Journal of Plankton Research Vol.20 no.l pp.145-168, 1998
Temporal patterns and variations in phytoplankton communityorganization and abundance in Narragansett Bay during1959-1980
Deneb Karentz1 and Theodore J.Smayda2
'Department of Biology, University of San Francisco, 2130 Fulton Street, SanFrancisco, CA 94117-1080 and 2Graduate School of Oceanography, University ofRhode Island, Narragansett, Rl 02882, USA2 To whom correspondence should be addressed
Abstract Incident irradiance, surface water temperature and phytoplankton species abundances weremeasured at weekly intervals in Narragansett Bay from 1959 through 1980. Stepwise discriminantanalyses (SDA) of this 22-year data set indicate that fundamental ecosystem changes occurredbetween the 1960s and 1970s, with 1969 being the key transitional year in these decadal shifts in phyto-plankton taxonomic structure and seasonal abundance. This decadal shift was accompanied by theincreased summer abundance of small Thalassiosira spp., which first appeared in 1966 and by 1969became the sixth most important phytoplankton component in this bay. Decadal trends in phyto-plankton community organization and abundance were also accompanied by distinct long-termclimatological gradients of temperature and light. The 1960s were generally colder and brighter thanthe 1970s. Prior to 1969, the annual phytoplankton maximum occurred most commonly during winter;in the 1970s, the annual maximum generally shifted to a summer event. Three 5-year phytoplanktoncycles occurred between 1959 and 1974. During each pentade, the phytoplankton community returnedto a similar taxonomic organization and abundance cycle after diverging in the intervening years.Pentade cycles did not occur after 1974; the phytoplankton community thereafter diverged signifi-cantly from each preceding year. Five species [Skeletonema costatum, Detonula confervacea, Asteri-onellopsis glacialis, Heterosigma akashiwo (= Olislhodiscus luteus) and Thalassiosira nordenskioeldii]dominated the phytoplankton over the 22-year period. SDA revealed a high degree of similarity andconstancy in the annual occurrence patterns of these taxa. The decadal shifts revealed by SDA weremore directly related to the considerable interannual variability that characterized the abundance andseasonality of the less abundant species.
Introduction
Temporal patterns of occurrence and abundance of numerically important phyto-plankton species in Narragansett Bay, Rhode Island (41°34'N, 71°23"W), over a22-year period have been described (Karentz and Smayda, 1984). R-type(species-oriented) principal components analysis (PCA) was used to evaluatespecies associations based on weekly quantitative sampling. The PCA brdinatedthe 30 most abundant species along a seasonal gradient of temperature andirradiance. While the in situ seasonal occurrence patterns of these species gener-ally agreed with those expected from laboratory experiments on tempera-ture-growth relationships, several lines of evidence suggested that the observedsuccessional patterns were not regulated by temperature.
In this paper, the Narragansett Bay phytoplankton series is re-examined usinga Q-type (sample-oriented) stepwise discriminant analysis (SDA) to investigatelong-term trends in species succession. We will focus on the variability in the taxon-omic structure of the phytoplankton assemblages over time. The results of theSDA were analyzed relative to specific occurrences and abundances of individual
species, rather than to generalized seasonal patterns identified previously by PCA.The degree to which annual and seasonal changes in taxonomic structure wererepetitive and predictable was also assessed. Finally, major shifts in successionalpatterns and probable relationships between phytoplankton events and changesin light, temperature and trophic level interactions have been identified.
Method
The phytoplankton data set from Narragansett Bay used in this analysis consistsof 1000 weekly samples collected over a 22-year period from 1959 through 1980.A total of 138 taxa have been recorded from surface samples. Methods used tomeasure incident light intensity, water temperature and phytoplankton speciesabundance, and procedures to log-transform species' cell counts and reduce therank of the species data matrix, were described previously (Karentz and Smayda,1984).
SDA program BMDP7M of the Biomedical Computer Programs P-series wasapplied to examine temporal changes in phytoplankton community structure(Cooley and Lohnes, 1971; Allen and Skagen, 1973; Dixon and Brown, 1979). SDAsummarizes variation within a data set, providing more concise and comprehens-ible presentations of large matrices than is possible with graphs and tables of rawfigures alone. In the SDA of the Narragansett Bay data, each sample date was pro-jected as a single point within a geometric hypervolume (n-dimensional space)defined by the number (n) of species included in a selected model. The individualabundances of all species (present or not) on any given sample date determine thelocation of each sample point. Additionally, each sample date is assigned to a pre-designated temporal group (a sub-matrix of the log-transformed cell counts).
SDA is an iterative procedure by which individual species are successivelyadded into the discriminant model so that the ratio of the variance betweengroups to the variance within each group is maximized. Species that have the mostvariable record of abundance and occurrence are added into the discriminantmodel first, followed by those which have more regular patterns of occurrence inthe phytoplankton. In this way, the SDA emphasizes differences between groupswhile maintaining the integrity of each group.
Canonical variables, i.e. vectors that represent the summarization of variabilityinherent in the data, are generated from the discriminant model (Cooley andLohnes, 1971; Dixon and Brown, 1979). The generated canonical variables arecalculated to maximize the percent of variance in the data explained by eachcanonical variable (CV), with CV1 explaining the greatest portion of the variance,CV2 explaining the next largest portion, etc. The number of canonical variablesderived is equal to the number of variables (species) entered. The first two canon-ical variables (CV1, CV2) usually accounted for at least half of the total variationwithin the data. The percent of variation in the data explained by the canonicalvariable(s) is shown in the corresponding graphical representations. Graphicalrepresentations of the correlation coefficients between samples and canonical vari-ables CV1 and CV2 facilitate comparison of the qualitative and quantitativeaspects of taxonomic structure among the designated temporal groups being
146
Phyioplankton in Narragansett Bay during 1959-1980
analyzed. In plots of the canonical variables, each group is represented by a singlepoint, which is a mean value. In reality, each mean is surrounded by a cloud ofpoints, one for each sample date in the group. Given the large number of samplesin the Narragansett Bay data set, we have plotted group means rather than theindividual points on the figures. The relative spatial distribution and the distancesbetween mean points indicate the degree of taxonomic similarity. The objective ofthe analyses is to separate (= discriminate) between the different annual and othersubgroups based on a series of seemingly continuously distributed measurementsof phytoplankton species occurrence and abundance, temperature and light. Thepercent variance explained indicates how well the SDA has succeeded in 'pullingapart' the various distinct subgroups hidden in the data.
