Journal of Marine Systems 94 (2012) 135–144

Contents lists available at SciVerse ScienceDirect

Journal of Marine Systems

j ourna l homepage: www.e lsev ie r .com/ locate / jmarsys

Physiochemical controls on phytoplankton distributions in the Ross Sea, Antarctica

Xiao Liu ⁎, Walker O. Smith Jr.Virginia Institute of Marine Science, College of William & Mary, Gloucester Pt., VA 23062, United States

⁎ Corresponding author at: Department of Earth SciCalifornia, Los Angeles, 90089, CA, United States. Tel.: +

E-mail address: liu284@usc.edu (X. Liu).

0924-7963/$ – see front matter © 2011 Elsevier B.V. Alldoi:10.1016/j.jmarsys.2011.11.013

a b s t r a c t

a r t i c l e i n f o

Article history:Received 6 December 2010Received in revised form 20 November 2011Accepted 22 November 2011Available online 28 November 2011

Keywords:Principle componentsTemperaturePhaeocystisVariabilityHydrography

The continental shelf of the Ross Sea, Antarctica, is characterized by extreme seasonal and interannualchanges in atmospheric and oceanographic processes, which result in distinct temporal patterns in phyto-plankton biomass and assemblage composition. However, the environmental forcing of these variations re-mains uncertain, especially when a series of correlated variables are considered. Hydrological profiles,dissolved nutrients, particulate matter, and phytoplankton pigments were measured in the southern RossSea in austral spring and summer during four years (1996–97, 2003–04, 2004–05, and 2005–06), and a seriesof multivariate analyses were conducted to assess the causative mechanisms in the control of phytoplanktondistributions in the Ross Sea. Our results demonstrate that the significant interannual, seasonal and spatialvariability that occurs in the southern Ross Sea in hydrographic and chemical properties is highly correlatedwith the variability in phytoplankton distributions. Although multiple controlling mechanisms were sug-gested, mixed layer depths did not appear to be a dominant factor regulating phytoplankton biomass or com-position; conversely, we found a significant role of water column temperature in structuring phytoplanktonassemblage composition in the southern Ross Sea, in that cooler water strongly selects for Phaeocystis antarc-tica, which is a dominant control of carbon flux to depth, and thus of substantial biogeochemical importance.

© 2011 Elsevier B.V. All rights reserved.

1. Introduction

The ocean and its plankton are changing as a result of anthropo-genically altered atmospheric conditions, which have resulted in in-creased oceanic surface temperatures, changed fluxes of visible andinfrared radiation, modified hydrological regimes (which in turnalter land inputs of nutrients to coastal systems), and increased oce-anic concentrations of carbon dioxide and decreased pH (Caldeiraand Wickett, 2003; Levitus et al., 2000; Montes-Hugo et al., 2009).These present and future changes have been incorporated intomodels to predict the impacts on oceanic biota and biogeochemicalcycles (e.g., Blackford, 2010; Orr et al., 2005). However, direct exper-imental verification of the influence of these variables is difficult, iffor no other reason that it is experimentally difficult to address bothindividual and interactive effects of multiple variables at one time. Afew investigations have examined the biological responses and inter-actions between temperature and micronutrients (Rose et al., 2009),CO2 and temperature (Feng et al., 2010; Fu et al., 2007, 2008;Hutchins et al., 2007; Hutchins et al., 2009), and CO2 and irradiance(Feng et al., 2008), but experimental manipulations of all potentiallychanging variables are not practical in plankton systems.

Another means of addressing the importance of climate-changevariables to phytoplankton growth, biomass and composition is by

ences, University of Southern1 213 740 5089.

rights reserved.

statistically analyzing multiple types of field observations using prin-cipal component analysis. This procedure detects patterns among asuite of variables, and generates components which describe a signif-icant fraction of the variability observed. The eigenvector of eachcomponent provides a measure of the variability explained by eachset of variables, as well as insights into the relationship among thevariables within each component. By using a large data set, robustcorrelations can provide insights into the causal mechanisms thatgenerate the correlations. Such an approach has been used in a varie-ty of oceanographic studies previously (e.g., Beaugrand et al., 2003;Edwards and Richardson, 2004; Massolo et al., 2009).

The Southern Ocean is changing rapidly, although the magnitudeand direction of the changes vary regionally. For example, the airtemperatures near the West Antarctic Peninsula are rapidly rising,ice concentrations have decreased dramatically (Vaughan et al.,2003), and ecosystem shifts have been noted (Fraser and Hofmann,2003; Montes-Hugo et al., 2009). Conversely, the Ross Sea sector isexperiencing increased ice cover (Parkinson, 2002; Stammerjohnet al., 2008) and a shortened open-water season, with substantialspatial variations in these trends, and biotic responses appear to becorrelated with these changes (Smith et al., 2007). The Ross Sea alsohas substantial spatial variability in physical, chemical and biologicaldistributions (Arrigo and van Dijken, 2004; Smith et al., 2003a,2010; Stammerjohn et al., 2008), and exhibits a large amount ofshort-term and interannual temporal variability as well (Arrigo andvan Dijken, 2004; Peloquin and Smith, 2007; Smith et al., 2000,2010; Smith et al., 2011a). The generalized seasonal pattern of

Table 1The years and dates of the cruises included in the analyses, and number of stations andsamples from each cruise. Growing season of 2003–04 includes the period of australspring in late 2003 and austral summer in early 2004.

Dates of cruises Growing season Number ofstations

Number ofsamples

13 Jan–06 Feb, 1997 1996–97, summer 14 11228 Dec–29 Dec, 2003 2003–04, spring 16 4403 Feb–06 Feb, 2004 2003–04, summer 20 6120 Dec–24 Dec, 2004 2004–05, spring 19 9728 Jan–31 Jan, 2005 2004–05, summer 21 4726 Dec–31 Dec, 2005 2005–06, spring 5 3001 Jan–23 Jan, 2006 2005–06, summer 13 154

136 X. Liu, W.O. Smith Jr. / Journal of Marine Systems 94 (2012) 135–144

phytoplankton composition is for the haptophyte Phaeocystis antarc-tica (P. antarctica) to grow and accumulate during austral springthrough late December, and diatoms to continue to grow in summer(Smith et al., 2010; Tremblay and Smith, 2007). This pattern hasbeen explained by the decreases in the mixed layer depths in theRoss Sea through the growing season (Smith and Asper, 2001),since P. antarctica is associated with deeper mixed layers than arediatoms (Arrigo et al., 1999; Smith et al., 2000, 2006), apparentlydue to its photosynthetic plasticity (Kropuenske et al., 2009). How-ever, it has proven difficult to quantify the differences in in situ pho-tosynthetic responses between the two assemblages (van Hilst andSmith, 2002). It is likely that other variables (such as micronutrients)significantly contribute to the spatial and temporal variability, butthis remains to be demonstrated.

Variability in the global ocean and the Ross Sea occurs at a varietyof time and space scales. Significant interannual variability in phyto-plankton biomass (Peloquin and Smith, 2007; Smith et al., 2010),productivity (Arrigo and van Dijken, 2004; Smith et al., 2006), thetiming of the initiation of the primary spring bloom (Arrigo et al.,1998), and vertical fluxes (Smith et al., 2011b) has been observed.Seasonal variability in chemical and biological properties has beenwell documented (Gordon et al., 2000; Smith et al., 2000; Sweeneyet al., 2000), and the short-term variations induced by physicalforcing have also been reported (Smith et al., 2011a). Interestingly,current models cannot adequately resolve this variability, and do

Fig. 1. Map of the station locations used in this analysis. Shaded areas indicate 76.5–77.5°S,and standard errors were extracted and binned.

not include the forcing generating the observed oceanic variability.As a result, alternative approaches to viewing planktonic systemscan provide new ways to understand both the variability and the pro-cesses occurring in natural systems.

