Unravelling microbial interactions in aquatic ecosystems ... · Unravelling microbial interactions...

Unravelling microbial interactions in aquatic

ecosystems: an improved model of

microbial controls on nutrient processing

Yu Li

B. Eng. - Shandong University of Science and Technology, China

B. Art. - Shandong University of Science and Technology, China

This thesis is presented for the degree of Doctor of Philosophy

at The University of Western Australia School of Earth and Environment

August 2013

iii

Abstract

In order to control algal blooms, it is necessary to better understand microbial

interactions in aquatic ecosystems. However, present ecological models, mainly based

on the simple ‘top-down’ (consumer) view of controlling algal blooms, do not always

provide an accurate picture of planktonic dynamics due to the complicated nature of

microbial interactions. This study has developed serial ecological models building on

the classic ‘Nutrient-Phytoplankton-Zooplankton-Detritus’ (NPZD) model to better

understand the significance of specific microbial interactions in aquatic ecosystems,

such as the microbial loop and the viral shunt. These interactions are relevant to

‘bottom-up’ (resource) control of algal blooms in aquatic ecosystems. Using Lake

Kinneret (Israel) as a study site, the significance of key microbial loop processes on

nutrient supply and stoichiometry is further examined by applying a one dimensional

coupled hydrodynamic-ecosystem model (DYRESM-CAEDYM) to a comprehensive

dataset (1997-2001).

In the first study, the potential significance two types of microbial interactions in

aquatic ecosystems have been theoretically explored. The improved serial models for

the microbial loop and the viral shunt illustrate the importance of ‘bottom-up’

(resource) control of algal blooms via these microbial interactions in aquatic

ecosystems.

In the second study, the relationship between phytoplankton internal nutrient

stoichiometry and water column N:P ratios has been investigated in a dynamic lake

environment. The results showed that the average internal N:P ratios of the

phytoplankton community followed the total carbon biomass seasonal patterns. The

seasonal patterns of the simulated dissolved inorganic N to total P (DIN:TP) ratios in

the water column were a useful indicator for reflecting the N:P stoichiometry of the

phytoplankton community and compared better than other indicators that were tested

including total N: total P (TN:TP) ratios and dissolved inorganic N to dissolved

inorganic P (DIN:DIP) ratios. However, the internal N:P ratio patterns of individual

phytoplankton groups did not always reflect the DIN:TP ratio patterns. The

stoichiometry of nutrient recycling pathways illustrated that the ability of bacteria to

regulate phytoplankton stoichiometry is a significant factor that has ecosystem wide

implications. The microbial loop has more considerable changes of the N:P ratios of the

iv

nutrient pools than the N:P ratios of the simulated phytoplankton groups, which further

indicate its importance in regulating the N:P stoichiometry of the nutrient fluxes

between bacteria, zooplankton, and inorganic and organic matters pools.

In the third study, the analysis of C:N:P stoichiometric variations demonstrated the

effect of bacterial competition for inorganic nutrients on the stoichiometry of

phytoplankton. In particular, bacterial competition with phytoplankton for inorganic

nutrients in the microbial loop plays a positive effect on phytoplankton primary

production rather than the traditional view of negative effect on primary production in

aquatic food webs.

In the fourth study, the microbial loop significantly affects phytoplankton growth rates

and succession patterns in Lake Kinneret. Dissolved organic phosphorous (DOP)

availability is critical in driving the microbial loop processes. When bacterial growth is

P limited, bacterial competition with phytoplankton for inorganic nutrients can switch

phytoplankton internal nutrient limitation, and thus change the predicted composition of

the phytoplankton community.

Overall, it is concluded that the microbial loop plays a crucial role in nutrient recycling

by regulating the quantity and stoichiometry of available nutrients. It is an important

model component that should be carefully parameterized when simulating

phytoplankton succession patterns and water quality dynamics in freshwater

ecosystems. This study improves the current understanding and management of

eutrophication and algal blooms by studying the microbial interactions and their role in

shaping the rates and pathways of nutrient cycling processes in aquatic ecosystems. The

improved model will provide an improved basis for water quality prediction and

ultimately help manage aquatic ecosystems in a changing climate.

v

Contents

Title page ................................................................................................................i

Abstract…............................................................................................................iii

Contents………....................................................................................................v

Figures…..............................................................................................................ix

Tables…..............................................................................................................xiii

Acknowledgements............................................................................................ xv

Statement of Candidate Contribution ............................................................xvii

Publications arising from this thesis ...............................................................xix

Chapter 1 Introduction..........................................................................................1

1.1 Background......................................................................................................1

1.2 Lake Kinneret (Israel) .............................................................................................. 3

1.3 Research objectives and approach...................................................................4

Chapter 2 The importance of model structural complexity when simulating

aquatic food webs- the case of Lake Kinneret (Israel)......................................7

2.1 Abstract............................................................................................................7

2.2 Introduction .............................................................................................................. 8

2.3 Methods.........................................................................................................10 2.3.1 Model structures........................................................................10

2.3.2 Model parameterisation............................................................12

2.3.3 Model setup...............................................................................17

2.3.4 Analysis approach.....................................................................18

2.4 Results and Discussion.................................................................................19

2.5 Conclusion ....................................................................................................28

Chapter 3 An analysis of the relationship between phytoplankton internal

stoichiometry and water column N:P ratios in a dynamic lake environment

................................................................................................................................31

3.1 Abstract ........................................................................................................31

3.2 Introduction ..................................................................................................32

vi

3.3 Methods........................................................................................................35

3.3.1 Study site...................................................................................35

3.3.2 Model overview........................................................................ 35

3.3.3 Validation approach................................................................ 46

3.3.4 Stoichiometric assessment....................................................... 47

3.4 Results .........................................................................................................49

3.4.1 Model performance..................................................................49

3.4.2 Temporal trends in N:P stoichiometry.................................... 58

3.4.3 Food web N:P stoichiometry ...................................................64

3.5Discussions...................................................................................................69 3.5.1 Model validation and nutrient ratios.......................................69

3.5.2 N:P stoichiometry of phytoplankton........................................71

3.5.3 Role of the microbial loop.......................................................72

Chapter 4 Bacterial competition with phytoplankton has a positive impact

on primary production of phytoplankton ………………………………….75

4.1 Abstract...................................................................................................... 75 4.2 Introduction................................................................................................ 76 4.3 Methods.......................................................................................................77

4.3.1 Study site..................................................................................77 4.3.2 Model overview........................................................................78 4.3.3 Bacterial sub-models...............................................................79 4.3.4 Model configuration................................................................80 4.3.5 Lake metabolism......................................................................80 4.3.6 Environmental factors.............................................................80

4.4 Results....................................................................................................... 82

4.4.1 Model evaluation.................................................................... 82 4.4.2 The effect of bacterial competition on primary production.....82 4.4.3 The impact of bacterial competition on ecological

stoichiometry of food web..................................................................83

4.4.4 The impact of bacterial competition on ecological

stoichiometry of phytoplankton.........................................................84

ix

vii

4.4.5 Lake metabolism.......................................................................87

4.4.6 Environmental factors.............................................................92

4.5 Discussion....................................................................................................96

Chapter 5 The role of the microbial loop in regulating nutrient availability

and phytoplankton dynamics.......................................................................... . 99

5.1 Abstract.........................................................................................................99 5.2 Introduction.................................................................................................100 5.3 Methods.......................................................................................................103

5.3.1 Site description.........................................................................103 5.3.2 Model overview and approach.................................................103 5.3.3 Analysis procedure...................................................................116

5.4 Results.........................................................................................................118

5.4.1 Comparison of model outputs..................................................118 5.4.2 Model parameter sensitivity analysis......................................122 5.4.3 Nutrient pools..........................................................................122 5.4.4 Nutrient fluxes.........................................................................126 5.4.5 Phytoplankton succession patterns.........................................129

5.5 Discussions.................................................................................................131

5.5.1 Model performance and sensitivity.........................................131 5.5.2 Role of the microbial loop in regulating nutrient flows.........133 5.5.3 Impact of the microbial loop on phytoplankton growth.........135

Chapter 6 Conclusions ................................................................................... 137

6.1 Summary of research findings…..................................................................137

6.2 Implications for water quality management…………..................................... 139

6.3 Recommendations for future work.............................................................140

References ........................................................................................................ 142

Appendixes ........................................................................................................ 154

ix

Figures

Figure 1.1 Bathymetric map for Lake Kinneret .............................................................. 5

Figure 2.1 Structure of NPZD (a), NPVD (b), NPZD+V (c), NPZD+B (d) and

NPZD+VB(e) models……………………. ...........................................................12

Figure 2.2 Light and temperature for Sim1 (a) and Sim2 (b) …...……….………….…18

Figure 2.3 Simulated phytoplankton of Sim1 (on the left) and Sim2 (on the right) with

different ecological models. ..................................................................................20

Figure 2.4 Simulated nutrient of Sim1 (on the left) and Sim2 (on the right) with

different ecological models. ..................................................................................20

Figure 2.5 Simulated detritus of Sim1 (on the left) and Sim2 (on the right) with different

ecological models. ................................................................................................21

Figure 2.6 Simulated zooplankton of Sim1 (on the left) and Sim2 (on the right) with

different ecological models. .................................................................................21

Figure 2.7 Simulated bacteria of Sim1 (on the left) and Sim2 (on the right) with

different ecological models. .................................................................................22

Figure 2.8 Simulated viruses of Sim1 (on the left) and Sim2 (on the right) with different

ecological models. ................................................................................................22

Figure 2.9 Ecosystem relationships of Sim1 (a) and Sim2 (b) of ecological models… 24

Figure 2.10 Summary of simulated nutrient fluxes (mmol/m3d-1) for NPZD (a) , NPVD

(b), NPZD+V(c), NPZD+B (d), and NPZD+VB(e) models.............................. 26

Figure 3.1 Conceptual diagrams outlining the configured microbial groups and

interactions in the Lake Kinneret DYRESM-CAEDYM model for the Microbial

Loop Absent Scenario, MLAS, (a) and the Microbial Loop Present Scenario,

MLPS, (b) configurations .....................................................................................46

Figure 3.2(a) Validation of nutrient variables in the surface water ...............................50

Figure 3.2(b) Validation of nutrient variables in the bottom water ...............................51

Figure 3.2(c) Validation of phytoplankton variables in the surface water .....................52

Figure 3.2(d) Validation of heterotrophic organism variables in the surface water....... 53

Figure 3.3 Simulated vs. observed monthly averaged time-series of a) DIN:DIP

x

ratios, b) DIN:TP ratios and c) TN:TP ratios in the surface water .........................56

Figure 3.4 Simulated vs. observed monthly averaged time-series of a) DIN:DIP

ratios, b) DIN:TP ratios and c) TN:TP ratios in the bottom water........................ 57

Figure 3.5 Comparison between the simulated C biomass and iN:iP ratios of

the combined phytoplankton community ..............................................................58

Figure 3.6 Comparison of the simulated water column DIN:TP ratios with a) the

simulated iN:iP ratios of the combined phytoplankton community, b) the bulk

nutrient uptake N:P stoichiometry, and c) the bulk excretion N:P stoichiometry ..

.................................... .........................................................................................59

Figure 3.7 Comparison of the simulated water column DIN:TP ratios with the iN:iP

ratios of a) Peridinium, b) Microcystis, c) Aphanizomenon, d)

nanophytoplankton, and e) Aulacoseira .............................................................. 63

Figure 3.8 Comparison of the simulated water column DIN:TP ratios with the iN:iP

ratios of heterotrophic organisms ...........................................................................64

Figure 3.9 Comparison of simulated average molar N:P stoichiometry of phytoplankton,

heterotrophic organisms and nutrient pools of the water column between the a)

microbial loop absent (MLAS) and b) microbial loop present (MLPS) simulations

.................................................................................................................................66

Figure 3.10 Frequency histograms of iN:iP ratios for a) the combined phytoplankton

community, b) Peridinium, c) Microcystis, d) Aphanizomenon, e)

nanophytoplankton, and f) Aulacoseira ...............................................................68

Figure 4.1 The influence of the microbial loop on phytoplankton……… ...................78

Figure 4.2 Conceptual diagram highlighting differences between B-N and B+N…... . 79

Figure 4.3 (a) Comparison of simulated C biomass of the combined phytoplankton

community and individual phytoplankton groups between B-N and B+N from

1997 to 2001; (b) Comparison of simulated C biomass of bacteria (BAC),

microzooplankton (ZOOP3), and seston between B-N and B+N from 1997 to

2001.............................................................................................................……83

Figure 4.4 Linear regression of simulated iC:iN:iP ratios of phytoplankton in B-N and

B+N.................................................................................................................... 85

Figure 4.5 The simulated iC:iN:iP of the phytoplankton community .........................87

Figure 4.6 Simulated primary production of different phytoplankton groups: (a)

Peridinium, (b) Microcystis, (c) Aphanizomenon ……………………………. 89

xi

Figure 4.7 Respiration of different phytoplankton groups: (a) Peridinium, (b)

Microcystis, (c) Aphanizomenon.......................................................................... 91

Figure 4.8 The emergent property of the simulated iC:iN:iP ratios of three

phytoplankton groups:(a) Peridinium, (b) Microcystis, (c) Aphanizomenon, in

response to environmental factors (light, temperature, N and

P)........................................................................................................................... 95

Figure 5.1Conceptual diagram highlighting the general ecosystem model configuration

for Lake Kinneret (top) and processes and feedbacks for the three microbial loop

models (bottom) explored in this study: (1) NOBAC (mineralization is not

dependent on the bacterial biomass), (2) BAC-DIM (bacteria only take up DOM),

and (3) BAC+DIM (bacteria not only take up DOM but also DIM) in with the

aquatic ecological model CAEDYM .................................................................. 107

Figure 5.2 Comparison of model simulations for a) nutrient variables in the surface 10m

(left) and bottom 10m (right) of the water column, and b) for the nine microbial

groups...................................................................................................................119

Figure 5.3 Sensitivity analysis of state variables and process rates for the C, N and P

cycles presented as the lake average absolute change after a +/-20% parameter

shift. .................................................................................................................... 124

Figure 5.4 Summary of a) NOBAC, b) BAC-DIM, and c) BAC+DIM (C pathways-

black values; N pathways-red values; P pathways-blue values), presented as the

lakewide average flux rates in brackets (×10-5mg L-1d-1). ................................. 128

Figure 5.5 Figure 5.5 Comparison of nutrient limitation functions fa(N) and fa(P)

defined in eqns (15) and (16) in the section 5.3.3.2 respectively for the five

simulated phytoplankton groups in a) NOBAC, b) BAC-DIM and c)

BAC+DIM................................................................................... ....................... 130

xiii

Tables

Table 2.1 The key plankton variables with the number of simulated groups in ecological

models. …………….…………………………………………………………12

Table 2.2 Comparison of the NPVD model, the NPZD+V model, the NPZD+B model,

and the NPZD+VB model. ………………………..…………………………15

Table 2.3 The parameters of the five ecological models…………………...……....16

Table 2.4 Summary of simulated annual average variables (mmol/m3) for ecological

models. …………………...…………………………………………….….....27

Table 2.5 Summary of simulated nutrient fluxes (mmol/m3d-1) for ecological models.

………………………………………………………………………….….....27

Table 3.1 List of biogeochemical and biological variables in DYRESM-CAEDYM.

………………..……………………………………………………………...39

Table 3.2 List of phytoplankton parameters used in DYRESM-CAEDYM

simulations of Lake Kinneret. .............………………..………… … ….…..41

Table 3.3 The equations for bacteria and microzooplankton in the two microbial

loop scenarios ............................................ ……………………………..…..43

Table 3. 4 Statistical comparison between model simulations and observed data in

the surface water.......................................................................………………54

Table 3.5 The impact of seasonality on the relationship between phytoplankton internal

nutrient ratios and DIN:TP ratios ……………………………………........... 61

Table 4.1 Stoichiometric comparison between B-N and B+N.................................. 86

Table 4.2 The Spearman rank correlation coefficients (Rs) between simulated iC:iN:iP

ratios of phytoplankton and lake metabolism processes................................. 91

Table 4.3 The Spearman rank coefficients (Rs) between simulated iC:iN:iP ratios of

phytoplankton and environmental factors....................................................... 92

Table 5.1 Overview of the variables configured with DYRESM-CAEDYM for Lake

Kinneret............................................................................................................. 105

Table 5.2 Equations for C, N and P within nutrients, organic matter, bacteria and

zooplankton pools............................................................................................. 109

Table 5.3 Microbial loop related parameters used in the three model simulations.... 111

xiv

Table 5.4 Statistical analysis of water quality variables comparing the three microbial

loop configurations by ANOVA and Multiple Comparisons.. .........................121

Table 5.5 Summary of average values (1997-2001) for C, N, and P contents (mg L-1)

and N:P molar ratios of the various food web components in different microbial

loop configurations……………………………………………………….…...125

vii

xv

Acknowledgements

I would like to gratefully thank my coordinating supervisor, Associate Professor

Matt Hipsey for introducing me into the ecological modelling area with his

enthusiasm. I would like also thank for my co-supervisor Winthrop Professor Anya

Waite, for her high level of professionalism as researchers.

I gratefully acknowledge the use of the models DYRESM and CAEDYM developed

the Centre for Water Research, University of Western Australia. We also thank the

Kinneret Limnological Laboratory (KLL) for making the field data available to us.

I benefitted enormously from the guidance, support and comments of Dr. Krys Haq,

Dr. Jo Edmondston, Dr. Michael Azariadis, and Associate Professor Andrew Rate,

especially at the most difficult times during my doctoral studies in Australia. I

appreciated their insightful comments and academic writing support.

I also acknowledge funding for my PhD study from China Scholarship Council

(CSC) and the other financial supports, the Grants for Research Student Training

in the University of Western Australia (GRST), the Gordon & Betty Moore

Foundation Award #1182, the U.S. National Science Foundation Grant DBI-

0639229 and DBI-0446017, the U.S. National Science Foundation Cyber-enabled

and Innovation (CDI) Program Award #941510, Australian Society of Limnology,

and the Postgraduate Students' Association Research Training and Development

Award in the University of Western Australia.

I am grateful for additional advice on my papers by Gideon Gal and Vardit Markler-

Pick on Lake Kinneret. I thank for selfless academic support from Prof. Robert

Sterner, Prof. David Hamilton, Prof. Paul Hanson, Prof. Lin Wang, Associate Prof.