Three temporal groupings of the Narragansett Bay data were established forSDA: 22 year-groups (January-December) to examine annual successionalpatterns; 44 semi-annual groups to examine 'summer' (June-October) and'winter' (November-May) patterns; and 88 quarterly-groups (January-March,April-June, July-September, October-December) to examine seasonal patternsand their annual variations. Various subsets of the species data bank in each ofthese temporal groupings were also subjected to SDA. Initially, the 49 most abun-dant species were used (Table I), followed by 30-, 10-, 6- and 5-species models.SDA was repeated on the 24 most frequently observed species (based on pres-ence and absence, rather than cell abundance) (Table I). Light and temperaturedata corresponding to the same temporal groupings were also subjected to SDA.Groups that had continuous segments of missing data (e.g. year 1963, 'summer'1977, 'winter' 1978) were not used to develop the discriminant model. However,the data available from these groups were evaluated relative to the model afterit was established. These points are identified on the accompanying figures by theuse of adjoining, broken lines.
Results
Temperature and light
The seasonal patterns of temperature and incident irradiance were described pre-viously (Karentz and Smayda, 1984). The 22-year quarterly mean temperaturewas lowest during the January-March quarter (1.8°C) and highest (20.7°C) duringJuly-September (Figure 1). The 1960s were generally colder than the 1970s(Table II). This was most pronounced for the January-March quarters; mean tem-perature was 32% lower during 1959-1969 than during 1970-1980. The long-termtrend in incident irradiance was the reverse of that for temperature; light levelswere higher in the 1960s than during the 1970s (Figure 2). Mean irradiance levelswere lowest (189 ly day-1) during autumn (October-December) and highest (460ly day"1) during the spring quarter (April-June) (Figure 2).
SDA of light and temperature
SDA of the annual (January-December) light and temperature data identifiedlight as the more discriminating of these two environmental variables relative to
147
D.Karentz and TJ^mayda
Table I. Rank order of the 49 most abundant species used in the SDA, and year of their maximummean weekly abundance. An asterisk identifies the 24 most frequently observed species. [See Table 1of Karentz and Smayda (1984) for additional quantitative data]
group separation (Figure 3A). Each year's correlation between the plotted posi-tion of CV1 and the corresponding mean light level was stronger than thatbetween CV1 position and mean temperature. Most of the annual data with high
148
Phytoplankton in Narragansett Bay during 1959-1980
Jon-Mor T= I.8°C
nil
Illllllnlllllll
12-
1 0 -
8 -
6 -
Oct-Dec 7 = 9.4°
1960 65 70
YEAR
75 8 0
Fig. 1. Annual quarterly means of surface water temperature in Narragansett Bay based on weeklymeasurements. The broken line indicates the mean quarterly value during the 22-year period.
CVl loadings are from the 1960s, while the cluster of years having low CVl load-ings are mostly from the mid- to late 1970s. The distribution of years along theCVl gradient in Figure 3A, therefore, represents a gradient of phytoplanktoncommunities extending from the 'low-light' years of the 1970s to the 'high-light'
149
D.Karentz and TXSmayda
Table II. Annual mean surface temperature (°C) and incident irradiance (/) (ly day"1)
years of the 1960s. Although the tight clustering of means reveals a high degreeof similarity, the segregation of the 1960s and 1970s year-groups is evident. Themajority of the 1960s group means were positively correlated with CV1, whereas1970s year-groups were negatively correlated. This decadal separation relative tolight and temperature reflects the trends observed from the 1960s through the1970s in which temperature increased and light decreased (Figures 1 and 2).
In SDA of 'summer' and 'winter' groupings, the two seasons clustered at oppo-site ends of CV1 (Figure 3B), a seasonal separation related primarily to temper-ature. Although the residual variation associated with CV2 was only 2%, decadalgroups within each seasonal cluster were segregated along this axis. SDA of thequarterly light and temperature data revealed that January-March groups exhibitthe greatest dispersion among years, July-September the least (Figure 4). As withthe annual and semi-annual groups, the 1960s and 1970s quarterly-groups are dis-tinctly segregated. For each quarter, yearly groups are widely distributed alongCV1. This chronological alignment reflects specific differences in light and tem-perature regimes for each time period.
Total phytoplankton
Mean annual and quarterly abundances of phytoplankton cells during the 22-yearperiod are shown in Figure 5. The highest (117 886 ml"1) and lowest (16 ml"1) cellconcentrations observed on a given sampling date were recorded during the 1979and 1966 summers, respectively. The mean annual maximum varied ~4-fold: from2820 cells ml"1 (1972) to 10 235 cells ml'1 (1980). From 1959 to 1969, the annual
150
Phytoplankton in Narragansett Bay during 1959-1980
Jan-Mar x = 222 ly day"1
300-
20O-
500-
400-
5 400 -)Q:
""~ 350-
300-
200-
100-
Jul-Sept 7=420
Oct-Oec 189
I960 65 70 75 80
Fig. 2. Annual quarterly means of incident irradiance over Narragansett Bay based on daily measure-ments. The broken line indicates the mean quarterly value during the 22-year period.
phytoplankton maximum usually occurred during the winter (January-March).During the following decade, the annual phytoplankton maximum was pre-dominantly a summer (June-September) event. In both decades, lowest cellnumbers generally were recorded in the autumn (October-December). Overall,mean weekly abundance ranged interannually from -1000 to 15 000 cells ml"1 inthe winter samples, from 1000 to 12 000 cells ml""1 during the spring, from 2000 to17 000 cells ml"1 during summer, and from 800 to 8000 cells ml"1 during autumn.These changes and their apparent environmental regulation will be presented ingreater detail elsewhere (TJ.Smayda, in preparation).
