Phytoplankton community reorganization driven by eutrophication and warming in Lake Biwa

Abstract

We compiled and analyzed long-term data, including chemical, physical and phytoplankton community data, for the Lake Biwa
ecosystem from 1962 to 2003. Analyses on environmental data indicate that Lake Biwa had experienced intensified eutrophication
(according to total phosphorus concentration) in the late 1960s and returned to a less eutrophic status around 1985, and then
exhibited rapid warming and thus increased water column stability since 1990. Total phytoplankton cell volume largely followed
the trend of total phosphorus concentration, albeit short-term fluctuations existed. However, phytoplankton community shifted
dramatically in response to those changes of environmental states. These shifts were cause by changes in trophic status driven
by phosphorus loadings and physical properties in the water column driven by warming. Moreover, most phytoplankton species
did not show a strong linear correlation with environmental variables, suggesting nonlinear transitions among different states.

KeywordsEcological characteristics-Multiple stable states-Nutrient loading-Total phytoplankton volume-Water column stratification

Full text

Page 1

RESEARCH ARTICLE
Phytoplankton community reorganization driven
by eutrophication and warming in Lake Biwa
Chih Hao Hsieh•Kanako Ishikawa•Yoichiro Sakai•Toshiyuki Ishikawa•
Satoshi Ichise•Yoshimasa Yamamoto•Ting Chun Kuo•Ho Dong Park•
Norio Yamamura•Michio Kumagai
Received: 13 November 2009/Accepted: 26 May 2010/Published online: 24 June 2010
? Springer Basel AG 2010
Abstract
including chemical, physical and phytoplankton commu-
nity data, for the Lake Biwa ecosystem from 1962 to 2003.
Analyses on environmental data indicate that Lake Biwa
had experienced intensified eutrophication (according to
total phosphorus concentration) in the late 1960s and
returned to a less eutrophic status around 1985, and then
exhibited rapid warming and thus increased water column
stability since 1990. Total phytoplankton cell volume
We compiled and analyzed long-term data,
largely followed the trend of total phosphorus concentra-
tion, albeit short-term fluctuations existed. However,
phytoplankton community shifted dramatically in response
to those changes of environmental states. These shifts were
cause by changes in trophic status driven by phosphorus
loadings and physical properties in the water column dri-
ven by warming. Moreover, most phytoplankton species
did not show a strong linear correlation with environmental
variables, suggesting nonlinear transitions among different
states.
Keywords
Multiple stable states ? Nutrient loading ?
Total phytoplankton volume ? Water column stratification
Ecological characteristics ?
Introduction
Anthropogenic impacts and climate effects on ecosystems
are pressing concerns. For lake systems, one of the most
serious anthropogenic impacts is eutrophication, and this
issue has been studied for a 100 years and remains a
serious concern today (Schindler 2006; Smith et al. 2006).
Eutrophication of lakes caused by domestic sewage was
well known in history and often was associated with an
increase in phytoplankton
Edmondson 1956; Davis 1964). In some lakes, phyto-
plankton abundance was found to track closely the
eutrophic condition of the lakes (often measured by the
total phosphorus, TP, in the water column). When TP
increased, phytoplankton abundance increased, and to
some extreme state, a phytoplankton bloom occurred; when
TP decreased, phytoplankton reversed to its original
abundance (Edmondson 1970; Ruggiu et al. 1998; Kohler
and Hoeg 2000). However, in many other lakes, when TP
abundance (Hasler1947;
Electronic supplementary material
article (doi:10.1007/s00027-010-0149-4) contains supplementary
material, which is available to authorized users.
The online version of this
C. H. Hsieh (&) ? T. C. Kuo
Institute of Oceanography
and Institute of Ecology and Evolutionary Biology,
National Taiwan University, No. 1, Sec. 4, Roosevelt Rd,
Taipei 10617, Taiwan
e-mail: chsieh@ntu.edu.tw
K. Ishikawa ? T. Ishikawa ? S. Ichise ? M. Kumagai
Lake Biwa Environmental Research Institute,
5-34 Yanagasaki, Otsu, Shiga 520-0022, Japan
Y. Sakai
Center for Ecological Research, Kyoto University,
Hirano 2-509-3, Otsu, Shiga 520-2113, Japan
Y. Yamamoto
Research Center for Environmental Changes, Academia Sinica,
No. 128, Sec. 2, Academia Rd, Taipei 11529, Taiwan
H. D. Park
Department of Environmental Science, Shinshu University,
Asahi 3-1-1, Matsumoto, Nagano 390-8621, Japan
N. Yamamura
Research Institute for Humanity and Nature, 457-4 Motoyama,
Kamigamo, Kita-ku, Kyoto 603-8047, Japan
Aquat. Sci. (2010) 72:467–483
DOI 10.1007/s00027-010-0149-4
Aquatic Sciences

