-
MARINE ECOLOGY PROGRESS SERIESMar Ecol Prog Ser
Vol. 497: 5167, 2014doi: 10.3354/meps10602
Published February 5
INTRODUCTION
Anthropogenic eutrophication of estuaries haslong been
recognized as a significant ecologicalproblem (Nixon 1995, Howarth
& Marino 2006). Nu -trients from agriculture, wastewater
treatment facil-ities, urban runoff, and septic systems can
stimulatephytoplankton blooms, which sink and enhance bio-logical
oxygen demand in deeper waters (Smith etal. 2003, Diaz &
Rosenberg 2008). Without physical
ventilation, water at depth becomes hypoxic or evenanoxic,
spatially compressing habitats for aerobicorganisms, including
commercially important finfishand shellfish (Howell & Simpson
1994). Coastal eu -trophication is a global concern, and many
estuaries,such as Chesapeake Bay, San Francisco Bay, andthose of
the Delaware, Neuse, Seine, and St. Law -rence Rivers, suffer from
seasonal hypoxia linked toanthropogenic nutrient loadings (Diaz
& Rosenberg2008).
Inter-Research 2014 www.int-res.com*Corresponding author:
[email protected]
Phytoplankton assemblage changes during decadal decreases in
nitrogen loadings to theurbanized Long Island Sound estuary,
USA
Elizabeth A. Suter1,*, Kamazima M. M. Lwiza1, Julie M. Rose2,
Christopher Gobler1, Gordon T. Taylor1
1School of Marine & Atmospheric Sciences, Stony Brook
University, Stony Brook, New York 11794-5000, USA2NOAA Fisheries,
Northeast Fisheries Science Center, Milford Laboratory, Milford,
Connecticut 06460, USA
ABSTRACT: Despite reductions in nitrogen loadings from
wastewater treatment plants (WWTPs)discharging into Long Island
Sound (LIS) over the last 15 yr, eutrophication and hypoxia remain
asevere problem. Here we used time series of hydrography,
meteorology, nutrients, and phyto-plankton pigments to explore the
relationships between planktonic biomass, nutrient stocks,
andphysical regimes in LIS. With the exception of the most
eutrophied station in the west, dissolvedinorganic nitrogen (DIN)
decreased between 1995 and 2009, likely resulting from
WWTPupgrades. However, total dissolved nitrogen increased during
this period, primarily driven by ris-ing organic nitrogen pools.
Simultaneous increases in inorganic phosphorus, silicate, and
chloro-phyll a (chl a) were also observed. Starting in 2002,
pigment-based phytoplankton communitycomposition revealed
systematic declines in diatom abundances coincident with increases
indinoflagellates and other flagellated phytoplankton groups.
Despite this, bottom water dissolvedoxygen concentrations did not
improve. The apparent paradox between increasing DIN limitationand
escalating chl a concentrations in LIS suggests a shifting nutrient
stoichiometry and an alteredphytoplankton community in which
phytoflagellates have increased in abundance relative todiatoms.
Despite these changes, diatoms remained the most abundant algal
group by the end ofthe study. In addition, a shift in chl a stocks
in the year 2000 coincided with decreases in temper-ature,
increases in salinity, and the proliferation of several algal
groups. These results reveal thecomplex nature of eutrophied
estuaries and indicate that policies targeting only inorganic
nitro-gen loadings may be insufficient to mitigate eutrophication
in systems such as LIS.
KEY WORDS: Nitrogen limitation Nutrients Eutrophication
Phytoplankton Nutrient ratio Hypoxia
Resale or republication not permitted without written consent of
the publisher
FREEREE ACCESSCCESS
-
Mar Ecol Prog Ser 497: 5167, 2014
Primary productivity is frequently limited by theavailability of
inorganic nutrients. In the 1970s, stud-ies revealed that
phosphorus (P) runoff was the pri-mary cause of eutrophication in
lakes, and legislationlimiting P loads soon followed (Howarth &
Marino2006). More recently, scientists and managers identi-fied
nitrogen (N) as the primary limiting factor forphytoplankton growth
in the coastal and estuarineenvironment, and concluded that
reductions in totalN loadings were needed to control
phytoplanktonblooms (Nixon 1995, Cloern 2001). Since then,
Nloadings have been the main target of legislationaimed at reducing
eutrophication in estuaries(Howarth & Marino 2006). Recent
publications haverecommen ded the inclusion of P in coastal and
estu-arine management programs in addition to N (Ho -warth &
Marino 2006, Conley et al. 2009). Other fac-tors such as water
turbidity, the distribution of plantsand macroalgae, sediment
chemistry, nutrient cyc -ling, and nutrient ratios have also been
identified asimportant factors controlling eutrophication
(Cloern2001).
Selective nutrient abatement can change nutrientratios, which
has important consequences for aplanktonic community. For example,
as N loadingsdecline, silica:N (Si:N) ratios should increase,
favor-ing diatoms over other phytoplankton taxa (Cloern2001).
