Thesis

Description and modeling of phytoplankton with an emphasis of cyanobacteria in deep peri-Alpine lakes under warmer climatic conditions

GALLINA, Nicole

Abstract

The aim was to investigate the impacts of climate change on the behavior of phytoplankton, with a special emphasis on harmful cyanobacteria in the peri-Alpine region. It was hypothesized that in this highly sensitive region more important episodes of harmful cyanobacteria outbreaks under warmer climatic conditions could lead to negative impacts on water quality. The objectives were i) to analyze if air temperature is able to influence the cyanobacteria community, ii) to define the main drivers for the phytoplankton/cyanobacteria community, iii) to predict the cyanobacteria biomass under warmer climatic. Multivariate analysis and statistical modeling were applied on data from seven peri-Alpine lakes. Air temperature significantly influences the phytoplankton community, which is mainly driven by nutrients and temperature. Oscillatoriales, may be favored with increased water temperature and a longer stratification period. Planktothrix rubescens biomass was projected to increase in abundance and frequency thus potentially could induce a community function change.

Reference

GALLINA, Nicole. Description and modeling of phytoplankton with an emphasis of cyanobacteria in deep peri-Alpine lakes under warmer climatic conditions. Thèse de doctorat : Univ. Genève, 2012, no. Sc. 4467

URN : urn:nbn:ch:unige-258603 DOI : 10.13097/archive-ouverte/unige:25860

Available at: http://archive-ouverte.unige.ch/unige:25860

Disclaimer: layout of this document may differ from the published version.

1 / 1 UNIVERSITÉ DE GENÈVE FACULTÉ DES SCIENCES Institut des sciences de l’environnement Professeur Martin Beniston

Description and modeling of phytoplankton with an emphasis of cyanobacteria in deep peri-Alpine lakes under warmer climatic conditions

THÈSE

Présentée à la Faculté des Sciences de l’Université de Genève pour obtenir le grade de Docteur ès Sciences, mention interdisciplinaire

par Nicole Gallina de Littau (LU)

Thèse N° XXXX

GENÈVE Ateliers d’impression ReproMail 2012

0

On fait la science avec des faits, comme on fait une maison avec des pierres: mais une accumulation de faits n'est pas plus une science qu'un tas de pierres n'est une maison.

Henri Poincaré

Pour mes enfants Elias et Liv, qui illuminent ma vie à chaque instant.

1

Remerciements

Une thèse est un travail scientifique, une œuvre artistique, qui se crée et s’accomplit avec l’aide et les réflexions de plusieurs personnes. Je tiens à remercier en tout premier lieu mes deux directeurs de thèse, Martin Beniston et Nico Salmaso. Merci à vous deux pour votre confiance sans faille pendant ces quelques années, votre aide et votre appui, ainsi que vos conseils que ce soit pour former ma réflexion ou pour me donner les moyens de travailler et accomplir cette recherche. Vous m’avez également et précieusement formé en tant que chercheuse. J’ai eu la chance d’avoir pu être à vos cotées, ce qui m’a permis d’enrichir considérablement mes connaissances scientifiques, tout comme vous étiez d’une grande valeur sur le plan humain et personnel. Merci Martin, tu étais présent à tout moment quand il le fallait. Merci Nico, tu m’as appris énormément sur le phytoplancton et je ne vais jamais oublier les moments de grandes réfections philosophiques que l’on a eu sur le sujet de ma thèse. Mes éternels remerciements vont également à Orlane Anneville, avec qui j’ai pu avoir des échanges forts intéressants sur le comportement du phytoplancton. J’ai apprécié ces moments de Brainstorming qui ont non seulement nourri ma compréhension mais aussi mon aptitude à faire de la recherche. J’ai beaucoup estimé ton humour et ta modestie. Ensuite, j’aimerais remercier l’équipe du Service de l’Ecologie d’Eau, de l’Etat de Genève, notamment Jean Perfetta, Arielle Cordonier, Sophie Lavigne et Vincent Ebener. C’est vous qui m’avait planté la graine pour que je puisse m’épanouir dans la limnologie et découvrir le monde merveilleux du phytoplancton. Je vous remercie aussi pour votre amitié qui m’est devenue chère. Le groupe climat, incluant mes collègues proches mais aussi les collaborateurs et les secrétaires, est, et on ne peut pas mieux le dire, une équipe fine de choc. J’ai pu travailler dans une magnifique ambiance, avec des personnes de grandes valeurs. Je tiens à spécialement remercier Kri in da Yourte, Thierry, Charly, Margot, Marjorie, Anna, Maura, Bastienne, Stéphane, Denis, Enrique, Francis et Anthony. C’était toujours un plaisir de venir travailler, car je savais que vous étiez présents. Une grande pensée et un éternel et profond remerciement à mes amis. Vous étiez là, c’est sûr, et comment. Sans vous, cette thèse n’aura jamais abouti de la même façon. Merci pour vos soutiens et les moments de grande joie que l’on a pu partager ensemble. Merci Pascale, Anne la Conj, Chanti, Ceska, Albertine, Lidia, Neyda, Nat, Sonja, Aline, Anne E., Anneso, Claudia, Lionel, Gregiboy, Pascal, Stroumpf, Alex, Pablo et Genti. Merci aussi à Mealy, à Oussa et à Kétia, ça fait du bien de vous savoir si purs et proches. Merci à ma famille qui me donne un amour inconditionnel, à ma mère qui m’a toujours fait confiance, à mon père qui m’a appris la persévérance, à mon frère qui veille toujours sur moi, à ma sœur qui est avec moi. Et surtout merci Elias et Liv, vous êtes mes rayons de soleil, vous me donnez le sourire chaque jour, je suis fière de vous.

2

Résumé

Les écosystèmes lacustres s’avèrent être particulièrement vulnérables aux changements climatiques, de façon telle que même de petits changements dans certains paramètres climatiques ont la capacité d’avoir des effets disproportionnés, non seulement sur leur composition chimique mais également sur leurs compartiments biologiques. Les lacs possèdent une valeur écologique, économique mais aussi sociale inestimable. Par conséquent, les effets du changement climatique sur les écosystèmes aquatiques, plus précisément sur la qualité et la disponibilité de l'eau, ont été déclarés menace majeure par l’Union européenne, soulignant l'urgence et la nécessité de pouvoir évaluer les impacts possibles liés à ces changements. En Europe centrale, les modèles climatiques régionaux prévoient une augmentation de la température de l'air jusqu’à 6 °C pour la période à venir de 2071 à 2100. Une attention particulière doit toutefois être portée sur la région alpine, car elle symbolise non seulement une région écologique importante, souvent qualifiée de ''château d'eau'' de l'Europe, mais elle représente également une région caractéristique face aux changements climatiques. Pour cette région distincte, les modèles climatiques ont prévu des températures plus élevées que la moyenne mondiale avec des réponses particulièrement sensibles à des modifications des conditions météorologiques à court terme. Le phytoplancton est constitué par les organismes photosynthétiques capables de vivre en suspension dans la colonne d'eau et il occupe un rôle central dans l’écosystème lacustre. Il est largement utilisé comme indicateur de la qualité de l'eau en raison de sa réponse rapide aux changements environnementaux. Parmi le phytoplancton, les cyanobactéries sont le seul groupe phytoplanctonique d’eau douce capable de produire des substances toxique qui ont le potentiel de contaminer non seulement l’eau potable, mais également de nuire à l'écosystème lacustre entière et de ce fait représentent une menace particulière. En outre, les cyanobactéries ont long histoire de vie sur terre étaient les premiers organismes répertoriés ayant la capacité de produire de l’oxygène et sont par conséquence responsables de l’évolution de la vie sur terre. Leur longue histoire leur a permis d'acquérir des traits éco-physiologiques adaptés qui leur attribuent la capacité d’être potentiellement compétitifs par rapport aux autres groupes phytoplanctoniques, en condition de changement et de stress environnementaux. Par conséquent, face aux changements climatiques, des efflorescences cyanobactériennes ont été présumées et déclarées se produire avec une biomasse et une fréquence plus élevées, ce qui provoque aujourd’hui pour les autorités de gestion de la qualité des eaux un problème considérable. En outre, l’étude de l'effet des changements climatiques sur la composante biologique lacustre présente des interactions très complexes et difficiles à isoler d'autres influences. Le but de cette thèse était d'étudier les impacts du changement climatique sur le comportement du phytoplancton, avec un accent particulier sur les cyanobactéries potentiellement nuisibles dans la région péri-Alpine. L’hypothèse formulée était que, dans cette région, les épisodes d’efflorescence de cyanobactéries sous conditions climatiques plus chaudes pourraient entraîner des impacts négatifs sur la qualité de l'eau et la santé publique. Les objectifs à atteindre étaient d'analyser, premièrement, si la température de l'air est capable d'influencer la communauté cyanobactérienne des lacs péri-Alpins, deuxièmement, de définir les principaux facteurs capables d’influencer significativement le phytoplancton et les cyanobactéries, et, troisièmement, de prédire la biomasse des cyanobactéries sous les conditions climatiques plus chaudes prévues pour les décennies à venir. Pour atteindre ces objectifs, une matrice a été construite, dérivée des sept lacs péri-Alpins et profonds suivants: le Lac de Constance, le Lac de Zürich (Haut et Bas), le Lac de Walen, le Lac de Lucerne, le Léman (Petit et Grand), le Lac Majeur et le Lac de Garde. Les variables environnementales et les données du phytoplancton à différentes périodes ont été fournies en vue d'obtenir une matrice couvrant le gradient trophique comprenant des lacs allant de l'état oligotrophe à l’état eutrophe. En outre, un gradient de

3 température de l'eau associé à un gradient altitudinal était inclus par rapports à la position des lacs qui sont situés dans la même région géographique et climatologique. Ces conditions ont permis une évaluation de la communauté phytoplanctonique d'une manière synoptique. Le phytoplancton et les cyanobactéries ont été classés en fonction de leurs groupes pigmentaires classiques, mais étaient aussi évalués en se basant sur les groupes morpho-fonctionnels (MFG), ce qui a permis d'inclure leur fonction écologique. Afin de prendre en compte l’absence de séries temporelles, des données de la température de l'air lors d'événements extrêmes du climat actuel étaient utilisés comme un proxy pour un réchauffement climatique futur. Ceci a permis d’évaluer ce qui pourrait être le potentiel d’impact des températures d’air sur le phytoplancton. En outre, pour définir quels sont les principaux facteurs responsables et comment ils influencent la composition de la communauté phytoplanctonique, une analyse multivariée a été employée (Non-Metric Multidimensional Scaling (NMDS). Le modèle statistique MARS (Multi Adaptive Regression Spline) a été utilisé pour prédire la biomasse de Planktothrix rubescense, une cyanobactérie potentiellement toxique, dans le Léman. Les résultats démontrent la capacité de la température de l'air d’influencer significativement la communauté phytoplanctonique et suggèrent qu’un climat futur plus chaud favorisera l'augmentation de la biomasse phytoplanctonique, notamment celle des cyanobactéries, ce qui entrainera une perte de diversité parmi les groupes taxonomiques. Des cyanobactéries productrices et non-productrices de toxines répondent de la même manière à la température de l’air plus chaude. La température (eau et air) et les nutriments (P, N) sont les facteurs majeurs contrôlant la communauté phytoplanctonique, selon une force équivalente mais indépendamment les uns des autres. La durée de la période de stratification et le broutage par les cladocères influencent aussi significativement le phytoplancton mais de façon moins forte. Les différents MFG des cyanobactéries réagissent inégalement aux facteurs environnementaux et une attention particulière devrait être accordée aux groupements filamenteux des Oscillatoriales qui semblent être spécialement favorisés avec l’augmentation de la température et une plus longue période de stratification. La concentration des éléments nutritifs au printemps a été démontrée comme étant décisive et capable de contrôler la succession saisonnière du phytoplancton au cours de l'année. L'étude synoptique a clairement montré que chaque lac étudié présentait une communauté phytoplanctonique significativement différente, ce qui a été désigné par « l'effet lac ». Il était présumé que cet « effet lac » est principalement dû à la position altidudinal, donc vincgraphique, de chaque lac et ainsi est responsable de générer une communauté phytoplanktonique bien distincte. Les résultats concernant la modélisation de P.rubescens dans le « Grand Lac Leman » ont démontré que les températures de l'air, saisonnières et extrêmement chaudes favorisaient une biomasse cyanobactérienne plus importante. Les prédicteurs sélectionnés du modèle de P. rubescens sont directement liés aux facteurs climatiques ce qui confirme la vulnérabilité de P. rubescens aux perturbations lorsqu'ils sont sous l'influence du changement climatique, indiquant ainsi la tendance de P. rubescens d'être affecté lors de ces scénarios. Le modèle a prévu non seulement une augmentation de la biomasse de P. rubescens, mais également de sa fréquence. De plus, une tendance de P. rubescens à induire un changement de la composition de la communauté a été démontré, ainsi qu’une période de croissance avec un développement estival plus tôt dans l’année. Il a été proposé que les printemps plus chauds aient pu jouer un rôle décisif favorisant une période de stratification plus longue et plus forte, entraînant ainsi un épuisement plus précoce des nutriments en été. Ce phénomène pourrait représenter les conditions idéales pour que P. rubescens surpasse les espèces concurrentes et se développe avec une biomase plus important en été. De la même manière, cette situation pourrait favoriser d’autres espèces adaptées à ces conditions metalimniques. Les autorités de gestion des eaux devraient être conscientes des périodes printanières plus chaudes que la normale, au cours desquelles une surveillance accrue devrait être entreprise afin de détecter plus rapidement les efflorescences de cyanobactéries potentiellement nuisibles, en particulier celles des Oscillatoriales. Cela pourrait aider les gestionnaires à mettre en œuvre des mesures pour prévenir les impacts négatifs possibles sur la qualité de l'eau et l'approvisionnement en eau potable.

4

Abstract

Lake ecosystems are stated to be particularly vulnerable to changes in climatic parameters, as even small changes have been demonstrated to have disproportionate effects on their chemical and biological components. As natural ecosystems, lakes have immeasurable ecological, economical but also social values. Therefore climate change impacts on lakes, more precisely on their water quality and availability, have been stated as a major threat, suggesting an urgency and necessity to understand the possible effects. In central Europe an increase in air temperature of as much as 6° C by 2071-2100 is forecasted, in which the alpine region has to be taken into special consideration, as it typifies an ecologically important region often stated as the “water–tower” of Europe. Furthermore, also it represents a region for which climate models project an even higher temperature increase than the global average, responding particularly sensitively to short- term changes in weather. Phytoplankton, photosynthetic organisms adapted to live in suspension in the water column, embody a central role in aquatic ecosystems and are widely used as indicators of water quality due to their rapid response to environmental changes. Among the different groups of phytoplankton, cyanobacteria represent a special threat as this group is the only freshwater phytoplankton capable of forming blooms and producing toxins, which have the potential to harm the entire lake ecosystem. Moreover, cyanobacteria have long life histories which allow them to acquire specific improved eco-physiological traits, thus enabling them to be potentially competitive over other phytoplankton groups under conditions of environmental changes and stress. Therefore cyanobacteria outbreaks have been stated as becoming increasingly important in biomass and frequency under climate change conditions, and represent a significant threat to lake water management authorities. However, the effect of changes in climatic conditions on the biological component, are complex and difficult to disentangle from other influences. The aim of this thesis was to investigate the impacts of climate change on the behavior of phytoplankton, with a special emphasis on harmful cyanobacteria in the peri-Alpine region. It was hypothesized that in this highly sensitive region more important episodes of harmful cyanobacteria outbreaks under warmer climatic conditions could lead to negative impacts on water quality and public health. The objectives were to first analyze if air temperature is able to influence the cyanobacteria community, second to define the main drivers for the phytoplankton/cyanobacteria community in peri-Alpine lakes, and third to predict the cyanobacteria biomass under warmer climatic conditions projected for the coming decades. To achieve these objectives, a matrix derived from seven deep peri-Alpine lakes was constructed: Lake Constance, Lake Zürich (Upper and Lower), Lake Walen, Lake Lucerne, Lake Geneva (Small and Big), Lake Maggiore and Lake Garda. Environmental variables and phytoplankton data from different time periods were provided in order to obtain a matrix that covered the entire trophic gradient from lakes in the oligotrophic state to those in a eutrophic state. Additionally, the lakes included an altitudinal gradient associated with a water temperature gradient, and moreover are situated in the same climatological and geographical region. These conditions allowed an assessment of the phytoplankton community in a general synoptical manner. The phytoplankton/cyanobacteria community was classified according their conventional pigmentary groups, and assessed using Morpho-Functional Groups (MFG), which enabled to include their ecological function. In order to overcome the challenge of missing long-term datasets, an appropriate solution was hypothesized as using data on air temperature during extreme events under current climate as a proxy for future mean climate change. Thus the potential of and the way how air temperature impacts the phytoplankton community could be investigated. Furthermore, in order to define the main drivers and how they influence the phytoplankton community composition, a multivariate analysis of Non-Metric Multidimensional scaling (NMDS) was employed. The statistical Multi Adaptive Regression Spline

5

(MARS) Model was used to predict the biomass of the potentially toxic species Planktothrix rubescens in Big Lake Geneva. The outcomes clearly show the capacity of air temperature to significantly influence the phytoplankton community and suggest that future warmer climate will enhance the increase of phytoplankton biomass, especially cyanobacteria in the peri-Alpine region, leading to a loss in diversity among the phytoplankton groups. Toxin and non-toxin producing cyanobacteria generally respond in the same manner to warmer air temperature. Throughout the year, the phytoplankton community is manly driven by temperature and nutrients with equal strength, but, independently to each other. The duration of the stratification period and grazing by the Cladocerans as well significantly impacted the phytoplankton community. The cyanobacteria MFG’s respond differently to the drivers and special attention should be paid to the filamentous Oscillatoriales, which may be favored with increased water temperature and a longer stratification period. The nutrient concentration in spring was revealed to be a decisive factor able to control the seasonal succession of phytoplankton during the remaining year. The synoptical study clearly showed that each lake studied had a statistically-significant different phytoplankton community. It was hypothesized that this differential behavior is mainly due to the geographic/altitudinal position of the lakes able to generate a different phytoplankton community composition. In Big Lake Geneva, seasonal extreme warm air temperatures led to higher cyanobacteria biomass. The predictors selected to model P. rubescens were climatologically related predictors, therefore vulnerable to disruption when under the influence of climate change, thus indicating the potential of P. rubescens to be affected under these scenarios. P. rubescens biomass was projected to increase not only in abundance but also in frequency thus potentially able to induce a community function change. Furthermore, the growth period is predicted to start earlier. It was suggested that a warmer spring might be the decisive factor, leading to a longer, stronger stratification period, thus resulting in earlier nutrient depletion during summer. This could represent not only an ideal circumstance for P. rubescense to outcompete other species and boost their growth earlier and with higher biomass, but also favor other species adapted to these metalimnic conditons.

Water management authorities should be aware of warmer than normal spring periods, during which increased monitoring should be undertaken to detect potential harmful cyanobacteria outbreaks earlier, especially Oscillatoriales. This could help water managers to implement measures to prevent possible negative impacts on water quality and drinking water supply.

6

Preface

This thesis is based on the following papers,

I Gallina N., O. Anneville, M. Beniston. Impacts of extreme air temperatures on cyanobacteria in five deep peri-Alpine lakes. Journal of Limnology, 70(2) : 186-196, 2011.

II Gallina N., N. Salmaso, G. Morabito, M. Beniston. Phytoplankton configuration in six deep lakes in the peri-Alpine region: Are the key drivers related to eutrophication and climate. (Submitted to Aquatic Ecology).

III Gallina N, M. Beniston, S. Jacquet. Will Lake Geneva turn red? (Submitted to Limnology & Oceanography).

7

Table of contents

1. INTRODUCTION ...... 9

1.1 Climate change and Lakes: An Overview ...... 9 1.1.1 Climate change in peri-Alpine Lakes: the importance of regional- geographical location of lakes ...... 10

1.2 Phytoplankton in lake ecosystems...... 11 1.2.1 Definition and importance ...... 11 1.2.2 Classifications ...... 11 1.2.3 The seasonal succession ...... 13

1.3 Cyanobacteria ...... 13 1.3.1 Origin, morphological characteristics, distribution and eco-physiological traits ...... 14 1.3.2 Factors influencing Cyanobacteria ...... 15 1.3.3 Environmental changes, causes for cyanobacteria outbreaks ...... 18 1.3.4 Consequences of Cyanobacteria blooms for Lake Ecosystems ...... 21

1.4 Aims and objectives ...... 21

1.5 Data and Methods used ...... 22 1.5.1 Data characteristics ...... 22 1.5.2 Methods ...... 27

2. IMPACTS OF EXTREME AIR TEMPERATURE ON CYANOBACTERIA IN FIVE DEEP PERI-ALPINE LAKES ...... 31

3. PHYTOPLANKTON CONFIGURATION IN SIX DEEP LAKES IN THE PERI-ALPINE REGION: ARE THE KEY DRIVERS RELATED TO EUTROPHICATION AND CLIMATE? ..... 49

4. WILL LAKE GENEVA TURN “RED” IN THE FUTURE? A POSSIBLE SCENARIO FOR THE DEVELOPMENT OF THE CYANOBACTERIUM PLANKTOTHRIX RUBESCENS...... 68

5. CONCLUSIONS AND PERSPECTIVES ...... 88

6. ANNEXES ...... 93

7. BIBLIOGRAPHY ...... 96

8

Chapter 1

1. Introduction 1.1 Climate change and Lakes: An Overview

Aquatic ecosystems are considered to be an integral part of human existence on earth that in addition provides an essential dimension to our lives (Williamson and al. 2009a). Lakes are freshwater bodies highly valued for recreational activities, providing drinking water supply and sustaining natural ecosystems that are home to many of species. Their economic importance is highlighted by the use of water for agricultural purposes (irrigation), their role as an important source of food (fishing), and their use for waste water and sewage treatment and in terms of hydro-power production. In Europe, the substantial concern to maintain or restore lake water quality was the keystone for the Water Frame Work Directive (WFD, Directive 2000/60/EC), which came into effect in 2000. The WFD represented a fundamental change in water management in Europe expressing the need by which aquatic ecosystems need to be assessed in a holistic way in order to achieve, by the end of 2015, a good ecological status for all ground and surface waters. However, at the time the WFD initiative was created, the climate change issues were not taken in account as they should have been. In 2008, a technical report of the Intergovernmental Panel on Climate Change (Bates and al., 2008) clearly stated with their following outcome the urgency and necessity to understand the effect of climate change on lakes: "Water and its availability and quality will be the main pressures on, and issues for, societies and the environment under climate change”.

Climate change is thus considered to be a genuine threat to many natural systems, and has become the body of research interest among limnologists around the world. Research in limnology was initiated to answer the question of how lake ecosystems may respond to a changing world, in which even small changes were demonstrated to have a disproportionate effect on their chemistry and biology (Wehenmeyer and al., 1999). Ten years later, lakes were stated to be effective sentinels, integrators and regulators of climate change (Adrian and al. 2009, Williamson and al., 2009a, 2009b), due to their rapid responses to environmental changes, and their capacity to integrate information about changes in the catchment zones (Adrian and al., 2009).

9

Changes in climate forcing have been found to affect the physical environment of lake ecosystems and thereby induce an alteration in their chemical and biological properties, which ultimately have the ability to jeopardize the ecosystem services that lakes provide (Vincent 2009). Of particular interest to limnologists are the interactions between variables, the feedback effects that accelerate or dampen environmental change.

1.1.1 Climate change in peri-Alpine Lakes: the importance of regional- geographical location of lakes

Of a particular interest when investigating lakes related to climate changes is the geographic location of the lakes. In Europe, different weather systems are used to classify weather types. George and al. (2010a) demonstrated by using three closely related weather systems, that each type is able to influence the observed variation in the surface temperature of lakes situated in Northern Europe, Western Europe and Central Europe.

In Central Europe, climate change scenarios suggest an increase by as much as 6°C by 2071-2100 (IPCC, 2007. Special attention should be paid to the Alpine region, where the temperature increase is projected by climate models may be even higher than the global average rise (Beniston, 1997). This region is defined as being particularly sensitive to short-term changes in the weather (Psenner, 2003; Thompson and al., 2005). Thus, these changes have a strong impact on the water bodies of the Alpine region, and because the Alps are often considered to be the ‘water tower’ of Europe, these lakes are therefore of major ecological importance.

Another important aspect in studying lakes on a same geographical region is the regional coherence in-between lakes, which was first demonstrated by the work of Magnuson and al. (1990). The features of lakes within the same region respond coherently to drivers such as climate forcing and catchment processes. Recent work from Livingstone and al. (2010) showed that spatial coherence for physical variables is the highest whereas for biological variables the weakest thereby emphasizing the difficulty to assess these variables in different lakes. In other words, the effect of changes in climatic conditions on the biological component, especially the phytoplankton community, are complex, difficult to disentangle from other influences and not easy to generalize (Livingstone and al., 2007). This complexity of interactions should be constantly taken into consideration if the growth, structure and dynamics of phytoplankton community are assessed.

10

1.2 Phytoplankton in lake ecosystems

1.2.1 Definition and importance The definition of phytoplankton is based on the term “phyto” meaning plants and designates organisms able to do the photosynthesis and “plankton”, organisms adapted to live in suspension in the water column. However, due to the widespread variety between species, the following re- definition from Reynolds (2006) was selected as follows: Phytoplankton is a collective of photosynthetic microorganisms adapted to live partly or continuously in open waters. As such it is the photoautotrophic part of the plankton and a major primary producer of organic carbon in the pelagic of the seas and of inland waters.

Phytoplankton plays an important role in lake ecology. In the process of photosynthesis, phytoplankton produces half of the world's oxygen (all kind of water bodies included). Moreover, by primary production, death and sinking they effectively transport carbon from surface layer to sediments, a process by which phytoplankton exert a global-scale influence on climate (carbon dioxide and the greenhouse effect). Furthermore, phytoplankton constitutes the bottom level of aquatic food webs (Arrigo, 2005). In addition, due to their rapid response to environmental changes, phytoplankton constitutes an important indicator of water quality. (Stoermer, 1978; Directive 2000/60/EC), and its use as an indicator of environmental changes became more important under the pressures of a changing climate.

1.2.2 Classifications

The classification of phytoplankton is a complicated and complex task for taxonomists, demanding a certain experience and knowledge in order to identify correctly these microorganisms, which have vast morphological and physiological characteristics. Their taxonomy has been continuously revised, improved with the discovering of new species or new traits on already discovered species and still remains subject to discussion (Komarék, 2003). Moreover, the advance of science added the knowledge of phylogenetic classification based on genomic resemblance in between species, which required updating the existing classification. The major taxonomic groups are based upon their primary characteristic, the photosynthetic pigments: chlorophylls, carotenoids, xanthophylls and the phycobiliproteins. In deep peri-Alpine lakes, the resulting seven algal groups were identified and considered following the more recent monographs of the series ‘’Süsswasserflora von Mitteleuropa” established by A. Pasher (Gustave Fisher Verlag) as well as The Süsswasserflora of the British Isles (John and al., 2002). The definition of the major groups of eukaryotic algae is coherent with the recent status in the systematic described by Krienitz (2009). However, the contribution of Chrysophyceae, and Bacillariophyceae, which were included, by Krienitz (2009) in the division Heterokontophyta, were considered separately during the thesis.

11

CYANOBACTERIA (blue green algae). They are structurally and physiologically like bacteria, but they photosynthesize functionally like plants in aquatic systems. An enhanced description will be given in a separate paragraph.

BACILLARIOPHYCEAE (Diatoms). A most important group of the phytoplankton and highly distinctive, through their characteristics including a unique cell wall of silica (termed as ‘frustule’). The beautiful structures of their frustule are complex and used as taxonomic characteristic. Diatoms are widely used as indicators of water quality in rivers (Gallina and al., 2009) since distinct nutritional requirements favor growth of one group over another (Wetzel 2001, Reynolds 2006).

CHLOROPHYCEAE (Green algae). This is an extremely large and morphologically diverse group of algae that is almost totally distributed in freshwater. Many members are flagellate, at least in the gametes stages (John 2002, Reynolds 2006).

CONJUGATOPHYCEAE (Charophyta – Green algae). Well defined group among the green algae’s and characterized by their sexual reproduction by the conjugation of the gametes. The genera Mougeotia bloomed in Lake Geneva and Lake Garda notably.

DINOPHYCEAE. Unicellular flagellated providing weak locomotion, which is relatively resistant to most grazing organisms, due to their size and are limited by weak nutrients concentrations. They are conspicuously represented in freshwaters (Reynolds, 2006)

CHRYSOPHYCEAE. Most are unicellular and only few are colonial; the presence of flagella is variable. Many species lack cell wall. They are major components in temperate oligotrophic lakes (Hutchinson 1967).

CRYPTOPHYCEAE. Most are naked, unicellular, and motile, occurring in most lakes, regardless their trophic state having a characteristic intermittent numerical dominance and qualified as ecologically significant internal stabilizing component of plankton communities (Wetzel, 2001).

1.2.2.1 An ecological way of classification to assess the phytoplankton community

Phytoplankton ecologists soon stated that the taxonomical classification of phytoplankton based upon their pigmentation characteristics, is not always convenient to use for the description of the ecological function and quality of the aquatic environment. With the idea to not only name the species, but also have a close idea of their affinities, in order to group species together with the same ecological traits and shared adaptive features, Reynolds and al. (2002) published a review in which they promote a new classification, a functional classification to study the phytoplankton community. The resulting function groups are based not only on individual functional traits, but also on the range of environmental conditions for which the species are found to occur. Salmaso

12 and Padisàk (2007) joined this idea and developed the morpho-functional groups (MFG), which include the morphological as well as the functional features of phytoplankton species. Recently, Kruk and al. (2011) tested all these classification methods, and stated that the community composition can be best predicted in using the new classification methods, namely the morphological groups.

1.2.3 The seasonal succession

The growth and seasonal successions of phytoplankton are regulated by variety of external as well as internal factors (Reynolds 2006). The main controlling factors are related to light, nutrients, water temperature, turbulence, and trophic interactions (grazing and competition) (Harris 1986). The Plankton Ecology Group of the international Society of Limnology proposed the conceptual Model (PEG- model) (Sommer and al. 1986), in which “24 Sequential Statements of Seasonal Succession of Plankton in Freshwater” (Annexe 1) were exposed for temperate lakes. The PEG-model defines the most important mechanisms governing the phytoplankton community during the year and clearly indicates the importance of physical forcing (light, temperature, and mixing) during autumn and winter, chemical conditions (depletion of nutrients) and biological interactions (grazing and competition) during spring and summer. Further work confirmed the role of physical constraints in shaping both the biomass and the composition of phytoplankton, especially in winter (Padisàk, 2010).

Therefore, as climate directly affects light, mixing and water temperature, its capacity to influence the phytoplankton community and its seasonal evolution becomes more obvious.

1.3 Cyanobacteria

The following chapter introduces the importance cyanobacteria present in lake ecosystems. It focuses on their origin, their morphological and eco-physiological traits which give cyanobacteria their competitive advantage over the rest of the phytoplankton community. Moreover, the environmental factors controlling their growth are delineated, including climate change and eutrophication. Finally, their potential harmful effects to lake ecosystems and the reasons why lake management authorities consider cyanobacteria as a major concern is discussed.

13

1.3.1 Origin, morphological characteristics, distribution and eco-physiological traits

Cyanobacteria are known to be the world’s oldest known oxygen-producing and nitrogen-fixing organism (Schopf 2000, Chorus and Bartram, 1999). They are considered as major actors responsible for the accumulation of oxygen in the Earth’s early atmosphere and are thus considered as being at the origin of the evolution of life on earth. Cyanobacteria have a long life history and their first appearance was confirmed to date back to about 2.15 billion years ago (Hoffmann 1975, Knopf 2006, Ramussen 2008).

Cyanobacteria are photosynthetic prokaryotes as they lack nuclei and other organelles, and have a peptidoglycan cell wall that is typical of gram-negative Eubacteria (Hoiczyk and Hansel, 2000). However, they are classified as blue-green algae because of their algal-like appearance, their possession of chlorophyll and their photosynthetic production of oxygen by a two photosystem process (I+II) (Castenholz, 2002). All cyanobacteria contain chlorophyll a and most comprise blue phycobiliproteins phycocyanon and phycocyanin, giving the cells their characteristic blue– green color (Grossman and al., 1995). Although cyanobacteria lack membrane-bound organelles, they have a variety of cellular structures and inclusions that have specialized functions that contribute to their ecological success, as these inclusions allow cells to accumulate energy and nutrients far in excess of their present requirement when they are under favorable conditions and to subsequently use these reserves for growth maintenance when they encounter source-poor conditions (Vincent, 2009). Some planktonic species contain in their cell up to several thousand gas vacuoles, providing buoyancy to the cells and colonies and thus float among the water column to require ideal light and nutrient conditions (Walsby and Hayes, 1988; Walsby, 1994; Walsby and al., 2004). Moreover cyanobacteria produce different cell types, like the heterocyte (e.g. genera Nostocales) which are the location of the enzyme nitrogenase for nitrogen fixation. Another specialized cell type is the akinete, formed under unfavorable conditions and that allows cyanobacteria to overwinter in the sediments (Mur and al., 1999).

