Impacts of dissolved organic carbon on productivity of fish and benthic macroinvertebrates in north temperate lakes

Nicola Craig

Department of Natural Resource Sciences, McGill University, Montreal

April 2016

A thesis submitted to McGill University in partial fulfillment of the requirements of the degree of PhD Renewable Resources

© Nicola Craig 2016

1

TABLE OF CONTENTS

Page

ABSTRACT 5

RÉSUMÉ 7

LIST OF TABLES 10

LIST OF FIGURES 11

PREFACE 14

Acknowledgements 14 Contributions of Authors 16 Novelty and Impact of Thesis Research 17

INTRODUCTION AND LITERATURE REVIEW 19

References 28 Figures 38

CHAPTER 1: HABITAT, NOT RESOURCE AVAILABILITY, LIMITS CONSUMER PRODUCTIVITY IN LAKE ECOSYSTEMS 39

Abstract 39 Introduction 40 Methods 42 Results 47 Discussion 51 Acknowledgements 56 References 56 Figure Captions 62 Tables and Figures 63 Appendix 1A 70

2

Appendix 1B 72

Connecting Statement 74

CHAPTER 2: DISSOLVED ORGANIC CARBON EFFECTS ON FISH FEEDING EFFICIENCY 75

Abstract 75 Introduction 76 Methods 78 Results 81 Discussion 82 Acknowledgements 84 References 85 Figure Captions 90 Tables and Figures 91 Appendix 2A 94 Appendix 2B 95 Appendix 2C 98

Connecting Statement 99

CHAPTER 3: LIFE HISTORY CONSTRAINTS EXPLAIN NEGATIVE RELATIONSHIP BETWEEN FISH PRODUCTIVITY AND DISSOLVED ORGANIC CARBON IN LAKES 100

Abstract 100 Introduction 101 Methods 103 Results 106 Discussion 107 Acknowledgements 111 References 112 Table and Figure Captions 118 Tables and Figures 119

3

Appendix 3A 123 Appendix 3B 124 Connecting Statement 131

CHAPTER 4: A TEMPORAL INCREASE IN DISSOLVED ORGANIC CARBON CONCENTRATIONS RESULTS IN AN UNEXPECTED INCREASE IN ZOOBENTHOS BIOMASS AND PRODUCTIVITY IN A WHOLE-LAKE EXPERIMENT 132

Abstract 132 Introduction 133 Methods 135 Results 140 Discussion 142 Acknowledgements 149 References 149 Table and Figure Captions 157 Tables and Figures 158 Appendix 4A 165 Appendix 4B 166 Appendix 4C 167 Appendix 4D 170

GENERAL CONCLUSIONS 171

Future directions 172 References 174

4

ABSTRACT

Over the past several decades, increases in concentrations of dissolved organic carbon (DOC) have been observed in many north temperate aquatic ecosystems, a phenomenon known as aquatic 'browning', however the ecological consequences of this increase are not well understood. DOC from terrestrial sources stains lake water a dark brown color, and can have strong effects on the physical and biological structure of lake ecosystems. This occurs through its associated light and heat attenuating properties, which reduce thermocline depths, and thus the area of light, warm, and well-oxygenated habitat in DOC-rich lakes. Recent spatial surveys, where consumers were sampled from various lakes spanning a DOC gradient, have shown that fish productivity decreases along a gradient of increasing DOC, however the mechanisms behind this have not been fully explored. This thesis demonstrates potential mechanisms for this loss in productivity by determining how DOC affects zoobenthos, a primary prey item for many fish, as well as how DOC affects fish feeding efficiency and life history strategies.

I begin by demonstrating, through the use of a spatial lake survey, that zoobenthos production declines over an increasing DOC gradient, and that this decline is due to limitations in oxygen-rich habitat availability, rather than the previously assumed mechanism of primary resource limitation. As many fish are visual predators, and high levels of DOC may reduce visibility, I then examined how DOC may affect fish feeding efficiency using mesocosm experiments and another gradient-based diet survey. I showed that bluegill (Lepomis macrochirus) feeding efficiency is not affected by DOC concentration, suggesting that these benthivores use cues other than vision to detect prey in darker, DOC-rich lakes. The reduction of fish productivity with increasing DOC is likely manifested through shifts in life history characteristics that are important to understand if we are to better manage fisheries with increasing browning. Again using bluegill as a model organism, I show that in low-DOC lakes, fish are able to attain enough energetic resources to reproduce as well as continue to grow after maturity. However, in high-DOC lakes, growth slows after maturity is reached, and so overall reproductive output and maximum size is reduced in these populations.

5

Currently, the majority of studies focusing on the effects of DOC on consumer productivity are based on spatial gradient surveys, models, and mesocosm experiments. However, browning is a temporal process, and these studies may not accurately reflect how consumers may react to increases of DOC over time. In my final chapter, I describe a whole- ecosystem experiment were we divided a lake in two and increased DOC in one basin, comparing the effects on zoobenthos biomass and productivity to a reference basin, as well as the spatial survey from the first chapter. I show that, contrary to the results of the spatial survey, zoobenthos productivity actually increased with increasing DOC concentrations in this temporal experiment. This result suggests that there may be transitional effects of DOC increases on zoobenthos communities, and that the relationship between DOC and ecosystem productivity may be non-linear. This thesis highlights the need for multiple approaches in order to untangle the complex effects of DOC in lake ecosystems. The observations within will help us better predict how consumers in lake ecosystems may react in the face of future increases in DOC, and how to manage them accordingly.

6

RÉSUMÉ

Depuis plusieurs décennies, une augmentation de la concentration en carbone organique dissous (COD) a été observée dans plusieurs écosystèmes aquatiques tempérés du Nord, un phénomène aussi appelé « brunissement », et dont les conséquences écologiques sont encore peu connues. Le COD d’origine terrestre donne une teinte brune foncée à l’eau des lacs et peut causer d’importantes répercussions sur la structure physique et biologique des écosystèmes lenthiques. Cet effet provient de la capacité du COD à réduire la pénétration de la lumière et de la chaleur à la surface des lacs, réduisant la profondeur de la thermocline dans les lacs riches en COD et diminuant par le fait même le volume d’habitat disponible où la lumière, la température et l’oxygène dissous sont plus élevés. Des relevés récents où les consommateurs ont été échantillonnés dans plusieurs lacs présentant une large répartition spatiale et couvrant une forte variation en COD ont montrés une diminution de la productivité des populations de poissons associée à une augmentation de la concentration en COD. Par contre, les mécanismes derrière cette relation n’ont pas été complètement explorés. Cette thèse représente une démonstration des mécanismes pouvant potentiellement expliquer cette réduction de la productivité, en montrant comment le COD influence le zoobenthos, une proie préférentielle pour plusieurs espèces de poissons, et en expliquant comment le COD affecte l’efficacité des poissons à exploiter les ressources ainsi que leur histoire de vie.

D’abord, à partir de données provenant de l’échantillonnage de plusieurs lacs présentant un large gradient spatial, je démontre que la productivité du zoobenthos diminue à travers l’augmentation d’un gradient de concentration en COD et que ce déclin est causé par une réduction du volume d’habitat riche en oxygène, plutôt que par le mécanisme de limitation des ressources primaires disponibles proposé jusqu’à maintenant. Comme plusieurs espèces de poissons sont des prédateurs visuels et qu’une concentration élevée en COD peut réduire la visibilité dans la colonne d’eau, j’ai ensuite examiné comment le COD est en mesure d’influencer l’efficacité des poissons à exploiter les ressources disponibles par la mise en place d’une expérimentation en mésocosmes incluant l’analyse des contenus stomacaux de poissons évoluant dans un gradient de COD. J’ai montré que l’efficacité d’exploitation des ressources du

7 crapet arlequin (Lepomis macrochirus) ne se trouve pas affectée par la concentration en COD, suggérant que cette espèce benthivore utilise d’autres sens que la vision pour détecter les proies dans les lacs foncés à haute concentration en COD. La réduction de la productivité des poissons associée à l’augmentation de la concentration en COD se manifeste probablement à travers de changements au niveau de certains traits d’histoire de vie, et une meilleure compréhension de ces changements est essentielle à une gestion efficace des pêches en vue de l’augmentation du brunissement de l’eau de certains lacs. En utilisant le crapet arlequin comme organisme modèle, je montre que dans les lacs avec une faible concentration en COD, les poissons sont en mesure d’obtenir suffisamment de ressources énergétiques pour se reproduire et continuer de croître une fois mature. Par contre, dans les lacs où la concentration en COD est élevée, la croissance ralentit après l’atteinte de la maturité sexuelle, résultant en une plus faible capacité reproductive et une taille maximale réduite pour ces populations.

Actuellement, la majorité des études qui se concentrent sur la relation entre la concentration en COD et la productivité des organismes aquatiques sont basées sur l’échantillonnage de gradients spatiaux en COD, la modélisation et des expérimentations en mésocosmes. Par contre, la brunissement est un processus temporel, et ces études pourraient ne pas refléter adéquatement la réaction des organismes face à une augmentation du COD dans le temps. Dans mon dernier chapitre, je présente une expérimentation à l’échelle écosystémique où nous avons divisé un lac en deux et augmenté la concentration en COD dans un des deux bassins, afin de mesurer les effets sur la biomasse du zoobenthos et de pouvoir effectuer la comparaison avec un bassin de référence et avec l’échantillonnage de gradients spatiaux présenté dans le premier chapitre. Je montre que, contrairement aux résultats obtenus à partir des gradients spatiaux, la productivité du zoobenthos a augmentée avec une augmentation de la concentration en COD dans cette expérimentation temporelle. Ce résultat suggère des effets transitoires potentiels de l’augmentation en COD sur les communautés de zoobenthos, et que la relation entre le COD et la productivité dans les écosystèmes aquatiques pourrait être non-linéaire. Cette thèse met en évidence le besoin d’utiliser des approches différentes afin de démêler les effets complexes du COD dans les écosystèmes lenthiques. Les observations mentionnées ici permettront de mieux prédire comment les organismes évoluant

8 dans ces écosystèmes pourraient réagir face aux futures augmentations en COD, et d’ajuster les méthodes de gestion conséquemment.

9

LIST OF TABLES

1.1 Summary of lake bathymetry and water chemistry for the ten survey lakes from chapter one. pg 63

1.2 Summary of bias-corrected linear measurement error models relating depth-specific zoobenthos production in ten lakes to mean depth-specific dissolved oxygen concentration, water temperature, and benthic primary production. pg 64

1A Length-mass equations used to calculate dry mass from head/shell width or body length of zoobenthos. pg 70

1B Benthic, pelagic, and terrestrial resource reliance by zoobenthos in ten lakes of varying dissolved organic carbon (DOC) concentration, estimated using a Bayesian stable isotope mixing model. pg 72

2.1 Summary of samples collected from each lake for chapter two. pg 91

2.2 Linear mixed effects model results for fish diet responses as a function of fish length and DOC. pg 91

2B Regression equations for estimating prey dry mass from body size measurements. pg 95

3.1 Summary of lake parameters for the eleven lakes surveyed for chapter three. pg 119

3A Information regarding fish available for various analyses in chapter three. pg 123

4.1 Average lake level environmental variables (DOC, thermocline depth, pelagic and benthic primary production), for a whole-lake DOC manipulation experiment. pg 158

4.2 Average depth-specific environmental variables (temperature, dissolved oxygen, and benthic primary production), for a whole-lake DOC manipulation experiment. pg 159

4D Taxa and abbreviations for the nMDS ordination in figure 4.5. pg 170

10

LIST OF FIGURES

1 Conceptual diagram of the physical effects of DOC on lake ecosystems. pg 38

1.1 Depth profiles of habitat and resource characteristics for the clearest and darkest lake sampled in chapter one. pg 65

1.2 Depth-specific zoobenthos production plotted against mean depth-specific dissolved oxygen concentration, water temperature, and benthic primary production. pg 66

1.3 Zoobenthos production across depths in lakes where DOC concentrations are low, intermediate, and high. pg 67

1.4 Relationships between zoobenthos production and DOC concentration, as well as catch-per- unit effort of zoobenthivorous fish. pg 68

1.5 Benthic reliance (proportion of biomass derived from benthic primary production) of zoobenthos in ten lakes across a gradient of DOC concentrations. pg 69

2.1 Number of chironomid larvae consumed by bluegill over a water color gradient in mesocosm feeding experiments. pg 92

2.2 Diet characteristics (stomach fullness, prey length, benthic proportion, and benthic selectivity) across a DOC gradient for bluegill and yellow perch. pg 93

2A Temperature, dissolved oxygen, and surface and bottom light levels for each trial in the benthic feeding efficiency experiments. pg 94

2C. Number of bluegill and yellow perch diets sampled for chapter two, broken down by fish length. pg 98

3.1 a) Bluegill maximum length and mass regressed with DOC concentration, b) Bluegill initial growth rates regressed with DOC concentration, c) Bluegill length-age growth curves for each lake sampled in chapter three. pg 120

11

3.2 a) Mean age-specific egg output of female bluegill, b) Female bluegill age at maturity as a function of DOC, c) Lifetime potential fecundity for female bluegills as a function of DOC, d) Lifetime realized fecundity of female bluegills as a function of DOC. pg 121

3.3 The intrinsic rate of natural population increase (r) for bluegill as a function of DOC concentration. pg 122

3B.1 Von-Bertalanffy growth curves for bluegill sampled for chapter three. pg 124

3B.2 Female bluegill age as a function of fish weight. pg 125

3B.3 Gonad mass of pre-spawn female bluegill plotted against body length. pg 126

3B.4 Female bluegill gonad mass as a function of body mass. pg 127

3B.5 Number of eggs per gonad as a function of gonad mass for female bluegill just before spawning. pg 128

3B.6 Egg number per gonad as a function of bluegill length. pg 129

3B.7 Egg width as a function of bluegill length. pg 130

4.1 Time series, and total differences of dissolved oxygen and temperature at 3m depth between basins over the course of a whole-lake DOC manipulation experiment. pg 160

4.2 Time series, and total differences of zoobenthos biomass at the whole lake level and at specific depths over the course of a whole-lake DOC manipulation experiment. pg 161

4.3 Average zoobenthos body size for the whole lake and for each individual sampling depth over the course of a whole-lake DOC manipulation experiment. pg 162

4.4 Zoobenthos production at the whole lake level and at each individual sampling depth over the course of a whole-lake DOC manipulation experiment. pg 163

4.5 nMDS ordination of zoobenthos assemblage structure at 3 m depth for one pre- and two post-manipulation years of a whole-lake DOC manipulation experiment. pg 164

12

4A Table showing the number of days that the water at 3 m depth was hypoxic over the course of a whole-lake DOC manipulation experiment. pg 165

4B Average zoobenthos abundance for each individual sampling depth over the course of a whole-lake DOC manipulation experiment. pg 166

4C Size-frequency plots of zoobenthos dry masses for each depth-year combination over the course of a whole-lake DOC manipulation experiment. pg 167

13

PREFACE

Acknowledgements

First and foremost, I have to extend a massive thank you to my advisor Chris for his unending positivity and encouragement. He took a chance by hiring an 'expensive' international student (as he so often reminded me), and I hope this thesis has done justice to that decision. Chris also helped a lot with some of the statistical processes and provided many useful comments on the manuscripts presented in this thesis. Thanks Chris, I couldn't have done this without you!

I was fortunate to be part of an amazing collaborative team during my PhD and had many fun, and informative discussions and meetings with Stuart Jones and Brian Weidel, both of whom have been an enormous help to me. Patrick Kelly and Jake Zwart also deserve special mention for great company, help across the board (data, R, field!), and generally making my time at UNDERC a blast.

My fellow lab members have been a great source of support, friendship, and advice. I especially thank P-O Benoit and Jake Ziegler who were there from the start and have helped me out in countless ways, as well as Shun Koizumi, Melissa Lenker, Raphaelle Thomas and Katrine Turgeon.

This thesis wouldn't have been possible without a legion of awesome undergraduates both from McGill, and the University of Notre Dame who helped out in the lab and field. Thanks to Katie Baglini, Caitlin Broderick, Ludovick Brown, Henri Chung, Brian Connor, Stephen Elser, Sean Godwin, Elizabeth Golebie, Erin Hanratty, CJ Humes, Ellen Mather, Shayna McCarthy, Mike O’Connor, Rachel Pilla, Lisa Raff, Alexandra Salcedo, Jessica Schaefer, Joey Vanderwall, and a special thanks to my bug and/or gonad nerds Alexandra Sumner, Karling Roberts, and Jacob Lerner.

All lot of the aquatic chemical and isotope sampling, as well as data entry and management occurred in Stuart Jones's lab and the Center for Environmental Science and Technology at the University of Notre Dame. With that in mind, I would like to thank Jim Coloso

14 for organising much of this, as well as all the graduate and undergraduate students there that worked on processing those samples.

I was fortunate to be able to conduct my field work at the University of Notre Dame Environmental Research Center which was a great place to be if you wanted to mess around with whole lake experiments... Thanks to all the staff who facilitated my research there, including Michael Cramer and Gary Belovsky. Also thanks to all the other graduate students, undergraduates and technicians that I met during my summers there. We had some awesome times!

The department of Natural Resource Sciences at McGill has been very accommodating to me during my time here. I would like to thank the administrative staff and technicians who helped me out at various stages along the way, in particular, Ian Ritchie, Susan Gregus, Peter Kirby and Ann Gossage.

My advisory, comprehensive, and defence committee members, as well as my thesis examiners provided many useful comments along the way and so I would like to thank Elena Bennett, Chris Buddle, Lauren Chapman, Andrew Hendry, Murray Humphries, and Mike Rennie for their time and insights.

Much of my funding came from Chris's NSERC Discovery grant. I also received various amounts of funding from NRS, QCBS, the CREATE-EI program, UNDERC, and a Schulich graduate fellowship.

Finally, I would like to thank my family back home, as well as my adopted family in Montreal who have supported me through this rollercoaster. I love you guys!

15

Contribution of Authors

This thesis is presented in a manuscript-based format, as each chapter consists of a data-based research project of which I am the primary author. One chapter is currently published, and one is currently under the review process. While each chapter provides a partial literature review in the introduction, I have also provided a more extensive literature review in the thesis introduction to give the reader a broader context of the effects of dissolved organic carbon on lake ecosystems, and familiarize them with concepts that I will return to often in the following chapters.

I have been fortunate enough to be a part of a collaborative project for the duration of my PhD, consisting of my supervisor, Chris Solomon, as well as Stuart Jones of the University of Notre Dame, and Brian Weidel of the US Geological Survey, all of whom are co-authors on my thesis chapters. In addition, Patrick Kelly and Jacob Zwart, graduate students at the University of Notre Dame are co-authors on Chapter 4. Their contributions are as follows:

Chapter 1 (published in Limnology & Oceanography). All authors (NC, SJ, BW, CS) conceived of and designed the study. NC collected the data. NC and CS analysed the data. NC wrote the manuscript and all authors contributed to revisions.

Chapter 2 (in preparation). NC conceived of and designed the study. NC collected and analysed the data. NC wrote the manuscript and all authors contributed to revisions.

Chapter 3 (undergoing revisions for submittal to the Journal of Animal Ecology. NC, CS, and BW conceived of and designed the study. NC collected the data. NC and CS analysed the data. NC wrote the manuscript and all authors contributed to revisions.

Chapter 4 (in preparation for submittal to Ecosystems). SJ, BW, and CS conceived of and designed the study. NC, PK and JZ collected the data. NC analysed the data. NC wrote the manuscript and all authors contributed to revisions.

16

Novelty and impact of thesis research

DOC concentration has been increasingly recognised as a driver of productivity in low nutrient freshwaters over the past few decades (Prairie 2008). However, the mechanisms behind this effect on productivity are still not well understood, and more attention has been paid to basal productivity, rather than higher level consumers such as zoobenthos or fish (Wetzel 2001). The chapters of this thesis demonstrate both the effects of DOC on the productivity of fish and zoobenthos, as well as providing some explanatory mechanisms for the observed patterns; substantially increasing our knowledge on how DOC affects aquatic ecosystems.

Chapter 1: This chapter describes one of the largest zoobenthos surveys conducted so far, detailing zoobenthos biomass and productivity with depth in ten lakes simultaneously over a growing season. The results refuted the widely held assumption that secondary production in DOC-rich lakes was limited by light-mediated bottom-up processes. Instead, we found that while benthic primary production had little effect on zoobenthos production, dissolved oxygen was a major predictor, and dissolved oxygen levels were influenced by thermocline depth, which was negatively influenced by DOC concentration. This chapter provides a new mechanism for understanding declines in secondary productivity across a DOC gradient.

Chapter 2: Experiments detailing how DOC affects fish feeding efficiency are rare, and exclusively focus on pelagic interactions. This chapter detailed the first mesocosm experiments conducted which focused on how DOC concentrations could affect benthic feeding behaviour in fish. Previous work showed that pelagic feeding efficiency could be negatively affected in higher DOC waters. Our study, however, showed that benthic feeding was not affected by DOC, which suggests that food webs in DOC-rich lakes may be more reliant on benthic resources.

Chapter 3: Life history theory provides one potential framework for understanding the mechanisms that link DOC to fish productivity. In this chapter, I applied this framework using bluegill (Lepomis macrochirus) as a model organism in order to determine how growth and reproduction vary across a DOC gradient. This study was the first to show how DOC in lakes can affect fish life histories through reducing adult growth, maximum size, and ultimately

17 reproductive output. This chapter contains valuable insights that will help in the management of fish populations in the face of increasing DOC levels in north-temperate lakes.

Chapter 4: The majority of research on DOC and productivity in lakes has been based on DOC-gradient surveys, modelling, and mesocosm experiments, which may not accurately reflect how a temporal increase in DOC could affect whole ecosystems. This chapter details the first whole-lake experiment focusing exclusively on how rising DOC levels affect zoobenthos biomass and productivity. We found that, in contrast to the results of chapter 1, zoobenthos biomass and productivity actually increased as DOC concentrations rose. This apparent disparity lends itself to the relatively novel suggestion that DOC-productivity relationships in lakes may be non- linear and complicated by lake nutrient status.

18

INTRODUCTION AND LITERATURE REVIEW

Dissolved organic carbon (DOC) can have profound effects on aquatic habitats and has been shown to be an important driver of productivity in lake ecosystems (Hessen & Tranvik 1998, Prairie 2008). Terrestrial DOC (hereafter referred to as DOC) is formed from the breakdown of plant material, and is particularly important as it constitutes the main DOC pool in the majority of lakes (Karlsson et al. 2003, Wilkinson et al. 2013). Its associated humic and fulvic acids have a characteristic yellow colour which stains the water of recipient aquatic ecosystems; 'browning' the waters (Jones 1992, Roulet & Moore 2006, Solomon et al. 2015). This means that waters with a high DOC concentration tend to be darker in color with reduced light penetration, which can have strong physical and biological effects on lake ecosystems (Jones 1992, Solomon et al. 2015, Fig. 1).

Terrestrial DOC enters lakes from the watershed mainly through surface and subsurface runoff which is intensified during precipitation events or with melt water, but can also enter through groundwater and atmospheric deposition (Hinton et al. 1997, Schiff et al. 1997). As leeching from the watershed is the main mode of entry, watershed characteristics such as vegetation type, and hydrological conditions are important factors contributing to the DOC concentration of a lake, as well as general climatic conditions (Hinton et al. 1997, Kortelainen 1999). For example, lakes surrounded by wetlands tend to have higher DOC inputs than lakes surrounded by well drained soils, or clear cut forests (Mulholland & Kuenzler 1979, Moore 1989). In addition, lakes with a high catchment area relative to lake size may be more likely to have higher DOC concentrations, and this suggests that larger lakes may not have as much capacity to reach high DOC concentrations compared to smaller lakes (Sobek et al. 2007).

Due to variability in watershed and climatic characteristics, DOC concentrations in freshwaters can vary substantially both spatially and temporally. Spatially, at a global/regional scale, DOC is controlled by climate and topography, where variation in vegetation, soil type, and hydrology can set the range of DOC available within a region (Sobek et al. 2007). Then, within each 'region', the remaining variation in DOC concentrations within lakes are generally set by catchment and lake size. Temporally, a recent increase in DOC levels has observed over

19 the past few decades in many freshwater systems in the Northern Hemisphere (Evans et al. 2005, Monteith et al. 2007, Weyhenmeyer et al. 2015). Many mechanisms for this shift in DOC concentration have been proposed and it is likely affected by several factors including climate change, land use change and recovery from acidification (Garnett et al. 2000, Freeman et al. 2001, Monteith et al. 2007, Oni et al. 2015). The fact that DOC concentrations are dynamic and are likely to continue to rise in certain regions means that it is important to understand how DOC affects aquatic food webs in order to make predictions for the future and manage these ecosystems appropriately.

DOC and ecosystem productivity

Over the past few decades, several papers based on stable isotope methods have shown that sometimes substantial proportions of aquatic consumer biomass are supported by terrestrial resources (Jones et al. 1998, Cole et al. 2006, Weidel et al. 2008, Cole et al. 2011, Solomon et al. 2011). This led some researchers to suggest that DOC may actually be a subsidy for aquatic consumers, as the proportion of terrestrial resource use increased with increasing DOC concentrations in lakes (Polis et al. 1997, Cole et al. 2006, Solomon et al. 2011). This may certainly be true for bacterial consumers, which are able to use the sometimes relatively recalcitrant terrestrial DOC (Tranvik 1988, Hessen et al. 1992). This often leads to net heterotrophy in DOC-rich lakes, and as such, increased inputs of terrestrial DOC can often lead to increased export of carbon dioxide in lakes (Tranvik 1992, Cole et al. 1994). However, the evidence for DOC as a subsidy to higher level consumers is mixed, with some research suggesting that zooplankton growth is actually reduced when they have to rely on DOC as a resource due to its relative lower quality as a food source (Brett et al. 2009, Kelly et al. 2014). It is becoming more apparent that inputs of terrestrial DOC can have a negative effect on ecosystem productivity, and while lake consumers may use terrestrial resources more in DOC- rich lakes, this actually has a detrimental effect on biomass production (Karlsson et al. 2015, Solomon et al. 2015).

While phosphorous levels have been traditionally viewed as the main driver of productivity in aquatic ecosystems, Prairie (2008) pointed out that in low nutrient systems, DOC

20 is the main variable affecting productivity. However, this relationship is only just beginning to be understood; for example, in Wetzel's 2001 Limnology textbook, there is much discussion around the role of organic carbon and allochthonous carbon in lakes, but very little focus on how increasing DOC may affect ecosystem productivity, except in the case of basal metabolism. The primary literature has only begun to pick apart the mechanisms by which DOC controls productivity in the past decade or so. Recent research has indicated that fish and zoobenthos biomass and production generally decreases along a gradient of increasing DOC concentration in lakes (Karlsson et al. 2009, Finstad et al. 2014, Karlsson et al. 2015), however, little is known about the mechanisms involved. The main objective of this thesis therefore, is to explore and determine potential mechanisms for the loss in productivity in zoobenthos and fish in high DOC systems. Outcomes from this work will advance fundamental knowledge of how aquatic ecosystems work, as well as provide information for how to manage these systems under conditions of varying DOC concentrations.

Potential effects of DOC on benthic productivity

The current assumption is that a reduction in basal and secondary production is the major contributor to the loss of fish productivity in DOC-rich lakes (Karlsson et al. 2009). The reduction in light penetration associated with high DOC levels restricts algal production to shallower depths as light is limited for photosynthesis, with a more dramatic effect on benthic algae which is restricted to lake substrates (Hanson et al. 2003, Vadeboncoeur et al. 2008, Godwin et al. 2014). This may affect the whole food web through a reduction in primary productivity. Indeed, primary production and zooplankton production have been shown to decrease with increasing DOC (Ask et al. 2012, Kelly et al. 2014, Godwin et al. 2014). However, there are no dedicated studies on the effects of DOC on benthic invertebrate production, a major diet item for many fish (Vander Zanden & Vadeboncoeur 2002, Weidel et al. 2008).

The benthic zone is extremely important in contributing to lake productivity (Schindler & Scheuerell 2002, Vander Zanden & Vadeboncoeur 2002). Benthic algae has been shown to contribute a high percentage of the autochthonous (within lake) production in clear water lakes (Vadeboncoeur et al. 2008, Ask et al. 2009), and can be an important food resource for benthic

21 invertebrates (Hecky & Hesslein 1995, Vadeboncoeur et al. 2003). Subsequently, benthic invertebrates can constitute a large proportion of fish diets, particularly for fish in early growth phases or insectivorous fish such as bluegill (Gerking 1962, Olson 1996, Vander Zanden & Vadeboncoeur 2002, Weidel et al. 2008). Therefore factors that affect the productivity in the benthos can have a consequential effects on fish production in both pelagic and benthic habitats (Vander Zanden & Vadeboncoeur 2002).

A recent paper focusing on the effects of DOC on fish productivity included some rudimentary estimates of zoobenthos production, showing a decrease in productivity with increasing DOC (Karlsson et al. 2009). The authors of this paper speculated that this decrease was due to light limitation of primary production, but the results were based from ~3 zoobenthos samples per lake, at a fixed depth, which may not be appropriate to determine the mechanisms behind a loss of productivity. Zoobenthos biomass and productivity can be affected by many other factors other than benthic primary production. While some zoobenthos depend on primary production as their main food source (Strayer & Likens 1986, Hecky & Hesslein 1995, Devlin et al. 2013), the zoobenthos are a diverse group of animals and feed on many other sources such as detritus, or even other animals (Cummins and Klug 1979, Merritt et al. 2008). Therefore zoobenthos biomass and productivity can depend on the availability of nutrients, as well as other potential resources such as detritus, settling pelagic production and terrestrial particulate matter such as leaves (Strayer & Likens 1986, Rasmussen 1988). In addition, abiotic factors such as sediment structure, lake bathymetry, temperature, and dissolved oxygen concentrations can affect zoobenthos populations (Rasmussen & Kalff 1987, Dermott 1988, Rasmussen 1988, Babler et al. 2008). This is particularly important to consider because lakes with high DOC concentrations absorb heat more rapidly with depth, and as such develop shallower thermoclines (Snucins and Gunn 2000, Read and Rose 2013, Fig. 1). This means that the area of cold hypolimnion is larger in these lakes, and often the lack of atmospheric exchange with surface waters results in hypoxic conditions due to microbial respiration (Arvola 1984, Wetzel 2001, Wissel et al. 2003). This lack of oxygen, as well as lower temperatures may also lead to reduced zoobenthos production in DOC-rich lakes. While some taxa are able to withstand prolonged periods of anoxia, this occurs at the expense of growth,

22 and the combination with colder temperatures may mean that productivity in the hypolimnion of dark lakes is extremely low (Jónasson 1984, Butler & Anderson 1990, Heinis & Davids 1993). Chapter 1 will address if higher DOC concentrations lead to a reduction in zoobenthos productivity, and if so, what are the major mechanisms behind it - i.e. is the previously hypothesized paradigm of light limitation of benthic primary production correct? This will be achieved by determining zoobenthos productivity extensively over several depths in ten lakes across a DOC gradient and identifying mechanisms for variations in this productivity.

Potential effects of DOC on fish productivity

Published relationships on the effects of DOC on fish productivity generally show a negative effect with increasing DOC, however, some studies also show low productivity at very low DOC concentrations - sometimes resulting in a unimodal relationship (Lester et al. 2004a, Finstad et al. 2014). Potential explanations for this lower productivity at very low DOC concentrations include low nutrient availability/basal production, and high UV irradiance which can damage eggs and larval fish (Stasko et al. 2012, Finstad et al. 2014). As the DOC concentrations in the lakes studied in this thesis are above this threshold of low DOC (~2mg/l), I am mainly concerned about the negative effects of high DOC concentrations on fish productivity.

Previous research (Kelly et al. 2014, Craig et al. 2015) has indicated that secondary productivity in the form of zooplankton and zoobenthos are reduced with increasing DOC. A reduction in this resource availability may increase energetic costs for fish populations if they have to expend more energy to find and obtain a standard amount of resource (Jobling 1994). Growth, reproduction and productivity may suffer as a result (Wootton 1990, Bromage et al. 1992). One mechanism for the reduction of fish productivity in darker lakes may be reduced foraging efficiency as a result of lower food availability.