However, at low dissolved inorganic N (DIN)concentrations, smaller
cells should proliferate be -cause of the ability of many small
taxa to outcompetelarger cells at low resource concentrations
(Sunda &Hardison 2007).
Long Island Sound (LIS), USA, located betweenLong Island, New
York (NY), and Connecticuts (CT)south shore, has experienced
eutrophication and sea-sonal hypoxia since at least the 1970s
(Parker &OReilly 1991). Eutrophication and hypoxia are
typi-cally most severe in western LIS due to its proximityto New
York City (NYC; Anderson & Taylor 2001)and the long hydraulic
residence times within theestuary (63 to 160 d; Turekian et al.
1996). In 1985, apartnership between the US Environmental
Protec-tion Agency (USEPA) and the states of NY and CTformed the
Long Island Sound Study (LISS) in orderto increase efforts and
collaboration among the 2states and a variety of federal, state,
and local part-ners to restore and protect LIS. One of the
manygoals of the program included a reduction in N loadings to LIS,
primarily through upgrades to waste-water treatment plants (WWTPs)
(NYSDEC & CT -DEEP 2000). LISS also funds the Connecticut De
-partment of Energy and Environmental Protection(CTDEEP) to manage
a water quality monitoring pro-
gram in which more than 40 stations are sampled ona monthly
basis. Based on early findings, a total ma x -i mum daily load was
approved for N in LIS in 2000,with the goal to reduce anthropogenic
N loadings by58.5% by 2014 (NYSDEC & CTDEEP 2000). By
2010,primarily through upgrades to WWTPs but alsothrough an
increasing number of projects targetingnonpoint sources of N, 70%
of this goal had beenachieved (41% total reduction; LISS 2011). By
limit-ing N, the LISS and managers hoped to decrease thefrequency
and intensity of phytoplankton blooms,and the subsequent sinking
and delivery of hypoxia-fueling organic matter to depth. However,
despitethe successful reduction of N loadings to the
estuary,eutrophication-driven hypoxia continues to be a
sig-nificant problem (LISS 2011).
LIS is not unique in that the implementation of Nsource control
programs has not mitigated eutrophi-cation and hypoxia to any
perceptible degree (Lee &Lwiza 2008, Kemp et al. 2009). The
comparison ofeutrophic estuaries worldwide has shown that
theefficacy of N source reduction efforts has variedwidely at the
ecosystem scale (Cloern 2001). Here weaimed to identify the impact
of the ongoing imple-mentation of N reductions on water quality in
LISover a 15 yr period by identifying changes in nutri-ent
concentrations, stoichiometry, phytoplanktonbiomass and community
structure, and hydrographicforcings that have occurred since N
reductions beganin the early 1990s.
MATERIALS AND METHODS
Biogeochemical and physical data
Annual rates of N discharge from all WWTPs in -to LIS were
obtained from the NY State Depart -ment of Environmental
Conservation (NYSDEC). Onlyannual discharge rates are available,
and thereforevariability in N discharge could not be compared
tomonthly fluctuations in other variables. In addition,annual N and
P discharge from all NYC WWTPs dis-charging into western LIS were
obtained from theInterstate Environmental Commission for New
York,New Jersey, and Connecticut. Water column
nutrientconcentrations, chlorophyll a (chl a), O2, and
hydro-graphic data were obtained from the CTDEEP data-base. Data
and analytical methods for collection ofbiogeochemical and physical
data are available atwww.lisicos.uconn.edu/ and
www.lisicos.uconn.edu/NutrientID.pdf. Biweekly to monthly data from
9 sta-tions along LISs primary axis were tabulated be -
52
-
Suter et al.: Phytoplankton assemblage changes in an urbanized
estuary
tween January 1995 and May 2009. The stations in -cluded A4, B3,
C2, D3, E1, H4, I2, J2, and M3 (Fig. 1).Each survey included
continuous depth profiles ofdensity, temperature, salinity, and
dissolved oxygen(DO). At each time point, , T, and S (bottomminus
surface density, temperature, and salinity)were calculated as
estimates of stratification condi-tions. Discrete Niskin bottle
samples from 2 depths(surface and near-bottom) at each station were
ana-lyzed for concentrations of chl a, nitrate + nitrite(NOx),
ammonium (NH4+), orthophosphate (dissolvedinorganic P, DIP),
dissolved silicate (DSi), dissolvedorganic carbon (DOC), total
dissolved N (TDN), totaldissolved P (TDP), particulate C, N, P (PC,
PN, PP),and biogenic Si (BioSi), and total suspended solids(TSS). C
and N content measurements did not dis-criminate between organic or
inorganic sources. DINwas calculated by summing NH4+ and NOx for
eachsample. Dissolved organic fractions of N (DON) andP (DOP) were
calculated by subtracting inorganicfractions from total dissolved
pools. Atomic ratios ofC:N and N:P were calculated for the
particulate anddissolved inorganic fractions.