Page 2

decreased, phytoplankton abundance remained high or
delayed its decline (Anneville and Pelletier 2000; Horn
2003; Dokulil and Teubner 2005). Recent studies suggest
that internal loading (release of phosphorus from the sed-
iments mediated by microbial activities under an anoxic
condition) is the main reason for the irreversible phenom-
enon or delayed response (Carpenter et al. 1998; Carpenter
2005). Theoretically, such phenomena are associated
with nonlinear transition among alternative stable states
(Scheffer et al. 2001; Hsieh et al. 2005).
Besides the bulk phytoplankton abundance, how phy-
toplankton communities respond to eutrophication and
later oligotrophication is also a subject of intensive
research. Some studies indicated that phytoplankton com-
munity changes responding to the trophic condition of lake
was symmetrical; that is, phytoplankton community would
return to its original structure when the trophic status of
lake recovered (Sommer et al. 1993; Kohler and Hoeg
2000). Other studies indicated that the changes in the
phytoplankton community were asymmetrical (Anneville
et al. 2002b) or the reversal was delayed (Dokulil and
Teubner 2005). Through trophic interactions, eutrophica-
tion effects also propagated to zooplankton (Molinero et al.
2006; Anneville et al. 2007).
In addition to eutrophication, warming in the past half
century has also drastically influenced lake ecosystems.
Increased water temperature resulted in intensified strat-
ification of lakes that caused strong hypolimnetic oxygen
depletion (Jankowski et al. 2006). Warming also changed
the phenology of lake processes, such as earlier onset of
stratification and consequent earlier spring phytoplankton
blooms (Winder and Schindler 2004). Furthermore,
warming altered phytoplankton communities, shifted the
community structure to small-sized species (Winder and
Hunter 2008), and favored the growth of cyanobacteria
(Elliott et al. 2006). Moreover, trophic status and
warming often had synergistic effects on the phyto-
plankton community as well as the whole ecosystem
(Elliott et al. 2006; Huber et al. 2008; Wilhelm and
Adrian 2008).
In this research, we investigated eutrophication and
warming effects on the phytoplankton community of Lake
Biwa. Lake Biwa is the largest lake in Japan; it contains a
high biodiversity, including more than 60 endemic species.
It also provides high economic values, including trans-
portation, drinking water, and fisheries (Kumagai 2008).
Along with urbanization, Lake Biwa had experienced
increasingly nutrient loading since 1960s and subsequently
suffered from blooms of Uroglena americana since 1977
and cyanobacteria since 1983 (Kumagai 2008). The ini-
tialization of eutrophication in the 1960s was reflected in
the sediment records, showing pronounced changes in
zooplankton composition (Tsugeki et al. 2003). The
progress of eutrophication from the 1960s to early 1980s
were revealed from stable nitrogen isotope analysis of
sediment core samples and preserved gobiid fish, Isaza,
Chaenogobius isaza (Ogawa et al. 2001). A water treat-
ment regulation was enforced in 1982, and nutrient loading
was progressively reduced and then stabilized after 1985
(Kumagai 2008). However since 1980, the air temperature
rose quickly, putting another threat on the Lake Biwa
ecosystem.
Due to its important ecological and economic value,
Lake Biwa has been intensively monitored by the envi-
ronmentalagency,fisheries
institutions. Despite of great amount of effort invested and
various data (including physical, chemical, and biological)
being collected, no attempt was ever made to compile those
data from different agencies. Particularly, detailed phyto-
plankton community data have been collected; however,
only total biovolume was reported in the Japanese literature
(Ichise et al. 2001; Ichise et al. 2007). Detailed ecosystem-
level time-series data from Lake Biwa are available, pro-
viding us a great opportunity to investigate eutrophication
and climate effects on a large temperate lake. In this
research, our purposes were to (1) compile and analyze
historical (1962–2003) chemical, physical, and phyto-
plankton data collected by various institutes, and (2)
investigate how eutrophication and climate warming
affected the phytoplankton community in Lake Biwa.
Although seasonal variation of lake environments and
seasonal succession of phytoplankton communities were
well known (Sommer et al. 1986; Anneville et al. 2002b;
Lau and Lane 2002), here we focus on long-term variation
and thus investigate only the annual average data in the
current study.
agency,andacademic
Materials and methods
Environmental data
We compiled long-term time-series data from four research
institutions. Chemical (concentration of phosphate, nitrate,
and ammonium) and physical (water temperature) data
from 1961 to 2005 (in Stations 1–5 in Fig. 1) were col-
lected by the Shiga Prefecture Fisheries Experimental
Station (SPFES). In addition, silicate data were collected
after 1978. Only data from the deepest station (Station 4)
are presented here to investigate the whole water column
properties. Chemical data were collected monthly at depths
of 0.5, 10, 20, 30 m and the bottom (*80 m). For each
month, we calculated the averaged value for the surface
layer (0–20 m); this surface layer approximates the living
habitat of phytoplankton in Lake Biwa. Monthly chemical
data for the surface layer are shown in Appendix 1 of
468 C. H. Hsieh et al.

Page 3

Supplementary material. Annual averaged values are
shown in Fig. 2. We further calculated monthly anomalies
by subtracting the long-term monthly mean and averaged
them into annual anomalies (Appendix 1 of Supplementary
material). In addition to the nutrient concentrations, we
also investigated the phosphate/total phosphorus ratio and
ammonium/nitrate ratio as indicators to nutrient regimes.
Water temperature data from the surface (0.5 m) to the
bottom were measured at 5-m intervals. Following Hsieh
et al. (2009b), we interpolated the temperature data into
1-m intervals using cubic spline. With the interpolated
data, we calculated the thermocline depth defined as the
depth where the greatest temperature gradient occurs and
computed the maximal buoyancy frequency representing
the buoyancy frequency at the thermocline depth, which is
indicative of water column stability (Hsieh et al. 2009b).
Monthly values, annual averages, and annual anomalies of
thermocline depth, maximal buoyancy frequency, and
surface (averaged 0–10 m) and bottom (80 m) temperature
were provided in Fig. 3 and Appendix 1 of Supplementary
material.
We further compiled chlorophyll a (chla) concentration
data from 1975 to 2005 collected by the Lake Biwa
Environmental Research Institute (LBERI, formerly Shiga
Prefecture Institute of Public Health and Environmental
Science) in Station L (Fig. 1). The chla data were treated in
the same manner and presented in Fig. 2 and Appendix 1 of
Supplementary material. In fact, the LBERI have also
collected similar physical and chemical data as the SPFES
at Station L since 1975. Patterns of these LBERI physical
and chemical data are consistent with those of SPFES
(correlation analyses for each of the aforementioned
physical and chemical variables, p\0.001). Because the
SPFES contain longer time series, only SPFES data were
used in further analyses.
We used total phosphorus (TP) in the surface layer
(\20 m) as a proxy to the trophic status of Lake Biwa. TP
data were collected by LBERI (Station L in Fig. 1) only
since 1978. To extend the time series backward, we com-
piled TP data from 1963 to 1980 collected by the Kyoto
University (Station I in Fig. 1). We combined these two
datasets to form a TP time series (Fig. 2, Appendix 1 of
Supplementary material). For the overlapped period (1978–
1980), the TP of these two datasets were at a similar
concentration, and thus we took the average of the series
for that period. We acknowledge that these two stations are
far apart, the two datasets overlap by only 3 years, and it is
difficult to test the consistency of these two datasets.
However, the combined TP series exhibits a history of
changes in trophic condition of Lake Biwa that is consis-
tent with that uncovered by Ogawa et al. (2001) based on
sediment records and isaza fish. Therefore, we consider
that the TP time series we reconstructed here is represen-
tative of the trophic status of Lake Biwa for the past half
century.
Since the chla data were only available after 1975, we
used 1/Secchi depth as a surrogate of phytoplankton bio-
mass. Monthly Secchi depth data from 1966 to 2000 were
collected by the Kyoto University at Station I (Fig. 1). The
inverse Secchi depth and chla data are significantly corre-
lated (correlation analysis, p\0.001). Inverse Secchi
depth data are shown in Fig. 2 and Appendix 1 of Sup-
plementary material.
To link the physical environment of the lake to atmo-
spheric forcing, we also compiled air temperature,
photosynthetic available radiation (PAR), wind velocity,
and precipitation from the Hikone Meteorological Station
(Station W in Fig. 1). Note that the PAR data were from
the Meteorological Station but not from the water column
and may not represent light availability in deeper layers.
The turbulent mixing condition of the upper water column
was estimated by a wind-mixing index, which is propor-
tional to the cube of the wind velocity (Bakun and Parrish
1991; Hsieh et al. 2009a). The wind mixing index was
Fig. 1 Map showing sampling stations in Lake Biwa. Stations 1–5
are the Shiga Prefecture Fisheries Experimental stations; station L is
the long-term monitoring station of the Lake Biwa Environmental
Research Institute; station I is the environmental monitoring station of
the Kyoto University; station W is the Hikone Meteorological Station
Environmental effects on phytoplankton469

Page 4

Fig. 2 Annual time series of chemical and biological variables:
a total phosphorus, b total phytoplankton cell volume for the SPFES
data, c total phytoplankton cell volume for the LBERI data, d total
phytoplankton carbon for the SPFES data, e total phytoplankton
carbon for the LBERI data, f chlorophyll concentration, g inverse
Secchi depth, h phosphate concentration, i nitrate concentration,
j silicate concentration (only available since 1978), k ammonium
concentration. For a, f, and h–k, shown are averaged value for the
0–20 m, while the whole water column average shows a similar
pattern
470 C. H. Hsieh et al.