151
D.Karentz and T-Umayda
O.2
O-
-O.2-
A
79
'?? 72
71
8°70R+ 6I
6 9 7667
Jan-Dec 1959-1980light + tempCVI * 8 3 %
5 9
6 0
62 646 8 65
66
-0.4 -0.2 0.2 0.4
1.0-
0.5-
0 -
0.5-
B
63
62
70 64
69 TN795
7278
winter
Summer-winter 1959-1light + tempCVI = 9 8 %
65
66°8• 64 6%l|9
72 75
summer
-2.0 -1.0 0CVI
1.0 2 .0 3.O
Fig. 3. Canonical variables 1 (CVI) and 2 (CV2) of stepwise discriminant analysis (= SDA) for lightand temperature data. (A) Annual groups (January-December); (B) 'summer' (June-October) and'winter' (November-May) groups. Encircled numbers represent years not used in the developmentof the discriminant model (see the text for further discussion).
SDA of annual successional patterns
The first model evaluated by SDA to assess annual (January-December) patternsin phytoplankton community structure was based on the 49 species most abun-dant numerically (Figure 6A; Table I). Although the plotted annual mean valuesare distinct, values of individual sample points (not shown) from within a givenyear overlapped those from other annual groups. This reflects the condition thatsummer assemblages from one year may be more similar to those of another yearthan to winter assemblages of the same year. The chronological continuumobserved along CVI reflects the continual change that characterizes communityorganization of the phytoplankton assemblages in Narragansett Bay. Two clus-ters of yearly means were generated. In the first cluster consisting of the
152
Phytoplankton in Narragansett Bay during 1959-1980
I -,
+ 1
0 -
§! -'
o-
- I+1
6 0
Jon-MarCVI =90%
73 76_ _66_ £167 69 70 63 63
^ , ; 6 8
73-
Apr-Jun
79 69 •-* C V . - 8 6 %_7 l_76_60 i&a62 ___ 63_74 , . 72 61
"80 « 7 5 '* - 6 667
66
80
Jul-Sept
677
73 75
60 64
76
6961^39 7372
Oct-Oec
79
-1.0 -0.5 0 0.5CVI
1.0
Fig. 4. CVI and CV2 of SDA for quarterly intervals of light and temperature data. Encircled years asin Figure 3.
1959-1968 year-groups, two pentade cycles are superimposed within this decade.The 1959-1963 year-groups complete one 5-year cycle; the 1964-1968 annualgroups the other. A significant taxonomic shift then occurred between 1968 and1969, which initiated the second clustering of annual groups. Year-groups1970-1974 are closely related and represent a third pentade cycle. After 1974, thecyclical pattern is broken, and the 1975-1980 year-groups then extend linearlyacross CVI. Within this linear array of annual means, 1975 and 1976 are closelyassociated, followed by a shift to, and clustering of, the 1977,1978 and 1979 year-groups.
When we redefined the year-groups as belonging to a 12-month calendar periodextending from May to April, the spatial relationships generated between yearswere nearly identical to those of the January-December annual groupings. Thissupports our view that the interannual differences suggested by SDA are real andnot biased by the temporal grouping procedure.
We then applied SDA to annual groups of progressively fewer species whoseinclusion in the models was based on their total maximum abundance, or
153
23000-
15000-
I If I 110000-
5000-
- 8000
- 6000
- 4000
- 2000
- 0
1959
6
0 61
63
64 1
965
66
67
68
69 1
970
71
72
73
74
1975
76
77
78
79 I
960
S I I
g. 1
Fig
. 5. A
nnua
l qu
arte
rly
mea
ns (
hist
ogra
m)
and
annu
al w
eekl
y m
eans
(so
lid li
ne)
of to
tal p
hyto
plan
kton
abu
ndan
ce. N
umbe
rs a
bove
sta
ndar
d er
ror
bars
ide
ntif
yth
e qu
arte
r of
hig
hest
mea
n ce
ll de
nsit
y fo
r ea
ch y
ear
(1 =
Jan
uary
-Mar
ch;
2 =
Apr
il-Ju
ne;
3 =
Jul
y-Se
ptem
ber;
4 =
Oct
ober
-Dec
emb
er).
<o
2- 1 -
0-
-1 -
-2-
c. 1 -
r\ U -i -
-2-
80
.
77.^
_^
A•A
-3
80
,. 7V
V«
77
,
B
49
Sp
tcll
l M
od
llC
Vl +
CV
2 .
56
%
|'T
y"7,
rop
-2
-1
0 1
224
S
pic
ill
Mo
dfl
CV
l +
C
V2
• 6
7%
6I*
^2
i/y
-2
-I
.5- 0- -.5-
.5-
0-
-.5-
7 c D
I 72
-1.5
-1
69..
67
~
.5
0
60<
--6
6?">
/
•68^ \& \A If 7
7
6
CV
l
/
.5 9
CV
l
V78
l/
Sp
acli
t+
CV
2 65
5964
y
•IM/ 60
Spic
ifll
• C
V2
-6
5
Mod
ll
• 82
%
Mod
i
• 69
%
-I
-.5
CV
l
o 1 a •z §.
Fig.
6. C
Vl a
nd
CV
2 o
f S
DA
fo
r Ja
nuar
y-D
ecem
ber
year
-gro
ups
of s
peci
es d
ata.
(A
) 4
9 m
ost
abun
dant
spe
cies
; (B
) 2
4 m
ost
freq
uent
ly o
bser
ved
spec
ies;
(C
)si
x m
ost
abun
dant
spe
cies
; (D
) fiv
e m
ost
abun
dant
spe
cies
. S
ee T
able
I f
or
spec
ies
used
.
D.Karentz and TJ.Smayda
frequency of occurrence. The spatial distributions of yearly means based on the24 most frequently observed species were similar to the model of the 30 mostabundant species (not shown) and relatively unchanged from the 49-speciesmodel (Figure 6B, Table I). However, the closer association of the 1978-1980year-groups (Figure 6B) suggests a greater degree of taxonomic similarity.
In restricting SDA to the six most abundant species, the clear separation of the1960s from the 1970s persists (Figure 6C). However, this 6-species model doesnot exhibit the three pentade cycles between 1959 and 1974 seen in the othermodels, and the 1970s year-groups exhibit a greater inter-annual similarity. A par-ticularly notable feature in the 6-species model is the significant departure fromprevious years initiated in 1967; in the other models, this occurred in 1969. Thisdivergence is related to the quantitative detection of small Thalassiosira spp. inthe Narragansett Bay samples beginning in 1966, and routinely found thereafter.By 1967, this assemblage established itself as a major component in the phyto-plankton community, and subsequently has become the sixth most numericallyabundant 'taxon' in the 22-year time series (Karentz and Smayda, 1984). Whenfirst observed in the samples, these cells were grouped as a single species. Subse-quent taxonomic studies indicate that several nanoplanktonic (<10 urn) Thalas-siosira species comprise this 'taxon' which are exceedingly difficult to distinguishduring routine phytoplankton enumeration. Electron microscopy is required,which Gayoso (1985) used to describe the new species, Thalassiosira solitaria.Their grouped inclusion here as Thalassiosira spp. does not alter the ensuing SDAresults and interpretations.