We conducted statistical analyses of a series of cruises to thesouthern Ross Sea to understand the causal mechanisms in the con-trol of phytoplankton assemblage composition. A similar approachwas recently conducted by Massolo et al. (2009), but they used adata set which was much more limited in time and space. We alsotreated the data to determine the effects of depth and season. Wehypothesized that temperature had an important influence on phyto-plankton composition, despite the modest (at most 5 °C) temperaturerange observed in the southern Ross Sea. Our results demonstratethat the significant interannual variability that occurs in the southernRoss Sea in hydrographic and chemical properties is highly correlatedwith the variability in plankton, and that water column temperatureexerts dominant control on structuring phytoplankton assemblagesduring the growing season.

2. Methods

2.1. In situ observations

CTD casts (continuous measurements of temperature, conductivi-ty and derived salinity and density, with data binned at 1 m intervals)and water samples were collected from the southern Ross Sea inlate spring (mid- to late December) and summer (early January toFebruary) on board the RVIB N.B. Palmer and USCGC Polar Star duringfour years (growing season of 1996–97, 2003–04, 2004–05, and2005–06; Table 1). Stations were occupied north of the Ross IceShelf and south of 75°S (Fig. 1). Mixed layer depth was estimated asthe depth where σT changed by 0.01 units from a stable surfacevalue (Smith et al., 2000). Sample collection and processing are de-scribed in detail elsewhere (Smith et al., 2000, 2006, 2010). Watersamples were collected from the upper 200 m of the water columnand analyzed for dissolved nutrients [NO3, NO2, PO4, Si(OH)4, andNH4], particulate matter [particulate organic carbon (POC), particu-late nitrogen (PN), and biogenic silica (BSi)], and phytoplankton pig-ments. Nutrients were analyzed using automated techniques, and

170–180°E, from which region MODIS eight-day averaged chlorophyll a concentrations

y,20

05an

dJanu

ary,

2006

).

hex

L−1)

NO3:PO

4

±2.71

12.2±

0.91

±0.76

14.9±

2.15

±0.80

12.8±

0.78

±0.13

12.2±

1.20

±0.73

17.9±

4.68

±0.17

14.1±

1.28

±0.59

13.9±

2.88

137X. Liu, W.O. Smith Jr. / Journal of Marine Systems 94 (2012) 135–144

POC and PN via pyrolysis on a Carlo-Erba 1108 elemental analyzer. Bio-genic silica (BSi) concentrations were determined by alkaline digestionand quantified spectrophotometrically (Nelson and Brzezinski, 1990).Samples for chlorophyll a (chl-a) were assessed fluorometrically(Smith et al., 2000), and accessory pigments were quantified usinghigh performance liquid chromatography (DiTullio et al., 2003; Smithand Asper, 2001).

rs(Jan

uary,1

997,

Februa

ry,2

004,

Februa

r

Chl-a

(μgL−

1)

Fuco

xanthin

(μgL−

1)

19′-

(μg

5.92

±3.68

1.27

±0.52

3.42

5.96

±3.42

2.35

±2.02

0.58

4.81

±1.16

0.32

±0.16

1.88

1.77

±0.95

0.62

±0.37

0.23

8.08

±5.16

4.49

±2.94

0.71

1.03

±0.60

0.32

±0.37

0.18

2.03

±1.44

0.48

±0.50

0.46

2.2. Remote sensing image analysis

Data from the Ocean Color Temperature Scanner (OCTS) andMod-erate Resolution Imaging Spectroradiometer (MODIS) were providedby NASA's ocean color processing group (http://oceancolor.gsfc.nasa.gov/). Level-3 global Standard Mapped Images at 9 km resolutionwere selected from the periods November 1996 to February 1997and November 2003 to February 2006. Eight-day averaged chl-a con-centrations and standard errors were extracted and binned from76.5–77.5°S, 170–180°E (Fig. 1).

prings

(Decem

ber20

03,2

004an

d20

05)an

dsu

mme

ples. SiO4

(μM)

PO4

(μM)

POC

(μM)

71.9±

4.67

1.84

±0.23

24.6±

15.1

62.8±

10.1

1.38

±0.34

30.4±

17.3

81.3±

4.07

1.56

±0.31

32.3±

8.00

78.3±

24.7

0.94

±0.17

31.0±

11.2

47.1±

11.2

1.29

±0.61

26.0±

15.6

67.0±

8.15

1.42

±0.16

6.78

±5.18

71.6±

14.3

1.56

±0.53

17.8±

10.9

2.3. Statistical analyses

2.3.1. Analysis of variance and Tukey's testOne-way analysis of variance (ANOVA, parametric) was used to

evaluate whether the mean values of the physical and biologicalvariables changed with time and depth (Addelman, 1970). The datawere divided into various groups (Table 2). All data were tested fornormality and homogeneity of variance using R (v. 2.8.1) beforeANOVA was applied. For the variables for which ANOVA indicatedthat the group means differed (i.e., pb0.05), Tukey's honestly signifi-cant difference test (Tukey's test) was used to compare the differencebetween each pair of means. The result for each pair is presented as ap-value.

Table2

Mea

nsan

dstan

dard

deviations

forph

ysical

prop

erties,n

utrien

ts,p

articu

late

matter,an

dpigm

entc

oncentration

sin

sBo

ldnu

mbe

rsindicate

asign

ificant

differen

cefrom

all/bo

thothe

rye

ars(Tuk

ey'stest;pb0.05

).n=

numbe

rof

sam

Cruises

Growingseason

nTe

mp

(°C)

Salin

ity

Z m (m)

NO3

(μM)

Decem

ber

2003

2003

–04

44−

1.05

±0.37

34.3±

0.36

20.0±

7.90

23.1±

3.64

2004

2004

–05

97−

0.59

±0.93

34.5±

0.06

22.1±

7.17

20.6±

5.83

2005

2005

–06

30−

0.91

±0.64

34.4±

0.07

19.8±

2.12

19.9±

4.87

Janu

ary/Februa

ry19

9719

96–97

112

−0.05

±0.38

34.2±

0.22

20.1±

6.18

11.3±

2.69

2004

2003

–04

61−

1.20

±0.26

34.0±

0.19

21.8±

7.20

20.7±

5.89

2005

2004

–05

47−

0.67

±0.95

34.5±

0.09

37.7±

16.4

20.1±

3.00

2006

2005

-06

154

−0.46

±0.95

34.3±

0.19

36.6±

17.1

21.4±

6.91

2.3.2. Principal components analysis (PCA)Principal components analysis (PCA) was used as a data reduction

technique to remove correlations among observed variables by trans-forming these original variables into a new set of uncorrelated vari-ables. The principal components (PCs) were derived in decreasingorder of importance, so that the first principal component accountedfor a majority of the variation in the original data, and each subse-quent component explained progressively less (Meglen, 1992). Theobjective of PCA is to determine whether the first few PCs accountfor most of the variation in the original data. If so, they can be usedto determine patterns within this set of data without loss of informa-tion. PC loadings were calculated to assess the relative contributionsof the original variables on each PC, while PC scores were derived asindicators of relationships between PCs and individual observations.Previous work has examined the relative importance of environmen-tal factors in the control of primary production and phytoplanktoncomposition using this technique (Adolf et al., 2006; Figueredo andGiani, 2009; Massolo et al., 2009). We used the following data as in-puts to the PCA: temperature (Temp), salinity (Sal), sampling depth(Z), mixed layer depth (Zm), nitrate (NO3), phosphate (PO4), silicicacid (SiO4), biogenic silica (BSi), particulate organic carbon (POC),particulate nitrogen (PN), chl-a, 19′-hexanoyloxyfucoxanthin (19′-hex), and fucoxanthin (fuco). The data were subdivided to explorethe vertical, seasonal and interannual variations and interactionsof physical and biological properties in the Ross Sea. Ordinationbi-plots were provided to decipher the patterns determined by thefirst two PCs. Because BSi concentrations were unavailable for the1996–97 cruise, we completed two parallel analyses, one that onlyincluded stations with BSi measurements, and the other that had allstations but omitted BSi as a variable. Analyses were also conductedusing depth criteria (0–25, 25–50, and >50 m depth bins).