When examining nutrient stoichiometry, DIN:DIPratios and N* (N*
= N [16 P] + 2.90 mol kg1;Gruber & Sarmiento 1997) were not
employed be -cause undetectable DIP concentrations would resultin
infinite DIN:DIP ratios, and stoichiometric correc-tions to N* are
not applicable to estuarine settings.Our solution was to use an
excess DIN index (DINxs),where:
DINxs = DIN (16 DIP) (1)
DINxs was originally developed to estimate rela-tive
contributions of N fixation and denitrification to
the N budget by estimating deviations from the Red-field ratio
(16N:1P) (Hansell et al. 2004). For LIS,anthropogenic N loading and
denitrification alsoaffect the N budget, so the DINxs parameter is
usedsolely as a Redfieldian index of DIN relative to
DIPconcentrations, i.e. a DINxs of 0 equates to a DIN:DIPratio of
16, negative DINxs suggests N-limitation,and positive values
suggest P-limitation.
Phytoplankton community data
In April 2002, CTDEEP augmented chl a measure-ments with
detailed pigment analyses of samplestaken from 2 m depth to monitor
phytoplankton com-munity composition. These analyses included bi
-weekly to monthly sampling to May 2010 fromStns A4, B3, C1, D3,
E1, F2, H4, I2, J2, and K2. Pig-ment concentrations were measured
using high per-formance liquid chromatography (HPLC) by theHorn
Point Analytical Services Laboratory (Cam-bridge, MD). Analyzed
photopigments included chla, divinyl-chl a, chl b, divinyl-chl b,
chl c1c2, chl c3,alloxanthin, antheraxanthin, -carotene,
canthaxan-thin, diadinoxanthin, diatoxanthin, echinenone,
but-fucoxanthin, fucoxanthin, hex-fucoxanthin, gyro -xanthin, and
lutein. CTDEEP compared pigmentconcentrations to known assemblages
from LIS usingCHEMTAX and reported phytoplankton taxon abundance
normalized to predicted chl a contri -bution (Li et al. 2004).
Taxonomic groups includeBacillariophyceae (diatoms), Dinophyceae
(dinofla-gellates), cyano bacteria, Prasinophyceae, Chlorophy
-ceae, Cry p to phyceae, Prymnesiophyceae, Raphido-phyceae,
Eustigmatophyceae, Chrysophyceae, and
Euglenophyceae. Diatoms frequently domi-nated total pigment
inventories while non-diatom groups had smaller and
variablecontributions. Thus derived chl a contribu-tion from the
non-diatom ensemble wassummed and considered as an
additionalstatistical variable. The CHEMTAX methodhas previously
been shown to underesti-mate the relative abundances of
certaindinoflagellates and haptophytes due tounique pigment markers
in some species ofthese groups (Irigoien et al. 2004).
Further-more, dia toms have a higher chl a:C ratiothan other
phytoplankton, and chl a esti-mates derived from the Chemtax
methodmay overestimate their biomass (Llewellynet al. 2005). We
therefore took a con -servative approach by only estimating
53
Fig. 1. Long Island Sound. A4 through M3 are stations at which
nutrientand phytoplankton data were collected. La Guardia Airport
(LGA) andFlushing (FWS) weather stations are meteorological data
collection sites
-
Mar Ecol Prog Ser 497: 5167, 2014
interannual changes (as opposed to within-year sea-sonal
evolution) using relative abundances in chl aunits and did not
attempt to estimate the overall con-tribution of taxa to total C
biomass.
Meteorological data
Monthly precipitation totals from La Guardia Airport (Fig. 1)
were acquired from the NationalOceanic and Atmospheric
Administration (NOAA,http://lwf.ncdc.noaa.gov/oa/ncdc.html).
Monthly cloudcover (%) and wind speed at the Flushing
weatherstation (Fig. 1) were acquired from Weather Under-ground
(www.wunderground.com). Wind energy atthe airwater interface was
calculated as the squareof the wind speed (m2 s2). Variability in
monthlyfreshwater discharge into LIS was assessed exclu-sively
using USGS tidal gauge data from the Con-necticut (CT) River,
because Hudson River dischargerecords from West Point, NY, were
discontinuous(http://waterdata.usgs. gov/nwis). Moreover, the
CTRiver contributes >70% of riverine freshwater dis-charge into
LIS, and thus represents a proxy for tem-poral variations and a
conservative estimate of totalfreshwater input (Lee & Lwiza
2005).
Data processing
To synchronize all measurements, data were lin-early
interpolated and resampled mid-month, at anevenly spaced gap of
30.5 d, resulting in contempora-neous monthly data from each
station and depth. Toremove seasonality, the average annual cycle
wascalculated for each variable and station, then sub-tracted from
respective monthly observations (e.g.Fig. 2). Seasonally-adjusted
monthly anomalies wereused for all statistics and comparisons.