Page 5

calculated from daily maximum wind velocity (extracted
from hourly measurements) and then averaged into
monthly means. In addition, we investigated the Arctic
Oscillation index (AO), a climate pattern defined by winds
circulating counterclockwise around the Arctic at about 55?
north latitude. The AO has been known to influence
weather condition of Japan (Thompson and Wallace 1998).
Time series of these atmospheric data are shown in Fig. 3
and Appendix 1 of Supplementary material. We investi-
gated also the Pacific Decal Oscillation (PDO) (Mantua
et al. 1997) and Southern Oscillation index (SOI) (Tren-
berth 1984) but found no significant correlation and do not
present these data.
Phytoplankton data
Phytoplankton community data include the time series
from 1978 to 2003 collected by LBERI (Stations L in
Fig. 1) and those from 1962 to 1991 collected by SPFES
(Stations 1–5 in Fig. 1). (In this paper, both eukaryotic
Fig. 3 Annual time series of physical variables: a air temperature,
b lake surface temperature (average 0–10 m), c lake bottom
temperature (80 m), d buoyancy frequency at the thermocline depth,
e wind mixing index, f thermocline depth, g precipitation,
h photosynthetic available radiation, i Arctic Oscillation index,
and j winter averaged air temperature (solid line) and Arctic
Oscillation index (dashed line)
Environmental effects on phytoplankton 471

Page 6

algae and prokaryotic cyanobacteria are referred to as
phytoplankton for simplicity.) Phytoplankton from the
LBERI were collected using a Van Dorn water sampler at
0.5 m depth. We acknowledge that this sampling depth
may not cover the habitats for all species in Lake Biwa;
however, this depth likely includes the representative
samples of phytoplankton cells. Samples were taken fort-
nightly (at the beginning and middle of each month during
day time). For each sampling date, all phytoplankton in a
1-ml water sample were identified and counted using a
microscope by a single expert (S. Ichise). Some colony-
forming cyanobacteria were counted as per colony unit and
the others were counted as per cell unit (detailed in
Appendix 2 of Supplementary material) to facilitate
counting. Average number of cells per colony was also
recorded (Appendix 2 of Supplementary material). Only
species with occurrence of more than 5 years during the
span were analyzed in this study (Appendix 3 of Supple-
mentary material).Thetaxonomy of phytoplankton
changed over time, and we have carefully adjusted the
database to ensure the consistency.
Phytoplankton from SPFES (Stations 1–5 in Fig. 1)
were collected monthly during daytime using a plankton
net of 95-lm mesh size towed at four depth intervals: 0–10,
10–20, 20–40, and 40–75 m. Experts carrying out species
identification changed in different periods. As in the
LBERI dataset, we have carefully adjusted the change of
taxonomy of phytoplankton in the SPFES database. To be
consistent with the LBERI data, only data from the 0–10 m
interval were analyzed here. Indeed, this surface layer
contained higher abundance of phytoplankton compared
with the other depths. The cell density in the surface layer
(0–10 m) is significantly higher than the average density of
other deep layers (with a grand average of all species
ranging 2.5 fold at Station 1 to 172 fold at Station 5). Only
species with occurrence of more than 5 years during the
span were analyzed (Appendix 4 of Supplementary mate-
rial). We averaged the data from the five stations to form a
single value. Data from the deepest station (Station 4, with
a water depth closest to Station L in LBERI) exhibited a
similar pattern with the averaged values for most species
we analyzed here (correlation analysis, p\0.00001) but
showed a higher variability. Only 4 species, Eudorina
elegans, Aphanothece nidulans, Fragilaria sp., and Syne-
dra sp. did not show a high correlation (p[0.05), and we
noticed that these species occurred in \10 years in the
database. We chose to show the results based on the
average data because these data exhibited less variance.
The results based on analyzing only Station 4 are qualita-
tively the same as the results based on the average data.
Missing data were imputed using the best fitting state-space
model based on the Kalman filter (Durbin and Koopman
2001; Hsieh et al. 2009a). We noticed that much fewer
phytoplankton taxa were identified in the SPFES data; this
is due to the large mesh size of the sampling gear that
misses most small phytoplankton species. This problem
makes it difficult to integrate the LBERI and SPFES
datasets, and thus these two datasets were analyzed
independently.
To compute total biomass of phytoplankton, we first
converted species abundance to biovolume and then to
carbon biomass. Biovolume of phytoplankton species
(Appendix 2 of Supplementary material) was measured by
LBERI and reported in Ichise et al. (2007). Biovolume of
colony-forming species was measured as the colony vol-
ume. Single cell volume was estimated as the colony
volume divided by averaged cell number per colony (Ichise
et al. 2007). Taxa-specific empirical equations (Rocha and
Duncan 1985; Verity et al. 1992; Menden-Deuer and
Lessard 2000) were then used to convert biovolume to
carbon biomass (Appendix 5 of Supplementary material).
We summed these values to construct the time series of
total biovolume and total carbon biomass for the SPFES
and LBERI datasets, respectively (Fig. 2).
Data analyses
Two main environmental issues are associated with the
Lake Biwa ecosystem: eutrophication and warming. To
understand these effects on the Lake Biwa ecosystem, first,
correlation analyses were used to investigate long-term
relationships among physical, chemical, and bulk biologi-
cal properties (shown in Figs. 2, 3) at an interannual scale.
Stationary bootstrap approach (Politis and Romano 1994)
with accelerated bias correction was used to compute 95%
confidence limits and to perform a hypothesis test in order
to account for serial dependence in the time-series data
(Hsieh et al. 2009a). For correlation analyses, only data in
the overlapped time period between pair-wise series were
considered. Consequently, sample sizes were not equal for
different comparisons, and thus paired series with a short
overlap tended to show no statistically significant correla-
tion when there was no sufficient variation within the short
overlap period.
Second, warming in water temperature in Lake Biwa
around 1990 (Fig. 3a–c) has caught particular attention,
because this warming event has resulted in significant
reduction of dissolved oxygen in bottom waters, which
might have negative effects on benthic organisms (Kumagai
2008). We tested whether significant changes in magnitude
and variance of physical variables had happened since
1990, using a permutation test. Specifically, we calculated
the difference in mean and variance for the two periods
(before and after 1990) for each variable, and then gener-
ated a null distribution by randomly shuffling the data to
perform hypothesis tests (Hsieh et al. 2009b). Prior to
472C. H. Hsieh et al.

