(PDF) Stratification strength and light climate explain variation in chlorophyll a at the continental scale in a European multil

Limnol. Oceanogr. 66, 2021, 4314–4333 © 2021 The Authors. Limnology and Oceanography published by Wiley Periodicals LLC on behalf of Association for the Sciences of Limnology and Oceanography. doi: 10.1002/lno.11963 Stratification strength and light climate explain variation in chlorophyll a at the continental scale in a European multilake survey in a heatwave summer Daphne Donis ,1* Evanthia Mantzouki,1 Daniel F. McGinnis,1 Dominic Vachon,1,2 Irene Gallego,1,a Hans-Peter Grossart,3,4 Lisette N. de Senerpont Domis,5,7 Sven Teurlincx,5 Laura Seelen,5,7 Miquel Lürling,5,6 Yvon Verstijnen,6 Valentini Maliaka,6,8,9 Jeremy Fonvielle,3 Petra M. Visser,10 Jolanda Verspagen,10 Maria van Herk,10 Maria G. Antoniou,11 Nikoletta Tsiarta,11 Valerie McCarthy,12 Victor C. Perello,12 Danielle Machado-Vieira,13 Alinne Gurja ~o de Oliveira,13 Dubravka Špoljaric Maronic,14 Filip Stevic,14 Tanja Žuna Pfeiffer,14 Itana Bokan Vucelic,15 Petar Žutinic,16 Marija Gligora Udovič,16 Anđelka Plenkovic-Moraj,16 Luděk Bla ha,17 Rodan Geriš,18 Markéta Fra nkova ,19 Kirsten Seestern Christoffersen, Trine Perlt Warming, Tõnu Feldmann, Alo Laas,21 Kristel Panksep,21 20 20 21 Lea Tuvikene,21 Kersti Kangro,21,22 Judita Koreiviene, _ 23 Ju rate_ Karosiene,_ 23 Ju rate_ Kasperovičiene, _ 23 Ksenija Savadova-Ratkus,23 Irma Vitonyte, _ 23 Kerstin Häggqvist,24 Pauliina Salmi,25 Lauri Arvola,26 Karl Rothhaupt, Christos Avagianos, Triantafyllos Kaloudis,28 Spyros Gkelis,29 Manthos Panou,29 27 28 Theodoros Triantis,30 Sevasti-Kiriaki Zervou,30 Anastasia Hiskia,30 Ulrike Obertegger,31 Adriano Boscaini,31 Giovanna Flaim,31 Nico Salmaso,31 Leonardo Cerasino,31 Sigrid Haande,32 Birger Skjelbred,32 Magdalena Grabowska,33 Maciej Karpowicz,33 Damian Chmura,34 Lidia Nawrocka,35 Justyna Kobos,36 Hanna Mazur-Marzec,36 Pablo Alcaraz-Pa rraga,37 Elżbieta Wilk-Woz niak,38 Wojciech Krzton , Edward Walusiak, Ilona Gagala-Borowska, Joana Mankiewicz-Boczek,39 38 38 39 Magdalena Toporowska,40 Barbara Pawlik-Skowronska,40 Michał Niedz wiecki,40 Wojciech Pęczuła,40 Agnieszka Napio rkowska-Krzebietke,41 Julita Dunalska,42 Justyna Sien ska,42 Daniel Szyman ski,42 43 Marek Kruk, Agnieszka Budzyn ska, Ryszard Goldyn, Anna Kozak, Joanna Rosin 44 44 44 ska,44 Elżbieta Szeląg-Wasielewska, Piotr Domek, Natalia Jakubowska-Krepska, Kinga Kwasizur,45 44 44 44 Beata Messyasz,45 Aleksandra Pełechata,45 Mariusz Pełechaty,45 Mikolaj Kokocinski,45 Beata Madrecka-Witkowska,46 Iwona Kostrzewska-Szlakowska,47 Magdalena Frąk,48 Agnieszka Ban kowska-Sobczak,49 Michał Wasilewicz,49 Agnieszka Ochocka,50 Agnieszka Pasztaleniec,50 51 Iwona Jasser, Ana M. Anta ~o-Geraldes,52 Manel Leira,53 Vitor Vasconcelos,54 Joao Morais,54 Micaela Vale, Pedro M. Raposeiro,55 Vítor Gonçalves,55 Boris Aleksovski,56 Svetislav Krstic,56 54 Hana Nemova,57 Iveta Drastichova,57 Lucia Chomova,57 Spela Remec-Rekar,58 Tina Elersek,59 Lars-Anders Hansson,60 Pablo Urrutia-Cordero,60,61 Andrea G. Bravo,61 Moritz Buck,61 William Colom-Montero,62 Kristiina Mustonen,62 Don Pierson,62 Yang Yang,62 Jessica Richardson,63 Christine Edwards,64 Hannah Cromie,65 Jordi Delgado-Martín,66 David García,66 Jose Luís Cereijo,66 Joan Gomà,67 Mari Carmen Trapote,67 Teresa Vegas-Vilarrúbia,67 Biel Obrador,67 Ana García-Murcia,68 Monserrat Real,68 Elvira Romans,68 Jordi Noguero-Ribes,68 David Parreño Duque,68 *Correspondence:

Author Contribution Statement: D.D. analyzed and worked on data visualization, coordinated feedback from coauthors, and wrote the manu- This is an open access article under the terms of the Creative Commons script. E.M. coordinated the EMLS, collected data, curated the dataset, Attribution-NonCommercial-NoDerivs License, which permits use and dis- analyzed the data, and contributed to writing the manuscript. tribution in any medium, provided the original work is properly cited, the B.I. conceived the idea for the EMLS, contributed to discussions through- use is non-commercial and no modifications or adaptations are made. out the study and to the writing of the manuscript. D.M., D.V., I.G., H.-P. G., L.N.d.S.D., S.T., L.S., N.C., A.G.B., M.B., P.V., and C.C. assisted in ana- Additional Supporting Information may be found in the online version of lyzing and interpreting the dataset. The rest of the coauthors were this article. responsible for finalizing the sampling protocols, organizing the local sur- veys, collecting data in their respective countries, and providing invalu- a Present address: Department Aquatic Ecology, Eawag Überlandstrasse, able feedback on the manuscript and data analysis. Dübendorf, Switzerland 4314 Donis et al. European lake survey: summer Chl-a drivers Elísabeth Fernandez-Mora n,68 Ba rbara Úbeda,69 José Angel Galvez,69 Núria Catala n,70 Carmen Pérez-Martínez, Eloísa Ramos-Rodríguez, Carmen Cillero-Castro, Enrique Moreno-Ostos,73 71 71 72 José María Blanco,73 Valeriano Rodríguez,73 Jorge Juan Montes-Pérez,73 Roberto L. Palomino,73 Estela Rodríguez-Pérez,73 Armand Herna ndez,74 Rafael Carballeira,75 Antonio Camacho,76 Antonio Picazo, Carlos Rochera, Anna C. Santamans,76 Carmen Ferriol,76 Susana Romo,77 76 76 Juan Miguel Soria,77 Arda Özen,78 Tünay Karan,79 Nilsun Demir,80 Meryem Bekliog lu,81 Nur Filiz,81 81 Eti Levi, Ug ur Iskin, Gizem Bezirci, Ülkü Nihan Tavşanog 81 81 lu, Kemal Çelik, Koray Ozhan,83 81 82 Nusret Karakaya,84 Mehmet Ali Turan Koçer,85 Mete Yilmaz,86 Faruk Maraşlıog lu,87 Özden Fakioglu,88 89 Elif Neyran Soylu, Meral Apaydın Yag cı, Şakir Çınar, Kadir Çapkın, Abdulkadir Yag 90 90 90 cı,90 Mehmet Cesur, Fuat Bilgin, Cafer Bulut, Rahmi Uysal, Köker Latife, Reyhan Akçaalan,91 90 90 90 90 91 Meriç Albay,91 Mehmet Tahir Alp,92 Korhan Özkan,93 Tug ba Ongun Sevindik,94 Hatice Tunca,94 Burçin Önem,94 Hans Paerl,95 Cayelan C. Carey,96 Bastiaan W. Ibelings1 1 Department F.-A. Forel for Environmental and Aquatic Sciences and Institute for Environmental Sciences, University of Geneva, Geneva, Switzerland 2 Department of Ecology and Environmental Sciences, Umeå University, Umeå, Sweden 3 Department of Experimental Limnology, Leibniz Institute of Freshwater Ecology and Inland Fisheries, Stechlin, Germany 4 Institute of Biochemistry and Biology, Potsdam University, Potsdam, Germany 5 Department of Aquatic Ecology, Netherlands Institute of Ecology (NIOO-KNAW), Wageningen, The Netherlands 6 Department of Environmental Sciences, Wageningen University & Research, Wageningen, The Netherlands 7 Department of Environmental Sciences, Aquatic Ecology and Water Quality Management group, Wageningen University, Wageningen, 6708 PB, The Netherlands 8 Society for the Protection of Prespa, Agios Germanos, Greece 9 Department of Aquatic Ecology and Environmental Biology, Institute for Water and Wetland Research, Radboud University Nijmegen, Nijmegen, The Netherlands 10 Department of Freshwater and Marine Ecology, Institute for Biodiversity and Ecosystem Dynamics, University of Amsterdam, Amsterdam, The Netherlands 11 Department of Chemical Engineering, Cyprus University of Technology, Lemesos, Cyprus 12 Centre for Freshwater and Environmental Studies, Dundalk Institute of Technology, Dundalk, Ireland 13 Departamento de Sistematica e Ecologia, Universidade Federal da Paraíba, Paraíba, Brazil 14 Department of Biology, Josip Juraj Strossmayer University of Osijek, Osijek, Croatia 15 Department for Ecotoxicology, Teaching Institute of Public Health of Primorje-Gorski Kotar County, Rijeka, Croatia 16 Department of Biology, Faculty of Science, University of Zagreb, Zagreb, Croatia 17 RECETOX, Faculty of Science, Masaryk University, Brno, Czech Republic 18 Department of Hydrobiology, Morava Board Authority, Brno, Czech Republic 19 Department of Paleoecology, Institute of Botany, The Czech Academy of Sciences, Brno, Czech Republic 20 Freshwater Biological Laboratory, Department of Biology, University of Copenhagen, Copenhagen, Denmark 21 Institute of Agricultural and Environmental Sciences, Estonian University of Life Sciences, Tartu, Estonia 22 Tartu Observatory, Faculty of Science and Technology, University of Tartu, Tartu, Estonia 23 Institute of Botany, Nature Research Centre, Vilnius, Lithuania 24 Department of Science and Engineering, Åbo Akademi University, Åbo, Finland 25 Department of Biological and Environmental Science, University of Jyväskylä, Jyväskylä, Finland 26 Lammi Biological Station, University of Helsinki, Lammi, Finland 27 Department of Biology, Limnological Institute, University of Konstanz, Konstanz, Germany 28 Water Quality Department, Athens Water Supply and Sewerage Company, Athens, Greece 29 Department of Botany, School of Biology, Aristotle University of Thessaloniki, Thessaloniki, Greece 30 Institute of Nanoscience and Nanotechnology, National Center for Scientific Research «DEMOKRITOS», Agia Paraskevi, Attiki, Greece 31 Research and Innovation Centre, Fondazione Edmund Mach, San Michele all’Adige, 38010, Italy 32 Department of Freshwater Ecology, Norwegian Institute for Water Research, Oslo, Norway 33 Department of Hydrobiology, University of Bialystok, Bialystok, Poland 34 Institute of Environmental Protection and Engineering, University of Bielsko-Biala, Bielsko-Biala, Poland 35 Institute of Technology, The State University of Applied Sciences, Elblag, Poland 36 Department of Marine Biotechnology, University of Gdansk, Gdynia, Poland 37 Department of Animal Biology, Plant Biology and Ecology, University of Jaen, Jaen, Spain 38 Institute of Nature Conservation, Polish Academy of Sciences, Krakow, Poland 4315 Donis et al. European lake survey: summer Chl-a drivers 39 European Regional Centre for Ecohydrology of the Polish Academy of Sciences, Lodz, Poland 40 Department of Hydrobiology and Protection of Ecosystems, University of Life Sciences in Lublin, Lublin, Poland 41 Department of Ichthyology, Hydrobiology and Aquatic Ecology, S. Sakowicz Inland Fisheries Institute, Olsztyn, 10-719, Poland 42 Department of Water Protection Engineering, University of Warmia and Mazury, Olsztyn, Poland 43 Department of Applied Computer Science and Mathematical Modelling, University of Warmia and Mazury, Olsztyn, 10-710, Poland 44 Department of Water Protection, Faculty of Biology, Adam Mickiewicz University, Poznan, Poland 45 Department of Hydrobiology, Faculty of Biology, Adam Mickiewicz University, Poznan, Poland 46 Institute of Environmental Engineering and Building Installations, Faculty of Environmental Engineering and Energy, Poznan University of Technology, Poznan, 60965, Poland 47 Faculty of Biology, University of Warsaw, Warsaw, Poland 48 Department of Remote Sensing and Environmental Assessment, Institute of Environmental Engineering, Warsaw University of Life Sciences - SGGW, Nowoursynowska Str. 166, Warsaw, 02-787, Poland 49 Department of Water Engineering and Applied Geology, Faculty of Civil and Environmental Engineering, Warsaw University of Life Sciences – SGGW, Warsaw, 02-787, Poland 50 Department of Freshwater Protection, Institute of Environmental Protection - National Research Institute, Warsaw, Poland 51 Department of Plant Ecology and Environmental Conservation, Faculty of Biology, University of Warsaw, Warsaw, 02-089, Poland 52 Centro de Investigaç~ao da Montanha (CIMO), Instituto Politécnico de Bragança, Bragança, Portugal 53 BioCost Research Group, Faculty of Science and Centro de Investigacio