Estuaries and Coasts (2010) 33:286–294
DOI 10.1007/s12237-009-9236-y
Long-term Field Observations on Seasonality in Chlorophyll-a
Concentrations in a Shallow Coastal Marine Ecosystem,
the Wadden Sea
Catharina Johanna Maria Philippart &
Jolanda Martine van Iperen &
Gerhard Cornelis Cadée & Alain François Zuur
Received: 21 July 2008 / Revised: 6 August 2009 / Accepted: 6 November 2009 / Published online: 8 December 2009
# The Author(s) 2009. This article is published with open access at Springerlink.com
Abstract Analyses of long-term field observations (1974–
2007) on chlorophyll-a concentrations in the western
Wadden Sea showed no long-term trends in the timing of
the wax and wane of phytoplankton spring blooms. There is
weak evidence, however, that the height of the autumn
bloom has decreased since the early 1990s. This fading of
the autumn bloom may have had consequences for the
carbon transfer to higher trophic levels, currently hampering
primary consumer species that mostly rely on food supply
during late summer. Current and other findings suggest a
shortening of the growing season due to the fading of the
autumn bloom in the Wadden Sea and a lengthening of the
growing season due to an advancement of the spring bloom
in the North Sea. These regionally different changes in
seasonality may have contributed to the coinciding decrease
in bivalve filtering capacity in the western Wadden Sea and
the large-scale offshore shift of juvenile plaice from the
Wadden Sea to the adjacent North Sea.
Keywords Long-term dynamics . Phytoplankton .
Chlorophyll-a . Phenology . Coastal marine ecosystems
C. J. M. Philippart (*) : J. M. van Iperen : G. C. Cadée
Department of Marine Ecology,
Royal Netherlands Institute for Sea Research,
P.O. Box 59, 1790 AB Den Burg, The Netherlands
e-mail: katja.philippart@nioz.nl
A. F. Zuur
Highland Statistics Ltd,
Laverock Road,
AB41 6FN Newburgh, UK
A. F. Zuur
Oceanlab, University of Aberdeen,
Main Street,
AB41 6AA Newburgh, UK
Introduction
In temperate marine waters that get stratified during
summer, time series of phytoplankton biomass show
recurring spring bloom events to begin anytime from late
winter to early spring (Townsend et al. 1994). Each bloom
is characterized by a period of rapid algae growth followed
by a crash in population numbers, after which there is a
period where phytoplankton numbers remain minimal.
Hereafter, phytoplankton may bloom again in late summer
before numbers get low for a longer period during the cold
and dark winter. Shallow non-stratified coastal waters, such
as estuaries, lagoons, and bays, however, are very diverse
systems, and the relative importance and interplay of the
various factors that can potentially affect bloom initiation
and development show much more variation. For example,
such waters may be characterized by series of pulses of
growth during spring and summer (e.g., Colebrook and
Robinson 1965; Riley 1967; Cadée 1986a). Typical patterns
of phytoplankton variability over years include episodic
blooms, cyclical blooms, seasonal shifts, long-term trends,
stochastic blooms, and novel appearances (Smayda 1998).
The underlying mechanisms that control the timing of onset
and duration of the spring bloom can be very different between
areas and between years (Townsend et al. 1994; Cloern 1996;
Smayda 1998), depending on the interplay of multiple
factors, such as day length, light conditions, nutrient
availability, and grazing (Iriogen et al. 2005). Although
observed interannual variation in timing of onset and duration
of blooms is more often attributed to external forcing such as
weather conditions and nutrient concentrations than internal
dynamics (e.g., Smayda 1998), modeling exercises suggest
that such year-to-year variations can be also be fully
explained by multispecies interactions (Dakos et al. 2009).
Estuaries and Coasts (2010) 33:286–294
287
The timing of onset and duration of blooms is
particularly important with respect to food web dynamics
in these shallow marine systems. Consumption of the
bloom by planktonic heterotrophs and filter-feeding
bivalves is expected to be much reduced when the bloom
occurs in relatively cold waters, e.g., when phytoplankton
blooms relatively early in the year (Prins et al. 1994;
Townsend et al. 1994). For invertebrates with a complex
lifecycle, optimizing the time of spawning is necessary to
match reproduction with the most optimal environmental
conditions for the first vulnerable life stages (i.e., the
settlement-timing hypothesis by Todd and Doyle 1981). In
temperate waters, many larvae of marine macrozoobenthos
species are planktotrophic, i.e., their food stores are not
sufficient to survive the pelagic phase without ingesting
exogenous food (Thorson 1950). Although there is still
much debate on whether growth and development of
invertebrate larvae are food-limited under natural conditions
(e.g., Strathmann 1996), there may be a causal relationship
between the timing of the planktonic food supply and
subsequent bivalve recruitment success (Philippart et al.
