Doktoritoo Pille Meinson.Indd

PILLE MEINSON

VIIS VIIMAST KAITSMIST

SANDRA METSLAID ASSESSMENT OF CLIMATE EFFECTS ON SCOTS PINE (PINUS SYLVESTRIS L.) GROWTH IN ESTONIA KLIIMA MÕJU HINDAMINE HARILIKU MÄNNI (PINUS SYLVESTRIS L.) KASVULE EESTIS Professor Andres Kiviste, dotsent Ahto Kangur 6. juuni 2017 HIGH-FREQUENCY MEASUREMENTS –ANEW APPROACH INLIMNOLOGY NELE NUTT THE PERCEIVABLE LANDSCAPE A THEORETICAL-METHODOLOGICAL APPROACH TO LANDSCAPE TAJUTAV MAASTIK TEOREETILIS-METODOLOOGILINE KÄSITLUS MAASTIKUST Dr. Juhan Maiste (Tartu Ülikool), professor Zenia Kotval (Michigan State University, USA; Tallinna Tehnikaülikool) 15. juuni 2017 HIGH-FREQUENCY MEASUREMENTS – A NEW APPROACH IN LIMNOLOGY TIIA DRENKHAN INTERACTION BETWEEN LARGE PINE WEEVIL (HYLOBIUS ABIETIS L.) PATHOGENIC AND SAPROTROPHIC FUNGI AND VIRUSES HARILIKU MÄNNIKÄRSAKA (HYLOBIUS ABIETIS L.) SEOS PATOGEENSETE PIDEVMÕÕTMISED – JA SAPROTROOFSETE SEENTE NING VIIRUSTEGA Dotsent Kaljo Voolma, dotsent Ivar Sibul, UUS LÄHENEMINE LIMNOLOOGIAS Risto Aarne Olavi Kasanen (Helsingi Ülikool, Soome) 19. juuni 2017

SEYED MAHYAR MIRMAJLESSI ASSESSMENT OF VERTICILLIUM DAHLIAE KLEB. AND SOIL FUNGAL COMMUNITIES ASSOCIATED WITH STRAWBERRY FIELDS MULLA PATOGEENI VERTICILLIUM DAHLIAE KLEB. JA MULLA PILLE MEINSON SEENEKOOSLUSTE ISELOOMUSTAMINE MAASIKAPÕLDUDEL Vanemteadur Evelin Loit, professor Marika Mänd 20. juuni 2017

RAIVO KALLE A Thesis CHANGE IN ESTONIAN NATURAL RESOURCE USE: for applying for the degree of Doctor of Philosophy in Hydrobiology THE CASE OF WILD FOOD PLANTS EESTI LOODUSLIKE RESSURSSIDE KASUTAMISE MUUTUS: LOODUSLIKE TOIDUTAIMEDE NÄITEL Professor Tiiu Kull, Dr. Renata Sõukand, Dr. Rajindra K Puri (University of Kent, UK) Väitekiri 5. september 2017 ﬁ losooﬁ adoktori kraadi taotlemiseks hüdrobioloogia erialal

ISSN 2382-7076 ISBN 978-9949-569-87-8 (trükis) ISBN 978-9949-569-88-5 (pdf) Tartu 2017 Trükitud taastoodetud paberile looduslike trükivärvidega © Kuma Print

Eesti Maaülikooli doktoritööd

Doctoral Theses of the Estonian University of Life Sciences

HIGH-FREQUENCYHIGH-FREQUENCY MEASUREMENTS MEASUREMENTS – – A NEWA NEW APPROACH APPROACH IN IN LIMNOLOGY LIMNOLOGY

PIDEVMÕÕTMISED – PIDEVMÕÕTMISED – UUS LÄHENEMINE UUS LÄHENEMINE LIMNOLOOGIAS LIMNOLOOGIAS

PILLE MEINSON

A thesis for applying for the degree of Doctor of Philosophy in Hydrobiology

Väitekiri filosoofiadoktori kraadi taotlemiseks hüdrobioloogia erialal

Tartu 2017

Institute of Agricultural and Environmental Sciences Estonian University of Life Sciences

According to verdict No 6-14/6-9 Doctoral Committee for Agricultural and Environmental Sciences of the Estonian University of Life Sciences has accepted this thesis for the defence of the degree of Doctor of Philosophy in Hydrobiology.

Opponent: Rafael Marcé, PhD Catalan Institute for Water Research (ICRA)

Supervisors: Peeter Nõges, PhD Institute of Agricultural and Environmental Sciences Estonian University of Life Sciences

Alo Laas, PhD Institute of Agricultural and Environmental Sciences Estonian University of Life Sciences

Defence of the thesis: Estonian University of Life Sciences, Festive Hall, Kreutzwaldi 1a, Tartu on September 5, 2017 at 14:00 pm.

The English and Estonian was edited by Tiina Nõges

Publication of this thesis is supported by the Estonian University of Life Sciences. This research was supported by European Social Fund’s Doctoral Studies and Internationalisation Programme DoRa and by institutional research funding IUT 21-02, PUT777, ETF8729, ETF8486 of the Estonian Ministry of Education and Research.

„Everybody comes to a point in their life when they want to quit. But it’s what you do at that moment that determines who you are“ (David Goggins)

This thesis is dedicated to my mom and grandparents with love

CONTENTS

LIST OF ORIGINAL PUBLICATIONS ...... 9 ABBREVIATIONS ...... 11 1. INTRODUCTION ...... 12 2. REVIEW OF THE LITERATURE ...... 15 2.1. Patterns in applying high-frequency measurements in limnology ...... 15 2.2. HFM versus conventional measurements ...... 16 2.3. Most used water parameters in high-frequency studies ...... 17 2.4. Th ermal stratifi cation of lakes ...... 18 2.5. Climate change eff ects on lakes ...... 19 2.6. Dissolved gases in lakes ...... 20 2.6.1. Dissolved oxygen ...... 20 2.6.2. Carbon dioxide ...... 21 3. HYPOTHESES AND AIMS OF THE STUDY ...... 23 4. MATERIAL AND METHODS ...... 25 4.1. Study sites ...... 25 4.2. Data collection ...... 25 4.3. Modelling lake temperature ...... 26 4.4. Statistical analyses ...... 26 5. RESULTS ...... 27 5.1. New insights into using continuous and HFM in limnology ...... 27 5.2. Has stratifi cation regime changed in Lake Võrtsjärv?...... 29 5.3. DO and CO2 regimes in 8 lake types in Estonia ...... 32 6. DISCUSSION ...... 34 CONCLUSIONS ...... 39

7 REFERENCES ...... 40 SUMMARY IN ESTONIAN ...... 54 ACKNOWLEDGEMENTS ...... 57 ORIGINAL PUBLICATIONS ...... 58 CURRICULUM VITAE ...... 105 LIST OF PUBLICATIONS ...... 107

8 LIST OF ORIGINAL PUBLICATIONS

This thesis is based on the following papers, which are referred to by their Roman numerals. The papers are reproduced by the kind permission of the publishers.

I Meinson, Pille; Idrizaj, Agron; Nõges, Peeter; Nõges, Tiina; Laas, Alo (2016). Continuous and high-frequency measurements in limnology: history, applications, and future challenges. Environmental Reviews 24, 52-62.

II Woolway, R. Iestyn; Meinson, Pille; Nõges, Peeter; Jones, Ian D.; Laas, Alo (2017). Atmospheric stilling leads to prolonged thermal stratification in a large shallow polymictic lake. Climatic Change 141, 759-773.

III Laas, Alo; Cremona, Fabien; Meinson, Pille; Rõõm, Eva- Ingrid; Nõges, Tiina; Nõges, Peeter (2016). Summer depth distribution profiles of dissolved CO2 and O2 in shallow temperate lakes reveal trophic state and lake type specific differences. Science of the Total Envrionment 566-567, 63-75.

9 CONTRIBUTIONS: I II III

Idea PM, PN PN, AL, PM AL, PM

Data collection PM, AI, PM, AL, PN AL, EIR, PM

Data analyses PM, AI RIW, IDJ, AL AL, FC, PM

Manuscript PM, AI, PN, RIW, PM, PN, AL, FC, PM, preparation TN, AL. IDJ, AL EIR, TN, TPPN

PM – Pille Meinson; AL – Alo Laas; PN – Peeter Nõges; TN – Tiina Nõges; FC – Fabien Cremona; RIW – R. Iestyn Woolway; IDJ – Ian D. Jones; EIR – Eva-Ingrid Rõõm; AI – Agron Idrizaj

10 ABBREVIATIONS

14C carbon 14 A1B a balanced emphasis on all energy sources ADCP Acoustic Doppler Current Profiler Alk alkalitrophic lakes CDOM coloured dissolved organic matter Chl a pigment chlorophyll a Coastal coastal lakes

CO2 carbon dioxide DO dissolved oxygen DOC dissolved organic carbon GLEON Global Lake Ecological Observatory Network HF high-frequency HFM high-frequency measurements Large Lake Võrtsjärv (as lake type) LightSoft light-coloured soft-water lakes MedAlk non-stratified, medium alkalinity lakes NETLAKE Networking Lake Observatories in Europe

NO3 nitrate PAR photosynthetically active radiation

PO4 phosphate SONAR hydrolocation using SOund Navigation And Ranging SRES special report on emissions scenarios StratMedAlk stratified, medium alkalinity.lakes V-Large Lake Peipsi (as lake type) 11 1. INTRODUCTION

Until recently, the majority of standard lake monitoring programmes were based on manual in situ measurements that can be time consuming and costly to procure and often lack both the needed spatial coverage as well as a suitable sampling frequency (Vos et al., 2003). The need for more and more detailed and high-quality monitoring data has contributed to sensor development. The main motivators of sensor development are intelligence, military and space exploration. From there all new developments infiltrate into civil science. Another field supporting the development of new and better sensors is water treatment which needs them to automatize the monitoring of both wastewaters and especially drinking waters (Ingildsen, 2002; Storey et al., 2011). In most cases, the concentrations of measured substances in wastewater are much higher than in natural waters (including drinking water), so these sensors do not have to be as accurate as those used in natural waters, therefore direct usage of the same sensors for both purposes is limited. One field requiring high sensor accuracy and stability is the detection of weak and noisy climate change signals in thermal, gas, and nutrient regimes of water bodies and distinguishing them from anthropogenic variability. Unfortunately, classical sciences are the last ones to motivate sensor development and usually have to rely on developments for other more applied fields, if these are suitable. To give a flavour of the main steps that several environmental analytical measurements have gone through, the advances in dissolved oxygen (DO) measurements are shortly described below.

Dissolved oxygen is one very important parameter measured in waterbodies as its quantity and changes in the environment reflect several biological and chemical conditions and processes in waterbodies. Over most of the 19th century, the test for determining DO involved boiling the water sample and collecting, over mercury, the gases released, for subsequent analysis (Wanklyn, 1907). A new era in the history of DO measurements in waters began with Winkler iodometric titration method (Winkler, 1888) which was based on the oxidizing properties of DO. For a long time, this has been the standard method for DO measurements because its accuracy, but it has its down sides, e.g. human error, release of supersaturated gas or aeration during sampling and sample handling.

12 Also titration itself is time consuming and difficult to execute in the field. The next step in DO measurements in water was made with the establishment of the colorimetric method (Oulman & Baumann, 1956) which offers a basic approximation of DO concentrations in a sample. The colorimetric method is quick and inexpensive but has its own limits and is subject to error. From here on, we needed methods which were easier to use and more accurate in most of applications. Modern technologies have introduced electrochemical/galvanic (Poole & Morrow, 1977) and optical methods (Quaranta et al., 2012) for measuring DO in waters. According to YSI user manual (YSI Inc., 2006), optical DO sensors measure the interaction between oxygen and certain luminescent dyes while electrochemical/galvanic sensors use two polarised electrodes and a cathode in an electrolyte solution. Optical DO sensors tend to be more accurate than their electrochemical counterparts are. Optical DO sensors are capable of accurately measure DO at very low concentrations, they are ideal for long time deployments due to their minimal maintenance requirements; can hold calibration for months compared to electrochemical sensors, which need more frequent calibration. Optical DO sensors also do not require as long warm-up time or stirring while taking measurements compared to electrochemical sensors.

Communication technologies between the sensor and data logger are divided into analogue and digital signals (Baher, 2001). Due to the advances in digital computation, digital signal is now commonly used in many applications, which were previously dominated by analogue techniques (Baher, 2001). Digital sensors have been developed to overcome the traditional disadvantages of analogue sensors. For instance, the signal that comes out is already digital and does not need any converting for later use with electronical devices; digital transmission is not sensitive to cable length (that with analogue sensors may be an issue); signal from many sensors can be sent through the same cable. Using several analogue sensors in one automated system (e.g., in automated buoys) requires a data logger with several analogue inputs. While sensing with digital sensors, we can connect several sensors to just one input on the data logger because the sensors send their number before measured data – which makes it easy to separate the data afterwards.

13 The present thesis gives a holistic overview of the application of high- frequency and continuous measurements in limnology which is followed by two case studies. As the first specific example, it introduces the combined use of high frequency and long-term water temperature and wind data for modelling the influence of a decade-scale atmospheric stilling on thermal stratification of a large and shallow lake. The second example analyses the effect of lake type-specific features and trophic state on summer distribution of dissolved gasses (DO and CO2) in eight Estonian lake types.

14 2. REVIEW OF THE LITERATURE

2.1. Patterns in applying high-frequency measurements in limnology

Because the global latitudinal distribution of lake numbers and lake areas peaks prominently between 40° and 70° North (Lehner & Döll, 2004), finding most lakes studied with high-frequency measurements (HFM) in North-America and Europe was an expected result. It was followed by Asia and worldwide studies. Review of most recent literature shows that during the last two years (2015-2016), 33 papers were published that fit our literature search criteria (I) and the leading position belongs to Europe, followed by North-America and global studies.

Figure 2.1.1. Distribution of the reviewed literature by year of publication and period of data collection (Fig. 2 (I) updated for years 2015 and 2016).

HFM data collected from 2008 to 2011 have been most represented in publications so far, whereas data collected during recent years are still in the preparation phase (Fig. 2.1.1). During the last two years, major emphasis of HFM studies in limnology has been put on different aspects of lake metabolism (Brighenti et al., 2015; Cremona et al., 2016b; Dugan

15 et al., 2016; Honti et al., 2016; Idrizaj et al., 2016; Tsai et al., 2016; Watras et al., 2016; Winslow et al., 2016), and the distribution of CO2 in lakes (Denfeld et al., 2015; III, Wilkinson et al., 2016).

2.2. HFM versus conventional measurements

Conventional measurement methods have evolved during decades and most commonly lakes are monitored on monthly or biweekly basis. Conventional limnological measurements including fieldworks and water sampling followed by lab analyses, require a lot of manpower and are time consuming. Usually, good weather conditions are needed to carry out fieldwork campaigns. Choosing good weather for sampling introduces a sampling bias as we likely miss important events taking place in harsh weather. Automated HFM instead gives us the opportunity to capture ecosystem changes constantly, in their full range and with a smaller effort compared to monthly manual monitoring.

With conventional measurements the horizontal distribution pattern of variables is often overlooked or studied at very low resolution given the unfeasibility of analysing a sufficiently high number of samples. Running HFM on ferryboats has become a widely used approach in marine monitoring (Rantajärvi et al., 1998; Kikas & Lips, 2016; Lips et al., 2016) however the same can be done in lakes using smaller vessels that gives us a cheap and easy opportunity to study horizontal distribution patterns of variables (e.g. the patchiness of plankton) at scales never achievable by means of conventional sampling (Anttila et al., 2008). High spatial resolution contact measurements by HFM if used for calibrating remote sensing data, allows to produce a broad and, at the same time, detailed picture of the processes taking place in a study area (Lindfors et al., 2005).

Using HFM has given considerable advantages compared to conventional sampling also in other fields of limnology, e.g. in lake metabolism studies. For decades, bottle incubation techniques based on either DO or 14C changes and encompassing only processes in the plankton community were dominating lake productivity studies

16 (Peterson, 1980). As a tempting alternative, the free-water, depth- weighted technique of estimating ecosystem metabolism covers primary production of all autotrophs and respiration of all lake biota, including nekton and benthos. Still, accounting for the effects of advective exchange of water masses at the measurement station remains one of the largest methodological challenges (Sadro et al., 2011a, 2011b) which can be overcome by increasing the number of measurement stations (Staehr et al., 2010). Another field of study in which the application of HFM has caused a real breakthrough, is the research of episodic events. HFM giving the opportunity to register and follow dynamical processes at timescales of minutes to hours allows to carry out detailed cause-effect relationships and discover fine scale reactions of biota to disturbances. The Intermediate Disturbance Hypothesis by (Connell, 1978) suggests that local species diversity is maximized when ecological disturbance frequency is at an intermediate level - neither too rare nor too frequent. Applying HFM opens new horizons for testing effects of various disturbance frequencies on biotic communities (e.g. Yang, 2015).

2.3. Most used water parameters in high-frequency studies

The fast growth of HFM applications in limnology mirrors the cumulative effect of methodological advancements made in a broad range of fields supporting various study purposes. The application of HF temperature measurements has expanded due to high reliability, low maintenance need, long autonomy and reasonable prices of temperature sensors. The use of data from low maintenance temperature monitoring ranges from describing thermal characteristics of lakes (Choiński & Łyczkowska, 2008) and detecting the presence of ice cover (Pierson et al., 2011) to studies on stratification (Song et al., 2013; Engelhardt & Kirillin, 2014; Pernica et al., 2014), lake mixing (Kulbe et al., 2008; Shade et al., 2010; Kimura et al., 2014; Bertone et al., 2015), effects of weather related episodic events such as heavy rainfalls (Jennings et al., 2012), summer heatwaves (Jöhnk et al., 2008) and storms (Tsai et al., 2008; Jennings et al., 2012; Klug et al., 2012).

17 Dissolved oxygen was the second most used parameter in HFM studies which has given a remarkable boost to open-water metabolism studies through calculating metabolic variables (gross primary production, net ecosystem production, and community respiration) in lakes, e.g. Coloso et al., 2008; Sadro et al., 2011a; Laas et al., 2012; Staehr et al., 2012; Van de Bogert et al., 2012. Additionally to lake metabolism studies, increasing usage of DO sensors in HFM have widened the knowledge about gas distribution and exchange in lakes (Read et al., 2012; Vachon & Prairie, 2013; Dugan et al., 2016; III).

Sensors measuring pH and conductivity are often included in a standard set of water monitoring probes. Both parameters are collected within standard water quality monitoring programs. There are several cases where pH plays an important role, e.g. in mineral carbon form equilibrium (Wetzel, 2001), ammonium/ammonia equilibrium (Emerson et al., 1975). It is also important to consider the corrosive effect of low pH on metals in drinking water supply systems (Volk et al., 2000) and its important role in environmental redox processes (Karakaya et al., 2011). Conductivity characterizes the ionic strength or the salt content of water being the best parameter to distinguish between different water masses in brackish waters, e.g. estuaries (Smith et al., 2008).

2.4. Thermal stratification of lakes

Globally increasing deployment of instrumented buoys over the last decade has given a strong impetus to comparative studies of lake thermal structure with increased temporal resolution and spatial extent (Porter et al., 2009; Pernica et al., 2014; Woolway et al., 2014). Thermal stratification is one of the best known thermal phenomena in water bodies that is represented by vertical layering of water masses of different temperature and density resulting from the absorption of solar irradiance in the surface layers and thermal expansion of water. The developed strata reflect the balance between turbulent mixing and buoyancy forces, which act to supress turbulence (Boehrer & Schultze, 2008). The upper layer called epilimnion is subject to direct atmospheric

18 forcing and is characterized by rather homogeneous temperature resulting from convective mixing. Hypolimnion, the coolest and densest stagnant water layer is constantly in contact with lake bottom but mostly isolated from direct atmospheric forcing and usually shows a small vertically decreasing temperature gradient. These two layers are separated from each other by a transitional layer called metalimnion in which the location of the maximum temperature gradient is termed the thermocline. Such compartmentalisation of the water column creates vertical differences in the availability of nutrients, light for phytoplankton, substrates for bacteria, and affects vertical distribution, migration, and feeding of organisms of higher trophic levels such as zooplankton and fish (Read et al., 2011).

Stratification can be transient or persistent, varying at time scales of hours in polymictic lakes (Pernica & Wells, 2012) to decades, e.g. in deep perialpine lakes (Salmaso, 2005). In deep tropical lakes like Tanganyika, constant stratification is supported besides thermal effects also by a salinity gradient (Spigel & Coulter, 1996). Even small changes in thermal stability as they affect the availability of light and nutrients, the distribution of dissolved gases, and vertical migrations of several planktonic organisms, may have major implications for food web interactions and the equilibrium state of the whole ecosystem. Using HFM we can now detect daily vertical mixing patterns in generally stratified lakes (Andersen et al., 2017), micro- stratification events in lakes that are considered well mixed (II), and even study thermal stratification under ice (Bruesewitz et al., 2015).

2.5. Climate change effects on lakes

Since the 1990s, climate change effects on lakes are widely studied and the findings summarized in a number of regional thematic overviews for North America (Cushing, 1997; Rouse et al., 1997), Europe (Blenckner et al., 2007; George, 2010), Africa (Magadza, 1994), and Australia (Jones et al., 2001). Most of the studies usually focus on the effects of increasing air temperature (e.g. Dokulil, 2014), although also changes in precipitation and wind may have notable impacts on lakes hydrological

19 and mixing regimes. Year-to-year changes in weather have major effects on thermal characteristics of lakes (Arvola et al., 2010). Besides secular climate change, seasonal and annual variations in atmosphere circulation patterns may have major effects on the physical dynamics of lakes (George et al., 2010). Anticyclonic days in summer characterised by very low rainfall, increased temperature and below average wind speeds have a major effect towards stronger thermal stratification of lakes. As shown by (Vautard et al., 2010), surface wind speeds have declined by 5 - 15% over almost all continental areas in the northern mid-latitudes. As a combined effect of increasing water temperatures and decreasing wind speed, the stability of the lake water column may increase even more that will reduce the mixing depth (O’Reilly et al., 2003). Still the changes may differ by region. A Europe-wide study assessing future changes in wind potential for power generation (Tobin et al., 2015) showed with a high level of confidence that under the SRES A1B emission scenario there will be a tendency toward a decrease of the wind power potential over Mediterranean areas and an increase over Northern Europe.

Given the high variability of environmental parameters and, hence, the low signal-to-noise ratio in secular climate change studies, massive databases are required to reach the necessary statistical significance for proving trends in time-series. HF data of climatic and aquatic parameters are also needed for revealing causal relationships through dynamic modelling of climate change impacts. A good example of combining HF and long-term measurements is given in the study of summer stilling effects on thermal regime of Võrtsjärv (Paper II).

2.6. Dissolved gases in lakes

2.6.1. Dissolved oxygen

DO is one of the most vital substances for life present in water. Being the most important oxidizing agent, it accepts an electron during cellular respiration. This reaction is vital not only because it fuels the energy systems of organisms, but also because it is involved in the natural decomposition of organic matter (mineralization) and biodegradation of

20 organic contaminants. Low concentrations of DO in natural waters may lead to death of biota, but also too high DO levels can harm aquatic life and affect water quality. Therefore, it is very important to be able to measure accurately DO concentrations in water. There are currently three mostly used methods for DO measurements: the Winkler iodometric titration method (Winkler, 1888), galvanic/electrochemical sensors (Taillefert et al., 2000), and optical sensors based on luminescence quenching (e.g. Sadro et al., 2011a).

In a constantly mixed water body with low biological activity, DO will remain at 100% air saturation, meaning that the water is holding as many dissolved gas molecules as it can in equilibrium at any given pressure and temperature conditions. In thermally stratified lakes, the distribution of oxygen is controlled by a combination of solubility conditions, hydrodynamics, inputs from photosynthesis, and losses to chemical and metabolic oxidation (Wetzel, 2001). The diffusive DO exchange at the air-water interface can be calculated with a reasonable accuracy with air- water gas transfer models using HF data (Holmén & Liss, 1984). The rest of DO changes can be attributed to metabolic processes in the lake that enables estimating ecosystem primary production and respiration from measurements of diel, open water changes in DO concentrations. This technique termed the open water method has become widely accepted in aquatic science and has been particularly popular in lakes (Gelda & Effler, 2002; Obrador et al., 2014).

2.6.2. Carbon dioxide

The distribution of carbon dioxide (CO2) in a lake’s water column is formed by a combination of inputs (dissolution from atmosphere, riverine and ground water loadings, in-lake release through respiration,) and outputs (evasion to the atmosphere, riverine or groundwater export, in-lake photosynthetic uptake, and methane oxidation) (Riera et al., 1999). Lakes with large inputs of allochthonous carbon from the catchment are often supersaturated with CO2 (Jonsson et al., 2003), however CO2 supersaturation can be caused also by other processes such as negative net ecosystem production (Cole et al., 2000),

21 photochemical degradation of dissolved organic carbon (DOC) (Vachon et al., 2016) or high dissolved inorganic carbon inflow from surface- or groundwater (Marcé et al., 2015; Weyhenmeyer et al., 2015). Lakes supersaturated with CO2 and release large amounts of CO2 to the atmosphere, particularly in winter when productivity is low (Maberly, 1996).

In several lakes, especially those with a high DOC concentration, the in- lake decomposition of allochthonous organic matter, both in the water column and sediment, plays an important role in the CO2 build-up (Jonsson et al., 2003). As photosynthetic CO2 uptake depends on diurnal and seasonal light availability, the rates are highly variable both vertically (Casper et al., 2000) and temporally (Dinsmore et al., 2009). In productive lakes photosynthetic CO2 depletion can occur leading to an influx of atmospheric CO2 (Maberly, 1996). Still, periods of net CO2 uptake in lakes seem to be limited to the growth rate peak of phytoplankton in early summer and on average 75% of lakes are supersaturated with CO2 and undersaturated with oxygen indicating predominant epilimnetic net heterotrophy in these lakes (Prairie et al., 2002).

Measuring CO2 concentrations in lakes gives us a better understanding of their metabolic status. Diurnal CO2 variability in lakes captured by HFM of CO2 (Dinsmore et al., 2009) allows to decide whether a lake functions as a carbon “sink” or “source”. Generally, lakes tend to be carbon “sources” during winter when productivity is low (Maberly, 1996). There are different ways to measure CO2 in fresh water systems. The most common method today is using none-dispersive infra-red based CO2 sensors (Baehr & DeGrandpre, 2004; Vachon & del Giorgio, 2014). Other measurement methods include the headspace equilibrium method (Kling et al., 1991) and indirect estimation of dissolved CO2 concentration from measured pH and alkalinity (Neal et al., 1998). Regarding the latter, however, (Johnson et al., 2010) note that small scale temporal changes in dissolved CO2 concentrations might be lost due to high uncertainty of CO2 estimates derived from pH measurements.

22 3. HYPOTHESES AND AIMS OF THE STUDY

The main objective of this thesis is to elucidate the present world-wide use and future perspectives of HFM in limnology and offer closer insights into two fields of application – using HFM in modelling climate change impacts on lakes and in describing type specific differences in lakes’ gas regime.

The specific aims of the thesis are:

1) to give an overview of continuous and high-frequency measurements in limnology (I)

a. highlighting the advantages of using HFM for various purposes compared to discrete or conventional measurements putting the main focus on HF data applications related to buoy and mooring systems (omitting other fields of HF technologies, e.g. SONAR, eco-sounding, ADCP or remote-sensing systems;

b. pointing out the universality and broad scale of applications of HFM in lakes by grouping the studies according to their limnological or lake management issues.

Hypothesis: As Paper I is a review paper, no real hypotheses were drawn. However, based on the published evidence, we expected to document the ever growing importance of HFM in limnological research and monitoring which in some areas are already replacing conventional field measurements (e.g. in lake metabolism studies).

2) based on HF weather and lake data to model water temperature distribution and stratification episodes in a large shallow polymictic Lake (Võrtsjärv)(II);

3) to apply the model to historical weather data to find out whether the magnitude of observed atmospheric stilling over 30 years could influence the frequency and duration of stratification episodes in Lake Võrtsjärv (II).

23 Hypothesis: Long-term decrease in wind speeds has caused substantial changes in stratification strength and frequency in Võrtsjärv.

4) to analyse how CO2 and DO vertical differences, saturation levels and their coupling are related to lake type and trophic state parameters (II);

5) to estimate the potential of HF vertical CO2 measurements as a method to enhance the understanding of carbon dynamics in aquatic environments. For this we compared the vertical distribution of dissolved CO2 and DO of 8 hemiboreal lake types based on obtained by direct continuous in situ measurements (III) in order to demonstrate

Hypothesis: we expected that DO and CO2 would be evenly distributed in homothermal conditions and that thermal stratification will cause opposite changes in the distribution of these two gases in lakes

24 4. MATERIAL AND METHODS

4.1. Study sites

The study was conducted in eight Estonian lakes each belonging to a different lake type according to the EU Water Framework Directive typology (Fig. 1; III). The typology is based on lake area, alkalinity, conductivity, chloride content, thermal stratification, and colour (Table 1; III). The general description of the lakes is given in papers II and III (Table 2).

4.2. Data collection

Literature search for the review was conducted by using the Google Scholar citation database. Search terms were “lake*” and “high- frequency data”. The timeframe was set to cover the last 15 years (i.e. articles published between 2000 and 2015 were considered). Screening of the 1480 papers retrieved revealed the main fields of HFM application for which we repeated more specific queries. We selected studies in which measurement intervals of ≤ 60 min were used and deployment time was at least 6 hours. Critical screening of the initial retrieval yielded a final list of 154 publications. Literature search and review table are thoroughly described in paper I.

High-frequency data for Lake Võrtsjärv surface and bottom temperatures was collected from 2013 to 2015. Long-term weather and Lake Võrtsjärv surface water temperature was measured by Estonian Environmental Agency and data since 1982 was used in current study. Instrumentation and data collection are described in paper II.

HF DO and CO2 data for eight lakes was collected from June to September 2014. Manual water samples for laboratory analysis were taken from all studied lakes once during the sensor deployment period. High-frequency instrumentation, manual and HF data collection are thoroughly described in paper III.

25 4.3. Modelling lake temperature

For lake temperature modelling we used the MyLake model (v1.2; (Saloranta & Andersen, 2007). The modelling principles are similar to other one-dimensional lake model codes, e.g. DYRESM-CADEYM (Hamilton & Schladow, 1997). A detailed description of the model is presented in paper II.

4.4. Statistical analyses

For assessing the significance of differences between proportions of various categories in the literature review results (I), a Z-test package was used available online at http://www.socscistatistics.com/.

An analysis of covariance (ANCOVA) was used (II) to compare the difference between regression slopes while testing the effect of a categorical factor on the dependent variable. All calculations were made by R (R Development Core Team, 2014).

Relationships between lake type/trophic state characteristics and measured gas distribution patterns (III) were evaluated using the basic statistics, multiple regression and principal component modules of STATISTICA analysis software (Dell Inc., 2015).

26 5. RESULTS

5.1. New insights into using continuous and high-frequencyHFM measurementsin limnology in limnology

The growing importance of HFM studies in limnology is reflected in publication statistics of ISI Web of Science, which shows that data derived from HFM were analysed in 384 publications since 1997 with a peak in 2011. The annual citation rate has exponentially increased from ca 30 in 1998 to nearly 950 in 2015.

In publication I, we distinguished 12 main purposes for which HFM have been used at different frequencies (Fig. 5.1.1). Most often HFM have been used for low-maintenance monitoring (e.g. water temperature) and collecting meteorological background data.

Figure 5.1.1. Frequency distribution by study purpose (publication I).

Lake metabolism studies using HF data have been growing in number in recent years and occupied the third position. There are still rare topics where HFM are applied (e.g. developing sampling strategies or calibrating remote-sensing data).

27 We also found that water temperature was the parameter most often measured, while measurements of DO concentration took the second place (Fig. 5.1.3). These two main parameters were followed by some meteorological (e.g. wind speed, air temperature, PAR) and water quality variables (conductivity, pH, turbidity). Chlorophyll a was more common in HFM programs than other biological parameters (e.g. phycocyanin or phycoerythrin). Measurements of chemical parameters (e.g. NO3, PO4) were rare.

