Hydrobiologia (2022) 849:3963–3978

https://doi.org/10.1007/s10750-021-04773-w (0123456789().,-volV)
(0123456789().,-volV)

ECOLOGY OF SHALLOW LAKES

Shorter blooms expected with longer warm periods

under climate change: an example from a shallow meso-
eutrophic Mediterranean lake
Gary Free . Mariano Bresciani . Monica Pinardi . Steef Peters .
Marnix Laanen . Rosalba Padula . Alessandra Cingolani . Fedra Charavgis .
Claudia Giardino

Received: 20 May 2021 / Revised: 20 October 2021 / Accepted: 20 November 2021 / Published online: 9 January 2022
Ó The Author(s) 2021

Abstract Satellite data from the Climate Change bloom related to the earlier onset of warmer temper-
Initiative (CCI) lakes project were used to examine the atures. Instead, a curvilinear relationship between Chl-
influence of climate on chlorophyll-a (Chl-a). Non- a and the temperature length of season above 20°C
parametric multiplicative regression and machine (LOS) was found with longer periods of warmer
learning were used to explain Chl-a concentration temperature leading to blooms of shorter duration. We
trend and dynamics. The main parameters of impor- suggest that a longer period of warmer temperatures in
tance were seasonality, interannual variation, lake the summer may result in earlier uptake of nutrients or
level, water temperature, the North Atlantic Oscilla- increased calcite precipitation resulting in a shortening
tion, and antecedent rainfall. No evidence was found of the duration of phytoplankton blooms. The current
for an earlier onset of the summer phytoplankton scenario of increasing LOS of temperature with
climate change may lead to an alteration of phyto-
plankton phenological cycles resulting in blooms of
Guest editors: José L. Attayde, Renata F. Panosso, Vanessa
Becker, Juliana D. Dias & Erik Jeppesen / Advances in the
shorter duration in lakes where nutrients become
Ecology of Shallow Lakes

G. Free (&)  M. Bresciani  M. Pinardi  C. Giardino R. Padula  A. Cingolani  F. Charavgis

Institute for Electromagnetic Sensing of the Environment, ARPA UMBRIA, Via Pievaiola, 207/B-3 Loc. S. Sisto,
National Research Council, Via Bassini 15, 20133 Milan, 06132 Perugia, Italy
Italy e-mail: [email protected]
e-mail: [email protected]
A. Cingolani
M. Bresciani e-mail: [email protected]
e-mail: [email protected]
F. Charavgis
M. Pinardi e-mail: [email protected]
e-mail: [email protected]
C. Giardino
e-mail: [email protected]

S. Peters  M. Laanen
Water Insight, Marijkeweg 22, 6709 PG Wageningen, The
Netherlands
e-mail: [email protected]
M. Laanen
e-mail: [email protected]

123
3964 Hydrobiologia (2022) 849:3963–3978

limiting. Satellite-derived information on lake tem- has been estimated as 0.34°C per decade but the
perature and Chl-a concentration proved essential in influence of lake specific parameters like morphology
detecting trends at appropriate resolution over time. frequently result in a lack of regional consistency to
these trends (O’Reilly et al., 2015). Recent projections
Keywords Shallow lakes  Cyanobacteria  have indicated that lakes are predicted to get warmer
Chlorophyll-a  Climate change  Remote sensing for longer periods, with heatwaves potentially extend-
ing across multiple seasons (Woolway et al., 2021).
Climate has been found to alter the seasonal pattern of
phytoplankton in several ways, for example, through
Introduction altering the physical environment with warmer win-
ters reducing the incidence of stratification overturn
Shallow lakes have been identified as providing up to and thereby altering nutrient cycling (Rogora et al.,
39 essential ecosystem services globally such as 2018; Free et al., 2021a). Whereas in high latitude
drinking water, biodiversity, and recreational use lakes an increase in the length of growing season of
which can be linked to sustainable development goals phytoplankton has been attributed to an increasing
(Janssen et al., 2021). Anthropogenic pressure can proportion of rain relative to snow resulting in earlier
cause a shift in state from being dominated by delivery of nutrient loads to lakes (Maeda et al., 2019).
macrophytes to phytoplankton with a resulting loss For cyanobacteria, warmer temperatures have been
in ecological functioning and degraded ecosystem found to result in the earlier occurrence of the growing
services (Scheffer, 1989; Scheffer & van Nes, 2007; season of Microsystis (Deng et al., 2014). Extreme
Janssen et al., 2021). The abundance of phytoplankton increases in rainfall followed by warmer winter
is a key component used in assessments of trophic temperatures can combine to increase external and
state and ecological quality of lakes (Council of the internal loading leading to earlier blooms of increased
European Communities, 2000; Phillips et al., 2013). duration and extent (Qin et al., 2021). Warmer
Some phytoplankton species, such as cyanobacteria, temperatures resulting from climate change have been
can form blooms or scum and can potentially produce predicted to lead to an increase in cyanobacteria
toxins (cyanotoxins) that can have adverse effects on through direct growth and also by promoting a
aquatic ecosystems and human health (Codd et al., stable water column under which buoyant cyanobac-
2005; Jöhnk et al., 2008; Ma et al., 2015). Physical and teria have a competitive advantage (Jöhnk et al.,
chemical parameters as well as seasonality affect 2008). Future scenarios have predicted an increase in
phytoplankton distribution, abundance, and species the intensity and frequency of cyanobacteria blooms
diversity (Raymond, 1983; Ezra & Nwankwo, 2001). resulting from warmer temperatures and eutrophica-
For example, cyanobacteria are typically more abun- tion (Paerl & Huisman, 2009; O’Neil et al., 2012;
dant in nutrient-rich freshwater and are regulated by Huisman et al., 2018).
seasonal drivers, such as warmer temperatures and Satellite remote sensing can enable the observation
calmer weather which influence water column thermal of a suite of indicators of water quality and ecosystem
stratification (Reynolds, 2006). In large shallow lakes condition for the study of long-term environmental
which are polymictic (lack stratification), calm warm trends in lakes (Tyler et al., 2016; Free et al., 2020) and
weather and directional light wind can be associated supports different studies on the consequences of
with surface accumulations of phytoplankton as well climate change on algal blooms (Qin et al., 2019; Shi
as promoting nutrient increase through internal load- et al., 2019). Phytoplankton is one the most frequently
ing that can further fuel blooms following storms studied parameters with chlorophyll-a (Chl-a) often
(Søndergaard et al., 2003; Shi et al., 2017; Bresciani used as a proxy of phytoplankton biomass. The spatial
et al., 2020; Free et al., 2021b). coverage and temporal sampling frequency achievable
Climate change has an important influence on with satellite images allow novel insights into phyto-
global biodiversity in lakes and has been listed as the plankton dynamic processes in lakes (Ho et al., 2019),
third most important driver after invasive species and especially when paired with high-frequency ground-
land-use change (Sala et al., 2000). The global based measurements (Bresciani et al., 2020) that
increase in surface temperature in lakes in summer cannot be easily captured through in situ sampling