Cyanobacteria are distributed worldwide and can be found in environments ranging from the tropics to the arctic, and from alkaline to acidic environments, and can proliferate in freshwater, estuarian and marine ecosystems (Chorus and Bartram, 1999, Mur and al. 1999).

Their long life history through which cyanobacteria had to evolve has resulted in an enhanced ability of adaptation, which is evidenced by their improved eco-physiological traits (Litchman and al., 2010; Cayelan 2011). These traits are hypothesized to confer cyanobacteria the ability to compete with other phytoplankton groups in situations of environmental stress. The following lists evoke these traits. As already mentioned, the first three traits result from the production of their differentiated cell structures.

14

a) Heterocysts, which enable some genera of cyanobacteria to fix the atmospheric nitrogen via the nitrogenase enzyme, when nitrogen is scarce in the environment (Ganf and Oliver, 2000). b) Akinetes, which are produced under stress situations (low light conditions and deficiency of nutrients) (Kaplan and Levy, 2010) and allow the cyanobacteria to survive under unfavorable environmental conditions. Akinetes are resistant to cold and dry weather (Adams et Duggan, 1999). c) Buoyancy enables cyanobacteria to migrate in stratified lakes between the surface where light is abundant to produce photosynthesis and to the nutrient rich metalimnion, where it absorbs the accumulated carbohydrates from the surface layer by respiration (Ganf and Oliver, 2000; Walsby 1994) d) UV radiation tolerance in developing sunscreen pigments that envelope the cell and function even when cells are at rest. Moreover cyanobacteria are able to develop efficient systems for repair of damaged DNA and for replacement of UVR-damaged compounds, and to implement directed gliding motility for escaping the diurnally high intensities of solar irradiance (Castenholz and Garcia-Pichel, 2002). e) Warmer temperature optima for growth are features of most cyanobacteria, and bloom- forming species prefer temperatures exceeding 15°C. The temperature at which maximum replication rates occurred for cyanobacteria varied from just over 20 °C for Aphanizomenon flos-aquae and Planktothrix agardhii, to 28 °C for Microcystis aeruginosa, and even 41 °C for Synechococcus sp. (Reynolds, 1989, 2006; Robarts and Zohary, 1987). f) Large light-harvesting antenna. Cyanobacteria, are reputed to be strong competitors for light due to their large light-harvesting antenna pigments (Reynolds 1997, Jacquet 2005). g) Toxin productions. Cyanobacteria are the only freshwater phytoplankton group to produce a diverse range of toxins (Codd and al., 1989, 1999, Carmichael 1997.), which have been hypothesized to confer allelopathic advantages to grazing and competition (e.g. Schatz, 2007, Roy, 2009).

1.3.2 Factors influencing Cyanobacteria

It is well established that the success of cyanobacteria is a result of complex and synergistic environmental factors, rather than a single dominant variable (Hyenstrand and al., 1998; Dokulil and Teubner, 2007). We introduce here the main factors which are supposed to govern significantly Cyanobacteria.

15

1.3.2.1 Physical forcing a) Thermodynamics and stratification

The thermodynamics of the water masses have an influence of the spatial distribution of phytoplankton communities (Hedger and al., 2004).

The horizontal distribution is a consequence of the wind, which is known to play an important role in creating advection in the water, thus accumulating mainly species capable of floating in matter to form scums, mats or blooms (Hedger and al., 2004).

The vertical distribution is a consequence of temperature and light. The atmosphere imposes a temperature signal on the lake surface inducing a density difference in water bodies, facilitating an evolution of chemical differences with many consequences for living organisms in lakes (Boehrer and Schulze, 2008). Peri-Alpine, temperate lakes are classified as warm monomictic lakes (Hutchinson, 1957), which are lakes that never freeze, and are thermally stratified throughout much of the year. The density difference between the warm surface waters (the epilimnion) and the colder bottom waters (the hypolimnion) prevents these lakes from mixing in summer. During winter the surface waters cool to a temperature equal to the bottom waters. Lacking significant thermal stratification, these lakes mix thoroughly in winter (holomictic lakes) and in deep lakes during very cold winters or to a certain extent during warmer winters (oligomictic lakes) (Lewis, 1983). Vertical movements favor species that are able to develop under low light conditions during mixing periods (Huismann 1999; Hedger, 2004) and for the opposite case, during stratification periods, species able to move vertically in the water column in order to seek ideal light and nutrient conditions (Huisman, 1999).

b) Temperature

Temperature is known to be the most important environmental driving factor affecting directly the metabolism, growth, reproduction and survival of living organisms, as well as the interaction among species (Ibelings and al., 2011) and therefore impacts the ecosystem functioning (De Stasio, 2009). Furthermore temperature is also indirectly involved in the strength and duration of the stratification period and enhanced availability of nutrients in the epilimnion and hypolimnion (Søndergraad and al. 2003). As already mentioned, most of the cyanobacteria are blooming under their optimal growth temperature of 20°-25°C (Reynolds, 1987; Robarts and Zohary, 1987), nevertheless their optimal growth varies among the cyanobacteria considered.

16 c) Light

The accessibility of light is an important factor for organisms performing photosynthesis. In the water column, the fraction of light energy available for photosynthesis, (PAR, Photosynthetically-Available Radiation) is situated around wavelengths between 400 and 700nm, and is highly influenced by the depth, turbidity, particles in suspension, chlorophyll concentrations, and also by the existing, self-shading phytoplankton community (Tilzer and al.,

1995). The depth of water transparency is measured by the help of the Secchi disc (zs), from which the euphotic depth (ze) can be derived as follows: ze= 2.7*zs. The euphotic depth is considerate as the limit of the photosynthetic activity, corresponding to 1% of the light intensity present at the lake surface (Lemmin, 1995). Light thus divides the lake water column into the euphotic zone, where the proliferation of phytoplankton takes place, and the aphotic zone. Another measure of the light availability used is the underwater light climate index, which is expressed by the ratio of the depth of the mixing layer and the euphotic depth (zm/zeu). The capacity to access and to use the available light is a determinant factor for cyanobacteria growth, which moreover have the capacity to modify, by expressing genes the quality of phycobilisomes, the light harvesting antennae of photosystem II (Kehoe et Gutu, 2006)

1.3.2.2 Nutrients

Tilman & Khilman (1976) and further Reynolds (2006) demonstrated that the differentiated abilities of phytoplankton species to gather resources necessary to support cell growth and replication might influence the dynamics and ecology of phytoplankton populations.

In addition to light, growth of cyanobacteria consumes carbon and iron (photosynthesis), ‘nutrients’ and equally, may often be constrained by their availability and fluxes. Since the earliest days of phytoplankton ecology, phosphorus (P) and nitrogen (N) have been emphasized to be the most important variables controlling the phytoplankton community structure and biomass (Hutchinson 1967), and their role as limiting nutrients was demonstrated (Tilman 1982). The term of limiting nutrients is due to the fact that it is not the quantity of these required elements that constrains the growth, but the ease with which they are obtained (Reynolds 2006). It is the demand relative to the supply that is ultimately critical, bearing in mind that the presence of nutrient should be in their assimilable form requested by the cells. Cyanobacteria have the ability to avoid periods of nutrient deficiency to allow achieving their need. The enzyme nitrogenase allows fixing the atmospheric nitrogen, as already evoked. In periods of phosphorus insufficiency, some cyanobacteria have the capacity to cleave the organic compounds by the activation of extracellular enzymes (Dyhrman and Ruttenberg, 2006). It is also shown that the stoichiometry of the primary nutrients present in the aquatic environment seems to influence the composition of the phytoplankton community (Smith and Bennet, 1999; Huisman and Hulot,

17

2005). Especially, a weak ratio TN:TP (total nitrogen (TN) and total phosphorus (TN)) is known to have the tendency to favor the cyanobacteria (Huisman and Hulot, 2005, Smith 2005).

1.3.2.3 Biological interactions

a) Competitons

With the famous “paradox of plankton”, published 50 years ago, Hutchinson (1961) raised the question of “how it is possible for a number of species to coexist in a relatively isotropic or unconstructed environment, all competing for the same sorts of materials”. With this statement, he highlighted the observed contradiction between the numbers of species competing for the same resource and the principle of competitive exclusion (Hardin 1960), in which the final equilibrium is attained if the population were to reduce to one species which outcompeted all the others. A number of investigations followed to describe the abiotic and biotic factors which prevent to attain the equilibrium (e.g. Scheffer and al., 2003, Benincà and al., 2008). The eco- physiological adaption mentioned above, clearly indicate the advantage of cyanobacteria in comparison to their competitors, therefore putting their hypothetical success in situations of environmental stress into perspective.

b) Grazing by zooplankton

For cyanobacteria, grazing seems not to be an important threat (Psenner, 1995). The reason of resistance to grazing can be summarized by first the important size of their colonial forms or filaments therefore difficult to ingest, second by the bad food quality they represent for grazers due to their weak content in polyunsaturated acid (DeMott and Müller-Navarra, 1997) and third, as already evoked, by their production of toxic allopathic compounds. On the contrary, even a negative impact of cyanobacteria on zooplankton needs to be discussed (Pearl, 2001), as well as the positive influence of zooplankton favoring cyanobacteria by reducing their competitors (Fulton and Pearl, 1987).

1.3.3 Environmental changes, causes for cyanobacteria outbreaks

During the last decades, an increasing number of publications have stated the rise in cyanobacteria outbreaks in freshwater ecosystems. This increase is partly due to more frequent surveys (Sellner and al., 2003), but several authors also observed cyanobacteria increases in biomass, duration and distribution (Anderson, 2002). The reasons for these increases are due to two anthropogenic related phenomena which are identified to be the most important threats in lake ecology: Eutrophication and Climate Change.

18

1.3.3.1 Eutrophication

Over several centuries human nutrient over-enrichement (particularly nitrogen and phosphorus), associated with urban, agricultural and industrial development, has promoted accelerated rates of primary production, or eutrophication (Pearl and Paul, 2011). Moreover eutrophication is projected to be favored in relation to climate change (Moss, 2011). A strong correlation in between eutrophication and blooms of cyanobacteria (Reynolds and Petersen, 2000; Downing and al., 2001) has been recorded. Some results also clearly point out that eutrophication favors the proliferation and dominance of harmful blooms of cyanobacteria (Fogg, 1969; Huismann and Hulot, 2005; Pearl and Fulton 2006, Moss 2011). Nevertheless, even after restoration of the largest natural French lake (Lake Bourget in Savoie), the proliferation of a cyanobacteria species was observed (Jacquet and al., 2005). Likewise Anneville and al. (2002) indicated in Lake Geneva higher phytoplankton biomass even though the nutrient loads decreased after management efforts. This non-linear relationship between nutrient and phytoplankton/cyanobacteria was termed ‘’Hysteresis”, translated by the resilience of the ecosystem to maintain its integrity once challenged to environmental modifications (Carpenter and Cottingham, 1997). Therefore the recovery of lakes after reductions in nutrient loading may be confounded by concomitant environmental changes such as global warming. However, effects of global change are likely to run counter to reductions in nutrient loading rather than reinforcing re-oligotrophication (Jeppesen and al., 2005).

1.3.3.2 Climate change

Cyanobacteria have been hypothesized to benefit from environmental changes associated with global warming (Paul, 2008; Paerl and Huisman 2008; Paerl and Huisman, 2009; Mooij and al., 2005; Pearl and Paul, 2011). Their increase may be mainly due to their high capacity to be well adapted when subjected to various stresses related to climate change (Cayelan and al., 2011). Moreover several authors observed the potential of cyanobacteria to become dominant (Elliot 2006, Jöhnk and al. 2008).), but also able to change in their phenology of bloom events (Elliot 201o, Zhang, 2011). Indirecty, deeper stratification (Anneville, 2002), favor cyanobacteria adapted to low light conditions. Moreover, the migration from tropical species to northern temperate situated lakes also has been observed (Briand and al., 2004; Wiedner and al., 2007). However many of these factors could be the direct consequences of temperature due to their optimal growth at higher temperatures compared to other phytoplankton groups (Reynolds, 1984; Robarts and Zohary, 1987), but also to indirect positive effects of temperature, which are multiple. First, buoyant cyanobacteria will be favored under thermal conditions that are likely to become more stable, prolonged and intense (Paerl 1988; Dokulil and Teubner 2000; Winder and Schindler 2004; Wagner and Adrian, 2009). Another important trait is that climate change may reinforce the symptoms of eutrophication that favor cyanobacteria (Moos, 2011; Adrian 2009).

19

Figure 2 summarizes the factors involved in the growth of cyanobacteria under a climate change scenario.

Figure 2 highlights the different factors and mechanisms involved in mediating the climatic response of cyanobacteria (adapted from Nõges and al., 2010).

1.3.3.3 Climate change and Nutrients: A synergistic effect?

When it comes to assess impacts on phytoplankton, identification of climate signals can be further complicated by the influence of other environmental changes, such as eutrophication (Adrian and al., 2009). The questions about the synergy and the relative importance of influence of these two elements of global changes arise, when affecting the phytoplankton community. Moss and al. (2003) found that warming had a considerably smaller effect on the phytoplankton community than did fish and nutrients through a mesocosm study. Research on sediment cores of Lake Biwa indicated a stronger effect of nutrient during the eutrophication period, whereas meteorological forcing was the driving factor during oligotrophicaton (Tsugeki and al, 2009). The results of Stich and al. (2009) indicate that oligotrophication outweights the effect of global warming in Lake Constance. The phytoplankton community model PROTECH (Phytoplankton Responses To Environmental Change; Reynolds and al. 2001) was used to investigate the impacts of changing water temperature and nutrient loading upon the phytoplankton in Loch Leven (Elliott and May 2008). Change in water temperature had relatively little effect on phytoplankton biomass and species diversity in comparison with changes in nutrient loading. However, phytoplankton varied according to the way in which nutrient loading changed. In another study, phytoplankton community models predicted a greater impact of nutrient loading over temperature, but with a large dominance of cyanobacteria when high water temperatures

20 faunawere combined with high nutrient loads (Elliott and al. 2006). Although these questions are important to answer when it comes to find the most adapted and efficient solution for water management authorities, the relative importance of nutrients versus temperature as the dominant impact on cyanobacteria still remains unclear.

1.3.4 Consequences of Cyanobacteria blooms for Lake Ecosystems

The consequences of cyanobacteria blooms and the way they may damage the lake ecosystem represent a major concern for lake management authorities.

The increased biomass during a bloom episode increases the turbidity of lakes which in turn affects the transparency and therefore restricts the light availability necessary for the aquatic vegetation, which in turn leads to habitat disappearance for fish and benthic flora and fauna (Scheffer and al., 1997). The phenomena of shading due to their outbreaks entrain the suppression of the phytoplankton via competition for light (Jöhnk and al., 2008). A drastic reduction of the phytoplankton diversity may be observed (Crosetti and al., 2008). Moreover, during nighttime, dense blooms of cyanobacteria are able to cause oxygen depletion through respiration and bacterial decomposition, which can results in massive fish mortality and loss of fauna and flora (Pearl and Fulton 2006), and to a complete imbalance in the entire trophic chain (Vanni and al., 1997).

The major concern of cyanobacteria is related to health risk and their ability to produce toxic compounds able to harm both animal and human health, as a result of drinking the water or swimming (Chorus and Bartram 1999, Dokulil and Teubner 2000, Briand and al. 2003). These toxins can be grouped into three families, depending on their toxic effects: Hepatotoxins, neurotoxins and dermatotoxins. Hence numerous hypotheses were formulated about the reasons underlying the production of these toxins (Vasconcelos, 2001; Wiegand and Pflugamcher, 2005; Schatz and al., 2007). The results suggest that the target of their allopatic effect may be to harm potential grazers or competitors (Paerl and Millie, 1996; Tillmans and al., 2008). The impact of these cyanotoxins on aquatic ecosystems still remains, however, largely misunderstood.

1.4 Aims and objectives The aim of this thesis is to investigate the impacts of climate change on the behavior of phytoplankton, with a special emphasis on harmful cyanobacteria in the peri-Alpine region, a region known for its vulnerability to global warming and its ecological importance as the “water- tower” of Europe. It is hypothesized that in this region more important episodes of harmful cyanobacteria outbreaks under warmer climatic conditions could lead to negative impacts on water quality and public health. The outcomes should lead to a better understanding of what water management authorities have to expect in order to avert the more negative risks arising

21 from a warming climate. The enhanced knowledge should therefore allow the responsible authorities to successfully maintain water quality and drinking water supply in a world in which climate is changing and population is growing. To achieve this goal the following objectives were defined:

I. To analyse of whether air temperature (a driver directly related to warming) is able to influence cyanobacteria in peri-Alpine lakes (cf. Chapter two).

II. To define which are the main drivers for the phytoplankton/cyanobacteria community in peri-Alpine lakes (cf. Chapter three).

III. To predict cyanobacteria biomass under warmer climatic conditions projected for the coming decades of the 21st century (cf. Chapter four).

The first two objectives (I and II) are meant to lead to a better understanding concerning the behavior of the phytoplankton, and especially the cyanobacteria, in the particular area of the peri- Alpine lakes. Even though some of the six lakes were investigated together (Anneville, 2004; Salmaso, 2006), this study represents the first synoptic investigation, involving a consistent dataset with an interesting number of lakes, focusing on cyanobacteria. This step was achieved by using descriptive statistics, with the aim to identify specific characteristic of the phytoplankton / cyanobacteria of peri-Alpine lakes. Further, the results obtained provide the necessary information needed to attempt to predict the evolution of cyanobacteria biomass, in a way that it represents realistic adapted features for the peri-Alpine region. This last step (III) was achieved with the help of statistical modeling methods.

1.5 Data and applied Methods

1.5.1 Data characteristics

The idea was to compile a unique and extensive matrix for the peri-Alpine region. When investigating which lake to include, three principle aims were pursued:

1) Having similar types of lakes in the same climatological and geographical regions, allowing for comparability and integration into one common matrix. Such a matrix moreover, thus enables to perform for a potential synoptic study in a defined region. 2) Having lakes covering the whole trophic gradient, from oligotrophic to eutrophic. As nutrients were supposed to be important drivers for the phytoplankton community, the assessment of phytoplankton at each trophic state could be integrated. 3) Having lakes at different altitudes, which potentially could facilitate the integration of a heterogenic phytoplankton community resulting from the temperature gradient.

22

In the end, an original matrix derived from seven lakes was able to be included into the final peri- Alpine Lake dataset, namely Lake Constance, Lake Zürich, Lake Walen, Lake Lucerne, Lake Geneva, Lake Maggiore, and Lake Garda. Data from Lake Geneva and Lake Zürich were collected at two points, referred to as “Small Lake Geneva” and “Big Lake Geneva”, and “Upper Lake Zürich” and “Lower Lake Zürich”, respectively. Consequently, nine datasets were derived from the seven lakes. The data for each lake were sampled at and during different time periods. The data were kindly provided by the state water authorities responsible for lake monitoring as well as from limnological research institutes, namely the LUBW for Lake Constance, the Wasserversorung Zürich and EAWAG for Upper and Lower Lake Zürich, Lake Walen and as well as for Lake Lucerne, SECOE for Small Lake Geneva, the CIPEL for Big Lake Geneva, the FEM-IASMA for Lake Garda and the CNR-ISE for Lake Maggiore. Figure 1 shows the geographical position of the considered lakes related to the alpine arc.

Figure 1: Geographical representation of the peri-Alpine region and the lakes which were taken into account during the thesis, with the length of the timeseries available for each lake. ZH is the abbreviation for Lake Zürich, while GE is the abbreviation for Lake Geneva.

23

The seven lakes are all deep, warm monomictic lakes (Hutchinson, 1957), belonging to the same geographical (peri-Alpine) and climatological (continental) region. Moreover, based on total phosphorus concentrations, and according to OECD (1982), the assessed lakes cover the entire trophic gradient, reaching from oligotrophic to eutrophic lakes. An altitudinal gradient could be derived from the highest lake situated at 434 m. a. s. l. (Lake Lucerne) to the lowest situated lake at 65 m. a. s. l. (Lake Garda). Figure 2 highlights the lake features with respect to the trophic state, the altitudinal position and the length of the time series. The hydro-morpho-metrical characteristics, and the time period during which the data were collected, figure in in chapter 3, table 1. Lake Lucerne is not included for reasons which will be explained below.

Figure 2: Lake data sets represented upon their trophic level, their altitudinal position, and the length of the timeseries, which is indicated propotionally by the length of the black bar. (O = oligotroph; O-M= oligo-mesotroph; M = mesotroph, M-E= meso-eutroph; E= eutroph).

However the matrix evolved and had to be adapted in advancing the objectives. To achieve the first objective (I), Big Lake Geneva, Lake Garda, and Lake Maggiore were not considered, as some data were not checked for homogeneity. To attain objective II the original matrix was used as the base, with the exception of Lake Lucerne. This lake could not be taken into consideration in this research because of lack of values, and a non-fitting between the sampling dates, which was a considerable loss in the matrix, as the data from this lake were quite precious, representing 17 years of already homogenized phytoplankton data from the highest-located lake of the matrix. Finally for objective III, we considered only Big Lake Geneva, based on the results obtained from

24 the second objective. The dataset for Lake Geneva was enhanced by data derived from the sampling period 2001-2008, as these years exhibited the best fit for the purposes of the present investigation. By this time, Big Lake Geneva recovered through oligotrophication to a meso- eutrophic lake. Further details are given in chapter 4.

The choices for potential drivers to explain the phytoplankton community configuration were made upon recent findings from the literature but strongly depend on the availability of data for each lake. It was essential, that only drivers, for which the same sampling strategies, standard measuring methods and no missing values could be guaranteed, were considered in order to ensure homogeneity of the data among the different datasets.  To evaluate if lake hydrological and morphological variables might have a potential influence on phytoplankton community (Blenckner, 2005), these variables were integrated in some of the analyses.  With the exception of Big Lake Geneva, Lake Garda, and Lake Maggiore, meteorological variables were downloaded from the Swiss Federal Office for Meteorology and Climatology (MeteoSwiss). For Lake Garda, the data were obtained from the meteorological station of Arco (ARC), at the northern border of the lake (Salmaso, 2010), for Lake Maggiore data measured at the meteorological station in Verbania-Pallanza (Ambrosetti and al., 2006), and for Big Lake Geneva from the meteorological station situated at Thonon-les-Bains in the southern part of Lake Geneva, and belonging to the INRA Research Institute (Quétin, 2004). The daily North Atlantic Oscillation (NAO) index was downloaded from the National Oceanic and Atmospheric Administration (NOAA) (http://www.cpc.ncep.noaa.gov/products/precip/CWlink/pna/nao.shtml).  Physical-chemical variables were sampled at the same day as the phytoplankton was collected. Data from discrete depths (0 m, 1 m , 2.5 m, 5 m, 7.5 m, 10 m, 12.5 m, 15 m, 20m) in Lake Walen, Upper and Lower Lake Zürich) were transformed into weighted averages and integrated over the water column from 0-20 m depth. As well the stratification length (month) and the water columns stability (using the local Brünt- Väisälä index) were included.

 As light related drivers, the euphotic depth (zeu) and the ratio between the mixing depth and the euphotic depth were chosen (zm /zeu), whereas zm stand for the depth of the mixed layer.  As proxy for the grazing pressure the Zooplankton represented by the Cladoceras and the Copepodes (without nauplii) was used. Furthermore, the numbers of genera present during the same sampling day as well as the total phytoplankton biomass at the previous sampling date were integrated as a proxy for competition

25

1.5.1.1 The phytoplankton dataset

The compilation of the phytoplankton matrix involved an enduring and meticulous work, which received a lot of attention and importance as it represented the core of this thesis. For all the monthly or bi-monthly (depending on the timing of the year) sampling data, the different institutes employed the same counting method designed by Uthermöhl 1958. The samples resulted from an integrated depth from 0-20m, with exception of Lake Geneva, for which until the year 2000, phytoplankton samples were obtained for depths from the first 10m. However comparative analysis with the 0-20m did not show any significant differences (CIPEL, 2003), therefore allowing the integration of these samples in the final matrix.

The principal challenge was due to the fact that the species nomenclature changed considerably during the assessed periods, and moreover was identified by different phytoplankton taxonomists. Therefore each species identified had to be homogenized for the latest nomenclature. This was done following the more recent monographs of the series ‘’Süsswasserflora von Mitteleuropa” established by A. Pasher (Gustave Fisher Verlag) as well as by help of the “Freshwater algal for of the British Isles” (John and al., 2002). Further to reduce possible counting biases between the different institutions, it was decided to sum the species biomass to the genera biomass, which was standardized and expressed in µg/L.

Figure 3 represents the biomass of the different phytoplankton groups present in the studied peri-Alpine lakes (3a), as well the distribution of the principal orders of cyanobacteria (3b). In peri-Alpine lakes the principal phytoplankton groups are presented and clearly dominated by diatoms. Oscillatoriales are the most present orders among the Cyanobacteria. In Lake Walen and Upper Lake Zürich only weak biomass of cyanobacteria were recorded.

Figure 3 represent the biomass of the different pigmentary groups of phytoplankton (3a), in figure 3b represent the different orders of cyanobacteria among the assessed peri-Alpine Lakes.

26

The analyses performed to define the main drivers influencing the phytoplankton/ cyanobacteria community was based upon the Morpho-Functional Groups classification, referred to as “MFG” and defined by Salmaso and Padisàk (2007). According to this classification, the genera were grouped following the criteria of motility, specific nutrient requirement (autotrophy, mixotrophy), size, shape and presence of gelatinous envelopes (Weithoff, 2003). This choice was primary based upon the fact that MFG were already valuated for deep peri-Alpine lakes and second that these functional classification represent a new innovative promising method to assess the phytoplankton community. The classification of the MFG is illustrated in the additional material in chapter 3.

Annexe 2 represents the final phytoplankton matrix, in which all genera identified in the assessed sub-Alpine lakes were listened, as well as their taxonomic classification and their subdivision into their respective Moropho-Funcional Groups.

1.5.2 Methods

1.5.2.1 Today’s extreme events as a proxy for future climate

Climatic effects on lakes (and ecosystems in general) might only be detected when long-term data (at least 20 years) are available, such as proxy data (sediment analyses, see for example Smol & Cumming 2000) or adequate and long-term monitoring. Similar conclusions were reached by Vincent (2009) indicating that paleolimological methods are required to understand the potential impact of future climate change on lakes, in order to obtain a long-term time series. However, complete time series beyond 20 years are very difficult to obtain in limnology, which is due to the fact that phytoplankton was not a major concern and therefore not investigated regularly in all lakes. Moreover, phytoplankton is not the only element that exhibits problems of data availability; the other required environmental factors related to the phytoplankton growth are often missing, or sparse, or available for different times.

To avoid the problem of missing long-term datasets, another method than time series analyses was also considered. An appropriate solution was hypothesized by using extreme events of air temperature under current climate as a proxy for future mean climate change and thus allows investigating its potential impact on phytoplankton. Luterbacher and al. (2005) stated that summer 2003 was considered as an ‘’extreme” summer, in which locally temperatures were up to 10°C above the long-term summer average. For future climate, climate models suggest that one summer in two may be at least as hot as the 2003 event (Beniston, 2007) by the end of the 21st century. In other words, what is today considered a hot extreme event could become the “normal” (average) climate in a "greenhouse atmosphere" by 2100. The use of current data on extremes as

27 a form of proxy for the future thus enables an assessment of the possible impacts on the natural environment (Beniston, 2007). The method of air temperature extreme events was applied twice in this thesis, to attain objective I, and second time to reconfirm this method, in applying it on a different dataset (Big Lake Geneva, objective III).

To define whether an event is extreme or not, the distribution of air temperature during a standard reference period was studied. A reference period is defined in climatology as a time interval of 30 years which averages are characterizing the present 'normal' climate conditions over a certain area (IPCC, 1994). It should be representative of the present-day or recent average climate in the study region and of a sufficient duration to encompass a range of climatic variations, including several significant weather anomalies (e.g., severe droughts or cool seasons). A popular climatological baseline period is a 30-year "normal" period, as defined by the World Meteorological Organization (WMO). The current WMO normal periods are 1961-1990, and 1971-2000, which provide standard references for many impact studies. Once the distribution for air temperature for the reference period was selected, the 1st and the 99th percentiles were calculated, representing the limit values for extreme events. Therefore, the percentiles of the reference period were further compared to the air temperature registered at the meteorological station for the lakes investigation. If today’s temperature were to be either above the 99th or below the 1st percentile, an event would be considered as an extreme warm or extreme cold event, respectively. This is illustrated in figure 3.

Figure 3: Definition of extreme events

1.5.2.2. Non-metric Multidimensional Scaling (NMDS)

NMDS is a descriptive statistical method in the area of multivariate data analysis. The NMDS was employed to define the main drivers and how they influence the phytoplankton community composition. (cf. Chapter 2). NMDS is an ordination technique that differs in several ways from nearly all other ordination methods, and is the most appropriate tool to define which factors are

28 implicated in changes of species composition. This technique is applied on dissimilarity matrices of species (Legendre & Legendre, 1998), in which samples having similar species composition are positioned close together, based on monotonic regression methods. The goodness-of-fit of the regression is then calculated and expressed by a stress factor. In a second step, the interpretation of the NMDS configurations based on the environmental variables was carried out by applying a vector fitting (Oksanan, 2011). Vector fitting methods find the maximum correlation of single variables with the set of phytoplankton configuration. The fitted vectors point to the direction of most rapid change of the environmental variable inducing a gradient on the ordination of the species composition, and therefore not only indicate the most significant variable responsible for the phytoplankton composition, but also in which manner this variable is able to influence it. The strength of the relationship of the fitted vectors is expressed through the correlation coefficient, whereas the significances are based on random permutations of the data.

1.5.2.3 Models

One of the major roles in ecology is to respond to the problems encountered by society when faced with a range of environmental issues. Ecology must therefore attempt to produce tools to describe, understand and predict the future behavior of our environment (Peters, 1991). In order to do so, models are developed which synthesize the knowledge acquired and related to the function of biotic communities with the aim of predicting their evolution.

There are two distinct types of model which can be applied to the prediction of phytoplankton biomass, namely dynamic models and statistic models. Dynamic models describe the processes involved and take into account their temporal evolution to model the phenomenon of interest. A statistical model corresponds to a mathematical formulation of a relationship or a series of relations between variables required to estimate or test the parameters. In other words, these are mathematical tools allowing a description and/or simplified representation of living systems (Legendre et Legendre, 1998). Dynamic models allow to achieve a high level of prediction of modeled phenomena but require time-consuming calibration of the model parameters and an important amount of measurements which are not always available. Statistical models, also easier to implement on new sites, have proved to be interesting alternatives and have already been described to enable good levels of prediction of the temporal dynamics of phytoplankton blooms (e.g. Jeong and al., 2006).

Many statistical methods are available for the construction of predictive models. The following two models were employed in this thesis:

 A mixed-effect model was applied to test whether extreme air temperature events have a significant influence on the phytoplankton community (cf. chapter 2). This model was chosen because of its capacity to handle unbalanced data, but also because of its ability to take into account the variability of the phytoplankton response for each lake in applying a

29

random effect. The mixed effect model was applied with the help of the lme4 package (Bates & Maechler 2009) in the R software (R Development Core Team, 2012).  MARS (multi-adaptive regression spline)-model was used to predict the biomass of Planktothrix rubescence in Big Lake Geneva (cf. chapter 4). The interest to use this model was its efficiency of prediction compared to other regression models (Elith and al., 2006) but also to test its capacity for the first time in the prediction of the temporal dynamic of phytoplankton biomass. MARS models represent a similar philosophy to the widely used GAM approaches, with the difference that it consists of a number of sections (hinge functions) leading to an over-paramerization of the model. The model is then simplified by a cross-validation procedure based on binary recursive partition of the data. This procedure allows selecting the terms that need to be retained to form the final model. The model was run under R, with the MARS function implemented in the mda-package (Hastie&Tibshirani 2001) and additional codes written by Leathwick and al (2005).

30

Chapter two

2. Impacts of extreme air temperature on cyanobacteria in five deep peri-Alpine lakes Nicole Gallina, Orlane Anneville and Martin Beniston

Paper published in Journal of Limnology. 70 (2): 186-196. 2011.