2003).
The present paper deals with the long-term dynamics
in seasonality of phytoplankton in the western Wadden
Sea. Data analysis is based on a long-term dataset
originating from one station near the Marsdiep tidal inlet
Fig. 1 Location of the sampling
station in the western part of the
Wadden Sea, a shallow coastal
sea at the southeastern border of
the North Sea (N Norway, DK
Denmark, G Germany, NL The
Netherlands, B Belgium, F
France, UK United Kingdom)
3˚0'0"W
over a study period of 34 years (1974–2007). Previous
analyses of parts of this dataset revealed that the
phytoplankton community changed drastically both
around 1978 and again around 1988 and that it was
relatively stable in-between and hereafter (Philippart et
al. 2000). The major changes in phytoplankton community structure coincided with changes in absolute and
relative nutrient concentrations (Philippart et al. 2000)
and were followed by changes in community structures of
macrozoobenthos and estuarine birds (Beukema et al.
2002; Philippart et al. 2007). In this paper, we focus on the
between-year variation in the timing of the wax and wane
of phytoplankton blooms as derived from chlorophyll-a
measurements.
Materials and Methods
Long-Term Field Observations
From 1974 onwards, bucket water samples have been
collected at high tide from the NIOZ sampling jetty
(53°00′06″ N 4°47′21″ E), located in the Marsdiep tidal
inlet between the North Sea and the Wadden Sea (Fig. 1).
Phytoplankton sampling frequency was 40–60 times per
year, varying from once or twice a month in winter up to
0˚0'0"
3˚0'0"E
6˚0'0"E
9˚0'0"E
N
58˚0'0"N
58˚0'0"N
North Sea
DK
56˚0'0"N
Wadden Sea
54˚0'0"N
NIOZ
sampling
station
UK
52˚0'0"N
3˚0'0"W
54˚0'0"N
G
52˚0'0"N
NL
B
F
0˚0'0"
3˚0'0"E
56˚0'0"N
6˚0'0"E
9˚0'0"E
288
Estuaries and Coasts (2010) 33:286–294
50
twice a week during phytoplankton spring blooms (e.g.,
Cadée and Hegeman 2002).
Chlorophyll-a concentrations were assessed from additional 0.5 to 1 l water samples, filtered over MgCO3-coated
filters (Cadée and Hegeman 2002). The concentrations in
previous publications were calculated from the difference
in spectrophotometric measurements of extinctions at
666 nm before and after acidification of the samples to
correct for the extinction by chlorophyll degradation
products (phaeopigments) in coastal waters at this
wavelength (Lorenzen 1967). Serious doubts on the
applicability of acidification to correct for the extinction
by phaeopigments (Stich and Brinker 2005) motivated us
to use the original data of non-acidified values (total
chlorophyll-a) in the present paper. On average, this backcalculation resulted in higher chlorophyll-a concentrations
(n=1,344, r2 =0.98, CHLnon-acidified =1.174 CHLacidified).
a
0
spring bloom
0
100
1.0
mg Chlorophyll-a m-3 d-1
b
Changes in seasonality of blooms were examined by means
of two approaches, viz. a first approach in which the data of
each year were analyzed separately for the timing of the
spring bloom by means of determination of maximum and
minimum daily changes in chlorophyll-a concentrations
and a second approach at which the full dataset was
examined which allowed for multiple blooms and possible
auto-correlation structures.
For the first approach, we applied a generalized additive
model (GAM) with a Gamma distribution and log-link
function on data of each year, provided there were
sufficient observations (>25) in that year. Hence, the
following model is applied:
max
0.5
200
300
400
300
400
0.0
-0.5
-1.0
min
0
100
200
0.04
c
max
-1
0.02
0.00
d
Chlorophylls Gammaðms ; vÞ
and varðChlorophylls Þ ¼
20
10
Data Analyses
E ðChlorophylls Þ ¼ ms
30
spring bloom
mg Chlorophyll-a m-3
40
m2s
v
-0.02
min
ms ¼ eaþf ðDayÞ
-0.04
where Chlorophylls is the observed chlorophyll-a at time s in
a particular year. This model estimated daily chlorophyll-a
concentrations by interpolating observed chlorophyll-a concentrations in that year (Fig. 2a).
