Chlorophyll Variability and Eddies in the Brazil–Malvinas Confluence

ARTICLE IN PRESS

Deep-Sea Research II 51 (2004) 159–172

Chlorophyll variability and eddies in the Brazil–Malvinas Conﬂuence region

Carlos A.E. Garciaa,*, Y.V.B. Sarmaa, Mauricio M. Mataa, Virginia M.T. Garciab a Department of Physics, Funda@ao* Universidade Federal do Rio Grande, Av. Italia,! km 8, Rio Grande RS 96201-900, Brazil b Department of Oceanography, Funda@ao* Universidade Federal do Rio Grande, Av. Italia,! km 8, Rio Grande RS 96201-900 Brazil

Received27 January 2003; receivedin revisedform 21 July 2003; accepted30 July 2003

Abstract

Ocean-color data from Sea-viewing Wide Field-of-view Sensor (SeaWiFS) have been used to investigate temporal andspatial variability of chlorophyll- a concentration in the Brazil–Malvinas Conﬂuence (BMC) region (30–50S and 30–70W). Our analysis is basedon 60 monthly averagedand230 weekly averagedimages (9 9km2 resolution) that span from October 1997 to September 2002 in the Southwestern Atlantic Ocean. A nonlinear model was used to ﬁt the annual harmonic of the chlorophyll concentration anomalies in the region. Analysis of the spatial variability of model parameters has shown the existence of a markedannual cycle at several locations in the studyregion. We also have observedshorter-periodoscillating features on smaller spatial scale, associatedwith the BMC dynamics.These mesoscale features have periods of about 9–12 weeks and have a marked northward propagation with higher chlorophyll values relative to the surrounding waters. Further investigation of these mesoscale features with an advanced very high-resolution radiometer thermal infrared and TOPEX/POSEIDON (T/P) altimeter data has unveiled interesting eddy-like surface structures in the BMC region. These multiple eddies are clearly visible on both thermal and color imagery off southern Brazil. Furthermore, the SeaWiFS images have shown a decrease in chlorophyll-a concentration as these eddies propagate northward, due to mixing with chlorophyll-poor tropical waters carriedsouthwardby the Brazil Current. Several rings along the shelf region, which are probably generatedby shear instabilities, also have been detectedby the ocean-color images. r 2004 Elsevier Ltd. All rights reserved.

1. Introduction eddies (e.g., Olson et al., 1988). This frontal zone is known as the Brazil–Malvinas Conﬂuence (BMC) The Brazil Current andthe Malvinas (also andis recognizedas one of the most energetic known as Falkland) Current converge along the regions of the worldoceans ( Gordon, 1989). The continental margin of South America within the strong mixing causes rapidcooling of subtropical latitude band 35–45S, where both currents ﬂow waters brought by the Brazil Current, making seawardin a complex pattern of meandersand this area very important for understanding circulation andheat transport processes ( Wainer et al., *Corresponding author. Tel.: +55-53-233-6643; fax: +55- 2000). 53-233-6652. The typical position of BMC is at approximately E-mail address: [email protected] (C.A.E. Garcia). 38S(Gordon and Greengrove, 1986), but several

160 C.A.E. Garcia et al. / Deep-Sea Research II 51 (2004) 159–172 studies have demonstrated remarkable variability Surface chlorophyll-a concentration over the of this position. For instance, Provost et al. (1992) shelf in the northern part of the area (31–34S) has studied the dominant periodic variations of been reportedto be high in coastal waters south of sea-surface temperature (SST) in the BMC region Patos Lagoon (>5 mgl1) andmuch lower (32–52S and50–70 W) from 202 advanced very (o0.5 mgl1) near the shelf break in early spring high-resolution radiometer (AVHRR) SST images (Ciotti et al., 1995). Farther south (approximately with 4 4km2 resolution. Although the annual 38–39S), between the coast andthe shelf signal dominates the SST time series, strong break, chlorophyll-a is highest in early spring departures from this signal also are found in the (B4–10 mgl1), associatedwith the coastal front BMC frontal region. The authors further showed andthe development of water-column stability that a semi-annual frequency is present in the (Carreto et al., 1995). This nucleus moves toward region, associatedwith Southern Hemisphere the shelf break during spring, where SAW provide atmospheric forcing. Garzoli andGiulivi (1994) a rich nutrient source. Lowest values have been studied the BMC frontal position and its varia- measuredin winter ( o1 mgl1), while in summer bility through modeling experiments forced by chlorophyll is generally low in the shelf, except for climatological winds. They concluded that the a few sites under influence of SAW (Carreto et al., main source for the BMC frontal variability is a 1995). The region between approximately local windforcing associatedwith the seasonal 39–46S, along approximately 56W was sampled cycle in the Southern Hemisphere. No apparent during four cruises of the Atlantic Meridional correlation was foundbetween wind-forcing Transect (AMT, April 1996–October 1997), and events in the Antarctic Circumpolar Current and total chlorophyll (chlorophyll-a+divinyl chloro- observedanomalous northwardpenetration of the phyll-a) was in the range 0.2–1.2 mgl1 (Gibb et al., Malvinas Current. 2000). Relevant biological processes take place in the Ocean-color data have seldom been used to BMC region, where photosynthetic uptake of investigate the temporal andspatial scales of carbon is enhancedby the nutrient-rich subantarc- dynamical processes in the South Atlantic Ocean. tic waters (SAW) of the Malvinas Current, However, previous works from other areas have contributing significantly with the global carbon shown that chlorophyll from ocean-color sensors budget (Moore andAbbott, 2000 ). This area is can be usedas a goodtracer of water masses (e.g., considered an important region of ocean CO2 Ginzburg et al., 2002) andsurface currents (e.g., uptake, due to high biological utilization of carbon Garcia andRobinson, 1989 ). Spatial andtime in spring andsummer ( Feely et al., 2001). In coverage by previous ocean-color sensors (e.g., addition to the SAW, other sources of nutrients in Coastal Zone Color Scanner) did not allow such the region are the freshwater runoffs from both La analysis to be made in the past. Signorini et al. Plata River (approximately 36S) andPatos (1999) investigatedthe variability of sea-surface Lagoon (approximately 32S), which also produce chlorophyll using first-year Sea-viewing Wide water-column stratification over the continental Field-of-view Sensor (SeaWiFS) data, in conjunc- shelf (Brandhorst and Castello, 1971). In general, tion with an ocean circulation model, TOPEX/ phosphate levels in the low-salinity coastal waters POSEIDON (T/P) altimetry andwind-stress are similar to SAW ranges (0.4–1.0 mM), while data sets, in the tropical and subtropical Atlantic silicate is usually higher, reaching 45 mM(Carreto Ocean during 1997–1998 El Nin˜ o event. They et al., 1986; Niencheski andFillmann, 1997 ; Ciotti concluded that the regions of high-chlorophyll et al., 1995). On the other hand, nitrate levels can concentration are correlatedwith mesoscale and be very low in the freshwater plumes (o1 mM), large-scale physical phenomena. In the eastern contrasting with variable andhigh values tropical Atlantic, for instance, the absence of fall (o1–20 mM) in the shelf break or deeper waters blooms was correlatedwith SST anomalies andEl under SAW influence (Carreto et al., 1995; Ciotti Nin˜ o Southern Oscillation (ENSO) whereas the et al., 1995). Ekman pumping was responsible for spreading ARTICLE IN PRESS

