Abstract An indoor mesocosm system was set up to
study the response of phytoplankton and zooplankton
spring succession to winter and spring warming of sea
surface temperatures. The experimental temperature
regimes consisted of the decadal average of the Kiel
Bight, Baltic Sea, and three elevated regimes with 2C,
4C, and 6C temperature difference from that at base-
line. While the peak of the phytoplankton spring bloom
was accelerated only weakly by increasing temperatures
(1.4 days per degree Celsius), the subsequent biomass
minimum of phytoplankton was accelerated more
strongly (4.25 days per degree Celsius). Phytoplankton
size structure showed a pronounced response to warm-
ing, with large phytoplankton being more dominant in
the cooler mesocosms. The first seasonal ciliate peak
was accelerated by 2.1 days per degree Celsius and the
second one by 2.0 days per degree Celsius. The over-
wintering copepod populations declined faster in the
warmer mesocosm, and the appearance of nauplii was
strongly accelerated by temperature (9.2 days per
degree Celsius). The strong difference between the
acceleration of the phytoplankton peak and the accel-
eration of the nauplii could be one of the ‘‘Achilles
heels’’ of pelagic systems subject to climate change,
because nauplii are the most starvation-sensitive life
cycle stage of copepods and the most important food
item of first-feeding fish larvae.
Keywords Plankton Æ Climate change Æ Seasonal
succession Æ Spring bloom
Introduction
Motivation
Climate change is already affecting a wide variety of
ecosystems (Walther et al. 2002), including aquatic
ones (Edwards et al. 2002; Edwards and Richardson
2004; Fromentin and Planque 1996; Straile 2000; Straile
and Adrian 2000), and will increasingly continue to
do so if the prevailing predictions of further green-
house warming are fulfilled. Already, now, the Baltic
Sea is characterised by a strong geographic, seasonal,
and interannual variability of all relevant hydrographic
variables, which are closely connected to the atmo-
spheric forcing in the region (Mattha
¨
us and Schinke
1994; Lehmann et al. 2002). Present day interannual
differences in surface temperature during winter may
amount to 5C. Seasonal temperature stratification
does not start before April, while reduced salinity
stratification persists throughout the winter. Based on
the current knowledge [summarised by the Interna-
tional Panel on Climate Change (IPCC) 2001] mainly
a pronounced winter warming is expected for north–
central Europe. For a doubling of CO
2
emissions
during the twenty-first century (‘‘business as usual
Communicated by Roland Brandl.
Priority programme of the German Research
Foundation— contribution 3.
U. Sommer (&) Æ N. Aberle Æ T. Hansen Æ K. Lengfellner Æ
M. Sandow Æ J. Wohlers Æ E. Zo
¨
llner Æ U. Riebesell
Leibniz Institute for Marine Sciences, Kiel University,
Du
¨
sternbrooker Weg 20, 24105 Kiel, Germany
e-mail: usommer@ifm-geomar.de
A. Engel
Alfred Wegener Institute, Bremerhaven, Germany
Oecologia
DOI 10.1007/s00442-006-0539-4
123
GLOBAL CHANGE AND CONVERSATION ECOLOGY
An indoor mesocosm system to study the effect of climate
change on the late winter and spring succession of Baltic Sea
phyto- and zooplankton
Ulrich Sommer Æ Nicole Aberle Æ Anja Engel Æ Thomas Hansen Æ
Kathrin Lengfellner Æ Marcel Sandow Æ Julia Wohlers Æ
Eckart Zo
¨
llner Æ Ulf Riebesell
Received: 7 February 2006 / Accepted: 9 August 2006
Springer-Verlag 2006
scenario’’, 192a, HadCM3) an increase of annual mean
surface temperatures by 3–5C can be predicted, while
winter temperature could increase by even 5–10C
(prediction for 2070–2100, compared with the refer-
ence period 1960–1990).
It is the ultimate research question of this study: how
will the spring succession of plankton change in re-
sponse to the forecasted climate changes at the
beginning of the growth season? Plankton from the
Kiel Bight (western Baltic Sea) is intended to serve as a
model system for moderately deep water bodies, where
the spring bloom of phytoplankton can start before the
onset of thermal stratification. This is a pronounced
contrast to deep water bodies (e.g. Lake Constance;
Scheffer et al. 2001), where there is strong coupling
between the light and the temperature, because phy-
toplankton receive too little light for the onset of
spring growth before temperature stratification begins
(‘‘critical mixing depth concept’’ sensu Sverdrup 1953).