123
Hydrobiologia (2022) 849:3963–3978 3965

(Palmer et al., 2015). Remote sensing has therefore for the years 2015–2017, while the annual average of
been identified as a key component of future high- Chl-a ranged from 5.37 to 14.73 mg m-3 for this
level lake water quality management strategies (Car- period (Cingolani & Charavgis, 2018). This can be
valho et al., 2019). compared with the background reference concentra-
Here we examine a 16-year (2003–2018) time tions of Chl-a assigned in legislation of 3.3 mg m-3
series for a shallow lake (Trasimeno) in central Italy in (DM, 2010). The values for Chl-a and TP would place
order to identify changes in Chl-a concentration trend the lake at the mesotrophic–eutrophic boundary
and dynamics in the context of climate change. according to the OECD trophic classification system,
Satellite observations of Chl-a provide higher tempo- with TP not significantly changing since the 1970s
ral resolution in addition to their capacity to integrate (OECD, 1982; Ludovisi & Gaino, 2010). Average
signals over a large spatial area of the lake. We expect alkalinity for the years 2015–2017 was 191 mg l-1
that warmer summers should increase the intensity and CaCO3. The water column is unstratified, with recur-
duration of summer blooms in the lake in line with rent sediment resuspension due to wind action.
climate change predictions. According to the Water Framework Directive
(WFD) system of classification, the lake is currently
classified at moderate ecological status (Council of the
Materials and methods European Communities, 2000, 2013; Cingolani &
Charavgis, 2017, 2018). Lake Trasimeno has a phy-
Study area toplankton assemblage dominated by chlorophytes
and dinoflagellates. Cryptophytes can also comprise a
This study focuses on Lake Trasimeno, a post- relatively large portion of the biomass, whereas
tectonic, shallow lake (maximum depth 6.3 m), situ- euglenophytes and diatoms are relatively scarce
ated at an altitude of 258 m in central Italy (43° 080 N; (Havens et al., 2009). In summer, the high nutrient
12° 060 E; Fig. 1). It is the fourth largest lake in the availability favors the occurrence of phytoplankton
country (surface area 120.5 km2) and has a circular blooms, including cyanobacteria species (e.g., Cylin-
shape (Ludovisi & Gaino, 2010). The catchment area drospermopsis raciborskii, Planktothrix agardhii)
(excluding the lake) is 264.5 km2 and given the small (Havens et al., 2009; Charavgis et al., 2020).
catchment area relative to that of the lake, the annual
input of water can be lower than that lost through WISPstation in situ data
evaporation with a residence time highly dependent on
rainfall with estimates ranging from [ 4 to [ 20 The WISPstation (Peters et al., 2018) is a fixed
years (Jørgensen et al., 2007; Frondini et al., 2019). position, above water autonomous spectroradiometer.
Therefore, the water balance of the lake is strongly It was installed on April 24th, 2018, 400 m north of
dependent on rainfall which is typically 700 mm— Polvese island in Lake Trasimeno (Fig. 1) and collects
just sufficient to keep the lake level stable, and with no remote sensing reflectance (Rrs) data every 15 min. It
significant outflow the lake can be classified as measures the radiance and irradiance in the spectral
endorheic (Dragoni, 2004; Frondini et al., 2019). range of 350–900 nm with a spectral resolution of
Land cover as a percentage of total catchment area 4.6 nm (full width at half maximum). The system is
(i.e., lake plus its catchment) was estimated as water calibrated relative to a reference instrument (cali-
(39.1), agricultural—mostly dominated by arable brated in a certified laboratory using a lamp and
(30.9), forest (22.3), baren land (4.9), and urban integrating sphere with National Institute of Standards
(2.9) (Giardino et al., 2010). Lake Trasimeno is of and Technology traceable calibrations). The WISPs-
considerable conservation importance, being part of tation is watertight and built into a highly climate-
the largest Natural Regional Park in peninsular Italy, a resistant case. Temperature of the sensor and humidity
Site of Community Interest, a Special Protection Zone in the case are registered with every measurement.
and contains two Natura 2000 sites (IT5210018 and Data are automatically sent using 3G connectivity to
IT5210070). Lake Trasimeno is generally turbid the database (WISPcloud). The instrument can be
(average Secchi disk depth 0.9 m) with an annual remotely accessed and, e.g., updated or configured to a
average total phosphorus concentration of 27 lg l-1 specific time interval or measurement frequency. It is

123
3966 Hydrobiologia (2022) 849:3963–3978

Fig. 1 True color composition of Lake Trasimeno from satellite (Sentinel-2 acquired on 5 June 2019) together with the WISPstation
position (red dot)

autonomously powered by a solar panel and internal Long-term EO dataset

large battery. Its hyperspectral features allow a
simulation of the band setting of satellite data and The European Space Agency’s Climate Change Ini-
hence provides valuable reference data. For water tiative (CCI) aims to exploit the long-term global
quality monitoring purposes, the remote sensing Earth Observation record to produce Essential Climate
reflectance observations are run through standard Variables (ECVs) supporting the United Nations
water quality algorithms to make a first estimate of Framework Convention on Climate Change
Chl-a (Gons, 1999). The WISPstation-derived data of (UNFCCC). The objective of the CCI dataset for the
Chl-a concentrations were validated against values ECV Lakes is to use satellite Earth Observation data to
measured in water samples analyzed by ARPA create the largest and longest possible consistent, open
Umbria (Bresciani et al., 2020). The WISPstation also global record of five lake thematic variables: lake
enables estimates of the concentration of phycocyanin water level, lake area extent, temperature, water-
using specific algorithms (Simis, 2006). A subset of leaving reflectance, and ice cover. The main charac-
the data from July 1st until September 30th, 2019, was teristics of this dataset (Crétaux et al., 2020) are as
used for analysis. follows:

123
Hydrobiologia (2022) 849:3963–3978 3967

– Spatial coverage: 250 globally distributed lakes, Service. Available online: https://annali.regione.
set to expand to around 2000 in the second phase umbria.it/# (accessed on 15 March 2021). Several
– Spatial resolution: 1/120° global grid studies have detected changes in Italian lakes linked to
– Temporal resolution: daily netCDF files contain- long-term climate change and fluctuations in large
ing all thematic variables including uncertainty scale regional climate drivers such as the North
– Temporal coverage: from 1992 up to 2019. Atlantic Oscillation index (NAO) and the East
Atlantic pattern (EA) during winter (Rogora et al.,
For Lake Trasimeno, lake surface water tempera-
2018; Salmaso et al., 2018). Daily values of the North
ture (LSWT) and Chl-a (derived from water-leaving
Atlantic Oscillation (NAO) were obtained from
reflectance data from MERIS and OLCI sensors) were
NOAA-CPC (https://www.cpc.ncep.noaa.gov/
extracted from the CCI Lakes database version 1.0.
products/precip/CWlink/pna/nao.shtml).
Quality level 4 and 5 were used to filter temperature
values; these represent the level of confidence that the
Analysis
LSWT is valid and are classified as suitable for climate
applications, similarly only a Chl-a uncertainty below
For the analysis, two approaches were trialed—
60% was used for analysis (Simis et al., 2020). The
Google AI (Artificial Intelligence) and Nonparametric
dataset for LSWT dates from 1993 while that for Chl-a
Multiplicative Regression (NPMR). NPMR (McCune,
starts in 2002. A total of 24 satellites were used to
2006) was used to estimate the response of average
compile data in the CCI Lakes project and are listed in
daily Chl-a concentration to climate and environmen-
the product user guide (Simis et al., 2020). The CCI
tal parameters listed above. NPMR can define
dataset presents an opportunity to examine what
response surfaces using predictors in a multiplicative
parameters are important in controlling the size of
rather than in an additive way. This method is
phytoplankton blooms and their interannual variation
progressive in better defining unimodal responses
and phenology. There is a significant data gap in Chl-a
than other methods such as multiple regression
from 2012 to 2015 when optical satellite data with the
(McCune, 2006). It has previously been applied to
same characteristics (i.e., MERIS and OLCI) were not
model tree species distribution (Yost, 2008), the
available, while other approaches, e.g., Huang et al.
response of lichens to climate change (Ellis et al.,
(2019) who used MODIS for long-term mapping of
2007), and in time-series analysis (Nicolaou &
phytoplankton in Lake Taihu, were not yet considered
Constandinou, 2016). NPMR was applied using the
suitable for Lake Trasimeno. In addition, gaps are also
software HyperNiche version 2.3 (McCune & Mef-
present due to satellite overpass limitations, where
ford, 2009). The response of Chl-a was estimated
images were contaminated by cloud cover or where
using a local mean multiplicative smoothing function
marked invalid for analysis (Simis et al., 2020). One of
with Gaussian weighting. NPMR models were pro-
the objectives of the work was to test the ECV Lakes
duced by adding predictors stepwise with fit expressed
dataset, exploring potential for scientific research and
as a cross-validated R2 (xR2) which can be interpreted
also the current limitations of the dataset.
in a similar way as a measure of fit as a traditional R2.
Few in situ environmental datasets are available to
The sensitivity, a measure of influence of each
match this temporal resolution so daily climatic data
parameter included in the NPMR model, was esti-
were obtained from ERA5, the fifth generation of the
mated by altering the range of predictors by ± 0.05
European Centre for Medium-Range Weather Fore-
(i.e., 5%) with resulting deviations scaled as a
casts (ECMWF) reanalysis for the global climate and
proportion of the observed range of the response
weather (https://cds.climate.copernicus.eu/cdsapp#!/
variable. Sensitivity can be used to evaluate the
home). Data used for analysis included the follow-
relative importance of variables included in models
ing: wind vectors to represent speed and direction, 2 m
because NPMR models are unlike linear regression
air temperature (temperature of air at 2 m above the
and have no fixed coefficients or slopes. Google AI
surface), total precipitation, and the sum of rainfall for
AutoML was also used to develop a model for Chl-a as
the preceding seven days. Lake level at S. Savino
a designated feature. The supervised learning model
station (Magione village) was obtained from the
followed a regression approach with chronological
regional authority (Umbrian Regional Hydrographic
assignment using the first 80% of the time series for