Collectively, the five species most abundant numerically accounted for 84% ofthe total cells represented by the 49-species model (Skeletonema costatum aloneaccounted for 64%). SDA of these five species produced a result strikingly differ-ent from that in the 49-, 24-, 10- (not shown) and 6-species models (Figure 6D).Asterionellopsis glacialis was entered into the model first because it was the mostvariable in terms of its annual occurrence patterns (i.e. it is the best discrimina-tor of the temporal groups). It was followed by Heterosigma akashiwo, Thalas-siosira nordenskioeldii, Detonula confervacea and S.costatum. The year-groups inthis 5-species model are tightly clustered; this reflects the repetitive nature of theirindividual annual occurrence patterns. Despite this tight clustering, the distinc-tion between the 1960s and 1970s year-groups persists. The 1960s year-groups aredistributed across CV1; those from the 1970s tend to be aligned along CV2. The5-species model contains some obvious outlying years (1960-1961, 1965,1967-1969, 1972 and 1977). In five of these eight years (1960-1961; 1967-1969),large populations of A.glacialis occurred during the fall and winter. These unusualyears in Narragansett Bay have been designated as 'Asterionellopsis years'(TJ.Smayda et ai, in preparation). The three other outlying year-groups (1967,1972,1977) correspond to years of peak abundance in the 5-year abundance cycleexhibited by T.nordenskioeldii (Karentz and Smayda, 1984).
The outlying years of the 5-species model also generally correspond to keyyears recognizable in the 49-species model. For example, 1965, which initiated thesecond pentade cycle within the 1960s cluster in the 49-species model, is also dis-tinct from the main 1960s cluster in the 5-species model. Total phytoplankton
156
Phytoplankton in Narragansett Bay during 1959-1980
abundance, particularly for diatoms, was low during 1965; the only year in the1960s when the annual maximum occurred during summer. Low mean annualabundances were recorded for four of the five species (D.confervacea was theexception). Skeletonema costatum and T.nordenskioeldii exhibited their 22-yearminima in 1965, and the annual maxima for S.costatum and A.glacialis thenshifted from the colder periods of the year to summer (Figures 7 and 8).
The years 1965-1969 are distributed horizontally across the entire range ofCV1 in the 5-species model. During this 5-year period, the abundance of severalspecies decreased sharply, including: H.akashiwo, Cerataulina pelagica, Kato-diniwn rotundatum, Chaetoceros gracilis and Gymnodinium sp. Chaetoceros com-pressus was not detected during 1965-1969, although very large populationsaccompanied the previous increase in A.glacialis during the 1960-1961 winter.Chaetoceros similis, very variable in the late 1960s, was not recorded after 1969.Between 1965 and 1969,14 of the 49 species attained their 22-year maximum inabundance (Table I), and five species attained their second largest mean weeklydensities: Leptocylindrus minimus (1965), Thalassionema nitzschioides (1965),Cylindrotheca closterium (1965), Phaeocystis pouchetii (1966) and Pyramimonastorta (1969). Six species first observed in Narragansett Bay during the late 1960ssubsequently established numerically important populations: Thalassiosira spp.(1966), Chaetoceros subtilis (1966), Ch.aflinis (1967), Ch.diadema (1968), Cryp-tomonas (?Chroomonas) amphioxeia (1968) and Thalassiosira pseudonana(1968).
SDA of the 'summer-winter' successional pattern
The 49- and 5-species models were used to evaluate semi-annual successional pat-terns along the interactive temperature and light continuum (Figure 9A). The'summer' and 'winter' groups of the 49-species model segregated along CV2 andaligned along CV1 in a chronological continuum, and generally exhibited cycli-cal trends similar to those which characterized the annual groups model (Figure6A). This response is consistent with expected summer-winter patterns in tem-perature and light. The two canonical variables account for 56% of the variationwithin the species data. 'Summer' patterns during the late 1970s and in 1980 areclosely related; 'winter' patterns are more variable and possibly cyclical. In thisSDA treatment, the shift in successional patterns between the 1960s and 1970scan be specifically identified as having occurred during the summer of 1969. Thiscommunity change preceded a similar shift evident in the 'winter' group (Novem-ber 1969-May 1970) community structure.
In the model with the five dominant species, 'summer' and 'winter' assemblagesare again clearly separated along CV1 (Figure 9B). Since three of the five specieshave distinct seasonal patterns of occurrence (Figures 7 and 8), this separationlinked to seasonal light and temperature patterns is not surprising. Detonula con-fervacea and T.nordenskioeldii occur only during winter, and H.akashiwo isobserved only in summer and autumn. Skeletonema costatum and A.glacialisoccur year round. Unlike the 49-species model, a linear chronology is not evident.'Summer' groups are more densely clustered, and short-term cycles, where
Fig. 7. Annual quarterly means of cell abundance for Skeletonema costatum, Detonula confervaceaand Thalassiosira nordenskioeldii during 1959-1980. Numbers above standard error bars are asdefined in Figure 5.
158
Phytoplankton in Narragansett Bay during 1959-1980
Fig. 8. Annual quarterly means of cell abundance for Asterionellopsis glacialis and Heterosigmaakashiwo (= Olisthodiscus luteus) during 1959-1980. Numbers above standard error bars are asdefined in Figure 5.
evident, are superimposed. Within the cluster of 'summer' groups, years ofhighest abundance of H.akashiwo (1959-1962, 1976, 1978-1980) have thestrongest positive correlations with CV1. Groups having high positive correla-tions with CV2 (1970,1973,1975) are years when the largest 'summer' blooms ofA.glacialis occurred. 'Winter' groups varied considerably between successiveyears and exhibited random chronological patterns. As in the summer cluster,'Asterionellopsis years' (fall and winter of 1961-1962 and 1967-1969) are groupedat the positive end of CV2. The two canonical variables account for 73% of thevariation within the species data.
