of the rest of theocean, which balances the downward diffusion of heatfrom the upper layer and thus accounts for the exist-ence of the main thermocline. The other is that themean horizontal deep flow in the open ocean is strictlygeostrophic, in the sense that streamlines parallel iso-bars, and the speed is inversely proportional to theirspacing and to the sine of the latitude. As indicated insection 1.1, the latter may not seem so obvious now asit once did, but there is yet no proof to the contrary.Stommel (1965, manuscript of first edition completedin 1955) perceived the dynamical implication of theseideas-that the vertically integrated deep flow shouldbe directed poleward-with the bizarre corollary thatdeep water in the open ocean of known polar originshould be moving on the whole toward its sourcesrather than away from them. Specifically, broad, deepequatorward flows in the North Atlantic, South Indian,and South Pacific Oceans, as envisaged in the Merz-Sverdrup picture, do not seem consistent with the pe-culiar dynamics of slow motions on a rotating sphere.

The physical argument is simple. The geostrophicvorticity balance is flv = fwz, where f is the Coriolisparameter, l3 its meridional derivative, v the northwardvelocity component and w the vertical component, andz the vertical coordinate (positive upward). Integratingthe equation vertically from the ocean bottom to the"top" of the deep water (however defined), and takingthe floor of the ocean to be level in the large-scalemean, so that the vertical velocity at the bottom canbe taken to be zero in a large-scale sense, then yieldsthe result that, with upward velocities at the top of thedeep water, the meridional component of flow is pole-ward, in both the northern and the southern hemi-spheres. This can also be seen in more rudimentaryterms: because geostrophic flow cannot cross isobars,the transport per unit depth between any two stream-lines of the horizontal flow increases downstream inflow toward lower latitudes, and decreases downstreamin flow toward higher latitudes. If water is also movingupward out of the deep layer, then the associated hor-izontal flow must be the convergent one--directedpoleward.

Clearly, however, there must be regions where thisconstraint breaks down, so that deep water of polarorigin can reach the low latitudes where it is found.For the large-scale circulation generally, the regionswhere the rules of the open ocean are relaxed are thewestern boundaries of basins, where relatively narrow,intense currents exist to satisfy whatever boundary orcontinuity conditions are not fulfilled by the interiorflow (see section 1.5.4). Stommel (1958) therefore hy-pothesized deep western boundary currents for all theoceans to carry deep water away from its sources, to

supply it to the interiors of oceans from their westernsides (rather than from their polar margins) and thusfeed the poleward interior flows, and to correct ocean-wide continuity imbalances between the upward fluxand the meridional flow.

The flow patterns thus constructed were so contraryto intuition that, to test their consistency, Stommel,Arons, and Faller (1958) developed a corresponding the-ory for circulation systems forced by a variety of pre-scribed distributions of sources and sinks in a rotatingtank. The excellent agreement with the experiments[continued further by Faller (1960)] encouraged Stom-mel and Arons to rework the theory in spherical co-ordinates (1960a), and they used that theory to developa model of the global deep circulation (1960b), includ-ing estimates of velocities and volume transports. Thethree oceans were represented as flat-bottom basinsbounded by either parallels or meridians, and con-nected in the south by an analogue to the Antarctic Cir-cumpolar Current. The flow was taken to be baro-tropic, driven by sources in the model counterparts ofthe Weddell Sea and northern North Atlantic (repre-senting sinking from the surface layer), and by thecompensatory upwelling, distributed uniformly overthe rest of the world ocean. The resulting circulationpattern in a slightly modified version of the model (Kuoand Veronis, 1973) is sketched in figure 1.1. The ana-logue to the Circumpolar Current is indicated along thesouthern margin; its transport through the counterpartof the Drake Passage is not determined by the modeldynamics, and was estimated roughly from observa-tions. The layer thickness is 3 km, the strength of eachsource is about 17 x 106 m3 s-', and the upwelling rateis 1.5 x 10-5 cms- 1; the horizontal speeds in the in-terior are of order 10-2 cm s- 1, and the transports of thewestern boundary currents, of order 10 x 106 m3 s-'.

The interior flows are everywhere poleward and east-ward, fed by intense currents along the western bound-aries. In the Atlantic, the boundary current emanatesfrom the northern source and flows to the southernlimit of the South Atlantic, where it, the interior flow,and the Weddell Sea source increase the transport ofthe deep Circumpolar Current. That in turn is thesource for northward-flowing western boundary cur-rents in the Indian and Pacific Oceans, entering at theirsouthwestern corners. At the northern margins of thosetwo oceans the interior flow forced by the upwellingfeeds into northern boundary currents, which flowwestward into southward-directed western boundarycurrents.

This model was never intended as a realistic descrip-tion of the deep-ocean circulation, and in at least tworespects it would be a qualitatively bad one. By treatingthe ocean floor as level, the model gives single circu-lation systems in the northern and southern hemi-spheres of each ocean, whereas the mid-ocean ridges

I2Bruce A. Warren

divide the oceans into separate basins with multiplecirculation systems and multiple entrance points fordeep water from the Antarctic. Also, because the modelis barotropic, it cannot allow layered deep flow, withopposed boundary currents one above the other, as oc-cur most spectacularly in the South Atlantic, whereAntarctic Bottom Water flows northward beneath thesouthward current of North Atlantic Deep Water. (Infact, there could be no direct, boundary-current flow ofAntarctic Bottom Water from the Weddell Sea into theSouth Atlantic anyway were it not that the Drake Pas-sage sill is substantially shallower than the floor of theAtlantic to the east.)

Nevertheless, it was a dynamically consistentscheme, which offered some drastically different ideasabout the deep circulation from those in the Merz-Sverdrup picture of basin-wide meridional flows, es-pecially the new physical concept of deep westernboundary currents. A search was immediately made forthe predicted deep current off the east coast of theUnited States, by means of direct current measure-ments with neutrally buoyant floats; its discovery wasreported by Swallow and Worthington (1957), and pro-vided dramatic support for the essential truth of thenew conceptual scheme. Considering what has beenlearned since then about the prevalence of strong, low-frequency motions in the deep water, one would prob-ably question now how sound a "proof" of the currentthose measurements were, given their restricted geo-graphical scope and brevity of duration (1 to 4 days);

but they were consistent among themselves and withlater observations (see section 1.5).