Fig. 2. Results from OCTS and MODIS satellite estimates of surface chlorophyll a(μg L−1) binned from 76.5–77.5°S, 170–180°E for 1996–97 (black square), 2003–04(blue filled circle), 2004–05 (red triangle) and 2005–06 (grey filled circle).

138 X. Liu, W.O. Smith Jr. / Journal of Marine Systems 94 (2012) 135–144

3. Results

3.1. Chlorophyll a estimates from satellites

Satellite (OCTS and MODIS) observations of the southern Ross Seacontinental shelf indicated a pattern characterized by rapid increasesin spring, maxima in December, and declines in chl-a concentrationsafter early January, as well as substantial interannual variations inthe bloom maxima (Fig. 2). Specifically, interannual variations inthe initiation date of the bloom were noted, with 2004–05 and2005–06 increasing approximately two weeks earlier than 1997–08and 2003–04. During the season of 2003–04, a secondary accumula-tion of biomass was observed in mid-February, as noted by previousin situ and remotely sensed observations (Peloquin and Smith,2007; Smith et al., 2006, 2011a). The magnitude of the secondarypeak was comparable to that of the December bloom.

3.2. Analysis of variance (ANOVA)

Differences among water column temperatures were detectedbetween December 2003, 2004, and 2005 (Table 2, ANOVA, tempera-ture, pb0.01). Specifically, temperatures were significantly higher inDecember 2004 (mean of −0.59±0.93 °C) than those in the othertwo years. In addition, the lowest annual mean, summer mean,and maximum of chl-a concentrations were found in this year aswell (Table 2, Fig. 2). Between-year differences were also detectedin phosphate, silicic acid and 19′-hex concentrations (phosphate,pb0.0001; silicic acid, pb0.0001; 19′-hex, pb0.0001). Concentrationsof both phosphate and silicic acid were lowest in December 2004,with values of 1.38±0.34 and 62.8±10.1 μM, respectively, the yearin which significantly higher levels of fucoxanthin were observed aswell (2.35±2.02 μg L−1). Despite the lack of a significant differencein nitrate and chl-a concentrations, 19′-hex concentrations (0.58±0.76 μg L−1) in December 2004, the year in which both temperatureand salinity were higher, were significantly lower than those in theother two years (difference between 2004 and 2003, pb0.001; 2004and 2005, pb0.001).

Fig. 3. PCA ordination bi-plots for the biological and environmental variables (biogenicsilica omitted) for the Ross Sea, Antarctica, in which a) all available data were analyzed,b) only data from spring cruises (December) were analyzed, and c) only data fromsummer cruises (January/February) were analyzed. Open circles represent samplescollected from spring cruises; filled circles represent samples collected from summercruises. Temp=temperature; sal=salinity; Z=sampling depth; Zm=mixed layer depth;NO3=nitrate; PO4=phosphate; SiO4=silicic acid; BSi=biogenic silica; POC=particulateorganic carbon; PN=particulate nitrogen; 19hex=19′-hexanoyloxyfucoxanthin; fuco=fucoxanthin; Chl=chlorophyll a.

Table 3Depth-binned (between 0–25, 25–50 and>50 m) means and standard deviations for physical properties, nutrients, particulate matter, and pigment concentrations averaged overgrowing seasons of 1996–97, 2003–04, 2004–05, and 2005–06. Bold numbers indicate a significant difference from both other depth groups (pb0.05). n=number of samples.

Depth(m)

n Temperature(°C)

Salinity NO3

(μM)SiO4

(μM)PO4

(μM)POC(μM)

Chl-a(μg L−1)

Fucoxanthin(μg L−1)

19′-hex(μg L−1)

NO3 : PO4

0–25 220 −0.18±0.81 34.2±0.25 16.9±4.34 61.5±21.2 1.16±0.47 28.7±15.6 4.95±3.91 2.07±2.48 0.64±0.90 14.8±1.3125–50 129 −1.09±0.51 34.5±0.16 22.6±4.64 70.0±14.8 1.62±0.47 18.4±11.2 3.75±2.78 1.05±1.42 0.92±1.41 14.0±2.02> 50 67 −1.53±0.28 34.4±0.14 28.3±3.28 76.0±7.30 2.01±0.21 10.4±8.43 1.56±1.48 0.18±0.12 0.57±0.64 14.1±3.90

139X. Liu, W.O. Smith Jr. / Journal of Marine Systems 94 (2012) 135–144

Interannual variability was also observed among the January/February data from 1997, 2004, 2005, and 2006 (Table 2). Signifi-cantly elevated values of chl-a (8.08±5.16 μg L−1), fucoxanthin(4.49±2.94 μg L−1), and 19′-hex (0.71±0.73 μg L−1) were ob-served in February 2004, during which period silicic acid concentra-tions were substantially reduced, with mean concentration being47.1±11.2 μM. Samples collected in January 2005 were character-ized by the highest salinity and the lowest POC and pigment concen-trations encountered, despite the significantly elevated values offucoxanthin observed during the previous December.

3.3. Principal component analysis (PCA)

Seasonal and interannual variability of environmental factors andphytoplankton composition was observed, as expected (Fig. 3). Thefirst two PCs in the three analyses accounted for 60, 58 and 70%of the total variance, respectively. Original variables with positiveloadings projected on the PC axes indicate positive relationshipswith each PC, and the magnitude of the loading is indicative of theinfluence of the variables on each PC. In the first analysis all availabledata were included in the PCA, with BSi omitted as a variable (Fig. 3a).Within these data, PC1 explains 43% of the total variance and is char-acterized by positive loadings of POC and PN, temperature, chl-a andfucoxanthin concentrations, and negative loadings of inorganic nutri-ents and sampling depth (Table 4). This suggests that phytoplanktonbiomass was negatively correlated with inorganic nutrients in gener-al; positive PC1 coordinate space embraces samples collected in themost productive areas, such as waters of the Ross Sea polynya,where increased availability of irradiance initiates phytoplanktongrowth, and consequently leads to an accumulation of particulate or-ganic matter and a significant removal of nutrients. This may explainthe positive PC1 scores for part of the samples collected in December2004, during which period waters were enriched with chl-a and fuco-xanthin. A portion of summer data were also distributed in this spacedue to elevated concentrations of POC, PN and chl-a, similar to those

Table 4Factor loading matrix from the principal component analyses (first two PCs only). Bold numbwith the highest absolute values. The analysis of the entire season includes data from bothporates data from both seasons (Fig. 4a-c).