ns Científicas Avanzadas (CICA), Department of Biology, Faculty of Science, University of A Coruña, A Coruña, 15071, Spain 54 Interdisciplinary Centre of Marine and Environmental Research (CIIMAR/CIMAR), University of Porto, Terminal de Cruzeiros do Porto de Leixões, Matosinhos, 4450-208, Portugal 55 Research Center in Biodiversity and Genetic Resources (CIBIO-Azores), InBIO Associated Laboratory, Faculty of Sciences and Technology, University of the Azores, Ponta Delgada, 9500-321, Portugal 56 Faculty of Natural Sciences and Mathematics, SS Cyril and Methodius University, Skopje, Macedonia 57 National Reference Center for Hydrobiology, Public Health Authority of the Slovak Republic, Bratislava, Slovakia 58 Department of Water Quality, Slovenian Environmental Agency, Ljubljana, Slovenia 59 Department of Genetic Toxicology and Cancer Biology, National Institute of Biology, Ljubljana, Slovenia 60 Department of Biology, Lund University, Lund, Sweden 61 Department of Ecology and Genetics, Limnology, Uppsala University, Uppsala, Sweden 62 Department of Ecology and Genetics, Erken Laboratory, Uppsala University, Norrtalje, Sweden 63 Department of Biological and Environmental Sciences, University of Stirling, Stirling, UK 64 School of Pharmacy and Life Sciences, Robert Gordon University, Aberdeen, UK 65 Agri-Food & Biosciences Institute, Belfast, UK 66 Department of Civil Engineering, University of A Coruña, A Coruña, Spain 67 Department of Evolutionary Biology, Ecology, and Environmental Sciences, University of Barcelona, Barcelona, Spain 68 Department of Limnology and Water Quality, AECOM U.R.S., Barcelona, Spain 69 Department of Biology, INMAR Marine Research Institute, University of Cadiz, Cadiz, 11510 Puerto Real, Spain 70 Catalan Institute for Water Research (ICRA), Girona, Spain 71 Department of Ecology and Institute of Water Research, University of Granada, Granada, Spain 72 R&D Department Environmental Engineering, 3edata, Lugo, Spain 73 Department of Ecology, University of Malaga, Malaga, Spain 74 Institute of Earth Sciences Jaume Almera, ICTJA, CSIC, Barcelona, Spain 75 Centro de Investigacio ns Cientificas Avanzadas (CICA), Facultade de Ciencias, Universidade da Coruña, A Coruña, Spain 76 Cavanilles Institute of Biodiversity and Evolutionary Biology, University of Valencia, Valencia, Spain 77 Department of Microbiology and Ecology, University of Valencia, Burjassot, Spain 78 Department of Forest Engineering, University of Cankiri Karatekin, Cankiri, Turkey 79 Department of Animal Nutrition and Zootechnics, Faculty of Veterinary Medicine, Yozgat Bozok University, Yozgat, Turkey 80 Department of Fisheries and Aquaculture Engineering, Ankara University, Ankara, 06110, Turkey 81 Department of Biological Sciences, Limnology Laboratory, Middle East Technical University, Ankara, Turkey 82 Department of Biology, Balikesir University, Balikesir, Turkey 83 Department of Oceanography, Institute of Marine Sciences, Middle East Technical University, Ankara, Turkey 84 Department of Environmental Engineering, Abant Izzet Baysal University, Bolu, Turkey 85 Department of Environment and Resource Management, Mediterranean Fisheries Research Production and Training Institute, Antalya, Turkey 86 Department of Bioengineering, Bursa Technical University, Bursa, Turkey 4316 Donis et al. European lake survey: summer Chl-a drivers 87 Department of Biology, Hitit University, Corum, Turkey 88 Department of Basic Science, Ataturk University, Erzurum, Turkey 89 Department of Biology, Giresun University, Giresun, Turkey 90 Republic of Turkey Ministry of Food Agriculture, Fisheries Research Institute, Isparta, Turkey 91 Department of Freshwater Resource and Management, Faculty of Aquatic Sciences, Istanbul University, Istanbul, Turkey 92 Faculty of Aquaculture, Mersin University, Mersin, Turkey 93 Institute of Marine Sciences, Marine Biology and Fisheries, Middle East Technical University, Mersin, Turkey 94 Department of Biology, Sakarya University, Sakarya, Turkey 95 Institute of Marine Sciences, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 96 Department of Biological Sciences, Virginia Tech, Blacksburg, Virginia Abstract To determine the drivers of phytoplankton biomass, we collected standardized morphometric, physical, and biological data in 230 lakes across the Mediterranean, Continental, and Boreal climatic zones of the European continent. Multilinear regression models tested on this snapshot of mostly eutrophic lakes (median total phos- phorus [TP] = 0.06 and total nitrogen [TN] = 0.7 mg L1), and its subsets (2 depth types and 3 climatic zones), show that light climate and stratification strength were the most significant explanatory variables for chloro- phyll a (Chl a) variance. TN was a significant predictor for phytoplankton biomass for shallow and continental lakes, while TP never appeared as an explanatory variable, suggesting that under high TP, light, which partially controls stratification strength, becomes limiting for phytoplankton development. Mediterranean lakes were the warmest yet most weakly stratified and had significantly less Chl a than Boreal lakes, where the temperature anomaly from the long-term average, during a summer heatwave was the highest (+4 C) and showed a signifi- cant, exponential relationship with stratification strength. This European survey represents a summer snapshot of phytoplankton biomass and its drivers, and lends support that light and stratification metrics, which are both affected by climate change, are better predictors for phytoplankton biomass in nutrient-rich lakes than nutrient concentrations and surface temperature. Globally, temperature, light, and nutrients are key drivers since the early days of eutrophication research (Mur of phytoplankton blooms, but their relative importance in et al. 1977). In general, by controlling light and nutrient avail- determining algal biomass strongly depends on the role of ability, the underwater light climate and stratification strength thermal stratification, that is, water column stability determine phytoplankton growth conditions. When stratifica- (Sverdrup 1953; Cloern 1996; Ptacnik et al. 2003; Carvalho tion is strong, thus suppressing fluxes from the deeper layers, et al. 2016). As a matter of fact, the relative importance of mixing is restricted to the surface layer. Under such condi- these drivers and interactive mechanisms between them can- tions, phytoplankton is constantly maintained within the not be fully resolved without including thermal stability euphotic zone, promoting algal growth until nutrients are (Winslow et al. 2017). This is particularly relevant under depleted or other factors as grazing and sedimentation take global processes of eutrophication and climate warming over in controlling phytoplankton biomass (Fig. 1a; Cam- (Sinha et al. 2017) as some research foresees an allied impact acho 2006; Reynolds 2006; Yankova et al. 2017). When strati- of eutrophication and climate change effects in promoting fication is weak, water column mixing can reach deep and harmful cyanobacterial blooms (Moss et al. 2011). nutrient rich waters, however potentially taking the algal com- Stratification suppresses the exchange of heat and dissolved munities beyond the euphotic zone that would limit their substances between the epi- and hypolimnion by reducing tur- growth (Ibelings et al. 1994; Fig. 1b). One other ecological bulent motions that otherwise would facilitate transport consequence of a strongly stratified lake is that phytoplankton (Wüest and Lorke 2003). While the vertical structure of the may have reduced access to nutrients that remain locked in water column constitutes the first response to temperature fluc- the hypolimnion (Nürnberg 1984; Posch et al. 2012; Sal- tuations (Sahoo et al. 2016), it also regulates the development maso et al. 2020). Yet, while the strength of stratification is of phytoplankton biomass by affecting light and nutrient avail- determined primarily by light climate and heat exchange, ability (Yang et al. 2016), as well as phytoplankton settling, and other factors too can affect the extent and duration of the therefore exerts a strong control on lake ecosystem functioning stratification, such as lake morphology (i.e., basin geome- (Scheffer et al. 2001; Bartosiewicz et al. 2015). try, maximum depth and surface area) (Thompson and Especially when nutrients are not limiting (e.g., in eutro- Schmidt 2005; Kirillin and Shatwell 2016; Magee and phic lakes), light climate and stratification strength likely play Wu 2017) as well as the dissolved organic and inorganic dominant roles in regulating phytoplankton biomass (Fig. 1), carbon content of the water, wind orientation and shelter- and this role of light as a limiting resource has been suggested ing. Dissolved organic matter in general can have a huge 4317 Donis et al. European lake survey: summer Chl-a drivers climatic gradients, the “grassroots” European Multi Lake Sur- vey (EMLS) was organized during summer 2015, which coin- cided with the period of maximum stratification in most of the examined lakes. Data from the EMLS are publicly available (Mantzouki et al. 2018). Here, we report on the difference in Chl a as a proxy for phytoplankton biomass between 230 of the EMLS lakes to: (1) determine the dependency of phyto- plankton biomass at the continental scale on a set of ecosys- tem drivers, including growth conditions (total phosphorous [TP], total nitrogen [TN], lake temperature, and light) and morphophysical properties (lake depth, surface area, light cli- mate, and stratification strength); and (2) investigate potential interactions between these predictors that influence phyto- plankton biomass. Fig 1. Schematic overview of how lake N2 and light climate (Zeu/Zmix) may define phytoplankton biomass in nutrient-rich lakes. (a) A strong stratification (> N2) allows phytoplankton to circulate well within the Methods euphotic zone (Zeu/Zmix ≥ 1)—promoting growth. (b) A weaker stratifica- tion (< N2) allows deeper mixing, hence phytoplankton communities are EMLS organization highly diluted—eventually below the euphotic zone (Zeu/Zmix < 1). During the EMLS in summer 2015, 230 lakes were sam- pled across major geographical and climatic regions in impact on stratification by influencing light penetration, and Europe for various chemical, physical, and biological consequently surface heating, as seen in humic boreal lakes parameters using highly standardized sampling protocols (Heiskanen et al. 2014). Wind and convection, acting on the sur- (Mantzouki et al. 2018; Mantzouki and Ibelings 2018). All face mixed layer (SML), control a lake’s interior diffusive fluxes key variables were analyzed centrally (by one person on one regulating the physical environment experienced by phyto- machine) in dedicated laboratories to ensure data compara- plankton. Important properties of the SML, such as its depth, bility and a fully integrated dataset. vary widely among lakes as the result of a specific balance The lake sampling site was selected as either the historical between factors that strengthen stratification (surface warming), sampling point, for which long-term records exist, or the geo- and factors that disrupt or deepen the layer, such as wind shear graphic center of the lake. The sampling period was defined as and surface cooling (Imberger 1985; Imboden and Wüest 1995; the warmest 2-week period of