Trina Machmon, Dr. Jason Antennucci, and Dr. Andrea Paparini. I would also like to

thank Zhenlin Zhang, Grace Hong, Emily Kara Read, Cayelan Carey, Dan Paraska,

Andrew Ong, and Ana laura Ruibal Conti for the friendship and discussions during

my PhD study, especially writing my PhD thesis.

xvi

Many thanks to my family who are always encouraging me and supporting my PhD

study in Australia. I appreciate their love and selfless support for my study and

love forever!

xvii

Statement of Candidate Contribution

The main body of this thesis is comprised of four research chapters, Chapters 2-5:

Chapter 2 is under review by MODSIM 2013, which will be prepared for

Environmental Modelling & Software;

Chapter 3 has been published in Ecological Modelling, 2013(252):196-213.

Chapter 4 is under review by Limnologica;

Chapter 5 has been accepted by Biogeosciences (BGD).

The content of this thesis is the author’s own work except where referenced and specific

acknowledgements are included in this thesis. Associate Professor Matthew Hipsey and

Winthrop Professor Anya Waite are my supervisors who have contributed a lot to

review and discussion about my PhD thesis. Specifically, in Chapter 2, I have obtained

modelling technical support and editorial help from Matt Hipsey and Anya Waite. In

Chapter 3, Chapter 4, and Chapter 5, the DYCD model was originally developed by

Matt Hipsey and Gideon Gal, which has been published in Gal et al. (2009). Based on

their work, I have modified and verified the model for ecological stoichiometry. In

Chapter 3, I have obtained editorial help from Matt Hipsey, Anya Waite, and Gideon

Gal. In Chapter 4, I have obtained editorial help from Vardit Makler-Pick and Emily

Read. In Chapter 5, I have obtained editorial help from Matt Hipsey, Vardit Makler-

Pick, Anya Waite, and Gideon Gal. In particular, Figure 5.1 was prepared by Matt

Hipsey and the Matlab scripts for Figure 5.2 were originally prepared by Vardit Makler-

Pick. For different papers, there are different co-authors listed in Publications.

Yu Li Dr. Matthew R. Hipsey

xix

Publications arising from this thesis

Yu LI, Matthew HIPSEY. The importance of model structural complexity when

simulating aquatic food webs. Under review, MODSIM2013, 20th International

Congress on Modelling and Simulation, 2013. (Chapter Two)

Yu LI, Vardit MAKLER-PICK, Emily READ, Gideon GAL, Matthew HIPSEY.

Bacterial competition with phytoplankton has a positive impact on primary

production of phytoplankton. Under review, Limnologica, 2013. (Chapter Four)

Yu LI, Gideon GAL, Vardit Makler-Pick, Anya WAITE, Louise BRUCE,

Matthew HIPSEY. The importance of the microbial loop in shaping

phytoplankton succession: A numerical analysis of Lake Kinneret, Israel.

Accepted, Biogeosciences(BGD), 2013. (Chapter Five)

Yu LI, Gideon GAL, Anya WAITE, Matthew HIPSEY. An analysis of the

relationship between phytoplankton internal stoichiometry and water column

N:P ratios in a dynamic lake environment. Ecological Modelling,

2013(252):196-213. (Chapter Three)

Yu LI, Lin WANG, Matthew HIPSEY. Ecological Stoichiometry of Microbial

Interactions in Aquatic Ecosystems. Proceedings of the 6th National Ph.D.

Candidates Academic Conference New Theories and New Technologies in

Environmental Science and Engineering. Beijing, China, October 2012, 6004.

(Chapter Two)

Yu LI, Vardit MAKLER-PICK, Gideon GAL, Anya WAITE, Matthew

HIPSEY. Exploring the microbial loop paradox with ecological stoichiometric

approach. In: Helmut Mader and Julia Kraml (eds) 9th International

Symposium on Ecohydraulics 2012 Proceedings. Vienna, Austria, September

2012, ISBN: 978-3-200-02862-3, 15391_2. (Chapter Four)

Yu LI, Anya WAITE, Gideon GAL, Matthew HIPSEY. Do phytoplankton

nutrient ratios reflect patterns of water column nutrient ratios? A numerical

stoichiometric analysis of Lake Kinneret. Procedia Environmental Sciences,

2012 (13):1630-1640. (Chapter Three)

xx

Yu LI, Gideon GAL, Anya WAITE, Matthew HIPSEY. Microbial loop

processes shape the food web stoichiometry in Lake Kinneret. In: Chan, F.,

Marinova, D. and Anderssen, R.S. (eds) MODSIM2011, 19th International

Congress on Modelling and Simulation. Modelling and Simulation Society of

Australia and New Zealand, December 2011 ISBN: 978-0-9872143-1-7, pp

3726-3732. (Chapter Three)

Yu LI, Matthew HIPSEY, Anya WAITE. Stoichiometric Modelling the impact

of microbial loop on patterns of phytoplankton community in Lake Kinneret. In.

The joint Australian Society for Limnology & New Zealand Freshwater Sciences

Society Congress: 26 September-30September 2011, Brisbane Convention &

Exhibition Centre, 50th Annual Congress, Delegate Handbook”. September 2011

ISSN: 1326-1142, pp 50. (Abstract)

1

1 Introduction

1.1 Background

With increasing human activities, large quantities of nutrients are mobilized from point

and diffuse sources into water bodies, which results in accelerated anthropogenic

eutrophication and nuisance algal blooms in aquatic ecosystems (Carpenter et al., 1998;

Rabalais et al., 2002; Smith, 2003; Bennett, 2003). As aquatic ecosystems play an

important role in supporting economic, recreational and ecological aspects of a

sustainable society (Hudnell, 2010), the water quality protection has become a popular

topic of public interest (Carpenter et al., 1999).

In the past decades, researchers have made great progress in understanding the

fundamental mechanisms of carbon (C) and nutrient cycles in aquatic ecosystems

(Thomson, 1998; Chan et al., 2001; Hellweger et al., 2008). Within this context, much

work has been conducted on the algal-based food web, that is, the classic ‘N-P-Z-D’

(Nutrients-Phytoplankton-Zooplankton-Detritus) paradigm, which assumes a relatively

simple flow of C and nutrients in biogeochemical cycles. However, it is now well

documented that the detrital-based food web can also influence the key C and nutrient

cycling processes in aquatic ecosystems, which mediates phytoplankton growth and

their succession patterns (Moore et al., 2004), such as the microbial loop and the viral

shunt. The microbial loop refers to the dynamics of the heterotrophic bacteria and

microzooplankton. As the key component of the microbial loop, bacteria regulate the

pelagic C, N and P recycling processes (Daufresne et al., 2001). This has been proven to

2

play an important role in shaping C fluxes in aquatic ecosystems (Stone et al., 1993;

Berman et al., 2010), and in enhancing nutrient cycling at the base of food webs (Hart et

al., 2000; Hambright et al., 2007). Furthermore, bacterial interactions with

phytoplankton can potentially influence the species composition of the plankton

community in aquatic ecosystems, which depends on environmental conditions and

stoichiometric imbalance (Cotner et al., 2002).

In addition to microbial loop processes, the viral shunt is another important microbial

interaction that can also play an important, but generally unquantified role in the flux

pathways of C and nutrients through planktonic systems (Suttle, 2005). The viral shunt

refers to the movement of C and nutrients from organisms to the dissolved organic

matter (DOM) and the particulate organic matter (POM) pools catalyzed by viral

infection and lysis (Suttle, 2005). Viruses infect both phytoplankton and bacteria, and

lyse their hosts to release DOM and POM, which affect nutrient availability and flux

pathways of C, N and P recycling processes in aquatic ecosystems.

Ecological stoichiometry has recently emerged as an underlying theory with an

improved mechanistic basis to study microbial interactions in aquatic ecosystems (Elser

et al., 2012). In 1958, Redfield’s work in the world’s oceans during the mid-1900s

pointed towards an almost ‘universal’ C: N: P ratio of marine seston for describing an

ecosystem by its stoichiometry (Redfield, 1958). This led to the assumption that the

biota had evolved to have a similar elemental composition to their aquatic medium, and

therefore that C, N and P cycles congruently throughout pelagic ecosystems. It is now

well known that the C: N: P ratios of the different heterotrophic and autotrophic aquatic

organisms need not conform to the Redfield’s fixed ratio (106:16:1). The field of

‘ecological stoichiometry’ has since developed and it is now recognized that organisms

simultaneously need a full suite of elements and the availability of one element may

control or influence the recycling processes of the others (Sterner and Elser, 2002; Li et

al., 2012). Stoichiometric analysis is the key for interpreting trophic interactions in

biogeochemical cycles when dynamic nutrient limitation occurs. Due to ecosystem

complexity, non-linear interactions and feedbacks between different elements, a system-

level approach is required to unravel the important physical, chemical, and (micro-)

biological processes occurring at a large ecosystem scale.

Modelling microbial interactions in aquatic ecosystems holds great promise for

understanding the C and nutrient flux pathways between different plankton groups

3

(Romero et al., 2004; León et al., 2005; Guven et al., 2006; Arhonditsis et al., 2007;

Burger et al., 2008). Some aquatic ecological models have been successfully applied to

understand the C, N and P cycles in the water column of aquatic ecosystems (Romero et

al., 2004; Bruce et al., 2006; Hipsey et al., 2008; Ng et al., 2011). Yet to date many

models over simplify the complexity of microbial population dynamics and the

corresponding nutrient fluxes. Even though bacteria and phytoplankton play important

roles in mediating nutrient flux pathways, as outlined above, few models parameterize

them as individual functional groups (Arhonditsis et al., 2006) and there has been little

research undertaken in understanding the role that the microbial loop and the viral shunt

processes play in shaping phytoplankton succession patterns in a dynamic aquatic

environment. It has been argued that a new conceptual modelling framework with an

improved mechanistic basis is therefore required for studying the microbial interactions

in aquatic ecosystems (Mooij et al., 2010; Trolle et al., 2012).

Coupled hydrodynamic-ecological models are useful tools to simulate the complicated

microbial processes in aquatic ecosystems and study how they interact with

environmental conditions (León et al., 2005; Arhonditsis et al., 2007; Burger et al.,

2008). Validation of these models is an effective approach to limit the range of

biogeochemical parameters and assess ecological representations in complex aquatic

ecosystems (Romero et al., 2004; Hipsey et al., 2008). However, complex ecosystem

models are generally largely over-parameterized. New approaches for validation and

sensitivity testing are also required, particularly within the context of microbial

interactions.

This research focus on addressing these challenges, by firstly developing serial

theoretical ecological models based on the classic NPZD model, to better depict

microbial interactions in aquatic ecosystems and understand how model complexity

impacts predictions ofplankton dynamics. The research then focuses on using a coupled

hydrodynamic-aquatic ecological model to link the biochemical composition of

bacterial and algal populations with dynamic variability in the ratios of C, N and P

elements. The model development conducted within this research has all been based on

Lake Kinneret (Israel), which serves as the visual laboratory to better understand these

ideas due to long history of experimental research and access to substantial ecological

data.

4

1.2 Lake Kinneret (Israel)

Lake Kinneret (Sea of Galilee) is a large monomictic lake located in the Syrian-African

Rift Valley in north-eastern Israel (Zohary et al., 1998; Parparov and Gal, 2012), which

covers an area of 170 km2, is 21 km long and 16 km wide and has a maximum depth of

43m (Figure 1.1). The lake is of critical importance to Israel since it supplies about one

third of the country’s drinking water. The deterioration of water quality in Lake

Kinneret has been the focus of considerable limnological research over the past few

decades (Berman et al., 1995; Zohary, 2004a; Roelke et al., 2007; Gal et al., 2009). As a

meso-eutrophic lake with annual primary production of approximately 650 gC m−2

(Berman et al. 1995), the lake was well known for the once regular occurrences of the

dinoflagellate Peridinium gatunese (Zohary et al., 1998). Because of a number of

significant changes since the mid-1990s (Gal and Anderson, 2010), its historically

stable phytoplankton assemblage was observed to be disrupted, and frequent

occurrences of nuisance cyanobacterial species have become a concern from a water

quality management perspective (Zohary, 2004; Zohary and Ostrovsky, 2011).

In order to simulate nutrient-phytoplankton dynamics to support decision-making for

water quality management in Lake Kinneret, the improvement of ecological model

application is required. A hydrodynamic model (Dynamic Reservoir Simulation Model,

DYRESM) coupled to an aquatic ecological model (Computational Aquatic Ecosystem

Dynamics Model, CAEDYM) has been successfully applied to identify the dominant

fate processes of the C, N and P cycles in the water column of Lake Kinneret (Bruce et

al., 2006). However, the model presented by Bruce et al. (2006) had a simplistic

representation of the microbial loop dynamics, and two important cyanobacterial

species, Microcystis sp. and Aphanizonmimen sp., were also not included within the

simulation, but continue to remain a concern to the overall health of the ecosystem

(Zohary, 2004). Gal et al. (2009) expanded this model to include a dynamic microbial

loop parameterization within Lake Kinneret. Following Bruce et al. (2006) and Gal et

al. (2009)’s modelling work, this study further explores the Lake Kinneret ecosystem to

study in details the microbial interactions with an ecological stoichiometric approach.

5

Figure 1.1: Bathymetric map for Lake Kinneret (Station A represents the sampling point where the observed data

were collected).

1.3 Research objectives and approach

The thesis is structured to sequentially develop the understanding of the importance of

microbial interactions in an aquatic environment.

Chapter 2 theoretically explores the microbial loop and the viral shunt in aquatic

ecosystems. The ‘Nutrients-Phytoplankton-Viruses-Detritus’ (NPVD) model has been

developed to compare the influence of zooplankton grazing mortality and viral induced

mortality on phytoplankton. The ‘Nutrients-Phytoplankton-Zooplankton-Detritus

+Viruses’ (NPZD+V) model and the ‘Nutrients-Phytoplankton-Zooplankton-Detritus

+Bacteria’ (NPZD+B) model have been respectively developed for describing the

microbial loop and the viral shunt in aquatic ecosystems. The ‘Nutrients-Phytoplankton-

Zooplankton-Detritus+Viruses+Bacteria’ (NPZD+VB) model has been further

developed for investigating how the viral shunt short-circuits the microbial loop in the

microbial food web.

Chapter 3 presents an analysis of two microbial loop configurations and assesses how

the internal nutrient ratios (N:P ratios) of several phytoplankton functional groups relate

to nutrient ratios within the water column, and examines if the microbial loop shapes the

stoichiometry of food web components in Lake Kinneret. This study explores how the

internal N:P stoichiometry of phytoplankton groups responds to variable patterns of

nutrient supply within a dynamic aquatic environment, as the assumption that

phytoplankton internal N:P stoichiometry matches the bulk properties of the water

6

column may not always be accurate. Although several types of N:P ratios have been

used to understand the nutrient limitation of phytoplankton, here the hypothesis has

been proposed that interactions between different ecosystem components result in

patterns of phytoplankton stoichiometry that are independent of bulk water column

indicators.

Chapter 4 presents two bacterial nutrient uptake sub-models for examining the impact

of bacterial uptake of inorganic nutrients on the internal C:N:P (iC:iN:iP) stoichiometry

of the phytoplankton community and detrital pools. When phytoplankton and bacteria

compete for the same limiting inorganic nutrients, it results in an increase of

phytoplankton biomass instead of a decline in biomass. While this seems paradoxical,

very little research has been directed towards resolving this paradox. This chapter re-

examines this paradox using ecological stoichiometric principles. It tests whether

bacterial competition with phytoplankton for inorganic nutrients has a positive effect on

the primary production of the phytoplankton community. The response of different

phytoplankton species to environmental factors (e.g., light, temperature, and nutrients)

has been explored individually for unraveling this phenomenon.

Chapter 5 compares the above three microbial loop sub-model configurations to

understand the mechanisms by which the microbial loop influences phytoplankton

succession patterns in Lake Kinneret. The hypothesis has been proposed that inclusion

of the microbial loop in a numerical model not only impacts our ability to directly

model the role of zooplankton and bacteria in lake ecosystems, but also impacts our

ability to simulate the ratios of inorganic nutrients available to primary producers, and

predict algal succession patterns. This chapter fully illustrates how microbial loop

processes regulate the nutrient fluxes between different groups of bacteria,

phytoplankton and zooplankton via nutrient recycling pathways and how these

processes shape the phytoplankton succession patterns in a freshwater ecosystem.

Chapter 6 summarises the main findings and points out what this thesis contributes to

the general use of biogeochemical models in studying microbial interactions in aquatic

systems. Based on an improved theoretical framework from this study, the outcomes are

an overall better understanding of the mechanisms that control variability in the flow

and recycling of C and nutrients, but also provide kinetic parameters for modelling

microbial community dynamics and the relevant C and nutrient transformation

processes.

7

2 The importance of model structural complexity when simulating aquatic food webs - the case of Lake Kinneret (Israel)

2.1 Abstract

With increasing occurrence of algal blooms in aquatic ecosystems, more and more

ecological models have been developed for depicting and forecasting algal blooms.

However, these ecological models often simplify microbial diversity and do not always

provide an accurate picture of the nutrient flux pathways that occur in food webs due to

the complicated nature of microbial interactions. This study developed several

ecological models by building on the classic ‘Nutrient-Phytoplankton-Zooplankton-

Detritus’ (NPZD) model to better understand the significance of specific microbial

interactions in aquatic ecosystems. The ‘Nutrient-Phytoplankton-Viruses-Detritus’

(NPVD) model has been developed to compare the influence of zooplankton mediated

mortality and virus mediated mortality on phytoplankton. The results showed that virus

mediated mortality on phytoplankton via infection and lysis of phytoplankton is as

important as zooplankton mediated mortality via grazing on phytoplankton. The results

of ‘Nutrient-Phytoplankton-Zooplankton-Detritus+Viruses’ (NPZD+V) model for the

viral shunt indicate that viruses catalyse the movement of nutrients from phytoplankton

to detritus. The results of the ‘Nutrient-Phytoplankton-Zooplankton-Detritus+Bacteria’

8

(NPZD+B) model for the microbial loop indicate the positive impact of the microbial

loop on phytoplankton growth in aquatic ecosystems. The results of the ‘Nutrient-

Phytoplankton-Zooplankton-Detritus+Viruses+Bacteria’ (NPZD+VB) model indicate

that the viral shunt short circuits the microbial loop via viral infection and lysis of

phytoplankton and bacteria, and thereby increased the transfer of C and nutrients to

detritus. Furthermore, these improved serial NPZD+V model for the viral shunt, the

NPZD+B model for the microbial loop, and the NPZD+VB model for the viral short

circuit of the microbial loop illustrate the importance of ‘bottom-up’ (resource) control

of algal blooms via microbial interactions in aquatic ecosystems. These results help

provide an improved mechanistic understanding for viral-bacterial-phytoplankton-

zooplankton interactions in aquatic ecosystems to control algal blooms for protecting

water quality.