Long-term trends of nutrients, total chl a, and therelative
abundances of phytoplankton taxa in chl aunits were analyzed using
the Theil-Sen estimator(Wilcox 2005, Cloern et al. 2007).
Associations be -tween variables were estimated using the
percentagebend correlation (rpb) (Wilcox 1994). These estimatorsare
more robust than ordinary least squares methodsand can be used on
non-normal and heteroscedasticdata (Wilcox 2005). Significance was
determined bycalculating 95% confidence intervals (CIs) from
boot-strapping 599 estimates of either the Theil-Sen esti-mated
trend or rpb, as suggested by Wilcox (2005) fora dataset of this
length (n = 173), and using themiddle 95% of these estimates. If
the upper and lower
limits were both positive (or both negative), the slopewas
considered significantly different than 0, demon-strating a trend
(Fig. 2B,C). Because of the distribu-tions of some variables, CIs
were not normally distrib-uted about the trend (or rpb), and
therefore both upperand lower limits are presented.
To determine whether any variables underwent asignificant shift
to a new mean condition, change-point analysis was performed
(Pettitt 1980). Cumula-tive sums (CUSUM) of deviations from the
meanwere calculated according to the equation:
54
Fig. 2. Demonstration of calculation of the seasonally-adjusted
anomalies and Theil-Sen trend. In (A), the monthlyinterpolated
total dissolved N (TDN) concentrations areplotted as dots, and the
bold black line is the average annualcycle. In (B), the anomalies
are shown, which are the resultof the average annual cycle from (A)
subtracted from themonthly data from (A). In (C), the y-axis from
(B) isexpanded to show the trend calculated using the
Theil-Senmethod (grey line; 0.39 mol N l1 yr1). The black
dashedlines are the lower and upper limits (0.34 and 0.44 mol N
l1
yr1) of the 95% confidence interval of the trend after
boot-strapping. The confidence interval does not straddle 0,
and
therefore this trend is considered significant
-
Suter et al.: Phytoplankton assemblage changes in an urbanized
estuary
Si = Si1 + (Xi ) (2)
where S1 = 0, Xi is the original time series anomalyvalue, and
is the mean of the variable being consid-ered. In addition, Sdiff
(maximum S minimum S)was calculated for each variable. S was then
plottedagainst time. Points at which S changed fromincreasing to
decreasing or decreasing to increasingwere considered as possible
change-points. Signifi-cance of each change-point was determined by
boot-strapping 1000 random reorderings of the originaldata and
calculating S and Sdiff for each reordering. If>95% of the
reordered Sdiff values were less than theactual Sdiff value, the
change-point was consideredsignificant.
Once a change-point in chl a was determined,canonical
correlation analysis (CCA) was used todetermine how phytoplankton
biomass covaried withother variables in time. This multivariate
approachwas chosen because it takes into account multiplepredictor
(x) and predictand (y) variables at once,and extracts a pattern of
interannual covariation,making the method superior to linear and
multiplelinear regressions and robust for ecological applica-tions
(Wilks 2006). CCA calculates a linear combina-tion from each of 2
datasets (called canonical vari-ables, or modes) such that the
correlation betweenthe 2 modes is maximized. It proceeds by
removingthe variability of the first modes to find the subse-quent
2 modes with the next highest correlation coef-ficient until the
number of pairs of modes equals thenumber of variables in the
smaller of the 2 originaldatasets (Barnett & Preisendorfer
1987). All variableswere standardized by subtracting the mean
anddividing by the standard deviation, then concate-nated into
matrices according to variable type:planktonic biomass indices (chl
a, TSS, PC, PN, PP,BioSi), nutrients (NH4+, NOx, DIN, TDN, DIP,
TDP,DSi, DOC, DON, DOP), hydrographic/meteorologi-cal variables
(temperature, salinity, density, T, S,, DO, cloud cover, wind
energy, precipitation, andCT River discharge) and phytoplankton
taxa fromHPLC pigment analysis (similar to Levine & Schind -ler
1999). Datasets were pre-filtered using principalcomponent analysis
(PCA), and the leading ortho -gonal components of each matrix
explaining >80%of the variance were retained as input for
CCA(Moron et al. 2006). Resulting canonical componentswere then
compared to original data using a hetero-geneous correlation
(Levine & Schindler 1999, Wilks2006), which correlates the
predictors with the modesfrom the predictands and the predictands
with themodes from the predictors in order to determine
which of the original variables were most associatedwith the
interannual patterns extracted by CCA. Percentage-bend correlation,
change-point analyses,and CCA were only performed with data from
Stn A4because this station exhibited the largest changes innutrient
concentrations and phytoplankton biomass.Furthermore, A4 is the
westernmost station in thisanalysis and is consequently the most
influenced bysewage-derived nutrients from NYC WWTPs andby seasonal
hypoxia (Sweeney & Saudo-Wilhelmy2004).