the summer, based on long-term Boehrer and Schultze 2008). (minimum 10 yr) air temperature data of each region. An in Stratification of lakes is changing under the impact of situ temperature profile carried out on the sampling day was eutrophication, re-oligotrophication and climate warming used to identify and characterize the thermocline as the point (Flaim et al. 2016). For instance, in recent decades, the where there was ≥ 1 C change of temperature per meter lake strength of stratification of lakes in northeastern North Amer- depth. An integrated water sample was obtained from 0.5 m ica has clearly increased (Richardson et al. 2017); a phenome- depth to the bottom of the thermocline using a water sampler non that might be further enhanced by a trend of that could effectively sample the whole volume without creat- atmospheric stilling (Woolway and Merchant 2019). Analyses ing intervals. In nonstratified shallow lakes, an integrated of the 2007 National Lake Assessment, NLA dataset (Pollard sample was drawn from 0.5 m below the lake surface to 0.5 m et al. 2018) showed that synergistic interactions between above the lake bottom. nutrients and temperature promoting algal or cyanobacterial developments are probable, especially in the eutrophic and hypereutrophic subsets of NLA lakes (Rigosi et al. 2014). Nutrient analyses Kosten et al. (2012) provided more support for synergistic Total phosphorus and nitrogen concentrations were interactions between nutrients and temperature in determin- assessed in unfiltered samples. Sample bottles were acid ing chlorophyll a (Chl a) and cyanobacterial dominance in a washed overnight in 1 M HCl and rinsed with demineralized multilake survey along a latitudinal gradient stretching from water before usage. Nutrients were measured using a Skalar the tip of South America to the equator. However, no lake SAN+ segmented flow analyzer (Skalar Analytical BV, Breda, physical variables other than surface temperature, such as den- the Netherlands) with UV/persulfate digestion integrated in sity gradient or stratification strength, were included in these the system. The limit of detection was 0.02 mg L1 for TP and large-scale studies on drivers of algal biomass. 0.2 mg L1 for TN. TP was analyzed following NEN (1986) and To further our understanding of the main drivers and their TN according to NEN (1990). All nutrient analyses were per- interactions on phytoplankton biomass across continental formed at the University of Wageningen, the Netherlands. 4318 Donis et al. European lake survey: summer Chl-a drivers Pigment analyses Lake groups Pigment analysis, modified from the method described by Lake classification was based on climatic zone and depth Van der Staay et al. (1992), was carried out to determine con- type. Predicted climatic zones based on different IPCC scenar- centrations of Chl a and Zeaxanthin (Zea). Measurement of ios (2000–2025; Rubel and Kottek 2010) were used to avoid Zea concentrations in the EMLS lakes were carried out with the inconsistency in available digital maps, especially for areas the aim of investigating cyanobacterial biomass, alongside to such as the Alpine region (Rubel et al. 2017). The climatic the general phytoplankton biomass estimate obtained with zones were defined using the Köppen-Geiger’s classification Chl a. Filters (45 mm diameter GF/C or /F) were freeze-dried (Köppen 1900). This classification regards the main climate of for 6 h and then cut in half, placed in separate Eppendorf the region (C = warm temperate, D = alpine), precipitation tubes, and kept on ice. A number of 0.5 mm beads and 600 μL levels (f = fully humid, s = summer dry), and mean tempera- of 90% acetone were added to each tube. To release the pig- ture (a = hot summers, b = warm summers). For easier inter- ments from the phytoplankton cells and increase the extrac- pretation and more statistical power, climatic regions that tion yield, tubes were placed on a bead-beater for 1 min and were of the same main climate and precipitation level were then in an ultrasonic bath for 10 min. To ensure complete combined in three main ones: Mediterranean (Csa and Csb, extraction of the total pigment content of the filters, the bead- n = 54 lakes), Continental (Cfa and Cfb, n = 128 lakes), and beater and ultrasonic bath steps were performed twice. To Boreal (Dfb and Dfc, n = 48 lakes) (Fig. 2). This way, only the achieve binding of the pigments during the high-performance mean temperature varied within each of the combined groups, liquid chromatography (HPLC) analysis, 300 μL of a tributyl which allowed for testing of a temperature gradient. The selec- ammonium acetate (1.5%) and ammonium acetate (7.7%) tion of climatic zones has a clear advantage over a latitudinal mix were added to each tube. Lastly, samples were centrifuged analysis, as several lakes within the Continental region are at 15,000 rpm and 4 C for 10 min. Next, 35 μL of the superna- classified as Boreal lakes based on their climatic characteristics tant from Eppendorf tubes were transferred into glass HPLC rather than their position on a latitudinal gradient (see sampling vials. Pigments were separated on a Thermo Scien- Table S1 for list of EMLS lakes and corresponding cli- tific ODS Hypersil column (250 mm 3 mm, particle size matic zone). 5 μm) in a Shimadzu HPLC, using a KONTRON SPD-M2OA The EMLS lakes were categorized into shallow (< 6 m maxi- diode array detector. The different pigments were identified mum depth, n = 93 lakes) and deep (> 6 m maximum depth, based on their retention time and absorption spectrum and n = 137 lakes). This classification was used in previous snap- quantified by means of pigment standards. Pigment analysis shot surveys as an approximation for weakly or strongly ther- was performed at the University of Amsterdam, the mally stratified systems (Kosten et al. 2012; Beaulieu Netherlands. et al. 2013). Fig 2. Location of the 230 EMLS lakes distributed over the main climatic zones of the European continent (Rubel and Kottek 2010). The Mediterranean region (n = 54) consists of Csa and Csb classes (C, warm temperate; s, summer dry; a, hot summer; b, warm summer), the Continental region (n = 128) of Cfa and Cfb (f, fully humid; rest as above), and the Boreal region (n = 48) of Dfb and Dfc (D, snow; c, cool summer; rest as above). 4319 Donis et al. European lake survey: summer Chl-a drivers Table 1. List of lake variables with their units, range of values, means, medians, and standard deviations for the 230 EMLS lakes. Vari- ables with * are included in the linear models. Variable Units Range Mean SD Median Maximum depth maxD m 1–310 2341 10.00 Surface area* SurfA km2 0.001–580 1969 5 Total nitrogen* TN mg L1 0.1–5 1.00.8 0.70 Total phosphorus* TP mg L1 0.02–1 0.10.1 0.06 Surface temperature* SurfT C 14.6–33 233.4 22.4 Average temperature AvT C 13.4–33 213.5 20.6 Secchi depth SD m 0.16–10 1.81.7 1.19 Light climate* Zeu/Zmix - 0.02–11 1.01.0 0.63 Stratification strength * N2 s2 3105–3102 51034103 4.103 Chlorophyll a* Chl a μg L1 0.03–933 44110 9.98 Zeaxanthin Zea μg L1 0.00–90 39.7 0.68 Statistical analysis the combined effects of salinity (set to 0) and water temperature To disentangle the importance of various drivers of phyto- based on the method of Millero and Poisson (1981). plankton biomass, we applied linear regression models to six Lake stratification is the density-induced layering of the lake groups: all, deep, shallow, Mediterranean, Continental, water column (Boehrer and Schultze 2008). Strength of water Boreal. We (1) assessed the quality of the statistical models column stratification was determined by the N2 given by the after excluding collinear and nonsignificant variables, Brunt Väisälä equation or buoyancy frequency, N (s1). (2) included groups of interactions and nominal variables as sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi environmental predictors, and (3) discussed the three most g ∂ρ g ∂ρ important predictors for each model. N¼ ;N 2 ¼ ð1Þ ρ ∂z ρ ∂z Buoyancy frequency is greater than zero, when a water vol- Response variable and environmental predictors ume (of density ρ) that is displaced vertically (z) from its initial The response variable of all regression models was the con- position without heat transfer, experiences a restoring force. If centration of Chl a obtained from the HPLC analysis, which N2 < 0 instead, the water parcel tends to be displaced away was used as a proxy for total phytoplankton biomass from its initial position and the vertical water column is (Pinckney et al. 2001; Tamm et al. 2015) and tested with the locally unstable. Here, we use the symbol N2 to indicate the following single predictors: maximum depth (maxD), surface maximum value over the entire water column. By suppressing area (SurfA), TN, TP, surface temperature (SurfT), average tem- vertical turbulent eddies, density stratification determines the perature (AvT), Secchi disk depth (SD), light climate (Zeu/ water column stability so that, in general, the greater the den- Zmix), and maximum buoyancy frequency (stratification sity gradient, the slower the diffusive exchange of water con- strength, N2) (Table 1). stituents between the hypolimnion and the epilimnion Surface and average temperatures were determined via a (Boehrer and Schultze 2008). water column profile with a temperature probe, taking respec- Three groups of interactions between some of the afore- tively the temperature of the top 0.5 m of the water column mentioned variables, selected based on ecological theory and and the average of the full profile. previous literature, were included as additional predictors in Light climate was defined as the ratio of euphotic depth over the models. Namely the interaction between (1) nutrients and mixing depth (Zeu/Zmix), which describes the light that phyto- surface temperature (Rigosi et al. 2014), (2) stratification plankton experience while circulating through the water col- strength and light climate (Graff and Behrenfeld 2018), and umn (Scheffer et al. 1997). The equation Zeu = 2 SD (Secchi (3) surface area and light climate. depth) was used to calculate Zeu (equation selected as an aver- age estimate from the range of constants reported in literature, Analysis of variance e.g., Koenings and Edmundson 1991; Salmaso 2002; Brentrup Differences in mean values of the selected variables within cli- et al. 2018). In stratified lakes, Zmix was determined as the matic zones and depth types were tested using one-way ANOVA. depth of the steepest density gradient (Winslow et al. 2017). In Homogeneity of variance was tested using the Levene’s test from nonstratified shallow lakes, Zmix matched the maximum depth the car R package (Fox and Weisberg 2011). In case of heteroge- and sampling depth. Water density was calculated according to neity, a Kruskal–Wallis test was used instead of ANOVA. Post hoc 4320 Donis et al. European lake survey: summer Chl-a drivers pairwise comparisons for unequal sample sizes were performed those differences included zero, it indicated that the predictors using Tukey HSD (Honest Significant Difference) or Games– were not significantly different from each other, meaning that Howell test (userfriendlyscience R package; Peters et al. 2018) for they contributed similarly to the interaction term. When the homogeneous or heterogeneous variance, respectively. interaction term had a significant value of p < 0.05 