123
3968 Hydrobiologia (2022) 849:3963–3978

model training with each subsequent 10% used for correspondence between the two estimates, for exam-
validation (a tuning of the models hyper parameters) ple, both detecting a steep increase around the 12th of
and testing (independent data used to derive the August. The WISPstation estimate of the concentra-
model’s performance statistics) (https://cloud.google. tion of phycocyanin rapidly increased around this
com/automl-tables). For AI and NPMR analysis, the period indicating that cyanobacteria were likely
Chl-a concentration and lake surface water tempera- causing the increase in Chl-a concentration. Counts
ture were linear-interpolated to daily values using the of cyanobacteria taxa are carried out under govern-
temporal smoothing and gap filling (tsgf) linear ment regulations (Ministry of Health, 1998) and were
function in the R package greenbrown (Forkel & reported by the regional authority (ARPA, Umbria;
Wutzler, 2015; R Core Team, 2019). Theil-Sen esti- Charavgis et al., 2020) at approximately weekly
mates for slope of air temperature were carried out in intervals during bathing season. These show the
the openair package in R (Carslaw & Ropkins, 2012). underlying changes in composition and abundance of
Speakman rank (rS) correlation was calculated using cyanobacteria during this period (Fig. 3). The
DataDesk (Velleman, 1989). cyanobacteria community was dominated by Cylin-
For phenological analysis, calculation of the start of drospermopsis raciborskii during the bloom period of
season (SOS), end of season (EOS), and length of mid-to-late August with its increase and decline being
season (LOS), a spline interpolation was used as this broadly matched by estimates of Chl-a and
smoothed the seasonal patterns. Interpolation and phycocyanin.
smoothing is a standard and essential step in pheno- In order to understand the factors influencing the
logical analysis in order to identify the timing of dynamics of Chl-a in Lake Trasimeno, we first carried
seasonal change and this technique has been used to out a NPMR including the variables day of year
determine the changes in land cover phenology (DOY), year, lake level, lake surface water tempera-
(Forkel et al., 2013, 2015; Garcı́a et al., 2019). For ture, wind vectors, the NAO, and the sum of the
phenological analysis, we used 15 mg m-3 of Chl-a as antecedent rain for 7 days (Table 1). The best model
a level that could be considered as a starting point for had an xR2 of 0.62 and included DOY, year, lake level,
the summer bloom for Lake Trasimeno, and this and the NAO; however, the NAO was interchangeable
concentration has been used as a trigger value for with lake surface water temperature. The sensitivity
instigating enhanced bloom monitoring (Touchette value provides an indication of the importance of the
et al., 2007). We calculated the SOS as the day of year variables in the models. The DOY had the highest
when the Chl-a concentration first reached 15 mg m3 sensitivity value (0.338) followed by year (0.065),
and the EOS when it subsequently declined below lake level (0.014), and the NAO (0.004), while
15 mg m3. The LOS was the difference between the interchanging the latter with LSWT in a second model
two values. For comparison with LSWT, we followed yielded a sensitivity of 0.006 (Table 1). Contour plots
the same approach using 20°C to define SOS, EOS, were produced to visualize estimates for Chl-a for
and LOS. This temperature was selected as the optima DOY with lake level, NAO, and LSWT (Fig. 4). The
for many cyanobacteria have been considered above contour lines and color intensity represent Chl-a in
20°C (Konopka & Brock, 1978; Jöhnk et al., 2008). 4 mg m-3 increments. Common to all the plots are the
higher concentrations centered around day 250 (7th
September). Chl-a concentration was estimated to be
Results higher at more positive values of the NAO, at lower
lake levels and at warmer temperatures. The similar
The Chl-a concentration in lake Trasimeno ranged responses and tight intertwining of seasonal parame-
from \ 1 to 130 mg m-3 with an annual pattern ters obviously present a difficulty in either deciding on
dominated by a summer bloom (Fig. 2). Focusing on a key factors or partitioning influence. We also applied
summer period from July to September in 2019, we Google AI AutoML to develop a model for Chl-a as a
can see that the increase in Chl-a starts in July and designated feature. Overall the model also had an R2 of
peaks in August and September (Fig. 3). Comparing 0.62 and the parameters with the highest estimated
the CCI estimates of Chl-a with that of the high- percentage feature importance were date (83.3), the
frequency WISPstation revealed a close seven-day antecedent sum of rainfall (6.5), LSWT

123
Hydrobiologia (2022) 849:3963–3978 3969

–3

Fig. 2 Satellite-derived Chl-a in Trasimeno from 2003 to 2019. characteristics of MERIS and OLCI sensors. One high value
Significant data gap from 2012 to 2015 is due to the of 130 mg m-3 on the 16/09/2008 was removed for clarity
unavailability of optical satellite data with the same