The onset of the spring bloom was defined as the date at
which the maximum daily increase in the interpolated
chlorophyll-a concentrations occurred, for the timing of the
breakdown of this bloom we used the date for which the
minimum value of the maximum daily increase in
chlorophyll-a concentrations was found (Fig. 2b). To
determine a more biologically realistic determination of the
onset of the bloom, it would have been better to calculate
specific net growth rates instead of daily increases. If we had
used these growth rates, however, we could not have
0
100
200
300
400
day number
Fig. 2 Example of determination of timing of annual start and
breakdown of phytoplankton bloom as derived from long-term
observations on chlorophyll-a concentrations, viz. a interpolation of
observations on chlorophyll-a concentrations (mg m−3) by means of a
model, b identification of maximum and minimum values of daily
change in chlorophyll-a concentrations (mg m−3 day−1), and c
identification of maximum and minimum values of specific growth
rates (day−1). The field observations on which these estimations are
based upon were made in 1988
determined the end of the spring bloom as the minimum
value of this parameter (Fig 2c). Results based upon daily
changes should, therefore, be considered as consistent but
rough approximations of the timing of the spring bloom.
Estuaries and Coasts (2010) 33:286–294
289
300
1975
0 20 40
1992
1984
1985
1976
0 100
2000
1993
1986
1978
300
2001
1994
1979
0 100
300
2002
10 30
1999
010 30
1991
2007
1995
0 50 100
1998
0 20 40 60
0 20 40
2006
0 20 60
10 30 50 0 20 40 60
300
0 20 60
0 20 40
0 100
1983
0 20 40 60 0 204060
1982
5 10 20
10 30
0 204060
1990
0 5 10 20
30
10
1997
1989
2005
0 20 60
2004
1996
1974
0 100
1987
1988
010 30
300
10 30
5 15 25
2003
0 20 40
mg chlorophyll-a m−3
5 15 2535
0 100
1980
1981
05 15 25
The dispersion is determined by v−1; a small value of v
(relative to μ2) implies that the spread in the data is large. We
used two different GAMM models; the first model contains a
2-dimensional smoother, and the second model consists of a
smoother for day of the year and a smoother for year. Thin
where εs is the residual at time s, k is the time lag (in days),
ρ is an unknown parameter that has to be estimated. This
approach allows for temporal correlation between 2 days
that gets smaller for increasing time lags. However, our data
are irregularly spaced in time; therefore, the AR-1 correlation had to be replaced by something more appropriate.
Pinheiro and Bates (2000) discussed various correlation
020 60
m2s
v
020 60
varðChlorophylls Þ ¼
0 204060
and
corð"s ; "s k Þ ¼ rk
0 2040 60
E ðChlorophylls Þ ¼ ms
The term µs is the expected chlorophyll-a concentration at
time s. The covariate Day is coded as the day of the year that
sampling took place and has values between 1 (in 2004) and
364 (in 1993 and 1996). Year has values from 1974 to 2007.
The term α is an intercept. Model 2 contains two smoothers;
the long-term trend f2, and the seasonal effect is modeled
with f1. This model assumes that the seasonal effect is the
same in each year. Model 1, on the other hand, allows for a
change of the seasonal effect over the years.
To allow for autocorrelation between sequential observations, a residual correlation structure was added to the
model. For regular spaced data, the autoregressive correlation of order 1 (AR-1) is a sensible choice (Pinheiro and
Bates 2000; Zuur et al. 2009) and is given by:
0 204060
where Ms is the maximum in year s (or the minimum).
Both time series did not exhibit any strong autocorrelation
as determined by an autocorrelation function, which
justifies the application of GAM without an autocorrelation structure.