Page 7

calculating the variance, we removed a long-term trend in
time-series data to avoid inflation in variance caused by the
trend (Hsieh et al. 2006, 2009a). To do this, we extracted
the trend by using the locally weighted scatter plot
smoother (LOWESS) and calculated the magnitudes of
residuals deviated from the trend (Hsieh et al. 2006). We
used a window of 20 years to estimate the LOWESS, but
sensitivity analysis indicated that our results were not
sensitive to the choice of window size, given that the
window is sufficiently large ([15 years).
Third, long-term variation of phytoplankton communi-
ties were examined using principal component analysis
(PCA) for the SPEFS and LBERI data separately. Prior to
analysis, abundance data were transformed as log2(X ? 1)
to account for the aggregation effects of phytoplankton. We
linked the temporal pattern of phytoplankton community to
environmental variables using redundancy analysis (RDA)
(Legendre and Legendre 1998). Stepwise procedure was
used to select the significant variables (using a = 0.05) to
exclude irrelevant variables, and such a procedure will
select only the variable that explains the highest variance
when multiple variables in the model are linearly depen-
dent (Peres-Neto et al. 2006). In addition, correlation
analyses were carried out to relate environmental variables
to individual species.
Fourth, we categorized phytoplankton species into
groups according to their time-series pattern (based on
scores extracted from the PCA), using K-means clustering
to obtain the optimal classification (Legendre and Legendre
1998). The optimal number of groups was determined by
the Davies-Bouldin index (Davies and Bouldin 1979). The
categorized phytoplankton species were plotted on the
RDA biplot to exhibit the loadings of phytoplankton spe-
cies on the community pattern and to link phytoplankton
species to environmental variables (Legendre and Legen-
dre 1998).
Results
Long-term environmental variation
The time series of total phosphorus (TP) in Lake Biwa
exhibited a history of eutrophication and a later return to a
less eutrophic state in the past half century. TP increased
quickly after 1967, reached a maximum in 1974 and then
declined until 1985, and fluctuated around a stable value
thereafter (Fig. 2a). This variation of trophic status also
was reflected in bulk phytoplankton biovolume (Fig. 2b)
and biomass (Fig. 2d) for the SPFES data (1962–1991) as
well as the water column transparency (Secchi depth) for
the whole period (Fig. 2g). Significant correlations were
found among these variables (Table 1). Correlations
between bulk phytoplankton variables of LBERI data
(1978–2003) and TP were not significant (Fig. 2c, e, f;
Table 1), presumably because the variation of trophic sta-
tus was much less pronounced during this later period.
Nevertheless, correlations among those bulk phytoplankton
variables (total biovolume and carbon) of SPFES and
LBERI data for the overlap period (14 years) were all
significant (Fig. 2b, c, d, e, f; Table 1), suggesting that the
bulk phytoplankton biomass from the two institutes showed
a similar response to environmental change, although these
two sets of data were collected with different methods and
in different locations. This supports our use of these two
datasets in parallel to investigate environmental effects on
phytoplankton in Lake Biwa.
Changes in the trophic status of Lake Biwa were also
seen in other chemical variables. Phosphate concentration
was low in the early period when TP started to increase
(Fig. 2h). Phosphate accumulated over time from 1972 to
1987, dropped rapidly since 1987, and then fluctuated
dramatically within a period of roughly 5 years. Nitrate
concentration was also low in the early period when TP
started to increase (Fig. 2i). It then accumulated over time
and leveled off after 1991. Silicate concentration was only
available after 1978, and it showed an increasing trend
(Fig. 2j). Ammonium concentration was almost undetect-
able in the early period when TP started to increase
(Fig. 2k). It increased suddenly in 1972, exhibited two
large peaks between 1972 and 1985, dropped to very low
value in 1985, and has increased since then.
Several chemical variables as well as their ratios were
correlated (Table 1). Phosphate/TP ratio correlated posi-
tively with nutrient concentrations and bulk phytoplankton
of SPFES series, and negatively correlated with mixing
condition and precipitation (Table 1). Ammonium/nitrate
ratio positively correlated with TP and ammonium con-
centration and bulk phytoplankton of SPFES series
(Table 1).
Physical data exhibited significant interannual variabil-
ity and a warming trend. Air temperature showed
interannual fluctuations (Fig. 3a), superimposed by a sig-
nificant warming trend (stationary bootstrapped corre-
lation; r = 0.641, p\0.05). Before 1990, air temperature
surpassing 15?C was only observed in 1979, but this high
value was observed frequently after 1990 (Fig. 3a). Albeit
strong short-term fluctuations, the average air temperature
after 1990 was significantly higher than that before 1990
(permutation test of the mean; p\0.001). One can also
notice a higher variability after 1990 (var = 0.311) than
that before 1990 (var = 0.166), although this difference
was only marginally significant (p = 0.082). A similar
situation was found in the surface water temperature
(Fig. 2b).A significantwarming
(r = 0.563, p\0.05), and the average surface water
trend wasfound
Environmental effects on phytoplankton473