(5.8), and NAO (3). While parameters estimated to quadratic equation (R2 = 0.92, P B 0.001, Fig. 6). It
have less importance included the v wind component appears that if the LOS for temperature is either too
(0.8), u wind component (0.6), and the lake level (0). short or too long, it may constrain the LOS of Chl-a
In order to examine the increase in temperature, we with an optimum of around 130 days (Fig. 6). The
needed a long time series and used the ERA5 data to Chl-a LOS was also related to the SOS for temperature
examine the increase in air temperature for Lake (rS = 0.72, P = 0.012) with years where 20°C was
Trasimeno. The Theil-Sen estimate for slope for reached earlier like in 2005 and 2007 having a shorter
annual average temperature between 1980 and 2019 Chl-a LOS (Fig. 5). A similar negative correlation
was 0.05 (P B 0.001) estimating an increase from (rS = - 0.811, P B 0.001) was found between LSWT
12.7 to 14.5°C. Similarly the increase in temperature LOS and average Chl-a concentration (calculated for a
for the summer period (June, July, August) had a slope 14-day window around DOY 250 when values are
of 0.06 (P B 0.001) increasing from 22 to 24°C. For typically at a maximum).
shallow Lake Trasimeno, the LSWT was highly In order to understand the implications of this, we
correlated with air temperature (R2 = 0.94) as previ- examined a longer time series, available for air
ously found (Ludovisi & Gaino, 2010). As the temperature, to see if there were any trends in the
temperature is increasing and its importance in summer temperature phenology that could have
influencing blooms is evident either directly or implications for lake phytoplankton (Fig. 7).
indirectly through the significance of the seasonal We found evidence for an increase in the LOS
component or via the NAO index, it deserves further (R2 = 0.28, P B 0.001), a decrease in SOS (R2 = 0.27,
analysis to determine if there is an influence on the P B 0.001) but not the EOS (R2 = 0.06, P B 0.12)
phenology of phytoplankton blooms. It had been with an estimated increase in 16 days LOS over the
postulated that perhaps blooms might occur earlier period. To answer the question as to why earlier and
with warmer temperatures or be of longer duration. longer occurrence of higher temperatures might lower
Using 15 mg m-3 to define the SOS for Chl-a, we the Chl-a LOS, we looked at the nutrient data from the
found that the timing was variable and ranged from local authority (https://apps.arpa.umbria.it/acqua/
DOY 196 (July 15, 2016) to 242 (August 30, 2007) qualita-acque-superficiali). Phosphate concentrations
(Fig. 5). Comparing with LSWT, the timing of the were occasionally below the level of detection in
SOS (when LSWT exceeded 20°C) ranged from DOY summer months; the concentrations for July for the
117 (April 27, 2005) to 155 (June 4,, 2004) (Fig. 5). last three years were 16, 18, and \ 10 lg l-1 PO4-P
The SOS for temperature was not significantly corre- for 2019, 2018, and 2017, respectively. However, both
lated with the SOS for Chl-a (rS, - 0.023, P = 0.95). the temporal resolution (bimonthly) and the spatial
However, the LOS of Chl-a appeared related to the coverage (two stations) prevented more detailed
LOS of temperature which could be described by a analysis. Another potential mechanism could be that

123
3970 Hydrobiologia (2022) 849:3963–3978

–3
–1

Fig. 3 (Top) estimates of Chl-a from the CCI project (2020). Note continuous and discontinuous x-axis on top and
(Chla_CCI) and WispStation estimated Chl-a (Chla_WISP) bottom figures, respectively. Counts of Microcystis sp.,
and phycocyanin (PC_WISP) in 2019. (Bottom) counts of Chrysosporum spp. and Aphanizomenon sp. were below 120
cyanobacteria cells per ml in 2019 reported by Charavgis et al. and are not visible on the graph

Table 1 Results of NPMR (nonparametric multiplicative regression) models for chlorophyll-a

xR2 Ave Variable Tol. Sen. Variable Tol. Sen. Variable Tol. Sen. Variable Tol. Sen. P
size 1 2 3 4

0.62 99.1 DOY 18.3 0.338 Year 1.2 0.065 Level 0.7 0.014 NAO 3.3 0.004 0.045
0.62 98.7 DOY 18.3 0.331 Year 1.2 0.064 Level 0.6 0.017 LSWT 13.0 0.006 0.045
First model (row 1) included NAO and second model (row 2) LSWT. The second model shows that NAO was interchangeable with
LSWT
xR2 cross-validated R2; Ave. size Average neighborhood size; Tol. tolerance; Sen. sensitivity; NAO North Atlantic Oscillation; Level
lake level; LSWT lake surface water temperature

123
Hydrobiologia (2022) 849:3963–3978 3971

Fig. 4 Nonparametric -0.2

multiplicative regression
-0.4
estimates for Chlorophyll-a
(mg m-3; contour lines) -0.6
against DOY (day of year) -0.8 8 8 12 16 20 16 12

Lake Level
for lake level (m), NAO
(North Atlantic Oscillation) -1.0
and lake surface water -1.2
temperature (°C)
-1.4
24
-1.6 8 8 12 16 20 16 12

-1.8
-2.0
0 50 100 150 200 250 300 350
DOY
3

2 12
8
1 8 12 16 20 16
24
NAO

-1
12
16
8 8 12 16 20
-2

-3
0 50 100 150 200 250 300 350
DOY

30
8 24
25
8
Temperature oC

20
12 16 20 16 12

15
24
10

8 16 20 16 12
5 8 12

0
0 50 100 150 200 250 300 350
DOY

the longer period of warmer temperature could be increasing temperature, for example, from 0.68 to
increasing the amount of calcium carbonate precipi- 0.76 mmol kg-1 from 15 to 25°C, an increase of 12%.
tation which could reduce resources through the
coprecipitation of phosphorus (Otsuki & Wetzel,
1972). Using chemistry data on major ions (Ludovisi Discussion
& Gaino, 2010), we followed the recommendations of
(Tang et al., 2021) in using the PHREEQC software to Examination of the satellite time series of Chl-a
calculate the calcium carbonate precipitation potential concentration showed that the annual pattern was
(CCPP) for an open system for various temperature dominated by a summer bloom that regularly occurred
increments (Fig. 8). Values of CCPP increased with over the years examined. There was good

123
3972 Hydrobiologia (2022) 849:3963–3978

Fig. 5 Start, end, and length of season for Chl-a (C 15 mg m3) and lake surface water temperature (LSWT) (C 20°C)

160 Google AI AutoML, and NPMR were used to

y = -0.714x2 + 183.63x - 11656
140 R² = 0.9204 examine the environmental parameters influencing
120 Chl-a concentration dynamics between 2003 and
Chlorophyll-a LOS

100
2011. Both modeling approaches found the time
component to be the most important accounting for
80
83.3% as feature importance in Google AI while in
60
NPMR, a sensitivity value was recorded of 0.338 and
40
0.065 for DOY and year, respectively. The seven-day
20 antecedent rainfall was the next most important for the
0 Google AI (6.5%) while this parameter was not
110 120 130 140 150
Temperature LOS
included in the NPMR. In NPRM, the lake water
level was the next most sensitive (0.014) followed by
Fig. 6 Chlorophyll-a length of season (LOS) ([ 15 mg m-3) either the NAO (0.004) or temperature (0.006). These
and temperature LOS ([ 20°C) with quadratic equation line fit variables were also included in the Google AI
AutoML ([ 3%) with the exception of lake level.
correspondence between the Chl-a estimated by the One of the drawbacks in the Google AI approach is the
WISPstation high-frequency monitoring and that of lack of interpretable diagnostic output to examine the
the lakes CCI project derived from Sentinel 3 OLCI behavior of individual parameters. For the NPMR, the
images. The increase, duration, and decline of the contour plots indicated higher concentrations centered
seasonal bloom were adequately delineated and were around day 250 (7th September). Chl-a concentration
also substantiated by the changes in abundance and was estimated to be higher at more positive values of
composition in cyanobacteria determined at weekly the NAO, at lower lake levels and at warmer
interval by the local authority (Charavgis et al., 2020). temperatures. It may be difficult to isolate the relative
Such rapid shifts in Chl-a concentration over a period importance of these parameters. The models already
of weeks have been found before (Bresciani et al., incorporate the seasonal and temporal components
2018) and underline the value of high-frequency and while the inclusion of other variables indicates
monitoring but also indicate that the data gathered by that they explain additional variation, a lot of the
the three optical sensors, including data obtained from variation may also be already explained by the
the lakes CCI project (Crétaux et al., 2020), perform seasonal component (also for other variables). How-
well in detecting dynamic changes underlain by a ever, the estimated response of Chl-a to higher
rapidly changing phytoplankton community. summer temperatures and a more positive NAO is

