Deep-Sea Research II 49 (2002) 2143–2162

Four-year study of large-particle vertical distribution (0–1000 m) in the NW Mediterranean in relation to hydrology, phytoplankton, and vertical flux Lars Stemmanna,*, Gabriel Gorskya, Jean-Claude Martya, Marc Picherala, Juan-Carlos Miquelb a

Laboratoire d’Oc!eanographie de Villefranche (LOV), Observatoire Oc!eanologique, BP 28, 06 234 Villefranche sur mer, Cedex, France b Marine Environment Laboratory, International Atomic Energy Agency, 4 Quai Antoine 1er, BP 800, MC98000 Monaco Accepted 17 November 2001

Abstract Data on large particles (LP; >0.15 mm), phytoplankton communities, vertical fluxes, and hydrology were collected between January 1992 and June 1996 in the NW Mediterranean Sea, during DYFAMED, an interdisciplinary program part of JGOFS France. LP concentrations at the study sites were typical for values found in other open-ocean studies. LP temporal evolution showed an annual cycle. Concentrations were the highest in winter/spring (20–120 l
1, 5– 280 mg m
3) and lowest in summer and autumn (0–20 l
1, 0.8–60 mg m
3). We estimated that LP accounted on average, for 2–30% of the total particulate (>0.7 mm) dry weight (DW). LP temporal evolution between 0–200 m was correlated with total Chl a and fucoxanthin (diatoms), and inversely correlated to zeaxanthin (cyanobacteria and prochlorophytes). Although diatoms were clearly associated to LP, prymnesiophytes were associated to the two highest accumulation of particles >1 mm. The DW fraction of particle >0.5 mm to total LP increased from 10% in regenerated systems dominated by picoplankton to 50% during spring blooms. LP concentrations in the upper 200 m were correlated to mass flux recorded in sediment trap at 200 and 1000 m. We defined three main periods for LP downward export related to physical stratification. (1) The major LP export occurred in winter and may be affected by deep vertical mixing; (2) in spring, at the onset of the thermal stratification LP downward export decreased, although pulses of phytoplanktonic production may have enhanced LP sedimentation over short time scales; (3) during the summer stratification, the deep water was generally depleted in LP. r 2002 Elsevier Science Ltd. All rights reserved.

1. Introduction

*Corresponding author. Department of Oceanography, Texas A&M University, College Station, TX, 77843, USA. Tel.: +1-979-845-1115; fax: +1-979-845-8219. E-mail address: [email protected] (L. Stemmann).

Marine aggregates harbour dense attached microbial communities and contain unique chemical microhabitats where photosynthesis, decomposition and nutrient regeneration can occur at highly elevated levels and serve as food for a variety of zooplankton (Alldredge and Silver, 1988; Lampitt et al., 1993a; Kirboe, 2000).

0967-0645/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved. PII: S 0 9 6 7 - 0 6 4 5 ( 0 2 ) 0 0 0 3 2 - 2

2144

L. Stemmann et al. / Deep-Sea Research II 49 (2002) 2143–2162

Aggregates may be a major transport agent of the surface-derived organic production to the deep sea (Fowler and Knauer, 1986). Studies on spatial distribution and temporal dynamic of large aggregates began two decades ago when adapted imaging systems were developed. Studies in the Pacific and Atlantic Oceans and in the Mediterranean Sea showed that marine aggregates larger than several hundred micrometers were ubiquitous in the world ocean, with concentrations ranging from almost 0 to 150 aggregates l
1 in the open ocean and from almost 0 to 500 aggregates l
1 in the nearshore environment. According to Alldredge and Silver (1988), Riebesell (1991b), Jackson et al. (1997), and Stemmann et al. (2000), these aggregates can constitute up to more than 70% of the total weight of particulate matter >0.7 mm. Because few data are available from the Mediterranean Sea, the primary objective of this work is to provide new insight into large particles (LP) (>0.15 mm) concentrations and size distributions in the Mediterranean Sea over several annual cycles. Aggregation processes, both physical and biological, are responsible for the packaging of the surface phytoplanktonic production into large aggregates. Biological aggregation is a direct function of zooplankton concentration and activity, while physical aggregation is a function of phytoplankton cell concentrations. Under low zooplankton concentrations, physical coagulation has been implicated as the main mechanism forming aggregates (Jackson, 1990). Numerous works have studied the role of diatoms in aggregation (Riebesell, 1991a, b; Alldredge and Jackson, 1995; Kirboe et al., 1996, 1998), but less information is available on the role of prymnesiophytes in aggregation and vertical flux of organic matter (Honjo, 1982; Wassmann, 1994). The second objective of this work is to relate the observed changes in LP distribution with the phytoplanktonic community structure (diatoms, prymnesiophytes, cyanobacteria and prochlorophytes). Once formed, phytoplankton aggregates settle rapidly because they have potentially higher settling speeds than single particles (Alldredge and Silver, 1988; Alldredge and Gotschalk, 1988).