For the second approach, we applied a generalized additive
mixed model (GAMM) with a Gamma distribution and loglink function (Wood 2006; Zuur et al. 2009) which was used
to model the chlorophyll-a data for the full dataset. The loglink ensures that fitted values are always positive and the
Gamma distribution uses the following mean and variance
Model 2 : ms ¼ eaþf1 ðDayÞþf2 ðYearÞ
0 40 80
varðMs Þ ¼ s
Model 1 : ms ¼ eaþf ðDay;YearÞ
0 50100
E ðMs Þ ¼ ms and
ms ¼ a þ f ðYearÞ
2
plate regression splines were used for the smoother. Hence,
these models are given by:
10 30
The day on which the maximum and minimum daily
changes were observed were extracted for each year, and
these were further analyzed for the presence of a long-term
trend using a Gaussian GAM with Year as a smoother.
Hence, the following model was applied:
Ms N ms ; s 2
0 100
300
day number
Fig. 3 Interpolation of annual observations on chlorophyll-a concentrations as measured in the Marsdiep tidal inlet at high tide between
1974 and 2007. Black dots indicate field observations and blue line
the interpolation. No data were available for 1977, and the model
could not fit the observations for 1976, 1981, 1983, 1984, and 1986
because not enough data were available for these years
290
2010
a
year
2000
1990
1980
1970
0
100
200
300
400
100
200
300
400
200
300
400
2010
b
2000
year
structures designed for (irregular-spaced) spatial data,
which can easily be adopted for irregular-spaced time series
(see Zuur et al. (2009) for various examples). We used the
continuous autoregressive correlation, and the exponential,
Gaussian, linear, and rational quadratic correlation structures (see Schabenberger and Pierce 2002 for their exact
definitions). The correlation structures require estimation of
two parameters: the range and the nugget. The range
specifies the time frame (in days) in which temporal
dependence can occur, and the nugget is the correlation
between observations separated less than 1 day.
The models were compared using the Akaike's information criterion (AIC), a tool to measure the goodness-of-fit of
an estimated statistical model and model complexity: if
competing models are ranked according to their AIC, the
one having the lowest AIC is the best. Calculations were
carried out in the mgcv (Wood 2006) package from the
software R (R Development Core Team 2008).
Estuaries and Coasts (2010) 33:286–294
1990
Results
1980
1970
0
2010
c
2000
year
For the first approach, years with insufficient information
(1976, 1981, 1983, 1984, and 1986) and no (1977) data on
chlorophyll-a concentrations had to be excluded from
interpolation of field observations for each year separately
by means of the GAM with a Gamma distribution and loglink function. The remaining dataset comprised 1,309
observations on chlorophyll-a concentrations in 30 years,
with an average number of 43.63±11.17 observations per
year (Fig. 3). Fitted values for those 30 years show year-toyear variation in height of the blooms as indexed by
chlorophyll-a concentrations (Figs. 3 and 4) and of the
timing of the spring bloom as described by daily change in
chlorophyll-a concentrations (Fig. 5).
The GAM in which the time series of maximum and
minimum daily changes are modeled as a function of year
gave nonsignificant smoother for the start of the spring
bloom (Table 1; Fig. 6). The GAM of the timing of the end
of the spring bloom gave a smoother with a p value of
0.033. According to Wood (2006), however, p values of a
smoother in the order of 0.01 to 0.05 should be interpreted
with care. Furthermore, the year 1974 was influential, and
the nonlinear pattern of the smoother was mainly determined by the 1974 value (Fig. 6). We, therefore, conclude
that no significant temporal changes in the timing of the
wax and wane of the spring bloom in the Marsdiep tidal
inlet occurred between 1974 and 2007.
The full dataset available for analysis by means of
GAMM with a Gamma distribution and log link function,
i.e., the second approach, consisted of 1,370 observations in
33 years, with an average number of 41.52 ± 12.72
observations per year. The two models without a correlation
1990
1980
1970
0
100
day number
Fig. 4 Days of the year at which chlorophyll-a concentrations were a
higher than 10 mg m−3, b higher than 15 mg m−3, and c higher than
20 mg m−3 as obtained by GAM with a Gamma distribution and loglink function using the annual datasets of field observations in the
Marsdiep tidal inlet between 1974 and 2007
structure have the highest value for AIC (Table 2),
indicating that these models give the worst fit compared
to models that do take correlation structures into account.
This result implies that autocorrelation between sequential
observations does exist.