The reduced light climate associated with high DOC concentrations may be also be a related factor in a reduction in fish productivity if visual predators find it harder to forage under these conditions. Several studies have shown that fish feeding efficiency decreases with increasing turbidity and/or decreasing light (e.g. Vinyard & O'Brien 1976, Diehl 1988, Utne-Palm

23

2002, de Robertis et al. 2003, Carter et al. 2010). However, less attention has been focused on DOC-mediated effects on fish feeding behavior. Light is absorbed in DOC-rich waters which may create more visual contrasts between the prey and the environment in comparison to turbid waters where light is scattered (Ranåker et al. 2012). The few existing studies that examine DOC-mediated effects on fish feeding efficiency focus on piscivore-prey relationships, and show a unimodal response of DOC on feeding efficiency (Ranåker et al. 2012, Jönsson et al. 2013). At the low end of the DOC gradient, predators have a better chance of spotting prey in these relatively clear waters. However, while this ability is diminished in dark, high DOC waters, the predator can only see prey that is relatively close, and so generally has higher capture success. Thus, it is clear that the impact of DOC on fish feeding efficiency may not be straightforward.

While all of the current research on DOC-mediated feeding behavior has concerned pelagic feeding interactions, benthic interactions are just as important in lake ecosystems (Vadeboncoeur et al. 2002, Vander Zanden & Vadeboncoeur 2002, Strayer 2009). Benthic feeding may be less reliant on vision compared to pelagic foraging (Rowe et al. 2003), and so it is possible that fish in DOC-rich lakes could be more reliant on this resource. If DOC increases cause shifts in selectivity between pelagic and benthic prey, this could alter the flow of energy through food webs (Covich et al. 1999, Vander Zanden & Vadeboncoeur 2002). In chapter 2, I will address how DOC affects benthic feeding interactions using mesocosm experiments and a diet survey across lakes of varying DOC concentrations.

DOC can also affect the physical structure of lakes in ways that may be detrimental to fish productivity. Heat radiation penetrating the surface is absorbed faster as it travels through the water in darker lakes which results in a shallower thermocline (Fee et al. 1996, Houser 2006, Read & Rose 2013). This in turn increases the proportional volume of the colder and relatively oxygen depleted hypolimnion in dark lakes which may be a less suitable habitat for many organisms (Wetzel 2001). The shifts in light climate, thermal structure and oxycline depth in lakes across a DOC gradient may influence the size of optimum habitat areas for many fish species in ways that are not yet fully understood (Lester et al. 2004a, Stasko et al. 2012, Solomon et al. 2015).

24

Fish growth and reproduction can be affected by food availability, light climate, temperature and oxygen levels (Jobling 1994). All of these variables are likely to decrease at the lake-level as DOC concentrations increase (Ask et al. 2009, Read & Rose 2013). Although many studies have determined how these individual factors affect fish life histories, we do not currently know how the shifts in lake structure and function associated with increasing DOC may affect growth and fecundity in fish communities. For example, a reduction in food availability and dissolved oxygen concentrations may slow growth and cause fish to mature at smaller sizes or at later ages in dark lakes (Roff 1983, Stearns & Koella 1986). If food resources are more limited in darker lakes, fish may not be able to grow as fast and may take longer to reach the same size as fish in a clearer lake (Berg & Finstad 2008). In addition, fish exposed to food limitation can have reduced reproductive output (Bromage et al. 1992). In this situation, the fish will have to make a 'choice' between continuing to grow to achieve greater reproductive output per spawning event but taking longer to reach maturity (with no reproductive output); or maturing at a smaller size and/or age and reducing future reproductive capacity but potentially have more years of reproduction (MacArthur & Wilson 1967, Pianka 1970, Lester et al. 2004b). The reality could be anywhere between these two extremes but in either case, it suggests that fish in darker lakes will have a net lower reproductive output than fish from clear lakes. This could be a problem in darker lakes if fishing regulations are based on information from clear-water populations, in that fish may be removed before they have a chance to reproduce. Chapter 3 will focus on how DOC affects life history strategies in fish. I use bluegill (Lepomis macrochirus) as a model organism and determine growth rates and fecundity in eleven lakes spanning a DOC gradient. Bluegill are an excellent model species for this type of work as they are distributed over large gradients of DOC and occur in fairly high abundances even in DOC-rich lakes. They also have fairly plastic life history strategies (Aday et al. 2003), are well represented in the literature (e.g. the papers of Mittelbach, Keast, Werner, and Hall), and are an important sport fish of economic value (Drake et al. 1997). Understanding how crucial life history characteristics vary with DOC will help further our understanding of why fish productivity is reduced is DOC-rich lakes.

The use of multiple approaches to determine the effects of DOC on ecosystem productivity

25

A combination of several approaches can lend support to each other in clarifying how ecosystems and organisms respond to a given variable and overcome limitations associated with using each approach alone (Kitchell et al. 1988). Most of the current studies on the effects of DOC on ecosystem productivity are based on spatial DOC-gradient lake surveys (Karlsson et al. 2009, Finstad et al. 2014, Godwin et al. 2014, Kelly et al. 2014, Craig et al. 2015), mesocosm experiments (Jones & Lennon 2015, Sanders et al. 2015, Weidel et al. in review), or models (Jones et al. 2012). Spatial surveys can be useful for describing patterns in response to environmental gradients, but do not show us how an ecosystem may respond to change over time and may be confounded by other variables that change across a given gradient and the adaptation of organisms to their environment over time (Carpenter 1998). Mesocosm experiments are appropriate for testing mechanisms and responses to perturbation but the results from these may not be meaningful at the ecosystem level and there are often issues in amplifying results from mesocosms up to this level (Schindler 1990; Carpenter 1998; Hanson et al. 2011). While these approaches are useful for describing broad patterns and mechanisms (Carpenter 1998), DOC can and is varying temporally and results from the aforementioned studies may not accurately reflect how organisms and whole ecosystems react as DOC increases over time. There may be periods of succession or adaptation in organisms that need to be determined in order to fully understand the consequences of increasing DOC. Chapter 4 addresses this through the use of a whole-lake experiment, where we measure how zoobenthos biomass and productivity varies with a temporal increase of DOC. While whole- ecosystem experiments have little or no replication, they have strength in their ability to capture complex interactions in response to manipulations at the ecosystem level that may not be covered by surveys or mesocosms (Likens 1985). Whole ecosystem experiments can compliment these other methods by observing responses to perturbation at an ecologically relevant, and more realistic scale (Carpenter 1998). Therefore, a comparison with the results from chapter 1 (a spatial observational study), and chapter 4 (a whole lake-experiment), may help us better understand how future lake browning may affect lake food webs.

Thesis outline and study site

26

To summarize, the main aim of this thesis is to determine mechanisms behind the recently reported negative relationship between DOC and fish productivity. I begin this thesis by examining the effects of DOC on secondary consumer productivity (zoobenthos, an important resource for fish), in chapter 1. The productivity of secondary consumers is likely to propagate up the food web and influence fish production. If resources are limited, or the ability of a predator to utilize them is inhibited by DOC, this may lead to lower food consumption, and thus lower productivity. This is focus of my second chapter, where I examine the relationship between fish and benthic feeding efficiency. Chapter 3 focuses on the effects of DOC on fish life history strategies. If fish are resource limited, this may affect growth, or reproductive rates. How these rates are affected will influence the best way to manage fisheries in DOC-rich lakes. As much of the current literature, as well as the previous three chapters, are based on observational studies of organisms in lakes over a DOC gradient, we may not be truly capturing how ecosystems may react to a temporal increase in DOC. In chapter 4, I again examine the effects of DOC on zoobenthos productivity, but this time in the context of a whole-ecosystem experiment where we increase DOC concentrations in a manipulated basin over several years.

The field work for this thesis was conducted at and around the University of Notre Dame Environmental Research Center, a research station located on the Wisconsin-Michigan border, USA. The property contains a range of lakes spanning a DOC gradient, some of which are available for experimental manipulation, as well as a wet lab facility for mesocosm experiments. This enabled me to use a suite of approaches in order to determine the mechanisms behind DOC-productivity relationships. Nailing down these mechanisms is fundamental both for understanding the basics of how lake ecosystems work, as well for making predictions regarding the fate of aquatic ecosystems in a browning world.

27

References

Aday, D. D., D. H. Wahl, and D. P. Philipp. 2003. Assessing population-specific and environmental influences on bluegill life histories: a common garden approach. Ecology 84: 3370-3375.

Arvola, L. 1984. Vertical distribution of primary production and phytoplankton in two small lakes with different humus concentration in Southern Finland. Holarctic Ecology 7: 390- 398.

Ask, J., J. Karlsson, L. Persson, P. Ask, P. Byström, and M. Jansson. 2009. Terrestrial organic matter and light penetration: Effects on bacterial and primary production in lakes. Limnology and Oceanography 54: 2034-2040.

Ask, J., J. Karlsson, and M. Jansson. 2012. Net ecosystem production in clear-water and brown- water lakes. Global Biogeochemical Cycles 26: GB1017.

Babler, A. L., C. T. Solomon, and P. R. Schilke. 2008. Depth-specific patterns of benthic secondary production in an oligotrophic lake. Journal of the North American Benthological Society 27: 108-119.

Berg, O. K., and A. G. Finstad. 2008. Energetic trade-offs in reproduction: cost-benefit considerations and plasticity. Pages 351-375 in Rocha, M. J, A. Arukwe and B. G. Kapoor (eds). "Fish reproduction". Science Publishers, Enfield.

Brett, M. T., M. J. Kainz, S. J. Taipale, and H. Seshan. 2009. Phytoplankton, not allochthonous carbon, sustains herbivorous zooplankton production. Proceedings of the National Academy of Sciences 106: 21197-21201.

Bromage, N, J. Jones, C. Randall, M. Thrush, B. Davis, J. Springate, J. Duston, and G. Barker. 1992. Broodstock management, fecundity, egg quality and the timing of egg production in the rainbow trout (Oncorhynchus mykiss). Aquaculture 100: 141-166.

28

Butler, M. G., and D. H. Anderson. 1990. Cohort structure, biomass, and production in a merovoltine Chironomus population in a Wisconsin bog lake. Journal of the North American Benthological Society 9: 180-192.

Carpenter, S. R. 1998. The need for large-scale experiments to assess and predict the response of ecosystems to perturbation. Pages 287-312 in Pace, M. L. and P. M. Groffman (eds). "Successes, limitations, and frontiers in ecosystem science". Springer-Verlag, New York.

Carter, M. W., D. E. Shoup, J. M. Dettmers, and D. H. Wahl. 2010. Effects of turbidity and cover on prey selectivity of adult smallmouth bass. Transactions of the American Fisheries Society 139: 353-361.

Cole, J. J., N. F. Caraco, G. W. Kling, and T. K. Kratz. 1994. Carbon dioxide supersaturation in the surface waters of lakes. Science 265: 1568-1570.

Cole, J. J., S. R. Carpenter, M. L. Pace, M. C. Van de Bogert, J. L. Kitchell, and J. R. Hodgson. 2006. Differential support of lake food webs by three types of terrestrial organic carbon. Ecology Letters 9: 558-568.

Cole, J. J., S. R. Carpenter, J. Kitchell, M. L. Pace, C. T. Solomon, and B. Weidel. 2011. Strong evidence for terrestrial support of zooplankton in small lakes based on stable isotopes of carbon, nitrogen, and hydrogen. Proceedings of the National Academy of Sciences 108: 1975-1980.

Covich, A. P., M. A. Palmer, and T. A. Crowl. 1999. The role of benthic invertebrate species in freshwater ecosystems. Zoobenthic species influence energy flows and nutrient cycling. Bioscience 49: 119-127.

Cummins, K. W. and M. J. Klug. 1979. Feeding ecology of stream invertebrates. Annual Review of Ecology and Systematics 10: 147-172.

Craig, N. C., S. E. Jones, B. C. Weidel, and C. T. Solomon. 2015. Habitat, not resource availability, limits consumer production in lake ecosystems. Limnology and Oceanography 60: 2079- 2089.

29 de Robertis, A., C. H. Ryer, A. Veloza, and R. D. Brodeur. 2003. Differential effects of turbidity on prey consumption of piscivorous and planktivorous fish. Canadian Journal of Fisheries and Aquatic Sciences 60: 1517-1526.

Dermott, R. M. 1988. Zoobenthic distribution and biomass in the Turkey Lakes. Canadian Journal of Fisheries and Aquatic Sciences 45: 107-114.

Devlin, S. P., M. J. Vander Zanden and Y. Vadeboncoeur. 2013. Depth-specific variation in carbon isotopes demonstrates resource partitioning among the littoral zoobenthos. Freshwater Biology 58: 2389-2400.

Diehl, S. 1988. Foraging efficiency of three freshwater fishes: Effects of structural complexity and light. Oikos 53: 207-214.

Drake, M. T., J. E. Claussen, D. P. Philipp, and D. L. Pereira. 1997. A comparison of bluegill reproductive strategies and growth among lakes with different fishing intensities. North American Journal of Fisheries Management 17: 496-507.

Evans, C. D., D. T. Monteith, and D. M. Cooper. 2005. Long-term increases in surface water dissolved organic carbon: Observations, possible causes and environmental impacts. Environmental pollution 137: 55-71.

Fee, E. J., R. E. Hecky, S. E. M. Kasian, and D. R. Cruikshank. 1996. Effects of lake size, water clarity, and climatic variability on mixing depths in Canadian Shield Lakes. Limnology and Oceanography 41: 912-920.

Finstad, A. G., I. P. Helland, O. Ugedal, T. Hesthagen, and D. O. Hessen. 2014. Unimodal response of fish yield to dissolved organic carbon. Ecology Letters 17: 36-43.

Freeman, C., C. D. Evans, and D. T. Monteith. 2001. Export of organic carbon from peat soils. Nature 412: 785.

Garnett, M. H., P. Ineson, and A. C. Stevenson. 2000. Effects of burning and grazing on carbon sequestration in a Pennine blanket bog, UK. The Holocene 10: 729-736.

30

Gerking, S. D. 1962. Production and food utilization in a population of bluegill sunfish. Ecological Monographs 32: 31-78.

Godwin, S. C., S. E. Jones, B. C. Weidel, and C. T. Solomon. 2014. Dissolved organic carbon concentration controls benthic primary production: Results from in situ chambers in north-temperate lakes. Limnology and Oceanography 59: 2112-2120.

Hanson, P. C., D. L. Bade, S. R. Carpenter, and T. K. Kratz. 2003. Lake metabolism: relationships with dissolved organic carbon and phosphorus. Limnology and Oceanography 48: 1112- 1119.

Hecky, R. E., and R. H. Hesslein. 1995. Contributions of benthic algae to lake food webs as revealed by stable isotope analysis. Journal of the North American Benthological Society 14: 631-653.

Heinis, F., and C. Davids. 1993. Factors governing the spatial and temporal distribution of chironomid larvae in the Maarsseveen Lakes with special emphasis on the role of oxygen conditions. Netherlands Journal of Aquatic Ecology 27: 21-34.

Hessen, D. O. 1992. Dissolved organic carbon in a humic lake: effects on bacterial production and respiration. Hydrobiologia 229: 115-123.

Hessen, D. O., and L. J. Tranvik (eds). 1998. Aquatic humic substances: ecology and biogeochemistry. Vol. 133. Springer-Verlag, Berlin.

Hinton, M. J., S. Schiff, and M. C. English. 1997. The significance of storms for the concentration and export of dissolved organic carbon from two Precambrian Shield catchments. Biogeochemistry 36: 67-88.

Houser, J. N. 2006. Water color affects the stratification, surface temperature, heat content, and mean epilimnetic irradiance of small lakes. Canadian Journal of Fisheries and Aquatic Sciences 63: 2447-2455.

Jobling, M. 1994. Fish bioenergetics. Chapman & Hall, London.

31

Jónasson, P. M. 1984. Oxygen demand and long term changes of profundal zoobenthos. Hydrobiologia 115: 121-126.

Jones, R. I. 1992. The influence of humic substances on lacustrine planktonic food chains. Hydrobiologia 229: 73-91.

Jones, R. I., J. Grey, D. Sleep, and C. Quarmby. 1998. An assessment, using stable isotopes, of the importance of allochthonous organic carbon sources to the pelagic food web in Loch Ness. Proceedings of the Royal Society of London. Series B: Biological Sciences 265: 105- 110.

Jones, S. E., C. T. Solomon, and B. Weidel. 2012. Subsidy or subtraction: How do terrestrial inputs influence consumer production in lakes? Freshwater Reviews 5: 37-49.

Jones, S. E., and J. T. Lennon. 2015. A test of the subsidy-stability hypothesis: the effects of terrestrial carbon in aquatic ecosystems. Ecology 96: 1550-1560.

Jönsson M., L. Ranåker, P. A. Nilsson, and C. Brönmark. 2013. Foraging efficiency and prey selectivity in a visual predator: differential effects of turbid and humic water. Canadian Journal of Fisheries and Aquatic Sciences 70: 1685-1690.

Karlsson, J., A. Jonsson, M. Meili, and M. Jansson. 2003. Control of zooplankton dependence on allochthonous organic carbon in humic and clear-water lakes in northern Sweden. Limnology and Oceanography 48: 269-276.

Karlsson, J., P. Byström, J. Ask, P. Ask, L. Persson, and M. Jansson. 2009. Light limitation of nutrient-poor lake ecosystems. Nature 460: 506-509.

Karlsson, J., A-K. Bergström, P. Byström, C. Gudasz, P. Rodríguez, and C. L. Hein. 2015. Terrestrial organic matter input suppresses biomass production in lake ecosystems. Ecology 96: 2870-2876.

Kelly, P. T., C. T. Solomon, B. C. Weidel, and S. E. Jones. 2014. Terrestrial carbon is a resource, but not a subsidy, for lake zooplankton. Ecology 95: 1236-1242.

32

Kitchell, J. F., S. M. Bartell, S. R. Carpenter, D. J. Hall, D. J. McQueen, W. E. Neill, D. Scavia and E. E. Werner. 1988. Epistemology, experiments, and pragmatism. Pages 263-280 in Carpenter, S. R. (ed). "Complex interactions in lake communities" Springer-Verlag, New York.

Kortelainen, P. 1999. Chapter 3: Occurrence of humic waters. Pages 41-57 in Keskitalo, J., and P. Eloranta (eds). "Limnology of humic waters". Backhuys Publishers, Leiden.

Lester, N. P., A. J. Dextrase, R. S. Kushneriuk, M. R. Rawson, and P. A. Ryan. 2004a. Light and temperature: key factors affecting walleye abundance and production. Transactions of the American Fisheries Society 133: 588-605.

Lester, N. P., B. J. Shuter, and P. A. Abrams. 2004b. Interpreting the von Bertalanffy model of somatic growth in fishes: the cost of reproduction. Proceedings of the Royal Society of London B: Biological Sciences 271: 1625-1631.

Likens, G. E. 1985. An experimental approach for the study of ecosystems: the fifth Tansley Lecture. Journal of Ecology 73: 381-396.

MacArthur, R. H., and E. O. Wilson. 1967. The theory of island biogeography. Princeton University Press, Princeton.

Merritt, R.W., K. W. Cummins, and M. B. Berg, editors. 2008. An introduction to the aquatic insects of North America. 4th edition. Kendall Hunt: Dubuque, Indiana.

Monteith, D. T., J. L. Stoddard, C. D. Evans, H. A. de Wit, M. Forsius, T. Høgasen, A. Wilander, B. L. Skjelkvåle, D. S. Jeffries, J. Vuorenmaa, B. Keller, J. Kopácek, and J. Vesely. 2007. Dissolved organic carbon trends resulting from changes in atmospheric deposition chemistry. Nature 450: 537-541.

Moore, T. R. 1989. Dynamics of dissolved organic carbon in forested and disturbed catchments, Westland, New Zealand: 1. Maimai. Water Resources Research 25: 1321-1330.

Mulholland, P. J., and E. J. Kuenzler. 1979. Organic carbon export from upland and forested wetland watersheds. Limnology and Oceanography 24: 960-966.

33

Olson, M. H. 1996. Predator-prey interactions in size-structured fish communities: implications of prey growth. Oecologia 108: 757-763.

Oni, S. K., T. Tiwari, J. L. J Ledesma, A. M. Ågren, C. Teutschbein, J. Schelker, H. Laudon, and M. N. Futter. 2015. Local- and landscape-scale impacts of clear-cuts and climate change on surface water dissolved organic carbon in boreal forests. Journal of Geophysical Research: Biogeosciences 120. DOI: 10.1002/2015/JG003190.

Pianka, E. R. 1970. On r- and K-selection. The American Naturalist 104: 592-597.

Polis, G. A., W. B. Anderson and R. D. Holt. 1997. Toward an integration of landscape and food web ecology: The dynamics of spatially subsidized food webs. Annual review of ecology and systematics 28: 289-316.

Prairie, Y. T. 2008. Carbocentric limnology: looking back, looking forward. Canadian Journal of Fisheries and Aquatic Sciences 65: 543-548.

Ranåker, L., M. Jönsson, P. A. Nilsson, and C. Brönmark. 2012. Effects of brown and turbid water on piscivore-prey fish interactions along a visibility gradient. Freshwater Biology 57: 1761-1768.

Rasmussen, J. B. and J. Kalff. 1987. Empirical models for zoobenthic biomass in lakes. Canadian Journal of Fisheries and Aquatic Sciences 44: 990-1001.

Rasmussen, J. B. 1988. Littoral zoobenthic biomass in lakes, and its relationship to physical, chemical, and trophic factors. Canadian Journal of Fisheries and Aquatic Sciences 45: 1436-1447.

Read, J. S., and K. C. Rose. 2013. Physical responses of small temperate lakes to variation in dissolved organic carbon concentrations. Limnology and Oceanography 58: 921-931.

Roff, D. A. 1983. An allocation model of growth and reproduction in fish. Canadian Journal of Fisheries and Aquatic Sciences 40: 1395-1404.

Roulet, N., and T. R. Moore. 2006. Environmental chemistry: Browning the waters. Nature 444: 283-284.

34

Rowe, D. K., T. L. Dean, E. Williams, and J. P. Smith. 2003. Effects of turbidity on the ability of juvenile rainbow trout, Oncorhynchus mykiss, to feed on limnetic and benthic prey in laboratory tanks. New Zealand Journal of Marine and Freshwater Research 37: 45-52.

Sanders, R. W,, S. L. Cooke, J. M. Fischer, S. B. Fey, A. W. Heinze, W. H. Jeffery, A. L. Macaluso, R. E. Moeller, D. P. Morris, P. J. Neale, M. H. Olson, J. D. Pakulski, J. A. Porter, D. M. Schoener, and C. E. Williamson. 2015. Shifts in microbial food web structure and productivity after additions of naturally occurring dissolved organic matter: results from large-scale lacustrine mesocosms. Limnology and Oceanography 60: 2130-2144.

Schiff, S. L., R. Aravena, S. E. Trumbore, M. J. Hinton, R. Elgood, and P. J. Dillon. 1997. Export of DOC from forested catchments on the Precambrian Shield of Central Ontario: clues from 13C and 14C. Biogeochemistry 36: 43-65.

Schindler, D. W. 1990. Experimental perturbations of whole lakes as tests of hypotheses concerning ecosystem structure and function. Oikos 57: 25-41.

Schindler, D. E. and M. D. Scheuerell. 2002. Habitat coupling in lake ecosystems. Oikos 98: 177- 189.

Snucins, E. and J. Gunn. 2000. Interannual variation in the thermal structure of clear and colored lakes. Limnology and Oceanography 45: 1639-1646.

Sobek, S., L. J. Tranvik, Y. T. Prairie, P. Kortelainen and J. J. Cole. 2007. Patterns and regulation of dissolved organic carbon: an analysis of 7,500 widely distributed lakes. Limnology and Oceanography 52: 1208-1219.

Solomon, C. T., S. R. Carpenter, M. K. Clayton, J. J. Cole, J. J. Coloso, M. L. Pace, J. M. Vander Zanden, and B. C. Weidel. 2011. Terrestrial, benthic, and pelagic resource use in lakes: results from a three-isotope Bayesian mixing model. Ecology 92: 1115-1125.

Solomon, C. T., S. E. Jones, B. C. Weidel, I. Buffam, M. L. Fork, J. Karlsson, S. Larsen, J. T. Lennon, J. S. Read, S. Sadro, and J. E. Saros. 2015. Ecosystem consequences of changing inputs of

35

terrestrial dissolved organic matter to lakes: current knowledge and future challenges. Ecosystems 18: 376-389.

Stasko, A. D., J. M. Gunn, and T. A. Johnston. 2012. Role of ambient light in structuring north- temperate fish communities: potential effects of increasing dissolved organic carbon concentration with a changing climate. Environmental Reviews 20: 173-190.

Stearns, S. C., and J. C. Koella. 1986. The evolution of phenotypic plasticity in life-history traits: predictions of reaction norms for age and size at maturity. Evolution 40: 893-913.

Strayer, D. and G. E. Likens. 1986. An energy budget for the zoobenthos of Mirror Lake, New Hampshire. Ecology 67: 303-313.

Strayer, D. L. 2009. Benthic invertebrate fauna, lakes and reservoirs. Pages 191-204 in Likens, E. (ed). "Encyclopedia of inland waters. Volume 2". Elsevier, Oxford.

Tranvik, L. J. 1988. Availability of dissolved organic carbon for planktonic bacteria in oligotrophic lakes of differing humic content. Microbial Ecology 16: 311-322.

Tranvik, L. J. 1992. Allochthonous dissolved organic matter as an energy source for pelagic bacteria and the concept of the microbial loop. Hydrobiologia 229: 107-114.

Utne-Palm, A. C. 2002. Visual feeding of fish in a turbid environment: physical and behavioural aspects. Marine and Freshwater Behaviour and Physiology 35: 111-128.

Vadeboncoeur, Y., M. J. Vander Zanden, and D. M. Lodge. 2002. Putting the lake back together: Reintegrating benthic pathways into lake food web models. Bioscience 52: 44-54.

Vadeboncoeur, Y., E. Jeppesen, M. J. Vander Zanden, H. Schierup, K. Christoffersen, and D. M. Lodge. 2003. From Greenland to green lakes: Cultural eutrophication and the loss of benthic pathways in lakes. Limnology and Oceanography 48: 1408-1418.

Vadeboncoeur, Y., G. Peterson, M. J. Vander Zanden, and J. Kalff. 2008. Benthic algal production across lake size gradients: Interactions among morphometry, nutrients, and light. Ecology 89: 2542-2552.

36

Vander Zanden, M. J., and Y. Vadeboncoeur. 2002. Fishes as integrators of benthic and pelagic food webs in lakes. Ecology 83: 2152-2161.

Vinyard, G. L., and W. J. O’Brien. 1976. Effects of light and turbidity on the reactive distance of bluegill (Lepomis macrochirus). Journal of the Fisheries Research Board of Canada 33: 2845-2849.

Weidel, B., S. Carpenter, J. Cole, J. Hodgson, J. Kitchell, M. Pace, and C. Solomon. 2008. Carbon sources supporting fish growth in a north temperate lake. Aquatic Sciences 70: 446-458.

Weidel, B. C., K. Baglini, S. E. Jones, P. T. Kelly, C. T. Solomon, and J. A. Zwart. In review. Differences in light climate due to dissolved organic carbon concentration drive species- specific changes in fish zooplankivory.

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

Wilkinson, G. M., M. L. Pace, and J. J. Cole. 2013. Terrestrial dominance of organic matter in north temperate lakes. Global Biogeochemical Cycles 27: 43-51.

Wissel, B., W. J. Boeing, and C. W. Ramcharan. 2003. Effects of water color on predation regimes and zooplankton assemblages in freshwater lakes. Limnology and Oceanography 48: 1965-1976.

Weyhenmeyer, G. A., R. A. Müller, M. Norman, and L. J. Tranvik. 2015. Sensitivity of freshwaters to browning in response to future climate change. Climatic Change, 1-15. DOI: 10.1007/s10584-015-1514-z.

Wootton, R. J. 1990. Ecology of teleost fishes. Chapman & Hall, London.

37

Figures

Figure 1. Conceptual figure showing how DOC affects the physical structure of lake ecosystems. In darker waters, light is extinguished more rapidly with depth, reducing the volume of water and bottom area capable of supporting photosynthesis. Heat penetration is also more limited in these systems, resulting in steeper and shallower thermoclines, and relatively larger areas of dark, cold, and potentially deoxygenated hypolimnion. Modified from Solomon et al. (2015).

38

Chapter 1

HABITAT, NOT RESOURCE AVAILABILITY, LIMITS CONSUMER PRODUCTION IN LAKE ECOSYSTEMS. Nicola Craig1*, Stuart E. Jones2, Brian C. Weidel3, and Christopher T. Solomon1

1 Dept. of Natural Resource Sciences, McGill University, Ste. Anne de Bellevue, QC, Canada.

2 Dept. of Biological Sciences, University of Notre Dame, Notre Dame, IN, USA.

3 U.S. Geological Survey, Great Lakes Science Center, Lake Ontario Biological Station, Oswego, NY, USA.

Status: Published in Limnology and Oceanography, 2015, DOI: 10.1002/lno.10153.

Abstract

Food web productivity in lakes can be limited by dissolved organic carbon (DOC), which reduces fish production by limiting the abundance of their zoobenthic prey. We demonstrate that in a set of ten small, north temperate lakes spanning a wide DOC gradient, these negative effects of high DOC concentrations on zoobenthos production are driven primarily by availability of warm, well-oxygenated habitat, rather than by light limitation of benthic primary production as previously proposed. There was no significant effect of benthic primary production on zoobenthos production after controlling for oxygen, even though stable isotope analysis indicated that zoobenthos do use this resource. Mean whole-lake zoobenthos production was lower in high-DOC lakes with reduced availability of oxygenated habitat, as was

39 fish biomass. These insights improve understanding of lake food webs and inform management in the face of spatial variability and ongoing temporal change in lake DOC concentrations.

Introduction

In lake ecosystems, consumers in benthic habitats – the zoobenthos – are a key food web link between basal resources and fishes. For instance, reliance of fishes on zoobenthic prey was 65% on average in a survey of 470 lacustrine fish populations (Vander Zanden and

Vadeboncoeur 2002), and varied from 60 to 80% in an intensive study of three species in a single lake (Weidel et al. 2008). Fishes have strong ecological effects on lakes (Carpenter et al.

2001, Vanni 2002) and support economically and culturally valuable fisheries. Although zoobenthos play a central role in structuring lakes in ecologically and societally important ways, surprisingly little is known about the ecological factors that limit their productivity. The most comprehensive analyses indicate that zoobenthos productivity may be regulated by lake trophic status and availability of resources such as detritus formed from settling particulates, and benthic algae (Strayer & Likens 1986, Rasmussen 1988); as well as abiotic factors like habitat structure, lake morphometry, humic water color, and dissolved oxygen concentration

(Rasmussen & Kalff 1987, Dermott 1988, Rasmussen 1988, Babler et al. 2008). However, the considerable effort required to quantify zoobenthos abundance, especially given their patchy distributions in space and time, means that robust comparative analyses are rare.

The current conceptual model of food web productivity in nutrient-poor lake ecosystems emphasizes light-mediated resource limitation of zoobenthos and ultimately fish populations (Karlsson et al. 2009, Finstad et al. 2014). Dissolved organic carbon (DOC) derived

40 from terrestrial organic matter stains the water, such that lakes with higher DOC concentrations have darker water (Jones 1992). This reduces light penetration and thereby also benthic primary production (Ask et al. 2012, Godwin et al. 2014). Karlsson et al. (2009) proposed that reduced benthic primary production in higher-DOC lakes limits zoobenthos production, which in turn limits fish production. This proposed causal chain fits with major established patterns in lake ecology including light limitation of benthic primary production (Hansson 1992,

Vadeboncoeur et al. 2008, Ask et al. 2012) and the importance of zoobenthos prey in supporting fish production (Vander Zanden & Vadeboncoeur 2002, Weidel et al. 2008).

However, there is limited evidence to support the idea that benthic primary production can limit zoobenthos production. While some zoobenthos can and do rely on benthic primary production (Strayer & Likens 1986, Hecky & Hesslein 1995, Devlin et al. 2013, Lau et al. 2014), others employ a diversity of other feeding strategies (Cummins and Klug 1979) and can feed on resources such as terrestrial particulates, settling phytoplankton and other animals (Merritt et al. 2008). In a survey of eight oligotrophic Arctic lakes, Northington et al. (2010) found only a weak relationship between zoobenthos production and benthic primary production.

An alternative mechanism by which DOC might limit zoobenthos production could be its effects on the availability of thermally suitable and well-oxygenated habitat. Because DOC absorbs incoming heat as well as light, darker lakes develop stronger, shallower thermal stratification (Snucins and Gunn 2000, Read and Rose 2013). This means that a greater proportion of their volume and sediment surface area lies in the hypolimnion, where the water is cold and isolated from light and atmospheric exchange such that respiration may drive dissolved oxygen to extremely low levels (Arvola 1984, Wetzel 2001, Wissel et al. 2003). Most

41 zoobenthos, like other aerobic poikilotherms, are strongly affected by both temperature and dissolved oxygen concentration (Dermott 1988, Plante and Downing 1989), although there are some taxa, such as Chironomus spp., which can tolerate periods of anoxia at the expense of growth (Jónasson 1984).