2.2 Introduction

With increasing human activities, the global problem of the accelerated eutrophication

of water bodies (e.g. algal blooms) has caused the public attention over past several

decades (Anderson et al., 2002). Modelling microbial interactions in aquatic ecosystems

is a useful tool for understanding the eutrophication processes that lead to algal blooms

by describing the nutrient flux pathways in the food web (Guven et al., 2006). Present

models usually focus on the modified prey-predator models (also known as Lotka-

Volterra equations), the classic ‘Nutrient-Phytoplankton-Zooplankton’ (NPZ) models

and ‘Nutrient-Phytoplankton-Zooplankton-Detritus’ (NPZD) models. These models

have been successfully developed and applied widely for aquatic ecosystem research

(eg. Popova et al. 2000; Edwards, 2001; Franks, 2002). However, these models assume

a relatively simple flow of nutrients between autotrophic and heterotropic pools

(Downing et al., 2001; Edwards et al., 2001). Therefore, they do not capture the

complexity of microbial food web interactions accurately.

To unravel the complexity in which nutrients move among microbial food webs, two

types of microbial interactions have been defined as the microbial loop and the viral

shunt (Hart et al., 2000; Suttle, 2005; Hambright et al., 2007). The microbial loop refers

to the dynamics of the heterotrophic bacteria and the microzooplankton grazers, which

9

is also known as the detrital-based food web (Moore et al., 2004). Moreover, the role of

viruses (V) in biogeochemical cycles has been well documented in aquatic ecosystems,

which challenges the traditional views of aquatic food webs (Wommack and Colwell,

2000; Clasen et al., 2008). The nutrients released by viral lysis are usually organically

bound, which affect nutrient availability and flux pathways of C, N and P cycling

processes (Gobler et al., 1997). In particular, viruses catalyse the movement of nutrients

from phytoplankton to detritus, termed the ‘viral shunt’ (Suttle, 2005). As 25% of the

primary production in the ocean ultimately follows through the viral shunt (Wilhelm

and Suttle, 1999), it is crucial to accurately quantify the role of the microbial

interactions, and incorporate it into biogeochemical models (Suttle, 2007).

As the microbial loop and the viral shunt are ‘bottom-up’ (resource) control processes,

the traditional models based more on ‘top-down’ (consumer) control of algal blooms

miss some nutrient flows between viruses, bacteria, phytoplankton and zooplankton.

These may respond non-linearly to environmental changes. Many ecological models

have begun to simulate plankton population dynamics and the corresponding nutrient

fluxes. In most cases, microbial interactions are parameterised using empirical or semi-

empirical relationships. However, these empirical parameters are not universal or are

often site-specific. A new conceptual modelling framework with an improved

mechanistic basis for studying microbial interactions in aquatic ecosystems is therefore

required (Arhonditsis and Brett, 2004; Mooij et al., 2010).

In this chapter, some exploratory work is conducted to understand the significance of

how our model conceptualisations impacts on water quality predictions. The analysis

starts with the development of a ‘Nutrients-Phytoplankton-Viruses-Detritus’ (NPVD)

model to compare the relative significance of zooplankton mediated mortality and virus

mediated mortality on phytoplankton. Furthermore, the NPZD+V model and the

NPZD+B model have been developed for the microbial loop and the viral shunt, and

their relation to the other compartments of the food web in aquatic ecosystems. The

improved NPZD+VB model allows us to further assess the impact of ‘viral shunt short

circuiting’ of the microbial loop through nutrient cycling processes. We compared the

simulated results with several biological variables and their ecosystem relationships

under different models to determine the impact of how microbial interactions may drive

changes in the food webs in aquatic ecosystems.

10

2.3 Methods

2.3.1 Model structures

Based on the NPZD model, the NPVD model was developed for virus mediated

mortality on phytoplankton. Furthermore, the NPZD+V model combined viruses and

zooplankton compartments to investigate the impact of the viral shunt on the food web

in aquatic ecosystems. The NPZD+B considered the role of bacteria to investigate the

impact of the microbial loop in aquatic ecosystems. Finally, the improved NPZD+VB

model combined viruses, bacteria, and zooplankton to investigate the impact of viral

shunt on the microbial loop. Their model structures were described in Figure 2.1. These

model compartments were assumed spatially homogeneous. The arrows indicated the

flows of nitrogen between different model compartments (Table 2.1).

NPZD: The original ‘Nutrient-Phytoplankton-Zooplankton’ (NPZ) models are universal

research tools in oceanography because they incorporate one of the simplest set of

oceanic plankton dynamics (Franks, 2002). Edwards (2001) has added the detritus

component into the simple NPZ models to form the four compartment model ‘Nutrient-

Phytoplankton-Zooplankton-Detritus’ (NPZD). This model consists of seven processes,

especially remineralisation (Figure 2.1a).

NPVD: The development of the ‘Nutrients-Phytoplankton-Viruses-Detritus’ (NPVD)

model was based on the traditional NPZD model to compare the influence of

zooplankton mediated mortality and virus mediated mortality on phytoplankton growth.

In the NPZD model, the phytoplankton mortality was only caused by zooplankton

grazing effect. In the NPVD model, the phytoplankton mortality was only caused by

viral infection and lysis (Figure 2.1b).

NPZD+V: The ‘Nutrient-Phytoplankton-Zooplankton-Detritus + Viruses’ (NPZD+V)

model combined the NPZD model and the NPVD model to investigate how viruses

short circuit the flow of C and nutrients from phytoplankton to higher trophic levels by

viral infection and lysis and shunt these fluxes into detritus (Figure 2.1c).

NPZD+B: The ‘Nutrient-Phytoplankton-Zooplankton-Detritus + Bacteria’ (NPZD+B)

model incorporated the bacterial compartment into the NPZD model to investigate how

the interactions between the heterotrophic bacteria and the microzooplankton grazers

influence phytoplankton via C and nutrient recycling processes (Figure 2.1d). Here we

11

divided zooplankton compartment into two sub-compartments: one was the normal

zooplankton (Z1), which only grazed on phytoplankton; the other was the

microzooplankton (Z2), which only grazed on bacteria.

NPZD+VB: The ‘Nutrient-Phytoplankton-Zooplankton-Detritus+Viruses+Bacteria’

(NPZD+ VB) model combined the NPZD+V model and the NPZD+B model to

investigate the impact of viral shunt on the microbial loop (Figure 2.1e). Here we not

only divided zooplankton compartment into two sub-compartments (Z1 and Z2), but also

divided viral compartment into two sub-compartments (V1 and V2). V1 refers to the

phytoplankton viruses; V2 refers to the bacteria viruses.

(a) (b)

(c) (d)

D P

N

Z1 B Z

2

D P

N

V

Z

D

P Z

N

P

D N

V

12

(e)

Figure 2.1 Structure of NPZD (a), NPVD (b), NPZD+V (c), NPZD+B (d) and NPZD+VB (e) models (Note that B refers to the bacterial functional group; D refers to bacteria/fungi-coated detritus; Z1 , Z2 , V1, and V2 refer to Table

2.1 in the caption for definition of variables).

Table 2.1 The key plankton variables with the number of simulated groups in ecological models.

Description

Symbol

Unit***

Number of the simulated plankton

NPZD NPVD NPZD+V NPZD+B NPZD+VB

Phytoplankton P mmol N m‐3 1 1 1 1 1

Zooplankton Z mmol N m‐3 1 0 1 2* 2*

Nutrients N mmol N m‐3 1 1 1 1 1

Detritus D mmol N m‐3 1 1 1 1 1

Bacteria B mmol N m‐3 0 0 0 1 1

Viruses V mmol N m‐3 0 1 1 0 2**

*Z1 refers to normal zooplankton; Z2 refers to the microzooplankton.

**V1 refers to the phytoplankton viruses; V2 refers to the bacteria viruses.

*** mmol N m‐3 refers to mmol Nitrogen m‐3.

2.3.2 Model parameterisation

The compartments and fluxes in the NPVD model were summarised as follows: the

nutrient uptake by phytoplankton was based on Michaelis–Menten kinetics, which was

also limited by light and nutrient availability; all other processes were based on linear

D P

N

Z1 B Z

2

V1 V

2

13

first-order kinematics.

Nutrient uptake for phytoplankton growth (dnp)

PN

N

I

I

I

Ird

opt

PAR

opt

PARnp

1expmax

With

min,

4

1max III PARopt

(1)

Phytoplankton excretion (dpn)

Prd pnpn (2)

Phytoplankton mortality (dpd)

Pmd ppd (3)

Viral production (dpv)

Vd vpv (4)

Viral decay (dvd)

Vrd vdvd

(5)

Remineralisation of detritus into nutrients (ddn)

Drd dndn (6)

Based on the equations of the NPVD model, the serial NPZD+V model for the viral shunt, the

NPZD+B model for the microbial loop, the NPZD+VB model for the viral shunt short circuit

the microbial loop were developed. The compartments and fluxes for zooplankton and

bacteria were summarised as follows:

Zooplankton production (dpz)

ZPIpd vpz ))exp(1( 22max

(7)

Ivlev constant (Iv) refers to the saturation rate with increasing food levels for zooplankton.

14

Zooplankton excretion (dzn)

Zrd znzn (8)

Zooplankton mortality (dzd)

Zmd zzd (9)

Bacterial production (ddb)

DKB

Bd

BBdb

(10)

Bacterial grazed by microzooplankton (dbz)

BBK

Bgd

zrbz

(11)

Bacterial excretion (dbn)

(12)

The equations of these five ecological models (NPZD model, NPVD model, NPZD+V

model, NPZD+B model, and NPZD+VB model) were compared in Table 2.2 to show

the similarities and the differences between key biological variables. The modelling

parameters of these ecological models were summarised in Table 2.3.

dbBebn dKd

15

Table 2.2 Comparison of the NPVD model, the NPZD+V model, the NPZD+B model, and the NPZD+VB model.


P

pzpdpnnp dddddt

dP

pvpdpnnp dddd

dt

dP

pv

pzpdpnnp

d

dddddt

dP

1pzpdpnnp dddd

dt

dP

11 pvpzpdpnnp ddddddt

dP

Z

znzdpz ddddt

dZ

Null

znzdpz ddddt

dZ nzdzpz ddd

dt

dZ111

1

nzdzbz sddd

dt

dZ

22

2

nzdzpz ddddt

dZ111

1

nzdzbz ddddt

dZ222

2

N

npzndn ddddt

dN

npzndn ddd

dt

dN npzndn ddd

dt

dN

np

nznzpnbn

d

dddddt

dN

21

np

nznzpnbn

d

dddddt

dN

21

D

dnzd

pdpn

dd

dddt

dD

dnvd

vpdpn

dd

Vdddt

dD

1

dnvdv

zdpdpn

ddV

ddddt

dD

)1(

db

dzdzpdpn

d

dddddt

dD

21

dbdvdvv

dzdzpdpn

dddV

dddddt

dD

21

21

)1(

B Null Null

Null

2bzbndb ddddt

dB

22 bvbzbndb dddddt

dB

V Null

vdpv dddt

dV vdpv dd

dt

dV

Null

dvpv dddt

dV11

1

dvbv dddt

dV22

2

16

Table 2.3 The parameters of five ecological models

Description Symbol Unit NPZD NPVD NPZD+V NPZD+B NPZD+VB Literature review

Maximum grazing rate on phytoplankton

pmax d‐1 0.2 0.2 0.2 0.2 0.2 1[2]0.5[3]

Zooplankton excretion rate

rzn d‐1 0.01 Null 0.01 0.01 0.01 0.01[3]

Zooplankton mortality rate

mz d‐1 0.02 Null 0.02 0.02 0.02 0.02‐0.07[1]

0.3[2]0.02[3]

maximum nutrient uptake rate

rmax d‐1 1.0 1.0 1.0 1.0 1.0 0.35‐3.6[1]

0.5‐1.5[2]

0.24‐4.56[4]

minimum photosynthetically active radiation (PAR)

Imin W/m2 25 25 25 25 25 25[3]

phytoplankton mortality rate

mp d‐1 0.02 0.02 0.02 0.02 0.02 0.02[3]

viral production rate (phytoplankton)

µv d‐1 Null 0.16 0.16 Null 0.16 0.16[5]

viral production rate (bacteria)

µvB d‐1 Null Null Null Null 0.1 0.5‐4.2 ×109

viruses l‐1 d‐1[7]

viral decay rate

(phytoplankton)

rvd d‐1 Null 1.23 1.23 Null 1.23 1.23[5]

viral decay rate

(bacteria)

rvdB d‐1 Null Null Null Null 0.05 0.025[6]

phytoplankton excretion rate

rpn d‐1 0.01 0.01 0.01 0.01 0.01 0.01[3]

mineralization rate

rdn d‐1 0.007 0. 007 0.007 Null Null 0.003[3]

half saturation constant

α mmol N m

‐3 1.35 1.35 1.35 1.35 1.35 1.35[3]

Ivlev constant Iv 1.1 1.1 1.1 1.1 1.1 1.1[3]

maximum bacterial DOM uptake rate

µB d‐1 Null Null Null 0.1 0.1 0.05[1]13.3[2]

DOM excretion KBe d‐1 Null Null Null 0.7 0.7 0.7[1]

half saturation constant for bacteria function

KB mmolN m‐3

Null Null Null 0.97 0.97 0.97[1]

grazing rate on bacteria

gr d‐1 Null Null Null 9 9 0.9[1]

Half saturation constant for grazing

Kz mmolN m‐3

Null Null Null 145.8 145.8 145.8[1]

[1] Gal et al. (2009)

[2] Van den Meersche et al. (2004)

[3]Burchard et al. (2005)

17

[4] Pollingher and Berman (1982)

[5] Rhodes and Martin (2010)

[6]Heldal and Bratbak (1991)

[7]Steward et al. (1996)

2.3.3 Model setup

The Framework for Aquatic Biogeochemical Models (FABM) is a recently developed

community modelling framework for simulating the biogeochemical and ecological

dynamics of aquatic ecosystems (Trolle et al., 2012). FABM supports coupling of a

diverse array of water quality and ecological models to various physical ‘driver’

models, ranging from a 0-dimensional box model to a suite of 1, 2 or 3-dimensional

hydrodynamic models. FABM has been applied to a variety of aquatic environments

including oceans, estuaries and lakes. Here we only adopted 0D FABM as modeling

platform.

In order to provide a typical and clearly defined physical environment for testing the

NPZD model, the NPVD model and the serial NPZD+V model, NPZD+B model, and

NPZD+VB model, we used an annual simulation of the water column in Lake Kinneret

for the period from 1997-01-01 to 1997-12-31.

The initial conditions were the same for all the variables for the following two

simulations based on the temperature and light field data of Lake Kinneret (Israel):

Sim1: There was no seasonality in 0D FABM with constant temperature at 15oC and

idealized light conditions with diel changes for a year (Figure 2.2a).

Sim2: There was seasonality in 0D FABM with field temperature and light obtained

from Lake Kinneret (Figure 2.2b).

All simulations were carried out with a time step of 12 h for the physical part. By doing

so, it would be possible to use exactly the same physical forcing for all ODE solvers and

all biogeochemical time steps.

18

(a)

(b)

Figure 2.2 Light and temperature for Sim1(a) and Sim2(b).

2.3.4 Analysis approach

To determine the difference of the viral shunt and the microbial loop on the key

biological variables (viruses, bacteria, phytoplankton, zooplankton) and the nutrient

pools (nutrient, detritus), each variable was compared under the different model

structures.

To gain a better understanding about the changes of state variable relationships based on

model structure, the phase space analysis was performed for the ecosystem relationships

(zooplankton vs. phytoplankton, viruses vs. phytoplankton, bacteria vs. zooplankton,

and bacteria vs. viruses) in the NPZD model, the NPVD model, the NPZD+V model,

the NPZD+B model, and the NPZD+VB model.

19

To determine the influence of the viral shunt and the microbial loop on the key nutrient

pathways of nutrient recycling processes between viruses, bacteria, phytoplankton,

zooplankton, the nutrient fluxes of the NPZD model, the NPVD model, the NPZD+V

model, the NPZD+B model, and the NPZD+VB model were averaged over the

simulation period of one year.

2.4 Results

2.4.1 Model comparison

The simulated results of the Sim2 were mainly similar to the simulated results of the

Sim1. The peak of phytoplankton bloom was both captured in the NPZD model and the

NPZD+B model (Figure 2.3). Although the Sim1 did not generate small peaks of

phytoplankton growth in the NPZD+V model and the NPZD+VB model like the Sim2.

The trends of nutrients, detritus, and zooplankton were similar in both Sim1 and Sim2,

although the trends fluctuated more in Sim2 (Figure 2.4-2.6), because of seasonality. It

illustrated these ecological models are not only suitable to the ideal environment (Sim1)

but also the real environment (Sim2).

It is clear that there were some differences in trends and magnitudes of some variables

between different model structures. The peak of phytoplankton growth in the NPZD+B

model was higher than the peak in the NPZD model both in Sim1 and Sim2 (Figure

2.3), which was relevant to the impact of the microbial loop on primary production. For

the simulated bacteria and viruses, another small peak of the NPZD+VB model

occurred in the Sim2, which was relevant to the impact of the viral shunt on the

microbial loop (Figure 2.7 and Figure 2.8).