Almost 60% of the studied lakes were described as thermally stratified and 24% as unstratified. Among the lakes with trophic state indicated, eutrophic and oligotrophic lakes formed equally 40% of studied lakes (Fig. 1e and 1f; I). Most often the studies were carried out in a timeframe of >1 month to 1 year (Fig. 5; I) and most of the studies were based on single point HFM in vertical profiles (Fig. 1d; I).

Figure 5.1.1. Parameters studied with high-frequency measurements (publication I).

Below the two case studies are introduced to offer closer insights into fields of application of HF data.

28 5.2. Has stratification regime changed in Lake Võrtsjärv?

In publication II, we found that over the last 28 years, wind speed has been declining in all seven meteorological stations in Estonia (Fig. 1; II) for which wind data were readily available. This is seen in the raw daily data from Tartu-Tõravere (Fig. 5.2.1a) and in the annually averaged wind speeds from all seven meteorological stations covering most of Estonia (Fig. 5.2.1b). Decline in wind speeds in all meteorological stations was

Figure 5.2.1. Observations of (a) daily-average wind speed measured at Tartu- Tõravere, (b) annually-averaged wind speed anomaly (relative to 1990 to 2010) and (c) occurrence frequencies (in %) for wind speeds > 3 m s-1. Grey lines illustrate the wind speeds for an individual meteorological station in Estonia, and the thick black line represents the regional average (computed as the arithmetic mean of all stations)(publication II).

29 statistically significant (p < 0.05). The observational data also suggested a decrease in the frequency of wind speeds > 3 m s-1 (calculated as the % of time in which wind speeds exceeded 3 m s-1 during a given year). Among other meteorological parameters measured in situ, only air temperature showed statistically significant (p < 0.05) change since 1982. None of the other atmospheric forcing data experienced a significant change in their annual averages during study period, nor did the lake water level (Fig. 2; II).

Figure 5.2.2. Comparison of modelled and observed (a) surface and (b) bottom water temperatures for 2013-2015. (c) Comparison of long-term modelled (grey line) and observed (black dots) lake surface water temperature, showing a comparison throughout the year (publication II).

The simulations of temperature in both surface and bottom waters of Võrtsjärv showed very good agreement with observed data during the three-year validation study (Fig. 5.2.2a, b). Using 28 years of atmospheric forcing data from Tartu-Tõravere, MyLake generally captured the temporal dynamics of lake surface temperatures with great success (Fig. 5.2.2c).

30 The model results demonstrate a significant relationship between wind speed and the number of stratified days in May, June, July, and August over the 28 years (Fig. 4b; II). Analysis of the model results for the past 28 years demonstrated an increase of over 80% in the number of stratified days in the second half (1995-2009) compared to the first half (1982-1995) of the study period (Fig. 4c; II).

Figure 5.2.3. Comparison of the number of stratified days from model runs with (a) constant 1982 air temperature annual cycle; (b)constant 1982 wind speed annual cycle; (c)constant 1987 air temperature annual cycle; and (d)constant 2009 air temperature annual cycle. Linear regressions of the statistically significant (p < 0.05) relationships are shown (publication II). 31 Our model sensitivity analysis illustrates that the decrease in wind speed is the key influence on the number of stratified days and that the influence of increasing air temperature was minimal (Fig. 5.2.3.). Running MyLake with a constant annual air temperature cycle demonstrates that the number of stratified days still increased markedly in recent years (Fig. 5.2.3. a), at a rate of 1.7 ± 0.6 days per decade (p < 0.001). In contrast, when the annual cycle in surface wind speed was kept constant, but air temperature allowed to vary, the model results show no statistically significant increase (p > 0.05) in the number of stratified days during the past 28 years (Fig. 5.2.3. b). We then evaluated further the impact of changing air temperature on the number of stratified days by running the model with air temperatures from 1987 (Fig. 5.2.3. c) and 2009 (Fig. 5.2.3. d), the coolest and warmest years, respectively, held constant. The calculated number of stratified days showed a statistically significant increase in both model runs, with neither of these simulations being statistically different from the model using observed meteorological forcing data.

5.3. DO and CO2 regimes in 8 lake types in Estonia

In publication III, we studied representatives of 8 Estonian lake types established according to European Water Framework Directive criteria (Directive, 2000; Table 1; III). Only one of those eight lake types has been described as a stratified one. Our results showed that 3 of the type representative lakes were strongly stratified and 3 others stratified occasionally and weakly during our study period (Fig. 2; III). Although some lakes did not exhibit a stable thermal stratification or were fully mixed during our study, they were all stratified for dissolved gases (Fig. 3; 4; III) with the strongest stratification occurring in two deeper lakes (StratMedAlk and LightSoft), but also in the shallow Alk.

Figure 5.3.1. Dissolved oxygen (DO) and carbon dioxide (CO2) distribution in Estonian lakes (publication III). We found several similarities between the occurrence of extremes of CO2 and DO in lakes (Fig. 5.3.1). Both gases had their highest surface layer saturation levels in Alk, the largest vertical difference occurred in the stratified LightSoft and the smallest vertical gradients in V-Large. Both gases showed the largest bottom layer standard deviation in MedAlk and the largest surface layer variation coefficients in Costal. Highest DO saturation levels were measured in the surface layer of the Alk (144%) but reached nearly 120% also in the eutrophic Costal, Large and MedAlk. Unlike other lakes where the highest values of DO were measured near the surface, in V-Large maximum DO levels occurred at 1.5 m depth.

33 6. DISCUSSION

Our review of HFM applications in limnology documented the status quo in this field showing that the development of sensor technologies during the last 15 years has opened new horizons for limnologists to study lake processes at short time scales that with conventional methods was almost impossible. Furthermore, continuous measurements can be done in extreme conditions and in places hardly accessible with conventional equipment. As the centres of mass in global distribution of both lakes (Lehner & Döll, 2004) and limnological research potential (King, 2004) are located in the Northern Hemisphere, so is the application of HFM in limnology still concentrated mostly in lowland areas between 40ºN - 70ºN although studies have expanded to cover tropical regions (e.g. Townsend et al., 2011), alpine areas (Sadro & Melack, 2012) and even both Antarctica and the Arctic (Hamilton et al., 2015). Automated buoys in remote lakes can record real-time impacts of typhoons on lake physics and biology (Porter et al., 2005; Hanson, 2007; Jones et al., 2008). The 15-year time span of our literature review enables, besides describing the status quo, reveal also a number of trends and tendencies in the use of HFM that allow some insight into future prospective of this technology:

1) improvement of technical parameters and dropping prices of sensors

Fast technical development in sensor technologies has rapidly improved the reliability, precision and autonomy of sensors hand in hand with less maintenance needs and dropping prices. A good example of that are temperature data loggers (e.g. HOBO®) measuring water temperatures from 0° to 50°C with a resolution of 0.02°C at 25°C, having a battery life of 6 years with 1 minute or greater logging interval, and a maximum drift of 0.1°C per year (http://www.onsetcomp.com/products/data-loggers/u22-001). Water temperature is a universal parameter that controls the rate of most chemical and biological processes in the environment (Poff et al., 2002) and that besides can be used as a marker to trace the movement of water masses (e.g. MacIntyre et al., 2002). Also the optical DO sensors have proven more reliable, maintenance-free and

34 stable compared with galvanic sensors and have allowed DO to become the second popular automatically measured parameter after water temperature in limnological studies. (Fig. 5.1.2). HF DO measurements are nowadays massively used in studying lake metabolism, i.e. gross primary production, community respiration, net ecosystem production (e.g. Cole et al., 2000; Tsai et al., 2008; Staehr et al., 2010; Laas et al., 2012) that has achieved a new level largely thanks to developments in sensor technologies.

2) improvement of spatial resolution

Metabolism studies can serve as a good example also for showing the advancement of spatial resolution of HF measurements. In these studies, a tendency of transition from single-point-single-depth measurements to multi-point (Coloso et al., 2008; Van de Bogert et al., 2012); I), transect or polygon measurements carried out over ranges of different depth can be noted. Installing sensors on board of ferries has become an increasingly popular option in marine monitoring but the small size of the sensors enables using them also on small boats in lakes. In this way, high temporal resolution is turned into high spatial resolution to study horizontal distribution of parameters not achievable with conventional sampling (e.g. Anttila et al., 2008; Kikas & Lips, 2016; Lips et al., 2016). In parallel with sensor technologies also the equipment for sensor deployment, such as buoys and platforms, and data transmission have developed. Profiling buoys equipped either with a chain of sensors or a mechanical winch moving sensors up and down have become a norm rather than exception in stratified lakes (Brentrup et al., 2016). This has enabled research-grade measurements at an affordable price and suggests that HFM have achieved a stable position in limnology and have replaced previously conventional methods in some fields, for example in lake metabolism studies.

3) global networking of measurements

The best example of global HFM networking is the Global Lake Ecological Observatory Network (GLEON). GLEON is a grass-root voluntary network of scientists, educators and community groups

35 interested in using, sharing and interpreting HF data for predicting the response of lakes to changes in global environment. In addition to limnologists and ecologists, GLEON involves experts of information technology and engineers having a common goal to develop theoretical models based on HF data and creating options for their more efficient use. GLEON has initiated a number of research projects in which lake data from various regions of the globe have been used in a shared way (e.g. Pierson et al., 2011; Jennings et al., 2012; Read et al., 2012; Woolway et al., 2014). GLEON have published a special issue in Inland Waters (volume 6, 2016). The Networking Lake Observatories in Europe (NETLAKE) joining European researchers and IT experts dealing with HFM in lakes can be considered a small brother of GLEON.

If in some fields, such as lake metabolism, the availability of HF data has supported the emergence of new methodologies, then in other fields using HF data in models has considerably improved model reliability that in some cases allows even retrospective analysis of data from periods when HF data were not yet available (II).

We have reached the era of Big data (e.g. Boyd & Crawford, 2012; Swan, 2013). Sensor networks produce continuously enormous data sets that are so large or complex that traditional data processing software becomes inadequate to deal with them. Capture, storage, transfer, sharing, quality check, visualization, querying and finally – analysis of these data volumes requires specific tools and a lot of experience. Regarding these issues, limnologists have got good support from the computing experts participating in GLEON and NETLAKE who have elaborated a number of specific solutions to facilitate and speed up data processing.

In several questions of exploratory research for which data accumulation has not yet reached the required critical mass to rise to a new qualitative level, single case studies may still play an important role. Our two case studies – one analysing the long-term changes in Lake Võrtsjärv stratification regime caused by multi-decadal wind stilling, and the other describing summer gas regime in Estonian lakes of different types – were both in many ways novel and pioneering in their fields.

36 So far, Võrtsjärv has been considered as a continuously fully mixed lake in which even ephemeral stratification is exceptional (Frisk et al., 1999). Wind has a strong impact on this large and shallow lake and, thus, any change in wind speed or direction is likely to influence the mixing regime. Analysis of wind data from the adjacent Tõravere meteorological station showed an increase in the number of calm days (average wind speed <3 m/s). (Jaagus & Kull, 2011) have explained the stilling tendency in Estonia by changes in atmospheric circulation patterns and/or changes in instrumentation, whereas (Suursaar & Kullas, 2006) pointed out the increased surface roughness in vicinity of some measuring stations. The results of Paper II showed that the stilling trend was common in all 7 wind time series that we checked which were obtained from different meteorological stations in Estonia located from each other at distances up to 200 km. This clearly contradicts the idea of increasing surface roughness as an explanation as it is unlikely that similar changes (e.g. increasing effect of housing or forest growth) have taken place at all meteorological stations. In fact, a similar stilling has been reported also elsewhere in the Northern Hemisphere (Vautard et al., 2010). Also the effect of changed measurement equipment can be eliminated as this could have resulted in a step change and not in a smooth trend as revealed in our study.

Only very few studies have addressed the effect of changing winds on thermal stratification of lakes and those have mostly analysed the effects of logging around lakes (France, 1997; Tanentzap et al., 2008) or the effect of seasonal wind patterns on sediment resuspension depending on lake water level (Tanentzap et al., 2008; Kerimoglu & Rinke, 2013). According to our knowledge, our study (Paper II) was the first to relate changes in thermal stratification of lakes with regional long-term atmospheric stilling. Model calculations based on HF wind and water temperature data showed that increasing number of calm days will bring about longer lasting stratification episodes in Võrtsjärv that may have strong impacts on its ecosystem.

Gas regime studies in lakes using HFM are in a rising trend (Dinsmore et al., 2009; Obrador et al., 2014; Vachon & del Giorgio, 2014; Dugan et al., 2016). The novelty of our study (III) was that we compared DO and CO2 summer distribution in a broad variety of lake types –

37 representatives of all 8 lake types distinguished in Estonia according to European Water Framework Directive (Directive, 2000). A comparison of this kind has neither been carried out in Estonia nor elsewhere in the world. The eight lake types, most of which being rather common within the Northern Hemisphere, showed striking differences both in thermal stratification and in vertical distribution of DO and CO2. In most lakes the surface layer was supersaturated both with DO and CO2 similarly to that described for many lakes around the world (Dinsmore et al., 2009; Vachon & del Giorgio, 2014), and thus contributed to CO2 atmospheric emission. As described by (Dinsmore et al., 2009) and (Ducharme-Riel et al., 2015), supersaturation with CO2 typically increased with depth reaching its maximum in the bottom layer of lakes. High values of DO% found in the alkalitrophic lake type (Alk) were rather unexpected as this karstic lake type has mostly groundwater feeding and low chlorophyll and primary production levels. As the bottom of the lake representing the Alk type was covered by Chara and the water had high transparency, the high DO% values could be attributed to this phytobenthic community (Cremona et al., 2016a). The gas distribution observed in the lake representing the medium alkaline type (MedAlk) was likely caused by its eutrophic state. Our measurements showed that in virtually homothermal conditions, strong gradients of both DO and CO2 may appear in the bottom layers overnight as a result of organic matter decomposition causing high sediment oxygen demand. Another factor causing inhomogeneity in gas distribution is the vertically attenuating photosynthetic activity taking up CO2 and enriching water with DO most intensively in the well illuminated surface layers. Given the strong wind exposure of Large and V-Large lake types, we expected vertically homogeneous temperature distribution and partial pressures of DO and CO2 in equilibrium with the atmosphere. The considerable vertical temperature difference due to calm weather occurring in the representative of large lakes, however, created atypically large inhomogeneity in DO and especially CO2 saturation levels. Taking into account the wind stilling trend observed over the last 30 years (II), it is likely that it affects not only the thermal regime but also the gas regime of lakes with implications for the functioning of the whole lake ecosystem.

38 CONCLUSIONS

The main objective of this thesis was to elucidate the present world-wide use and future perspectives of HFM in limnology and offer closer insights into two fields of application – using HFM in modelling climate change impacts on lakes and in describing type specific differences in lakes’ gas regime.

We documented the ever growing importance of HFM in limnological research and monitoring which in some areas (e.g. in lake metabolism studies) are already replacing conventional field measurements. Sensor technologies developing towards increasing sensitivity, reliability and autonomy, have opened new horizons for studying lakes and their success is likely to continue. However, it is clear that automated and conventional measurements complement each other and the wider use of HFM will never fully replace classic field works with conventional measurements in which the experienced eye and intuition of the researcher have crucial roles (I).

Our analysis showed that long-term decrease in wind speeds has caused substantial changes in stratification strength and frequency in Võrtsjärv, while increasing air temperatures had a negligible effect to it (II).

We expected DO and CO2 to be evenly distributed in homothermal conditions in lakes and that thermal stratification will cause opposite changes in the distribution of these two gases. Although the second hypothesis was mostly supported by our study, our measurements showed that in homothermal conditions, strong gradients of both DO and CO2 could appear as a result of organic matter decomposition causing high sediment oxygen demand in the bottom layers and intensive photosynthesis taking up CO2 and enriching the water with DO in the surface layers (III).

39 REFERENCES

Andersen, M. R., K. Sand-Jensen, R. Iestyn Woolway & I. D. Jones, 2017. Profound daily vertical stratification and mixing in a small, shallow, wind-exposed lake with submerged macrophytes. Aquatic Sciences 79: 395–406.

Anttila, S., T. Kairesalo & P. Pellikka, 2008. A feasible method to assess inaccuracy caused by patchiness in water quality monitoring. Environmental Monitoring and Assessment 142: 11–22.

Arvola, L., G. George, D. M. Livingstone, M. Järvinen, T. Blenckner, M. T. Dokulil, E. Jennings, C. N. Aonghusa, P. Nõges, T. Nõges & G. A. Weyhenmeyer, 2010. The Impact of the Changing Climate on the Thermal Characteristics of Lakes. In George, G. (ed), The Impact of Climate Change on European Lakes. Springer Netherlands, Dordrecht: 85–101.

Baehr, M. M. & M. D. DeGrandpre, 2004. In situ pCO2 and O2 measurements in a lake during turnover and stratification: Observations and modeling. Limnology and Oceanography 49: 330–340.

Baher, H. 2001. Analog and Digital Signal Processing. John Wiley & Sons, Chichester, New York.

Bertone, E., R. A. Stewart, H. Zhang & K. O’Halloran, 2015. Analysis of the mixing processes in the subtropical Advancetown Lake, Australia. Journal of Hydrology 522: 67–79.

Blenckner, T., R. Adrian, D. M. Livingstone, E. Jennings, G. A. Weyhenmeyer, D. G. George, T. Jankowski, M. Järvinen, C. N. Aonghusa, T. Nõges, D. Straile & K. Teubner, 2007. Large-scale climatic signatures in lakes across Europe: a meta-analysis. Global Change Biology 13: 1314–1326.

Boehrer, B. & M. Schultze, 2008. Stratification of lakes. Reviews of Geophysics 46: RG2005.

40 Boyd, D. & K. Crawford, 2012. Critical questions for Big Data: Provocations for a cultural, technological, and scholarly phenomenon. Information, Communication & Society 15: 662– 679.

Brentrup, J. A., C. E. Williamson, W. Colom-Montero, W. Eckert, E. de Eyto, H. P. Grossart, Y. Huot, P. D. F. Isles, L. B. Knoll, T. H. Leach, C. G. McBride, D. Pierson, F. Pomati, J. S. Read, K. C. Rose, N. R. Samal, P. A. Staehr, & L. A. Winslow, 2016. The potential of high-frequency profiling to assess vertical and seasonal patterns of phytoplankton dynamics in lakes. Inland Waters 6: 565–580.

Brighenti, L. S., P. A. Staehr, L. M. Gagliardi, L. P. M. Brandão, E. C. Elias, N. A. S. T. de Mello, F. A. R. Barbosa & J. F. Bezerra-Neto, 2015. Seasonal changes in metabolic rates of two tropical lakes in the Atlantic forest of Brazil. Ecosystems 18: 589–604.

Bruesewitz, D. A., C. C. Carey, D. C. Richardson & K. C. Weathers, 2015. Under-ice thermal stratification dynamics of a large, deep lake revealed by high-frequency data: Lake under-ice mixing and stability. Limnology and Oceanography 60: 347–359.

Casper, P., S. C. Maberly, G. H. Hall & B. J. Finlay, 2000. Fluxes of methane and carbon dioxide from a small productive lake to the atmosphere. Biogeochemistry 49: 1–19.

Choiński, A., & G. Łyczkowska, 2008. Thermal Characteristics of Waters of Wielki Staw in the Karkonosze Mountains and Morskie Oko in the Tatras, July 2006. Polish Journal of Environmental Studies17: 835–840.

Cole, J. J., M. L. Pace, S. R. Carpenter & J. F. Kitchell, 2000. Persistence of net heterotrophy in lakes during nutrient addition and food web manipulations. Limnology and Oceanography 45: 1718–1730.

Coloso, J. J., J. J. Cole, P. C. Hanson & M. L. Pace, 2008. Depth- integrated, continuous estimates of metabolism in a clear-water lake. Canadian Journal of Fisheries and Aquatic Sciences 65: 712– 722.

41 Connell, J. H. 1978. Diversity in Tropical Rain Forests and Coral Reefs. Science 199: 1302–1310.

Cremona, F., A. Laas, L. Arvola, D. Pierson, P. Nõges & T. Nõges, 2016a. Numerical exploration of the planktonic to benthic primary production ratios in lakes of the Baltic Sea catchment. Ecosystems 19: 1386–1400.

Cremona, F., A. Laas, P. Nõges, & T. Nõges, 2016b. An estimation of diel metabolic rates of eight limnological archetypes from Estonia using high-frequency measurements. Inland Waters 6: 352–363.

Cushing, C. E. (ed), 1997. Freshwater ecosystems and climate change in North America: a regional assessment. Wiley, Chichester, New York.

Dell Inc. 2015. Dell Statistica (Data Analysis Software System). Version 13, https://www.quest.com/products/statistica/.

Denfeld, B. A., M. B. Wallin, E. Sahlée, S. Sobek, J. Kokic, H. E. Chmiel & G. A. Weyhenmeyer, 2015. Temporal and spatial carbon dioxide concentration patterns in a small boreal lake in relation to ice cover dynamics. Boreal environment research 20: 679–692.

Dinsmore, K. J., M. F. Billett & T. R. Moore, 2009. Transfer of carbon dioxide and methane through the soil-water-atmosphere system at Mer Bleue peatland, Canada. Hydrological Processes 23: 330–341.

Directive, 2000. Directive 2000/60/EC of the European Parliament and of the council of 23 October 2000 establishing a framework for community action in the field of water policy. Official Journal of the European Communities L 327: 1–72.

Dokulil, M. 2014. Impact of climate warming on European inland waters. Inland Waters 4: 27–40.

Ducharme-Riel, V., D. Vachon, P. A. del Giorgio & Y. T. Prairie, 2015. The relative contribution of winter under-ice and summer hypolimnetic CO2 accumulation to the annual CO2 emissions from Northern lakes. Ecosystems 18: 547–559.

42 Dugan, H. A., R. I. Woolway, A. B. Santoso, J. R. Corman, A. Jamies, E. R. Nodine, V. P. Patil, J. A. Zwart, J. A. Brentrup, A. L. Hetherington, S. K. Oliver, J. S. Read, K. M. Winters, P. C. Hanson, E. K. Read, L. A. Winslow & K. C. Weathers, 2016. Consequences of gas flux model choice on the interpretation of metabolic balance across 15 lakes. Inland Waters 6: 581–592.

Emerson, K., R. C. Russo, R. E. Lund & R. V. Thurston, 1975. Aqueous ammonia equilibrium calculations: effect of pH and temperature. Journal of the Fisheries Research Board of Canada 32: 2379–2383.

Engelhardt, C. & G. Kirillin, 2014. Criteria for the onset and breakup of summer lake stratification based on routine temperature measurements. Fundamental and Applied Limnology 184: 183– 194.

France, R. 1997. Land water linkages: influences of riparian deforestation on lake thermocline depth and possible consequences for cold stenotherms. Canadian Journal of Fisheries and Aquatic Sciences 54: 1299–1305.

Frisk, T., Ä. Bilaletdin, H. Kaipainen, O. Malve & M. Möls, 1999. Modelling phytoplankton dynamics of the eutrophic Lake Võrtsjärv, Estonia. Hydrobiologia 414: 59–68.

Gelda, R. K. & S. W. Effler, 2002. Estimating oxygen exchange across the air–water interface of a hypereutrophic lake. Hydrobiologia 487: 243–254.

George, D. G. (ed), 2010. The impact of climate change on European lakes. Springer Verlag, Dordrecht, New York.

George, G., U. Nickus, M. T. Dokulil & T. Blenckner, 2010. The influence of changes in the atmospheric circulation on the surface temperature of lakes. In George, G. (ed), The Impact of Climate Change on European Lakes. Springer Netherlands, Dordrecht: 293–310.

Hamilton, D., C. Carey, L. Arvola, P. Arzberger, C. Brewer, J. Cole, E. Gaiser, P. Hanson, B. Ibelings, E. Jennings, T. Kratz, F.-P. Lin, C. McBride, D. de Motta Marques, K. Muraoka, A. Nishri, B. Qin, J. Read, K. Rose, E. Ryder, K. Weathers, G. Zhu, D. Trolle & J.

43 Brookes, 2015. A Global Lake Ecological Observatory Network (GLEON) for synthesising high–frequency sensor data for validation of deterministic ecological models. Inland Waters 5: 49– 56.

Hamilton, D. P. & S. G. Schladow, 1997. Prediction of water quality in lakes and reservoirs. Part I — Model description. Ecological Modelling 96: 91–110.

Hanson, P. C. 2007. A grassroots approach to sensor and science networks. Frontiers in Ecology and the Environment 5: 343–343.

Holmén, K. & P. Liss, 1984. Models for air-water gas transfer: an experimental investigation. Tellus B: Chemical and Physical Meteorology 36B: 92–100.

Honti, M., V. Istvanovics, P. A. Staehr, L. S. Brighenti, M. Zhu, & G. Zhu, 2016. Robust estimation of lake metabolism by coupling high frequency dissolved oxygen and chlorophyll fluorescence data in a Bayesian framework. Inland Waters 608–621.

Idrizaj, A., A. Laas, U. Anijalg & P. Nõges, 2016. Horizontal differences in ecosystem metabolism of a large shallow lake. Journal of Hydrology 535: 93–100.

Ingildsen, P. 2002. Realising full-scale control in wastewater treatment systems using in situ nutrient sensors. PhD thesis, Lund University, Sweden.

Jaagus, J. & A. Kull, 2011. Changes in surface wind directions in Estonia during 1966–2008 and their relationships with large-scale atmospheric circulation. Estonian Journal of Earth Sciences 60: 220.

Jennings, E., S. Jones, L. Arvola, P. A. Staehr, E. Gaiser, I. D. Jones, K. C. Weathers, G. A. Weyhenmeyer, C.-Y. Chiu & E. De Eyto, 2012. Effects of weather-related episodic events in lakes: an analysis based on high-frequency data: episodic events in lakes. Freshwater Biology 57: 589–601.

44 Jöhnk, K. D., J. Huisman, J. Sharples, B. Sommeijer, P. M. Visser & J. M. Stroom, 2008. Summer heatwaves promote blooms of harmful cyanobacteria. Global Change Biology 14: 495–512.

Johnson, M. S., M. F. Billett, K. J. Dinsmore, M. Wallin, K. E. Dyson & R. S. Jassal, 2010. Direct and continuous measurement of dissolved carbon dioxide in freshwater aquatic systems-method and applications. Ecohydrology 3: 68–78.

Jones, R. N., T. A. McMahon & J. M. Bowler, 2001. Modelling historical lake levels and recent climate change at three closed lakes, Western Victoria, Australia (c.1840–1990). Journal of Hydrology 246: 159– 180.

Jones, S. E., C.-Y. Chiu, T. K. Kratz, J.-T. Wu, A. Shade, & K. D. McMahon, 2008. Typhoons initiate predictable change in aquatic bacterial communities. Limnology and Oceanography 53: 1319– 1326.

Jonsson, A., J. Karlsson & M. Jansson, 2003. Sources of carbon dioxide supersaturation in clearwater and humic lakes in Northern Sweden. Ecosystems 6: 224–235.

Karakaya, N., F. Evrendilek, & K. Güngör, 2011. Modeling and validating long-term dynamics of diel dissolved oxygen with particular reference to pH in a temperate shallow lake (Turkey). CLEAN - Soil, Air, Water 39: 966–971.

Kerimoglu, O. & K. Rinke, 2013. Stratification dynamics in a shallow reservoir under different hydro-meteorological scenarios and operational strategies. Water Resources Research 49: 7518–7527.

Kikas, V. & U. Lips, 2016. Upwelling characteristics in the Gulf of Finland (Baltic Sea) as revealed by Ferrybox measurements in 2007–2013. Ocean Science 12: 843–859.

Kimura, N., W.-C. Liu, C.-Y. Chiu & T. K. Kratz, 2014. Assessing the effects of severe rainstorm-induced mixing on a subtropical, subalpine lake. Environmental Monitoring and Assessment 186: 3091–3114.

45 King, D. A. 2004. The scientific impact of nations. Nature 430: 311– 316.

Kling, G. W., G. W. Kipphut & M. C. Miller, 1991. Arctic lakes and streams as gas conduits to the atmosphere: implications for tundra carbon budgets. Science 251: 298–301.

Klug, J. L., D. C. Richardson, H. A. Ewing, B. R. Hargreaves, N. R. Samal, D. Vachon, D. C. Pierson, A. M. Lindsey, D. M. O’Donnell, S. W. Effler & K. C. Weathers, 2012. Ecosystem effects of a tropical cyclone on a network of lakes in northeastern North America. Environmental Science & Technology 46: 11693– 11701.

Kulbe, T., D. M. Livingstone, P. Guilizzoni & M. Sturm, 2008. The use of long-term, high-frequency, automatic sampling data in a comparative study of the hypolimnia of two dissimilar Alpine lakes. Verhandlungen, Internationale Vereinigung für Theoretische und Angewandte Limnologie 30: 371–376.

Laas, A., P. Nõges, T. Kõiv & T. Nõges, 2012. High-frequency metabolism study in a large and shallow temperate lake reveals seasonal switching between net autotrophy and net heterotrophy. Hydrobiologia 694: 57–74.

Lehner, B. & P. Döll, 2004. Development and validation of a global database of lakes, reservoirs and wetlands. Journal of Hydrology 296: 1–22.

Lindfors, A., K. Rasmus & N. Strömbeck, 2005. Point or pointless – quality of ground data. International Journal of Remote Sensing 26: 415–423.

Lips, U., V. Kikas, T. Liblik & I. Lips, 2016. Multi-sensor in situ observations to resolve the sub-mesoscale features in the stratified Gulf of Finland, Baltic Sea. Ocean Science 12: 715–732.

Maberly, S. C. 1996. Diel, episodic and seasonal changes in pH and concentrations of inorganic carbon in a productive lake. Freshwater Biology 35: 579–598.

46 MacIntyre, S., J. R. Romero & G. W. Kling, 2002. Spatial-temporal variability in surface layer deepening and lateral advection in an embayment of Lake Victoria, East Africa. 47: 656–671.

Magadza, C. H. D. 1994. Climate change: some likely multiple impacts in Southern Africa. Food Policy 19: 165–191.

Marcé, R., B. Obrador, J.-A. Morguí, J. Lluís Riera, P. López & J. Armengol, 2015. Carbonate weathering as a driver of CO2 supersaturation in lakes. Nature Geoscience 8: 107–111.

Neal, C., W. House & K. Down, 1998. An assessment of excess carbon dioxide partial pressures in natural waters based on pH and alkalinity measurements. Science of The Total Environment 210– 211: 173–185.

Obrador, B., P. A. Staehr & J. P. C. Christensen, 2014. Vertical patterns of metabolism in three contrasting stratified lakes. Limnology and Oceanography 59: 1228–1240.

O’Reilly, C. M., S. R. Alin, P.-D. Plisnier, A. S. Cohen & B. A. McKee, 2003. Climate change decreases aquatic ecosystem productivity of Lake Tanganyika, Africa. Nature 424: 766–768.

Oulman, C. S. & E. R. Baumann, 1956. A colorimetric method for determing dissolved oxygen. Sewage and Industrial Wastes 28: 1461–1465.

Pernica, P. & M. Wells, 2012. Frequency of episodic stratification in the near surface of Lake Opeongo and other small lakes. Water Quality Research Journal of Canada 47: 227.

Pernica, P., M. G. Wells & S. MacIntyre, 2014. Persistent weak thermal stratification inhibits mixing in the epilimnion of north-temperate Lake Opeongo, Canada. Aquatic Sciences 76: 187–201.

14 Peterson, B. J. 1980. Aquatic primary productivity and the C-CO2 method: a history of the productivity problem. Annual Review of Ecology and Systematics 11: 359–385.

Pierson, D. C., G. A. Weyhenmeyer, L. Arvola, B. Benson, T. Blenckner, T. Kratz, D. M. Livingstone, H. Markensten, G. Marzec, K.

47 Pettersson & K. Weathers, 2011. An automated method to monitor lake ice phenology: automated monitoring of lake ice. Limnology and Oceanography: Methods 9: 74–83.

Poff, N. L., M. M. Brinson & J. W. Day, 2002. Aquatic ecosystems and global climate change. Pew Center on Global Climate Change, Arlington, U.S.A.

Poole, R. & J. Morrow, 1977. Improved galvanic dissolved oxygen sensor for activated sludge. Journal Water Pollution Control Federation 49: 422–428.

Porter, J., P. Arzberger, H.-W. Braun, P. Bryant, S. Gage, T. Hansen, P. Hanson, C.-C. Lin, F.-P. Lin, T. Kratz, W. Michener, S. Shapiro & T. Williams, 2005. Wireless sensor networks for ecology. BioScience 55: 561.