and was positive, it was interpreted as a synergistic interaction. Multiple linear regression model To avoid multicollinearity between the interactions and All variables were log-transformed (natural logarithm) to their main effects, we checked the variance inflation factor obtain a normal and homogeneous distribution. Stepwise (VIF). If VIFs were exhibiting high numbers (VIF > 3, threshold selection (backwards and forward) was used for model selec- according to (Zuur et al. 2010), we centered the interaction tion where the AIC scores were compared based on a modified term with the mean of the raw variables which alleviated the equation that corrects for unequal sample size among catego- collinearity problem. ries (R code provided by Statoo Consulting, Switzerland). If We applied multiple linear regression models to test the rel- the interaction term was significant (p ≤ 0.05), the lower order ative importance of the selected response variables in terms were included in the equation. The most parsimonious explaining Chl a variance. The model applied was: model, in which elimination or addition of any other predic- tors would not improve the model by ΔAIC > 2, was used for Chl a ¼ A0 þ A1 XSurfA þ A2 XN2 þ A3 XSurfT þ A4 XTN þ A5 XZeu=Zmix the ANOVA. The metric “lmg” of the relaimpo R package þ A6 XN2Zeu=Zmix þ ε (Gro} mping 2006) was used to decompose the overall R2 of ð2Þ each final model into the absolute contributions of each pre- dictor term and their interaction terms (similarly done in where A0 represents the intercept term, A1–A6 are model parame- Rigosi et al. 2014). The relative contribution of each predictor ters for each respective predictor in the models, “*” denotes the was normalized, by forcing the sum to 100%. A bootstrapping interaction between two terms, and ε is an error term. Two multi- approach was used to replicate the observed data 9999 times ple linear regression models were applied to the entire EMLS and determine if there were any clear differences between the group of lakes. Apart from the full set of environmental predictors, predictors of the interaction terms with regards their relative each of these two models included the nominal variable “depth contribution to the interaction term (Gro } mping 2006). If type” or “climatic zone” (see Supplementary Material for more Table 2. List of applied models and relative metrics. AIC does not apply correctly if number of observations is not the same, for which we rely on R2. Lake group Multilinear model N lakes R2 AIC (1) All-a Chl a = 9.06 0.23 (SurfA) 0.31 (N ) + 3.36 (SurfT) 2 230 35% 842.43*** + 0.46 (TN) + 0.47 (Zeu/Zmix) + 0.18 (N2*Zeu/Zmix) 0.90 (Cont) 2.06 (Med) (2) All-b Chl a = 2.44 0.15 (SurfA) 0.11 (N2) + 1.12 (SurfT) 230 30% 856.65** + 0.31 (TN) + 0.45 (Zeu/Zmix) + 0.19 (N2*Zeu/Zmix) + 1.17 (Shallow) (3) Shallow Chl a = 0.33–0.05 (SurfA) 0.17 (N2) + 0.48 (SurfT) 93 31% Na + 0.78 (TN) 0.09 (Zeu/Zmix) + 0.12 (N2*Zeu/Zmix) (4) Deep Chl a = 0.65 0.14 (SurfA) + 0.07 (N2) + 0.84 (SurfT) 137 12% Na + 0.01 (TN) + 1.13 (Zeu/Zmix) + 0.29 (N2*Zeu/Zmix) (5) Med. Chl a = 17.035 0.23 (SurfA) 0.029 (N2) + 5.26 54 45% Na (SurfT) + 0.40 (TN) + 0.83 (Zeu/Zmix) + 0.22 (N2*Zeu/ Zmix) (6) Cont. Chl a = 5.40 0.26 (SurfA) 0.33 (N2) + 1.88 (SurfT) 128 25% Na + 0.47 (TN) + 0.23 (Zeu/Zmix) + 0.12 (N2*Zeu/Zmix) (7) Bor. Chl a = 15.65 0.15 (SurfA) 0.005 (N2) + 6.05 48 43% Na (SurfT) + 0.41 (TN) + 2.77 (Zeu/Zmix) + 0.62 (N2*Zeu/ Zmix) * P ≤ 0.05. ** P ≤ 0.01. *** P ≤ 0.001. 4321 Donis et al. European lake survey: summer Chl-a drivers Fig 3. EMLS log-transformed response variable, (a) Chl a, and significant Fig 4. EMLS log-transformed response variable (a) Chl a, and significant predictors: (b) surface temperature, (c) total nitrogen, (d) maximum predictors: (b) surface temperature, (c) total nitrogen, (d) maximum buoyancy frequency (stratification strength), (e) light climate (Zeu/Zmix), buoyancy frequency (stratification strength), (e) light climate (Zeu/Zmix), and (f) surface area, averaged over climatic zones. Significant differences and (f) surface area, averaged over depth type. Significant differences at at the 0.05 level are marked with *. Different italic letters indicate signifi- the 0.05 level are marked with *. cant differences among categories (Tukey test; p < 0.05). detail). These nominal variables comprehend the lake subsets to Significant collinearity was observed between maximum which the same multilinear regression model that was further depth and surface area, and between surface temperature and applied, that is, deep, shallow, Mediterranean, Continental, Boreal average temperature (Fig. S1). VIFs of maximum depth and (Table 2). average temperature were higher than 3, thus they were removed from subsequent analyses. Secchi depth was also removed in favor of using the light climate variable, Zeu/Zmix. Results All the variables were found to be significant, except for TP, and the interactions TN*SurfT and SurfA*Zeu/Zmix, which Response variable and environmental predictors therefore never appeared as Chl a variance predictors. The EMLS lake data cover a wide range of morphological, physical, chemical, and biological values (Table 1). The median measured TP was 60 μg L1, and according to Carlson Lake groups: Climatic zone and depth type trophic state index (TSI) 85% of the lakes were classified as Climatic zone eutrophic (TSI > 50). EMLS lakes were largely represented by ANOVA was performed on the three climatic zone groups, eutrophic conditions (70%) also when calculating the TSI on composed by 54 Mediterranean, 128 Continental, and basis of Secchi disk depth (median SD = 1.2 m), while TSI 48 Boreal lakes (Fig. 3; Table S2). based on Chl a concentration (median Chl a = 10 μg L1) Mean Chl a concentrations were significantly higher in the leads to 54% of lakes being classified as eutrophic. Boreal lakes (mean ln 1 SD, 3 1 μg L1) compared to 4322 Donis et al. European lake survey: summer Chl-a drivers Continental (2.2 1 μg L1) and Mediterranean the lowest variability (R2 = 12%), compared to R2 = 31% for 1 (1.7 2 μg L ), while no significant difference was found shallow lakes. Based on AIC comparison, the nominal variable between Continental and Mediterranean lakes (Fig. 3a; “climatic zone” is more significant than “depth type” in Table S2). explaining the variance of algal biomass (Table 2). Nevertheless, the lake group “depth type” explained more of the overall R2 Depth type compared to “climatic zone” (37% vs. 26%; Table S3). The EMLS dataset is composed of 93 shallow and 137 deep When the model included the nominal variable “climatic lakes (> 6 m). Response variable, Chl a, and all of the predic- zone” among the predictors, it resulted as the strongest predic- tors used in the statistical models of the EMLS significantly tor for algal biomass with 26% of the model R2 explained, differed between deep and shallow lakes (Fig. 4; Table S2). Chl closely followed by stratification strength (24%), and with a a, SurfT, TN, and Zeu/Zmix were all higher for shallow lakes, significant but smaller contribution of TN (13%; Table S3). whereas deep lakes showed a stronger stratification strength Similarly, when “depth type” was included, it resulted as the (N2) and greater surface area than shallow ones. most significant predictor (37%); however, it was much more important than the second most significant predictor (stratifi- Drivers explaining Chl a at the continental scale cation strength, 18%), that was closely followed by light cli- The applied models significantly explain a proportion of the mate (15%; Table S3). variability in Chl a (p ≤ 0.001; Table 2), with the model applied The 230 lakes dataset allows us to carry out the same analy- to Mediterranean lakes explaining the highest variability sis separately on each group of lakes corresponding to the (R2 = 45%), closely followed by the model applied to Boreal explanatory categories, climatic zone, and depth type, to gain lakes (R2 = 43%) with Continental lakes further behind more insights on the summer drivers of phytoplankton bio- (R2 = 25%), while the model applied to deep lakes explained mass for this set of lakes. Fig 5. First two significant predictors for Chl a in lake group model 3 (shallow) and model 4 (deep). (a, b) Light climate, Zeu/Zmix, and TN explain respectively 46% and 33% of Chl a variance in EMLS shallow lakes (model R2 = 31%). (c, d) Light climate, Zeu/Zmix, and its interaction with stratification strength, N2*Zeu/Zmix, explain respectively 32% and 29% of Chl a variance in EMLS deep lakes (model R2 = 12%). All variables are plotted as to the statis- tical models, that is, natural logarithm (ln). See Table S4 for relative contribution and significance of all predictors. 4323 Donis et al. European lake survey: summer Chl-a drivers Fig 6. First two significant predictors for Chl a in lake group model 5 (Mediterranean), model 6 (Continental), and model 7 (Boreal). (a, b) Stratification strength, N2, and surface temperature, SurfT, explain respectively 46% and 24% of Chl a variance in EMLS Med lakes (model R2 = 45%). (c, d) TN and N2 explain respectively 35% and 29% of Chl a variance in EMLS Cont lakes (model R2 = 25%). (e, f) Light climate, interaction with stratification strength, N2*Zeu/Zmix, and SurfT explain respectively 34% and 21% of Chl a variance in EMLS Boreal lakes (model R2 = 43%). All variables are plotted as to the sta- tistical models, that is, natural logarithm (ln). See Table S5 for relative contribution and significance of all predictors. Shallow vs. deep lakes for shallow lakes, TN played a more significant role than strati- Light climate was the most important variable explaining fication (33%; Fig. 5b) while not appearing as a significant pre- Chl a variance in both shallow and deep lake subsets (46% dictor of algal biomass in the deep lakes subset. and 32%, respectively; Fig. 5a,c). Stratification strength was also a significant contributor for both lake types, either indi- Mediterranean vs. Continental vs. Boreal lakes vidually (14%, shallow lakes; Table S4), or in synergistic inter- When applying the model to the different climatic zones, action with light climate (29%, deep lakes; Fig. 5d). However, the strength of the stratification appeared as a strong 4324 Donis et al. European lake survey: summer Chl-a drivers Fig 7. Relationships between N2 and (a) surface temperature, 2nd-order polynomial fit, R2 = 0.14; (b) euphotic depth, 2nd-order polynomial fit, R2 = 0.06; (c) surface area, 3rd-order polynomial fit, R2 = 0.08; (d) maximum depth, 4th-order polynomial fit, R2 = 0.2. All polynomial fit are signifi- cantly better than function y = constant at the 0.05 level. predictor, either individually (Med. 46% and Cont. lakes 29%; (R2 = 0.08) than between maximum N2 and maximum lake Fig. 6a,d) or in interaction with light climate (Boreal lakes depth (R2 = 0.2). Here, the 4th-order polynomial fit followed 34%; Fig. 6e). In Mediterranean and Boreal lakes, surface tem- the effect of temperature on N2 for increasing lake depths, perature was also a strong predictor of algal biomass (24% and reaching a plateau for lakes deeper than 20 m (Fig. 7d). 