RESULTS
Fifteen-year analysis of nutrients
Total annual N loads from all WWTPs discharginginto LIS declined
from 36.8 103 in 1994 to 24.7 103 standard tons N in 2010 (Fig. 3,
grey bars). Themajority of this decline occurred between 1994
and2002. Between 2002 and 2010, annual WWTP-sourced N loads
remained relatively constant, vary-ing between 24.7 103 and 29.6
103 tons N. Data onP loadings from NYC WWTPs are available
startingin 2005 (Fig. 3, green line). For comparison, total Nloads
from only NYC WWTPs are also presented for2005 to 2010 (Fig. 3, red
line). From 2005 to 2010,total annual N and P loadings from NYC
showed nomajor changes.
Interannual trends in some deseasonalized vari-ables displayed
reversals in the year 2000 (also discussed by LISS:
http://longislandsoundstudy.net/2010/07/chlorophyll-a-abundance/).
However, trendsprior to and after this reversal were
unidirectional.
55
Fig. 3. Total annual N loads (standard tons, grey bars) fromall
waste water treatment plants (WWTPs) from New York(NY) and
Connecticut (CT) discharging into Long IslandSound (LIS). Red
dashed line is the 2014 goal of 58.75% ofthe total maximum daily
loads. Annual N loads from the 4WWTPs discharging into western LIS
from New York City(NYC) are shown with red circles, and annual P
loads from
the 4 plants are shown with green diamonds
-
Mar Ecol Prog Ser 497: 5167, 2014
Therefore rates of change in this section reflect theentire
observation period, while the shift that oc-curred in 2000 is
discussed below under Change-point analysis. Fifteen-year
interannual rates ofchange in selected variables are presented
inFigs. 49. TDN concentrations increased at most sta-tions in both
surface and bottom waters, especially inwestern LIS, as a result of
increased DON stocks(Fig. 4A,B). DIN concentrations decreased at
all sta-tions except at A4 (Fig. 4AC). The modest decreasesin DIN
at most stations were driven by decreases inconcentrations of NOx
(Fig. 4D). There were no con-sistent trends in NH4+ concentrations,
except at Stn A4,where they significantly increased (Fig. 4E).
UnlikeN, dissolved P and Si stocks increased between 1995and 2009.
TDP concentrations increased at all centraland western stations,
with DIP stocks driving 82 to100% of those increases (Fig. 5). DSi
concentrationsincreased at all stations in surface and bottom
waters(Fig. 6A) while particulate BioSi decreased (Fig.
6B).Increases in DSi were generally larger than decreasesin BioSi,
and so total Si significantly increased at moststations (Fig.
6C).
Chl a increased significantly in surface and bottomwaters
between 1995 and 2009, with trends varyingfrom 0.09 to 0.56 g chl
l1 yr1 among the 9 stations(Fig. 7A). In contrast, TSS decreased at
most stationsbut with no consistent spatial pattern (Fig. 7B).
PCand PN concentrations, which include both livingand detrital C
and N, significantly decreased at moststations, with the most
pronounced decreases in thewest (Figs. 7C,D). Measurable changes in
PP werenot apparent (not shown). Differential rates ofchange among
PC, PN, and PP caused increases inPC:PN ratios (315% yr1) and
decreases in PN:PPratios (1750% yr1 among all stations).
Interannualchanges in the PN:BioSi ratio were not evident at
anystation (not shown), because both PN and BioSideclined at
similar rates. Collectively, variations inparticulate nutrient
ratios indicate that the plank-tonic community has become
increasingly depletedin N and Si relative to C and P.
Nutrient stoichiometry necessarily changed as aconsequence of
varying interannual trends amongnutrients. The DINxs index
significantly decreased insurface and bottom waters, with the
largest changesin the west (Fig. 8). Dissolved nutrients
remainednear the Redfield ratio only at the eastern-most sta-tion,
while DIN depletion accelerated in a westerlydirection. Overall,
changes in the DINxs parameterwere driven by differential changes
in total DIN andDIP pools. DIN stocks declined from an average
of10.1 0.6 (SE) mol l1 among all stations in the first
56
Fig. 4. Annual rates of change in concentrations of (A)
totaldissolved nitrogen (TDN), (B) dissolved organic nitrogen(DON),
(C) dissolved inorganic nitrogen (DIN), (D) nitrate +nitrite (NOx),
and (E) NH4+ in surface and bottom waters atthe 9 stations
analyzed. (d) surface samples; (m) bottomsamples. (s) trends that
are not significant at the 95% level,i.e. the confidence interval
(CI) straddles 0. The x-axis shows
distance from Stn A4. Error bars represent the 95% CI
-
Suter et al.: Phytoplankton assemblage changes in an urbanized
estuary
year of analysis, to an average of 6.7 0.4 mol l1 inthe last 12
mo of analysis, while DIP stocks increasedfrom 1.1 0.02 to 1.2 0.03
mol l1.