Porter, J. H., E. Nagy, T. K. Kratz, P. Hanson, S. L. Collins & P. Arzberger, 2009. New eyes on the world: advanced sensors for ecology. BioScience 59: 385–397.

Prairie, Y. T., D. F. Bird & J. J. Cole, 2002. The summer metabolic balance in the epilimnion of southeastern Quebec lakes. Limnology and Oceanography 47: 316–321.

Quaranta, M., S. M. Borisov & I. Klimant, 2012. Indicators for optical oxygen sensors. Bioanalytical Reviews 4: 115–157.

R Development Core Team, 2014. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, http://www.R-project.org.

Rantajärvi, E., R. Olsonen, S. Hällfors, J.-M. Leppänen & M. Raateoja, 1998. Effect of sampling frequency on detection of natural variability in phytoplankton: unattended high-frequency measurements on board ferries in the Baltic Sea. ICES Journal of Marine Science 55: 697–704.

Read, J. S., D. P. Hamilton, A. R. Desai, K. C. Rose, S. MacIntyre, J. D. Lenters, R. L. Smyth, P. C. Hanson, J. J. Cole, P. A. Staehr, J. A. Rusak, D. C. Pierson, J. D. Brookes, A. Laas & C. H. Wu, 2012.

48 Lake-size dependency of wind shear and convection as controls on gas exchange. Geophysical Research Letters 39: L09405.

Read, J. S., D. P. Hamilton, I. D. Jones, K. Muraoka, L. A. Winslow, R. Kroiss, C. H. Wu & E. Gaiser, 2011. Derivation of lake mixing and stratification indices from high-resolution lake buoy data. Environmental Modelling & Software 26: 1325–1336.

Riera, J. L., J. E. Schindler & T. K. Kratz, 1999. Seasonal dynamics of carbon dioxide and methane in two clear-water lakes and two bog lakes in northern Wisconsin, U.S.A. Canadian Journal of Fisheries and Aquatic Sciences 56: 265–274.

Rouse, W. R., M. S. V. Douglas, R. E. Hecky, A. E. Hershey, G. W. Kling, L. Lesack, P. Marsh, M. Mcdonald, B. J. Nicholson, N. T. Roulet & J. P. Smol, 1997. Effects of climate change on the freshwaters of arctic and subarctic North America. Hydrological Processes 11: 873–902.

Sadro, S. & J. M. Melack, 2012. The effect of an extreme rain event on the biogeochemistry and ecosystem metabolism of an oligotrophic high-elevation lake. Arctic, Antarctic, and Alpine Research 44: 222–231.

Sadro, S., J. M. Melack & S. MacIntyre, 2011a. Spatial and temporal variability in the ecosystem metabolism of a high-elevation lake: integrating benthic and pelagic habitats. Ecosystems 14: 1123– 1140.

Sadro, S., J. M. Melack & S. MacIntyre, 2011b. Depth-integrated estimates of ecosystem metabolism in a high-elevation lake (Emerald Lake, Sierra Nevada, California). Limnology and Oceanography 56: 1764–1780.

Salmaso, N. 2005. Effects of climatic fluctuations and vertical mixing on the interannual trophic variability of Lake Garda, Italy. Limnology and Oceanography 50: 553–565.

Saloranta, T. M. & T. Andersen, 2007. MyLake—A multi-year lake simulation model code suitable for uncertainty and sensitivity analysis simulations. Ecological Modelling 207: 45–60.

49 Shade, A., C.-Y. Chiu & K. D. McMahon, 2010. Seasonal and episodic lake mixing stimulate differential planktonic bacterial dynamics. Microbial Ecology 59: 546–554.

Smith, C. G., J. E. Cable & J. B. .Martin, 2008. Episodic high intensity mixing events in a subterranean estuary: effects of tropical cyclones. Limnology and Oceanography 53: 666–674.

Song, K., M. A. Xenopoulos, J. M. Buttle, J. Marsalek, N. D. Wagner, F. R. Pick & P. C. Frost, 2013. Thermal stratification patterns in urban ponds and their relationships with vertical nutrient gradients. Journal of Environmental Management 127: 317–323.

Spigel, R. H. & G. W. Coulter, 1996. Comparison of hydrology and physical limnology of the East African great lakes: Tanganyika, Malawi, Victoria, Kivu and Turkana (with reference to some North American Great Lakes). In Johnson, T. C. & E. O. Odada (eds), The limnology, climatology and paleoclimatology of the East African lakes. Gordon and Breach Science Publishers, Amsterdam: 103–138.

Staehr, P. A., D. Bade, M. C. Van de Bogert, G. R. Koch, C. Williamson, P. Hanson, J. J. Cole & T. Kratz, 2010. Lake metabolism and the diel oxygen technique: state of the science. Limnology and Oceanography: Methods 8: 628–644.

Staehr, P. A., J. P. A. Christensen, R. D. Batt & J. S. Read, 2012. Ecosystem metabolism in a stratified lake. Limnology and Oceanography 57: 1317–1330.

Storey, M. V., B. van der Gaag & B. P. Burns, 2011. Advances in on-line drinking water quality monitoring and early warning systems. Water Research 45: 741–747.

Suursaar, Ü. & T. Kullas, 2006. Inﬂuence of wind climate changes on the mean sea level and current regime in the coastal waters of west Estonia, Baltic Sea. Oceanologia 48: 361–383.

Swan, M. 2013. The Quantified Self: Fundamental Disruption in Big Data Science and Biological Discovery. Big Data 1: 85–99.

50 Taillefert, M., G. W. Luther & D. B. Nuzzio, 2000. The application of electrochemical tools for in situ measurements in aquatic systems. Electroanalysis 12: 401–412.

Tanentzap, A. J., Y. Norman D., B. Keller, R. Girard, J. Heneberry, J. M. Gunn, D. P. Hamilton & P. A. Taylor, 2008. Cooling lakes while the world warms: Effects of forest regrowth and increased dissolved organic matter on the thermal regime of a temperate, urban lake. Limnology and Oceanography 53: 404–410.

Tobin, I., R. Vautard, I. Balog, F.-M. Bréon, S. Jerez, P. M. Ruti, F. Thais, M. Vrac & P. Yiou, 2015. Assessing climate change impacts on European wind energy from ENSEMBLES high-resolution climate projections. Climatic Change 128: 99–112.

Townsend, S. A., I. T. Webster & J. H. Schult, 2011. Metabolism in a groundwater-fed river system in the Australian wet/dry tropics: tight coupling of photosynthesis and respiration. Journal of the North American Benthological Society 30: 603–620.

Tsai, J.-W., T. K. Kratz, P. C. Hanson, J.-T. Wu, W. Y. B. Chang, P. W. Arzberger, B.-S. Lin, F.-P. Lin, H.-M. Chou & C.-Y. Chiu, 2008. Seasonal dynamics, typhoons and the regulation of lake metabolism in a subtropical humic lake. Freshwater Biology 53: 1929–1941.

Tsai, J.-W., T. K. Kratz, J. A. Rusak, W.-Y. Shih, W.-C. Liu, S.-L. Tang & C.-Y. Chiu, 2016. Absence of winter and spring monsoon changes water level and rapidly shifts metabolism in a subtropical lake. Inland Waters 6: 295–307.

Vachon, D. & P. A. del Giorgio, 2014. Whole-lake CO2 dynamics in response to storm events in two morphologically different lakes. Ecosystems 17: 1338–1353.

Vachon, D., J.-F. Lapierre & P. A. del Giorgio, 2016. Seasonality of photochemical dissolved organic carbon mineralization and its relative contribution to pelagic CO2 production in northern lakes. Journal of Geophysical Research: Biogeosciences 121: 864–878.

Vachon, D. & Y. T. Prairie, 2013. The ecosystem size and shape dependence of gas transfer velocity versus wind speed

51 relationships in lakes. Canadian Journal of Fisheries and Aquatic Sciences 70: 1757–1764.

Van de Bogert, M. C., D. L. Bade, S. R. Carpenter, J. J. Cole, M. L. Pace, P. C. Hanson & O. C. Langman, 2012. Spatial heterogeneity strongly affects estimates of ecosystem metabolism in two north temperate lakes. Limnology and Oceanography 57: 1689–1700.

Vautard, R., J. Cattiaux, P. Yiou, J.-N. Thépaut & P. Ciais, 2010. Northern Hemisphere atmospheric stilling partly attributed to an increase in surface roughness. Nature Geoscience 3: 756–761.

Volk, C., E. Dundore, J. Schiermann & M. LeChevallier, 2000. Practical evaluation of iron corrosion control in a drinking water distribution system. Water Research 34: 1967–1974.

Vos, R. ., J. H. . Hakvoort, R. W. . Jordans & B. . Ibelings, 2003. Multiplatform optical monitoring of eutrophication in temporally and spatially variable lakes. Science of The Total Environment 312: 221–243.

Wanklyn, J. A. 1907. Water-analysis, a practical treatise on the expremimentation of portable water. Kegan Paul, Trench, Trübner, and Co. Ltd.: 131–134.

Watras, C. J., K. A. Morrison, N. R. Lottig, T. K. Kratz & R. Smith, 2016. Comparing the diel cycles of dissolved organic matter fluorescence in a clear-water and two dark-water Wisconsin lakes: potential insights into lake metabolism. Canadian Journal of Fisheries and Aquatic Sciences 73: 65–75.

Wetzel, R. G. 2001. Limnology: lake and river ecosystems. Academic Press, San Diego.

Weyhenmeyer, G. A., S. Kosten, M. B. Wallin, L. J. Tranvik, E. Jeppesen & F. Roland, 2015. Significant fraction of CO2 emissions from boreal lakes derived from hydrologic inorganic carbon inputs. Nature Geoscience 8: 933–936.

Wilkinson, G. M., C. D. Buelo, J. J. Cole & M. L. Pace, 2016. Exogenously produced CO2 doubles the CO2 efflux from three

52 north temperate lakes. Geophysical Research Letters 43: 1996– 2003.

Winkler, L. W. 1888. Die Bestimmung des im Wasser gelösten Sauerstoffes. Berichte der deutschen chemischen Gesellschaft 21: 2843–2854.

Winslow, L. A., J. A. Zwart, R. D. Batt, H. Dugan, R. I. Woolway, J. Corman, P. C. Hanson & J. S. Read, 2016. LakeMetabolizer: an R package for estimating lake metabolism from free-water oxygen using diverse statistical models. Inland Waters 6: 622–636.

Woolway, R. I., S. C. Maberly, I. D. Jones & H. Feuchtmayr, 2014. A novel method for estimating the onset of thermal stratification in lakes from surface water measurements. Water Resources Research 50: 5131–5140.

Yang, Y., 2015. Phytoplankton and Physical Disturbance: Seasonal dynamics in temperate Lake Erken, Sweden. PhD thesis, Uppsala University, Sweden.

YSI Inc., 2006. 6-Series Multiparameter Water Quality Sondes User Manual (item #069300). Yellow Springs, USA.

53 SUMMARY IN ESTONIAN

PIDEVMÕÕTMISED – UUS LÄHENEMINE LIMNOLOOGIAS

Viimase ajani on enamus järvede seiretöid/uuringuid toimunud traditsioonilisi, aga oma olemuselt ajamahukaid ja kulukaid proovivõtumeetodeid kasutades. Tihtipeale ei anna sellised uuringud piisavat ettekujutust mõõdetud näitajate ruumilisest jaotusest ja ajalisest muutlikkusest. Suure arenguhüppe mitmetes limnoloogilistes uurimisvaldkondades on kaasa toonud pidevmõõtmiste kasutuselevõtt, mis annavad võimaluse uurida järvi sellistes ajaskaalades, mis traditsiooniliste proovivõtumeetoditega pole võimalikud. Sensortehnoloogiate töökindlus, täpsus ja autonoomsus on kiiresti tõusnud, hinnad aga langenud, mis muudab nad laialdaselt kättesaadavaks. Lisaks kvaliteetsete andmete pakkumisele vabastab automatiseeritud mõõtesüsteemide kasutamine uurijad rutiinsetest välitöödest võimaldades aega efektiivsemalt kasutada. Käesoleva töö peamiseks eesmärgiks on anda ülevaade pidevmõõtmiste praegustest limnoloogilistest kasutusvaldkondadest, eesmärkidest ja kasutuse ulatusest maailmas ning hinnata võimalikke tulevikuväljavaateid. Elust võetud näidetena on töösse lisatud kaks uuringut, millest ühes kasutasime veetemperatuuri ja ilmastikunäitajate pidevmõõtmisi kliimamuutuse mõju modelleerimiseks järve termilisele kihistumisele ning teises analüüsisime järvede tüübiomaseid gaasirežiimi erinevusi.

Ülevaateartikli (I) eesmärgiks oli anda ülevaade pidevmõõtmiste kasutamisest limnoloogias läbi ajaloo, kuid eelkõige keskenduti kirjanduse otsingul 21. sajandil publitseeritule. Kuna selle artikli puhul on tegemist ülevaateartikliga, siis otsest hüpoteesi ei püstitatud. Üritasime hinnata pidevmõõtmiste limnoloogias rakendamise geograafiat ja arenguid viimase 15 aasta jooksul, kategoriseerida senised ja lootustandvad tulevikuvaldkonnad ja tuua esile pidevmõõtmiste eelised, millel põhineb nende aina kasvav tähtsus limnoloogilistes uuringutes ja seires, kus mõnes valdkonnas (nt metabolismi uuringutes) on automatiseeritud mõõtmised juba asendamas või asendanud traditsioonilisi välimõõtmiste meetodeid. Keskendusime oma ülevaates peamiselt mõõtmispoidel kasutatavatele sensoritele ja jätsime oma

54 analüüsist teadlikult välja mõned pidevmõõtmistega seotud tehnoloogiad nagu hüdrolokatsiooniseadmed (sonarid ja kajaloed), Doppleri efektil põhinevad hoovusemõõtjad (ADCP) ja kaugseire seadmed (nt lidarid). Meie analüüs näitas, et pidevmõõtmiste kasutamine limnoloogilistes uuringutes on viimase 15 aasta jooksul märkimisväärselt tõusnud. Enim on levinud veetemperatuuri ja lahustunud hapniku (DO) mõõtmine sensorite abil, kuid aina enam kasutust on leidmas ka süsihappegaasi (CO2) sensorid. Üksikasjalisi veetemperatuuri andmeid vajatakse väga erinevate klimatoloogia, hüdrofüüsika, hüdrokeemia ja hüdrobioloogia küsimuste lahendamiseks ja sensormõõtmised on temperatuurimõõtmistes saavutanud silmapaistva töökindluse, vastupidavuse ja autnoomsuse. Lahustunud hapniku sensorite laialdane rakendamine on viinud järvede ainevahetuse uuringud täiesti uuele tasemele ja suuresti välja vahetanud varasemad proovivõtu meetodid. Sensortehnoloogiad võimaldavad üksikute punktmõõtmiste asemel aina rohkem kasutada mõõtmisvõrgustikke, mis toovad esile näitajate ruumilise jaotuse. Kirjandusest lähtuvalt võib täheldada aina enam ka liikumist profileerivate poide suunas, mis lisavad järve andmetele nii olulise vertikaalse mõõtme. Sensorite edulugu kindlasti jätkub. Hoolimata sellest on ilmselge, et automaatsed ja traditsioonilised proovivõtumeetodid täiendavad teineteist ning pidevmõõtmiste laialdane kasutuselevõtt ei asenda kunagi täielikult traditsioonilisi mõõtmisi, milles teadlase kogenud silmal ja kaemusel on tähtis roll.

Kliimamuutused mängivad olulist rolli ka järvede ökosüsteemides. Võrreldes viimase 30 aasta tuulekiiruse andmeid erinevates Eesti ilmajaamades, täheldasime tuulekiiruse langust, mille üheks avalduseks on kevadsuvisel perioodil tõusnud tuulevaiksete päevade arv. Püstitasime hüpoteesi, et pikaajaline tuulekiiruse langus on põhjustanud märkimisväärsed muutused polümiktilise Võrtsjärve episoodilise termilise kihistumise esinemissageduses ja tugevuses. Ülesande lahendasime kahes etapis (II). Esmalt, kasutades „MyLake“ mudelit ja aastatel 2013-2015 tiheda sagedusega mõõdetud ilmastikuandmeid, modelleerisime veetemperatuuri jaotust ja kihistumise episoode Võrtsjärves. Teises etapis kasutasime kalibreeritud mudelit ajalooliste ilmastikuandmetega, et leida, kas vaadeldud ulatuslik tuulte taltumine viimase 30 aasta jooksul võib olla mõjutanud termilise kihistumise tugevust ja esinemissagedust Võrtsjärves (II). Uurimise tulemusena

55 saime teada, et järves, mida seni on peetud pidevalt segunenuks, on tõenäoliselt alati esinenud termilise kihistumise episoode, mis aga tuulevaikuse süvenemisega on oluliselt sagenenud ja nende kestus pikenenud. Uuring näitas ka, et õhutemperatuuri tõusul ei ole kihistumise muutustele olnud seni nimetamisväärset mõju.

Teises uuringus võrreldi 8 parasvöötme järvetüübi gaasirežiimi, lahustunud hapniku ja süsihappegaasi sügavusjaotuse alusel, mis saadi in situ pidevmõõtmiste tulemusena (III). Eesmärgiks oli vaadata, kuidas järvede kesksuvine gaasirežiim sõltub nende tüübiomadustest ja toitelisusest ja millisel määral on DO ja CO2 dünaamika omavahel seotud. Järvede peamiste ainevahetusprotsesside – fotosünteesi ja hingamise – vastassuunalist mõju arvestades eeldasime tugevat pöördvõrdelist sõltuvust nende gaaside dünaamikates. Samuti eeldasime, et termilise kihistumise korral esinevad gaasijaotuses tugevad vertikaalsed gradiendid ning et termiliselt kihistumata järvedes on ka gaasid ühtlaselt jaotunud. Mõõtmisperioodi jooksul täheldasime märkimisväärseid erinevusi nii järvede termilises kihistumises kui ka DO ja CO2 ööpäevases dünaamikas. Need erinevused olid peamiselt tingitud järvede erinevast toitelisusest ja vähemal määral ka järvede tüübispetsiifilistest omadustest nagu morfomeetria ja vee keemia. Uuringu tulemused üldiselt toetasid meie hüpoteesi DO ja CO2 pöördvõrdelisest seosest, kuid vastupidiselt oodatule täheldasime tugevaid DO ja CO2 gradiente ka termilise kihistumise puudumisel. Gradiendid kujunesid ühelt poolt orgaanilise aine lagunemise tulemusena, mis põhjustas sette kõrge hapnikutarbe järve põhjakihtides ning teisalt intensiivse fotosünteesi tõttu, mis kasutab CO2 ja rikastab vett lahustunud hapnikuga pinnakihis. Uuring kinnitas vertikaalsete CO2 mõõtmiste potentsiaali täiustamaks arusaama süsiniku dünaamikast veekeskkonnas.

56 ACKNOWLEDGEMENTS

“A scientist is a child who’s never grown up” Neil deGrasse Tyson

I should start with a phrase “Never say never”. Four years ago I was asked, if I’m planning to do a PhD. My answer was a fast and quick “No”. Yet one day I received a phone call from Alo with an offer for a 4 year adventure and my answer then was “Yes”. And here I am - finishing my PhD, and I have a lot of people to thank for this journey.

I owe my deepest gratitude to my two supervisors Alo Laas and Peeter Nõges. Alo, you offered me this 4 year journey, all your support, ideas, guidance, enthusiasm and the craziest adventures of all and Peeter, I am forever grateful for your guidance and support, for reading everything I wrote, making it red as hell and saying “you did well”.

I wish to express my sincere thanks to my co-authors of all the published papers. My special thanks will go to Iestyn and Ian, without you this thesis wouldn’t be possible. I’m grateful to people in our institute and co-workers for the fun coffee breaks. I am especially thankful to Sirje, Ingmar and Fabien, for always asking how things were going and encouraging me to pursue this path. I want to thank Tiina Nõges, for the support through my PhD studies.

I want to thank my friends Merit, Katrin, Maili, Ronald and Yang for listening my complaining and encouraging me to go forward. I want to thank all of Imaginaarkuivikud, for inspiration and fun gatherings.

The warmest gratitude goes to my mom. You have supported me through the seemingly endless studies and listened endless talks on topics you barely understood. Urmeli, Jaan and Kristjan Kalev have been there for me, whenever I needed. I’m happy that I have Urmas in my life, whose endless love, support and motivation have helped me through thick and thin.

57 ORIGINAL PUBLICATIONS I

58 Meinson, Pille; Idrizaj, Agron; Nõges, Peeter; Nõges, Tiina; Laas, Alo (2016). Continuous and high-frequency measurements in limnology: history, applications, and future challenges. Environmental Reviews 24, 52-62.

59 52 REVIEW

Continuous and high-frequency measurements in limnology: history, applications, and future challenges Pille Meinson, Agron Idrizaj, Peeter Nõges, Tiina Nõges, and Alo Laas

Abstract: Over the past 15 years, an increasing number of studies in limnology have been using data from high-frequency measurements (HFM). This new technology offers scientists a chance to investigate lakes at time scales that were not possible earlier and in places where regular sampling would be complicated or even dangerous. This has allowed capturing the effects of episodic or extreme events, such as typhoons on lakes. In the present paper we review the various fields of limnology, such as monitoring, studying highly dynamic processes, lake metabolism studies, and budget calculations, where HFM has been applied, and which have benefitted most from the application. Our meta-analysis showed that more than half of the high-frequency studies from lakes were made in North America and Europe. The main field of application has been lake ecology (monitoring, lake metabolism) followed by physical limnology. Water temperature and dissolved oxygen have been the most universal and commonly measured parameters and we review the various study purposes for which these measurements have been used. Although a considerable challenge for the future, our review highlights that broadening the spatial scale of HFM would substantially broaden the applicability of these data across a spectrum of different fields.

Key words: lake metabolism, temporal variability, spatial variability, extreme events, hydrodynamics. Résumé : Au cours des 15 dernières années, on observe un nombre croissant d’études en limnologie ayant utilisé des données provenant de mesures a` hautes fréquences (HFM). Cette nouvelle technologie offre aux scientiﬁques la possibilité d’étudier les lacs a` des échelles de temps qui étaient impossibles auparavant et dans des endroits où l’échantillonnage serait compliqué ou même dangereux. Ceci a permis de saisir les effets d’événement épisodiques ou extrêmes, tels que des typhons sur des lacs. Les auteurs passent ici en revue différents champs de la limnologie tels que le suivi, l’étude de processus hautement dynamiques, l’étude du métabolisme des lacs ainsi que le calcul des budgets, où on a appliqué la HFM et qui ont le plus bénéﬁcié de cette application. La méta-analyse des auteurs montre que plus de la moitié des études conduites en hautes fréquences dans des lacs l’ont été en Amérique du Nord. Le principal champ d’application a concerné l’écologie des lacs (suivi, métabolisme des lacs), suivi de la limnologie physique. La température et l’oxygène dissout constituent les paramètres les plus universels et communément mesurés et les auteurs passent en revue les objectifs des diverses études pour lesquels ces mesures ont été utilisées. Bien que ceci

For personal use only. constitue un déﬁ considérable pour le futur, cette revue souligne que l’élargissement de l’échelle spatiale des HFM élargirait substantiellement l’applicabilité des données pour un ensemble de champs différents. [Traduit par la Rédaction]

Mots-clés : métabolisme des lacs, variabilité temporelle, variabilité spatiale, évènements extrêmes, hydrodynamiques.

1. Introduction automatic high-frequency measurements (HFM). Even though automatic recording is vulnerable to vandalism, biofouling, Today, global environmental change and the increasing exploi- and occasional failures in the systems, and maintenance issues tation of ecosystem services by man has created complex multiple may causes gaps in time series data (Dur et al. 2007), a fast transi- stressor situations for most inland water bodies (Ormerod et al. tion to automatic HFM in environmental monitoring systems is 2010). This has obviated the need for more and better monitoring inevitable. data and is a prerequisite for understanding the often synergistic, The need for continuous observations has led to automating cumulative, and non-linear impacts of combined stressors (Brown measurements since the early times of limnology. In the ﬁrst et al. 2013) and for making adequate management decisions report of the Indiana University Turkey Lake Biological Station (Hering et al. 2015). Until recently, the majority of standard lake (the ﬁrst inland biological station in America), its director,

Environ. Rev. Downloaded from www.nrcresearchpress.com by Miss Pille Meinson on 03/02/16 monitoring programs were based on manual in situ measure- C.H. Eigenmann, mentioned, among other equipment, an “auto- ments that can be time-consuming and costly to procure and matic recording apparatus to observe seiches” (Eigenmann 1895, often lack both the necessary spatial coverage as well as an appro- p. 207). In the same year, Warren and Whipple (1895, p. 639) in- priate sampling frequency (Vos et al. 2003). The latter is especially troduced the thermophone, “a new instrument for obtaining the important for detecting the effects of hydrology or weather- temperature of a distant or inaccessible place,” and provided related episodic events, from which biological consequences can some temperature profiles measured in Lake Cochituate, Mass. range from short-term, reversible changes to those that are more Simple physical parameters, such as lake water levels and water persistent (Jennings et al. 2012). Time and reliability issues can be temperature, were among the first limnological variables for efficiently addressed by replacing manual measurements with which measurement could be automated. Parameters recorded in

Received 26 May 2015. Accepted 5 October 2015. P. Meinson,* A. Idrizaj, P. Nõges, T. Nõges, and A. Laas. Centre for Limnology, Institute of Agricultural and Environmental Sciences, Estonian University of Life Sciences, 61117 Rannu, Tartu County, Estonia. Corresponding author: Pille Meinson (e-mail: [email protected]). *Present address: Kreutzwaldi 5, Tartu 51014, Estonia.

Environ. Rev. 24: 52–62 (2016) dx.doi.org/10.1139/er-2015-0030 Published at www.nrcresearchpress.com/er on 6 October 2015.

60 Meinson et al. 53

physical meteorology, such as air temperature, humidity, wind confront the information management and analytical challenges speed and direction, and the amount of precipitation served as posed by massive volumes of data, while Porter and Lin (2013) important background data supporting lake research (Porter et al. specifically focused on available hybrid wireless sensor network 2005). However, soon the development of electrochemistry — the technologies. Johnson et al. (2007) reviewed the chemical sensing invention of the glass electrode by Cremer (1906) and the intro- capabilities with a special focus on such chemical sensor net- duction of the concept of pH by Sørensen (1909) — gave an impe- works that can be deployed on autonomous platforms in aquatic tus to the development of various chemical sensors initially environments and then operated without significant human in- enabling automatic recording of a number of single charged ions, tervention for extended periods. Crawford et al. (2015) analyzed + + + such as H ,Na ,NH4 , proved soon as limnological key variables. the potential of using advanced sensors to investigate spatial vari- The oxygen electrode widely used in aquatic studies measuring ability in biogeochemistry and hydrology. Besides Porter et al. oxygen on a catalytic platinum surface was invented by Leland (2005), which gives a broad picture of HFM applications in envi- Clark in 1954, initially for blood gas analysis (Severinghaus and ronmental studies, some reviews give more specific insights into Astrup 1986). While chemical sensors still cannot fully compete to the use of high-frequency (HF) data in lakes. For example, with physical sensors regarding their cost or reliability, a variety Jennings et al. (2012) analyzed the potential of HFM for identifying of chemical sensing systems (e.g., for carbon dioxide, pH, and the effects of weather-related episodic events in lakes, whereas oxygen and a number of ions) are now continuously deployed in Staehr et al. (2010) reviewed the use of the diel oxygen technique aquatic environments (reviewed by Johnson et al. 2007). for studying lake metabolism. A big leap in environmental measurement and monitoring In the present paper we review various fields of limnology and technologies came with the transition from analogue to digital lake management that have benefitted or potentially may benefit technologies from the late 1990s to the early 2000s (Hilbert and from using HF measurements, in the hope of stimulating addi- López 2011). Thanks to fast technological development, the auton- tional use of this promising technology. We highlight the advan- omy of different measurement systems has increased. Using wire- tages of using HFM for various purposes compared to discrete or less systems allows retrieving data from the weather and water manual sampling, but also discuss the challenges. We put the monitoring stations in near-real time and without manually vis- main focus on HF data applications related to buoy and mooring iting the study site (Porter et al. 2005). The development of sensors systems and do not discuss other fields of HF technologies, such as has made progress by taking under consideration the protection SONAR, eco-sounding, ADCP or remote-sensing systems. We have from biofouling (Manov et al. 2004) using self-cleaning systems reviewed 154 papers using HFM in lakes and grouped the studies (e.g., wipers or pressure). After reviewing the recent develop- according to their limnological or lake management issues (i.e., ments in HFM systems, sensor technologies and networking in measurement objectives rather than technical parameters of the limnology, Crawford et al. (2015) emphasized the new and unex- study design). Although the list of papers is not exhaustive, it is pected insights into ecosystem processes that would have been representative and sufficiently large to give a broad overview of impossible with previous techniques. Nevertheless, the authors the various applications using HFM in limnology. show that despite the fact that sensor technology is becoming common in limnological research, current applications focus al- 2. Literature search most entirely on temporal patterns and variation while spatial For the literature search, we used the Google Scholar citation variability is rarely documented because of the high investment database. Queries were made using search terms “lake*” and costs for the spatial replication of such infrastructure. “high frequency data”. To focus on recent research only, we re-

For personal use only. The amount of data that can be generated from HFM is huge stricted the time window to the past 15 years (i.e., only articles even if only one set of sensors is deployed in a lake for time published between 2000 and 2015 were considered). Screening of periods of several days or months. Globally, the HFM data from the 1480 papers retrieved revealed the main fields of HFM appli- lakes add volume, velocity, variety, resolution, and relational na- cation for which we repeated more specific queries using as ture, the so-called three Vs and two Rs of big data (Kitchin 2013), search terms “lake*” and “high frequency” in combination with and so dealing with these data are challenging. At the same time, one of the following terms: “monitoring”, “early warning”, “me- more data may lead to more accurate analyses and the high reso- tabolism”, “budget”, “flux”, “migration*”, “sampling design”, “ex- lution allows insights into highly dynamic processes (Coloso et al. treme event*”, and “patchiness”. The first 100 results of each 2008; Schwientek et al. 2013). retrieval were screened for their relevance resulting in a total of Besides challenges related to costs of assessing spatial variabil- 1730 records. Further, we selected studies in which measurement ity and organizing, cleaning, and processing large datasets, other intervals of ≤60 min were used and the deployment time was at challenges exist for HFM including resolution and response time least 6 h. As we could not find any commonly agreed-upon defini- of certain electrochemical sensors (e.g., for pH and dissolved tion for HFM, we set the arbitrary 60 min upper limit for the oxygen (DO)) creating problems for using them in profiling instru- measurement interval considering this the coarsest resolution ments (Taillefert et al. 2000; Tengberg et al. 2006), and the photo- still allowing to follow the diurnal dynamics of processes. Papers Environ. Rev. Downloaded from www.nrcresearchpress.com by Miss Pille Meinson on 03/02/16 bleaching issues with fluorescence sensors (Johnson et al. 2007). using measurement intervals longer than1hordeployment times As Cushing (2013) noticed, the exponential growth in data ac- less than 6 h were excluded from our results. This screening quisition and progress in data documenting, archiving, valida- yielded a final list of 154 papers that met our criteria and were tion, and retrieval have made more (in quantity and diversity) and used as the basis of the present review. better (less “dirty”) data available to scientists. The Global Lake Ecological Observatory Network (GLEON) is a prime example of a 3. Review table lake and data network, sharing and interpreting high-resolution Relevant information from the papers was extracted to an Excel sensor data from a broad spectrum of lakes across the globe to table to enable future search by meta-analysis. Each article was understand, predict, and communicate the role and response of described in a separate row of the table (154 papers) starting with lakes in a changing global environment (Weathers et al. 2013). a full reference in the first column. A set of 11 parameters, some of Various aspects of using HFM in environmental sciences, in- them with predefined categories, was used to extract important cluding limnology, have been analyzed in a number of earlier information from the papers: review papers. Porter et al. (2005) reviewed some existing uses of wireless sensor networks in environmental sciences, possible ar- 1. Number of lakes and lake type including stratification (strat- eas of application, and the underlying technologies. Porter et al. ified, non-stratified) and trophic state (5 cat.) (2012) provided a synopsis of innovative approaches being used to 2. Geographic location (continent, 7 cat.)