20

Figure 2.3 Simulated phytoplankton of Sim1(on the left) and Sim2 (on the right) with different ecological models

Figure 2.4 Simulated nutrient of Sim1(on the left) and Sim2 (on the right) with different ecological models

21

Figure 2.5 Simulated detritus of Sim1(on the left) and Sim2 (on the right) with different ecological models

Figure 2.6 Simulated zooplankton of Sim1(on the left) and Sim2 (on the right) with different ecological models

22

Figure 2.7 Simulated bacteria of Sim1(on the left) and Sim2 (on the right) with different ecological models

Figure 2.8 Simulated viruses of Sim1(on the left) and Sim2 (on the right) with different ecological models

23

2.4.2 Phase space analysis

The ecosystem relationship of ecological models in Sim1 and Sim2 further showed that

seasonal changes in light intensity influence the interaction between phytoplankton and

zooplankton, and the interaction between phytoplankton and viruses (Figure 2.9). When

the ecosystem relationship between phytoplankton, zooplankton, bacteria, and viruses

were plotted into phase space, the trajectories of phytoplankton vs. zooplankton in the

NPZD model and the NPZD+B model, zooplankton vs. bacteria in the NPZD+B model

and the NPZD+VB model moved into a single point in phase space (Figure 2.9). When

the phytoplankton biomass increased fast, the zooplankton biomass increased slowly.

When phytoplankton stopped increasing at around 2 mmol/m3 in the NPZD model and

2.5 mmol/m3 in the NPZD+B model and began to decrease, the zooplankton increased

fast. Later when zooplankton reached at the range of 1.5-2 mmol/m3 in the NPZD

model or 2.5-3 mmol/m3 in the NPZD+B model, phytoplankton stopped decreasing and

zooplankton also correspondingly decreasing very fast. When the microbial loop is

included, the phytoplankton vs. zooplankton relationship of NPZD+B is larger than that

of NPZD, which illustrates that the microbial loop plays a positive effect on the primary

production. The trajectory of zooplankton vs. bacteria in the NPZD+B model was

approaching a constant. This reflects the predation relationship between phytoplankton

and zooplankton in the traditional food web chain, bacteria and microzooplankton in the

microbial loop.

The dynamic behavior of parasitism between viruses, bacteria, and phytoplankton were

also reflected in the phase space diagram. Both of the phytoplankton biomass and the

viral biomass simultaneously increased. However, when the phytoplankton biomass

decreased, the viral biomass still kept a high increasing rate. This resulted in the

phytoplankton always stayed in the same low level (0.1-0.3 mmol/m3), which was

unstable. Therefore when the viral shunt is included, there was no obvious trend in the

phytoplankton vs. zooplankton relationship and bacteria vs. zooplankton relationship,

which illustrates the viruses diverted the nutrient movement from phytoplankton into

the detritus pool. Similarly, when the viral biomass increased from 0 to 0.3 mmol/m3,

the bacteria biomass also kept around 0.1 mmol/m3. The viruses can shunt the bacteria

biomass into detritus, which short circuits the microbial loop.

24

(a) (b)

Figure 2.9 Ecosystem relationship of Sim1(a) and Sim2 (b) of ecological models

25

2.4.3 Nutrient fluxes

The difference between viral infection and zooplankton grazing illustrates the

phytoplankton mortality caused by viral infection and lysis is as important as the

phytoplankton mortality caused by zooplankton grazing. From the perspective of

nutrient recycling processes in aquatic ecosystems, in the NPZD model, mineralization

recycled 79.3% of total nutrient taken up by phytoplankton, and zooplankton excretion

returned 16.9% (Figure 2.10a). In the NPVD model, however, mineralization recycled

308.5%, with viral mortality contributing 101.5% (Figure 2.10b). The phytoplankton

mortality was102.2% from viral infection in the NPVD model, but it was only 54.2%

from zooplankton grazing in the NPZD model. In the NPZD+V model, because of viral

infection, a large amount of nutrients were stored in the virus pool (Table 2.4) so that

the viral infection was almost 0.0%. However, the phytoplankton mortality was 10.9%

(Figure 2.10c), which was much higher than zooplankton grazing (0.34%) because the

phytoplankton biomass was converted into viral biomass.

When bacterial compartment was incorporated into the NPZD models, the microbial

loop had a significant influence on mineralization and the zooplankton pool (Table 2.5).

In the NPZD+B model, the bacterial mineralization increased to 120.1% compared to

the other mechanisms, such as zooplankton grazing on bacteria (34.1%) and

zooplankton excretion (31.9%) (Figure 2.10d), which is the primary mechanism of the

microbial loop influencing the phytoplankton growth. In the NPZD+VB model,

although the viruses mainly infected phytoplankton (86.7%), viruses infected bacteria

(55.0%), which increased bacterial uptake nutrient (193.0%) and bacterial excretion

(135.1%) compared to 120.1% and 84.1% in the NPZD+B model. Meanwhile, the

zooplankton grazing on bacteria decreased from 34.1% in the NPZD+B model to 3.5%

in the NPZD+VB model (Figure 2.10e). This illustrated that the viral shunt can short

circuit the microbial loop via increasing the ‘bottom-up’ control bacterial mineralization

but decreasing the influence of the ‘top-down’ control via zooplankton grazing.

26

(a) (b)

(c) (d)

(e)

Figure 2.10 Summary of simulated nutrient fluxes (mmol/m3d-1) for NPZD (a), NPVD (b), NPZD+V(c), NPZD+B (d), and NPZD+VB (e) (Values in () are provided as % of total nutrient taken up by phytoplankton).

27

Table 2.4 Summary of simulated annual average variables (mmol/m3) for ecological models


N 1.9307 6.6111 4.1239 4.0534 4.9804

P 0.3940 0.0069 0.0809 0.4302 0.0940

Z 0.7567 null 0.0017 1.3354(Z1) 0.0021(Z1)

0.6514(Z2) 0.0191(Z2)

D 5.9186 2.3778 2.3406 2.0495 3.7128

B null null null 0.5800 0.0956

V null 0.0042 2.4530 null 0.0122(V1)

0.1837(V2)

Table 2.5 Summary of simulated nutrient fluxes (mmol/m3d-1) for ecological models.


dnp 0.0448 0.0045 0.0148 0.0623 0.0174

dpn 0.0039 6.1017×105 8.1019×10-4 0.0043 9.4170×104

dpd 0.0079 1.2203×104 0.0016 0.0086 0.0019

dpz 0.0243 null 5.0598×105 0.0430 6.6439×105

dbz null null null 0.0212 6.0797×104

dzn 0.0076 null 1.6569×105 0.0199 2.1197×104

dzd 0.0151 null 3.3138×105 0.0397 4.2394×104

ddn 0.0355 0.0138 0.0140 null null

dpv null 0.0046 1.5004×107 null 0.0151

dbv null null null null 0.0096

dvd null 0.0045 3.4913×105 null 0.0243

dbd null null null 0.0749 0.0335

dbn null null null 0.0524 0.0235

28

2.5 Discussion

According to the results, virus-mediated mortality on phytoplankton via infection and

lysis is as important as zooplankton-mediated mortality on phytoplankton via grazing

when lysis and grazing rates reported in the general literature are used under identical

conditions. By comparing nutrient fluxes between the NPZD model and the NPVD

model, the virus-mediated mortality on phytoplankton speeds up the mineralization

process, which helps terminate algal blooms via the ‘bottom-up’ control. However,

zooplankton grazing on phytoplankton only provides organic matter to detritus by their

own mortality. From this point view, the recycling of organic nutrients via

mineralization is facilitated by viruses (‘bottom-up’ control) more than zooplankton

(‘top-down’ control).

Two types of microbial interactions, the microbial loop and the viral shunt, have been

incorporated into the NPZD model to explore the interactions between different

plankton populations (esp., viruses, bacteria, and phytoplankton). The NPZD+V model

further illustrated the impact of the viral shunt on phytoplankton growth in aquatic

ecosystems, that is, viruses catalyse the net movement of nutrients from phytoplankton

to the detritus pool. The NPZD+B model for the microbial loop showed the magnitude

of the peak of phytoplankton in NPZD+B was larger than the magnitude of the peak of

phytoplankton growth in the NPZD model, which indicates that overall, the microbial

loop has a positive impact on the primary production of phytoplankton. In the

NPZD+VB model, viruses infected and lysed both phytoplankton and bacteria, and

thereby reduced the transfer of C and nutrients to higher trophic levels. Therefore, the

viral shunt short circuits the microbial loop via increasing the turnover of ‘bottom-up’

control, that is, bacterial mineralization. When viruses are bacteriophages, they also

decrease the influence of the ‘top-down’ control via zooplankton grazing. We therefore

conclude that the activity of viruses can dominate the microbial community and play a

significant role in shifting the structure of aquatic ecosystems. These modeling results

highlight the importance of the microbial loop on developing algal blooms and the viral

shunt on terminating algal blooms.

The increase in model complexity from the NPZD model to the NPZD+VB model helps

to unravel how microbial interactions influence on algal blooms. Under similar physical

conditions, the serial NPZD model, the NPVD model, the NPZD+V model, and the

NPZD+VB model generate different chemical and biological results, especially for

29

phytoplankton compartment.

In the 0D FABM platform, the present simulations are mostly steady. If it links to

hydrodynamics such as inflow events and mixing, this would change the seasonality of

the majority of chemical and biological variables. This also would change the simulated

magnitude of algal blooms in aquatic ecosystems.

Scientists need to make an appropriate choice of model complexity depending on the

type of aquatic ecosystems. The three compartment NPZ model or the four

compartment NPZD model have been applied and coupled to the hydrodynamic model

for exploring physical-chemical-biological interactions in aquatic ecosystems (Edwards,

2001; Frank, 2002). For example, the classic NPZ model has been successful for

studying spring algal blooms in North Pacific Sea (Parslow, 1985; McGillicuddy et al.,

1995). However, these models must be used carefully and appropriately after checking

if the ecosystem is dominated by simple interactions between phytoplankton and

zooplankton. When the microbial loop has significant impact on regulating the C and

nutrient fluxes in aquatic ecosystems, such as Lake Kinneret (Hart et al., 2000;

Hambright et al., 2007), the NPZD+B model can capture the planktonic dynamics

successfully (Gal et al., 2009; Makler-Pick et al., 2011). When the microbial community

is dominated in some aquatic ecosystems, such as Antarctic lakes (Laybourn-Parry et

al., 2001; Madan et al., 2005), the NPZD+V model or the NPZD+VB model is able to

capture the pivotal role of viruses in C and nutrient recycling, especially bacteriophage.

When bacteria and viruses parameter and data are not available for modelling the viral

shunt and the microbial loop in some aquatic ecosystems, scientists need to constrain

some parameters with literature review data in these models to test their scientific

hypothesis with the improved NPZD+V, NPZD+B or NPZD+VB modelling

framework.

For different types of aquatic ecosystems, the improved serial NPZD ecological models

also facilitate scientists to further develop ecological stoichiometry (C:N:P ratios)

information about the microbial interactions within specific model configurations.

Ecological stoichiometric modelling has been recently developed to study the

interactions between phytoplankton and zooplankton in aquatic ecosystems (Elser et al.,

2012). The stoichiometric requirements of different organisms are constrained by mass

conservation of elements (Elser and Urabe, 1999), which leads to stoichiometric control

of some microbial processes (e.g., bacterial mineralization). Therefore, stoichiometric

30

analysis helps interpret microbial processes in biogeochemical cycles, and whilst not

included in this chapter, should be further investigated.

Based on the mathematical basis and detailed parameterisation for developing the

NPVD model, the NPZD+V model, the NPZD+B model, and the NPZD+VB model, in

the future, we can design different sub-models for specific scientific questions and

provide dynamic scenarios of microbial interactions with appropriate field dataset.

Take the NPZD+B model for the microbial loop as an example. It is important to

consider predation between carnivorous zooplankton, microzooplankton and

bacteriovorous heterotrophic nanoflagellates. In this chapter, the model is simplified for

comparison with other models, but they are addressed in more details in the following

Chapters 3, 4, and 5 with different scientific questions for investigating the microbial

loop in Lake Kinneret. To explore how the ecological stoichiometry of phytoplankton

groups respond to nutrient pools within a dynamic aquatic environment, we can test if

interactions between ecosystem pools will result in patterns of phytoplankton

stoichiometry that are independent of bulk water column indicators. Moreover, we can

also define quantitatively how these microbial interactions influence on phytoplankton

growth in aquatic ecosystems, especially the role of the microbial loop on regulating the

nutrient fluxes between bacteria, phytoplankton, and zooplankton to shape

phytoplankton succession patterns in Lake Kinneret.

31

3 An analysis of the relationship

between phytoplankton internal

stoichiometry and water column N:P

ratios in a dynamic lake environment

3.1 Abstract

The N:P stoichiometry of a water body is one of the most commonly used indicators of

its nutrient status and algal growth. However, in a dynamic aquatic ecosystem the N:P

stoichiometry of phytoplankton is highly variable and depends on environmental

conditions and key microbial interactions that influence their growth, such as grazing

pressures and the microbial loop. Here we determine the influence of the nutrient-

dependent microbial interactions between zooplankton, phytoplankton and bacteria on

the ecological stoichiometry at different trophic levels and how they relate to water

column properties. A 1D hydrodynamic-ecological model (DYRESM-CAEDYM) was

applied to Lake Kinneret (Israel) for examining how the internal nutrient ratios of

several phytoplankton functional groups correlate with nutrient ratios within the water

column, and further explore how the microbial loop shapes the patterns of stoichiometry

within the food web by testing two microbial loop configurations. The results showed

that the average internal N:P ratios of the phytoplankton community followed their total

32

carbon biomass patterns, and that seasonal patterns of simulated dissolved inorganic N

to total P (DIN:TP) ratios in the water column were a useful indicator for reflecting the

bulk phytoplankton N:P stoichiometry as compared with total N to total P (TN:TP)

ratios and dissolved inorganic N to dissolved inorganic P (DIN:DIP) ratios. However,

the internal N:P ratio patterns of individual phytoplankton groups did not necessarily

correlate with DIN:TP ratio patterns in the water column. This was because different

microbial processes regulate nutrient flows to individual phytoplankton groups. Our

simulations with the microbial loop highlight the ability of bacteria to regulate

phytoplankton stoichiometry. These results provide an improved mechanistic

understanding of the food web in aquatic ecosystems.

3.2 Introduction

With an increase in human activities, large quantities of nutrients have been mobilized

into freshwater and coastal ecosystems. It is now well documented that eutrophication

has become widespread around the world (Smith, 2003). In such instances, excess

nitrogen (N) and phosphorus (P) are primarily responsible for fuelling primary

production and organic matter accumulation. This has resulted in frequent occurrences

of algal blooms and the associated deterioration of water quality in aquatic ecosystems.

In particular, the prevalence of harmful algal blooms are an increasing issue of concern

(Howarth 2008), because many nuisance species display high growth rates when

nutrients are in excess (Reynolds, 1984). The management of algal blooms and

eutrophication typically attempts to reduce nutrient loading to water bodies, with a

general focus on reducing N in marine systems (Howarth and Marino, 2006), and

reducing P in freshwater systems (Schindler et al., 2008). Nonetheless, when managing

a specific aquatic system it is necessary to determine the degree of N or P limitation for

setting nutrient reduction targets and assessing management initiatives. Therefore, in

conjunction with nutrient and chlorophyll-a concentrations, the N:P ratio is often used

as a basis to guide water quality management efforts to reduce algal blooms in aquatic

ecosystems (Smith, 1983; Sterner and Elser, 2002; Gal et al., 2009; Ptacnik et al., 2010;

Bergström, 2010).

The application of the N:P ratio is based on Alfred Redfield’s early work, which pointed

33

towards an almost ‘universal’ C:N:P molar ratio (106:16:1) of marine seston (Redfield,

1958), such that when the N:P ratio is less than 16, the growth of phytoplankton is N

limited and when the N:P ratio is more than 16, the growth of phytoplankton is P

limited. However, it is now well established that there is a great deal of variability in

phytoplankton internal N:P ratios (Saxton et al., 2012), and the optimal N:P ratio of

phytoplankton can range from 8:1 to 45:1 depending on the species of interest and the

prevailing environmental conditions (Klausmeiser et al., 2004). Therefore, the

assumption that phytoplankton internal N:P stoichiometry matches the bulk properties

of the water column may not always be accurate.

Several types of N:P ratios have been used to understand the nutrient limitation of

phytoplankton, such as dissolved inorganic N: total P (DIN:TP) ratios (Morris and

Lewis, 1988, Ptacnik et al., 2010; Bergström, 2010), dissolved inorganic N: dissolved

inorganic P (DIN:DIP) ratios (Rhee, 1978; Sterner and Elser, 2002), and total N:total P

(TN:TP) ratios (Gal et al., 2009; Gillor et al., 2010). Whilst these ratios are easily

calculated from monitoring data, the abundance of phytoplankton is usually only

represented by carbon (C) biomass or Chlorophyll-a (Chl-a), and the internal N (iN) and

internal P (iP) content of phytoplankton are seldom measured at a high enough

frequency to assess how the changes of phytoplankton internal N:P (iN:iP) ratios relate

to these nutrient ratios of the water column. Currently, there have been limited

investigations exploring inter-relationships between these bulk nutrient indicators of

ecosystem conditions and the physiological response of phytoplankton communities and

associated food webs. Of particular interest is how the internal N:P stoichiometry of

phytoplankton groups respond to variable patterns of nutrient supply within a dynamic

aquatic environment.

Phytoplankton stoichiometry is influenced by a range of environmental factors relevant

to their growth, and their C:N:P ratios vary considerably in space and time. The

dynamics of algal blooms are known to be controlled by ‘bottom-up’ factors of light,

temperature and physical processes such as mixing (Ng et al., 2011) and in some cases

diurnal vertical migration (Regel et al., 2004). They are also influenced by energy,

nutrients (Kooijman et al., 2004), complex life cycles , and species-specific intrinsic

physiological processes (Michaels et al., 2001; Karl et al., 2001; Vrede et al., 2004;

Frost et al., 2005). For example, if inorganic N is insufficient to satisfy their N:P ratios,

some phytoplankton species will supplement N through N2 fixation, and others can

store P in the form of polyphosphate (Sterner and Elser, 2002). They may also be

34

influenced by the community structure since microbial interactions, such as competition

for nutrient resources and grazing pressure from higher consumers, can regulate nutrient

uptake and storage kinetics. It is therefore difficult to accurately determine the uptake of

different forms of N and P from the water column into phytoplankton (Ballantyne et al.,

2008; Gillor et al., 2010) and this interplay of different processes may lead to organism-

specific patterns of internal N:P ratios that are decoupled from the water column N:P

indicators.