123
Hydrobiologia (2022) 849:3963–3978 3973

Fig. 7 Start, end, and length of season where average daily air temperature was C 20°C in lake Trasimeno from 1979 to 2019 (SOS,
EOS ,and LOS, respectively)

Criado-Aldeanueva & Soto-Navarro, 2020). The

incorporation of indices such as the NAO likely
reflects the importance of other parameters associated
with summer heatwaves such as low cloud cover,
reduced wind speed, as well as higher temperatures
that promote cyanobacteria growth directly or indi-
rectly through water column stability (Jöhnk et al.,
2008). The positive influence of lower lake levels may
reflect increased suspension of sediments by wave
action potentially increasing nutrient levels and pro-
moting bloom development (Søndergaard et al., 2003;
Ludovisi & Gaino, 2010). In addition, lower lake
levels can be associated with higher phytoplankton
biomass in shallow lakes through increasing light
Fig. 8 Increase in the calcium carbonate precipitation potential availability alone (Nõges et al., 2003). While the
(CCPP) with increasing temperature estimated for Lake seven-day antecedent sum of rainfall was not included
Trasimeno
in NPMR, its inclusion in Google AI may reflect the
importance of inflow events transporting nutrients to
not surprising. Cyanobacteria are well known to
the lake from the catchment. Some visual evidence of
respond to higher temperatures and a more positive
the importance of this in bloom development has been
NAO around this time period reflects high pressure
seen in satellite images in 2018 for Lake Trasimeno
with warm sunny weather (Konopka & Brock, 1978;

123
3974 Hydrobiologia (2022) 849:3963–3978

(Bresciani et al., 2020). However, significant storm increasingly more optimal climatic conditions (tem-
events and flushing can suppress blooms in shallow perature, water column stability, cloud cover) for
lakes (Havens et al., 2019). cyanobacteria is to be expected in lakes that are not
The examination of the phytoplankton phenology nutrient rich. This may cover the majority of moni-
found that there was no evidence for an earlier increase tored lakes in Europe as only 16% had phosphorus
in summer blooms that was related to the earlier onset conditions classified as less than good (although 36%
of higher temperatures. It was also anticipated that a were not classified) (Kristensen et al., 2018).
longer warmer period with temperatures above 20°C The potential for warmer temperatures to increase
would lead to a longer bloom period but the relation- the rate of calcite precipitation was also examined, and
ship was more complex. Contrary to expectations, an an increase from 15 to 25°C estimated an increase of
earlier onset and longer duration of warmer temper- 12% in the calcium carbonate precipitation potential
atures appeared to be related to a shorter phytoplank- (CCPP). This may lead to more coprecipitation of
ton bloom. The relationship could be described by a phosphorus leading to smaller blooms, especially
quadratic equation where initially Chl-a increased considering the longer period of warmer weather,
with a longer LOS for temperature but started to and also the fact that pH may be higher with increased
decline after 130 days. The relationship likely reflects photosynthesis (Otsuki & Wetzel, 1972; Homa &
aspects of resource availability. The decline after Chapra, 2011). It may be possible to examine the
130 days could be the result of a longer LOS, typically presence and extent of this process in future as
with an earlier SOS, leading to more or earlier uptake methods have recently been developed to use remote
of nutrients resulting in lower amounts to support later sensing to detect calcite spatially in lakes (Heine et al.,
bloom development or duration. A significant propor- 2017).
tion of the phytoplankton biomass was found to be A more systematic examination of the response of
comprised of chlorophytes in early summer (Havens Chl-a concentrations to climate change in summer is
et al., 2009) and these could play a role in reducing needed. While the increasingly more optimal condi-
resources for other succession species. While the tions for cyanobacteria has been predicted to lead to an
frequency of chemical monitoring did not allow increase in frequency and abundance of blooms, it
incorporation into analysis, phosphate concentrations may also be plausible that higher temperatures for
were occasionally below the limit of detection longer periods may alter the phenology of many
(\ 10 lg l-1 PO4-P) and monitoring was not carried limnetic organisms and catchment processes, altering
out for the month of August which may be the critical nutrient and other geochemical cycles and resulting in
period for bloom development. However, for at least unanticipated responses in summer phytoplankton,
some years, phosphate is likely to limit bloom including blooms of shorter duration. There may be a
development and overall, there is a close relationship tendency for publications to confirm the former
between total phosphorus and summer Chl-a anticipated response (Simundic, 2013). Given the
(R2 = 0.65) (Bresciani et al., 2021, in press). During increase in LOS observed using the ERA air temper-
the 2003 heatwave in Europe, examples were recorded ature data since 1980, climate change may continue to
of very large cyanobacteria blooms but this was under alter the phenology of the phytoplankton, even if
hypertrophic conditions (Jöhnk et al., 2008). Simi- larger or more prolonged blooms are not expected, this
larly, in shallow lake Taihu in China, high total can lead to asynchronization with other groups such as
phosphorus concentrations ([ 130 lg l-1) have led to zooplankton thereby altering energy transfer between
extensive blooms, despite restoration attempts (Qin trophic levels (Winder & Sommer, 2012). The lakes
et al., 2019, 2021). In addition, earlier onset of CCI database may be suited to exploring the incidence
Microcystis blooms has been found to be dependent on of such patterns as it contains information on lake
sufficient nutrient concentrations (Deng et al., 2014). level, temperature, Chl-a, turbidity, lake area, and ice
Furthermore, the fact that SOS can vary spatially cover and is due to expand to about 2000 lakes in 2021
within a lake underlines the importance of other local (https://climate.esa.int/en/projects/lakes/). One of the
factors such as nutrients in addition to climate limitations of the current dataset used here is the
warming (Shi et al., 2019). An interplay during presence of gaps—especially between 2012 and 2015.
summer between nutrient availability and the One potential solution would be to use data from the

123
Hydrobiologia (2022) 849:3963–3978 3975

MODIS satellite which has been used successfully to Data availability All CCI and ERA data are available online
define bloom phenology previously in Lake Taihu in as cited.
China (Shi et al., 2017, 2019). This solution is cur-
Code availability Not applicable.
rently being evaluated for incorporation into later
versions of the lakes CCI database, but would have to Declarations
be evaluated for lake Trasimeno. Lake Trasimeno may
have provided us with a ‘tidy’ example in that there is Conflict of interest The authors declare that they have no
conflict of interest.
essentially one well-timed and well-defined Chl-a
peak per year and innovative ways may be needed for Ethical approval Not applicable.
more widespread application of this approach to lakes
with more complex phytoplankton dynamics. Consent to participate All authors have agreed to participate.