Trends in hydrographic and meteorological param-eters were
analyzed to determine possible drivers forobserved changes.
However, relative to the nutrientand phytoplankton biomass
parameters, most hydro-graphic parameters displayed little to no
change overthe 15 yr observation period (Table 1). Overall,
indicated small but significant decreases in stratifica-tion at Stn
A4 and increases at stations in the centraland eastern basins.
Small, significant decreases inbottom water DO concentrations were
also observedat the 3 westernmost stations and 1 easternmost
sta-tion, while no significant changes were observed forcentral
stations.
Eight-year phytoplankton community analyses
Of the major phytoplankton groups, diatoms exhib-ited the most
pronounced long-term change, withconsistent declines at all western
and central sta-tions. Diatom-chl a decreased by 41% at Stn
A4,while the chl a contribution from all other groupsincreased
(Fig. 9). The increase in chl a contribu-tions from the non-diatom
ensemble was driven bygrowing inventories of dinoflagellates,
Prymnesio -phy ceae, Cryptophyceae, Raphidophyceae, and Eu g -
le no phyceae (not shown). For all other groups,changes were
small to undetectable. As a portion oftotal chl a, diatom
abundances decreased from 62.3 2.0% in the first 12 mo of analysis
to 52.5 2.2% inthe last 12 mo of analysis, among all stations.
There-fore, despite large declines, diatoms remained themost
abundant phytoplankton taxa on averagethroughout the entire period
of analysis. However,the frequency of events in which diatoms were
thedominant taxa declined. In the first 12 mo of analysis,diatoms
were
-
Mar Ecol Prog Ser 497: 5167, 2014
positively correlated with those for diatom-chl a.Therefore,
despite decreasing contributions of dia -tom-chl a during this
period, diatoms remained moreclosely correlated with chl a than any
other singlephytoplankton taxon. This also supports the use ofBioSi
as a proxy for trends in diatoms pre-2002 (whenHPLC data were not
available). In addition, Crypto-phyceae and Raphidophyceae were
negatively cor-related with total chl a, and BioSi and
Prymnesio-phyceae were positively related to TSS, suggesting
that this group contributes to variability in total sus-pended
biomass, but not consistently to total chl a.
Correlation analyses also showed that diatom abun -dances in Stn
A4 surface waters were strongly associ-ated with concentrations of
all inorganic nutrients andDINxs. Interannual variations in diatom
abundanceanomalies were also strongly related with those of PC,PN,
and PP, suggesting that diatoms can explainmuch of the variability
in total particulate pools. Totalchl a apportioned to the
non-diatom ensembleshowed no relationships with any nutrients
(Table 2).However, Prasinophyceae concentrations were posi-tively
correlated with TDN, and Cryptophyceae co-varied with most
inorganic N and P pools. Further-more, DON anomalies were
positively correlated withanomalies in the individual
Prasinophyceae andCryptophyceae groups. PC and PN anomalies
wereweakly correlated with chl a contributed by the non-diatom
ensemble. Together, these results suggest thatdiatoms are limited
by inorganic N and P, and controlthe variability in most planktonic
particulate pools,while the non-diatom ensemble is not as sensitive
toinorganic nutrient dynamics and only covaries slightlywith some
particulate pools.
Diatom-chl a at Stn A4 surprisingly also covariedwith salinity
stratification (S). To determine whetherthis result was unique to
Stn A4, we also performedrpb analyses at Stn J2 in the eastern
basin of LIS,which has a more offshore character. At J2, diatom-chl
a was negatively correlated with T, indicating anegative
relationship with thermal stratification (rpb =0.23; CI = 0.39 to
0.02). This relationship con-firmed that in eastern LIS, as
observed elsewhere,diatoms tend to respond negatively to
stratificationand are displaced by mixed assemblages of
flagel-lated taxa. The unexpected dynamic between dia -toms and
stratification at Stn A4 is discussed later.
58
Fig. 7. Annual rates of change in concentrations of (A) chl
a,(B) total suspended solids (TSS), (C) particulate carbon (PC),and
(D) particulate nitrogen (PN) in surface and bottomwaters at the 9
stations analyzed. Symbols and format are
same as for Fig. 4
Fig. 8. Annual rates of change in the excess dissolved
inor-ganic nitrogen index (DINxs) parameter in surface and bot-tom
waters at the 9 stations analyzed. Symbols and format
are same as for Fig. 4
-
Suter et al.: Phytoplankton assemblage changes in an urbanized
estuary 59
(A)
Tem
per
atu
reS
alin
ity
Den
sity
DO
Sta
tion
(C
yr
1 )(y
r1 )
(kg
m
3yr
1 )
(mg
l
1yr
1 )
A4
2.
