JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 93, NO. C7, PAGES 8111-8118, JULY 15, 1988

SeasonalChanges in the North Atlantic Subtropical Gyre

LOTHAR STRAMMA AND GEROLD SIEDLER

Institutfiir Meereskunde,Universitdt Kiel, Kiel, Federal Republicof Germany

The easternpart of the North Atlantic subtropicalgyre is foundin the regionbetween the Azoresand the Cape VerdeIslands. A studyof the gyrestructure in the area eastof 35øWbetween 8øN and 41øN is presented.The geostrophicflow field determinedfrom historicaltemperature-salinity data setsby objec- tive analysisindicates seasonal variations in shapebut no significantchanges in the magnitudeof volume transports.The easternpart of the gyre has a larger east-westand smallernorth-south extension in summercompared with the winter season.The centershifts by about 2ø latitude to the southfrom winter to summer.Long-term temperature time series(6.5 years)from a mooring near the Azoresare consistent with theseresults, showing always a consistenttemperature increase at the beginningof the year which is apparentlydue to the displacementof the northeasternpart of the gyre.A comparisonbetween the mean flow fields and fields obtained from individual zonal sectionsindicates large deviations north and south of the gyre but small deviations within the gyre.

1. INTRODUCTION strong from July through December,with a related shift of the The subtropical gyre in the North Atlantic includes a recir- to the north. Also, indications for culation regime with large transports toward the equator in seasonal variations in the northern part of the gyre, namely the western basin ['Worthington, 1976] and also smaller but the and the related subtropical front, were significanttransports in the easternbasin ['Strarnrna,1984a]. It found by Siedler et al. [1985]. Furthermore, Strarnrnaand is this eastern part in the Canary and Cape Verde basins Iserner [1988] noted a southward shift of the Azores Current which will be consideredhere (Figure 1). The large-scalemean in summer in their investigation of meridional temperature structureof the gyre flow as determinedin densityfield studies fluxes. In the present study we will use temperature-salinity by Wunsch and Grant [1982] using inverse modeling, by data obtained from ships as well as data from moorings to Olbers et al. [1985] employing the fi spiral method, and in determine seasonalchanges in the subtropical gyre in the east- drifter observation studies by Krauss [1986]. The data base ern Atlantic and to look for deviations from the annual mean. for the eastern part of the subtropicalgyre has considerable 2. DATA gapscompared with that for the westernpart. Only for certain areasdoes a high data densityexist. These include the African An edited data set was prepared from historical hydro- coastalupwelling regions[Mittelstaedt, 1983], the area of the graphic data of the eastern North Atlantic archived at the fi triangle near 27øN, 33øW [Armi and Stommel,1983], the World Oceanographic Data Center A (WODC, status 1980) area of the "Kiel Warmwassersph•re Programm" between the for the computation of temperature (T) and salinity (S) curves Azores and the Canary Islands, the area along 30øW of the in the eastern subtropical Atlantic I-Siedler and Strarnrna, Alexander yon Humboldt stations ['Schernaindaet al., 1976], 1983]. Stations with dubious data apparent from spikes or and the area of the USSR POLYGON project near 17øN, systematicdeviations were rejected.This data set was supple- 33øW [Brekhovskikh et al., 1971]. Owing to the use of mented by conductivity-temperature-depth(CTD) profiles of a smoothedclimatological data setsin the above mentioned in- R/V Meteor cruise with two long north-south sectionsalong vestigations,small-scale temporal and spatial features cannot 27øW and 33øW in March and April 1982 I-Strarnrna,1984b], be resolved. However, when only the strongly variable upper CTD data from a R/V Meteor cruise along about 27øW in ocean down to the lower main thermocline is being con- April 1981 [Thiele et al., 1986], and CTDs from two of the fl sidered,data quality requirementsare less stringent than in triangle cruises[Arrni and Stornrnel,1983] made on R/V At- studiesfor the deep ocean.In the case of a strongly variable lantis in October 1979 and R/V Atlantis in July and August upper ocean,a broader data basewith lesssmoothing can be 1981. Subsetsin 3ø x 3ø squares were used to compute mean used. The time-averaged southward transports east of 35øW profiles for the periods January-March (winter), April-June for the upper 1000 m were determinedwith high spatial reso- (spring), July-September (summer), and October-December lution by Strarnrna[1984a] from the international data set. He (fall). From all the singleprofiles, mean profiles were obtained obtaineda net transportof 11 x 106 m3 s-x, whichis in good for the mean position within each 3ø x 3ø square. The sets of agreementwith resultsfrom individual zonal sectionspresent- four mean profiles for spring, summer, fall, and winter were ed by Saunders[1982] and Gould[1985]. first used to compute the mean profile for the entire year. In We will attempt here to study the temporal variation of the this way, the seasonswere equally weighted. In the case of gyre with the aim of verifying the existenceof seasonalvari- missing data for one season for a 3ø x 3ø square, only data ations in the gyre structure. Seasonalchanges in the southern from the other seasons were used. part of the gyre have been known for a long time, and a recent We prefer not to excludeprofiles taken within any particu- analysis of ship drift data by Richardsonand Walsh [1986] lar square during a single cruise because it is important to showed the North Atlantic Equatorial Countercurrent to be have as much spatial information as possible within each square when estimating the mean profiles. However, the use of Copyright1988 by the AmericanGeophysical Union. all data could lead to a bias in the interannual variability Paper number 8C0254. toward one year. We inspected the entire data set (Figure 2) 0148-0227/88/008C-0254505.00 and found that in most squaresthe information is spread over