DOI: 10.3274/Jl11-70-2-04

31

Impacts of extreme air temperatures on cyanobacteria in five deep peri-alpine lakes Nicole Gallina*1, Orlane Anneville2 and Martin Beniston

1Climate Change and Climate Impacts Research Group, Institute for Environmental Sciences (ISE), University of Geneva; 7 rte de Drize, CH-1227 Carouge – Switzerland. 2Orlane Anneville, INRA, UMR CARRTEL, Stn Hydrobiol Lacustre, 75 Ave Corzent, F-4203 Thonon Les Bains, France. *e-mail corresponding author: [email protected]

ABSTRACT Cyanobacteria are of major interest in freshwater ecosystems, since they are able to produce toxins with potentially negative impacts on the environment, health and thus on economics and society. It is therefore important for water management authorities to assess the manner in which cyanobacteria may evolve under climate change, especially in the alpine region where warming is projected by climate models to be more important than the global average. In this study, air temperature extremes under current climate were used as a proxy for future “average” climate forced by enhanced greenhouse gas concentrations. The impacts of extreme temperature events on cyanobacteria were analyzed in five deep peri-alpine lakes, covering the entire trophic gradient and using a synoptic approach. Extreme air temperatures were observed to alter the biomass of the cyanobacteria community. In general, extreme hot events are associated with high biomass while extreme cold events are characterised by low biomass. However, the assessed air temperature extremes did not lead to a dominance of cyanobacteria over the other phytoplankton groups, which also showed responses in relative biomass change during extreme events. These results suggest that the critical threshold, inducing a community change, was not attained in the studied period. Both extreme hot and extreme cold events were seen to generate a loss of diversity among cyanobacteria. In addition, the use of extreme events as a proxy to “average” future climates is a useful approach to enhance possible impacts of future global warming on the biota in freshwater systems. The outcomes of a synoptic approach provide general responses and are a useful tool for further modelling purposes. Keywords: Climate change, Biodiversity, Phytoplankton, Meteorological forcing, synoptic study.

INTRODUCTION Under current warming trends at the global and regional scales, extreme events such as heat waves are becoming more frequent (IPCC, 2007). The summer of 2003 was probably the hottest summer in Europe in the past 500 years, where locally temperatures were 5 °C above the long-term summer average (Beniston, 2004; Luterbacher and al., 2004). For future climate, climate models suggest that one summer in two may be at least as hot as the 2003 event (Beniston, 2007) by the end of the 21st century. In other words, what is today considered a hot extreme event could become the norm in a “greenhouse climate” by 2100. The use of current data on extremes as a form of proxy for the future thus enables an assessment of the possible impacts on the natural environment (Beniston, 2007). Cyanobacteria, which are a pigmentary group of phytoplankton, can be found in all aquatic ecosystems, ranging from hydrothermal springs to Arctic zones. Cyanobacteria are of importance since they are the major phytoplankton group in freshwater ecosystems capable of producing toxic blooms (Carmichael and al., 1990). Such toxins can become a major problem for public health if contaminated water is stored in reservoirs, and used in irrigation, fishing, cultivation and recreational purpose (Chorus, 2001; Codd, 1995; Codd and al., 2005). As the Earth’s oldest known oxygen-producing organism (Schopf, 2000), cyanobacteria have played a key role in evolution since their first appearance 2.15 billion years ago (Ramussen, 2008; Knopf, 2006; Hoffmann, 1975). The long life history of cyanobacteria is responsible of their capacity to be well adapted to environmental stress including scarce and abundant nutrients (Paerl, 2006), exposure to UV, high solar radiation and above all high temperatures (Paerl and al., 1985; Robarts & Zohary, 1987; Briand, 2004). These particular environmental conditions may favour the dominance of cyanobacteria in many aquatic habitats. The specific ability of cyanobacteria to be very tolerant when subjected to various stress factors suggest that cyanobacteria are likely to benefit from environmental changes associated with global warming (Paerl & Huisman, 2008; Paerl 2009).

32

These finding also lead to questions related to the capacity of cyanobacteria to become dominant under future climate regimes. It seems that under warming in shallow non-stratified lakes, the phytoplankton community is not changed and its diversity is maintained (Moss and al. 2003) whereas Elliot and al. (2006) suggest that the potential of cyanobacteria to dominate the community is greatest with higher water temperatures combined with more elevated nutrient loads. It has been hypothesised that the proliferation of cyanobacteria may be more important under the warmer climatic conditions projected for the coming decades of the 21st century. A further paradigm is that the impact of warming may lead to an alteration of the phytoplankton development (Paerl & Huisman, 2009; Paul, 2008), and therefore to a possible change in the community structure. Warming patterns directly influence air temperature, which in turn represents an indirect link between atmospheric conditions and aquatic ecology. Shifts in air temperatures are probably the first step in a causal chain of processes, which leads to hydrological and ecological changes in lakes (Straile, 2000), as for example warmer water temperatures (George & Hewitt 1999, Staile 2000), a more stable water column, a stronger and longer- lasting stratification with probable impacts on the availability of nutrients in the epilimnion (Perroud and al., 2009) as well as in the hypolimnon (Søndergaard and al. 2003). Mean lake temperatures have been increasing in responses to warmer weather. However, precise predictions of all possible responses of lake temperatures are difficult because of the complex interaction between aquatic environments and the atmosphere. Data indicate that in some cases, there may also have been direct responses of ecosystem functions to a warming climate. (De Stasio and al., 2009). The model results of Perroud & Goyette (2010) on the prediction of the water-temperature profiles of Lake Geneva indicate that there will be a rising of epilimnic temperatures corresponding to 55%- 98% of the monthly increase in air temperature under the IPPC 2A scenario. It is well known that multitude of factors affecting the growth of cyanobacteria that exhibit a complex and synergistic relation to each other (Dokulil and Teubner 2000). An important driver for phytoplankton growth is nutrient load, that itself can be partly driven by changes in inflow related to runoff in rivers upstream of the lakes. While changes of nutrient loads are not the main focus of the present paper, this topic will be briefly addressed at the discussion part. Because of the numerous impacts that warmer air temperatures is capable of exerting on physical, chemical and biological factors in lake ecosystems, the aim of this study is firstly to assess whether extreme air temperatures recorded during past years have been able to affect the biomass of cyanobacteria. If this is the case, the second goal consists to analyse, if the changes in the cyanobacteria biomass has the potential to influence the diversity of the phytoplankton community. The principle idea was not only to have the response of cyanobacteria in one lake, but to analyse their response in several lakes, in order to obtain a broader picture of what is taking place. In order to achieve this intention, data from five lakes was used; they are all deep lakes located within the peri-alpine region. The five lakes have different trophic status and cover the entire trophic gradient from eutrophic to oligotrophic.

MATERIALS AND METHODS

2.1 Locations and data description

For this study, phytoplankton data from five, deep peri-alpines lakes were analysed (Lake Constance, Lake Zurich, Lake Walen, Lake Lucerne and Lake Geneva). Lake Lucerne was split into two different data sets, as the sampling point for Lake Lucerne changed from Kreuztrichter (used until 1997) to Obermatt (since 1998). The geographic locations of these lakes are shown in figure 1. Table 1 summarises the hydro-morphological characteristics, the trophic level as well as the time period used for each lake. All the water bodies sampled in this study are deep, monomictic lakes and belong to the same geographical (peri-alpine) and climatological (continental) region. These common characteristics enable all lakes to be studied together. The trophic state of each lake was derived from the classification of the total phosphorus concentration according to the Organisation for Economic Cooperation and Development (OECD, 1982). Oligotrophic lakes have a total phosphorus concentration lower than 10 mg m-3, mesotrophic lakes are classified as having a total phosphorus concentration in between 10-35 mg m-3 and if the lakes reach a total phosphorus concentration in the range 35-100 mg m-3, they are considered eutrophic. The lakes cover the entire trophic gradient, which provides an opportunity to assess the overall phytoplankton response (Winder & Schindler, 2004) to the impact of extreme air temperature in this particular region, whatever the trophic state of the lake. The conclusions drawn for deep lakes may not necessarily hold for shallow lakes, however.

33

Figure 1: Map of Switzerland indicating the lakes and the meteorological stations referred to in the text. See Tables 1 and 2 for the names and characteristics of the selected lakes.

2.2 Phytoplankton data

Monthly or bi-monthly phytoplankton sampling data were obtained from state water authorities responsible for lake monitoring and from limnological research institutes. Table 2 gives an overview of the survey methods and sampling points applied during the different monitoring programs. All institutes employed the same counting method designed by Utermöhl (1958). Due to the fact that the species nomenclature changed during the periods under consideration, species identification followed the more recent monographs of the series “Süsswasserflora von Mitteleuropa” established by A. Pascher (Gustav Fischer Verlag, and Elsevier, Spectrum Akademischer Verlag) as well as “The freshwater algal flora of the British Isles” (John and al., 2002). To further reduce the counting biases between different institutions, the species biomass was added to the genera biomass, which was finally expressed in μg L-1. For further statistical analyses, we attempt to distinguish between the well-known genera able of producing toxins and the genera for which there have been very sporadic or no records of toxicity up till now. We divided the cyanobacteria genera in two categories: the first includes the total cyanobacteria biomass and the second the potentially-toxic cyanobacteria biomass. This allowed to study separately the cyanobacteria biomass responsible for generating potentially critical situations for water management. The key genera known for their potential ability to produce toxic substances include Anabaena, Aphanizomenon, Coelosphaerium, Oscillatoria, Plankthothrix, Pseudoanabaena, Synechococcus, Woronichinia (Carmichael, 2001). The Cyanobacteria genera presented in the lakes investigated here and not known to produce any toxins are: Aphanocapsa, Aphanothece, Chroococcus, Cyanobacterium, Cyanothece, Dactylococcopsis, Gleitlerinema, Gomphosphaeria, Jaaginema, Leptolyngbya, Limnothrix, Lyngbya, Merismopedia, Pannus, Phormidium, Planktolyngbya, Pleurocapsa, Rhabdogloea, Snowella, Spirulina, Synechocystis.

34

The uppermost water layers from 0-20 m were taken into account in the present study. Data from discrete depths have been transformed into weighted averages integrated over the water column from 0 – 20 m depth.

Reten- Mean Max Time Trophic Altitude Surface Volume tion Lake Name Depht Depht Period state (m) (km2) (km3) Time (m) (m) (yrs) 1985- Constance Eutroph 395 472 101 252 48 4.3 1990 1981- Meso- Lower Zurich 406 65 51 136 3.3 1.4 1990 Eutroph 1991- Oligo- Walen 420 24 100 145 2.42 1.4 2000 troph Oligo- Lucerne 1987- Meso- 434 113 104 214 11.8 3.4 Kreuztrichter 1996 troph Oligo- Lucerne 1998- Meso- 434 113 104 214 11.8 3.4 Obermatt 2004 troph Oligo- Small 2001- Meso- 372 81 41 76 3 0.4 Geneva 2005 troph Table 1 : List of lakes studied in the present paper, summarized by their time period, trohphic levels and hydromorphological caracteristics.

Lake Institution Sampling point Sampling Depht Name

LUBW (Landesanstalt für Umwelt, Messungen FU (Fischbach- Constance und Naturschutz Baden- Wurtenberg; Institut für 0-20 m Uttwil) Seenforschung; Langenargen 0 m, 1 m ,2.5 m, 5 m, Lower Wasserversorgung Zürich TH (Thalwil) 7.5 m, 10 m, 12.5 m, Zürich 15 m, 20m 0 m, 1 m ,2.5 m, 5 m, Walen Wasserversorgung Zürich MU (Murg) 7.5 m, 10 m, 12.5 m, 15 m, 20m KZ EAWAG (Swiss Federal Institute of Aquatic Lucerne (Kreuztrichter), 0-20 m Science and Technology ) OB (Obermatt)

Small SECOE (Service Cantonal de I'Ecologie de GE 3 0-20 m Geneva l'Eau) ;Geneva

Table 2 : Summary of the phytoplankton survey in the different lakes.

35

2.3 Meteorological data

The daily average air temperature data were downloaded from the digital database of the Swiss Federal Office for Meteorology and Climatology (MeteoSwiss). Five meteorological stations have been chosen (Fig. 1), based upon their proximity to the phytoplankton sampling points, namely Güttingen (GUT) and Kreuzlingen (KRZ) for Lake Constance; Zürich/Fluntern (SMA) for lower Lake Zürich; Glarus (GLA), for Lake Walen; Lucerne (LUZ) for Lake Lucerne and Changins for the the narrow western segment of Lake Geneva (“Small Lake Geneva” or “Petit Lac”). The dataset has been quality-checked for homogeneity in the records (Bergert and al., 2005). Cross- correlation analyses based on phytoplankton distribution showed that the best responses to meteorological data were obtained when taking the temperature values on the fifth day before the sampling day. Therefore, air temperatures five days prior to plankton sampling were used, in order to respect the response time of phytoplankton to meteorological events.

2.4 Statistical Methods

Extreme air temperature (hereafter referred to as “ET”) enabled the derivation of a number of sample classes (ET-Classes). Each phytoplankton sampling date was recorded into 3 different classes, each of which was defined with the help of percentiles. For each lake, the monthly 1st percentile and the 99th percentile of air temperature were computed for the common reference period from 1961-1990. If the corresponding air temperature of the sampling date was between the 1st and the 99th percentile of the reference period, the sampling date was classified into the “normal” or “non-extreme” class. For the case where the air temperature was below the 1st percentile of the reference period, the sampling date was classified as an extreme cold sample; conversely, extreme hot samples were dates for which the air temperature exceeded the 99th percentile. For the total of 705 samples among the lakes, we classified 419 events into “normal” events, 193 samples into “extreme hot” events and 93 samples into “extreme cold” events. Use of the percentiles rather than the two original sets of data provide a measure of objective interpretation of what may constitute a cold or a warm extreme, that can be transposed from one lake to another even if the true temperature values may be different for these selected thresholds. Due to a sparsity of data for Lake Constance, two meteorological stations were taken into account to compute the reference period. From 1961 to 1975 the air temperature was downloaded from the station of Kreuzlingen (KRZ) and from 1976 to 1990 from Güttingen (GUT) (Fig. 1). In order to assess the impact of extremes on the seasonal cyanobacteria biomass, the average of the sampling months corresponding to the same seasons was calculated, within the lakes and affiliated to the same ET-Class. The cyanobacteria biomass for each ET-Class and each season was derived, thus representing a form of “virtual, season” in peri-alpine lakes. The winter season included samples from December, January and February; spring was defined from March to May, summer from June to August and autumn from September to November.

2.5 Model

A linear mixed-effect model was used to test the influence of temperature and the significance of the different ET-Classes on cyanobacteria biomass and on phytoplankton community diversity. For this purpose we employed the “lmer” function from the “lme4” package (Bates & Maechler 2009) in the R software (R Development Core Team, 2009). The “lme4” Model is very robust when using S4 classes (Chambers, 1998). The S4 class is an object oriented programming originally used by the S-Language and adopted by the R software. Contrary to the old S3 class system, the S4 object system is much richer, thus allowing an easier implementation of functions and is robust enough to support real software engineering. Applying S4 objects, classes and methods become much more formal and rigorous. The “lme4” Model is based on the restricted maximum likelihood estimation method and able to handle unbalanced data. ET-Classes were defined as model predictors, while the total cyanobacteria biomass, the potentially toxic cyanobacteria biomass as well as diversity were defined as the response variables. As each lake has its specific effect on the phytoplankton growth as does water chemistry and environmental factors (Ryding & Rast 1989; Reynold & Walsby, 1975; Lung & Perl, 1988; Blenckner 2005), the lake identity was set as random effect’, which allows the model to take into account the variability that can exist between the different lakes.

36

The significance (p-value) of the fixed effects was tested with the Markov Chain Monte Carlo (MCMC) method on 10,000 permutations.

2.6 Biodiversity

The goal of using biodiversity was to assess possible changes in the phytoplankton community as a result of impacts related to extreme air temperatures. As the outcomes of several studies indicate changes of biodiversity in lakes ( Sala and al., 2010; Magnusson 1997 ), this raises the question of whether the phytoplankton community analysed in the sampled lakes result in a loss or in a gain in biodiversity, as there are opposite number of counteracting effects that can occur (Heino 2009). Indeed, a dominance of cyanobacteria may well occur as a result of climate change (Wagner & Adrian, 2009; Elliot 2006), but it is still unclear whether these changes may lead to a loss (Elliot. 2006) or to a gain in diversity (Magnuson 1997). As different groups are likely to respond differently to temperature change (Moss, 2003), we calculated the genera diversity (H”) among phytoplankton groups, using the Shannon-Weaver function (Shannon & Weaver 1949) implemented in the ‘’vegan’’ package (Oksanen, 2009) available from the R Software (R Development Core Team, 2009),

) H'' (bi B)log2(bi B)

where B is the total biomass of each phytoplankton group, bi is the biomass of the ith genera present. The diversity index was applied to the main phytoplankton groups present in the studied lakes, which included in addition to Cyanobacteria, Chlorophyceae, Diatoms, Dinophyceae, Chrysophyceae, Conjugatophyceae and Cryptophyceae. The monthly biomass average was thus derived for each lake and each phytoplankton group. The effect of each ET-Class, as well as its significance, to phytoplankton group diversity was calculated in using the “lmer” function implemented in the “lme4” Model. The lake identity was used as random effect. ET-Classes were defined as predictors and biodiversity as the response variable.

37

Figure 2: Mean and standard error for each ET-Class derived from A) the log-transformed total cyanobacteria biomass (Log (TotCyano)) and B) the log-transformed total toxic cyanobacteria biomass (Log (TotCyanoToxic)).

RESULTS

3.1 Effects of extreme air temperatures on the total and on the potentially toxic cyanobacteria biomass

Each ET-Class has a highly significantly effect on total and potentially toxic cyanobacteria (Tab. 3). Warm temperatures have a positive effect, inducing an increase in biomass while cold temperatures lead to a decrease. The impact of these effects on cyanobacteria biomass can be visualised with the help of the bar graphs in Figure 2. Bar graphs highlight the mean and the standard error of all samples belonging to the same ET-Class. It appears that hot events are characterised by significantly higher cyanobacteria biomass, while cold events are characterized by lower biomass. Graphs A and B in Figure 2 show an increase of the same order of magnitude for both total and potentially toxic cyanobacteria in relation to air temperatures.

38

Total cyanobacterial biomass (µg/l) Total potentioally toxic cyanobacterial biomass (µg/l)

Estimates p-value Estimates p-value

Hot 3.055 0.001 2.497 0.001

Cold -0.961 0 -0.794 0

Normal -0.447 0.006 0.332 0.001

Table 3: Significance of the different ET-Classes (Hot, Cold, Normal) to the total and the potentially-toxic cyanobacteria biomass, evaluated by the “lme4” Model. The significance of the model was tested with help of the MCMC method (p-value). All ET-Classes are significant for values of p < 0.05.

3.2 Distributions of temperature inducing a change in the development of cyanobacteria

The box-plot graphical method was used to display the distribution of air temperature measured during the periods of phytoplankton sampling, grouped by class and seasons. The aim was to define the range of recorded air temperatures that are capable of inducing an increase in cyanobacteria biomass. For this reason, we focused on measurements for which we identified a high rate of proliferation of cyanobacteria during extreme events, as was the case for hot extremes in general, and more specifically during hot summers and autumns, as well as for cold winters. We compared these temperature spreads with air temperatures recorded during normal events (Fig. 3). During extreme hot events, the recorded air temperature had a median value of 14.8 °C, and 50% of the values were distributed between 9.1 °C and 19 °C. On the other hand, samples within the normal classes exhibited an air temperature distribution with a median value of 9.9 °C and an inter-quartile range between 3.3 °C to 15.9 °C. In summer, the median air temperature for normal events and extreme hot events was 17.4°C and 21.3°C, respectively. Under normal condition, the 25th percentile was 16.3 °C and the 75th percentile was 18.6 °C, while under extreme hot summer conditions, these percentiles increased to 20.1 °C and 22.5 °C, respectively. In autumn, temperatures tend to be more variable than in summer, such that the inter-quartile range under normal conditions was between 6.9°C and 14°C, and the median air temperature was 10.9°C. Hot extremes ranged from 9.6 °C to 17.4 °C, with a median value of 12.7 °C. Strong cyanobacteria proliferation has also been observed even during cold winter extremes. The air temperature has a median value of -6.1 °C, with a small inter-quartile range between -6.4 °C and -5.3 °C. Normal air temperatures in winter, on the other hand, exhibit a median value of 1°C, where 50% of the values were distributed between -0.7 °C and 2.7 °C.

39

Figure 3: A) Box-plot graphics displaying the median (thick line), the upper (75th percentile) and lower (25th percentile) quartiles, the minimum and maximum values from Normal (N) and Hot (H) ET-Classes. B) The seasonal air temperature distribution of Hot and Normal ET-classes for summer and autumn are also shown, as well as for Normal and Cold (C) winters.

3.3 Consequence of cyanobacteria biomass increase on the phytoplankton community

The relative seasonal biomass of ET-Classes was analysed taking into consideration the major phytoplankton groups. The aim of assessing the relative cyanobacteria biomass was to emphasize whether the cyanobacteria, which under normal conditions exhibit a low relative biomass contribution, have the potential to become dominant under extreme events (Fig. 4). Even though their biomass has been observed to increase under such events, cyanobacteria do not appear to dominate the other phytoplankton groups during extreme hot (mainly summer and autumn) and extreme cold (mainly winter) events. Moreover, it seems that the relative proportion of biomass between the different ET-Classes and within the same season remains similar. Nevertheless, the graphical

40 presentation of figure 4 indicates that the different algal groups change their relative biomass in different manners according to the assessed ET-classes and seasons.

100%

80%

60%

40% Log(Biomass) µg/L

20%

0%

C_Winter N_Winter H_Winter C_Spring N_Spring H_Spring C_SummerN_SummerH_Summer C_AutumnN_AutumnH_Autumn

Cyanobacteria Chlorophyceae Chrysophyceae Conjugatophyceae Cryptophyceae Diatoms Dinophyceae

Figure 4: Relative seasonal biomass of the major phytoplankton groups for each ET-Class (C = Cold; N = Normal, H = Hot).

Cyanobacteria Chlorophyceae Dinophyceae

6

10

4

5

8

4

3

6

3

2

4

2

Shannon diversityShannon index diversityShannon index diversityShannon index

1

2

1

0 0 0

c h n c h n c h n

Chrysophyceae Diatoms Cryptophyceae

8

7

2.5

6

2.0

6

5

4

1.5

4

3

1.0

2

2

Shannon diversityShannon index diversityShannon index diversityShannon index

0.5

1

0 0 0.0 c h n c h n c h n

ET-Classes

41

Figure 5: Average and standard error of the biodiversity of the major phytoplankton groups derived for each ET- Class (c = Cold; h = Hot, n = Normal).

3.4 Effects of extreme air temperature on phytoplankton community diversity

Biodiversity among phytoplankton groups change for the different ET-classes were investigated, with a special focus on cyanobacteria. Table 4 summarises the results obtained by means of the “lmer” method. Surprisingly, air temperature affected every phytoplankton group under extreme cold events. Air temperature was also able to significantly influence the biodiversity of all phytoplankton groups under normal conditions. This outcome indicates that biodiversity is statistically different, and significantly so, under normal conditions compared to cold extreme events, for all phytoplankton groups with the exception of Conjugatophyceae. However, an unexpected result of this study is that there is a significant loss in biodiversity only for cyanobacteria during extreme hot events. In order to measure the biodiversity change for every phytoplankton group among ET-classes, we used the bar graph method derived from the average of the Shannon-Weaver Index for each class among each phytoplankton group (Fig. 5). The Conjugatophyceae were not taken into account, because of the general lack of significant changes during normal as well as extreme hot temperature conditions. An overall loss of biodiversity in the phytoplankton community is seen to take place whenever extreme cold events occur compared to normal events. Caution is necessary, however, in the interpretation of the biodiversity effect within the warm ET-Class, because of the low p-values among the different phytoplankton groups. Nevertheless, if air temperature rises, biodiversity tends to be lower than under normal conditions. The cyanobacteria community showed on the other hand a significant loss of biodiversity for both of the extreme cases considered. Cyanobacteria seem to be the most vulnerable to changes in temperature. Figure 6 shows the same effects in summer, autumn and winter, where a loss of diversity occurs during extreme warm as well as during extreme cold events. Under the latter conditions, biodiversity decreased more than under the former one. In spring, the seasonal pattern is the reverse of the other seasons; cyanobacteria gain in biodiversity, for both extreme hot and cold spring events.

Cold Normal Hot Estimates Estimates Estimates p-value p-value p-value µg/L µg/L µg/L Cyanobacteria 3.323 0 1.347 0 0.809 0.016 Chlorophyceae 7.452 0 3.269 0 1.475 0.089 Cryptophyceae 2.446 0 0.209 0.005 0.127 0.075 Dinophyceae 4.223 0 1.31 0 0.519 0.09 Diatoms 6.666 0 1.412 0 0.345 0.29 Chrysophyceae 5.427 0 1.707 0 0.563 0.152 Conjugatophyceae 2.252 0 0.262 0.09 0.278 0.083 Table 4: Significance of the effect of ET-Class to biodiversity of the major phytoplankton groups evaluated by the “lme4” Model and tested with the MCMC method (p-value). The effects are significant for p<0.05 and are highlighted in bold.

42

Cyanobacteria Chlorophyceae Dinophyceae

6

10

4

5

8

4

3

6

3

2

4

2

Shannon diversityShannon index diversityShannon index diversityShannon index

1

2

1

0 0 0

c h n c h n c h n

Chrysophyceae Diatoms Cryptophyceae

8

7

2.5

6

2.0

6

5

4

1.5

4

3

1.0

2

2

Shannon diversityShannon index diversityShannon index diversityShannon index

0.5

1

0 0 0.0 c h n c h n c h n

ET-Classes

Figure 5: Average and standard error of the biodiversity of the major phytoplankton groups derived for each ET- Class (c = Cold; h = Hot, n = Normal).

DISCUSSION

The purpose of this study was to use extreme air temperatures as observed in recent years as a proxy for a warmer climate that is projected to occur over the next 50-100 years (e.g., Beniston and al., 2007; IPCC, 2007), in order to assess the response of cyanobacteria biomass and diversity in deep continental lakes. The altered response of the cyanobacteria community to the different ET-Classes suggests that air temperature has an impact on cyanobacteria biomass in deep peri-alpine lakes. Given the time lag of 5 days used in this study, which is considered to be the response time of phytoplankton to meteorological changes, it is suggested here that air temperature is able to affect phytoplankton biomass in a rather rapid way. The use of air temperature as a proxy for water temperature, in the uppermost meters of the lake waters and during summer, was also discussed by Livingstone (1998). Based on these findings, air temperature can be considered as a reliable indicator of biotic processes, such as cyanobacteria behaviour within the epilimnion. Similar results were found for Microcystis, in a eutrophic lake during the summer heatwave of 2003, by means of a lake experiment and with the help of a biological-physical model (Jöhnk and al., 2008). The conclusions of that work suggest that air temperature directly affects cyanobacteria through an increased growth rate, and indirectly via the stabilisation of the water column favouring buoyant cyanobacteria over other non-buoyant phytoplankton groups. Other similar results were found in a subalpine lake in northern California through a long-term study by Park and al. (2004), which showed that cyanobacteria biomass increases during warmer years. Furthermore, it was concluded that higher water temperatures were accompanied by increasing cyanobacteria biomass in summer. De Senerpont and al. (2007) observed through experiments with mesocosms and mechanistic models that the growth rates as well the

43 abundance of cyanobacteria were higher during warm springs than cold ones, particularly in lakes whose conditions were phosphorus-limited. Nevertheless, the relative proportion of the different phytoplankton groups during extreme events is, unexpectedly, well maintained. It seems that a constant composition governs phytoplankton groups, which is not able to be disturbed by the extreme events recorded. Neither cyanobacteria nor any other phytoplankton group became dominant under extreme events nor was overall phytoplankton development over the season and among phytoplankton groups well maintained. Furthermore, atmospheric warming does not appear to affect any particular pattern in taxonomy (Moss and al., 2003). These findings suggest that the constant composition under both types of extreme events seems to be well maintained. This may be due to the fact the different genera respond in different ways to shifts in temperature (Heino, 2009; Moss and al., 2003). These altered responses are, however, not strong enough to disturb the phytoplankton composition. Using the findings by Elliot and al. (2006) on cyanobacteria dominance, it is suggested that for the present study, summer and autumn heat-waves probably did not reach critical thresholds beyond which a collapse of the constant composition that would favour the dominance of cyanobacteria over other species would be observed. Under extreme hot seasons, compared to normal season, median air temperature increased in autumn by 4°C, by 2.9°C in summer, and decreased by 5.1°C in winter. These air temperature differences are able to trigger an increase in cyanobacteria biomass; however, temperature changes alone are not sufficient to induce the domination of cyanobacteria. It can be hypothesised, however, that once these thresholds are exceeded, a possible community functional change is likely to take place. Biodiversity among the different phytoplankton groups is surprisingly affected more by cold extreme events than by warm ones. All of the major phytoplankton groups respond significantly to these cold extremes. A particular case relates to the Conjugatophyceae, whose diversity seems to change only under extreme cold events. A decrease in biodiversity can be expected under extreme cold events for all phytoplankton groups. Hot extreme events, in contrast, influence only cyanobacteria in a significant way. This emphasizes the point that cyanobacteria are perhaps the most sensitive phytoplankton group when it comes to biodiversity under hot temperature regimes. Although the shifts in the biodiversity of the other major phytoplankton groups under hot events are low, a decrease in phytoplankton biodiversity may occur under such conditions. These finding confirm Elliot’s and al. (2006) results, derived for only eight dominant species belonging to different phytoplankton groups. With higher air temperatures, cyanobacteria biomass increases, but its diversity decreases. This suggests that fewer genera lead to higher biomass, which is only possible if one or several genera begin to proliferate excessively. Similar results were obtained by Duarte and al. (2006), who studied the relationship between ecosystem functioning (production) and structure (biodiversity) in coastal lagoons, suggesting a decrease in diversity is associated with an increase in biomass and production. However, the impact of extreme warm air temperatures to overall phytoplankton community structure could not be ascertained from this study. This is because of the lack of a significant response of the major phytoplankton groups to heat waves, with the exception of cyanobacteria. Further work will be necessary to assess the questions related to the interaction between community structure and function. It should be stressed that nutrient loads are also a major driving force, in addition to water temperatures (Moss and al., 2003; Elliot and al., 2006, Stich & Brinker, 2010). In a future climate, however, warmer summer temperatures will not have the same impacts on discharge as they do under current climate, because up to 90% of current alpine glacier mass may be lost by 2100, according to future levels of greenhouse-gas emissions. With such reductions, and the projected seasonal shifts in precipitation towards much drier summers (e.g., Beniston, 2006), many rivers originating in the Alps may see sharp decreases in runoff compared to current climate (e.g., Beniston, 2010). Nutrient availability will clearly be influenced by these long-term, sustained changes in river discharge, thus requiring further investigations to include the influence of nutrient availability in addition to that of temperature discussed in the present paper. It is important to reveal that each lake is reflects its particular environment, which is defined in Bleckner’s (2005) conceptual model of climate-related effects on lake ecosystems as the “Internal Lake Filter”. It is, however, interesting to emphasize that the conclusions for different lakes under different trophic statuses, based on the different methods outlined in the literature, from in-situ studies, mesocosms experiments, and model simulations, all point to the similar response of cyanobacteria biomass increase with higher air and water temperatures. This common conclusion confirms the validity of the use of the synoptic approach applied in this study. Finally, this study shows that the use of extreme events is a reasonable proxy to assess future climate change, particularly in cases where time series are not available. As these extreme events are likely to become more common, it can be expected that the potentially toxic as well as the total cyanobacteria biomass should also increase with a corresponding loss in biodiversity. Such a conclusion should help to develop possible adaptation strategies that could help reduce some of the associated risks of climatic change (Beniston, 2007), in particular on water quality and human health.

44

CONCLUSIONS

Extreme air temperatures are a useful proxy to investigate cyanobacteria development in deep peri-alpine lakes in a changing climate, and could be applied as possible indicator of changes in biological processes in the upper layers of the water column. Extreme warm events favour a development of cyanobacteria biomass, whereas cold extreme temperatures tend to induce a decrease in biomass. However, even under conditions of increasing biomass during extreme conditions, cyanobacteria were not able to dominate the phytoplankton community. Surprisingly, extreme cold events also had a significant impact on the major phytoplankton groups investigated, resulting in an overall community diversity loss. Hot extreme events had a significant impact only on cyanobacteria. A loss of diversity could be expected for the cyanobacteria community under both extreme hot and extreme cold events. As more frequent extreme events are expected in a warmer global climate, an enhanced biomass increase and diversity loss among the cyanobacteria community can be anticipated. This study concludes that synoptic in situ approaches are appropriate tools for a general assessment of the impacts of meteorological forcing on similar lakes located in the same geographical and/or climatological region and provide necessary information for further investigations. It will be necessarily to improve the understanding of the strength of meteorological effects on phytoplankton behaviour, compared to the internal physico-chemical and biological influences, such as grazing pressure and competition. This understanding is of primary importance when it comes to model cyanobacteria development under scenarios of global warming.

ACKNOWLEDGEMENTS

The authors wish to thank N. Salmaso for his useful advice and S. Lavigne for helping us with the identification of the phytoplankton species. Special thanks go to the different institutes providing the phytoplankton data, namely: Dr. R. Kümmerlin from the LUBW, Dr. R. Forster from the Wasserversorung Zürich, Dr. H. R. Bürgi from the EAWAG Dübendorf, Dr. J. Perfetta from the SECOE from Geneva, as well as the CIPEL. In addition, the authors extend their thanks to the Swiss Office for Meteorology and Climatology (MeteoSwiss) for allowing the use of their meteorological data. This work was supported by the University of Geneva and contributes indirectly to the research objectives of the EU/FP7 “ACQWA” project (www.acqwa.ch) funded by the European Union under Grant Nr. 212250.