21%, respectively; Fig. 6b,f) but not for Continental lakes (Table S5). Instead, nutrients were the most significant predic- tors of Chl a (34%) for Continental (Fig. 6c), while being less important for Boreal lakes (14%) and not important for Medi- terranean lakes (Table S5). Relationship between stratification metrics Within the EMLS lakes, we analyzed the relationship between stratification strength and some of the drivers, that is, temperature, light penetration, and lake morphology. As already shown, Mediterranean lakes, while being on average the warmest, did not have the highest average stratification strength (Fig. 3d). When looking at the entire dataset (Fig. 7), the polynomial fit between the maximum N2 and surface tem- perature was significant (p < 0.001) but weak (R2 = 0.14; Fig. 7a), indicating that only for a relatively small number of the EMLS lakes, higher surface water temperatures at the sam- pling time corresponded to a stronger stratification. An even weaker relationship (R2 = 0.06) was observed between stratifica- Fig 8. Eight-day average temperature anomaly at the sampling site and sampling period in relation to the lake stratification strength for the EMLS tion strength and light penetration depth (Zeu; Fig. 7b). As for climatic zone subsets, Continental (gray), Mediterranean (light pink), and the morphological features, the relationship observed between Boreal (blue). All best fits are given by exponential curves, with the one maximum N2 and surface area (Fig. 7c), was much weaker for Boreal lakes being the most significant (R2 = 0.4). 4325 Donis et al. European lake survey: summer Chl-a drivers Air temperature anomaly properties of a lake, such as stratification strength and light The summer of 2015 was the third warmest summer (after climate (expressed as the ratio of euphotic to mixing depth), 2003 and 2010) since 1880 in Europe (GISTEMP, NOAA are the strongest ecosystem drivers for phytoplankton biomass online data). During the sampling period in 2015, 70% of for this set of mostly nutrient-rich lakes, at the sampling time. EMLS lakes experienced a positive temperature anomaly of It is possible, however, that a different result would be 1.9 C 3.4 C (average 1 SD, based on each lake 8-d temper- obtained from the same dataset in a different time of the year. ature average compared to 10-yr average for the same 8 d). In a similar fashion to the present work, an earlier study on However, when looking at each climatic zone separately, 96% 1076 US lakes (Rigosi et al. 2014), showed that surface temper- of Continental lakes and 87% of the Boreal lakes experienced ature, nutrients, and their interaction were the main phyto- a positive temperature anomaly of 3.8 C 2.6 C and plankton biomass predictors. Interestingly, their results 3.7 C 2.9 C, respectively. In contrast, only 53% of the Med- showed that the largest part of the variance in Chl a for the iterranean lakes experienced a positive temperature anomaly, subset of eutrophic and hypereutrophic lakes was explained of 1.4 C 1 C. The remaining 30% of the total, 4% of Conti- by a synergistic interaction between nutrients and tempera- nental, 13% of Boreal and 47% of Mediterranean are lakes that ture. Our study moves a step forward and highlights the fact showed a negative deviation from the long term average. that additional variables need to be considered when collect- Hence, at the time of sampling, the great majority of Conti- ing lake “snapshots” at a continental scale. The analysis pres- nental and Boreal lakes experienced a strong temperature ented here indicates that nutrients, temperature, and light increase compared to the long-term average levels, which was should not be the only algal growth conditions to be consid- not the case for Mediterranean lakes. Compared to the other ered. We show that when lake stratification metrics are regions, Boreal lakes as well showed the strongest exponential included, we can gain insights into the lake physics mecha- relationship between the experienced temperature anomaly nisms that promote phytoplankton biomass growth and and stratification strength (Fig. 8). potentially improve the development of predictive tools. Moreover, our statistical analysis indicates that surface tem- Pigments analysis perature alone should not be used as a proxy for stratification Measurement of Zea concentrations in the EMLS lakes were strength. Indeed for a multilake survey, it is necessary to esti- carried out with the aim of investigating cyanobacterial bio- mate lake stability (N2) as a variable that comprises the lake mass. A strong linear relationship was found between Zea and thermal “history,” and therefore gives insight into the environ- Chl a (R2 = 0.6; Fig. S3) indicating that higher Chl mental conditions that the phytoplankton have experienced a concentrations systematically corresponded with higher during the recent past. Such information is easily attained with concentrations of Zea. a temperature profile and is extremely relevant when looking at ecosystem functioning, as thermal structure and light penetra- tion determine the physical constraints of the photosynthetic Discussion biomass distribution in the water column. These constraints Drivers explaining phytoplankton biomass at the also determine to what extent specific phytoplankton features continental scale adapted to life in a stable water column, such as the pigment Several studies have focused on the effects of nutrients and composition (e.g., presence of phycoerythrin in deep chloro- warming on phytoplankton in more than one lake (table S1 in phyll maxima), and buoyancy regulation (e.g., gas vesicles, Salmaso and Tolotti 2021). This is of particular concern for motility, shape adaptations) may favor specific algal groups. resolving the climate warming effect on lakes and the positive feedbacks on eutrophication of lakes (Sinha et al. 2017; Deng Shallow vs. deep lakes et al. 2018). However, thermal stratification, which will likely In the EMLS, most of the lakes were eutrophic which may increase with climate warming (Woolway and Merchant 2019), explain the predominant importance of light climate (Zeu/ is an important feature governing lake ecosystems as it affects Zmix) for algal biomass variance in both shallow and deep both nutrient availability and light climate (Schwefel lakes. We therefore assume that, for nutrient-rich lakes, phyto- et al. 2016), generating complex feedbacks for the biota plankton rather than inorganic suspended solids determine (Mesman et al. 2021). The importance of these factors may underwater light extinction (Scheffer et al. 1997), which sub- dominate when lakes are not nutrient limited. sequently determines phytoplankton biomass. We have applied a set of multiple regression models to Although light climate was the most important factor for 230 European lakes (54–85% of which were eutrophic both EMLS depth types, we observed a relatively greater depending on the criterion applied) to test the dependency of importance of light climate in shallow rather than deep lakes Chl a on phytoplankton growth resources (nutrients, tempera- (explaining 46% and 32% of the variation, respectively), ture, and light climate) and morphophysical lake properties which may be explained by the fact that shallow lakes exist in (surface area, stratification strength), including interactions two clearly distinct states, clear vs. turbid. Mechanisms between specific predictors. Our results indicate that physical directly linked to the underwater light climate, for example, 4326 Donis et al. European lake survey: summer Chl-a drivers high cyanobacterial biomass and benthivorous fish stirring up and mixing can reach deeper layers, it will take the algal com- the sediment, provide varying degrees of resilience to the tur- munities beyond the euphotic zone reducing algal growth bid state (Scheffer et al. 1997). In contrast, macrophytes stabi- (Zeu/Zmix < 1; Fig. 1b). A deeper mixed layer will allow light to lize the clear water state, and light penetration that reaches reach greater depths by diluting epilimnetic phytoplankton the sediment is vital for their development (Ibelings over a larger volume of lake water, thus increasing light pene- et al. 2007). With 72% of the shallow EMLS lakes having a tration. This extended euphotic depth will likely, however, Secchi depth of less than 0.8 m, we could argue that the not make up for light limitation due to a deeper mixing depth, majority are in a turbid state, be it stable or not. This may go so the ratio Zeu/Zmix would still decrease when water column some way to explain the critical role of light in determining stability decreases (Fig. 1b), exacerbating the light limitation biomass of algae in EMLS shallow lakes. of phytoplankton growth. TN is the second-most important predictor for Chl a in In contrast to the shallow lakes, in EMLS deep lakes neither shallow lakes (33%) which, together with the general absence TP nor TN appeared as a significant predictor of algal biomass, of TP as significant predictor for Chl a variance, suggests that possibly because of the higher likelihood of light limitation for the 230 EMLS lakes, the commonly found linear relation- mentioned above. Interestingly, another difference between ship between TP and Chl a does not hold true EMLS shallow and deep lakes was that the surface area (Vollenweider 1968). This is in line with previous studies on explained a significant 22% of the overall Chl a variance of nutrient-rich lakes suggesting that (1) a positive linear TP–Chl deep lakes, while did not explain the Chl a variance for the a relationship exists only at intermediate concentrations of TP shallow lakes (Table S4). This might be due to the fact that the (0.004–0.23 mg L1; Quinlan et al. 2020) and (2) nitrogen surface area becomes important considering its direct relation- becomes limiting for phytoplankton under high TP, especially ship with lake wind exposure, which can influence the water over shorter temporal scales (Filstrup et al. 2014). column mixing depth in deep lakes, hence the availability of A eutrophic status of a lake, however, does not mean that light and nutrients for phytoplankton (Fig. 1). Although wind nutrients cannot be limiting for dense phytoplankton, with a exposure was not included in this study, EMLS lake area corre- large demand to sustain a high biomass. Yet, the condition of lated with depth (Fig. S1a), and was therefore indirectly related nutrient limitation (in our case nitrogen) could be seen as an to the water column thermal structure. Indeed, EMLS lakes effect driven by Zeu/Zmix (first predictor). Especially for shal- with larger surface areas tended to be deeper (Fig. 4f) and more low lakes, when this ratio becomes smaller, the mixed layer stable (Fig. 4d), and this may have favored phytoplankton’s exceeds the euphotic zone and nutrients from the sediment access to light, in particular when nutrients are not—or less are likely to be resuspended. We may easily see a more direct of—a limiting factor, for example, when DCMs are formed relationship between Chl a and light climate than with the where phytoplankton has access to nutrients in the hypolim- nutrient abundance, because light climate, by revealing the nion (Leach et al. 2018). recent mixing history, is a more integrative indicator of nutri- ent availability than the nutrient content of a single water Mediterranean vs. Continental vs. Boreal lakes sample, especially for productive shallow lakes. When EMLS lakes were clustered by climatic zone, stratifi- For deep lakes, light climate and its synergistic interaction cation strength appeared as a strong predictor for Chl a, either