C.A.E. Garcia et al. / Deep-Sea Research II 51 (2004) 159–172 161 surface blooms over the region (Signorini et al., other instabilities observedoff southern Brazil 1999). coast. Raw images (HRPT) were processedto level In this study, we first investigate the spatial and L1A, andsubscenes of 900 pixels by 2100 lines temporal variability of chlorophyll-a concentra- from these full-resolution (1 1km2) images were tion in the BMC region. Then, we explore the selectedfor further processing. Each subscene possibility of using the pigment concentration as a covers the geographical region between 27–50S passive tracer to study surface dynamics in the and37–58 W in the South Atlantic Ocean. The Southwestern Atlantic Ocean. We also present a subscenes were then processedto Level L2 using case study of the evolution of eddies in the area OC2 algorithm provided in SEADAS (version 4.1) during January 2001. The importance of ocean- software. The Level-2 images have a spatial color data in this context is highlighted by close resolution of 1 km. examination of temporally coincident SeaWiFS, AVHRR andT/P satellite data. 2.1.2. Advanced very high-resolution radiometer images AVHRR images, also receivedat Rio Grande 2. Data and methods station, were processedto generate SST fields (Kidwell, 1995). Both the AVHRR andSeaWiFS 2.1. Satellite data images from January 12, 2001 were analyzedin combination to study the mesoscale eddy-structure 2.1.1. Sea-viewing Wide Field-of-view Sensor fieldin the BMC. images We used60 monthly averagedand230 weekly 2.1.3. Altimetry data averagedSeaWiFS Level-3 StandardMapped Along-track sea-level anomalies (SLA) from the Images (SMI) with 9 9km2 resolution to study T/P MergedGeophysical Data records( AVISO, the variability of chlorophyll-a concentration (in 1998) were usedto estimate the dynamic height logarithm scale)—log½C—in the BMC region. The signal associatedto the mesoscale features ob- data set corresponds to the period from October servedby the SeaWiFS sensor. The along-track 1997 to September 2002 (5 years) that were SLA usedhere refers to ascending track 239 on the available by NASA after reprocessing #4. The January 12, 2001 (cycle 306). The corrections original SMI, which have dimensions of 4096 needed to minimize the environmental effects on (longitude) 2048 (latitude) pixels, were converted the altimeter signal (tides, water vapor, iono- into subsamples of an area corresponding to sphere, etc.) were appliedaccording to the AVISO 30–50S and30–70 W. Then, the original SMI (1998) standards. data of 0.089 0.089 spatial gridwere re-binned to 0.52 0.52 grid. The mean images were 2.2. Empirical model of the annual cycle calculatedfrom the weekly averagedre-binned images. Weekly image anomalies, relative to the The methodemployedto modelthe annual cycle total periodof 5 years, were calculatedby is basedon a nonlinear least-squares procedureto subtracting the mean image from each image to model the weekly averaged log½C anomalies. We examine periodic signals for time-series analysis. usedlog chlorophyll values because they were Landandcloudypixels were flaggedto zero and nearly normally distributed, which is a pre- were not taken into account for computations condition for time-series and spectral analyses. carriedout in this work. Mean values of weekly averagedlog ½C were Local area coverage (LAC) SeaWiFS scenes, computedfor each pixel andwere subsequently receivedat Rio Grandegroundreceiving station removedfrom the correspondingtime series to (Brazil), were processedwith SeaWiFS Data obtain the anomalies. The model is based on the % e Analysis System (SeaDAS). Images for January following equation: Yij ¼ Yj þ Yij; where Yij 2001 were selected to study the eddy fields and stands for the observed weekly averaged log½C ARTICLE IN PRESS