SDA of the seasonal successional pattern
SDA of the seasonal successional patterns based on quarterly groupings of the 49species also separates the 1960s year-groups from those during the 1970s (Figure10). The July-September groups show a significant shift in 1969 away from theearly 1960s to form a very compact cluster of year-groups ending with 1974(Figure IOC). This quarterly model confirms that a precipitous change in theannual successional patterns observed during the 22-year observational period
159
D.Karentz and XXSmayda
occurred in 1969 (Figure 6), and indicates that this change was initiated duringthe summer of 1969. The close association of the 1969-1974 year-groups showsthe large degree of similarity in the taxonomic structure and successionalsequences of the summer phytoplankton communities during this period. The dis-tinction between the two decades initiated during the 1969 summer persisted inthe October-December groups (Figure 10D). The marked deviation in taxon-omic structure during 1976 accompanied one of the coldest seasons in the NewEngland area (see also Figure 4) (Stone et al., 1980). The mean temperature forOctober-December 1976 was 3.8°C below the 22-year average. These decadaldifferences in the taxonomic structure and abundance of the phytoplankton com-munities reflected in their 'summer-winter' (Figure 9) and seasonal (Figure 10)successional patterns accompanied the significant long-term changes in tempera-ture and light which characterize the 22-year time series (Figures 1 and 2; Karentzand Smayda, 1984; Smayda, 1984). However, the two canonical variables accountfor <50% of the variation within the species data.
Discussion
Discriminant analyses have previously been used to evaluate stratification ofspecies composition in benthic communities (Walker et al., 1979), to describerelationships between surf-zone diatom blooms and hydrographic factors(Garver and Lewin, 1981), to examine the role of environmental factors in thespatial and temporal changes of both planktonic and benthic diatom communitystructure (Karentz and Mclntire, 1977; Amspoker and Mclntire, 1978; Zurlini etal., 1983; Goodman et al., 1984), and in the prediction of red tide occurrences(Ouchi, 1984). Advantages of using discriminant analysis with time series phyto-plankton data include: changes in community taxonomic structure can be quan-tified relative to defined temporal categories; SDA facilitates comparison of thelong-term trends, cycles and successional patterns exhibited by selected groupsof species; and SDA results can be presented parsimoniously. In the presentstudy, we were able to condense 22 years of data onto one plot that summarizesthe relative degree of similarity in species successional patterns within a givenchronology (annual, semi-annual and seasonal) and of selected phytoplanktonassemblages.
Segregation of the 1960s and 1970s year-groups was the most obvious and con-sistent feature of the phytoplankton and environmental data sets. These decadaldifferences were not intuitively obvious prior to SDA evaluation. Only after thelatter consistently indicated that these year-groups were separable did we re-examine the data from a different perspective. This re-examination confirmed theSDA results evident in all temporal subgroupings of the data set: 1969 was thekey transitional year in the observed decadal shift in phytoplankton taxonomicstructure and seasonal abundance. Subsequent review of the raw data revealedthat spring 1969 was not only the second wannest spring in the 22-year time series,but the annual phytoplankton maximum in 1969 occurred during April, ratherthan during winter, as in previous years (excluding 1966), or during summer, asin the majority of later years. The significant shift in the taxonomic character of
160
Phytoplankton in Narragansett Bay during 1959-1980
49 Species 1959-1980
CVl + CV2 • 40%
- I
-2
-3
<O
I -
0 -
- I -
-vr-79
B
66*67
1—
-4— i —-3 - I I 2
5 Species 1959-1980
CVl + CV2 = 73%
59
WINTER
SUMMER
-I
CVl
Fig. 9. SDA for 'summer' (June-October) and 'winter' (November-May) groupings of species data,with 40% of the variation accounted for by canonical variables CVl and CV2. (A) 49-species model;(B) 5-species model.
161
D.Karentz and TJ^mayda
49 Sp.cl.t 1959-1980Joo-Uor
CVI+CV2 - 4 6 %
Apr*Jun«CVl + CV2 - 4 6 %
CVI+CV2-48*
iQm -T%
0cl-0«cCVI+CVZ-39%
-8 - 6 - 4 - 2 0 2 4
CVl
Fig. 10. SDA for quarterly groupings of 49 species, and percentage of its variation accounted for bycanonical variables CVl and CV2.
162
Phytoplankton in Narragansett Bay during 1959-1980
the phytoplankton community between the 1960s and 1970s can also be attrib-uted to a 1969 inception, specifically to the summer of 1969. This appears to bepartly linked to the increased summer occurrence of small Thalassiosira spp., firstdetected in 1966 during the weekly phytoplankton surveys, and by 1969 becameestablished as the sixth most important taxon in Narragansett Bay (Karentz andSmayda, 1984). The apparent impact of this assemblage on altered phytoplanktoncommunity structure is verified by the conspicuous SDA differences between the5-species model, in which Thalassiosira spp. are omitted, and the 6-, 10-, 24-, 30-and 49-species models, in which it was included. An influence of other species onthis decadal shift in taxonomic character is less obvious. A number of significantchanges in species occurrence and abundance began prior to the decadal shift in1969, notably in 1965, as described above. Thus, several simultaneous trends andcycles of irradiance, temperature and phytoplankton community structure pre-ceded the conspicuous shift that commenced between decades.
We previously examined the relationship between the annual temperature andirradiance gradients in Narragansett Bay and the seasonal occurrence and abun-dance of a number of dominant phytoplankton species (Karentz and Smayda,1984). Despite general agreement between their in situ occurrences and responsesin laboratory experiments assessing their temperature-growth relationships, tem-perature was not a useful predictor in modeling the initiation, development ormaintenance of individual populations for most of these species in NarragansettBay. In contrast, the SDA results suggest that the warm spring temperaturesduring 1969 influenced the remarkable and unusual events in the annual phyto-plankton cycle that year, namely, the occurrence of the annual maximum inspring, not winter, and the significant floristic shift observed that summer. More-over, the warm winter temperatures subsequently during the 1970s generallywere accompanied by low winter phytoplankton abundance, with the annualphytoplankton maximum frequently occurring during summer (Smayda, 1984).These events, directly or indirectly, apparently altered subsequent phytoplanktonsuccessional cycles in Narragansett Bay, and suggest a cause-and-effect relation-ship between the decadal patterns of phytoplankton succession and light and tem-perature.