Page 8

Table 1 Result of correlation analysis among environmental variables
TP
Phosphate
Nitrate
Ammonium
Silicate
Air T
Lake
surface
T
Lake
bottom
T
Wind
mixing
Thermocline
depth
Buoyancy
frequency
AO
Shiga
cell
volume
LBERI
cell
volume
Shiga
total
carbon
LBERI
total
carbon
1/
Secchi
depth
Chla
Precip-
itation
PAR
Phosphate/
TP
Ammonium/
nitrate
TP
NA
0
0
1
-1
0
0
0
0
0
0
0
1
0
1
0
1
0
0
-1
0
1
Phosphate
-0.102
NA
1
0
1
0
0
0
-1
-1
1
0
1
0
1
1
0
0
-1
0
1
0
Nitrate
-0.101
0.562
NA
0
1
1
1
1
-1
0
1
1
0
0
0
0
0
0
-1
-1
1
0
Ammonium
0.446
0.101
0.159
NA
0
0
0
0
0
0
0
0
1
0
1
0
1
0
0
-1
0
1
Silicate
-0.467
0.462
0.662
-0.286
NA
0
0
0
0
-1
0
0
0
0
0
0
0
0
0
0
1
0
Air T
-0.212
0.399
0.583
0.025
0.315 NA
1
1
0
0
1
1
0
0
0
0
0
0
-1
0
1
0
Lake surface
T
-0.020
0.371
0.507
0.134
0.260
0.868
NA
1
0
0
1
1
0
0
0
0
0
0
-1
0
0
0
Lake bottom
T
-0.344
0.194
0.601
0.135
0.343
0.627
0.613
NA
0
0
0
1
0
0
-1
0
0
0
-1
0
0
0
Wind mixing
-0.194
-0.585
-0.663
-0.202
0.264
-0.292
-0.245
-0.184
NA
1
-1
0
-1
-1
-1
-1
0
-1
0
1
-1
0
Thermocline
depth
-0.066
-0.473
-0.298
0.109
-0.350
-0.203
-0.265
0.188
0.358
NA
-1
0
0
0
0
0
0
1
1
0
-1
0
Buoyancy
frequency
0.121
0.349
0.508
0.070
0.236
0.561
0.557
0.089
-0.348
-0.448
NA
0
1
0
1
0
0
-1
-1
-1
1
0
AO
0.022
-0.061
0.349
-0.049
-0.017
0.388
0.291
0.401
-0.247
0.163
0.107
NA
0
0
0
0
0
0
0
-1
0
0
Shiga cell
volume
0.596
0.441
0.356
0.368
-0.256
-0.038
0.048
-0.271
-0.552
-0.177
0.516
-0.087
NA
1
1
1
1
1
0
-1
1
1
LBERI cell
volume
0.260
0.253
-0.155
0.430
-0.104
-0.406
-0.192
-0.240
-0.418
0.067
-0.231
-0.434
0.629 NA
1
1
0
1
0
0
0
0
Shiga total
carbon
0.594
0.437
0.328
0.371
-0.232
-0.059
0.028
-0.284
-0.538
-0.169
0.496
-0.100
0.999
0.628
NA
1
1
1
0
-1
1
1
LBERI total
carbon
0.290
0.354
-0.219
0.436
-0.086
-0.408
-0.149
-0.267
-0.404
-0.007
-0.203
-0.553
0.797
0.947
0.788 NA
0
1
0
0
0
0
1/Secchi
depth
0.448
-0.137
-0.196
0.479
-0.454
-0.139
-0.242
-0.128
0.075
0.170
-0.150
-0.049
0.326
0.420
0.327
0.366
NA
0
0
0
0
1
Chla
0.053
0.148
-0.182
0.377
-0.385
-0.292
-0.180
-0.034
-0.409
0.345
-0.458
-0.171
0.425
0.724
0.434
0.769
0.343
NA
0
0
0
0
Precipitation
0.341
-0.491
-0.401
0.253
-0.171
-0.526
-0.486
-0.310
0.378
0.324
-0.407
-0.139
0.028
0.356
0.039
0.297
0.408
0.148
NA
0
-1
0
PAR
-0.509
-0.332
-0.604
-0.370
-0.242
-0.135
-0.134
-0.159
0.701
0.150
-0.394
-0.212
-0.652
0.021
-0.639
0.060
-0.191
0.040
0.039
NA
0
0
Phosphate/
TP
-0.143
0.989
0.564
0.050
0.527
0.408
0.373
0.242
-0.552
-0.464
0.317
-0.068
0.368
0.202
0.364
0.300
-0.177
0.100
-0.509
-0.289
NA
0
Ammonium/
nitrate
0.424
-0.057
-0.094
0.936
-0.412
-0.059
0.051
0.046
-0.013
0.221
-0.049
-0.080
0.291
0.375
0.297
0.403
0.572
0.346
0.315
-0.232
-0.104
NA
Correlation coefficients are shown in the lower triangle, and results of stationary bootstrap tests (a = 0.05) are shown in the upper triangle with 1 indicating a significant positive correlation, -1 indicating a significant negative correlation, and 0 indicating no
significant correlation
474C. H. Hsieh et al.

Page 9

temperature after 1990 was significantly higher than that
before 1990 (p\0.001). A higher variability of surface
water temperature was seen in the later period (var =
0.326) compared with the earlier period (var = 0.201), but
therewas onlyamarginally
(p = 0.110). The warming signal was especially strong in
the bottom water temperature (Fig. 3c). The average bot-
tom temperature after 1990 was significantly higher
(p\0.001), but the variability was significantly lower
(p = 0.048).
Along with warming, the buoyancy frequency across the
thermocline also increased over time (r = 0.606, p\0.05;
Fig. 3d). The mean value and variance in the later period is
significantly higher than that in the earlier period
(p = 0.002 and 0.021, respectively). Wind mixing strength
declined initially from 1961 to 1975 and then leveled off
(Fig. 3e), and the long-term declining trend was significant
(r = -0.597, p\0.05). The mean mixing index was
slightly higher in the earlier period (p = 0.078), and no
significant difference in variance was found (p = 0. 226).
The mean and variance in thermocline depth in the earlier
period is slightly higher than the later period (p = 0.107
and p = 0.051, respectively), but no significant long-term
declining trend existed (r = -0.333, p[0.05; Fig. 3f).
Precipitation showed a long-term declining trend (r =
-0.546, p\0.05; Fig 3g), and its magnitude in the earlier
period was significantly higher than the latter (p = 0.003);
however, the variance was not different for the two periods
(p[0.05). The photosynthetic available radiation (Fig. 3h)
was high before 1970 and remained steady after 1970 (no
significant trend and no difference in mean magnitude and
variance between the two periods, p[0.05).
Most physical variables were correlated. Both air and
water temperatures were significantly correlated with the
Arctic Oscillation index (Table 1). Air and surface tem-
peratures had a positive effect on maximal buoyancy
frequency (Table 1), as one would expect. However, tem-
perature did not significantly affect thermocline depth
(Table 1); instead, it was the wind mixing strength that
determined thermocline depth (Table 1). In addition, the
wind mixing strength showed a negative effect on maximal
buoyancy frequency (Table 1). Temperature and tempera-
ture related variables were negatively correlated with
precipitation (Table 1). PAR was negatively correlated
with several nutrients, AO, and SPFES (1962–1991) phy-
toplankton abundance (Table 1).
significantdifference
Responses of phytoplankton community
Variation in trophic status and warming considerably
affected the phytoplankton community in Lake Biwa. The
increase in TP around 1967 was detected by the PC2 of
SPFES phytoplankton data, and the decrease in TP around
1985 was reflected in the PC2 of SPFES data and the PC1
and PC2 of LBERI data (Fig. 4). The PC1 of SPFES data
seemed to follow the variation of TP but with a time delay
(Fig. 4). The dramatic warming around 1990 also signifi-
cantly altered the phytoplankton community (with 1 year
delay), as shown by the PC1 and PC2 of LBERI data
(Fig. 4). The phytoplankton community clearly signaled
environmental variations.
The linkage between temporal variation of the phyto-
plankton community and environmental variables are
illustrated in RDA plots (Fig. 5). We separated the ordi-
nation of years and species into two plots for better
visualization. When overlapped, one can find the relation-
ships among sampling years, phytoplankton taxa, and
environmental factors. For the SPFES data, the change in
the phytoplankton community followed mainly the trophic
status of the lake, with a strong correlation with TP, nitrate
and total phytoplankton carbon (Fig. 5a). In addition, the
Fig. 4 Principal components based on phytoplankton data from a the
SPFES and b the LBERI. Vertical dashed lines indicate roughly the
years when the lake environment changed
Environmental effects on phytoplankton 475