It is hypothesised that increased temperatures will
accelerate heterotrophic processes more strongly than
light-limited phytoplankton growth. This should lead
to more profound community level changes than to a
simple seasonal advancement of events. Different
temperature sensitivities of seasonal growth patterns
and activity patterns in food webs could lead to a loss
in synchrony between prey supply and predator de-
mand with far reaching ecosystem consequences (cf.
the ‘‘match–mismatch’’ hypothesis; Cushing 1975).
Traditional field mesocosms would have been a
logical choice for the scale of experimentation needed,
but temperature control of such systems was beyond
our capacity. Therefore, we developed a new type of
indoor mesocosm, which combined a plankton con-
tainer and a benthos container serving as a source for
meroplanktonic larvae and benthic resting stages of
plankton organisms. The proximate goal of this study
was a feasibility test of our experimental systems,
concentrating on three questions:
• Are we able to reproduce the natural pattern of
plankton spring succession in our mesocosms?
• How long can the mesocosms be operated before
containment artefacts become too strong?
• Given the usual variability between replicate mes-
ocosms, will it be possible to obtain statistically
significant temperature effects?
Plankton seasonal succession
According to the predominant paradigm, seasonal
succession of plankton is initiated by the spring blooms
of autotrophic phytoplankton. In temperate and boreal
waters this spring bloom is almost a start from zero,
because only few phytoplankton have survived winter.
The spring bloom is initiated by the improvement of
light supply (for reviews cf. Greve and Reiners 1995;
Sommer et al. 1986; Sommer 1996). The direct, physi-
ological consequences of temperature play no promi-
nent role in the initiation of the phytoplankton spring
bloom, because of the well-known temperature inde-
pendence of light-limited photosynthetic rates at tem-
peratures >2C (Tilzer et al. 1986).
Zooplankton spring growth follows after the phyto-
plankton spring bloom, the usual sequence being first a
bloom of fast-growing protozoans followed by slower
growing metazoans. This sequence can be reversed, if
there are strong over-wintering mesozooplankton
populations (often the case in copepods). After a few
weeks of zooplankton increase, grazing rates exceed
phytoplankton production and lead to a decline in
phytoplankton biomass and a subsequent biomass
minimum in late spring/early summer (called ‘‘clear-
water phase’’ in the limnological literature). While the
causation of the clear-water phase by grazing has been
well accepted in limnology for two decades, it is still
controversial in biological oceanography (as an exam-
ple for a grazing-induced clear-water phase see Bautista
et al. 1992). Obviously, there are cases where nutrient
limitation and subsequent aggregation of phytoplank-
ton lead to major sinking losses prior to the onset of
heavy grazing (Smayda 1971; Smetacek et al. 1984).
The spring development of phytoplankton and
zooplankton depends differently on physical conditions
during early spring: phytoplankton growth depends on
light (and stratification in deep waters), while zoo-
plankton growth depends on food availability and
temperature. With identical food supply, zooplankton
population growth will become faster, the warmer
spring temperatures are. This temperature dependence
can be accentuated if the spring population depends on
the germination of resting stages. Madhudatrap et al.
(
1996) triggered the germination of six Baltic Sea
zooplankton species (four copepods, two cladocerans)
by temperature increase in the laboratory. The tem-
perature dependence of zooplankton spring growth
must have consequences for the timing and extent of
the clear-water phase, as predicted by a recent model
for Lake Constance (Scheffer et al. 2001) and by an
analysis of field data using the present day climate
variability (Straile 2000).
However, those results cannot be simply extrapo-
lated to marine food webs with a more complex me-
sozooplankton structure, particularly because copepods
are not as herbivorous as previously assumed (e.g.
Oecologia
123
White 1979). They are omnivores, which feed on pro-
tozoans and large phytoplankton, while phytoplankton
<10 lm are not taken if there is enough large food
(Katechakis et al. 2002; Kleppel 1993; Sell et al. 2001;
Sommer and Stibor 2002). This could lead to the fol-
lowing consequences of warming to the spring succes-
sion in the Baltic Sea: an earlier onset of calanoid
copepod grazing would reduce the biomass of large
phytoplankton while releasing small phytoplankton
from grazing pressure by protozoans. Thus, phyto-
plankton size and species composition would change
without change of the overall seasonal biomass pattern.
Methods
Mesocosms
Eight mesocosms were set up in temperature-con-
trolled culture rooms of the IfM-GEOMAR. The
experimental period lasted from 4 February to 4 May
in order to encompass the winter–spring transition.