8111 8112 STRAMI•IA AND SIEDLER' NOi•T• ATLANTIC SEASONAL CHANGES

j i [1984a]. He selecteda level of no motion using mean density | (G' ,oø'••,-eAzoresdo7'4o7 profiles, combining information on the advection of water N • massesidentified by oxygen and salinity maxima and minima, and a mass conservation scheme developed by Fiadeiro and 1- 2000,,.•,q Veronis [1982, 1983]. The resulting zero reference plane is found between 1500-m depth in the tropical eastern North 4000 Atlantic and 1200-m depth in the subtropics[Stramma, 1984a,

Canary Is. Figure 6]. The referencedepth was verified by Miiller [1987] at the mooring location Kiel 276 (see Figure 4a) from direct current measurementsover 5.5 years. Vertical shear is small at 30eI • 4000m these levels below the lower boundary of the main thermocline and much larger in the upper ocean. As a result, deviations from the levels given above will not result in large changesin .:.Z.Z.Z':':':';':'Z':':':'Z':';" the transport in the upper ocean. If, for instance, the above 20' level of no motion is replacedby a level of 2000 m, the change Verde Is. in the 0 to 1000 m transport across a zonal section in the area AFRICA of this study is less than 12%. On the other hand, the total transport changes are less than 5% when noise-simulating :':':':':':':i!i!ii!!iii!i:!:...... alterations are made to the mean temperature and salinity profiles.The total error for the 0- to 1000-m transport, caused by the uncertainty in the referencelevel depth and data noise, 10" is 15%. Since it is the aim of this study to determine seasonal changes in the transport field, the question might be asked 3O ø W 20' 10 ø whether a seasonal change could occur in the level of no Fig. 1. Area of this study with selecteddepth contours. motion. The Fiadeiro and Veronis [1982, 1983] method of computing transport imbalances was used by Stramma [1984a] for triangular columns given by the mean profiles in several years, the years being different from season to season, neighboring 3øx 3ø squares. There was only one triangle and that in adjacent squares the data are from differing years where the profiles reached the seafloor for all four seasons.In also. This much reduces the likelihood of a large bias appear- that case, the results for the annual mean and for the four ing in the interannual variability. For example, in the area 32ø seasons were almost identical. This suggests that temporal to 35øN, 29 ø to 32øW, which is important for the shift of the changesin the level of no motion are negligibly small for the Azores Current, we included profiles from the years 1964, present analysis. 1971, and 1971 for winter; 1965, 1969, 1969, 1971, and 1971 for spring; 1960, 1967, 1971, and 1981 for summer; and 1947, 1957, and 1979 for fall. 12 In addition to the hydrographic data set, data from selected 12 191 ;11 ' moorings were used to study the seasonal variability. Mooring 10 4 5 • 11 9 15- 11 1 11 2 27_ 20 2 11 1 15 3 21 ? • 11 117 68 Kiel 276 with a nominal position at 33øN, 22øW, at the earlier 5 I j • S 9 2 4 1 5 2 7 3 9 6 21 10 • 1 3 5 2 4 1 10 1 S 2 9 9 23 6 12 9 North-East Atlantic Dynamic Studies (NEADS) site 1 [Dick- 1 2 3 2 2 1 4 1 4 1 5 5 8 3 7 5 son et al., 1985], has been in continuous operation since April 1 21 4 43 1115 1 23 23 45 23 156 35c ' 177 8 4 •5 • '2711 1980. The records being considered here cover eight consecu- tive periods, with the mooring having been replaced seven 3 2 3 4 2 1 2 3 2 4 4 3 7 4 , .. times. We used the temperature time series of one Aanderaa 2 1 2 3 4 1 3 1 a 7• • current meter per mooring period in the upper ocean. The 5 4 5 6 1 2 3 periods of the mooring records and the nominal depths of the .:..?'.'.'..' temperature sensorsare given in Table 1. In addition to this temperature record of 6.5-year duration, the temperatures of • 3 • • the thermistor chain mooring Kiel 297 at 28øN, 26ø30'W, from October 24, 1983, to November 1, 1984, at depths between 300 m and 700 m are used. Finally, current measurementsat 165 20"a • m depth of mooring Kiel 303 in the at 26øN, 18øW, from November 12, 1984, to October 7, 1985, are in- 4 3 c 14 cluded. Details of the mooring data can be found in data 912•:;- reports [Miiller, 1981, 1984; Miiller and Zenk, 1983; Miiller et 48 23 73 68 29 43 1425 2 • • •126 I'•F.' al., 1987]. 1 1 3 3 2 2 • 1 • . oo. ; 5 7 3 5. 3 3 5'•' ' 3. GEOSTROPHIC TRANSPORTS ;_ •_ ; • 142 .., ß 30" W •" 10" 3.1. Transport Computations Fig. 2. Number of stations used in each 3ø x 3ø square for each The method used here for determining the velocity and season (top number) and number of different years represented transport field in the upper ocean from the density distri- (bottom number) for winter (top left area), spring (top right area), bution is the same as in an earlier approach by Stramma summer (bottom left area) and fall (bottom right area). STRAMMA AND SIEDLER' NORTH ATLANTIC SEASONAL CHANGES 8113