Actually, as pointed out in section 1.2, the relativelyintense deep meridional flows adjacent to South Amer-ica had been recognized long before, and Stommel(1950a) had already suggested that they might be dy-namically akin to the Gulf Stream. Moreover, as thetheory was being developed, Wiist (1955) constructedgeostrophic velocity sections for two of the Meteortransects, one of which vividly illustrated these twocurrents, the one directed northward, the other south-ward. Thus the general idea of deep western boundarycurrents was not without observational support evenbefore its confirmation by Swallow and Worthington(1957).

It was only for simplicity and clarity that the Stom-mel-Arons model was presented as a barotropic one.It can be brought somewhat closer to the real stratifiedocean by regarding it instead as a theory for verticallyintegrated flow. Equatorward interior flow at some lev-els is therefore not excluded, so long as the verticalintegral is poleward. Layered interior flows might bedifficult to detect in tracer distributions, though, be-cause mixing of properties can mask the flow patterns(see below, this section).

The theory is clear, however, in allowing no hemi-spheric exchange of water (of the sort that Merz em-phasized) through the interior flow, because at all lev-els the meridional component of flow vanishes withthe Coriolis parameter at the equator: hemispheric ex-

WESTERN BOUNDARYTRANSPORT

no- 1 2 3 4

600

300

U-

30

550

70OU- VV U- DU'L 4U'_ 1UVW YVUW

0 3

zo

-2

TRANSPORT OF ANTARCTIC CIRCUMPOLAR CURRENT (without recirculation)

Figure I . Deep circulation in a schematic world ocean drivenby uniform upwelling with sources at the counterparts of the

North Pole and Weddell Sea. See text. Transports measuredin units of about 6 x 106 m s-'. (Kuo and Veronis, 1973.)

I3Deep Circulation of the World Ocean

-

I

I,

-L F

�_j L-------- - ----- ___1

I - - - - - - - - - - - - -

, - . ..

v^

change by the mean flow is limited to the boundarycurrents. There can still be hemispheric exchange ofproperties all along the equator, of course, through lat-eral mixing.

A more fundamental novelty of the Stommel-Aronsmodel is in its looser connection between specific hor-izontal flows and specific sinking phenomena than wasimplied in the earlier schemes. The net interior flowat any position is directly coupled to the local upwardmovement of the deep water, but the amplitude of thatupwelling (as distinguished from its geographical var-iation) is related to the world-ocean integral of the rateof sinking into the deep water, rather than to the down-ward flux in any one sinking region. Nor is there anyone-to-one relation between the transport of a deepwestern boundary current and the rate of sinking in aregion from which it may emanate, even after subtract-ing losses to interior upwelling-as may be seen byconsidering the water budget for the portion of a deepocean north of an arbitrary parallel of latitude. Theboundary-current transport T out of that volume (pos-itive southward) can be calculated as

T=f wdl- fwdo- + S,

where S is the rate of sinking into that volume, and wis the vertical velocity at the top of the deep water; thefirst term on the right, expressing the net interior trans-port into the volume, includes an integral of w alongthe arbitrary latitude, and the second term is an inte-gral over the top of the volume. Not only can T be lessthan S, as is obvious without the formula, but T canalso be greater than S, signifying recirculation in anocean or basin, because the boundary current mustcompensate not only the sinking but also any excessof interior transport into the volume over upward fluxout of it. Extreme examples of this effect are illustratedin the northern Pacific and Indian areas of figure 1.1,where southward flowing boundary currents are re-quired even though no sinking occurs to the north.

Despite the physical consistency of the deep-circu-lation model, the Merz-Sverdrup picture still looksintuitively more believable, in that poleward interiorflows are not obvious in the property distributions.Stommel and Arons (1960b) suggested that lateral mix-ing of tracer properties might be so intense as to maskthe patterns of such slow mean interior flow; to explorethat effect, Kuo and Veronis (1973) made calculationsof dissolved-oxygen distributions based on the circu-lation field of figure 1.1 and ranges of values for themixing coefficient and oxygen-consumption rate. Theyprescribed the oxygen concentration at the two sinkingpoints, and compared the resulting fields with the ob-served distribution at 4 km.

Two extreme cases are instructive. With no mixing,high oxygen concentrations are introduced to theoceans from their western boundaries, and the valuesdiminish eastward on account of consumption, withthe effect that the lowest values are found along theeastern boundaries; the western boundary currents arediscernible in the field, as well as the eastward com-ponent of interior flow, but the meridional componentof the latter is not particularly evident. With mixingbut no advection, however, the oxygen concentrationis simply diffused zonally in the circumpolar belt, andrelatively high values are introduced to the Indian andPacific Oceans all along their southern boundaries; theconcentrations decrease northward and the lowest val-ues in those oceans are obtained at their northernboundaries. (In the Atlantic, with northern and south-ern sources, the lowest values are found in the inter-tropical zone.)

The best match with observation (figure 1.2) was anintermediate case, representing approximately "equal"effects of advection and mixing. The mixing coefficientin this case was 6 x 106 cm2 s-', and the oxygen con-sumption rate, 2 x 10-3 mll- 1 yr-'-values roughlyconsistent with those from other studies. The patternswere found to depend on the speed of the CircumpolarCurrent, since that affected the "source" values for theIndian and Pacific Oceans; the optimal transport valuefor the Drake Passage was 35 x 106 m3 s- '. That figurefrom the model is difficult to compare precisely withobservation because, given that the "top" of whatought to be considered deep water is much shallowerin the Drake Passage than in the tropics, it is not clearwhat is the most appropriate depth interval to choosefor comparison; density sections combined with year-long current measurements made during the ISOS pro-gram gave estimates of the net transport below 1000 mof about 47 x 106 m 1s-1, and below 2000 m of 12 x106 m3 s- ' (H. Bryden and R. D. Pillsbury, personalcommunication).

Certainly there is nothing in figure 1.2 to indicatethe character of the underlying flow field, except per-haps some hint of the western boundary currents. Themost obvious interpretation of the pattern-and onequite wrong-would be of slow, ocean-wide, trans-equatorial movement northward from the Antarcticand southward from the northern North Atlantic, inthe manner of Merz and Sverdrup, though perhapssomewhat stronger in the west than in the east. Eventhough, as Kuo and Veronis (1973) pointed out, thesecalculations were based on a highly simplified flowscheme (especially as not constrained by the ridge sys-tem), the results do suggest that, with recognition ofa moderate degree of lateral mixing in the ocean, ob-served tracer distributions can probably be rationalizedin terms of more realistic flow fields constructed fromthe Stommel-Arons dynamics.