The mesocosms consisted of a two-chamber system
(Fig. 1), with a 1,400 l plankton chamber and a smaller
(300 l) benthos chamber, which served as a source of
meroplanktonic larvae of zoobenthos and of plank-
tonic organisms germinating from benthic resting
stages. The plankton was gently stirred by a propeller.
The benthos chamber was filled with sediment from
the Kiel Fjord and 20 adult blue mussels Mytilus edulis.
There was a continuous, but small, exchange (on
average ca. 60 l, with some variability from 30 l to 90 l)
of the water between both chambers. This was suffi-
cient for feeding the mussels with phytoplankton but
was an insignificant loss for phytoplankton, even if all
the phytoplankton would have been consumed by the
mussels (<5% loss per day). Temperature and light
regimes in both chambers were identical. The experi-
ment was run as an almost closed system. Only the
sample volume was replaced by unfiltered water from
the Kiel Fjord.
Temperature regime
There were four temperature regimes (each dupli-
cated), defined by the initial temperature difference
from the decadal mean 1993–2002 in Kiel Bight, called
0, +2, +4, +6 treatments. The initial temperature dif-
ferences between treatments of 2C were maintained
until the end of February and were reduced by 0.25C
per month thereafter, in order to mimic the less pro-
nounced warming later in the year. Actual tempera-
tures measured in the mesocosms deviated only slightly
from planned ones, particularly in one of the +6C
treatments (Fig. 2).
Light regime
Light was supplied by computer-controlled aquarist
light units (GHL Groß Hard- und Softwarelo
¨
sungen,
Lampunit HL3700 and ProfiluxII). Each light unit
contained six fluorescent tubes [T5, types 5· JBL Solar
Tropic (4,000 K), 1· JBL Solar Natur (9,000 K)]. This
setup allowed the simulation of daily triangular light
Fig. 1 Scheme of mesocosms
Oecologia
123
curves. Timing of sunrise and sundown and the maxi-
mum light intensity was daily supplied by a specialised
database computer program (GHL, Prometeus). Sea-
son-dependent database values were derived from a
model that is based on astronomic formulae (Brock
1981). The astronomic peak-shaped light curve was
transformed into a triangular light curve by calculating
sunrise and sundown to preserve daily integrated light
intensities. Light attenuation by the cloud cover was
superimposed by a randomised cloud cover generator
(ProfiluxII), assuming an average 80% cloud cover. We
made a further reduction in light intensity , in order to
account for water column light attenuation, by calcu-
lating the mean light intensity (I
mix
) of a 12 m mixed
water column (z; here depth of the halocline) and an
average attenuation coefficient (k) of 0.5 m
–1
accord-
ing to the equation of Riley (1957):
I
mix
¼ I
0
ð1 e
kz
ÞðkzÞ
1
:
Stocking with organisms, and water exchange
Initially, the mesocosms were filled with unfiltered
water from Kiel Bight containing the over-wintering
populations of phytoplankton, bacteria, and protozoa.
Mesozooplankton was added from net catches at nat-
ural densities (ca. 20 ind l
–1
) which conforms to usual
February values (Behrends 1996). It consisted mainly
of the cyclopoid copepod Oithona similis and the cal-
anoid copepod Pseudocalanus/Paracalanus spp.
Copepod survival experiment
From 3 June to 13 July we performed a ‘‘copepod
survival experiment’’ in one of the warmest and one of
the coldest mesocosms in order to test whether the
copepod mortality observed during the main experi-
ment could be explained by containment artefacts,
stirring or otherwise unfavourable mechanical condi-
tions. In order to enhance food supply, the experiment
was run under high light conditions, i.e. 100% surface
light with the day length of the season.
Samples
Samples for nutrient chemistry, phytoplankton, and
protozoa were taken three times per week (Monday,
Wednesday, Friday), while samples for mesozoo-
plankton were taken once per week. Phytoplankton
and protozoan samples for microscopic counts were
fixed with Lugol’s iodine, while samples for flow
cytometric analysis were processed immediately.
Mesozooplankton samples were taken with a bucket
(three times 5 l, once per week), filtered onto a 64 lm
sieve and fixed with industrial methylated spirit.
Phytoplankton >5 lm and protozoa were counted
by the inverted microscope method (Utermo
¨
hl 1958)
and distinguished at the genus level in most cases.