The extent to which herbivory and allelopathy influenced these altered long-term trends and interannual variations in phytoplankton assemblages isunknown. Previous studies in Narragansett Bay have shown that herbivorousgrazing modifies phytoplankton species composition and abundance (Martin,1965,1970; Smayda, 1973b; Tomas and Deason, 1981), and that allelopathic inter-actions between phytoplankton species occur (Pratt, 1966; Smayda, 1973b). Onewould expect that the temporal displacement of a phytoplankton bloom and itscommunity restructuring, such as reported here, would partially alter interactionsbetween various trophic levels. Indeed, correspondence analysis of fish popu-lations in Narragansett Bay indicates that a similar decadal shift in communitystructure occurred between the 1960s and 1970s, with 1969 being the transitionalyear (Jeffries and Terceiro, 1985). A peak period for winter flounder, Pseudo-pleuronectes americanus, the most common fish species in Narragansett Bay,occurred during the 'Asterionellopsis years' of 1967-1969. As found for the
163
D.Karentz and TJ.Smayda
phytoplankton, the fish data during these years deviated significantly from thecyclical patterns characterizing other years. More than 75% of the interannualvariations in winter flounder abundance in Narragansett Bay could be accountedfor by very small changes in mean annual water temperature (Jeffries andJohnson, 1974,1976; Johnson, 1980). These variations were attributed to thermaleffects on the metamorphosis of winter flounder larvae during April. Duringyears when this month was warmer (such as 1969), the larvae developed morerapidly, but had a higher mortality rate, resulting in decreased adult populations2-3 years later (Johnson, 1980). Decreases in both larval and adult flounder wouldhave immediate and long-term repercussions in the trophic structure of Narra-gansett Bay. Any alteration in grazing pressure on zooplankton would influencephytoplankton composition, abundance and their interannual and seasonal vari-ations, such as reported here.
Clearly, a fundamental ecosystem change occurred in Narragansett Baybetween the 1960s and 1970s, with 1969 the transitional year: changes beingevident in temperature, irradiance, phytoplankton and winter flounder levels.Although their quantitative inter-relationships remain obscure, available evidencecollectively suggests that the warm spring temperatures of 1969 may have initiatedsignificant changes in the biology of Narragansett Bay. The shift in timing of theannual phytoplankton maximum from being predominantly a winter event in the1960s to, increasingly, a summer occurrence in the 1970s is another manifestationof this ecosystem change. In this case, the limited data on ctenophore-zooplank-ton-phytoplankton interactions in Narragansett Bay suggest that the increased,though episodic, abundance of the ctenophore Mnemiopsis leidyi during the 1970smay have initiated and regulated this seasonal shift in the timing of the annualphytoplankton maximum. Abundance of this carnivorous ctenophore andsummer phytoplankton levels were directly related during 1972-1977 (Deasonand Smayda, 1981 a,b). Mnemiopsis leidyi grazes on copepods, which reduces thelatter's grazing pressure on the phytoplankton. Summer populations of the diatomS.costatum were lowest during 1972,1976 and 1977, years when mean populationdensities of Mnemiopsis were also lowest and zooplankton abundance high;during these years, the annual phytoplankton maximum occurred during winter.During 1973-1975, high ctenophore densities and summer abundances of S.costa-tum co-occurred; in those years, the annual phytoplankton maximum occurred inAugust. Although quantitative data on ctenophore populations prior to 1972 donot exist, Martin (1970) noted that a large increase in ctenophores during the 1965summer relative to the previous 3 years was accompanied by decimation of thenon-gelatinous zooplankton populations. Interestingly, 1965 was the only year inthe 1960s when the annual phytoplankton maximum occurred in the summer. Pre-sumably, increased ctenophore grazing on herbivorous zooplankton reduced theirgrazing pressure on the phytoplankton and allowed for a large summer bloom.Lindahl and Hernroth (1983) have suggested that episodic incursions of thescyphomedusan, Aurelia aurelia, a zooplankton carnivore, were responsible forsome unusual phytoplankton blooms in Gullmar Fjord, Sweden.
Our final major finding is that three 5-year phytoplankton cycles occurredbetween 1959 and 1974. During each pentade, the phytoplankton community
164
Phytoplankton in Narragansett Bay during 1959-1980
returned to a similar taxonomic organization after diverging in the interveningyears. Thus, year-groups 1959 and 1963 were similar to each other during the firstpentade, as were 1964 and 1968 during the second pentade, and 1969 and 1974during the third pentade. The distinction between the first two pentades during1959-1968 is less pronounced than that between the second and third pentades.The significant difference between 1968 and 1969 heralded the third 5-year cycle,but the significant community shift between 1974 and 1975 did not result in a new5-year cycle; rather, each year thereafter through 1980 (the final year of the timeseries analyses) diverged considerably from the preceding year. Communitystructure and dynamics, at least as described by SDA, differed considerablyduring this period from those occurring during the 1960s through mid-1970s. Thecauses of the remarkable 5-year cycling patterns characterizing the first 15 yearsof the data set, and the subsequent disappearance of this cyclicity, are unknown.It is notable that these 5-year cycling patterns are embedded within the 22-yearlong-term trend in which phytoplankton abundance progressively decreasedduring the winter quarter and increased during the summer quarter (Smayda,1984).