For all models that do include correlation structures,
models with one 2-dimensional smoother have lower values
Estuaries and Coasts (2010) 33:286–294
291
0 100200300
-0.2 0.2
-0.5
-0.2 0.2
-0.2 0.2
1990
1979
-0.5 0.5
0 100200300
1997
-0.5 0.5
1991
1980
1982
-0.4 0.2
1978
2003
1996
1989
-0.5 0.5
-1.0 0.0
0 100200300
1995
-0.5 0.5
1988
0.5
-0.5
-1 0 1 2
1994
2002
-0.10 0.05
1987
1975
-0.020.01
1974
1993
2001
-1.0 0.0
2000
-0.5 0.5
1999
-0.6 0.0 0.6 -0.4 0.2 0.8 -0.1 0.1
1985
2007
0.5
2006
-0.2 0.2
-0.4 0.20.6 -0.40.0 0.4
-0.2 0.2 0.6 -0.5 0.5
1992
-0.5 0.5
mg chlorophyll-a m −3 d −1
1998
2005
-0.2 0.2
-0.2 0.2
-1.0 0.51.5
2004
0 100200300
0 100200300
day number
Fig. 5 Daily changes in chlorophyll-a concentrations as derived from
interpolation of observations in the Marsdiep tidal inlet at high tide
between 1974 and 2007 for each year separately. No interpolations
could be made for the years 1976, 1977, 1981, 1983, 1984, and 1986
due to insufficient field data
for AIC than those with two 1-dimensional smoothers
(Table 2). This result implies that the seasonal effect was
not the same in each year but has changed between 1974
and 2007.
The model with a 2-dimensional smoother and a rational
quadratic correlation structure (Schabenberger and Pierce
2002) has the lowest AIC and is selected as the best model
(Tables 2 and 3). For this model, the correlation structure
(the estimated range is 9.70 and the nugget is 0.47) indicate
that observational data are no longer correlated if 10 or
more than 10 days apart. A model validation of this model
did not show any patterns in the residuals (data not shown).
The results of the GAMM model fitted to the long-term
field observations on chlorophyll-a concentrations show a
fading of the autumn bloom from the 1970s towards the
2000s (Fig. 7). During the 1970s and 1980s, fitted
chlorophyll-a concentrations were relatively high between
April and June and between mid-July and mid-August
(Fig. 8). The relatively low values during midsummer did
no longer occur since the early 1990s. Since the mid-1990s,
chlorophyll-a concentrations higher than 10 mg m−3 do no
longer occur during August (Fig. 8a). These findings
suggest that fading of the autumn bloom occurred around
the early 1990s.
Table 1 Generalized additive model analysis of the day numbers at
which maximum and minimum daily change occurred, indicating the
respective wax and wane of the spring bloom, as derived from
interpolation of long-term field observations of chlorophyll-a concentrations in the Marsdiep tidal inlet between 1974 and 2007
Timing of spring bloom
Wax
Wane
Estimated df
1.661
2.254
F
1.173
1.921
p Value
0.346
0.033
Discussion
The limits of the dataset should be taken into account when
interpreting the results of the analyses. First, the datasets of
some years have fewer values than others. Second, even
with a frequency up to 60 measurements a year, the precise
timing of the onset of a bloom is very difficult to determine
because wax and wane of phytoplankton may take place at
a much shorter time scales, viz. days instead of weeks (e.g.,
Cadée 1986b). The resulting resolution in the dynamics
implies that our findings should only be considered as an
indication of actual phytoplankton dynamics (Rolinski et al.
2007).
For the western Wadden Sea, the timing of wax and
wane of the phytoplankton spring bloom varied strongly
but did not show mean long-term advancements or delays,
while the autumn bloom appeared to have weakened
between 1974 and 2007 (this paper). Our findings on
timing of the wax of the spring bloom are in agreement
292
Estuaries and Coasts (2010) 33:286–294
1980
1990
2000
Wane
200
120
80
150
100
day number
140
250
Wax
1980
1990
2000
Year
Fig. 6 Annual timing of wax and wane of spring phytoplankton
bloom in the Marsdiep tidal inlet at high tide between 1974 and 2007,
as derived from maximum and minimum values of daily changes in
chlorophyll-a concentrations. A LOESS smoother with a span width
of 0.5 was added to aid visual interpretation
with observations for the eastern (Wiltshire et al. 2008) and
central North Sea (Edwards and Richardson 2004) where
the timing of the onset of the bloom was fairly constant
during the last decades. There is a suggestion, however, of
an advancement of the start and peak of the phytoplankton
spring bloom by 1 month along the Dutch coastline in the
southeastern North Sea (Gieskes et al. 2007; BarettaBekker et al. 2009) since the late 1990s.