Published by NRC Research Press

61 54 Environ. Rev. Vol. 24, 2016

Fig. 1. Split-off of HF studies by object, subject, and space–time scale. Geographical distribution (a) and ﬁeld of study (b), types of measured parameters (c), space–time scale (d), lake categories by stratiﬁcation (e), and trophic state (f). For personal use only.

3. Field of study (physical limnology, geochemical limnology or 2D horizontal, time series measured at one depth along a tran- lake ecology) sect (e.g., the FerryBox systems)

Environ. Rev. Downloaded from www.nrcresearchpress.com by Miss Pille Meinson on 03/02/16 4. Main purpose of applying HFM (12 cat.) 3D, time series measured at one depth over an area (e.g., poly- 5. Type of parameters measured (meteorological, hydrophysical/ gon measurements) or at several depths along a transect (e.g., optical, hydrochemical, biological) robot fish measurements) 6. Parameters measured (23 cat.) 7. Last year of data collection The table included three types of entries: 8. Year of publication 1. (Free) text columns were used for the full reference and the 9. Temporal scale of study (5 cat. from <1 day to multiannual) lake names, comments. Several initially free text variables, 10. Temporal resolution (measurement and aggregation interval, such as temporal resolution, measured parameters, geograph- minutes) ical distribution, and trophic state, were categorized after- 11. Space–time scale wards and included as multiple-choice variables. To describe the space–time scale of the measurement design, we 2. Numerical entries were used for publication year, length of used the following categories: publishing cycle, and temporal resolution of study. 3. Number 1 was used as a tick-mark denoting “Yes” for selecting 1D, time series measured at one station at a single depth one (or more) of the multi-choice columns under different 2D vertical, time series measured at one station in a vertical categories (year of study, temporal and spatial scale of study, profile (e.g., profiling buoys and sensor chains) geographical distribution, field and main purpose, trophic

Published by NRC Research Press

62 Meinson et al. 55

Fig. 2. Distribution of the reviewed literature by year of publication and period of data collection.

state and stratification, measured parameters, and temporal Fig. 3. Length of the period from the end of data collection to resolution of study). publication. Cells were left empty if the choice was “No”. For trophic state and stratification we added field “NA” (not applicable). 4. Meta-analysis The meta-analysis was based on numerical variables, such as the year of publication, and counting the numbers of entries in

For personal use only. categorical multiple-choice variables. To assess the signiﬁcance of differences between proportions of various categories within a population, a Z-test was used, available online at http://www. socscistatistics.com/. 5. Patterns in applying HFM in limnology From a geographical point of view, almost 40% of HF studies were undertaken in North America, which was 1.4 times more than in Europe (Fig. 1a) followed by Asia with 18%. Review papers including data from several continents made up 5% of the studies, whereas the smallest number of studies originated from Africa. The number of publications per year analyzed applying HFM had an increasing trend and the maximum was reached in 2012 (Fig. 2). In 2008, the number of HF studies was signiﬁcantly higher compared to both neighboring years (z-score > 2.3; p < 0.05). The The majority of studies using HFM were made in lake ecology

Environ. Rev. Downloaded from www.nrcresearchpress.com by Miss Pille Meinson on 03/02/16 citation report generated for our search terms by the ISI Web of (56%) covering ﬁelds, such as low-maintenance monitoring and Knowledge showed similar patterns, although the number of HF lake metabolism measurements, followed by physical limnology studies in 2008 was signiﬁcantly higher only compared to 2007 (28%) dealing with hydrological processes and different water (z-score > 2.2; p < 0.05) while the difference with 2009 remained movements (e.g., internal waves, currents, and seiches), and stud-

non-signiﬁcant. ies in geochemical limnology (sensor measurements of CO2, ni- Since the 1990s until 2009, there has been an increasing trend in trates, etc.; 16%) (Fig. 1b). the amount of collected HFM data represented in the publications A substantial part of measured parameters were meteorological analyzed. HFM data collected in 2005 and in the period from 2008 variables (44%; e.g., air temp, wind speed, etc.) (Fig. 1c). A notice-

to 2010 have been most represented in publications so far, able 30% of measured parameters were hydrochemical (e.g., CO2, whereas data collected during recent years are still in the prepa- etc.). Hydrophysical and optical parameters formed 19% of ration phase (Fig. 2). the parameters measured while biological parameters, such as Most often it took between 2 and 3 years after the study period chlorophyll a, phycocyanin, and phycoerythrin, were least until HFM data ﬁrst appeared in publications, although 1 year and studied (7%). 4–5 year publishing cycles were rather common too (Fig. 3). Al- We distinguished 12 main purposes for which HFM have been though older data were often used as part of the database, only a used at different frequencies (Fig. 4). Most often HFM have been few papers were entirely based on data older than 5 years. used for low-maintenance monitoring (e.g., water temperature)

Published by NRC Research Press

63 56 Environ. Rev. Vol. 24, 2016

Fig. 4. Frequency distribution by study purpose.

Fig. 5. Temporal scales used in studies. Fig. 6. Temporal resolution by measurement interval (minutes). For personal use only.

quality variables (conductivity, pH, turbidity). Chlorophyll a was and collecting meteorological background data. Lake metabolism more common in HFM programs than other biological parame- studies using HF data have been growing in number in recent ters (e.g., phycocyanin or phycoerythrin). Measurements of chem- years and occupied the third position. There are still rare topics ical parameters (e.g., NO3,PO4) were rare. where HFM are applied (e.g., developing sampling strategies or Almost 60% of the studied lakes were described as thermally calibrating remote-sensing data). stratified and 24% as unstratified, while in 17% of the lakes the Most often studies were performed in time scales from >1 month stratification remained ambiguous. Among the lakes with trophic

Environ. Rev. Downloaded from www.nrcresearchpress.com by Miss Pille Meinson on 03/02/16 to 1 year, being about twice as frequent as multi-annual studies state indicated, eutrophic and oligotrophic lakes formed equally and those lasting from >1 day to 1 month (Fig. 5). Diurnal and 40% followed by mesotrophic (15%), hypereutrophic (5%), and dys- shorter time scales were least used. trophic (3%) lakes (Figs. 1e and 1f). Most of the studies were based on single point HFM in vertical profiles (2D vertical in a space–time scale), 31% on single point – 6. Explanation for the patterns revealed single depth time series (1D), followed by profile (3D) and single According to our meta-analysis, more than two-thirds of the HF depth measurements along transects (2D horizontal) (Fig. 1d). studies in lakes were carried out either in North America or Eu- The length of the measurement interval used in studies had a rope (Fig. 1a) that, on one hand, reflects the global distribution of nearly bimodal distribution with peaks at <1 min and at 10 and lakes peaking in the northern temperate zone between 40°N and 15 min, but also 5, 30, and 60 minute intervals were widely used 70°N (Lehner and Döll 2004). Studies involving up to 25 lakes have (Fig. 6). In some papers HFM were aggregated, most often to 10 and been carried out both in the U.S.A. (Hanson et al. 2003; Langman 60 min or 24 h. et al. 2010) and in Europe (Staehr et al. 2012). On the other hand, Water temperature was the parameter most often measured, this distribution is strongly influenced by the global distribution while measurements of DO concentration took the second posi- of scientific research potential. As shown by King (2004) in a com- tion (Fig. 7). These two main parameters were followed by some prehensive review of the scientific impact of nations for the pe- meteorological (e.g., wind speed, air temperature, PAR) and water riod 1993–2001, USA and the EU15 of that time headed the list,

Published by NRC Research Press

64 Meinson et al. 57

Fig. 7. Parameters studied with HFM.

whereas Japan had the highest number of full-time researchers ment intervals <1 min (Fig. 5) are often used to study dynamic per 1000 employed. Nowadays, the contribution of China to world hydrological processes using sensitive water temperature sensors. science is in a fast growing phase (Zhou and Leydesdroff 2006). MacIntyre et al. (2002) measured water temperature every 30 s Among HF studies in lakes that we analyzed, Asian papers occu- with 14 sensors to study surface layer deepening and lateral ad- pied a decisive third position with 18%. Studies including data vection in Lake Victoria (Africa). Internal waves with periods of from several continents (termed by us as ‘Global’, e.g., Read et al. 5–45 min were generated during relaxation from wind forcing 2012; Solomon et al. 2013) constituted 5% of studies. and as the thermocline rapidly downwelled. HF water tempera- As our analysis revealed, more studies used 2D vertical mea- ture measurements have also been used in other large lakes for surements compared to just single depth sensors. One could ex- studying internal waves (e.g., Boegman et al. 2003; Lorke et al. pect the latter to be more common given their likely lower cost 2006; Lorke 2007), benthic boundary mixing (Hondzo and Haider, and greater reliability due to the lack of a cable winding system 2004) and to develop circulation models for lakes (Laval et al. often used to raise and lower sensors in the water column. Verti- 2003). Using HFM of temperature profiles for studying lake mixing For personal use only. cal 2D measurements involve not only profiling buoys (that are (Kulbe et al. 2008; Shade et al. 2010; Read et al. 2011a; Kimura et al. really expensive and therefore not so widely used), but also simple 2014; Pernica et al. 2014; Bertone et al. 2015) and stratification sensor chains (e.g., thermistor chains) and all cases where sensors processes (Pernica and Wells 2012; Song et al. 2013; Sullivan et al. were deployed at more than one depth, being a common practice 2013) has been common practice. in stratified lakes strongly predominating in the selected studies. Besides hydrological processes, HF temperature data allow cal- Another reason explaining the predominance of 2D studies is that culating heat flux and energy balance of lakes. Lakes affect the if some parameters were measured vertically and some at a single climate at scales ranging from local to global, but these effects are depth, we indicated the spatio-temporal dimension as 2D. Analys- often neglected in models (Ljungemyr et al. 1996). To improve the ing the use of 1D and 2D vertical data in lake metabolism calcula- understanding of the intra- and inter-annual dynamics of the heat tions, Staehr et al. (2012) pointed out that measurements with a balance components on a boreal lake, Nordbo et al. (2011) studied single sensor in the surface layer tend to overestimate gross pri- the thermal structure of a small lake in Finland over several years mary production and community respiration while measure- using HF temperature profile measurements to calculate the heat ments in two layers give more truthful results. A single sensor storage change of the lake. Heat and mass balance calculations are may overlook important heterogeneity within lake processes and also important for monitoring volcanoes with heated crater lakes, may not accurately represent system-wide values of metabolism

Environ. Rev. Downloaded from www.nrcresearchpress.com by Miss Pille Meinson on 03/02/16 as they reflect the mass and thermal fluxes from the volcanic (Van de Bogert et al. 2012). vents into the bottom of the lake, as an indication of the state of 6.1. Versatile and universal use of simple parameters the underlying volcano (Hurst et al. 2012). Because evaporation Our review of the published literature found that HFM of water can be substantially affected by the lake’s influence on the air- temperature were most common. High reliability, long auton- mass above it, HF air and water temperature measurements can omy, low maintenance need, and reasonable prices of tempera- help to specify the seasonality of evaporation in crater lakes ture sensors have certainly contributed to their wide application; (Redmond 2007; Hurst et al. 2012). Climate warming increases the however, the main reasons lie in the possibility of using temper- thermal stability of lakes with implications for chemical and ature as a marker of water masses and in the universal nature of biological processes (Weinberger and Vetter 2014). Using HF temperature as a controlling factor of chemical and biological moored thermistor records and meteorological data, Churchill processes. and Kerfoot (2007) studied the impact of surface heat flux and Because of the high specific heat content of water, it warms up wind on thermal stratification in Portage lake, Michigan. Pierson and cools down relatively slowly, which allows using water tem- et al. (2011) developed a simple method to automatically detect the perature as a marker of water masses to study various water move- presence of ice cover by continuously recording water tempera- ments (currents, internal waves, and seiches), but also mixing, ture just below the ice–water interface and just above the lake stagnation, and stratification of water masses. HFM with measure- bottom using moored temperature sensors.

Published by NRC Research Press

65 58 Environ. Rev. Vol. 24, 2016

Only by applying HF measurements has it become possible to nighttime mixing (Branco and Torgersen 2009). Other examples follow in detail the effects of extreme weather events on lakes. For of using HF DO data in lakes include its application as an indicator example, Jennings et al. (2012) analyzed the effects of weather- of the presence of biodegradable organic pollution (Ansa-Ansare related episodic events in seven lakes from Europe, USA, and et al. 2000), studying the efficiency of artificial destratification Taiwan based on high- and moderate-frequency data. They distin- (Read et al. 2011b), describing temporal dynamics of DO in a float- guished two classes of abiotic effects of weather events: the gen- ing leaved macrophyte bed (Goodwin et al. 2008), and assessing erally short-lived effects of storms affecting lake thermal the effect of summer warming on DO conditions (Wilhelm and structure and the more prolonged effects of high rainfall events Adrian 2008). on dissolved organic matter levels and water clarity. Episodic Other sensors often included in a standard set of water moni- events, such as hurricanes and storms have a great impact on both toring probes are pH and conductivity sensors. The relationship coastal areas (Chang and Dickey 2001) and lakes (Tsai et al. 2008, between DO and pH depends upon the system and the time scale: 2011; Shade et al. 2010) by disrupting the thermal stratification in bottom waters of relatively deep regions in Lake Victoria structure and redistributing biogenic and nonbiogenic matter. Alexander and Imberger (2013) found similar trends in these two Using a network of HF, in situ, automated sensors, Klug et al. variables, while Karakaya et al. (2011) and Reeder (2011) described (2012) document the regional effects of Tropical Cyclone Irene on strong correlation between diel changes in DO and pH, corre- thermal structure and ecosystem metabolism in nine lakes and spondingly, in a shallow lake in Turkey and in open-water habi- reservoirs in northeastern North America. The fast decline of ther- tats in the Beaver Creek Wetlands Complex in Kentucky (USA). mal stability was related to the amount of precipitation on the Data on pH are used as a background characteristic (e.g., Hemond lake and the catchment area while the temperature change pre- et al. 2008; Bertone et al. 2015), as an environmental factor (e.g., dicted the change in primary production across all systems. HF Rubbo et al. 2006; Karakaya 2011) or used in equations for car- temperature measurements have also provided important details bonate balance (e.g., Rudorff et al. 2011), redox calculations of the effects of the 2003 and 2006 heatwaves in Europe on the (Hamilton-Taylor et al. 2005) or the proportion of toxic unionized thermal structure of lakes (Choin´ ski and Łyczkowska 2008; Jöhnk ammonia (Gelda and Effler 2003). Conductivity data often col- et al. 2008; Kulbe et al. 2008; Wilhelm and Adrian 2008). lected within a standard water quality monitoring program Water temperature is a ubiquitous factor affecting most chem- belong mostly to descriptive background data and have been ical and biological processes in aquatic environments (Regier et al. addressed independently in few lake studies. Conversely, in 1990; Poff et al. 2002), and thus combining chemical or biological brackish environments, such as estuaries, conductivity is the measurements with HFM of temperature opens up new perspec- main parameter to distinguish between different water masses tives for interpretation. Jones et al. (2008) explored patterns of (Trevethan et al. 2007; Smith et al. 2008). Temporally and spatially change in bacterioplankton and phytoplankton community com- dynamic salinity levels in a reservoir resulting from the mixing of position in response to typhoons in a freshwater humic lake in waters from two different tributaries, were reported by Atkinson Taiwan and found that after each typhoon-induced mixing event, and Mabe (2006) using a mobile water quality probe from a small the bacterial community composition revealed a deterministic boat and recording the sampling trajectory by the global position- pattern of recovery with distinct bacterial assemblages being as- ing satellite (GPS) system. Also Lindfors et al. (2005) used a flow- sociated with epilimnion and hypolimnion. In contrast, phyto- through system operated from a boat to map the water masses in plankton communities did not recover in a predictable way after their study area. typhoons. Baulch et al. (2005) studied whether warming could HF recordings of dissolved nutrient levels, such as nitrates

For personal use only. stimulate respiration and light-saturated photosynthetic rates in (NO −) and phosphates (PO 3−), in lakes are scarce, although ni- the epilithon. Water temperatures were monitored every 15 min 3 4 trate levels are often measured in rivers (Sherson 2012; Feng et al. in experimental enclosures and in the lake using continuously 2013; Sherson et al. 2015). The advantage of using continuous mea- recording thermocouples. Good agreement found between long- surements for nitrates stems from their remarkable diurnal vari- term and experimental results suggests that increased tempera- ability (Feng et al. 2013) that can be easily overlooked with discrete tures will increase metabolic rates of the epilithon. measurements (Schwientek et al. 2013). The detection limit of Fishes often exhibit clear temperature preferences. Comparing temperature profiles with fish distribution data in Lake Lesjask- phosphate sensors is mostly around 50–100 ppb but has been ogsvatnet (Norway), Bass et al. (2014) showed that during stratifi- lowered to 5 ppb (Karube and Nomura 2000) or likely even more cation, European grayling, Thymallus thymallus preferred to stay (Ohio Lake Erie Protection Fund 2010) and that of nitrate to withina2mzone around the thermocline supposedly driven by 6.2 ppb (Myers et al. 2012) making them potentially applicable for foraging opportunity. lakes. Nitrate measurements with an optical sensor in a reservoir DO was the second most often measured parameter using HFM where nitrate concentrations ranged from <0.1 mg N/l to >4 mg N/l, (Fig. 7). Continuous online DO measurements are routinely ap- gave highly consistent results with laboratory measurements. plied in aquaculture (e.g., Xu et al. 2006) and for effluent monitor- 7. HFM versus conventional sampling Environ. Rev. Downloaded from www.nrcresearchpress.com by Miss Pille Meinson on 03/02/16 ing (e.g., Bourgeois et al. 2001), remaining, however, outside the scope of our review. In lakes, calculating lake metabolic variables HFM nicely illustrates Hegel’s dialectical principle of the (gross primary production, community respiration, and net eco- transition from quantity to quality. Increasing measurement system production) from diel DO curves in the open water has frequency has enabled accounting for natural variability in mon- become increasingly popular and has been used in a large number itoring data and selecting optimum sampling frequencies for cost- of studies (e.g., Cole et al. 2000; Lauster et al. 2006; Staehr et al. effective representative sampling (Anttila et al. 2012). Although 2010; Tsai et al. 2011; Laas et al. 2012; Klotz 2013; Alfonso et al. manual monitoring frequencies may prove sufficient for certain 2015). Recently, the uncertainty issues of lake metabolism mea- purposes, a comparison with HFM data enables an uncertainty surements have been addressed in a number of studies (Batt and estimate even for historical data that improves data reliability. Carpenter 2012; Cremona et al. 2014). Based on DO observations Besides the already mentioned studies on episodic events and the derived from in situ sensors in 25 northern temperate lakes, opportunities to register and follow dynamic processes at time Langman et al. (2010) showed that at scales, ranging from minutes scales of hours and minutes, there is clear evidence of HFM data to days DO patterns were affected by a number of physical and providing a greater understanding of certain processes, first of all biological processes, such as internal waves, mixing, and ecosys- in lake metabolism. Compared to bottle incubation techniques tem metabolism. In small, shallow, inland water bodies, a diurnal encompassing only processes in the plankton community, the DO pattern can be affected by daytime thermal stratification and freewater, depth-weighted approach in estimating ecosystem me-

Published by NRC Research Press

66 Meinson et al. 59

tabolism takes into account primary production of all autotrophs 9. New challenges and directions and respiration of all lake biota, including benthic and possibly 9.1. Broadening the spatial scale of HFM littoral habitats depending on the location of the buoy, however, Application of sensor technology has considerably shifted the accounting for the effects of turbulence within the water column lower-end boundaries of temporal resolution in limnological remains one of the largest methodological challenges (Sadro et al. studies, shedding light upon diurnal, hourly, and even shorter a b 2011 , 2011 ). Furthermore, HFM can be used in remote locations scale processes. The length of time series from automated buoy and severe conditions where conventional measurements would stations has in places exceeded 10 years (e.g., Pierson et al. 2011) be complicated. For example, Sadro and MacIntyre (2014) mea- allowing assessment of year-to year variability of physical and sured spatial and temporal patterns in the ecosystem respiration chemical lake environments. However, as shown by Crawford of five Alaskan lakes that stay covered by up to3moficeandsnow et al. (2015), spatial variability is rarely documented with sensors for approximately two-thirds of the year, and described two mech- because of the high investment costs for the spatial replication of anisms operating in tandem accounting for the DO drawdown such infrastructure. Still the use of profiling buoys is an emerging patterns found within lake basins during the winter. trend over the past years that has likely been promoted by de- Applying HFM on boats or other moving vessels allows turning creasing instrumentation prices. A slow transition from one- the high temporal resolution into high spatial resolution provid- station measurements to multi-station measurements per lake in ing unique tools for studying horizontal distribution patterns of metabolism studies can also be seen (e.g., Van de Bogert et al. variables at scales never achievable by means of conventional 2007, 2012; Coloso et al. 2008; Sadro et al. 2011b) allowing to at least field measurements. These studies can address specific scientific partially include spatial variability aspects. Broader use of mobile issues, such as plankton patchiness (Anttila et al. 2008), elucidate on-board novel flow-through systems, such as the fast limnology basic hydrological mechanisms at relatively low costs of invest- automated measurement (FLAMe) platform (Crawford et al. 2015), ments and manpower (Schwientek et al. 2013) or can be used for equipped with various sensors and combined with GPS and an calibrating remote sensing data to produce a better picture of the acoustic eco-sounder, allows for fast surveys of extensive water processes taking place in the study area (Lindfors et al. 2005). bodies that can serve for calibrating remote sensing data (e.g., Combined with modelling, HF data can inform early warning sys- Östlund et al. 2001; Lindfors et al. 2005), studying patchiness, and tems of the occurrence of cyanobacteria in drinking water sources optimizing monitoring networks (Anttila et al. 2008). A combina- using phycocyanin probe (Izydorczyk et al. 2005; Zamyadi et al. tion of short-term mobile surveys with stationary buoy measurements together with sensor technologies developing towards 2012) or can be used as input data to follow the spread of sub- increasing sensitivity, reliability, and autonomy, will likely open stances in a water body (Abell and Hamilton 2015). new horizons for HF measurements in lakes. A comprehensive description of the advantages and limitations of HFM in surface water monitoring is given by Rinke et al. (2013) Acknowledgements based on Rappbode Reservoir, where a set of online-sensors for This research was supported by the target-financed projects the measurement of physical, chemical, and biological variables SF0170011s08 and IUT 21-2, and by personal research grant PUT777 was complemented by a biweekly limnological sampling sched- of the Estonian Ministry of Education and Research, by the Esto- ule. The authors note that the HFM provide a deeper insight into nian Science Foundation grants ETF8729 and ETF9102, by the EU ecosystem dynamics and lake metabolism and are a powerful tool through European Regional Development Fund, program Envi-

For personal use only. for assessing matter fluxes and establishing precise biochemical ronmental Conservation and Environmental Technology R&D budgets. Online measurements offer data for developing and val- Programme project VeeOBS (3.2.0802.11-0043) and MARS project idating state-of-the-art lake models and to improve their predic- (Managing Aquatic ecosystems and water Resources under multi- tive capabilities. As a fundamental limitation, the authors point ple Stress) funded under the 7th EU Framework Programme, out the lack of reliable online sensors for several important water Theme 6 (Environment including Climate Change), Contract No. quality variables (e.g., phosphorus compounds) making it neces- 603378 (http://www.mars-project.eu). sary to realise lab-based measurements of these variables by means of regular field samplings. Besides this, the regular field References sampling was used for checking the measurement quality and Abell, J.M., and Hamilton, D.P. 2015. Biogeochemical processes and phytoplankton nutrient limitation in the inflow transition zone of a large eutrophic lake accuracy of the online sensors and for the identification of their during a summer rain event. Ecohydrology, 8: 243–262. doi:10.1002/eco.1503. calibration intervals. Rinke et al. (2013) conclude that the imple- Alexander, R., and Imberger, J. 2013. Phytoplankton patchiness in Winam Gulf, mentation of in situ sensors, therefore, can never fully substitute Lake Victoria: a study using principal component analysis of in situ fluores- classical field sampling approaches as the latter will remain im- cent excitation spectra. Freshwater Biol. 58: 275–291. doi:10.1111/fwb.12057. Alfonso, M.B., Vitale, A.J., Menendez, M.C., Perillo, V.L., Piccolo, M.C., and portant at least for calibration and quality control of the sensor Perillo, C.M. 2015. Estimation of ecosystem metabolism from diel oxygen

Environ. Rev. Downloaded from www.nrcresearchpress.com by Miss Pille Meinson on 03/02/16 data. technique in a saline shallow lake: La Salada (Argentina). Hydrobiologia, 752: 223–237. doi:10.1007/s10750-014-2092-1. 8. Have HFM applications in limnology reached a Ansa-Ansare, O.D., Marr, I.L., and Cresser, M.S. 2000. Evaluation of modelled and measured patterns of dissolved oxygen in a freshwater lake as an indicator of plateau after the boom? the presence of biodegradable organic pollution. Water Resour. Res. 34: The publication record of HF studies in lakes (Fig. 2) shows that, 1079–1088. doi:10.1016/S0043-1354(99)00239-0. Anttila, S., Kairesalo, T., and Pellikka, P. 2008. A feasible method to assess inac- over the last 2 years, the number of papers has not reached 2011– curacy caused by patchiness in water quality monitoring. Environ. Monit. 2012 levels. Although automatic HFM potentially can replace dis- Assess. 142: 11–22. doi:10.1007/s10661-007-9904-y. PMID:17891528. crete manual measurements in many ﬁelds of lake research and Anttila, S., Ketola, M., Vakkilainen, K., and Kairesalo, T. 2012. Assessing temporal management, the leveling off of the publication records can be a representativeness of water quality monitoring data. J. Environ. Monit. 14: 589–595. doi:10.1039/C2EM10768F. PMID:22159426. sign of conceptual stagnation, as hardly any new HFM applica- Atkinson, S.F., and Mabe, J.A. 2006. Near real-time monitoring and mapping of tions have reached a level comparable to that of lake metabolism speciﬁc conductivity levels across lake Texoma, U.S.A. Environ. Monit. As- studies. Partly the slowdown in the publication activity could be sess. 120: 449–460. doi:10.1007/s10661-005-9072-x. PMID:16741798. caused by the economic recession that has delayed advances, al- Bass, A.L., Haugen, T.O., and Vøllestad, L.A. 2014. Distribution and movement of European grayling in a subarctic lake revealed by acoustic telemetry. Ecol. though, it may just be a sign of “natural” variability as the publi- Freshw. Fish, 23: 149–160. doi:10.1111/eff.12056. cation record was very low also in 2009. Batt, R.D., and Carpenter, S.R. 2012. Free-water lake metabolism: addressing

Published by NRC Research Press

67 60 Environ. Rev. Vol. 24, 2016

noisy time series with a Kalaman filter. Limnol. Oceanogr. Methods, 10: Communicate, and Compute Information. Science, 332: 60–65. doi:10.1126/ 20–30. doi:10.4319/lom.2012.10.20. science.1200970. PMID:21310967. Baulch, H.M., Schindler, D.W., Turner, M.A., Findlay, D.L., Paterson, M.J., and Hondzo, M., and Haider, Z. 2004. Boundary mixing in a small stratified lake. Vinebrooke, R.D. 2005. Effects of warming on benthic communities in a Water Resour. Res. 40: W03101. doi:10.1029/2002WR001851. boreal lake: Implications of climate change. Limnol Oceanogr. 50: 1377–1392. Hurst, T., Christenson, B., and Cole-Baker, J. 2012. Use of weather buoy to derive doi:10.4319/lo.2005.50.5.1377. improved heat and mass balance parameters for Ruapehu Crater Lake. J. Bertone, E., Stewart, R.A., Zhang, H., and O’Halloran, K. 2015. Analysis of the Volcanol. Geotherm. Res. 235–236: 23–28. doi:10.1016/j.jvolgeores.2012.05. mixing processes in the subtropical advancetown Lake, Australia. J. Hydrol. 004. 522: 67–79. doi:10.1016/j.jhydrol.2014.12.046. Izydorczyk, K., Tarczynska, M., Jurczak, T., Mrowczynski, J., and Zalewski, M. Boegman, L., Imberger, J., Ivey, G.N., and Antenucci, J.P. 2003. High-frequency 2005. Measurement of phycocyanin fluorescenceas an online early warning internal waves in large stratified lakes. Limnol. Oceanogr. 48: 895–919. doi: system for cyanobacteria in reservoir intake water. Environ. Toxicol. 20: 10.4319/lo.2003.48.2.0895. 425–430. doi:10.1002/tox.20128. PMID:16007662. Bourgeois, W., Burgess, J.E., and Stuetz, R.M. 2001. On-line monitoring of waste- Jennings, E., Jones, S., Arvola, L., Staehr, P.A., Gaiser, E., Jones, I.D., water quality: a review. J. Chem. Technol. Biotechnol. 76: 337–348. doi:10. Weathers, K.C., Weyhenmeyer, G.A., Chiu, C.-Y., and De Eyto, E. 2012. Effects 1002/jctb.393. of weather-related episodic events in lakes: ananalysis based on high- Branco, B.F., and Torgersen, T. 2009. Diurnal sediment resuspension and set- frequency data. Freshwater Biol. 57: 589–601. doi:10.1111/j.1365-2427.2011. tling: impact on the coupled physical and biogeochemical dynamics of dis- 02729.x. solved oxygen and carbon in shallow water body. Mar. Freshwater Res. 60: Jöhnk, K.D., Huisman, J., Sharples, J., Sommeijer, B., Visser, P.M., and 669–679. doi:10.1071/MF08113. Stroom, J.M. 2008. Summer heatwaves promote blooms of harmful cyano- Brown, C.J., Saunders, M.I., Possingham, H.P., and Richardson, A.J. 2013. Manag- bacteria. Glob. Change Biol. 14: 495–512. doi:10.1111/j.1365-2486.2007.01510.x. ing for Interactions between Local and Global Stressors of Ecosystems. PLoS Johnson, K.S., Needoba, J.A., Riser, S.C., and Showers, W.J. 2007. Chemical Sensor One, 8: e65765. doi:10.1371/journal.pone.0065765. PMID:23776542. Networks for the Aquatic Environment. Chem. Rev. 107: 623–640. doi:10. Chang, G.C., and Dickey, T.D. 2001. Optical and physical variability on timescales 1021/cr050354e. PMID:17249737. from minutes to the seasonal cycle on the New England shelf: July 1996 to Jones, S.E., Chiu, C.-Y., Kratz, T.K., Wu, J.-T., Shade, A., and McMahon, K.D. 2008. June 1997. J. Geophys. Res. 106: 9435–9453. doi:10.1029/2000JC900069. Typhoons initiate predictable change in aquatic bacterial communities. Lim- Choin´ ski, A., and Łyczkowska, G. 2008. Thermal characteristics of water of nol. Oceanogr. 53: 1319–1326. doi:10.4319/lo.2008.53.4.1319. Wielki Staw in the Karkonosze mountains and Morskie Oko in the Tatras, Karakaya, N. 2011. Does different versus equal daytime and night-time respira- July 2006. Pol. J. Environ. Stud. 17: 835–840. tion matter for quantification of lake metabolism using diel dissolved oxy- Churchill, J.H., and Kerfoot, C.W. 2007. The impact of surface heat flux and wind gen cycles? Ann. Limnol. Int. J. Lim. 47: 251–257. doi:10.1051/limn/2011042. on thermal stratification in Portage lake, Michigan. J. Great Lakes Res. 33: Karakaya, N., Evrendilek, F., and Güngör, K. 2011. Modeling and Validating Long- 143–155. doi:10.3394/0380-1330(2007)33[143:TIOSHF]2.0.CO;2. Term Dynamics of Diel Dissolved Oxygen with Particular Reference to pH in Cole, J.J., Pace, M.L., Carpenter, S.R., and Kitchell, J.F. 2000. Persistence of net a Temperate Shallow Lake (Turkey). Clean – Soil, Air, Water, 39: 966–971. heterotrophy in lakes during nutrient addition and food web manipulations. doi:10.1002/clen.201100051. Limnol. Oceanogr. 45: 1718–1730. doi:10.4319/lo.2000.45.8.1718. Karube, I., and Nomura, Y. 2000. Enzyme sensors for environmental analysis. J. Coloso, J.J., Cole, J.J., Hanson, P.C., and Pace, M.L. 2008. Depth-integrated, con- Mol. Catal. B Enzym. 10: 177–181. doi:10.1016/S1381-1177(00)00125-9. tinuous estimates of metabolism in a clear-water lake. Can. J. Fish. Aquat. Sci. Kimura, N., Liu, W.-C., Chiu, C.-Y., and Kratz, T.K. 2014. Assessing the effects of 65(4): 712–722. doi:10.1139/f08-006. severe rainstorm-induced mixing on a subtropical, subalpine lake. Environ. Crawford, J.T., Loken, L.C., Casson, N.J., Smith, C., Stone, A.G., and Winslow, L.A. Monit. Assess. 186: 3091–3114. doi:10.1007/s10661-013-3603-7. PMID:24415132. 2015. High-Speed Limnology: Using Advanced Sensors to Investigate Spatial King, D.A. 2004. The scientific impact of nations. Nature, 430: 311–316. doi:10. Variability in Biogeochemistry and Hydrology. Environ. Sci. Technol. 49: 1038/430311a. PMID:15254529. 442–450. doi:10.1021/es504773x. PMID:25406073. Kitchin, R. 2013. Big data and human geography: Opportunities, challenges and risks. Cremer, M. 1906. U¨ ber die Ursache der elektromotorischen Eigenschaften der Dialogues in Human Geography, 3: 262–267. doi:10.1177/2043820613513388. Gewebe, zugleichein Beitrag zur Lehre von Polyphasischen Elektrolytketten. Klotz, R.L. 2013. Factors driving the metabolism of two north temperate ponds. Z. Biol. 47: 56. Hydrobiologia, 711: 9–17. doi:10.1007/s10750-013-1450-8. Cremona, F., Laas, A., Nõges, P., and Nõges, T. 2014. High-frequency data within Klug, J.L., Richardson, D.C., Ewing, H.A., Hargreaves, B.R., Samal, N.R.,