Through their interactions, microbes in the water column influence overall patterns of

ecological stoichiometry at different trophic levels. Heterotrophic and autotrophic

microorganisms regulate C:N:P ratios by coupling carbon-to-nutrient recycling

processes (Thingstad et al., 2008). Since bacteria and zooplankton acquire the majority

of their C, N, and P supply from the same source of organic material, they are

stoichiometrically homeostatic and generally have a more constant C:N:P ratio (Sterner

and Elser, 2002; Makino et al., 2003). In contrast, phytoplankton have a different

mechanism for acquiring their source of C, from atmospheric CO2, compared to their

sources of N and P, and their C:N:P stoichiometry is more highly variable and

independent (Kooijman et al., 2004). In order to maintain mass balances, these

differences lead to variable patterns of C:N:P ratios in trophic transfers (Elser and

Urabe, 1999). For example, the physiological constraints of a constant C:N:P ratio in

bacteria has been shown to regulate nutrient fluxes to phytoplankton in experimental

cultures (Danger et al., 2007), suggesting that we must consider bacterial recycling of

organic matter when understanding the impact of trophic interactions on the

stoichiometry of phytoplankton.

Traditional studies of aquatic ecosystems have focused on the classic ‘nutrients-

phytoplankton-zooplankton’ paradigm whilst often ignoring the detrital-based pathway

and the role of organic matter cycling in regulating nutrient recycling and the

stoichiometry of nutrient flows between trophic levels (Ballantyne et al., 2008). It is

now accepted that predators, such as crustacean zooplankton or fish, can be supported

by the detrital-based pathway of aquatic ecosystems, also known as the ‘microbial loop’

(Azam et al., 1983; Moore et al., 2004). As the size range of heterotrophic flagellates

and microzooplankton are similar to phytoplankton, they facilitate the movement of

energy, carbon and nutrients from the microbial loop to the conventional food chain

(Stone et al., 1993; Hart et al., 2000; Hambright et al., 2007; Pomeroy et al., 2007).

Currently, little is understood about how the microbial loop moves nutrients between

35

organic matter, bacteria and microzooplankton, and its ability to regulate the

stoichiometry of available nutrients to phytoplankton within the food web in aquatic

ecosystems. In earlier analyses, we identified that complex patterns of nutrient flows

between microbial groups emerge using a dynamic, stoichiometry-based ecosystem

model of the Lake Kinneret (Israel) ecosystem (Gal et al., 2009; Makler-Pick et al.,

2011a).

The aim of this study was to apply this model to investigate patterns of algal internal

nutrient stoichiometry compared with water column properties. By considering the

variability in stoichiometry of nutrient flows between organisms and the constraints of

mass conservation, we hypothesised that interactions between ecosystem pools would

result in patterns of phytoplankton stoichiometry that were independent of bulk water

column indicators. To test this hypothesis, we applied ecological stoichiometry

principles to results from a coupled hydrodynamic-ecological model (DYRESM-

CAEDYM) of Lake Kinneret (Israel), and explored the relationship between several

N:P ratios of the water column (TN:TP ratios, DIN:DIP ratios, DIN:TP ratios) and the

internal N:P ratios of several phytoplankton groups. Furthermore, we used the model to

examine how the microbial loop can influence N:P stoichiometry of food web

components, especially the various phytoplankton functional groups.

3.3 Methods

3.3.1 Study site

Please refer to the section1.2 about the site description of Lake Kinneret.

3.3.2 Model overview

The coupled Dynamic Reservoir Simulation Model-Computational Aquatic Ecosystem

Dynamics Model (DYRESM-CAEDYM) was run for the period from 1997 to 2001 and

validated against data from a substantial monitoring program (Zohary et al., 2006). All

the parameters for running DYRESM-CAEDYM in Lake Kinneret were assigned based

on available field or experimental data from previous work (e.g., Hambright et al., 2007;

Gal et al., 2009). Several other related papers have been published on the lake

ecosystem, including application of DYRESM alone (Gal et al., 2003), and in

combination with CAEDYM (Bruce et al., 2006; Makler-Pick et al., 2011a; Makler-

Pick et al., 2011b) to which the reader is also referred.

36

In summary, the model was configured to simulate the C, N and P content of three

groups of zooplankton, five groups of phytoplankton, and one functional group of

bacteria, in addition to organic and inorganic nutrient pools within the water column. In

particular, the model simulated the carbon and intracellular nutrient stores of five main

phytoplankton taxa (A1: Peridinium; A2: Microcystis; A3: Aphanizomenon; A4:

nanophytoplankton; A5: Aulacoseira), adopting a modified Droop kinetic N and P

uptake model that sets lower and upper limits on C:N and C:P ratios for each group

based on the available empirical data. The model can therefore capture the dynamic

response of phytoplankton stoichiometry to environmental conditions and food web

structure. This in turn provides a means for evaluating the relationship between the

internal nitrogen (iN) to internal phosphorous (iP) ratios of phytoplankton and the N:P

ratios of the water column in the lake. The heterotrophic groups, including the three

zooplankton groups and one bacteria group, have fixed C:N:P stoichiometry in line with

earlier studies (Gal et al., 2009). Whilst the model equations are the same as has been

previously reported, a summary of the essential algorithms relevant to the present study

are outlined below.

3.3.2.1 Phytoplankton model approach

The simulated growth rate (μg) of each of the five phytoplankton groups in CAEDYM is

multiplied by a minimum expression for light, N, and P (and Si for the diatom group),

and is scaled according to a species specific temperature function:

(1)

where μMAX (day−1) is the maximum growth rate at 20oC, f(I), f(N) and f(P) represent

limitation by light, nitrogen and phosphorus respectively, f(Si) represents limitation by

silica for diatoms, and f(T) is a temperature function, which allows for inhibition of

phytoplankton at higher temperatures (Hipsey and Hamilton, 2008; Özkundakci et al.,

2011). Each of the functions is described below and the relevant variables and

parameters are summarised in Table 3.1 and Table 3.2:

(2)

(3)

)()(),(),(),(min TfSifPfNfIfMAXg

bTf aTkT )(20)(

ss I

I

I

IIf 1exp)(

37

(4)

(5)

Based on the modified Droop model (Hipsey and Hamilton, 2008), nutrient uptake for

different phytoplankton groups (Aa, a is algal group index =1, ..., 5) is modelled

dynamically according to:

(6)

(7)

where PNa is the preference of the phytoplankton group for NH4 (between 0 and 1) and

is defined according to the relative abundance of the inorganic N species:

(8)

Nitrogen fixation is simulated for Aphanizomenon (A3) according to:

(9)

and P uptake for all groups is modelled as:

(10)

Phytoplankton nutrient loss (E) through mortality and excretion is dynamically

calculated as:

PNa

NH4NO3

NH4 KNa NO3 KNa

NH4KNa

NH4 NO3 NO3 KNa

aaNFN ANfkUa

12

a

PMINMAX

aMAX A MAXFRP A

KFRP

FRP

IPIP

IPIP T f UP U

aaa

a

aa

a

NMINMAX

aMAX

AMAXNNO AKNONH

NONH

ININ

ININTUN PU

aaa

a

a a a

34

3413

f

a

NMINMAX

aMAX AMAX NNH

AKNONH

NONH

ININ

ININT f UNPU

aaa

a

a aa

34

34

4

a

MIN

MIN MAX

MAX

IP

IP

IPIP

IPP f a

aa

a 1)(

a

MIN

MIN MAX

MAX

IN

IN

IN IN

IN N f a

a a

a 1)(

38

(11)

(12)

(13)

(14)

where fDOM is the fraction of mortality and excretion to the dissolved organic pool with

the remainder going to the particulate organic pool.

Vertical migration is quantified based on the modified model of Kromkamp and Walsby

(1990), including the light response (Webb, 1974), for phytoplankton groups A1 and

A2, according to Stokes Law for A3 and A4, and based on a constant settling rate for

A5:

(15)

V s a

V I MAX a1 exp

I

I K a

VNMAX a

1IN a

IN min a

IN max a IN min a

a = 1, 2

g A a w d Aa 2

18 a = 3, 4

constant a = 5

a ArDOM POP

IP T f k f Eaa a

1

a ArDOM DOP IPT f k f Eaaa

a A rDOM PON

IN T f k f Eaaa

1

aArDOM DON IN T k f Eaaa

f

39

Table 3.1: List of biogeochemical and biological variables in DYRESM-CAEDYM

Notation CAEDYM Name

Description Units

PHYSICAL VARIABLES

I PAR Light intensity uE m-2

T Temperature C

Computational time step days

Vertical thickness of computational cell m

Vertical thickness of computational cell overlying sediment

m

Vertical thickness of computational cell adjacent to water-atmosphere interface

m

Kd EXTC m-1

DENSITY kg m-3

BIOGEOCHEMICAL VARIABLES

DOC DOCL Dissolved organic carbon concentration mg C L-1

POC POCL Detrital particulate organic carbon concentration mg C L-1

TN Total nitrogen concentration mg N L-1

PON PONL Detrital particulate organic nitrogen concentration mg N L-1

DON DONL Dissolved organic nitrogen concentration mg N L-1

NH4 NH4 Ammonium concentration mg N L-1

NO3 NO3 Nitrate concentration mg N L-1

TP Total phosphorus concentration mg P L-1

POP POPL Detrital particulate organic phosphorus concentration mg P L-1

DOP DOPL Dissolved organic phosphorus concentration mg P L-1

FRP PO4 Filterable reactive phosphorus mg P L-1

DO DO Dissolved oxygen concentration mg O L-1

BIOLOGICAL VARIABLES

NA Number of algal groups being simulated (=5) -

a Algal group index (1… NA) -

t

z

zbot

zsurf

40

A1 DINOF Algae #1 (Dinoflagellate: Peridinium gatunense the main, bloom-forming species) C biomass concentration

mg C L-1

A2 CYANO Algae #2 (Cyanobacteria: Non N2 fixing group represented by Microcystis, toxin-producing species) C biomass concentration

mg C L-1

A3 NODUL Algae #3 (Cyanobacteria: Filamentous N2 fixing group represented mostly by Aphanizomenon ovalisporum and Cylindrospermopsis cuspis) C biomass concentration

mg C L-1

A4 CHLOR Algae #4 (nanophytoplankton: A large suite of species that are nanoplanktonic in size and are readily grazed by zooplankton) C biomass concentration

mg C L-1

A5 FDIAT Algae #5 (Diatom: Aulacoseira granulata, a winter bloom forming filamentous diatom) C biomass concentration

mg C L-1

AIN a IN_XXX Algae #a(a=1,2,3,4,5) internal N concentration mg N mgC-1

AIP a IP_XXX Algae # a(a=1,2,3,4,5) internal P concentration mg P mgC-1

NZ Number of zooplankton groups being simulated (=3) -

z Zooplankton group index (1… NZ) -

Z1 ZOOP1 Zooplankton #1 (Predators: adult copepods, predatory rotifers) C biomass concentration

mg C L-1

Z2 ZOOP2 Zooplankton #2 (Large herbivores: cladocerans, copepodites) C biomass concentration

mg C L-1

Z3 ZOOP3 Zooplankton #3 (Microzooplankton: copepod nauplii, most rotifers, ciliates, heterotrophic flagellates) C biomass concentration

mg C L-1

B BAC Heterotrophic bacterial C biomass concentration mg C L-1

41

Table 3.2: List of phytoplankton parameters used in DYRESM-CAEDYM simulations of Lake Kinneret.

Parameter Description Units Assigned values: Values from field/literature

Per

idin

ium

Mic

rocy

stis

Aph

aniz

omen

on

Nan

opla

nkto

n

Aul

acos

eira

Peri

dini

um

Mic

rocy

stis

Aph

aniz

omen

on

Nan

opla

nkto

n

Aul

acos

eira

MAX Maximum potential growth rate day-1 0.35 0.70 0.41 2.70 3.60 0.24-4.56 2.4-8.57 0.715

Is Light saturation for maximum production Em-2s-1 600 150 80 400 200 130 75 440-710

KeA Specific attenuation coefficient m-1 (gC m-3)-1 0.1 0.1 0.1 0.1 0.1 0.449 0.448

KP Half saturation constant for phosphorus uptake

g P m-3 0.0024 0.0018 0.0012 0.0014 0.0050 0.001-0.0048

0.0011 0.0028-0.0111

KN Half saturation constant for nitrogen uptake

g N m-3 0.100 0.081 0.001 0.038 0.050 0.38

INMIN Minimum internal N:C ratio g N (g C)-1 0.030 0.025 0.020 0.084 0.050 0.0448 0.125

INMAX Maximum internal N:C ratio g N (g C)-1 0.070 0.060 0.110 0.330 0.150 0.09 0.146

UNMAX Maximum rate of nitrogen uptake g N (g C)-1 day-1 0.20 0.12 0.20 0.13 0.15 0.0043

IPMIN Minimum internal P:C ratio g P (g C)-1 0.0003 0.0081 0.0155 0.0067 0.0090 0.0040 0.0119

IPMAX Maximum internal P:C ratio g P (g C)-1 0.003 0.050 1.300 0.030 0.060 0.0187 0.0850

UPMAX Maximum rate of phosphorus uptake g P (g C)-1 day-1 0.010 0.080 0.250 0.050 0.400 0.0006-0.0060

0.0074 0.0031-0.0187

kNF N fixation rate g N (g C)-1 day-1 0 0 0.15 0 0

fNF Growth reduction under N fixation - 1.00 1.00 0.67 1.00 1.00

Ag Temperature multiplier for growth - 1.07 1.07 1.10 1.07 1.08 1.08 1.06

42

TSTDA Standard temperature C 19 19 24 20 13

TOPTA Optimum temperature C 24 26 29 27 15 22 20-30 16-17

TMAX A Maximum temperature C 32 35 34 35 22 28 >35 26-27

kr Metabolic loss rate coefficient day-1 0.039 0.060 0.118 0.065 0.085 0.03 0.039-0.051

Ar Temperature multiplier for metabolic loss - 1.05 1.05 1.05 1.06 1.12

kpr - 0.014 0.014 0.014 0.014 0.014

fres Fraction of respiration relative to total metabolic loss

- 0.25 0.25 0.25 0.25 0.25

fDOM Fraction of metabolic loss rate that goes to DOM

- 0.2 0.2 0.2 0.2 0.05

VI MAX Maximum migration velocity towards depth of optimum light

m s-1 0.0003 0.0003 N/A N/A N/A

VNMAX Maximum migration velocity towards depth of optimum N

m s-1 5.5e-5 5.5e-5 N/A N/A N/A

dA Cell diameter m N/A N/A 5.0e-7 1e-5 N/A

VSA Settling velocity m s-1 N/A N/A N/A N/A 1.0e-5 7e-6-1.2e-5

Sources Parameters and values from field and literature are based on the values used by Gal et al. (2009).

43

Table 3.3: Equations for bacteria and zooplankton in CAEDYM model for Microbial 'Loop Absent Scenario (MLAS) and Microbial Loop Present Scenario (MLPS)

MLAS MLPS

Bacteria fB(B) =1

BK

BBf

BB

)(

Hydrolysis:

POMDOfTfD DOBB

TBPOM )(min )( max

Hydrolysis:

POMBfDOfTfD BDOB

BT

BPOM )(min )( max

Mineralization:

)(min )( 1 DOMDOfTfU DOBB

TBDECDOM

Mineralization:

)(min )( 1 DOMBfDOfTfU BDOB

BT

BDECDOM

Organic nutrient uptake:

tkUDONDON

tkUDONkUU

BIN

BINBINDON <

>

tkUDOPDOP

tkUDOPkUU

BIP

BIPBIPDOP <

>

Organic nutrient uptake:

tkUDONDON

tkUDONkUU

BIN

BINBINDON <

>

tkUDOPDOP

tkUDOPkUU

BIP

BIPBIPDOP <

>

Inorganic nutrient uptake:

tkUU

tkUUUkU

tNHDONkUNH

U

BINDON

BINDONDONBIN

BIN

NH

= 0

< -

44

4

tkUUU

tkUUUUUkU

tNONHDONkUNO

U

BINNHDON

BINNHDONNHDONBIN

BIN

NO

0

< + --

4

443

343

44

tkUU

tkUUUkUU

BIPDOP

BIPDOPDOPBIPFRP = 0

< -

Micro‐zooplankton POM

POMK

POMgPOMG MAXZ

3 BBK

BgBG MAXZ

3

Note: As for the description of bacteria and zooplankton parameters used in the Lake Kinneret DYRESM‐CAEDYM simulations and definitions of the components in these equations, readers are referred to Gal et al. (2009) and Li et al. (submitted).

3.3.2.2 Microbial loop configurations

The following two configurations were used to assess the impact of the microbial loop

on the N:P stoichiometry of phytoplankton (Figure 3.1):

(1) Microbial Loop Absent Scenario (MLAS): This simulation assumed organic matter

was mineralized at a rate that was not dependent on the bacterial biomass (ie., the

bacterial biomass was assumed constant, and fB(B) for POM hydrolysis and DOM

mineralization was fixed at 1 in Table 3). Therefore this approach moved C, N and P

between DOM and DIM proportionally, varying only according to oxygen and

temperature. The microzooplankton consumed POM in place of bacteria at an

equivalent rate as POM and bacteria were assumed to be lumped together in this

configuration.

(2) Microbial Loop Present Scenario (MLPS): This simulation assumed organic matter

was mineralized by the simulated bacteria group. In addition, bacteria could supplement

their internal nutrient requirement by taking inorganic nutrients, thereby competing with

phytoplankton for nutrients. This configuration has been reported in Gal et al. (2009).

Z1

Z2

DIM

DOM

Z3

A1

POM (B)

A2

A3

A4

A5

(a)MLAS

46

Figure 3.1: Conceptual diagrams outlining the configured microbial groups and interactions in the Lake Kinneret

DYRESM-CAEDYM model for the Microbial Loop Absent Scenario, MLAS, (a) and the Microbial Loop Present

Scenario, MLPS, (b) configurations. (Note that solid lines indicate the common processes of MLAS and MLPS;

dashed lines indicate bacterial uptake of DOM and DIM in MLPS; the arrows on this diagram correspond to the flows

of C, N, and P).

A comparison of the equations for bacteria (B) and microzooplankton (Z3) under the

two configurations are summarised in Table 3.3.