The fading of the autumn bloom is in agreement with
observations in the northern Wadden Sea where
chlorophyll-a concentrations in summer (May–Sept) had
decreased since the mid-1990s (Loebl et al. 2008, van
Beusekom et al. 2009). This decrease was explained by a
combination of an enhanced grazing pressure by benthic
filter feeder activity and mesozooplankton due to an
increase in water temperatures and a decrease of the
riverine nitrogen supply (van Beusekom et al. 2009). In the
southwestern part of the Wadden Sea, however, the summed
filtering capacity of the major benthic grazers declined
during the last decades (Philippart et al. 2007). Unfortunately, we do not have long-term field observations on grazing
pressure by zooplankton in this area. Assuming that
Table 2 Estimated AICs for various generalized additive models, with one and two dimensions and a number of correlation structures, of longterm field observations of chlorophyll-a concentrations in the Marsdiep tidal inlet between 1974 and 2007
Model Code
M1
M2
M3
M4
M5
M6
M7
M8
M9
M10
M11
M12
No. of dimensions
2
1
2
1
2
1
2
1
2
1
2
1
Smoothers
Correlation structure
AIC
f(Day, Year)
f1(Day)+f2(Year)
f(Day, Year)
f1(Day)+f2(Year)
f(Day, Year)
f1(Day)+f2(Year)
f(Day, Year)
f1(Day)+f2(Year)
f(Day, Year)
f1(Day)+f2(Year)
f(Day, Year)
f1(Day)+f2(Year)
No correlation
No correlation
Autoregressive correlation
Autoregressive correlation
Exponential correlation
Exponential correlation
Gaussian correlation
Gaussian correlation
Linear correlation
Linear correlation
Rational quadratic correlation
Rational quadratic correlation
2,679.164
2,705.32
2,561.595
2,652.674
2,483.745
2,528.89
No convergence
2,708.841
2,682.103
2,710.160
2,475.328
2,514.791
Estuaries and Coasts (2010) 33:286–294
293
Table 3 Generalized additive mixed model analysis (n=1,370, r2 =0.37) with a rational quadratic correlation structure of long-term field
observations of chlorophyll-a concentrations in the Marsdiep tidal inlet between 1974 and 2007
Intercept (α)
2.33483
SE
t
0.02695
86.64
p Value
<0.0001
zooplankton grazing did not increase since the late 1990s,
the fading of the autumn bloom could be reflecting the
decline of the summer productivity of the phytoplankton as
a result of the decrease in riverine nutrient inputs into the
western Wadden Sea.
These findings point to regional differences in long-term
trends in seasonality, viz. a shortening of the growing
season due to the fading of the autumn bloom in the
Wadden Sea and a lengthening of the growing season due
to an advancement of the spring bloom in the open North
Sea. If changes in chlorophyll-a concentrations reflect
changes in primary productivity, the disappearance of the
relatively high chlorophyll-a concentrations in late summer
in the Wadden Sea may have resulted in decreased growth
rates in primary consumers and have been restrictive to
species that depended on food supply during this season
(e.g., to build up reserves to reproduce or to survive low
food conditions in winter). These changes in seasonality in
the phytoplankton blooms may have contributed to the
coincident shift in bivalve species composition, the decrease in bivalve filtering capacity since the late 1980s
(Philippart et al. 2007), and the large-scale offshore shift of
juvenile plaice from the Wadden Sea to the adjacent North
Sea (Pastoors et al. 2000, Grift et al. 2004).