Page 10

early period was affected by mixing condition (Fig. 5a).
The associations between phytoplankton taxa and the
environmentalfactorswere
Fig. 5b). The phytoplankton taxa can be statistically cate-
gorized into five groups, and these groups largely depicted
the evolution of the environmental state of the lake from
1962 to 1991 (Fig. 5b). For the LBERI data, the phyto-
plankton community was related to TP and precipitation in
the early period (1978–1985) and then to maximum
buoyancy frequency and nitrate concentration in a later
period (Fig. 5c). The associations between phytoplankton
taxa and the environmental factors are shown in Fig. 5d.
The phytoplankton taxa can be categorized into six groups;
however, the separation of these groups was less distinct
(Table 2, Fig. 5d). The group L6 showed no significant
variation throughout the period, while group L1 to L5
revealed a transition (Fig. 5d).
alsorevealed(Table 2,
The signals of phytoplankton in response to environ-
mental events remained strong even when only presence/
absence data of species were considered (Fig. 6). As can be
seen in Fig. 6a (SPEFS data), some species mainly
occurred during the eutrophic period, while others only
appeared when the eutrophic condition diminished. Simi-
larly in Fig. 6b (LBERI data), some species mainly
occurred between 1978 and 1985, while others only
appeared when the eutrophication faded away. Interest-
ingly, some species emerged after the temperature
increased.
Species-specific responses to each environmental factor
corroborated with the results of RDA. To visualize this,
we plotted correlation structures between each phyto-
plankton species and environmental variables categorized
by group (Fig. 7). For groups S1, S2, and S3 of SPFES
data, many species showed no significant correlation with
Fig. 5 Ordination biplots exhibiting the relationships among years
and phytoplankton species and their linkage to environmental
variables for (a, b) SPFES and (c, d) LBERI data. a and c represents
ordination of years, and b and d represents ordination of species. Only
significant (p\0.05) environmental variables (red arrows) were used
based on stepwise selection. Color coding indicates phytoplankton
groups classified from K-means classification: black, group 1; blue,
group 2; red, group 3; green, group 4; magenta, group 5; cyan, group
6. See Table 2 for species names
476 C. H. Hsieh et al.

Page 11

Table 2 List of frequent species (occurring [5 years in samples)
used in the analyses. Shown are species ID along with group classi-
fications. The ID corresponds to the order in Fig. 6
SpeciesID Group
SPFES
Staurastrum dorsidentiferum
1 S4
Pediastrum biwae
2 S4
Closterium aciculare
3S4
Asterionella formosa
4S5
Melosira solida
5 S4
Stephanodiscus carconensis
6 S5
Melosira varians
7S5
Melosira granulata
8 S5
Sphaerocystis schroeteri
9S3
Ceratium hirundinella
10 S2
Eudorina elegans
11S2
Xanthidium antilopaeum
12S3
Staurastrum limneticum
13S2
Aphanocapsa elachista
14S3
Cosmocladium constrictum
15 S1
Synedra ulna
16S3
Fragilaria capucina
17S3
Aphanothece nidulans
18 S3
Aphanocapsa sp. 19 S1
Staurastrum sp. 20 S1
Dinobryon sp. 21S2
Spirogyra sp.22S5
Fragilaria sp.23S2
Synedra sp.24 S1
Oscillatoria sp.25S2
Aphanothece sp.26 S2
Diatoma sp.27 S1
Microcystis sp. 28S3
Staurastrum arctiscon
29 S4
Melosira italica
30 S5
Spondylosium moniliforme
31S4
Oocystis sp. 32S4
Chroococcus dispersus
33S4
Fragilaria crotonensis
34S5
Xanthidium hastiferum
35S5
Mougeotia sp. 36S5
LBERI
Stephanodiscus carconensis
1 L3
Staurastrum dorsidentiferum
2 L6
Melosira granulata
3 L4
Fragilaria crotonensis
4 L5
Cryptomonas sp.5 L4
Coelastrum cambricum
6 L6
Closterium aciculare
7 L6
Chroococcus dispersus
8 L4
Chlamydomonas sp.9 L3
Table 2 continued
SpeciesIDGroup
Aphanothece clathrata
10L3
Ankistrodesmus falcatus
11L2
Uroglena americana
12 L6
Nitzschia sp. 13 L5
Asterionella formosa
14L4
Pediastrum biwae
15 L1
Nitzschia acicularis
16 L5
Melosira solida
17 L1
Gymnodinium helveticum
18 L5
Cyclotella glomerata
19 L6
Cosmocladium constrictum
20 L4
Elakatothrix gelatinosa
21 L2
Closterium sp. 22L5
Ceratium hirundinella
23L5
Oocystis lacustris
24L1
Staurastrum pingue
25L6
Mougeotia sp. 26 L1
Gloeocystis sp. 27 L5
Kirchneriella contorta
28 L2
Cyclotella stelligera
29L2
Scenedesmus sp.30 L5
Melosira italica
31 L1
Mallomonas fastigata
32L6
Oocystis parva
33 L2
Spondylosium moniliforme
34L6
Oocystis sp. (a)
35L1
Actinastrum hantzschii
36 L1
Planktosphaeria gelatinosa
37 L1
Gloeocystis vesiculosa
38 L1
Gloeocystis gigas
39L1
Glenodinium sp. 40L1
Cryptomonas erosa
41L1
Quadrigula lacustris
42L1
Coelastrum microporum
43L5
Chodatella citriformis
44 L1
Staurastrum longiradiatum
45 L6
Scenedesmus denticulatus
46L1
Synedra acus
47L5
Gymnodinium sp.48L5
Cyclotella sp.49L5
Eudorina elegans
50L6
Tetraspora lacustris
51L6
Oocystis submarina
52L2
Staurastrum arctiscon
53 L5
Aphanocapsa elachista
54L2
Navicula sp. 55L6
Ankistrodesmus sp.56L6
Dictyosphaerium pulchellum
57 L5
Environmental effects on phytoplankton477