9
10
27.
6
10
35.
6
10
3
0.04
3.
7
10
2 ,
2.0
10
2
2.3
10
3 ,
1.2
10
2
1.5
10
3 ,
9.2
10
3
0.
05,
0.04
B3
ns
6.1
10
3
ns
0.
016.
8
10
4 , 1
.1
10
2
0.02
, 3.
8
10
3
C2
7.
3
10
35.
6
10
34.
7
10
3
0.01
1.
5
10
2 ,
6.5
10
4
8.1
10
4 ,
1.1
10
2
9.3
10
4 ,
8.9
10
3
0.
02,
4.7
10
3
D3
1.
0
10
2n
sn
sn
s
1.7
10
3 ,
1.
5
10
3
E1
2.
1
10
2n
sn
sn
s
3.0
10
2 ,
1.
2
10
2
H4
3.1
10
2
2.
0
10
2
2.3
10
2
ns
2.3
10
2 ,
4.0
10
2
2.
5
10
2 ,
1.6
10
2
2.
8
10
2 ,
1.9
10
2
I23.
8
10
2n
s
1.9
10
2
ns
2.9
10
2 ,
4.7
10
2
2.
4
10
2 ,
1.3
10
2
J2n
s1.
6
10
28.
0
10
3n
s1.
0
10
2 , 2
.0
10
22.
7
10
3 , 1
.2
10
2
M3
ns
ns
8.
7
10
3
0.01
1.
3
10
2 ,
4.2
10
3
0.
02,
6.8
10
3
(B)
Tem
per
atu
reS
alin
ity
Den
sity
TS
D
OS
tati
on(
C y
r1 )
(yr
1 )(k
g m
3
yr
1 )(m
g l
1
yr
1 )
A4
3.
3
10
2n
s6.
4
10
3
4.2
10
3
3.
4
10
3
3.8
10
3
0.
02
4.1
10
2 ,
2.6
10
2
2.4
10
3 ,
1.1
10
2
8.
3
10
3 ,
3.5
10
4
5.
8
10
3 ,
1.4
10
3
5.
9
10
3 ,
2.2
10
3
0.
02,
0.01
B3
ns
ns
ns
0.
02
4.6
10
3
ns
ns
0.
03,
0.02
2
6.
6
10
3 ,
2.6
10
3
C2
2.
2
10
2n
s4.
5
10
3
0.02
3.
8
10
3n
s0.
01
2.8
10
2 ,
1.
5
10
21.
5
10
3 , 8
.3
10
3
0.02
, 0.
01
5.4
10
3 ,
2.
3
10
34.
6
10
3 , 0
.01
D3
2.
4
10
25.
0
10
38.
8
10
3
0.02
3.8
10
3
6.0
10
3
0.01
3.
2
10
2 ,
1.7
10
2
1.1
10
3 ,
9.1
10
3
5.0
10
3 ,
1.2
10
2
0.
03,
0.01
2.2
10
3 ,
5.2
10
3
4.6
10
3 ,
8.0
10
3
3.8
10
3 ,
0.0
1E
1
4.8
10
2
1.1
10
2
1.8
10
2
0.
029.
5
10
30.
020.
01
5.5
10
2 ,
4.
1
10
25.
6
10
3 , 1
.6
10
21.
4
10
2 , 2
.2
10
2
0.02
, 0.
017.
7
10
3 , 0
.01
0.01
, 0.0
21.
3
10
3 , 0
.01
H4
2.
9
10
21.
5
10
21.
5
10
2
0.04
1.7
10
2
0.03
0.01
3.
5
10
2 ,
2.0
10
2
9.6
10
3 ,
1.9
10
2
1.1
10
2 ,
1.8
10
2
0.
05,
0.04
1.5
10
2 ,
2.0
10
2
0.02
, 0.0
30.
01, 0
.02
I2
7.3
10
4
1.7
10
2
1.1
10
2
0.
031.
7
10
20.
020.
01
6.3
10
3 ,
5.8
10
3
1.2
10
2 ,
2.2
10
2
7.5
10
3 ,
1.5
10
2
0.
04,
0.03
1.5
10
2 ,
1.9
10
2
0.02
,0.0
32.
8
10
3 , 0
.01
J24.
4
10
31.
4
10
26.
8
10
3n
s5.
6
10
30.
01n
s
2.4
10
3 ,
1.1
10
2
9.6
10
3 ,
1.9
10
2
3.6
10
3 ,
1.1
10
2
2.3
10
3 ,
9.2
10
3
7.2
10
3 ,
0.0
2M
3n
s1.
2
10
24.
4
10
30.
020.
017.
4
10
34.
0
10
3
7.4
10
3 ,
1.6
10
2
6.4
10
4 ,
7.4
10
3
2.5
10
3 ,
0.0
39.
0
10
3 , 0
.02
4.2
10
3 ,
0.0
18.