GAMM
Source
Estimated df
F
p Value
s(Day, Year)
18.82
25.01
<0.0001
2010
a
2000
year
Coefficient
1990
1980
1970
0
100
200
300
400
100
200
300
400
200
300
400
2010
b
2000
year
Variable
1990
1980
1970
40
0
c
2000
20
year
10
1990
m
0
be
1 97 0
40
0
r
30
nu
20
y
0
10
da
0
0
mg Chlorophyll-a m-3
2010
30
19 8 0
1 9 90 r
20 1 0
20 00
1980
yea
Fig. 7 Fitted values on long-term field observations on
chlorophyll-a concentrations, as measured in the Marsdiep tidal
inlet at high tide between 1974 and 2007, obtained by GAMM with
a 2-dimensional smoother. The vertical axis shows the fitted values
(mg chlorophyll-a m−3) and the other two axes the day number (with
values between 1 and 364) and the year (from 1974 to 2007)
1970
0
100
day number
Fig. 8 Days of the year at which chlorophyll-a concentrations were a
higher than 10 mg m−3, b higher than 11 mg m−3, and c higher than
12 mg m−3 as obtained by GAMM with a 2-dimensional smoother
using the full dataset of field observations in the Marsdiep tidal inlet
between 1974 and 2007
294
Acknowledgements We thank Jan Hegeman for his help in
collecting long-term field data, Thalia Watmough for her help in
preparing the map, and Jan Beukema, Justus van Beusekom, Corina
Brussaard, Louis Peperzak, and three anonymous referees for their
comments on previous versions of the manuscript.
Open Access This article is distributed under the terms of the
Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any
medium, provided the original author(s) and source are credited.
References
Baretta-Bekker, J.G., J.W. Baretta, M.J. Latuhihin, X. Desmit, and T.C.
Prins. 2009. Description of the long-term (1991–2005) temporal and
spatial distribution of phytoplankton carbon biomass in the Dutch
North Sea. Journal of Sea Research 61: 50–59.
Beukema, J.J., G.C. Cadée, and R. Dekker. 2002. Zoobenthic biomass
limited by phytoplankton abundance: evidence from parallel
changes in two long-term data series in the Wadden Sea. Journal
of Sea Research 48: 111–125.
Cadée, G.C. 1986a. Recurrent and changing seasonal patterns in
phytoplankton of the westernmost inlet of the Dutch Wadden Sea
from 1969 to 1985. Marine Biology 93: 281–289.
Cadée, G.C. 1986b. Increased phytoplankton primary production in
the Marsdiep area (western Dutch Wadden Sea). Netherlands
Journal of Sea Research 20: 285–290.
Cadée, G.C., and J. Hegeman. 2002. Phytoplankton in the Marsdiep at
the end of the 20th century; 30 years monitoring biomass,
primary production, and Phaeocystis blooms. Journal of Sea
Research 48: 97–110.
Cloern, J. 1996. Phytoplankton bloom dynamics in coastal ecosystems: a
review with some general lessons from sustained investigation of
San Francisco Bay, California. Reviews of Geophysics 34: 127–168.
Colebrook, J.M., and G.A. Robinson. 1965. Continuous plankton
records: seasonal cycles of phytoplankton and copepods in the
north-eastern Atlantic and the North Sea. Bulletin of Marine
Ecology 6: 123–139.
Dakos, V., E. Benincà, E.H. van Nes, C.J.M. Philippart, M. Scheffer,
and J. Huisman. 2009. Interannual variability in species
composition explained as seasonally entrained chaos. Proceedings of the Royal Society B 276: 2871–2880.
Edwards, M., and A.J. Richardson. 2004. Impact of climate change on
marine pelagic phenology and trophic mismatch. Nature 430:
881–884.
Gieskes, W.W.C., S.C. Leterme, H. Peletier, M. Edwards, and P.C.
Reid. 2007. Phaeocystis colony distribution in the North Atlantic
Ocean since 1984, and interpretation of long-term changes in the
Phaeocystis hotspot in the North Sea. Biogeochemistry 83: 49–
60. doi:10.1007/s10533-007-9082-6.
Grift, R.E., I. Tulp, L. Clarke, U. Damm, A. McLay, S. Reeves, J.
Vigneau, and W. Weber. 2004. Assessment of the ecological
effects of the Plaice Box. Report of the European Commission
Expert Working Group to evaluate the Shetland and Plaice boxes.
Brussels. 121 pp.
Iriogen, X., K.J. Flynn, and R.P. Harris. 2005. Phytoplankton blooms:
a ‘loophole’ in microzooplankton grazing impact? Journal of
Plankton Research 27: 313–321.
Loebl, M., F. Colijn, and J.E.E. van Beusekom. 2008. Increasing
nitrogen limitation during summer in the List Tidal Basin
(Northern Wadden Sea). Helgoland Marine Research 65: 59–65.
Lorenzen, C.J. 1967. Determination of chlorophyll and phaeopigments. Spectrophotometric equations. Limnology and Oceanography 12: 343–346.