Page 12

environmental variables, as did many species in the group
L6 of LBERI data. Further, the patterns were clear in the
LBERI data but not very strong in the SPFES data.
Discussion
Ecosystem changes
Dramatic changes of trophic status and water column
physical properties have occurred during the past half
century in Lake Biwa. The start of eutrophication in the
late 1960s and return to a less eutrophic state in the mid
1980s had been suggested according to anecdotal evidence,
sediment records, and museum fish samples (Ogawa et al.
2001; Nakazawa et al. 2010). However, here for the first
time, the history of the trophic status of Lake Biwa is
documented based on direct measurements of water col-
umn chemical data. These chemical data overcome the
issues of coarse temporal resolution and high uncertainty
associated with sediment records and museum fish
samples.
The total phytoplankton biomass was mainly determined
by trophic status. The TP data clearly showed a rapid
increase after 1967, reached maximum in 1974, then
declined until 1985, and fluctuated around a stable value
thereafter (Fig. 2a). The return to a less eutrophic condition
in the mid 1980s was due to a successful governmental
Fig. 6 Species occurrence for phytoplankton species collected from
(a) the SPFES and (b) the LBERI. The order (top–down) follows the
first occurrence of species in the data series. Only species occurred for
more than five sampling years are included. See Table 2 for species
names
Table 2 continued
Species IDGroup
Schroederia judayi
58L1
Staurastrum sp.59L5
Micractinium pusillum
60L6
Crucigenia irregularis
61L1
Cocconeis placentula
62 L5
Synedra sp. 63L6
Chroococcus sp.64L6
Peridinium sp. 65L5
Melosira distans
66 L6
Mallomonas tonsurata
67 L6
Dinobryon bavaricum
68 L4
Microcystis sp.69 L5
Gomphonema sp.70 L1
Chodatella subsalsa
71L1
Xanthidium hastiferum
72L6
Mallomonas akrokomos
73 L6
Chrysamoeba radians
74L3
Chromulina sp.75L2
Trachelomonas sp. 76 L6
Gomphosphaeria lacustris
77L4
Mallomonas sp. 78L5
Rhodomonas sp.79 L4
Planktosphaeria sp. 80L3
Peridinium berolinense
81L4
Pseudokephyrion gallicum
82 L3
Monoraphidium tortile
83 L3
Chlorocloster sp. 84L3
Oocystis sp. 85L5
Oocystis solitaria
86L4
Microcystis wesenbergii
87L5
Peridinium elpatiewskyi
88L5
Microcystis incerta
89 L5
Microcystis aeruginosa
90L5
Quadrigula chodatii
91 L5
478C. H. Hsieh et al.

Page 13

mitigation enforced since 1982 (Kumagai 2008). The total
phytoplankton biomass (or its surrogate) largely followed
the trend of TP (Table 1, Fig. 2b, d, g), at least for the
period from 1962 to 1991 (SPFES data). Further, cross-
correlation analyses between TP and phytoplankton bio-
mass indicators showed that the best correlation occurred
when phytoplankton biomass lagged 1 year behind TP,
suggesting a quick response of phytoplankton total biomass
to the control of nutrient loading. This is in contrast to
many studies (e.g. Anneville and Pelletier 2000; Horn
2003; Dokulil and Teubner 2005) that discovered none or
delayed response of phytoplankton biomass to nutrient
mitigation. The negative correlation between total phyto-
plankton biomass and light (photosynthetic available
radiation measured from the meteorological station) may
be a spurious result; however, we do not have measure-
ments of light intensity in the water column to clarify this
issue. The negative correlation between mixing strength
and SPFES phytoplankton biomass implies negative effects
of strong mixing on phytoplankton, perhaps by destructing
Fig. 7 Correlation structures
between phytoplankton species
and environmental variables
categorized by group for
a SPFES and b LBERI data.
Filled and open circles indicate
significant positive and negative
correlations, respectively. See
Table 2 for species ID
Environmental effects on phytoplankton479

Page 14

vertical structure of phytoplankton that is critical for their
light acclimation (Tirok and Gaedke 2007; Winder and
Hunter 2008).
The correlations between TP and LBERI phytoplankton
bulk measurements (Fig. 2c, e, f) were not significant
(Table 1). This is maybe because the variation of TP has
been at a low level during this period (1978–2003) and
factors in addition to nutrients had also contributed to
affecting phytoplankton, such as mixing regime (Table 1).
The negative correlation between mixing strength and
LBERI phytoplankton biomass also existed, as in the
SPFES data series. The differential responses of the SPFES
and LBERI data series to environmental variables may be
because they covered different periods. The critical envi-
ronmental forcing on phytoplankton may have changed
through time, and the phytoplankton species compositions
were different for these two periods (Table 2, Figs. 4, 6).
Inorganic nutrients exhibited complex variation irre-
spective of TP. Phosphate concentration was low during
the early period of eutrophication presumably due to rapid
uptake of phytoplankton, and then accumulated gradually
up to 1987 (Fig. 2h). Such kinds of progression were
commonly observedduring
(Anneville et al. 2002a). However after 1987, phosphate
fluctuated with a high amplitude despite TP remaining
stable during that period (Fig. 2a, h); that is, even though
nutrient loading had been controlled, phosphate still
showed episodic high pulses. Preliminary analyses sug-
gested this is related to internal loading from the sediments
(Murphy et al. 2001; Kumagai 2008). The nitrate (Fig. 2i)
and silicate (Fig. 2i) accumulating over time was likely due
to both external and internal loading (Kumagai 2008).
Ammonium (Fig. 2k) was low during the early period of
eutrophication presumably due to rapid uptake of phyto-
plankton, and showed two strong peaks between 1972 and
1985, which was likely associated with eutrophication. The
fluctuation of ammonium is not a result of measurement
error, as the ammonium data from LBERI also showed a
similar pattern. We found a significant correlation between
the phosphate/TP ratio and air temperature, mixing index,
and precipitation (Table 1), suggesting climate variations
might have effects on nutrient dynamics through regional
physical processes.
Global climate changes have affected Lake Biwa
through atmospheric forcing. This can be seen from sig-
nificant correlations between the Arctic Oscillation index
versus air and water temperature (Table 1, Fig. 3). This
association is especially strong in winter (Fig. 3j). It is
worth noting that the temperature of Lake Biwa has
warmed significantly in the past half century, with a
warming rate of 0.031?C/year for air temperature (Fig. 3a)
and 0.028?C/year for surface water temperature (Fig. 3b).
Increased temperature enhanced water column stability as
eutrophicationin lakes
signified by the max buoyancy frequency (Table 1,
Fig. 3d), which potentially reduced nutrient mixing and
upwelling (Hsieh et al. 2009b). Temperature did not
directly affect the thermocline depth (Table 1, Fig. 3f).
Rather, thermocline depth was influenced by the strength of
wind mixing (Table 1, Fig. 3e), as also shown in other
lakes (Coats et al. 2006). The strength of wind mixing has
weakened along with warming (Fig. 3e), and further rein-
forced water column stability (Table 1). The enhanced
water column stability due to the combination of increased
water temperature and reduced wind mixing is attributed to
the reduced bottom water renewal and consequently anoxic
benthic condition, which might lead to catastrophic mass
mortality of benthic fishes (Kumagai 2008).
Air and water temperature and max buoyancy frequency
after 1990 was significantly higher than that before 1990
(Fig. 3a, b, c). In addition, the variance was also higher
during the later period, suggesting that the lake physical
environment in this later period was unstable. The cause of
this high variability is not known at this time. However,
such increased variance has been suggested to be indicative
of ecosystem transition, often to undesired states (Brock
and Carpenter 2006; Carpenter and Brock 2006). Thus, the
effects of high variability of the living environment on
organisms warrant further investigation.
Reorganizations of phytoplankton community
Variation in the phytoplankton community was clearly
affected by changes in trophic status and the physical
environment of lake Biwa (Fig. 5). For the SPFES data, the
changes in community were mainly caused by nutrient-
related variables through time (roughly clock-wise evolu-
tion in Fig. 5a, b). Some species developed after
eutrophication started, and others occurred before eutro-
phicationandafterrecovery.
community showed a clear change as revealed by the sta-
tistically defined five groups (Figs. 5b, 6a, Appendix 3 of
Supplementary material). For the LBERI data, the changes
in community were caused by nutrient-related and tem-
perature-related variables (roughly clock-wise evolution in
Fig. 5c, d). One can replace the max buoyancy frequency
with lake water temperature or AO to obtain qualitatively
similar RDA results, indicating that climate driven changes
in physical properties of the water column have had sig-
nificant impacts on the phytoplankton community in Lake
Biwa. One can replace nitrate concentration with ammo-
nium/nitrate ratio and obtain qualitatively similar RDA
results, indicating that nutrient dynamics also contributed
to variations in the phytoplankton community. In addition,
we found precipitation was a significant variable in RDA.
The precipitation might affect phytoplankton community
by changing nutrient input or turbidity of the water column.
Thephytoplankton
480C. H. Hsieh et al.