In this study, we tested the importance of the DOC-mediated resource-limitation and habitat-limitation mechanisms for controlling zoobenthos production. We estimated zoobenthos production and resource use in ten low-nutrient lakes spanning a wide DOC gradient, and related rates of production to measurements of benthic primary production, dissolved oxygen, and temperature. Our results confirm a strong negative effect of DOC on zoobenthos production. However, they indicate that this effect occurs mainly via reductions in dissolved oxygen and available habitat, not via resource limitation driven by light availability and benthic primary production.

Methods

Study site and sample collection

We estimated zoobenthos production in ten lakes at the University of Notre Dame

Environmental Research Centre located on the Wisconsin-Michigan border, USA (46.228° N

89.524° W). The lakes spanned broad environmental gradients; for instance, DOC ranged from

5.3 to 23.0 mg/L and total phosphorous (TP) ranged from 11.4 to 33.9 μg/L (Table 1.1).

We collected zoobenthos samples from each lake on three occasions across the summer growing period, in late May, late June and early August 2012. On each sampling occasion samples were taken at 4-5 depths, depending on the maximum depth of the lake, along four

42 replicate transects. All lakes were sampled at 0.5, 1, and 3 meters; deep lakes were also sampled at 8 and 12 meters and shallower lakes (<8 m) were sampled at the deepest depth available. We used a push corer to sample sediment at depths of 1 m or less (5 cores per sample, 0.017 m2 total) and an Ekman grab (0.023 m2) to sample at all deeper depths. Sediment samples were seived through 250 µm mesh bags and organisms were sorted visually from the debris on the same day as collection and stored in 70% ethanol. Spot checks on ~10% of samples indicated that we were effective at picking macrozoobenthos from the samples, but it is likely that some small macrozoobenthos were missed. In addition, this method was not designed to collect small meiofauna. Thus our production estimates should be viewed as macroinvertebrate-specific and as minimal estimates of total zoobenthos production (Strayer and Likens 1986).

Zoobenthos production

We identified zoobenthos to genus or the lowest possible taxonomic level using a stereo microscope, following the keys of Holsinger (1972), Stern (1990), and Merritt et al. (2008). We photographed each individual using a digital microscope camera and measured head capsule width, body length, or shell width from the images using ImageJ software (National Institutes of

Health, USA). Zoobenthos dry mass was calculated from these measurements following published length-mass relationships (Appendix 1A). We calculated production for each taxon at each sample site using the Plante and Downing (1989) predictive regression model, which uses mean annual biomass, maximum individual body mass, and mean annual surface water temperature as predictors. We used data from the summer rather than the entire year to estimate these predictors, which probably biased our production estimates towards higher

43 values. To check the magnitude of this bias, we applied this method to the data of Babler et al.

(2008), who previously estimated zoobethos production using the size-frequency method in one of the lakes that we studied here. Estimates of zoobenthos production from the Plante and

Downing method with summer data were 10 to 42% higher than those from the size-frequency method.

We summed production across taxa and reported it in two different manners in the results. Depth-specific production is the average production from the four replicate samples taken at each depth within each lake. Whole-lake average production was calculated by dividing each lake up into depth bands centered around our sampling depths, multiplying the area of each depth band by the average depth-specific production, summing across the depth bands, and dividing by the total area.

Enviromental variables

We measured depth-specific water temperatures and dissolved oxygen concentrations on each of the three zoobenthos sampling occasions using a handheld polarographic sensor (YSI

Pro 20, Yellow Springs Instruments, USA), and used the mean of these three measurements as an indicator of average summertime conditions at each depth. We similarly calculated average light availability at each depth based on profile measurements of photosynthetically active radiation (PAR) with an underwater quantum PAR sensor and light meter (LICOR LI-192SA and

LI-250A, LICOR, USA).

We measured depth-specific benthic primary production using in situ benthic chambers and the diel oxygen method, as described in Godwin et al. (2014). Briefly, we placed optical

44 dissolved oxygen sensors inside clear skylight domes deployed for 3-7 days on the bottom at depths corresponding to the 60%, 25%, and 5% light level in each lake, and calculated daily esimates of primary production based on the rate of change in dissolved oxygen in each dome using a modified version of the method described by Cole et al. (2000). We repeated these deployments three times over the course of the summer. For every set of deployments in a given lake, benthic primary production estimates were linearly interpolated at 0.5 m depth intervals from the rates calculated at the deployment depths. To make these interpolation calculations, we assumed that benthic primary production was zero below the 1% light level and that it was constant between the minimum deployment depth and the surface of the lake.

We calculated the mean depth-specific benthic primary production by averaging across all of the available measurements for each lake-depth.

In addition to these depth-specific measures, we calculated average whole-lake temperature, dissolved oxygen, light availability, and benthic primary production by weighting depth-specific values by the area of each depth band, as described above for zoobenthos production calculations. We also measured water column DOC, chlorophyll a, total phosphorous, and total nitrogen concentrations in each lake based on epilimnion samples collected on the three sampling occassions across the summer, and processed as described in

Kelly et al. (2014). We estimated fish biomass in six of our lakes in 2013 using catch per unit effort of zoobenthivorous fishes from overnight fyke net sets to estimate the relationship between benthic invertebrate production and fish biomass.

45

Resource use

We estimated the reliance of zoobenthos on benthic, pelagic, and terrestrial sources of primary production using measurements of C, N, and H stable isotope ratios and a Bayesian mixing model, following the methods in Solomon et al. (2011). For the benthic end member of the mixing model we scraped periphyton from ceramic tiles positioned at 0.5 m depth in each lake. For the pelagic end member, we estimated the stable isotope ratios of phytoplankton following the method of Cole et al. (2011). For the terrestrial end member, we used published stable isotope ratios for leaf litter collected within 5 km of our study lakes (Cole et al. 2011).

Finally, for the environmental water end member in the H equation of the mixing model, we measured the δD of surface water in each lake on three occasions throughout the summer.

Zoobenthos for stable isotope analysis were collected simultaneously with the other zoobenthos samples from four sites on each lake at 1 m depth. We picked chironomids, odonates, and trichopterans from these samples, held them overnight to allow for gut evacuation, then pooled them by taxon and sample date, dried them at 60ºC, and ground them to a fine powder. We measured stable isotope ratios on isotope ratio mass spectrometers at the University of Notre Dame (δ13C and δ15N) and Northern Arizona University (δD). Stable isotope ratios of the three zoobenthos taxa were similar within each lake, so we fit a single mixing model for each lake using all of the available zoobenthos samples.

Identifying predictors of zoobenthos production

We fit linear regression models, including lake as a blocking factor, to describe the effects of benthic primary production, dissolved oxygen, and water temperature (singly and in

46 all possible combinations; seven candidate models in all) on depth-specific zoobenthos production. We compared the candidate models using AICc to identify the model(s) with the best predictive ability. Our predictors (particularly benthic primary production) are themselves measured with error, which can lead to biased parameter estimates and loss of power in classical regression models (Fuller 1987). To avoid these issues we used a measurement error model (also known as an error-in-variables model). Specifically, we calculated the maximum likelihood estimates of the parameters, and the associated likelihood, following Equations

2.2.10 through 2.2.12 of Fuller (1987). We bootstrapped the residuals of our fitted models

10,000 times to determine 95% confidence intervals for the parameters (Carroll et al. 2006,

Manly 2007). All predictors were log(x+1) transformed and converted to Z-scores prior to analysis, to normalize distributions and facilitate comparisons among the coefficients. We present the coefficient of determination (R2) for each model, but note that these should be used only as rough aids to interpretation because the R2 calculation is not strictly valid in the presence of measurement error (e.g. Cheng et al. 2014). While we report results from these error-in-variables models here, we note that results were qualitatively similar when we used standard least-squares regression. We also used standard least-squares regression to ask whether there was a relationship between zoobenthos production at the whole-lake level and

DOC concentration or other lake-level predictors.

Results

We collected a total of 529 zoobenthos samples and measured 10,950 individuals from

11 orders and 32 families. Total zoobenthos biomass per sample averaged 0.85 g dry mass m-2 and ranged between 0 and 9.87 g dry mass m-2. Chironomidae was the dominant taxon in all

47 lakes, constituting 13.1-100% of the biomass in shallow samples (0.5-1 m) and an even greater portion of biomass (39.7-100%) in deeper samples (3-12 m). We did not see strong evidence for a shift in the proportion of biomass associated with any particular order of invertebrates along the DOC gradient although some genera appeared to be more associated with clear or dark lakes.

Depth gradients and DOC

Differences in DOC concentration between lakes were associated with large differences in environmental depth gradients within lakes. To illustrate this, we show depth profiles for two lakes representing the clear and dark ends of our DOC gradient (Figure 1.1; values for lakes with intermediate DOC concentration fall between the two extremes displayed). Depth profiles of limnological variables showed that high-DOC “dark” lakes become colder, darker and oxygen depleted at much shallower depths than do low-DOC “clear” lakes. For example, the dissolved oxygen concentration at 3 m depth was 0.1 mg/L in our darkest lake, compared to 8.5 mg/L in our clearest lake. Similar comparisons can be made for temperature (7.4 °C compared to 21.4

°C) and benthic primary production (0 mg C m-2 d-1 compared to 575 mg C m-2 d-1). Pearson's correlation coefficients between dissolved oxygen, temperature, benthic primary production, and light ranged from 0.59 to 0.88. Both the thermocline depth (the depth at which the change in temperature with depth, between warm surface waters and cold hypolimnetic waters, is most rapid) and the oxycline depth (similarly defined for dissolved oxygen concentration) were

2 shallower in lakes with higher DOC concentrations (F1,8 = 20.13, p = <0.01, R = 0.72, and F1,8 =

10.54, p = 0.01, R2 = 0.57 respectively).

48

Depth-specific zoobenthos production

Dissolved oxygen was the best single predictor of within- and among-lake variation in depth-specific zoobenthos production that we considered, while benthic primary production was the worst (Figure 1.2, Table 1.2). The best model as indicated by AICc was the one that contained only dissolved oxygen and the lake term as predictors, and this model was clearly better than the next-best model (ΔAICc=11), which contained only temperature and the lake term. Individually, benthic primary production was the least effective predictor of zoobenthos production, and its estimated coefficient did not differ from zero in any model that included either dissolved oxygen or temperature as a predictor. We observed that sites with zero benthic primary production had zoobenthos production ranging between 0 and 17.7 g m-2 y-1

(i.e. including the lowest and nearly the highest rates of zoobenthos production that we measured); furthermore, there was little or no trend in zoobenthos production with increasing benthic primary production above 0 (Figure 1.2c).

Patterns of zoobenthos production with depth mirrored the depth response of dissolved oxygen and other environmental gradients observed within the lakes, which were strongly influenced by DOC. In clear (low DOC) lakes, zoobenthos production was fairly constant across depths, with high production even at the deepest sites (Figure 1.3a). In stark contrast, production in the darkest (high DOC) lakes was negligible at 4-5 m and deeper (Figure 1.3c).

Considering that thermocline and oxycline depths are reduced in darker lakes (Figure 1.1) and that zoobenthos production is related to these variables (Figure 1.2), it appears that zoobenthos production is limited in dark lakes by availability of suitable habitat because oxygen is depleted at shallower depths.

49

Whole-lake zoobenthos production

Whole-lake average zoobenthos production was negatively related to DOC concentration, reflecting the loss of productivity at deeper depths in darker lakes (Figure 1.4a,

2 F1,8 = 7.89, p = 0.02, R = 0.50, log-log regression). While zoobenthos production varied substantially among the lakes with low DOC concentrations, it was low in all of the lakes with high DOC concentrations (Figure 1.3a, 1.3b). Whole-lake zoobenthos production was positively

2 related to whole-lake average estimates of dissolved oxygen (F1,8 = 9.25, p = 0.02, R = 0.54, log-

2 log regression), temperature (F1,8 = 20.73, p = <0.01, R = 0.72, log-log regression) and benthic

2 primary production (F1,8 = 13.4, p = <0.01, R = 0.62, log-log regression), all of which were also negatively related to DOC concentration. There was no relationship between water column nutrient concentrations (total phosphorus and total nitrogen) and whole-lake zoobenthos production (p > 0.13, R2 < 0.26). Fish biomass (measured as biomass catch per unit effort) was positively related to zoobenthos production for the six lakes in which data was available (Figure

2 1.4b, F1,4 = 18.46, p = 0.01, R = 0.82).

Resource use

Reliance of zoobenthos collected from 1 m depth on benthic primary production was moderate to high in the clearest lakes, declined over a DOC range of ~ 7 to 9 mg/L, and was generally lower in the dark lakes with DOC >9 mg/L (Figure 1.5). Zoobenthos reliance on terrestrial and pelagic primary production (medians of posterior distributions; see Appendix 1B for additional detail) ranged from 19-65% and from 6-46%, respectively; neither was clearly related to DOC concentration.

50

Discussion

A revised conceptual model for the effects of DOC on food web productivity

The current conceptual model of lake food webs in nutrient-poor settings posits that the effect of DOC on light availability limits resource availability in the form of benthic primary production, thereby limiting zoobenthos and ultimately fish production (Karlsson et al. 2009,

Finstad et al. 2014). This conceptual model assumes that benthic primary production is the essential resource that limits zoobenthos production, such that zoobenthos production increases when benthic primary production increases. Our results do not support that assumption. While benthic primary production by itself is significantly related to zoobenthos production, this model had much lower predictive ability than models that included water temperature or, especially, dissolved oxygen as predictors. Furthermore, confidence intervals for this effect overlapped zero whenever water temperature or dissolved oxygen also appeared in the model. Many sites with zero benthic primary production had quite high rates of zoobenthos production, up to 75% of the maximum zoobenthos production that we observed at any site. Similarly to the results of Northington et al. (2010), only a weak relationship existed between zoobenthos production and benthic primary production at sites with non-zero benthic primary production. If we consider patterns at 1 m depth across lakes of increasing DOC, we

2 observed decreases in benthic primary production (F1,8 = 14.52, p = <0.01, R = 0.64) and reliance of zoobenthos on benthic primary production (Fig. 1.5), yet no decrease in production

2 of zoobenthos (F1,8 = 0.44, p = 0.52, R = 0.05). Taken as a whole, our results do not support the idea that benthic primary production is the key limiting factor for zoobenthos production.

51

Instead, our results suggest that DOC affects zoobenthos and ultimately fish production largely by limiting the availability of warm and well-oxygenated habitat. Dissolved oxygen concentration was by far the strongest single predictor of depth-specific zoobenthos production in these lakes, followed by temperature. Zoobenthos production was generally very low at sites with hypoxic conditions (near-zero dissolved oxygen; Figure 1.2). Because hypoxic conditions persisted across most of the potentially available habitat in high-DOC lakes, these lakes had much lower whole-lake zoobenthos production than did low-DOC lakes. Hypoxia has been identified as a limiting factor for zoobenthos production in previous studies, usually in cases where hypoxia results from eutrophication (e.g. Jónasson 1984, Dermott 1988). Some authors have also suggested that a similar effect might occur as a result of high DOC concentrations (Rasmussen & Kalff 1987, Estlander et al. 2010). However, our study is the first clear demonstration that DOC-driven variability in dissolved oxygen concentrations can control zoobenthos production. One simple way to conceptualize this pattern is as a DOC-induced

“squeeze” of the habitat available to support appreciable zoobenthos production. A similar habitat squeeze has been suggested to drive DOC effects on zooplankton production (Kelly et al. 2014), and could also play a role in limiting fish production in high-DOC lakes (Coutant 1985,

Finstad et al. 2014).

The lack of a strong connection between zoobenthos production and benthic primary production suggests that additional basal resources must also be important for supporting zoobenthos. Evidence from our study and a considerable body of literature support this idea.

The zoobenthos is a diverse group consisting of specialist and generalist feeders of many guilds, including shredders, predators, collectors, and others (Cummins and Klug 1979, Merritt et al.

52

2008, Strayer 2009). It is likely that in light-limited sites with low benthic primary production but high zoobenthos production, the zoobenthos are feeding on other resources such as sedimented phytoplankton, bacteria, terrestrial particulate matter, or other invertebrates

(Hecky & Hesslein 1995, James et al. 2000, Solomon et al. 2008, Premke et al. 2010, Lau et al.

2014). This idea is supported by our stable isotope results, which showed that zoobenthos reliance on combined terrestrial and pelagic primary production was, in the majority of cases, higher than their reliance on benthic primary production (Appendix 1B). While benthic primary production clearly can be an important food resource for some zoobenthos in some locations

(Strayer and Likens 1986, Hillebrand & Kahlert 2001, Vadeboncoeur & Steinman 2002, Solomon et al. 2011, Devlin et al. 2013), our results, like those of Northington et al. (2010), demonstrate that it is not the strongest predictor of zoobenthos biomass and production.

It is important to note that our results do support the central idea of the conceptual model proposed by Karlsson et al. (2009) – that DOC concentrations can limit zoobenthos and fish production – as well as most of the mechanisms within that model. We saw that high-DOC lakes had more rapid light extinction, lower benthic primary production, and lower zoobenthos production. Furthermore, the data that we have available on fish populations in these lakes suggests that fish biomass is positively related to zoobenthos production (Figure 1.4b), although other mechanisms such as the oxygen-mediated habitat squeeze might also affect fish biomass.

This result is consistent with the patterns in fish production across DOC gradients observed by

Karlsson et al. (2009) and Finstad et al. (2014), and with the idea that zoobenthos are a major contributor to fish production in lakes (Vander Zanden & Vadeboncoeur 2002, Weidel et al.

2008). Our major contribution is to show that this DOC effect on food web productivity of

53 higher consumers is primarily a function of oxygen and habitat limitation rather than light limitation.

Unexplained variability in zoobenthos production

While the environmental factors that we considered explain a great deal of the within- and among-lake variation in zoobenthos production, there is appreciable variation not explained by our statistical models. Two aspects of this unexplained variation are particularly instructive to consider. First, we observed considerable variation in zoobenthos production at shallow depths that was not related to dissolved oxygen concentration, benthic primary production, or temperature. Factors such as sediment and macrophyte structure, wave action, and predation have been shown to have important influences on zoobenthos biomass in littoral zones, and likely played a role in our study as well (Rasmussen & Kalff 1987, Moss & Timms

1989, Tolonen et al. 2001). Lake depth and stratification strength also seem to play a role; the two lakes with the highest rates of zoobenthos production at shallow sites, Brown and Inkpot, were the only two in our study that never fully stratified, due to their relatively large size and shallow depth. Because periodic mixing of the water column reduces hypoxia and regenerates nutrients, shallow systems like these may have higher zoobenthos productivity than would be predicted on the basis of DOC alone (Finstad et al. 2014). Second, zoobenthos production varied appreciably (from 0 to 5.2 g m-2 y-1) even among hypoxic sites where mean dissolved oxygen concentrations were < 0.05 mg L-1. Previous work has shown that the duration of hypoxia can be an important control on zoobenthos; for instance, Jónasson (1984) found that chironomid growth and reproduction in a eutrophic lake was negatively related to the duration of summer stratification in which oxygen concentrations over sediments were depleted.

54

Logistical constraints prevented us from obtaining oxygen profiles for our ten lakes on more than the three sampling occasions, but if we regress site-specific zoobenthos production against the number of sampling dates per site that experienced hypoxic conditions (dissolved oxygen

2 <0.05 mg/L), we see a significant negative trend (F10,36 = 11.82, p = <0.001, R = 0.7, including lake as a blocking factor). Future studies might be able to better resolve this effect. It is interesting to note the parallels between high nutrient and high DOC concentrations, both of which are increasing globally (Bennett et al. 2001, Monteith et al. 2007), affect water clarity, and can increase hypolimnetic hypoxia and therefore decrease zoobenthos production.

Conclusions

Ecologists have long debated the role of resources and habitat in controlling consumer populations. In lake ecosystems, nutrient limitation of primary producers is often recognized as a major control on the production of higher trophic levels. More recently it has become clear that, in the relatively nutrient-poor systems that dominate northern landscapes, DOC is a major regulator of productivity from the base of the food web to its apex (Prairie 2008, Karlsson et al.

2009, Finstad et al. 2014, Kelly et al. 2014). It is important to note that this study focused on a restricted number of relatively small, shallow lakes, which although numerically dominant

(Verpoorter et al. 2014), may react differently compared to large, deep, highly mixed ones.

Nonetheless, our results confirm that DOC concentrations can limit productivity of zoobenthos and ultimately fishes, and demonstrate that this occurs mainly via oxygen-mediated habitat limitation rather than light-mediated resource limitation. In the face of considerable spatial variability and ongoing temporal change in DOC concentrations in northern lake ecosystems

55

(Monteith et al. 2007, Sobek et al. 2007), this mechanistic insight may improve understanding and management of lake food webs.

Acknowledgements

Funding was provided by the Natural Sciences and Engineering Research Council of

Canada. The staff of the University of Notre Dame Environmental Research Center facilitated our field work there. Technical assistance was provided by James Coloso, Alexandra Sumner,

Katherine Baglini, Sean Godwin, Patrick Kelly, Rachel Pilla, and Jacob Zwart. Comments from the anonymous reviewers improved the manuscript. Mention of specific products does not constitute endorsement by the U.S. Government. This is contribution number 1940 to the Great

Lakes Science Center.

References

Arvola, L. 1984. Vertical distribution of primary production and phytoplankton in two small lakes with different humus concentration in Southern Finland. Holarctic Ecology 7: 390- 398.

Ask, J., J. Karlsson, and M. Jansson. 2012. Net ecosystem production in clear-water and brown- water lakes. Global Biogeochemical Cycles 26: GB1017.

Babler, A. L., C. T. Solomon, and P. R. Schilke. 2008. Depth-specific patterns of benthic secondary production in an oligotrophic lake. Journal of the North American Benthological Society 27: 108-119.

Bennett, E. M., S. R. Carpenter, and N. F. Caraco. 2001. Human impact of erodible phosphorus and eutrophication: a global perspective. Bioscience 51: 227-234.

56

Carpenter, S. R., J. J. Cole, J. R. Hodgson, J. F. Kitchell, M. L. Pace, D. Bade, K. L. Cottingham, T. E. Essington, J. N. Houser and D. E. Schindler. 2001. Trophic cascades, nutrients, and lake productivity: whole-lake experiments. Ecological Monographs 71: 163-186.

Carroll, R. J., D. Ruppert, L. A. Stefanski, and C. M. Crainiceanu. 2006. Measurement error in non-linear models: A modern perspective. 2nd ed. Chapman & Hall, New York, pp 55-57.

Cheng, C. -L., Shalabh, and G. Garg. 2014. Coefficient of determination for multiple measurement error models. Journal of Multivariate Analysis 126: 137-152.

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

Cole, J. J., S. R. Carpenter, J. Kitchell, M. L. Pace, C. T. Solomon, and B. Weidel. 2011. Strong evidence for terrestrial support of zooplankton in small lakes based on stable isotopes of carbon, nitrogen, and hydrogen. Proceedings of the National Academy of Sciences 108: 1975-1980.

Coutant, C. C. 1985. Striped bass, temperature, and dissolved oxygen: a speculative hypothesis for environmental risk. Transactions of the American Fisheries Society 114: 31-61.

Cummins, K. W. and M. J. Klug. 1979. Feeding ecology of stream invertebrates. Annual Review of Ecology and Systematics 10: 147-172.

Dermott, R. M. 1988. Zoobenthic distribution and biomass in the Turkey Lakes. Canadian Journal of Fisheries and Aquatic Sciences 45: 107-114.

Devlin, S. P., M. J. Vander Zanden and Y. Vadeboncoeur. 2013. Depth-specific variation in carbon isotopes demonstrates resource partitioning among the littoral zoobenthos. Freshwater Biology 58: 2389-2400.

Estlander, S., L. Nurminen, M. Olin, M. Vinni, S. Immonen, M. Rask, J. Ruuhijärvi, J. Horppila and, H. Lehtonen. 2010. Diet shifts and food selection of perch Perca fluviatilis and roach

57

Rutilus rutilus in humic lakes of varying water colour. Journal of Fish Biology 77: 241- 256.

Finstad, A. G., I. P. Helland, O. Ugedal, T. Hesthagen, and D. O. Hessen. 2014. Unimodal response of fish yield to dissolved organic carbon. Ecology Letters 17: 36-43.

Fuller, W. A. 1987. Measurement error models. Wiley.

Godwin, S. C., S. E. Jones, B. C. Weidel, and C. T. Solomon. 2014. Dissolved organic carbon concentration controls benthic primary production: Results from in situ chambers in north-temperate lakes. Limnology and Oceanography 59: 2112-2120.

Hansson, L.-A. 1992. Factors regulating periphytic algal biomass. Limnology and Oceanography 37: 322-328.

Hecky, R. E. and R. H. Hesslein. 1995. Contributions of benthic algae to lake food webs as revealed by stable isotope analysis. Journal of the North American Benthological Society 14: 631-653.

Hillebrand, H. and M. Kahlert. 2001. Effect on grazing and nutrient supply on periphyton biomass and nutrient stoichiometry in habitats of different productivity. Limnology and Oceanography 46: 1881-1898.

Holsinger, J. R. 1972. The freshwater amphipod crustaceans (Gammaridae) of North America. Biota of Freshwater Ecosystems, Identification Manual 5. U.S. EPA, 89p.

James, M. R., I. Hawes, M. Weatherhead, C. Stanger, and M. Gibbs. 2000. Carbon flow in the littoral food web of an oligotrophic lake. Hydrobiologia 441: 93-106.

Jónasson, P. M. 1984. Oxygen demand and long term changes of profundal zoobenthos. Hydrobiologia 115: 121-126.

Jones, R. I. 1992. The influence of humic substances on lacustrine planktonic food chains. Hydrobiologia 229: 73-91.

58

Karlsson, J., P. Byström, J. Ask, P. Ask, L. Persson, and M. Jansson. 2009. Light limitation of nutrient-poor lake ecosystems. Nature 460: 506-509.

Kelly, P. T., C. T. Solomon, B. C. Weidel, and S. E. Jones. 2014. Terrestrial carbon is a resource, but not a subsidy, for lake zooplankton. Ecology 95: 1236-1242.

Lau, D. C. P., I. Sundh, T. Vrede, J. Pickova, and W. Goedkoop. 2014. Autochthonous resources are the main driver of consumer production in dystrophic boreal lakes. Ecology 95: 1506-1519.

Manly, B. F. J. 2007. Randomization, bootstrap and Monte Carlo methods in biology. 3rd edition. Chapman & Hall, New York. pp 41-80.

Merritt, R.W., K. W. Cummins, and M. B. Berg, editors. 2008. An introduction to the aquatic insects of North America. 4th edition. Kendall Hunt: Dubuque, Indiana.

Monteith, D. T., J. L. Stoddard, C. D. Evans, H. A. de Wit, M. Forsius, T. Høgasen, A. Wilander, B. L. Skjelkvåle, D. S. Jeffries, J. Vuorenmaa, B. Keller, J. Kopácek, and J. Vesely. 2007. Dissolved organic carbon trends resulting from changes in atmospheric deposition chemistry. Nature 450: 537-541.

Moss, B. and M. Timms. 1989. Predation, sediment stability and food availability as determinants of the benthic invertebrate fauna in two shallow lakes. Hydrobiologia 185: 249-257.

Northington, R. C., M. D. Keyse, S. R. Beaty, S. C. Whalen, E. R. Sokol and A. E. Hershey. 2010. Benthic secondary production in eight oligotrophic arctic Alaskan lakes. Journal of the North American Benthological Society 29 (2): 465-479.

Plante, C. and J. A. Downing. 1989. Production of freshwater invertebrate populations in lakes. Canadian Journal of Fisheries and Aquatic Sciences 46: 1489-1498.

Prairie, Y. T. 2008. Carbocentric limnology: looking back, looking forward. Canadian Journal of Fisheries and Aquatic Sciences 65: 543-548.

59

Premke, K., J. Karlsson, K. Steger, C. Gudasz, E. von Wachenfeldt, and L. J. Tranvik. 2010. Stable isotope analysis of benthic fauna and their food sources in boreal lakes. Journal of the North American Benthological Society 29: 1339-1348.

Rasmussen, J. B. and J. Kalff. 1987. Empirical models for zoobenthic biomass in lakes. Canadian Journal of Fisheries and Aquatic Sciences 44: 990-1001.

Rasmussen, J. B. 1988. Littoral zoobenthic biomass in lakes, and its relationship to physical, chemical, and trophic factors. Canadian Journal of Fisheries and Aquatic Sciences 45: 1436-1447.

Read, J. S. and K. C. Rose. 2013. Physical responses of small temperate lakes to variation in dissolved organic carbon concentrations. Limnology and Oceanography 58: 921-931.

Snucins, E. and J. Gunn. 2000. Interannual variation in the thermal structure of clear and colored lakes. Limnology and Oceanography 45: 1639-1646.

Sobek, S., L. J. Tranvik, Y. T. Prairie, P. Kortelainen, and J. J. Cole. 2007. Patterns and regulation of dissolved organic carbon: an analysis of 7,500 widely distributed lakes. Limnology and Oceanography 52: 1208-1219.

Solomon, C. T., S. R. Carpenter, J. J. Cole, and M. L. Pace. 2008. Support of benthic invertebrates by detrital resources and current autochthonous primary production: results from a whole-lake 13C addition. Freshwater Biology 53: 42-54.

Solomon, C. T., S. R. Carpenter, M. K. Clayton, J. J. Cole, J. J. Coloso, M. L. Pace, J. M. Vander Zanden, and B. C. Weidel. 2011. Terrestrial, benthic, and pelagic resource use in lakes: results from a three-isotope Bayesian mixing model. Ecology 92: 1115-1125.

Stern, E. M. 1990. An illustrated key to the freshwater mussels (Bivalvia: Unionacea) of Wisconsin. Reports of the Museum of Natural History, University of Wisconsin-Stevens Point, Wisconsin.

Strayer, D. and G. E. Likens. 1986. An energy budget for the zoobenthos of Mirror Lake, New Hampshire. Ecology 67: 303-313.

60

Strayer, D. L. 2009. Benthic invertebrate fauna, lakes and reservoirs. Pages 191-204 in Likens, E. (ed). "Encyclopedia of inland waters. Volume 2". Elsevier, Oxford.

Tolonen, K. T., H. Hämäläinen, I. J. Holopainen, and J. Karjalainen. 2001. Influences of habitat type and environmental variables on littoral macroinvertebrate communities in a large lake system. Archiv für Hydrobiologie 152: 39-67.

Vadeboncoeur, Y. and A. D. Steinman. 2002. Periphyton function in lake ecosystems. The Scientific World 2: 1449-1468.

Vadeboncoeur, Y., G. Peterson, M. J. Vander Zanden, and J. Kalff. 2008. Benthic algal production across lake size gradients: Interactions among morphometry, nutrients, and light. Ecology 89: 2542-2552.

Vander Zanden, M. J. and Y. Vadeboncoeur. 2002. Fishes as integrators of benthic and pelagic food webs in lakes. Ecology 83: 2152-2161.

Vanni, M. J. 2002. Nutrient cycling by animals in freshwater ecosystems. Annual Review of Ecology and Systematics 33: 341-370.

Verpoorter, C., T. Kutser, D. A. Seekell, and L. J. Tranvik. 2014. A global inventory of lakes based on high resolution satellite imagery. Geophysical Research Letters 41: 6396-6402.

Weidel, B., S. Carpenter, J. Cole, J. Hodgson, J. Kitchell, M. Pace, and C. Solomon. 2008. Carbon sources supporting fish growth in a north temperate lake. Aquatic Sciences 70: 446-458.

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

Wissel, B., W. J. Boeing, and C. W. Ramcharan. 2003. Effects of water color on predation regimes and zooplankton assemblages in freshwater lakes. Limnology and Oceanography 48: 1965-1976.

61

Figure captions

Figure 1.1. Depth profiles of habitat and resource characteristics for the clearest (grey circles; Crampton Lake, DOC=5.26 mg/L) and darkest (black circles; Reddington Lake, DOC=23.0 mg/L) lakes in this study. BPP represents benthic primary production.

Figure 1.2. Depth-specific zoobenthos production plotted against mean depth-specific (a) dissolved oxygen concentration, (b) water temperature, and (c) benthic primary production. Each point represents a depth-lake combination; error bars are ± 1 SE of production estimates across four replicate sites per depth. The best model describing variation in zoobenthos production included dissolved oxygen and a lake blocking term, but not temperature or benthic primary production.