However, mass aggregation of diatom blooms is not always followed by mass sedimentation due to the high biological remineralization that takes place in the euphotic zone (Kirboe et al., 1998). Moreover, retention of marine snow in the surface layer can be linked to the physical stratification (MacIntyre et al., 1995) and also to the formation of gas bubbles within the aggregates (Riebesell, 1992). According to Kirboe (2000), most of the aggregated material may be degraded within the euphotic zone, in several days, by colonizing invertebrates and microbes. The last objective of this work is to determine the distinct biological and hydrological conditions, which might affect LP export to the deep sea.

2. Methods 2.1. Field sampling between 1992 and 1996 Data on LP, phytoplankton communities, vertical fluxes, and hydrology were collected during the DYFAMED program between 1992 and 1996 (Dynamics of Fluxes in the MEDiterranean). The studied station was sampled monthly since 1991 in the course of the DYFAMED long-term survey program, part of the France-JGOFS activities (Buat-Menard and Lambert, 1993). The study site was located at 431250 N, 071520 E in the Ligurian Sea, 52 km off Nice-France, and 2350 m deep (Fig. 1). In the NW Mediterranean Sea the continental shelf is narrow and the slope is steep (off Nice, the 1000 m isobath is located only 9 km offshore). The Underwater Video Profiler (Gorsky et al., 1992; Stemmann et al., 2000) was deployed during 46 cruises at the DYFAMED site between January 1992 and July 1996 (Table 1). The average time interval between deployments was 30722 days. All profiles were performed during daytime in order to avoid diel variation that can be important (Lampitt et al., 1993b; Graham et al., 2000; Stemmann et al., 2000). CTD data, nutrients, phytoplanktonic biomass and pigments were obtained most of the time simultaneously (on average 2.5 days before or after; Table 1). Major delays between UVP casts

L. Stemmann et al. / Deep-Sea Research II 49 (2002) 2143–2162

and other sampling occurred on July and November 1995 (13–15 days). From 1992 to 1993 the CTD system used was a SeaBird SBE 9 underwater unit. Water for pigment (and other parameters) analyses was obtained from hydrocasts using Niskin bottles. After 1993 the CTD system consisted of a SeaBird SBE911+ unit equipped with additional sensors (dissolved O2 and fluorescence) mounted on a SeaBird rosette equipped with 12 12-l Niskin bottles. Salinity and dissolved oxygen data from CTD sensors were calibrated with discrete salinity and oxygen data from Niskin bottles. The samples were analyzed for nitrate, nitrite, phosphate, and silicate using standard automated colorimetric methods (Tre! guer and Le Corre, 1975). Phytoplankton pigments were analyzed by HPLC (see Marty et al., 2002). Fucoxanthin, 190 -Hexanoyloxyfucoxanthin (19HF0 ), and zeaxanhtin were used to monitor diatom, prymnesiophyte and cyanobacteria communities respectively (see in Marty et al., 2002).

2145

The vertical mass flux at 200 and 1000 m depths was recorded from May 1993 to March 1994 and from June 1994 to June 1995 in PPS5 sediment traps (Miquel and La Rosa, 1999). Traps were deployed at 200 and 1000 m depth and were preprogrammed for 15-day collection periods. Two Aanderaa current meters with temperature sensors were fixed 5 m below the traps. Most of the time, currents at 200 m depth were lower than 5– 6 cm s
1 but could reach 12–14 cm s
1 during short

Table 1 Sampling periods for LP, phytoplankton and sediment trap at DYFAMED between 1992 and 1996 Date

Date

Sediment

Date

Date

Sediment

(UVP)

(Phyto.)

trap

(UVP)

(Phyto.)