I4Bruce A. Warren

Figure I.2 Model distribution of dissolved-oxygen concentra-tion (mll - }) associated with the flow field in figure 1.1. See

Although the Stommel-Arons (1960b) model itselfis not a theory for the actual mean deep circulation,the model does provide a dynamical framework inwhich to think about aspects of the circulation. Evi-dence bearing on the verification of elements of themodel as applied to the real ocean is discussed in sec-tion 1.5.

1.4 Sources of Deep Water

Even though the individual sites where water sinks togreat depth do not seem so important for driving theoverall deep circulation as they once did, it is desirableto know the rates at which sinking occurs, because thecompensating upwelling from the rest of the deepwater is thought to force its mean horizontal circula-tion. Moreover, the climatic conditions at the specificsinking regions determine the characteristics of thedescending water, and thereby the properties and lay-ering of deep water through the world ocean. The sink-ing phenomena are an integral part of the deep circu-lation as well, and the various processes by which near-surface water is brought to depth are of considerableinterest in themselves.

As noted in section 1.2, in the southern hemispheredeep sinking is limited to the waters around Antarc-tica. In the northern hemisphere it occurs only in thenorthern North Atlantic, but there are also outflowsfrom marginal seas: the Mediterranean Sea, the RedSea, and the Persian Gulf. These outflows descend prin-cipally to mid-depths, but affect the water character-istics at deeper levels too.

1.4.1 Sinking around AntarcticaThe only location in the Southern Ocean where sinkingto the bottom through convective overturning has beenidentified is the Bransfield Strait (figure 1.3), whichseparates the northern tip of the Antarctic Peninsula

text for "best-fit" values of controlling parameters. (Kuo andVeronis, 1973.)

from the South Shetland Islands {(Clowes, 1934; Gordonand Nowlin, 1978). There a somewhat isolated troughof depth 1100-2800 m is filled with a mass of nearlyhomogeneous water very different from the surround-ing Circumpolar Deep Water: much higher oxygen con-centration, and notably lower temperature, salinity,and nutrient concentrations. Values of tritium concen-tration, increasing toward the bottom below 300 m in1975, are further evidence for recent contact betweenthe bottom water and the sea surface. Although noobservations have been made there in the winter sea-son, these circumstances strongly imply that local win-ter convection renews the water column through itsfull depth. So far as one can tell from the propertydistributions, however, none of this bottom waterflows out of the trough into the rest of the ocean,probably because of the topographic barriers.

Reduced vertical stratification northeast of theBransfield Strait, in a zone separating the Scotia andWeddell Seas, raises the possibility of some deep con-vection there too (G. E. R. Deacon and Moorey, 1975),but the evidence to date is less clear-cut than that fromthe Bransfield Strait.

The bottom water that does enter the rest of theocean originates in several areas of the Antarctic con-tinental shelf, where water is made sufficiently coldand saline that, in flowing down the continental slopeand mixing with the surrounding deep water, it is denseenough to reach the floor of the ocean. In order ofdecreasing amount and extent of influence on the deep-water property distributions-and presumably, there-fore, of rate of bottom-water production-these regionsare the Weddell Sea, the Ross Sea, and the Adlie Coast(figure 1.3; A. L. Gordon, 1974); perhaps there is someproduction off Enderby Land too.

Because of its predominant role, the Weddell Sea hasattracted the most attention, although the extremeharshness of working conditions there has discouraged

ISDeep Circulation of the World Ocean

1800

Figure i.3 Index map identifying Antarctic place names men-tioned in text.

I6Bruce A. Warren

any wintertime observations since the Deutschlandwas frozen into the pack ice in 1912 (Brennecke, 1921).Nevertheless, summer data suggest that flow down thecontinental slope takes place throughout the year. Sec-tions of temperature, salinity, and dissolved-silica con-centration, as constructed from observations made in1968 on a station line running eastward from the Ant-arctic Peninsula (figure 1.5; see figure 1.4 for stationpositions) show a 200-m thick layer of relatively cold,fresh, low-silica bottom water on the slope, extendingonto the floor of the Weddell Sea (Carmack, 1973).Although the station line did not reach the continentalshelf ice cover has prevented any shelf-water obser-vations in that region), the evidence points plainly toa flow of shelf water down the slope, probably quiteobliquely, entraining, and being diluted by, the sur-rounding water along the way. Similar sections nearlongs. 40°W and 29°W (Foster and Carmack, 1976a), aswell as 10-20°W (Carmack and Foster, 1975), showdiminishing evidence for downslope flow to the east-ward, and the distribution of bottom potential temper-ature in the Weddell Sea (figure 1.6; Foster and Car-mack, 1976a) demonstrates that the newly formedbottom water leaves the continental slope mainly atthe northern tip of the Antarctic Peninsula, in lats.63-65°S.

This Weddell Sea Bottom Water-defined by Car-mack and Foster (1975) as of potential temperature 0<-0.7°C-is widespread in the Weddell Sea, but since

60" 50" 40" 30e 20"

60' 50' 40- 30' 20'

Figure 1.4 Positions of stations occupied by U.S.C.G.C. Gla-cier in the Weddell Sea, 26 February- 13 March 1968, that areused for the construction of the sections in figure 1.5. Isobathslabeled in meters. (After Carmack, 1973.)

water of such extreme characteristics is not found tothe north in the Argentine Basin (Reid, Nowlin, andPatzert, 1977), it must be mixed fairly rapidly into thewater above. That overlying water in the Weddell Seais itself also "Antarctic Bottom Water," but "older"water, whose, properties have been modified by mixingelsewhere; it flows westward into the Weddell Seaclose to the continent in the clockwise Weddell gyre(e.g., Brennecke, 1918; G. E. R. Deacon, 1976).

By combining station data with current records ofduration between 3 and 4 weeks, Carmack and Foster(1975) estimated that the net flow of Weddell Sea Bot-tom Water out of the Weddell Sea is about 16 x106 m3 s -'. At their stations, that water had a meanpotential temperature of -0.786°C, brought about, pre-sumably, by mixing overlying water of 0 = -0.6 to-0.7°C into water that initially, at the edge of thecontinental shelf, was of 0 = -1.2 to -1.4°C. Theyestimated, therefore, that the rate of sinking from theshelf in the Weddell Sea is 2-5 x 106 m3 s-1.