Page 15

Some species only appeared during eutrophication (1978–
1985 for the LBERI series), and others only occurred after
warming (1990) (Figs. 5d, 6b, Appendix 2 of Supplemen-
tary material). Interestingly, several cyanobacteria taxa
(such as Microcystis spp.) established large concentrations
during the warming period (Appendix 2 of Supplementary
material). This may be because they prefer stable water
conditions (Elliott et al. 2006), and at the same time, the
nutrient supply was sufficient (Fig. 2h, i) for them to
prosper. In fact, nutrients could come from internal loading
due to anoxic bottom conditions caused by warming and
thus stronger water column stability. These results suggest
synergistic effects of eutrophication and warming in Lake
Biwa in driving phytoplankton community dynamics.
The changes in trophic status and warming have driven
reorganization of the phytoplankton community in Lake
Biwa. On the one hand, while the total phytoplankton
biomass has largely followed TP (Fig. 2), the community
structure has changed in the SPFES data (Figs. 4, 5, 6). On
the other hand, while the total phytoplankton biomass
remained at a stable level, the phytoplankton community
almost completely reorganized in response to eutrophica-
tion and warming in the LBERI data (Figs. 4, 5, 6). Such a
phenomenon was also observed in other lakes (Zohary
2004; Winder and Hunter 2008), and we believe this is the
norm rather than the exception. One might consider these
reorganizations as nonlinear transitions among multiple
stable states (Scheffer et al. 2001; Hsieh et al. 2008).
Conceptually, total phytoplankton live at the limit of the
carrying capacity of the lake, determined by resource
availability (e.g. nutrients). Phytoplankton species compete
for resources; some species were sensitive to different
nutrients, whereas others were sensitive to temperature or
the mixing condition of the water column, as exemplified
in the current study (Figs. 5, 6, 7). The winners are decided
by a certain threshold of resource factors, or more likely by
multiplication of several factors; that is, several conditions
(both environmental and biological) need to be fulfilled
simultaneously for a new state to emerge (Dixon et al.
1999; Hsieh et al. 2005). Once the tipping point is reached,
the winners are determined. Often through some kind of
positive feedback, the winner can stay at the stage for a
prolonged period until a different threshold is reached,
owing to external forcing such as a nutrient pulse or change
in temperature (Scheffer and Carpenter 2003).
Nonlinear transitions among stable states might have
happened in the Lake Biwa phytoplankton community.
Significant changes in relative abundance (Fig. 4) and even
species members (Fig. 6) in different trophic regimes or
physical conditions were observed. Although community
patterns showed correlations with environmental factors
(Fig. 5), most of the correlations were weak. When indi-
vidual species was considered, the correlations were even
weaker or non-existent (Fig. 7). Nonetheless, the changes
of community as an ensemble (Fig. 4) were consistent with
the eutrophication and warming in timing. The weak linear
correlations and threshold responding of the community
implies regime shifts of phytoplankton in Lake Biwa
(Hsieh and Ohman 2006). However, we cannot rule out the
possibility that such results may arise due to the change in
vertical distribution of phytoplankton species (Anneville
et al. 2002a).
It is likely that the trophic status and phytoplankton
community in the large north basin was influenced by the
small, shallow and eutrophic south basin, because flows
between two basins have been observed (Kumagai 2008). It
was speculated that cyanobacteria (Microcystis) observed
in the north basin might be seeded from the south basin
(Kumagai 2008). However, to what extent the north basin
was influenced by the south basin needs further study.
Limitations of the data
We had hoped to completely integrate the historical data
collected in Lake Biwa when we started to compile those
time series. However, our data contain several caveats.
First, the discrepancy between the sampling methods for
phytoplankton by the two institutes makes data integration
difficult. Second, the phytoplankton sampling in the
LBERI was limited to the surface layer, which prevented
us from investigating the vertical distribution of species.
Third, the large mesh size used in sampling phytoplankton
in the SPFES data likely missed important small species.
Fourth, the TP data are compiled from Station I and L
(Fig. 1), which are far apart, and these two series have
limited overlap to facilitate rigorous calibration. Here, we
faithfully present the data (Appendices 1–4 of Supple-
mentary material). Readers can judge our analyses with
those caveats in mind.
Conclusion
We analyzed long-term time-series data for the Lake Biwa
ecosystem from 1962 to 2003. Analyses on environmental
data indicate that Lake Biwa had experienced intensified
eutrophication in the late 1960s and return to a less
eutrophic status around 1985, and then exhibited rapid
warming since 1990. The phytoplankton clearly responded
to the change in trophic status and to more recent warming
in Lake Biwa. Generally, the total phytoplankton biomass
followed the rise and fall of total phosphorus loadings in
the Lake (Fig. 2), exhibiting a symmetrical behavior. This
suggests a quick response of phytoplankton total biomass
to the control of nutrient loading. However, the phyto-
plankton community changed dramatically (Figs. 4, 5, 6),
Environmental effects on phytoplankton 481