Figure 1.3. Zoobenthos production across depths in lakes where DOC concentrations are (a) low, (b) intermediate, and (c) high. The average position of the metalimnion (the depth zone where temperature is changing at >1°C m-1) in each set of lakes is indicated by the horizontal grey bands.

Figure 1.4. Relationships between zoobenthos production and, (a) DOC concentration 2 for the ten survey lakes (F1,8 = 7.89, p = 0.02, R = 0.50, log-log regression), (b) Catch-per-unit effort of zoobenthivorous fish (CPUE, in biomass units) for the 6 survey lakes where fish data 2 were available (F1,4 = 18.46, p = 0.01, R = 0.82).

Figure 1.5. Benthic reliance (proportion of biomass derived from benthic primary production) of zoobenthos in ten lakes across a gradient of DOC concentrations. Box plots show the 2.5th, 25th, 50th, 75th, and 97.5th percentiles of the posterior probability distributions for benthic reliance, based on a Bayesian stable isotope mixing model. The prior probability distribution is shown at the right edge of the figure. Zoobenthos for this analysis were collected from 1 m depth in each lake.

62

Tables and Figures

Table 1.1. Summary of lake bathymetry and water chemistry for the ten survey lakes during the study period. DOC is dissolved organic carbon. Average light climate is the average percentage of surface light available in each lake, calculated using weighted depth bands (see whole-lake zoobenthos production methods). Metalimnion depth was characterised as the shallowest depth at which the change in temperature exceeded 1 °C m-1.

Average Average Average benthic Mean Max light Surface Average lake Metalimnion lake Total Area DOC Total nitrogen primary Lake depth depth climate temperature temperature depth dissolved phosphorus (ha) (mg/L) (μg/L) production (m) (m) (% surface (°C) (°C) (m) oxygen (μg/L) (mg C m¯² light) (mg/L) d¯¹)

Crampton 25.81 5 15.2 5.3 14.2 21.6 16.8 3.5 7.1 11.4 343.2 286.4 Bay 67.3 4.3 13.7 6.1 17.7 21.4 17.3 3.7 6.2 12.4 406.1 184.6 Raspberry 4.63 3.1 6 7.3 9.1 22.2 16.7 1.8 4.2 25.9 552.0 71.7 Brown 32.57 2.8 5.5 7.4 7.3 21.2 19.7 2.5 4.3 33.7 623.8 190.5 Long 7.87 3.9 14 7.6 10.6 21.9 15.1 1.8 5.7 21.0 441.5 103 Bergner 17.85 3.6 12 8.1 13.7 21.3 17.7 2.5 5 18.8 487.2 233.9 Inkpot 6.61 2.9 5.2 9.1 9.1 21.7 18.2 2.3 4.9 28.3 540.7 79 Morris 5.93 2.6 6.7 15.7 6.4 21.4 15.3 1.2 3.5 29.5 772.7 31.8 Hummingbird 0.76 3.7 7 19.9 0.7 20.9 9.6 0.7 1.4 32.5 880.5 9.3 Reddington 1.24 2.5 16 23.0 5.2 21.4 10.9 0.8 2.5 33.9 798.3 13.6

63

Table 1.2. Summary of bias-corrected linear measurement error models relating depth- specific zoobenthos production in ten lakes to mean depth-specific dissolved oxygen concentration (DO, mg/L), water temperature (Temp, °C), and benthic primary production (BPP, mg C m-2 d-1). All predictor variables were log(x+1) transformed and converted to Z-scores. Models are sorted by ΔAICc, with the best model in the top row. Estimates of the coefficients included in the model (with bootstrapped 95% confidence intervals) are given in the first three columns; each model also included lake as a blocking term. Bold text indicates coefficients for which the 95% CI does not include zero. R2 estimates are not valid in the presence of measurement error and are provided only as rough aids to interpretation.

DO Temp BPP R² ΔAICc 3.4 (2.5, 4.1) - - 0.76 0 - 3.0 (1.9, 4.0) - 0.66 11 - - 3.0 (1.6, 3.8) 0.65 18 4.2 (2.3, 5.9) -0.9 (-2.8, 1.0) - 0.77 52 2.9 (1.6, 4.4) - 0.7 (-1.1, 2.0) 0.77 93 - 1.6 (0.3, 3.5) 1.8 (-0.3, 3.0) 0.69 105 3.7 (1.8, 5.8) -1.1 (-3.0, 0.9) 0.8 (-1.0, 2.1) 0.77 143

64

Figure 1.1

65

Figure 1.2

66

Figure 1.3

67

Figure 1.4

68

Figure 1.5

69

Appendix 1A

Table 1A. Length-mass equations used to calculate dry mass from head/shell width or body length of zoobenthos.

Subclass Order Family Genus Reference Bivalvia Veneroida Sphaeriidae A Bivalvia Veneroida Corbiculidae A Bivalvia Veneroida Unionidae A Hirudinea B Oligochaeta C Amphipoda Gammaridae Crangonyx/Gammarus A Coleoptera Chrysomelidae A Coleoptera Dytiscidae Neoporus A Coleoptera Gyrinidae Dineutus A Coleoptera Haliplidae Peltodytes A Diptera Ceratopogonidae A Diptera Chironomidae A Diptera Tabanidae Chrysops A Diptera Tipulidae Erioptera A Diptera Tipulidae Tipula A Ephemeroptera Caenidae Caenis A Ephemeroptera Ephemeridae Hexagenia A Gastropoda D Hemiptera A Lepidoptera Crambidae A Megaloptera Sialidae Sialis A Odonata Aeshnidae Aeshna A Odonata Coenagrionidae Enallagma A Odonata Corduliidae Cordulia A Odonata Corduliidae Epitheca A Odonata Corduliidae Somatochlora A

70

Odonata Gomphidae Gomphus A Odonata Libellulidae Celithemis A Odonata Libellulidae Erythemis A Odonata Libellulidae Ladona A Odonata Libellulidae Leucorrhinia A Odonata Libellulidae Pachydiplax A Odonata Libellulidae Tramea A Trichoptera Dipseudopsidae Phylocentropus A Trichoptera Hydroptilidae Oxyethira A Trichoptera Leptoceridae Ceraclea A Trichoptera Leptoceridae Mystacides A Trichoptera Leptoceridae Oecetis A Trichoptera Leptoceridae Triaenodes A Trichoptera Limnephilidae Limnephilus A Trichoptera Limnephilidae Platycentropus A Trichoptera Limnephilidae Pycnopsyche A Trichoptera Molannidae Molanna A Trichoptera Philopotamidae A Trichoptera Phryganeidae Agrypnia A Trichoptera Phryganeidae Banksiola A Trichoptera Phryganeidae Fabria inornata A Trichoptera Phryganeidae Phryganea A Trichoptera Polycentropodidae Nyctiophylax A Trichoptera Polycentropodidae Polycentropus A

A - Benke, A.C., Huryn, A.D., Smock, L.A. & Wallace, J.B. (1999). Length-mass relationships for freshwater macroinvertebrates in North America with particular reference to the Southeastern United States. J. N. Am. Benthol. Soc., 18, 308-343.

B - Edwards, F.K., Lauridsen, R.B., Armand, L., Vincent, H.M. & Jones, I.J. (2009). The relationship between length, mass and preservation time for three species of freshwater leeches (Hirudinea). Fund. Appl. Limnol., 173, 321-327.

71

C - Woodward, G. & Hildrew, A.G. (2002). Body-size determinates of niche overlap and intraguild predation within a complex food web. J. Anim. Ecol., 71, 1063-1074.

D - Baumgärtner, D. & Rothhaupt, K. (2003). Predictive length-dry mass regressions for freshwater invertebrates in a pre-alpine lake littoral. Int. Rev. Hydrobiol., 88, 453-463.

Appendix 1B

Table 1B. Benthic, pelagic, and terrestrial resource reliance by zoobenthos in ten lakes of varying dissolved organic carbon (DOC) concentration. We used a Bayesian stable isotope mixing model to estimate zoobenthos reliance on each of those three resources in each lake, using C, N, and H stable isotope ratio data. The data in the tables are the percentiles of the posterior probability distributions for resource reliance. The last row in each table gives the percentiles of the prior probability distribution. Zoobenthos included in this analysis were Trichoptera, Odonata and Chironomidae collected from 1 m depth.

Benthic Reliance

Lake DOC (mg/L) 2.5% 25% 50% 75% 97.5% Crampton 5.26 0.15 0.3 0.38 0.45 0.63 Bay 6.15 0.04 0.36 0.48 0.6 0.88 Raspberry 7.31 0.01 0.49 0.7 0.87 0.99 Brown 7.4 0.003 0.08 0.13 0.2 0.38 Long 7.62 0.01 0.15 0.23 0.32 0.55 Bergner 8.06 0.19 0.38 0.44 0.5 0.62 Inkpot 9.06 0.001 0.01 0.04 0.09 0.3 Morris 15.71 0.001 0.04 0.13 0.3 0.62 Hummingbird 19.93 0.001 0.02 0.11 0.41 0.97 Reddington 23.05 0.001 0.04 0.15 0.3 0.7 Prior - 0.001 0.03 0.17 0.59 0.98

72

Pelagic Reliance

Lake DOC (mg/L) 2.5% 25% 50% 75% 97.5% Crampton 5.26 0.002 0.06 0.17 0.27 0.43 Bay 6.15 0.001 0.05 0.2 0.35 0.62 Raspberry 7.31 0.001 0.01 0.06 0.18 0.41 Brown 7.4 0.05 0.32 0.46 0.58 0.81 Long 7.62 0.004 0.07 0.15 0.22 0.36 Bergner 8.06 0.001 0.03 0.13 0.28 0.55 Inkpot 9.06 0.03 0.2 0.29 0.36 0.49 Morris 15.71 0.27 0.38 0.43 0.48 0.56 Hummingbird 19.93 0.002 0.06 0.2 0.34 0.55 Reddington 23.05 0.03 0.22 0.29 0.34 0.43 Prior - 0.001 0.03 0.19 0.67 0.99

Terrestrial Reliance

Lake DOC (mg/L) 2.5% 25% 50% 75% 97.5% Crampton 5.26 0.19 0.37 0.46 0.53 0.66 Bay 6.15 0.01 0.19 0.31 0.41 0.57 Raspberry 7.31 0.002 0.04 0.19 0.36 0.78 Brown 7.4 0.03 0.28 0.41 0.53 0.76 Long 7.62 0.29 0.5 0.61 0.71 0.89 Bergner 8.06 0.06 0.3 0.41 0.5 0.62 Inkpot 9.06 0.39 0.57 0.65 0.73 0.91 Morris 15.71 0.03 0.28 0.42 0.51 0.64 Hummingbird 19.93 0.004 0.34 0.58 0.73 0.96 Reddington 23.05 0.12 0.42 0.54 0.64 0.8 prior - 0.001 0.02 0.18 0.64 0.98

73

Connecting Statement

In chapter 1, I showed that zoobenthos productivity decreases over a DOC gradient, due to a reduction of suitable habitat in DOC-rich lakes. In addition, Kelly et al. (2014) showed a similar pattern for zooplankton. Zoobenthos and zooplankton are primary food sources for juvenile fish in lake ecosystems, and the reduction in productivity of these organisms in high DOC lakes may contribute to the concurrent reduction in fish productivity observed by Karlsson et al. (2009). Potential mechanisms contributing to this may include both the lack of resources available, and the difficulty in obtaining those resources under the low light conditions associated with high levels of terrestrial DOC.

In chapter 2, I describe how benthic feeding success is affected by DOC using a series of mesocosm experiments. Studies of how DOC affects fish feeding behaviour are rare, and of those available, none have focused on benthic interactions. I compliment this with fish diet data collected from five lakes over a DOC gradient. The aim of this chapter is to determine if benthic feeding efficiency is negatively affected by DOC concentration, which may help explain negative relationships between DOC and fish productivity.

74

Chapter 2

DISSOLVED ORGANIC CARBON EFFECTS ON FISH FEEDING EFFICIENCY.

Nicola Craig1*, Stuart E. Jones2, Brian C. Weidel3, and Christopher T. Solomon1

1 Dept. of Natural Resource Sciences, McGill University, Ste. Anne de Bellevue, QC, Canada.

2 Dept. of Biological Sciences, University of Notre Dame, Notre Dame, IN, USA.

3 U.S. Geological Survey, Great Lakes Science Center, Lake Ontario Biological Station, Oswego, NY, USA.

Status: In preparation

Abstract

Dissolved organic carbon (DOC) may affect the feeding behavior of visual predators through its negative effects on both secondary production, and light penetration. Few studies have focused on the role of DOC in feeding interactions, and those that do focus exclusively on pelagic linkages, ignoring the important role of the benthos in fueling fish productivity. Benthic feeding may be less vision orientated and so fish in high DOC environments may rely more on zoobenthic prey. In this study, we test the hypothesis that benthic feeding is unaffected by DOC-induced reductions in light climate through the use of lab-based feeding experiments where bluegill benthic feeding efficiency is measured over a DOC gradient. We compliment these experiments with a diet survey where we observe the variation in stomach fullness, prey size, and benthic selectivity in juvenile bluegill and yellow perch across five lakes of varying DOC concentration. We found that bluegill foraging success did not change along the experimental DOC gradient, and was fairly high regardless of light climate, suggesting that this species can utilize sensory mechanisms such as the lateral line, or olfactory system to detect prey under

75 sub-optimal visual conditions. The field study suggested that feeding behavior was similar across the DOC gradient, with high benthic selectivity all round. These results suggest that juvenile fish are able to feed effectively in high DOC environments, despite a reduction in food availability and light climate. As lakes with high DOC concentrations have been shown to have lower fish productivity, this result is somewhat unexpected. There is potential that adult feeding behavior may be more affected by lower secondary production in high DOC lakes and further studies would be needed to resolve this disparity.

Introduction

The study of feeding behavior is an important field as foraging choices and success are crucial to the fitness of organisms, and can also affect whole ecosystems. For example, in aquatic ecosystems, fish feeding behavior can alter the physical states of lakes through trophic cascades, or at a smaller scale, transfer energy between habitats (Carpenter et al. 1985, Vander Zanden and Vadeboncoeur 2002). Fish growth and productivity, and thus the success of fisheries, depends on food intake, and this depends on both the availability of resources as well as the ability of fish to utilize them (Jobling 1994).

Dissolved Organic Carbon (DOC) concentration is an environmental factor that will likely influence fish feeding behavior due to its associated effects on the availability of secondary production and light (Stasko et al. 2012). High levels of terrestrial DOC in lakes stain the water a characteristic brown color and reduce light penetration (Jones 1992), which may limit the ability of visual predators to obtain prey. In addition, recent studies have shown that secondary production of zooplankton and zoobenthos is negatively affected by higher DOC concentrations, resulting in lower prey availability for fish in these lakes (Kelly et al. 2014, Craig et al 2015). Fish productivity has been shown to decline with increasing DOC in several gradient studies (Karlsson et al. 2009, Finstad et al. 2014, Karlsson et al. 2015), and it is possible that this decline in productivity could be due to a reduction in food uptake. As DOC concentrations can vary spatially, and have been rising temporally over the past few decades (Monteith et al. 2007, Solomon et al. 2015), it is important to understand how this variable relates to fish feeding behavior in order to better manage aquatic resources through environmental changes.

76

There are numerous studies focusing on the effects of light and/or turbidity on fish feeding efficiency (e.g. Vinyard & O'Brien 1976, Diehl 1988, Aksnes & Giske 1993, Utne-Palm 2002, De Robertis et al. 2003), however, much less attention has been paid to the effects of DOC. There is an important distinction because DOC absorbs incoming light, whereas light entering turbid waters is scattered, leading to differing optical properties (Ranåker et al. 2012). Due to these differences, more contrast may be obtained between prey and environment in high DOC environments compared to turbid environments, and so feeding behavior may differ accordingly (Ranåker et al. 2012). Of the limited available studies focusing on DOC effects on feeding behavior, all have been focused on pelagic interactions (Estlander et al. 2012, Ranåker et al. 2012, Jönsson et al .2013, Weidel et al. in review). While these studies provide valuable insight on how DOC may affect feeding behavior, they ignore benthic feeding interactions which may be of great importance in some species (Vadeboncoeur et al. 2002, Weidel et al. 2008). As benthic resources sustain a large proportion of fish productivity (Vander Zanden & Vadeboncoeur 2002), it is important to know how DOC may affect benthivore - prey interactions.

Benthic feeding may be less reliant on vision compared to pelagic feeding (Rowe et al. 2003), and if so, may be less affected by lack of light in dark lakes. If this is the case, we may find that fish depend more on benthic resources in darker lakes and may preferentially select for this resource (Estlander et al. 2012). In addition, while many studies have shown that fish can select for larger zooplankton prey in clear waters where they can be visually discerning (e.g. Brooks & Dodson 1965, Hall et al. 1970), the same effect may not be observed with benthic interactions if they are based on non-visual detection methods. These variations in resource selectivity may alter the flow of energy within the food web (Covich et al. 1999, Vander Zanden & Vadeboncoeur 2002), and depending on the comparative quality, size, and availability of benthic vs. pelagic resources in darker lakes, may negatively affect productivity in predator populations.

In this study, we conducted a lab-based feeding experiment to test if benthic feeding ability is affected by DOC concentration, with the hypothesis that DOC would not have an effect as benthic feeding is less reliant on vision. We compliment this with a field survey, observing

77 the stomach fullness and benthic component of fish diets across a DOC gradient. We expect that stomach fullness will decline across the gradient as food availability is reduced, and also that reliance on benthic resources will increase across this gradient as fish depend less on visual, pelagic predation.

Methods

Benthic feeding experiments

Experimental Design

The benthic feeding experiments were conducted during the summer of 2014, at the lab facilities of the University of Notre Dame Environmental Research Center (UNDERC) on the Wisconsin-Michigan border, USA. We conducted trials in nine 262 L circular, metal cattle tanks. Each tank had a sandy substrate on the bottom (~5mm depth) and was filled with filtered lake water from nearby Tenderfoot Lake (except for the clearest treatment, which had to be diluted with well water to achieve the appropriate water color). Nine DOC treatments over a gradient from approximately 4-28 mg/L were created using Super Hume (United Agricultural Services of America, Lake Panasoffkee, Florida, USA), a commercially available humic concentrate which approximates natural terrestrial DOC (Lennon et al. 2013). A DOC gradient of approximately 4- 28 mg/L was used as this is similar to the range of DOC concentrations found in surrounding lakes and examined in spatial surveys for secondary consumer production (Kelly et al. 2014, Craig et al. 2015). Each time the experiment was run, the nine DOC concentration treatments were randomly assigned to the nine tanks, and we changed the water and reassigned DOC concentrations between each individual trial. As we could not measure DOC directly at the lab facilities, we report the results in terms of water color which we measured through a spectrophotometer, recording absorbance using a 100mm cuvette at 440nm. Super Hume based water color is directly related to DOC concentration (Lennon et al. 2013). Before each trial, we measured water temperature and dissolved oxygen saturation using a handheld polarographic sensor (YSI Pro 20, Yellow Spring Instruments, USA), to check for similar conditions between treatments (Appendix 2A). We also measured the surface light and light

78 levels at the bottom of each tank using an underwater quantum PAR sensor and light meter (LICOR LI-192SA and LI-250A, LICOR, USA).

We used juvenile bluegill (Lepomis macrochirus, ~60-80mm) as the predator in these trials, and Chironomus spp. larvae as the prey, all collected from nearby lakes. In order to determine if the DOC concentration that a fish came from had an effect on feeding ability, we conducted separate trials using fish from a relatively clear lake (Deadwood Lake, ~ 8 mg/L DOC) and a much darker lake (Hummingbird, ~ 20 mg/L DOC). The chironomid larvae (~12mm) were collected using an Ekman dredge from Bay Lake. Fish were collected using minnow traps set overnight (from approximately 4pm to 9am) and kept in buckets of water from their original lake with a bubbler until water temperatures in the bucket reached the same as those in the tanks (~1-2 hours). The fish were then transferred to a holding tank of intermediate DOC concentration (i.e. the middle of the gradient, ~16 mg/L) and fed chironomid larvae to acclimate to the experimental conditions. After feeding, fish were starved for 2 days so that they would be hungry, and have empty stomachs for the trials.

We conducted five replicate trials at each treatment level for the clear lake fish and four replicates per treatment for the dark lake fish. For each trial, we randomly placed ten chironomid larvae into the experimental tanks and allowed them 15 minutes to burrow into the substrate. At the start of each trial, a bluegill was removed from the acclimation tank, placed into one of the treatments and left for 15 minutes. The fish was then removed, euthanised using an overdose of MS-222, and weighed and measured. The stomach was then removed and examined, and the number of larvae within was recorded. Several fish showed symptoms of some sort of infection, with a slight swelling occurring around the caudal peduncle. These fish never consumed more than one prey item and so were excluded from the analysis (a total of 1 from Hummingbird Lake, and 19 from Deadwood Lake, resulting in 35 and 26 useable fish for each lake respectively).

Data Analysis

Linear regressions were used to test if water color (a proxy for DOC) had an effect on the number of prey consumed. Within each treatment, water color varied slightly from trial to

79 trial, and so we treated each individual trial separately, instead of grouping into treatments. We tested the residuals of these regressions against fish length, to ensure that this did not have a confounding effect on the results.

Field based diet survey

Sample collection

Diets were collected from juvenile bluegill and yellow perch (Perca flavescens) in five UNDERC lakes of varying DOC concentrations over the period on one week in July 2012 (Table 2.1). In each lake, fish were sampled at least once in the morning (8-9am) and afternoon (4- 5pm) using seine nets, or minnow traps set for one hour to ensure minimal digestion of stomach contents. Diets were collected by gastric lavage from fish larger than 60mm using a small syringe and tubing, and for smaller fish, stomachs were dissected. We euthanized and conducted spot checks on ~5% of fish that were sampled through gastric lavage in order to ensure that the results were comparable, i.e. that we obtained all the stomach content through the lavage method. Diets were stored in 70% ethanol for analysis.

Diet processing

Stomach contents were examined under a dissecting microscope and prey items were enumerated and measured. Benthic invertebrates and zooplankton were generally identified to order, but sometimes family-level identifications were possible (e.g. Chironomidae, Chaoboridae). There were also occasional terrestrial invertebrates present in the diets, but these were insignificant in comparison to aquatic prey (~1.4 % of diet biomass). We randomly selected eight prey items per category for length measurements, measuring all items if there were less than eight individuals, and recorded average length of each prey item per diet.

Data Analysis

Biomass of each prey category was estimated using published length - dry mass equations (Appendix 2B). Average prey dry mass was multiplied by the count data to obtain a total mass per prey item per fish, and these were added to obtain total diet mass per fish. The average prey length per fish was obtained by multiplying average length of each prey item per

80 stomach by the count of those items in the stomach, then the sum of all lengths was divided by n items in the diet. The proportion of diet mass consisting of benthic items was calculated by dividing total stomach dry mass by mass of benthic items. Due to the low occurrence of terrestrial items in the diets, whatever was not benthic tended to be pelagic biomass, i.e. zooplankton. Selectivity for benthic prey was calculated by the following from Ivlev (1961):

E = (ri - pi)/(ri + pi) where E is the index of benthic selectivity, ri is the relative abundance of benthic prey in the stomach and pi is the relative abundance of benthic prey in the environment. Data on the proportion of benthic (vs. pelagic) biomass in the environment was obtained from Craig et al. (2015) and unpublished data collected by Patrick Kelly (see Kelly et al. (2014) for sampling details).

To test for effects of DOC on the diet variables (total stomach biomass, mean prey size, benthic proportion of diet, and benthic selectivity), we used linear mixed effects models including a fish length term and a random lake effect. This separated the effect of fish length, which had varied patterns with the diet variables for each lake and species, from DOC. These analyses were conducted using the lme4 package in R (R Core Team 2014, Bates et al. 2014), and data were transformed and centered before analysis to meet assumptions of normality.

Results

Benthic feeding trials

We tested a total of 35 bluegill between 67 and 90 mm from the darker, Hummingbird Lake, and 26 bluegill between 66 and 80 mm from the clearer Deadwood Lake. We found no significant effect of DOC related water color on the benthic feeding ability of bluegill from either lake (Fig. 2.1). We tested the residuals of the regressions against fish length, to ensure that this did not have a confounding effect on the ability of fish to consume the chironomids 2 and found no significant effects (Hummingbird: F1,33 = 2.67, R = 0.07, p = 0.1; Deadwood: F1,24 = 0.56, R2 = 0.02, p = 0.5).

81

Field based diet survey

We collected 270 diets from bluegill and yellow perch across five lakes over a fairly even size range (Table 2.1, Appendix 2C). The majority of diet items were either of benthic or pelagic origin, with only 2.2% in terms of dry mass coming from terrestrial or unidentified sources.

Generally, we found that DOC had little effect on any of the diet variables measured, namely stomach biomass (a proxy for fullness), prey size (length), benthic proportion of stomach biomass and selectivity for benthic prey (Table 2.2, Figure 2.2). However, there was one exception where there was a significant positive effect of DOC on bluegill stomach fullness. Fish size generally had a positive effect on the diet parameters, indicating that as fish grow, they can consume a higher biomass of prey, larger prey items, and more benthic prey in general (Table 2.2).

Discussion

Both the feeding experiments and field survey suggest that generally DOC has little effect on benthic feeding efficiency in bluegill. The feeding trail results in particular indicate that bluegill can feed on zoobenthos regardless of light, or lake of origin, strongly implying that they use a non-visual method of detecting prey such as the lateral line, or olfactory or acoustic cues. Feeding in the absence of light has been shown for several fish species thought to be primarily visual predators, such as rainbow trout, yellow perch, and bluegill (Diehl 1988, Enger et al. 1989, Rowe et al. 2003), however, it is primarily benthic feeding that occurs in the darkness, and usually at a reduced rate. It could be the case that a benthic substrate provides a narrower search area for fish using the lateral line system to detect prey, which is generally only possible over short distances, e.g. a body length (Enger et al. 1989, Webb et al. 2008). If this is the case, then we would expect fish to become more reliant on benthic resources in lakes with higher DOC concentrations.

Both bluegill and yellow perch strongly relied on benthic resources across the DOC gradient in the diet survey, but the magnitude of this reliance did not vary, suggesting that, feeding behavior is not strongly affected by DOC in these lakes. This result is contrary to our expectation that pelagic feeding would be more difficult in darker waters, but is not altogether

82 surprising considering the limited scope of this survey, and the mixed results of similar previous studies. For example, Ranåker et al. (2012) and Jönsson et al. (2013) found that that pike feeding success on roach prey had a unimodal relationship to DOC with high success rates in both clear, and very dark waters. Although predator reaction distances were shorter in darker waters, prey were less likely to escape once spotted, leading to high capture success. While Estlander et al. (2010, 2012) found that European perch consumed fewer zooplankton in dark waters in both lab and field experiments, Weidel et al. (in review) found that the relationship between DOC level and planktivore feeding success varied by species, with largemouth bass and bluegill feeding less in darker treatments, but fathead minnows consuming more. These mixed results suggest that the effects of DOC on predator-prey interactions are complex and may vary from species to species.

The fact that feeding behavior did not vary along the DOC gradient, and that bluegill stomach mass actually increased with DOC concentration is surprising when we consider that both benthic and pelagic secondary production tend to be negatively affected by DOC (Kelly et al. 2014, Craig et al. 2015). Fish productivity has been shown to decline with higher DOC levels and the current theory is that this is primarily due to resource limitation (Karlsson et al. 2009, Finstad et al. 2014, Craig et al. 2015). One explanation for this disparity could be related to fish size and the distribution of prey within lakes across a DOC gradient. Due to the limited number of fish available in Hummingbird Lake, and lack of other suitable high DOC lakes on the UNDERC property, we could only obtain sufficient numbers of juveniles (as opposed to larger fish) for both the feeding experiment and diet survey, and so these were the focus of our study. The juveniles of many fish species tend to reside in vegetated littoral zones (Werner et al. 1977, Mittelbach 1981) where zooplankton and zoobenthos abundances are relatively high even across a DOC gradient (Kelly et al. 2014, Craig et al. 2015). As juvenile fish are smaller, they need fewer resources than larger fish in order to fulfill their energy requirements, and our results suggest that they seem to be able to obtain sufficient resources even in the darkest lake. Two recent spatial surveys suggest that initial growth rates of several fish species, including bluegill and yellow perch are unaffected by DOC concentration (Benoit et al. in review, Craig et al. in review), again suggesting that food availability is sufficient at early ages in these lakes.

83

However, Craig et al. (in review) showed that later growth of bluegill was negatively affected by DOC, and suggested that this may be due to the fish outgrowing the prey availability in dark lakes with comparatively reduced littoral zones. To confirm that fish become food limited at larger sizes in high DOC lakes, a more detailed diet survey, perhaps complimented by reciprocal transplant experiments, focusing on adult fish across a DOC gradient would be required.

A second potential explanation for the lack of relationships between overall feeding efficiency and DOC, which has been touched on in Weidel et al. (in review), is the 'sheltering' effect of DOC. Juvenile bluegill and yellow perch are vulnerable to predation until they become larger than the gape limit of their predators. However, high DOC environments may reduce the chances of being seen, unless the predator is relatively close (Ranåker et al. 2012, Jönsson et al. 2013). This extra cover may allow prey fish to spend less time scanning for potential predators, and more time foraging for their own prey. This may make up for the potentially lower food availability in darker lakes, if the fish can spend more time searching for these limited resources.

An issue with this study, as well as many others focusing on fish feeding behavior is that diet content and foraging success can vary greatly between individuals (Magurran 1986, Wolf and Weissing 2012), and only supplies a snap shot of feeding behavior at the time (Vander Zanden et al. 1997). Thus an extremely large number of samples are needed in order to discern patterns from the resultant noisy data. Future additions to the data presented in this paper from other lakes may help us strengthen or refute the patterns already observed in terms of benthic feeding and foraging success, and answer important questions regarding specific diet composition and how fish bioenergetics may shift over a DOC gradient. Understanding how DOC concentrations affect fish resource uptake is crucial in order to determine the future effects of DOC increases on aquatic food webs and fishery productivity.

Acknowledgements

Funding was provided by the Natural Sciences and Engineering Research Council of Canada. The staff of the University of Notre Dame Environmental Research Center facilitated

84 our field work there. Technical assistance was provided by Jacob Lerner, Karling Roberts, Alexandra Sumner, and Jacob Zwart. Mention of specific products does not constitute endorsement by the U.S. Government.

References

Aksnes, D. L. and J. Giske. 1993. A theoretical model of aquatic visual feeding. Ecological Modelling 67: 233-250.

Bates, D., M. Maechler, B. Bolker, and S. Walker. 2015. lme4: linear mixed-effects models using Eigen and S4. R package version 1.1-8, http://CRAN.R-project.org/package=lme4.

Benoît, P-O., B. E. Beisner, and C. T. Solomon. In review. Growth rate and abundance of common fishes is negatively related to dissolved organic carbon concentration in lakes.

Brooks, J. L., and S. I. Dodson. 1965. Predation, body size, and composition of plankton. Science 150: 28-35.

Carpenter, S. R., J. F. Kitchell, and J. R. Hodgson. 1985. Cascading trophic interactions and lake productivity. Bioscience 35: 634-639.

Covich, A. P., M. A. Palmer, and T. A. Crowl. 1999. The role of benthic invertebrate species in freshwater ecosystems. Zoobenthic species influence energy flows and nutrient cycling. Bioscience 49: 119-127.

Craig, N. C., S. E. Jones, B. C. Weidel, and C. T. Solomon. 2015. Habitat, not resource availability, limits consumer production in lake ecosystems. Limnology and Oceanography 60: 2079- 2089.

Craig, N. C., S. E. Jones, B. C. Weidel, and C. T. Solomon. In review. Life history constraints explain negative relationship between fish productivity and dissolved organic carbon in lakes.

85 de Robertis, A., C. H. Ryer, A. Veloza, and R. D. Brodeur. 2003. Differential effects of turbidity on prey consumption of piscivorous and planktivorous fish. Canadian Journal of Fisheries and Aquatic Sciences 60: 1517-1526.

Diehl, S. 1988. Foraging efficiency of three freshwater fishes: Effects of structural complexity and light. Oikos 53: 207-214.

Enger, P. S., A. J. Kalmijn, and O. Sand. 1989. Behavioral investigations on the functions of the lateral line and inner ear in predation. Pages 575-587 in Coombs, S., P. Görner, and H. Münz (eds). "The Mechanosensory Lateral Line: Neurobiology and Evolution". Springer- Verlag, New York.