In accounting for the formation of Weddell Sea Bot-tom Water, three distinct processes need to be treated:(1) how the density of the shelf water is increased tothe point where it can sink through the relativelywarm, saline water at mid-depths; (2) how the watermoves off the shelf; and (3) how it descends with en-trainment to the floor of the Weddell Sea. The firstproblem has been considered at length by Gill (1973).Although near-surface water in the open Weddell Seahas a temperature close to the freezing point, it isgenerally too fresh to be dense enough to sink throughthe water below. On the shelf, however, the salinitycan be increased by salt release during ice formation,and there more saline water is, in fact, found, which isdense enough to sink to depth. By itself the annualcycle of freezing and melting at-the sea surface cannotaccount for the greater salinity of the water on theshelf, because whatever salt were added to the waterduring winter freezing would be mixed back into themelt water during summer. Nor can it be supposedthat the salt-enriched water formed during winter im-mediately leaves the shelf, because it is found duringsummer both on the shelf and flowing down the con-tinental slope. Seabrooke, Hufford, and Elder 11971)therefore suggested that the salinity enhancementtakes place year-round through freezing at the base ofthe Filchner and Ronne Ice Shelves (figure 1.3). Gill(1973) showed that this was unlikely, however: becausethe heat flux through the ice shelf is too small to allowenough salt release for a significant rate of bottom-water production; because the evidence is for meltingrather than freezing under the ice shelf anyway; andbecause the temperature of the most saline shelf wateris commonly well above the freezing point at the depthof the ice-shelf base. It seems instead that the higher

I7Deep Circulation of the World Ocean

stations

29 30 31 35 37 47 58 55 0

1000

E

-va-o

2000

3000

4000

(1.5B)

0

1000

2000

-Ca.

'o

3000

4000

stations

29 30 31 35 37 47 58 55

stations

29 30 31 35 37 47 55

(1.5C)

Figure .5 Sections of (A) potential temperature (°C), (B) sal-inity (oo), and (C) dissolved silica concentration (M l-') alonga line running eastward from the Antarctic Peninsula (left) inthe southwestern Weddell Sea, illustrating the descent ofdense shelf water along the Antarctic continental slope. (Car-mack, 1973.) See figure 1.4 for positions of stations.

I8Bruce A. Warren

0

1000

2000E

c

Q0.

' -Smax.----:C

I I I | I I I I

- \ ) ) *ze°x ' -2 .2 ·.

-0.2

-0.4'- -0.8 ._-3000

4000

(1.5A)

I I I I I I

i* * ..

I

- - -

_

_

_

-

_

-

Figure I.6 Distribution of bottom potential temperatures inthe Weddell Sea, illustrating the eastward spreading of newlyformed bottom water from the northern tip of the AntarcticPeninsula. (Foster and Carmack, 1976a.)

I9Deep Circulation of the World Ocean

salinities are indeed due to the seasonal freezing-melt-ing cycle at the sea surface: evidently there is a pre-vailing offshore movement of the pack ice that trans-ports potential melt water locked up in ice out of theregion, thus allowing a net brine production there. Thatmovement also increases the wintertime brine releaseby opening up leads in the ice field where much morerapid freezing occurs. A plausibly estimated value forthe net effective rate of ice formation on the continen-tal shelf, through the complete seasonal cycle, is about1 myr - ' (Gill, 1973).s

Given denser water on the shelf than offshore, it isstill not immediately obvious why it moves toward thecontinental slope, because even if the density differ-ence were imparted impulsively, the water should onlyadvance offshore a distance of order of the deformationradius (10-20 km), and the ensuing steady motionshould be geostrophic flow along the isobaths, ratherthan across them. In zonally symmetric conditionswithout longshore pressure gradients, the only possiblesteady efflux of bottom water from the shelf would bein a frictional boundary layer, but the likely transportthere is much too small to be realistic (Gill, 1973).Antarctica is far from zonally symmetric, of course,and the existence of large-scale indentations in thecoastline, with much broadened continental shelves,as in the Weddell Sea, allows much more substantialtransport off the shelf.

In particular, the salinity of the shelf water in theWeddell Sea increases markedly from east to west, byabout 0.4%o, and the associated longshore density gra-dient implies a thermal-wind shear, with surface watermoving onshore and the deeper shelf water movingoffshore. Combining these density data with a valuefor the onshore Ekman flux associated with easterlywinds, Gill (1973) estimated the offshore transport ofdense water along the length of the shelf to be about1 x 106 m3 s- , of which half is compensation for thesurface Ekman flux. The net offshore salt transportassociated with this vertical circulation is only abouthalf that required to remove the amount of salt addedat the surface by net ice formation of 1 myr -1. It seemslikely, though, that the longshore salinity gradient isitself maintained by a slow westward flow along theshelf-part of the Antarctic coastal current-wherebythe salinity of the water is gradually raised to the west-ward by the widespread freezing there. There are nodirect measurements of mean velocities on the shelf,but a longshore flow of 1 x 106 m 3 s- l, carrying salt-enriched water off the shelf at its western end on theAntarctic Peninsula, would be adequate to remove theremaining salt produced during the annual ice-forma-tion cycle (Gill, 1973). These estimates, derived en-tirely from conditioning on the shelf, thus give a figurefor the total flux of dense water off the shelf of about

2 x 106 m3 s- l, which is in fairly good agreement withthe value of 2-5 x 106 m3 s- ' obtained by Carmackand Foster (1975) from direct measurements of the out-flow of Weddell Sea Bottom Water.

The offshore-moving shelf water encounters thewaters of the open Weddell Sea in a frontal zone at theedge of the continental shelf. Detailed studies of thiszone demonstrate a rather complicated mixing process,involving both the warm deep water of the WeddellSea and the winter-cooled near-surface water, throughwhich the characteristics of the shelf water are modi-fied to those of the descending Weddell Sea BottomWater (Foster and Carmack, 1976a).

As noted, there are subtleties as well in the dynamicsof the descent on account of the rotational constraint,which tends to orient the flow westward along theisobaths rather than across them. In order to identifythe conditions that bring about flow down the slope tothe floor of the Weddell Sea, with the amount of en-trainment observed, Killworth (1977) considered a se-ries of simplified models of turbulent plumes in a sta-bly stratified environment. Bottom friction wasessential, of course, for any downslope flow at all, butthe mere fact of shelf water being denser than its sur-roundings was not in itself sufficient for flow to reachthe sea floor with the right entrainment, whether theplume was two-dimensional, three-dimensional, or in-termittent. Because the thermal-expansion coefficientfor sea water increases with pressure, however, verycold descending water can draw on its own internalenergy to increase its kinetic energy; thus if a relativelydense falling parcel is sufficiently colder than its sur-roundings, the buoyancy force acting on it can increaseduring the descent, even though the parcel is falling instably stratified water. With this effect taken into ac-count, Killworth (1977) was able to construct modelplumes (three-dimensional, steady as well as intermit-tent) that both descended far enough to reach the floorof the Weddell Sea and entrained roughly the observedamount of surrounding water along the way. These hada substantial component of flow along the isobaths too,consistent with the bulk of the bottom water enteringonto the sea floor in the northwestern comer of theWeddell Sea.