Variable BSi_omitted

Entire season Spring Summer 0–25 m 25–5

PC1 PC2 PC1 PC2 PC1 PC2 PC1 PC2 PC1

Temp 11.7 −20.9 10.1 −0.62 7.01 −20.4 0.34 −16.7 5.3Sal −6.31 −0.70 −0.19 −1.92 −7.56 −3.20 −4.75 −1.59 −2.0Z −9.74 7.31 2.72 2.78 9.44 −0.06 −1.48 6.43 −5.3Zm −4.31 −5.83 −0.09 −7.63 −3.76 −1.07 −6.30 3.20 −1.8NO3 −16.6 12.5 −10.4 −5.52 −12.8 13.2 −8.30 13.2 −7.3SiO4 −13.3 −1.90 −10.1 −3.06 −11.3 12.0 −7.42 −2.60 1.7PO4 −11.2 6.58 −6.24 1.91 −8.51 5.97 −8.20 6.98 −5.3POC 16.3 8.36 9.37 5.28 14.5 2.00 12.9 −0.21 8.3PN 11.5 5.13 −7.35 13.3 −10.9 −2.80 13.3 −0.21 9.0Chl 8.89 13.2 4.36 3.38 7.52 9.60 8.77 8.33 6.2Fuco 9.23 9.46 5.27 −0.97 7.29 9.16 8.79 6.88 4.619′-hex 0.59 20.3 −3.44 15.6 4.98 16.2 0.57 11.1 3.8BSi

collected in February 2004, the year in which an intensive secondarybloom occurred later in the growing season (Fig. 2). A much smallerloading of 19′-hex on PC1 relative to that of fucoxanthin indicates areduced contribution of P. antarctica relative to diatoms with respectto total phytoplankton biomass, and no direct correlation betweenP. antarctica and environmental variables which are dominant onPC1, such as silicic acid and PN. However, a negative correlation wasfound between 19′-hex and temperature on the basis of positive load-ing of 19′-hex and negative loading of temperature on PC2, whichaccounts for 17% of the total variance. A majority of the samples col-lected in December are located in the positive PC2 space, suggesting astrong correlation between elevated values of 19′-hex (i.e., higherbiomass of P. antarctica) and lower water column temperatures. Therelationship between temperature and fucoxanthin appears to bemore complicated, in terms of their positive correlation on PC1 andnegative correlation on PC2. Mixed layer depth and salinity explainedless variation than other original variables, as shown by their smallerloadings on both axes. This implies that the availability of irradiancemay have less influence, relative to temperature, on the variationsin phytoplankton biomass and nutrient distributions.

To specifically assess the interannual variability of physical andbiological factors in late spring and summer, data were subdividedby season. Only spring (December) or summer (January/February)data were included in the calculation of variable loadings and samplescores associated with the first two PCs. In the spring analysis(Fig. 3b), PC1 is described by positive loadings of temperature andPOC concentrations, and negative loadings of nitrate and samplingdepth, which suggests strong variations associated with vertical dis-tribution of phytoplankton biomass and nutrients. Negative correla-tion between 19′-hex and water temperature was observed on PC1,but is far less noticeable on PC2, indicating that the generallyobserved negative correlation between 19′-hex and temperaturein the entire data set (Fig. 3a) might be largely determined by pat-terns associated with seasonal variations which were minimized bydata subdivision. Variable loadings projected on PC2 are primarily

ers denote the dominant environmental variables on each PC, indicated by the loadingsspring and summer, and from all three depth bins (Figs. 3a, 5a); each depth bin incor-

BSi_included

0 m > 50 m Entire season Spring Summer

PC2 PC1 PC2 PC1 PC2 PC1 PC2 PC1 PC2

3 −7.78 2.32 −8.55 10.4 −13.7 9.86 1.37 5.70 −12.77 3.16 0.06 4.92 −6.10 −3.04 −0.17 −2.96 −6.94 −5.497 3.86 −1.55 2.54 −8.27 13.6 −10.0 −4.49 −12.0 8.177 −5.48 1.08 −6.44 −6.08 −6.41 0.26 −8.58 −5.54 0.109 6.85 −5.49 5.48 −15.9 7.87 −10.0 2.61 14.1 1.473 2.26 1.88 0.54 −15.4 0.22 −4.14 3.73 7.06 14.15 7.01 −2.97 3.05 −9.85 11.5 −5.87 −2.68 6.93 13.50 5.77 7.68 2.62 15.7 4.94 9.41 7.60 4.95 8.809 3.51 7.87 1.90 6.20 4.82 2.50 4.89 8.23 −0.394 5.69 5.42 2.52 8.47 14.2 5.10 3.73 7.06 14.19 0.06 1.20 −1.20 8.90 12.2 −3.80 1.37 5.70 −12.78 8.95 7.15 3.81 0.24 −3.04 −0.17 −2.96 −6.94 −5.49

17.3 1.88 11.0 −2.75 13.9 2.14

140 X. Liu, W.O. Smith Jr. / Journal of Marine Systems 94 (2012) 135–144

determined by interannual variations, as suggested by the deeplymixed and nitrate enriched surface waters sampled in December2004 and silicic acid and 19′-hex enriched waters collected in Decem-ber 2003 and 2005. Although no apparent difference was detected onPC1 between the three years for which December data were available,a portion of the data collected in December 2004 exhibited highertemperatures, chl-a and fucoxanthin concentrations than others, sug-gested by the adjacency of these individual observations and originalvariables on the ordination bi-plot. December data from 2003 and2005 fall into the same quadrant with 19′-hex, silicic acid and phos-phate in the coordinate plate, showing that the phytoplankton assem-blage during those periods was dominated by P. antarctica, despitelower biomass present, and that water samples were enriched (rela-tive to nitrate) with silicic acid and phosphate.

Compared to spring, the correlations among fucoxanthin, 19′-hexand chl-a concentrations in summer appeared to be stronger (Fig. 3c).The variation of chl-a concentrations was tightly correlated with thatof fucoxanthin, as suggested by their comparable loadings on bothaxes. A negative correlation between 19′-hex and temperature wasobserved on PC2 but not on PC1, which may suggest that the varia-tions of temperature, and consequently of 19′-hex, are not directand maybe associated with both spatial and vertical distributions.Data from January 2004 and 2006 exhibited strong variability onPC1, which represent samples with a wide range of POC, PN and ni-trate concentrations. Despite the complexity, differences were foundbetween these two years, in that a majority of 2004 data have positivescores on PC2, which can be explained by higher values of nitrate and19′-hex, while a significant portion of 2006 data have negative PC2scores, which indicate higher temperatures but lower pigmentconcentrations.

To reduce the effects associated with vertical variability of biolog-ical and physical factors in the water column, data were subdividedby depth (0–25, 25–50, and>50 m, corresponding to light-saturating, light-limiting, and sub-euphotic depths; Fig. 4). Comparedto the analyses with pooled depth data, the relative contributions ofPC1 are lower, with only 36, 30 and 36% of the variance accountedfor, respectively. This suggests that the significant importance ofPC1 in previous analyses may be attributed to the strong variabilityassociated with vertical distribution of phytoplankton biomass andnutrients. Within the shallowest depth bin, the pattern of variableloadings and sample scores (Fig. 4a) are consistent with those of theentire data set (Fig. 3a). This suggests that the studied parametersexhibited stronger variability in the upper 25 m than that at greaterdepths, and the magnitude and direction of changes likely dominatedthe water column. At depths greater than 25 m, silicic acid displayedreduced variability on both PC1 and PC2; decoupling also appears onPC2 between fucoxanthin and the cluster of chl-a, 19′-hex, POC andPN (Fig. 4b and c), indicating that the role that diatoms play in totalphytoplankton biomass and nutrient distributions decreased marked-ly with depth. The inverse correlation between 19′-hex and tempera-ture is less obvious at light-limiting depths (Fig. 4b) than that withinthe surface bin, and nearly disappears at sub-euphotic depths(Fig. 4c).