Pratt (1959), based on a 5-year (1952-1956) data set, described four 'naturalphases' of phytoplankton species succession in Narragansett Bay, characterizedby an alternation of diatom and flagellate dominance. Diatoms dominated thewinter-spring and late-summer periods; flagellates predominated during latespring-early summer and during autumn. These temporal recurrence patterns ofthe numerically dominant species in Pratt's 5-year study persisted through 1980,and subsequently (TJ.Smayda, unpublished). The small Thalassiosira species,first detected in 1966, are a notable addition to Pratt's list of dominants. Thecausative factors of their emergence, persistence and overall ecological signifi-cance in Narragansett Bay remain to be established, including the enigmatic roleof nutrients in these occurrences. Bioassay of the surface waters over an annualcycle using the closely related nanodiatom Thalassiosira oceanica grown at 16different nutrient enrichment treatments revealed that nitrogen limitation usuallyimpeded its growth (Smayda, 1974). In contrast, similar, small Thalassiosira spp.dominated the community in experimental mesocosms enriched daily at rates 16and 32 times greater than the daily riverine inputs of sewage into NarragansettBay, but not in the four lower nutrient treatments ranging from 1 to 8 times(Oviatt et al., 1989). This raises the issue of whether the observed proliferation ofthe Thalassiosira spp. since 1969 represents a response to progressive nutrientbuild-up in the waters at the collection site, even though ambient nutrient levelsdo not signal such an increase (TJ.Smayda, unpublished). Such observationsneed not be contradictory, since the initial response of the ecosystem to nutrientenrichment may be an increase in cellular abundance (i.e. the excess nutrientresults in greater population size), and only after the carrying capacity is reachedmay evidence of increased nutrient loading be detected in the chemical measure-ments. Furnas (1982), using diffusion chamber methodology, established thatsustained, high in situ growth rates (often exceeding 2.0 divisions day"1) charac-terized summer populations of Thalassiosira spp. in Narragansett Bay. Thisbehavior certainly suggests an adequate summer nutrient flow. Related to this
165
D.Karentz and TJ-Smayda
issue is the role of grazing. Growth rates of natural tintinnid populations in Nar-ragansett Bay, unlike for other micro-zooplankton components, were repressedduring summer blooms of the Thalassiosira spp. (Verity, 1986a,b). Experimentsshowed that this inhibition was attributed to the presence of gelatinous threadsextruded by these Thalassiosira spp. which inhibited their ingestion (Verity andVillareal, 1986). Cells lacking these threads were readily ingested by tintinnidsand supported good growth. Thus, whatever the reasons for this enigmaticappearance and subsequent importance of Thalassiosira spp. in NarragansettBay, nutrients and micro-zooplankton grazing effects undoubtedly influenceobserved annual and interannual variability in their blooms and role in alteringphytoplankton community organization.
The observed interannual variations, long-term trends and 5-year cycles ofphytoplankton species in Narragansett Bay (as revealed by SDA) primarilyreflect variable patterns in species combinations, their seasonal occurrence andabundance characteristics, and the timing of their blooms. Such variability is pre-sumably under environmental regulation, although variable seeding mechanismscould also be involved (see Karentz and Smayda, 1984). The greatly reduced andrelatively modest interannual variation in SDA of the five most dominant speciesrelative to the 49-, 30-, 24-, 10- and 6-most-dominant-species models is a particu-larly interesting result. In Narragansett Bay, these five species must be those bestadapted autecologically to local environmental conditions, especially S.costatumand A.glacialis which tend to occur year round. Skeletonema costatum frequentlycontributes >75% of the total annual phytoplankton cell density in NarragansettBay and contiguous waters (Smayda, 1957, 1973a,b). This ubiquity has beenattributed to the existence of genetically differentiated clones which have specificpatterns of seasonal occurrence (Gallagher, 1980). .
The marked contrast in the SDA results between the 5-species and the 49-, 24-and 6-species models suggests that, on a long-term basis, less abundant speciesmay be more important than the major species in determining and modifying thecommunity organization of the phytoplankton assemblages, and their succes-sional and overall bloom patterns. However, temporal shifts in the occurrenceand/or abundance of numerically dominant taxa may be useful indicators ofchanges occurring within the phytoplankton community as a whole. This isespecially true for A.glacialis, whose episodic major blooms are accompanied bynumerous alterations in the occurrence and abundance of less dominant species.Such blooms may also be the diatom analogue of unpredictable flagellate blooms,some of which are harmful in their trophic impact (Smayda, 1992).
Acknowledgements
Dr David Pratt provided phytoplankton data for 1959-1963, Drs Betty Mitchell-Innes, Miles Furnas and Gabriel Vargo assisted with the phytoplankton counts,and David Borkman provided useful information. Drs James Yoder and DeborahFrench assisted with the computer analyses. This research was supported by thefollowing grants awarded to T.J.S.: NSF grants OCE 68-1500, OCE 71-00556,
166
Phytoplankton in Narragansett Bay during 1959-1980
OCE 76-22563, OCE 95-30200 and Department of Commerce (NOAA) grant no.NA80RA-D0006.
ReferencesAllen,T.F.H. and Skagen.S. (1973) Multivariate geometry as an approach to algal community analysis.
Br. Phycol. J., 8, 267-287.Amspoker.M.C. and McIntire,C.D. (1978) Distribution of intertidal diatoms associated with the
sediments in Yaquina Estuary, Oregon. / Phycol., 14,387-395.Cooley.W.W. and Lohnes,P.R. (1971) Multivariate Data Analysis. Wiley & Sons, New York.Deason.E.E. and Smayda.TJ. (1982a) Ctenophore-zooplankton-phytoplankton interactions in
Narragansett Bay, Rhode Island, USA, during 1972-1977. /. Plankton Res., 4,203-217.Deason,E.E. and Smayda.TJ. (1982b) Experimental evaluation of herbivory in the ctenophore
Mnemiopsis leidyi relevant to ctenophore-zooplankton-phytoplankton interactions in NarragansettBay, Rhode Island, USA. /. Plankton Res., 4,219-236.
Dixon.WJ. and Brown.M.B. (1979) Biomedical Computer Programs p-Senes. University of CaliforniaPress, Berkeley, CA.
Furnas.MJ. (1982) Growth rates of summer nanoplankton (< 10 urn) populations in lower Narra-gansett Bay, Rhode Island, USA. Mar. Bio/., 70, 105-115.
GallagherJ.C. (1980) Population genetics of Skeletonema costatum (Bacillariophyceae) in Narra-gansett Bay. J. Phycol., 16, 464-474.
GarverJ.L. and LewinJ. (1981) Persistent blooms of surf diatoms along the Pacific Coast, USA. I.Physical characteristics of the coastal region in relation to the distribution and abundance of thespecies. Estuarine Coastal Shelf Set., 12, 217-229.
Gayoso,A.M. (1985) Thalassiosira solitaria, sp. nov., from Narragansett Bay. Bot. Mar., 28, 477^184.Goodman.D., EppleyJR.W. and ReidJr.M.H. (1984) Summer phytoplankton assemblages and their
environmental correlates in the southern California Bight. /. Mar. Res., 42, 1019-1049.Jeffries.H.P. and Johnson,W.C. (1974) Seasonal distributions of bottom fishes in the Narragansett Bay
area: seven-year variations in the abundance of winter flounder (Pseudopleuroneaes americanus).J. Fish. Res. Board Can., 31,1057-1066.