trap

13/01/92* 20/01/92

04/05/94*

28/04/94

21/02/92

09/05/94*

26/03/92* 30/03/92

06/06/94*

02/06/94

08/04/92

30/06/94*

07/07/94

26/05/92* 21/05/92

22/08/94*

29/06/92

29/06/92

20/08/94 25/09/94

07/07/92* 17/07/92

17/10/94*

15/10/94

15/09/92* 21/09/92

24/11/94*

20/11/94

13/10/92

03/12/94*

13/10/92

19/11/92* 09/11/92

15/01/95*

17/01/95

16/12/92

12/02/95*

10/02/95

27/01/93* 14/01/93

06/03/95*

09/03/95

19/02/93

03/04/95*

10/04/95

10/03/93* 09/03/93

12/04/95*

23/03/93

12/05/95

21/04/93* 22/04/93

16/05/95

17/05/93*

26/05/95

11/05/95

24/05/93* 25/05/93 23/06/93 08/07/93

31/05/95

06/07/93

21/07/93*

08/06/95 22/07/95*

08/09/93* 07/09/93

14/09/95

05/10/93* 04/10/93 16/10/93

Fig. 1. The sampling zone. The central and coastal zones are delimited by the frontal zone (grey line). The flow of the Ligurian Current is represented by the arrows. The black line represents the transects performed in the framework of the MBP-Front program to monitor the extension of the coastal water. The asterisk marks the position of the DYFAMED station.

05/10/95 29/11/95

09/11/93* 08/11/93

03/12/95

10/12/93* 06/12/93

09/02/96*

01/02/94

02/04/96*

05/04/96 23/05/96

08/03/94* 04/03/94 08/04/94* 07/04/94

15/11/95

01/03/96

12/01/94* 12/01/94 03/02/94

07/07/95

18/06/96

The * marks period for which a CTD transect was completed.

2146

L. Stemmann et al. / Deep-Sea Research II 49 (2002) 2143–2162

periods in winter (always o6 consecutive days). At 1000 m depth, currents were o3–4 cm s
1 although currents of 10–12 cm s
1 were occasionally observed in winter. The 0–200 m integrated load of LP, phytoplankton pigments (including Chl a) at 200 and 1000 m depth were compared by cross correlation (Spearman rank correlation). All the data were interpolated over a period of 1 month assuming that the LP and phytoplankton monthly measurements were representative for the whole month. Periods with missing data were excluded from the analysis. 2.2. The Underwater Video Profiler data The Underwater Video Profiler (UVP model II, Gorsky et al., 1992) records and provides information on abundance, size and shape of objects larger than 0.1 mm situated in its field of vision at a rate of 25 Hz. For the quantitative study of particles, the camera recorded objects illuminated in a 1.5-cmthick volume of water (0.28 l). Because the lowering speed was 1 m s
1, one frame was recorded every 4 cm and successive images did not overlap. The recorded images are digitized and automatically treated by the image analysis software. Calibration experiments in a sea-water tank, using measured particles of two types, dense marine aggregates (fecal pellets, copepods molts) and transparent aggregates (mucous particles, gelatinous plankton), showed that the UVP was able to detect particles larger than 100 mm in equivalent spherical diameter (ESD). The metric surface (Y ) as a function of the pixel surface (X ) is Y ¼ 0:00139  X 1:43 (Stemmann, 1998). The area of each particle on the video profile was converted to ESD (in mm) and equivalent spherical volume (ESV in ppm). In order to evaluate the quantity of LP relative to the total particulate pool (>0.7 mm), we have calculated LP dry weight (DW) using the equations of Alldredge and Gotschalk (1988) obtained in Monterey Bay, on aggregates sampled in the euphotic zone. Alldredge (1998) recently confirmed the algorithms for particles of different origins. In this work we present the results concerning particles larger than 0.15 mm for which the