Estlander, S., L. Nurminen, M. Olin, M. Vinni, S. Immonen, M. Rask, J. Ruuhijärvi, J. Horppila, and H. Lehtonen. 2010. Diet shifts and food selection of perch Perca fluviatilis and roach Rutilus rutilus in humic lakes of varying water colour. Journal of Fish Biology 77: 241- 256.

Estlander, S., J. Horppila, M. Olin, M. Vinni, H. Lehtonen, M. Rask, and L. Nurminen. 2012. Troubled by the humics - effects of water colour and interspecific competition on the feeding efficiency of planktivorous perch. Boreal Environment Research 17: 305-312.

Finstad, A. G., I. P. Helland, O. Ugedal, T. Hesthagen, and D. O. Hessen. 2014. Unimodal response of fish yield to dissolved organic carbon. Ecology Letters 17: 36-43.

Hall, D. J., W. E. Cooper, and E. E. Werner. 1970. An experimental approach to the production dynamics and structure of freshwater animal communities. Limnology and Oceanography 15: 839-928.

Ivlev, V. W. 1961. Experimental ecology of the feeding of fishes. Yale University Press, New Haven.

Jobling, M. 1994. Fish bioenergetics. Chapman & Hall, London.

Jones, R. I. 1992. The influence of humic substances on lacustrine planktonic food chains. Hydrobiologia 229: 73-91.

86

Jönsson M., L. Ranåker, P. A. Nilsson, and C. Brönmark. 2013. Foraging efficiency and prey selectivity in a visual predator: differential effects of turbid and humic water. Canadian Journal of Fisheries and Aquatic Sciences 70: 1685-1690.

Karlsson, J., P. Byström, J. Ask, P. Ask, L. Persson, and M. Jansson. 2009. Light limitation of nutrient-poor lake ecosystems. Nature 460: 506-509.

Karlsson, J., A-K. Bergström, P. Byström, C. Gudasz, P. Rodríguez, and C. L. Hein. 2015. Terrestrial organic matter input suppresses biomass production in lake ecosystems. Ecology 96: 2870-2876.

Kelly, P. T., C. T. Solomon, B. C. Weidel, and S. E. Jones. 2014. Terrestrial carbon is a resource, but not a subsidy, for lake zooplankton. Ecology 95: 1236-1242.

Lennon, J. T., S. K. Hamilton, M. E. Muscarella, A. S. Grandy, K. Wickings, and S. E. Jones. 2013. A source of terrestrial organic carbon to investigate the browning of aquatic ecosystems. PLoS ONE 8: e75771.

Magurran, A. E. 1986. Individual differences in fish behaviour. Pages 338-365 in Pitcher, T. J. (ed). "The behaviour of teleost fishes". Springer-Verlag, US.

Mittelbach, G. G. 1981. Foraging efficiency and body size: a study of optimal diet and habitat use by bluegills. Ecology 62: 1370-1386.

Monteith, D. T., J. L. Stoddard, C. D. Evans, H. A. de Wit, M. Forsius, T. Høgasen, A. Wilander, B. L. Skjelkvåle, D. S. Jeffries, J. Vuorenmaa, B. Keller, J. Kopácek, and J. Vesely. 2007. Dissolved organic carbon trends resulting from changes in atmospheric deposition chemistry. Nature 450: 537-541.

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

Ranåker, L., M. Jönsson, P. A. Nilsson, and C. Brönmark. 2012. Effects of brown and turbid water on piscivore-prey fish interactions along a visibility gradient. Freshwater Biology 57: 1761-1768.

87

Rowe, D. K., T. L. Dean, E. Williams, and J. P. Smith. 2003. Effects of turbidity on the ability of juvenile rainbow trout, Oncorhynchus mykiss, to feed on limnetic and benthic prey in laboratory tanks. New Zealand Journal of Marine and Freshwater Research 37: 45-52.

Solomon, C. T., S. E. Jones, B. C. Weidel, I. Buffam, M. L. Fork, J. Karlsson, S. Larsen, J. T. Lennon, J. S. Read, S. Sadro, and J. E. Saros. 2015. Ecosystem consequences of changing inputs of terrestrial dissolved organic matter to lakes: current knowledge and future challenges. Ecosystems 18: 376-389.

Stasko, A. D., J. M. Gunn, and T. A. Johnston. 2012. Role of ambient light in structuring north- temperate fish communities: potential effects of increasing dissolved organic carbon concentration with a changing climate. Environmental Reviews 20: 173-190.

Utne-Palm, A. C. 2002. Visual feeding of fish in a turbid environment: physical and behavioural aspects. Marine and Freshwater Behaviour and Physiology 35: 111-128.

Vadeboncoeur, Y., M. J. Vander Zanden, and D. M. Lodge. 2002. Putting the lake back together: Reintegrating benthic pathways into lake food web models. Bioscience 52: 44-54.

Vander Zanden, M. J., G. Cabana, and J. B. Rasmussen. 1997. Comparing trophic position of freshwater fish calculated using stable nitrogen isotope ratios (δ15N) and literature dietary data. Canadian Journal of Fisheries and Aquatic Sciences 54: 1142-1158.

Vander Zanden, M. J. and Y. Vadeboncoeur. 2002. Fishes as integrators of benthic and pelagic food webs in lakes. Ecology 83: 2152-2161.

Vinyard, G. L., and W. J. O’Brien. 1976. Effects of light and turbidity on the reactive distance of bluegill (Lepomis macrochirus). Journal of the Fisheries Research Board of Canada 33: 2845-2849.

Webb, J. F., J. C. Montgomery, and J. Mogdans. 2008. Bioacoustics and the lateral line system of fishes. Pages 145-182 in Webb, J. F., R. R. Fay, and A. N. Popper (eds). "Fish bioacoustics". Springer-Verlag, NewYork.

88

Weidel, B., S. Carpenter, J. Cole, J. Hodgson, J. Kitchell, M. Pace, and C. Solomon. 2008. Carbon sources supporting fish growth in a north temperate lake. Aquatic Sciences 70: 446-458.

Weidel, B. C., K. Baglini, S. E. Jones, P. T. Kelly, C. T. Solomon, and J. A. Zwart. In review. Differences in light climate due to dissolved organic carbon concentration drive species- specific changes in fish zooplanktivory.

Werner, E. E., D. J. Hall, D. R. Laughlin, D. J. Wagner, L. A. Wilsmann, and F. C. Funk. 1977. Habitat partitioning in a freshwater fish community. Journal of the Fisheries Research Board of Canada 34: 360-370.

Wolf, M., and F. J. Weissing. 2012. Animal personalities: consequences for ecology and evolution. Trends in Ecology and Evolution 27: 452-461.

89

Figure captions

Figure 2.1. Number of chironomid larvae consumed by bluegill (max: 10) over a water color gradient in mesocosm feeding experiments. DW represents bluegill from the relatively 2 clear Deadwood Lake (F1,24 = 0.4, R = 0.02, p = 0.5). HB represents bluegill from the much 2 darker Hummingbird Lake (F1,33 = 0.12, R = <0.01, p = 0.7).

Figure 2.2. Diet characteristics across the DOC gradient. Stomach biomass represents the dry mass of prey in each fish stomach. Prey length is the average prey length (mm) per fish. Benthic diet fraction is the proportion of dry mass in each stomach that consisted of benthic prey. Benthic selectivity is the index of selectivity for benthic items depending on proportion of benthic vs pelagic biomass in the environment, with positive numbers representing selectivity for benthic resources. Bluegill data is presented in the left column, and yellow perch in the right.

90

Tables and Figures

Table 2.1. Summary of samples collected from each lake including DOC concentration, and the size range of fish that diets were collected from.

Mean size Min size Max size Lake DOC (mg/L) Species n Sampled (mm) (mm) (mm) Crampton 5.3 Bluegill 31 66 38 109 Yellow Perch 50 75 38 123 Raspberry 7.3 Bluegill 11 98 64 156 Yellow Perch 22 96 64 105 Brown 7.4 Bluegill 9 49 39 85 Yellow Perch 54 101 39 221 Bergner 8.1 Bluegill 37 96 51 180 Yellow Perch 14 95 51 150 Hummingbird 19.9 Bluegill 30 79 45 153 Yellow Perch 12 86 45 228

Table 2.2. Linear mixed effects model results (fixed slopes) for fish diet responses as a function of fish length and DOC, with lake as a covariate blocking factor. Significant results are in bold.

Response Species Intercept Fish length DOC Stomach mass Bluegill 0.000 (0.071) 0.623 (0.071) 0.161 (0.071) Yellow perch -0.032 (0.188) 0.183 (0.081) -0.001 (0.144) Average prey size Bluegill -0.052 (0.213) 0.722 (0.075) -0.060 (0.233) Yellow perch -0.048 (0.256) 0.219 (0.086) -0.075 (0.197) Benthic diet fraction Bluegill -0.907 (0.088) 0.127 (0.035) -0.063 (0.095) Yellow perch -0.862 (0.073) -0.039 (0.033) 0.054 (0.058) benthic selectivity Bluegill -1.423 (0.089) 0.038 (0.023) -0.098 (0.098) Yellow perch -1.076 (0.105) 0.017 (0.023) -0.063 (0.079)

91

Figure 2.1

10

8

6

4 Chironomids consumed (n) Chironomids consumed

2

HB 0 DW

5 10 15 20 25 30

Color (g440)

92

Figure 2.2

93

Appendix 2A

Figure 2A. Temperature (°C), dissolved oxygen (% saturation), and surface and bottom light levels (photosynthetically active radiation - PAR, μmol m-2 sec-1) plotted with the water color treatments. Temperature was slightly lower in the clearest treatments as we had to mix some colder well water in with the lake water to achieve the correct water color.

25 110

105

100 20 95

90 15 Temperature (C) Temperature 85

80 HB Dissolved oxygen (% saturation) HB DW DW 10 75

5 10 15 20 25 30 5 10 15 20 25 30

Color (g440) Color (g440)

20 HB 7 HB DW DW 6 15 5

4 10

3 PAR tank bottom PAR tank surface 2 5 1

0 0

5 10 15 20 25 30 5 10 15 20 25 30

Color (g440) Color (g440)

94

Appendix 2B

Table 2B. Regression equations for estimating prey dry mass from body size measurements. DM = dry mass (mg), L = body length (mm).

Class Prey Item Equation Reference 3.211 Zoobenthos Amphipoda DM=0.002*L A 2.809 Anisoptera naiad DM=0.0076*L A 2.617 Benthic invertebrate unidentifiable DM=0.0018*L A 2.617 Benthic invertebrate unidentifiable larvae DM=0.0018*L A 2.617 Benthic invertebrate unidentifiable pupae DM=0.0018*L A 2.871 Ceratopogonidae larvae DM=0.00022*L A 2.617 Chironomidae larvae DM=0.0018*L A 2.53 Chironomidae pupae DM=0.004571*L B 2.91 Coleoptera DM=0.0077*L A 2.692 Diptera larvae DM=0.0025*L A 2.692 Ephemeroptera naiad DM=0.0025*L A 2.734 Hemiptera DM=0.0108*L A 1.66 Hydrachnidae DM=0.13265*L C 2.918 Lepidoptera larvae DM=0.0027*L A 2.809 Odonata naiad DM=0.0076*L A 1.54 Oligochaeta DM=0.005888*L B 2.754 Plecoptera naiad DM=0.0094*L A 2.801 Sialidae larvae DM=0.0031*L A 2.839 Trichoptera nymph DM=0.0056*L A 2.839 Trichoptera pupae DM=0.0056*L A 2.785 Zygoptera naiad DM=0.0051*L A 3.23 Zooplankton Copepoda DM=0.00927*L D 3.599 Chaoborus larvae DM=0.001425*L E 3.599 Chaoborus pupae DM=0.001425*L E 2.52 Cladocera DM=0.011705*L D Ln(DM)= -7.849+ Terrestrial 2 prey Arachnid 0.49335*L+0.0080448*L F Ln(DM)= -9.314+0.66297*L 2 Diptera adult -0.016486*L F 2.696 Hymenoptera adult DM=0.01379*L G

95

2.27 Odonata adult DM=0.14*L H Ln(DM)= -7.761+ 2 Terrestrial invertebrate 0.34975*L+0.0039315*L F 3.044 Trichoptera adult DM=0.00995*L G 2.27 Zygoptera adult DM=0.14*L H Ln(DM)= -7.761+ 2 Unknown Invertebrate unidentifiable 0.34975*L+0.0039315*L F

References:

A: Benke, A. C., A. D. Huryn, L. A. Smock and J. B. Wallace. 1999. Length-mass relationships for freshwater macroinvertebrates in North America with particular reference to the Southeastern United States. Journal of the North American Benthological Society 18 (3): 308-343.

B: Methot, G., C. Hudson, P. Gagnon, B. Pinel-Alloul, A. Armellin, and A. M. Tourville Poirier. 2012. Macroinvertebrate size-mass relationships: how specific should they be? Freshwater Science 31: 750-764.

C: Baumgärtner D. and K. Rothhaupt. 2003. Predictive length-dry mass regressions for freshwater invertebrates in a pre-alpine lake littoral. International Review of Hydrobiology 88 (5): 453-463.

D: McCauley, E. 1984. Chapter 7. The estimation of the abundance and biomass of zooplankton in samples. In Downing, J. A., and F. H. Rigler (eds). "A manual on methods for the assessment of secondary production in fresh waters". Blackwell Scientific Publications.

E: Ramcharan, C. W., D. J. McQueen, A. Perez-Fuentetaja, N. D. Yan, E. Demars, and J. Rusak. 2001. Analyses of lake food webs using individual-based models to estimate Chaoborus production and consumption. Archive for Hydrobiology: Issues in Advanced Limnology 56: 101-126.

F: Sage, R. D. 1982. Wet and dry-weight estimates on insects and spiders based on length. American Midland Naturalist 108: 407-411.

96

G: Sample, B. E., R. J. Cooper, R. D. Greer, and R. C. Whitmore. 1993. Estimation of insect biomass by length and width. American Midland Naturalist 129: 234-240.

H: Sabo, J. L., J. L. Bastow, and M. E. Power. 2002. Length-mass relationships for adult aquatic and terrestrial invertebrates in a California watershed. Journal of the North American Benthological Society 21: 336-343.

97

Appendix 2C

Figure 2C. Number of diets sampled, broken down by fish length. Lakes in order of increasing DOC. BLG=bluegill, YWP=yellow perch. Lake legends: CR = Crampton, RB = Raspberry, BR = Brown, BE = Bergner, HB = Hummingbird.

15 15 CR_BLG CR_YWP 10 10

5 5

Frequency Frequency 0 0 0 50 100 150 200 0 50 100 150 200

15 length (mm) 15 length (mm) RB_BLG RB_YWP 10 10

5 5

Frequency Frequency 0 0 0 50 100 150 200 0 50 100 150 200

15 length (mm) 15 length (mm) BR_BLG BR_YWP 10 10

5 5

Frequency Frequency Frequency 0 0

0 50 100 150 200 0 50 100 150 200

15 length (mm) 15 length (mm) BE_BLG BE_YWP 10 10

5 5

Frequency Frequency 0 0

0 50 100 150 200 0 50 100 150 200

15 length (mm) 15 length (mm) HB_BLG HB_YWP 10 10

5 5

Frequency Frequency 0 0 0 50 100 150 200 0 50 100 150 200

length (mm) Fish length (mm) length (mm)

98

Connecting Statement

In the previous chapters I explored how DOC affects resource availability and the ability of fishes to utilize these resources. Karlsson et al. (2009) demonstrated a negative relationship between DOC and fish productivity, and suggested that this was due to resource limitation in lakes of higher DOC concentrations. In the next chapter, I determine how DOC-mediated resource limitation affects fish life history strategies, in order to better understand how the negative DOC-fish productivity relationship is manifested. Using bluegill (Lepomis macrochirus) as a model organism, I determine how growth and maturity trade off and vary along a DOC gradient. Understanding how DOC affects the growth rates, and size and age of maturity of fish is critical in managing fisheries effectively in the event of rising DOC concentrations (Stasko et al. 2012).

99

Chapter 3

LIFE HISTORY CONSTRAINTS EXPLAIN NEGATIVE RELATIONSHIP BETWEEN FISH PRODUCTIVITY AND

DISSOLVED ORGANIC CARBON IN LAKES.

Nicola Craig1*, Stuart E. Jones2, Brian C. Weidel3, and Christopher T. Solomon1

1 Dept. of Natural Resource Sciences, McGill University, Ste. Anne de Bellevue, QC, Canada.

2 Dept. of Biological Sciences, University of Notre Dame, Notre Dame, IN, USA.

3 U.S. Geological Survey, Great Lakes Science Center, Lake Ontario Biological Station, Oswego, NY, USA.

Status: Undergoing revisions for submittal to the Journal of Animal Ecology.

Abstract

Resource availability constrains the life history strategies available to organisms and may thereby limit population growth rates and productivity. We used this conceptual framework to explore the mechanisms driving recently reported negative relationships between fish productivity and dissolved organic carbon (DOC) concentrations in lakes. We studied populations of bluegill (Lepomis macrochirus) in a set of lakes with DOC concentrations ranging from 3 to 24 mg L-1; previous work has demonstrated that primary and secondary productivity of food webs is negatively related to DOC concentration across this gradient. For each population we quantified growth rate, age at maturity, age-specific fecundity, maximum age, length-weight and length-egg size relationships, and other life history characteristics. We observed a strong negative relationship between maximum size and DOC concentration; for instance, fish reached masses of 225 to 350 g in low-DOC lakes but < 150 g in high-DOC lakes. Relationships between fecundity and length, and between egg size and length, were constant

100 across the DOC gradient. Because fish in high-DOC lakes were smaller but had similar fecundity and egg size at a given size, their total lifetime fecundity was as much as two orders of magnitude lower than fish in low-DOC lakes. High DOC concentrations appeared to constrain the range of life history strategies available; populations in high-DOC lakes always had low initial growth rates and older ages at maturity, whereas populations in low-DOC showed higher variability in these traits. This was also the case for the intrinsic rates of natural increase of these populations, which were always low at the high end of the DOC gradient. This has clear implications for the sustainable management of recreational fisheries in the face of considerable spatial heterogeneity and ongoing temporal change in lake DOC concentrations.

Introduction

Resource availability is a major driver for selection of various life history strategies in organisms that can ultimately affect population growth and productivity (MacArthur & Wilson 1967, Wilbur et al. 1974, Begon et al. 2006). When resources are limited less energy is available for growth and reproduction, necessitating reduced allocation to one or the other with implications for lifetime reproductive output and fitness (Fisher 1930, Stearns & Koella 1986). In unproductive environments organisms may grow more slowly and allocate their limited energy to growth for longer periods until they become large enough to produce sufficient numbers of offspring (MacArthur & Wilson 1967, Pianka 1970). In such cases organisms are likely to mature later than those in more productive environments, which can grow faster and get larger quicker. Understanding how resource limitation affects life history traits and ecosystem productivity is important when managing populations of organisms such as fish, particularly in an era of unprecedented environmental changes (Winemiller & Rose 1992, Jennings et al. 1998, Steffen et al. 2011).

In lake ecosystems, high levels of dissolved organic carbon (DOC) reduce ecosystem productivity and resource availability through the limitation of light and habitat availability (Karlsson et al. 2009, Craig et al. 2015). Terrestrially derived DOC, which makes up the majority of the DOC pool in many lakes, is flushed in from the surrounding landscape and stains the water, such that high DOC lakes have a dark brown color (Jones 1992, Wilkinson et al. 2013).

101

This darkening of the water has physical and biological effects on lakes by reducing light and heat penetration and thus reducing thermocline depths, restricting the area of well-oxygenated epilimnion (Wetzel 2001, Read & Rose 2013). As a result of these effects, primary production is limited due to the diminished light climate (Ask et al. 2009, Godwin et al. 2014) and secondary production is limited due to both reduced primary production and area of suitable habitat (Karlsson et al. 2009, Kelly et al. 2014, Craig et al. 2015). DOC-mediated reductions in resource and habitat availability will likely affect fish life history strategies in ways that are currently poorly understood (Stasko et al. 2012, Solomon et al. 2015).

Several recent studies have shown that high levels of DOC can have a negative effect on fish production and abundance (Karlsson et al. 2009, Finstad et al. 2014), however the mechanisms behind this are not well understood. Understanding how this reduction in productivity is manifested through life history characteristics may help us better manage populations in high DOC environments. This is especially important as DOC concentrations have been rising in northern hemisphere freshwaters over the past two decades (Evans et al. 2005, Monteith et al. 2007, Solomon et al. 2015). In addition, lake DOC concentrations are spatially heterogeneous and can span wide gradients across the landscape (Hanson et al. 2007). Understanding how DOC affects critical life history traits, such as growth rates and age at maturity, can help form better management regulations both spatially, and as lake ecosystems become darker (Jennings et al. 1998).

In this study, we quantified how fish life histories vary with DOC and lead to decreased productivity by estimating initial growth, reproductive output and age at maturity of female bluegill (Lepomis macrochirus) across an eleven lake DOC gradient. In high DOC (dark) environments, fish populations are likely to be resource limited and so we predict they will have less surplus energy available to allocate to growth and reproduction, which may lead to slower somatic growth, smaller maximum size, later age at maturity, and thus a lower overall fecundity than populations in low DOC (clear) lakes (Roff 1983, Stearns & Koella 1986, Charnov 1993).

102

Methods

Study system and data collection

Our study focused on bluegill as they are widely distributed over gradients of DOC, have variable life history strategies, and are a popular sport fish of economic value (Drake et al. 1997). We collected the fish from eleven lakes near the Wisconsin-Michigan border, including the University of Notre Dame Environmental Research Centre and the surrounding area of Vilas County, Wisconsin. The lakes spanned broad gradients of size, nutrients and DOC (Table 3.1).

Bluegill were collected just before spawning (May-June) in 2013 and 2014, with the exception of Erickson Lake and Birch Lake which were only sampled in 2013. Fish were collected using littoral fyke nets set overnight (2-6 net nights per lake per year), with the exception of McCullough where we were restricted to angling. We collected a wide range of sizes (from 75 mm to maximum size) in order to estimate size at maturity, maximum size, and the relationship between gonad mass and fish size.

Euthanized fish were frozen until they could be dissected. Once thawed, the fish were measured (mm) and weighed (g) and then the gonads were removed and weighed whole on a microbalance. Otoliths were also removed at this point. In 2014, we measured egg size and abundance for female fish. We stored the gonads in formalin (1:10 formaldehyde dilution) until analysis. The gonad was removed from formalin, reweighed, and three replicate sub-samples were taken and weighed. Each sub-sample was then placed on a microscope slide and two drops of glycerol were applied to help separate the eggs. The number of mature eggs per sub- sample were counted under a stereo microscope and then this was scaled up to estimate the number of mature eggs for the whole gonad. A photograph was also taken of each slide using a digital microscope camera and egg widths were measured using ImageJ software (National Institutes of Health, USA).

The ages of a subset of fish were determined using otoliths in order to understand the age structure of the populations and to estimate age at maturity. Between 27 and 67 otoliths were analyzed per lake (mean = 44, see Appendix 3A for full details of sample sizes for various analyses). Sagittal otoliths were mounted in resin and a transverse section (~300 μm thick) was

103 cut through the otolith origin using a pair of diamond blades and a low-speed IsoMet saw (Buehler, USA). Sections were polished with successively finer polishing pads, adhered to slides, and annuli were interpreted and counted under a compound microscope. Otoliths were interpreted by two individuals and assigned ages were compared until agreement was reached for all otoliths.

Environmental variables

DOC samples were collected twice during summer 2013 and 3 times during summer 2014. Water samples were taken using a Van Dorn bottle at 3 points in the epilimnion (bottom, middle, top) after which the pooled samples were filtered through 0.7 μm GF/F filters and DOC concentrations were measured using a Shimadzu TOC-V total organic carbon analyzer (Shimadzu Scientific Instruments, Kyoto, Japan).

Growth, maximum age and size

We estimated initial growth rates (ω; mm y-1) for each population by fitting the Gallucci and Quinn (1979) parameterization of the von Bertalanffy growth model to the length at age data for females in each lake, via maximum likelihood with a Gamma likelihood (Appendix 3B.1). We estimated confidence intervals for this parameter by quadratic approximation from the information matrix.

We estimated maximum length, mass, and age in each lake as the 95th percentile of the size and age distributions. For age, we first estimated the age of each captured fish from its size, using the lake-specific age-mass relationship. We estimated maximum age in this way rather than from the subset of fish that we directly aged because that subset was a non- random sample intended to evenly span the entire size range in each lake. We included male and female fish in maximum size and age estimates because size at age was similar between the sexes (Appendix 3B.1).

Size and age at maturity

We estimated size at maturity using a broken-stick regression of gonad mass on fish length (Appendix 3B.3). We set the slope of gonad mass on length to zero for the first piece of

104 the regression (immature fish), and estimated the remaining parameters by maximum likelihood using a Gamma likelihood. We converted the estimated size at maturity to age at maturity using the lake-specific age-length relationships (Appendix 3B.1). There are two lakes for which this model did not fit well. Tenderfoot Lake had few data points for immature fish and a shallow slope for mature fish which resulted in an estimated size at maturity that is probably too low. Also, no immature fish were collected in Deadwood Lake, and so the intercept of immature gonad mass was fixed as an average of the other lakes (0.43 ± 0.07 standard error).

Fecundity and intrinsic rate of increase

We estimated fecundity in two different manners; the potential lifetime fecundity of an individual if they reach maximum age, and the realized lifetime fecundity which incorporates the probability of mortality.

We calculated potential lifetime fecundity for females in each lake based on lake- specific relationships between age and body mass, body mass and gonad mass, and gonad mass and number of eggs, summed between the lake-specific age at maturity and maximum age (Appendices 3B.2, 3B.4, and 3B.5). For Birch and Erickson lakes we did not have data on the gonad mass to egg number relationship so we used the across-lake average in our calculations for these two lakes.

Realized lifetime fecundity (R0), i.e. the mean number of eggs produced by an average individual accounting for mortality, was estimated using the formula:

R0 = Σ lxmx

Where lx is the proportion of the population surviving to age x, and mx is the number of eggs produced at age x (Connell 1970). We estimated mortality from the fitted Von Bertalanffy growth function for each lake (m=K*1.6, Jensen 1996). We also used the R0 estimates to calculate the intrinsic rate of natural increase, r, which is the change in population size per individual per unit time (T, in this case average generation time):

r = ln R0/T

105

Testing effect of DOC on life history parameters

We used linear regression models to test for effects of DOC concentration on life history parameters. We included a random lake effect when the dependent variable was observed at the level of individual fish. These analyses were conducted in R using the lm function and lme4 package (R Core Team 2014, Bates et al. 2014).

Results

Bluegill in clear lakes attained larger sizes than those in dark lakes (Fig. 3.1). The 95th percentiles of both length and mass were negatively related to DOC concentration (Fig. 3.1a). The differences in maximum length and especially mass between clear and dark lakes were substantial; for instance, in clear lakes the largest fish sampled were generally between 225 and 350 g, whereas in dark lakes the largest fish were < 150 g. Initial growth rate was not significantly related to DOC concentration, although growth was slow in the darkest lakes and 2 sometimes fast in clear lakes (F1,9 = 0.76, R = 0.08, p = 0.4; Fig. 3.1b). Instead, fish in clear lakes apparently achieved large maximum size either by sustaining high growth rates for a longer period of time, or by growing slowly for many years (Fig. 3.1c). Mass was strongly related to length (linear mixed effects (LME) slope = 0.99 ± 0.02) and this relationship was fairly constant across the DOC gradient (effect of DOC: LME slope = -0.04 ± 0.06).

Because fish in clearer lakes reached larger sizes, they also had much greater fecundity than those in dark lakes (Fig. 3.2). Age-specific fecundity was positively related to fish size, and this relationship did not vary among lakes as a function of DOC concentration (LME slope of egg number ~ length: 0.73 ± 0.3, DOC slope: -0.01 ± 0.24, Appendix 3B.6). Fecundity therefore increased much more quickly with age in clear lakes than in dark ones (Fig. 3.2a). Egg size was not related to fish size except in Red Bass Lake, nor was it related to DOC concentration (LME length slope: 0.3 ± 0.3, DOC slope: -0.16 ± 0.24, Appendix 3B.7). Populations with high initial 2 growth rates matured at earlier ages (F1,9 = 8.5, R = 0.48, p = 0.02). Given the relationship between initial growth and DOC (Fig. 3.1b) this meant that populations in dark lakes tended to mature late but that there was no significant linear relationship between DOC and age at 2 maturity (F1,9 = 2.26, R = 0.2, p = 0.17; Fig. 3.2b). Size at maturity, which ranged from

106 approximately 95 to 155 mm, and reproductive lifespan, which ranged from 3 to 10 years, were 2 2 also not related to DOC concentration (F1,9 = 0.22, R = 0.02, p = 0.6 and F1,9 = 1.21, R = 0.12, p = 0.3, respectively). Differences in age at maturity and especially age-specific fecundity translated into large differences in lifetime potential fecundity, which was strongly negatively related to DOC concentration (Fig. 3.2c). Realized fecundity was not significantly related to DOC, but this relationship improved when we removed from the analysis two heavily-fished lakes (Fig. 3.2d). The intrinsic rate of natural increase was variable at low DOC concentrations and tended to be low at high DOC concentrations, although there was no significant linear relationship (Fig. 3.3).

Discussion

The effects of DOC-induced resource limitation on fish life histories

Our results demonstrate that life history mechanisms, such as reductions in size at age leading to reduced reproductive output, may help explain the negative effect of DOC on fish productivity. Fish in darker lakes are resource limited and thus have less surplus energy to allocate to growth and reproduction (Diana 1987, Lester et al. 2004, Karlsson et al. 2009). Gonad mass at a given size does not change with DOC; instead, DOC-mediated reductions in resource availability seem to affect reproductive output by limiting post-maturation growth and ultimately maximum size. Further support for the idea of limited post-maturation growth in dark lakes comes from the fact that the relationship between initial growth and DOC is weak (Fig. 3.1b); greater differences in growth are only observed at later ages, after the fish have reached maturity (Fig. 3.1c). Larger female fish have been shown to be disproportionally valuable in fisheries due to their ability to produce greater numbers of eggs (Green, 2008) so the lack of large fish in high DOC environments is contributing to a loss of reproductive output and thus productivity.

Resource limitation has been found to reduce maximum size and reproductive output in other systems and species. For example, Grether et al. (2001) and Reznick et al. (2001) found that guppies living in resource-rich, low-competition environments were able to grow faster, reach larger sizes and allocate more resources to reproduction. Resource limitation has also

107 resulted in smaller maximum sizes for other populations of bluegill (Aday et al. 2006), as well as for yellow perch (Heath & Roff, 1996) and northern pike (Diana 1987). As reproductive output is strongly related to body size in fishes (Roff 1983), we could assume that these species were also limited in their reproductive potential. This evidence suggests that DOC-mediated changes in resource availability and life history patterns resulting in reduced fecundity may occur in other fish species.

Low energy availability and/or high DOC levels have been shown to reduce initial growth in fish populations such as bluegill (Gerking 1962), perch (Horppila et al. 2010), and rainbow smelt (King et al. 1999), however, we did not observe a strong relationship between DOC and initial growth in our study (Fig. 3.1b). We did however observe that fish in the darker lakes grew more slowly at later ages and this resulted in lower maximum sizes (Fig. 3.1a, 3.1c). Similarly, Rask & Tunnainen (1990) found that European perch and roach were smaller in length at a given age in darker Finnish lakes. This suggests that bluegill, and perhaps some other species, are not resource limited at small sizes across a DOC gradient but may become so in darker lakes as they get bigger and require more energy to maintain a larger body size, resulting in slower adult growth.

There was also a trend of later age at maturity in the darker lakes suggesting that the DOC-mediated resource limitation restricts the age at which females can produce eggs. This fits with established patterns as it has been shown that resource limitation can delay maturity as females must wait longer until they have amassed sufficient body mass and energy reserves to make reproduction viable (Tyler & Dunn 1976, Roff 1982, Stearns & Koella 1986, Drake et al. 1997). Interestingly, despite bluegill in darker waters maturing later, maximum age and number of spawning years showed no trend with DOC, suggesting that even though fish mature later, it does not necessarily reduce the number of spawning years. This again reinforces the importance of maximum body size, as even though bluegill from high DOC lakes may live and/or spawn for equivalent periods to low DOC populations, they are still limited in their potential and realized reproductive output.