9
10
4 , 0
.01
Tab
le 1
. Tre
nd
s in
hyd
rog
rap
hic
var
iab
les
in (
A)
surf
ace
wat
er a
nd
(B
) b
otto
m w
ater
fro
m 1
995
to 2
009
at a
ll s
tati
ons.
Th
e u
pp
er a
nd
low
er li
mit
s of
th
e 95
% c
onfi
den
cein
terv
al a
re p
rese
nte
d b
elow
th
e tr
end
. n
s: t
ren
ds
that
are
not
sig
nif
ican
t. D
O:
dis
solv
ed o
xyg
en; T
, S
, an
d :
str
atif
icat
ion
cal
cula
ted
as
bot
tom
min
us
surf
ace
in
tem
per
atu
re, s
alin
ity,
an
d d
ensi
ty, r
esp
ecti
vely
-
Mar Ecol Prog Ser 497: 5167, 2014
Lastly, variations in diatom populations as well as allother
phytoplankton taxa did not relate to those inbottom water DO
concentrations. However, bottomwater DO concentrations did
positively correlatewith total chl a (rpb = 0.31; CI = 0.16, 0.47),
TDP (rpb =0.22; CI = 0.37, 0.09), DIP (rpb = 0.24; CI = 0.42,0.08),
and temperature (rpb = 0.31; CI = 0.43,0.16). Surprisingly, bottom
water DO did not corre-late with any measure of stratification.
Comparisons between phytoplankton taxa chl acontribution and
planktonic biomass indices usingCCA resulted in 2 modes that were
significantly cor-related: modes 1 (r = 0.84; p < 0.0001) and 2
(r = 0.44;p < 0.0001). The first mode showed correlations
simi-lar to those found by rpb analysis, further validating
60
Diatoms Prasino- Crypto- Prymnesio- Raphido- All non-diatom
phyceae phyceae phyceae A phyceae ensembles
Chl a (g l1) 0.72 ns 0.25 ns 0.20 ns0.53, 0.82 0.45, 0.09 0.44,
7.1 104
TSS (mg l1) 0.34 ns ns 0.33 ns ns0.09, 0.51 0.07, 0.53
TDN 0.41 0.21 0.32 ns ns ns0.56, 0.23 0.02, 0.42 0.12, 0.52
TDP 0.42 ns 0.26 ns ns ns0.60, 0.18 0.06, 0.46
NH4+ 0.45 ns 0.28 0.22 ns ns0.60, 0.27 0.09, 0.46 0.42,
1.2103
NO 0.24 ns ns ns ns ns0.41, 0.04
DIN 0.37 ns 0.30 ns ns ns0.56, 0.19 0.09, 0.49
DIP 0.44 ns 0.23 ns ns ns0.60, 0.21 0.05, 0.44
DSi 0.42 ns ns ns ns ns0.57, 0.25
PC 0.71 ns ns 0.30 ns 0.240.53, 0.83 0.05, 0.52 0.02, 0.45
PN 0.65 ns ns 0.30 ns 0.210.44, 0.79 0.07, 0.50 4.5103, 0.43
PP 0.58 0.24 ns ns ns ns0.38, 0.74 0.09, 0.43
Biogenic silica 0.60 ns 0.26 ns 0.25 ns0.46, 0.73 0.43, 0.06
0.43, 0.05
DON ns 0.35 0.30 ns ns ns0.16, 0.56 0.07, 0.50
DINxs 0.35 ns ns ns ns ns0.07, 0.49
S 0.26 ns ns ns ns ns0.10, 0.41
Table 2. Percentage bend correlation coefficients (rpb) between
chl a contribution from relevant phytoplankton taxa, andselected
biological, nutrient, and hydrographic data between 2002 and 2009.
The upper and lower limits of the 95% confi-dence interval of the
rpb are presented below the coefficient. ns: correlations that are
not significant. Groups of phytoplanktonpigments or field data that
did not correlate with any other variables are omitted. Values of
phytoplankton pigments are ing l1. All other values are in mol l1
unless shown otherwise. TSS: total suspended solids; TDN (TDP):
total dissolved nitrogen(phosphorus); DIN (DIP): dissolved
inorganic nitrogen (phosphorus); DSi: dissolved silicate; PC:
particulate carbon; PN: partic-
ulate nitrogen; PP: particulate phosphorus; DON: dissolved
organic nitrogen; S: change in salinity
Fig. 9. Annual rates of change in chl a from diatoms (d) andfrom
all other phytoplankton groups (m). (s) trends that arenot
significant at the 95% level. The x-axis shows distance
from Stn A4. Error bars represent the 95% CI
-
Suter et al.: Phytoplankton assemblage changes in an urbanized
estuary
those results (not shown). In addition, heterogeneouscorrelation
of the second mode from CCA revealedsome new relationships that
were not apparent withunivariate statistics: dinoflagellates (r =
0.44, p