The Weddell Sea Bottom Water is fresher than theoverlying water, but in the southwestern Pacific-spe-cifically, north of the Ross Sea (figure 1.3) to the mid-ocean ridge (about lat. 65°S), and close to Antarcticasouth of Australia-the salinity increases with depthnear the bottom to values >34.72%o (Gordon, 1971,1975a; Gordon and Molinelli, 1975). This high-salinitybottom water must have a different source, and it hasbeen traced to the Ross Sea (e.g., A. L. Gordon, 1974),where cold, saline shelf water has been observed to

20Bruce A. Warren

descend the western continental slope in a mannersimilar to that of the downslope flow in the WeddellSea (Gordon, 1975a). Although the effect of the RossSea bottom water on the deep temperature distributionis barely noticeable, and its effect on the salinity dis-tribution only became clear through the precision af-forded by conductivity measurements, its productionrate is not necessarily a great deal less than that ofWeddell Sea Bottom Water, because the water withwhich the shelf water mixes to form bottom water isboth more saline and much warmer than in the Wed-dell Sea. No transport measurements like those of Car-mack and Foster (1975) have been made, however.

The oceanographic conditions on the Ross Sea con-tinental shelf are roughly similar to those in the Wed-dell Sea, in that there is westward flow along the shelf,and the salinity of the shelf water increases to the westby 0.3-0.4%bo (Jacobs, Amos, and Bruchhausen, 1970;Jacobs, Gordon, and Ardai, 1979). This correspondencesuggests that the mechanisms sketched by Gill (1973)are also operative in the Ross Sea in moving densewater off the shelf onto the continental slope. A ver-tical circulation (involving the Ekman transport andgeostrophic shear) about 60% as large as that on theWeddell Sea shelf is indicated (Gill, 1973); that is, anoffshore flux along the length of the Ross Sea shelf ofroughly 0.6 x 106 m3 s-'. No estimate has been madeof the rate of removal of shelf water by the longshoreflow.

The water deeper than 300-500 m on the westernpart of the Ross Sea shelf is the most saline (34.8-35.0°bo) and the most dense (or = 28.0-28.1) waterfound in the Antarctic (Jacobs, Amos, and Bruchhau-sen, 1970; Gordon, 1971). One wonders to what extentit may be involved in the formation of bottom waterin the Ross Sea, but no very satisfactory account seemsto have been given for the existence of this very salinewater. Because the high salinities of the layer are highlycorrelated with low temperatures, and because itsthickness and breadth increase southward to the RossIce Shelf, it is tempting to think that the characteristicsof the layer may be imparted through freezing at thebase of the ice shelf. Near the front of the ice shelf,though, the evidence is of melting at the base (Jacobs,Amos, and Bruchhausen, 1970). Moreover, vertical var-iations of temperature and salinity observed beneaththe ice shelf far to the south at 82°S, 169°W demon-strated melting there too at the time of observation,but because the lowest 6 m of ice was of marine originrather than terrestrial, it is not certain what the pre-vailing condition is at that point (Gilmour, 1979; Ja-cobs, Gordon, and Ardai, 1979). Even if freezing is oc-curring, the maximum possible rate at the base of theice shelf is very small (3.5 cm yr-1; Jacobs, Gordon, andArdai, 1979), and any outflow from beneath the ice

shelf consistent with the required salinity enhance-ment would be much less than the estimated offshoretransport of water on the continental shelf. The bulkof this very saline water is found in a large topographicdepression on the shelf, whose sill depths near the edgeof the shelf are not well known (Jacobs, Amos, andBruchhausen, 1970); perhaps most of that dense layeris indeed renewed only slowly through freezing underthe western Ross Ice Shelf, is trapped in the depression,and does not contribute significantly to the local bot-tom-water formation (Gordon, 1971).

On the other hand, the high salinity of this westernshelf water might be due entirely to annual sea-icefreezing, as proposed by Gill (1973). Unusually densewater formed in that process could contribute substan-tially to bottom water, or it might form infrequentlyand merely accumulate in the topographic depressionif the sill is actually shallow enough to allow muchcontainment. The situation is unclear.

Several similar deep depressions-though of muchsmaller area-exist on the continental shelf off theAd6lie Coast of Antarctica (figure 1.3). They are filledbelow their sill depths with water of temperature closeto the freezing point, and of salinity 34.4-34.7oo, prob-ably as a result of deep convection associated withwinter sea-ice formation (Gordon and Tchernia, 1972).Evidently such water spills over the sills, because it isfound as a layer a few tens of meters thick on thenearby continental slope, underlying the high-salinitybottom water from the Ross Sea. This low-salinity bot-tom water mixes into the Ross Sea water rapidlyenough that it has not been detected far from the con-tinental rise near its points of origin. No estimates ofthe rate of production of the Ad6lie Coast water havebeen made, though the rate must be relatively small,given the small volumes of such water that have beenfound away from the shelf. The flow down the conti-nental slope appears intermittent, inasmuch as dupli-cated stations in February 1969 and December 1971showed that the comparatively fresh layer was morewidespread at the latter time (Gordon and Tchemia,1972).

Evidence has also been cited by Jacobs and Georgi(1977) for some small production of bottom water offEnderby Land. They suggest that new source regionsare revealed almost with each new exploration of theAntarctic continental slope.

Bottom water from these several sources-the Wed-dell Sea being the paramount one-mixes with thewarmer, more saline water above to form the "Antarc-tic Bottom Water" of the world ocean. Since all thesubtypes originate through freezing at the sea surface,it is the coldest, and thereby the densest, deep waterin the open ocean, detectable into northern latitudesby low temperatures close to the bottom.

21Deep Circulation of the World Ocean

Not all deep water formed in the Antarctic is bottomwater, however. Carmack and Killworth (1978) havedescribed a layer of relatively cold, fresh water atdepths of about 2 km off Wilkes Land (figure 1.3; long.150-170°E), which they demonstrated through Kill-worth's (1977) plume theory to have originated mostprobably from shelf water that is not quite denseenough to sink to the bottom. Instead, it sinks to adepth where its density matches that of the surround-ing water, in the manner of the Mediterranean and RedSea outflows. This particular instance of sinking tomid-depths was discernible because the shelf waterintruded into saline deep water from the Ross Sea, andhence could be identified by its relatively low salinity.Such phenomena seem possible elsewhere around Ant-arctica, but in regions where the surrounding deepwater is fresher-more like that in the Weddell Sea-they would be far more difficult to detect in distribu-tions of temperature and salinity, though perhaps lessso in distributions of nutrients (e.g., silica). As Car-mack and Killworth (1978) pointed out, if the salinityof shelf water really is increased almost everywherearound Antarctica by annual freezing, then some proc-ess must operate to remove the extra salt from theshelf; that process may be sinking of water to mid-depths, in the manner that was observed off WilkesLand. If so, then the rate of deep sinking in the Ant-arctic might be much greater than that involved in theimmediate production of bottom water; but there is nobasis at present for estimating how much greater.