with water column stability had a similarly important contri- individually (Mediterranean 46% and Continental lakes 29% bution to the overall R2, explaining Chl a variance (32% and variation explained) or in synergistic interaction with light cli- 29% for Zeu/Zmix and N2, respectively). High algal biomass mate (Boreal lakes 34%). Stratification strength was thus a increases turbidity, which can increase water temperature in dominant factor promoting phytoplankton optimal growth the surface layer through increased heat absorption (Ibelings conditions, interacting with the availability of nutrients and et al. 2003), and thus reinforce stratification (Paerl and light, as discussed. Light climate interaction with water col- Huisman 2008). Reinforced stratification through increased umn stratification was a strong factor for Boreal lakes phyto- turbidity implies that phytoplankton is maintained within the plankton growth (Fig. 6e) possibly due to their tendency to be euphotic zone offering a potential explanation of how light richer in humic substances and consequently darker (Kutser climate can interact synergistically with water column stability et al. 2005; Kelly et al. 2018). (Zeu/Zmix > 1; Fig. 1a). However, in a strongly stratified lake, Phytoplankton biomass in Continental lakes seemed to nutrients may remain available in the hypolimnion even exhibit a higher degree of nitrogen dependency (Fig. 6c); how- when they are depleted in the epilimnion, so that deeper ever, we cannot exclude that those lakes in other regions were mixing, also of short duration, enhances the likelihood that in a similar state, since as discussed above, predictors like light phytoplankton gains access to this pool of nutrients. In deep, climate can possibly encompass nutrient limitation. On the well-stratified lakes, it is also relatively common to find algal other hand, the comparatively lower Chl a content of Medi- biomass maxima (a.k.a. deep chlorophyll maximum [DCM]) terranean lakes (Fig. 3a) seems to indicate that, at the time of at the crossroads of light from above and nutrients from below sampling, these lakes were experiencing a better nutrient– (Leach et al. 2018). On the other hand, if stratification is weak phytoplankton balance than Continental lakes. 4327 Donis et al. European lake survey: summer Chl-a drivers The best predictor for algal biomass: Stratification strength temperature (Fig. 4b,d). The absence of a strong correlation or lake depth? between stratification strength and surface temperature is fur- Stratification strength decreased in importance when split- ther confirmed by the absence of any trend between stratifica- ting the dataset into depth types, which may indicate that tion strength and climatic zone (Fig. 4d). depth-type itself explained algal biomass variance. This is Moreover, the fact that shallow and Mediterranean lakes also suggested by the fact that all predictors were signifi- had the highest surface temperature, but the weakest stratifica- cantly different between deep and shallow lakes (Fig. 4), and tion confirms that surface temperature can be a misleading some important environmental factors have a different effect indicator for stratification strength, especially for snapshot on these two clusters. Wind has generally a larger effect on surveys as shown in previous studies on large datasets (Read temperature structure and stability of shallow lakes, because et al. 2014; Winslow et al. 2017). the wind-induced mixing allows heat to be transferred throughout the entire water column (Nõges et al. 2011). Fur- Light penetration depth thermore, shallow lakes respond more directly to short-term Changes in light absorption by the dissolved and weather variations (Arvola et al. 2009; Deng et al. 2018). For suspended content of a lake affect the vertical distribution of deep lakes that have a higher heat retention and potential heat and resulting stratification (Andrew et al. 2008; Rinke energy, greater wind speeds are required to drive mixing dur- et al. 2010). We did not observe, however, a distinct rela- ing the summer months, resulting in greater stability tionship between stratification strength and light penetra- (Boehrer and Schultze 2008). Fetch and dominant wind tion (Fig. 7b). The reason why this relationship is not direction and intensity are also important in determining stronger may be that the effect of light on stratification is stratification strength in deep lakes (Wetzel 2001), although more evident in time series than in spatial gradients. This is these data were not collected for this study. However, given because light-induced heat diffusion in the water column the consistently higher N2 observed for EMLS deep lakes and its temporal variability has a stronger effect on the dura- (Fig. 4d), we can assume that sufficiently strong and long- tion of the stratification than on its absolute value. Indeed, lasting winds were not present at each sampling site during— more transparent lakes (Secchi transparency > 5 m) tend to or shortly prior to—the sampling period to modify the afore- maintain a seasonal thermal stratification for a longer dura- mentioned scenario of deep lakes that are more strongly tion than more turbid ones (Richardson et al. 2017), there- stratified than shallow lakes. fore remaining stable longer. Assessing whether this is the Depth and N2 may therefore be confounding variables case is not possible with a summer snapshot sampling because, at least for this dataset, lake depth can explain most design, although it was observed that light penetration can of the variation in stratification trends. Nevertheless, whether drive the depth of the mixed layer. This is suggested for the lake maximum depth or stratification strength is actually the EMLS dataset by a moderate linear relationship (R2 = 0.35) most significant predictor of Chl a in the overall EMLS dataset, between the depth of the epilimnion, or mixed layer, and the message remains unchanged: lake morphophysical proper- the euphotic depth (Fig. S2). ties are essential when investigating phytoplankton biomass responses to environmental changes. Maximum depth and surface area We observed a relationship between N2 and both the lake Relationship between stratification metrics maximum depth and lake surface area (Fig. 7c,d). The shape of Given the importance of stratification strength as a predic- the polynomial fit shows that the linearity of stratification tor of Chl a variance at the European continental scale, we strength with lake maximum depth holds until lake depths analyzed the relationship between this variable, represented of 20 m, because of the physical limit dictated by the ther- by maximum N2, and the environmental and morphological mal diffusivity of water. This relationship may confirm that characteristics that act on the density gradient of a lake (sur- N2 and depth are interdependent in determining resource face temperature, light penetration depth, maximum depth, availability for the algal communities. and surface area). EMLS lakes with a greater area were on average deeper and had a more stable water column (Fig. 4d,f). Therefore, surface Surface temperature area of the EMLS dataset was directly correlated with maxi- Stratification strength responds directly to changes in water mum depth and was used as the only morphological variable temperature, yet each lake will need a certain number of warm in the statistical models. However, the relationship between days with relatively low wind speed to develop stratification, surface area and N2 was not strong (Fig. 7c), possibly because a which also depends on lake morphological factors. The reason larger surface area does not necessarily mean a greater wind for the weak correlation observed between maximum N2 and exposure, which is largely determined by the lake’s orienta- surface temperature (Fig. 7a) may be that deep and shallow tion toward the dominant wind direction and lake topogra- lakes are equally represented, and while deep lakes are more phy. Clearly, as for the underwater light regime, analyzing the strongly stratified, the shallow lakes had the highest surface effect of wind exposure on the thermal structure is not 4328 Donis et al. European lake survey: summer Chl-a drivers possible with a single observation, but would require a water lakes, as they are for overall phytoplankton. Moreover, cyano- column temperature time series. bacteria have evolved specific traits like buoyancy and acces- sory pigments that renders them specifically well adapted to Temperature anomaly: 2015, an unusually hot summer stably stratified conditions (Huisman et al. 2018). Conse- As expected, Mediterranean lakes had higher surface tem- quently, the fact that light climate was the main driver for peratures than Boreal ones (Fig. 3a,b). However, Boreal lakes both lake depth types (Fig. 4) may confirm that at high nutri- exhibited a significantly higher Chl a concentration than ent levels, light becomes limiting for cyanobacterial develop- Mediterranean lakes, while lakes in both climatic zones had ment (Ganf and Oliver 1982; Bouterfas et al. 2002; Huisman comparable nutrient concentrations (Table S2). This seems to et al. 2004). confirm the importance of factors other than temperature Considering the high temperature anomaly experienced in (lower in Boreal lakes) or nutrients (similar) driving phyto- Boreal regions at the sampling time, the significance of a posi- plankton biomass, especially water column stability in relation tive interaction between water column stability and light cli- to the light climate. Indeed, while Boreal lakes are known to mate in promoting cyanobacteria in the Boreal lakes during a stratify intermittently during summer (Kirillin and Shat- record hot summer supports the general observation that well 2016; Woolway and Merchant 2019), the heat wave likely “blooms like it hot” (Paerl and Huisman 2008). In the context intensified the stratification strength in the Boreal lakes more of climate change—and rapid warming at high latitudes— strongly than in other regions, given that the region experi- perhaps a more appropriate rephrasing is “blooms like it enced the largest temperature anomaly (Fig. 8). This may have warmer than usual,” since the Boreal lakes were still cooler favored the conditions shown in Fig. 1a and supported by our than the Mediterranean lakes, yet the temperature anomaly model results, that is, the interaction between light climate was higher as were Zea levels. Among the EMLS subset of lakes and stratification strength is the main Chl a driver for Boreal with detectable Zea (which are 172 over the total 230 of this lakes (Fig. 6e). study), almost all (95%) of the Boreal lakes experienced a As several studies addressed the relationship between light, higher positive T-anomaly ( 4 C 2.5 C). Evidently, more nutrients, and temperature effects on primary producers in detailed integrated lab-field studies, including both ecological Boreal regions (Zwart et al. 2016; Bergstrom and Karlsson 2019), and evolutionary aspects, are needed to resolve this issue. alternative explanations may apply too. Although it is not pos- sible to generalize, such observations are crucial to generate Future scenario and management strategies ideas and stimulate future research. It is possible that a higher Among the Chl a predictors of this study, lake surface tem- abundance of mixotrophs in