In the analyses of all available data with biogenic silica (BSi)included (2003–2006), the complete data set was first analyzed(Fig. 5a), and then subdivided by season (Fig. 5b and c) and depth(figures not shown) according to the same criteria from the previousanalyses. High loadings of BSi on PC1, and the adjacency between BSi

Fig. 4. PCA ordination bi-plots for the biological and environmental variables for theRoss Sea, Antarctica, in which data collected from (a) 0–25 m; (b) 25–50 m; (c)>50 mwere analyzed. Open circles represent samples collected from spring cruises; filled circlesrepresent samples collected from summer cruises. Temp=temperature; sal=salinity;Z=sampling depth; Zm=mixed layer depth; NO3=nitrate; PO4=phosphate; SiO4=silicic acid; BSi=biogenic silica; POC=particulate organic carbon; PN=particulatenitrogen; 19hex=19′-hexanoyloxyfucoxanthin; fuco=fucoxanthin; Chl=chlorophyll a.

141X. Liu, W.O. Smith Jr. / Journal of Marine Systems 94 (2012) 135–144

and POC were observed, suggesting that the variation of accumulatedbiomass during the study period was largely attributable to diatoms.Addition of BSi reduced the loadings of 19′-hex on PC2, indicatingthe relatively smaller variation of 19′-hex compared to BSi. A tightnegative correlation was observed on PC1 between BSi and silicicacid when the complete data set was analyzed (Fig. 5a), which sug-gests that biogenic silica in the water column might be generated di-rectly by biological removal of silicic acid rather than other processessuch as lateral advection. In the spring analysis, BSi accumulation wasuncorrelated with salinity and mixed layers, but positively related totemperature (Fig. 5b), while in the summer, it was more strongly cor-related, although inversely, with depth, nitrate, and salinity than withtemperature (Fig. 5c). Patterns of other variable loadings are general-ly consistent to those within the entire dataset with BSi excluded.

4. Discussion

Phytoplankton assemblages of the Ross Sea are generally dominat-ed by two functional groups: haptophytes (Phaeocystis antarctica)and diatoms. Substantial taxonomic heterogeneity exists temporallyand spatially, although P. antarctica generally accumulates in australspring in the south central portion of Ross Sea polynya, whereas dia-toms usually dominate along the coast of Victoria Land and near iceedges in the summer (Arrigo et al., 1999; DiTullio and Smith, 1996;Garrison et al., 2003; Smith and Asper, 2001; Smith et al., 2010). Ithas been hypothesized that this general pattern of heterogeneity iscontrolled by a combination of environmental and biological factors.Arrigo et al. (1999) and Smith and Asper (2001) suggested that dee-per mixed layers favored the growth of P. antarctica, whereas shal-lower ones were often dominated by diatoms, a difference whichwas primarily driven by the difference in efficiency of light harvest-ing. Smith et al. (2010) generated a seasonal “climatology” of pigmentdistribution in the Ross Sea, and confirmed the general seasonal pro-gression in which diatom accumulation occurs in the later growingseason after spring blooms of P. antarctica, and suggested thatbiomass of all functional groups in summer is primarily limited bymicronutrient (iron) supplies. It has also been suggested that ironlimitation may constrain P. antarctica colony formation and result inan increase in the contribution of solitary forms during summer, de-spite the fact that colonial cell numbers were always greater thanthose of solitary cells during the growing season (Smith et al., 2003b).

4.1. Interannual and seasonal variability

As expected, seasonal differences were detected in water columntemperatures, nutrient concentrations, and pigment concentrationsbetween austral spring and summer (Table 2). Seasonal averages ofpigments confirmed that P. antarctica blooms dominated the phyto-plankton field in the southern Ross Sea through December, consistentwith previous observations and the “climatology” of pigment distri-butions (e.g., Arrigo et al., 1999; El-Sayed et al., 1983; Smith andGordon, 1997; Smith et al., 2010). The pigment climatology (Smithet al., 2010) suggested that standing stocks of diatoms in summerwere generally much lower than those of P. antarctica in spring.However, a secondary diatom bloom, which was characterized bysignificantly elevated fucoxanthin concentrations, was found in Feb-ruary 2004, with chl-a maxima exceeding the spring peak observed

Fig. 5. PCA ordination bi-plots for the biological and environmental variables for theRoss Sea, Antarctica (1996–97 data omitted), in which a) all available data wereanalyzed; b) data from spring cruises were analyzed; c) data from summer cruiseswere analyzed. Open circles represent samples collected from spring cruises; filledcircles represent samples collected from summer cruises. Temp=temperature; sal=salinity; Z=sampling depth; Zm=mixed layer depth; NO3=nitrate; PO4=phosphate;SiO4=silicic acid; BSi=biogenic silica; POC=particulate organic carbon; PN=particulatenitrogen; 19′-hex=19hexanoyloxyfucoxanthin; fuco=fucoxanthin; Chl=chlorophyll a.

142 X. Liu, W.O. Smith Jr. / Journal of Marine Systems 94 (2012) 135–144

in December 2003 (Smith et al., 2006, 2010). This year had substan-tially different physical properties (i.e., salinity and temperature),nutrients, and pigment concentrations, with reduced temperature,salinity and silicic acid concentrations through both December andFebruary, and elevated values of pigments in February (Table 2);however, the direct environmental forcing of this exceptional diatomgrowth has not yet been elucidated.

Previous studies suggested that diatoms have lower iron half sat-uration constants relative to P. antarctica (Coale et al., 2003); conse-quently, summer diatom blooms may be enhanced by intrusions ofa warmer water mass-modified circumpolar deep water (MCDW),which could subsequently introduce iron into surface layers(Hiscock and Millero, 2005; Peloquin and Smith, 2007). MCDW ischaracterized by being warmer and more saline, and has been consid-ered a potential source of iron to the Ross Sea (Carmack and Aagaard,1977). Although there is no evidence of an intrusion of MCDW intothe surface waters based on seasonally averaged temperature andsalinity during season of 2003–04 (Table 2; 2003–04 had reducedtemperatures and salinity rather than elevated ones), we do not ex-clude the possibility of short-term intrusions or advective eventswhich were under-sampled during field studies, but were capturedby continuous in situ fluorescence and CTD measurements in a morerecent study (Smith et al., 2011a). Assuming that growth of diatomsin summer is controlled by iron, alternatively, it may be inferredthat the substantial growth of diatoms in February 2004 was initiatedby iron supply from lateral advection in addition to ice melting, pro-cesses which could result in reduced salinity in the surface waters.Another explanation is that stratification was intensified by enhancedsurface freshwater inputs in late December of 2003; consequently,diatoms, with a photosynthetic “advantage” under high irradianceconditions (Kropuenske et al., 2009), grew rapidly and accumulatedlater in the season. This hypothesis does not, however, explain howthe elevated iron concentrations necessary for the observed large bio-mass were introduced. December 2004 deviated substantially fromthe climatological means as well, in that it was a period in whichaverage fucoxanthin concentrations (mean of 2.35±2.02 μg L−1;Table 2) were substantially higher than 19′-hex, indicating an earliertransition from P. antarctica to diatom-dominated assemblages in thatyear. Significantly reduced silicic acid concentrations (mean of 62.8±10.1 μg L−1), compared to the other two years, were also notedduring this period. December 2004 was also characterized by the low-est phosphate concentrations, the highest NO3 : PO4 ratios, and thelowest 19′-hex concentrations, despite a ‘normal’ mean value ofchl-a (5.96±3.42 μg L−1). This exception from the climatologicalmeans was associated with warmer and yet deeply mixed water col-umns; therefore, we suggest that the elevated temperature, ratherthan stronger stratification, played a significant role on diatom selec-tion. Alternatively, diatom growth could have been seeded by meltingice, resulting from the substantial diatom accumulation in February2004, the previous summer.