108

Our results suggest that life history parameters are more constrained at the higher end of the DOC gradient and more flexible in clearer waters. This is true for growth and for size and age at maturity, suggesting that fish in clear waters can take advantage of several different strategies such as growing fast, maturing early, and dying relatively young, or growing more slowly, maturing later and larger, and living longer. On the other hand, fish in darker waters can only grow relatively slowly and mature relatively late; they can live for a variable amount of time but never reach the large sizes seen in the clearer lakes. Similar wedge-shaped patterns of biomass and production across DOC gradients have been observed in other studies of primary producers, primary consumers, and fishes (Kelly et al. 2014, Craig et al. 2015, Karlsson et al. 2015, Seekell et al. 2015a). Some of this variation may be explained through the dynamics of nutrient and light limitation (Seekell et al. 2015b, Solomon et al. 2015). In clear waters there may be high variability in nutrient concentrations which may stimulate or constrain growth and productivity with a minimal effect of light (Seekell et al. 2015b). However dark lakes may reach a threshold of light limitation after which an increase in nutrients will no longer stimulate productivity (Seekell et al. 2015b, Solomon et al. 2015). This variability in clear lakes explains why some of our results (e.g. initial growth and age at maturity) showed trends, but were not significant. Bluegill in clear lakes are subjected to a gradient of productivity levels which allow for variation in life history strategies, whereas in the dark lakes, fish growth and thus other life history characteristics are constrained by low levels of productivity. This may make it easier to predict how fish populations may respond to increasing DOC, but less so in situations where DOC levels are decreasing (e.g. Schindler et al. 1996).

Alternative mechanisms for life history patterns

The declines in bluegill growth, maximum size and fecundity with DOC have been discussed thus far in the context of resource availability, however, there may be other mechanisms at play which are currently not well understood. Population dynamics such as survivorship and competition could also play a role and not necessarily in a mutually exclusive manner. Lower light climates in darker lakes may provide additional shelter for larval and juvenile fish and increase survivorship, which would in turn increase competition for limited resources (Ylikarjula et al. 1999, Ranåker et al. 2012). Although our study shows that potential

109 egg output declines along the DOC gradient, we do not know how many of these eggs, and subsequent juveniles, actually survive and if this varies across the lakes. Oplinger & Wahl (2015) showed that juvenile survival in bluegill was positively related to egg size, but not the size of the mother. We saw that egg size was not related to DOC, nor to female size (except in one lake). This suggests that chances of juvenile survival may not vary in this respect. Determining how egg and juvenile survival varies with DOC could help us tease out the influence of density- dependent effects from environmental resource limitation on life history strategies such as growth. Similarly, it has been shown that DOC-induced shallowing of thermoclines reduces habitat availability for secondary consumers (Kelly et al. 2014, Craig et al. 2015) and likely fish populations as well (Finstad et al. 2014), which may also result in higher competition for resources in a more limited space. It is generally difficult to pick apart mechanisms for life history variation (Wilbur et al. 1974) and in this case the potential mechanisms are so intimately linked that further research into survivorship, energy intake and population densities with DOC is required.

Implications for fish populations

Fishing pressure is another factor that can strongly influence fish life histories (Drake et al. 1997, Law 2000) and it had a strong negative effect in the two most heavily fished lakes in this study. Allequash and Big Arbor Vitae were the only two lakes with high fishing pressure and they stood out as having fast growth rates and higher mortality rates than the other lakes. This reduced realized fecundity (Fig. 3.2d) which incorporates the probability of survival into the potential lifetime egg output, such that the fish in these lakes have high potential to produce eggs but are harvested before they get the chance to do so. These two populations had relatively high intrinsic rates of increase (r > 1.48), which suggests that they are better able to withstand fishing pressure. However, the fish populations at the higher end of the DOC gradient tended to have lower r values, indicating that these populations would find it harder to sustain substantial fishing pressure. In addition, populations with the life history characteristics common to dark lakes (e.g. slower growth and later age at maturity) tend to be more susceptible to fishing pressure (Jennings et al. 1998), increasing the likelihood that they may collapse, and may do so under lower levels of fishing pressure than in many clearer lakes. Rypel

110

(2015) showed that decreasing bag limits from 25 to 10 fish per day had a positive effect on bluegill maximum size, particularly in more productive lakes. If this were to work in high-DOC lakes, some of the pressure may be relieved if females can get larger and so produce more eggs. Modeling the effects of increased fishing pressure on the resilience of fish populations in darker lakes could be beneficial, and allow the tailoring of fisheries management strategies both spatially and under future browning scenarios.

Conclusions

DOC has many complex physical and biological effects on lakes, and understanding how DOC concentrations impact fish ecology is crucial in order to predict how ecosystems work both spatially, and as lakes get browner. DOC-mediated resource limitation appears to reduce post- maturation growth rates and thus maximum sizes of bluegill in darker lakes, which in turn decreases lifetime fecundity. This could be a major factor in the reduction of fish productivity that has been observed with DOC in previous studies (Karlsson et al. 2009, Finstad et al. 2014). Further increases in DOC in north temperate lakes could result in undesirable size structures in fisheries, and there may be a threshold of DOC above which it may be hard for fish to recover from high fishing pressure. Therefore lakes with moderate DOC concentrations may need to be monitored more closely and stricter fishing regulations may be needed in these lakes to maintain productive fisheries in the future.

Acknowledgements

Funding was provided by the Natural Sciences and Engineering Research Council of Canada. The staff of the University of Notre Dame Environmental Research Center facilitated our field work there. Technical assistance was provided by Pierre-Olivier Benoit, Ludovick Brown, Patrick Kelly, Jacob Lerner, Karling Roberts, Greg Sass, Jacob Ziegler, and Jacob Zwart. Members of the Wisconsin DNR, as well as Andrew Hendry and Melissa Lenker provided useful comments that improved the execution of this paper. Mention of specific products does not constitute endorsement by the U.S. Government. This is contribution number xxxx to the Great Lakes Science Center.

111

References

Aday, D. D., D. P. Philipp, and D. H. Wahl. 2006. Sex-specific life history patterns in bluegill (Lepomis macrochirus) interacting mechanisms influence individual body size. Oecologia 147: 31-38.

Ask, J., J. Karlsson, L. Persson, P. Ask, P. Byström and M. Jansson. 2009. Terrestrial organic matter and light penetration: Effects on bacterial and primary production in lakes. Limnology and Oceanography 54 (6): 2034-2040.

Bates, D., M. Maechler, B. Bolker, and S. Walker. 2015. lme4: linear mixed-effects models using Eigen and S4. R package version 1.1-8, http://CRAN.R-project.org/package=lme4.

Begon, M., Townsend, C. R., and J. L. Harper. 2006. Ecology: from individuals to ecosystems. 4th edition. Blackwell Publishing Ltd, Oxford.

Charnov, E. L. 1993. Life history invariants. Oxford University Press, Oxford.

Connell, J. H. 1970. A predator-prey system in the marine intertidal region. I. Balanus glandula and several predatory species of Thais. Ecological Monographs 40: 49-78.

Craig, N. C., S. E. Jones, B. C. Weidel, and C. T. Solomon. 2015. Habitat, not resource availability, limits consumer production in lake ecosystems. Limnology and Oceanography 60: 2079- 2089.

Diana, J. S. 1987. Simulation of mechanisms causing stunting in northern pike populations. Transactions of the American Fisheries Society 116: 612-617.

Drake, M. T., J. E. Claussen, D. P. Philipp and D. L. Pereira. 1997. A comparison of bluegill reproductive strategies and growth among lakes with different fishing intensities. North American Journal of Fisheries Management 17: 496-507.

Evans, C. D., D. T. Monteith and D. M. Cooper. 2005. Long-term increases in surface water dissolved organic carbon: Observations, possible causes and environmental impacts. Environmental pollution 137: 55-71.

112

Finstad, A. G., I. P. Helland, O. Ugedal, T. Hesthagen and D. O. Hessen. 2014. Unimodal response of fish yield to dissolved organic carbon. Ecology Letters 17 (1): 36-43.

Fisher, R. A. 1930. The genetical theory of natural selection. Dover Publications Inc, New York.

Gallucci, V. F., and T. J. Quinn. 1979. Reparameterizing, fitting, and testing a simple growth model. Transactions of the American Fisheries Society 108: 14-25.

Gerking, S. D. 1962. Production and food utilization in a population of bluegill sunfish. Ecological Monographs 32: 31-78.

Godwin, S. C., S. E. Jones, B. C. Weidel, and C. T. Solomon. 2014. Dissolved organic carbon concentration controls benthic primary production: Results from in situ chambers in north-temperate lakes. Limnology and Oceanography 59: 2112-2120.

Green, B. S. 2008. Maternal effects in fish populations. Advances in Marine Biology 54: 1-105.

Grether, G. F., D. F. Millie, M. J. Bryant, D. N. Reznick, and W. Mayea. 2001. Rain forest canopy cover, resource availability, and life history evolution in guppies. Ecology 82: 1546-1559.

Hanson, P. C., S. R. Carpenter, J. A. Cardille, M. T. Coe, and L. A. Winslow. 2007. Small lakes dominate a random sample of regional lake characteristics. Freshwater Biology 52: 814- 822.

Heath, D. D., and D. A. Roff. 1996. The role of trophic bottlenecks in stunting: a field test of an allocation model of growth and reproduction in yellow perch, Perca flavescens. Environmental Biology of Fishes 45: 53-63.

Horppila, J., M. Olin, M. Vinni, S. Estlander, L. Nurminen, M. Rask, J. Ruuhijärvi and H. Lehtonen. 2010. Perch production in forest lakes: the contribution of abiotic and biotic factors. Ecology of Freshwater Fish 19: 257-266.

Jennings, S., J. D. Reynolds, and S. C. Mills. 1998. Life history correlates of responses to fisheries exploitation. Proceedings of the Royal Society of London B: Biological Sciences 265: 333- 339.

113

Jensen, A. L. 1996. Beverton and Holt life history invariants result from optimal trade-off of reproduction and survival. Canadian Journal of Fisheries and Aquatic Sciences 53: 820- 822.

Jones, R. I. 1992. The influence of humic substances on lacustrine planktonic food chains. Hydrobiologia 229: 73-91.

Karlsson, J., P. Byström, J. Ask, P. Ask, L. Persson and M. Jansson. 2009. Light limitation of nutrient-poor lake ecosystems. Nature 460 (7254): 506-509.

Karlsson, J., A-K. Bergström, P. Byström, C. Gudasz, P. Rodríguez, and C. L. Hein. 2015. Terrestrial organic matter input suppresses biomass production in lake ecosystems. Ecology 96: 2870-2876.

Kelly, P. T., C. T. Solomon, B. C. Weidel, and S. E. Jones. 2014. Terrestrial carbon is a resource, but not a subsidy, for lake zooplankton. Ecology 95: 1236-1242.

King, J. R., B. J. Shuter and A. P. Zimmerman. 1999. Empirical links between thermal habitat, fish growth, and climate change. Transactions of the American Fisheries Society 128: 656- 665.

Law, R. 2000. Fishing, selection, and phenotypic evolution. ICES Journal of Marine Science 57: 659-668.

Lester, N. P., B. J. Shuter and P. A. Abrams. 2004. Interpreting the von Bertalanffy model of somatic growth in fishes: the cost of reproduction. Proceedings of the Royal Society of London B: Biological Sciences 271: 1625-1631.

MacArthur, R. H., and E. O. Wilson. 1967. The theory of island biogeography. Princeton University Press, Princeton.

Monteith, D. T., J. L. Stoddard, C. D. Evans, H. A. de Wit, M. Forsius, T. Høgasen, A. Wilander, B. L. Skjelkvåle, D. S. Jeffries, J. Vuorenmaa, B. Keller, J. Kopácek and J. Vesely. 2007. Dissolved organic carbon trends resulting from changes in atmospheric deposition chemistry. Nature 450: 537-541.

114

Oplinger, R. W., and D. H. Wahl. 2015. Egg characteristics and larval growth of bluegill from stunted and non-stunted populations. Journal of Freshwater Ecology 30: 299-309.

Pianka, E. R. 1970. On r- and K-selection. The American Naturalist 104: 592-597.

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

Ranåker, L., M. Jönsson, P. A. Nilsson, and C. Brönmark. 2012. Effects of brown and turbid water on piscivore-prey fish interactions along a visibility gradient. Freshwater Biology 57: 1761-1768.

Rask, M., and P. Tuunainen. 1990. Acid-induced changes in fish populations of small Finnish lakes. Pages 911-927 in Kauppi, P., P. Anttila, and K. Kenttämies (eds.) “Acidification in Finland”. Springer-Verlag, New York.

Read, J. S. and K. C. Rose. 2013. Physical responses of small temperate lakes to variation in dissolved organic carbon concentrations. Limnology and Oceanography 58: 921-931.

Reznick, D., M. J. Butler IV, and H. Rodd. 2001. Life-history evolution in guppies. VII. The comparative ecology of high- and low-predation environments. The American Naturalist 157: 126-140.

Roff, D. A. 1982. Reproductive strategies in flatfish: a first synthesis. Canadian Journal of Fisheries and Aquatic Sciences 39: 1686-1698.

Roff, D. A. 1983. An allocation model of growth and reproduction in fish. Canadian Journal of Fisheries and Aquatic Sciences 40: 1395-1404.

Rypel, A. L. 2015. Effects of a reduced daily bag limit on bluegill size structure in Wisconsin lakes. North American Journal of Fisheries Management 35: 388-397.

Schindler, D. W., P. J. Curtis, B. R. Parker, and M. P. Stainton. 1996. Consequences of climate warming and lake acidification for UV-B penetration in North American boreal lakes. Nature 379: 705-708.

115

Seekell, D. A., J-F. Lapierre, J. Ask, A-K. Bergström, A. Deininger, P. Rodríguez, and J. Karlsson. 2015a. The influence of dissolved organic carbon on primary production in northern lakes. Limnology and Oceanography 60: 1276-1285.

Seekell, D. A., J-F. Lapierre, and J. Karlsson. 2015b. Trade-offs between light and nutrient availability across gradients of dissolved organic carbon concentrations in Swedish lakes: implications for patterns in primary production. Canadian Journal of Fisheries and Aquatic Sciences 72:1663-1671.

Solomon, C. T., S. E. Jones, B. C. Weidel, I. Buffam, M. L. Fork, J. Karlsson, S. Larsen, J. T. Lennon, J. S. Read, S. Sadro, and J. E. Saros. 2015. Ecosystem consequences of changing inputs of terrestrial dissolved organic matter to lakes: current knowledge and future challenges. Ecosystems 18: 376-389.

Stasko, A. D., J. M. Gunn and T. A. Johnston. 2012. Role of ambient light in structuring north- temperate fish communities: potential effects of increasing dissolved organic carbon concentration with a changing climate. Environmental Reviews 20 (3): 173-190.

Stearns, S. C., and J. C. Koella. 1986. The evolution of phenotypic plasticity in life-history traits: predictions of reaction norms for age and size at maturity. Evolution 40: 893-913.

Steffen, W., Å. Persson, L. Deutsch, J. Zalasiewicz, M. Williams, K. Richardson, C. Crumley, P. Crutzen, C. Folke, L. Gordon, M. Molina, V. Ramanathan, J. Rockström, M. Scheffer, H. Joachim Schellnhuber, and U. Svedin. 2011. The Anthropocene: from global change to planetary stewardship. Ambio 40: 739-761.

Tyler, A. V., and R. S. Dunn. 1976. Ration, growth and measures of somatic and organ condition in relation to meal frequency in winter flounder, Pseudopleuronectes americanus, with hypotheses regarding population homeostasis. Journal of the Fisheries Board of Canada 33: 63-75.

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

116

Wilbur, H. M., D. W. Tinkle, and J. P. Collins. 1974. Environmental certainty, trophic level, and resource availability in life history evolution. The American Naturalist 108: 805-817.

Wilkinson, G. M., M. L. Pace, and J. J. Cole. 2013. Terrestrial dominance of organic matter in north temperate lakes. Global Biogeochemical Cycles 27: 43-51.

Winemiller, K. O., and K. A. Rose. 1992. Patterns of life-history diversification in North American fishes: implications for population regulation. Canadian Journal of Fisheries and Aquatic Sciences 49: 2196-2218.

Ylikarjula, J., M. Heino, and U. Dieckmann. 1999. Ecology and adaptation of stunted growth in fish. Evolutionary Ecology 13: 433-453.

117

Figure captions

Figure 3.1. a) 95th percentile of bluegill maximum length and mass regressed with DOC 2 2 concentration (length: F1,9 = 5.73, R = 0.39, p = 0.04; mass: F1,9 = 9.99, R = 0.53, p = 0.01), b) Von Bertalanffy Gallucci-Quinn initial growth rate (ω, mm per year) regressed with DOC 2 concentration (F1,9 = 0.76, R = 0.08, p = 0.4), c) Von Bertallanffy length-age growth curves for bluegill for each lake, lines are color-coded according to DOC concentration.

Figure 3.2. a) Mean age-specific egg output of female bluegill. Each line indicates the relationship for one lake, and extends from the age of maturity to maximum age. Lines are color-coded according to the DOC concentration of the lake. b) Female bluegill age at maturity 2 as a function of DOC (F1,9 = 2.26, R = 0.2, p = 0.17). Error bars are 95% confidence intervals. c) Lifetime potential fecundity for female bluegills as a function of DOC (log(potential fecundity) vs 2 DOC, F1,9 = 8.11, R = 0.47, p = 0.02). d) Lifetime realized fecundity (i.e. average egg number per age accounting for mortality) of female bluegills as a function of DOC (log(realized fecundity) vs 2 DOC, F1,9 = 2.75, R = 0.2, p = 0.13). This relationship improved when the heavily fished lakes 2 (Big Arbor Vitae and Allequash – grey points) were removed (F1,7 = 5.16, R = 0.42, p = 0.06).

Figure 3.3. The intrinsic rate of natural population increase (r) plotted against DOC 2 concentration (F1,9=1.45, R =0.14, p=0.26).

118

Tables and Figures

Table 3.1. Summary of lake parameters for the eleven lakes surveyed during the study period. DOC is dissolved organic carbon. Standard error (n samples = 5) is in parentheses.

Total Area Max DOC Lake Phosphorus (ha) depth (m) (mg/L) (μg/L) Big Arbor Vitae 433 12.5 3.1 (0.2) 29.1 (10.0) Crampton 25.9 18.5 5.0 (0.2) 9.4 (0.6) Allequash 168.4 8 5.7 (0.4) 17.7 (2.1) Erickson 44.5 5.5 6.2 (0.2) 19.6 (7.5) Bay 67.3 12.2 7.4 (0.4) 12.4 (1.5) Deadwood 9.7 8.8 9.7 (0.7) 12.4 (1.2) Tenderfoot 194.2 9.14 12.0 (0.8) 15.5 (1.4) Birch 204.7 13.7 12.5 (0.4) 12.1 (2.3) McCullough 89.4 8.2 14.3 (0.6) 14.0 (3.2) Red Bass 10.9 6.7 18.9 (0.8) 45.0 (15.4) Hummingbird 0.8 7.6 24.5 (0.8) 20.0 (1.5)

119

Figure 3.1

a) c)

length (mm) 250 weight (g) 250 200

150 Maximum size 200 100

5 10 15 20 25 150 DOC (mg/L)

b) (mm) Length

100 80

70

60 50

50 0-5mg/L 5-10mg/L 40 10-15mg/L 0 15-25mg/L

30 Initial growth rate (omega, mm/year) (omega, Initialrate growth 5 10 15 20 25 0 2 4 6 8 10 12 14

DOC (mg/L) Age (years)

120

Figure 3.2

a) b)

80000 8 0-5mg/L 5-10mg/L 60000 10-15mg/L 6 15-25mg/L

40000 4

20000 maturity at Age 2 Mean egg output egg Mean

0 0

0 2 4 6 8 10 12 14 5 10 15 20 25

Age (years) DOC (mg/L)

c) d)

50000 2e+05 20000 1e+05 10000

5e+04 5000

2000 2e+04 1000

1e+04 500 Realized lifetime fecundity (n eggs) (n Realized fecundity lifetime Potential lifetime fecundity (n eggs) (n fecundity lifetimePotential 5 10 15 20 25 5 10 15 20 25

DOC (mg/L) DOC (mg/L)

121

Figure 3.3

2.0

1.8

1.6

r 1.4

1.2

1.0

0.8 5 10 15 20 25

DOC (mg/L)

122

Appendix 3A.

Table 3A. Data on number of fish available for various analyses. F = female, M = male. n bluegill measured refers to total number of bluegill collected in fyke nets for which we obtained length-weight data, whether they were euthanized or released. n bluegill gonads examined refers to the total number of females for which we collected gonad weight from. n gonads for maturity analysis refer to the number of gonads used in the size at maturity analysis – i.e. from the date that gonad mass was at its peak in each lake. Asterisk (*) indicates that there were fish in this group for which sex could not be determined.

n gonads n bluegill n gonads n bluegill for n otoliths gonads used for Lake measured maturity analysed examined egg (F & M) analysis (F/M) (F) analysis (F) (F) Big Arbor Vitae 368 104 32 24 67 (51/16) Crampton 311 92 40 27 53 (38/15) Allequash 222 53 53 14 54 (36/18) Erickson 160 72 43 - 29 (19/10) Bay 214 79 25 35 50 (32/17)* Deadwood 209 71 35 23 39 (30/9) Tenderfoot 222 67 27 24 47 (30/16)* Birch 151 55 40 - 36 (18/18) McCullough 58 24 11 7 27 (21/6) Red Bass 271 75 26 22 40 (31/9) Hummingbird 93 38 20 6 44 (27/17)

123

Appendix 3B.

Supplementary figures of lake level fish data. Lake code key: BV – Big arbor Vitae, CR – Crampton, AQ – Allequash, ER – Erickson, BA – Bay, DW – Deadwood, BH – Birch, TF – Tenderfoot, MC – McCullough, RS – Red Bass, HB – Hummingbird.

Figure 3B.1. Von-Bertalanffy growth curves fit to female length at age data (black points and lines), as well as combined female and male data (black and grey points, grey lines) generated from otolith analysis of a subset of fish collected from each lake. Lakes are in order

of increasing DOC.

250 250 250

250 BV CR AQ ER

200 200 200 200

150 150 150 150

100 100 100 100

50 50 50 50

0 0 0 0

0 5 10 15 0 5 10 15 0 5 10 15 0 5 10 15

250 250 250

250 BA DW TF BH

200 200 200 200

150 150 150 150

100 100 100 100

Length(mm)

50 50 50 50

0 0 0 0

0 5 10 15 0 5 10 15 0 5 10 15 0 5 10 15

250 250

250 MC RS HB

200 200 200

150 150 150

100 100 100

50 50 50 female

male

0 0 0

0 5 10 15 0 5 10 15 0 5 10 15 Age (years) 124

Figure 3B.2. Female bluegill age determined by otolith analysis as a function of fish weight. Lakes are in order of increasing DOC.

BV CR AQ ER

15 15 15 15

10 10 10 10

5 5 5 5

0 0 0 0

0 100 200 300 0 100 200 300 0 100 200 300 0 100 200 300

BA DW TF BH

15 15 15 15

10 10 10 10

Age(years)

5 5 5 5

0 0 0 0

0 100 200 300 0 100 200 300 0 100 200 300 0 100 200 300

MC RS HB

15 15 15

10 10 10

5 5 5

0 0 0

0 100 200 300 0 100 200 300 0 100 200 300 Weight (g)

125

Figure 3B.3. Gonad mass of pre-spawn females plotted against body length. Lakes are in order of increasing DOC. These data were used to calculate size at maturity, as the break point between the slope of the regression lines for immature and mature fish. The model did not fit well to the Tenderfoot Lake (TF) data, for which the estimated size at maturity may therefore be too low. There were no immature fish collected in Deadwood Lake (DW) so the immature gonad mass for this lake was estimated as an average of the other lakes.

35 35 35 35 BV CR AQ ER 30 30 30 30

25 25 25 25

20 20 20 20

15 15 15 15

10 10 10 10

5 5 5 5

0 0 0 0

50 100 150 200 250 50 100 150 200 250 50 100 150 200 250 50 100 150 200 250

35 35 35 35 BA DW TF BH 30 30 30 30

25 25 25 25

20 20 20 20

15 15 15 15

10 10 10 10

5 5 5 5 Female gonadFemale weight (g) 0 0 0 0

50 100 150 200 250 50 100 150 200 250 50 100 150 200 250 50 100 150 200 250

35 35 35 MC RS HB 30 30 30

25 25 25

20 20 20

15 15 15

10 10 10

5 5 5

0 0 0

50 100 150 200 250 50 100 150 200 250 50 100 150 200 250 Length (mm)

126

Figure 3B.4. Female gonad mass as a function of fish mass. Lakes are in order of increasing DOC. This figure is for female bluegill collected just before spawning.

35 35 35 35 BV CR AQ ER 30 30 30 30

25 25 25 25

20 20 20 20

15 15 15 15

10 10 10 10

5 5 5 5

0 0 0 0

0 100 200 300 0 100 200 300 0 100 200 300 0 100 200 300

35 35 35 35 BA DW TF BH 30 30 30 30

25 25 25 25

20 20 20 20

15 15 15 15

10 10 10 10

5 5 5 5 Female gonadFemale weight (g) 0 0 0 0

0 100 200 300 0 100 200 300 0 100 200 300 0 100 200 300

35 35 35 MC RS HB 30 30 30

25 25 25

20 20 20

15 15 15

10 10 10

5 5 5

0 0 0

0 100 200 300 0 100 200 300 0 100 200 300 Weight (g)

127

Figure 3B.5. Number of eggs per gonad as a function of gonad mass for female bluegill just before spawning. Lakes are in increasing order of DOC.

BV CR AQ 40000 40000 40000

30000 30000 30000

20000 20000 20000

10000 10000 10000

0 0 0

0 5 10 15 20 25 30 0 5 10 15 20 25 30 0 5 10 15 20 25 30

BA DW TF 40000 40000 40000

30000 30000 30000

20000 20000 20000

10000 10000 10000 neggsgonad per

0 0 0

0 5 10 15 20 25 30 0 5 10 15 20 25 30 0 5 10 15 20 25 30

MC RS HB 40000 40000 40000

30000 30000 30000

20000 20000 20000

10000 10000 10000

0 0 0

0 5 10 15 20 25 30 0 5 10 15 20 25 30 0 5 10 15 20 25 30 Female gonad w eight (g)

128

Figure 3B.6. Egg number per gonad as a function of fish length. Lakes are in order of increasing DOC. Egg number was significantly related to fish length, but not DOC (LME length slope: 0.73 ± 0.3, DOC slope: -0.01 ± 0.24).

BV CR AQ 40000 40000 40000

30000 30000 30000

20000 20000 20000

10000 10000 10000

0 0 0

100 150 200 250 100 150 200 250 100 150 200 250

BA DW TF 40000 40000 40000

30000 30000 30000

20000 20000 20000

10000 10000 10000 neggsgonad per

0 0 0

100 150 200 250 100 150 200 250 100 150 200 250

MC RS HB 40000 40000 40000

30000 30000 30000

20000 20000 20000

10000 10000 10000

0 0 0

100 150 200 250 100 150 200 250 100 150 200 250 Fish length (mm)

129

Figure 3B.7. Egg width as a function of fish length. Lakes are in order of increasing DOC. Egg width was not significantly related to fish length or DOC for 8 out of 9 lakes examined (LME length slope: 0.3 ± 0.3, DOC slope: -0.16 ± 0.24, with the exception being Red Bass (RS) Lake.

BV CR AQ 1.4 1.4 1.4

1.2 1.2 1.2

1.0 1.0 1.0

0.8 0.8 0.8

0.6 0.6 0.6

0.4 0.4 0.4

100 150 200 250 100 150 200 250 100 150 200 250

BA DW TF 1.4 1.4 1.4

1.2 1.2 1.2

1.0 1.0 1.0

0.8 0.8 0.8

0.6 0.6 0.6

Average egg width Average (mm) 0.4 0.4 0.4

100 150 200 250 100 150 200 250 100 150 200 250

MC RS HB 1.4 1.4 1.4

1.2 1.2 1.2

1.0 1.0 1.0

0.8 0.8 0.8

0.6 0.6 0.6

0.4 0.4 0.4

100 150 200 250 100 150 200 250 100 150 200 250 Fish length (mm)

130

Connecting Statement

In the preceding chapters, I have used observational spatial DOC-gradient surveys and mesocosms to investigate how DOC affects consumer behaviour and productivity in lakes. However, while DOC is variable across spatial scales, it has also been varying temporally, with a net increase observed in north temperate freshwaters over the past two decades. In order to truly understand how these changes in DOC may affect lake food webs, we need to use experimental approaches at the appropriate temporal scale. In the next chapter I return to zoobenthos again, measuring biomass and productivity in the context of a whole ecosystem experiment where we manipulate DOC concentrations over a period of several years in an experimental lake. The variation in zoobenthos dynamics resulting from this temporal increase in DOC is compared to the results from the spatial survey in chapter 1, and the differing outcomes of both these studies showcase why multiple approaches are needed to fully determine the effects of environmental variables on ecosystems.

131

Chapter 4

A TEMPORAL INCREASE IN DISSOLVED ORGANIC CARBON CONCENTRATION RESULTS IN AN

UNEXPECTED INCREASE IN ZOOBENTHOS BIOMASS AND PRODUCTIVITY IN A WHOLE-LAKE

EXPERIMENT.

Nicola Craig1*, Stuart E. Jones2, Patrick T. Kelly2, Brian C. Weidel3, Jacob A. Zwart2, and Christopher T. Solomon1

1 Dept. of Natural Resource Sciences, McGill University, Ste. Anne de Bellevue, QC, Canada.

2 Dept. of Biological Sciences, University of Notre Dame, Notre Dame, IN, USA.

3 U.S. Geological Survey, Great Lakes Science Center, Lake Ontario Biological Station, Oswego, NY, USA.

Status: In preparation for submittal to Ecosystems.

Abstract

Over the past several decades, dissolved organic carbon (DOC) concentrations have been rising in many north-temperate freshwaters, however the ecological consequences of this increase are not well understood. Spatial DOC-gradient studies suggest that consumer productivity should decrease in lakes with higher DOC concentrations due to a reduction in available habitat for primary producers and secondary consumers. However, analysis of static systems may not inform us how productivity may be affected by a temporal increase in DOC, and we may be missing valuable information regarding transitional periods which are essential to our understanding of how future rises in DOC may affect aquatic ecosystems. Here, we describe the results of a whole-lake ecosystem experiment where we increased DOC over time in a treatment basin, relative to a reference, and documented the effects on zoobenthos

132 biomass and productivity. Contrary to the results of a recent spatial survey, we found that zoobenthos productivity actually increased with a temporal increase in DOC concentration in the treatment basin. This pattern was largely due to the increase in the proportion of large- bodied Chironomidae in the metalimnion of the treatment basin, which coincided with the largest changes in environmental variables such as temperature and oxygen, and we discuss potential explanations for these observations. The increase in DOC in the treatment basin coincided with an increase in the availability of nutrients which entered the lake along with the DOC. We discuss the potential for non-linear relationships between DOC and productivity, where an increase of DOC-associated nutrients may actually increase productivity in nutrient- limited systems, provided the light-limiting effect of DOC is not strong. The effects of DOC on ecosystem productivity are evidently complex, and this study demonstrates that we cannot substitute temporal studies for DOC-gradient observations if we truly want to understand how future increases in DOC may affect north-temperate lake ecosystems.

Introduction

Over the past several decades, increases in concentrations of dissolved organic carbon (DOC) have been observed in many north temperate aquatic ecosystems (Evans et al. 2005, Monteith et al. 2007). As terrestrially derived DOC tends to stain waters, this process has sometime been referred to as lake 'browning' (Roulet & Moore 2006). There are several proposed mechanisms driving this phenomenon including land use change (Oni et al. 2015), recovery from acidification (Monteith et al. 2007), climate change (Freeman et al. 2001, Oni et al. 2015), and nitrogen deposition (Findlay 2005). While the debate regarding this mechanism continues, less attention has been paid to the ecological consequences of temporal variations in DOC (Solomon et al. 2015). This knowledge gap needs to be addressed in order to fully understand what may happen in aquatic ecosystems in the event of further DOC increases.

DOC plays a major role in the physical and biological structure of lake ecosystems (Solomon et al. 2015). The dark color of DOC stains the water and reduces light and heat penetration (Jones 1992, Read & Rose 2013). In DOC-rich lakes, this results in reduced thermocline depths, restricting the area of warm and well-oxygenated water to relatively

133 shallower epilimnions (Read & Rose 2013). The reduction in light penetration also restricts primary productivity, and can result in net decreases in pelagic and benthic algal production (Hanson et al. 2003, Ask et al. 2012, Godwin et al. 2014). DOC can also reduce the productivity of secondary consumers, with production of zooplankton, zoobenthos, and fish all decreasing over a DOC gradient in various spatial lake surveys (Karlsson et al. 2009, Finstad et al. 2014, Kelly et al. 2014, Craig et al. 2015, Karlsson et al.2015).