45

REFERENCES

Bates, D. & M. Maechler, 2009. lme4: Linear mixed-effects models using S4 classes. R package version 0.999375- 31. http://CRAN.R-project.org/package=lme4

Begert, M., T. Schlegel., & W. Kirchhofer, 2005. Homogenous temperature and precipitation series of Switzerland from 1864 to 2000. International journal of climatology 25: 65-80.

Beniston, M., 2004. The 2003 heat wave in Europe: a shape of things to come? An analysis based on Swiss climatological data and model simulations. Geophysical Research Letters, 31, L02202, doi: 10.1029/2003 GL018857.

Beniston, M., 2006: The August 2005 intense rainfall event in Switzerland: not necessarily an analog for strong convective events in a greenhouse climate. Geophysical Research Letters, 33, L5701

Beniston, M., 2007. Entering into the "greenhouse century": recent record temperatures in Switzerland are comparable to the upper temperature quantiles in a greenhouse climate. Geophysical Research Letters, 34, L16710.

Beniston, M., 2010: Impacts of climatic change on water and associated economic activities in the Swiss Alps. Journal of Hydrology, doi:10.1016/j.jhydrol.2010.06.046

Beniston, M., Stephenson, D. B., Christensen, O. B., Ferro, C. A. T., Frei, C., Goyette, S., Halsnaes, K., Holt, T., Jylhä, K., Koffi, B., Palutikoff, J., Schöll, R., Semmler, T., and Woth, K., 2007: Future extreme events in European climate; an exploration of Regional Climate Model projections. Climatic Change, 81, 71-95.

Blenckner, T., 2005. A conceptual model of climate-related effects on lake ecosystems Hydrobilogia, 533: 1-14.

Briand, J. F., C. Leboulanger & J. F. Humbert, 2004. Cylindrospermopsis raciborskii (cyanobacteria) invasion at mid-latitudes: selection, wide physiological tolerance, or global warming. Journal of Phycologie 40: 231-238.

Carmichael, W.W., N. A. Mahmood & E. G. Hyde, 1990. Natural toxins from cyanobacteria (blue-green algae). In Hall S., Strichartz G. (eds.) Marine toxins: origin, structure, and molecular pharmacology. American Chemical Society, Washington, d.C. Pp. 87-106.

Carmichal, W.W., 2001. A mini-review of cyanotoxins; toxins of cyanobacteria (blue-greenalgae), in W.J. de Koe, R.A Samson, H.P. van Egmond, J. Gilbert, M. Sabino and W.J. de Koe (eds.), Microcystin and Phycotoxins in Perspective and the Turn of the Millennium. Ponson &Looyen, Wageningen, The Netherlands, pp. 495-504.

Chambers J. M., 1998. Programming with Data - A Guide to the S Language, Springer-Verlag.

Chorus, I., 2001. Cyanotoxins: Occurrence, Causes, Consequences. Berlin: Springer-Verlag.

Codd, G. A., 1995. Cyanobacterial toxins: Occurrence, properties and biological significance. Water Science and Technology 32:149-156.

Codd, G. A., L. F. Morrison & J. F. Metcalf, 2005. Cyanobacterial toxins: risk management for health protection. Toxicology and applied Pharmacology 203: 264-272.

De Stastio B.T., T. Golemgeski, & D.M. Livingstone, 2009. Temperature as a Driving Factor in Aquatic Ecosystems. Encyclopedia of Inland Waters, 690-698 pp.

De Senerpont Domis, L. N., W. M. Mooij & J. Huisman, 2007. Climate-induced shifts in an experimental phytoplankton community: a mechanistic approach. Hydrobiologia 584: 403-413.

Duarte, P., M. F. Macedo & L. Cancela de Fonseca, 2006. The relationship between phytoplankton diversity and community function in coastal lagoon. Hydrobiologia 555: 3-18.

46

Elliot, J. A., I. D. Jones & S. J. Thackeray, 2006. Testing the sensitivity of phytoplankton communities to changes in water temperature on nutrient load, in a temperate lake.

George, D. G. & D. P. Hewitt, 1999. The influence of year-to-year variations in winter weather on the dynamic of Daphnia and Eudiaptomus in Estwaite Water, Cumbria.

Heino, J., R. Virkkala & H. Toivonen, 2009. Climate change and freshwater biodiversity: detected patterns, future trends and adaption in northern regions. 94: 39-54

Hofmann, H. J., 1976. Precambrian microflora, Belcher Islands, Canada: Significance and systematics. Journal of Paleontologie 50, 1040–1073.

IPCC, 2007: Climate Change 2007: Synthesis Report. Contribution of Working Groups I, II and III to the Fourth Assessment. Report of the Intergovernmental Panel on Climate Change [Core Writing Team, Pachauri, R.K and Reisinger, A.(eds.)]. IPCC, Geneva, Switzerland, 104 pp.

John, D. M., A. J. Brooks & B. A. Whitton 2002. The Freshwater Algal Flora of the British Isles. An Identification Guide to Freshwater and Terrestrial Algae. Cambridge University press. 702pp.

Jöhnk, K. D., J. Huisman, J. Sharples, B. Sommeijer, P. M. Visser and J. M. Stroom, 2008. Summer heatwave promote bloom of harmful Cyanobacteria, Global change biology 14: 495-512.

Knopf, A. H., E. J. Javaux, D. Hewitt & P. Cohen, 2006. Eukaryotic organisms in Proterozoic oceans. Philosophical Transactions of the Royal Society B 361: 1023-1038.

Livingstone, D. M. & A. F. Lotter, 1998. The relationship between air and water temperatures in lakes of the Swiss Plateau: a case study with palaeolimnological implications. Journal of Paleolimnology 19: 181-198.

Lung, W. S. & H. W. Pearl, 1988. Modelling green-blue algal blooms in the Lower Neuse River. Water Research 22: 895-905.

Luterbacher, J., D. Dietrich, E. Xoplaky, M. Grossjean & H. Wanner, 2004. European seasonal and annual temperature variability, trends, and extremes since 1500. Science 303: 1499 -1503.

Magnuson, J.J., K.E. Webster, R.A. Assel, C.J. Bowser, P.J. Dillon, J.G. Eaton, H.E. Evans, D.J. Fee, R.I. Hall, L.R. Mortsch, D.W. Schindler and F.H. Quinn. 1997. “Potential effects of climate change on aquatic systems: Laurentian Great Lakes and Precambrian Shield Region,” pp. 7-53 in C.E. Cushing (ed.), Freshwater Ecosystems and Climate Change in North America: A Regional Assessment. Advances in Hydrological Processes. John Wiley& Sons. (Also as an Issue of the Journal Hydrological Processes 11(6) 1997.)

Moss B., D. McKee, D. Atkinson, S. E. Collings, J. W. Eaton, A. B. Gill, I. Harvey, K. Hatton, T. Heyes & D. Wilson, 2003. How important is climate? Effects of warming, nutrient addition and fish on phytoplankton in shallow lake microcosms. Journal of applied Ecology. 40: 782-792.

OECD, 1982. Eutrophicaton of waters. Monitoring, assessment and control. Organisation for Economic Cooperation and Development. 193pp.

OFEV, 2009: Hydrological data online at URL: http://www.hydrodaten.admin.ch/f/index.htm?lang=fr

Oksanen J., R. Kindt, P. Legendre, B. O'Hara, G. L. Simpson, P. Solymos, M. H. Stevens & H. Wagner, 2009. Vegan: Community Ecology Package. R package version 1.15-3. http://CRAN.R-project.org/package=vegan

Paerl, H. W., P. T. Bland, N. D. Bowles & M. E. Haibach, 1985. Adaptation to high intensity, low wavelength light among surface blooms of the cyanobacterium Microcystis aeruginosa.

47

Paerl, H. W., 2008. Nutrient and other environmental controls of harmful cyanobacteria blooms along the freshwater-marine continuum. Advances in Experimental Medicine and Biology 619 : 216-241.

Paerl, H. W. & J. Huisman, 2008. Blooms like it hot. Science 320: 57-58.

Paerl, H. W. & J. Huisman, 2009. Climate change: a catalyst for global expansion of harmful algal blooms. Environnemental Microbilogy Reports 1(1), 27-37.

Park, S., M. T. Brett, A. Müller-Sorger, C. R. Goldmann, 2004. Climatic forcing and primary productivity in a subalpine lake: Interannual variability as a natural experiment. Limnology and Oceanography 49: 614-619.

Paul, V.J., 2008. Global warming and cyanobacterial harmful algal booms. In: K.H. Hudnell, Editor, Cyanobacterial Harmful Algal Blooms: State of the Science Research Needs Series: Advances in Experimental Medicine and Biology vol. 619. XXIV, 950 pp.

Perroud, M. & S. Goyette, 2009: Impacts of a warmer climate on Lake Geneva water temperature profiles. Submitted to Boreal Environment Research

R Development Core Team, 2009. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, ISBN 3-900051-07-0. http://www.R-project.org.

Rasmussen, B., I. R. Fletcher1, J J. Brocks & Matt R. Kilburn, 2008. Reassessing the first appearance of eukaryotes and cyanobacteria. Nature 455: 1101-1104 Reynolds, C. S. & A. E. Walsby, 1975. Water blooms. Biological Review 50: 437-481.

Robarts, R. D. & T. Zohary, 1987. Temperature effects on photosynthesis capacity, respiration, and growth rates of bloom-forming cyanobacteria. New Zealand Journal of Marine and Freshwater Research 21: 391-399.

Ryding, S. O. & W. Rast, 1989. The control of eutrophication of lakes and reservoirs. United Nations Educational Scientific and Cultural Organization and Partenon.

Sala O.E, F. S. Chapin, J.J.Armesto, E. Berlow, J. Bloomfield, R. Dirzo, E. Huber-Sanwald, L. F. Huenneke, R.B. Jackson, A. Kinzig, Rik Leemans, D.M. Lodge, H. A. Mooney, M. Oesterheld, N. Leroy Poff, M. T. Sykes, B. H. Walker, M. Wallker, D. H. Wall, 2000. Global Biodiversity Scenarios for the Year 2100. Science. Review : Biodiversity. 1770-1774

Schopf, J.W., 2000. The fossil record: Tracing the roots of the cyanobacterial lineage. In The Ecology of Cyanobacteria. Whitton, B.A. and Potts, M. (eds.) Dordecht, the Netherlands: KluwerAcademic Publishers, pp. 13- 15. Shannon, C. E. & W. Weaver, 1949. The Mathematical Theory of Communication. University of Illionois, Urbana.

Søndergaard, P. A. 1997. Factors affecting the timing of surface scums and epilimnetic blooms of blue-green algae ina a eutrophic lake. Hydrobiologia 506: 135-145. Straile, D., 2000. Meteorological forcing of plankton dynamics in a large and deep continental European lake. Oecologia 122: 44-50.

Utermöl, H., 1958. Zur Vervollkommenung der quantitativen Phytoplankton Methodik. Mitteilungen der Internationalen Vereinigung für Limnologie 9 : 1-38.

Wager, C. & Adrian R., 2009. Cyanobacteria dominance: Quantifiying the effects of climate change. Limnology and Oceangraphy 54: 240-2468.

Winder, M. & D. E. Schindler, 2004. Climatic effects on the phenology of lake processes. Global Change Biology 10 : 1844-

48

Chapter three

3. Phytoplankton configuration in six deep lakes in the peri-Alpine region: are the key drivers related to eutrophication and climate?

Nicole Gallina, Nico Salmaso, Giuseppe Morabto, Martin Beniston

Paper submitted in Aquatic Ecology

49

Phytoplankton configuration in six deep lakes in the peri-Alpine region: Are the key drivers related to eutrophication and climate?

Nicole Gallina1, Nico Salmaso2, Giuseppe Morabito3 and Martin Beniston1

1 University of Geneva, Climate Change and Climate Impacts, Institute of Environmental Science, Campus Battelle, Building D, 7 Rte de Drize, CH-1227 Carouge, Switzerland. 2 IASMA Research and Innovation Centre, Istituto Agrario di S. Michele all'Adige - Fondazione E. Mach, Sustainable Agro-ecosystems and Bioresources Department, Via E. Mach 1, 38010 S. Michele all’Adige, Trento, Italy. 3 CNR – Istituto Studio Ecossistemi, Largo Tonolli 50, I-28922 Verbania-Pallanza (VB), Italy.

Corresponding author: Nicole Gallina, e-mail: [email protected]

Keywords: Cyanobacteria; Temperature; Morpho-Functional Groups, Phosphorus loads, Global warming

Abstract

The aim of this study was to draw a general picture of the phytoplankton community in an enlarged dataset of six deep peri-Alpine lakes, belonging to the same geographical region. The objective was to define the main key drivers that influence the phytoplankton community composition in this particular vulnerable region, since the impacts of climate change have been demonstrated to be stronger than on a global average. The phytoplankton was investigated with a particular focus on cyanobacteria and using a classification approach based on Morpho- Functional Groups (MFG). We hypothesized that phytoplankton in peri-Alpine lakes are mainly driven by nutrients loads as well as by water temperatures, variables that are strongly influenced by climate change and eutrophication. Though different phytoplankton configurations among lakes were partly due to their geographical (altitude) position, assemblages were mostly linked to temperature and nutrients. Furthermore, the results confirmed the significant role of the spring fertilization on the seasonal phytoplankton development. Cyanobacteria were related to the increasing temperature gradient and therefore might become more important under future warming scenario. Air temperatures have a significant impact on water temperature in the uppermost meters of the water column, with a stronger influence in warmer lakes.

Introduction

Physical, hydro-dynamical, and ecological changes are occurring in lake ecosystems; these changes are forecasted to become greater in the future as global temperatures are increasing partly as a result of human activities (IPCC, 2007; Williamson et al.;, 2009, George, 2010;). In the course of the 20th century, the Alpine region has been shown to be particularly vulnerable and sensitive to climate change (Beniston et al., 1997; Beniston 2006), with higher rate of warming, at least double of the observed global average. The question of particular interest is to assess how global warming may affect the phytoplankton community composition, which plays an important role as it is the basis of nearly all aquatic ecosystems’ food supply (Arrigo, 2005). Changes in the phytoplankton community exert an effect on the higher food chain and can potentially disrupt the ecological equilibrium of lakes. Moreover, particular attention is paid to the group of cyanobacteria, the only freshwater phytoplankton group that not only form blooms, but is also able to produce a variety of toxic compounds (Sivonen & Jones, 1999). Cyanobacteria thus can potentially harm lake ecosystems through food web disturbance

50 and anoxia due to massive bloom events, affect human health (water supply, food contamination), the economy (fishing industry), and finally, social activities, as for example the recreational use of lakes. As the world’s oldest known organisms (Schopf, 2000), cyanobacteria have always been capable of adapting to environmental change (Huisman et al., 2005; Paul, 2008). They have relatively high temperature optima for growth (Reynolds 2006), and are able to migrate vertically in the water column to better compete for light and nutrients (Ibelings et al. 1991; Walsby et al. 1997; Reynolds 2006). A number of studies have predicted that cyanobacteria are likely to become more abundant in the future as climate continues warming (Paerl & Huismann, 2008; Paul, 2008; Gallina et al. 2011; Paerl & Paul, 2011). The common phylogenetic classification of phytoplankton is based on taxonomy. However, since a traditional taxonomy does not reflect the ecological function of phytoplankton, other classifications where proposed to aggregate species / genera having similar traits, for example functional-based classifications (Reynolds, 2002), the morpho-functional-based classification (Salmaso & Padisàk, 2007) and the morphological-based classification (Kruk et al., 2010). These aggregations have the advantages to better interpret environmental mechanisms and conditions, and are easier to deduce as they enclose common affinities and considerably reduce the numbers of interpreting components. In deep lakes under a continental climate, phytoplankton communities exhibit strong seasonality in their behavior, which is summarized by the broadly accepted Plankton Ecology Group (PEG) model (Sommer et al., 1986). The PEG model identifies the main influencing factors of the seasonal behavior as climate, weather, grazing, and water chemistry. Several studies debate the duality and the interaction between nutrients (mainly phosphorus) and temperature (reflecting climatological and meteorological forcing) in affecting phytoplankton communities (Moss, 2003; Elliot, 2006; Salmaso, 2009; Stich, 2010). Both these factors have been shown to considerably influence the phytoplankton seasonal community change. As several lakes are implicated, it’s noteworthy that every lake is a mirror of its environment, and therefore the phytoplankton community growth can be presumed to differ in-between lakes, as does the water chemistry and environmental factors (Reynold & Walsby 1975; Lung & Pearl 1988; Ryding & Rast 1989; Blenckner, 2005). This phenomenon implies that the hydro-morphometrical features, the surrounding landscape, origin and history of a lake may play an important role (Ryding & Rast 1989; Blenckner 2005). Recent studies demonstrate even different response of lakes to climate change (Blenckner, 2005; George, 2010). The underlying assumption is that phytoplankton compositions among peri-Alpine lakes mainly represent different responses to nutrient concentrations and temperature. These factors are strongly affected by the influence of climate change and eutrophication. In deep lakes, a further element to take into account is the spring replenishment of nutrients from the deeper to the surface layer. We hypothesize that this pool could represent an important source of nutrients for phytoplankton throughout the year. After previous research based on a smaller number of Northern peri-Alpine lakes (cf. Anneville et al., 2004; 2005), this study reports a synoptic assessment of phytoplankton assemblages and their main driving factors across six peri-Alpine lakes situated North and South of the alpine chain. The main key-drivers of phytoplankton composition changes will be evaluated on yearly and seasonal time-scales (summer-autumn). A matrix including eight different datasets generated from six lakes (two of the lakes were sampled at two different locations) was compiled. For reasons of comparability, the six lakes are all deep and warm monomictic lakes, situated in the same geographical region. Additionally, these lakes have the interesting features to cover the entire trophic gradient and to be located along an altitudinal gradient.

51

Materials and Methods

Lake characteristics and data sources

Figure 1 shows the locations of the lakes. The matrix includes six different lakes at different time periods, namely Lake Constance, Lake Zürich, Lake Walen, Lake Geneva, Lake Maggiore, and Lake Garda. Data from Lake Geneva and Lake Zürich were collected at two sampling points, named as “Small Lake Geneva” and “Big Lake Geneva”, as well as “Upper Lake Zürich” and “Lower Lake Zürich”, respectively. Consequently, eight datasets were derived from the six lakes. The six lakes have a number of features in common: they are all deep, warm monomictic lakes (Hutchinson, 1957), belonging to the same geographical (peri-Alpine) and climatological (continental) region (see Table 1). These similarities allow for comparability and integration of the eight datasets into one data matrix, which builds the basis for the further analysis presented here. Monthly or bi-monthly environmental drivers and phytoplankton biomass recorded in the layer from 0 - 20 m were obtained from state water authorities responsible for lake monitoring as well as from limnological research institutes, namely the LUBW for Lake Constance (1980-1989), the Wasserversorung Zürich and EAWAG for Upper (1980-1990) and Lower Lake Zürich (1980-1990) as well as for Lake Walen (1991-200), SECOE for Small Lake Geneva (2001-2005), the CIPEL for Big Lake Geneva (1977-2000), the FEM-IASMA for Lake Garda (1997-2003) and the CNR for Lake Maggiore (1997-2000). Data from discrete depths (0m, 1m , 2.5m, 5m, 7.5m, 10m, 12.5m, 15m, 20m) in Lake Walen, Upper and Lower Lake Zürich) have been transformed into weighted averages and integrated over the water column from 0-20 m depth. An exception for the sampling depth of the phytoplankton data represented “Big Lake Geneva”, for which the sampling strategy method was applied on an integrated depth from 0- 10 m. However, a three years comparison strategy with the 0-20m layer did not reveal any significant difference in the species composition and total biomass (Cipel, 2002). Same results were found for “Small Lake Geneva” (CIPEL, 2002), which allowed us for integration of those data. Based on total phosphorus concentrations, and according to OECD (1982), the assessed lakes cover the entire trophic gradient (cf. next paragraph).

Figure 1: Location of the Lakes in the peri-Alpine arc which were considered in this study. Lake Constance (CONST), Lower Lake Zürich (LOZH), Upper Lake Zürich (UPZH), Lake Walen (WAL), Small Lake Geneva (SGE) and Big Lake Geneva (BGE) are situated in the northern part of the alpine arc, whereas Lakes Maggiore (MAG) and Garda (GARD) are located on the southern part.

52

Environmental and biological data The choices for potential drivers to explain the phytoplankton community configuration were done upon recent findings but strongly depend on the availability of data. In the present study those drivers are considered, for which the same sampling strategies, standard measuring methods and no missing values could be guaranteed, in order to ensure homogeneity of the data. Hydro-morphometrical descriptors include lake altitude, surface, watershed area, ratio of the watershed area and surface, maximum and mean depth; lake volume, residence time, and outflow (cf. table 1). Physical-chemical descriptors (0-20 m) include conductivity (as a proxy for runoff), nitrate, total phosphorus, soluble reactive phosphorus concentrations and water temperature. The euphotic depth was obtained in multiplying the Secchi depth by a factor of 2.5 (Vollenweider, 1982). The trophic state of a lake was defined by the total phosphorus concentration (OECD, 1982). The time series of each lake were selected to guarantee that the entire trophic gradient was presented and well balanced between the different archives: Lake Walen and Lake Maggiore are oligotrophic (O), Upper Lake Zürich and Small Lake Geneva mesotrophic (M), Lake Constance and Big Lake Geneva eutrophic (E). Lower Lake Zürich and Lake Garda are classified respectively as meso-eutrophic (ME) and oligo-mesotrophic (OM). For computational reasons, the trophic states were additionally assigned in classes ranging from 1 to 5, class 1 corresponding to the oligotrophic state and class 5 to the eutrophic state (Table 1). For the Swiss lakes, the daily average air temperature data were downloaded from the digital database of the Swiss Federal Office for Meteorology and Climatology (MeteoSwiss). Five meteorological stations were chosen, based upon their proximity to the phytoplankton sampling stations, namely Güttingen and Kreuzlingen for Lake Constance; Zürich/Fluntern for lower Lake Zürich; Glarus, for Lake Walen; and Changins for Small Lake Geneva. For Big Lake Geneva data were received from the meteorological station situated at Thonon-les-Bains belonging to the INRA Research Institute. As for Lake Garda, daily averaged air temperature was measured at the meteorological station of Arco (ARC), at the northern border of the lake (Salmaso, 2010). For Lake Maggiore, the daily averaged air temperature was measured at the meteorological station in Verbania-Pallanza (Ambrosetti et al., 2006). The datasets derived from MeteoSwiss have been quality-checked for the homogeneity of these records (Bergert et al., 2005). The meteorological and climatological data were measured at the same days as the phytoplankton samples were taken. Zooplankton was selected as a proxy for grazing pressure on phytoplankton. Two main groups of zooplankton were distinguished: Cladocerans and Copepods (without nauplii) (both measured as individuals m-2). To estimate phytoplankton abundance, standard sampling and counting methods were employed (Utermöhl, 1958; Lund et al., 1958; see also Straile, 2000; Morabito, 2002; Anneville and al. 2002, 2004; Lavigne and al., 2006; Salmaso, 2011). For every single taxon, biovolumes were calculated from recorded abundances and specific biovolumes approximated to simple geometrical solids. These procedures were standardised (Rott, 1981), therefore minimising the errors associated to the estimates of single specific biovolumes in different laboratories. Since the species nomenclature for phytoplankton changed considerably during the different periods in which the data were collected and counted, it was necessary to continuously update and quality check with the species nomenclature, before proceeding with the analysis. The update was completed following the more recent monographs of the series “Süsswasserflora von Mitteleuropa” established by A. Pasher (Gustav Fisher Verlag, and Elsevier, Spectrum Akademischer Verlag). Subsequently the species biomass was summed to the genera biomass (µg L-1). The obtained 233 genera were afterwards grouped into the Morpho Functional Groups (MFGs) defined by Salmaso and Padisák (2007). All in all, 25 MFGs were identified. Based on this classification, phytoplankton genera were grouped following the criteria of motility, specific nutrient requirement (autotrophy, mixotrophy), size, shape, and presence of gelatinous envelopes (Weithoff, 2003). This work will focus particularly on Cyanobacteria, which in these lakes are mainly represented by three Morpho-Functional Groups, namely MFG 5a (thin filaments, Oscillatoriales), MFG 5c (other large colonies, mostly non-vacuolated Chroococcales), MFG 5d (small Colonies, Chroococcales) and MFG 5e (Nostocales).

53

Data analysis Annual averages were computed from the original monthly and bi-monthly dataset (phytoplankton MFG abundance and descriptors). The resulting matrix contained 73 averaged yearly data, which was employed to evaluate the difference in between phytoplankton configuration in peri-Alpine lakes, and to define the key factors driving phytoplankton community. Moreover, seasonal averages were used to assess the role played by the spring replenishment of P on the phytoplankton assemblages during the main growth period, from summer to autumn (June to November). Therefore, in this specific analysis, variables were averaged for the summer-autumn period, with the exception of the nutrients (N and P), which were averaged also for the spring period. Data were further analyzed by Non-Metric Multidimensional Scaling (NMDS) and applied to Bray & Curtis’ dissimilarity matrices (Legendre & Legendre, 1998) computed on MFGs biomass values. NMDS ordination can be rotated, inverted, or centered to any desired configuration. The accuracy of fit of the projections is measured by “stress” estimates (Kruskal & Wish, 1978). Environmental variables were related to the strongest gradients in species composition by fitting environmental vectors to the NMDS configurations. In the present analysis, vector fitting finds the maximum correlation of the single variables with the set of lakes in the configurations. Fitted vectors point to the direction of most rapid change in the environmental variables, whereas their length is proportional to the correlation between the environmental variable and the ordination. The significance of vectors was based on 999 random permutations of the data (Oksanen et al, 2011). Before the NMDS ordination, MFGs were double-square rooted to reduce the weight of the more abundant groups (Salmaso, 1996). The descriptors were log-transformed, Yi = log (Xi + 1) prior to vector-fitting. The data analyses were performed with specific packages in R (ade4 and vegan; R Development Core Team, 2011). The relationship between the descriptors was analyzed through a Spearman (ρ) rank correlation matrix (stats package in R, R Development Core Team, 2011).

The phytoplankton composition between lakes To examine if each peri-Alpine lake dataset has its own phytoplankton community, a NMDS was applied upon the “Bray-Curtis” dissimilarities matrix, permuted 999 times on the yearly averaged and double root squared MFG. Following this step, the categorical variable ‘’lake’’ was built to partition the rows in classes belonging to the same lake dataset. The functions “ordispider” (stars) and “ordihulls” (convexhulls) from the vegan package (Oksanen et al., 2011) were applied, in which the categorical variable “lake” grouped the samples belonging to the same lake (Legendre and Legendre, 1998). The lakes are labeled at the centroid of each convex hull. To further test if the centroids have different position, we performed the “Adonis” function with 999 permutations (Oksanen et al., 2011).

Results

Hydro-morphometrical similarities and differences between peri-Alpine lakes. Table 1 presents the hydro-morphometrical characteristics of the eight analyzed datasets. Lake Maggiore reaches a maximum depth of 370 m whereas the most shallow, Upper Lake Zürich, has a depth of 48 m. The lakes are located at altitudes between 420 m a.s.l. (Walen) and 65 m a.s.l. (Garda). Lake Geneva, the largest lake in central Europe, has a surface area 20 times that of Lake Walen (24 km2). The theoretical residence time ranges between 2 months (Upper Lake Zürich) and 27 years (Garda). Upon hydro-morphometrical resemblance, namely trophic state, altitude and size, the lakes can be differentiated in: 1) Larger lakes at lower altitudes and low trophic states (Lakes Garda and Maggiore); 2) Larger lakes at intermediate altitudes and high trophic states (Lake Constance and Big Lake Geneva); 3) Smaller lakes at higher altitude and low trophic states (Lake Walen); 4) Smaller lakes at higher altitude and higher trophic states (Upper and Lower Zürich). Small Lake Geneva is situated in-between ranges of trophic state, altitude and trophic levels.

54

Max Mean Lake Altitude Surface Watershed Ratio Volume Outflow Trophic Trophic Depth Depth Residence Abbreviation (m a.s.l.) (km2) area (km2) WSA/SURF (km3) (m3s-1) state class (m) (m) time (yrs)

TIME Lake Name ALT SURF WSA RATIO maxD meanD VOL RT OF TS TC PERIOD

Constance CONST 80-89 395 472 10900 23.1 252 101 48 4.3 750 E 5

Lower Zürich LOZH 80-90 406 65 1740 26.8 136 51 3.3 1.4 89 ME 4

Upper Zürich UPZH 80-90 406 20 1564 77.2 48 23 0.5 0.18 76 M 3

Walen WAL 91-00 420 24 1061 44.2 145 100 2.4 1.4 57 O 1

Small Geneva SGE 01-05 372 81 7395 91.3 76 41 3 0.4 252 M 3

Big Geneva BGE 77-00 372 499 7395 14.8 310 172 86 11.4 252 E 5

Maggiore MAG 97-03 193 213 6599 31.0 370 178 37.5 4.1 291 O 1

Garda GARD 97-00 65 368 2290 6.2 350 133 49 26.6 58 OM 2

Table 1: Hydro-morphometrical characteristics of the analyzed peri-Alpine lakes are shown. The time period in which each lake was sampled, the unities and the abbreviations of Lakes and variables are also given. The trophic state are classified in E = eutrophic, ME= meso-eutrophic, M= mesotrophic, OM = olig- mesotrophic, and O = oligotrophic state.

55

Phytoplankton MFGs assemblages Figure 2 shows the Non-Metric Multidimensional Scaling (NMDS) configuration based on the MFGs annual averages. The years belonging to the same lake are grouped together by convex hulls and labeled at their weighted centroid. The outcomes of the “adonis” function on 999 permutation indicate a significant different and non-random position of each centroid of the lakes (R2 = 0.598, p<0.001). This highlights the existence of significant differences in the assemblages of the different lakes. The phytoplankton communities of Lake Constance and Small Lake Geneva overlap for some years with the one of Big Lake Geneva indicating some similarity in their composition during these years. Likewise, Lower Lake Zürich’s position also suggests similarity to the phytoplankton configuration of Big Lake Geneva. In contrast, Lake Walen, Lake Maggiore and Lake Garda display different positions, indicating a differentiated phytoplankton composition. Figure 3 represents the biomass composition of the 11 main MFGs for each lake. For graphical reasons, the subordinated MFGs were not displayed. The lakes are ordered along the trophic gradient (see Table 1), from oligotrophy to eutrophy. Oligotrophic and oligo-mesotrophic states (Maggiore, Walen and Garda) had lower phytoplankton biomass than eutrophic lakes. However, the largest quantity of biomass was recorded in the mesotrophic Small Lake Geneva, whereas Lake Constance’s phytoplankton biomass was relatively low compared to its trophic state. The MFG1 represents large flagellates including the potential mixotrophs like large Chrysophytes/Haptophytes (Dinobryon, Mallomonas), Dinophytes and Euglenophytes (Gymnodinium, Ceratium). The MFG1 is present in all lakes, with highest biomass in Big Lake Geneva. Small potentially mixotrophes flagellates (MFG2) are present in all lakes but have higher biomass in lakes that exhibit higher trophic levels, and are therefore almost absent in Lake Garda. The representative genera of MFG2 are Rhodomonas, Cryptomonas and Erkenia. The MFG3, mostly autotrophs flagellates represented by Phytomonadina (especially Chlamydomonas, Pandorina and Phacotus), are practically absent in Lake Maggiore and Garda, at the Southern part of the alpine chain, reaching relatively high biomass in Small Lake Geneva. Non flagellated generas are represented by MFG5 to MFG11. The MFG5 comprises the colonial Cyanobacteria, with the most representative cyanobacteria genera being Planktothrix, Pseudoanabaena, Aphanothece, Aphanocapsa and Aphanizomenon. This group is present in all the lakes, however with quite different biomass. MFG5 are most abundant in Small Lake Geneva and Lower Lake Zürich, have important biomass in Lakes Maggiore, Garda and Big Geneva, and low biomass in lakes Walen, Upper Zürich and Constance. MFG5 tends to prefer lakes on lower altitude and higher water temperature. Diatoms were subdivided, into large Diatoms and small Diatoms, represented respectively by MFG6 and MFG7. Large Diatoms, mainly Fragilaria, Asterionella, Diatoma and Nitzschia, are prevalent in all lakes, representing a very important contribution to biomass. Smaller diatoms (especially Stephanodiscus, Aulacoseira, Cyclotella and Navicula) were also present in all lakes, but with lower biomass, and with a positive tendency in biomass towards higher trophic lakes. MFG8 represents the “Other Large Unicellular” Genera and MFG9 the “Other Small Unicellular” Genera. These two groups both exhibit relatively unimportant biomass. MFG8 (Closterium, Staurastrum and Cosmarium) was particularly more abundant in Big Lake Geneva, whereas MFG9 (Chlorella, Ankyra) in Lake Constance. MFG10 and MFG11 represent “Others Colonials”. Filamentous colonies are grouped in MFG10 and non-filamentous colonies in MFG11. Filamentous colonies (MFG10, mostly Mougeotia) reach very high biomass in Lake Garda and Small Lake Geneva, but also have important biomass in Big Lake Geneva, whereas in the remaining lakes their biomass is unimportant. MFG11, basically Oocystis, Elakatothrix and Scenedesmus, showed important biomasses in Small Lake Geneva and in Lower Lake Zürich. To summarize, peri-Alpine Lakes are characterized by high biomass of MFG1 and MFG6. MFG2 and MFG7 prefer lakes with higher nutrient loads, whereas the colonial Cyanobacteria (MFG5) are more frequently found in lakes at lower altitude where water temperatures are higher.