162 C.A.E. Garcia et al. / Deep-Sea Research II 51 (2004) 159–172

% at time i (week) andlocation j; Yj is the mean and for the period( Figs. 1A and3 ), the mean position e Yij is the ﬂuctuation (anomaly). The last term is a where continental shelf waters with relatively high- e em sum of the two terms, Yij ¼ Yij þ eij; or chlorophyll content are advected by the western e boundary current system toward open areas is at Yij ¼ aj cos½ð2p=TÞðti fjÞ þ eij: 39.5S–54.25W. We consider this as a mean

em position of the encounter of the two currents, The term Yij stands for model fluctuations, aj is where chlorophyll levels are aroundlog ½C=0.2 the amplitude of the periodic function, fj is the (C ¼ 0:63 mg m3) at the crossing of the 1000 m phase and eij is the residual. T is period(46 weeks) isobath. Those waters then flow seawardin a and ti varies from 1 to 230 weeks (October 1997– complex pattern of menders and eddies carrying September 2002). We useda nonlinear fitting chlorophyll-rich shelf waters. In situ observations procedure to model the periodical signal, andP for 2 along a transect from 30 to 62 S showedhigher each time-series we minimizedthe quantity eij: values of chlorophyll, associatedwith the BMC The root mean square (rms) of residuals of the zone, as comparedwith Brazil Current or Mal- annual model for each location was then esti- vinas Current waters, in three consecutive years mated. We also calculated the ratio of the variance (Brandini et al., 2000). explainedby the modelto the total variance 2 Along the Argentinean continental shelf, the (coefficient of determination, r ) for each time chlorophyll signal is particularly strong. The peak series to verify the goodness of the model. values for chlorophyll concentration are foundat the entrance of La Plata estuary, spreading north- wardalong the Uruguayan andthe Brazilian 3. Results and discussion coastlines. Freshwater discharges by La Plata River andPatos Lagoon (32 S–52W) strongly 3.1. Simple statistics influence the backscattering of radiation in visible bandover the inner continental shelf, as these The mean and standard deviation weekly waters carry large quantities of suspended matter. averagedlog ½C images for the entire periodcan Chlorophyll concentration values given by Sea- be seen in Figs. 1A andB . The minimum and WiFS algorithm in these case-2 waters are maximum values for mean weekly averagedlog ½C reportedto be non-valid( Gonzalez et al., 2000; were 1.85 (0.014 mg m3) and1.09 (12.3 mg m 3), Omachi andGarcia, 2000 ). respectively. The standard deviation of weekly Another striking feature is the strong signal averagedof log ½C variedfrom 0.02 to 1.58. along the Argentinean shelfbreak, with higher The number of cloudless pixels is relatively high values over the continental shelf andBMC regions (Fig. 1C), with mean and standard deviation (Fig. 1A). These high-reflectance streaks are values of 179.5 and26.8, respectively. We also regularly observedalong the Argentinean shelf- computedthe simple statistics of monthly aver- break, associatedwith coccolithophore blooms, agedSeaWiFS data,andsimilar results were which produce large quantities of calcite carbon obtained(not shown here). As seen in Fig. 1A, (Brown andPodest a,! 1997). The southwardBrazil higher mean values are observedover the Brazilian Current shows lower chlorophyll values, asso- andArgentinean continental shelves, as compared ciatedwith the low-nutrient waters of tropical to open-ocean waters. The lower values of origin. The northwardMalvinas Current also chlorophyll offshore in the northern regions are showeda tongue of relatively low chlorophyll due to the oligotrophic waters of the South values (south of approximately 42S). This is in Atlantic Ocean gyre. The transport of surface agreement with results in Brandini et al. (2000), chlorophyll eastwardinto the interior of the ocean who foundconsistently low chlorophyll values also can be clearly seen on the mean image as part (o0.4 mg m3) in the Malvinas Current during of the BMC extension andthe flow of the South November of three consecutive years. Those low Atlantic Current (SAC). Basedon the mean image values were attributedto excessive turbulence and, ARTICLE IN PRESS

C.A.E. Garcia et al. / Deep-Sea Research II 51 (2004) 159–172 163

(A) MEAN IMAGE FOR THE PERIOD OCT97-SEP02

1 - 1 - BC -1 - 1

La Plata 1 - -35 River

-40 SAC

LATITUDE (DEGREE) -45

-0 .4 -50 MC -65 -60 -55 -50 -45 -40 -35 LONGITUDE (DEGREE) Log[C]

-1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8

(B) STANDARD DEVIATION IMAGE FOR THE PERIOD OCT97-SEP02 (C) NUMBER OF CLOUDLESS DATA FOR THE PERIOD OCT97-SEP02

-35 -35

-40 -40

-45 LATITUDE (DEGREE) -45 LATITUDE (DEGREE)

-50 -50 -65 -60 -55 -50 -45 -40 -35 -65 -60 -55 -50 -45 -40 -35 LONGITUDE (DEGREE) LONGITUDE (DEGREE)

Log [C] Number of Data

0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 100 120 140 160 180 200 Fig. 1. (A) Mean chlorophyll concentration, (B) standard deviation and (C) number of cloudless pixels obtained from the weekly averagedSeaWiFS images for periodOctober 1997–September 2002. The main current systems in the Southwestern Atlantic Ocean (BC, Brazil Current; MC, Malvinas Current; SAC, South Atlantic Current) also are shown in (A). The two black lines are transects usedto extract chlorophyll anomalies (see Fig. 2) from the mean monthly chlorophyll ﬁeldover the periodOctober 1997–September 2002. The insert indicates the study area. therefore, instability in the upper layers, particu- like structure associatedwith the BMC, observed larly in the euphotic zone. on the mean chlorophyll image (Fig. 1A). The zonal Hovmoller. diagram unveils strong and clear 3.2. Time–space diagrams of the anomalies annual signal, which has a bandwidth of 5–6 months over the whole area (Fig. 2A). Further- The time history of monthly averagedlog ½C more, near the western boundary the diagram anomalies was examinedalong two zonal transects shows a well-deﬁned westward propagating signal, (see Fig. 1A) in order to depict latitudinal- and which resembles intraseasonal Rossby wave-like longitudinal-variability of the chlorophyll signal. structures, which have been already observed in The transect along 43S was from 58Wto33W the area (Campos andOlson, 1991 ). The same (25 wide), while the transect along 48W ex- signal is not clearly observedfurther east, prob- tended from 35Sto50S (15 wide). Both ably as a result of the markeddecayin chlorophyll transects were chosen so as to cover the tongue- concentration seawardfrom the BMC zone. ARTICLE IN PRESS