The reduction of fish productivity in DOC-rich lakes has been partially attributed to the concurrent reduction of zoobenthic prey (Karlsson et al. 2009, Craig et al. 2015). Zoobenthos are an important element of aquatic food webs and provide prey for both juvenile fish, as well as adult benthivores (Vander Zanden and Vadeboncoeur 2002, Weidel et al. 2008). Zoobenthos are also important in structuring lake sediments, recycling nutrients, and as a subsidy to terrestrial ecosystems during emergence events (Covich et al. 1999, Strayer 1991, Nakano and Murakami 2001). Many factors can affect the biomass and productivity of zoobenthos including, but not limited to temperature, oxygen, substrate type, resource availability, and predation (Strayer and Likens 1986, Rasmussen 1988, Strayer 2009). Previous studies have indicated that DOC may limit the productivity of zoobenthos through reductions in primary production (Karlsson et al. 2009), and suitable, well-oxygenated habitat (Craig et al. 2015).

While the aforementioned studies give us a reasonable idea of how DOC affects zoobenthos productivity, they are based on spatial surveys which sample communities from lakes spanning a DOC-gradient, and as a result, may not reflect how these organisms may react to a temporal variation in DOC concentration. As we move towards understanding how recent and future increases in DOC affect ecosystem functioning, we need to incorporate studies with a temporal dynamic which may reveal succession effects that would be otherwise missed in spatial surveys (Carpenter 1998). Here, we use a whole-lake experiment as a tool to observe how a temporal increase in DOC concentration affects zoobenthos biomass and productivity. Based on results from a recent spatial survey (Craig et al. 2015), we expected productivity to decline as DOC concentration increased in an experimental basin. However, we found the opposite pattern, illustrating how valuable a combination of approaches can be in understanding fundamental dynamics in ecosystems (Kitchell et al. 1988, Carpenter 1998).

134

Methods

Study system and whole-ecosystem experiments

In order to determine how zoobenthos respond to a temporal increase in DOC, we conducted a whole lake experiment in Long Lake, Michigan, USA (46° 13'N, 89° 32'W), located at the University of Notre Dame Environmental Research Center. Whole ecosystem experiments are extremely useful tools in terms of unearthing complexities arising from environmental manipulations that may not be observed in equivalent spatial studies or mesocosm experiments (Carpenter 1998, Schindler 1998). However, this advantage comes with a cost in that these types of experiments often lack replication and may be affected by environmental variables beyond the researchers control (Carpenter 1989, Schindler 1998). In order to partially overcome these difficulties, a similar reference system is used which can help control for environmental variability. Any differences arising between the treatment and reference systems after an experimental manipulation can then be attributed to the effect of the manipulated variable. In the case of this experiment, we compare the differences between the reference and treatment before the manipulation, to the differences between them post- manipulation, to see if the magnitude of these differences increases significantly, which would indicate an effect of the manipulation.

Long Lake is a small (8.1 ha) oligo-mesotrophic hour-glass shaped lake that has two similar basins of a maximum depth of 14m. More detail on the watershed characteristics of the lake and experimental set up is described in Zwart et al. (2016), who detailed the effects of the manipulation on ecosystem metabolism. Briefly, in the spring of 2011, we separated the lake into two basins with a permeable curtain in order to collect two years of pre-manipulation data from both sides before installing an impermeable curtain in the fall of 2012. The hydrology of Long Lake is such that the majority of the DOC input comes through a small inlet stream in the East basin. Therefore, when the lake was divided in half with an impermeable curtain, most of the DOC load remained trapped in this Eastern basin, with DOC concentrations doubling over two years in a previous experiment that was focusing on nutrient manipulations (Christensen et al. 1996). We collected a total of two summers of pre-manipulation data, and three summers of

135 post-manipulation data, after DOC concentrations increased in the treatment basin relative to the reference (Zwart et al. 2016). Having pre-manipulation data from both basins allows us to attribute post-manipulation differences between them as an effect of the treatment rather than natural variation (Stewart-Oaten et al. 1986, Carpenter 1990).

Zoobenthos biomass, productivity, and species composition can vary significantly with depth (Babler et al. 2008). To adequately represent this potential variability in the experiment, we sampled zoobenthos in the epilimnion (1 m depth), the metalimnion (3 m depth), and the hypolimnion (8 m depth), and used these samples to generalize about productivity for each of these depth bands.

Environmental variables

Environmental variables for 2011-2014 were collected and summarized in Zwart et al. (2016), and additional 2015 data was collected in the same manner and summarized in this paper (Tables 4.1 and 4.2). Briefly, temperature and oxygen profiles were taken weekly (May- August) from the deepest point of both basins using a handheld polarographic sensor (YSI Pro 20, Yellow Springs Instruments, USA) and an integrated epilimnetic water sample was taken to estimate DOC concentration. Thermocline depth was estimated from a fixed chain of high- frequency temperature sensors using the function ts.thermo.depth() from the R package rLakeAnalyzer (Winslow et al. 2014). Pelagic primary production was estimated using high- frequency dissolved oxygen sensors placed at 0.7m in each basin and the metabolism model described in Solomon et al. (2013).

Other variables not summarized in Zwart et al. (2016) include benthic primary production and depth specific temperature and dissolved oxygen. Benthic primary production was estimated for ~4 days in May, June, and July, in 2012-2015 using in situ benthic chambers and the diel oxygen method as described in Godwin et al. (2014), and Craig et al. (2015). In 2011, benthic primary production was estimated using benthic core incubation methods. Cores were collected through scuba (two time zero (T0), two dark, and two light cores per sample site), transported to the lab and incubated at in situ temperatures. Change in core water CO2 concentrations through time was calculated from mean concentrations across duplicate cores

136 at T0 and the light and dark incubated cores. Respiration rates were calculated as the increase in CO2 divided by incubation time in dark cores. Net CO2 Production was the decrease in CO2 in light incubated cores divided by incubation time. Gross Primary Production was calculated as

Net CO2 Production plus the respiration rate. These rates were converted to areal rates using the area of the corer (0.00196 m2) and an approximate water volume in each core of 0.14 L. The benthic chamber and benthic core methods have been shown to be comparable (Godwin et al. 2014). Depth specific (1, 3, and 8 m) temperature and dissolved oxygen were determined by averaging the output from the weekly profiles described above. The water at 3 m depth in Long Lake was generally oxygenated for most of the summer, but sometimes dipped below 0.5 mg/L around August. This low oxygen period may affect zoobenthos productivity and so we attempted to quantify the length of it for each year. As we often did not sample the lake in late- August - early September, we only used estimates of the days dissolved oxygen was <0.5 mg/l for which we had data, and so they should be viewed as minimal estimates (Appendix 4A).

Zoobenthos collection, biomass and production

We collected zoobenthos in sediment samples six times during the summer (May- August) for the two pre- and three post-manipulation years (2011-2015). For each basin and sampling occasion, two samples were taken at 8 m and four samples at 3 m using an Ekman dredge (0.023 m²), and four samples were taken at 1 m depth using a push core (5 cores per sample, total area: 0.017 m²). Samples were sieved on the day of collection through a 250μm mesh bag and invertebrates were picked from the samples by eye and stored in 70% ethanol.

Zoobenthos were identified to genus or the lowest possible taxonomic level using the keys of Holsinger (1972), Stern (1990), and Merritt et al. (2008). Chironomidae were an exception to this, and usually were not identified beyond the family level. However, a subset of the Chironomidae individuals collected from 2012-2014 (~100 randomly sampled per basin, per year, from the 3 m depth band) were classified at the Genus level using a partial sequence of the 28S ribosomal RNA gene (Cranston et al. 2012). These individuals were stored at room temperature in DESS solution (Yoder et al. 2006) until analysis. DNA was extracted from Chironomidae individuals according to manufacturer’s recommendations using the E-Z 96 tissue

137

DNA kit from Omega bio-tek and an epMotion 5075 Eppendorf liquid handling robot. A portion of the 28S ribosomal RNA gene was amplified from DNA extracts according to methods presented in Cranston et al. (2012) and Sanger sequenced using BigDye v3.1 and an Applied Biosystems AB3730XL sequencer. The resultant sequences were classified to the Genus level using the reference 28S ribosomal RNA gene sequence set published by Cranston et al. (2012) and the classify.seqs() function in mothur (Schloss et al. 2009). Only individuals that were identified at a confidence level of 50% or greater were used in species composition analysis.

Each organism collected was photographed under a stereo microscope using a digital microscope camera and head capsule or body length was measured from the images using ImageJ (National Institutes of Health, USA). Using published length-mass regression equations (Appendix 1A), we calculated dry mass (mg) from the recorded lengths for every individual.

Mean biomass (g m-2) was estimated for each depth for each sampling occasion as the average of the two or four samples taken. Mean invertebrate body size (mg dry mass) was taken as the average across all samples per depth per year. Production (g m-2 y-1) was estimated for each depth using the predictive regression model of Plante and Downing (1989), which uses mean annual biomass, maximum individual body mass, and mean annual water surface temperature as predictors. As in Craig et al. (2015), we only had summer data available to use in this model, and this may bias our estimates towards higher values. In addition, as this model uses surface temperature as a predictor for invertebrate production from deeper sites which may be colder, we believe that production estimates from these colder sites may also be biased high, and thus the production estimates presented here should be interpreted with caution.

As well as these depth-specific estimates of zoobenthos biomass and production, we also estimated these values at the whole lake level. This was achieved by dividing the lake into three depth bands representing the areas around the 1, 3 and 8 m sampling depths. We multiplied the area of each depth band by the average depth-specific zoobenthos biomass or production, summed these depth band values, and divided by the total area of the lake.

138

Statistical analysis

We chose to use Welch t-tests to analyze the impact of our manipulation on our biological and environmental variables as it is robust to changes in variances in the data among years, and has been recommended for use in whole-ecosystem experiments (Stewart-Oaten et al. 1992). In these analyses, we used the Welch t to compare the differences in paired sampling events of variables (i.e. the treatment basin - reference basin variable) from the pre- manipulation period, against the post-manipulation period. If the direction of these differences were significantly different post-manipulation compared to pre-manipulation, this would indicate that the manipulation had an effect on the given variable. We conducted these analyses for zoobenthos biomass for each depth, as well as at the whole lake level, and also tested for differences in the associated environmental variables at depth. We used an autocorrelation function (ACF) to check the Welch t assumption of a lack of temporal autocorrelation in the data. ACF values were within the 95% confidence bounds at lags up to 18 time steps for zoobenthos biomass, as well as temperature and dissolved oxygen, indicating no significant temporal autocorrelation in the data.

We incorporated multivariate analysis to explore whether there were significant changes in invertebrate assemblage structure as a result of the manipulation. As the majority of the shift in zoobenthos biomass occurred in the 3 m stratum, we focused on this depth band for the years 2012-2014 when a subset of Chironomidae identifications were available. The relative proportions of the subset of Chironomidae Genera identified were scaled to the number of chironomidae sampled in each year, and then combined with the non-chironomid data to estimate the total proportion abundance of each taxa. We then applied a non-metric multidimensional scaling ordination using the metaMDS() function within the vegan package in R, applying a Bray-Curtis similarity matrix to square root transformed data (Oksanen et al. 2015, R Core Team 2015). Rare species (i.e. only found in one year-basin combination) were excluded from the analysis.

139

Results

Environmental variables

As previously reported in Zwart et al. (2016), the manipulation caused DOC to increase in the treatment basin by 32% and decrease in the reference basin by 19% (Table 4.1). As a result of the manipulation, the thermocline depth of the treatment basin was reduced by 0.3 m relative to the reference (Table 4.1). Pelagic primary production was similar between the basins before the manipulation but became ~32% higher in the treatment basin relative to the reference post-manipulation (Table 4.1). This increase in pelagic production was partially attributed to an increase in available nutrients which entered the lake along with the DOC (Zwart et al. 2016).

Due to the aforementioned shallowing of the thermocline depth, water temperature was significantly lower at 3 m depth in the treatment basin after the manipulation while remaining fairly constant in the reference (Welch t = 3.92, df = 20.4, p = 0.001, Fig. 4.1). There was no significant effect of the manipulation on temperatures at 1 m (Welch t = 1.97, df = 20.7, p = 0.06), however temperatures at 8 m were higher in the reference before the manipulation, and become more similar between the basins post-manipulation (Welch t = -5.83, df = 27.5, p = <0.001). Dissolved oxygen was also significantly lower at 3 m depth in the treatment basin compared to the reference after the manipulation (Welch t = 6.86, df = 27.9, p = <0.001, Fig. 4.1), but this trend was partially due to an increase in metalimnetic oxygen concentrations in the reference basin. Dissolved oxygen remained similar between the basins at 1 m and 8 m (Welch t, p > 0.49). The reference basin did not appear to have long periods of anoxia at 3 m either before or after the manipulation. However, the treatment basin had a minimum of 11.5 ± 5 days of anoxia at 3 m at the end of summer before the manipulation, and this rose to a minimum of 33.7 ± 17 days in the three post-manipulation years (Appendix 4A).

There was not a large enough sample size to compare benthic primary production between the basins statistically, however the treatment basin as a whole had much lower levels of benthic primary production post-manipulation compared to the reference (Table 4.1). While comparable before the manipulation, post-manipulation benthic primary production was 58%

140 lower in the treatment basin at 1 m compared to the reference (Table 4.2). In general, benthic -1 -1 primary production was low at 3 m, averaging around 111 mg O2 m d in the reference basin, and negligible values in the treatment basin post-manipulation.

Zoobenthos biomass and production

Whole lake zoobenthos biomass became significantly higher in the treatment basin compared to the reference post-manipulation (Welch t = -4.85, df = 27.5, p = <0.001, Fig. 4.2). This difference was attributable to an increase in biomass at 3m depth, with 1 m and 8 m remaining fairly unchanged (3 m Welch t = -6.32, df = 27.9, p = <0.001; 1 m Welch t = 0.05, df = 27.7, p = 0.96; 8 m Welch t = -0.11, df = 27.3, p = 0.9, Fig. 4.2). Zoobenthos abundance did not differ between the basins at any depth or as a whole throughout the whole manipulation (Welch t p = >0.2, Appendix 4B). However, individual body mass became significantly higher at 3 m in the treatment basin post-manipulation (Welch t = -5.07, df = 25.6, p = <0.001, Fig. 4.3, Appendix 4C) suggesting that the increase in biomass observed here was due to increases in body size rather than abundance (there was no change in average body mass at 1 and 8 m, p = >0.1).

Whole lake zoobenthos production was ~32.5% lower in the treatment basin pre- manipulation (2011-12), but eventually became ~48.5% higher relative to the reference post- manipulation (2014-15) with a lag effect evident in 2013 where the treatment was still 19.5% less productive (Fig. 4.4). Again, this result was largely driven by shifts in productivity at the 3 m depth band with the 1 m and 8 m sites showing no consistent patterns in productivity throughout the experiment (Fig. 4.4).

Zoobenthos assemblage structure

As the majority of the shift in zoobenthos biomass post-manipulation occurred at 3 m depth, we decided post hoc to assess whether changes in assemblage structure might have contributed to this pattern. As part of this, we genetically identified some of the Chironomidae from this stratum, which represented over 90% of the zoobenthos population from this depth. We also incorporated the other taxa present into the analysis which included members of

141

Trichoptera, Odonata, Amphipoda, Oligochaeta, Ephemeroptera, Megaloptera, Veneroida, Hirudinea, and other Diptera (Appendix 4D).

Our ordination, which incorporated one year of pre- and two years of post-manipulation data, indicated that community composition was not affected by the experiment in a predictable way (Fig. 4.5). The two basins were very dissimilar in 2012 before the manipulation, and then became more similar in the following two post-manipulation years. In the pre- manipulation year (2012), the reference basin was dominated by Chironomus spp. (82%), while species composition in the treatment basin was more spread out, with the most dominant taxa being Djalmabatista spp. (17%), and Chironomus spp. (15%). However, in the two post- manipulation years, the dominant species became more similar within and between basins, consisting mostly of Glyptotendipes spp. (reference: 53 and 50%, treatment: 42 and 67% in 2013 and 2014 respectively), and Chironomus spp. (reference: 11 and 10%, treatment: 21 and 12%). These two genera are closely related members of the subfamily Chironominae and typically contain high levels of hemoglobin in order to cope with low oxygen conditions.

Discussion

Dissolved organic carbon can have profound effects on the structure and functioning of aquatic ecosystems, and recent DOC increases in north temperate freshwaters have led to the publication of several lake surveys with the aim of understanding how these increases may affect consumer productivity in lakes over a spatial DOC gradient (Karlsson et al. 2009, Finstad et al. 2014, Kelly et al. 2014, Craig et al. 2015). However, while these studies are a useful starting point for understanding mechanisms behind DOC-mediated shifts in productivity, they may not accurately reflect how consumers may react to temporal increases in DOC (Carpenter 1998). Temporal ecosystem studies focusing on the effect of DOC on aquatic food webs are rare, and those available either focus solely on pelagic processes (Christensen et al. 1996, Blomqvist et al. 2001), or detail an unrealistically large and rapid increase in DOC (Brothers et al. 2014). Here, we address this imbalance through the use of a whole-ecosystem experiment where we increased DOC in an experimental basin over time and measured the response in terms of zoobenthos biomass and productivity.

142

Disparity in the DOC-productivity relationship between spatial and temporal scales

While recent spatial observation surveys suggested that zoobenthos productivity should decrease as DOC increases due to reductions in resource and habitat availability (Karlsson et al. 2009, Craig et al. 2015), we found that the opposite was true over the temporal DOC increase in this study. There could be several possibilities for this disparity, the first being the occurrence of a transition, or lag period, that may be longer than the three post-manipulation years of our experiment, after which productivity may decline as expected. The second possibility is that there may be a non-linear relationship between DOC and ecosystem productivity, and the DOC concentration shift in our experiment was not large enough to see the expected effect.

One issue with using spatial gradient surveys to determine the effect of a variable that can vary temporally is the potential to miss valuable information from transitional periods where organisms may behave in an unpredictable manner (Carpenter 1998). Lakes used in spatial surveys may be at a stable state in which organisms have had many years to adapt to their surroundings, or leave the system altogether if conditions are unsuitable, whether through death or migration. However, when a variable changes significantly over time, there may be a succession period in which the organisms within the system have to react and adapt, and the length of an organisms lifecycle may determine how long this transitional period lasts. Zoobenthos are a diverse group of organisms with lifecycles that can vary from a few weeks, to several years (Lindegaard 1992, Armitage et al. 1995). In particular, zoobenthos living in deeper meta- and hypo-limnetic waters can take several years to complete their life cycle (Butler & Anderson 1990, Armitage et al. 1995), and potentially the three years of post-manipulation data we have from this experiment details somewhat of a lag effect of these incomplete invertebrate life cycles rather than the final result of a DOC increase. We touch on this point in more detail later on in the discussion, however this possibility highlights the importance of considering the time scale of whole-ecosystem experiments, and determining what is appropriate in order to ensure that the true long-term effects of a manipulation are captured (Schindler 1998). Unfortunately, due to the nature of funding and graduate cycles, as well as the considerable costs involved, it is not always possible to conduct such long-running experiments (Carpenter 1998).

143

Another theory that has been considered over the past few years is the possibility of a non-linear relationship between DOC and ecosystem productivity (Finstad et al. 2014, Solomon et al. 2015). In a recent series of spatial lake surveys, Finstad et al. (2014) demonstrated a unimodal relationship between fish productivity and DOC, and both Kelly et al. (2014) and Craig et al. (2015) observed high variability in zooplankton and zoobenthos production respectively at the lower end of a DOC gradient, yet consistently low levels of productivity in DOC-rich lakes. A potential explanation for these patterns is a trade-off between nutrient availability, and light limitation. Nutrient and DOC inputs are often linked (Dillon & Molot 2005, Kopáček et al. 2015), and so lakes with low DOC concentrations (e.g. <5 mg/L) may also be low in nutrients, limiting ecosystem productivity. On the other end of the gradient, lakes with high DOC concentrations (e.g. >15 mg/L) are generally nutrient-rich, but light limitation and associated shallowing of thermocline depths limit productivity through the reduction of suitable habitat availability for both primary producers, and secondary consumers (Hanson et al. 2003, Godwin et al. 2014, Kelly et al. 2014, Craig et al. 2015). It has been posed that there may be a threshold DOC concentration somewhere around 10-14 mg/L where the negative effects of light limitation start to strongly outweigh the positive effects of DOC-mediated nutrient availability, leading to consistently low levels of productivity beyond this threshold (Solomon et al. 2015). Lakes of intermediate DOC, including both the treatment and reference basins in this experiment, exist along a spectrum of nutrient regimes that may depend on surrounding watersheds and land use, which could result in variability in productivity both spatially, and temporally. As DOC increased in our treatment basin, so did nutrient concentrations, which may have released primary producers from nutrient limitation and resulted in the observed increase of primary and secondary productivity (Zwart et al. 2016, Kelly et al. 2016). Potentially, if the DOC concentrations in the treatment basin had surpassed the proposed light limitation threshold, we may have observed the expected reduction in ecosystem productivity. Further ecosystem experiments would be valuable in testing this hypothesis and determining if and at what DOC concentration this threshold may exist (Solomon et al. 2015).

Mechanisms for an increase in zoobenthos biomass and production in the metalimnion.

144

The relative increase in zoobenthos biomass in the treatment basin was mainly due to the occurrence of larger individuals around the 3 m, i.e. metalimnetic, depth band. This also coincided with relatively lower temperatures, dissolved oxygen concentrations, and benthic primary production at 3 m, but higher levels of pelagic primary production at the whole-basin level. As zoobenthos biomass, as well as the depth-specific environmental variables recorded remained largely similar at the other two depths sampled, we will focus the next section of discussion on the events occurring at the 3 m depth zone.

There are several possibilities as to what was driving this pattern in zoobenthos biomass in the treatment basin. The first is decreased predation, as reductions of dissolved oxygen concentrations in the metalimnion of the treatment basin may have deterred fish predators from foraging there and depleting zoobenthos biomass (Strayer 1991). The second is a shift in community composition towards larger bodied species. Thirdly, the increase in pelagic primary production in the treatment basin may have subsidized zoobenthos biomass as pelagic particles sink, and settle on deeper sediments (Strayer 2009). Finally, the decreases in light, temperature and oxygen may have affected growth and emergence patterns (Oliver 1971, Jónasson 1984, Strayer 2009), resulting in a build up of larger bodied individuals. We consider each of these possibilities in turn in the paragraphs below.

Decreased predation risk in the metalimnion is an unlikely explanation for the increase in zoobenthos biomass in the treatment basin for several reasons. Firstly, we collected diet data from largemouth bass, (Micropterus salmoides, the main species present) monthly throughout the five summers of the experiment and rarely found chironomids, the main prey item at 3 m, in the stomachs even during emergence events where they appeared as pupae. Chironomid larvae constituted 1% of the treatment, and 3.9% of the reference diet items collected, and pupae constituted 2.6, and 2.2% respectively (N. Craig, unpublished data). Secondly, in June of 2014, we conducted a pilot tagging experiment on some largemouth bass from each basin (three in the reference, and six in the treatment), where we tracked external temperatures of the fish every ten minutes over one to two weeks per individual. We were able to convert these temperatures to approximate swimming depths by matching them with data from the high frequency temperature sensor chains situated in the middle of each basin. During this study

145 period, we never observed a bass traveling below the thermocline depth, which at the time was ~2.1 m in the treatment, and ~2.3 m in the reference. This indicates that bass were not preying upon deeper invertebrates in either basin, at least during summer stratification, and so this should not have affected summer biomasses of zoobenthos in the metalimnion.

Due to the reduction in dissolved oxygen concentrations at 3 m in the treatment, we considered that there might be a species composition shift towards larger bodied, hypoxic- tolerant chironomids such as Chironomus spp. (Jónasson 1984, Armitage et al. 1995). However, while zoobenthos composition was fairly different between the basins before the manipulation (2012), after the manipulation (2013-14) communities in both basins were dominated by similar genera of hypoxic-tolerant members of the sub-family Chironominae. This suggests that zoobenthos assemblage structure was not the driver of increased biomass in the treatment basin. However, because of the limited (three year) scope of the data, and aforementioned potential lag effects of zoobenthos transition, we cannot completely rule out this possibility.

The relative increase in pelagic primary production in the treatment basin may have subsidized biomass production of zoobenthos in the metalimnion (Strayer 2009). Some zoobenthos attain their energy requirements through consumption of flocculating algal and detrital particles that descend upon the sediments (Pinder 1986, Armitage et al. 1995). Members of the subfamily Chironominae, which was the most abundant taxa at 3 m, include filter and deposit feeders which can utilize this resource (Jónasson & Kristiansen 1967). This explanation may also account for the slight decrease in zoobenthos production in the reference basin, which coincided with a similar small reduction in pelagic algal production. One issue with this theory is that we do not see a similar increase of zoobenthos biomass in the 1 m and 8 m depth zones in the treatment basin. This could be a result of fish predation, lower deposition rates, and a more diverse invertebrate community in the shallower sites (Strayer 1991). In the deeper sites, there may be a reduction in the amount and quality of particulate matter reaching the bottom as it is consumed by pelagic organisms during sinking (Meyers & Eadie 1993, Strayer 2009). Another potential resource that was affected by the experiment was benthic primary production. However, this decreased in the treatment basin, and increased in the reference post-manipulation, so is less likely to be a factor in the increase of zoobenthos biomass in the

146 manipulated basin. This observation supports the findings of the spatial survey by Craig et al. (2015) where it was suggested that benthic algal production is not the best predictor of zoobenthos production.

Many zoobenthos taxa, including the Chironomidae, have complex life cycles including a long aquatic larval phase, followed by emergence as adults when they reach a sufficient size and are triggered by specific cues. The variation in environmental conditions as a result of the manipulation may have affected the growth phase and emergence timings of the zoobenthos in a way that encouraged a build up of larger individuals in the metalimnion of the treatment basin (Jónasson 1984). There is also the possibility that the experiment coincided with a natural variation in zoobenthos life cycles between the basins, however, this is unlikely due to the dramatic divergence of biomass between the basins in the final two years of the experiment which was much greater than in any of the preceding years (Figure 4.2, 3 m). Within reasonable limits, higher temperatures allow invertebrate development to proceed at a faster rate than growth (Oliver 1971). Due to the relatively lower temperatures in the treatment basin, zoobenthos may have had slower growth rates, and attained larger size-at-maturity, relative to the warmer reference basin. In addition, slower growth rates as a result of reduced temperatures and dissolved oxygen concentrations can lead to increased life cycle length, which may have resulted in a build up of large final-instar larvae in the treatment basin (Jónasson 1972). The increased period of late-summer hypoxia in the metalimnion of the treatment basin likely decreased growth rates of zoobenthos also, as even chironomid species that are adapted to low oxygen conditions are forced to go dormant when oxygen levels decrease below 0.5 mg/L (Jónasson 1984, Heinis & Crommentuijn 1989, Hamberger et al. 1994). Our production estimates were largely based on biomass rather than cohort growth rates, and so may have overestimated productivity in the treatment basin if unfavorable conditions resulted in a standing crop of slow growing final-instar larvae.

From the above possibilities, it appears that fish predation and zoobenthos assemblage structure were less likely to be behind the higher metalimnetic zoobenthos productivity in the treatment basin, compared to resource availability and shifts in growth cycles. The fact that temperature and oxygen concentrations decreased significantly at this depth stratum likely had

147 an effect, but it is less clear how this was manifested over the course of this study. An observational study focusing on zoobenthos productivity across ten lakes over a DOC-gradient demonstrated a positive relationship between dissolved oxygen and zoobenthos biomass. Lower dissolved oxygen concentrations in the treatment did not seem to have this same negative effect on zoobenthos production, however the period of late-summer hypoxia did increase. If these periods of hypoxia, which may be increasing in north temperate lakes with rising DOC concentrations (Couture et al. 2015), become too prolonged, even specially adapted Chironomidae may be less able to survive them (Butler & Anderson 1990, Heinis & Davids 1993). This suggests that productivity in the treatment may have eventually decreased if DOC concentrations rose further, and adds to the possibility that our experiment was either demonstrating a transitional phase after which zoobenthos may settle to a new stable state, or that our manipulation did not reach a threshold DOC concentration after which limitations in light and habitat availability become so strong as to reduce productivity.

Concluding remarks

The results of recent spatial, DOC-gradient surveys suggest that zoobenthos productivity decreases with increasing DOC due to a reduction in resource and habitat availability (Karlsson et al. 2009, Craig et al. 2015), however in this temporal study, we found the opposite pattern. One explanation for this disparity is a non-linear relationship between DOC and ecosystem productivity, where lakes that are nutrient limited may actually benefit from additional DOC inputs provided light limitation below a certain level. In this case, it seems that the fertilizing effect of DOC-associated nutrients outweighed the potential negative effects of reduced habitat availability for zoobenthos in the treatment basin. This would account for the variability in zoobenthos productivity found in lakes of intermediate DOC concentration described by Craig et al. (2015), as the nutrient status, or other variables such as size or depth of a lake may determine how much DOC may impact productivity. Although, our reference basin was not a true reference in the sense that it was slightly impacted by the manipulation, it actually lends to this theory, as the reduction of DOC and associated nutrients led to a concurrent decline in zoobenthos biomass and productivity. In a sense the reallocation of resources caused the basins to diverge, but in the opposite direction from what we originally hypothesized. This

148 experiment demonstrates that spatial studies cannot always be substituted for equivalent temporal designs when considering the effect of a changing environmental variable on ecosystems. The effects of DOC on ecosystem productivity are evidently complex, and future studies that attempt to untangle the associated effects of nutrient availability, as well as the potential for light limitation thresholds, are essential in order to determine how future DOC increases may affect aquatic ecosystems.

Acknowledgements

Funding was provided by the Natural Sciences and Engineering Research Council of Canada, and the Quebec Centre for Biodiversity Science. The staff of the University of Notre Dame Environmental Research Center facilitated our field work there. The chemical analyses were conducted at the Center for Environmental Science and Technology, at the University of Notre Dame. Technical assistance was provided by J. Coloso, K. Baglini, H. Chung, B. Conner, K. Creamer, S. Elser, S. Godwin, E. Golebie, C.J. Humes, S. Koizumi, J. Lerner, E. Mather, S. McCarthy, M. O'Connor, R. Pilla, L. Raaf, K. Roberts, J. Schaefer, A. Searle, A. Sumner, J. Vanderwall, and J. Ziegler. Mention of specific products does not constitute endorsement by the U.S. government. This is contribution number XXXX to the Great Lake Science Center.

References

Armitage, P. D., P. S. Cranston, and L. C. V. Pinder (eds). 1995. The Chironomidae: biology and ecology of non-biting midges. 1st edition. Springer, UK.

Ask, J., J. Karlsson, and M. Jansson. 2012. Net ecosystem production in clear-water and brown- water lakes. Global Biogeochemical Cycles 26: GB1017.

Babler, A. L., C. T. Solomon, and P. R. Schilke. 2008. Depth-specific patterns of benthic secondary production in an oligotrophic lake. Journal of the North American Benthological Society 27: 108-119.

149

Blomqvist, P., M. Jansson, S. Drakare, A. –K. Bergström, and L. Brydsten. 2001. Effects of additions of DOC on pelagic biota in a Clearwater system: results from a whole lake experiment in Northern Sweden. Microbial Ecology 42: 383-394.

Brothers, S., J. Köhler, K. Attermeyer, H. P. Grossart, T. Mehner, N. Meyer, K. Scharnweber, and S. Hilt. 2014. A feedback loop links brownification and anoxia in a temperate, shallow lake. Limnology and Oceanography 59: 1388-1398.

Butler, M. G., and D. H. Anderson. 1990. Cohort structure, biomass, and production in a merovoltine Chironomus population in a Wisconsin bog lake. Journal of the North American Benthological Society 9: 180-192.

Carpenter, S. R. 1989. Replication and treatment strength in whole-lake experiments. Ecology 70 (2): 453-463.

Carpenter, S. R. 1990. Large-scale perturbations: opportunities for innovation. Ecology 71: 2038-2043.

Carpenter, S. R. 1998. The need for large-scale experiments to assess and predict the response of ecosystems to perturbation. Pages 287-312 in Pace, M. L. and P. M. Groffman (eds). "Successes, limitations, and frontiers in ecosystem science". Springer-Verlag, New York.