56

Figure 2: Non-Metric Multidimensional Scaling (NMDS) of the double square roots transformed Morpho-Functional Groups (MFG), which are yearly averaged. The label is situated at the centroid of each convex hull grouping the phytoplankton community for each year belonging to the same lake together (G=Garda, M=Maggiore, W=Walen, U=Upper Zürich, L=Lower Zürich, C=Constance, B=Big Geneva, S=Small Geneva).

Figure 3: Diversity of phytoplankton composition upon their Morpho-Functional Groups (MFG) (Salmaso &Padisàk, 2007). The lakes are ordered following the trophic gradient form oligotrophic (Lake Maggiore, MAG) to eutrophic (Big Lake Geneva, BLG). The abbreviations of the MFG are given in the legend, the full names are listened hereafter: 1) Large colonial or unicellular, potential mixotrophs; 2) Small potential mixotrophes (unicellular); 3) Phytomonadina; 5) Colonial Cyanobacteria; 6) Large Diatoms; 7) Small Diatoms; 8) Other large unicellular; 9) Other small unicellular; 10) Other filamentous colonies; 11) Other non filamentous colonies;

57

Correlations between the environmental descriptors Table 2 reports the Spearman correlation coefficients between the annual averages of the descriptors. Significant correlations (p < 0.05, N = 73) are highlighted in bold text with a larger font. Both TP and SRP were strongly correlated with the conductivity, the duration of the stratification period and the cladocera. Moreover, cladocera displayed a significant negative relationship with the water and air temperature and a positive relationship with the nitrate concentration. As for the copepods, they only correlated significantly with the euphotic depth. The relationship between air temperatures and water temperature (annual averages) were adequately described by the following equation: y = 0.06x2 – 0.93x + 12.5 (r = 0.8) (Figure 4). The lakes located at higher elevations and therefore with colder temperature regimes formed a group distinct from warmer lakes at lower altitudes.

Figure 4: Polynominal regression model in between the yearly averaged air and water temperature (y = 0.06x2 – 0.93x + 12.5; r = 0.8; N=73). The dashed line separates lakes with higher and lower water temperature.

58

) )

2 2

Air

(m)

(°C)

(°C)

20°C 20°C

Water Water

(µg/L) (µg/L)

Nitrate

Soluble Soluble

reactive reactive

(µS/cm)

(month)

(Ind/m (Ind/m

Duration

tot (µg/L)

Copepodes

Cladoceras Cladoceras

Descriptors

phosphorus phosphorus

Temperatur

Phosphorus

Temperature Temperature

Stratification Stratification

Conductivity at at Conductivity Depth Eupohic Abbre- CLAD COND NO3N TP SRP EuD WT COPE AT StratD O viation

200’36 Mean 248.8 544.8 24.4 13.9 6.9 10.5 370’504 10.5 6.72 9

153’81 SD 42.5 176.91 15.6 12.4 1.67 1.36 216’953 1.95 0.88 5

13’621 137.8- 209.7- 5.7- 1.39- 3.9- 9.1- - 94’764- 8.08- 5- Range 292.3 779.3 51.2 36.9 10.9 13.9 495’01 807’471 13.5 8 6

COND 1 -0.34 0.61 0.61 -0.02 0.04 0.09 -0.21 0.00 -0.25

NO3N 1 0.05 -0.060 -0.23 -0.43 0.39 0.34 -0.38 -0.14

TP 1 0.98 0.05 -0.22 0.58 0.11 -0.15 -0.41

SRP 1 0.01 -0.18 0.54 0.14 -0.1 -0.38

EuD 1 0.4 -0.29 0.43 0.25 -0.33

WT 1 -0.49 0.26 0.81 0.14

CLADO 1 0.22 -0.38 0.01

COPE 1 0.12 -0.15

AT 1 0.25

StratD 1

Table 2: Mean, standard deviation (SD) and the range (minimum and maximum values) of the yearly averaged phytoplankton descriptors are shown in the first 3 lines in italic, followed by the spearman rank correlation coefficient matrix. The significant correlations (α < 0.001) are represented in bold and larger font.

59

Phytoplankton configurations and relationships with the environmental variables a) Analyses based on annual averages In this section we will interpret two NMDS configurations obtained from annual (Fig. 5a) and seasonal (Fig. 5c) averages of MFGs by vector fitting using the available environmental data. The configurations of lakes in Fig. 5a coincide with Fig. 2; however data have been presented differently, and in a similar way as Fig. 5c. Total phosphorus was not considered because it was highly correlated with soluble reactive phosphorus (Table 2). Only SRP was used as it represents the chemical form for nutrient uptake by phytoplankton. In Figure 6a the lowest stress factor (0.14) was reached after 6 runs and permuted 999 times, which guarantees a good configuration and confident interpretation (Zuur, 2007). The phytoplankton configuration showed a significant link with the water temperature as well as the air temperature (WT, AT), nutrients (SRP, NO3-N) and conductivity (COND) (p<0.0001). The grazing by cladocera and the duration of the stratification period also were linked (0.0001

b) Analyses carried out in the main growth period (summer-autumn) The aim was to determine whether the spring nutrients represented an important source for phytoplankton during the main growth season (summer-autumn). The more important variables linked with the NMDS configuration were, besides nutrient concentrations (SRP, NO3N) the water and air temperature (WT, AT), are as well as the spring pool of nutrients present during the maximum spring replenishment of euphotic layers (SRPsp and NO3sp) (Figure 5d; p<0.01). The direction and strength of the variables are practically coincident with those reported in Fig. 5b. Cladocerans and the duration of stratification did not show a link with the configuration in Fig. 5c. During the summer-autumn period, the phytoplankton composition of Lake Constance was more similar to that of Lower Lake Zürich (Fig. 5c). However, the MFG did not change the position of the lakes Constance and Lower Zürich significantly. MFGs belonging to cyanobacteria were anew dispersed, with the difference that during the main growth season, MFG5c is more linked to the phosphorus gradient. MFG11c (other non-filamentous colonies) were more associated with the nutrient gradient.

60

Figure 5a – 5d: Output on the NMDS analyses based on yearly averaged MFG (a,b) and on summer-autumn averaged MFG (c,d) . The samples belonging to the same lake are labeled at their centroid (G=Garda, M=Maggiore, W=Walen, U=Upper Zürich, L=Lower Zürich, C=Constance, B=Big Geneva, S=Small Geneva) (6a,c).The environmental variables were fitted and the significant vectors (bold p<0.001, normal p<0.01) show the direction of the environmental gradient and its strengths (length of the vector).

Discussion

Nutrients and temperature influenced differently the deep peri-Alpine lakes. SRP and conductivity (as a proxy for the intensity of runoff), and air and water temperature had opposite directions in the NMDS configurations (Figs 5a, b). The interpretation could not be univocal, because it could be simply linked to the characteristics of the lakes included in the dataset. The warmer lakes, Garda and Maggiore, shared oligo - oligo- mesotrophic characteristics, while Big Lake Geneva and Constance were both eutrophic and colder. This study reveals an equivalent importance but a different effect on the phytoplankton community. However, both variables are interconnected variables, as warming affects the stratification of lakes, on the other hand, the deepening and duration of the stratification period impacts the nutrient concentration available for the biotic

61 compartment in lakes (Wetzel, 2001; Anneville and al., 2005, George, 2010,). Besides difficulties to generalize the results, linked to the number of available lakes, the analysis allowed to appreciate the strong peculiarities in the composition of phytoplankton in the single lakes. These results support Bleckner’s (2005) argument that differences and changes of phytoplankton result not only from the “landscape filter” but also from the so-called “internal lake filter” (abiotic/biotic interactions and lake history). Therefore, future studies on peri-Alpine lake are advised to further specify the nature and consequences of this differential behavior of phytoplankton communities, since it can be argued that the phytoplankton community of each lake will respond differently to the effects of eutrophication and climate change in the future (Gerten & Adrian, 2001; Georg et al., 2004). The eutrophication state based upon phosphorus concentration (OECD, 1982) remains a reliable indicator, as higher trophic levels have higher phytoplankton biomass. An exception, however, was Small Lake Geneva, which followed a non-linear hysteresis pattern (Gawler, 1988). Reduced phosphorus levels resulting from lake recovery management, do not result in equal levels of reduced phytoplankton biomass. Small Lake Geneva shows that its route to recovery seems to differ significantly from the route to eutrophication (Ibelings, 2007), which was already observed in other deep peri-Alpine lakes (Anneville, 2002, 2004). The colonial cyanobacteria (MFG 5) were present in all the considered lakes, and with higher frequency in lakes at intermediate and lower altitudes where water temperatures were higher. The filamentous cyanobacteria (Oscillatoriales MFG5a) and other large, mostly non vacuolated colonies (MFG5c) were related to the increasing temperature. The specific ability of Cyanobacteria to adapt to high temperatures has been reported in a number of studies (e.g., Paerl et al. 1985; Robarts & Zohary 1987; Briand 2004; Reynolds, 2006). MFG groups are an effective way to assess the ecology of Lakes. For example, Lake Garda and Small Lake Geneva were characterized by MFG 10b, colonial filamentous genera, represented by Mougeotia. Actually, it was found that in Lake Garda, in the period of its maximum dominance, Mougeotia showed higher biomasses after cold winters and deep mixing, i.e. after episodes of greater replenishment of phosphorus in the euphotic layers (Salmaso, 2002). In summer 2001 the small Lake Geneva showed a massive development of Mougeotia gracillima bloom (Lavigne & Druart, 2002). Therefore, these results demonstrated that the use of MFGs has a potential advantage, not only in comparison to, but also as complementary to the use of basic taxonomic units to reveal different ecological aspects. Using exclusively a few basic morphological groups, Kruk (2002) stated that phytoplankton composition can be better predicted by a morphological approach. In the case of multiple phytoplankton assemblages recorded in different lakes, MFGs may represent a useful tool to concurrently investigate the evolution of different lakes, overcoming problems related to the comparison of different taxa and to the existence of possible differences in taxonomic accuracy (Salmaso & Padisák, 2007; Reynolds et al., 2002; Kruk et al., 2011). The relationship between air and water temperature has its significance in the light of changing climate in the peri-Alpine region. Air temperature, which is directly affected by global warming, and water temperature were significantly correlated. The use of air temperature as a proxy for water temperature in peri-Alpine lakes, in the uppermost meters of the lake waters and during summer, has been discussed by Livingstone (1998). Moreover, formerly Gallina and al. (2011) were able to demonstrate the capacity of air temperature to affect the phytoplankton communities in the epilimnion in peri-Alpine lakes. Remarkably, the effect of air temperature on water temperature has been proven to be even stronger in the case of lakes in lower altitudes. As previously stated, nutrients and conductivity (NO3-N, SRP and COND), temperatures (AT and WT) showed a strong link with the configurations of lakes obtained from MFG groups. Cladocerans and the duration of the stratification period had an important role, even though less prominent. Generas, which belong to the larger forms (colonies and filaments) of green algae, followed the phosphorus gradient. Smaller forms, however, such as the MFG3a (Chlamydomonas), were an exception. Chlamydomonas is known to be abundant in extremely nutrient rich waters (Reynolds, 2006). Padisák and Tóth (1991) also noticed that these small green algae seem to benefit from the environmental heterogeneity. The temperature gradient mainly pointed towards MFG5a (Oscillatoriales), with the potentially toxic genera Planktothrix and Pseudanabaena and partly of MFG5d (Aphanothece). Under future climate warming scenario, modifications in freshwater communities could favour morphotypes of colonial phytoplankton, especially colonial cyanobacteria (MFG5). Since colonial forms have evolutionary and/or ecophysiological advantages over unicellular forms, with regard to, i.e., predation, viral mortality, specialization (Beardall et al., 2008), it can be hypothesized that colonial forms will become in the future even more dominant in the peri-Alpine region that is very sensitive to short-changes in weather (Thompson et al., 2005). Likewise, Shatwell et al. (2008) showed how warming promoted the colonial filamentous Cyanobacteria (Oscillatoriales) in Müggelsee (Germany), a shallow temperate lake. Contrariwise, several studies and ecological rules dealing with the effect of temperature-size relationship forecast that in aquatic systems warming benefits the small forms (Daufresne, 2009; Winder & Hunter, 2008). If analysed more in detail, these considerations contrast only apparently. Large colonial species and filamentous cyanobacteria possessing gas vesicles have the ability, unlike other eukariotic algae, to overcome the environmental constrains originated by the increasing water stability, which is an important side effect of higher water temperatures (Walsby, 1997). The results presented here confirm Shatwell’s (2008) rather general arguments and further

62 suggest a strong connection between global warming and the development of cyanobacteria (Shatwell 2008; Paerl & Huisman 2008; Paerl 2009, O’Neil 2011). In the middle of the temperature and phosphorus gradient, and opposite of the NO3N gradient, MFG5c included Anabaena and Aphanizomenon. Both generas are potentially toxic, nitrogen-fixing cyanobacteria (Reynolds, 2006). Therefore these genera are tolerant in nitrogen-poor waters. Interestingly, this study found no cyanobacteria situated along the cold, NO3N rich gradient, further supporting the importance of both phosphorus and higher temperature in supporting the growth of this algal group. During the growth period, nutrients and temperature were the main key factors linked to phytoplankton. In contrast, the grazing and the duration of the stratification period did not seem to affect the summer-autumn population. This is in agreement with the PEG Model, which indicates that grazing is mostly prevalent during the spring period (Sommer et al., 1986). Overall, these results demonstrate the crucial importance of the concentration of spring nutrients as fertilizers able to affect the phytoplankton composition, not only during the summer-autumn period, but also throughout the year. The nutrient enrichment in spring highly depends on the depth of the mixing layer (Sommer, 1986; George, 2010), which in turn depends on the climatological/meteorological conditions encountered during winter (Salmaso et al., 2003). Future climate change scenarios however, predict milder winters with less deep mixing (Perroud, 2009). Consequently, this will provide a less nutrient enriched epilimnon with important effects on the seasonal phytoplankton growth.

Conclusions

As hypothesized, differences in phytoplankton composition in the deep peri-Alpine lakes were mainly driven with similar strengths by temperature (air and water) and nutrients (P, N), and secondarily by the duration of the stratification period and grazing by the Cladocerans. Cyanobacteria responded differently to these gradients. Oscillatoriales were closely related to higher water temperatures and longer stratification period; Chroococcales and small colonies were found were temperatures were higher and nitrates more high; lower N-concentrations were linked with a larger presence of large non-vacuolated cyanobacteria colonies and Nostocales. Most notably, during the growth season nutrients and temperature showed a strong link with the distribution of lakes in the NMDS configurations based on MFGs dissimilarities. In a future scenario, where temperatures in the Alpine region are predicted to increase, it is hypothesized that the effects of global warming will mostly impact cyanobacteria .This study gives strong support to the hypothesis that the fertilization of the epilimnion in deep lakes is a crucial stage in the phytoplankton growth cycle, controlling seasonal development throughout the rest of the year. Since air temperature seems to have a stronger impact on water temperature in lower altitude lakes, the predicted consequences of global warming are assumed to have a stronger impact on lowland large waterbodies. Furthermore, the results support the existence of specific phytoplankton morpho-functional groups in different lakes, partly unrelated to the climatic and trophic gradients. Future efforts should focus on further defining those factors that can explain these intrinsic differences. In the broader context of research on the effects of climate change, the present study represents only a first step towards the knowledge that is needed to model the impacts of climate change on phytoplankton assemblages.

Acknowledgements The authors wish to thank S. Lavigne for helping with the identification of the phytoplankton species. Special thanks are due to the support we received from our collaborators at the different institutes in providing the data, namely: Dr. R. Kümmerlin from the LUBW, Dr. R. Forster from the Wasserversorung Zürich, Dr. H. R. Bürgi from the EAWAG Dübendorf , Dr. J. Perfetta from the SECOE in Geneva, Dr. O. Anneville for the the CIPEL. In addition, the authors extend their thanks to the Swiss Office for Meteorology and Climatology (MeteoSwiss) for allowing the use of their data.

63

References

Arrigo K (2005) Marine microorganism and global nutrient cycles. Nature 437: 349-355,

Anneville O, Souissi S, Ibanez F, Ginot V, Druart JC, Angeli N (2002) Temporal Mapping of Phytoplankton Assemblages in Lake Geneva: Annual and Interannual Changes in Their Patterns of Succession. Limnol Oceanogr 47: 1355-1366.

Anneville O, Souissi S, Gammeter S, Straile D (2004) Seasonal and inter-annual scales of variability in phytoplankton assemblages: comparison of phytoplankton dynamics in three peri-alpine lakes over a period of 28 years. J Freshwat Biol 49: 98–115.

Anneville O, Gammeter D, Straile D (2005) Phosphorus decrease and climate variability: mediators of synchrony in phytoplankton changes among European peri-alpine Lakes. J Freshwat Biol 50: 1731-1746.

Ambrosetti W, Barbant L, Rolla A, (2006) The climate of Lago Maggiore area during the last fifty years. J Limnol 65: Suppl. 1: 1-62.

Begert M, Schlegel T, Kirchhofer W (2005) Homogenous temperature and precipitation series of Switzerland from 1864 to 2000. Int J Climatol: 2565-80.

Beniston M, Diaz HF, Bradley RS (1997) Climatic change at high elevation sites; a review. Clim Change, 36, 233 - 251

Beniston M (2006) Mountain weather and climate: A general overview and a focus on climatic change in the Alps. Hydrobiologia 526: 3-16

Berdall J, Allen D, Braag J, Finkel ZV, Flynn KJ, Quigg A, Rees TAV. Richardson A , Raven JA (2009) Allometry and stoichiometry of unicellular, colonial and multicellular phytoplankton. New Phytol 181: 295-309.

Bleckner T (2005). A conceptual model of climate- related effects on lake ecosystems. Hydrobiologia 533: 1-14.

Briand JF, Leboulanger C, Humbert JF (2004) Cylindrospermopsis raciborskii (cyanobacteria) invasion at mid- latitudes: selection, wide physiological tolerance, or global warming. J Phycol 40: 231-238.

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, OJ L 327, 22.12.2000, p. 1–73

Dodson S (2004) Introduction to Limnology. McGraw-Hill Science, Engineering , Math 416pp.

Dray S, Dufour AB (2007) The ade4 package: implementing the duality diagram for ecologists. J Stat Softw 22(4): 1-20.

Elliot JA, Jones ID, Thackeray SJ (2006) Testing the sensitivity of phytoplankton communities to changes in water temperature on nutrient load, in a temperate lake. Hydrobiologia 559:401-411.

Gallina N, Anneville O, Beniston M (2011) Impacts of extreme air temperatures on cyanobacteria in five deep peri-Alpine lakes. J Limnol 70(2): 186-196.

Gawler M, Balvay G, Blanc P, Druart P, Pelletier C (1988) Plankton ecology of Lake Geneva : a test of the PEG- Model. Arch Hydrobiol 114: 161-174.

George G, Marbery SC, Hewitt DP (2004) The influence of the North Atlantic Oscillation in the physical, chemical and biological characteristics of four lakes in the English Lake District. J Freshwat Biol 49:760-774.

George G (2010) The impact of Climate Change on European Lakes. Aquatic Ecology Series, Vol.4. Springer edition.

Gerten DR, Adrian R (2001) Differences in the persistency of the North Atlantic Oscillation. Limnol Oceanogr, 46:448-455.

64

Holland SM (2008) Non-metric multidimensional scaling (MDS). Athens: R forge.

Hutchinson GE (1957) A treatis on limnology Vol. l. John Wiley and Sons. New York.

Huisman JM, Matthijs HPC, Visser PM (2005) Harmful Cyanobacteria. Springer aquatic Ecology Series. Springer, Dordrecht, The Netherlands.

Ibelings BW, Mur LR, Walsby AE (1991) Diurnal changes in bouncy and vertical distribution in populations of Microcystis in two shallow lakes. J Plankton Res 13: 419-436.

Ibelings BW, Portielje R, Lammens E, Noordhuis R, van den Berg MS, Joosse W, Meijer ML (2007) Resilience of Alternative Stable States during the Recovery of Shallow Lakes from Eutrophication: Lake Veluwe as a Case Study. Ecosystems 10, 4–16.

Intergovernmental Panel on Climate Change (IPPC). Climate change (2007) Synthesis Report. Contribution of Working Groups I, II and III to the Fourth Assessment Report of the Interngovernemental Panel on Climate Change. Core writing team, Pachaure, R.K., Reisinger, A.(Eds.) Geneva: IPCC; 2007.

Jongman RHG, Ter Braak CJF, Van Tongeren OFR (1995) Data analysis in community and landscape ecology. Cambridge University Press, New York.

Kruk C, Huszar VLM, Peeters ETHM, Bonilla S, Costa LS, Lürling M, Reynolds CS, Scheffer M (2010) A morphological classification capturing functional variation in phytoplankton. J Freshwa. Biol 55 (3): 614-627.

Kruk C, Peeters ETHM, Van Nes EH, Huszar VML, Costa LS, Scheffer M (2011) Phytoplankton community compostion can be predicted best in terms of morphological groups Limnol Oceanogr 56(3):110-118.

Kruskal J.B. & Wish M. (1978) Multidimensional Scaling. Sage Publications, Beverly Hills and London.

Lavigne S, Druart JC (2002) Evolution du phytoplancton du Léman. Campagne 2001. CIPEL 2002, 71-84.

Lavigne S, Cordonier A, Gallina N, Perfetta J, (2006) Campagne 2005. CIPEL, 101-116.

Legendre P, Legendre L (1998) Numerical ecology. 2nd English edition, Elsevier Science, Amsterdam. 853 pp.

Livingstone DM , Lotter AF (1998) The relationship between air and water temperatures in lakes of the Swiss Plateau: a case study with palaeolimnological implications. J. Paleolimnol. 19: 181-198.

Lund JWG, Kipling C, Lecren EO (1958) The inverted microscope method of estimating algal numbers and the statistical basis of estimations by counting. Hydrobiologia, 11: 143-170.

Lung WS, Pearl HW (1988) Modeling green-blue algal blooms in the Lower Neuse River. Water Res 22: 895- 905.

Moe JS, Dudley B, Ptacnik R (2008) REBECCA databases: experiences from compilation and analyses of monitoring data from 5,000 Lakes in 20 European countries. Aquat Ecol 42:182-201.

Moss B, McKee D, Atkinson D, Collings SE, Eaton JW, Gill AB, Harvey K, Hatton T, Heyes D, Wilson (2003) How important is climate? Effects of warming, nutrient addition and fish on phytoplankton in shallow lake microcosms. J Appl Eco 40: 782-792.

Morabito G, Ruggio D, Panzani P (2002). Recent dynamics (1995-1999) of the phytoplankton assemblages in Lago Maggiore as a basic tool for defining association patterns in the Italian deep lakes. J Limnol 61(1) 129-145

Tsugeki NK, Urabe J, Hayami Y, Kuwae M, Nakanishi M (2010) Phytoplankton dyanmics in Lake Biwa during the 20th century: complex responses to climate variation and changes in nutrient status. J. Paleolimnol 44:69-83 (15).

65

Nordhausen K, Sirkia S, Oja H, Tyler DE (2010) ICSNP: Tools for Multivariate Nonparametrics. R package version, 1.0-7. http://CRAN.R-project.org/package=ICSNP.

OECD (1982) Eutrophicaton of waters. Monitoring, assessment and control. Organisation for Economic Cooperation and Development. 193pp.

Oksanen J, Blanchet FG, Kindt R, Legendre P, O'Hara RB, Simpson GL, Solymos P, Stevens MHH, Wagner H (2011) vegan: Community Ecology Package. R package version 1.17-10. http://CRAN.R- project.org/package=vegan.

Paul VJ (2008) Global warming and cyanobacterial harmful algal blooms. In: K.H. Hudnell (Ed.), Cyanobacterial harmful algal blooms: state of the science research needs series. Springer Advances in biology and medicine 619: 239-257.

Padisák J, Tóth LG (1991) Some aspects of the ecology of the subdominant green algae in a large nutrient limited shallow lake (Balaton, Hungary). Arch Protistenkunde 139: 225-242.

Paerl HW, Bland PT, Bowles ND, Haibach ME (1985) Adaptation to high intensity, low wavelength light among surface blooms of the cyanobacterium Microcystis aeruginosa. Appl Environ Microbiol 49:1046-1052.

Paerl HW, Huisman J (2008) Blooms like it hot. Science 320: 57-58

Paerl HW, Huisman J (2009) Climate change: a catalyst for global expansion of harmful algal blooms. Appl Environ Microbiol. 1: 27-37.

Pearl HW, Paul VJ (2011) Climate change: Links to Global Expansion of Harmful Cyanobacteria, Water Res. Doi: 10.1016/j.watres.2011.08.002

Rott, E., 1981. Some results from phytoplankton counting intercalibrations. Schweiz Z Hydrol 43: 34–63.

Ryding SO, Rast W (1989) The control of eutrophication in lakes and reservoirs. Unesco and Parthenon Publishing Group, 314 pp.

Reynolds CS, Walsby AE (1975) Water blooms. Biol Rev, 50: 437-481.

Reynolds CS, Huzar V, Kruk C, Naselli-Flores L, Melo S (2002) Towards a functional classification of the freshwater Phytoplankton. Review. J Plankton Res 24:417-428.

Reynolds CS (2006) Ecology of Phytoplankton. Ecology, Biodiversity and Conservation. Cambridge University Press, Cambridge, UK.

Robarts RD, Zohary T (1987) Temperature effects on photosynthesis capacity, respiration, and growth rates of bloom-forming cyanobacteria. New Zeal J Mar Fresh 21: 391-399.

R Development Core Team (2011) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-project.org/.

Schopf JW (2000) The fossil record: Tracing the roots of the cyanobacterial lineage. In: B.A. Whitton & M. Potts (Eds), The ecology of cyanobacteria. Dordecht, the Netherlands: KluwerAcademic Publishers: 13-15.

Salmaso N (1996) Seasonal variation in the composition and rate of change of the phytoplankton community in a deep subalpine lake (Lake Garda, Northern Italy). An application of non-metric multidimensional scaling and cluster analysis. Hydrobiologia 337:49-68.

Salmaso N, Mosello R, Garibaldi L, Decet F, Brizzio MC, Cordella P (2003) Vertical mixing as a determinant of trophic status in deep lakes: a case study from two lakes south of the Alps (Lake Garda and Lake Iseao). J Limonl 62(Suppl.1): 33-41.

Salmaso N, Padisàk J (2007) Morpho- Funcional Groups and phytoplankton development in two deep lakes (Lake Garda, Italy and Lake Stechlin, Germany). Hydrobiologia 578: 97-112.

66

Salmaso N, Boscaini A, Cappelletti C, Ciutti F (2009) Le condizioni di salute del Lago di Garda: aggiornamento dello stato delle conoscenze su carichi di nutrienti algali e sulle componenti biologiche della zona pelagica e litorale. In: F. Bertin & A. Bortoli (Eds), Problematiche ambientali del Lago di Garda. Approfondimenti e proposte di risanamento. Libro degli Atti: Torri del Benaco: 49-88.

Salmaso N, (2010) Long-term phytoplankton community changes in a deep subalpine lake: responses to nutrient availability and climatic fluctuations. J Freshwat Biol 55: 825–846.

Salmaso N, (2011) Interactions between nutrient availability and climatic fluctuations as determinants of the long term phytoplankton community changes in Lake Garda, Northern Italy. Hydrobiologia 660: 59–68.

Shatwell T, Köhler J, Nicklisch A (2008) Warming promotes cold-adapted phytoplankton in temperate lakes and opens a loophole for Oscillatoriales in spring. Global Change Biol 14: 1-7.

Sivonen K, Jones G (1999) Cyanobacterial toxins, I. Chorus, J. Bartram, Editors , Toxic cyanobacteria in water: a guide to their public health consequences, monitoring and management, E&FN Spon, London, 41–111.

Sommer U, Gliwicz ZM, Lampert W, Duncan A (1986) PEG-model of Seasonal Succession of Planktonic Events in Fresh Waters. Arch Hydrobiol 106(4): 433-471.

Stich HB, Brinker A (2011) Oligotrophication outweighs effect of global warming in a large, deep, stratified lake ecosystem. Global Change Biol 16:877-888.

Straile D (2000) Meteorological forcing of plankton dynamics in a large and deep continental European lake. Oecologia 122: 44-50.

Thackeray JS, Jones ID, Marberly SC (2008) Long-therm change in the phenology of spring phytoplankton: species-specific responses to nutrient enrichemnt and climatic change. Acta Oecol 2008, 98: 523-535.

Utermöl H (1958) Zur Vervollkommenung der quantitativen Phytoplankton Methodik. Mitteilungen der Internationalen Vereinigung für Limnologie 9 : 1-38.

Vollenweider RA, Kerekes J (1982) Eutrophication of waters. Monitoring, assessment and control. OECD Cooperative programme on monitoring of inland waters (Eutrophication control), Environment Directorate, OECD, Paris. 154p.

Weithoff G (2003) The concepts of “plant functional types” and “functional diversity” in lake phytoplankton- a new understanding of phytoplankton ecology? Freshwater Biol 48:1669-1675.

Walsby AE, Hayes PK, Boje R, Star LJ (1997) The Selective advantage of buoyancy provided by gas vesicles for planktonic cyanobacteria in the Baltic sea. J Phycol 136:407-417

Wilkinson GN, Rogers CE (1973) Symbolic descriptions of factorial models for analysis of variance. Appl Stat 22: 392–9.

Williamson CE, Saros JE, Vincent WF, Smol JP (2009) Lakes and reservoirs as sentinels, integrators and regulators of climate change. Limnol Oceanogr 54: 2273-82.

Winder M, Hunter DA (2008) Temporal organization of phytoplankton communities linked to chemical and physical forcing. Oecologia 156: 179–192.

Wetzel R.G., 2001. Limnology: lake and river ecosystems. 3rd edition, Academic Press.

Zuur AF, Ieno EN, Smith GM (2007) Analyzing Ecological Data (Statistics for Biology and Health); 672pp

67

Chapter four

4. Will Lake Geneva turn “red” in the future? A possible scenario for the development of the cyanobacterium Planktothrix rubescens

Gallina Nicole, Bensiton Martin, Jacquet Stéphan

Paper submitted in Limnology and Oceanography

68

Will Lake Geneva turn “red” in the future?

A possible scenario for the development of the cyanobacterium Planktothrix rubescens

Gallina Nicole1*, Bensiton Martin1, Jacquet Stéphan2

1. University of Geneva, Climatic Change and Climate Impacts Research Group, Institute for Environmental Sciences, Site de Battelle, route de Drize, CH 1227 Carouge, Geneva, Switzerland 2. INRA, UMR 042 CARRTEL, 75 Avenue de Corzent, 74203 Thonon-les-Bains cx, France

* Corresponding author: [email protected]

Keywords

Lake Geneva, Global warming, Statistical model, Extreme Events, Cyanobacteia, Planktothrix rubescens

Abstract Among the multiple forms of cyanobacteria, the phycoerythrin-rich species Planktothrix rubescens is well adapted to temperate, deep and large lakes. In Lake Geneva (the largest lake of Western Europe), the biomass of this filamentous and microcystin-producing species has become the dominant cyanobacteria species during the last decade. Air as well as water temperature influences on the occurrence and development of cyanobacteria are also particularly relevant to consider in the context of the climate global change, which may be particularly marked for lakes in the Alpine region, with a rate of warming twice as large as the global average. The impact of climate change on the cyanobacteria, especially P. rubescens, was analysed through two different approaches, by using extreme air temperature events as a proxy for future climate and the Multi Adaptive Regresssion Splines (MARS) model to predict future P. rubescens biomass. These methods allowed us to figure out whether Lake Geneva will turn red (owing to the color of P. rubescens) by the end of this century. The outcomes strongly suggest that cyanobacteria may gain in contribution by respect to the total phytoplankton community. Moreover, following expected temperature increase, the biomass of P. rubescens could be 30% more important by the end of this century, provided nutrients remain in the same range level. Additionally, the results point out that spring is the key season, during which air temperature and nutrients become the determinant factors for cyanobacteria outbreaks for the following seasons.