3.2. Seasonal Variability rent between20øW and the Portuguesecoast is 2 x 106 m 3 The geostrophic transport integrated from the surface to s- x, comparedwith 4 x 106 m 3 s- x for the AzoresCurrent. 200-m depth is shown in Figures 3a-3d for the four seasons A seasonal cycle is also evident in the position of the re- and in Figure 4a for the annual mean. The contours were versal in east-west direction at the western boundary of the determined by using an objective analysis scheme. Since the area of investigation. We call it the "center," although the real correlation scale was not known, a value of 500 km was center of the subtropical gyre will be located to the west of the chosen arbitrarily. It is about 1.5 times the typical distance area studied here. In winter the center is located at 29øN and between the mean profiles and is larger than the typical in spring at 30øN. The center shifts south to 27øN in summer scale. The error variance was chosen as 30% of the total and back northward to 29øN in fall. Related to this shift of the variance. Tests with different correlation scales and error vari- center in summer is a southward displacement of the northern ances were made. The resultant flow fields show some differ- core of the gyre, the Azores Current (Figure 3c). South of the ences in the smoothness of the flow contours and the width of Azores the core of the Azores Current shifts between 35øN in the current bands, but the information on seasonal variations winter and 33øN in summer. Synoptic observations of the does not change. Barred areas in Figures 3 and 4 indicate Azores Current have also shown a north-south shift [Kiise et error valueslarger than 0.6 x 106m 3 s-•. Areasof large error al., 1986]. In this casethe Azores Current at 22øW was found values reflect some 3øx 3ø squares with no profile in this in October 1984 to be 3ø to the south of the spring 1982 season. This occurs mostly at the western and the southern location. The data from cruises of R/V Meteor and R/V boundary and in the central region between the Canary Is- Oceanusin 1984 and R/V Poseidonin 1982 used by Kiise et al. lands and the Cape Verde Islands. [1986] are not included in the mean profiles. Only the R/V The upper ocean is expected to display the largest seasonal Meteor data of spring 1982 are included. The synoptic obser- signal. Since the errors in transport computations increase vations of Kiise et al. [1986], therefore, provide independent with depth and may finally mask the seasonal signal, the evidence of the southward shift of the Azores Current in Oc- analysis of seasonal transport changes was restricted to the tober. upper 200 m. But we also present the annual mean field for We also combined some historical ship observations near the upper 800 m. We searched for spatial displacements of 30øW from different vesselsand years but from the same gyre-related structures which have scales comparable to the season, which also demonstrated a southward shift of the correlation scale normal to the transport direction and ex- Azores Current in summer. In reality, the Azores Current is a ceeding the correlation scale by a factor of 2 to 3 in the deep, jetlike current system [Kiise et al., 1985, Sledlet et al., transport direction. 1985] with horizontalscales much smaller then thoseshown A seasonal variation of the eastern part of the subtropical in Figure 3. The amount of geostrophic transport derived gyre is recognized. To ensure that the seasonal cycle in the from the mean profiles will not be influenced by the smaller gyre is not due to effects of data outside of the subtropical scales,but the position of the main horizontal gradients is less gyre region, we repeated the computations with data between well determined. In summer the Azores Current is oriented 17øN and 38øN only, the area where we see the subtropical west-east' in the other seasons the Azores Current shows a gyre. The resulting seasonality of the subtropical gyre was the larger southward component. This west-east-oriented strong same as for the area 8øN to 41øN, indicating that the infor- narrow band of the Azores Current is seen much more clearly mation on the seasonal cycle results from the data within the in summer in the 200- to 800-m depth contours (not presented smaller area. In all seasonsthe 0- to 200-m transport isolines here). display a separation of the Azores Current into two south- Seasonalchanges in the southern part of the gyre have been ward flowing current bands near 30øN, south of Madeira observed before. Richardson and Walsh [1986] computed (Figure 3). The boundaries of this separation area are indicat- near-surface currents from ship drift data south of 20øN and ed by heavy lines which facilitate tracing the displacementsof found the North Equatorial Countercurrent to be strong in the flow field in the seasonalcycle. the North Atlantic from July to December. The counter- The gyre transportamounts to approximately4 x 106 m3 current transports equatorial water to the tropical eastern At- s- • east of 35øW between 0-m and 200-m depth. The main lantic, and related to its temporal changes is a shift of the inflow enters south of the Azores as a relatively narrow cur- North Equatorial Current to the north. In the tropical eastern rent, the Azores Current. In all seasonsthe Portugal Current North Atlantic, hydrographic data are sparse. Therefore the is weak compared with the Azores Current. The Portugal Cur- transports shown in Figure 3 for the tropics must be regarded rent is directed toward the entrance of the Mediterranean Sea. with some caution. In summer and fall (Figures 3c and 3d), The annual mean transport (Figure 4a) of the Portugal Cur- equatorial water is transported to the north and contributes to the North Equatorial Current. The salty water from the subtropicscarried in the subtropi- TABLE 1. Periods of Mooring Records and Nominal Depths of cal gyre is shifted to the north in the North Equatorial Cur- Temperature Sensors rent. The northward shift can be seenin Figure 5 in the salini- Period Depth, m ty isohaline35.9 on the densitysurface y• = 26.8 kg m -3 (y = p --10 3, p is density, and 0 is potential temperature). April l, 1980, to Oct. 17, 1980 379 This density surface is 200 to 300 m deep in that area. The Oct. 17, 1980, to July 27, 1981 196 contours are taken from seasonalsalinity distributions on this July 27, 1981, to March 2, 1982 245 March 5, 1982, to April 17, 1983 195 density surface derived with the objective analysis scheme.In April 22, 1983, to Oct. 15, 1983 245 summer the 35.9 isohaline is north of its winter and spring Oct. 20, 1983, to Oct. 25, 1984 330 location. This isohaline has its northernmost location in the 330 Oct. 26, 1984, to Nov. 16, 1985 fall near 25øN. Less salty tropical water lies to the south of the Nov. 17, 1985, to Nov. 1, 1986 302 35.9 isohaline. The northeast-to-southwest slope in salinity is 8114 STRAMMA AND SIEDLER' NORTH ATLANTIC SEASONALCHANGES