For personal use only. a modeling framework: On the benefit of assessing uncertainties of lake Vachon, D., Pierson, D.C., Lindsey, A.M., O’Donnell, D.M., Effler, S.W., and metabolism. Ecol. Model. 294: 27–35. doi:10.1016/j.ecolmodel.2014.09.013. Weathers, K.C. 2012. Ecosystem Effects of a Tropical Cyclone on a Network of Cushing, J.B. 2013. Beyond Big Data? Comput. Sci. Eng. 15: 4–5. doi:10.1109/MCSE. Lakes in Northeastern North America. Environ. Sci. Technol. 46: 11693–11701. 2013.102. doi:10.1021/es302063v. PMID:23016881. Dur, G., Schmitt, F.G., and Souissi, S. 2007. Analysis of high frequency tempera- Kulbe, T., Livingstone, D.M., Guilizzoni, P., and Sturm, M. 2008. The use of ture time series in the Seine estuary from the Marel autonomous monitoring long-term, high-frequency, automatic sampling data in a comparative study buoy. Hydrobiologia, 588: 59–68. doi:10.1007/s10750-007-0652-3. of the hypolminia of two dissimilar Alpine lakes. Verh. Int. Verein. Limnol. Eigenmann, C.H. 1895. First report of the Indiana University Biological Station. 30: 371–376. Proc. Indiana Acad. Sci. 5: 203–216. Laas, A., Nõges, P., Kõiv, T., and Nõges, T. 2012. High-frequency metabolism study Feng, Z., Schilling, K.E., and Chan, K.-S. 2013. Dyanmic regression modeling of in a large and shallow temperate lake reveals seasonal switching between net daily nitrate-nitrogen concentrations in a large agricultural watershed. En- autotrophy and net heterotrophy. Hydrobiologia, 694: 57–74. doi:10.1007/ viron. Monit. Assess. 185: 4605–4617. doi:10.1007/s10661-012-2891-7. PMID: s10750-012-1131-z. 23054269. Langman, O.C., Hanson, P.C., Carpenter, S.R., and Hu, Y.H. 2010. Control of Gelda, R.K., and Effler, S.W. 2003. Application of a Probabilistic Ammonia dissolved oxygen in northern temperate lakes over scales ranging from min- Model: Identification of Important Model Inputs and Critique of a TMDL utes to days. Aquat. Biol. 9: 193–202. doi:10.3354/ab00249. Analysis for an Urban Lake. Lake Reserv. Manage. 19: 187–199. doi:10.1080/ Lauster, G.H., Hanson, P.C., and Kratz, T.K. 2006. Gross primary production and 07438140309354084. respiration differences among littoral and pelagic habitats in northern Wis- Goodwin, K., Caraco, N., and Cole, J. 2008. Temporal dynamics of dissolved consin lakes. Can. J. Fish. Aquat. Sci. 63(5): 1130–1141. doi:10.1139/f06-018. oxygen in a floating leaved macrophyte bed. Freshwater Biol. 53: 1632–1641. Laval, B., Imberger, J., Hodges, B.R., and Stocker, R. 2003. Modeling circulation in Environ. Rev. Downloaded from www.nrcresearchpress.com by Miss Pille Meinson on 03/02/16 doi:10.1111/j.1365-2427.2008.01983.x. lakes: Spatial and temporal variations. Limnol. Oceanogr. 48: 983–994. doi: Hamilton-Taylor, J., Smith, E.J., Davison, W., and Sugiyama, M. 2005. Resolving 10.4319/lo.2003.48.3.0983. and modelling the effects of Fe and Mn redox cycling on tracemetal behavior Lehner, B., and Döll, P. 2004. Development and validation of a global database of in a seasonally anoxic lake. Geochim. Cosmochim. Acta, 69: 1947–1960. doi: lakes, reservoirs and wetlands. J. Hydrol. 296: 1–22. doi:10.1016/j.jhydrol.2004. 10.1016/j.gca.2004.11.006. 03.028. Hanson, P.C., Bade, D.L., Carpenter, S.R., and Kratz, T.K. 2003. Lake metabolism: Lindfors, A., Rasmus, K., and Strömbeck, N. 2005. Point or pointless- quality of Relationships with dissolved organic carbon and phosphorus. Limnol. Ocean- ground data. Int. J. Remote Sens. 26: 415–423. doi:10.1080/01431160410001720261. ogr. 48: 1112–1119. doi:10.4319/lo.2003.48.3.1112. Ljungemyr, P., Gustafsson, N., and Omstedt, A. 1996. Parameterization of lake Hemond, H., Cheung, J., Mueller, A., Wong, J., Hemond, M., Mueller, D., and thermodynamics in a high-resolution weather forecasting model. Tellus Ser. Eskesen, J. 2008. The NERUS in-lake wireless/acoustic chemical data network. A, 48: 608–621. doi:10.1034/j.1600-0870.1996.t01-4-00002.x. Limnol. Oceanogr. Methods, 6: 288–298. doi:10.4319/lom.2008.6.288. Lorke, A. 2007. Boundary mixing in the thermocline of a large lake. J. Geophys. Hering, D., Carvalho, L., Argillier, C., Beklioglu, M., Borja, A., Cardoso, A.C., Res. 112: C09019. doi:10.1029/2006JC004008. Duel, H., Ferreira, T., Globevnik, L., Hanganu, J., Hellsten, S., Jeppesen, E., Lorke, A., Peeters, F., and Bäuerle, E. 2006. High-frequency internal waves in the Kodeš, V., Solheim, A.L., Nõges, T., Ormerod, S., Panagopoulos, Y., littoral zone of a large lake. Limnol. Oceanogr. 51: 1935–1939. doi:10.4319/lo. Schmutz, S., Venohr, M., and Birk, S. 2015. Managing aquatic ecosystems and 2006.51.4.1935. water resources under multiple stress—An introduction to the MARS project. MacIntyre, S., Romero, J.R., and Kling, G.W. 2002. Spatial-temporal variability in Sci. Total Environ. 503–504: 10–21. doi:10.1016/j.scitotenv.2014.06.106. PMID: surface layer deepening and lateral advection in an embayment of Lake 25017638. Victoria, East Africa. Limnol Oceanogr. 47: 656–671. doi:10.4319/lo.2002.47.3. Hilbert, M., and López, P. 2011. The World’s Technological Capacity to Store, 0656.

Published by NRC Research Press

68 Meinson et al. 61

Manov, D.V., Chang, G.C., and Dickey, T.D. 2004. Methods for Reducing Biofoul- #B43B-0245. Available from http://adsabs.harvard.edu/ [accessed 11 Septem- ing of Moored Optical Sensors. J. Atmos. Oceanic Technol. 21: 958–968. doi: ber 2015]. 10.1175/1520-0426(2004)021<0958:MFRBOM>2.0.CO;2. Sadro, S., Melack, J.M., and MacIntyre, S. 2011a. Spatial and temporal variability Myers, M., Podolska, A., Pope, T., Khir, F.M.L., Mishra, U.K., Nener, B.D., Baker, in the ecosystem metabolism of a high-elevation lake: integrating benthic M.V., and Parish, G. 2012. Nitrate-selective gallium nitride transistor-based and pelagic habitats. Ecosystems, 14: 1123–1140. doi:10.1007/s10021-011-9471-5. ion sensors with low detection limit. In Proceedings of the 14th International Sadro, S., Melack, J.M., and MacIntyre, S. 2011b. Depth-integrated estimates of Meeting on Chemical Sensors – IMCS 2012, 2012-05-20 – 2012-05-23, Nürnberg/ ecosystem metabolism in a high-elevation lake (Emerald Lake, Sierra Nevada, Nuremberg, Germany. Chapter 8.1. Chemical Sensors based on III-V Semicon- California). Limnol. Oceanogr. 56: 1764–1780. doi:10.4319/lo.2011.56.5.1764. ductors, pp. 671–673. doi:10.5162/IMCS2012/8.1.5. Schwientek, M., Osenbrück, K., and Fleischer, M. 2013. Investigating hydrologi- Nordbo, A., Launiainen, S., Mammarella, I., Leppäranta, M., Houtari, J., Ojala, A., cal drivers of nitrate export dynamics in two agricultural catchments in and Vesala, T. 2011. Long-term energy flux measurements and energy balance Germany using high-frequemcy data series. Environ. Earth Sci. 69: 381–393. over a small boreal lake using eddy covariance technique. J. Geophys. Res. doi:10.1007/s12665-013-2322-2. 116: D02119. doi:10.1029/2010JD014542. Severinghaus, J.Q., and Astrup, P.B. 1986. History of blood gas analysis. IV. Leland Ohio Lake Erie Protection Fund. 2010. Improving Detection Limit of Phosphate Clark’s oxygen electrode. J. Clin. Monit. 2: 125–139. doi:10.1007/BF01637680. Microsensor. Technical Report, Small Grant Program Project Number: SG 385-10 PMID:3519875. [online]. Available from http://lakeerie.ohio.gov/Portals/0/Closed%20Grants/ Shade, A., Chiu, C.-Y., and McMahon, K.D. 2010. Seasonal and Episodic lake small%20grants/SG%20385-10.pdf [accessed 24 April 2015]. mixing stimulate differential planktonic bacterial dynamics. Microb. Ecol. Ormerod, S.J., Dobson, M., Hildrew, A.G., and Townsend, C.R. 2010. Multiple 59: 546–554. doi:10.1007/s00248-009-9589-6. PMID:19760448. stressors in freshwater ecosystems. Freshwater Biol. 55: 1–4. doi:10.1111/j.1365- Sherson, L. 2012. Nutrient dynamics in a headwater stream: use of continuous 2427.2009.02395.x. water quality sensors to examine seasonal, event, and diurnal processes in Östlund, C., Flink, P., Strömbeck, N., Pierson, D., and Lindell, T. 2001. Mapping of the east fork Jemez river, NM. Ph.D. thesis, University of Oregon, U.S.A. the Water Quality of Lake Erken, Sweden from Imaging Spectrometry and Sherson, L.R., Van Horn, D.J., Comez-Velez, J.D., Crossey, L.J., and Dahm, C.N. Landsat Thematic Mapper. Sci. Total Environ. 268: 139–154. doi:10.1016/S0048- 2015. Nutrient dynamics in an alpine headwater stream: use of continuous 9697(00)00683-5. PMID:11315737. water quality sensors to examine responses to wildfire and precipitation Pernica, P., and Wells, M. 2012. Frequency of episodic stratification in the near events. Hydrol. Process. 29: 3193–3207. doi:10.1002/hyp.10426. surface of Lake Opeongo and other small lakes. Water Qual. Res. J. Can. 47: Smith, C.G., Cable, J.E., and Martin, J.B. 2008. Episodic high intensity mixing 227–237. doi:10.2166/wqrjc.2012.001. events in a subterranean estuary: effects of tropical cyclones. Limnol. Ocean- Pernica, P., Wells, M.G., and MacIntyre, S. 2014. Presistent weak thermal strati- ogr. 53: 666–674. doi:10.4319/lo.2008.53.2.0666. fication inhibits mixing in the epilimnion of north-temperate Lake Opeongo, Solomon, C.T., Brueswitz, D.A., Richardson, D.C., Rose, K.C., Van de Bogert, M.C., Canda. Aquat. Sci. 76: 187–201. doi:10.1007/s00027-013-0328-1. Hanson, P.C., Kratz, T.K., Larget, B., Adrian, R., Leroux Babin, B., Chiu, C.-Y., Pierson, D.C., Weyhenmeyer, G.A., Arvola, L., Benson, B., Blenckner, T., Kratz, T., Hamilton, D.P., Gaiser, E.E., Hendricks, S., Istvánovics, V., Laas, A., Livingstone, D.M., Markensten, H., Marzec, G., Petterson, K., and O’Donnell, M., Pace, M.L., Ryder, E., Staehr, P.A., Torgersen, T., Vanni, M.J., Weathers, K. 2011. An automated method to monitor lake ice phenology. Weathers, K.C., and Zhu, G. 2013. Ecosystem respiration: Drivers of daily Limnol. Oceanogr. Methods, 9: 74–83. doi:10.4319/lom.2010.9.0074. variability and background respiration in lakes around the globe. Limnol. Poff, N.L., Brinson, M.M., and Day, J.W. 2002. Aquatic ecosystems and global Oceanogr. 58: 849–866. doi:10.4319/lo.2013.58.3.0849. climate change. Technical Report, Pew Center on Global Climate Change, Song, K., Li, L., Tedesco, L., Clercin, N., Hall, B., Li, S., Shi, K., Liu, D., and Sun, Y. Arlington, U.S.A. 2013. Remote estimation of phycocyanin (PC) for inland waters coupled with Porter, J.H., and Lin, C.C. 2013. Hybrid networks and ecological sensing. In Wire- YSI PC fluorescence probe. Environ. Sci. Pollut. Res. 20: 5330–5340. doi:10. less Sensor Networks & Ecological Monitoring, Springer, pp. 99–124. 1007/s11356-013-1527-y. Porter, J., Arzbergen, P., Braun, H.-W., Bryant, P., Gage, S., Hansen, T., Hanson, P., Sørensen, S.P.L. 1909. “Enzymstudien. II: Mitteilung. U¨ ber die Messung und die Lin, C.-C., Lin, F.-P., Kratz, T., Michener, W., Shapiro, S., and Williams, T. 2005. Bedeutung der Wasserstoffionenkoncentration bei enzymatischen Proz- Wireless Sensors Networks for Ecology. BioScience, 55: 561–572. doi:10.1641/ essen”. Biochemische Zeitschrift, 21: 131–304. [In German.] 0006-3568(2005)055%5B0561:WSNFE%5D2.0.CO;2. Staehr, P.A., Bade, D., Van de Bogert, M.C., Koch, G.R., Williamson, C., Porter, J., Hanson, P.C., and Lin, C.-C. . 2012. Staying afloat in the sensor data Hanson, P., Cole, J.J., and Kratz, T. 2010. Lake metabolism and the diel oxygen deluge. Trends Ecol. Evol. 27: 121–129. doi:10.1016/j.tree.2011.11.009. PMID: technique: State of the science. Limnol. Oceanogr.: Methods, 8: 628–644. 22206661.

For personal use only. doi:10.4319/lom.2010.8.0628. Read, J.S., Hamliton, D.P., Jones, I.D., Muraoka, K., Winslow, L.A., Kroiss, R., Staehr, P.A., Baastrup-Spohr, L., Sand-Jensen, K., and Stedmon, C. 2012. Lake Wu, C.H., and Gaiser, E. 2011a. Derivation of lake mixing and stratification metabolism scales with lake morphometry and catchment conditions. indices from high-resolution lake buoy data. Environ. Model. Softw. 26: Aquat. Sci. 74: 155–169. doi:10.1007/s00027-011-0207-6. 1325–1336. doi:10.1016/j.envsoft.2011.05.006. Sullivan, T., Broszeit, S., O’Sullivan, K.P., McAllen, R., Daevenport, J., and Read, J.S., Shade, A., Wu, C.H., Gorzalski, A., and McMahon, K.D. 2011b. “Gradual Regan, F. 2013. High resolution monitoring of episodic stratification events Entrainment Lake Inverter” (GELI): A novel device for experimental lake in an enclosed marine system. Estuar. Coast. Shelf Sci. 123: 26–33. doi:10.1016/ mixing. Limnol. Ocenaogr. Methods, 9: 14–28. doi:10.4319/lom.2011.9.14. j.ecss.2013.02.012. Read, J.S., Hamilton, D.P., Desai, A.R., Rose, K.C., MacIntyre, S., Lenters, J.D., Taillefert, M., Luther, G.W., and Nuzzio, D.B. 2000. The application of electro- Smyth, R.L., Hanson, P.C., Cole, J.J., Staehr, P.A., Rusak, J.A., Pierson, D.C., chemical tools for in situ measurements in aquatic systems. Electroanalysis, Brookes, J.D., Laas, A., and Wu, C.H. 2012. Lake-size dependency of wind shear 12: 401–412. doi:10.1002/(SICI)1521-4109(20000401)12:6%3C401::AID-ELAN401% and convection as controls on gas exchange. Geophys. Res. Lett. 39: L09405. 3E3.0.CO;2-U. doi:10.1029/2012GL051886. Tengberg, A., Hovdenes, J., Andersson, J.H., Brocandel, O., Diaz, R., Hebert, D., Redmond, K.T. 2007. Evaporation and the hydrologic budget of Crater Lake, Arnerich, T., Huber, C., Körtzinger, A., Khripounoff, A., Rey, F., Rönning, C., Oregon. Hydrobiologia, 574: 29–46. doi:10.1007/s10750-006-2603-9. Schimanski, J., Sommer, S., and Stangelmayer, A. 2006. Evaluation of a Reeder, B.C. 2011. Assessing constructed wetland functional success using diel lifetime-based optode to measure oxygen in aquatic systems. Limnol. Ocean- changes in dissolved oxygen, pH, and temperature in submerged, emergent, ogr. Methods, 4: 717. doi:10.4319/lom.2006.4.7. and open-water habitats in the Beaver Creek Wetlands Complex Kentucky Trevethan, M., Chanson, H., and Takeuchi, M. 2007. Continuous high-frequency (U.S.A.). Ecol. Eng. 37: 1772–1778. doi:10.1016/j.ecoleng.2011.06.018. turbulence and suspended sediment concentration measurements in an up- Environ. Rev. Downloaded from www.nrcresearchpress.com by Miss Pille Meinson on 03/02/16 Regier, H.A., Holmes, J.A., and Pauly, D. 1990. Influence of temperature change per estuary. Estuar. Coast. Shelf Sci. 73: 341–350. doi:10.1016/j.ecss.2007.01. on aquatic ecosystems: an interpretation of empirical data. Trans. Am. Fish. 014. Soc. 119: 374–389. doi:10.1577/1548-8659(1990)119<0374:IOTCOA>2.3.CO;2. Tsai, J.-W., Kratz, T.K., Hanson, P.C., Wu, J.-T., Chang, W.Y.B., Arzberger, P.W., Rinke, K., Kuehn, B., Bocaniov, S., Wendt-Potthoff, K., Büttner, O., Tittel, J., Lin, B.-S., Lin, F.-P., Chou, H.-M., and Chiu, C.-Y. 2008. Seasonal dynamics, Schultze, M., Herzprung, P., Rönicke, H., Rink, K., Rinke, K., Dietze, M., typhoons and the regulation of lake metabolism in a subtropical humic lake. Matthes, M., Paul, L., and Friese, K. 2013. Reservoirs as sentinels of catch- Freshwater Biol. 53: 1929–1941. doi:10.1111/j.1365-2427.2008.02017.x. ments: the Rappbode Reservoir Observatory (Harz Mountains, Germany). Tsai, J.-W., Kratz, T.K., Hanson, P.C., Kimura, N., Liu, W.-C., Lin, F.-P., Chou, H.-M., Environ. Earth Sci. 69: 523–536. doi:10.1007/s12665-013-2464-2. Wu, J.-T., and Chiu, C.-Y. 2011. Metabolic changes and the resistance and Rubbo, M.J., Cole, J.J., and Kiesecker, J.M. 2006. Terrestrial Subsidies of Organiv resilience of subtropical heterotrophic lake to typhoon disturbance. Can. J. Carbon Support Net Ecosystem Production in Temporary Forest Ponds: Evi- Fish. Aquat. 68(5): 768–780. doi:10.1139/f2011-024. dence from an Ecosystem Experiment. Ecosystems, 9: 1170–1176. doi:10.1007/ Van de Bogert, M.C., Carpenter, S.R., Cole, J.J., and Pace, M.L. 2007. Assessing s10021-005-0009-6. pelagic and benthic metabolism using free water measurements. Limnol. Rudorff, C.M., Melack, J.M., MacIntyre, S., Barbosa, C.C.F., and Novo, E.M.L.M. Oceanogr. Methods, 5: 145–155. doi:10.4319/lom.2007.5.145. 2011. Seasonal and spatial variability of CO2 emission from a large floodplain Van de Bogert, M.C., Bade, D.L., Carpenter, S.R., Cole, J.J., Pace, M.L., lake in the lower Amazon. J. Geophys. Res. 116: G04007. doi:10.1029/ Hanson, P.C., and Langman, O.C. 2012. Spatial heterogeneity strongly affects 2011JG001699. estimates of ecosystem metabolism in two north temperate lakes. Limnol. Sadro, S., and MacIntyre, S. 2014. Life Under the Ice: Spatial and Temporal Oceanogr. 57: 1689–1700. doi:10.4319/lo.2012.57.6.1689. Patterns in Rates of Water Column and Sediment Respiration in 5 Alaskan Vos, R.J., Hakvoort, J.H.M., Jordansand, R.W.J., and Ibelings, B.W. 2003. Multi- Arctic Lakes. American Geophysical Union, Fall Meeting 2014, abstract platform optical monitoring of eutrophication in temporally and spatially

Published by NRC Research Press

69 62 Environ. Rev. Vol. 24, 2016

variable lakes. Sci. Total Environ. 312: 221–243. doi:10.1016/S0048-9697(03) Wilhelm, S., and Adrian, R. 2008. Impact of summer warming on the thermal 00225-0. PMID:12873412. characteristics of a polymictic lake and consequences for oxygen, nutrients Warren, H.E., and Whipple, G.C. 1895. The thermophone, a new instrument for andphytoplankton.FreshwaterBiol.53:226–237.doi:10.1111/j.1365-2427.2007. obtaining the temperature of a distant or inaccessible place, and some ob- 01887.x. servations on the temperature of surface waters. Am. Meteorol. J. 12: 35–50. Xu, J., Liu, Y., Cui, S., and Miao, X. 2006. Behavioural responses of tilapia (Oreo- Weathers, K., Hanson, P.C., Arzberger, P., Brentrup, J., Brookes, J., Carey, C.C., chromisniloticus) to acute ﬂuctuations in dissolved oxygen levels as monitored Gaiser, E., Hamilton, D.P., Hong, G.S., Ibelings, B., Istvánovics, V., by computer vision. Aquacult. Eng. 35: 207–217. doi:10.1016/j.aquaeng.2006. Jennings, E., Kim, B., Kratz, T., Lin, F.-P., Muraoka, K., O’Reilly, C., Piccolo, C., Rose, K.C., Ryder, E., and Zhu, G. 2013. The Global Lake Ecological Observa- 02.004. tory Network (GLEON): the evolution of grassroots network science. Limnol. Zamyadi, A., McQuaid, N., Prévost, M., and Dorner, S. 2012. Monitoring of poten- Oceanogr. Bull. 22: 71–73. tially toxic cyanobacteria using an online multi-probe in drinking water Weinberger, S., and Vetter, M. 2014. Lake heat content and stability variation sources. J. Environ. Monit. 14: 579–588. doi:10.1039/c1em10819k. due to climate change: coupled regional climate model (REMO)-lake model Zhou, P., and Leydesdroff, L. 2006. The emergence of China as a leading nation in (DYRESM) analysis. J. Limnol. 73: 109–121. doi:10.4081/jlimnol.2014.668. science. Res. Policy, 35: 83–104. doi:10.1016/j.respol.2005.08.006. For personal use only. Environ. Rev. Downloaded from www.nrcresearchpress.com by Miss Pille Meinson on 03/02/16

Published by NRC Research Press

70 II

71 Woolway, R. Iestyn; Meinson, Pille; Nõges, Peeter; Jones, Ian D.; Laas, Alo (2017). Atmospheric stilling leads to prolonged thermal stratifi cation in a large shallow polymictic lake. Climatic Change 141, 759-773.

72 Climatic Change DOI 10.1007/s10584-017-1909-0

Atmospheric stilling leads to prolonged thermal stratification in a large shallow polymictic lake

R. Iestyn Woolway1 & Pille Meinson2 & Peeter Nõges2 & Ian D. Jones3 & Alo Laas2

Received: 16 September 2016 /Accepted: 21 January 2017 # The Author(s) 2017. This article is published with open access at Springerlink.com

Abstract To quantify the effects of recent and potential future decreases in surface wind speeds on lake thermal stratification, we apply the one-dimensional process-based model MyLake to a large, shallow, polymictic lake, Võrtsjärv. The model is validated for a 3-year period and run separately for 28 years using long-term daily atmospheric forcing data from a nearby meteorological station. Model simulations show exceptionally good agreement with observed surface and bottom water temperatures during the 3-year period. Similarly, simulated surface water temperatures for 28 years show remarkably good agreement with long-term in situ water temperatures. Sensitivity analysis demonstrates that decreasing wind speeds has resulted in substantial changes in stratification dynamics since 1982, while increasing air temperatures during the same period had a negligible effect. Atmospheric stilling is a phenomenon observed globally, and in addition to recent increases in surface air temperature, needs to be considered when evaluating the influence of climate change on lake ecosystems.

1 Introduction

Thermal stratification is a natural phenomenon that occurs in lakes as a result of the thermal expansion properties of water. It is determined by the balance between turbulence, which acts to enhance mixing, and buoyancy forces, which act to suppress turbulence and result in a vertical layering (Boehrer and Schultze 2008). The epilimnion is defined as that part of the water column immediately below the water surface, directly influenced by atmospheric

R. Iestyn Woolway and Pille Meinson are considered to be joint lead authors.

* R. Iestyn Woolway [email protected]

1 Department of Meteorology, University of Reading, Reading, UK RG6 6BB 2 Institute of Agricultural and Environmental Sciences, Centre for Limnology, Estonian University of Life Sciences, Tartu, Estonia 3 Centre for Ecology & Hydrology, Lancaster Environment Centre, Lancaster LA1 4AP, UK 73 Climatic Change forcing. The hypolimnion, the coolest and densest layer, lies in contact with the bottom of the lake and is separated from the epilimnion above by a temperature-driven density gradient known as the thermocline. The vertical layering that exists during stratification has several implications for the ecosystem, as it not only inhibits the downward penetration of direct vertical mixing and thus influences temperatures at depth (Livingstone 2003)butitalso separates processes of production and nutrient depletion in the epilimnion from processes of decomposition and nutrient regeneration in the hypolimnion and sediment. Stratification can be transient or persistent, and its duration can vary from hours to months or even be permanent in some lakes. For lakes of sufficient depth, the water column typically evolves seasonally from being isothermal in the early spring, developing stratification as the weather warms and then overturning sometime in the autumn. Some shallow lakes, however, do not develop a full seasonal stratification but rather experience alternate periods of mixing and stratification driven strongly by meteorological conditions. These lakes are often termed polymictic and, according to Hutchinson’s(1957) mixing classification, are among the most abundant at mid-latitudes. Polymictic lakes can be categorised further into continuous and discontinuous polymictic, depending on whether stratification occurs at most on a daily basis or for periods of several days to weeks, but with irregular interruption by mixing events (Lewis 1983). The mixing class a lake belongs to is mostly determined by lake morphometry, although the strength and extent of stratification are also influenced by extrinsic features of a lake, such as altitude (Woolway et al. 2015a), inflows (Rimmer et al. 2011), and meteorological conditions (Churchill and Kerfoot 2007), and intrinsic factors such as water clarity (Heiskanen et al. 2015). Mixing regime shifts, the term used to describe a change in the mixing classification of lakes, have been reported (Shatwell et al. 2016), and polymictic lakes have been described as particularly susceptible to mixing regime shifts due to climate change (Kirillin 2010). Despite the ubiquitous recognition that climate change affects lake stratification (Verburg and Hecky 2009), most studies typically only consider the response of a lake to increasing air temperature (Elo et al. 1998). Climate, however, is much more than temperature and studies have found climate-induced changes in a number of other meteorological variables known to influence lake stratification, including cloud cover (Eastman and Warren 2013) and solar radiation (Wild 2012). In addition, several studies have shown that, in recent years, wind speeds from around the world have been decreasing, in a phenomenon termed atmospheric stilling, where observations indicate that the annual land 10-m wind speed has decreased during the past few decades (Vautard et al. 2010), attributed to, among other things, increased surface roughness (Vautard et al. 2010;Bichetetal.2012), and changes in urbanisation (Xu et al. 2006). Despite the recognition that wind stress is one of the most important factors driving mixing in lakes, the response of lake stratification to atmospheric stilling has not yet been investigated. In this contribution, we aim to address this research gap by investigating the response of a large (average surface area 270 km2) and shallow (mean depth of 2.8 m; maximum depth = 6.1 m) lake, Võrtsjärv, to decreasing surface wind speeds. Traditionally, Võrtsjärv has been classified as continuous polymictic (Jaani 1973; Nõges & Nõges 2012;Laasetal.2012), but recent high frequency in situ temperature measurements demonstrate that the lake now stratifies occasionally, thus suggesting a shift to a discontinuous polymictic state. The main aim of this study is therefore to investigate if stratification in Võrtsjärv has changed in response to the large-scale decline in surface wind speed or to some other factor, and if the magnitude of observed atmospheric stilling is sufficient to influence stratification dynamics. Due to the lack of available long-term depth-resolved temperature data for Võrtsjärv, we restore a multi- 74 Climatic Change decadal stratification history of the lake with the use of a numerical model and evaluate, in terms of a sensitivity analysis, changes in stratification during the past ∼30 years, in response to atmospheric stilling.

2 Study site

This study took place in a large and shallow lake, Võrstjärv, situated in Central Estonia (58.29 °N, 26.03 °E; Fig. 1a). The lake is eutrophic, characterised by the following mean concentrations: total phosphorus (TP) 54 μgl−1, total nitrogen (TN) 1.6 mg l−1, chlorophyll-a 24 μgl−1 (Tuvikene et al. 2004; Nõges et al. 2007), and typically has a Secchi depth of less than 1 m.

3 Methods

3.1 Lake water temperature observations

Lake water temperatures used in this investigation were as follows: (i) high-resolution lake surface (0.5 m from surface) and bottom (0.5 m above bed) temperatures recorded with a PME

° ° 15 0 30 E

a ) b ° N -1 60 10

° N 50 5

° N Wind speed (m s 40 0 2 )

-1 c

° 60 N 1

° 59 N Wind anomaly (m s -1

25 d ° 58 N 15 5 -5 ° 0 40 80 120 km 57 N -15 Wind frequency (%)

° ° 2 ° ° ° ° E -25 3 E 24 E 25 E 26 E 27 E 28 1980 1990 2000 2010 Fig. 1 a Map of study site. Shown is the location of Võrtsjärv (dark grey), the location of the meteorological station at Tõravere (58.26 °N, 26.47 °E; triangle), and the stations in which wind speed measurements were available (circles): Pärnu-Sauga (58.37 °N, 24.50 °E), Tallinn-Harku (59.47 °N, 24.82 °E), Jõgeva (58.75 °N, 26.39 °E), Türi (58.80 °N, 25.42 °E), Valga (57.78 °N, 26.05 °E), and Viljandi (58.37 °N, 25.59 °E). Also shown are observations of b daily-averaged wind speed measured at Tartu-Tõravere; c annually averaged wind speed anomaly (relative to 1990–2010 annual average = 3 ms−1); and d occurrence frequencies (in %) for wind speeds >3 m s−1. Grey lines illustrate the wind speeds for an individual meteorological station in Estonia, and the thick black line represents the regional average (computed as the arithmetic mean of all stations) 75 Climatic Change

MiniDOT logger at 10 min intervals from 2013 to 2015; and (ii) long-term observations of surface temperatures measured from 1982 to 2009 at approximately monthly intervals.