3.3.3 Validation approach

Samples were collected with a 5-L vertical Rohde sampler from the deepest point

(Station A, Figure 1.1) of Lake Kinneret at depths of 0, 1, 2, 3, 5, 7, 10, 15, 20, 30 and

40 m. The details for analysis and determination of nutrient concentrations (NO3, NH4,

TN, PO4, and TP) and C biomass phytoplankton have been described previously

(Pollingher, 1986; Zohary, 2004; Gal et al., 2009). Note that the detection limit of PO4

was approximately 2 µg/L, potentially a little lower. The database, however, included

values of 1 µg/L which indicated samples that had some color in them but not enough to

be read by the spectrophotometer, i.e. just under the detection limit. Therefore, monthly

average values may have values as low as 1 µg/L or lower if the 4 weekly

measurements (included in the monthly mean values) had values of 1 and 0 which was

possible for the summer months when values were very low. Values of 0 in the database

Z1

Z2

DIM

DOM

Z3

A1

POM B

A2

A3

A4

A5

Microbial loop

(b) MLPS

47

were values that no color could be seen.

Parameter values of the equations in Section 3.2.2 are provided in Table 3.2. For further

parameter justifications the reader is referred to Gal et al. (2009). Model tests were

performed to assess the simulated results in terms of the inter-annual and intra-annual

variability in nutrient variables and the peaks and seasonality of the biological variables

in the water column relative to the field data. In this analysis we further validated the

nutrient ratios in the water column and compared the range of the simulated internal N:P

ratios of different phytoplankton groups against available literature values for Lake

Kinneret. Model performance was assessed based on inspection of simulated and

observed data with the maximum and minimum values, correlation coefficient (r), and

Spearman rank correlation coefficient (Rs) on the nutrient variables and ratios. The

correlation coefficient (r) shows the degree of overlap in the seasonal trends and timing

with the actual magnitudes of the peak values. The Spearman rank correlation

coefficient (Rs) gives the nature of the seasonal changes with less regard to the

magnitude of the variation for relative assessment between the model simulation results

and field data.

3.3.4 Stoichiometric assessment

Simulated nutrient concentrations of DIP, DIN (NH4+NO3), TN and TP variables, were

vertically integrated over the top and bottom 10m, and monthly averaged over the five

year simulation period (1997-2001). The relevant variables were then converted into

molar DIN:DIP, DIN:TP, and TN:TP ratios. The simulated ratios were compared with

average of the observed values in the surface and bottom 10 m of the water column. The

C, N and P biomass of phytoplankton was similarly vertically integrated and averaged.

The average iN:iP ratios for individual phytoplankton groups were calculated. For the

combined phytoplankton community and the combined heterotrophic community, their

iN:iP ratios were also determined by integrating the N:P stoichiometry of the relevant

organisms weighted by their biomass.

To determine the relationship between the simulated iN:iP ratio patterns of

phytoplankton and the DIN:TP ratio patterns of the water column, we conducted a

48

simple linear correlation analysis between the peak ratio values with different monthly

time lag values after checking that the assumptions of linear correlation were met (ie.

normality, independence, and linearity). If these assumptions were not met, data were

log transformed prior to analysis. Accounting for seasonal differences in these patterns,

the above ratios were also grouped in two classes: winter–spring (January–June) and

summer–autumn (July–December). This time lag analysis on monthly and seasonal

values was done for the combined phytoplankton community and for each specific

phytoplankton group.

To further explore the variability of the phytoplankton stoichiometry, a frequency

analysis was conducted for the distribution of combined and individual phytoplankton

iN:iP values. Due to the boom-bust nature of many phytoplankton groups, the analysis

was limited to the periods in which phytoplankton biomass was considered to be above

the numerical lower biomass limit of the model. Therefore, the data were filtered above

the threshold of 0.05 mgC L-1 for Aulacoseira, or 0.01 mgC L-1 for Peridinium,

Microcystis, Aphanizomenon, and nanophytoplankton.

In order to explore how variability in phytoplankton stoichiometry relates to the food

web, nutrient pools and fluxes were averaged over the simulation period for the MLAS

and MLPS configurations. Volume weighted values were assigned to the concentrations

of nutrient variables and biological variables at each depth. These values were then

summed to provide the lake wide values in ITS files of CAEDYM. Nutrient variables

and biological variables were integrated over each time step to get the lake-wide long-

term averages, and converted to molar ratios. The iN:iP ratios for the biological

variables were categorized into three different groups: constant stoichiometry (bacteria

and zooplankton), variable stoichiometry (within a user defined range for

phytoplankton), and freely varying stoichiometry (according to microbial interactions

and other ecosystem processes for POM, DOM, and DIM pools).

49

3.4 Results

3.4.1 Model performance

The simulated values of key chemical variables (NO3, NH4, DIN, TN and TP) in the

water column matched the seasonal patterns and nutrient dynamics of the lake

observations in both the surface water and the bottom water (Figure 3.2a,b). The Rs

values ranged between 0.52 and 0.84 in the surface layer (Table 3.4), which suggests

seasonal and inter-annual variability was well captured. However, the simulated PO4

concentrations in the surface water were lower than the field data in the lake. Since the

field PO4 concentrations in the surface water were around the detection limit, there was

a relatively large error associated with these PO4 values. Although Rs of PO4 was 0.751,

the magnitude of the simulated PO4 could not be overlapped with the actual magnitude

of the field PO4 (r =0.031). However, the seasonal trends of TP in the surface water

were successfully simulated in visual comparison with field data (Figure 3.2a).

The simulated key biological variables matched the seasonal patterns of the field data in

the surface water (Figure 3.2c,d). For the combined phytoplankton community, their

biomass and seasonal trends were successfully captured, although some discrepancies

existed in their peak values. For example, the simulated biomass peak of Peridinium

bloom in 1998 was much lower than the corresponding field data (Figure 3.2c) because

of model limitation in capturing the phytoplankton patchiness during the big algal

bloom event. However, the magnitude of the variation of Peridinium spring blooms was

successfully captured in Lake Kinneret, especially the timing of the blooms.

Furthermore, in most cases the model successfully captured the large extent of inter-

annual variation observed in the lake.

Compared with phytoplankton, the seasonal variability of three zooplankton groups

were more significant in model, although the simulated magnitude did not always match

the observed data (Figure 3.2d). In particular, the seasonal trends of the simulated

microzooplankton biomass were similar to the observations in the lake, although the

simulated values were higher than the field data. In addition, obvious seasonal

variations of the simulated bacteria biomass exhibited over the simulated period but this

50

variability was only partially reflected in the bacterial field data (Figure 3.2d).

Figure 3.2(a): Validation of nutrient variables in the surface water (0-10 m). Solid line represents simulated results

and symbols are lake based monthly mean data.

51

Figure 3.2(b): Validation of nutrient variables in the bottom water (30-40 m). Solid line represents simulated results

and symbols are lake based monthly mean data.

52

Figure 3.2(c): Validation of phytoplankton variables in the surface water. Solid line represents simulated results and

symbols are lake based monthly mean data.

53

Figure 3.2(d): Validation of heterotrophic organism variables in the surface water. Solid line represents simulated

results and symbols are lake based monthly mean data.

54

Table 3.4: Statistical comparison between model simulations and observed data in the surface water

Variable

r Rs

Simulated m

in N:P

Simulated m

ax N:P

Values from filed and

literature

NO3 0.580 0.708 N/A N/A N/A

NH4 0.688 0.528 N/A N/A N/A

PO4(DIP) 0.031 0.751 N/A N/A N/A

TN 0.750 0.842 N/A N/A N/A

TP 0.194 0.528 N/A N/A N/A

DIN 0.741 0.751 N/A N/A N/A

DIN: DIP 0.722 0.645 203.51 22701.16 13.80‐1261.22

TN:TP 0.436 0.383 33.44 86.52 36.45‐71.16

DIN:TP 0.577 0.675 0.51 20.59 0.95‐33.08

iN:iP ratio(Peridinium) N/A N/A 2.21 196.95 9.4‐49.6 a,b; 22.1‐516. 7d

iN:iP ratio(Microcystis) N/A N/A 2.21 12.38 22‐47a; 1.1‐16.4d

iN:iP ratio (nanophytoplankton)

N/A N/A 2.21 87.48 39.7‐47.1a; 6.2‐109.1d

iN:iP ratio

(Aphanizomenon)

N/A N/A 0.33 7.18 16a,c; 0.03‐15.7d

iN:iP ratio(Aulacoseira) N/A N/A 2.21 18.74 7.08‐9.4a; 1.8‐36.9d

iN:iP ratio( phytoplankton community)

N/A N/A 0.38 57.78 N/A

a Zohary (2004)

b Zohary (1998)

c Pollingher (1986)

d Gal (2009)

55

Based on a qualitative comparison of the different N:P ratios in the surface water and

the bottom water (Figure 3.3 and Figure 3.4), the simulated DIN:TP ratios matched the

seasonal patterns of the field DIN:TP ratios best, particularly in the surface water layer

well in magnitude and timing of bloom occurrences. The DIN:TP ratios and DIN:DIP

ratios successfully reproduced the seasonal patterns based on r values and Rs values

(Table 3.4). There was also a clear match between the simulated DIN:DIP ratios and the

field DIN:DIP ratios in the seasonal trends and inter-annual patterns in the surface water

(Figure 3.3a) and the bottom water (Figure 3.4a). However, the magnitude of the

simulated DIN:DIP ratios exceeded the observed values considerably, with values of the

observed DIN:DIP ratios in the surface water were on average two orders of magnitude

lower than the simulated values (Figure 3.3a). The range of the simulated DIN:DIP

ratios was 203.51-22701.16, which deviated from the range of the observed DIN:DIP

ratios (13.80-1261.22). In the bottom water, the values of the simulated DIN:DIP ratios

were on average one order of magnitude higher than the field values (Figure 3.4a),

although the simulated and field values matched the troughs well. Since the field PO4

concentrations in the surface water were around or below the threshold for P detection

limit in the lake, these extremely low simulated PO4 concentrations were inaccurate,

which further exaggerated the errors associated with the simulated DIN:DIP values.

Another factor that may account for the discrepancy between simulated and observed

DIN:DIP ratios is the relative higher DIN (NH4 and NO3) concentrations compared to

the low DIP (PO4) concentrations at some time points, even though DIN had good

validation results. In particular, compared to the magnitude of DIN:DIP ratio peaks,

some smaller discrepancies existed in the DIN:TP ratio peak values (Figure 3.3b), and

the range of the simulated DIN:TP ratios (0.51-20.59) also matched the range of the

field DIN:TP ratios (0.95-33.08). The main patterns of the simulated TN:TP ratios in the

bottom water had similar seasonal patterns compared to the field TN:TP ratios (Figure

3.4c), but their patterns were not as obvious as the main seasonal patterns of the

simulated DIN:TP ratios in the surface water (Figure 3.3c), although the range of the

simulated TN:TP ratios (33.44-86.52) matched the range of the field TN:TP ratios

(36.45-71.16). However, the simulated TN:TP ratios in the surface water did not always

56

succeed in matching the observed variation and timing of peaks, which results in the

low r value and Spearman rank correlation value (r = 0.436 and Rs = 0.383).

Figure 3.3: Simulated vs. observed monthly averaged time-series of a) DIN:DIP ratios, b) DIN:TP ratios and c)

TN:TP ratios in the surface water.

57

Figure 3.4: Simulated vs. observed monthly averaged time-series of a) DIN:DIP ratios, b) DIN:TP ratios and c)

TN:TP ratios in the bottom water.

Since phytoplankton usually exist within the upper 10 m layer of the water column,

based on visual inspection of simulated results and observed data (Figure 3.3) along

with statistical analysis of correlation (Table 3.4), the DIN:TP ratio was adopted as the

indicator for reflecting phytoplankton nutrient limitation in the surface water for the

following stoichiometric analysis.

58

3.4.2 Temporal trends in N:P stoichiometry

3.4.2.1 N:P stoichiometry of the phytoplankton community

The simulated seasonal C biomass patterns of the combined phytoplankton community

followed their iN:iP ratio patterns (Figure 3.5). Furthermore, the magnitude of the C

biomass variation of the combined phytoplankton community matched the magnitude of

the changes in their simulated iN:iP ratios. While these two patterns were similar, the

peaks of the phytoplankton C biomass slightly lagged behind their iN:iP ratio peaks.

Figure 3.5: Comparison between the simulated C biomass and iN:iP ratios of the combined phytoplankton

community.

The simulated iN:iP ratio patterns of the combined phytoplankton community also

followed the DIN:TP ratio patterns of the water column with a variable time lag

between peaks of these two ratios in different years (Figure 3.6a). The time lag in 1998

was smallest, and largest in 2001. Overall, the time lag that gave the highest correlation

between the DIN:TP ratios of the water column and the iN:iP ratios of the

59

phytoplankton community was two months (Table 3.5). Furthermore, the iN:iP ratio

magnitude of the combined phytoplankton community matched reasonably well with

the DIN:TP ratio magnitude of the water column in different years (r = 0.60 and

Rs=0.79). For example, when the DIN:TP ratio of the water column in April 1998 was

20.59 (the maximum DIN:TP ratio), the iN:iP ratio of the phytoplankton community

was 57.78 in May 1998, which was also the maximum iN:iP ratio of the combined

phytoplankton community. Considering the seasonality, we identified that the

correlation between DIN:TP ratios and iN:iP ratios in summer-autumn for the combined

phytoplankton community (r=0.67 and Rs=0.86) was higher than in winter-spring

(r=0.59 and Rs=0.67).

Figure 3.6: Comparison of the simulated water column DIN:TP ratios with a) the simulated iN:iP ratios of the

combined phytoplankton community, b) the bulk nutrient uptake N:P stoichiometry, and c) the bulk excretion N:P

stoichiometry.

The phytoplankton uptake and excretion nutrient ratios link the water column

60

stoichiomtery and the phytoplankton stoichiometry. There was significant seasonal

variation in the N:P ratios of these nutrient flux pathways (Figure 3.6 b,c), with both the

uptake and excretion N:P ratio patterns of the combined phytoplankton community

following the water column DIN:TP ratio patterns. The difference, however, was that

the excretion N:P ratio patterns of the combined phytoplankton community exhibited a

minor time lag (1-2 months) similar to the iN:iP ratio patterns of the combined

phytoplankton community.

3.4.2.2 N:P stoichiometry of individual phytoplankton groups

Although the simulated iN:iP ratio patterns of the combined phytoplankton community

followed the DIN:TP ratio patterns of the water column, this was not the case for

individual phytoplankton groups. The individual phytoplankton groups had various

seasonal iN:iP ratio patterns and different degrees of similarity with the DIN:TP ratio

peaks. For Peridinium (Figure 3.7a), Microcystis (Figure 3.7b), and nanophytoplankton

(Figure 3.7d), their iN:iP ratio patterns had double peaks within each year: a major peak

and a minor peak. The major iN:iP ratio peaks occurred after the water column DIN:TP

ratios peaked. Conversely, the minor peaks in iN:iP ratios of these groups occurred after

the DIN:TP ratios were at their lowest level. This double peak feature of their patterns

was in contrast to the peak features of the combined phytoplankton community (Figure

3.6a), Aphanizomenon (Figure 3.7c), and Aulacoseira (Figure 3.7e), which all showed

only a single peak each year.

The time lags with the highest correlation between the simulated iN:iP ratio patterns of

the individual phytoplankton groups and the DIN:TP ratio patterns of the water column

are summarised in Table 3.5. Although they varied between the different phytoplankton

groups, the highest correlated time lags for the simulated phytoplankton groups ranged

from 0 to 2 months. The highest correlation was found for Aphanizomenon (A3) with r

value of 0.71 and Rs value of 0.69 at a time lag of one month compared to other four

individual phytoplankton groups. However, the iN:iP ratio patterns of Aphanizomenon

were not as highly correlated with the DIN:TP ratio patterns as the combined

phytoplankton community. In contrast to Aphanizomenon, there was no time lag

between Aulacoseira’s (A5) iN:iP ratio peaks and the DIN:TP ratio peaks, and the

61

correlation was weak (r=0.32 and Rs=0.14).

Table 3.5: The impact of seasonal changes on the relationship between phytoplankton internal nutrient ratios and

DIN:TP ratios

Phytopalnkton

community

Aphanizomenon

Aulacoseira

Microcystis

Peridinium

nanophytoplankton

Time lag (months) 2 1 0 2 2 0

r (annual) 0.60 0.71 0.32 0.28 0.32 0.46

r (winter‐spring) 0.59 0.64 0.43 0.50 0.51 0.59

r (summer‐autumn) 0.67 0.68 0.50 0.65 0.43 0.19

Rs (annual) 0.79 0.69 0.14 0.24 0.34 0.58

Rs (winter‐spring) 0.67 0.70 0.49 0.57 0.59 0.66

Rs (summer‐autumn) 0.86 0.52 0.76 0.60 0.49 0.44

The seasonality had a different impact on the correlation between the iN:iP ratios of

phytoplankton and the DIN:TP ratios (Table 3.5). From the r values, the correlation for

Aphanizomenon in summer-autumn was almost the same as in winter-spring. The

variation in the magnitude of Aphanizomenon did not track the observed inter-annual

variation in the water column. The correlation values between DIN:TP ratios of the

water column and the iN:iP ratios for the combined phytoplankton community,

Aulacoseira and Microcystis in summer-autumn were higher than in winter-spring.

However, the correlation values for Peridinium and nanophytoplankton in summer-

autumn were lower than in winter-spring.

62

0

5

10

15

20

25

0

50

100

150

200

250

0 6 12 18 24 30 36 42 48 54

DIN:TP ratio

IN:IP ratio

month from 1997

(a)

IN:IP ratio DIN:TP ratio

0

5

10

15

20

25

0

2

4

6

8

10

12

14

0 6 12 18 24 30 36 42 48 54

DIN:TP ratio

IN:IP ratio

month from 1997

(b)


0

5

10

15

20

25

0

1

2

3

4

5

6

7

8

0 6 12 18 24 30 36 42 48 54

DIN:TP ratio

IN:IP ratio

month from 1997

(c)


63

Figure 3.7: Comparison of the simulated water column DIN:TP ratios with the iN:iP ratios of a) Peridinium, b)

Microcystis, c) Aphanizomenon, d) nanophytoplankton, and e) Aulacoseira.