Furthermore, a set of closely spaced observations inthe eastern Weddell Sea in February 1977 revealed acolumn of relatively cold, fresh water some 30 km indiameter, extending from near the surface through thewarm, saline deep water to about 4000 m (Gordon,1978). It suggests a relic of deep convective overturn-ing, such as has been well documented in the north-western Mediterranean Sea in regions of similarlysmall area (MEDOC Group, 1970). It may represent yetanother mode of deep sinking in the Antarctic, if notto the bottom at least to mid-depths, that had beenmissed in the past, perhaps because of difficulty ofaccess to critical areas or too coarse a spacing of sta-tions. The specific conditions for overturning in theWeddell Sea have been carefully considered by Kill-worth (1979), but again there is unsufficient basis atpresent for estimating how significant a contributionsuch events may make.

1.4.2 Sinking in the Northern North AtlanticNext to the Antarctic Bottom Water, the densest waterin the open ocean is what Wiist (1935) called LowerNorth Atlantic Deep Water. It does not originate, how-ever, as Nansen (1912) first proposed, and Wiist (1935)and Sverdrup (Sverdrup, Johnson, and Fleming, 1942)

maintained, from wintertime convection to the bottomof the Irminger or Labrador Seas,9 but from dense Nor-wegian Sea water that flows into the North Atlanticover three sills on the ridge connecting Greenland andthe British Isles (figure 1.7). These overflows entrainresident North Atlantic water in the course of theirdescent, and join together to form the bottom water ofthe northern North Atlantic.

It is curious that the significance of these overflowsfor the large-scale circulation was not appreciated untilthe late 1950s. Dr. Carpenter (Carpenter, Jeffreys, andThomson, 1869) had hypothesized long before thatoverflow east of the Faroe Islands was the source ofbottom water in the northern North Atlantic, and Ti-zard (1883) actually observed some such flow over theWyville Thomson Ridge, which is the southwesternboundary of the Faroe Bank Channel (figure 1.7). Knud-sen (1899) discovered the western overflow, and alsofound overflow water on the southern side of the Ice-land-Faroe Ridge, which he supposed had come southacross that ridge; in hindsight, his station positionssuggest that this water was really outflow from theFaroe Bank Channel, and had thus derived from theeastern, rather than the central, overflow. In any case,with better station coverage, Nielsen (1904) unambig-uously established the existence of the central over-flow, and Nansen (1912), with additional observations,discussed all three of them, presenting illustrative sec-tions of water properties across the ridge. Yet somehowall four observers, with the possible exception of Niel-sen, believed that the overflows could make only aminor contribution to the deep water of the NorthAtlantic-an opinion also expressed by Wiist (1935).Although Jacobsen (1916), Brennecke (1921), and De-fant (1938) all briefly suggested somewhat greater sig-nificance for the overflows, they were generally ignoreduntil Cooper (1952, 1955a,b) and Dietrich (1956,1957a,b), more or less independently, recalled them toattention." [Evidence for overflow was also noted byVinogradova, Kislyakov, Litvin, and Ponomarenko(1959) in new data obtained at this time, but they, likethe writers at the turn of the century, did not recognizeits large-scale significance.]

The deepest passage across the ridge is the FaroeBank Channel, which has a sill depth of about 800 m.Evidently some small amounts of Norweigan Sea waterin this passage-do spill intermittently over the WyvilleThomson Ridge (Ellett and Roberts, 1973), but most ofthis overflow continues northwestward through thechannel to its sill, where the measured shear of geo-strophic velocity in combination with analysis of watercharacteristics indicates an outflow of Norwegian Seawater of roughly 1 x 106 m3s - (Crease, 1965). Afterexiting from the Faroe Bank Channel, this current joinsa second overflow, which passes over the ridge betweenthe Faroe Islands and Iceland, where the sill depth is

22Bruce A. Warren

300-400 m (figure 1.7).11 By referencing geostrophicvelocity calculations for sections southeast of Icelandaccording to the water characteristics, Steele, Barrett,and Worthington (1962) estimated the volume trans-port of this combined Iceland-Scotland overflow, afterit has left the ridge, to be about 5 x 106 m3 s- l. Analysisof that overflow water in terms of its component watertypes showed that about 3 x 106 m3 s - l of the net flowis entrained Atlantic water, leaving 1 x 106 m3 s- l forthe transport of "pure" Norwegian Sea water acrossthe Iceland-Faroe sill (Worthington, 1970). With dif-ferent data, Hermann (1967) also estimated roughly2 x 106 m3 s-

1 for the net outflow of Norwegian Seawater between Iceland and Scotland.

This overflow continues southward against the east-ern side of the Mid-Atlantic Ridge, and passes throughthe Gibbs Fracture Zone near lat. 53°N into the Labra-dor Basin (figure 1.7). Its westward volume transportthere was calculated by Worthington and Volkmann

50"1

400N

(1965)-on the same basis as by Steele, Barrett, andWorthington (1962) farther north-to be still about5 x 106 m3 s- x, suggesting little direct flow into lowlatitudes of the eastern North Atlantic, although Leeand Ellett (1965) have detected overflow influence onwater characteristics as far south as lat. 47°N in theeastern Atlantic. This current then flows northward inthe eastern Labrador Basin, and, south of Greenland,joins the third overflow from the Norwegian Sea. Thatoverflow comes through the Denmark Strait (Dietrich,1957b; Mann, 1969), where the sill depth is about600 m, and is easily identified as cold water bankedagainst Greenland (figure 1.8).

Geostrophic velocity calculations, referenced bywater characteristics as for the Iceland-Scotland over-flow, gave a figure of about 10 x 106 m 3 s - for thevolume transport of the full overflow current south ofGreenland (Swallow and Worthington, 1969), therebyindicating a contribution of 5 x 106 m 3 s -l from the

Figure I.7 Index map identifying the overflow currents fromthe Norwegian Sea and local place names mentioned in thetext. (After Worthington, 1970.)