Consent for publication All authors have read and agreed to

this version of the manuscript.
Conclusions
Open Access This article is licensed under a Creative Com-
mons Attribution 4.0 International License, which permits use,
Lake Trasimeno has a well-defined summer bloom of
sharing, adaptation, distribution and reproduction in any med-
phytoplankton and its seasonal timing could be ium or format, as long as you give appropriate credit to the
described using satellite data dating back to 2003. original author(s) and the source, provide a link to the Creative
Larger summer blooms were associated with warmer Commons licence, and indicate if changes were made. The
images or other third party material in this article are included in
temperatures, a lower lake level and a more positive
the article’s Creative Commons licence, unless indicated
NAO, indicative of high pressure and warm sunny otherwise in a credit line to the material. If material is not
calm weather. The relative role of these parameters included in the article’s Creative Commons licence and your
and other factors in influencing Chl-a is difficult to intended use is not permitted by statutory regulation or exceeds
the permitted use, you will need to obtain permission directly
apportion because they are correlated seasonally.
from the copyright holder. To view a copy of this licence, visit
Interestingly, blooms were not occurring earlier but http://creativecommons.org/licenses/by/4.0/.
the length of season where water temperatures were
above 20°C appeared to play a role in controlling the
duration of blooms. A longer warmer season, typically References
commencing earlier in the year, was found to lead to a
shorter duration of blooms likely owing to seasonal Bresciani, M., I. Cazzaniga, M. Austoni, T. Sforzi, F. Buzzi, G.
nutrient limitation and perhaps increased calcite Morabito & C. Giardino, 2018. Mapping phytoplankton
blooms in deep subalpine lakes from Sentinel-2A and
precipitation. As the duration of warmer temperatures Landsat-8. Hydrobiologia 824: 197–214.
is increasing, this is likely to be more important in Bresciani, M., G. Free, M. Pinardi, M. Laanen, R. Padula, A.
future with implications for species phenology and Fabbretto, S. Mangano, & C. Giardino, 2021. Optical
lake ecosystem functioning. remote sensing in shallow lake Trasimeno: understanding
from applications across diverse temporal, spectral and
spatial scales Springer ‘‘Water’’ book series.
Acknowledgements We would like to thank the
Bresciani, M., M. Pinardi, G. Free, G. Luciani, S. Ghebrehiwot,
‘‘Cooperativa dei Pescatori del Trasimeno’’ for the support in
M. Laanen, S. Peters, V. Della Bella, R. Padula & C.
field campaigns. Many thanks to Luca Nicoletti and Luca Galli
Giardino, 2020. The use of multisource optical sensors to
from Polvese Center ‘‘Climate Change and Biodiversity in
study phytoplankton spatio-temporal variation in a shallow
Lakes and Wetlands’’ from ARPA Umbria for collecting water
turbid lake. Water 12: 284.
samples. We also thank the Province of Perugia for the
Carslaw, D. C. & K. Ropkins, 2012. Openair—an R package for
availability to use the platform in Polvese Island. We would
air quality data analysis. Environmental Modelling &
like to acknowledge the two anonymous reviewers for valuable
Software 27: 52–61.
comments and suggestions that helped improve this study.
Carvalho, L., E. B. Mackay, A. C. Cardoso, A. Baattrup-Ped-
ersen, S. Birk, K. L. Blackstock, G. Borics, A. Borja, C.
Funding This research is part of the ESA CCI LAKES project K. Feld & M. T. Ferreira, 2019. Protecting and restoring
(Contract No. 40000125030/18/I-NB) and of the H2020 projects Europe’s waters: an analysis of the future development
Water-ForCE (GA No. 101004186) and EOMORES (GA No. needs of the Water Framework Directive. Science of the
730066). Total Environment 658: 1228–1238.

123
3976 Hydrobiologia (2022) 849:3963–3978

Charavgis, F., A. Cingolani, M. Di Brizio, E. Rinaldi, G. Tozzi, water control on global interannual variability and trends in
& P. Stranieri, 2020. Qualita’ delle acque di balneazione land surface phenology and greenness. Global Change
dei laghi Umbri, stagione balneare 2019. ARPA, Umbria. Biology 21: 3414–3435.
Cingolani, A., & F. Charavgis, 2017. Valutazione dello stato Forkel, M., & T. Wutzler, 2015. greenbrown - land surface
ecologico e chimico dei corpi idrici lacustri (2013–2015). phenology and trend analysis. A package for the R soft-
ARPA, Umbria, Italy: 25. ware. Max Planck Institute, Germany, http://greenbrown.r-
Cingolani, A., & F. Charavgis, 2018. Valutazione dello stato forge.r-project.org/.
ecologico e chimico dei corpi idrici lacustri (2015–2017). Free, G., M. Bresciani, M. Pinardi, N. Ghirardi, G. Luciani, R.
ARPA, Umbria, Italy: 26. Caroni & C. Giardino, 2021a. Detecting climate driven
Codd, G. A., L. F. Morrison & J. S. Metcalf, 2005. Cyanobac- changes in chlorophyll-a in deep subalpine lakes using long
terial toxins: risk management for health protection. Tox- term satellite data. Water 13: 866.
icology and Applied Pharmacology 203: 264–272. Free, G., M. Bresciani, M. Pinardi, C. Giardino, K. Alikas, K.
Council of the European Communities, 2000. Directive Kangro, E.-I. Rõõm, D. Vaičiūt_e, M. Bučas, E. Tiškus, A.
2000/60/EC of the European Parliament and of the Council Hommersom, M. Laanen, & S. Peters, 2021b. Detecting
of 23 October 2000 establishing a framework for com- climate driven changes in chlorophyll-a using high fre-
munity action in the field of water policy. L327: 72. quency monitoring: the impact of the 2019 European
Council of the European Communities, 2013. Commission Heatwave in Three Contrasting Aquatic Systems. Sensors
Decision of 20 September 2013 establishing pursuant to 21
Directive 2000/60/EC of the European Parliament and of Free, G., M. Bresciani, W. Trodd, D. Tierney, S. O’Boyle, C.
the Council, the values of the Member State monitoring Plant & J. Deakin, 2020. Estimation of lake ecological
system classifications as a result of the intercalibration quality from Sentinel-2 remote sensing imagery. Hydro-
exercise and repealing Decision 2008/915/EC. Official biologia 847: 1423–1438.
Journal of the European Communities 480: 1–47. Frondini, F., W. Dragoni, N. Morgantini, M. Donnini, C. Car-
Crétaux, J.-F., C. J. Merchant, C. Duguay, S. Simis, B. Cal- dellini, S. Caliro, M. Melillo & G. Chiodini, 2019. An
mettes, M. Bergé-Nguyen, Y. Wu, D. Zhang, L. Carrea, X. endorheic lake in a changing climate: geochemical inves-
Liu, N. Selmes, & M. Warren, 2020. ESA Lakes Climate tigations at lake Trasimeno (Italy). Water Multidisciplinary
Change Initiative (Lakes_cci): Lake products, Version 1.0. Digital Publishing Institute 11: 1319.
Centre for Environmental Data Analysis (CEDA), https:// Garcı́a, M. A., H. Moutahir, G. M. Casady, S. Bautista & F.
catalogue.ceda.ac.uk/uuid/ Rodrı́guez, 2019. Using hidden Markov models for land
3c324bb4ee394d0d876fe2e1db217378. surface phenology: an evaluation across a range of land
Criado-Aldeanueva, F. & J. Soto-Navarro, 2020. Climatic cover types in southeast Spain. Remote Sensing 11: 507.
indices over the Mediterranean sea: a review. Applied Giardino, C., M. Bresciani, P. Villa & A. Martinelli, 2010.
Sciences Multidisciplinary Digital Publishing Institute 10: Application of remote sensing in water resource manage-
5790. ment: the case study of lake Trasimeno, Italy. Water
Deng, J., B. Qin, H. W. Paerl, Y. Zhang, J. Ma & Y. Chen, 2014. Resources Management 24: 3885–3899.
Earlier and warmer springs increase cyanobacterial (Mi- Gons, H. J., 1999. Optical teledetection of chlorophyll a in
crocystis spp.) blooms in subtropical Lake Taihu, China. turbid inland waters. Environmental Science & Technol-
Freshwater Biology 59: 1076–1085. ogy 33: 1127–1132.
DM, 2010. Ambiente 8 Novembre 2010, N.260. ‘‘Regolamento Havens, K. E., A. C. Elia, M. I. Taticchi & R. S. Fulton, 2009.
recante i criteri tecnici per la classificazione dello stato dei Zooplankton–phytoplankton relationships in shallow sub-
corpi idrici superficiali, per la modifica delle norme tec- tropical versus temperate lakes Apopka (Florida, USA) and
niche del decreto legislativo 3 aprile 2006, n. 152, recante Trasimeno (Umbria, Italy). Hydrobiologia 628: 165–175.
norme in materia ambientale, predisposto ai sensi dell’ar- Havens, K. E., G. Ji, J. R. Beaver, R. S. Fulton & C. E. Teacher,
ticolo 75, comma 3, del medesimo decreto legislativo’’. 2019. Dynamics of cyanobacteria blooms are linked to the
S.O. n. 31 alla GU 7 febbraio 2011 n. 30. hydrology of shallow Florida lakes and provide insight into
Dragoni, W., 2004. Il Lago Trasimeno e le variazioni cli- possible impacts of climate change. Hydrobiologia 829:
matiche. Provincia di Perugia, Perugia, Italy. 43–59.
Ellis, C. J., B. J. Coppins, T. P. Dawson & M. R. Seaward, 2007. Heine, I., A. Brauer, B. Heim, S. Itzerott, P. Kasprzak, U. Kienel
Response of British lichens to climate change scenarios: & B. Kleinschmit, 2017. Monitoring of calcite precipita-
trends and uncertainties in the projected impact for con- tion in hardwater lakes with multi-spectral remote sensing
trasting biogeographic groups. Biological Conservation archives. Water 9: 15.
140: 217–235. Ho, J. C., A. M. Michalak & N. Pahlevan, 2019. Widespread
Ezra, A. G. & D. I. Nwankwo, 2001. Composition of phyto- global increase in intense lake phytoplankton blooms since
plankton algae in Gubi reservoir, Bauchi, Nigeria. Journal the 1980s. Nature 574: 667–670.
of Aquatic Sciences 16: 115–118. Homa, E. S. & S. C. Chapra, 2011. Modeling the impacts of
Forkel, M., N. Carvalhais, J. Verbesselt, M. D. Mahecha, C. calcite precipitation on the epilimnion of an ultraolig-
S. R. Neigh & M. Reichstein, 2013. Trend change detection otrophic, hard-water lake. Ecological Modelling 222:
in NDVI time series: effects of inter-annual variability and 76–90.
methodology. Remote Sensing 5: 2113–2144. Huang, C., Y. Zhang, T. Huang, H. Yang, Y. Li, Z. Zhang, M.
Forkel, M., M. Migliavacca, K. Thonicke, M. Reichstein, S. He, Z. Hu, T. Song & A. Zhu, 2019. Long-term variation of
Schaphoff, U. Weber & N. Carvalhais, 2015. Codominant phytoplankton biomass and physiology in Taihu lake as