40 ø

N

30" 30'

20' 20"

10' 10"

30 ø W 20' 10ø 30 ø W 20' 10ø

40' 40'

30'

20' 20'

10' 10'

3(7' W 20" 10ø 3(7' W 20" 10ø

Fig. 3. Integrated volume transport (0 to 200 m) from mean densityprofiles for (a) January through March (winter), (b) April through June(spring), (c) July throughSeptember (summer), and (d) October throughDecember (fall). Contour intervalsrepresent 0.5 x 106m 3 s-x. The error field largerthan 0.6 x 106m 3 s- x is barred.Heavy lines represent limits of the separationarea betweenthe two transportbands extending from the AzoresCurrent and are given as indicatorsfor gyre displacments.

the well-known transition between the North and the South large error fields computedby objectiveanalysis. However, Atlantic Central Water [Tomczak, 1984; Emery and Meincke, mooringKiel 303, deployedfor 1 year in this area, also shows 1986]. A ship drift data analysis for the eastern a northward flow in the fall of 1985, as will be discussedin (G. Jeckstr6m, unpublished manuscript, 1983) also indicates a section 4. Ship drift data from the eastern Atlantic (G. northward shift of the North Equatorial Current in the period Jeckstr/Sm,unpublished manuscript, 1983) indicate the small- July to September. est southwestwardship drift south of the Canary Islands for From April to September the southward transport near the the period October through December.Geostrophic compu- African coast is larger than that from October to March, when tations,ship drift data, and resultsfrom a mooringall indicate the subtropical gyre is stronger west of ,theCanary Islands. In a weaker Canary Current south of the Canary Islands in fall. the fall (Fig. 3d) there is even a northward componentsouth of Near the Canary Islands, some transport lines are directed the Canary Islands, although only in the barred area, with toward the African coast and reappear north of 20øN. This is STRAMMAAND SIEDLER:NORTH ATLANTICSEASONAL CHANGES 8115

40'

30'

20' 20'

10' 10'

30' W 20' 10' 30' W 20' 10' Fig.4. Integratedvolume transport asin Figure3 butfor (a) the yearly mean of 0 to 200m and(b) the yearly mean of 200 to 800 m. Dots and numbersindicate the Kiel mooring locations.