Introduction Long-term monitoring of aquatic ecosystems is likely to provide useful information to address key questions dealing with causes and consequences of changes in water quality. Lake Geneva, which is the largest natural lake in Western Europe, has been surveyed since 1957 as part of the international programme of water protection lead by the “Commission Internationale pour la Protection des Eaux du Léman (CIPEL; International Commission for the Protection of Lake Geneva Waters)”. A few long-term studies of Lake Geneva water quality and biology changes have already been published (Anneville & Leboulanger 2001, Anneville and al. 2002, 2007, Gillet & Quétin 2006, Tadonléké 2010, Jacquet and al. 2012). These studies have clearly revealed major changes in the chemical properties, especially phosphorus, and also largely debated on seasonal changes of plankton, more especially the phytoplanktonic community in relation to phosphorus concentrations and temperature increase, but also of the relationships between primary production and nutrients or global warming. None of these studies however focused on a particular group with the exception of the studies of Druart & Rimet (2009) and Rimet and

69 al. (2009) that reported the dynamics of pelagic diatoms in Lake Geneva over a >30-year period, from 1974 to 2007, and revealed the structural changes of this community along the re-oligotrophication of the lake. To the best of our knowledge, nothing has yet been published on a special group such as cyanobacteria (especially potentially toxic species) although they can be a key concern for lake management authorities and for which a large debate exists nowadays related to their development in relation to global climate change (Shatwell and al. 2008, Pearl & Huisman 2009, Gallina and al. 2011). Cyanobacteria are oxygenic photosynthetic prokaryotes and these microorganisms have often been reported to proliferate (i.e. form blooms) in response to ecosystem eutrophication. This often leads to oxygen depletion, a reduction in biodiversity resulting in the disruption of the trophic chain, a negative impact on the image of the lake, and possible problems for local fisheries as well as potential toxic risks, for both animal and human health, as a result of drinking the water or swimming (Chorus & Bartram 1999, Dokulil & Teubner 2000, Briand and al. 2003). As cyanobacteria are able to produce numerous toxins (hepatotoxins, neurotoxins, cytotoxins and lipopolysaccharide endotoxins), they can affect the structure and functioning of ecosystems and thus have a negative impact on animal and human health (Carmichael and al. 2001, Briand and al. 2003). Indeed, cyanobacteria are likely to have been responsible for the death of fishes (Rodger and al. 1994), wild birds (Matsunaga and al. 1999), cattle (Mez and al. 1997), dogs (Nehring 1993) and human beings (Jochimsen and al. 1998). This has created the need for a better understanding of the conditions favouring cyanobacterial growth and development, with the goal of providing a basis for the control and management of their occurrence and abundance (Chorus & Bartram 1999). As cyanobacteria are very tolerant to environmental changes, they could be favored under a changing climate, which is indeed expected for Lake Geneva in the course of the 21st century (Perroud and al. 2009). The question is thus whether climate changes (warming or extreme events) will favour the development of the cyanobacteria and counteract the re-oligotrophication of this ecosystem. This is a critical aspect for Lake Geneva which is ecologically and socio-economically important, because it supplies drinking water and contains important stocks of commercially important fishes such as the Arctic char, perch and whitefish. Among the different varieties of cyanobacteria, Planktothrix rubescens (Anagnostidis & Komarek 1988, synonym Oscillatoria rubescens Gomont 1892) is a red-colored, filamentous cyanobacterium with gas vesicles, which is able to adjust its buoyancy and thus its position in the water column, in order to obtain its preferred low light environment (and possibly also in response to other resources). It often forms deep water maxima in the stratified lakes of temperate latitudes (Feuillade 1994, Micheletti and al. 1998, Bright & Walsby 2000, Jacquet and al. 2005). Reynolds and al. (2002) described it as an R-strategist, i.e. belonging to 'attuning' or 'acclimating' strategists especially tolerant to physical perturbations, or even requiring turbulence to remain in suspension. They achieve relatively high growth rates even at low or variable light intensities, and their morphology (frequently elongated or trichal) provide a high surface/volume ratio, thus enabling rapid exchange of metabolites through the cell surface. P. rubecscens constitutes a perfect biological model to test the impact of global warming. Under current warming trends at the global and regional scales, extreme events such as heat waves are becoming more frequent (IPCC 2007) and are forecast to be the norm in a greenhouse climate by 2100. Therefore the use of current climate extremes as a proxy for tomorrow climate enables to assess possible impacts on the natural environment (Beniston 2007), and such a methodology has already been applied successfully to other deep sub-alpine lakes (Gallina and al. 2011). If many studies have pointed out the increasing risk of cyanobacterial blooms with climate change and that recent accumulating evidence suggested indeed that global warming will favour cyanobacteria (e.g. Paerl & Huisman 2008, Elliott 2010), a recent study also reported that heat waves may not necessarily promote cyanobacterial blooms (Huber and al. 2012) because of the importance of critical thresholds for some abiotic drivers. In this paper, we tested the hypothesis that there will be more cyanobacteria (exemplified by the toxic and filamentous species P. rubescens) in Lake Geneva under warmer climatic conditions projected for the coming decades of the 21st century. To address this goal, we employed two methods based on (i) a descriptive approach in which we used current climate extremes as a proxy for future mean climate and (ii) a quantitative analysis in which we used a statistical model (Multivariate Adaptive Regression Splines, MARS) (Leathwick and al. 2005) to predict the possible evolution of the biomass in the future. To our knowledge, this study presents the first application of MARS to model phytoplankton biomass on a temporal scale.

Methods

Study site and sampling strategy Lake Geneva is a deep peri-Alpine lake, located on the border between France and Switzerland. It is the largest natural western European lake situated at an altitude of 372 m, with an area of 582 km2, a maximum width and length of 13 km and 72 km respectively, and a volume of approximately 86x109 m3 for an average residence time of about 11 years. Lake Geneva is composed of two basins, a deep central eastern basin, also

70 called the “large lake” (“Grand Lac”), which registers 309 m at its deepest point, and a western and more shallow basin, the “small lake” (“Petit Lac”), with a maximum depth of 74 m (Figure 1). Defined as a warm monomictic lake, temperatures seldom drop below 5°C and this lake never freezes. Between 1974 and 2010, full mixing of the water column occurred only 7 times (1979, 1981, 1984, 1986, 1999, 2005 and 2006). Lake Geneva underwent rapid eutrophication starting in the 1960’s, reaching app. 90 µgP.l-1 around 1980. Following the program of reduction of phosphorus loading, initiated in the early 1970’s, it recovered and is an example of a restored ecosystem “in progress”. In 2011, Lake Geneva could be defined as a mesotrophic ecosystem (Jacquet and al. 2012). Water sampling was performed at the reference station called “SHL2”, situated at the deepest point (309 m) of the lake (Figure 1). Even though the data cover the period since 1974 for the phytoplankton time series, we considered in this study only the 2001-2008 period in order to avoid bias in the data, since in 2000 the sampling strategy changed (see explanations below, Figure 2). Samples were obtained each month from 1974 to 1980 and from 1980 onwards semi-monthly from March to November, when the biological activity is the highest (Tadonléké 2010). Water samples were collected in the upper euphotic layer (0-10 m until the end of 2000 and 0-20 m since 2001) using an integral sampler (Pelletier & Orand 1978). After Lugol fixation and sedimentation in Uthermöhl chambers, the phytoplankton samples were identified and counted using an inverted ZEISS microscope (Uthermöhl 1958). Physical and chemical parameters were sampled at discrete depths (0 m, 1 m, 2.5 m, 5 m, 7.5 m, 10 m, 12.5 m, 15 m, 20 m) at the same date and frequency than the phytoplankton samples and they were transformed into weighted averages and integrated over the 0-20 m water column. All these data were obtained from the CIPEL data set.

Venoge Aubonne Switzerland Morges Lausanne

300 100 150 200 250 Veveyse SHL2 250 Nyon 300 200 150 100 50 Evian 50 Thonon & N INRA meteo station Dranse Rhône

Large lake

Rhône Small lake France 0 5 10 km Geneva

Fig. 1 Study site: Lake Geneva separating Switzerland to France. The reference station named SHL2 corresponds to the deepest zone of the Lake. The meteorological station is also indicated and is situated at the INRA limnological research station in Thonon-les-Bains.

71

Data The phytoplankton biomass was expressed in µg L-1, whereas to assess the behaviour of the cyanobacteria during extreme events, their biomass was converted into the percentage relative to the total phytoplankton community to better state their contribution to the whole phytoplankton community. For further modelling purposes, we analysed the biomass of Planktothrix rubescens, as it was the most abundant species among cyanobacteria in Lake Geneva, representing over 45% of the total cyanobacteria biomass during the selected period (i.e. from 2001 to 2008) of this study. For modelling, predictors were created based on parameters known to influence the phytoplankton community. To exclude co-linearity among predictors and to enable no inference with the fitting process, only 13 predictors were selected from an initial set of 22 constructed predictors. The selection was performed upon the graphical outcome of a PCA analysis and the correlation coefficients between the descriptors. Descriptors having a correlation coefficient of r>0.8 were excluded, as for descriptors showing in the results of the PCA analyses an overlapping gradient, pointing either to the same or to the opposite direction (negative correlation) with the same strength. PCA analysis was performed using the ade4-package in R (Dray & Dufour 2007). The selected predictors could be gathered into five groups, namely into predictors related to nutrients (total Nitrogen and total Phosphorus, and the nitrogen to phosphorus ratio), to physical (temperature, water column stability), to light availability (euphotic depth and underwater light climate index), competition (number of species present in the sample, total phytoplankton biomass of the previous sample) and meteorological predictors (rainfall and wind speed) (Table 1). The stability of the water column was calculated using the Brunt–Väisälä frequency, according to the following formula:

2 g  d  N      dz  where N2 is the stability coefficient (in s-2), g is the acceleration parameter (in m s-2), which we assumed to be 9.81 (m s-2),  is the water density (no unit), z is the depth (in m), and the pressure calculated according to Lemming (1978) as follows:(T) = 1000 – 7x10-3 (T-4)2 and where T is the temperature (°C). We used the log of the maximum value of N2. To include light condition and limitation for phytoplankton growth, the euphotic depth (Zeu) was calculated by multiplying the Secchi depth by 2.5 (Vollenweider 1982), which corresponds to the depth at which the light irradiance is below 1% of that of the surface (Talling 1957). The mixing depth referred to as Zm , which was undertaken in help of the graphical and the numerical outputs of the temperatures profiles, was defined as the depth where the temperature gradient change suddenly (1°C per m) following Wetzel (2001). This led to the possibility to calculate the underwater light climate index translated by the Zm/Zeu ratio. To take into account the interspecific competition of phytoplankton, we counted the species present in the samples at each sampling date and labeled this index as DivNB, representing diversity in the numbers of species present. Another index used was the biomass which was present at the previous sampling date and which was labeled as totalBM-1. The meteorological predictors were recorded from the meteorological station situated at Thonon-les-Bains in the southern part of Lake Geneva, and belonging to the INRA Research Institute (Figure 1). Rainfall (mm), wind speed (log transformed) as well as the maximum wind speed (m/s) were considered. As air temperature was highly correlated with the water temperature, this parameter was not integrated into the model. In order to undertake predictions, simulated water temperature profiles for the period 2082-2089 were used. They were obtained from the model results of Perroud and al. (2009), in which future temperature profiles for Lake Geneva were simulated under the IPCC A2 Scenario, using a one dimensional lake model. As in modeling time-series, the main problem is obtaining accurate statistical values. Therefore to avoid misinterpretation, the auto-correlation structure in the data has to be taken into account. A standard method was used to integrate a synthetic auto-regressive variable, which is a time-lag of P. rubescens ((LAG-Pthrix), corresponding to the biomass of P. rubescens from the previous sampling.

Method used to define extreme temperatures Our goal was to define the extreme seasons during the period 2001-2008. Therefore, a 30-year reference period from 1971 to 2000 was first defined, in which air temperature averages of the common meteorological seasons were calculated, i.e. winter (December to February), spring (March to May), summer (June to August) and autumn (September to November). Secondly, the 99th percentile of the averaged seasonal air temperature distribution of the reference period was computed. These percentiles were then compared to the averages of the seasons from the assessed period 2001-2008. A season was defined as extremely hot, when its average value was beyond the percentile value. Once the extreme seasons defined, we analyzed the cyanobacterial biomass

72 contributions during these seasons in comparison to the “normal’’ season of the selected period using boxplot graphical presentation.

Multiple Adaptive Regression Splines (MARS) model description MARS is a flexible non-parametric regression modeling method introduced by Friedmann (1991), with the capacity to model complex, non-linear relationships between response and explanatory variables with similar levels of complexity to that of Generalized Additive Models (GAM) (Hastie & Tibshirani 1990). Its originality consists in the hinge functions breaking the full range of the predictors into subsets of linear segments using knots. The model fitting in MARS has the advantage to be very rapid and consists in two main phases: the forward and the backward pass, the latter was evaluated by using generalized cross-validation (GCV). This two- stage approach is the same as that used by recursive partitioning trees. The MARS function, implemented in the “mda”-package (Hastie & Tibshirani 2011), was applied using the free statistical software R, version 2.1.14 (R Development Core Team, 2012). In our study the code written by Leathwick and al. (2005) was used, having the added ability to fit a least-squares MARS model, followed by the extraction of the basis functions and the refitting of the model as a glm, adding the ability to specify a “family” distribution. Further statistical details can be found in Friedmann (1991) and Leathwick and al. (2006).

73

AUTOCORRELA- NUTRIENTS LIGHT RELATED CLIMATOLOGICAL PHYSICAL RELATED PREDICTORS COMPETITION RELATED PREDICTORS

TION CORRECTION RELATED PREDICTORS PREDICTORS RELATED PREDICTORS

–

-

Brunt

climate climate

-

light

windspeed

(m)

(°C)

LAG

index (mm) (m/s) (m/s)

depth

(mg/L) (mg/L)

(µg/L)

Rainfall

Euphotic

PREDICTORS

phorusratio

P.rubescence

to total phos total to

Total nitrogen Total

Total Nitrogen Total

Total Phosphorus Total väisäläfrequency

previoussampling

Numbers of species Numbersspecies of

Water temperatures Water Phytoplankton Total

Log of the the ofLog

Maximumwindspeed

present in the sample the in present

biomass present at the the at present biomass

Log of maximumofLog

Underwater

LAG-Pthrix TN TP TP:TN WT LogN2x Zeu Zm/Zeu DivNB Total BM-1 RF LogV Vx

Mean 201.5 0.49 0.004 38.4 12.3 5.21E-04 17.69 0.7 19.7 3.29 0.42 6.9

Range 0-784 0.2-0.8 0.027 38.4-82.9 5.7-19.3 1.36E-06 -2.61E-03 5.5-37.5 0-2.55 4-33 1.61-4.25 0-1.19 0-0.071 1.5-12.17

LAG-Pthrix 1 0.11 -0.12 0.23 0.25 0.38 0.15 0.19 0.02 0.12 0.05 -0.13 -0.13

TN 1 0.51 0.28 -0.68 -0.60 0.49 0.46 0.56 -0.24 0.32 0.15 0.1

TP 1 0.51 0.61 -0.43 0.13 -0.51 -0.63 -0.35 0.38 0.14 -0.02

TP/TN 1 0.09 -0.07 -0.13 0.19 0.12 0.01 0.21 -0.01 -0.04

WT 1 0.79 0.74 0.51 0.67 0.46 0.29 -0.19 0.02

LogN2x 1 -0.62 0.31 0.51 -0.26 0.2 -0.12 0.14

Zeu 1 -0.52 -0.48 -0.54 -0.13 -0.06 -0.17

Zm/Zeu 1 0.41 0.33 0.36 -0.12 0.12

DivNB 1 0.36 0.23 -0.18 0.04

Total 1 0 -0.1 0.01 BM-1

RF 1 0 0.29

LogV 1 0.78

Vx 1

Table 1: Predictors used to fit P. rubescens. Mean, range (minimum and maximum values) of the predictors from August to February 2001-2008 are shown in the first two first lines in italic, followed by the Spearman rank correlation coefficient matrix. The significant correlations (α < 0.01, n=64) are represented in bold and larger font.

74

The model was fitted on P. rubescens biomass from August to February, as the aim was to assess the leading predictors involved during the main growth period and to predict its variation of biomass under warmer water temperatures. The Poisson Error distribution family was applied. The selected model was evaluated by the correlation coefficient derived from the relationship between the observed and the fitted values as well as a k- fold (k = 10) cross validation algorithm of the MARS/GLM model (Hastie and al. 2001), which is better at estimating the expected prediction error across training sets. The residuals were checked for homogeneity and plotted against the fitted values. The prediction of P. rubescens biomass was applied to a dataset, in which all predictors remained constant except for water temperature. For water temperature, we used the outcomes of Perroud and al. (2009) as already mentioned above. The simulated water profiles were used and further averaged for the 0-20m layer, on the same days as the phytoplankton was sampled during the 2001-2008 period. In this manner, the influence on P. rubescens of predicted water temperature, i.e. under future warming scenario, could also be investigated.

Results

Long-term phytoplankton biomass variation in Lake Geneva (1974-2008) The yearly averages calculated for the biomass of total phytoplankton and cyanobacteria are represented in Figure 2A. The total phytoplankton biomass has an overall average of 1,840 µg L-1. Its biomass displayed a slight decrease at the start of the sampling period and management program in 1974 and seems to be stabilized from the beginning of the 1980s till the mid-1990s, when an increase began. Such increases became important from 2001 onwards, with two important peaks, namely in 2001 and 2007, respectively. It is noteworthy that this abrupt biomass alteration in 2001 coincided with the change operated in the sampling procedure (increase of the water depth sampling from 0-10 to 0-20 m) so that interpretations relating to this increase need to be taken with precaution. The average biomass from 2001 onwards was almost 2,900 µg L-1, corresponding to nearly twice the average recorded during the 1974 - 2000 period (1,500 µg L-1); diversity (species number) also increased in the same range (not shown). The cyanobacteria are not an abundant group in Lake Geneva and represent around 9.7% of the total phytoplankton biomass. However, its importance rose from 2001 and its biomass average recorded till 2008, was 2.5 times higher (with a value of 334 µg L-1) than the biomass averaged from 1974 - 2000 (130 µg L-1). This resulted in a contribution increase of 2% in relation to the phytoplankton community, leading to a final contribution of 11%. Figure 2B illustrates the variation of P. rubescens among the cyanobacterial community from 1974 to 2008. Until the beginning of the 1990s, P. rubescens was sporadic and even absent some years, but reached important concentrations during the autumns of 1972, 1982, and 1985. From the early 1990s till 2001, P. rubescens became more abundant and from 2001 onwards, its biomass increased and P. rubescens became the dominant cyanobacteria species among the community (45%). Since 2001, the biomass peaks of P. rubescens have coincided with the peaks of the total cyanobacterial community. The monthly variation of P. rubescens form 2001 to 2008 is illustrated in Figure 2C. In Lake Geneva, P. rubescens starts its growing phase gently in the middle of the summer and its biomass gradually increases until it reaches its highest biomass concentration in November, with a median value of about 600 µg L-1, and decreases progressively thereafter, and disappears by the end of March.

Assessed extreme seasons during the period of 2001-2008 The variation of the yearly averaged seasonal air temperature is shown in Figure 3. The dotted lines represent the 99th percentile value of the air temperature distribution from the reference period 1971-2000 for each season, i.e. 5.08°C, 12.51°C, 21.51°C and 14.17°C for winter, spring, summer and autumn respectively. Three out of the 34 seasons (i.e. summer 2003, autumn 2006 and winter 2007) were above these fixed limits and were defined as extreme seasons. The summer of 2003 and the winter of 2006/2007 were clearly above the threshold. Furthermore, the winter of 2001, the springs of 2001, 2003 and 2007, and the summer of 2006 were considered as relatively warm seasons. Finally, it is to be noted that for the years 2003 and 2007, springs were warmer than autumns.

Cyanobacterial behaviour during extreme seasons Cyanobacteria in Lake Geneva start their growing phase slowly in summer (10%) and contribute to 20% of the total phytoplankton biomass in autumn and decrease in winter by half, to reach less than 5% in spring. Their contribution to the remaining phytoplankton communities varies from less than 5 to 20% during normal seasons, as illustrated in Figure 4A. Interestingly, during the three selected extreme hot seasons, the contribution of the cyanobacteria was much more important compared to the seasons in which the air temperature was defined as

75

“normal”; it doubled during autumn 2006 and summer 2003 and a contribution three times more important than normal was recorded during the winter of 2006/2007.

Fig. 2 (A) Annual mean concentrations of the total phytoplankton vs. the cyanobacteria biomass and (B) monthly evolution of the total cyanobacteria vs. Planktothrix rubescens biomass, in Lake Geneva between 1974 and 2008. The dashed vertical lines show the period when the sampling strategy was changed (from 0-10 m to 0-20 m). Fig. 2C illustrates the monthly biomass variation of P. rubescens assessed from 2001 to 2008, indicating the main growth period starting at the end of the summer to reach the maximum biomass by the end of autumn to decrease and disappear in spring.

76

25

sum

20 C) ° aut 15

spr

10 Air Temperature Air ( Temperature

win 5

0 2001 2002 2003 2004 2005 2006 2007 2008

Winter Spring Sommer Automn

Fig. 3 Bar charts of the yearly seasonal mean air temperature registered at the weather station at INRA Thonon-les-Bains. The dashed horizontal grey line visualizes the value of the 99th percentile of the seasonal air temperature computed from a 30 yrs reference period (sum = value of the 99th summer percentile value, aut = same for autumn, spr = same for spring, win = same for winter). The bars black rimmed in bold, indicated with an arrow, are the three seasons defined as extreme seasons during the 2001-2008 period.

In order to analyse the behaviour of the cyanobacterial community in the season preceding and following an extreme hot event, biomass variation was investigated (Figure 4B). After the summer heat wave in 2003, cyanobacteria maintained a significantly high contribution in autumn 2003, even though air temperature was not recorded to be higher (Figure 3). The spring of 2007 looked particular, as it was a season that followed two consecutive extreme warm seasons (autumn 2006, winter 2006/2007) and was in itself a relatively warm spring. The cyanobacteria were present with almost 20% of the total phytoplankton biomass during this season, which is a very important contribution as in spring normally their presence is close to zero. The season prior to an extreme season was also analysed in order to investigate for the presence of a cyanobacterial inoculum that could potentially contribute to their further development. Spring 2003 was a warm spring season and the cyanobacteria were already present with a higher percentage than during normal springs, namely around 10%. Therefore the cyanobacterial pool was already present before the summer 2003 heat wave began to impact the community. The situation during summer 2006, preceding the anomalously warm autumn 2006 and winter 2006/2007 seasons was completely different. The cyanobacterial community contribution increased from just above 0% to about 40%. Moreover, during summer 2006 their contributions were even smaller compared to summers in which air temperatures were “normally” distributed.

Predictors and their contribution to the fitted P. rubescens model Table 1 represents the statistical summary of the predictors used to fit P. rubescens to the MARS model, namely the mean and the range of the values with their maxima and minima for the period of August to February, 2001 to 2008. The spearman rank correlation coefficient r in between the predictors is listed on the second half of table one, showing in larger bold font significant correlations (α < 0.01, n = 64). As already stated, correlations having a coefficient r > 0.8 were excluded in order to avoid co-linearity and miss fitting. Therefore 13 predictors finally were selected and subdivided into different classes.

77

Fig. 4 (A) Seasonal variation of the cyanobacterial biomass among the phytoplankton during the “normal” seasons related to the extreme hot seasons. Fig 4B extends Fig 4A by adding seasons which proceed and succeed the extreme seasons. The numbers stand for the year and the letters are the abbreviations for the seasons (a = autumn, sp = spring, su = summer, w = winter).

The integrated lag of P. rubescens (LAG-Pthrix), used to accurate the auto-correlative behaviour of the response variable, was not strongly related to any of the other predictors, even though it significantly correlated to the physically-related predictors and the TP:TN ratio. As expected, the nutrients (TP, TN) were correlated significantly with all predictors except to wind speed and the LAG-Pthrix. Surprisingly, TP was not correlated significantly with the euphotic depth, while TN was. The TN:TP ratio only correlated with the nutrients and the LAG-Pthrix predictor. Water temperature was highly correlated with the water column stability, nutrients and light as well as the competition-related predictors. It had, however, a lower relation with rainfall and LAG- Pthrix, and no relation with wind speed. The stability of the water column was highly and negatively correlated with the euphotic depth, but highly and positively with the DivNB present in the sample and, as already stated, with the water temperature. A smaller and negative relationship was found for Total BM-1. The light-related predictors did not significantly correlate with the meteorological predictors, and the euphotic index did not interact with the total phosphorus concentration. The pressure of competition did not seem to depend on the meteorological predictors. However, rainfall was positively correlated with the underwater light climate index, both TN and TP concentration, water temperature, and wind speed.

Selection and validation of the P. rubescens model The results indicate that the MARS model was able to reproduce P. rubescens biomass during its “growing” period. The final model selected the LAG-Pthrix, the underwater light climate index (Zm/Zeu), the water temperature (WT) and the rainfall (RF) as the predictors which contribute the most to fit the response of P. rubescens (figure 5A). The highest contribution to the model was achieved by the LAG-Pthrix, followed by WT, Zm/Zeu and RF. The selected predictor’s response curves characterised P. rubescens as having a strong positive

78 response to warmer water temperature, and a higher under-water light climate index, but a declining response when rainfall increases.

Fig. 5 Illustrations of the MARS model outputs of the selected predictors fitting best P. rubescens biomass. For each predictor, the form of the response function of P. rubescens is shown. The four predictors selected are the auto-correlative lag of P. rubescens biomass (LAG-Pthrix), the rainfall (RF), the underwater light climate index Zm/Zeu, and the water temperature (WT).

The goodness-of-fit of the model was assessed by the correlation between the actual response and the fitted response and reached a r = 0.75, indicating that 57% (R2 = 0.57) of P. rubescens variation could be explained by the model (figure 5B). Further, the accuracy estimation of the selected model was tested via a 10-fold cross- validation test, and the full test score achieved an r = 0.75 whereas the mean score of the subsets reached a score of r = 0.58.

Prediction of P. rubescens P. rubescens under a future climate scenario was predicted by using the warmer water temperatures simulated by Perroud and al. (2009) for the period 2082-2089 in Lake Geneva. The output of the predictive model (2082-2089) was further compared to the output of the calibrated model on the observed data (2001- 2008). The results are shown in Figure 6, the bold line, with a slope of one and crossing the origin represents the limit in which no changes occur. Clearly, higher biomass of P. rubescens under future water temperature was forecasted, provided the other predictors do not change. The biomass of P. rubescens increased in frequency but also in intensity. August and September were the months in which P. rubescens biomass almost doubled, and in October the increase was still significant. However, from November to February the increase was more modest. The proliferation exceeded the maximum value of 800 µg L-1 during the observed period, to attain a concentration over 1000 µg/l. For the end of the century, a total biomass difference of 5100 µg L-1 was simulated for P. rubescens, corresponding to an increase of 32% related to the 2001-2008 period.

79

Fig. 6 (A) Model results using the observed temperature data from 2001-2008 vs. the model results using the water temperature predicted by the end of this century. The bold regression line shows the limit if no changes in biomass of P. rubescens occur, whereas the dashed line represents the regression in-between both model outcomes. Fig 6B highlights the observed yearly variation of P. rubescens biomass (black bars) and the monthly averaged differences in biomass predicted by the MARS model for the period 2082-2089.

80

Discussion

The results of this research confirm our initial hypothesis and clearly suggest that cyanobacteria, especially the dominant species P. rubescens in Lake Geneva for the period of study, may become more abundant and frequent in this lake under future climate conditions, in which air and water temperatures have been projected to rise (Beniston 2007, Perroud and al. 2009). Even though comparative analyses did not reveal any significant difference in the species composition and biomass when changing the sampling strategy from 0-10 m to 0-20 m (CIPEL, 2002), our research clearly indicate a difference not only in biomass, but also a change in the phytoplankton community. The investigated period from 2001-2008 was marked with two exceptional biomass peaks which occurred in 2001 and 2007, both due to Mougeotia gracillima, a filamentous conjugate which proliferated excessively from July to August 2001 and from July to November 2007. For both years, a warmer than usual spring was recorded, which led to an earlier onset and longer duration and particularly a deeper water column stratification, resulting in an earlier and deeper depletion of the nutrients in the water column, as already documented on surface waters of Lake Geneva (Anneville & Leboulanger 2001, Nõges 2004). This situation may have favoured M. gracililma, known to be tolerant for low light levels and capable of developing deeper in the water column, where phosphorus is abundant (Anneville and al. 2002). Therefore, we may state the hypothesis that warm spring events could be favourable to the massive M. gracillima outbreaks that occurred in 2001 and 2007 in Lake Geneva. Interestingly, even though management efforts to reduce phosphorus loads in Lake Geneva have been undertaken for quite a long time, the total phytoplankton biomass started to increase in the beginning of the 1990s. This contradictive non-linear pattern called hysteresis (Gawler 1988), has already been reported for deep and large peri-alpine lakes (Anneville 2002, Jacquet and al. 2012). This recovering period, which differs from the route to eutrophication (Ibelings 2007), is also characterized by a positive and significant trend of air temperature increase (non-parametric Mann-Kendall time series trend analysis, p = 0.004). Concerning P. rubescens, this species was more present in the 1990s when P loads started to decrease, and this is in concordance with its optimal growth when phosphorus concentration is intermediate and typical of mesotrophic conditions (Jacquet and al. 2005). Since 2001, P. rubescens developed more frequently and became the dominant species among the cyanobacterial community. The development has been shown to be higher near the thermocline indicating the metalimnic preference of P. rubescens between 10 and 20 m, as already observed elsewhere (Reynolds 1984, Dokulil & Teubner 2000, Jacquet 2005), and thus emphasized the fact that during the previous sampling strategy, P. rubescens, together with other metalimnic (M. gracillima) important species, have probably not been collected adequately. The impact of global warming on the stratification period, together with the decreasing phosphorus loads due to improved management efforts, may have favored the increased growth of P. rubescens during the analyzed period. Therefore, nowadays it is an important concern to evaluate the indistinct effect of global warming on this hysteresis process, as it might potentially contribute through a synergistic effect to the increase of the phytoplankton biomass. This investigation clearly showed the discrepancy of the data sets resulting from both sampling strategies and therefore the absolute need to apply statistical techniques to adjust the collected dataset in order to use with confidence these entire long-term records, which unfortunately are still surprisingly rare. The general potential of extreme air temperature to impact cyanobacterial growth in five peri-alpine lakes, excluding the largest basin of Lake Geneva, was also highlighted by Gallina and al. (2011) through a synoptic analysis in other peri-alpine lakes, in British and Irish lakes (George and al. 2010), boreal lakes (Grahm & Vinebrooke 2009), Dutch lakes (Jöhnk and al. 2008). However the effect of air temperature has different forms and depends on the characteristics of each lake, such as its morphology, geographic position, or mixing regime (Livingstone 1993, 1997, Blenckner 2005). Moreover, each lake has its own specific phytoplankton community (Gallina and al. submitted), so that the effect of extreme events may have different repercussions on the cyanobacteria community, depending on the investigated lake. The present research was a concrete case study and may serve as an example to investigate specific lake type. In Lake Geneva during each extreme hot season, namely in summer 2003, autumn 2006 and winter 2006/2007, the percentage of the cyanobacteria increased significantly with respect to “normal” seasons. This suggested the potential of air temperature to impact the phytoplankton community in creating more favorable conditions for this group. However, Huber and al. (2012) clearly documented that extreme summer temperatures may not always be the cause for cyanobacterial biomass increase, and highlighted the importance of the differences in the intensity and duration of thermal stratification, related to differences in short-term weather variability. In our study, the same conclusion could be drawn. Each of the investigated extreme hot seasons was preceded by a warmer spring, known to influence positively the stratification strength and duration. Therefore it is suggested that in years where the air temperatures may be ‘extremely hot”, a warm spring might be the decisive factor to evaluate if an enhanced cyanobacterial proliferation may take place, by metalimnic species, adapted to low underwater light climate and furthermore having the ability to be buoyant and able to move

81 through the water column to search for optimal nutrient and light conditions in the deeper metalilmnion (Jacquet and al. 2005). The increase of contribution as well as the persistence of this increase until the following season suggested the capacity of air temperature to induce a community change. Changes in species composition are often observed in phytoplankton communities and result from a complex interplay between physical and chemical properties of the aquatic environment on the one hand (Wagner & Adrian 2011) and the responsiveness of the individual species on the other hand (Dignum and al. 2005). However, these results clearly highlight the additional important role of the climatological predictors that are capable, not only to influencing biological in- lake processes, but also to altering the phytoplankton community composition. These conclusions for Lake Geneva are in concordance with the above mentioned research on the synoptic study in which other peri-alpine lakes were assessed (Gallina and al. 2011) with the potential of air temperature to impact the cyanobacteria. However through the synoptic study, any community change was suggested, as the phytoplankton proportion of the different group stayed maintained. This difference could originate by the fact that the present study focus on one specific lake, having its own response, whereas through the synoptic study by Gallina and al. (2011), the individual responses per lake could have been balanced out, as each lake may responds differently to climate variation (Bleckner 2005). Of the numerous factors influencing cyanobacterial dynamics, nutrients, water column stability and competition-related predictors did not show the capacity to contribute in an important manner to the behavior during their growth periods. Nutrients have been invoked to be one of the variables controlling cyanobacteria since the earliest days of phytoplankton ecology (Whipple and al. 1899, Husted 1939). It has been confirmed by many studies and nutrients can be responsible for the seasonal phytoplankton succession (Killham and al. 1982, Reynolds 1984, Sommer 1986, Anneville and al. 2002) and cyanobacterial growth (Pearl 2001, Briand and al. 2005, Havens 2008, Gallina and al. submitted). Additionally the N:P ratio has been shown to have a dramatic effect on the algal species composition, and even though later controversially disputed, a “N:P” rule emerged, indicating that if the ratio is lower than 29, cyanobacteria will be favored (Smith 2005). However, in this investigation nutrients have not been selected to contribute as predictors to the final model, which fitted the best P. rubescens growth from August to February. Likewise, recent studied suggested the importance of the nutrient concentration during the spring season, as being the main nutrient pool responsible for the phytoplankton growth during its main growth period in summer and autumn (Gallina and al. submitted). Interestingly, Salmaso (2012) stated that nutrients will change their status during the seasonal succession depending on the season: In spring, nutrients could be considered as explaining variable and once absorbed in summer - autumn by the phytoplankton become a response variable. These could explain why nutrients were not selected by the MARS model, as during the analyzed season, they were already transformed and thus considered as a response variable. Similar results were achieved by Dupuis and al. (2009) who stated that of several factors influencing phytoplankton dynamics, total dissolved nutrients [nitrogen (N), phosphorus (P) and N:P] and water column stability did not show important changes between years. To predict future growth of P. rubescens, the most contributing predictors selected by the MARS model were climatological variables (rainfall) or directly related climatological variables namely water temperature (influenced by air temperature in the upper layers of the water column) and the underwater light climate index by the incident light. This leads to the conclusion that a changing climate will inevitably impact P. rubescens growth, competition and development. In detail, P. rubescens responded positively to an increase of the water climate index, indicating the tolerance to low insolation of their autumnal environments due to their large light antennae (Reynolds 1997), as suggested previously (Micheletti and al. 1998, Jacquet and al. 2005). In contrast, the response of P. rubescens to higher rainfall decreases. As a metalimnic species (Davis and al. 2003, Walsby and al. 1999, Reynold 1984) favored by a stable water column (Dokulil & Teubner 2000), heavy rainfall (that was always associated to strong wind) could play a disturbance factor of its habitat preference. A strong positive response was assessed on the P. rubescens biomass once the water temperature increased. Water temperature is a driving factor in aquatic ecosystems (Paerl and al. 1985, Robarts & Zohary 1987, De Stasio, 2009, Gallina and al. submitted) and Dubuis (2009) also stated that water temperature was related to changes in general phytoplankton and the percentage of cyanobacterial biovolume. The analogy between the conclusions of other studies with the selection of the most contributing variables added with the good results of the model validation suggested that the MARS model is an appropriate and reliable prediction tool for the growth of P. rubescens. In the future, under higher water temperature regimes, a phenology change is predicted to take place for P. rubescens, in which the most important change (increase) will take place in late summer and early autumn. The advances of the growth season from autumn to summer was already seen by R strategists in the nineties due to warmer water temperatures, as they are well adapted to the deeper metalimnic with low light conditions (Anneville and al. 2002). Moreover, P. rubescens is forecasted to be not only more frequent but also present under higher biomass under future climate scenario in Lake Geneva. The increased growth of P. rubescens in stratified lakes under warmer conditions in lakes was already reported, and is the cause of the synergetic effect of increased transparency due to the reduction in the phosphorus loads, the deepening of the P-depleted zone and

82 water column stability (Jacquet and al. 2005, Nõges and al. 2010). Moreover, in general, under climate change scenarios, harmful cyanobacteria, known as temperature-sensitive groups, have already been reported in recent studies to become more important (Pearl & Huisman, 2008, 2009, Jöhnk and al. 2008). With the potential to be harmful and toxic and therefore generating numerous sanitary and economic problems, these results need to be taken into consideration by water management authorities to find proper solutions, in a world where the ecology and the quality of aquatic ecosystems are shown to be negatively influenced by global warming.