3.2 Lake temperature model

In this investigation, we use the one-dimensional process-based lake model, MyLake (v1.2; Saloranta and Andersen 2007). MyLake has been designed to simulate accurately the daily vertical profiles of lake water temperature and thus thermal and density stratification, and has been used successfully in numerous lakes from around the world (Dibike et al. 2012; Pätynen et al. 2014; Couture et al. 2015). The modelling principles of MyLake are similar to other one- dimensional lake model codes, such as DYRESM-CAEDYM (Hamilton and Schladow 1997). The model code is written in MATLAB and has been designed to be computationally efficient and easily adaptable to different lakes. In addition to the model code, three different parameter and input data files are required to run MyLake. These include (i) time series of daily atmospheric data, (ii) lake morphometry and initial vertical water temperature profiles, and (iii) model parameter values. The model also requires information on water clarity, included in −1 terms of the light attenuation coefficient (Kd,m ). In this study, we followed the methods of Woolway et al. (2015b) and estimated Kd as a function of secchi depth (zsecchi)as:Kd =1.75/ zsecchi.Anaveragezsecchi was estimated from monthly averaged observations from Võrtsjärv. Initial profiles of water temperature were set equivalent to the surface air temperature at the start of the investigation. Thus, the water column was assumed vertically mixed. Model parameters were kept similar to those suggested in the user manual. MyLake also contains five switches, which can be used to disable some particular model processes if desired. These include (i) snow compaction, (ii) river inflow, (iii) sediment heat flux, (iv) self-shading, and (v) tracer simulation. Each of these features was disabled in this investigation. Simulated temperatures were validated for a 3-year period using the high-resolution lake surface and bottom temperatures described above.

3.3 Atmospheric forcing data

MyLake requires time series of atmospheric forcing data. Specifically, data for air pressure at station level (hPa), air temperature at 2 m (°C), relative humidity at 2 m (%), precipitation (mm day−1), solar radiation (MJ m−2 day−1), wind speed at 10 m (m s−1), and cloud cover fraction (ccf, 0–1) are required. Atmospheric pressure, p, was assumed constant in this modelling study and was estimated based on the surface elevation of the lake. We calculated pressure as (Woolway et al. 2015c): hi. 5:25588 p ¼ 101325 1−h 2:25577 10−5 100; where h is altitude (m). Using this calculation, we used a constant value of 1009 hPa. The only atmospheric forcing parameter that was not measured in situ was ccf. Therefore, we estimated ccf as ccf = 1 − s,wheres is the ratio of the measured solar radiation to the estimated clear-sky solar radiation at the lake surface, which was estimated following Woolway et al. (2015c). Specifically, the algorithm requires information on air temperature and relative humidity, and can be used to estimate clear-sky radiation at hourly resolution. Daily-averaged ccf was then calculated and used as an input in MyLake. 76 Climatic Change

Atmospheric forcing data used in this investigation were available from two different sources: (i) a monitoring buoy situated on Võrtsjärv, equipped with a Vaisala multi-weather station (WXT520) and Li-Cor pyranometer at a height of 2 m above the lake surface; and (ii) a meteorological station situated in Tõravere (58.26 °N, 26.47 °E), 30 km from the lake, data for which were provided by the Estonian Environmental Agency. These two datasets were used in the modelling study. In addition to the observations available from Tõravere, we also analysed wind speed observations from six other meteorological stations situated throughout Estonia. Wind speed measurements from Pärnu-Sauga (58.37 °N, 24.50 °E; 1973–2014) and Tallinn- Harku (59.47 °N, 24.82 °E; 1973–2016) were available from HadISD (Dunn et al. 2012), which is a quality-controlled synoptic meteorological dataset used for climate applications at sub-daily resolution. Wind speed measurements from Jõgeva (58.75 °N, 26.39 °E), Türi (58.80 °N, 25.42 °E), Valga (57.78 °N, 26.05 °E), and Viljandi (58.37 °N, 25.59 °E) meteorological stations were available from the Estonian Environmental Agency.

3.3.1 Scaling of atmospheric data

To scale the land-based measurements to be representative of over-lake conditions, we compared the recent (2013–2015) in situ observations with those measured at Tartu-Tõravere station. Observations of solar radiation, relative humidity, and air temperature were similar among the sites. Specifically, solar radiation had an R2 of 0.96 and a root mean square error (RMSE) of 22 W m−2; air temperature had an R2 of 0.98 and a RMSE of 0.88 °C; and relative humidity had an R2 of 0.86 and a RMSE of 1.65%, all of which were statistically significant (two-tail t test p < 0.001). To compare wind speed measurements between land and lake, we first converted the over-lake measurements to a standard height of 10 m, similar to that measured over land. Numerous methods have been proposed to convert wind speed, uz,at the measurement height, z, to wind speed at 10 m, u10 (e.g., Amorocho and DeVries 1980; Large and Pond 1981). Most of these methods, however, assume a neutral atmospheric boundary layer, which, in lakes, is not often true and can therefore bias u10 estimates. In this study, we not only corrected for measurement height and wind speed but also corrected for atmospheric stability, using the algorithm developed by Zeng et al. (1998) and included in the Lake Heat Flux Analyzer program (http://heatfluxanalyzer.gleon.org/; Woolway et al. 2015c).

After converting uz to u10, we compared the lake and land-based wind measurements, ulake and 2 uland (R = 0.6, two-tail t test p < 0.001), and found a relationship of

ulake ¼ 2:17 uland; which was then used to scale the long-term wind measurements to be representative of over- lake conditions.

3.4 MyLake experiments

To evaluate the relative contribution of changes in air temperature and changes in wind speed, the two variables that varied significantly in recent years (see Section 4), on stratification dynamics in Võrtsjärv, we performed a series of MyLake simulations. Firstly, lake temperatures were simulated using the observed atmospheric forcing data over the 28-year period. These simulations were then repeated, except using the wind speed observations for 1982, the first year of available meteorological data, for every year, thereby removing any effect of wind 77 Climatic Change speed change from the results. A similar set of simulations were then performed but with the air temperatures, rather than the wind speeds, held at the 1982 levels in order to remove the impact of systematic changes in air temperature. The first year of observations (i.e., 1982) was chosen to represent conditions that have not varied since the start of the investigation, and thus to investigate how would stratification in the lake have evolved had air temperatures or wind speed not changed since 1982. To ensure that using the data from 1982 to determine a constant annual cycle in both wind speed and air temperature did not bias our modelling results, we repeated all simulations but replacing the 1982 constant annual cycle with the climatological average daily meteorological forcing for both these variables. This represents the mean atmospheric conditions for each day throughout the 28 years. These model results were then compared to those produced from the first sensitivity analysis (i.e., using the 1982 data as a constant annual cycle) for validation. Though vertical temperature differences are often used to define stratification (Stefan et al. 1996;Woolwayetal.2014), stratification is the result of the associated density differences and these vary non-linearly with temperatures. Thus, at 5 °C, a 1 °C temperature difference is equivalent to less than a 0.025 kg m−3 density difference, while at 19 °C a 1 °C temperature difference represents more than a 0.2 kg m−3 density difference. To capture all stratification periods here, we therefore used a threshold of a 0.025 kg m−3 density difference between the top and the bottom of the lake to define stratification. For all model experiments, the effects of wind speed and air temperature on stratification were determined by firstly calculating the number of stratified days per year (May–August, see Section 4). A linear regression model was then used to evaluate the rate of change in the number of stratified days during the 28-year period. To evaluate if the regression slopes were different among model runs, and thus determine if the influence of air temperature and/or wind speed on stratification were statistically distinguishable, we followed an analysis of covariance (ANCOVA) approach. Specifically, ANCOVA was used to compare the different regression slopes by testing the effect of a categorical factor (e.g., different model) on the dependent variable (e.g., the number of stratified days; y-variable) while controlling for the effect of the continuous co-variable (e.g., time; x-variable). A statistically significant interaction between the categorical factors demonstrates that the regressions have statistically different slopes. In contrast, if the interaction is not statistically significant, the covariate has the same effect for all levels of the categorical factor, meaning that a single regression line can be used to represent the different relationships. All calculations in this study were performed in R (R Development Core Team 2014).

4Results

4.1 Long-term atmospheric forcing in Võrtsjärv

Over the last 28 years, wind speed has been declining in all seven of the meteorological stations in Estonia (Fig. 1). This is seen in the raw daily data from Tartu-Tõravere (Fig. 1b)and in the annually averaged wind speeds from all seven meteorological stations situated throughout Estonia (Fig. 1c). The annually averaged wind speeds are shown in Fig. 1c as anomalies relative to the 1990–2010 average to account for the different magnitude of wind speeds observed throughout Estonia. The regional average change is −0.29 ± 0.03 m s−1 decade−1, varying from −0.11 ± 0.05 to −0.68 ± 0.05 m s−1 decade−1 in Tartu-Tõravere and Valga, respectively (Fig. 1c), with declines in all stations being statistically significant (two-tail t test 78 Climatic Change p < 0.05). In addition, the observational data suggests a decrease in the frequency of wind speeds >3 m s−1 (calculated as the % of time in which wind speeds exceeded 3 m s−1 during a given year. This threshold was chosen based on the mean wind speed calculated in Tartu- Tõravere throughout the study period = 3 m s−1), demonstrating an increase in the frequency of calm days (shown as negative values in Fig. 1d) leading to a regionally averaged change of −6.25 ± 0.77% decade−1 (Fig. 1d). Among the other meteorological data measured in situ, only air temperature experienced a statistically significant (p < 0.05) change since 1982, where the annually averaged air temperature increased at a rate of 0.52 ± 0.21 °C decade−1 (p = 0.0185). None of the other atmospheric forcing data experienced a significant change in their annual averages during the study period, nor did the lake water level (Fig. 2).

4.2 Lake temperature modelling

The simulations of temperature in both surface and bottom waters of Võrtsjärv showed very good agreement with observed data during the 3-year validation study (Fig. 3a, b). In general,

Fig. 2 Observations of annually 9 a averaged a surface air temperature;

C) 6 b relative humidity; c global solar o ( radiation; d precipitation; and e a 3 mean depth, in Võrtsjärv. T Meteorological data were 0 measured 30 km from the lake at 1980 1985 1990 1995 2000 2005 2010 Tartu-Tõravere meteorological 85 station. Linear regressions of the b statistically significant (p <0.05) relationships are shown 80 RH (%) 75 1980 1985 1990 1995 2000 2005 2010 12 c )

-2 11 10 9 G (MJ m G (MJ 8 1980 1985 1990 1995 2000 2005 2010 3 d ) -1 2 (mm d Precipitation 1 1980 1985 1990 1995 2000 2005 2010 4 e 3 2

Mean 1 depth (m) 0 1980 1985 1990 1995 2000 2005 2010

79 Climatic Change

30 30 a b C) o C)

o 25 ( 25 (

top 20 20 bottom

15 15

Modeled T o 10 o 10 RMSE = 0.63 C T Modeled RMSE = 0.71 C 10 15 20 25 30 10 15 20 25 30 Observed T (oC) Observed T (oC) top bottom

30 c

20 C) o

T ( 10

0 1980 1982 1985 1987 1990 1992 1995 1997 2000 2002 2005 2007 2010

30 30 d e C) C) o 20 o 20

10 10 Modeled( T Modeled( T RMSE = 1.75 oC RMSE = 1.37 oC 0 0 01020300 102030 Observed T (oC) Observed T (oC) Fig. 3 Comparison of modelled and observed a surface and b bottom water temperatures for 2013–2015. c Comparison of long-term modelled (grey line) and observed (black dots) lake surface water temperature, showing a comparison d throughout the year and e during spring/summer (MJJA) lake surface water temperatures were simulated with the higher accuracy, although the mean absolute difference (MAD) and RMSE of bottom water temperature simulations were under 1 °C. The mean MAD among the 3 years were 0.48 °C (surface) and 0.53 °C (bottom), and the mean RMSE among the 3 years were 0.63 °C and 0.71 °C for surface and bottom waters, respectively. The maximum difference between observed and simulated surface water temperatures was 1.75 °C. Bottom water temperature simulations were marginally better in terms of maximum difference, which was calculated as 1.29 °C. Using 28 years of atmospheric forcing data from Tartu-Tõravere, MyLake generally captured the temporal dynamics of lake surface temperatures with great success (Fig. 3c). The model performed less well in winter, occasionally resulting in large temperature differences between observed and modelled values (Fig. 3d), in particular during periods of ice cover, which is an important factor for lakes in the northern hemisphere, having decreased in recent decades (Magnuson et al. 2000; Kheyrollah Pour et al. 2012;Duguayetal.2013). During the May–August (MJJA) months, modelled temperatures closely matched those observed (MAD = 0.97 °C; RMSE = 1.37 °C; R2 = 0.88; Fig. 3e). These months were used to assess changes in stratification from the model. 80 Climatic Change

4.2.1 Modelled changes in thermal stratification

Simulated water density profiles for the 28-year period reveal a noticeable change in recent years with a substantial increase in the average top-bottom density difference (Fig. 4a). During the start of the investigation, the density difference between top and bottom waters was minimal, with bottom waters only marginally exceeding those at the surface. However, in recent years, the density difference increased noticeably. The model results demonstrate a significant relationship between wind speed and the number of stratified days in MJJA over the 28 years (Fig. 4b). Specifically, we find that the number of stratified days per year in Võrtsjärv increased at a rate of 3.16 ±0.86 (p = 0.001) days with a 1 m s−1 decrease in annually averaged wind speed. Analysis of the model results for the past 28 years demonstrated an increase of over 80% in the number of stratified days in the second half of the study period (1995– 2009) compared to the first half (1982–1995) (Fig. 4c) and that the overall number of stratified days increased at a rate of 2.1 ± 0.6 days per decade (two-tail t test p < 0.001).

0 14 a b -1 12

10 -2 8 -3 1982 1986 6 Depth (m) 1990 -4 1994 4 1998 Number of stratified days -5 2002 2 2007 -6 0 -6 -4 -2 0 2 4 6 2 2.5 3 3.5 4 4.5 kg m-3 (x 10-3, anomaly) Wind speed (m s-1)

15 c

5 Number of stratified days

0 1980 1985 1990 1995 2000 2005 2010 Fig. 4 a Simulated water column density profiles averaged for 4 months during spring/summer (MJJA) for selected example years. Shown are the density profile anomalies relative to the 1982 to 2010 average density profile. b Relationship between the average wind speeds measured at Tartu-Tõravere meteorological station and the simulated number of stratified days from model runs with observed meteorological data. c Time series of the simulated number of stratified days from model runs with observed meteorological data 81 Climatic Change

4.2.2 Impacts of wind change versus air temperature change

Our model sensitivity analysis illustrates that the decrease in wind speed is the key influence on the number of stratified days and that the influence of increasing air temperature was minimal. Running MyLake with a constant annual air temperature cycle demonstrates that the number of stratified days still increased markedly in recent years (Fig. 5a), at a rate of 1.7 ± 0.6 days per decade (two-tail t test p < 0.001). This calculated rate of change in the number of stratified days is statistically indistinguishable compared to the model run using the full observational data, as evaluated via ANCOVA between regression slopes. In contrast, when the annual cycle in surface wind speed was kept constant, but air temperature allowed to vary, the model results show no statistically significant increase (p > 0.05) in the number of stratified days during the past 28 years (Fig. 5b). We evaluated further the impact of changing air temperature on the number of stratified days by running the model with air temperatures from 1987 (Fig. 5c) and 2009 (Fig. 5d), the coolest and warmest years, respectively, held constant. The calculated number of stratified days showed a statistically significant increase in both

Fig. 5 Comparisons of the 15 number of stratified days from a model runs with a constant 1982 10 air temperature annual cycle; b constant 1982 wind speed annual 5 cycle; c constant 1987 air temperature annual cycle; and d constant 2009 air temperature 0 annual cycle. Linear regressions of 1980 1985 1990 1995 2000 2005 2010 the statistically significant 15 (p < 0.05) relationships are shown b 10

0 1980 1985 1990 1995 2000 2005 2010

15 c

Number of stratifiedNumber of days 10

0 1980 1985 1990 1995 2000 2005 2010

15 d 10

0 1980 1985 1990 1995 2000 2005 2010

82 Climatic Change model runs, with neither of these simulations being statistically different from the model using observed meteorological forcing data. In particular, the interaction term in the ANCOVA model demonstrated that they were not significant (p > 0.1), meaning that the regression slopes were statistically indistinguishable. We repeated all MyLake simulations using the daily climatological annual cycle in both wind speed and air temperature instead of using the 1982 values held constant. Similar to the other model runs, the regression slopes were not statistically different and wind speed was again found to be the dominant driver for the increased number of stratified days in recent years.

5 Discussion

Based on long-term meteorological data available from in situ meteorological stations in Estonia, we found a large-scale decrease in surface wind speeds during the past ∼30 years, consistent with other locations globally (Vautard et al. 2010). Previous investigations attribute this regional (i.e., in Estonia) decrease to changes in atmospheric circulation (Jaagus and Kull 2011) with others suggesting that changes in surface roughness surrounding the meteorological stations (Suursaar and Kullas 2006), and inconsistencies introduced by instrumental change (Jaagus and Kull 2011), may bias the observed trends. Analysis of wind speed data from seven meteorological stations, some of which were separated by up to 200 km, provide evidence against the latter and suggest that the observed decrease in surface wind speed is not confined to a particular location or meteorological station in Estonia. In this investigation, we examined the influence of this large-scale decrease in surface wind speed on thermal stratification in Võrtsjärv, a large and shallow lake situated in Central Estonia. High-resolution in situ temperature measurements from Võrtsjärv illustrate that, in recent years, the lake stratifies for several days during spring/summer. This is unexpected, as previous investigations have described Võrtsjärv as traditionally having no thermal stratification (Frisk et al. 1999). In this study, we found a strong influence of atmospheric stilling on stratification dynamics and suggest that a decrease in surface wind speed has resulted in an increase in the number of stratified days in recent years. In particular, our model results illustrate that the number of stratified days in Võrtsjärv has increased at a rate of 2 days per decade. Few studies have examined impacts of systematic wind speed changes on lake stratification, but those that do have tended to focus on the influence of local changes such as deforestation and forest regrowth (France 1997; Tanentzap et al. 2007). To our knowledge, this is the first study to relate regional atmospheric stilling to changes in lake stratification. In contrast, many studies have discussed the likely effects of rising air temperatures on lake stratification and the focus on this one parameter has drawn attention away from the possible influences of other aspects of directional climate changes, with only a very few studies having investigated the response of lake thermal dynamics to changing surface wind speeds (Tanentzap et al. 2008; Desai et al. 2009; Kerimoglu and Rinke 2013). Despite air temperatures increasing in Estonia (0.52 °C decade−1), at a rate twice that of the global average (0.25 °C decade−1) (Hartmann et al. 2013) over a similar time period (1979–2012), our modelling suggests this increase has not been sufficient to nudge the large, shallow, typically mixed, Võrtsjärv to a different mixing regime. The contemporaneous changes in wind speed, though, have been sufficient to do this. This suggests that ongoing changes in wind speed resulting from atmospheric stilling may be a more potent driver of stratification change in 83 Climatic Change some lakes than the changes in air temperature, despite the overwhelming preference for studying the latter. Other factors are also known to influence stratification in some lakes. For example, water level has previously been shown to be an important factor influencing thermal stratification in Lake Kinneret, Israel (Rimmer et al. 2011). Moreover, lake water level has been described as one of the most influential factors affecting the physics (Nõges & Järvet 1995), chemistry (Nõges & Nõges 1999), and biology (Nõges et al. 2003; Järvalt et al. 2005) of Võrtsjärv. However, in this study we found no systematic change in water level during the past 28 years (e.g., Fig. 2e). Our study highlights the importance of taking high frequency measurements in mixed and polymictic systems, without which transient periods of stratification and long-term changes to stratification may be missed. Such measurements are proliferating globally (e.g., Woolway et al. 2016) enabling more detailed studies of stratification to now take place than would have been possible just a decade ago. High frequency, in situ measurements are being further complemented by advancements in Earth Observation. Projected changes in lake stratification under future climate are likely to induce a variety of impacts on the ecology and chemistry of lakes, which in turn will have implications for water quality and for adaptation strategies and management of lakes. For example, thermal stratification has been shown to restrict the supply of oxygen to deep waters (Foley et al. 2012;North et al. 2014), leading to deoxygenation in productive lakes and potential increases in the rate of recycling of phosphorus from the sediment to the water column (Søndergaard et al. 2003), with consequences for lake productivity (O’Reilly et al. 2003; Verburg et al. 2003). Also, ecosystem properties such as metabolism (Staehr et al. 2010) and gas flux at the air-water interface (Coloso et al. 2011) are highly influenced by lake thermal dynamics, thus being important for the global carbon cycle. Changing stratification can affect the likelihood of cyanobacterial bloom formation (Paerl and Huisman 2008; Jöhnk et al. 2008) and, through the impact on both temperature and oxygen, can impact the available habitat for fish species (Jones et al. 2008; Kangur et al. 2016). Our results illustrate that if wind speeds continue to decrease, Võrtsjärv may stratify for longer, having large consequences for the ecosystem. With lakes worldwide being subjected to increased air temperature against the potential backdrop of decreasing wind speed, future investigations should aim to understand the relative importance of these factors and how they interact to influence thermal stratification and potentially impair water quality.

Acknowledgements RIW was funded by EUSTACE (EU Surface Temperature for All Corners of Earth) which received funding from the European Union’s Horizon 2020 Programme for Research and Innovation, under Grant Agreement no 640171. Data collection in this manuscript was collected under the Estonian Ministry of Education and Research grants (IUT 21–02 and PUT 777). Initial discussions for this work took place at a NETLAKE (Networking Lake Observatories in Europe) working group meeting in Riga, Latvia. The authors are grateful for the support of NETLAKE and its chair, Eleanor Jennings. We thank three anonymous reviewers for suggestions that improved an earlier version of this manuscript.

Authorship statement PN, AL, and PM developed the concept of the study; RIW, IDJ, and AL designed the model experiments; RIW and IDJ analysed the results; RIW and PM led the paper writing; all authors contributed to discussions, revisions and the production of the final manuscript.

Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and repro- duction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

84 Climatic Change

References

Amorocho J, DeVries JJ (1980) A new evaluation of the wind stress coefficient over water surfaces. J Geophys Res 85:433–442 Bichet A, Wild M, Folini S, Schär C (2012) Causes for decadal variations of wind speed over land: sensitivity studies with a global climate model. Geophys Res Lett 39(11). doi:10.1029/2012GL051685 Boehrer B, Schultze M (2008) Stratification of lakes. Rev Geophys 46(2). doi:10.1029/2006RG000210 Churchill JH, Kerfoot WC (2007) The impact of surface heat flux and wind on thermal stratification in Portage Lake, Michigan. J Great Lakes Res 33:143–155 Coloso JJ, Cole JJ, Pace ML (2011) Short-term variation in thermal stratification complicates estimation of lake metabolism. Aquat Sci 73:305–315 Couture R-M, de Wit HA, Tominaga K, Kiuru P, Markelov I (2015) Oxygen dynamics in a boreal lake responds to long-term changes in climate, ice phenology, and DOC inputs. J Geophys Res Biogeosci 120:2441–2456 Desai A et al (2009) Stronger winds over a large lake in response to weakening air-to-lake temperature gradient. NatGeosci2:855–858 Dibike Y, Prowse T, Bonsal B, Rham LD, Saloranta T (2012) Simulation of North American lake-ice cover characteristics under contemporary and future climate conditions. Int J Climatol 32:695–709 Dunn RJH, Willett KM, Thorne PW et al (2012) HadISD: a quality-controlled global synoptic report database for selected variables at long-term stations from 1973–2011. Clim Past 8:1649–1679. doi:10.5194/cp-8-1649- 2012 Duguay C, Brown L, Kang K-K, Kheyrollah Pour H (2013) Arctic—lake ice, in BState of the Climate in 2012^. Bull Am Meteorol Soc 94:S124–S126 Eastman R, Warren SG (2013) A 39-yr survey of cloud changes from land stations worldwide 1971–2009: long- term trends, relation to aerosols, and expansion of the tropical belt. J Clim 26:1286–1303 Elo A-R, Huttula T, Peltonen A, Virta J (1998) The effects of climate change on the temperature conditions of lakes. Boreal Environ Res 3:137–150 Foley B, Jones ID, Maberly SC, Rippey B (2012) Long-term changes in oxygen depletion in a small temperate lake: effects of climate change and eutrophication. Freshw Biol 57:278–289 France R (1997) Land-water linkages: influences of riparian deforestation on lake thermocline depth and possible consequences for cold stenotherms. Can J Fish Aquat Sci 54:1299–1305 Frisk T, Bilaletdin Ä, Kaipainen H, Malve O, Möls M (1999) Modelling phytoplankton dynamics of the eutrophic Lake Võrtsjärv, Estonia. Hydrobiologia 414:59–69 Hamilton DP, Schladow SG (1997) Prediction of water quality in lakes and reservoirs. Part 1—model description. Ecol Model 96:91–110 Hartmann DL (2013) IPCC fifth assessment report, climate change 2013. In: Stocker TF (ed) The physical science basis. Cambridge Univ. Press, Cambridge Heiskanen J, Mammarella I, Ojala A, Nordbo A (2015) Effects of water clarity on lake stratification and lake- atmosphere heat exchange. J Geophys Res - Atmos 120:7412–7428 Hutchinson E (1957) A treatise on limnology vol 1. Wiley, New York, 1015p Jaagus J, Kull A (2011) Changes in surface wind directions in Estonia during 1966–2008 and their relationship with large-scale atmospheric circulation. Estonian J Earth Sci 60:220–231 Jaani A (1973) Hydrobiology. In: Timm, T. (Ed.), L. Võrtsjärv (in Estonian) Valgus, Tallinn, pp. 37–60 Järvalt A, Laas A, Nõges P, Pihu E (2005) The influence of water level fluctuations and associated hypoxia on the fishery of Lake Vortsjarv, Estonia. Ecohydrol Hydrobiol 4:487–497 Jöhnk KD, Huisman J, Sharples J, Sommeijer B, Visser PM, Stroom JM (2008) Summer heatwaves promote blooms of harmful cyanobacteria. Glob Change Biol 14:495–512 Jones ID, Winfield IJ, Carse F (2008) Assessment of long-term changes in habitat availability for Arctic charr (Salvelinus alpinus) in a temperate lake using oxygen profiles and hydroacoustic surveys. Freshwater Biol 53:393–402 Kangur K, Ginter K, Kangur P, Kangur A, Nõges P, Laas A (2016) Changes in water temperature and chemistry preceding a massive kill of bottom-dwelling fish: an analysis of high-frequency buoy data of shallow Lake Võrtsjärv (Estonia). Inland Waters 6:535–542. doi:10.5268/IW-6.4.869 Kerimoglu O, Rinke K (2013) Stratification dynamics in a shallow reservoir under different hydro- meteorological scenarios and operational strategies. Wat Resour Res 49:7518–7527 Kirillin G (2010) Modeling the impact of global warming on water temperature and seasonal mixing regimes in small temperate lakes. Boreal Environ Res 15:279–293 Kheyrollah Pour H, Duguay C, Martynov A, Brown LC (2012) Simulation of surface temperature and ice cover of large northern lakes with 1-D models: a comparison with MODIS satellite data and in situ measurements. Tellus A 64:17614. doi:10.3402/tellusa.v64i0.17614

85 Climatic Change

Laas A, Nõges P, Kõiv T, Nõges T (2012) High frequency metabolism study in a large and shallow temperate lake revealed seasonal switching between net autotrophy and net heterotrophy. Hydrobiologia 694:57–74 Large WG, Pond S (1981) Open ocean momentum flux measurements in moderate to strong winds. J Phys Oceanogr 11:324–336 Lewis WMJR (1983) A revised classification of lakes based on mixing. Can J Fish Aquat Sci 40:1779–1787 Livingstone DM (2003) Impact of secular climate change on the thermal structure of a large temperate central European lake. Clim Chang 57:205–225 Magnuson JJ et al (2000) Historical trends in lake and river ice cover in the Northern Hemisphere. Science 289: 1743–1746 Nõges P, Järvet A (1995) Water level control over light conditions in shallow lakes. Report Series in Geophysics. University of Helsinki 32:81–92 Nõges P, Nõges T (2012) Lake Võrtsjärv. In: Bengtsson L, Herschy R, Fairbridge R (eds) Encyclopedia of lakes and reservoirs. Springer, Dordrecht, pp 850–863 Nõges T, Järvet A, Kisand A, Laugaste R, Loigu E, Skakalski B, Nõges P (2007) Reaction of large and shallow lakes Peipsi and Võrtsjärv to the changes of nutrient loading. Hydrobiologia 584:253–264 Nõges T, Nõges P (1999) The effect of extreme water level decrease on hydrochemistry and phytoplankton in a shallow eutrophic lake. Hydrobiologia 408:277–283 Nõges T, Nõges P, Laugaste R (2003) Water level as the mediator between climate change and phytoplankton composition in a large shallow temperate lake. Hydrobiologia 506:257–263 North RP, North RL, Livingstone DM, Köster O, Kipfer R (2014) Long-term changes in hypoxia and soluble reactive phosphorus in the hypolimnion of a large temperate lake: consequences of a climate regime shift. Glob Change Biol 20:811–823 O’Reilly CM, Alin SR, Plisner P, Cohen AS, McKee BA (2003) Climate change decreases aquatic ecosystem productivity of Lake Tanganyika. Nature 424:766–768 Paerl HW, Huisman J (2008) Blooms like it hot. Science 320:57–58 Pätynen A, Elliott JA, Kiuru P, Sarvala J, Ventelä A-M, Jones RI (2014) Modelling the impact of higher temperature on the phytoplankton of a boreal lake. Boreal Environ Res 19(1):66–78 R Development Core Team (2014) R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna [Available at http://www.R-project.org/.] Rimmer A, Gal G, Opher T, Lechinsky Y, Yacobi YZ (2011) Mechanisms of long-term variations in the thermal structure of a warm lake. Limnol Oceanogr 56:974–988 Saloranta TM, Andersen T (2007) MyLake—a multi-year lake simulation model code suitable for uncertainty and sensitivity analysis simulations. Ecol Model 207:45–60 Shatwell T, Adrian R, Kirillin G (2016) Plantonik events may cause polymictic-dimictic regime shifts in temperate lakes. Sci Reports 6(24361). doi:10.1038/srep24361 Søndergaard M, Jensen JP, Jeppesen E (2003) Role of sediment and internal loading of phosphorus in shallow lakes. Hydrobiologia 506:135–145 Staehr PA et al (2010) Lake metabolism and the diel oxygen technique: state of the science. Limnol Oceanogr Methods 8:628–644 Stefan HG, Hondzo M, Fang X, Eaton JG, McCormick JH (1996) Simulated long-term temperature and dissolved oxygen characteristics of lakes in the north-central United States and associated fish habitat limits. Limnol Oceanogr 41:1124–1135 Suursaar U, Kullas T (2006) Influence of wind climate on the mean sea level and current regime in the coastal waters of west Estonia, Baltic Sea. Oceanologia 48:361–383 Tanentzap AJ et al (2007) Cooling lakes while the world warms: effects of forest regrowth and increased dissolved organic matter on the thermal regime of a temperate, urban lake. Limnol Oceanogr 53:404–410 Tuvikene L, Kisand A, Tõnno I, Nõges P (2004) In: Pihu E, Raukas A, Haberman J (eds) Chemistry of lake water and bottom sediments. Estonian Encyclopaedia Publishers, Tallinn, pp 89–101 Vautard R, Cattiaux J, Yiou P, Thepaut J, Ciais P (2010) Northern Hemisphere atmospheric stilling partly attributed to an increase in surface roughness. Nat Geo Sci 3:756–761 Verburg P, Hecky RE (2009) The physics of the warming of Lake Tanganyika by climate change. Limnol Oceanogr 54:2418–2430 Verburg P, Hecky RE, Kling H (2003) Ecological consequences of a century of warming in Lake Tanganyika. Science 301:505–507 Wild M (2012) Enlightening global dimming and brightening. Bull Am Meteorol Soc 93(1):27–37 Woolway RI, Livingstone DM, Kernan M (2015a) Altitudinal dependence of a statistically significant diel temperature cycle in Scottish lochs. Inland Waters 5:311–316. doi:10.5268/IW-5.4.854 Woolway RI, Jones ID, Hamilton DP, Maberly SC, Muraoka K, Read JS, Smyth RL, Winslow LA (2015b) Automated calculation of surface energy fluxes with high-frequency lake buoy data. Environ Model Softw 70:191–198. doi:10.1016/j.envsoft.2015.04.013