suggestiveof a Canary Current flowing in a narrow current [1983] in the /• triangle data. In the 200- to 800-m layer, band near the shelf. therefore,the subtropicalgyre doesnot extendas far southas In winter and spring the currents are weak southwestof it doesin the top 200 m. The seasonalcycle of the center of Madeira. Siedler et al. [1987] observedmode water formation the gyre, as well as the southward shift of the Azores Current, in this area. This Madeira Mode Water is transported to the is apparent in the seasonalflow contours of the 200- to 800-m west and south during the course of the year. The low trans- depth. Since there is no new information on the seasonal ports, observedin winter and spring near Madeira, will pro- changesin the resultsfor the 200- to 800-m layer and the error mote the formation of a homogeneouswater layer due to fields increase considerably, the figures for this layer are not thermal convection in this area, and the stronger transport presentedhere. fields in summer and fall will support the advection and In summary, we observe a part of the subtropical gyre in mixing of the mode water. The potentialvorticity distribution the eastern North Atlantic with larger east-westextension and on ,the R/V Meteor section east of, or along, 27øW in April smaller north-south extension in summer than in winter. In 1981 (not presentedhere) showsa potential vorticity minimum the North Equatorial Current the errors of the geostrophic at Y0values of about26.5 kg m-3 between25øN and 32øN, flow fields are too large to prove the north-south shift, but it is indicatingthe presenceof the Madeira Mode Water. seenin the salinity distribution (Figure 5). A schematicpresen- The 200- to 800-m transport contours of the annual mean tation of the shape of the subtropical gyre in summer .and are presented in Figure 4b. In comparison with the 0- to winter in the top 200 m of the eastern North Atlantic is given 200-m contours (Figure 4a), it is obvious that the axis of the in Figure 6. The center is also indicated in this figure. The gyre is inclined to the north with increasingdepth. This north- magnitudeof the transport in the upper 200 m of the ocean ward deflection was already observed by Arrni and Stornrnel doesnot changesignificantly and is approximately4 x 106 m3 s- x in all seasons,but the structure of the gyre changessea- sonally. I ' I ' I

3.3. Deviations From the Annual Mean 30 ø The historical oceanographic data base is too small to in- •' / ....-.v.v.v.v...-. N vestigate interannual variability from mean profiles. We can, however, use some single sectionsto investigate deviations from the seasonalmean. In Table 2 we present a comparison between the geostrophic transport from International Geo- physical Year (IGY) sectionseast of 35øW and the seasonal 20 ø • "i:i:!:i:i:i:i:i:i:i:i:!:i:i:i:i:i:i:i:i:i:i:i:i:i:!:!:i:i:i:mean transports. The Crawford section in October 1959 at 40ø15'N showsa southwardtransport of 0.5 x 106 m3 s-1 east of 35øW including the coastal area of the Portugal Cur- 30ø W 20ø 10• rent, lessthan the 1.4 x 106 m 3 s- 1 for the mean fall value in the upper 200 m. The higher transport indicated in Figure 4a Fig. 5. Geographicallocation of the isohaline35.9 on the surface of potentialdensity Y0 = 26.8 kg m-3 (about 200 to 300 m deep) is introduced by the objective analysis scheme, close to the shownfor the four seasonsas definedin the legend of Figure 3. strong Azores Current at the northern boundary of the select- 8116 STRAMMA AND SIEDLER: NOP, TI-I ATLANTIC SEASONAL CHANGES

Azores and in the tropics at 16ø15'N, variability is large. It is unclear at this point whether the transport deviations in indi- vidual sectionsare typical and suggestiveof interannual vari- ations of similar magnitude. At least there appears to be a consistentspatial pattern. Another indicationin this direction is given by the results of Roemmichand Wunsch [1985]. After a repetition of the sections at 24ø30'N and 36ø15'N in mid- 1981, they found that the repeated sections have similar fea- tures in their large-scalevelocity and similar zonally averaged meridional transports, supporting our findings of low varia- bility in the subtropical gyre.

4. DIRECT MEASUREMENTS The mooring Kiel 276 with the nominal position at 33øN, 22øW, is located in the eastern extension of the Azores Cur- rent. Since this current was found in historical hydrographic