Estuaries and Coasts (2010) 33:286–294
Pastoors, M., L. Bolle, K. Groeneveld, P. Groot, P. van Leeuwen, G. Piet,
A. Rijnsdorp, and G. Rink. 2000. Evaluatie van de scholbox. RIVO
rapport C002/00. Rijksinstituut voor Visserijonderzoek, IJmuiden.
Philippart, C.J.M., G.C. Cadée, W. van Raaphorst, and R. Riegman.
2000. Long-term phytoplankton-nutrient interactions in a shallow
coastal sea: algal community structure, nutrient budgets, and
denitrification potential. Limnology and Oceanography 45: 131–
144.
Philippart, C.J.M., H.M. van Aken, J.J. Beukema, O.G. Bos, G.C.
Cadée, and R. Dekker. 2003. Climate-related changes in
recruitment of the bivalve Macoma balthica. Limnology and
Oceanography 48: 2171–2185.
Philippart, C.J.M., J.J. Beukema, G.C. Cadée, R. Dekker, P.W.
Goedhart, J.M. van Iperen, M.F. Leopold, and P.M.J. Herman.
2007. Impacts of nutrient reduction on coastal communities.
Ecosystems 10: 96–119.
Pinheiro, J.C., and D.M. Bates. 2000. Mixed-effects models in S and SPLUS. New York: Springer. 528 pp.
Prins, T.C., N. Dankers, and A.C. Smaal. 1994. Seasonal variation in
the filtration rates of a semi-natural mussel bed in relation to
seston composition. Journal of Experimental Marine Biology and
Ecology 176: 69–86.
R Development Core Team. 2008. R: A language and environment for
statistical computing. R Foundation for Statistical Computing,
Vienna, Austria. ISBN 3-900051-07-0, URL http://www.Rproject.org.
Riley, G.A. 1967. The plankton of estuaries. In Estuaries, ed. G.
Lauff, 316–326. Washington: American Association for the
Advancement of Science.
Rolinski, S., H. Horn, T. Petzholdt, and L. Paul. 2007. Identifying
cardinal dates in phytoplankton time series to enable analysis of
long-term trends. Oecologia 153: 997–1008.
Schabenberger, O., and F.J. Pierce. 2002. Contemporary statistical
models for the plant and soil sciences. Boca Raton: CRC.
Smayda, T.J. 1998. Patterns of variability characterizing marine
phytoplankton, with examples from Narragansett Bay. ICES
Journal of Marine Science 55: 562–573.
Stich, H.B., and A. Brinker. 2005. Less is better: uncorrected versus
phaeopigment-corrected photometric chlorophyll-a estimation.
Archives of Hydrobiology 162: 111–120.
Strathmann, R.R. 1996. Are planktonic larvae of marine benthic
invertebrates too scarce to compete within species? Oceanologia
Acta 19: 399–407.
Thorson, G. 1950. Reproductive and larval ecology of marine bottom
invertebrates. Biological Reviews 25: 1–45.
Todd, C.D., and R.W. Doyle. 1981. Reproductive strategies of marine
benthic invertebrates: a settlement-timing hypothesis. Marine
Ecology Progress Series 4: 75–83.
Townsend, D.W., L. Cammen, P.M. Holligan, D.E. Campbell, and N.
R. Pettigrew. 1994. Causes and consequences of variability in the
timing of spring phytoplankton blooms. Deep-Sea Research I 41:
747–765.
Van Beusekom, J.E.E., M. Loebl, and P. Martens. 2009. Distant
riverine nutrient supply and local temperature drive the long-term
development in a temperate coastal basin. Journal of Sea
Research 61: 26–33.
Wiltshire, K.H., A.M. Malzahn, K. Wirtz, W. Greve, S. Janisch, P.
Mangelsdorf, B.F.J. Manly, and M. Boersma. 2008. Resilience of
North Sea phytoplankton spring bloom dynamics: An analysis of
long-term data at Helgoland Roads. Limnology and Oceanography 53: 1294–1302.
Wood, S.N. 2006. Generalized additive models: an introduction with
R. New York: Chapman and Hall.
Zuur, A.F., E.N. Ieno, N.J. Walker, A.A. Saveliev, and G.M. Smith.
2009. Mixed effects models and extensions in ecology with R.
New York: Springer.