detection is accurate. Data treatment was performed for the portion of the profile displaying images with constant dark background only. Sunlight interfered with the collimated light beam. Therefore, most of the data were only obtained below 50 m depth and in this paper we present vertical profiles in the 60–1000 m depth range. However, in order to compare LP DW temporal evolution with the integrated load of Chl a (0–200 m depth) and vertical flux, we have estimated LP DW in the 0–200 m layer by calculating the mean LP concentration between 60 and 200 m and extrapolating the result to 0– 200 m depth. To estimate the bias induced by the extrapolation, we performed 10 comparisons using day/night profiles obtained at DYFAMED site during May 1995 for which we had measurements in the full 0–200 m layer. These comparisons indicated that the extrapolation induced an underor over-estimation ranging from 0.5% to 17.2% (Stemmann, 1998). We conducted a hierarchical flexible clustering in order to define typical LP size distributions between 1992 and 1996. Each profile obtained during this period was divided vertically into 22 layers. Each layer was 20 m thick in the superficial 150 m (70 l sampling volume) and 50 m thick below this depth (175 l sampling volume). Thus, we could consider 1012 layers, each with the corresponding size distribution determined with a minimum count of 1000 particles. As particles >1.5 mm were rare (o0.01% of total count), they were summed in the last size class. The hierarchical flexible clustering was performed on a matrix of Kolmogorovdistances among the size distributions (Legendre and Legendre, 1984; Stemmann et al., 2000). For the quantitative study of large amorphous aggregates (LAA >1 cm) and zooplankton, the camera recorded objects illuminated in 70 l of water. Since the lowering speed in this case was 0.5 m s
1, one frame is recorded every 2 cm and the successive images overlaped. The LAA vertical distribution was visually estimated in layers 50 m thick between 0 and 200 m depth and 100 m thick between 200 and 1000 m. Although, the size resolution is B0.5 mm, we estimated semi-quantitatively only the LAA larger than 1 cm. The occurrence of LAA was coded in 4 abundance

L. Stemmann et al. / Deep-Sea Research II 49 (2002) 2143–2162

ranges (no observations, o1, from 1 to 10, and more than 10 aggregate m
3).

3. Results 3.1. LP abundance, volume and DW between 1992 and 1996 LP abundance, ESV and calculated DW vertical distributions at the DYFAMED site between 1992 and 1996 are shown in Fig. 2. The site was not sampled for LP in 1995 between July 22 and November 29. In the 60–200 m layer, LP abundance and DW were highest from January to May (100–120 part l
1 and 60–280 mg m
3) and decreased from June to December (0–10 part l
1 and 0–60 mg m
3). In the deeper layers, down to 1000 m depth, the pattern was similar although the concentrations were smaller (20–30 part l
1 in winter–spring and 0–20 part l
1 in summer– autumn). The peaks duration were also shorter (between February and March in most of the year). LP ESV ranged from 0.07 to 3 ppm and showed high values (1–3 ppm) between the end of December and end of May in the upper 150 m. A second peak was observed in the upper 150 m depth in June 1992 and September 1993. Summer and autumn concentrations were very low (0.07–0.5 ppm). The deeper layers down to 1000 m depth showed the same temporal evolution but with lower concentrations (0.07–0.1 ppm). Transition periods between high winter–spring concentrations and summer low concentrations were short (o1month) as suggested by the sharp decrease in the entire water column between two sampling. 3.2. LP abundance for three size classes between 1992 and 1996 Table 3 gives the mean concentrations, over the whole period, of three LP size classes (0.15–0.5, 0.5–1 and >1 mm in ESD). The smallest particles were dominant but their variability estimated by the coefficient of variation was the lowest, suggesting that they were more uniformly distrib-

2147

uted. The temporal evolution of three size classes showed that the smaller particles numerically dominate in the water column (Fig. 3). The temporal evolutions were roughly similar for the three size classes, showing higher concentration during winter-spring time and lower the rest of the year. The concentrations were greater in the 60–300 m depth layer than deeper. The LP >1 mm concentrations below 200 m depth showed more variability than the two smaller size classes. Spatial and temporal differences between the three size classes were particularly marked in 1994 during the winter and spring. On 2 February 1994, the deep maxima of the three size classes were located at three different depths, the small particles between 400 and 700 m (peak at 500 m), the middle-sized particles between 660 and 900 m (peak at 720 m), and the large particles below 800 m depth (peak at 800 m). Assuming that these particles were simultaneously exported from the surface after the previous sampling performed on 12 January, their different vertical distributions suggest that settling speed increased with LP size. It is possible to calculate three minimum sinking speeds of 20, 30, and 40 m day
1, respectively, for LP 0.15–0.5 mm, 0.5–1 mm, and >1 mm. 3.3. LP size distribution between 1992 and 1996 We calculated 1012 size distributions from the 46 profiles divided into 22 layers between 60 and 1000 m depth (see Section 2). Classification of the 1012-particle size distributions yielded four groups (dendrogramm not shown here). The common feature in all cases was the general decrease in particle abundance with increasing size (Fig. 4). The difference in the size distribution is mainly due to the increase of the frequency of particles >0.5 mm that contributed up to 25% of the total abundance in group 1 and o5% in group 4. The difference was even greater when the frequencies were expressed in terms of DW and volume. Particles >0.5 mm corresponded to 50% of the total DW in group 1 and to 10% in group 4. Particles >0.5 mm corresponded to more than 90% of the total ESV in group 1 and o40% in group 4.