10' data to shift with the seasonalcycle, we searchedfor a possible seasonal cycle in the mooring data time series. The long 3O ø W 20' record of 6.5 years is advantageous,but the mooring's location in the eastern extension of the Azores Current is unfortunate Fig. 6. Schematicpresentation of the subtropical gyre in summer and winter in the top 200 m of the ocean. The reversal of the flow becausethe total transport is weaker in this region compared direction on the western boundary is shown for summer (CS) and with the transport south of the Azores. The currents in the winter (CW). upper ocean are governed by strong-current events on the scaleof 1-3 months[Siedler et al., 1985]which mask the small signal of the seasonal cycle. Using a 180-day mean every 90 ed area. The Chain section along 36ø15'N in May 1959 pre- days, Miiller [1987] found a northward velocity component in sents a southward transport of 3.5 x 106 m 3 s-•, and our the winters and a southward component in the summers of springmean transportat this latitude is 2.7 x 106 m3 s- • to 1980 and 1981. However, the overall results on seasonal cur- the south. The difference might be variability in the current rent variations are inconclusive. strength or the influence of the shift of the Azores Current. Temperature records may be better suited for a study of The three sections at 32ø15'N, 28ø15'N, and 24ø30'N give variations caused by seasonal changes in the position of the transport values which are within 0.5 x 106 ma s-• of the water mass boundary. A displacement of the boundary in the seasonal mean for the respective season. This indicates a sub- direction of the horizontal temperature gradient will generate tropical gyre with stable transport valuesand weak variability. a strong temperature signal at a fixed position in the bound- The Crawford section in November 1957 at 16ø15'N shows a ary region but may well cause only small variations in the northward transport of 1.8 x 106 m a s-•. The data base for baroclinic currents which are predominantly normal to the the seasonalmean at this latitude is very limited and the mean horizontal temperature gradient and constant in the case of a failtransport of 1.0x 106m a s- • to thenorth might be influ- constant density gradient. As a result, the signal-to-noiseratio enced considerably by the Crawford section. The data from may be better in the caseof seasonaltemperature variations. the IGY sections are included in the mean profiles, and the We therefore attempted to obtain a temperature time series seasonal mean and IGY transports are therefore not indepen- at a fixed depth, using appropriate vertical interpolation and dent of each other. On the other hand, the data base in the smoothing in time. For this purpose the annual mean temper- subtropical area is still large enough to distinguish transport ature profile from historical data for the 3ø • 3ø square in- deviations of single sectionsfrom the mean transports. cluding the mooring site was approximated by a second-order in thesubtropical gyre, the IGY sectionsindicate only small polynomial in the upper 500 m. The depths of the temperature deviations from the annual mean, while at the latitude of the sensorson the mooring were first corrected for vertical dis- placementsdue to mooring inclination by current drag, using TABLE 2. Meridional Geostrophic Transports in Sverdrups pressure sensor records. The temperatures at 200-m depth From IGY Sections and From the Related Seasonal Means for were then obtained by applying the above polynomial for the 0 to 200 m East of 35øW interpolation in the vertical. No corrections were made for the different seasonaltemperature amplitudes at different depths. Individual Seasonal Section Mean They are difficult to estimate. For this reason the amplitude at Ship Latitude Transport Transport the end of the record is lower becausethe actual temperature sensor-was at a greater depth during this time. During the first Crawford (October 1959) 40ø15'N -0.5 -1.4 period of April 1, 1980, to October 17, 1980, the pressure Chain (May 1959) 36ø15'N -3.5 -2.7 sensor at 379-m depth had failed, and the pressurechanges at Discovery (December 32ø15'N -3.8 -4.0 1957) 673 m had to be used. This sensor showed an amplitude simi- Atlantis (February 1959) 28ø15'N -4.4 -4.2 lar to that of another pressure sensor at a depth of 24 m on Discovery (October 24ø30'N -4.2 -3.7 the same mooring. During the last period, from November 1957) 1985 to November 1986, the only pressuresensor in the moor- Crawford (November 16ø15'N + 1.8 + 1.0 1957) ing failed. From the current measurementsat different depths it is obvious in this case that no strong barotropic flow had Positivesign represents northward transport (1 Sv = 106 m3 s-l). reached the mooring during this period and no large incli- STRAMMA AND SIEDLER' NORTH ATLANTIC SEASONAL CHANGES 8117

T/"C

14 i J i i i i i i i i i i i • i i i i i I i • i i i ii i I i i i .... , , i i i , i i i , i i i i i i i i 1980 • 1981 •= 1982 •- 1983• ,-•1984 -•J•'-1985 ----•- .---- 1986 Fig. 7. Temperaturetime seriesat 33øN, 22øW. The curve was interpolatedto 200-m depth and low-passfiltered over 179days. The temperatureincrease at the turn of the yearis indicatedby the heavylines. nation of the mooring due to current drag was probable. (seeFigure 4), are presentedin Figure 8. The temperature time Therefore the depth correctionwas only made for the differ- seriesare given between 300 and 700 m, in intervals of 40 m. encebetween the nominal depth (302 m) and 200 m. The curves were smoothed with a 10-day running mean. The Furthermore, the temperaturetime seriesat 200-m depth location of this mooring is close to the center of the subtropi- was linearly interpolated in time for the gaps between moor- cal gyre, and the temperature signal is expectedto be weak. ing recovery and launching and finally low-passfiltered over The temperature curvesare much smoother than at the moor- 179 days. With this filter we loose 6 months of data at the ing in the Azores Current, but the curves in Figure 8 also beginningand the end of the time series,but we also deletethe show some tendency to higher temperatures in January and short-term signalsof mesoscaleeddies or other events. February 1984. The resulting temperature time seriesis given in Figure 7. Finally, current measurementsfrom mooring Kiel 303 at The temperature at 200 m always increasesin January except 165 m depth on 26øN, 18øW (see Figure 4) from November for January 1984, when the maximum winter temperature was 1984 to October 1985 are shown in Figure 9. This mooring is reached earlier (in December 1983). For the period shown here located in the Canary Current south of the Canary Islands. there is always a temperature increasein winter which is ap- The seasonalgeostrophic computations showed a weakening parently related to the northward shift of the Azores Current of the Canary Current in winter and spring. This 1-year cur- in winter known from the geostrophic currents obtained from rent record shows a northward flow in November and in the historical data (Figure 3). This temperature increasein winter beginningof December 1984 and then again at the end of is in contrast to the normal seasonal temperature cycle near Septemberand the beginningof October 1985. The strongest the surface at northern latitudes. When the subtropical gyre in southward flow is observed in June and July of 1985. From the Azores Current region shifts to the north, salty and warm January to May the current is weak and variable, with north- water advances to the north. Some of the Aanderaa current westward flow at the end of April and the beginning of May meters were also equipped with conductivity cells.The result- 1985. For this 1-year period in Figure 9 the measured current ant salinity recordsdo indeed display saltier water in winter, velocity at 165-m depth is consistentwith the basic featuresof but there is no completerecord of salinity for the upper ocean seasonal variations found in the geostrophic computations over the 6.5-yearmooring period. The accuracyof the salinity based on historical data. measurementson the moorings is marginal, and therefore no figure is presentedhere. In 1982, 1984, and 1985 in late 5. CONCLUSION summer a secondtemperature maximum is seenat the moor- We can state that the seasonal variations in the structure of ing location. We can only speculatethat this may be caused the easternpart of the North Atlanticsubtropical gyre, as by a meandering of the Azores Current on shorter than determined from historical density data, is consistent with re- annual time scales. sults from long-term observations with moored temperature For a comparisonwith this time seriesthe temperaturedata sensors,with seasonal changes of the salinity at the central from the thermistor chain mooring Kiel 297 at 28øN, 26ø30'W water boundary off West Africa, and with ship drift data re- sults. It remains to be demonstrated by numerical modeling whether the observed changes are caused solely by the large 17