23Deep Circulation of the World Ocean

ZIZ

QQQ1%Qz

0 km 10 20 30 40 50

Figure .8 Temperature (C) section across the DenmarStrait in lats. 65-66°N (left, west; see figure 1.7), illustratinthe southward flow of cold water from the Norwegian Se;

Denmark Strait overflow. Worthington (1970) has e,timated that about one-fifth of this is Atlantic wattentrained by the overflow as it descends the Greenlancontinental slope, but the irregularity of the overfloiis so great that the fraction may generally be muclarger. For example, from a series of hydrographic sections reported by Mann (1969), Smith (1975) calculatean overflow transport of 1.3 x 106 m3 s- ' near the silwhich increased fourfold toward the southern tip cGreenland. Even though the overflows on either sidof Iceland originate from the same Norwegian Sewater, the Denmark Strait overflow is distinctly coldsand fresher than the Iceland-Scotland overflow (potertial temperature 0.0-2.0°C and salinity 34.88-34.93%(as contrasted with 1.8-3.0°C and 34.98-35.03%0o; Woithington, 1976), mainly because the upper-kilometeAtlantic water that they entrain is warmer and morsaline to the east.

Although the gross dimensions, course, and entrairment of the descending Denmark Strait overflow cabe fairly well rationalized in terms of a steady turbulenplume model (Smith, 1975), it is, in fact, a highly vaiiable flow: its water properties differ markedly frorone time to another (e.g., StefAnsson, 1968; Lachenbruch and Marshall, 1968; Mann, 1969), and monthlong near-bottom records of velocity and temperatur(not taken, it was thought, from the core of the currenlshowed that overflow was occurring in bursts 1 or '

60 70 80 90 100 110 120 130

C.S.S. Hudson cruise BI 0267, stations 3-13, 28-29 January1967. (Worthington, 1969.)

days long, with speeds up to 140 cm s - , separated byintervals of 1 to several days (Worthington, 1969). Var-iability in water characteristics suggests that the Ice-land-Scotland overflow is equally unsteady (Tait, 1967;Lee, 1967).

The process by which dense water is supplied tothese overflows has not been investigated in as muchdetail as the formation of Antarctic Bottom Water. TheNorwegian Sea appears to be an area of large mean-annual heat loss to the atmosphere, however, and Wor-thington (1970) worked out a heat budget for it inwhich he characterized it as a mediterranean basinwhere relatively light incoming water is made denseby strong local winter cooling, then sinks, and runs outbeneath the inflow.12 The overflow water is not deepNorwegian Sea water, which is colder and more densestill; tritium concentrations show that the overflowwater comes from depths no greater than 1000 m, andprobably from levels still shallower in the pycnocline(Peterson and Rooth, 1976). The outflow process thusseems to be separate from that of deep-water renewal;perhaps the overflow phenomenon is akin to the shal-low buoyancy-driven flow over a sill modeled by O. M.Phillips (1966a) for the Red Sea. As with the salinewater on the Antarctic continental shelf, the rotationalconstraint would tend to confine newly formed densewater to the Norwegian Sea, so that it would be ex-pected to leak out throughout the year(s), rather thanjust when cooling was most intense.

24Bruce A. Warren

The other location in the northern North Atlanticwhere deep sinking occurs is the Labrador Sea. Northof roughly lat. 40°N there is a distinct salinity mini-mum at a potential temperature of about 3.5°C (Wor-thington and Metcalf, 1961), which grows more pro-nounced with approach to the Labrador'Sea, wherenearly homogeneous water (temperature 3.5°C, sal-inity -34.9%o) is found in winter from the sea surfaceto about 1500 m (e.g., Worthington and Wright, 1970).The salinity minimum is associated with an oxygenmaximum, which was Wiist's (1935) diagnostic for hisMiddle North Atlantic Deep Water, and these charac-teristics must be imparted in the Labrador Sea, presum-ably through deep convection driven by intense wintercooling at the surface. Despite extensive wintertimesurveys in the area, however, no well-documenteddeep-overturning events have yet been reported, butLazier (1973a) has found a few examples in the recordsof Weather Ship Bravo (56°30'N, 51°00'W) of deep iso-pycnals rising to the surface layer for brief intervals oftime, in association with virtually homogeneous watercolumns 1500 m deep. Most likely, the convection oc-curs like that in the northwestern Mediterranean Sea(MEDOC Group, 1970), in small patches and very in-termittently (perhaps not even every winter), so thatthe convection could escape notice unless the obser-vations were frequent and closely spaced.13

Because of these small space and time scales, it isdifficult to make much more than a guess at the long-term mean rate of deep sinking in the Labrador Sea.From an estimate of the net annual heat flux acrossthe sea surface, Wright (1972) suggested 3.5 x106 m3 S- 1 .

Both the Labrador Sea water and the overflow waterare more saline at their sources than the AntarcticBottom Water, and their salinities are increased some-what in middle latitudes of the North Atlantic throughmixing with the outflow from the Mediterranean Sea(see below, this section). Although they are both madedense by cooling, their temperatures are not reduced tothe freezing point before sinking, but remain highenough that, despite the salinity difference, they areless dense in situ than the Antarctic Bottom Water,and therefore override it.14 This high-salinity waterspreads throughout the South Atlantic, is carried east-ward around Antarctica in the Circumpolar Current,and its elevated salinities can be followed into lowlatitudes of the South Indian and South Pacific Oceans(Reid and Lynn, 1971). The effect of North AtlanticDeep Water in maintaining high oxygen concentrationsin the deep ocean is familiar, but its source waters arealso exceptionally low in dissolved silica: <8 M 1- inthe Norwegian Sea overflow (Stefansson, 1968),roughly 10 pl.M1-' in the Labrador Sea water (Mann,Coote, and Garner, 1973), and about 6 M1-1 in theMediterranean outflow (Schink, 1967). Consequently,

the deep North Atlantic has the lowest silica concen-trations of all the oceans (Metcalf, 1969), there is apronounced deep silica minimum in the South Atlanticbetween the Antarctic Intermediate and BottomWaters (Mann, Coote, and Garner, 1973), and there iseven a weak silica minimum in the deep southwesternPacific due to the influence of North Atlantic sourcewaters (Warren, 1973).