164 C.A.E. Garcia et al. / Deep-Sea Research II 51 (2004) 159–172

(A) MONTHLY ANOMALIES FOR A TRANSECT AT 43 S blooms tendto occur ﬁrst in subtropical zones, 9 0.8 6 where markedincreases in SST result in thermal 3 0.6 01 12 stratiﬁcation andthus in water-column stability. 9 0.4 Towardhigher latitudes, those crucial factors need 6 3 some more time to turn the environmental 00 12 0.2 9 conditions from winter to the spring/summer 6 0 regime. This is clearly seen in the annual cycle 3 99 12 plots of northern sites, as comparedto southern

TIME (MONTHS) -0.2 9 sites (Fig. 4). 6 3 -0.4 98 12 9 -0.6 6 3.3. The annual cycle 3 -0.8 97 12 -58 -53 -48 -43 -38 -33 In order to understand the role of periodical LONGITUDE (DEGREES) signals in the chlorophyll concentration, we

(B) MONTHLY ANOMALIES FOR A TRANSECT AT 48 W computedthe variance preserving spectra for six 9 0.8 selectedsites in the studyregion as seen in Fig. 3. 6

3 0.6 A strong annual signal is present in the Argenti- 01 12 9 nean continental (site 1) andshelf-break waters 0.4 6 (site2) andBrazilian shelf waters (site 3) ( Figs. 4A– 3 00 12 0.2 C). In Fig. 4C a positive outlier is seen in June 9 1998, which may be relatedto interannual signals, 6 0 3 such as ENSO, which the model is unable to solve. 99 12 TIME (MONTHS) -0.2 9 Thus, the spuriously high-chlorophyll signal 6 3 3 -0.4 (>50 mg m ) over the southern Brazilian shelf 98 12 at this occasion is probably causedby the strong 9 -0.6 6 reﬂectance signal of suspended matter from the 3 97 12 -0.8 -35 -40 -45 -50

LATITUDE (DEGREES) MEAN IMAGE FOR THE PERIOD OCT1997-SEP2002 Fig. 2. Chlorophyll concentration anomalies (in log scale) for -1 - the (A) zonal (at 43 S) and(B) meridional (at 48 W) transects 3 0.9 -35 for the periodOctober 1997–September 2002. The log ½C -0.8 anomalies were obtainedfrom the monthly averagedimages. -0.7 5 6 -0. - The exact positions of transects can be seen in Fig. 1A.InFig. -40 0.6 -0 .4 2A, the black lines show intraseasonal Rossby wave-like 2 .4 structures propagating westward. .1 5 -0

LATITUDE (DEGREE) 0 .5 -45 -0

6 4 0. 1 - The meridional Hovmoller. of monthly anoma- -65 -60 -55 -50 -45 -40 -35 lies (Fig. 2B) shows evidence of propagation of LONGITUDE (DEGREE) chlorophyll signal as low-frequency wave from Log[C] north to south. As this transect was taken over the 48 W meridian, the signal is not associated with -1.5 -1 -0.5 0 0.5 the polewardadvectionof the Brazil Current itself. Fig. 3. Mean log½C image for the October 1997–September One possible explanation for the observedfeature 2002 period. Modeled and observed log½C anomaly, in is the natural time lag in the annual increase of conjunction with variance preserving spectra, are discussed for six sites shown in this ﬁgure: (1) Argentinean Continental chlorophyll concentration following the seasons, Shelf, (2) Argentinean shelfbreak, (3) Brazilian shelf, (4) as the classical spring bloom process (Sverdrup, Malvinas Current, (5) Conﬂuence Zone and (6) Shed Eddy 1953). As spring approaches, phytoplankton Zone. ARTICLE IN PRESS

C.A.E. Garcia et al. / Deep-Sea Research II 51 (2004) 159–172 165

(A) 1- Argentinean Continental Shelf (B) 2-Argentinean Shelf break 1.5 1.5 Observed Observed Modeled Modeled

1 1

0.5 0.5

0 0 LOG[C] ANOMALY LOG[C] ANOMALY

-0.5 -0.5

-1 Mean log[C] = 0.22 -1 Mean log[C] = 0.08

r =0 .81 r =0 .59 Amplitude (in log[C]) = 0.66 Amplitude (in log[C]) = 0.5 -1.5 Jan 98 Jul 98 Jan 99 Jul 99 Jan 00 Jul 00 Jan 01 Jul 01 Jan 02 Jul 02 -1.5 Jan 98 Jul 98 Jan 99 Jul 99 Jan 00 Jul 00 Jan 01 Jul 01 Jan 02 Jul 02 TIME TIME

1 1

0.5 0.5

0 0 LOG[C] ANOMALY LOG[C] ANOMALY

-0.5 -0.5

Mean log[C] = -0.45 -1 Mean log[C] = 0.067 -1 r = 0.41 r =0 .57 Amplitude (in log[C]) = 0.32 Amplitude (in log[C]) = 0.56

-1.5 -1.5 Jan 98 Jul 98 Jan 99 Jul 99 Jan 00 Jul 00 Jan 01 Jul 01 Jan 02 Jul 02 Jan 98 Jul 98 Jan 99 Jul 99 Jan 00 Jul 00 Jan 01 Jul 01 Jan 02 Jul 02 TIME TIME