The maximum biomass observed in the four years analyzed wasobserved during the year in which the bloom was initiated latest —1996. Conversely, the smallest maximum was observed during2004, the year that exhibited the most elevated chlorophyll concen-trations earliest. Arrigo et al. (1998) argued that bloom initiation(and demise) was regulated by processes regulating polynya forma-tion, in that early polynya formation (caused by extensive winterwinds breaking the ice cover and advecting it northward) resultedin deeper mixed layers and a delay in bloom initiation. Conversely,a later bloom initiation resulted in greater maximum biomass dueto seasonally decreased winds and lower advective removal. Ourdata appears to be consistent with these hypotheses, in that the max-imum biomass and latest bloom initiation were both observed in1996. However, the relationship was not completely coherent in allyears, in that in 2003 was initiated at a similar time as the bloom in1996, but failed to reach the same biomass; similarly, in 2005 the

bloom was initiated early (similar to 2004), but attained a far greaterbiomass. Our analysis did not include ice concentrations, but a com-plete analysis of the relationship of polynya formation and bloomdynamics using satellite data might provide evidence of the role ofenvironmental variables in regulating blooms in the southern RossSea.

Temporal patterns of phytoplankton distributions can be seenfrom PCA results as well (Fig. 3a). Besides February 2004, an over-whelming majority of samples collected in December show positivescores on PC2, which are primarily explained by variations of 19′-hex and water column temperature, indicating warmer waters andthe dominance of P. antarctica in spring. Interestingly, data collectedin February 2004 substantially deviated from the cluster of mixedlayer depth and January/February data of all other years, indicatingstronger stratification during February 2004, similar to the compari-sons of variable means. Although it has been suggested that deeplymixed layers favor the growth of P. antarctica (Arrigo et al., 1999;Kropuenske et al., 2009), the inverse correlation between 19′-hexconcentrations and mixed layers was not stronger than that betweenfucoxanthin and mixed layers, consistent with the findings of vanHilst and Smith (2002) who concluded that multiple factors regulatethe phytoplankton assemblage in the Ross Sea rather than simplyirradiance. Indeed, mixed layer depth had a relatively minor contribu-tion to the total variance, suggesting that other factors (such astemperature) may have played a stronger role in controlling thetaxonomic composition of the assemblage.

4.2. Vertical variability

The vertical variability within the Ross Sea is appreciable. ANOVAclearly confirmed the expected decreases in temperature, phyto-plankton biomass and chl-a concentrations, and increases in inorgan-ic nutrients with increased depth (Table 3). However, 19′-hexaccumulated at the medium depth bin (25–50 m; correspondingto the light-limiting environment, albeit with enough light to allowpositive growth; Kropuenske et al., 2009) and well adapted to lowlevels of irradiance present. Although 19′-hex concentrations werereduced at depths greater than 50 m, the ratios of 19′-hex to chl-aincreased (12.9, 24.5, and 36.5% corresponding to 0–25, 25–50 and>50 m depth bins, respectively), which suggests that the contribu-tion of P. antarctica to total biomass were greater at depth. Waterlayers dominated by diatoms and P. antarctica displayed distinctnutrient drawdown characteristics (Table 3). Disappearance ratiosof NO3 : PO4 (ΔNO3 : ΔPO4, relative to values at depths greater than50 m) at the shallowest and medium depth bins were 13.4 and 14.6,respectively, in response to the different Redfield uptake ratios ofdiatom-dominated and the two taxa mixed assemblages. Patterns ofmultiple relationships among variables differed with depth as well,as demonstrated from the PCA analyses. Chl-a, POC and PN show pos-itively higher loadings on PC1 in the individual analyses of all thethree depth bins (Fig. 4); however, their variations were less con-trolled by the distribution of fucoxanthin than by 19′-hex at depth,when compared to the surface bin. This can be explained by the rela-tively reduced contribution of diatoms (and relatively elevated con-tribution of P. antarctica) to total biomass with depth. Additionally,loadings of phosphate on PC1 decreased with depth, indicating rela-tively less removal of phosphate compared to nitrate. This was corre-lated with an increased contribution of P. antarctica, and is consistentto pooled depth data (Table 3) and other studies in which lower NO3 :PO4 ratios were observed in diatoms relative to P. antarctica (Arrigoet al., 2000; Dunbar et al., 2003).

4.3. Relationship between temperature and P. antarctica

Although the relationship between temperature and P. antarcticacannot be quantitatively deduced from data solely from four years,

143X. Liu, W.O. Smith Jr. / Journal of Marine Systems 94 (2012) 135–144

our statistical analyses do suggest an inverse correlation between P.antarctica biomass and water column temperature during the grow-ing seasons. This relationship was reflected both spatially and tempo-rally. In December, mean temperature was lowest in 2003 andhighest in 2004 (Table 2), and was associated with highest and lowestvalues, respectively, of 19′-hex observed during the same growingseasons. Despite the exceptional relative dominance of diatoms,absolute concentrations of 19′-hex were remained elevated in theFebruary 2004, during which period the lowest water column tem-peratures were detected as well. Similarly, PCA analyses demonstratea strong correlation between P. antarctica and temperature. Com-pared to the analysis of the entire data set (Fig. 3a), data from theupper 25 m exhibited a substantially reduced loading of samplingdepth on PC1 (Fig. 4a), which suggests that the variation of PC1 isstrongly associated with vertical distributions of phytoplankton bio-mass and nutrients. However, 19′-hex and temperature are inverselycorrelated on PC2 but uncorrelated on PC1 (Fig. 3a), which indicatesthat the role of water column temperature on P. antarctica biomasswas less likely determined by vertical patterns than by seasonal pat-terns (i.e., P. antarctica blooms dominated the phytoplankton fieldthrough December when it was colder, and diatoms accumulated insummer as it warms), as suggested by reduced negative correlationin individual seasonal analyses where seasonal variations were ex-cluded (Fig. 3b, c). It is also possible that the temperature controlsof P. antarctica distribution vary spatially, meaning that biomass ofP. antarctica accumulates in regions which are generally colder,whereas diatoms are relatively more abundant and accumulate in re-gions that are warmer. However, the Ross Sea is historically known tohave two taxonomically and spatially distinct regimes: the south-central polynya (with relatively deeply mixed layers and dominatedby haptophytes) and the western portions (with shallower mixedlayers and dominated by diatoms; DiTullio and Smith, 1996; Arrigoet al., 1999; Smith and Asper, 2001; Garrison et al., 2003; Smith etal., 2010); since the stations visited during our cruises were primarilyoccupied in the south-central portion, we believe that the P. antarcticaand temperature correlation that we found was less likely driven byphysical heterogeneity at regional scale than by seasonality.

5. Implications

Phytoplankton distributions in the southern Ross Sea exhibit sub-stantial interannual, seasonal and spatial variability, as well as vari-ability with respect to bloom timing, biomass and composition.Environmental factors that control phytoplankton biomass and com-position are often correlated with each other, difficult to separate andcomplex. Our multivariate analysis approach deciphered the patternsof correlations among observed variables, and suggested a significantrole of water temperature as a control of haptophyte and diatom dis-tributions in the Ross Sea. We also detected a large diatom bloom inFebruary 2004, which was considered as an example of deviationsfrom the climatological means. We attribute this observed bloom tointensified stratification or iron inputs supplied by short-term physi-cal forcing. Changes in climatic conditions have modified a varietyof oceanographic properties, such as sea surface temperatures andwater column stability, which we showed to be highly correlatedwith the variability in phytoplankton composition in the Ross Sea. Be-cause this composition plays a critical role in food web structure andbiogeochemical cycles (e.g., diatoms are readily grazed and theirenergy is transferred to higher trophic levels, whereas P. antarcticacolonies contribute to carbon flux via passive sinking), our findingsrepresent a significant contribution to current understanding ofclimate-change-associated impacts on phytoplankton distribution inthe Ross Sea, and thus provide insights into the biogeochemical pro-cesses that are occurring and potential future changes which are rea-sonably expected in the near future.