Christensen, D. L., S. R. Carpenter, K. L. Cottingham, S. E. Knight, J. P. LeBouton, D. E. Schindler, N. Voichick, J. J. Cole, and M. L. Pace. 1996. Pelagic responses to changes in dissolved organic carbon following division of a seepage lake. Limnology and Oceanography 41: 553-559.

Couture, R-M, H. A. de Wit, K. Tominaga, P. Kiuru, and I. Markelov. 2015. Oxygen dynamics in a boreal lake responds to long-term changes in climate, ice phenology, and DOC inputs. Journal of Geophysical Research: Biogeosciences 120. DOI: 10.1002/2015JG003065.

Covich, A. P., M. A. Palmer, and T. A. Crowl. 1999. The role of benthic invertebrate species in freshwater ecosystems. Zoobenthic species influence energy flows and nutrient cycling. Bioscience 49: 119-127.

150

Craig, N. C., S. E. Jones, B. C. Weidel, and C. T. Solomon. 2015. Habitat, not resource availability, limits consumer production in lake ecosystems. Limnology and Oceanography 60: 2079- 2089.

Cranston, P. S., N. B. Hardy, and G. E. Morse. 2012. A dated molecular phylogeny for the Chironomidae (Diptera). Systematic Entomology 37: 172-188.

Dillon, P. J., and L. A. Molot. 2005. Long-term trends in catchment export and lake retention of dissolved organic carbon, dissolved organic nitrogen, total iron, and total phosphorus: the Dorset, Ontario, study, 1978-1998. Journal of Geophysical Research: Biogeosciences 110: G01002.

Evans, C. D., D. T. Monteith, and D. M. Cooper. 2005. Long-term increases in surface water dissolved organic carbon: Observations, possible causes and environmental impacts. Environmental pollution 137: 55-71.

Findlay, S. E. G. 2005. Increased carbon transport in the Hudson River: unexpected consequence of nitrogen deposition? Frontiers in Ecology and the Environment 3: 133- 137.

Finstad, A. G., I. P. Helland, O. Ugedal, T. Hesthagen, and D. O. Hessen. 2014. Unimodal response of fish yield to dissolved organic carbon. Ecology Letters 17: 36-43.

Freeman, C., C. D. Evans, and D. T. Monteith. 2001. Export of organic carbon from peat soils. Nature 412: 785.

Godwin, S. C., S. E. Jones, B. C. Weidel, and C. T. Solomon. 2014. Dissolved organic carbon concentration controls benthic primary production: Results from in situ chambers in north-temperate lakes. Limnology and Oceanography 59: 2112-2120.

Hamberger, K., P. C. Dall, and C. Lindegaard. 1994. Energy metabolism of Chironomus anthracinus (Diptera: Chironomidae) from the profundal zone of Lake Esrom, Denmark, as a function of body size, temperature and oxygen concentration. Hydrobiologia 294: 43-50.

151

Hanson, P. C., D. L. Bade, S. R. Carpenter, and T. K. Kratz. 2003. Lake metabolism: relationships with dissolved organic carbon and phosphorus. Limnology and Oceanography 48: 1112- 1119.

Heinis, F., and T. Crommentuijn. 1989. The natural habitat of the deposit feeding chironomid larvae Stictochironomus histrio (Fabricius) and Chironomus anthracinus Zett. in relation to their responses to changing oxygen concentrations. Acta Biologica Debrecen, Oecologica Hungarica 3: 135-140.

Heinis, F., and C. Davids. 1993. Factors governing the spatial and temporal distribution of chironomid larvae in the Maarsseveen Lakes with special emphasis on the role of oxygen conditions. Netherlands Journal of Aquatic Ecology 27: 21-34.

Holsinger, J. R. 1972. The freshwater amphipod crustaceans (Gammaridae) of North America. Biota of Freshwater Ecosystems, Identification Manual 5. U.S. EPA, 89p.

Jónasson, P. M., and J. Kristiansen. 1967. Primary and secondary production in Lake Esrom. Growth of Chironomus anthracinus in relation to seasonal cycles of phytoplankton and dissolved oxygen. Internationale Revue der Gesamten Hydrobiologie 52: 163-217.

Jónasson, P. M. 1972. Ecology and production of the profundal benthos in relation to phytoplankton in Lake Esrom. Oikos Supplement 14: 1-148.

Jónasson, P. M. 1984. Oxygen demand and long term changes of profundal zoobenthos. Hydrobiologia 115: 121-126.

Jones, R. I. 1992. The influence of humic substances on lacustrine planktonic food chains. Hydrobiologia 229: 73-91.

Karlsson, J., P. Byström, J. Ask, P. Ask, L. Persson, and M. Jansson. 2009. Light limitation of nutrient-poor lake ecosystems. Nature 460: 506-509.

Karlsson, J., A-K. Bergström, P. Byström, C. Gudasz, P. Rodríguez, and C. L. Hein. 2015. Terrestrial organic matter input suppresses biomass production in lake ecosystems. Ecology 96: 2870-2876.

152

Kelly, P. T., C. T. Solomon, B. C. Weidel, and S. E. Jones. 2014. Terrestrial carbon is a resource, but not a subsidy, for lake zooplankton. Ecology 95: 1236-1242.

Kelly, P. T., N. Craig, C. T. Solomon, B. C. Weidel, J. A. Zwart, and S. E. Jones. 2016. Experimental whole-lake increase of dissolved organic carbon concentration produces unexpected increase in zooplankton biomass and production. Global Change Biology: doi: 10.1111/gcb.13260.

Kitchell, J. F., S. M. Bartell, S. R. Carpenter, D. J. Hall, D. J. McQueen, W. E. Neill, D. Scavia and E. E. Werner. 1988. Epistemology, experiments, and pragmatism. Pages 263-280 in Carpenter, S. R. (ed). "Complex interactions in lake communities" Springer-Verlag, New York.

Kopáček, J., J. Hejzlar, J. Kaňa, S. A. Norton, and E. Stuchlík. 2015. Effects of acidic deposition on in-lake phosphorus availability: a lesson from lakes recovering from acidification. Environmental Science and Technology 49: 2895-2903.

Lindegaard, C. 1992. Zoobenthos ecology of Thingvallavatn: Vertical distribution, abundance, population dynamics and production. Oikos 64: 257-304.

Merritt, R.W., K. W. Cummins and M. B. Berg (eds). 2008. An introduction to the aquatic insects of North America (4th edition). Kendall Hunt: Dubuque, Indiana. 1158pp.

Meyers, P. A., and B. J. Eadie. 1993. Sources, degradation and recycling of organic matter associated with sinking particles in Lake Michigan. Organic Geochemistry 20: 47-56.

Monteith, D. T., J. L. Stoddard, C. D. Evans, H. A. de Wit, M. Forsius, T. Høgasen, A. Wilander, B. L. Skjelkvåle, D. S. Jeffries, J. Vuorenmaa, B. Keller, J. Kopácek, and J. Vesely. 2007. Dissolved organic carbon trends resulting from changes in atmospheric deposition chemistry. Nature 450: 537-541.

Nakano, S., and M. Murakami. 2001. Reciprocal subsidies: dynamic interdependence between terrestrial and aquatic food webs. PNAS 98: 166-170.

153

Oksanen, J, F. G. Blanchet, R. Kindt, P. Legendre, P. R. Minchin, R. B. O'Hara, G. L. Simpson, P. Solymos, M. H. H. Stevens, and H. Wagner. 2015. vegan: Community Ecology Package. R package version 2.3-2. http://CRAN.R-project.org/package=vegan.

Oliver, D. R. 1971. Life history of the Chironomidae. Annual Review of Entomology 16: 211-230.

Oni, S. K., T. Tiwari, J. L. J Ledesma, A. M. Ågren, C. Teutschbein, J. Schelker, H. Laudon, and M. N. Futter. 2015. Local- and landscape-scale impacts of clear-cuts and climate change on surface water dissolved organic carbon in boreal forests. Journal of Geophysical Research: Biogeosciences 120. DOI: 10.1002/2015/JG003190.

Pinder, L. C. V. 1986. Biology of freshwater Chironomidae. Annual Review of Entomology 31: 1- 23.

Plante, C., and J. A. Downing. 1989. Production of freshwater invertebrate populations in lakes. Canadian Journal of Fisheries and Aquatic Sciences 46: 1489-1498.

R Core Team. 2015. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/.

Rasmussen, J. B. 1988. Littoral zoobenthic biomass in lakes, and its relationship to physical, chemical, and trophic factors. Canadian Journal of Fisheries and Aquatic Sciences 45: 1436-1447.

Read, J. S., and K. C. Rose. 2013. Physical responses of small temperate lakes to variation in dissolved organic carbon concentrations. Limnology and Oceanography 58: 921-931.

Roulet, N., and T. R. Moore. 2006. Environmental chemistry: Browning the waters. Nature 444: 283-284.

Schindler, D. W. 1998. Replication versus realism: the need for ecosystem-scale experiments. Ecosystems 1: 323-334.

Schloss, P. D., S. L. Westcott, T. Ryabin, J. R. Hall, M. Hartmann, E. B. Hollister, R. A. Lesniewski, B. B. Oaklet, D. H. Parks, C. J. Robinson, J. W. Sahl, B. Stres, G. G. Thallinger, D. J. Van Horn, and C. F. Weber. 2009. Introducing mothur: open-source, platform-independent,

154

community-supported software for describing and comparing microbial communities. Applied and Environmental Microbiology 75: 7537-7541.

Solomon, C. T., D. A. Bruesewitz, D. C. Richardson, K. C. Rose, M. C. Van de Bogert, P. C. Hanson, T. K. Kratz, B. Larget, R. Adrian, B. Leroux Babin, C. Chiu, D. P. Hamilton, E. E. Gaiser, S. Hendricks, V. Istvànovics, A. Laas, D. M. O'Donnell, M. L. Pace, E. Ryder, P. A. Staehr, T. Torgersen, M. J. Vanni, K. C. Weathers, and G. Zhu. 2013. Ecosystem respiration: drivers of daily variability and background respiration in lakes around the globe. Limnology and Oceanography 58: 849-866.

Solomon, C. T., S. E. Jones, B. C. Weidel, I. Buffam, M. L. Fork, J. Karlsson, S. Larsen, J. T. Lennon, J. S. Read, S. Sadro, and J. E. Saros. 2015. Ecosystem consequences of changing inputs of terrestrial dissolved organic matter to lakes: current knowledge and future challenges. Ecosystems 18: 376-389.

Stern, E. M. 1990. An illustrated key to the freshwater mussels (Bivalvia: Unionacea) of Wisconsin. Reports of the Museum of Natural History, University of Wisconsin-Stevens Point, Wisconsin.

Stewart-Oaten, A., W. W. Murdoch, and K. R. Parker. 1986. Environmental impact assessment: “pseudoreplication” in time? Ecology 67: 929-940.

Stewart-Oaten, A., J. R. Bence, and C. W. Osenberg. 1992. Assessing effects of unreplicated perturbations: no simple solutions. Ecology 73: 1396-1404.

Strayer, D., and G. E. Likens. 1986. An energy budget for the zoobenthos of Mirror Lake, New Hampshire. Ecology 67: 303-313.

Strayer, D. L. 1991. Perspectives on the size structure of lacustrine zoobenthos, its causes, and its consequences. Journal of the North American Benthological Society 10: 210-221.

Strayer, D. L. 2009. Benthic invertebrate fauna, lakes and reservoirs. Pages 191-204 in Likens, E. (ed). "Encyclopedia of inland waters. Volume 2". Elsevier, Oxford.

155

Vander Zanden, M. J., and Y. Vadeboncoeur. 2002. Fishes as integrators of benthic and pelagic food webs in lakes. Ecology 83: 2152-2161.

Weidel, B., S. Carpenter, J. Cole, J. Hodgson, J. Kitchell, M. Pace, and C. Solomon. 2008. Carbon sources supporting fish growth in a north temperate lake. Aquatic Sciences 70: 446-458.

Winslow, L., J. Read, R. Woolway, J. Bentrup, T. Leach, and J. Zwart. 2014. rLakeAnalyzer: package for the analysis of lake physics. R package version 1.4. http://CRAN.R- project.org/package=rLakeAnalyzer.

Yoder, M., I. Tandingan De Ley, I. W. King, M. Mundo-Ocampo, J. Mann, M. Blaxter, L. Poiras, and P. De Lay. 2006. DESS: a versatile solution for preserving morphology and extractable DNA of nematodes. Nematology 8: 367-376.

Zwart, J. A., N. Craig, P. T. Kelly, S. D. Sebestyen, C. T. Solomon, B. C. Weidel, and S. E. Jones. 2016. Metabolic and physiochemical responses to a whole-lake experimental increase in dissolved organic carbon in a north-temperate lake. Limnology and Oceanography 61: 723-734.

156

Figure captions

Figure 4.1. Left panels: Time series of dissolved oxygen (mg/L) and temperature (°C) at 3m depth over the course of the experiment (sampling took place between May and August of each year). 2011-12 were pre-manipulation years, with DOC concentration increasing in the treatment basin from 2013 onwards (marked by vertical black dashed lines). Right panels: The difference in average dissolved oxygen and temperature at 3m in the treatment basin compared to the reference (represented by the grey dashed line) over the course of the experiment.

Figure 4.2. Left panels: Time series of zoobenthos biomass (g-m2) at the whole lake level and at specific depths over the course of the experiment (sampling occurred between May and August of each year). The beginning of the manipulation is represented by the vertical dashed black lines, with DOC concentration increasing in the treatment basin from 2013 onwards. Right panels: The difference in average zoobenthos biomass in the treatment basin compared to the reference (represented by the grey dashed line) over the course of the experiment.

Figure 4.3. Average zoobenthos body size (mg dry mass ± standard error) for the whole lake and for each individual sampling depth over the course of the experiment. The DOC manipulation began between 2012 and 2013 after which the DOC concentration increased in the treatment basin relative to the reference.

Figure 4.4. Zoobenthos production (g-m2-y) at the whole lake level and at each individual sampling depth over the course of the experiment. The DOC manipulation began between 2012 and 2013 after which the DOC concentration increased in the treatment basin.

Figure 4.5. nMDS ordination of zoobenthos species composition at 3 m depth for one pre- (2012) and two post-manipulation years (2013-14). A convergent two-dimensional solution was found after the first try, with a stress of 0.02. Full species names, and orders, can be found in Appendix 4D.

157

Tables and Figures

Table 4.1. Lake level environmental variables, averaged between May and August, with standard deviations. Data from 2011-2014 are also summarized in Zwart et al. (2016). 2015 data, as well as all benthic primary production values, are newly summarized for this paper. Pre-manipulation data was recorded in 2011 and 2012, after this DOC concentrations were increased in the treatment basin compared to the reference. At the time of writing, the DOC samples from 2015 had not yet been analyzed, but we expect concentrations to be similar to 2013-14. Whole lake benthic primary production was not calculated in 2011 and in the treatment basin in 2012 due to insufficient data.

*The 2015 pelagic primary production data is missing 22 observations from the dates 21st June - 13th July and such may be a slight underestimate as this tends to be one of the more productive periods of the summer.

Dissolved Organic Thermocline Pelagic Primary Production Benthic Primary Production Year Carbon (mg/L) depth (m) (mg O2 L-1 d-1) (mg O2 m-1 d-1)

Reference Treatment Reference Treatment Reference Treatment Reference Treatment 2011 7.7 (1.8) 7.3 (1.5) 2.6 (0.2) 2.7 (0.0) 0.52 (0.37) 0.48 (0.35) - - 2012 7.7 (0.7) 7.9 (1.1) 2.4 (0.2) 2.3 (0.2) 0.72 (0.41) 0.69 (0.26) 73.1 (33.9) - 2013 7.0 (0.6) 11.4 (1.1) 2.1 (0.6) 1.8 (0.6) 0.42 (0.28) 0.65 (0.59) 107.4 (48.3) 79.3 (42.9) 2014 6.7 (0.7) 10.9 (1.3) 2.3 (0.8) 2.0 (0.8) 0.46 (0.26) 0.73 (0.5) 204.7 (88.7) 108.8 (71.6) 2015 - - 2.8 (0.5) 2.5 (0.5) 0.47 (0.33)* 0.60 (0.42)* 189.8 (75.0) 109.3 (38.4)

158

Table 4.2. Table of depth specific environmental variables, averaged between May and August, with standard deviations. Pre-manipulation data was recorded in 2011 and 2012, after this DOC concentrations were increased in the treatment basin compared to the reference.

* 2011 benthic primary production was estimated from core samples, while the remaining years were estimated through in situ benthic chambers (see methods).

Depth Dissolved Oxygen Benthic Primary Production Year Temperature (°C) (m) (mg/L) (mg O2 m-1 d-1)

Reference Treatment Reference Treatment Reference Treatment 2011 1 7.5 (2.6) 7.1 (1.1) 19.0 (4.4) 19.9 (3.4) 286.3 (39.0)* 258.5 (32.1)* 2012 1 7.5 (0.7) 7.6 (0.8) 21.3 (3.2) 21.8 (3.3) 314.3 (105.1) - 2013 1 7.9 (3.0) 7.3 (2.8) 18.3 (5.3) 18.8 (5.1) 317.5 (47.2) 157.5 (72.3) 2014 1 8.0 (1.7) 7.6 (1.3) 18.9 (4.2) 18.2 (4.2) 614.2 (51.1) 163.1 (152.8) 2015 1 7.4 (1.1) 7.4 (1.0) 19.6 (3.3) 19.6 (3.3) 550.4 (134.2) 308.2 (222.1) 2011 3 5.0 (2.5) 4.5 (3.1) 10.2 (2.4) 12.1 (2.9) - - 2012 3 8.4 (2.7) 7.1 (4.8) 15.2 (3.4) 14.5 (3.3) 0.0 (0.0) 0.0 (0.0) 2013 3 10.6 (2.8) 4.5 (4.0) 11.6 (3.3) 10.2 (2.7) 33.3 (51.6) 0.0 (0.0) 2014 3 12.1 (2.5) 3.0 (2.9) 11.4 (4.1) 9.8 (2.8) 5.0 (12.2) 0.0 (0.0) 2015 3 8.6 (2.3) 2.3 (2.3) 16.0 (3.8) 11.1 (2.8) 295.4 (144.7) 0.0 (0.0) 2011 8 3.4 (2.3) 4.3 (2.8) 4.9 (0.1) 4.6 (0.1) - - 2012 8 0.1 (0.1) 0.1 (0.1) 5.8 (0.2) 5.2 (0.2) - - 2013 8 0.4 (0.7) 0.6 (1.1) 4.7 (0.2) 4.5 (0.2) - - 2014 8 0.1 (0.0) 0.5 (1.1) 4.3 (0.2) 4.2 (0.2) - - 2015 8 0.2 (0.1) 0.3 (0.4) 5.1 (0.2) 5.3 (0.1) - -

159

Figure 4.1

160

Figure 4.2

161

Figure 4.3

162

Figure 4.4

163

Figure 4.5

treatment lim reference mol hir epipol 2012t 0.5 phr cla cry dic

s_dae 2014r mic.1mic 2013t beakan dja 2013r sia Axis2 0.0 ser gam bez gly phy nil 2014t lau

oli agr

chi -0.5 2012r end cae

-0.5 0.0 0.5 1.0

Axis 1

164

Appendix 4A.

Figure 4A. Table showing the number of days that the water at 3 m depth was hypoxic (<0.5 mg/L dissolved oxygen). Each column represents one day. We usually did not have measurements for the end of August and beginning of September. If conditions were similar between the last August and first September measurement, we assumed they were uniform in the intervening period. If conditions differed between these measurements, we entered a '?' and only included days that we were sure were hypoxic in the results. Therefore, we present minimum estimates of the number of days the 3 m zone was hypoxic. EL refers to the treatment basin, WL refers to the reference basin. The manipulation began after the 2012 measurements.

165

Appendix 4B.

Figure 4B. Average zoobenthos abundance (n per m2) for each individual sampling depth over the course of the experiment. The DOC manipulation began between 2012 and 2013 (represented by the black dashed line) after which the DOC concentration increased in the treatment basin relative to the reference.

166

Appendix 4C.

Figure 4C. Size-frequency plots of zoobenthos dry masses (mg) for each depth-year combination. The DOC manipulation began between 2012 and 2013 after which the DOC concentration increased in the treatment basin relative to the reference.

treatment - 2011 - 1m reference - 2011 - 1m

200 200

100 100

Frequency Frequency

0 0

0 1 2 3 4 0 1 2 3 4

treatment - 2012 - 1m reference - 2012 - 1m

200 200

100 100

Frequency Frequency

0 0

0 1 2 3 4 0 1 2 3 4

treatment - 2013 - 1m reference - 2013 - 1m

200 200

100 100

Frequency Frequency

0 0

0 1 2 3 4 0 1 2 3 4

treatment - 2014 - 1m reference - 2014 - 1m

200 200

100 100

Frequency Frequency

0 0

0 1 2 3 4 0 1 2 3 4

treatment - 2015 - 1m reference - 2015 - 1m

200 200

100 100

Frequency Frequency

0 0

0 1 2 3 4 0 1 2 3 4

Size (mg dry mass)

167

treatment - 2011 - 3m reference - 2011 - 3m

400 400

200 200

Frequency Frequency

0 0

0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0

treatment - 2012 - 3m reference - 2012 - 3m

400 400

200 200

Frequency Frequency

0 0

0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0

treatment - 2013 - 3m reference - 2013 - 3m

400 400

200 200

Frequency Frequency

0 0

0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0

treatment - 2014 - 3m reference - 2014 - 3m

400 400

200 200

Frequency Frequency

0 0

0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0

treatment - 2015 - 3m reference - 2015 - 3m

400 400

200 200

Frequency Frequency

0 0

0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0

Size (mg dry mass)

168

treatment - 2011 - 8m reference - 2011 - 8m

60 60

40 40

20 20

Frequency Frequency

0 0

0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 2.5

treatment - 2012 - 8m reference - 2012 - 8m

60 60

40 40

20 20

Frequency Frequency

0 0

0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 2.5

treatment - 2013 - 8m reference - 2013 - 8m

60 60

40 40

20 20

Frequency Frequency

0 0

0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 2.5

treatment - 2014 - 8m reference - 2014 - 8m

60 60

40 40

20 20

Frequency Frequency

0 0

0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 2.5

treatment - 2015 - 3m reference - 2015 - 3m

60 60

40 40

20 20

Frequency Frequency

0 0

0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 2.5

Size (mg dry mass)

169

Appendix 4D.

Table 4D. Taxa and abbreviations for nMDS ordination (Figure 4.5, main text).

Abbreviation Taxa Order Family Subfamily agr Agrypnia Trichoptera Phryganeidae - bea Beardius Diptera Chironomidae Chironominae bez Bezzia or Palpomyia Diptera Ceratopogonidae - cae Caenis Ephemeroptera Caenidae - chi Chironomus Diptera Chironomidae Chironominae cla Cladotanytarsus Diptera Chironomidae Chironominae cry Cryptochironomus Diptera Chironomidae Chironominae dic Dicrotendipes Diptera Chironomidae Chironominae dja Djalmabatista Diptera Chironomidae Tanypodinae end Endochironomus Diptera Chironomidae Chironominae epi Epitheca Odonata Corduliidae - gam Gammaridae Amphipoda Gammaridae - gly Glyptotendipes Diptera Chironomidae Chironominae hir Hirudinea Hirudinea - - kan Kaniwhaniwhanus Diptera Chironomidae Orthocladiinae lau Lauterborniella Diptera Chironomidae Chironominae lim Limnephilus Trichoptera Limnephilidae - mic Microtendipes Diptera Chironomidae Chironominae mol Molanna Trichoptera Molannidae - nil Nilobezzia Diptera Ceratopogonidae - oli Oligochaeta Oligochaeta - - phr Phryganeidae Trichoptera Phryganeidae - phy Phylocentropus Trichoptera Dipseudopsidae - pol Polycentropus Trichoptera Polycentropodidae - s_dae Sphaeriidae Veneroida Sphaeriidae - ser Serromyia Diptera Ceratopogonidae - sia Sialis Megaloptera Sialidae -

170

GENERAL CONCLUSIONS

DOC concentration has been increasingly recognised as a driver of productivity in low nutrient freshwaters over the past few decades (Prairie 2008). However, the mechanisms behind this effect on productivity are still not well understood, and more attention has been paid to basal productivity, rather than higher level consumers such as zoobenthos or fish (Wetzel 2001). DOC has been, and may continue to increase temporally in north-temperate freshwaters (Monteith et al. 2007, Weyhenmeyer et al. 2015) and so it is more important than ever to understand the effects it has on lake ecosystems. The outcomes of these thesis chapters indicate that the relationship between DOC and ecosystem productivity is not straightforward and so it is still difficult to predict how future increases in DOC may affect consumers in lakes. While productivity appears to always be relatively lower in DOC-rich systems (~ > 15 mg/L), there is much more variability at the lower end of the gradient in spatial observation surveys (Finstad et al. 2014, Kelly et al. 2014, Craig et al. 2015), and in the whole lake experiment from chapter 4. How an ecosystem may respond to an increase (or decrease) in DOC concentration may therefore depend on the starting point of the system along a DOC spectrum, as well as individual ecosystem characteristics such as size and nutrient status (Finstad et al. 2014, Solomon et al. 2015, Kelly et al. 2016).

I did however, pinpoint some potential mechanisms for the loss of productivity observed in high DOC systems. It is highly likely that DOC-induced reductions in thermocline depth, and thus habitat availability plays an important role in regulating ecosystem productivity in dark lakes. Zoobenthos had much lower productivity rates in the low oxygen environments in the hypolimnion, and their productivity in high DOC lakes was reduced as this type of habitat dominated in these systems. The same effect is likely for fish populations too as they are unlikely to spend much, or any, time in the hypoxic hypolimnion of these lakes, thus reducing the amount of habitat available.

While I did not find compelling direct evidence for a link between DOC concentration and fish feeding efficiency, I did observe an effect of DOC on fish life history strategies. The limitation in maximum body size a bluegill could attain in DOC-rich lakes indicates a potential

171 impact of resource limitation in later life. I also found that realized fecundity in these bluegill populations decreased along an increasing DOC gradient, and that fishing pressure had a potential additive effect on this relationship. All these factors combined resulted in low rates of intrinsic increase for populations in the darker lakes, indicating that these populations may be more susceptible to collapse at lower levels of fishing pressure compared to clearer lakes. This kind of information on the potential impacts of DOC on fish communities may be useful when designing management strategies for DOC-rich lakes, or those that are getting darker - particularly those that are near the potential threshold of DOC beyond which productivity is always low (Stasko et al. 2012).

Future directions

While this thesis has determined several mechanisms behind the effects of DOC on ecosystem productivity, there are also unanswered and new questions remaining. In many cases, it is difficult to understand the relationships between zoobenthos communities and environmental variables. This is often due to the difficulty in identifying zoobenthos to a fine enough taxonomic resolution (even Genus level can be challenging), and also a lack of information available on taxon specific lifecycles (Covich et al. 1999). Some detailed studies on the effects of DOC (and DOC-mediated shifts in temperature and dissolved oxygen) on the life cycles, and especially growth rates of specific taxa across gradients of space and time would be a first step in better understanding these relationships. In addition, both the zoobenthos survey, and the whole-lake experiment indicated that oxygen is a strong driver of zoobenthos production, but it is still unclear how extended periods of hypoxia - that become more prevalent at shallower depths as DOC increases - affects zoobenthos growth and survival. Jónasson (1984) showed that if a habitat is hypoxic for over two months, it is unlikely that most zoobenthos can survive, even with excess haemoglobin; the metalimnion hypoxic period in the whole-lake experiment reached 50 days in 2013, but we saw no decrease in abundance. A better understanding of this threshold would help us predict how increasing DOC may affect zoobenthos productivity in lakes with moderate DOC, that may be more susceptible to collapse.

172

We also were not able to fully determine the effects of DOC on fish feeding behaviour as we did not have appropriate data from adult fish. It is possible that adult fish are disproportionally affected by DOC-mediated resource limitation due to their larger sizes and energy requirements. More extensive diet studies over a DOC-gradient, including the entire size range of a given fish species would be needed to confirm or refute this speculation. Bioenergetics studies in this context would also be informative, as well as observing how fish diets may vary in the context of further whole-lake experiments.

The results of our whole-ecosystem experiment were not truly satisfying in that we could not fully determine why we observed the unexpected increase in zoobenthos biomass and productivity in the treatment basin. Part of the issue was the short timeframe of the experiment which meant we could not truly separate effects of the manipulation from potential lag effects of invertebrate life cycles. A longer term experiment would be useful in picking apart these effects. Another consideration would be a more dramatic DOC increase in a whole-lake experiment to test the threshold of productivity hypothesis posed in chapter 4. However, perhaps a more practical solution would be to manipulate DOC concentrations over time in several small pond experiments with a natural zoobenthos assemblage, reducing the costs associated with whole-ecosystem experimentation, but keeping some of the realism.

As DOC is intrinsically linked to other environmental variables such as nutrients, and temperature and dissolved oxygen depth gradients, some studies that focus on these interactions could be interesting. For example, in chapter 4 I suggested a potential non-linear relationship between DOC and productivity due to interacting factors of nutrients in low to moderate DOC lakes. This hypothesis could be tested through the use of mesocosms where both DOC and nutrients could be manipulated across treatments, with the response variable of consumer productivity. Some recent studies have also suggested that there may be cumulative effects of rising temperatures with DOC increases, which may have implications for oxygen dynamics, and thus suitable habitat availability for certain fish species that require cold, oxygenated waters, such as lake trout (Stasko et al. 2012, Couture et al. 2015). Determining these effects will be challenging, but will surely be possible through modelling, survey, mesocosm, and whole-ecosystem approaches.

173

References

Couture, R-M, H. A. de Wit, K. Tominaga, P. Kiuru, and I. Markelov. 2015. Oxygen dynamics in a boreal lake responds to long-term changes in climate, ice phenology, and DOC inputs. Journal of Geophysical Research: Biogeosciences 120. DOI: 10.1002/2015JG003065.

Covich, A. P., M. A. Palmer, and T. A. Crowl. 1999. The role of benthic invertebrate species in freshwater ecosystems. Zoobenthic species influence energy flows and nutrient cycling. Bioscience 49: 119-127.

Craig, N. C., S. E. Jones, B. C. Weidel, and C. T. Solomon. 2015. Habitat, not resource availability, limits consumer production in lake ecosystems. Limnology and Oceanography 60: 2079- 2089.

Finstad, A. G., I. P. Helland, O. Ugedal, T. Hesthagen, and D. O. Hessen. 2014. Unimodal response of fish yield to dissolved organic carbon. Ecology Letters 17: 36-43.

Jónasson, P. M. 1984. Oxygen demand and long term changes of profundal zoobenthos. Hydrobiologia 115: 121-126.

Kelly, P. T., C. T. Solomon, B. C. Weidel, and S. E. Jones. 2014. Terrestrial carbon is a resource, but not a subsidy, for lake zooplankton. Ecology 95: 1236-1242.

Kelly, P. T., N. Craig, C. T. Solomon, B. C. Weidel, J. A. Zwart, and S. E. Jones. 2016. Experimental whole-lake increase of dissolved organic carbon concentration produces unexpected increase in zooplankton biomass and production. Global Change Biology: doi: 10.1111/gcb.13260.

Monteith, D. T., J. L. Stoddard, C. D. Evans, H. A. de Wit, M. Forsius, T. Høgasen, A. Wilander, B. L. Skjelkvåle, D. S. Jeffries, J. Vuorenmaa, B. Keller, J. Kopácek, and J. Vesely. 2007. Dissolved organic carbon trends resulting from changes in atmospheric deposition chemistry. Nature 450: 537-541.

Solomon, C. T., S. E. Jones, B. C. Weidel, I. Buffam, M. L. Fork, J. Karlsson, S. Larsen, J. T. Lennon, J. S. Read, S. Sadro, and J. E. Saros. 2015. Ecosystem consequences of changing inputs of

174

terrestrial dissolved organic matter to lakes: current knowledge and future challenges. Ecosystems 18: 376-389.

Stasko, A. D., J. M. Gunn, and T. A. Johnston. 2012. Role of ambient light in structuring north- temperate fish communities: potential effects of increasing dissolved organic carbon concentration with a changing climate. Environmental Reviews 20: 173-190.

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

Weyhenmeyer, G. A., R. A. Müller, M. Norman, and L. J. Tranvik. 2015. Sensitivity of freshwaters to browning in response to future climate change. Climatic Change, 1-15. DOI: 10.1007/s10584-015-1514-z.

175