Boreal lakes may help to explain perature and water column stratification are expected to have the higher Chl a in that region, since Hansson et al. (2019) the strongest impact on lake ecosystems in a warming future demonstrated that the success of mixotrophs is correlated with (O’Reilly et al. 2015; Kraemer et al. 2017). Both variables were the elevated colored dissolved organic matter content of Boreal significant drivers for trends in phytoplankton biomass across lakes. It is also interesting to note that Mantzouki et al. (2018) climatic gradients in Europe. Thus, since lake water column found that for the same EMLS dataset, the variety of toxins pro- stability will likely increase with a warming climate (Oleksy duced by cyanobacteria increased with latitude, which possibly and Richardson 2021), bloom-forming cyanobacteria in partic- may have reduced the grazing pressure in Boreal lakes, contrib- ular will be further promoted given their typical dependence uting to higher Chl a. on buoyancy that makes them particularly well adapted to a stable water column (Steinberg and Hartmann 1988; Paerl and Cyanobacteria like it warmer? Paul 2012). Zea is frequently used as a pigment to indicate cyanobacterial Although we concur with Ibelings et al. (2016) that any biomass (Bianchi et al. 2000; Glibert et al. 2004; Przytulska sustainable approach controlling cyanobacterial blooms has to et al. 2017; Ewing et al. 2020), although it is found both in cya- be rooted in nutrient reduction, our present analysis under- nobacteria and in chlorophytes (Deshpande et al. 2014; Ibelings lines the potential effectiveness of additional measures that et al. 2016). In this study, we do not provide microscopy results weaken the future strengthening of lake stratification, which to confirm the correspondence between cyanobacteria and Zea; is demonstrated here to play such a critical role in determin- therefore, the following discussion is presented with a degree of ing differences in lake phytoplankton and cyanobacterial bio- caution, and mainly serves to stimulate further ideas, eventually mass. It may be essential, for instance, to include measures contributing to a deeper understanding of the worldwide like artificial lake mixing (Visser et al., 2016) to mitigate algal increase in cyanobacterial blooms. (and cyanobacterial) blooms. The strong correlation between Zea and Chl a EMLS (Fig. S3) indicated that higher Chl a concentrations systemati- cally corresponded with higher concentrations of Zea, which Conclusions may suggest that water column stratification and light climate Nutrients and light are the fundamental resources for phy- are the main drivers for cyanobacterial growth in eutrophic toplanktonic biomass, even in nutrient rich lakes, such as the 4329 Donis et al. European lake survey: summer Chl-a drivers ones represented in this study; however, results from the cyanobacterial biomass in a 1147 lakes data set. Limnol. EMLS analysis show that Chl a variance is better predicted by Oceanogr. 58: 1736–1746. light climate and stability metrics. These predictors are also Bergstrom, A. K., and J. Karlsson. 2019. Light and nutrient strong indicators of the epilimnetic nutrient load and of the control phytoplankton biomass responses to global light experienced by the algal biomass prior to sampling. This change in northern lakes. Glob. Chang. Biol. 25: 2021– explains why in this nutrient-rich lake dataset, light climate 2029. was the most important variable explaining Chl a variance in Bianchi, T. S., E. Engelhaupt, P. Westman, T. Andren, C. Rolff, both shallow and deep lakes, with the difference that, only for and R. Elmgren. 2000. Cyanobacterial blooms in the Baltic deep lakes the optimum condition for photosynthetic biomass Sea: Natural or human-induced? Limnol. Oceanogr. 45: was obtained when stratification operated in a synergistic 716–726. interaction with light climate. The dominance of light climate Boehrer, B., and M. Schultze. 2008. Stratification of lakes. Rev. and the absence of TP as significant predictor for Chl Geophys. 46: 1–27. a variance confirms that: (1) when TP levels are high as in the Bouterfas, R., M. Belkoura, and A. Dauta. 2002. Light and tem- average EMLS, light and nitrogen become limiting resources perature effects on the growth rate of three freshwater algae for phytoplankton and (2) light climate, as metric for the isolated from a eutrophic lake. Hydrobiologia 489: 207–217. recent history of water column mixing, is a powerful indicator Brentrup, J. A., and others. 2018. The potential of high- for nutrient availability, and needs to be included in similar frequency profiling to assess vertical and seasonal patterns studies. of phytoplankton dynamics in lakes: An extension of the Furthermore, our analysis of this pan-European dataset Plankton Ecology Group (PEG) model. Inland Waters 6: shows that shallow and Mediterranean lakes exhibit the 565–580. highest surface temperature, although the weakest stratifica- Camacho, A. 2006. On the occurrence and ecological features tion, confirming that lake surface temperature does not neces- of deep chlorophyll maxima (DCM) in Spanish stratified sarily correlate with lake stratification strength. Consequently, lakes. Limnetica 25: 453–478. especially for snapshot surveys, a lake temperature profile Carvalho, F., J. Kohut, M. J. Oliver, R. M. Sherrell, and O. should be preferred over surface temperature data as it is a Schofield. 2016. Mixing and phytoplankton dynamics in a more sensible indicator for stratification strength and ecologi- submarine canyon in the West Antarctic Peninsula. cal response to warming. J. Geophys. Res. Oceans 121: 5069–5083. Finally, among the 230 European lakes, we observed a signif- Cloern, J. E. 1996. Phytoplankton bloom dynamics in coastal icant exponential correlation between temperature anomaly ecosystems: A review with some general lessons from and the stratification strength only for Boreal lakes that, inci- sustained investigation of San Francisco Bay, California. dentally, had the highest Chl a concentrations; a notion deserv- Rev. Geophys. 34: 127–168. ing further attention in light of most rapid increases in Deng, J., and others. 2018. Climatically-modulated decline in warming taking place at high latitude regions. This, coupled wind speed may strongly affect eutrophication in shallow with the fact that, for mildly to hyper-eutrophic lakes, light cli- lakes. Sci. Total Environ. 645: 1361–1370. mate and water column stratification are the most important Deshpande, B. N., R. Tremblay, R. Pienitz, and W. F. Vincent. drivers determining phytoplankton biomass, may serve to bet- 2014. Sedimentary pigments as indicators of cyanobacterial ter plan and implement lake management and mitigation dynamics in a hypereutrophic lake. J. Paleolimnol. 52: strategies. 171–184. Ewing, H. A., and others. 2020. “New” cyanobacterial blooms are not new: Two centuries of lake production are related to References ice cover and land use. Ecosphere 11(6): e03173. Andrew, J. T., and others. 2008. Cooling lakes while the world Filstrup, C. T., and others. 2014. Regional variability among warms: Effects of forest regrowth and increased dissolved nonlinear chlorophyll-phosphorus relationships in lakes. organic matter on the thermal regime of a temperate, urban Limnol. Oceanogr. 59: 1691–1703. lake. Limnol. Oceanogr. 53: 404–410. Flaim, G., E. Eccel, A. Zeileis, G. Toller, L. Cerasino, and U. Arvola, L., and others. 2009, 2010, p. 85–101. In D. G. George Obertegger. 2016. Effects of re-oligotrophication and cli- [ed.], The impact of the changing climate on the thermal mate change on lake thermal structure. Freshw. Biol. 61: characteristics of lakes. Springer Science+Business Media 1802–1814. B.V. Fox, J., and S. Weisberg. 2019. An R Companion to Applied Bartosiewicz, M., I. Laurion, and S. MacIntyre. 2015. Green- Regression, Third edition. Sage, Thousand Oaks CA. house gas emission and storage in a small shallow lake. Ganf, G. G., and R. L. Oliver. 1982. Vertical separation of light Hydrobiologia 757: 101–115. and available nutrients as a factor causing replacement of Beaulieu, M., F. Pick, and I. Gregory-Eaves. 2013. Nutrients green-algae by blue-green-algae in the plankton of a strati- and water temperature are significant predictors of fied lake. J. Ecol. 70: 829–844. 4330 Donis et al. European lake survey: summer Chl-a drivers Glibert, P. M., and others. 2004. Evidence for dissolved Koenings, J. P., and J. A. Edmundson. 1991. Secchi disk and organic nitrogen and phosphorus uptake during a photometer estimates of light regimes in Alaskan lakes - cyanobacterial bloom in Florida Bay. Mar. Ecol. Prog. Ser. effects of yellow color and turbidity. Limnol. Oceanogr. 36: 280: 73–83. 91–105. Graff, J. R., and M. J. Behrenfeld. 2018. Photoacclimation Kosten, S., and others. 2012. Warmer climates boost responses in subarctic Atlantic phytoplankton following a cyanobacterial dominance in shallow lakes. Glob. Chang. natural mixing-restratification event. Front. Mar. Sci. 5: Biol. 18: 118–126. 209. doi:10.3389/fmars.2018.00209 Kraemer, B. M., T. Mehner, and R. Adrian. 2017. Reconciling Gro} mping, U. 2006. Relative Importance for Linear Regression the opposing effects of warming on phytoplankton bio- in R: The Package relaimpo. Journal of Statistical Software, mass in 188 large lakes. Sci. Rep. 7: 1–7. Foundation for Open Access Statistics vol. 17(i01). Köppen, W. 1900. Versuch einer klassifikation der klimate, Hansson, T. H., H.-P. Grossart, P. A. del Giorgio, N. F. St- vorzugsweise nach ihren beziehungen zur pflanzenwelt. Gelais, and B. E. Beisner. 2019. Environmental drivers of Geographische Zeitschrif 6: 657–679. mixotrophs in boreal lakes. Limnol. Oceanogr. 64: 1688– Kutser, T., D. C. Pierson, L. Tranvik, A. Reinart, S. Sobek, and 1705. K. Kallio. 2005. Using satellite remote sensing to estimate Heiskanen, J. J., and others. 2014. Effects of cooling and inter- the colored dissolved organic matter absorption coefficient nal wave motions on gas transfer coefficients in a boreal in lakes. Ecosystems 8: 709–720. lake. Tellus B 66: 1–16. Leach, T. H., and others. 2018. Patterns and drivers of deep Huisman, J., G. A. Codd, H. W. Paerl, B. W. Ibelings, J. M. H. chlorophyll maxima structure in 100 lakes: The relative Verspagen, and P. M. Visser. 2018. Cyanobacterial blooms. importance of light and thermal stratification. Limnol. Nat. Rev. Microbiol. 16: 471–483. doi:10.1038/s41579-018- Oceanogr. 63: 628–646. 0040-1 Magee, M. R., and C. H. Wu. 2017. Response of water tempera- Huisman, J., and others. 2004. Changes in turbulent mixing tures and stratification to changing climate in three lakes shift competition for light between phytoplankton species. with different morphometry. Hydrol. Earth Syst. Sci. 21: Ecology 85: 2960–2970. 6253–6274. Ibelings, B. W., M. Bormans, J. Fastner, and P. M. Visser. 2016. Mantzouki, E., and B. W. Ibelings. 2018. The principle and CYANOCOST special issue on cyanobacterial blooms: value of the European Multi Lake Survey. Limnol. Ocean- Synopsis—a critical review of the management options for ogr. Bull. 27: 82–86. their prevention, control and mitigation. Aquat. Ecol. 50: Mantzouki, E., and others. 2018. A European Multi Lake Sur- 595–605. vey dataset of environmental variables, phytoplankton pig- Ibelings, B. W., B. M. A. Kroon, and L. R. Mur. 1994. Acclima- ments and cyanotoxins. Sci. Data 5: 180226. tion of photosystem II in a cyanobacterium and a eukary- Mesman, J. P., and others. 