Acknowledgments

This research was supported by NSF grants OPP 0087401 and0338157 to WOS. Helpful comments on an earlier draft of this manu-script were provided by M. Friedrichs from VIMS. We acknowledgethe invaluable help from the crew on the research vessel RVIB N.B.Palmer and USCGC Polar Star. We also thank Amy Shields and JenniferDreyer who assisted laboratory analysis. This is VIMS contributionNumber XXXX.

References

Addelman, S., 1970. Variability of treatments and experimental units in the design andanalysis of experiments. J. Amer. Stat. Assoc. 65, 1095–1108.

Adolf, J.E., Yeager, C.L., Miller, W.D., Mallonee, M.E., Harding, L.W., 2006. Environmentalforcing of phytoplankton floral composition, biomass, and primary productivity inChesapeake Bay, USA. Estuar. Coast. Shelf Sci. 67, 108–122.

Arrigo, K.R., van Dijken, G.L., 2004. Annual changes in sea-ice, chlorophyll a, and prima-ry production in the Ross Sea, Antarctica. Deep-Sea Res. II 51, 117–138.

Arrigo, K.R., Weiss, A.M., Smith Jr., W.O., 1998. Physical forcing of phytoplanktondynamics in the southwestern Ross Sea. J. Geophys. Res. 103, 1007–1021.

Arrigo, K.R., Robinson, D.H., Worthen, D.L., Dunbar, R.B., DiTullio, G.R., van Woert, M.,Lizotte, M.P., 1999. Phytoplankton community structure and the drawdown ofnutrients and CO2 in the Southern Ocean. Science 283, 365–367.

Arrigo, K.R., DiTullio, G.R., Dunbar, R., Robinson, D., van Woert, M., Worthen, D., Lizotte,M., 2000. Phytoplankton taxonomic variability in nutrient utilization and primaryproduction in the Ross Sea. J. Geophys. Res. 105, 8827–8846.

Beaugrand, G., Brander, K.M., Lindley, J.A., Souissi, S., Reid, P.C., 2003. Plankton effect oncod recruitment in the North Sea. Nature 426, 661–664.

Blackford, J.C., 2010. Predicting the impacts of ocean acidification: challenges from anecosystem perspective. J. Mar. Syst. 81, 12–18.

Caldeira, K., Wickett, M.E., 2003. Anthropogenic carbon and ocean pH. Nature 425, 365.Carmack, E.C., Aagaard, K., 1977. Volumetric considerations of upwelling in Benguela

current. Estuar Coast Mar Sci 5, 135–142.Coale, K.H., Wang, X.J., Tanner, S.J., Johnson, K.S., 2003. Phytoplankton growth and bio-

logical response to iron and zinc addition in the Ross Sea and Antarctic Circumpo-lar Current along 170 degrees W. Deep-Sea Res II 50, 635–653.

DiTullio, G.R., Smith Jr., W.O., 1996. Spatial patterns in phytoplankton biomass and pig-ment distributions in the Ross Sea. J. Geophys. Res. 101, 18467–18477.

DiTullio, G.R., Geesey, M.E., Leventer, A.R., Lizotte, M.P., 2003. Algal pigment ratiosin the Ross Sea: implications for CHEMTAX analysis of southern ocean data. In:DiTullio, G., Dunbar, R. (Eds.), Biogeochemistry of the Ross Sea. : Ant. Res. Ser.,78. American Geophysical Union, Washington, DC, pp. 279–293.

Dunbar, R., Arrigo, K.R., Lutz, M., DiTullio, G.R., Leventer, A.R., Lizotte, M., van Woert, M.,Robinson, D., 2003. Non-Redfield production and export of marine organic matter:a recurrent part of the annual cycle in the Ross Sea, Antarctica. In: DiTullio, G.,Dunbar, R. (Eds.), Biogeochemistry of the Ross Sea. : Ant. Res. Ser., 78. AmericanGeophysical Union, Washington, DC, pp. 179–196.

Edwards, M., Richardson, A.J., 2004. Impact of climate change on marine pelagicphenology and trophic mismatch. Nature 430, 881–884.

El-Sayed, S.Z., Biggs, D.C., Holmhansen, O., 1983. Phytoplankton standing crop, primaryproductivity, and near-surface nitrogenous nutrient fields in the Ross Sea, Antarc-tica. Deep-Sea Res I 30, 871–886.

Feng, Y., Warner, M.E., Zhang, Y., Sun, J., Fu, F., Rose, J.M., Hutchins, D.A., 2008. Interac-tive effects of increased pCO2, temperature and irradiance on the marine cocco-lithophore Emiliania huxleyi (Prymnesiophyceae). Eur. J. Phycol. 43, 87–98.

Feng, Y., Hare, C.E., Rose, J.M., Handy, S.M., DiTullio, G.R., Lee, P.A., Smith Jr., W.O.,Peloquin, J., Tozzi, S., Sun, J., Zhang, Y., Dunbar, R.B., Long, M.C., Sohst, B., Lohan,M., Hutchins, D.A., 2010. Interactive effects of iron, irradiance and CO2 on RossSea phytoplankton. Deep-Sea Res. I 57, 368–383.

Figueredo, C.C., Giani, A., 2009. Phytoplankton community in the tropical lake of LagoaSanta (Brazil): conditions favoring a persistent bloom of Cylindrospermopsis raci-borskii. Limnol. 39, 264–272.

Fraser, W.R., Hofmann, E.E., 2003. A predator's perspective on causal links between cli-mate change, physical forcing and ecosystem response. Mar. Ecol. Prog. Ser. 265,1–15.

Fu, F., Warner, M.E., Zhang, Y., Feng, Y., Hutchins, D.A., 2007. Effects of increased tem-perature and CO2 on photosynthesis, growth, and elemental ratios in marine Syne-chococcus and Prochlorococcus (Cyanobacteria). J. Phycol. 43, 485–496.

Fu, F., Mulholland, M.R., Garcia, N.S., Beck, A., Bernhardt, P.W., Warner, M.E., Sanudo-Wilhelmy, S.A., Hutchins, D.A., 2008. Interactions between changing pCO2, N2 fixa-tion, and Fe limitation in the marine unicellular cyanobacterium Crocosphaera.Limnol. Oceanogr. 53, 2472–2484.

Garrison, D.L., Jeffries, M.O., Gibson, A., Coale, S.L., Neenan, D., Fritsen, C., Okolodkov,Y.B., Gowing, M.M., 2003. Development of sea ice microbial communities duringautumn ice formation in the Ross Sea. Mar. Ecol. Prog. Ser. 259, 1–15.

Gordon, L.I., Codispoti, L.A., Jennings, J.C.,Millero, F.J.,Morrison, J.M., Sweeney, C., 2000. Sea-sonal evolution of hydrographic properties in the Ross Sea, Antarctica, 1996–1997.Deep-Sea Res II 47, 3095–3117.

Hiscock, W.T., Millero, F.J., 2005. Nutrient and carbon parameters during the SouthernOcean iron experiment (SOFeX). Deep-Sea Res. I 52, 2086–2108.