86 Climatic Change

Woolway RI, Maberly SC, Jones ID, Feuchtmayr H (2014) A novel method for estimating the onset of thermal stratification in lakes from surface water measurements. Wat Resour Res 50. doi:10.1002/2013WR014975 Woolway RI, Jones ID, Feuchtmayr H, Maberly SC (2015c) A comparison of the diel variability in epilimnetic temperature for five lakes in the English Lake District. Inland Waters 5:139–154. doi:10.5268/IW-5.2.748 Woolway RI et al (2016) Diel surface temperature range scales with lake size. PLoS ONE 11(3):e0152466. doi:10.1371/journal.pone.0152466 Xu M, Chang CP, Fu C, Qi Y, Robock A, Robinson D, Zhang HM (2006) Steady decline of east Asian monsoon winds, 1969–2000: evidence from direct ground measurements of wind speed. J Geophys Res 111:D24111. doi:10.1029/2006JD007337 Zeng X, Zhao M, Dickinson RE (1998) Intercomparison of bulk aerodynamic algorithms for the computation of sea surface fluxes using TOGA COARE and TAO data. J Clim 11:2628e2644

87 III

88 Laas, Alo; Cremona, Fabien; Meinson, Pille; Rõõm, Eva-Ingrid; Nõg- es, Tiina; Nõges, Peeter (2016). Summer depth distribution profi les of dissolved CO2 and O2 in shallow temperate lakes reveal trophic state and lake type specifi c diff erences. Science of the Total Envrionment 566- 567, 63-75. 89 Science of the Total Environment 566–567 (2016) 63–75

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Summer depth distribution proﬁles of dissolved CO2 and O2 in shallow temperate lakes reveal trophic state and lake type speciﬁc differences

Alo Laas ⁎, Fabien Cremona, Pille Meinson, Eva-Ingrid Rõõm, Tiina Nõges, Peeter Nõges

Centre for Limnology, Institute of Agricultural and Environmental Sciences, Estonian University of Life Sciences, Kreutzwaldi 5, 51014 Tartu, Estonia

HIGHLIGHTS GRAPHICAL ABSTRACT

• We measured CO2 and DO proﬁles in 8 lake types at sub-hourly intervals over aweek. • Surface layers of alkaline and dystrophic lake were steadily supersaturated

with CO2. • 3 lake types acted as CO2 sinks, another 3 were in equilibrium with atmospheric

CO2. • Vertical dissolved gas gradients occurred even in thermally non-stratiﬁed lakes. • Differences in trophic state and depth accounted most for gas regime differences.

article info abstract

Article history: Knowledge about dissolved oxygen (DO) and carbon dioxide (CO2) distribution in lakes has increased considerably Received 12 February 2016 over the last decades. However, studies about high resolution dynamics of dissolved CO2 in different types of lakes Received in revised form 5 May 2016 over daily or weekly time scales are still very scarce. We measured summertime vertical DO and CO2 proﬁles at sub- Accepted 6 May 2016 hourly intervals during one week in eight Estonian lakes representing different lake types according to European Available online xxxx Water Framework Directive. The lakes showed considerable differences in thermal stratiﬁcation and vertical distri- Editor: D. Barcelo bution of dissolved oxygen and CO2 as well as different diurnal dynamics over the measurement period. We observed a continuous CO2 supersaturation in the upper mixed layer of the alkalitrophic (calcareous groundwater- “ ” Keywords: fed) lake and the dark soft-water lake showing them as CO2 emitting chimneys although with different underlying Dissolved oxygen mechanisms. In three lake types strong undersaturation with CO2 occurred in the surface layer characterising them

Dissolved carbon dioxide as CO2 sinks for the measurement period while in another three types the surface layer CO2 was mostly in equilib- Proﬁles rium with the atmosphere. Factor analysis showed that DO% in the surface layer and the strength of its relationship WFD lake types with CO2% were positively related to alkalinity and negatively to trophic state and DOC gradients, whereas deeper

lakes were characterised by higher surface concentration but smaller spatial and temporal variability of CO2.Multi- ple regression analysis revealed lake area, maximum depth and the light attenuation coefﬁcient as variables affecting the largest number of gas regime indicators. We conclude that the trophic status of lakes in combination with type speciﬁcfeaturessuchasmorphometry,alkalinityandcolour(DOC)determinesthedistributionanddynamicsofdis-

⁎ Corresponding author. E-mail address: [email protected] (A. Laas).

90 64 A. Laas et al. / Science of the Total Environment 566–567 (2016) 63–75

1. Introduction In this paper we present the vertical distribution of dissolved CO2 and DO in 8 hemiboreal lake types obtained by direct continuous in

Because aquatic primary producers consume CO2 for photosynthesis situ measurements using optical sensors. Each lake in our selection rep- at roughly the same rate as they release DO into water, and respiration resents one of the eight lake types in Estonia according to the European can be seen as a reverse process of that, a strong coupling between DO Water Framework Directive (WFD) typology (Table 1). Despite Estonia and dissolved CO2 in lake water could be assumed, at least in productive being a small country, the diverse geological setting supports a broad systems where the intensity of both processes is higher, making their ef- variety of natural lake types. Based on six indicators of the gas regime, fect on the concentration dynamics of dissolved gases more visible. Field we study how the broad range of observed saturation levels of both measurements in lakes show that in some cases the dynamics of dis- CO2 and DO and their coupling is related to lake type and trophic state solved gases are indeed caused mainly by metabolism (Johnson et al., parameters. We demonstrate the existence of gradients in the distribu- 2010.). In most cases, however, the distributions of DO and dissolved tion of dissolved gases even under isothermal conditions of lakes and

CO2 are strongly decoupled for several reasons including CO2 additions discuss the potential of vertical CO2 measurements as a method to en- from allochthonous organic matter degradation (Jonssonetal.,2003)or hance understanding of carbon dynamics in aquatic environments. volcanism (Jones, 2010), compartmentalisation of lake environments fi by thermal strati cation (Baehr and DeGrandpre, 2004), pH depen- 2. Material and methods dence of dissolved CO2 concentrations resulting from the functioning of the carbonate buffer (Marcé et al., 2015; Weyhenmeyer et al., 2.1. Study sites 2015), nitrate or sulphate respiration in anoxic conditions (Liikanen et al., 2002), methanogenesis and methane release by ebullition (Casper This study was conducted in eight Estonian lakes (Fig. 1) each be- et al., 2000), and anoxygenic photosynthesis (Bryant and Frigaard, longing to a different lake type according to the EU WFD typology 2006). Consistent low frequency oscillation in the CO2 partial pressure, which is based on lake area, alkalinity, conductivity, chloride content, DO, and water temperature time-series may also be caused by hydrody- thermal stratification, and colour (Table 1). The two largest lakes in Es- namic processes such as seiches (Baehr and DeGrandpre, 2002). Hence, tonia, Peipsi (3555 km2, fifth largest lake in Europe) and Võrtsjärv the question remains, to what extent the distribution and dynamics of (270 km2), form individual types referred to, correspondingly, as V- dissolved CO2 and DO are controlled by trophic state determining the Large and Large. They were allocated to separate lake types in the fi intensity of lake metabolism and what is the role of lake type speci c WFD compliant lake typology (ME, 2009), because strong wind induced features such as lake morphometry, alkalinity or water colour in modi- mixing makes them incomparable with smaller lakes in the region, fying the gas regime. whereas stronger sediment resuspension in the shallower Võrtsjärv Lakes are frequently super-saturated with CO2 relative to the atmo- causes higher turbidity and light limitation clearly distinguishing it sphere (Cole et al., 1994; Prairie et al., 2002; Jonsson et al., 2003; from the deeper Peipsi. Kortelainen et al., 2006). CO2 supersaturation can be caused by several The remaining ~1200 small lakes are grouped into six types (Ott, alternative processes: by negative net ecosystem production (NEP; 2006; ME, 2009). To make the text easier to follow, the eight lakes in Cole et al., 2000), photochemical degradation of dissolved organic car- this study are further referred to by their abbreviated type names bon (DOC) (Vachon et al., 2016) or high dissolved inorganic carbon (Table 1). In addition, the lakes in this study differed also by catchment fl (DIC) in ow from surface- or groundwater (Marcé et al., 2015; land use, trophic status, and water retention time (Table 2). All lakes Weyhenmeyer et al., 2015). Excluding the temperature effect on gas sol- were rather shallow with a mean depth b10m. Measurements in all ubility, DO supersaturation in lakes can only be reached by positive net lakes were carried out within a 2-month period from July to September N ecosystem production (NEP 0 if primary production exceeds respira- 2014. (See Table 3.) tion) occurring during limited time in the growing season when chem- Our selection of lakes represented rather adequately the broad range ical and physical conditions support intensive photosynthesis. of lake characteristics of the region both by type specific features and Therefore, supersaturation of the surface layer of lakes with dissolved along the trophic scale. The size of lakes ranged from 2 ha to N200.000 CO2 is more common in various geographic regions than supersatura- ha, i.e. over 5 orders of magnitude, Chl a over 3 orders, Kd and TP over tion with DO. – 2 orders, and DOC and HCO3 over one order of magnitude. According Nowadays advanced sensor technologies are widely applied in to the trophic scale, the lakes ranged from oligotrophic to hypertrophic aquatic studies. Diurnal variation in water temperature and DO pro- category. Also the indicator values characterising CO and DO distribu- fi 2 les in different lake types is already well-known (Smith and Bella, tion had similarly broad ranges. This extensive variability included in fl 1973; Melack, 1982; Gelda and Ef er, 2002; Sadro et al., 2011; all variables was certainly instrumental for the analysis enabling a Obrador et al., 2014). However, direct observations of daily dynamics strong manifestation of their impacts although determined a transfor- of dissolved CO2 in lakes are still rare, probably because of the rela- mation of most of the variables to reduce the skewness of their tively low reliability and accuracy and high cost of CO2 sensors. A ro- distribution. bust, accurate and responsive sensor-based method for direct and continuous measurement of dissolved CO2 has been elaborated and tested in tropical, temperate and boreal streams and ponds 2.1.1. Alkalitrophic lakes (Alk) (Johnson et al., 2010). Regarding lakes, sensor measurements of Äntu Sinijärv is a source lake fed by karstic ground waters and repre- – −1 sents highly alkalitrophic (N290 HCO3 mg L ) lakes in Estonia. Lakes of CO2 have been reported for the upper mixed layers of some lakes (Dinsmore et al., 2009; Johnson et al., 2007; Vachon and del this type are very local in the Pandivere Upland area (only a few else- Giorgio, 2014) or for the surface and bottom layers (Baehr and where) and there are only 60 lakes of that type in Estonia (Ott and −1 DeGrandpre, 2002, 2004),whilewehavenotfounddataoncontinu- Kõiv, 1999). With a mean light attenuation (Kd) of 0.16 ± 0.02 m for the photosynthetically active spectral region measured in 1995–96, ous profile measurements of CO2. In most of in situ studies (Frankignoulle et al., 2001; Jones and Äntu Sinijärv was the most transparent lake in Estonia (Nõges, 2000). Mulholland, 1998; Hope et al., 2001; Billett and Moore, 2008;

Sakagami et al., 2012) dissolved CO2 concentration is still determined 2.1.2. Non-stratiﬁed lakes with medium alkalinity (MedAlk) by the commonly used headspace method of Kling et al. (1991).As This is the most abundant lake type in Estonia, comprising approxi- noted by Johnson et al. (2010), attempts to use automated sampling mately 1/3 of our lakes (Ott and Kõiv, 1999). These lakes are relatively to increase sampling frequency for this method face the problem of shallow, with medium water retention times and may exhibit only epi- degassing of dissolved CO2 in the sample bottle. sodic thermal stratiﬁcation.

91 A. Laas et al. / Science of the Total Environment 566–567 (2016) 63–75 65

Table 1 Estonian lake types according to European Water Framework Directive.

– − Type name Abbreviation Area Alkalinity HCO3 mg Conductivity μS Cl mg Thermal Colour Pt\\Co Representative in current 2 km L−1 cm−1 L−1 stratifycation scale study

Alkalitrophic Alk b10 N240 N400 b25 Non-stratified N/A Äntu Sinijärv Non-stratified, medium MedAlk b10 80–240 165–400 b25 Non-stratified N/A Ülemiste alkalinity Stratified, medium StratMedAlk b10 80–240 165–400 b25 Stratified N/A Saadjärv alkalinity Dark-coloured soft-water DarkSoft b10 b80 b165 b25 Non-stratified ≥100° Valguta Mustjärv lakes Light-coloured soft-water LightSoft b10 b80 b165 b25 Non-stratified b100° Erastvere lakes Lake Võrtsjärv Large 100–300 80–240 165–400 b25 Non-stratified b100° Võrtsjärv Lake Peipsi V-Large N1000 80–240 165–400 b25 Non-stratified b100° Peipsi Coastal lakes Coastal N25 Non-stratified Mullutu Suurlaht

2.1.3. Stratiﬁed lakes with medium alkalinity (StratMedAlk) The only thermally stratiﬁed lake in our selection – Lake Saadjärv – rep-

This is the second most abundant lake type in Estonia, which in- resents deep clear water lakes in Estonia (Zmax =25m). cludes about 1/4 to 1/5 of the total number of lakes. Due to larger depth, these lakes provide more diverse habitats than the non-stratiﬁed 2.1.4. Dark-coloured soft-water lakes (DarkSoft) lakes but the lakes are also more sensitive to human impacts due to po- It has been estimated that about 8% of Estonian lakes are dark- tential build-up of anoxic conditions in the benthal leading to release of coloured soft-water lakes (Ott and Kõiv, 1999). Lakes under this type phosphorus from the sediment and thus a positive feedback to eutro- are shallow and acidic, contain large amounts of humic matter that con- phication. The character and functioning of these lakes are largely deter- tributes to radiative heating in summer (Ott, 2010). Strong light attenu- mined by the thermally stratifying larger volume of water (Ott, 2010). ation (up to 11 m−1) restricts the euphotic layer and limits

Fig. 1. Location of the studied lakes in Estonia.

92 66 A. Laas et al. / Science of the Total Environment 566–567 (2016) 63–75

Table 2 General description of the study lakes. Values of chemical and optical parameters are the averages of our measurements and earlier data included in the database of the Centre of Limnology.

Variable Alk (Äntu Medalk StratMedalk DarkSoft (Valguta LightSoft Large V-Large (Peipsi) Coastal (Mullutu Sinijärv) (Ülemiste) (Saadjärv) Mustjärv) (Erastvere) (Võrtsjärv) Suurlaht)

Trophic status Alkalitrophic Eutrophic Mesotrophic Hypertrophic Hypertrophic Eutrophic Eutrophic Eutrophic Mixing regime Polymictic Polymictic Dimictic Polymictic Dimictic Polymictic Polymictic Polymictic Area (ha) 2.1 944 724.5 20.4 16.3 27000 261100 412.7 Mean depth (m) 6 2.5 8 b1 3.5 2.8 8.3 b1 Max depth (m) 8 4.2 25 1 9.7 6 12.9 1.7 TP (μgL−1) 9 48 22 242 137 48 47 60 TN (μgL−1) 345 723 414 670 1126 910 375 1000 Chl a (μgL−1) 1 24.7 5.62 23.19 125.64 35.71 13.4 9.04 DOC (mg L−1) 4.72 13.7 9.2 35.2 12.3 11.8 12 18.1 – −1 HCO3 (mg L ) 292.8 201.3 150.47 30.5 99.6 211 170.8 109.8 −1 Kd (m ) 0.25 3.5 0.42 10.34 2.96 2.76 1.6 0.58 Secchi depth (m) bottom 1.4 4.9 0.15 1.25 0.55 2.9 bottom Water residence source lake 0.33 0.13 source lake 0.5 1 2 0.2 time (y) Watershed size 1,37 98.8 28.4 1.34 5.2 3116 47800 (10489 in 238 (km2) Estonia)

– TP — total phosphorous, TN — total nitrogen, Chl a — chlorophyll a,DOC— dissolved organic carbon, HCO3 — alkalinity, Kd — vertical light-attenuation coefficient. photosynthesis (Reinartetal.,2000). Among our study this lake type is 2.2. Manual sampling represented by Valguta Mustjärv, a very shallow 20-ha lake that has partly recovered from a heavy nutrient loading in the 1980s. Manual water samples for laboratory analyses of dissolved organic carbon (DOC), total phosphorus (TP) and nitrogen (TN), carbonate alka- – linity (HCO3), chlorophyll a (Chl a), and phytoplankton were taken from 2.1.5. Light-coloured soft-water lakes (LightSoft) all studied lakes once during the sensor deployment period. For deter- The study of Ott and Kõiv (1999) shows that light-coloured soft- mination of DOC concentrations in lakes the carbon content of the fil- water lakes comprise a little more that 5% of all Estonian lakes. Those trate was measured according to Toming et al. (2013).TPwas lakes are originally oligotrophic or semi-dystrophic, mainly determined with C. Zeiss spectrophotometer, according to Estonian na- characterised by low productivity, small catchment area implying tional standard EVS-EN 1189, TN with Bran + Luebbe autoanalyser, ac- – slow water exchange, low buffering capacity, and weak stress tolerance cording to EVS-EN ISO 13395. HCO3 was determined colorimetrically (Ott, 2010). Because of originally high water transparency, these lakes using 0.02% methyl orange test. For Chl a, 0.1–1 l of water was passed do not develop stable stratification. If impacted, e.g. by eutrophication, through Whatman GF/F glass microfiber filter and concentrations the ecosystems of light-coloured soft-water lakes become strongly were measured spectrophotometrically in 96% ethanol extracts at a destabilized characterised by frequent algal blooms and onset of strati- wavelength of 665 nm (Edler, 1979). fication. In the current study this lake type was represented by the impacted Lake Erastvere which by now has become dimictic and hypertrophic and, unfortunately, cannot characterize the reference 2.3. Sensor deployments and monitoring stations state of this lake type. All lakes were equipped with a high frequency monitoring platform or small lake buoy (OMC-7012 data-buoy) for a 6 to 12 full day period. 2.1.6. Coastal lakes (Coastal) Before sensor deployment (described in detail further), water tempera- Coastal lakes constitute approximately 10% of the total number of ture (T, °C), DO and electrical conductivity profiles were measured with Estonian lakes. Their functioning depends strongly on the very irregular handheld multiparametric sonde (YSI ProPlus). The location of the marine water inflow, which creates highly variable conditions and un- metalimnion was determined from water temperature profile accord- stable biota. Estonian coastal lakes are shallow, with high pH, water ing to Wetzel (2001). transparency and water temperature in summer. The main primary Continuous monitoring of DO concentration and water temperature producers are charophytes not phytoplankton. While those lakes are was performed by an automated station equipped with a shallow and clear, there is no temperature stratification. In our study multiparametric sonde (Yellow Springs Instruments (YSI) 66,002–4) this category is represented by Mullutu Suurlaht. at one-meter depth. Additional sensors for DO/temperature (Ponsel

Table 3 Data collection periods and weather conditions during measurements.

Lake Measurement period Full days of measurement Full days with parallel DO and CO2 data Range (and average) of daily mean meteorological variables during measurementsa

Wind speed, m s−1 Air temperature, o C PAR, μmol m−2 s−1

Alk 15.07–22.07.2014 7 6 1.2–3.4 (2.4) 17.4–19.8 (18.7) 324–619 (517) MedAlk 16.07–23.07.2014 6 1 1.6–2.1 (1.9) 20.1–22.2 (21.2) 578–608 (593) StratMedAlk 18.08–27.08.2014 9 5 1.7–4.1 (2.6) 13.0–14.3 (13.5) 55–397 (270) DarkSoft 23.07–05.08.2014 12 4 2.5–3.4 (2.9) 22.6–24.7 (23.7) 337–452 (419) LightSoft 19.08–29.08.2014 10 9 1.4–4.4 (2.8) 12.4–15.7 (13.9) 55–391 (263) Large 23.07–05.08.2014 12 4 2.5–3.4 (2.9) 22.6–24.7 (23.7) 337–452 (419) V-Large 09.09–16.09.2014 7 6 0.6–2.8 (1.3) 9.9–16.6 (13.8) 179–350 (251) Coastal 07.08–14.08.2014 7 6 1.9–7.0 (4.5) 17.1–21.9 (20.4) 152–525 (358)

a Data from closest meteorological station to the lake.

93 A. Laas et al. / Science of the Total Environment 566–567 (2016) 63–75 67

Table 4

Observed extreme values of gas regime indicators in lakes over the selected 24h period. Cases where CO2 and DO showed similar patterns are highlighted.

Characteristic Indicator Minimum Lake Maximum Lake

CO2% surface 1 Coastal 1835 Alk DO% surface 80 DarkSoft 144 Alk Saturation level CO2% bottom 114 V-Large 7047 LightSoft DO% bottom 0 LightSoft 114 Coastal

CO2% vertical difference 67 V-Large 6991 LightSoft DO% vertical difference 0 Coastal 102 LightSoft Spatial

variability CO2% vertical gradient 8 V-Large 4372 Coastal DO% vertical gradient 0.2 V-Large 33.5 DarkSoft

CO2% stdevsurface 4 Coastal 74 DarkSoft Absolute DO% stdev surface 2 StratMedAlk 9 Coastal temporal

variability CO2% stdevbottom 18 V-Large 768 MedAlk (St.dev) DO% stdevbottom 0 LightSoft 38 MedAlk

Relative CO2% cv surface 2 Alk 276 Coastal temporal DO% cv surface 2 StratMedAlk 8 Coastal variability

(CV%) CO2% cv bottom 3 StratMedAlk 67 Large DO% cv bottom 0 LightSoft 115 Medalk

Correlation DO%–CO2%correlation surface -0.1 Coastal -0.7 DarkSoft (Pearson r)

OPTOD) and dissolved CO2 concentration (AMT Analysenmesstechnik decided depending on stratiﬁcation to adequately characterize all dif- GmbH) were used at several depths. In all lakes the upper sensors ferent layers (see Figs. 2–4 for details). In the deepest lake were placed at 0.5 m depth and the position of other sensors was (StratMedalk) a chain of 12 HOBO Pendant temperature loggers was

Table 5 Multiple regression results over the selected 24h period. X — signiﬁcant exploratory variables (highlighting is used to better visualize the pattern). n.s. — non-signiﬁcant variables selected by forward stepwise procedure. Abbreviations: DO% — dissolved oxygen saturation, CO2% — dissolved carbon dioxide saturation, DOC — dissolved organic carbon, TP — total phosphorus,

Chl a — chlorophyll a,Kd — vertical light-attenuation coefficient, cv — coefficient of variation, corr — Pearson correlation coefficient, vert. diff. — vertical, diff. — difference, grad — gradient (change m−1).

Dependent variable Explanatory variables r2 p α – Characteristic Location Indicator LogArea LogMax d LogTP LogChl LogDOC HCO3 LogKd LogCO % 2 n.s. n.s. n.s. n.s. 0.626 0.145 Surface DO% n.s. X 0.752 0.132 Saturation

level LogCO2% X n.s. X X 0.915 0.017 Bottom DO% X X X 0.673 0.061

CO % 2 n.s. n.s. n.s. X 0.809 0.056 Surface DO % X 0.746 0.004 Spatial variability CO % 2 n.s. X n.s. 0.806 0.022 Bottom DO% n.s. n.s. n.s. 0.203 0.324

CO % vert. 2 X n.s. X X 0.908 0.019 Water diff. column DO% vert. Temporal X X X 0.681 0.058 diff. variability CO % vert. (CV%) 2 n.s. 0.199 0.149 Water grad. m–1 column DO% vert. X n.s. X 0.765 0.032 grad. m–1 Correlation Surface DO% & X X 0.894 0.002 (Pearson) CO2%

94 68 A. Laas et al. / Science of the Total Environment 566–567 (2016) 63–75 ter-column. Horizontal bold line above date represents the typical 24-h period that was represent sensor position in the wa gure legend, the reader is referred to the online version of this chapter.) our scales differ by lakes. Black dots fi Lake water temperature distribution over the study period. Note that the col Fig. 2. selected for statistical analysis. (For interpretation of the references to color in this 95 A. Laas et al. / Science of the Total Environment 566–567 (2016) 63–75 69 ntal bold line above date represents the typical 24-h period that was selected for statistical y lakes. Black dots represent sensor position in the water-column. Horizo gure legend, the reader is referred to the online version of this chapter.) fi Distribution of DO saturation in lakes. Note that the colour scales differ b Fig. 3. analysis. (For interpretation of the references to color in this 96 70 A. Laas et al. / Science of the Total Environment 566–567 (2016) 63–75 e date represents the typical 24-h period that was selected for differ by lakes. Black dots represent sensor position in the water-column. Horizontal bold line abov gure legend, the reader is referred to the online version of this chapter.) fi saturation in lakes. Note that the colour scales 2 or interpretation of the references to color in this Distribution of dissolved CO statistical analysis. (F Fig. 4.

97 A. Laas et al. / Science of the Total Environment 566–567 (2016) 63–75 71 used reaching from 0.5 to 18 m depth. The YSI multiparametric sondes candidate variables for the multiple regression analysis (see Supple- were equipped with a self-cleaning optical sensor for DO and turbidity, mentary materials, Table S1). We used the forward stepwise procedure all other sensors were cleaned manually after one-week period. All sen- for selecting the variables for the model. As several explanatory vari- sors were calibrated at the beginning and at the end of deployments. No ables such as TP, Chl a,Kd and the different expressions for depth drifts in the sensors were observed between calibrations. Sensors per- were strongly correlated, we carried out a Principal Component Analysis formed automatic measurements every 10 to 30 min (depending on (PCA). All statistical analyses were performed using the STATISTICA power availability at each lake) at up to four different depths during analysis software (Dell Inc., 2015). Contour graphs were produced the study period. Data collection on buoys and platforms was controlled using SigmaPlot 12.0 (Systat Software, Inc. GmbH). with OMC-045-II GPRS data loggers. Automated stations were mostly placed near the deepest area in each lake. In Võrtsjärv (Large) our 3. Results buoy was located close to our long-term monitoring point which, according to Nõges and Tuvikene (2012), is representative for N90% of 3.1. Stratiﬁcation in studied lakes the lake area. In Peipsi (V-Large), because of security reasons, measurements were made at a station in the Mustvee bay approximately 1 km No stable thermal stratiﬁcation was observed in most lakes in sum- from the western shore. mer because of their shallowness. Although 7 out of 8 Estonian lakes types according to the WFD typology should not stratify (Table 1), 3 of

2.4. Measurement and calculation of dissolved CO2 values the type representative lakes were strongly stratiﬁed and 3 others strat- iﬁed occasionally and weakly during our study period (Fig. 2 and Fig. S1

Membrane covered optical CO2 sensors (AMT Analysenmesstechnik in Supplementary materials). Only the V-Large and the Coastal were al- GmbH) with measuring ranges of 30 mg L−1 and 80 mg L−1 were used ways mixed. Due to exceptionally calm weather during measurements to record dissolved CO2 partial pressure (pCO2) values in lakes. Accord- in Large (Table 2), considerable vertical temperature differences oc- ing to the sensors' manual (http://www.amt-gmbh.com/), the inner curred in this lake ranging from 1.3 to 5.75 °C. sensor volume is separated from the sample by means of a gas perme- Although some lakes did not exhibit a stable thermal stratification or able silicone membrane, non-passable for liquids and solids. If the sen- were fully mixed during our study, they were all stratified for dissolved sor is immersed into a sample, a pCO2 equilibration is achieved gases (Fig. 3; 4; Figs. S2; S3) with the strongest stratification occurring between the inner sensor volume and the sample. A Single-Beam Dual in two deeper lakes (StratMedAlk and LightSoft), but also in the shallow Wavelength nondispersive infrared (NDIR) optical sensor mounted in- Alk. In the stratified conditions in LightSoft there was a strong bloom side the sensor detects the dissolved CO2 gas, but is insensitive to car- caused by the cyanobacterium Aphanizomenon gracile Lemm at the be- bonate and bicarbonate. Measurement of pCO2 is accompanied by ginning of our study. water temperature and air pressure measurements to calculate CO2 concentrations in the lakes. Increases in water temperature cause a de- 3.2. Gas distribution in studied lakes crease in sensor output, while increases in atmospheric pressure cause an increase in sensor output. All sensors had a fixed measuring depth, Comparing the gas distribution indicators for the selected most typ- therefore we could do the depth correction for each measurement ical 24h period (Table 4; Fig. S4), the highest concentration of dissolved time interval once for all. We assumed a constant atmospheric pCO2 of CO2 was measured at the bottom of LightSoft lake (7047% relative to 400 μatm (http://co2now.org/) which was taken as the equilibrium equilibrium concentration with atmosphere), followed by Coastal lake value for the air-water interface. (4373%, not shown). The lowest bottom concentrations of dissolved

Although DO and CO2 measurements in all lakes lasted between 6 CO2 (114%) but also the smallest vertical difference and gradient were and 12 days, in some lakes (DarkSoft and Large) we could not capture measured in V-Large. We observed a continuous CO2 supersaturation parallel data on both gases for N4 days because of malfunctioning of de- in the upper mixed layer of Alk reaching 1835% and DarkSoft (607%); vices. In MedAlk parallel measurements of CO2 and DO at all depths all other lakes had at least short periods when the surface of the lake succeeded for one full day only (Table 3). In StratMedAlk with a maxi- was undersaturated. All eight lakes had at least a short-term stratiﬁca- mum depth of 25 m we could measure temperature down to 18 m tion for dissolved CO2 (Fig. 3) while in the 4 shallower lakes the concen- and DO and CO2 only down to 10 m depth due to limited length of the trations were temporarily fully uniform. The largest difference between cables. CO2 saturation levels at the surface and near bottom occurred in the LightSoft but the gradient was larger in the shallow Coastal lake 2.5. Statistical analysis (Table 4; Fig. S4).