2021. The role of internal feedbacks otic green alga to high and fluctuating photosynthetic in shifting deep lake mixing regimes under a warming cli- photon flux densities, simulating light regimes induced by mate. Freshw. Biol. 66: 1021–1035. mixing in lakes. New Phytol. 128: 407–424. Millero, F. J., and A. Poisson. 1981. International one- Ibelings, B. W., M. Vonk, H. F. J. Los, D. T. van der Molen, atmosphere equation of state of seawater. Deep-Sea Res. A and W. M. Mooij. 2003. Fuzzy modeling of cyanobacterial Oceanogr. Res. Pap. 28: 625–629. surface waterblooms: Validation with NOAA-AVHRR satel- Moss, B., and others. 2011. Allied attack: Climate change and lite images. Ecol. Appl. 13: 1456–1472. eutrophication. Inland Waters 1: 101–105. Ibelings, B. W., and others. 2007. Resilience of alternative stable Mur, L. R., H. J. Gons, and L. Van Liere. 1977. Some experi- states during the recovery of shallow lakes from eutrophica- ments on the competition between green algae and blue tion: Lake Veluwe as a case study. Ecosystems 10: 4–16. green bacteria in light-limited environments. FEMS Imberger, J. 1985. The diurnal mixed layer. Limnol. Oceanogr. Micobiol. Lett. 1: 335–338. 30: 737–770. Nõges, P., T. Nõges, M. Ghiani, B. Paracchini, J. Pinto Grande, Imboden, D. M., and A. Wüest. 1995. Mixing mechanisms in and F. Sena. 2011. Morphometry and trophic state modify lakes. In Physics and chemistry of lakes. Springer-Verlag. the thermal response of lakes to meteorological forcing. Kelly, P. T., C. T. Solomon, J. A. Zwart, and S. E. Jones. 2018. A Hydrobiologia 667: 241–254. framework for understanding variation in pelagic gross pri- Nürnberg, G. K. 1984. The prediction of internal phosphorus mary production of lake ecosystems. Ecosystems 21: 1364– load in lakes with anoxic hypolimnial. Limnol. Oceanogr. 1376. 29: 111–124. Kirillin, G., and T. Shatwell. 2016. Generalized scaling of sea- Oleksy, I. A., and D. C. Richardson. 2021. Climate change and sonal thermal stratification in lakes. Earth Sci. Rev. 161: teleconnections amplify lake stratification with differential 179–190. local controls of surface water warming and deep water 4331 Donis et al. European lake survey: summer Chl-a drivers cooling. Geophys. Res. Lett. 48: e2020GL090959. doi:10. Köppen-Geiger climate zones 1800–2100. Meteorol. Z. 26: 1029/2020GL090959 115–125. O’Reilly, C. M., and others. 2015. Rapid and highly variable Rubel, F., and M. Kottek. 2010. Observed and projected cli- warming of lake surface waters around the globe. Geophys. mate shifts 1901-2100 depicted by world maps of the Res. Lett. 42: 10773–10781. Köppen-Geiger climate classification. Meteorol. Z. 19: Paerl, H. W., and J. Huisman. 2008. Climate. Blooms like it 135–141. hot. Science 320: 57–58. Sahoo, G. B., A. L. Forrest, S. G. Schladow, J. E. Reuter, R. Paerl, H. W., and V. J. Paul. 2012. Climate change: Links to Coats, and M. Dettinger. 2016. Climate change impacts on global expansion of harmful cyanobacteria. Water Res. 46: lake thermal dynamics and ecosystem vulnerabilities. 1349–1363. Limnol. Oceanogr. 61: 496–507. Peters, G. J. Y. 2018. Userfriendlyscience: Quantitative analysis Salmaso, N. 2002. Ecological patterns of phytoplankton made accessible (R package version 0.7-1071). doi:10. assemblages in Lake Garda: Seasonal, spatial and historical 17605/osf.io/txequ features. J. Limnol. 61: 95–115. Pinckney, J. L., T. L. Richardson, D. F. Millie, and H. W. Paerl. Salmaso, N., and M. Tolotti. 2021. Phytoplankton and anthro- 2001. Application of photopigment biomarkers for quanti- pogenic changes in pelagic environments. Hydrobiologia fying microalgal community composition and in situ 848: 251–284. growth rates. Org. Geochem. 32: 585–595. Salmaso, N., and others. 2020. Responses to local and global Pollard, A. I., S. E. Hampton, and D. M. Leech. 2018. The stressors in the large southern perialpine lakes: Present sta- promise and potential of continental-scale limnology using tus and challenges for research and management. J. Great the U.S. Environmental Protection Agency’s National Lakes Lakes Res. 46: 752–766. Assessment. Limnol. Oceanogr. Bull. 27: 36–41. Scheffer, M., S. Carpenter, J. A. Foley, C. Folke, and B. Walker. Posch, T., O. Köster, M. M. Salcher, and J. Pernthaler. 2012. 2001. Catastrophic shifts in ecosystems. Nature 413: Harmful filamentous cyanobacteria favoured by reduced 591–596. water turnover with lake warming. Nat. Clim. Change 2: Scheffer, M., S. Rinaldi, A. Gragnani, L. R. Mur, and E. H. van 809–813. Nes. 1997. On the dominance of filamentous cyanobacteria Przytulska, A., M. Bartosiewicz, and W. F. Vincent. 2017. in shallow, turbid lakes. Ecology 78: 272–282. Increased risk of cyanobacterial blooms in northern high- Schwefel, R., A. Gaudard, A. Wüest, and D. Bouffard. 2016. latitude lakes through climate warming and phosphorus Effects of climate change on deepwater oxygen and winter enrichment. Freshw. Biol. 62: 1986–1996. mixing in a deep lake (Lake Geneva): Comparing observa- Ptacnik, R., S. Diehl, and S. Berger. 2003. Performance of sink- tional findings and modeling. Water Resour. Res. 52: 8811– ing and nonsinking phytoplankton taxa in a gradient of 8826. mixing depths. Limnol. Oceanogr. 48: 1903–1912. Sinha, E., A. Michalak, and V. Balaji. 2017. Eutrophication will Quinlan, R., and others. 2020. Relationships of total phospho- increase during the 21st century as a result of precipitation rus and chlorophyll in lakes worldwide. Limnol. Oceanogr. changes. Science 357: 405–408. 66: 392–404. Steinberg, C. E. W., and H. M. Hartmann. 1988. Planktonic Read, J. S., and others. 2014. Simulating 2368 temperate lakes bloom-forming cyanobacteria and the eutrophication of reveals weak coherence in stratification phenology. Ecol. lakes and rivers. Freshw. Biol. 20: 279–287. Model. 291: 142–150. Sverdrup, H. U. 1953. On conditions for the vernal blooming Reynolds, C. S. 2006. The ecology of phytoplankton. of phytoplankton. ICES J. Mar. Sci. 18: 287–295. Cambridge Univ. Press. Tamm, M., R. Freiberg, I. Tonno, P. Noges, and T. Noges. Richardson, D. C., and others. 2017. Transparency, geomor- 2015. Pigment-based chemotaxonomy—a quick alternative phology and mixing regime explain variability in trends in to determine algal assemblages in large shallow eutrophic lake temperature and stratification across northeastern lake? PLoS One 10: e0122526. North America (1975-2014). Water 9: 442. Thompson, R. K. C., and R. Schmidt. 2005. Ultra-sensitive Rigosi, A., C. C. Carey, B. W. Ibelings, and J. D. Brookes. 2014. alpine lakes and climate change. J. Limnol. 64: 139–152. The interaction between climate warming and eutrophica- Van der Staay, G. W. M., A. Brouwer, R. L. Baard, and F. tion to promote cyanobacteria is dependent on trophic state van Mourik. 1992. Separation of photosystems I and II and varies among taxa. Limnol. Oceanogr. 59: 99–114. from the oxychlorobacterium (prochlorophyte) Rinke, K., P. Yeates, and K. O. Rothhaupt. 2010. A simulation Prochlorothrix hollandica and association of Chlb binding study of the feedback of phytoplankton on thermal struc- antennae with PS II. Biochim Biochim. Biophys. Acta ture via light extinction. Freshw. Biol. 55: 1674–1693. 1102: 220–228. Rubel, F., K. Brugger, K. Haslinger, and I. Auer. 2017. The cli- Visser, P. M., B. W. Ibelings, M. Bormans, and J. Huisman. mate of the European Alps: Shift of very high resolution 2016. Artificial mixing to control cyanobacterial blooms: A 4332 Donis et al. European lake survey: summer Chl-a drivers review. Aquatic Ecology 50: 423–441. doi:10.1007/s10452- Acknowledgments 015-9537-0 The authors acknowledge COST Action ES 1105 “CYANOCOST – Vollenweider, R. A. 1968. Scientific fundamentals of the eutro- Cyanobacterial blooms and toxins in water resources: Occurrence impacts phication of lakes and flowing waters with particular refer- and management” and COST Action Global Change Biology ES 1201 NETLAKE – Networking Lake Observatories in Europe” for contributing to ence to nitrogen and phosphorus as factors in this study through networking and knowledge sharing with European eutrophication. OECD. Paris. Tech. Report DA5/SCII/68 27, experts in the field. We acknowledge the members of the Global Lake 250 p. Ecological Observatory Network (GLEON) for their collaborative spirit and Wetzel, R. G. 2001. Water movements. Limnology (Third Edi- enthusiasm that inspired the grassroots effort of the EMLS. E.M. was tion). Academic Press; 93–128. doi:10.1016/B978-0-08- supported by a grant from the Swiss State Secretariat for Education, Research and Innovation to Bas Ibelings and by supplementary funding 057439-4.50011-3 from University of Geneva. We thank Wendy Beekman for the nutrient Winslow, L. A., J. S. Read, G. J. A. Hansen, K. C. Rose, and analysis. We thank Pieter Slot for assisting with the pigment analysis. We D. M. Robertson. 2017. Seasonality of change: Summer thank Dr. Ian Jones for valuable feedback on an earlier version of the man- warming rates do not fully represent effects of climate uscript. We thank the Leibniz Institute of Freshwater Ecology and the Aquatic Microbial Ecology Group for logistic and technical support of change on lake temperatures. Limnol. Oceanogr. 62: 2168– J. Fonvielle and H.-P. Grossart, and the Leibniz Association for financial 2178. support. H.P. was supported by the US National Science Foundation Woolway, R. I., and C. J. Merchant. 2019. Worldwide alter- (1840715, 1831096). A.C.’s work was funded by the Spanish Agencia ation of lake mixing regimes in response to climate change. Estatal de Investigacion and EU funds through the project CLIMAWET Nat. Geosci. 12: 271–276. (CGL2015-69557-R). The collection of data for Lough Erne and Lough Wüest, A., and A. Lorke. 2003. Small-Scalehydrodynamics Neagh were funded by the Department of Agriculture, Environment and Rural Affairs, Northern Ireland. We are grateful to Kristiina Vuorio from the Inlakes. Annu. Rev. Fluid Mech. 35: 373–412. Freshwater Centre of the Finnish Environment institute for her help in Yang, Y., W. Colom, D. Pierson, and K. Pettersson. 2016. organizing, collecting and analysing samples by the University of Jyväskylä Water column stability and summer phytoplankton and to Gerald Dörflinger from the Water Development Department of dynamics in a temperate lake (Lake Erken, Sweden). Inland Cyprus for his assistance with the sampling in Cyprus and for granting the Waters 6: 499–508. CUT team permission to use WDD’s equipment. Finally, we would like to thank the numerous other assistants that helped realizing each local Yankova, Y., S. Neuenschwander, O. Koster, and T. Posch. survey. Open access funding provided by Universite de Geneve. 2017. Abrupt stop of deep water turnover with lake warming: Drastic consequences for algal primary producers. Sci. Rep. 7: 13770. Conflict of Interest Zuur, A. F., E. N. Ieno, and C. S. Elphick. 2010. A protocol for None declared. data exploration to avoid common statistical problems. Methods in Ecology and Evolution, 1: 3–14. Submitted 24 September 2020 Zwart, J. A., and others. 2016. Metabolic and physiochemical Revised 01 July 2021 responses to a whole-lake experimental increase in dis- Accepted 08 October 2021 solved organic carbon in a north-temperate lake. Limnol. Oceanogr. 61: 723–734. Associate editor: Catherine M. O’Reilly 4333