Hutchins, D.A., Fu, F., Zhang, Y., Warner, M.E., Feng, Y., Portune, K., Bernhardt, P.W.,Mulholland, M.R., 2007. CO2 control of Trichodesmium N2 fixation, photosynthesis,

144 X. Liu, W.O. Smith Jr. / Journal of Marine Systems 94 (2012) 135–144

growth rates, and elemental ratios: implications for past, present, and future oceanbiogeochemistry. Limnol. Oceanogr. 52, 1293–1304.

Hutchins, D.A., Mulholland, M.R., Fu, F., 2009. Nutrient cycles and marine microbes in aCO2-enriched ocean. Oceanogr. 22, 128–145.

Kropuenske, L.R., Mills, M.M., van Dijken, G.L., Bailey, S., Robinson, D.H., Welschmeyer,N.A., Arrigo, K.R., 2009. Photophysiology in two major Southern Ocean phytoplank-ton taxa: photoprotection in Phaeocystis antarctica and Fragilariopsis cylindrus.Limnol. Oceanogr. 54, 1176–1196.

Levitus, S., Antonov, J.I., Boyer, T.P., Stephens, C., 2000. Warming of the world ocean.Science 287, 2225–2229.

Massolo, S., Messa, R., Rivaro, P., Leardi, R., 2009. Annual and spatial variations of chem-ical and physical properties in the Ross Sea surface waters (Antarctica). Cont. ShelfRes. 29, 2333–2344.

Meglen, R.R., 1992. Examining large databases — a chemometric approach using prin-cipal component analysis. Mar. Chem. 39, 217–237.

Montes-Hugo, M., Doney, S.C., Ducklow, H.W., Fraser, W., Martinson, D., Stammerjohn,S.E., Schofield, O., 2009. Recent changes in phytoplankton communities associatedwith rapid regional climate change along the western Antarctic Peninsula. Science323, 1470–1473.

Nelson, D.M., Brzezinski, M.A., 1990. Kinetics of silicic-acid uptake by natural diatomassemblages in 2 gulf-stream warm-core rings. Mar. Ecol. Prog. Ser. 62, 283–292.

Orr, J.C., Fabry, V.J., Aumont, O., Bopp, L., Doney, S.C., Feely, R.A., Gnanadesikan, A.,Gruber, N., Ishida, A., Joos, F., Key, R.M., Lindsay, K., Maier-Reimer, E., Matear, R.,Monfray, P., Mouchet, A., Najjar, R.G., Plattner, G.K., Rodgers, K.B., Sabine, C.L.,Sarmiento, J.L., Schlitzer, R., Slater, R.D., Totterdell, I.J., Weirig, M.F., Yamanaka, Y.,Yool, A., 2005. Anthropogenic ocean acidification over the twenty-first centuryand its impact on calcifying organisms. Nature 437, 681–686.

Parkinson, C.L., 2002. Trends in the length of the Southern Ocean sea-ice season,1979–99. Annals Glaciol. 34, 435–440.

Peloquin, J.A., Smith Jr., W.O., 2007. Phytoplankton blooms in the Ross Sea, Antarctica:interannual variability in magnitude, temporal patterns, and composition. J. Geo-phys. Res. 112, C08013. doi:10.1029/2006JC003816.

Rose, J.M., Feng, Y., DiTullio, G.R., Dunbar, R.B., Hare, C.E., Lee, P.A., Lohan, M., Long, M.,Smith Jr., W.O., Sohst, B., Tozzi, S., Zhang, Y., Hutchins, D.A., 2009. Synergistic ef-fects of iron and temperature on Antarctic phytoplankton and microzooplanktonassemblages. Biogeosciences 6, 3131–3147.

Smith Jr., W.O., Asper, V.L., 2001. The influence of phytoplankton assemblage composi-tion on biogeochemical characteristics and cycles in the southern Ross Sea,Antarctica. Deep-Sea Res. I 48, 137–161.

Smith Jr., W.O., Gordon, L.I., 1997. Hyperproductivity of the Ross Sea (Antarctica)polynya during austral spring. Geophys. Res. Lett. 24, 233–236.

Smith Jr., W.O., Marra, J., Hiscock, M.R., Barber, R.T., 2000. The seasonal cycle of phyto-plankton biomass and primary productivity in the Ross Sea, Antarctica. Deep-SeaRes. II 47, 3119–3140.

Smith Jr., W.O., Dennett, M.R., Mathot, S., Caron, D.A., 2003a. The temporal dynamics ofthe flagellated and colonial stages of Phaeocystis antarctica in the Ross Sea. Deep-Sea Res. II 50, 605–617.

Smith Jr., W.O., Dinniman, M.S., Klinck, J.M., Hofmann, E., 2003b. Biogeochemical clima-tologies in the Ross Sea, Antarctica: seasonal patterns of nutrients and biomass.Deep-Sea Res. II 50, 3083–3101.

Smith Jr., W.O., Shields, A.R., Peloquin, J.A., Catalano, G., Tozzi, S., Dinniman, M.S., Asper,V.L., 2006. Interannual variations in nutrients, net community production, and bio-geochemical cycles in the Ross Sea. Deep-Sea Res. II 53, 815–833.

Smith Jr., W.O., Ainley, D.G., Cattaneo-Vietti, R., 2007. Trophic interactions within theRoss Sea continental shelf ecosystem. Phil. Trans. R. Soc. Lond. B 362, 95–111.

Smith Jr., W.O., Dinniman, M.S., DiTullio, G.R., Tozzi, S., Mangoni, O., Modigh, M.,Saggiomo, V., 2010. Phytoplankton photosynthetic pigments in the Ross Sea: pat-terns and relationships among functional groups. J. Mar. Systems 82, 177–185.

Smith Jr., W.O., Asper, V., Tozzi, S., Liu, X., Stammerjohn, S.E., 2011a. Surface layer var-iability in the Ross Sea, Antarctica as assessed by in situ fluorescence measure-ments. Prog. Oceanogr. 88, 28–45.

Smith Jr., W.O., Shields, A.R., Dreyer, J.C., Peloquin, J.A., Asper, V., 2011b. Interannualvariability in vertical export in the Ross Sea: magnitude, composition, and environ-mental correlates. Deep-Sea Res. I 58, 147–159.

Stammerjohn, S.E., Martinson, D.G., Smith, R.C., Yuan, X., Rind, D., 2008. Trends inAntarctic annual sea ice retreat and advance and their relation to El Nino-Southern Oscillation and Southern Annular Mode variability. J. Geophys. Res. 113,C03S90. doi:10.1029/2007JC004269.

Sweeney, C., Smith Jr., W.O., Hales, B., Bidigare, R.R., Carlson, C.A., Codispoti, L.A.,Gordon, L.I., Hansell, D.A., Millero, F.J., Park, M.O., Takahashi, T., 2000. Nutrientand carbon removal ratios and fluxes in the Ross Sea, Antarctica. Deep-Sea Res. II47, 3395–3421.

Tremblay, J.-E., Smith Jr., W.O., 2007. Phytoplankton processes in polynyas. In: SmithJr., W.O., Barber, D.G. (Eds.), Polynyas: Windows to the World's Oceans. Elsevier,Amsterdam, pp. 239–270.

van Hilst, C.M., Smith Jr., W.O., 2002. Photosynthesis/irradiance relationships in theRoss Sea, Antarctica, and their control by phytoplankton assemblage compositionand environmental factors. Mar. Ecol. Prog. Ser. 226, 1–12.

Vaughan, D.G., Marshall, G.J., Connolley, W.M., Parkinson, C., Mulvaney, R., Hodgson,D.A., King, J.C., Pudsey, C.J., Turner, J., 2003. Recent rapid regional climate warmingon the Antarctic Peninsula. Clim. Change 60, 243–274.