Acknowledgements We are grateful to J.-C. Druart and F. Rimet who realised phytoplankton determination and counts. The long- term record of data on Lake Geneva was financially supported by CIPEL and comes from the sampling and work of many people including J.C. Druart, F. Rimet, J.C. Hustache, P. Chifflet, J. Lazzarotto, P. Perney, and G. Monet. This project was supported through an assistantship of the University of Geneva and contributes indirectly to the research objectives of the EU/FP7 “ACQWA” project (www.acqwa.ch) funded by the European Union under Grant Nr. 212250.

83

References

Anagnostidis, K., J. Komarek. 1988. Modern approach to the classification system of cyanophytes. 3. Oscillatoriales. Arch. Hydrobiol. 80, 327–472.

Anneville, O. and C. Leboulanger. 2001. Long-term changes in the vertical distribution of phytoplankton biomass and primary production in Lake Geneva: a response to the oligotrophication. Alti Associazione Italiano Oceanologia Limnologia 14:25-35

Anneville, O., V. Ginot, J.-C. Druart, N. Angeli. 2002. Long-term study (1974–1998) of seasonal changes in the phytoplankton in Lake Geneva: a multitable approach. J. Plank. Res. 24, 993–1007.

Anneville, A., J.C. Mollinero, S. Souissi, G. Balvay and D. Gerdeaux. 2007. Long-term changes in the copepod community of Lake Geneva. J. Plank. Res. 29:49-59.

Beniston, M. 2007. Entering into the "greenhouse century": recent record temperatures in Switzerland are comparable to the upper temperature quantiles in a greenhouse climate. Geophysical Research Letters, 34, L16710.

Briand, J.-F., S. Jacquet, C. Bernard and J.-F. Humbert. 2003. Health hazards for terrestrial vertebrates from toxic cyanobacteria in surface water ecosystems. Vet. Res. 34:361-378.

Briand, J.-F., S. Jacquet, C. Flinois, C. Avois-Jacquet, C. Maisonnette, B. Leberre and J.-F. Humbert. 2005. Variations in the microcystins production of Planktothrix rubescens (cyanobacteria) assessed by a four years in situ survey of Lac du Bourget (France) and by laboratory experiments. Microb. Ecol. 50:418-428.

Bright, D. I. And A. E. Walsby. 2000. The daily integral of growth by Planktothrix rubescens calculated from growth rate in culture and irradiance in Lake Zurich. New phytol. 146:301-316.

Carmichael, W.W., S.M.F.O. Azevedo, J.S. An, R.J.R. Molica, E.M. Jochimsen, S. Lau, K:L Rinehart, G.R. Shaw, G.K. Eaglesham, G.K., 2001. Human fatalities from cyanobacteria: chemical and biological evidence for cyanotoxins. Environ. Health Perspect. 109 (7):663–668.

Chessel, D. and A.B. Dufour, and J. Thioulouse. 2004. The ade4 package-I- One-table methods. R News. 4: 5- 10.

Chorus, I. and J. Bartram. 1999. Toxic Cyanobacteria in Water: a Guide to Public Health Significance, Monitoring and Management.

Davis, P.A, M. Dent, J. Parker, C.S Reynolds and A.E Walsby. 2003. The annual cycle of growth rate and biomass change in Planktothrix spp. In Blelham Tarn, English Lake District.

Dignum, M, H.C.P Matthijs, R. PEL, H.J. Laanbroek, R. Luuc. 2005. Nutrient limitation of freshwater Cyanobacteria. Tools to monitor Phosphorus limitation at the individual level. In J.Huisman, H.C.P Matthijs and P.M. Viser (eds), Harmful Cyanobacteria, 65-86. 2005 Springer. Printed in the Netherlands.

Dupuis, F. and B. J. Hahn. 2009. Warm spring and summer water temperatures in small eutrophic lakes of the Canadian prairies: potential implications for phytoplankton and zooplankton. Journal of Plankton Research. Volume 31(5): 489-502.

Dokulil, M. T. and K. Teubner. 2000. Cyanobacterial dominance in lakes. Hydrobiol. 438:1-12.

Dray, S. and Dufour, A.B. 2007. The ade4 package: implementing the duality diagram for ecologists. Journal of Statistical Software. 22(4): 1-20.

Druart, J.C. and F. Rimet. 2009. Dynamique du peuplement de diatomées pélagiques du Léman de 1974 à 2007. Archives des Sciences, 61: 17-32.

Elliott, J.A. 2010. The seasonal sensitivity of cyanobacteria and other phytoplankton to changes in flushing rate and water temperature. Glob Change Biol 16:864–876

84

Feuillade, J. 1994. The cyanobacterium (blue-green alga) Oscillatoria rubescens D.C. Arch. Hydrobiol. 41:77- 93.

Friedman, J.H. 1991. Multivariate Adaptive Regression Splines. Ann. Statist. 19:1-67.

Gallina, N., O. Anneville, M. Beniston. 2011. Impacts of extreme air temperatures on cyanobacteria in five deep peri-Alpine lakes. Journal of Limnology 70(2): 186-196.

Georg, G., E. Jennings, N. Allot. 2010. The impact of climate change on lakes in Britain and Ireland. D.G. George. (ed.). The impact of Climate Change on European Lakes. Aquatic Ecology Series 4. Springer Science + Buisness Media B.V. 2010.

Graham, M. D and Rolf Vinebrooke (2009). Extreme weather events alter planktonic communities in boreal lakes. Limnology and Oceanography. 54 (6 part 2): 2481-2492.

Hastie, T. and Tibshirani, R. 1990. Generalized Additive Models. Chapman & Hall.

Hastie, T. and Tibshirani R.(S Language) . Original R port by Friedrich Leisch, Kurt Hornik and Brian D. Ripley. (2011). mda: Mixture and flexible discriminant analysis. R package version 0.4-2. http://CRAN.R- project.org/package=mda

Havens, K.E. Cyanobacteria blooms: effect on aquatic ecosystems. 2008; 619:799-47

Huber, V, C. Wagner, D. Gerten D, Adrian R. 2012. To bloom or not to bloom: contrasting responses of cyanobacteria to recent heat waves explained by critical thresholds of abiotic drivers. Oecologia 169:245–256

IPCC. 2007. Climate Change 2007: Synthesis Report. Contribution of Working Groups I, II and III to the Fourth Assessment. Report of the Intergovernmental Panel on Climate.Change [Core Writing Team, Pachauri, R.K and Reisinger, A.(Eds)]. IPCC, Geneva, Switzerland: 104 pp.

Jacquet, S., J.-F. Briand, C. Leboulanger, C. Avois-Jacquet, G. Paolini, L. Oberhaus, B. Tassin, B. Vinçon-Leite, J.-C. Druart, O. Anneville and J.-F. Humbert. 2005. The proliferation of the toxic cyanobacterium Planktothrix rubescens following restoration of the largest natural French lake (Lac du Bourget). Harmful Algae 4:651-672

Jacquet, S., I. Domaizon and O. Anneville. 2012. Evolution de paramètres clés indicateurs de la qualité des eaux et du fonctionnement écologique des grands lacs péri-alpins (Léman, Annecy, Bourget): Etude comparative de trajectoires de restauration post-eutrophisation. Archives des Sciences (accepted)

Jochimsem, E. M., W. W. Carmicheal., J. D. M. Ancardo, S. T. Cookson; C. E. M. Holmes, M. B. Antunes, T. M. Lyra, V. S. T. Barreto, S. M. F. O .Azevedo, W. R. Jarvis. 1998. Liver failure and death after exposure to microcystins at a hemodyalisis center in Brazil. New Engl J. Med., 338:873-878.

Leathwick, D. Rowe, J. Richardson, J. Elith, and T. Hastie. 2005 Using Multivariate Adaptive Regression Splines to Predict the Distributions of New Zealand's Freshwater Diadromous Fish. Freshwater Biology, 50, 2034-2052.

Leathwick, J.R., J. Elith, and T. Hastie. 2006. Comparative performance of generalized additive models and multivariate adaptive regression splines for statistical modelling of species distributions. Ecological Modeling, 199: 188-196.

Matsunaga, H, K. Harada, M. Senma, Y. Ito , N. Yasuda, S. Ushida, Y.Kimura. 1999. Possible cause of unnatural mass death of wild birds in a pond in Nishinomiya, Japan: sudden appearance of toxic cyanobacteria. Nat Toxins 7(2):81– 84

Mez, K, K. A. Beattie, G. A. Codd, K. Hanselmann , B. Hauser , H.P. Naegeli, H.R. Preisig. 1997. Identification of a microcystin in benthic cyanobacteria linked to cattle deaths on alpine pastures in Switzerland. European Journal of Phycology. 32: 111–117.

85

Nehring, S. 1993. Mortality of dogs associated with a mass development of Nodularia spumigena (Cyanophyceae) in a brackish lake at the German North Sea coast. J Plankton Res 15: 867-872.

Paerl ,H.W. and J. Huisman. 2008. Blooms like it hot. Science 320:57–58

Pearl, HW, R.S. Fulton, P.H. Moisander, J. Dyble. 2001. Harmful freshwater algal bloom, with an emphasis on cyanobacteria. Scientific World journal, 1: 76-113.

Paerl, H.W. and J. Huisman, 2009. Climate change: a catalyst for global expansion of harmful cyanobacterial blooms. Environmental Microbiology Reports 1: 27-37

Pelletier, J. & A. Orand. 1978. Apareil de prélèvement d’un échantillon dans un fluide. INRA patent Perroud, M., S. Goyette, A. Martynov, M Beniston and O. Anneville. 2009. Simulation of multiannual thermal profiles in deep Lake Geneva: A comparison of one-dimensional lake models. Limnol. Oceanogr. 54: 1574- 1594.

R Development Core Team (2012). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-project.org/.

Reynolds C.S., and Walsby A.E. 1975. Water blooms. Biol. Rev. 50:437–81.

Reynolds, C., V. Huszar, C. Kruk, L. Naselli-Flores and S. Melo. 2002. Towards a functional classification of the freshwater phytoplankton J. Plankton Res. 24 (5): 417-428.

Reynolds, C.S. 2006. Ecology of Phytoplankton. Ecology, Biodiversity and Conservation. Cambridge. University Press, Cambridge, UK.

Rimet, F., J.C. Druart and O. Anneville. 2009. Exploring the dynamics of plankton diatom communities in Lake Geneva using emergent self-organizing maps (1974-2007). Ecological Informatics, 4: 99-101.

Rodger, H. D., T. Turnbull, C. Edwards, and G.A. Codd. 1994. Cyanobacterial (blue-green algal) bloom associated pathology in brown trout, Salmo trutta L., in Loch Leven, Scotland. Journal of Fish Diseases, 17: 177–181.

Shatwell, T., J. Köhler and A. Nicklish. 2008. Warming promotes cold-adapted phytoplankton in temperate lakes and opens a loophole for Ocillatoriales in spring. Global Change Biology 14: 1-7.

Shapiro, J. 1972. Blue-green algae: why they become dominant. Science 179:382–4.

Sommer, U. 1986. The periodicity of phytoplankton in Lake Constance (Bodensee) in comparison to other deep lakes of central Europe. Hydrobiologia.138: l-7.

Strawcynski, A. and F. Pasquini . 2002. Analyses comperatives interlaboratoires 2001. Rapport Commission internationale pour la protection des eaux du Léman (CIPEL), Campagne 2001, 155-165.

Smith, V.H. 2005. Low nitrogen to phosphorus ratios favor dominance by blue-green algae in lake phytoplankton. Science 983;221:669–71.

Tadonléké, R.T. 2010. Evidence of warming effects on phytoplancton productivity rates and their dependence on eutrophication status. Limnology & Oceanographie 55: 973–982

Talling J. F. 1957. The phytoplankton population as a compound photosynthetic system. New phytologist, 56, 133-149.

Uthermöhl H. 1958. Zur vervollkommnung der quantative phytoplankton-methodik. Mitt. Inst. Verh. Limn. 9: 1- 38

Utermöl, H., 1958. Zur Vervollkommenung der quantitativen Phytoplankton Methodik. Mitteilungen der Internationalen Vereinigung für Limnologie 9 : 1-38.

86

Vollenweider, R.A., and J. Kerekes, 1982. (OECD, 1982). Eutrophication of waters. Monitoring, assessment and control. OECD Cooperative programme on monitoring of inland waters (Eutrophication control), Environment Directorate, OECD, Paris. 154p

Wagner C, R. Adrian R. 2011. Consequences of changes in thermal regime for plankton diversity and trait composition in a polymictic lake: a matter of temporal scale. Freshwater biology 56, 1949–1961.

Wetzel, 2001. Limnology: lake and river ecosystems. 3rd edition, Academic Press

Whipple G.C., Jackson D.D. 1899. Asterionella – its biology, its chemistry and its effect on water supplies. New English water work association 1-25.

87

Chapter five

5. Conclusions and perspectives

The aim of this study was to draw a picture of what might be the behavior of phytoplankton, with an emphasis on the cyanobacteria community in lakes within the sensitive peri-Alpine region. The aims and objectives of this study were accomplished and the following findings should give useful advice and insights to water management authorities to help these adapt strategies to handle future warmer temperature which have been projected by different climate change scenarios (IPCC, 2007).

 The results clearly indicate that air temperatures have the capacity to influence phytoplankton groups in general, and suggest that future warmer climate enhance the increase of phytoplankton biomass, especially cyanobacteria, which have been seen to be particularly sensitive to temperature increase. However, with respect to heat waves, that were chosen as a means of representing a proxy for a future climate, no differences in response between toxin producing and non-toxin-producing species were established. Furthermore, it is suggested that the diversity among phytoplankton groups will more likely decrease, with a pronounced effect on cyanobacteria, which were shown to be the most sensitive group to respond to warm but also cold extreme events. This loss in diversity could be due to the emergence of genera well adapted to a changing environment.

 In a general manner, it can be excluded that cyanobacteria have the potential to dominate the phytoplankton community in peri-Alpine lakes under a future climate scenario. However, the case study on Big Lake Geneva reveals that in the future cyanobacteria will gain in contribution and therefore be able to introduce a shift in the community composition.

 The main drivers of phytoplankton configuration in peri-Alpine lakes were investigated on different levels:

a) As hypothesized, the phytoplankton configuration throughout the year was mainly driven with similar strengths by temperature (air and water) and nutrients (P, N), but also responded significantly to the duration of the stratification period and grazing by the Cladoceran (herbivorous zooplankton). Cyanobacteria behaved differently to these gradients, Oscillatoriales were closely related to higher water temperature and longer stratification period, Chroococcales and small colonies were found were temperatures were higher and nitrates more concentrated, whereas as large non-vacuolated cyanobacteria colonies and Nostocales preferred condition with lower nitrate concentration.

88

b) During the growth season (summer-autumn) only nutrients (spring nutrient concentrations, summer-autumn nutrient concentrations) and temperature (water and air) influenced significantly the phytoplankton configuration. Similarly, during the growth period, Oscillatoriales were driven by the water temperature gradient, Nostocales by lower nitrate concentration and Chroococcales and small colonies by higher nutrients concentration. However, during the growth period, large non-vacuolated cyanobacteria colonies still prefer low nitrate concentration, but were also driven by higher phosphorus concentration.

o These outcomes clearly highlight that among the cyanobacteria, especially the filamentous cyanobacteria, Oscillatoriales will be favored with increased water temperature and longer stratification period. As both mechanisms are predicted to occur under future climate change scenarios, Oscillatoriales may thus become more abundant.

o Moreover, the potential of spring nutrient concentrations to impact significantly the phytoplankton growth during its main growth period in summer-autumn highlights their important role as fertilizers, and is able to control the seasonal succession of phytoplankton during the remaining year.

o The cyanobacteria community responds differently to the environmental drivers taken into account in these studies. Hence, for future phytoplankton assessment, it seems more judicious to sub-divide the cyanobacteria community into different groups with common morpho-functional traits, to obtain a more ecological and functional response of their behavior.

o Phosphorus and nitrate concentration, together with temperature (air and water) have the greatest impact, as hypothesized, on the phytoplankton community, but this study reveals that these drivers act independently to each other on the community configuration.

 Additionally, this synoptic study on peri-Alpine lakes brought into light the important finding that each of the lakes studied has a statistically-significant different phytoplankton community, which was much more important than previously supposed, as these lakes were all deep lakes, situated in the same region. Among the investigated hydro-mophometrical, climatological and in-lake drivers, the altitudinal position of the lakes, causing a water temperature gradient, seems to an important factors differentiating the phytoplankton communies among peri-alpine lakes. For future synoptic studies, it is therefore advised to well discern lakes at higher altitudes (northern part of the alpine arc, Chapter two) with lakes situated at lower elevations (on the southern alpine arc). Interestingly, it was found that the phytoplankton community composition between lakes is thus primarily influenced by this temperature gradient. Hence, the trophic level of those lakes, thus the P concentration do not seems to play a primordial role of the phytoplankton composition of each lake.

89

These conclusions have in turn stimulated the investigation of what might be the behavior of Planktothrix rubescens under warmer climatic conditions; this because of Planktothrix belongs to the Oscillatoriales, which were found to be favored when temperatures are rising. Only one single lake was considered, as the previous studies revealed an important differentiated behavoiur of the phytoplankton communities inbetween lakes. Thus, Big Lake Geneva was further investigated and represented a case study, in which extreme warm air temperatures were for a second time applied as proxy for future climate conditions, together with the statistical model MARS. MARS allowed to define the most important predictors for their behavior, and furthermore permitted to model Planktothrix rubescens biomass under warmer water temperature conditions.

 Seasonal extreme warm air temperatures led to higher cyanobacteria biomass. Additionally, it was suggested that in years where the air temperatures may be ‘extremely hot”, a warmer spring might be the decisive factor. Warmer springs lead to a longer, stronger stratification period, thus resulting in lower nutrient concentrations in summer in the epilimnion, as depleted earlier. Warmer springs could therefore play an important role to an enhanced cyanobacteria proliferation in early summer, by Planktothrix rubescens, a metalimnic species, adapted to the low underwater light climate and furthermore having the ability to be buoyant and able to move through the water column to search for optimal nutrient and light conditions in the deeper metalimnion. Moreover, due to its filaments and its potential capacity to produce toxic compounds, P.rubescens is not affected by the grazing pressure occurring in spring and summer.

 The selected predictors to model P. rubescens were climate related predictors: rainfall,

the ratio zm/zeu,, and water temperature. These predictors are vulnerable to be disrupted when under the influence of climate change, thus indicating the potential of P. rubescens to be affected under these scenarios. Surprisingly, nutrients do not play an important role during their growth period in summer-autumn. It thus was proposed what Salmaso and al. (2012) already hypothesized, that nutrients are decisive predictors during the spring period when having a role as fertilizers, but become a response variable once taken up by the phytoplankton during their main growth period in summer-autumn.

 Under future water temperature scenarios (2082-2089), P. rubescens biomass increased not only in frequency but also attained higher biomass. Their growth period was predicted to start earlier, which is translated into an almost doubled biomass during August and September.

 In addition, during warmer spring years in 2001 and 2007, an excessive proliferation of the filamentous conjugate Mougeotia gracillima was recorded in Lake Geneva (Big and Small), leading to important ecological problems not only in relation to the water quality, but also affected the fishing industry on the Lake. During this investigation, it was concluded that these warmers in spring lead to the optimal condition for this

90

metalimnic species to be well adapted under low underwater light climate conditions and having low grazing pressure.

 In general, the strength and duration of the stratification period was demonstrated to have essential physical role in shaping the biomass, the composition, and phenology of cyanobacteria. As warming has been shown to enhance the stratification period, cyanobacteria are consequently suggested to be favored under these conditions.

The methods applied during the thesis seemed to be well adapted to the issues investigated, thus allowing to fulfill the principal objectives that were initially defined. The assessment of phytoplankton through Morpho-Functional Groups (MFG) was efficient and allowed easier interpretation as less functional units compared to genera were involved and moreover had the aptitude to show the same functional behavior. MFG thus highlights the evolution of traits under environmental drivers and therefore may enrich the knowledge in terms of fundamental plankton research. The use of extreme events as proxies for future climate conditions was seen to be an effective substitute when long term data are missing. This method not only has the advantage to be fairly simple to use, but also was improved in its reproducibility because it was employed for different datasets. Moreover, the use of the percentiles of the reference period rather than the original sets of data provide a measure of objective interpretation of what may constitute a cold or a warm extreme, that can be transposed from one lake to another even if the absolute temperature values may be different for these selected thresholds.

The satisfactory results in the model validation suggest that MARS is an appropriate prediction tool that can be used with confidence to assess the temporal succession of phytoplankton species, even though it is a rather difficult goal to assess biological variables in different lakes, due to their complex interrelation with other variables. However, in a next step, it might be interesting to apply the community function of MARS which allows the assessment of up to nine different species. Moreover, as each lake has been shown to respond differently to temperature, and for reasons of a better model validation, it would be necessary to apply MARS on other peri-Alpine lakes.

The assembled peri-Alpine matrix including nine different datasets belonging to seven deep lakes covering the whole trophic gradient and including an altitudinal gradient represents in scientific terms, a highly interesting base for additional research, and therefore further analyses are under preparation and should be continued in order to respond to advanced questions related to phytoplankton behavior in deep peri-Alpine lakes. The differential phytoplankton compositions in between lakes should be analyzed in more details with the aim to better define its origin. This question is important when it comes to assess ‘’synoptically’’ the impact of future global warming on the phytoplanktonic behavior over a certain region.

However the ‘’static’’ statistical modeling exhibits certain limits. To complete and validate the results it is suggested here to apply powerful, dynamic deterministic models, which are able to take into account a majority of the key processes governing the complex behavior of the phytoplankton/cyanobacteria community. These dynamic models should be primarily

91 applied to the lakes located at lower altitudes which were shown in this study to represent a special threat to cyanobacteria outbreaks.

Warmer air temperatures have different repercussions on diverse groups among the phytoplankton/cyanobacteria community. This study clearly indicates Oscillatoriales being the most favored cyanobacteria group under warmer temperatures. Moreover as the relationship between air and water temperatures indicated a stronger impact of air temperature on water temperature in lakes situated at lower altitudes, those lakes have the inherent capacity of meeting the ideal conditions for enhanced growth of Oscillatoriales. It is thus suggested that Oscillatoriales might proliferate with higher biomass and earlier in summer, bringing along the capacity to introduce a community composition change. It is therefore advised that water management authorities should pay special attention to lakes at lower altitudes in the peri-Alpine regions, which potentially have an enhanced sensitivity to filamentous cyanobacteria outbreaks under future climate conditions.

It is advised that as soon as warmer than normal spring occur, increased monitoring should be undertaken in order to allow detecting potential harmful cyanobacteria outbreaks earlier. This would help to take adapted measures to prevent possible negative impacts on water quality and drinking water supply

It is primordial to maintain both, P and N nutrient inputs as small as possible, as these factors have seen to impact significantly the phytoplankton composition. This is even more so as eutrophication tends to favor cyanobacteria outbreaks and was stated to increase under future climate impact scenarios. However, to restore the water quality of lakes, the control of nutrient inputs are of a major concern for lake managements authorities, as confronted by climate change affecting negatively the oligotrophication efforts undertaken. Additionally, the non-linear phenomenon of hysteresis tends to add ambiguity in the complex interactions between climate change impacts, nutrients and phytoplankton biomass. Therefore it might be recommended to further define a range of nutrient concentrations which will tend to maintain a constant phytoplankton biomass.

Mitigation strategies to counter the possibility of harmful cyanobacteria outbreaks in sub- alpine lakes should be seriously considered in policy terms, but certainly represent a significant challenge. Global warming is the reality of today and to halt climate change is an additional challenge in a world where population is growing and CO2 emissions are increasing. Mitigation solutions need to be integrated into collective thinking, in order to highlight the importance of freshwater ecosystems as the only source for drinking water supply and to preserve and reinstating as many of our natural environments as possible.

92

6. Annexes

Annexe 1: The PEG Model: 24 Sequential Statements of Seasonal Succession of Plankton in Freshwater (Sommer and al., 1986.)

1. Towards the end of winter, nutrient availability and increased light permit unlimited growth of the phytoplankton. A small crop of fast-growing algae such as Cryptophyceae and centric diatoms develops.

2. This crop of small algae is grazed upon by herbivorous zooplanktonic species which become abundant due both to hatching from resting eggs and to high fecundity by high levels of edible algae.

3. Planktonic herbivores with short generation duration times increase their populations first and are followed by slower growing species.

4. The herbivore populations increase exponentially up to the point at which their density is high enough to produce a community filtration rate, and so cropping rate that exceeds the reproduction rate of the phytoplankton.

5. As a consequence of herbivore grazing, the phytoplankton biomass decreases rapidly to very low levels. There then follows a ‘clear-water’ equilibrium phase which persists until inedible algae species develop in significant numbers. Nutrients are re-cycled by the grazing process and may accumulate during the ‘clearwater’ phase.

6. Herbivorous zooplanktonic species become food-limited and both their body weight per unit length and their fecundity declines. This results in a decrease in their population densities and biomasses.

7. Fish predation accelerates the decline of herbivorous planktonic populations to very low levels and this trend is accompanied by a shift towards a smaller average body size amongst the surviving crustaceans.

8. Under the conditions of reduced grazing pressure and sustained non-limiting concentrations of nutrients, the phytoplankton summer crops start to build up. The composition of the phytoplankton becomes complex both due to the increase in species richness and due to the functional diversification into small ‘undergrowth’ species (which are available as food for filter-feeders) and large ‘canopy’ species (which are only consumed by specialists such as raptors or parasites).

9. At first, the edible Cryptophyceae and inedible colonial green algae become predominant. They deplete the soluble reactive phosphorus to nearly undetectable levels.

10. From this time onwards, the algal growth becomes nutrient-limited and this prevents an explosive growth of ‘edible’ algae. Grazing by predator-controlled herbivores balances the nutrient-limited growth rate of edible algal species.

11. Competition for phosphate leads to a replacement of green algae by large diatoms, which are only partly available to zooplankton as food.

93

12. Silica-depletion leads to a replacement of the large diatoms by large Dinoflagellates and/or Cyanophyta.

13. Nitrogen depletion favors a shift to nitrogen-fixing species of filamentous blue-green algae.

14. Larger species of crustacean herbivores are replaced by smaller species and by rotifers. These small species are less vulnerable to fish predation and are less affected by interference with their food collecting apparatus which can be caused by some forms of inedible algae. Accordingly, their population mortality is lower and their fecundity is higher than that of the larger species.

15. The smaller species of herbivores coexist under a persistent fish predation pressure and the increased possibility of food partitioning which is associated with the greater species complexity of the phytoplankton.

16. The population densities and species composition of the zooplankton fluctuate throughout the summer, the latter being also influenced by temperature.

17. The period of autogenic succession is terminated by factors related to physical changes which include increased mixing depth resulting in nutrient replenishment and a deterioration of the effective underwater light climate.

18. After a minor reduction in algal biomass, an algal community develops which is adapted to being mixed. Large unicellular or filamentous algal forms appear. Among them diatoms become increasingly important with the progress of autumn.

19. This association of poorly-ingestible algae is accompanied by a variable biomass of small, edible algae.

20. This algal composition together with some reduction in fish predation pressure leads to an autumnal maximum of zooplankton which includes larger forms and species.

21. A reduction of light energy input results in a low or negative net primary production and an imbalance with the algal losses which causes a decline of algal biomass to the winter minimum.

22. Herbivore biomass decreases as a result of reduced fecundity due both to lower food concentrations and to decreasing temperature.

23. Some species in the zooplankton produce resting stages at this time, whereas other species produced resting stages earlier.

24. At this period in the year, some cyclopoid species ‘awake’ from their diapause and contribute to the over-wintering populations in the zooplankton.