3.4.2.3 N:P stoichiometry of heterotrophic organisms

The simulated patterns of the combined heterotrophic organism group (three

zooplankton groups and one bacteria group) followed the DIN:TP ratio patterns of the

water column without a time lag between these two ratio peaks in different years

(Figure 3.8). Furthermore, the magnitude of their iN:iP ratio variation was similar to the

0

5

10

15

20

25

0

10

20

30

40

50

60

70

80

90

100

0 6 12 18 24 30 36 42 48 54

DIN:TP ratio

iN:iP ratio

month from 1997

(d)


0

5

10

15

20

25

0

2

4

6

8

10

12

14

16

18

20

0 6 12 18 24 30 36 42 48 54

DIN:TP ratio

iN:iP ratio

month from 1997

(e)


64

magnitude of the changes in the simulated DIN:TP ratios. The good match between

iN:iP ratio patterns and DIN:TP ratio patterns is due to a shift in the abundance of the

various groups within the combined group, as the iN:iP ratios of the individual groups

were constant in the model.

Figure 3.8: Comparison of the simulated water column DIN:TP ratios with the iN:iP ratios of the heterotrophic

organisms (three zooplankton groups and one bacteria group).

3.4.3 Food web N:P stoichiometry

The internal N:P stoichiometry of the simulated phytoplankton groups varied not only

between groups but also depending on the microbial loop model (Figure 3.9). In MLAS,

the iN:iP ratio averages over the simulation period were separately 150:1 (Peridinium),

9:1 (Microcystis), 3:1 (Aphanizomenon), 55:1 (nanophytoplankton), and 15:1

(Aulacoseira) (Figure 3.9a). In MPLS, the average iN:iP ratio for Peridinium and

nanophytoplankton decreased to 107:1 and 47:1, and the remaining groups only had

minor changes. Therefore, with the microbial loop present and bacteria competing for

inorganic nutrients, the iN:iP ratios of Peridinium, Microcystis, and nanophytoplankton

decreased, whereas the iN:iP ratios of Aphanizomenon and Aulacoseira increased

slightly.

65

35:1

72:1

Zooplankton without Microbial Loop

Constant

Z3

28:1

Z2

20:1

Z1

27:1 Phytoplankton

Variable

A3

3:1

A2

9:1

A1

150:1

Nutrient pools

Adaptable

DIM

109:1

POM

57:1

DOM

307:1

A4

55:1

A5

15:1

31:1

6:1

28:1

(a)

WATER COLUMN

SEDIMENT LAYER 33:1

66

Figure 3.9: Comparison of simulated average molar N:P stoichiometry of phytoplankton, heterotrophic organisms and

nutrient pools of the water column between the a) microbial loop absent (MLAS) and b) microbial loop present

(MLPS) simulations (Note that the arrows on this diagram correspond to flow of C, N, and P).

The inorganic and detrital nutrient pools were different for MLPS compared with

MLAS. Given that the stoichiometries of DIM, DOM and POM were free to change,

they were quite different from the stoichiometry of the individual phytoplankton groups,

based on mass-balance constraints. In MLAS, the N:P ratios of the DOM and DIM

pools were 307:1 and 109:1, respectively. In MLPS, the N:P ratio of the DOM pool

increased dramatically to 3543:1, but the N:P ratio of the DIM pool decreased to 67:1.

The microbial loop thus had a greater impact on the nutrient pools than on the internal


5:1

0.27:1

5:1

20:1

13:1

Zooplankton with Microbial Loop

Constant

Z3

28:1

Z2

20:1

Z1

27:1 Phytoplankton

Variable

A3

4:1

A2

8:1

A1

107:1

Nutrient pools

Adaptabl

e DIM

67:1

POM

119:1

DOM

3543:1

A4

47:1

A5

16:1

22:1

12:1

2372:1

B

5:1

5:1

(b)

WATER COLUMN

SEDIMENT LAYER 33:1

67

The differences in N:P stoichiometry between the nutrient pools resulted in different

N:P ratios of the nutrient flux pathways between these ecosystem components. In

MLAS, for internal transformations, the average N:P ratio of algal nutrient uptake was

31:1, and the N:P ratio of algal excretion was 35:1. When bacteria competed with

phytoplankton for DIM in MLPS, the N:P ratios of algal nutrient uptake and excretion

decreased to 22:1 and 20:1, respectively. As a result, the N:P ratio of zooplankton

excretion decreased from 72:1 in MLAS to 13:1 in MLPS. Because the N:P ratios of

bacterial DOM uptake, the remineralization rate, and bacterial grazing rate were all 5:1,

bacteria were nutrient-balanced. The average N:P ratio of the sediment release rate was

33:1, and the net inflow and outflow contribution was negligible and only a small

percentage to the overall processes.

For all individual phytoplankton groups, the frequency histograms for the simulated

iN:iP ratios distribution of phytoplankton were analyzed (Figure 3.10) and also

compared to the long term average iN:iP ratios of the individual phytoplankton groups

in Figure 3.9b. The distribution of the simulated iN:iP ratios of the combined

phytoplankton community ranged from 4:1 to 85:1; their simulated iN:iP ratio peaks

with the highest frequency (20:1-30:1) were slightly higher than the Redfield ratio

(16:1), suggesting the lake is generally P limited. Though the simulated iN:iP ratios of

Peridinium and nanophytoplankton ranged widely from 50:1 to 210:1 and from 15:1 to

100:1, their five year simulated average iN:iP ratios (Peridinium: 107:1

nanophytoplankton: 47:1) fell within the filtered range of iN:iP ratio peaks with high

frequency (Peridinium: 55:1-140:1 nanophytoplankton: 20:1-85:1). In addition, the

iN:iP ratio distributions of Microcystis (3:1-13:1), Aphanizomenon (1:1-7:1), and

Aulacoseira (6:1-18:1) had a narrow range, their five year simulated average iN:iP

ratios (Microcystis:8:1 Aphanizomenon: 4:1 Aulacoseira: 16:1) also fell within the

filtered ranges of their iN:iP ratio peaks with the highest frequency (Microcystis:4:1-

12:1 Aphanizomenon: 3:1-7:1 Aulacoseira: 16:1-20:1).

68

Figure 3.10: Frequency histograms of iN:iP ratios for a) the combined phytoplankton community, b)

Peridinium, c) Microcystis, d) Aphanizomenon, e) nanophytoplankton, and f) Aulacoseira. (Note that the

shaded area indicates the user defined iN:iP range configured for each group.)

69

3.5 Discussion

3.5.1 Model Validation and Nutrient Ratios

Given the complexity of environmental factors affecting phytoplankton dynamics, the

model successfully captured the seasonal variability in nutrient dynamics and

phytoplankton biomass to a suitable level for the purposes of this investigation. The Rs

values demonstrated that the match between the simulated results and observed data in

both the seasonal trends and timing of peaks thereby indicating successful reproduction

of the observed seasonal patterns and inter-annual variability for the key state variables.

The simulated results for these chemical state variables overlapped with the observed

seasonal trends in the lake, except the magnitude of the PO4 values, which was under-

predicted by the model. As the molybdate method was used for measuring PO4

concentrations, the field PO4 concentrations were lower than or around the detection

limit of this method (Gal et al., 2009). Therefore, a relatively large difference between

the simulated and observed PO4 concentrations was unavoidable.

The simulated biological values matched the values for the field except some peaks in

the monitoring data. It is important to note that the model is laterally averaged and does

not account for the patchy nature of phytoplankton (Ng et al., 2011) and more complex

biological processes (Gal et al., 2009). For example, the simulated biomass peak of

Peridinium bloom in 1998 was much lower than the corresponding field data. In the

model assessment, we averaged the Peridinium biomass over a month, which smoothed

out some of the short terms peaks in the simulated phytoplankton C biomass. These

peaks were noticeable at the sub-daily to weekly time scale due to periodic

concentration build-up and strong vertical migration of this species. The laterally

averaged model output is also known to be lower than the field biological biomass from

Station A since it is not able to account for the patchy nature of this species linked to the

Jordan River inflow (Hillmer et al., 2008).

70

In the present study, three common types of N:P ratios (DIN:DIP ratios, TN:TP ratios,

and DIN:TP ratios) were compared for discriminating nutrient limitation of

phytoplankton growth. As nutrient supply ratios, DIN:DIP ratios are the best

phytoplankton nutrient limitation indicator in experimental work (Rhee 1978; Goldman

et al. 1979; Sterner and Elser, 2002). However, in field environments, when lakes are

usually P limited (Hart et al., 2000; Thingstad et al., 2005; Schindler et al., 2008), the

assessment of P limitation has become a challenge (Beardall et al., 2001b; Gillor et al.,

2010). One reason is that the complexity of phosphate chemistry causes difficulties in

measuring techniques and methodological biases for the available inorganic P at low

ambient concentrations (Bjoerkman and Karl, 1994; Rose and Axler, 1998; Beardall et

al., 2001b). The other reason is that the physiological status of phytoplankton groups

can be affected by perturbation of the photosynthetic rate with nutrient resupply and

partitioning into surface-adsorbed and intracellular P pools (Beardall et al., 2001a;

Sanudo-Wilhelmy et al., 2004; Gillor et al., 2010; Saxton et al., 2012). Considering that

some simulated DIN:DIP values were too high, the deviation from field DIN:DIP ratios

became magnified not only due to under predictions of the PO4 concentrations but also

due to the relative higher DIN concentrations compared to the low DIP (PO4)

concentrations, even where DIN had good validation results. Therefore, a relatively

large error was associated with the DIN:DIP values predicted by the model, which was

not practical for further comparisons in this study.

The DIN:TP ratio has been proposed as the best index for discriminating nutrient

limitation of phytoplankton in lakes (Morris and Lewis, 1988). Ptacnik et al. (2010) and

Bergström (2010) further compared a large range of different nutrient limitation

indicators and confirmed that the DIN:TP ratio was the best indicator and better than

TN:TP ratios for discriminating nutrient limitation of phytoplankton growth. In this

study, we further identified the DIN:TP ratios for phytoplankton nutrient limitation and

demonstrated the usefulness of the DIN:TP indicator for reflecting the relationship

between iN:iP ratios of phytoplankton and water column N:P ratios in a dynamic

aquatic environment. The other N:P ratios investigated in this study, which are usually

used as indicators to understand nutrient limitation of phytoplankton in lakes

(Bergström, 2010), showed less obvious seasonal patterns and weaker correlations with

71

phytoplankton. However, the simulated TN:TP ratios exhibited marked variations in the

surface water over the period and this variability was not reflected in observations.

3.5.2 N:P stoichiometry of phytoplankton

Currently, the abundance of phytoplankton in lake ecosystem models is usually

represented by C biomass or Chl-a, and the internal N:P ratios of phytoplankton are

seldom considered. The model DYRESM-CAEDYM showed that the simulated

abundance of phytoplankton represented by C biomass in Lake Kinneret followed the

changes in the simulated iN:iP ratios of the phytoplankton community with a small time

lag.

Many factors may limit the accuracy of model predictions of the iN:iP ratios of

phytoplankton blooms including the inappropriate use of experimental data for

modeling parameters, the complexity and scale of ecosystems, the level and type of the

nutrient inputs, and the spatial heterogeneity in environmental conditions (Ballantyne et

al., 2010). Therefore, it is important for lake modellers to conduct more rigorous

investigation of the correlation between the iN:iP ratios of phytoplankton groups and

different types of the water column N:P ratios.

Our use of the coupled model to analyze the stoichiometric variations in the

phytoplankton community shows that the iN:iP ratios of the combined phytoplankton

community reflect the patterns of the water column nutrient ratios in a dynamic

freshwater environment. This further supports Sterner and Elser’s hypothesis based on

laboratory experimental results (Goldman et al., 1979; Sterner and Elser, 2002), and

suggests that the iN:iP ratios of phytoplankton match the nutrient ratios of this meso-

eutrophic lake ecosystem across the temporal scale of seasons to years. Here, we also

confirm that the DIN:TP ratio is closely correlated to the nutrient ratio of phytoplankton

when a time lag is considered between nutrient uptake and biomass accumulation at the

community level.

However, the iN:iP ratio patterns of individual phytoplankton groups did not necessarily

relate to DIN:TP ratio patterns, since group specific seasonal iN:iP ratio shifts were

predicted to emerge. The iN:iP ratio patterns simulated for Aphanizomenon matched the

72

DIN:TP ratio patterns more closely than for the other groups. Aphanizomenon is a N2

fixing cyanobacteria dependent on nutrient ratios (Smith, 1983), for example, low N:P

ratios have been shown to contribute to Aphanizomenon blooms in Lake Kinneret

(Berman, 2001). Aphanizomenon may take approximately one month to adjust its

internal nutrient ratios to nutrient changes in the water column based on the time lag

noted here. This may explain why the iN:iP ratio peaks of Aphanizomenon occur when

the DIN:TP ratio reaches the minimum level. In contrast, Aulacoseira takes a shorter

period of time to adjust its iN:iP ratios in response to the changes of DIN:TP ratios of

the water column which fits with the observation of intensified winter Aulacoseira

blooms in Lake Kinneret (Zohary, 2004). The most significant change in the

phytoplankton community is Peridinium. Because Peridinium is one kind of

phytoflagellate, different specific growth rates of phytoflagellate manifests in different

iN:iP ratios (Sterner and Elser, 2002). Therefore Peridinium has a wide iN:iP ratio

range: at high growth rates, N:P ratios were close to the Redfield ratio; under P

limitation, as the growth rate of Peridinium declined, their iN:iP ratios can reach

extremely high values (>100), increasingly deviating from the Redfield ratio (Sterner

and Elser, 2002). Furthermore, the iN:iP ratio patterns of Peridinium, Microcystis, and

nanophytoplankton are characterised by annual double peaks, which suggests that their

internal nutrient ratios not only reflect the nutrient supply ratios in the water column but

also other factors that can influence the iN:iP ratios of these phytoplankton groups.

Other environmental factors can mediate the internal nutrient limitation patterns of

phytoplankton, such as temperature (Wohlers-Zollner et al., 2011), light (Sanches et al.,

2011), food web structure (Danger et al., 2008) and anthropogenic factors (Zohary,

2004).

3.5.3 Role of the microbial loop

The microbial loop is an important component of the food web that can influence the

iN:iP ratios of phytoplankton by regulating nutrient fluxes in the water column. In the

microbial loop, bacteria are a nutrient-rich source of food consumed by

microzooplankton. As bacteria are more like animals than plants in terms of N:P

homeostasis (Makino et al., 2003), their stoichiometric regulation affects the N:P ratios

73

of mineralised nutrients, and the N:P ratios of the materials grazed by

microzooplankton. When bacteria are consumed by microzooplankton, an input of

nutrient stoichiometry is in excess of the sotichiometry required for zooplankton.

Because zooplankton have fairly constant internal N:P stoichiometry (Sterner and Elser,

2002), this ultimately results in an enhancement of the nutrient excretion by

zooplankton to the food web. When the N:P ratio imbalance occurs within the food web,

C and P-depleted organic compounds may be accumulated in the organic matter pool

(Frost et al., 2002). This is reflected in our simulations in the form of an increase in the

N:P ratio of DOM by an order of magnitude in the presence of the microbial loop. In

this case, the P limitation in bacterial growth induced by the microbial loop, as observed

by Caron (1994), further propagates through the food web to make the availability of

DOP to bacteria an indirect limiting factor for phytoplankton growth.

The stoichiometry of nutrient recycling pathways as controlled by the stoichiometry of

the heterotrophic organisms illustrates the ability of bacteria to impact phytoplankton

stoichiometry, which, in turn, has ecosystem wide implications. Bacteria can switch the

nutrient limitation of phytoplankton growth and change their population structure

(Danger et al., 2007). Similarly, our results show that the microbial loop may further

impact the bulk iN:iP ratios of the phytoplankton community since phytoplankton

growth becomes nutrient limited at different levels when bacteria regulate the nutrient

recycling processes. This influence is demonstrated through modest changes to

phytoplankton, which are able to control their stoichiometry to a certain degree,

compared to significant changes in the DIM/DOM/POM pools of the water column.

Therefore, the microbial loop has a more significant impact on the N:P stoichiometry of

nutrient pools in the water column than the N:P stoichiometry of phytoplankton.

Mechanistically, some phytoplankton groups maintain their iN:iP ratios independent of

water column N:P ratio patterns. As a result, their physiological stress is caused by

nutrient limitation via regulatory processes occurring within the microbial loop. The

microbial loop regulates the stoichiometry of the nutrient fluxes between bacteria,

zooplankton, and pools of inorganic and organic matters thereby leading to different

nutrient limitations for different phytoplankton groups. The variations in the iN:iP ratios

74

of the five simulated phytoplankton groups suggest that the microbial loop has a more

significant impact on the N:P stoichiometry of Peridinium and nanophytoplankton than

on cyanobacteria (Microcystis and Aphanizomenon) and diatoms (Aulacoseira). Based

on the optimal allocation strategy between cellular P-poor resource-acquisition

machinery and P-rich assembly machinery of phytoplankton (Klausmeiser et al., 2004),

the iN:iP ratios of the individual phytoplankton groups vary according to their growth

status: the exponential growth phase and the competitive equilibrium phase. Although

Lake Kinneret is generally considered to be P limited (Hart et al., 2000), the real

ecosystem is generally a mixture of equilibrium and exponential growth phases of

different phytoplankton groups. Peridinium and nanophytoplankton have a wide range

of iN:iP ratios at their low growth rates, which facilitates the competitive equilibrium

phase to maintain their higher iN:iP ratios. In particular, the most significant change to

the phytoplankton community in response to the microbial loop is Peridinium, because

different specific growth rates caused by the microbial loop results in their different

iN:iP ratios. Conversely, Microcystis and Aphanizomenon have a narrow range of iN:iP

ratios at their high growth rates, which facilitates the growth phase to develop the lower

N:P ratios, although some N-fixing species often have high iN:iP stoichiometry. In

contrast to the other groups, the average iN:iP ratio of Aulacoseira closely matched the

Redfield ratio regardless of the overall P limitation. The results we present reveal the

pivotal role that the microbial loop plays in the lake food web by regulating the N:P

stoichiometry of the nutrient supply and uptake rates to determine phytoplankton

stoichiometry.