(E) 5-Malvinas-Brazil Confluence (F) 6-Eddy Shedding Zone 1.5 1.5 Observed Observed Modeled Modeled

1 1

0.5 0.5

0 0 LOG[C] ANOMALY LOG[C] ANOMALY

-0.5 -0.5

Mean log[C] = -0.26 -1 Mean log[C] = -0.2 -1 r =0 .13 r = 0.032 Amplitude (in log[C]) = 0.19 Amplitude (in log[C]) = 0.098 -1.5 -1.5 Jan 98 Jul 98 Jan 99 Jul 99 Jan 00 Jul 00 Jan 01 Jul 01 Jan 02 Jul 02 Jan 98 Jul 98 Jan 99 Jul 99 Jan 00 Jul 00 Jan 01 Jul 01 Jan 02 Jul 02 TIME TIME Fig. 4. Modeled and observed log½C anomaly for the six sites shown in Fig. 3. ARTICLE IN PRESS

166 C.A.E. Garcia et al. / Deep-Sea Research II 51 (2004) 159–172

VARIANCE PRESERVING SPECTRUM 0.2 || annual semi-annual 0.18

1- Argentinean Continental Shelf 0.16 2- Argentinean Shelfbreak 3- Brazilian Shelf Waters 4- Malvinas Current

) 0.14 5- Malvinas-Brazil Confluence 2 6- Eddy Shedding Zone

0.12

0.1

0.08

POWER SPECTRUM0.06 (LOG[C]

0.04

0.02

0 -3 -2 -1 0 10 10 10 10 FREQUENCY( CYCLES/WEEK) Fig. 5. Variance preserving spectra (log½C squared) for the six sites shown in Fig. 3. anomalously high La Plata River discharge the year (Bertolotti et al., 1996), andwater-column following the 1997–1998 El Nin˜ o event. stabilization must occur later (late spring–sum- The variance preserving spectra for the selected mer), as result of relaxation in southwestern winds. sites confirms the strong annual signal (Fig. 5), The Malvinas Current (site 4) also presents a although differences in phase exist between loca- moderately strong annual chlorophyll anomaly tions. For instance, the peaks in chlorophyll signal (Figs. 4D and5 ). Here, summer peaks are anomaly values in site 1 (Fig. 4A) occur in austral probably associatedwith lower water turbulence summer (January–February) while in the Brazilian at this season, coupledwith across-shelf mixture shelf (Fig. 4C) the peaks are observedat the endof with high-chlorophyll shelf-break waters. In win- the winter (August–September). As statedabove, ter, chlorophyll values are generally low due to the this time lag in pigment peaks probably reflects the high turbulence of these waters (Brandini et al., onset time of phytoplankton blooms in both 2000). In the Confluence Zone (site 5), a broad- regions, driven by water-column stabilization. frequency signal that encompasses annual and Indeed, in the south Brazilian waters ( B32S) semi-annual cycles is present (Figs. 4E and5 ). high chlorophyll andprimary productionrates are Here, variations in chlorophyll levels are influ- observedduringlate winter andspring, associated encedby inter-annual changes in dynamics of both with stabilization of water column with con- currents andtherefore of the confluence itself. In comitant nutrient enrichment from freshwater this region, which is associatedwith the shelf- outflow andprevious nearshore bottom turbulence break front, Carreto et al. (1995) foundhigh levels (Odebrecht and Castello, 2001). On the other of nitrate supporting a high phytoplankton bio- hand, on the Argentinean shelf, south of 45S, mass throughout summer andautumn. However, phytoplankton biomass is mainly regulatedby they pointedout that the shelf-break system varies light availability, determined by water-column in time andspace (either shorewardor seaward) stability, since nutrient levels are high throughout depending on cyclical advance or reverse ARTICLE IN PRESS

C.A.E. Garcia et al. / Deep-Sea Research II 51 (2004) 159–172 167 movements of the Malvinas Current. Thus, a (A) AMPLITUDE OF ANNUAL CYCLE considerable degree of variation departing from an annual signal can be expectedat this site. -45 Basedon 6-year time series of SST fieldsfrom AVHRR images, Lentini et al. (2002) evaluated -40 the lifetime andaverage speedof warm-core rings shedby the Brazil Current. They foundthat ring -35 lifetime ranges from 11 to 95 days, with a LATITUDE (DEGREE) translational mean speedof 13.1 km day 1.We also investigatedthe chlorophyll variability at the -65 -60 -55 -50 -45 40 -35 LONGITUDE (DEGREE) mean position of these eddies (site 6). In the Log [C] Warm-Core Eddy Zone (Fig. 4F), the energy is distributed in a broader spectrum of frequencies; 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 therefore, the annual cycle is not so clear (Fig. 5). This is due to the increase of the relatively higher PHASE OF ANNUAL CYCLE frequency mesoscale energy available in the area, (B) which acts like a noise input in the clean deterministic seasonal cycle. -45 As a synthesis of the annual cycle model performance for the whole study region, a spatial -40 8 distribution of the model parameters is shown in -35 Fig. 6. The amplitude aj (Fig. 6A) andphase ( Fig. LATITUDE (DEGREE) 6B) of the annual cycle show a markeddifference 2 in some locations. The mosaic image of r of fitting -65 -60 -55 -50 -45 40 -35 model to observed log½C anomalies also was LONGITUDE (DEGREE) produced (Fig. 7) for each time series, which Weeks from 1st October explains the goodness of the model for the whole study area. 5 10 15 20 25 30 35 40 In general, the amplitude of the anomalies is Fig. 6. Mosaic images of the estimates of amplitude (A) and small in the inner shelf region all along the South phase (B) of the modeled log½C annual cycle. In the case of American coast, especially in the La Plata estuary phase values, day zero means 1 October. region. This signal of low anomaly amplitude continues across the shelf andinto the interior of the ocean, along the BMC front. The outer shelf SYNTHETIC IMAGE OF R2 andslope regions along the South American coast show higher amplitude (0.4–0.6) except off La -35 Plata estuary, where the continuity is broken by a cross-shelf signal of low anomaly amplitudes. The -40 annual cycle seems to dominate in the outer shelf in the northern region, due to its proximity to the LATITUDE (DEGREE) Brazil Current andto the seasonal role of -45 stratification on phytoplankton production. In the inner shelf, the annual amplitude is small and -65 -60 -55 -50 -45 -40 -35 LONGITUDE (DEGREE) strongly influencedby the La Plata andPatos r2 Lagoon (32S–52W). The fresh andnutrient-rich outflow from these water bodies does not reveal a 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 dominant annual cycle. This is probably partially Fig. 7. Mosaic image of r2 of fitting model to observed log½C due to persistent presence of suspended sediments, anomalies. ARTICLE IN PRESS