94

Annexe 2 : Assessed phytoplankton genera and their corresponding Morpho- Functional Groups (MFG)

CHLOROPHYTA CHRYSOPHYSEAE CYANOBACTERIA GROUPE GENERA MFG GROUPE GENERA MFG GROUPE GENERA MFG CHL Acanthosphaera 8a CHR Amphirhiza 2a CYA* Anabaena 5e CHL Actinastrum 11b CHR Bicosoeca 2a CYA* Aphanizomenon 5e CHL Ankistrodesmus 9d CHR Bitrichia 1a CYA Aphanocapsa 5d CHL Ankyra 9b CHR Chromulina 2a CYA Aphanothece 5d CHL Asterococcus 11b CHR Chrysamoeba 2a CYA Chamaesiphon 4 CHL Botryococcus 11b CHR Chrysarachnion 1a CYA Chroococcus 5d CHL Carteria 3a CHR Chrysidiastrum 1a CYA* Coelosphaerium 5d CHL Characium 8a CHR Chrysococcus 2a CYA Cyanobacterium 4 CHL Chlamydocapsa 11c CHR Chrysolykos 2a CYA Cyanodictyon 5d CHL Chlamydomonas 3a CHR Codosiga 2a CYA Cyanothece 4 CHL Chlorella 9b CHR Desmarella 2a CYA Dactylococcopsis 5d CHL Chlorochytrium 9b CHR Dinobryon 1a CYA DivChroocSmall 5d CHL Chlorococcum 9b CHR DivChrysoflagBig 1a CYA DivOscillatoriales 5a CHL Chlorolobion 9b CHR DivChrysoflagMed 2a CYA Gleitlerinema 5a CHL Choricystis 9d CHR DivChrysoflagSmall 2a CYA Gloeocapsa 5c CHL Coelastrum 11b CHR DivRhizocrysidales 2a CYA Gomphosphaeria 5d CHL Coenochloris 11b CHR Epipyxis 1a CYA Jaaginema 5a CHL Coenocystis 11b CHR Erkenia 2a CYA Komvophoron 5a CHL Coronastrum 11a CHR Kephyrion 2a CYA Lemmermaniella 5d CHL Crucigenia 11b CHR Mallomonas 1a CYA Leptolyngbya 5a CHL Crucigeniella 11b CHR Monochrysis 2a CYA Limnothrix 5a CHL Dictyosphaerium 11b CHR Monosiga 2a CYA Lyngbya 5a CHL Didymocystis 11b CHR Ochromonas 2a CYA Mantellum 4 CHL DivChloroBig 8a CHR Pseudokephyrion 2a CYA Merismopedia 5d CHL DivChloroMed 9b CHR Pseudopedinella 2a CYA Microcystis 5c CHL DivChloroSmall 9b CHR Salpingoeca 2a CYA* Oscillatoria 5a CHR Sphaeroeca 1a CYA Pannus 5c CHL Elakatothrix 11b CHR Spiniferomonas 2a CYA Phormidium 5a CHL Eudorina 3b CHR Spumella 2a CYA Planktolyngbya 5a CHL Eutetramorus 11b CHR Stelexomonas 1a CYA* Planktothrix 5a CHL Fotterella 11c CHR Stylochrysalis 2a CYA Pleurocapsa 5c CHL Geminella 10a CHR Syncrypta 1a CYA* Pseudanabaena 5a CHL Gloeotila 10a CHR Synura 1a CYA Rhabdoderma 5d CHL Golenkinia 8a CHR Uroglena 1a CYA Rhabdogloea 5d CHL Gonium 3b CYA Snowella 5d CHL Hyaloraphidium 8a CYA Spirulina 5a CHL IndChlamydo 3a DIATOMS CYA* Synechococcus 4 CHL Keratococcus 9b GROUPE GENERA MFG CYA Synechocystis 4 CHL Kirchneriella 9b DIAT Acanthoceras 6a CYA Trichonema 5a CHL Klebsormidium 10a DIAT Achnanthes 7b CYA* Woronichinia 5c CHL Koliella 8a DIAT Amphora 7a * Potential toxines producing species CHL Korschpalmella 11b DIAT Asterionella 6b CHL Korshikoviella 8a DIAT Aulacosira 7a DINOPHYCEAE CHL Lagerheimia 8a DIAT Caloneis 6b GROUPE GENERA MFG CHL Lobocystis 11b DIAT Campylodiscus 6b DIN Amphidinium 2b CHL Micractinium 11c DIAT CentrDiatBig 6a DIN Ceratium 1b CHL Monoraphidium 9b DIAT CentrDiatMed 7a DIN Diplopsalis 1b CHL Nephrochlamys 11b DIAT CentrDiatSmall 7a DIN DivDinoMed 2b CHL Nephrocytium 11b DIAT Chaetoceros 7a DIN DivDinoSmall 2b CHL Nephroselmis 3a DIAT Cocconeis 6b DIN Glenodinium 1b CHL Oedogonium 10a DIAT Cyclostephanos 7a DIN Gonyaulax 1b CHL Oocystis 11b DIAT Cyclotella 7a DIN Gymnodinium 1b CHL Pandorina 3b DIAT Cymatopleura 6b DIN Katodinium 2b CHL Papenfussiomonas 3a DIAT Cymbella 6b DIN Peridiniopsis 1b CHL Paradoxia 11a DIAT Denticula 6b DIN Peridinium 1b CHL Paramastix 3a DIAT Diatoma 6b DIN Woloszynskia 1b CHL Paulschulzia 11b DIAT Diploneis 6b CHL Pediastrum 11a DIAT Epithemia 6b CHL Phacotus 3a DIAT Eunotia 7b CHL Planktonema 10a DIAT Fragilaria 6b CHL Planktosphaeria 11b DIAT Gomphonema 6b CHL Pseudoquadrigula 11b DIAT Gyrosigma 6b CHL Pseudosphaerocystis 11b DIAT Meridion 6b CHL Pteromonas 3a DIAT Navicula 7b CHL Quadrigula 11b DIAT Neidium 7b CHL Rayssiella 11b DIAT Nitzschia 6b CHL Scenedesmus 11a DIAT PennDiatMed 7b CHL Schroederia 8a DIAT PennDiatSmall 7b CHL Scourfieldia 3a DIAT Pinnularia 7b CHL Selenochloris 3a DIAT Rhizosolenia 6a CHL Stichococcus 10a DIAT Rhoicosphenia 6b CHL Tetrachlorella 11b DIAT Stauroneis 6b CHL Tetraedron 9b DIAT Stephanocostis 7a CHL Tetraselmis 3a DIAT Stephanodiscus 7a CHL Tetrastrum 11a DIAT Surirella 6b CHL Thorakochloris 11b DIAT Synedra 7b CHL Treubaria 9b DIAT Tabellaria 6b CHL Trochiscia 9b DIAT Thalassiosira 7a CHL Ulothrix 10a CHL VolvocalesMed 3a CHL VolvocalesSmalll 3a CRYPTOPHYCEAE CHL Volvox 3b GROUPE GENERA MFG CHL Westella 11b CRY Chilomonas 2d CHL Willea 11b CRY Chroomonas 2d CRY Cryptaulax 2d CRY Cryptomonas 2d CONJUGATOPHYCEAE CRY Cyathomonas 2d GROUPE GENERA MFG CRY DivCryptoBig 2d CON Actinotaenium 8a CRY DivCryptoSmall 2d CON Closterium 8a CRY Katablepharis 2d CON Cosmarium 8a CRY Plagioselmis 2d CON Mougeotia 10b CRY Rhodomonas 2d CON Spirogyra 10b CON Spondylosium 10b CON Staurastrum 8a CON Staurodesmus 8a CON Zygnema 10b

95

7. Bibliography

Adams D.G., Duggan P.S., 1999. Heterocyst and akinete differentiation in cyanobacteria. New Phytol. 144 : 3-33

Adrian R., O’Reilly C.M, Zagarese H., Bainesd S.B., Hessene D., Keller W., Livingstone D.M., Sommaruga R., Straile D., Van Donk E., Weyhenmeyer G.A., Winder M., 2009. Lakes as sentinels of Climate change, 2009. Limnology and Oceanographie. 6: 2283–2297.

Ambrosetti W., Barbant L., Rolla A., 2006. The climate of Lago Maggiore area during the last fifty years. Journal of Limnology 65: Suppl. 1: 1-62.

Anderson D., Gilipert P., Burkolder J., 2002. Harmful algal blooms and eutrophication: nutrient sources, composition and consequences. Estuar. Coast 25:704-726.

Anneville O., Ginot V., Druart J.-C., Angeli N., 2002. Long-term study (1974–1998) of seasonal changes in the phytoplankton in Lake Geneva: a multitable approach. J. Plank. Res. 24, 993– 1007.

Anneville O., Souissi S., Gammeter S., Straile D., 2004. Seasonal and inter-annual scales of variability in phytoplankton assemblages: comparison of phytoplankton dynamics in three peri-alpine lakes over a period of 28 years. Freshwater Biology 49: 98–115.

Arrigo K. 2005. Marine microorganism and global nutrient cycles. Nature 437: 349-355.

Bates, B.C., Kundzewicz Z.W., Wu S., and Palutikof J.P., 2008. Climate Change and Water. Technical Paper of the Intergovernmental Panel on Climate Change, IPCC Secretariat, Geneva, 210 pp.

Bates, D. & Maechler M., 2009. lme4: Linear mixed-effects models using S4 classes. R package version 0.999375-31. http://CRAN.R-project.org/package=lme4.

Benincà E., Huisman J., Heerkloss R., Jöhnk K.D, Branco P., Van Nes E.H, Scheffer M. and Ellner S. 2008. Chaos in a long-term experiment with a plankton community. Nature (Letters) 451: 822-825.

Beniston M., Diaz, H.F., R.S. Bradley. 1997. Climate change at high elevation sites: an overview. Climatic change 36: 233-251.

Beniston M. 2004. The 2003 heat wave in Europe: a shape of things to come? An analysis based on Swiss climatological data and model simulations. Geophysical Research Letters,31, L02202, doi: 10.1029/2003 GL018857.

Beniston M. 2007. Entering into the "greenhouse century": recent record temperatures in Switzerland are comparable to the upper temperature quantiles in a greenhouse climate. Geophysical Research Letters, 34, L16710.

Bleckner T. 2005. A conceptual model of climate related effects on lake ecosystems. Hydrobiologia 533: 1-14.

Boehrer B., and Schultze M.. 2008. Stratification of lakes, Rev. Geophys., 46:1-27.

Briand, J.-F., Jacquet S., Bernard C., and Humbert J.-F., 2003. Health hazards for terrestrial vertebrates from toxic cyanobacteria in surface water ecosystems. Vet. Res. 34:361-378.

96

Briand J.-F., Leboulanger C., Humbert J.-F., Bernard C. and Dufour P., 2004. Cylindrospermopsis raciborskii invasion at mid-latitudes: Selection, wide physiological tolerance or global warming? Journal of Phycology, 40(2): 231-238.

Cayelan C.C., Ibelings B.W, Hoffmann E.P, Hamilton D.P, Brooks J.D., 2011. Eco-physiological adaptions that favor freshwater cyanobacteria in a changing climate. Water research. Doi:10.1016/j.watres.2011.12.016

Castenholz R.W., 2001. Oxygenic photosynthetic bacteria. In: Boone D.R. and Castenholz R.W. (eds.), Bergey’s manual of systematic bacteriology Vol.1 (2nd edn). Springer-Verlag, New York, pp 473-599.

Castenholz R.W, Garcia-Pichel F. Cyanobacterial Responses to UV-Radiation. 2002. In: The Ecology of Cyanobacteria. Springer Netherlands. 2002.

Carpenter S. R. et Cottingham K.L. 1997. Resilience and Restoration of lakes. Conservation ecology. Vol. 1.

Carey C.C, Ibelings B.W., Hoffmann E.P., Hamilton D.P., Brookes J.D., 2011. Eco-physiological adaptations that favour freshwater cyanobacteria in a changing climate. Water Research. 46 : 1394- 1407.

CIPEL, 2003. Rapp. Comm. Int. prot. Eaux Léman contre pollution, champagne 2002. p.69-83.

Codd G.A and Beattie, K.A., 1989. Cyanobacterial toxinsin water. Water Science and Technology. 21:1-13.

Codd G.A., Bell, S.G., Kaya k., ward, C.J, Beattie K.A. and Meatcalf, j.S. 1999. Cyanobacterial toxins, exposures routes and human health. European journal of Phycology 34, 405-415.

Crossetti L.O., Bicudo D.C., Bicudo C.E.M. and Bini L.M., 2008. Phytoplankton biodiversity changes in a shallow tropical reservoir during the hypertrophication process. Brazilian Journal of Biology, 68(4): 1061-1067.

Chorus I. and Bartram J. (eds.), 1999. Toxic cyanobacteria in water: a guide to public health significance, monitoring and management. E & FN Spon/Chapman & Hall, London, united Kingdom, 416 p.Clark J.S., Carpenter S.R., Barber M..

Dokulil M. T. and K. Teubner. 2000. Cyanobacterial dominance in lakes. Hydrobiol. 438:1-12.

Downing J.A., Watson S.B., and McCauley E., 2001. Predicting cyanobacteria dominance in lakes. Canadian Journal of Fisheries and Aquatic Sciences, 58(10): 1905-1908

De Mott W.R., and Müller-Navarra D.C., 1997. The importance of highly unsaturated fatty acids in zooplankton nutrition: evidence from experiments with Daphnia, a cyanobacterium and lipid emulsions. Freshwater Biology, 38(3): 649-664.

De Stasio B.T., and Golemgeski T., 2009. Temperature as a driving Factor in aquatic Ecosystems. Encyclopedia of Inland Waters. 690-698.

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, OJ L 327, 22.12.2000, p. 1–73.

97

Dyhrman S.T. and Ruttenberg K.C., 2006. Presence and regulation of alkaline phosphatase activity in eukaryotic phytoplankton from the coastal ocean: implications for dissolved organic phosphorus remineralization. Limnology and Oceanography, 51(3): 1381-1390.

Elliott J.A., Jones I.D., Thackeray S.J 2006. Testing the sensitivity of phytoplankton communities to changes in water temperature and dnutreinet load, in a temperate lake. Hydrobiologia, 559:401:411.

Elliott J.A. and May L. (2008). The sensitivity of phytoplankton in Loch Leven (UK) to changes in nutrient load and water temperature. Freshwater Biology, 53, 32-41.

Elliott J.A., 2010. The seasonal sensitivity of Cyanobacteria and other phytoplankton to changes in flushing rate and water temperature. Global Change Biology, 16, 864-876.

Fogg G.E., Stewart W.D.P., Fay P. and Walsby A.E., 1973. The blue-green algae. Academic Press, London, United Kingdom, 459 p.

Fulton III R.S., and Pearl H.W., 1987. Toxic and inhibitory effect of the blue-green alga Microcystis aeruginosa on herbivorous zooplankton. Journal of Plankton Research, 9(5):837-855.

Gallina N., and Cordonier A., Nirel P.M.; 2009. Essay on the characterisation of environmental factors structuring communities of epilithic Diatoms in the major rivers of the canton of Geneva, Switzerland. 2009. Milieu&Vie.

Georg G., U. Nickus, Dokulil M.T. and Blenckner T., 2010a. The influence of changes in the atmospheric Circulation on the Surface Temperature of Lakes. In D.G George (ed.). The impact of climate change on European Lakes (p. 293-310). Aquatic Ecology Series 4. Springer Science +Buisness media B.V. 2

Grossman A.R., Bhaya D. and He Q., 2001. Tracking the light environment by cyanobacteria and the dynamic nature of light harvesting. The Journal of Biological Chemistry, 276(15): 11449-11452.

Hardin G. 1960. The competitive exclusion principle. Science 131:1292-1298.

Harris G. P. 1986. Phytoplankton ecology: structure, function and fluctuation. Chapman and Hall, London.

Hastie T. and Tibshirani R.(S Language) . Original R port by Friedrich Leisch, Kurt Hornik and Brian D. Ripley. (2011). mda: Mixture and flexible discriminant analysis. R package version 0.4-2. http://CRAN.R-project.org/package=mda

Hedger R.D., Olsen N.R.B., George D.G., Malthus T.J. and Atkinson, P.M., 2004. Modelling spatial distributions of Ceratium hirundinella and Microcystis spp. In a small productive British lake. Hydrobiologia, 528: 217-227.

Hyenstrand P., P. Nyvali, A. Peterson, and P. Blomquist. 1998. Regulation of non-nitrogen-fixing cyanobacteria by inorganic sources – experimants from Lake Erken.

Hoffmann, H.J. 1976. Precambrian microflora, Belcher Islands, Canada: Significance and systematics. J. Paleontol., 50:1040-1073.

Hoiczyk E. and Hansel A., 2000. Cyanobacterial cell walls: news from an unusual prokaryotic envelope. Journal of Bacteriology, 182(5): 1191-1199

Huisman J., van Oostveen P. and Weissing F.J., 1999. Species dynamics in phytoplankton blooms: incomplete mixing and competition for light. The American Naturalist, 154(1): 46-68.

98

Huisman J. and Hulot F., 2005. Population dynamics of harmful cyanobacteria. Factorsaffecting species composition. In: Huisman J., Matthijs H.C.P. and Visser P.M. (eds.) Harmful cyanobacteria. Springer-Verlag, Dordrecht, The Netherlands, pp 143-176.

Hutchinson G.E. 1961. The paradox of the plankton. The American naturalist. 95:137-145.

Hutchinson G.E. 1967. A treatise on Limnology. Vol 2. NY. Wiley.

Ibelings B. W., Gsell A.S., Mooij W.M, Vand Donk E., Van den Wyngaert, S., de Senerpont Dormis, L.N., 2011. Chytrid infections of diatoms spring blooms: Paradoxial effects of climate warming on epidemics in lakes. Freshw. Biol. 56, 756-766.

IPCC, 2007. Climate Change 2007: Synthesis Report. Contribution of Working Groups I, II and III to the Fourth Assessment. Report of the Intergovernmental Panel on Climate Change [Core Writing Team, Pachauri, R.K and Reisinger, A.(Eds)]. IPCC, Geneva, Switzerland: 104 pp.

Jacquet S., Briand J.-F., Leboulanger C., Avois-Jacquet C., Paolini G., Oberhaus L., Tassin B., Vinçon-Leite B., Druart J.-C., Anneville O., Humbert J.-F., 2005. The proliferation of the toxic cyanobacterium Planktothrix rubescens following restoration of the largest natural French lake (Lac du Bourget). Harmful Algae 4:651-672.

Jeong K.-S., Kim H.-W., Whigham P.A. and Joo G.-J., 2003. Modelling Microcystis aeruginosa bloom dynamics in the Nakdong River by means of evolutionary computation and statistical approach. Ecological Modelling, 161(1-2): 67-78.

Jeppesen, E., Søndergaard M., Jensen J.P., Havens K., Anneville O., Carvalho L., Coveney M.F., Deneke R., Dokulil M., Foy B., Gerdeaux D., S.E. Hampton, K. Kangur, J. Köhler, S. Körner, E. Lammens, T.L. Lauridsen, M. Manca, R. Miracle, B. Moss, P. Nõges, G. Persson, G. Phillips, R. Portielje, S. Romo, C. L. Schelske, D. Straile, I.Tatrai, E. Willén & M. Winder, 2005: Lake responses to reduced nutrient loading – an analysis of contemporary long-term data from 35 case studies. - Freshwat. Biol. 50: 1747-1771.

John, D.M., Brooks A.J. & B.A. Whitton. 2002. The freshwater algal flora of the British Isles. An identification guide to freshwater and terrestrial algae. Cambridge University Press: 702 pp.

Jöhnk, K.D., Huisman J., Sharples J., Sommeijer B., Visser P.M., Stroom J.M.. 2008. Summer heatwave promote bloom of harmful Cyanobacteria. Global change biology, 14:495-512

Kaplan-Levy R.N., Hadas O., Summers M.L., Rucker J., Sukenik A., 2010. Akinetes: dormant cells of cyanobacteria. In: Lubzens, E., Cerda, J., Clark, M.S. (Eds.), Dormancy and Resistance in Harsh Environments. Springer, Heidelberg.

Kehoe D.M., and Gutu A., 2006. Responding to color: the regulation of complementary chromatic adaptation. Annual Review of Plant Physiology, 57: 127-150.

Komárek J., 2003. Problem of the taxonomic category “species” in cyanobacteria. Archiv für Hydrobiologie Supplementband - Algological Studies, 109: 281-297.

Knopf A.H., E. J. Javaux, D. Hewitt & P. Cohen. 2006. Eukaryotic organisms in Proterozoic oceans. Philosophical Transactions of the Royal Society B, 361: 1023-1038.

Kruk C, Peeters ETHM, Van Nes EH, Huszar VML, Costa LS, Scheffer M (2011) Phytoplankton community composition can be predicted best in terms of morphological groups. Limnology & Oceanography 56(3):110-118.

99

Krienitz L. (2009) Algae. In: Likens GE (ed), Encyclopedia of Inland Waters, vol. 1. Elsevier, Oxford, pp 103–113

Leathwick, D.R, Richardson J., Elith J., Hastie and T., 2005. Using Multivariate Adaptive Regression Splines to Predict the Distributions of New Zealand's Freshwater Diadromous Fish. Freshwater Biology, 50, 2034-2052.

Legendre P, Legendre L (1998) Numerical ecology. 2nd English edition, Elsevier Science, Amsterdam. 853 pp.

Lemmin U., 1995. Limnologie physique. In: Limnologie générale, R. Pourriot and M. Meybeck, collection d’écology 25. Masson. Paris 60-114.

Lewis W.M., 1983. "A revised classification of lakes based on mixing”. Canadian Journal of Fisheries and Aquatic Sciences 40 (10): 1779–1787

Litchman E., de Tezanos Pinto P., Klausmeier C.A., Thomas, M.K., Yoshiyama, K., 2010. Linking traits to species diversity and community structure in phytoplankton. Hydrobiologia 653, 15-28.

Livingstone, D. M. and J. Padisàk . 2007. Large-scale coherence in the response of lake surface-water temperatures to synoptic-scale climate forcing during summer. Limnology and Oceanography 50: 1313-1325.

Magnuson J.J., B.J. Benson, J.D. Lenters and Kratz T.K. 1990. Temporal coherence in the limnology of a suit of lakes in Wisconsin, USA. Freshwater Biology 23: 145-149.

Mur L.R., Skulberg O.M. and Utkilen H., 1999. Cyanobacteria in the environment. In:Chorus I. and Bartram J. (eds.), Toxic cyanobacteria in water: A guide to their public health consequences, monitoring and management. Spoon Press, on Behalf of WHO, London, pp 15-40.

Mooij W.M., Hülsmann S., De Senerpont Domis L.N., Nolet B.A., Bodelier P.L.E., Boers P.C.M., Dionisio Pires L.M., Gons H.J., Ibelings B.W., Noordhuis R., Portielje R., Wolfstein K. & Lammens E.H.R.R., 2005. The impact of climate change on lakes in the Netherlands: a review, Aquatic Ecology 39: 381-400.

Moss B., McKee D., Atkinson D., Collings S. E., Eaton J.W., Gill A.B., Harvey I., Hatton K., Heyes T. & Wilson D., 2003. How important is climate? Effects of warming, nutrient addition and fish on phytoplankton in shallow lake microcosms. Journal of applied Ecology. 40: 782-792.

Moss B.., 2011. Cogs in the endless machine: Lakes, climate change and nutrient cycles: A review. Science of total environment. Doc:10.1016/j.scitotenv.2011.07.069.

OECD, 1982. Eutrophicaton of waters. Monitoring, assessment and control. Organisation for Economic Cooperation and Development: 193 p.p.

Oksanen J., Blanchet F.G., Kindt R., Legendre P., O'Hara R.B., Simpson G.L., Solymos P., Stevens M.H.H., Wagner H.,2011. vegan: Community Ecology Package. R package version 1.17-10. http://CRAN.R-project.org/package=vegan.

Oliver R.L. and Ganf G.G., 2000. Freshwater blooms. In: The ecology of cyanobacteria. Springer Netherlands. 2002.

100

Padisák J., Hanjnal E., Naselli-Flores L., Dokulil M.T., Peeter Nõges, Zohary T.. 2010. Convergences and divergence in organization of phytoplankton communities under various regimes of physical and biological control. Hydrobiologia. 639:205-220.

Paerl H.W. and Millie D.F., 1996. Physiological ecology of toxic cyanobacteria. Phycologia, 35(6): 160-167.

Paerl H.W., 1998. Nuisance phytoplankton blooms in coastal, estuarine and inland waters. Limnology and Oceanography 33: 823-847.

Paerl H.W., Fulton III. R.S., 2006. Ecology of harmful cyanobacteria. In: Granelli E. and Turner J. Ecology of harmful marine Algae. Springeer-Verlag, Berlin.

Paerl H.W, Huisman J., 2008. Blooms like it hot. Science 320, 57-58.

Paerl H.W. and Paul V.J., 2011. Climate change: Links to Global Expansion of Harmful cyanobacteria. Water research. doi: 10.1016/j.watres.2011.08.002.

Paerl H.W, Fulton III. R.S., Moisander P. and Dyble J. 2001. Harmful freshwater algae blooms with an emphasis to cyanobacteria. The scientific World 1:76-113.

Paul V.J. 2008. Global warming and cyanobacterial harmful algal blooms. In: K.H. Hudnell (Ed.), Cyanobacterial harmful algal blooms: state of the science research needs series. Springer Adv. Exp. Med. Biol., 619: 239-257.

Peters R.H., 1991. A critique for ecology. Cambridge University Press, Cambridge, USA, 366 pp.

Psenner R. 2003. Alpine lakes: extreme ecosystem under the pressure of global change, EAWAG news 55: 12-14.

Quétin P., Météorology, 2004. Papp. Comm. Int. prot.eaux. Léman contre pollut. Campagne 2003, 2004. pp. 19-29.

Rasmussen B., I.R. Fletcher, J.J. Brocks & M.R. Kilburn.2008. Reassessing the first appearance of eukaryotes and cyanobacteria. Nature, 455: 1101-1104.

Reynolds C.S., 1989. Physical determinants of phytoplankton succession. In: Sommer, U. (Ed.), Plankton Ecology: Succession in Plankton Communities. Springer Verlag, Berlin.

Reynolds C.S. and Peterson A.C., 2000. The distribution of planktonic cyanobacteria in Irish lakes in relation to their trophic states. Hydrobiologia, 424: 91-99.

R Development Core Team, 2012. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, ISBN 3-900051-07-0. http://www.R-project.org

ReynoldsC.S., Irish A.E. and Elliott J.A., 2001. The ecological basis for simulating phytoplankton responses to environmental change (PROTECH). Ecological Modelling, 140, 271-291.

Reynolds CS, Huzar V, Kruk C, Naselli-Flores L, Melo S., 2002. Towards a functional classification of the freshwater Phytoplankton. Review. Journal of Plankton Research, 24(5):417-428.

Reynolds CS., 2006. Ecology of Phytoplankton. Ecology, Biodiversity and Conservation. Cambridge 717 University Press, Cambridge, UK.

101

Robarts R.D. & T. Zohary. 1987. Temperature effects on photosynthesis capacity, respiration, and growth rates of bloom-forming cyanobacteria. N.Z.J. Mar. Freshwat. Res.,21: 391-399.

Roy S., 2009. The coevolution of two phytoplankton species on a single resource: allelopathy as a pseudo-mixotrophy. Theoretical Population Biology, 75(1): 68-75.

Schatz D., Keren Y., Vardi A., Sukenik A., Carmeli S., Börner T., Dittman E. and Kaplan A., 2007. Towards a clarification of the biological role of microcystins, a family of cyanobacterial toxins. Environmental Microbiology, 9(4): 965-970.

Scheffer M., Rinaldi S., Gragnani A., Mur R.L., van Nes E.H., 1997. On the dominance of filamentus cyanobacteria in shallow, turbid lakes. Ecology 78, 272-282.

Søndergaard M., Jensen J.P., Jeppesen E., 2003. Role of sediment and internal loading of phosphorus in shallow lakes. Hydrobiologia 506-509.

Schopf J.W., 2000. The fossil record: Tracing the roots of the cyanobacterial lineage. In: B.A. Whitton &728 M. Potts (Eds), The ecology of cyanobacteria. Dordecht, the Netherlands: KluwerAcademic Publishers: 13-15.

Sellner K.G., Doucette, G.J., Kirkpatrick, G.J., 2003. Harmful algal blooms: causes, impacts and detection. J. Ind. Microbiol. Biotechnol. 30: 383-406.

Smol J. P. & Cumming B.F. (2000). Tracking long-term changes in climate using algal indicators in lake sediments. J. Phycol. 36: 986-1011.

Sommer, U., ZM. Gliwicz, W. Lampert, A. Duncan. 1986. PEG-model of Seasonal Succession of Planktonic Events in Fresh Waters. Archives of Hydrobiology. 106(4): 433-471.

Stoermer E.F., 1978. Transactions of the American Microscopical Society. 97.1: 2-16.

Salmaso N., J. Padisàk. 2007. Morpho- Funcional Groups and phytoplankton development in two deep lakes (Lake Garda, Italy and Lake Stechlin, Germany). Hydrobiologia 578: 97-112.

Salmaso N., 2010. Long-term phytoplankton community changes in a deep subalpine lake: responses to nutrient availability and climatic fluctuations. Freshwater Biology 55: 825–846.

Salmaso N., 2011. Interactions between nutrient availability and climatic fluctuations as determinants of the long term phytoplankton community changes in Lake Garda, Northern Italy. Hydrobiologia 660: 59–68.

Smith V.H. and Bennet S.J., 1999. Nitrogen:phosphorus supply ratios and phytoplankton community structure in lakes: nutrient ratios. Archiv für Hydrobiologie, 146(1): 37-53

Smith V.H. 2005. Low nitrogen to phosphorus ratios favor dominance by blue-green algae in lake phytoplankton. Science 983; 221: 669–71.

Scheffer M., Rinaldi S., Huisman J. and Weissing F.J., 2003. Why plankton communities have no equilibrium: solution to the paradox. Hydrobiologia, 491: 9-18.

Schatz D., Keren Y., Vardi A., Sukenik A., Carmeli S., Börner T., Dittman E. and Kaplan A., 2007. Towards a clarification of the biological role of microcystins, a family of cyanobacterial toxins. Environmental Microbiology, 9(4): 965-970.

102

Stich HB, Brinker A (2011) Oligotrophication outweighs effect of global warming in a large, deep, stratified lake ecosystem. Global Change Biology, 16:877-888.

Tilzer M.M., Strambler N., Lovengreen C.. 1995. The role of phytoplankton in determining the underwater light climate in Lake Constance. Hydrobiologia, 316:161-172.

Tilman D. and Kilham S.S., 1976. Phytoplankton community ecology: the role of limiting nutrients. Rev. Ecol. System.

Tilman D., Kilham S.S., Kilham P., 1982. Phytoplankton Community Ecology: The role of Limiting Nutrients. Annual Review of Ecology and Systematics, 13: 349-372.

Tillmans A.R., Wilson A.E., Pick F.R. and Sarnelle O., 2008. Meta-analysis of cyanobacterial effects on zooplankton population growth rate: species-specific responses. Fundamental and Applied Limnology, 171(4): 285-295.

Thompson R., Kamenik C., Schmidt R., 2005. Ultra-sensitive Alpine lakes and climate change. Journal Limnology 64: 139-152.

Tsugeki N.K., Urabe J., Hayami Y, Kuwae M., Nakanishi M., 2010. Phytoplankton dynamics in Lake Biwa during the 20th century: complex responses to climate variation and changes in nutrient status. Journal of Paleolimnology, 44:69-83 (15).

Vanni M.J., Layne C.D., Arnott, S.E, 1997. “Top-down” trophic interactions in lakes: effects of fish on nutrient dynamics. Ecology 78: 1-20.

Vasconcelos V., 2001. Cyanobacterial toxins: diversity and ecological effects. Limnetica, 20(1): 45- 58.

Vincent W.F., 2009. Effects of Climate Change on Lakes. W. F. Vincent, Laval University, Quebec City, QC, Canada 2009 Elsevier Inc

Wagner C., Adrian R., 2009. Cyanobacteria dominance: quantifying the effects of climate change. Limnology. Oceanogr. 54:2460-2468.

Walsby A.E. and Hayes P.K., 1988. The minor cyanobacterial gas vesicle protein, GVPc, is attached to the outer surface of the gas vesicle. Journal of General Microbiology, 134(10): 2647-2657.

Walsby A.E., 1994. Gas vesicles. Microbiological Reviews, 58(1): 94-144.

Walsby A.E., Ng G., Dunn C. and Davis P.A., 2004. Comparison of the depth where Planktothrix rubescens stratifies and the depth where the daily insulation supports its neutral buoyancy. New Phytologist, 162(1): 133-145.

Walsby A.E., 2005. Stratification by cyanobacteria in lakes: a dynamic buoyancy model indicates size limitations met by Planktothrix rubescens filaments. New Phytologist, 168(2): 365-376.

Wehenmeyer G.A., Bleckner T. and Petterson, K (1999). Changes in the plankton spring outburst related to the North Atlantic Oscillation. Limnology and Oceanography 44, 1788-1792.

Weithoff G. 2003. The concepts of “plant functional types” and “functional diversity” in lake phytoplankton- a new understanding of phytoplankton ecology? Freshwater Biology 48:1669-1675.

Wetzel R.G., 2001. Limnology: lake and river ecosystems. 3rd edition, Academic Press.

103

Winder M., Schindler D.E., 2004. Climatic effects on the phenology of lake processes. Global Change Biology, 10: 1844–1856.

Wiedener C., Rücker J. Brüggemann R., Nixdorf B. 2007. Climate change affects timing and size of population of an invasive cyanobacterium in temperate regions. Oecologia 152:473-484.

Wiegand C. and Pflugmacher S., 2005. Ecotoxicological effects of selected cyanobacterial secondary metabolites a short review. Toxicology and Applied Pharmacology, 203(3): 201-218.

Williamson C.E., Saros J.E., Vincent W.F., J.P. Smol. 2009a. Lakes and reservoirs as sentinels, integrators, and regulators of climate change. Limnology & Oceanography. 54 (2) 2273-2282.

Williamson C.E., Saros J.E., Schindler D.W., 2009b, Sentinels of Change. Sciences 323:887-888.

Zhang M., Duand H., Shi Yu., Kong F., 2001. Contributions of metrology to the phenology of cyanobacteria bloom: implication for future climate change. Water Research. Doi: 10.1016/j.Watres.2011.11.013.

104