The largest diurnal variability in CO2 saturation (%) in the surface In order to analyse relationships between lake type/trophic state layer (standard deviation, SD) occurred in DarkSoft (74%) and Alk characteristics and the measured gas distribution patterns, we first se- (46%) where the surface waters were constantly supersaturated with lected visually the most typical 24-h period from each lake (marked in CO2. The relatively strongest diurnal variability was measured in the Figs. 2–4 with a bold horizontal line above date). This was done to re- Coastal lake (CV = 276%) followed by that in LightSoft (45%) and V- duce noise caused by meteorological disturbances and possible techni- Large (30%). cal issues concerning single lakes and single days. For this 24h period We found several similarities between the occurrences of extremes we calculated the following indicators for both DO and CO2 to character- of CO2 and DO in lakes. Both gases had their highest surface layer satu- ize the gas regime of lakes: average saturation level (DO%, CO2%) and co- ration levels in Alk, the largest vertical differences occurred in the strat- efficient of variation (DO% cv, CO2% cv) at the surface (0.5 m), and at the ified LightSoft and the smallest vertical gradients in V-Large. Both gases bottom (data from the deepest sensor), vertical difference (DO% vert showed the largest bottom layer SD in MedAlk and the largest surface diff, CO2% vert diff) and vertical gradient (DO% vert grad, CO2%vert layer variation coefficients in Coastal. Highest DO saturation levels grad, m−1) of the average saturation level, and the Pearson correlation were measured in the surface layer of the Alk (144%) but reached nearly coefficient between DO% and CO2% at the surface. Among lake type spe- 120% also in the eutrophic Coastal, Large, and MedAlk. Unlike other cific characteristics, we selected lake area, mean, maximum and relative lakes where the highest values of DO were measured near the surface, – depth (calculated according to Wetzel (2001)), HCO3 and DOC. The tro- in V-Large the maximum occurred at 1.5 m depth. phic state of lakes was characterised by TP, TN, Chl a and Kd. Variables The highest vertical difference of DO was measured in LightSoft, with a log-normal distribution were included in the analysis as logarith- which was strongly stratified for both gases. DO saturation in this lake mic values. A correlation analysis was carried out to select the best ranged from 102% at the surface to anoxia at 2.8 m depth. The lowest

98 72 A. Laas et al. / Science of the Total Environment 566–567 (2016) 63–75

signiﬁcant variable for any of the gas distribution indicators. The best

models, including the ones for CO2 saturation in the bottom layer, CO2 vertical difference and the strength of the DO–CO2 relationship, explained approximately 90% of the variance in these dependent variables using morphometric and trophic state variables. The two ﬁrst factors of the PCA (Fig. 6) explained, respectively, 42%

and 21% of the total variance included in the data. Kd, TN, TP, Chl a and DOC all had highest positive loadings to F1 that justiﬁed calling it a “trophic state factor”, whereas alkalinity had the highest negative loading to F1. Among gas distribution indicators, vertical gradients of

both CO2 and DO were positively related to the trophic state factor. Among lakes, the two hypertrophic softwater lakes (DarkSoft and LightSoft) were positively related to the trophic state factor, while the Alk, which can be considered oligotrophic by most trophic state indicators, remained on the negative side of F1. Total nitrogen and lake area had the highest (negative) loadings to the second PCA factor F2. The “ ” 2 size factor was positively associated with DO saturation both at the Fig. 5. Strength of the relationship (R ) between DO and CO2 concentrations in the surface layer (0.5 m) and the trophic state parameters total phosphorus (TP), chlorophyll a (Chl) surface and bottom and negatively with CO2 saturation in the surface and light attenuation coefficient (Kd) in the studied lakes. layer. Among lakes, Large, V-Large and Coastal were positively associated with F2 and all small lakes remained on its negative side. The only exception was StratMedAlk, which grouped together with smaller lakes. vertical DO gradient was captured in V-Large where the saturation level fluctuated slightly around 100%. The highest DO temporal variation 4. Discussion both in absolute and relative terms occurred in Coastal for surface layers and MedAlk for bottom layers. 4.1. Type and state characteristics The strength of the negative correlation between DO and CO2 in the surface layer scaled well with the trophic state of lakes being the stron- The European Water Framework Directive (EU, 2000) clearly distin- gest in hypertrophic DarkSoft (r = −0.73) and decreasing towards the guishes between natural lake type descriptors that characterize type oligotrophic Alk (r = −0.22). The only exception was the eutrophic specific biological reference conditions, and status descriptors describ- Coastal lake where the correlation was weak too (Fig. 5). ing the anthropogenic impact which tends to deviate the ecosystem According to the multiple regression analysis (Table 5), K and lake d from its pristine state. Lake type descriptors are divided into obligatory area appeared among significant determinants for the largest number descriptors, such as geographic coordinates, depth, size and basin geol- of gas distribution indicators, and were included in the models of, corre- ogy (calcareous, siliceous, organic) and optional descriptors which may spondingly, 6 and 5 out of 13 indicators tested. Also the maximum include for instance lake shape, residence time, mixing characteristics, depth of lakes explained a large proportion of variability included in – alkalinity etc. As the mixing type is defined for lakes in reference condi- the gas distribution data. Both HCO and Chl a were significant determi- 3 tions, it may happen that with increasing K due to eutrophication, sil- nants in two models, DOC in one, while TP was not selected as a d tation or brownification, the mixing type changes (Heiskanen et al., 2015). A predominant change would be that polymictic lakes start developing a more stable thermal stratification. Among our study lakes, this was certainly the case for LightSoft which being oligotrophic and polymictic by origin (reference conditions) has turned into a stratified hypertrophic lake. We observed stronger stratification than expected by type description also in Alk and MedAlk. In order to avoid mixing up lake type specific effects with trophic state effects and concentrate only on the former, an alternative would have been to study only reference lakes, i.e. the pristine lakes without human impacts. That, however, would have deviated us from the real life where most of the water bodies are more or less impacted by cultural eutrophication (Smith and Schindler, 2009).

4.2. General patterns

The “trophic state” factor remained clearly the strongest determi- nant for the gas distribution in our set of lakes. However, certain type- speciﬁc peculiarities can be drawn out for individual lakes. The multiple regression analysis revealed lake area as one of the key variables determining the dynamics of several gas distribution indicators. Thus its only moderate loading to F2 in the PCA was rather unexpected. In summer, the surface layers of most Estonian lakes were supersat-

urated for both DO and CO2 similarly to several lakes elsewhere Fig. 6. Results of the Principal Component Analysis relating indicators of DO and CO2 (Dinsmore et al., 2009; Vachon and del Giorgio, 2014) and therefore distribution with lake type and trophic state parameters. Both groups of variables are contributed to CO2 emissions to the atmosphere. In line with patterns projected on the 1 × 2 factor plane as empty dots, lakes as ﬁlled dots. Abbreviations: DO described by Dinsmore et al. (2009) and Ducharme-Riel et al. (2015), — dissolved oxygen, DOC — dissolved organic carbon, TN — total nitrogen, TP — total the supersaturation with CO in Estonian lakes typically increased grad- phosphorus, Chl a — chlorophyll a,Kd — diffuse light attenuation coefﬁcient, grad — 2 −1 gradient (change m ), surf — surface. ually with depth, reaching its maximum near the lake bottom. High CO2

99 A. Laas et al. / Science of the Total Environment 566–567 (2016) 63–75 73

concentration near the bottom suggests CO2 diffusion from the sedi- responsible for the highest and most stable CO2 supersaturation levels ment (Rantakari and Kortelainen, 2005). (up to 1800%) in the surface layer of Alk compared to our other studied

The surface layer was undersaturated with CO2 in three of our stud- lakes. Reverse weathering of calcium carbonate going on in the lake is ied lakes (V-Large, LightSoft and Coastal) during most of the study peri- evidenced by strong calcite precipitation (a visual observation of the od likely caused by photosynthetic CO2 uptake that enables a lake bottom revealed that submerged plants were covered by a thick downward flux of CO2 from the atmosphere to the lake. As the Estonian calcite crust). Calcite precipitation releases CO2 in amounts exceeding part of V-Large with an area of 1442 km2 makes up about 2/3 of the total by large the uptake abilities of the vegetation, turning the lake into an- 2 area of Estonian lakes (2130 km ), we can claim that during our study, other CO2 emitting “chimney” types (Marcé et al., 2015; Weyhenmeyer in a major part of Estonian lake surface area CO2 uptake was et al., 2015). predominating over release. Our earlier studies (Rõõm et al., 2014) Gas distribution in MedAlk was likely mostly determined by its eu- have shown a mid- and late summer CO2 uptake also in the pelagic trophic state. Our study showed that even in virtually homothermal part of Large in case of similar weather conditions as in 2014. conditions, both DO and CO2 concentration gradients could build up near the bottom of the lakes overnight, probably because of high respi- 4.3. Individual features of lakes ration rates near the sediment. Another reason for DO gradients in these lakes is the uneven vertical distribution of photosynthesis: most of the A clear drawback of this selection of lakes was that the small number DO that is produced near the surface may not reach the bottom layers of lakes predetermined a unique combination of type and trophic state during windless days. characteristics. So for example, the clustering of DOC (type characteris- A deeper and stratified water column determined the main charac- tic) together with TP and Chl a (trophic state characteristics) in the fac- teristics of StratMedAlk. Despite the size of the lake (7 km2), the PCA tor analysis was clearly caused by the specific features of the DarkSoft showed that its gas regime was more similar to smaller lakes that lake type. Although a large number of dystrophic (humic) lakes have could be attributed to its higher water column stability. During most been affected by eutrophication over the last half century (Ott and of the time, DO showed a clinograde profile being uniformly distributed Kõiv, 1999), the combination of dystrophic and eutrophic features can within the epilimnion and declining rapidly to zero within the upper 1 still be due to the specific characteristics of the lakes in this study. The m of the metalimnion. We could capture most of the changes, except highest trophic state DarkSoft combined with the highest DOC content perhaps the possible increase in CO2 concentration in the unmeasured among the study lakes determinedmainly its gas regime and the strong hypolimnion. We observed also a short deep mixing episode in this synchrony of the opposite changes in DO and CO2 dynamics. Fast radia- lake (on 25th of august) after which the stratification was fully restored tive heating and cooling of the dark water caused the surface tempera- (Figs. 3 & 4). ture to fluctuate by 4 °C daily that likely contributed to the large The cyanobacteria bloom in LightSoft was the likely reason that in amplitude of DO and CO2 saturation through changes in gas solubility. good light conditions caused the DO saturation in the epilimnion to in- Decomposition of allochthonous organic matter was the likely reason crease up to 120% and fully depleted this layer of CO2. Because of strong causing low DO and continuous high supersaturation with CO2 even in stratification, the hypolimnetic CO2 was unavailable for photosynthesis the surface layer. This kind of lakes form a “chimney” type in which which had to rely mostly on CO2 from respiratory release and diffusion −1 CO2 supersaturation is based on high organic C respiration rates (Cole from air. Stronger winds on 25–26 August 2014 (up to 10 m s )deep- et al., 1994; Jonsson et al., 2003). ened the thermocline bringing up less oxygenized water but not consid-

A phenomenon that also can be attributed to Alk is the positioning of erably alleviating the lack of CO2 in the epilimnion. – HCO3 to the negative end of the “trophic state” factor in the PCA. High Given the large wind exposed surface area of the Large and V-Large alkalinity itself is not contradicting with high trophic state as two eutro- lakes, we expected vertically isothermal conditions in these lakes and – phic lakes included in our selection, MedAlk and Large, had HCO3 values fast equilibration of gas partial pressures with the atmosphere. Howev- exceeding 200 mg L−1. However, the main feature that makes the Alk er, the considerable vertical temperature differences occurring in Large distinct is the predominant ground water feeding of this lake type. Dur- due to calm weather created atypically large inhomogeneity also in DO ing percolation of water through the karstic area, virtually all phospho- and especially CO2 saturation levels. In V-Large where the measure- rus is bound to calcium maintaining the oligotrophic character of this ments were carried out in September, the temperature and gas distribu- lake type (Ott and Kõiv, 1999) despite the vicinity of areas of intense ag- tions were more homogeneous, typically to large lakes. The contrasting riculture. Even high TN values sometimes measured in this lake (Ott and CO2 ranges between the lakes could partially be explained by different Kõiv, 1999) have not changed its nature. sediment composition at measuring stations: sand in V-Large and or- Although the Alk lake type is deﬁned as non-stratiﬁed, also earlier ganic-rich sediments in Large. studies have revealed considerable vertical temperature differences Both lakes showed marked diurnal patterns in temperature and gas ranging from 2.7 to 10.5 °C with a maximum gradient of 7.3 °C m−1 lo- regimes more manifested, however, in the V-large than in Large. Both cated mostly within the upper 1–2 m layer (Nõges and Nõges, 1998). large shallow lakes consumed most of the dissolved CO2 available for High surface DO% values in this lake were unexpected as the spring photosynthesis, creating an equilibrium state (100% saturation) at the water feeding the karstic lake is not saturated with oxygen and lake air-water interface in Large but undersaturated conditions (10–70% sat- has very low phytoplankton biomass, Chl a and planktonic primary pro- uration) in V-Large. The difference in CO2 saturation levels showed that duction. Previous measurements (Nõges and Nõges, 1998) showed that during the measurement period, V-Large was likely taking up CO2 from −3 −1 the extremely low primary production (b2mgCm h ) was usually the atmosphere, but in Large the respiratory CO2 release provided restricted to 1–1.5 m layer, although the irradiance level in the transpar- enough carbon to support photosynthesis, i.e. the two processes were ent water could enable photosynthesis in the whole water column. Be- more or less balanced. Our earlier studies (Rõõm et al., 2014)have cause the bottom of the representative of the Alk type is covered with shown a mid- and late summer CO2 uptake also in the pelagic part of Chara, a large part of primary production and water oxygenation can Large, like we found in this study for lake V-Large. Laas et al. (2012) be attributed to benthic communities (Cremona et al., in press). Besides and Cremona et al. (2014) showed that Large turns from autotrophic photosynthesis, the DO supersaturation in the surface layer can be par- to heterotrophic type of metabolism around mid-summer, but autotro- tially caused by the warming of the surface layer in which the equilibra- phic periods in early August are still commonplace in this lake. tion with the atmospheric partial pressure is not immediate in sheltered The coupling of the “shallowness factor” with the large vertical gra- conditions created by the forest surrounding the small lake. dient of CO2 can obviously be attributed to the Coastal where, despite of It is probable that dissolved inorganic carbon which enters the lake small depth, a strong vertical gradient in CO2 was likely caused by respi- through a number of carbonate-rich groundwater springs was ration of the rich Chara mat in this eutrophic lake. Weyhenmeyer et al.

100 74 A. Laas et al. / Science of the Total Environment 566–567 (2016) 63–75

Billett, M.F., Moore, T.R., 2008. Supersaturation and evasion of CO and CH in surface wa- (2015) conclude that in boreal lakes CO2 is concentrated in bottom wa- 2 4 ters at Mer Bleue peatland, Canada. Hydrol. Process. 22, 2044–2054. ters throughout the year, although these lakes are typically shallow with Bryant, D.A., Frigaard, N.-U., 2006. Prokaryotic photosynthesis and phototrophy illuminat- short water retention times. ed. Trends Microbiol. 14, 488–496. http://dx.doi.org/10.1016/j.tim.2006.09.001. Our measurements showed neither temperature nor DO stratifica- Casper, P., Maberly, S.C., Hall, G.H., Finlay, B.J., 2000. Fluxes of methane and carbon dioxide from a small productive lake to the atmosphere. Biogeochemistry 49, 1–19. tion in the Coastal lake, although the diurnal variation of both gases Cole, J.J., Caraco, N.F., Kling, G.W., Kratz, T.K., 1994. Carbon dioxide supersaturation in the was among the highest. The strong vertical gradient in CO2 was likely surface waters of lakes. Science 265, 1568–1570. caused by the dense Chara mat on the bottom of this lake (Cremona et Cole, J.J., Pace, M.L., Carpenter, S.R., Kitchell, J.F., 2000. Persistence of net heterotrophy in al., in press). The strong CO accumulation in the bottom layer caused lakes during nutrient addition and food web manipulations. Limnol. Oceanogr. 45, 2 1718–1730. a decoupling of the DO CO2 relationship which was strong in other eu- Cremona, F., Laas, A., Nõges, P., Nõges, T., 2014. High-frequency data within a modelling trophic lakes. framework: on the benefit of assessing uncertainties of lake metabolism. Ecol. Model. 294, 27–35. Cremona, F., Laas, A., Arvola, L., Pierson, D., Nõges, P., Nõges, T., 2016. Numerical explora- 5. Conclusions tion of the planktonic to benthic primary production ratios in lakes of the Baltic Sea catchment. Ecosystems (in press). The eight lakes representing different lake types in Estonia, most of Dell Inc., 2015. Dell Statistica (Data Analysis Software System), Version 12. Software.Dell.Com. them widespread in the whole northern temperate region, showed con- Dinsmore, K.J., Billett, M.F., Moore, T.R., 2009. Transfer of carbon dioxide and methane siderable differences in thermal stratification and vertical distribution of through the soil–water–atmosphere system at Mer Bleue peatland, Canada. Hydrol. – dissolved oxygen and CO as well as different diurnal dynamics over the Process. 23, 330 341. 2 Ducharme-Riel, V., Vachon, D., del Giorgio, P.A., Prairie, Y.T., 2015. The relative contribu- 1–2 weeks of high-frequency measurements. These differences could tion of winter under-ice and summer hypolimnetic CO2 accumulation to the annual mostly be attributed to different trophic state of the lakes and to a lesser CO2 emissions from northern lakes. Ecosystems 18, 547–559. extent to lake type specific characteristics such as morphometry and Edler, L., 1979. Recommendations for Marine Biological Studies in the Baltic Sea. Phyto- plankton and Chlorophyll: The Baltic Marine Biologists 5, pp. 1–38. water chemistry. Frankignoulle, M., Borges, A., Biondo, R., 2001. A new design of equilibrator to monitor With increasing trophic state, the negative coupling between DO carbon dioxide in highly dynamic and turbid environments. Water Res. 35, 1344–1347. and CO2 grew stronger suggesting trophic state was a good proxy of Gelda, R.K., Effler, S.W., 2002. Metabolic rate estimates for a eutrophic lake from diel dis- gas uptake and release processes. Strong dependence of several gas dissolved oxygen signals. Hydrobiologia 485 (1–3), 51–66. tribution indices on lake area and depth refers to morphometry of the Heiskanen, J.J., Mammarella, I., Ojala, A., Stepanenko, V., Erkkilä, K.-M., Miettinen, H., lakes as a complex of factors affecting the redistribution of gases within Sandström, H., Eugster, W., Leppäranta, M., Järvinen, H., Vesala, T., Nordbo, A., 2015. fi the water column and the exchange rate with the atmosphere. Effects of water clarity on lake strati cation and lake-atmosphere heat exchange. J. Geophys. Res. Atmos. 120, 7412–7428. http://dx.doi.org/10.1002/2014JD022938. Stronger stratification of impacted lakes compared to those in refer- Hope, D., Palmer, S.M., Billett, M.F., Dawson, J.J.C., 2001. Carbon dioxide and methane eva- ence conditions has strong implications on the gas regimes exacerbating sion from a temperate peatland stream. Limnol. Oceanogr. 46, 847–857. anoxia, internal phosphorus release and accumulating large amounts of http://co2now.org/ (06.04.2016) http://www.amt-gmbh.com/ (06.04.2016) CO2 in the bottom layers of lakes, especially in the light of an ongoing Johnson, M.S., Weiler, M., Couto, E.G., Riha, S.J., Lehmann, J., 2007. Storm pulses of dis- global climate warming. solved CO2 in forested headwater Amazonian stream explored using hydrograph sep- Among other type specific factors alkalitrophic groundwater feeding aration. Water Resour. Res. 43, W11201. http://dx.doi.org/10.1029/2007WR006359. Johnson, M.S., Billett, M.F., Dinsmore, K.J., Wallin, M., Dyson, K.E., Jassal, R.S., 2010. Direct and high DOC loads in the dystrophic lake were the likely reasons caus- and continuous measurement of dissolved carbon dioxide in freshwater aquatic sys- – – ing continuous CO2 supersaturation in the upper mixed layer of these tems method and applications. Ecohydrology 3, 68 78. Jones, N., 2010. Battle to degas deadly lakes continues. Nature 466 (7310), 1033. http:// lake types showing them as CO2 emitting “chimneys” but with totally dx.doi.org/10.1038/4661033a. different underlying mechanisms. Jones, J.B., Mulholland, P.J., 1998. Methane input and evasion in a hardwood forest stream: Supplementary data to this article can be found online at http://dx. effects of subsurface flow from shallow and deep pathways. Limnol. Oceanogr. 43, doi.org/10.1016/j.scitotenv.2016.05.038. 1243–1250. Jonsson, A., Karlsson, J., Jansson, M., 2003. Sources of carbon dioxide supersaturation in clearwater and humic lakes in northern Sweden. Ecosystems 6, 224–235. Acknowledgements Kling, G.W., Kipphut, G.W., Miller, M.C., 1991. Arctic lakes and streams as gas conduits to the atmosphere—implications for tundra carbon budgets. Science 251, 298–301. Kortelainen, P., Rantakari, M., Huttunen, J.T., Mattsson, T., Alm, J., Juutinen, S., Larmola, T., This research was inspired by GLEON (Global Lake Ecological Obser- Silvola, J., Martikainen, P.J., 2006. Sediment respiration and lake trophic state are im- vatory Network) and NETLAKE (Networking Lake Observatories in Eu- portant predictors of large CO2 evasion from small boreal lakes. Glob. Chang. Biol. 12, rope, COST Action) and was funded by the Estonian Ministry of 1554–1567. Laas, A., Nõges, P., Kõiv, T., Nõges, T., 2012. High frequency metabolism study in a large Education and Research (IUT 21-02; PUT 777), Estonian Science Foun- and shallow temperate lake revealed seasonal switching between net autotrophy dation (grant 9102, ETF8486), MARS project (Managing Aquatic ecosys- and net heterotrophy. Hydrobiologia 694, 57–74. tems and water Resources under multiple Stress) funded under the 7th Liikanen, A., Flöjt, L., Martikainen, P., 2002. Gas dynamics in eutrophic lake sediments af- – EU Framework Program, Theme 6 (Environment including Climate fected by oxygen, nitrate, and sulfate. J. Environ. Qual. 31, 338 349. Marcé, R., Obrador, B., Morguí, J.-A., Riera, J.L., López, P., Armengol, J., 2015. Carbonate

Change), Contract No.: 603378 (http://www.mars-project.eu)and weathering as a driver of CO2 supersaturation in lakes. Nat. Geosci. 8, 107–111. Swiss Grant for Program “Enhancing public environmental monitoring Melack, J.M., 1982. Photosynthetic activity and respiration in an equatorial African soda – capacities”. lake. Freshw. Biol. 12, 381 399. Nõges, P., 2000. Euphotic holding capacity for optically active substances. Geophysica 36 (1–2), 169–176. References Nõges, P., Nõges, T., 1998. Stratification of Estonian lakes studied during hydro optical ex- peditions in 1995–97. Proc. Estonian Acad. Sci. Biol. Ecol. 47, 268–281. [EU] European Union, 2000. Directive 2000/60/EC of the European Parliament and of the Nõges, P., Tuvikene, L., 2012. Spatial and annual variability of environmental and phyto- council of 23 October 2000 establishing a framework for community action in the plankton indicators in Võrtsjärv: implications for water quality monitoring. Est. field of water policy. Off. J. L 327, 1.71. J. Ecol. 61, 227–246. [ME] Ministry of the Environment, 2009. Procedure for the establishment of bodies of sur- Obrador, B., Staehr, P.A., Christensen, J.P.C., 2014. Vertical patterns of metabolism in three face water and a list of the bodies of surface water the state of which is to be contrasting stratified lakes. Limnol. Oceanogr. 59 (4), 1228–1240. established, classes of the states and the values of quality indicators corresponding Ott, I., 2006. Some principles of ecological quality classification in Estonian lakes. In: de to these state classes, and the procedure for the establishment of the classes of Wit, H., Skjelkvale, B.L. (Eds.), Proceedings of the 21th Meeting of the ICP Waters Pro- state (RTL 2009, 64, 941). Order from 28.07.2009 no. 44. Ministry of the Environment gramme Task Force in Tallinn, Estonia, October 17.19, 2005. Norwegian University of (www.riigiteataja.ee/akt/13210253) [in Estonian], 11 pp. Science and Technology (84/2006, 8.14).

Baehr, M.M., DeGrandpre, M.D., 2002. Under-ice CO2 and O2 variability in a freshwater Ott, I., 2010. Pinnavee seisundi hindamine, võrdlus veekogumid ja pinnavee seisundi lake. Biogeochemistry 61, 95–113. klassipiirid bioloogiliste kvaliteedielementide järgi. Evaluation Of Surface Waters,

Baehr, M.M., DeGrandpre, M.D., 2004. In situ pCO2 and O2 measurements in a lake during Reference Waterbodies And The Status Of Class Boundaries By The Biological Quality turnover and stratiﬁcation: observations and modeling. Limnol. Oceanogr. 49 (2), Elements. Estonian University of Life Sciences, Centre for Limnology report to Esto- 330–340. nian Ministry of the Environment (222 pp.).

101 A. Laas et al. / Science of the Total Environment 566–567 (2016) 63–75 75

Ott, I., Kõiv, T., 1999. Eesti väikejärvede eripära ja muutused. Estonian Small Lakes: Special Smith, V.H., Schindler, D.W., 2009. Eutrophication sciences: where do we go from here? Features And Changes. Estonian Ministry of the Environment Information Centre, Tal- Trends Ecol. Evol. 24 (4), 201–207. linn (127 pp. in Estonian). Toming, K., Tuvikene, L., Vilbaste, S., Agasild, H., Kisand, A., Viik, M., Martma, T., Jones, R., Prairie, Y.T., Bird, D.F., Cole, J.J., 2002. The summer metabolic balance in the epilimnion of Nõges, T., 2013. Contributions of autochthonous and allochthonous sources to dis- southeastern Quebec lakes. Limnol. Oceanogr. 47, 316–321. solved organic matter in a large, shallow, eutrophic lake with a highly calcareous Rantakari, M., Kortelainen, P., 2005. Interannual variation and climatic regulation of the catchment. Limnol. Oceanogr. 58 (4), 1259–1270.

CO2 emission from large boreal lakes. Glob. Chang. Biol. 11, 1368–1380. Vachon, D., del Giorgio, P.A., 2014. Whole-lake CO2 dynamics in response to storm events Reinart, A., Arst, H., Nõges, P., Nõges, T., 2000. Comparison of euphotic layer criteria in in two morphologically different lakes. Ecosystems 17, 1338–1353. lakes. Geophysica 36 (1–2), 141–159. Vachon, D., Lapierre, J.-F., del Giorgio, P.A., 2016. Seasonality of photochemical dissolved

Rõõm, E.-I., Nõges, P., Feldmann, T., Tuvikene, L., Kisand, A., Teearu, H., Nõges, T., 2014. organic carbon mineralization and its relative contribution to pelagic C02 production Years are not brothers: two-year comparison of greenhouse gas ﬂuxes in large shal- in northern lakes. J. Geophys. Res. Biogeosci. 121, 1–11. http://dx.doi.org/10.1002/201 low Lake Võrtsjärv, Estonia. J. Hydrol. 519, 1594–1606. 5JG003244. Sadro, S., Melack, J.M., Macintyre, S., 2011. Depth integrated estimates of ecosystem me- Wetzel, R.G., 2001. Limnology. Lake and River Ecosystems, third ed. Academic Press, Lon- tabolism in a high elevation lake (Emerald Lake, Sierra Nevada, California). Limnol. don (1006 pp.). Oceanogr. 56, 1764–1780. Weyhenmeyer, G.A., Kosten, S., Wallin, M.B., Tranvik, L.J., Jeppesen, E., Roland, F., 2015.

Sakagami, H., Takahashi, N., Hachikubo, A., Minami, H., Yamashita, S., Shoji, H., Khlystov, Signiﬁcant fraction of CO2 emissions from boreal lakes derived from hydrologic inor- O., Kalmychkov, G., Grachev, M., De Batist, M., 2012. Molecular and isotopic composi- ganic carbon inputs. Nat. Geosci. 8, 933–936. tion of hydrate-bound and dissolved gases in the southern basin of Lake Baikal, based on an improved headspace gas method. Geo-Mar. Lett. 32, 465–472. Smith, S.A., Bella, D.A., 1973. Dissolved oxygen and temperature in a stratiﬁed Lake. J. Water Pollut. Control Fed. 45 (1), 119–133.

102 CURRICULUM VITAE

Name: Pille Meinson Date of birth: 21.09.1989 E-mail: [email protected]

Education: 2013 – 2017 PhD studies in Estonian University of Life Sciences 2011 – 2013 MSc in applied hydrobiology, Estonian University of Life Sciences 2008 – 2011 BSc in applied hydrobiology, Estonian University of Life Sciences 1996 – 2008 Tartu Forseliuse Gymnasium

Research interests: High frequency changes of biological and physical parameters in aquatic ecosystems

Membership: 2016 - Member of Association for the Science of Limnology and Oceanography (ASLO) 2013 - Member of Global Lake Ecological Observatory Network (GLEON) 2013 - Member of Networking Lake observatories in Europe (NETLAKE)

Professional training: 2015 NETLAKE Short Time Scientific Meeting in Edmund Mach Foundation of San Michele all'Adige, Italy. 2015 NETLAKE Training School in Analysis of High- Frequency Data from Lake Monitoring Systems, Centre for Limnology, Estonian University of Life Sciences, Estonia. 2014 NETLAKE Training School in Automated Monitoring and High Frequency Data Analysis, Erken Laboratory, Uppsala University, Sweden.

103 2014 International Summer School of Limnology - Lake Ecology and temporal Dynamics, Evian- les-Bains, France. 2014 GLEON 16 workshop „Open source science tools in GLEON“, Jouvence, Orford, Quebec, Canada.

104 ELULOOKIRJELDUS

Nimi: Pille Meinson Sünniaeg: 21.09.1989 E-post: [email protected]

Haridus: 2013 – 2017 PhD õpe, Eesti Maaülikool 2011 – 2013 MSc rakendushüdrobioloogia erialal, Eesti Maaülikool 2008 – 2011 BSc rakendushüdrobioloogia erialal, Eesti Maaülikool 1996 – 2008 Tartu Forseliuse Gümnaasium

Teadustöö põhisuunad: Bioloogiliste ja füüsikaliste näitajate pidevmõõtmised, erinevas ajaskaalas toimuvate muutuste ja nende põhjuste selgitamine veeökosüsteemides

Teadusorganisatsiooniline tegevus: 2016 - Rahvusvahelise ühenduse Association for the Science of Limnology and Oceanography (ASLO) liige 2013 - Rahvusvahelise ühenduse Global Lake Ecological Observatory Network (GLEON) liige 2013 - Rahvusvahelise ühenduse Networking Lake observatories in Europe (NETLAKE) liige

Erialane enesetäiendamine: 2015 täiendõpe “NETLAKE Short Time Scientific Meeting”, Edmund Mach Foundation, San Michele all'Adige, Itaalia. 2015 koolitus “NETLAKE Training School in Analysis of High-Frequency Data from Lake Monitoring Systems”, Limnoloogiakeskus, Eesti Maaülikool, Eesti.

105 2014 koolitus “NETLAKE Training School in Automated Monitoring and High Frequency Data Analysis”, Erken laboratoorim, Uppsala Ülikool, Rootsi. 2014 Suvekool “International Summer School of Limnology - Lake Ecology and temporal Dynamics”, Evian-les-Bains, Prantsusmaa. 2014 Koolitus „Open source science tools in GLEON“ GLEON 16 koosolekul, Jouvence, Orford, Quebec, Kanda.

106 LIST OF PUBLICATIONS

Publications indexed in the ISI web of Science database

Woolway, R. Iestyn; Meinson, Pille; Nõges, Peeter; Jones, Ian D.; Laas, Alo (2017). Atmospheric stilling leads to prolonged thermal stratification in a large shallow polymictic lake. Climatic Change, 1−15, 10.1007/s10584-017-1909-0.

Meinson, Pille; Idrizaj, Agron; Nõges, Peeter; Nõges, Tiina; Laas, Alo (2016). Continuous and high-frequency measurements in limnology: History, applications and future challenges. Environmental Reviews, 24, 52−62, 10.1139/er-2015-0030.

Laas, Alo; Cremona, Fabien; Meinson, Pille; Rõõm, Eva-Ingrid; Nõges, Tiina; Nõges, Peeter (2016). Summer depth distribution profiles of dissolved CO22 and O22 in shallow temperate lakes reveal trophic state and lake type specific differences. Science of the Total Environment, 566–567, 63−75, 10.1016/j.scitotenv.2016.05.038.

Articles published in conference proceedings not listed by Thomson Reuters or Scopus

Laas, Alo; Nõges, Peeter; Cremona, Fabien; Rõõm, Eva-Ingrid; Meinson, Pille; Nõges, Tiina (2015). Distribution of dissolved CO22 and O22 in the 8 European Water Framework Directive lake types in Estonia. Abstract book: 9th Symposium for European Freshwater Sciences; July 5-10, 2015, Geneva, Switzerland. Geneva, Switzerland, 201−201.

Articles published in local conference proceedings

Meinson, P. 2014. Fütoplankton kui partikulaarse orgaanilise aine kandja Võrtsjärve sissevooludes ja väljavoolus. Kadri Jõks (eds.) TalveAkadeemia 2014 publications. Tartu: Ecoprint, 123–132.

107 PILLE MEINSON

VIIS VIIMAST KAITSMIST

ISSN 2382-7076 ISBN 978-9949-569-87-8 (trükis) ISBN 978-9949-569-88-5 (pdf) Tartu 2017 Trükitud taastoodetud paberile looduslike trükivärvidega © Kuma Print