168 C.A.E. Garcia et al. / Deep-Sea Research II 51 (2004) 159–172 which result in high mean values (>4 mg m3) and variance of observedfluctuations. We also have standard deviation of computed log½C compared forced the model with other periodic signals such to the surrounding shelf waters (Figs. 1A andB ). as semi-annual, but no significant improvement The long-wave structure associatedwith the was achieved. western boundary current separation and therefore high mesoscale energy can be clearly recog- 3.4. Shorter time-scale features nizedin Fig. 6A. Thus, the region has a small annual amplitude signal (B40W) with high Several processes like the polewardmigration of relative errors (Fig. 7). These frontal motions warm Brazil Current, generation of eddies and andadvection of chlorophyll from coastal areas by signatures of instability along continental shelf the Brazil Current produce intense mesoscale margin, as well as equatorwardpropagation of features at shorter time scales, which also are Malvinas Current, are clearly visible in the reflectedin the log ½C anomaly signal. Therefore, sequence of SeaWiFS images of January 2001 an annual cycle wouldnot be expectedto (not shown here). While the polewardmovement dominate the chlorophyll signal in this area. with subsequent retroflection of the Brazil Current Chlorophyll anomaly fieldin this region fluctuated during summer is well known, the eddy-shedding at broader high-frequency values as will be nature is seldom observed in the ocean-color described below. The model is unable to reproduce images. Fig. 8 shows the distribution of chlor- the annual cycle at several locations in the study ophyll concentration on January 12, 2001 where region and, consequently, cannot explain the the track of the T/P altimeter is also shown.

Fig. 8. SeaWiFS image (1 km resolution) of January 12, 2001. ARTICLE IN PRESS

C.A.E. Garcia et al. / Deep-Sea Research II 51 (2004) 159–172 169

Typically, eddy formation processes have been of eddies from the BMC region. This process associatedwith the classical scenario observedin weakenedafter 26 January andthe northwardjet the Gulf Stream andthe Kuroshio, where zonal gradually disappeared thereafter. We noticed that instabilities generate meanders along the current the development of eddies occurred around begin- path that grow, close upon themselves, andthen ning of January, at a periodof 3–4 days.All the shed eddies (or rings). However, the BMC eddies three eddies visible in Fig. 8 have developed and shown in Fig. 8 are illustrative of retroflection movedout from the generation area between 3 and eddies; associated with currents that retroflect (i.e., 12 January. Later by 26 January, the first two turn back on itself) near their separation latitude eddies lost their identity and only a weak signature (e.g., Olson, 1991). The retroflection eddies are of the tertiary eddy appears in the northern region. generatedclose to the coast roughly at the same During the course of their migration, these eddies location, suggesting a geographically controlled are clearly recognizable by their corresponding formation (Richardson et al., 1994). Recently, core values in chlorophyll concentration. As they Silveira et al. (1999) confirmedthe models movedaway from the generation region, chlor- proposedby Ou anddeRuijter (1986) and ophyll concentration progressively reduced. The Campos andOlson (1991) , stating that the coast- AVHRR image (Fig. 9) shows the cold-core line tilt with respect to the north plays a major role eddies, and the temperature at the core gradually in the type of separation of a particular western increases as they moved northward. Correspond- boundary current. When the coastline has west- ingly, the altimeter sea-level height data along the wardtilt relative to the directionof the current a path of these eddies also show fall in sea level at retroflection type of separation, typically observed the centers of these eddies (Fig. 10). It may be in the Southern Hemisphere western boundary noticedthat the fall in sea-level height anomaly currents (Brazil Current, Agulhas Current, East decreased as these eddies moved northward and Australian Current (EAC)), occurs. Conversely, gainedheat. when the coastline tilts eastwardit results in a much Assuming that at the core of the eddy the smoother separation (Gulf Stream, Kuroshio). chlorophyll-a concentration remains higher than The eddy-structure observed in the BMC region in surrounding waters, we attempted to determine (Fig. 8) resembles the mechanism proposedby the velocity of these eddies by tracking the position Nilsson andCresswell (1981) for the formation of of the peak value of chlorophyll-a at the core of EAC warm-core (anticyclonic) eddies. Their model these eddies. We found that they are being implies that the westwardphase propagation of advected at an average velocity of 12 km day1, baroclinic Rossby waves along the Tasman front, which agrees with the observations in Lentini et al. associatedwith a strong polewardexcursion of the (2002). EAC, is responsible for the eddy detachment from the main flow through the constriction of the EAC retroflection meander. In the same way, there 4. Conclusion couldbe constriction of the first coldmeander(the first meander seaward from the retroflection) to A time series of monthly andweekly averaged form cold eddies in the warm (north) side, as binnedSeaWiFS images is closely examinedin the observedin the BMC case studiedhere.In the Brazil–Malvinas Confluence region. Simple statis- sequence, these eddies may be advected toward the tics performedon the images have shown that the equator following the Brazil Current return flow. mean position where continental shelf waters with Furthermore, it is reasonable to assume that those relatively high-chlorophyll content are advected by eddies decay in size and strength as they travel the western boundary current system toward open away from the source region in a pattern similar to areas is at 39.5S–54.25W. This high-chlorophyll that shown in Fig. 8. signal is usedhere as a quasi-conservative tracer to The ocean-color images of January 2001 show verify time-spacing variability of the BMC. Ana- continuous generation andnorthwardmigration lyses of zonal (at 43S) andmeridional(at 48 W) ARTICLE IN PRESS