_ - seasonal wind stress variation [Hellerman and Rosenstein, T/•C 1983; Isemer and Hasse, 1987] in the northern and southern 16 part of the area investigatedhere. The data base is not suf- ficiently complete at this time to determine interannual vari- 15 ations in the structure of the subtropical gyre. --_ • ••••••x.r• 340m-- 14 _ "-"-•• 380m_ _ ,"'"--'",,,•• 420m _ 13 _ •••_••v--.-••-•" 460m_ _ ,,,,,--,,.,,,,.•• 500m _ oE 20 12 ••.•..•--•-,--• 540rn _ 10 •• 580m ••-__.•.•._,•-•• 620rn - 0 11 - 10 • 700 rn _ _•...•_.• 660m_ -20

10 -30 O N D J F M A M J J A S O -40 I I I I I I I I I I I I I I -- 1983 =•-• 1984 O N D J FM A MJ J A S O N --1984 •'• 1985 Fig. 8. Temperature time seriesat 28øN, 26ø30'W, between 300-m and 700-m depth from October 1983 to October 1984. The data are Fig. 9. Current vector time seriesat 26øN, 18øW, at 165-m depth smoothedwith a 10-day running mean. from November 1984 to October 1985. 8118 STRAMMA AND SIEDLER: NORTH ATLANTIC SEASONAL CHANGES