123
Hydrobiologia (2022) 849:3963–3978 3977

observed via MODIS satellite. Water Research 153: Organisation for Economic Co-operation and Develop-
187–199. ment, Paris.
Huisman, J., G. A. Codd, H. W. Paerl, B. W. Ibelings, J. M. H. O’Neil, J. M., T. W. Davis, M. A. Burford & C. J. Gobler, 2012.
Verspagen & P. M. Visser, 2018. Cyanobacterial blooms. The rise of harmful cyanobacteria blooms: the potential
Nature Reviews Microbiology 16: 471–483. roles of eutrophication and climate change. Harmful Algae
Janssen, A. B. G., S. Hilt, S. Kosten, J. J. M. de Klein, H. 14: 313–334.
W. Paerl & D. B. V. de Waal, 2021. Shifting states, shifting O’Reilly, C. M., S. Sharma, D. K. Gray, S. E. Hampton, J.
services: linking regime shifts to changes in ecosystem S. Read, R. J. Rowley, P. Schneider, J. D. Lenters, P.
services of shallow lakes. Freshwater Biology 66: 1–12. B. McIntyre, B. M. Kraemer, G. A. Weyhenmeyer, D.
Jöhnk, K. D., J. Huisman, J. Sharples, B. Sommeijer, P. Straile, B. Dong, R. Adrian, M. G. Allan, O. Anneville, L.
M. Visser & J. M. Stroom, 2008. Summer heatwaves Arvola, J. Austin, J. L. Bailey, J. S. Baron, J. D. Brookes, E.
promote blooms of harmful cyanobacteria. Global Change de Eyto, M. T. Dokulil, D. P. Hamilton, K. Havens, A.
Biology 14: 495–512. L. Hetherington, S. N. Higgins, S. Hook, L. R. Izmesteva,
Jørgensen, S. E., S. Bastianoni, F. Müller, B. C. Patten, B. K. D. Joehnk, K. Kangur, P. Kasprzak, M. Kumagai, E.
D. Fath, J. C. Marques, S. N. Nielsen, E. Tiezzi & R. Kuusisto, G. Leshkevich, D. M. Livingstone, S. MacIntyre,
E. Ulanowicz, 2007. 9 - Ecosystem principles have appli- L. May, J. M. Melack, D. C. Mueller-Navarra, M. Nau-
cations. In Jørgensen, S. E., S. Bastianoni, F. Müller, B. menko, P. Noges, T. Noges, R. P. North, P.-D. Plisnier, A.
C. Patten, B. D. Fath, J. C. Marques, S. N. Nielsen, E. Rigosi, A. Rimmer, M. Rogora, L. G. Rudstam, J.
Tiezzi & R. E. Ulanowicz (eds), A New Ecology Elsevier, A. Rusak, N. Salmaso, N. R. Samal, D. E. Schindler, S.
Amsterdam: 199–241. G. Schladow, M. Schmid, S. R. Schmidt, E. Silow, M.
Konopka, A. & T. D. Brock, 1978. Effect of temperature on E. Soylu, K. Teubner, P. Verburg, A. Voutilainen, A.
blue-green algae (cyanobacteria) in lake Mendota. Applied Watkinson, C. E. Williamson & G. Zhang, 2015. Rapid and
and Environmental Microbiology 36: 572. highly variable warming of lake surface waters around the
Kristensen, P., C. Whalley, F. N. N. Zal, & T. Christiansen, globe. Geophysical Research Letters 42: 10773–10781.
2018. European waters assessment of status and pressures Otsuki, A. & R. G. Wetzel, 1972. Coprecipitation of phosphate
2018. EEA Report European Environment Agency. with carbonates in a Marl lake. Limnology and Oceanog-
Ludovisi, A. & E. Gaino, 2010. Meteorological and water raphy 17: 763–767.
quality changes in Lake Trasimeno (Umbria, Italy) during Paerl, H. W. & J. Huisman, 2009. Climate change: a catalyst for
the last fifty years. Journal of Limnology 69: 174–188. global expansion of harmful cyanobacterial blooms.
Ma, X., Y. Wang, S. Feng & S. Wang, 2015. Vertical migration Environmental Microbiology Reports 1: 27–37.
patterns of different phytoplankton species during a sum- Palmer, S. C. J., P. D. Hunter, T. Lankester, S. Hubbard, E.
mer bloom in Dianchi Lake, China. Environmental Earth Spyrakos, A. N. Tyler, M. Présing, H. Horváth, A. Lamb,
Sciences 74: 3805–3814. H. Balzter & V. R. Tóth, 2015. Validation of Envisat
Maeda, E. E., F. Lisboa, L. Kaikkonen, K. Kallio, S. Koponen, MERIS algorithms for chlorophyll retrieval in a large,
V. Brotas & S. Kuikka, 2019. Temporal patterns of phy- turbid and optically-complex shallow lake. Remote Sens-
toplankton phenology across high latitude lakes unveiled ing of Environment 157: 158–169.
by long-term time series of satellite data. Remote Sensing Peters, S., M. Laanen, P. Groetsch, S. Ghezehegn, K. Poser, A.
of Environment 221: 609–620. Hommersom, E. De Reus, & L. Spaias, 2018. WISPstation:
McCune, B., 2006. Nonparametric Multiplicative Regression a new autonomous above water radiometer system. Ocean
for Habitat Modeling, Oregon State University, Oregon: Optics XXIV conference. Dubrovnik, Croatia. Available
McCune, B., & M. J. Mefford, 2009. HyperNiche. Nonpara- online: https://doi.org/10.5281/zenodo.2533079. Accessed
metric Multiplicative Habitat Modeling. MjM Software, 1 July 2020.
Oregon, U.S.A. Phillips, G., A. Lyche-Solheim, B. Skjelbred, U. Mischke, S.
Ministry of Health, 1998. Circolare del ministero della sanità del Drakare, G. Free, M. Järvinen, C. de Hoyos, G. Morabito,
09 aprile 1998, N. 400.4/13.1/3/562. Aggiornamento delle S. Poikane & L. Carvalho, 2013. A phytoplankton trophic
metodiche analitiche per la determinazione dei parametri index to assess the status of lakes for the Water Framework
previsti nel decreto interministeriale 17 giugno 1988 con- Directive. Hydrobiologia 704: 75–95.
cernente i criteri per la definizione del programma di Qin, B., J. Deng, K. Shi, J. Wang, J. Brookes, J. Zhou, Y. Zhang,
sorveglianza di cui all’art. 1 del D.L.14 maggio 1988 n. 155 G. Zhu, H. W. Paerl & L. Wu, 2021. Extreme climate
convertito con legge del 15 luglio 1988 n. 271, 9 pp. anomalies enhancing cyanobacterial blooms in eutrophic
Nicolaou, N. & T. G. Constandinou, 2016. A nonlinear causality lake Taihu, China. Water Resources Research 57:
estimator based on non-parametric multiplicative regres- e2020WR029371.
sion. Frontiers in Neuroinformatics 10: 19. Qin, B., H. W. Paerl, J. D. Brookes, J. Liu, E. Jeppesen, G. Zhu,
Nõges, T., P. Nõges & R. Laugaste, 2003. Water level as the Y. Zhang, H. Xu, K. Shi & J. Deng, 2019. Why Lake Taihu
mediator between climate change and phytoplankton continues to be plagued with cyanobacterial blooms
composition in a large shallow temperate lake. Hydrobi- through 10 years (2007–2017) efforts. Science Bulletin 64:
ologia 506: 257–263. 1.
OECD, 1982. Eutrophication of waters. Monitoring, assessment R Core Team, 2019. R: A Language and Environment for Sta-
and control. Final report, OECD cooperative programme tistical Computing. R Foundation for Statistical Comput-
on monitoring of inland waters (eutrophication control). ing, Vienna, Austria, https://www.R-project.org/.