We aimed at counting 100 individuals per taxonomic
unit, which gives 95% confidence limits of ±20%, but
this standard could not be attained with rare species.
Small phytoplankton were counted by a flow cytom-
eter (FACScalibur, Becton Dickinson) and distin-
guished by size and fluorescence of the pigments
chlorophyll a and phycoerythrin. Three flow cytometer
categories were matched to taxa identified micro-
scopically (the small flagellates Chrysochromulina,
Plagioselmis, Teleaulax). Flow cytometry counts were
consistently higher than were microscopic ones, indi-
cating incomplete sedimentation in the Utermo
¨
hl
counting chambers. Phytoplankton cell volumes were
calculated from linear measurements after approxi-
mation to the nearest geometric standard solid
(Hillebrand et al. 1999) and converted into carbon
content according to Menden-Deuer and Lessard
(2000). Phytoplankton were grouped in four functional
categories: autotrophic picoplankton (<3 lm), nanofla-
gellates (3–20 lm), nanodiatoms (3–20 lm), and mic-
rodiatoms (>20 lm). Flagellates >20 lmwerefound
occasionally but never exceeded 1% of total biomass.
For ciliate counts the samples were transferred to
100 ml sedimentation chambers, and, for each sample,
we counted the whole area of the bottom plate in
order to guarantee precise data. For bio-volume cal-
culations geometric proxies were used according to
Hillebrand et al. (1999), and ciliate carbon biomass
was calculated using the conversion factors given in
Putt and Stoecker (1989).
20 40 60 80 100 120 140 160
Julian da
y
0
2
4
6
8
10
12
14
16
temperature (˚C)
Fig. 2 Temperature regime in the mesocosms
Oecologia
123
Mesozooplankton samples were counted with a
binocular microscope (Leica MS5). Copepod adults
and copepodites were distinguished by genus. Copepod
nauplii were not distinguished taxonomically. The rest
of the mesozooplankton was separated into larval types
(e.g. polychaete larvae, cirripedia larvae, etc.). As the
experiment was planned to run for several months we
had to keep sample sizes small in order to diminish the
mesozooplankton populations as little as possible.
Therefore, our samples (three times 5 l out of each
mesocosm per week) did not contain enough individ-
uals to have similarly high counting standards as for
phytoplankton.
Water samples for the determination of inorganic
nutrient concentrations were taken after filtration
through 0.65 lm cellulose acetate filters. The mea-
surements of nitrate, nitrite and phosphate were car-
ried out following the standard protocols by Hansen
and Koroleff (1999). Ammonium concentrations were
determined from unfiltered samples according to the
protocol described by Holmes et al. (1999). All
analyses were performed on the day of sampling.
Particulate organic carbon (POC) and particulate
organic nitrogen (PON), as well as particulate organic
phosphorus (POP), were determined from 500 ml
samples filtered onto pre-combusted Whatman GF/F
filters. After filtration the samples were immediately
frozen and stored at –20C. Analysis of POC and
PON were carried out by a gas chromatograph after
Sharp (1974) on a EuroVector elemental analyser,
whereas the measurement of POP was conducted
colorimetrically after oxidation with potassium per-
oxodisulphate, as described by Hansen and Koroleff
(1999). Samples for particulate nutrients were not
measured before the beginning of the phytoplankton
spring bloom.
Results
Phytoplankton
Initially, phytoplankton biomass declined until Julian
days 54–63. The decrease was steeper in the warmer
treatments (Fig. 3). Thereafter, phytoplankton biomass
increased to form a spring bloom. The timing of the
0
0.5
1
1.5
2
2.5
3
log
10
(µg l
-1
)
30 50 70 90 110 130
91 91
124 126
+ 0˚C
0
0.5
1
1.5
2
2.5
3
log
10
(µg l
-1
)
30 50 70 90 110 130
Julian day
Julian day
84 91
103 110
+ 2˚C
a
b
0
0.5
1
1.5
2
2.5
3
30 50 70 90 110 130
Julian day
Julian day
84 84
96 101
+ 4˚C
0
0.5
1
1.5
2
2.5
3
30 50 70 90 110 130
77 89
96 101
+ 6˚C
log
10
(µg l
-1
) log
10
(µg l
-1
)
c
d
Fig. 3 Phytoplankton biomass (lgCl
–1
) in the mesocosms. Numbers indicate day of spring bloom (biomass maximum) and clear-
water phase (biomass minimum); treatments: a +0C, b +2C, c +4C, d +6C
Oecologia
123