2148

L. Stemmann et al. / Deep-Sea Research II 49 (2002) 2143–2162

Fig. 2. Multi-annual variation of LP concentrations in the 60–1000 m water column expressed in (a) abundance per liter, (b) Equivalent spherical volume, and (c) dry weight (mg m
3). The WIW is figured by the isotherm of 13.11 in graphs a, b and c (see text for details).

The spatial and temporal locations of the four separated groups are given in Fig. 4. Large particles (group 1) were found in the 60–600 m layer, whereas small particles (groups 3 and 4) occupied the entire water column. Group 1 (large particles) was only observed from December to June and group 3 (small particles) from May to December.

3.4. Large aggregates vertical distribution The vertical distribution of LAA larger than 1 cm is presented in Fig. 5. A seasonal pattern was observed with the highest concentrations in winter and spring time and the lowest in summer and autumn. LAA highest concentrations were found

L. Stemmann et al. / Deep-Sea Research II 49 (2002) 2143–2162

2149

Fig. 3. Multi-annual variation of three size classes of LP concentrations in the 60–1000 m water column (aggregates in 70 l). (a) From 0.15 to 0.5 mm, (b) from 0.5 to 1 mm, and (c) >1 mm.

L. Stemmann et al. / Deep-Sea Research II 49 (2002) 2143–2162

2150

(a) 90

80

80

80

70

70

70

G4

50

G3

40

G2

30

60 50 40

60 50 40

30

30

20

20

10

10

10

0

0

20

Depth (m)

0.6 0.8 1.0 ESD (mm)

1.2

1.4

1.6

0 0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

ESD (mm) d120194 d030294 d080394 d080494 d040594 d090594 d060694 d300694 d220894 d171094 d241194 d031294

Date

(b)

0.4

d270193 d190293 d100393 d230393 d170593 d240593 d080793 d210793 d080993 d051093 d161093 d091193 d101293

0.2

d130192 d260392 d260592 d070792 d150992 d191192

0.0

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

ESD (mm)

d150195 d120295 d060395 d120495 d120595 d160595 d260595 d290595 d310595 d220795 d291195

G1

d090296 d020496 d180696

60

DW (%)

100

90

Volume (%)

100

90

Abundance (%)

100

0-60 60-80 80-100 100-125 125-150 150-200 200-300 300-400 400-500 500-600 600-700 700-800 800-900 900-1000 G1 (large particles) G2

G3 G4 (small particles)

Fig. 4. Multi-annual variation of LP size distributions in the 60–1000 m water column. (a) Mean cumulative size distribution of four groups isolated by clustering and (b) spatio-temporal location of the four groups during DYFAMED.

between 50 and 600 m depth, except on 24 May 1993 and 8 April 1995 when high concentration were observed below 600 m depth. Although low LAA concentrations were typical for summer, some exceptions were noted. For example, on 8 September 1993, high concentrations of LAA were observed from surface to 500 m depth and on 7 July 1994, high LAA concentrations were found between 100 and 200 m depth. 3.5. Temperature, nutrients and phytoplankton biomass An annual thermal evolution was observed through the 4 years of the study at DYFAMED (Fig. 6A). The water column was homogeneous

during the winter and stratified thermally in summer with a thermocline at B25 m depth in May and June and deeper in summer and autumn. The highest superficial temperatures (>251C) were observed between July and August. The vertical stratification generally disrupted during December each year. Between 1992 and 1996, winter intermediate water (WIW, temperature o13.11C and salinities of 38.1–38.3) was observed from January to March at different depths. The maximum deepening of the 13.1 isotherm generally took place in March, except in February 1996. During summer the stratification isolated the surface layer from the subjacent waters and stopped nutrient influx. Hence, the lowest nitrate concentrations (NO3o0.1 mM) were found in

D101293

D211194

D201192

D091193

D161093

D131094

D051093

Autumn D240994

D100992

D80993

D220894

D210793

D070794

D180696

D060694

D240593

D150593

Summer D040594

D220493

D080495

D080494

D030495

D020496

D100393

D070394

D060395

Spring D120295

D090296

D270193

D160195

D150192

D120194

Depth\Dates

Winter

0-50 LP

LP

LP

LP

LP

LP

LP

100-150 150-200 200-300 300-400 400-500 500-600 600-700 700-800 800-900 900-1000

Abundance

LAA max > 10 per m3 LAA 0,1