123
3978 Hydrobiologia (2022) 849:3963–3978

Raymond, J. E. G., 1983. Plankton and Productivity in the Climate Change Initiative (Lakes_cci): Product User
Oceans Zooplankton, Pergamon Press, Oxford: Guide. European Space Agency, https://climate.esa.int/
Reynolds, C. S., 2006. Ecology of Phytoplankton, Cambridge documents/360/CCI-LAKES-0029-PUG_v1.1_signed_
University Press, Cambridge: CA.pdf.
Rogora, M., F. Buzzi, C. Dresti, B. Leoni, F. Lepori, R. Mosello, Simundic, A.-M., 2013. Bias in research. Biochemia Medica
M. Patelli & N. Salmaso, 2018. Climatic effects on vertical Medicinska Naklada 23: 12–15.
mixing and deep-water oxygen content in the subalpine Søndergaard, M., J. P. Jensen & E. Jeppesen, 2003. Role of
lakes in Italy. Hydrobiologia 824: 33–50. sediment and internal loading of phosphorus in shallow
Sala, O. E., F. Stuart Chapin, J. J. Armesto, E. Berlow, J. lakes. Hydrobiologia 506: 135–145.
Bloomfield, R. Dirzo, E. Huber-Sanwald, L. F. Huenneke, Tang, C., B. Godskesen, H. Aktor, M. van Rijn, J. B. Kristensen,
R. B. Jackson, A. Kinzig, R. Leemans, D. M. Lodge, H. P. S. Rosshaug, H.-J. Albrechtsen & M. Rygaard, 2021.
A. Mooney, M. Oesterheld, N. L. Poff, M. T. Sykes, B. Procedure for calculating the calcium carbonate precipi-
H. Walker, M. Walker & D. H. Wall, 2000. Global biodi- tation potential (CCPP) in drinking water supply: impor-
versity scenarios for the year 2100. Science 287: 1770. tance of temperature. Ionic Species and Open/closed
Salmaso, N., A. Boscaini, C. Capelli, & L. Cerasino, 2018. System. Water 13: 42.
Ongoing ecological shifts in a large lake are driven by Touchette, B. W., J. M. Burkholder, E. H. Allen, J. L. Alexan-
climate change and eutrophication: evidences from a three- der, C. A. Kinder, C. Brownie, J. James & C. H. Britton,
decade study in Lake Garda. Hydrobiologia 824: 177–195. 2007. Eutrophication and cyanobacteria blooms in run-of-
Scheffer, M., 1989. Alternative stable states in eutrophic, shal- river impoundments in North Carolina, U.S.A. Lake and
low freshwater systems: a minimal model. Hydrobiological Reservoir Management 23: 179–192.
Bulletin 23: 73–83. Tyler, A. N., P. D. Hunter, E. Spyrakos, S. Groom, A. M. Con-
Scheffer, M. & E. H. van Nes, 2007. Shallow lakes theory stantinescu & J. Kitchen, 2016. Developments in Earth
revisited: various alternative regimes driven by climate, observation for the assessment and monitoring of inland,
nutrients, depth and lake size. In Gulati, R. D., E. Lam- transitional, coastal and shelf-sea waters. Science of the
mens, N. De Pauw & E. Van Donk (eds), Shallow Lakes in Total Environment 572: 1307–1321.
a Changing World Springer, Dordrecht: 455–466. Velleman, P. F., 1989. Data Desk: Handbook, Volume 1 (1),
Shi, K., Y. Zhang, Y. Zhang, N. Li, B. Qin, G. Zhu & Y. Zhou, Data Description, Inc., New York:
2019. Phenology of phytoplankton blooms in a trophic lake Winder, M. & U. Sommer, 2012. Phytoplankton response to a
observed from long-term MODIS data. Environmental changing climate. Hydrobiologia 698: 5–16.
Science & Technology American Chemical Society 53: Woolway, R. I., E. Jennings, T. Shatwell, M. Golub, D.
2324–2331. C. Pierson & S. C. Maberly, 2021. Lake heatwaves under
Shi, K., Y. Zhang, Y. Zhou, X. Liu, G. Zhu, B. Qin & G. Gao, climate change. Nature 589: 402–407.
2017. Long-term MODIS observations of cyanobacterial Yost, A. C., 2008. Probabilistic modeling and mapping of plant
dynamics in Lake Taihu: responses to nutrient enrichment indicator species in a Northeast Oregon industrial forest,
and meteorological factors. Scientific Reports 7: 40326. USA. Ecological Indicators 8: 46–56.
Simis, S. G. H., 2006. Blue-green catastrophe: remote sensing of
mass viral lysis of cyanobacteria. Ph.D. Thesis, Vrije
Publisher’s Note Springer Nature remains neutral with
University.
regard to jurisdictional claims in published maps and
Simis, S., N. Selmes, B. Calmettes, C. Duguay, C. J. Merchant,
institutional affiliations.
E. Malnes, H. Yésou, & P. Blanco, 2020. ESA Lakes

123