Acknowledgments. We would like to acknowledgethe assistanceof 1982--A data report, Ber. Inst. Meereskd. Christian Albrechts Univ. the data processingstaff of the Marine Physics Group at the Institut Kiel, 114, 145 pp., 1983. f/Jr Meereskunde, Kiel. We want to thank L. Armi for providing the Miiller, T. J., M. Finke, W. Dasch, and R.-R. Wittstock, Hydro- tapes of two of the beta triangle cruises,T. J. Miiller for assistancein graphic and current measurements in the north-east Atlantic the mooring data processing,and E. Tidmarsh for helpful comments Ocean,Ber. Inst. Meereskd.Christian Albrechts Univ. Kiel, 166, 106 on the text. This work was supportedby the DeutscheForschungsge- pp., 1987. meinschaftgrant SFB 133. Olbers, D. J., M. Wenzel, and J. Willebrand, The inferenceof North Atlantic circulation patterns from climatologicalhydrographic data, Rev. Geophys.,23, 313-356, 1985. REFERENCES Richardson,P. L., and D. Walsh, Mapping climatologicalseasonal Armi, L., and H. Stommel, Four views of a portion of the North variationsof surfacecurrents in the tropical Atlantic usingship Atlantic subtropicalgyre, d. Phys. Oceanogr.,15, 828-857, 1983. drifts, J. Geophys.Res., 91, 10,537-10,550, 1986. Brekhovskikh,L. M., K. N. Fedorov, L. M. Fomin, M. N. Koshlya- Roeromich, D., and C. Wunsch, Two transatlantic sections: Meridio- kov, and A.D. Yampolsky, Large scale multi-buoy experimentin nal circulation and heat flux in the subtropicalNorth Atlantic the tropical Atlantic, Deep Sea Res., 18, 1189-1206, 1971. Ocean, Deep Sea Res., 32, 619-664, 1985. Dickson, R. R., W. J. Gould, T. J. Miiller, and C. Maillard, Estimates Saunders,P.M., Circulation in the easternNorth Atlantic, J. Mar. of the mean circulation in the deep ( > 2000 m) layer of the eastern Res., 40, suppl., 641-657, 1982. North Atlantic, Prog. Oceanogr.,14, 103-127, 1985. Schemainda,R., W. Kaiser, D. Nehring, and S. Schulz, Ozeanolo- Emery, W. J., and J. Meincke, Global water masses:Summary and gischeUntersuchungen im tropischenNordatlantik auf 30øW zwis- review, Oceanol. Acta, 9, 383-391, 1986. chen2øN-15øN, Geod. Geophys. Vergefl., Reihe IV, Heft 15, 52 pp., Fiadeiro, M. E., and G. Veronis, On the determination of absolute 1976. velocitiesin the ocean,J. Mar. Res.,40, suppl.,159-182, 1982. Siedler,G., and L. Stramma,The applicabilityof the T/S methodto Fiadeiro, M. E., and G. Veronis, Circulation and heat flux in the geopotential anomaly computations in the northeast Atlantic, Bermuda triangle, J. Phys. Oceanogr.,13, 1158-1169, 1983. Oceanol.Acta, 6, 167-172, 1983. Gould, W. J., Physical oceanography of the Azores front, Prog. Siedler,G., W. Zenk, and W. J. Emery, Strongcurrent events related Oceanogr., 14, 167-190, 1985. to a subtropicalfront in the northeastAtlantic, J. Phys.Oceanogr., Hellerman, S., and M. Rosenstein,Normal monthly wind stressover 15, 885-897, 1985. the world ocean with error estimates,J. Phys. Oceanogr.,13, 1093- Siedler,G., A. Kuhl, and W. Zenk, The Madeira Mode Water, J. 1104, 1983. Phys. Oceanogr.,17, 1561-1570, 1987. Isemer, H.-J., and L. Hasse, The Bunker Climate Atlas of the North Stramma,L., Geostrophictransport in the warm water sphereof the Atlantic Ocean, vol. 2, Air-Sea Interactions, 252 pp., Springer- eastern subtropical North Atlantic, J. Mar. Res., 42, 537-558, Verlag, New York, 1987. 1984a. K/ise, R. H., W. Zenk, T. B. Sanford, and W. Hiller, Currents, fronts, Stramma,L., Potentialvorticity and volumetransport in the eastern and eddy fluxes in the Canary Basin, Prog. Oceanogr.14, 231-257, North Atlanticfrom two long CTD sections,Dtsch Hydrogr. Z., 37, 1985. 147-155, 1984b. K/ise, R. H., J. F. Price, P. L. Richardson, and W. Zenk, A quasi- Stramma, L., and H.-J. Isemer, Seasonalvariability of meridional synoptic survey of the thermocline circulation and water mass dis- temperaturefluxes in the eastern North Atlantic Ocean, J. Mar. tribution within the Canary Basin, J. Geophys.Res., 91, 9739-9748, Res., 46, 281-289, 1988. 1986. Thiele, G., W. Roether, P. Schlosser,R. Kuntz, G. Siedler,and L. Krauss, W., The , J. Geophys.Res., 91(C4), Stramma,Baroclinic flow and transient-tracerfields in the Canary- 5061-5074, 1986. Cape-Verde Basin,J. Phys. Oceanogr.,16, 814-826, 1986. Mittelstaedt,E., The upwellingarea off northwestAfrica--A descrip- Tomczak,M., Ausbreitungund Vermischungder Zentralwassermass- tion of phenomena related to coastal , Prog. Oceanogr., en in den Tropengebietender Ozeane, 1, AtlantischerOzean, Ocea- 12, 307-331, 1983. nol. Acta, 7, 145-158, 1984. Miiller, T. J., Current and temperature measurementsin the north- Worthington,L. V., On the North Atlantic circulation,Johns Hopkins east Atlantic during NEADS, Ber. Inst. Meereskund.,Christian Al- Stud. Oceanogr.6, 110 pp., 1976. brechtsUniv. Kiel, 90, 100 pp., 1981. Wunsch, C., and B. Grant, Towards the general circulation of the Miiller, T. J., Eulerian current measurements from the north east North Atlantic Ocean, Prog. Oceanogr.,11, 1-59, 1982. Atlantic, March 1982-October 1983--A data report, Ber. Inst. Meereskd.Christian Albrechts Univ. Kiel, 127, 99 pp., 1984. G. Siedler and L. Stramma, Institut fiir Meereskunde, Universit/it Miiller, T. J., Analyse niederfrequenterStr6mungsschwankungen im Kiel, DiisternbrookerWeg 20, D-2300 Kiel 1, Federal Republic of Nordostatlantik, Doctoral dissertation, 134 pp., Kiel Univ., Kiel, Germany. Federal Republic of Germany, 1987. Miiller, T. J., and W. Zenk, Some Eulerian current measurements and (Received December 18, 1987; XBT-sections from the north east Atlantic October 1980-March acceptedJanuary 26, 1988.)