1.4.3 Outflows from Marginal SeasExcess evaporation makes water in the MediterraneanSea, the Red Sea, and the Persian Gulf more saline andthereby more dense than that in the outlying ocean,and circulations are set up in the shallow connectingstraits in which near-surface water flows in, and dense,saline water flows out beneath. The outflows do notfall to the floor of the open ocean, however, because indescending the continental slopes they entrain enoughlight thermocline water to become less dense than thedeep water of the open ocean, and they therefore spreadout at mid-depths. They are thus not important sourcesfor the deep circulation, but through vertical mixingthey do at least influence the properties of the deepwater; hence they are noted briefly here.

The salinity maximum marking the core of the Med-iterranean outflow [Wiist's (1935) Upper North Atlan-tic Deep Water] is found at a depth of about 1200 m inthe southeastern North Atlantic (e.g., Fuglister, 1960),but relatively high salinities resulting from the outflowcan be detected at least as deep as 3000 m (Worthingtonand Wright, 1970), and the effect of the outflow inraising the salinity of the Labrador Sea water and Nor-wegian Sea overflow water was remarked above. Therate of outflow through the Straits of Gibraltar hasbeen estimated from the overall salt budget for theMediterranean Sea and by current measurements ofduration a few days to 2 weeks; both methods agree ona mean value of about 1 x 106 m 3s- ' (Lacombe, 1971).Calculations by Smith (1975) show that by the timethe outflow current completes its descent and departsfrom the continental slope, its volume transport hasincreased, by entrainment of Atlantic water, to roughly10 x 106 m3 s- . To what extent any of this transportcontributes to the flow of the deep water below-asdistinct from its properties-is not known.

The salinity maximum associated with the Red Seaoutflow occurs at depths of 500-600 m in the northernArabian Sea, but relatively high salinities, due presum-ably to mixing with the outflow water, are found thereat least as deep as 2500 m (Wyrtki, 1971). A salt-budgetcalculation based on climatological data cited by Sied-ler (1968) gives a mean outflow through Bab-el-Mandebof 0.4 x 106 m3 s-h, which is the same value that hederived from direct current measurements there of 2.5

25Deep Circulation of the World Ocean

days duration. Such close agreement was probably for-tuitous, however, because his directly measured inflowvalue was 40% greater. The entrainment during thedescent into the Gulf of Aden has not been estimated.

The Persian Gulf outflow (salinity maximum atdepths of 300-400 m in the Gulf of Oman; Wyrtki,1971) has a small, patchy effect on intermediate-depthproperty fields in the North Indian Ocean (Rochford,1964), but any influence on the deeper distributionsseems to be masked by that of the Red Sea outflow(Wyrtki, 1971). It is probably only slight, however, be-cause the outflow through the Strait of Hormuz is verysmall: only 0.1 x 106 m 3 s- l according to the salt budget(Koske, 1972).

1.5 Deep Western Boundary Currents in the WorldOcean

Many more hydrographic stations have been occupiedsince Sverdrup wrote chapter XV of The Oceans. Theyhave filled in blank areas in the world-ocean coverage,added details to property distributions, and sharpenedthe definition of features, but our broad picture of theproperty fields has not changed qualitatively. What haschanged greatly is the conception of their basis: thatthe underlying mean deep circulation, largely maskedby lateral mixing of the water characteristics, is a sys-tem of western boundary currents and poleward inte-rior flows, as in the dynamically consistent models ofStommel and Arons (1960a,b)-rather than the Merziansystem of basin-wide, slow meridional flows fromsource regions.

Although this conception makes good sense, it is notso easy to tell from observation how correct it is. Thekey element in the circulation theory is the supposedgeneral upward movement of deep water, but the hy-pothesized mean rate (-10-5 cms -') is so small as tobe utterly beyond present measurement capability.Probably measurement of the deep interior flow, to seewhether it is directed poleward in vertical average, istechnically feasible, but certainly not economicallyand socially so, given the immensity of the current-meter program that would be required. Large-scaletracer fields are not of much help in this respect sincemixing effects can make them ambiguous with regardto circulation patterns for the small mean speeds(-10 -2 cm s- ') of the interior flow (section 1.3; Kuo andVeronis, 1973).

What can be observed, through current measure-ments and short lines of hydrographic stations, are thenarrow, relatively swift, deep western boundary cur-rents predicted by the circulation theory. To be sure,some other theory, with different forcing, might alsorequire such boundary currents, but they are a neces-sary element of the Stommel-Arons dynamics, and

demonstrations of their existence in much of the worldocean have contributed strong support-indeed, theprimary observational support-for its essential ideas.This section will therefore assess the evidence for deepwestern boundary currents in the world ocean, withcomment on a few related features of the circulation.

1.5.1 Atlantic OceanThere are two deep boundary currents in the westernAtlantic, the one flowing southward from the Norwe-gian and Labrador Seas, and the other flowing north-ward from the Antarctic [the latter not included in thebarotropic model of Stommel and Arons (1960b) northof the counterpart of the Antarctic Peninsula].

After the Norwegian Sea overflow current passes thesouthern tip of Greenland, it takes a counterclockwisecourse around the Labrador Sea (figure 1.7) and flowssoutheastward along the North American continentalslope by Labrador and Newfoundland (Swallow andWorthington, 1969). Figure 1.9, for example, is a tem-perature section running from southern Labradornortheast to the southern tip of Greenland. The coldoverflow water may be seen pressed up against thelower Greenland slope, and the associated density gra-dient (mainly determined by temperature there) is con-sistent geostrophically with northwestward flow in-creasing in strength from mid-depths to the bottom.Adjacent to the North American continental slope, thesoutheastward-flowing deep boundary current is evi-dent in similar fashion, 200-300 km wide, probablyincluding Labrador Sea water above the overflow water.Three neutrally buoyant floats tracked nearby at depthsof 1600-2400 m all registered southeastward flow and,in combination with station data, suggested extensionof such flow up to about 1200 m. Though the meas-urements were of insufficient duration (12-31 hours)for very satisfactory results, nevertheless the indicatedvolume transport for the deep boundary current wasroughly 10 x 106 m 3 s- ' (Swallow and Worthington,1969).

High-oxygen and low-silica values close to the con-tinental slope in a similar depth interval show that thiscurrent continues around the Grand Banks of New-foundland, at least in the long-term mean, but recordsfrom moored current meters in the area have not yetgiven clear, direct evidence for such flow (Clarke andReiniger, 1973; Clarke, Hill, Reiniger, and Warren,1980).

Westward and southward in the North Atlantic, pro-nounced tracer characteristics marking the deepboundary current are lost, and its cross-stream densitygradient is not readily distinguishable from that of thenearby Gulf Stream; hence evidence for its existencecomes mainly from velocity measurements. On thecontinental slope north of the Gulf Stream near long.

26Bruce A. Warren