170 C.A.E. Garcia et al. / Deep-Sea Research II 51 (2004) 159–172

Eddy 2

Eddy 1

Fig. 9. AVHRR image (1 km resolution) of January 12, 2001. transects of the log½C anomalies have revealeda 0.3 0.5 wave-like strong signal that oscillates approxi- Core of Eddy 1 0.2 mately between 35S and50 S. The zonal Hovm- 0.4 oller. diagram unveils a strong and clear annual 0.1 Core of Eddy 2 signal that has a bandwidth of 5–6 months over 0.3 ) the whole area. The annual cycle can be clearly 0

SLA (m) SLA 0.2

seen on the time–spacing diagrams. (mg.m Chla A simple sinusoidal model was produced to -0.1

0.1 solve the annual signal from the log½C anomaly -0.2 time series. The model explained reasonably well the observedanomalies only in certain regions of -0.3 0 -36 -35.5 -35 -34.5 -34 -33.5 -33 the study area. Analysis of power spectrum in Latitude several zonal transects andat speciﬁc locations has Fig. 10. Chlorophyll concentration (from SeaWiFS, right axis, conﬁrmedthat mesoscale activity dominates over dashed line) and sea-surface height (from T/P, left axis, solid the BMC zone, where periodic signals are about line) along a track (red line) in the eddy propagation zone as 9–12 weeks. shown in Fig. 8. Ocean-color (SeaWiFS LAC) andinfrared thermal images (AVHRR) together with T/P have shown that chlorophyll-a concentration altimetry data were examined to verify mesoscale reduces rather rapidly upon mixing with warmer activities in the region. The eddy-like structures northern waters. During mid-summer, the ARTICLE IN PRESS

C.A.E. Garcia et al. / Deep-Sea Research II 51 (2004) 159–172 171 wind-driven boundary currents along the east phyll in the southwestern Atlantic (30–62S). Deep-Sea coast of Brazil manifest several features of Research I 47, 1015–1033. ! current–current interaction. The separation of Brown, C.W., Podesta, G.P., 1997. Remote sensing of coccolithophore blooms in the Western South Atlantic eddies from the mean flow is rather rapid in this Ocean. Remote Sensing of Environment 60, 83–91. region. Zones of frontal instability also can be seen Campos, E.P.D., Olson, D.B., 1999. Stationary Rossby waves along the boundary between the western branch of in western boundary current extensions. Journal of Physical Brazil Current andthe continental shelf waters. Oceanography 21, 1202–1224. Despite the dynamics of this process being far Carreto, J.I., Negri, R.M, Benavides, H.R., 1986. Algunas from understood, this study shows that SeaWiFS caracteristicas del florecimiento del fitoplancton en el frente del R!ıo de la Plata I: los sistemas nutritivos. Revista de images of surface chlorophyll concentration can be Investigacion y Desarrollo Pesquero, Mar del Plata 5, 7–29. used as tracer to study the bio-physical dynamics Carreto, J.I., Lutz, V.A., Carignan, M.O., Colleoni, A.D.C., in the Southwestern Atlantic andBMC region. Marcos, S.G.D., 1995. Hidrography and chlorophyll-a in a Results in this work are in fairly goodagreement transect from the coast to the shelf-break in the Argentinean between ocean-color andother physical datalike Sea. Continental Shelf Research 15, 315–336. Ciotti, A.M., Odebrecht, C., Fillmann, G., Moller Jr., O.O., sea-surface temperature andsea-level change. 1995. Freshwater outflow andsubtropical convergence Furthermore, the ocean-color data provide sub- influence on phytoplankton biomass on the southern stantial knowledge of signal variability at annual Brazilian continental shelf. Continental Shelf Research 15, andother higher frequencies. 315–336. Feely, R.A., Sabine, C.L., Takahashi, T., Wanninkhof, R., 2001. Uptake and storage of carbon dioxide in the ocean:

the global CO2 survey. Oceanography 14, 18–32. Acknowledgements Garcia, C.A.E., Robinson, I.S., 1989. Sea surface velocities in shallow seas extractedfrom sequential coastal zone color scanner data. Journal of Geophysical Research 94, The authors thank the Goddard Space Flight 12681–12691. Center (GSFC/NASA) for providing the SeaWiFS Garzoli, S.L., Giulivi, C., 1994. What forces the variability of data used in this work. We also knowledge the the southwestern Atlantic boundary currents. Deep-Sea comments made by two anonymous reviewers. Research I 41, 1527–1550. This work was supportedby the Brazilian Ant- Gibb, S.W., Barlow, R.G., Cummings, D.G., Rees, N.W., arctic Program (PROANTAR) andthe National Trees, C.C., Holligan, P., Suggett, D., 2000. Surface phytoplankton pigment distributions in the Atlantic Ocean: Council for Research andScientific Development an assessment of basin scale variability between 50N and (CNPq) from Brazil. 50S. Progress in Oceanography 45, 339–368. Ginzburg, A.I., Kostianoy, A.G., Nezlin, N.P., Soloviev, D.M., Stanichny, S.V., 2002. Anticyclonic eddies in the north- western Black Sea. Journal of Marine Systems 32, 91–106. References Gonzalez, A.S., Santamaria-del-Angel, E., Garcia, V.M.T., Garcia, C.A.E., 2000. Biogeochemical regions of the AVISO, 1998. User’s Handbook for Sea Level Anomalies Atlantic Ocean between 5N and40 S basedon Coastal (SLAs). CLS Report AVI-NT-011-312-CN, Toulouse, Zone Color Scanner imagery. Proceedings of the XV France. International Conference on Ocean Optics, Monaco, Bertolotti, M.I., Brunetti, N.E., Carreto, J.I., Prenzki, L.B., 16–20 October 2000, CD-ROM, Paper 1317, 11pp. Sanchez,! R.P., 1996. Influence of shelf-break fronts on Gordon, A.L., 1989. Brazil–Malvinas Confluence—1984. Deep- shellfish andfish stocks off Argentina. International Council Sea Research 36, 359–384. for the Exploration of the Sea, CM 1996/S:41, 23pp. Gordon, A.L., Greengrove, C., 1986. Geostrophic circulation Brandhorst, W., Castello, J.P., 1971. Evaluacion! de los recursos of the Brazil–Falklandconfluence. Deep-Sea Research 33, de anchoita (Engraulis anchoita) frente a la Argentina y 573–585. Uruguay I: las condiciones oceanograficas,! sinopsis del Kidwell, K.B., 1995. NOAA Polar Orbiter Data (TIROS-N, conocimiento actual sobre la anchoita y el plan para su NOAA-6, NOAA-7, NOAA-8, NOAA-9 NOAA-10, evaluacion.! Proyecto de Desarrollo Pesquero, Serie In- NOAA-11, NOAA-12 andNOAA-14) Users Guide. formes Tecnicos, Mar del Plata 29, 1–63. NOAA-NESDIS, Washington, DC. Brandini, F.P., Boltovskoy, D., Piola, A., Kocmur, S., Lentini, C.A.D., Olson, D.B., Podesta, G.P., 2002. Statistics of Rottgers, R., Abreu, P.C., Lopes, R.M., 2000. Multi-annual Brazil current rings observedfrom AVHRR: 1993–1998. trends in fronts and distribution of nutrients and chloro- Geophysical Research Letters 29, 1–4. ARTICLE IN PRESS

172 C.A.E. Garcia et al. / Deep-Sea Research II 51 (2004) 159–172

Moore, J.K., Abbott, M.R., 2000. Phytoplankton chlorophyll Ou, H.W., de Ruijter, W.P., 1986. Separation of an inertial distribution and primary production in the Southern Ocean. boundary current from a curved coastline. Journal of Journal of Geophysical Research 105, 28709–28722. Physical Oceanography 16, 189–280. Niencheski, L.F., Fillmann, G., 1997. Chemical characteristics. Provost, C., Garcia, O., Gar@on, V., 1992. Analysis of sea In: Seeliger, U., Kjerfve, B. (Eds.), Coastal Marine surface temperature time series in the Brazil–Malvinas Ecosystems of Latin America. Ecological Studies Series Currents confluence region: dominance of the annual and 144. Springer, Berlin, pp. 96–98. semi-annual periods. Journal of Geophysical Research 97, Nilsson, C.S., Cresswell, R.G., 1981. The formation and 17841–17858. evolution of East Australian Current warm-core eddies. Richardson, P.L., Hufford, G.E., Limeburner, R., Brown, Progress in Oceanography 9, 133–183. W.S., 1994. North Brazil Current retroflection eddies. Odebrecht, C., Castello, J.P., 2001. The convergence ecosystem Journal of Geophysical Research 99, 5081–5093. in the Southwest Atlantic. In: Seeliger, U., Kjerfve, B. Signorini, S.R., Murtuggude, R.G., McClain, C.R., Christian, (Eds.), Coastal Marine Ecosystems of Latin America. J.R., Picault, J., Busalacchi, A.J., 1999. Biological and Ecological Studies Series 144. Springer, Berlin, pp. 147–165. physical signatures in the tropical andsubtropical Atlantic. Olson, D.B., 1991. Rings in the ocean. Annual Review of Earth Journal of Geophysical Research 104, 18367–18382. andPlanetary Sciences 19, 283–311. Silveira, I.C.A.da, Flierl, G.R., Brown, W.S., 1999. Dynamics Olson, D., Podesta, G.P, Evans, R.H., Brown, O., 1988. of separating western boundary currents. Journal of Temporal variations in the separation of Brazil and Physical Oceanography 29, 119–144. Malvinas Currents. Deep-Sea Research 35, 1971–1990. Sverdrup, H.U., 1953. On conditions for the vernal blooming of Omachi, C.Y., Garcia, C.A.E., 2000. Analysis of empirical phytoplankton. Journal du Conseil permanent international algorithms of surface chlorophyll-a for SeaWiFS in d’ Exploration de la Mer 18, 287–295. the Southwestern Atlantic Ocean. Proceedings of the Wainer, I., Gent, P., Goni, G., 2000. Annual cycle of the Brazil XIV International Conference on Ocean Optics, confluence region in the National Center for Atmosphere 16–20 October, Monaco, France, CD-ROM, Paper 1184, andResearch Climate System Model.Journal of Geophy- 13pp. sical Research 105, 26167–26177.