Jeffries.H.P. and Johnson.W.C. (1976) Petroleum, temperature, and toxicants: examples of suspectedresponses by plankton and benthos on the continental shelf. In Manowitz,B. (ed.), Effects ofEnergy-Related Activities on the Atlantic Continental Shelf. Brookhaven National Laboratory,Assoc. Univ., Inc., BNL50484, pp. 96-108.
Jeffries.H.P. and Terceiro.M. (1985) Cycle of changing abundances in the fishes of the NarragansettBay area. Mar. Ecol. Prog. Ser., 25,239-244.
Johnson.W.C. (1980) The response of selected marine species to subtle long-term climatic variations.PhD Dissertation, University of Rhode Island, 137 pp.
Karentz,D. and McIntire.C.D. (1977) Distribution of diatoms in the plankton of Yaquina Estuary,Oregon. / Phycol., 13, 379-388.
Karentz.D. and Smayda.TJ. (1984) Temperature and the seasonal occurrence pattern of 30 dominantphytoplankton species in Narragansett Bay over a 22-year period (1959-1980). Mar. Ecol. Prog.Ser, 18, 277-293.
Lindahl.O. and Hernroth.L. (1983) Phyto-zooplankton community in coastal waters of westernSweden—an ecosystem off balance? Mar. Ecol. Prog. Ser., 10,119-126.
MartinJ.H. (1965) Phytoplankton-zooplankton relationships in Narragansett Bay. Limnol.Oceanogr., 10,185-191.
MartinJ.H. (1970) Phytoplankton-zooplankton relationships in Narragansett Bay. IV. The seasonalimportance of grazing. Limnol. Oceanogr., 15,413-418.
Ouchi,A. (1984) Prediction of red tide occurrence by means of discriminant analysis. Bull. Jpn Soc.Sci. Fish., 50, 1647-1651.
Oviatt.C, Lane.P., French,F.,III and DonaghayJ?. (1989) Phytoplankton species and abundance inresponse to eutrophication in coastal marine mesocosms. / Plankton Res., 11,1223-1244.
Pratt,D.M. (1959) The phytoplankton of Narragansett Bay. Limnol. Oceanogr., 4, 425-440.Pratt,D.M. (1966) Competition between Skeletonema costatum and Olisthodiscus luteus in Narra-
gansett Bay and in culture. Limnol. Oceanogr., 11, 447-455.Smayda.TJ. (1957) Phytoplankton studies in lower Narragansett Bay. Limnol. Oceanogr., 2,342-359.Smayda.TJ. (1973a) A survey of phytoplankton dynamics in the coastal waters from Cape Hatteras
to Nantucket. In Coastal and Offshore Environmental Inventory, Cape Hatteras to Nantucket.University of Rhode Island Marine Publication Series, 2, pp. 1-100.
167
D.Karentz and TJ^mayda
Smayda.TJ. (1973b) The growth of Skeletonema costatum during a winter-spring bloom in Narra-gansett Bay, Rhode Island. Norw. J. Bot., 20,219-247.
Smayda.TJ. (1974) Bioassay of the growth potential of the surface water of lower Narragansett Bayover an annual cycle using the diatom Thalassiosira pseudonana (oceanic clone, 13-1). LimnoLOceanogr., 19, 889-901.
Smayda.TJ. (1984) Variations and long-term changes in a phytoplankton-based coastal marineecosystem: relevance to field monitoring for pollution assessment. In White^H.H. (ed.), Conceptsin Marine Pollution Measurements. Maryland Sea Grant Publication, University of Maryland,College Park, MD, pp. 663-679.
Smayda.TJ. (1992) Global epidemic of noxious phytoplankton blooms and food chain consequencesin large ecosystems. In Sherman,K., Alexander.L.M. and Gold,B.D. (eds), Food Chains, Yields,Models and Management of Large Marine Ecosystems. Westview Press, Boulder, CO, pp. 275-307.
Smayda.TJ. and Villareal.T. (1989) The 1985 'brown tide' and the open phytoplankton niche in Narra-gansett Bay during summer. In CosperJ£.M., Bricelj.V.M. and Carpenter,EJ. (eds). Novel Phyto-plankton Blooms. Springer-Verlag, New York, pp. 159-187.
Stone J.H., DayJ.W. Jr, Bahr,L.M. Jr and Muller.R.A. (1978) The impact of possible climatic changeson estuarine ecosystems. In Wiley.M.L. (ed.), Estuarine Interactions. Academic Press, New York,pp. 305-322.
Tomas.C.R. and Deason.E.E. (1981) The influence of grazing by two Acartia species on Olisthodis-cus luteus Carter. Mar. Ecoi, 2, 215-233.
Verity,P.G. (1986a) Growth rates of natural tintinnid populations in Narragansett Bay. Mar. EcoLProg.Ser., 29, 117-126.
Verity.P.G. (1986b) Grazing of phototrophic nanoplankton by microzooplankton in NarragansettBay. Mar. Ecol. Prog. Ser., 29,105-115.
Verity.P. and Villareal.T.A. (1986) The relative food value of diatoms, dinoflagellates, flagellates, andcyanobacteria for tintinnid ciliates. Arch. Protistenkd., 131, 71-84.
Walker.H.A., Saila.S.B. and Anderson.E.L. (1979) Exploring data structure of New York Bightbenthic data using past-collection stratification of samples, and linear discriminant analysis forspecies composition comparisons. Estuarine Coastal Mar. Sci., 9,101-120.
Zurlini.G., Zattera,A. and Bruschi.A. (1983) Structural analysis of phytoplankton communitiesvariation in the archipelago of la Maddalena (north Sardinian coast): a canonical correlationapproach. /. Exp. Mar. Biol. EcoL, 70, 227-248.
Received on March 7, 1995; accepted on September 15, 1997
![Variations in the light absorption coefficients of ...geoconvention.org/wp-content/uploads/2016/10/organic-content-variations.pdf[1] We measured the absorption properties of phytoplankton,](https://static.documents.pub/doc/80x56/60afe7c04047e43c6d2daccb/variations-in-the-light-absorption-coefficients-of-1-we-measured-the-absorption.jpg)