75

4 Bacterial competition with

phytoplankton has a positive impact

on primary production of

phytoplankton

4.1 Abstract

The carbon to nitrogen to phosphorus (C:N:P) stoichiometry of ecosystems is known to

influence broad-scale processes (e.g. carbon cycles) as well as the structure and the

function of food webs. However, very little is known about the C:N:P stoichiometry of

the interactions between the microbial loop and the phytoplankton community within

freshwater ecosystems. In order to better understand these interactions, we examined the

impact of bacterial uptake of inorganic nutrients, in the microbial loop, on the internal

C:N:P (iC:iN:iP) stoichiometry of the phytoplankton community and detritus pools. We

incorporated two bacterial nutrient uptake sub-models into a one-dimensional coupled

76

hydrodynamic-ecological model (DYRESM-CAEDYM) and applied this model to Lake

Kinneret (Israel) over a five year period (1997-2001). The iC:iN:iP stoichiometry of

microzooplankton and bacteria in the microbial loop was fixed to test the effect of

bacterial competition for inorganic nutrients on the variable stoichiometry of

phytoplankton. We found bacterial competition with phytoplankton for inorganic

nutrients in the microbial loop has a positive effect on the primary production of the

phytoplankton community, which contrasts with the traditional view of the negative

effect on primary production in aquatic food webs. However, not all simulated

individual phytoplankton groups necessarily increased their C biomass (e.g. N-fixation

species). These results provide an improved mechanistic understanding of bacterial-

phytoplankton interactions in aquatic ecosystems.

4.2 Introduction

Human activities have led to an increase in algal blooms in lakes and reservoirs around

the world affecting the food webs of these freshwater ecosystems. However, complex

microbial interactions in aquatic ecosystems result in planktonic diversity (Li et al.,

2012). The concept ‘microbial loop’ has been proposed to describe one of the complex

interactions between microzooplankton and bacteria as a key component in freshwater

ecosystems, which implies a strong modification of the classical aquatic food web

‘nutrients-phytoplankton-zooplankton’ paradigm (Azam et al., 1983; Moore et al.,

2004), because it affects the recycling of nutrients between different organisms to shape

phytoplankton succession patterns.

Conventionally, bacteria mainly remineralize organic matter (Ferrier and

Rassoulzadegan, 1994; Vadstein 2000). However, bacteria can also compete with

phytoplankton for inorganic nutrients (Currie and Kalff, 1984; Cotner and Wetzel,

1992; Kirchman, 1994). Under nutrient limiting conditions, bacteria compete with

phytoplankton for inorganic nutrients, and indirectly limit the primary production of

phytoplankton (Joint and Morris, 1982). In turn, microzooplankton graze on bacteria in

the microbial loop (Thingstad and Lignell, 1997). Therefore, nutrient availability for

phytoplankton becomes complicated by nutrient recycling processes. In particular,

77

phytoplankton and bacteria compete for the same limiting inorganic nutrients, which

results in an increase of phytoplankton biomass (Bratbak and Thingstad, 1985;

Brussaard and Riegman, 1998), instead of a decline in biomass (Joint et al., 2002).

While this seems paradoxical (Stone, 1990; Kirchman, 1994), very little research has

been directed towards resolving this paradox from perspective of ecological

stoichiometry.

Ecological stoichiometry provides the key for interpreting microbial interactions in

aquatic ecosystems (Sterner and Elser, 2002). Phytoplankton vary in terms of their

internal nutrient ratios because of many factors relevant to their growth (Michaels et al.,

2001; Karl et al., 2001; Frost et al., 2005). When inorganic N is insufficient to satisfy

their iN:iP ratios, some phytoplankton species will supplement N through N2 fixation

(Tyrrell, 1999). Moreover, the physiological constraints of bacteria have been shown to

regulate nutrient fluxes to phytoplankton in experimental cultures (Danger et al., 2007).

Based on Stone (1990)’s modeling framework and methodology, it is possible to study

the paradox about the competition between bacteria and phytoplankton for inorganic

nutrients. In this study we incorporated two sub-models into the coupled hydrodynamic-

ecological model (DYRESM-CAEDYM) configured for Lake Kinneret (Gal et al.,

2009) to re-examine and explore this paradox. From the perspective of ecological

stoichiometry, this study has tried to test if bacterial competition with phytoplankton for

inorganic nutrients has a positive effect on the C biomass (or primary production) of the

phytoplankton community, find out how different phytoplankton species respond to

environmental factors, and unravel why the competition has this positive effect.

4.3 Methods

4.3.1 Study site

Lake Kinneret (Sea of Galilee) is a large monomictic lake located in the Syrian-African

Rift Valley in north-eastern Israel. It is a meso-eutrophic lake with annual primary

production of 650 gC m−2 (Berman et al., 1995). Due to increased anthropogenic

78

stresses, the frequent occurrence of nuisance cyanobacterial blooms has become a

concern (Ballot et al., 2011). Major phytoplankton groups present in the lake include

Peridinium sp., Aphanizomenon sp., Microcystis sp., Aulacoseira sp., and

nanophytoplankton. Lake Kinneret was once well known for seasonal blooms of

Peridinium that regularly occurred until the late 1990s (Zohary et al., 1998; Zohary,

2004). However, observations over the last decade have seen a remarkable decline in

Peridinium due to fungal epidemics and a disruption in the historically stable

phytoplankton (Zohary, 2004); the contribution of cyanobacteria to the total

phytoplankton biomass has increased in summer.

4.3.2 Model overview

The coupled Dynamic Reservoir Simulation Model (DYRESM)-Computational Aquatic

Ecosystem Dynamics Model (CAEDYM) was run and validated against data during the

period from 1997 to 2001. Parameters for running DYRESM-CAEDYM in Lake

Kinneret were adopted from previous work on the lake (Gal et al., 2009; Li et al., 2012).

This hydrodynamic-ecological modelling platform was used for investigating the

influence of bacterial competition with phytoplankton for inorganic nutrients on the

primary production of phytoplankton (Figure 4.1).

Figure 4.1 The influence of the microbial loop on phytoplankton (the dash line refers to the traditional

algal-based pathways; the solid line refers to the microbial loop pathways).

79

4.3.3 Bacterial sub-models

Based on this platform, two bacterial sub-models are developed (Figure 4.2).

Sub-model 1 B-N (Bacteria without the uptake of inorganic nutrient):

In this sub-model, bacteria (BAC) were configured to only consume dissolved organic

matter (DOM) during the mineralization process. Under this condition, bacterial

biomass and mineralisation rates changed depending on temperature and organic matter

availability, but bacteria obtained the necessary amount of C, N and P exclusively from

the dissolved organic matter (DOM) pool.

Sub-model 2 B+N (Bacteria with the uptake of inorganic nutrient):

In this sub-model, bacteria not only obtained the necessary amount of C, N and P from

the DOM pool but also supplemented their internal nutrient requirement for their growth

by taking up dissolved inorganic matter (DIM) pool. In doing so, they competed with

phytoplankton for nutrient resources.

B‐N: B+N:

Figure 4.2 Conceptual diagrams highlighting the difference between B-N and B+N (PHYTO: the

phytoplankton community; DIM: dissolved inorganic matter; DOM: dissolved organic matter; BAC:

bacteria; the solid line refers to the cycling pathways between phytoplankton, bacteria, and nutrient pools;

the dot line refers to the inorganic nutrient uptake pathway by bacteria).

DIM DOM

BAC

PHYTO

DIM DOM

BAC

PHYTO

80

4.3.4 Model configuration

Nutrient and biological variables were depth-averaged over the five year period (1997-

2001). The averages of the C, N and P pools and fluxes were then converted to molar

ratios. The internal C:N:P (iC:iN:iP) stoichiometry of microzooplankton and bacteria in

the microbial loop was fixed to test the effect of bacterial competition for inorganic

nutrients on the stoichiometry of phytoplankton. Therefore, the iC:iN:iP ratios for the

biological variables were categorized into two types: (1) the constant stoichiometry for

bacteria and zooplankton; (2) the variable stoichiometry for phytoplankton within a user

defined range (Li et al., 2013). The chemical variables (POM, DOM, and DIM) were

regulated accordingly by microbial interactions.

4.3.5 Lake metabolism

To examine how the iC:iN:iP ratio of phytoplankton was relevant to lake metabolism

process variables, the Spearman rank coefficient (Rs) was used to explore the

correlation between the iC:iN:iP ratios of different individual phytoplankton groups and

their primary production and respiration. For primary production calculation, the

shortwave intensity at the surface water was converted to the photosynthetically active

component (PAR) based on Jellison and Melack (1993)’s assumption about the incident

spectrum and the Beer-Lambert Law about the light extinction coefficients for

phytoplankton, detritus, and dissolved organic matter (DOM). Since the respiration term

in CAEDYM lumps together respiration, mortality and excretion, a constant (Rresp) was

used to isolate respiration losses from the lumped parameterization (Hipsey and

Hamilton, 2008). Other constants (Rmor and Rexc) were also used to isolate the fraction

of mortality and excretion that goes into the DOM pool, and the fraction that goes into

the detritus pool.

4.3.6 Environmental factors

To further examine how the iC:iN:iP ratio of phytoplankton responds to environmental

factors, such as, light (I), nutrients (N and P), and temperature (T), the Spearman rank

coefficient (Rs) was also used to explore the correlation between the iC:iN:iP ratios of

different phytoplankton groups and environmental factors limitation. The environmental

81

limitation factors were defined as follows:

I limitation:

(1)

N limitation:

(2)

P limitation:

(3)

T limitation:

(4)

These limitation functions range from 0 (extreme limitation) to 1 (no limitation). As for

the description of these variables and parameters used in the above equations, readers

are referred to Li et al. (2013).

ss I

I

I

IIf 1exp)(

a

MIN

MINMAX

MAX

AIN

IN

ININ

INNf a

aa

a 1)(

a

MIN

MINMAX

MAX

AIP

IP

IPIP

IPPf a

aa

a 1)(

bTf aTkT )(20)(

82

4.4 Results

4.4.1 Model evaluation

In the water column, the simulated phytoplankton community and the main individual

phytoplankton groups in B+N had a good fit of the field data (Figure3.2c). The

validation details were in Chapter 3 (or refer Li et al., 2012). For the combined

phytoplankton community, the magnitude of C biomass and the timing of blooms were

successfully captured, although some discrepancies existed in their peak values and

timing. Furthermore, the model was successful in capturing the inter-annual variation

observed of different phytoplankton species in the lake (Figure 3.2c), except

nanophytoplankton. Because there was no significant pattern of the observed data for

nanophytoplankton, the majority of the simulated values were different from the field

data.

4.4.2 The effect of bacterial competition on primary production

Bacterial competition with phytoplankton for inorganic nutrients via the microbial loop

(B+N) increased the primary production of the phytoplankton community. The C

biomass of Microcystis (CYANO), Perdinium (DINOF), nanophytoplankton (CHLOR),

and Aulacoseira (FDIAT) increased in B+N compared to B-N (Figure 4.3a). In

particular, the C biomass of Peridinium (DINOF), Aulacoseira (FDIAT), and the

phytoplankton community (PHYTO) doubled. Moreover, the net C biomass of

Peridinium (DINOF), the main component phytoplankton species in Lake Kinneret,

increased most among the five simulated phytoplankton species. However, the C

biomass of Aphanizomenon (NODUL), N-fixation species, decreased in B+N compared

to B-N.

83

Figure 4.3 (a) Comparison of simulated C biomass of the combined phytoplankton community and individual

phytoplankton groups between B-N and B+N from 1997 to 2001; (b) Comparison of simulated C biomass of bacteria

(BAC), microzooplankton (ZOOP3), and seston between B-N and B+N from 1997 to 2001.

Bacterial competition with phytoplankton for inorganic nutrients via the microbial loop

not only had a positive influence on primary production but also the microbial loop

components themselves (Figure 4.3b). The C biomass of bacteria, microzooplankton

and seston increased in B+N compared to B-N. In addition, the C biomass of

microzooplankton tripled.

4.4.3 The effect of bacterial competition on ecological stoichiometry of food web

When bacteria had the ability to take up DIM in competition with phytoplankton, DOM

became enriched in N and P, while DIM became more limited in N and P. When the

iC:iP ratios and iN:iP ratios of DOM decreased from B-N to B+N, the iC:iP ratios and

iN:iP ratios of DIM increased from B-N to B+N (Table 4.1). When bacteria competed

with phytoplankton for inorganic nutrients, the iC:iP ratio and the iN:iP ratio of DIM

increased, which suggests that bacteria have an advantage in taking up PO4 from the

water column compared to phytoplankton.

While the iC:iN:iP ratios of bacteria and zooplankton were fixed, the iC:iN:iP

stoichiometry of the five simulated phytoplankton groups changed between B-N and

B+N. For Perdinium (DINOF), Microsystis (CYANO), and nanophytoplankton

(CHLOR), the iC:iP and iN:iP ratios increased from B-N to B+N; but the C:N ratio

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

C biomass (m

gC/L) B‐N B+N

0.00190.0071

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

BAC ZOOP3 Seston

C biomass (m

gC/L)

B‐N B+N(a) (b)

84

decreased from B-N to B+N For Aulacoseira (FDIAT), the iC:iP and iN:iP ratios also

increased from B-N to B+N; but the iC:iN ratios remained unchanged. For

Aphanizomenon (NODUL), the iC:iP ratio increased from B-N to B+N; but the iN:iP

and iC:iN ratios did not change.

The changes in the iC:iN:iP ratios of the bulk phytoplankton community (PHYTO) were

similar to the changes in the above iC:iN:iP ratios of Perdinium (DINOF), Microsystis

(CYANO), and nanophytoplankton (CHLOR), which are the most abundant

phytoplankton species in Lake Kinneret. The iC:iP and iN:iP ratios increased from B-N

to B+N; but the iC:iN ratio decreased.

When bacteria competed with phytoplankton for inorganic nutrients, the changes in

iC:iN:iP ratios of the combined phytoplankton community (PHYTO), Perdinium

(DINOF), Microsystis (CYANO), and nanophytoplankton (CHLOR) were similar to the

changes in the iC:iN:iP ratios of DIM of the water column. Among the five simulated

phytoplankton species, the iC:iN:iP ratios of the nanophytoplankton species (123:18:1)

was closest to the Redfield ratio (106:16:1). Although the changes in the iN:iP ratios of

the different phytoplankton groups showed a different trend to the change between B-N

and B+N, all the iC:iP ratios of phytoplankton increased from B-N to B+N.

4.4.4 The impact of bacterial competition on ecological stoichiometry of

phytoplankton

There was a linear relationship between iC:iP ratios and the iN:iP ratios of

phytoplankton in B-N (R2=0.9215) and B+N (R2=0.8592) (Figure 4.4). Their equations

were as follows:

B‐N: 3.196.8 iP

iN

iP

iC

(5)

B+N:

8.425.6 iP

iN

iP

iC

(6)

85

Figure 4.4 Linear regression of simulated iC:iN:iP ratios of phytoplankton in B-N and B+N.

As Lake Kinneret is P limited, the slopes of these equations between iC:iP ratios and

iN:iP ratios can roughly represent iC:iN ratios, especially at extremely low P

concentrations. The bacterial competition with phytoplankton for inorganic nutrients

also has a significant impact on the iC:iN ratios of phytoplankton.

R² = 0.8592

R² = 0.9215

0

100

200

300

400

500

600

700

800

900

0 20 40 60 80 100 120

iC:iP ratios

iN:iP ratios

B+N B‐N Linear (B+N) Linear (B‐N)

Table 4.1 Stoichiometric comparision between B-N and B+N.

Configurations

iC:iN:iP iC:iN iC:iP iN:iP

B-N B+N B-N B+N B-N B+N B-N B+N

DIM 9936:23:1 12423:67:1 425:1 186:1 9936:1 12423:1 23:1 67:1

DOM 616684:28475:1 148805:3543:1 22:1 42:1 616684:1 148805:1 28475:1 3543:1

POM 125:86:1 169:119:1 1:1 1:1 125:1 169:1 86:1 119:1

Bacteria (BAC) 45:5:1 45:5:1 9:1 9:1 45:1 45:1 5:1 5:1

Microcystis (CYANO) 51:4:1 70:8:1 11:1 9:1 51:1 70:1 4:1 8:1

Peridinium (DINOF) 530:59:1 779:107:1 9:1 7:1 530:1 779:1 59:1 107:1 Aphanizomenon (NODUL) 24:4:1 28:4:1 6:1 6:1 24:1 28:1 4:1 4:1 nanophytoplankton (CHLOR) 123:18:1 219:47:1 7:1 5:1 123:1 219:1 18:1 47:1 Aulacoseria (FDIAT) 204:10:1 322:16:1 20:1 20:1 204:1 322:1 10:1 16:1 Phytoplankton community (PHYTO) 93:11:1 207:28:1 8:1 7:1 93:1 207:1 11:1 28:1

Predators 207:27:1 207:27:1 8:1 8:1 207:1 207:1 27:1 27:1

Macrograzers 108:20:1 108:20:1 5:1 5:1 108:1 108:1 20:1 20:1

Microzooplankton 161:28:1 161:28:1 6:1 6:1 161:1 161:1 28:1 28:1

Total dissolved nutrients 13024:168:1 13232:88:1 77:1 151:1 13024:1 13232:1 168:1 88:1

Seston 93:31:1 112:36:1 3:1 3:1 93:1 112:1 31:1 36:1

When bacteria compete with phytoplankton for inorganic nutrients, the iC:iN:iP

stoichiometric dynamics of the phytoplankton community has been illustrated in Figure

4.5. The seasonal patterns of iN:iP ratios and iC:iP ratios were similar but different in

their magnitudes. Compared to the temporal trends of iC:iP ratios and iN:iP ratios, the

seasonal patterns of the iC:iN ratios were inverse, that is, the magnitude without large

changes between different years.

Figure 4.5 The simulated iC:iN:iP ratios of the phytoplankton community ( ‘·’ represents C:N ratios, ‘*’ represents

C:P ratios, ‘○’ represents N:P ratios).

4.4.5 Lake metabolism

Based on visual inspection (Figure 4.6 and Figure 4.7) and statistic analysis (Table 4.2

and Table 4.3), the iC:iN:iP dynamics of different phytoplankton groups in B+N

correlated variably to primary production and respiration. Both of the correlation

between the iC:iP ratios of Peridinium sp. and their primary production (Rs=0.129) and

the correlation between the iN:iP ratios of Peridinium sp. and their primary production

(Rs=0.042) were poor. There was a regular time lag between the major peaks of iC:iP

ratios and primary production; however, there were sometimes no time lags between the

major peaks of iN:iP ratios and the peaks of primary production but sometimes time

lags existed (Figure 4.6a). The correlation between the iC:iP ratios of Microsystis sp.

and the peaks of their primary production (Rs=0.340) was almost the same as the

correlation between the iN:iP ratios of Microsystis sp. and their primary production

(Rs=0.384). There was a time lag between both the peaks of iC:iP ratios and iN:iP ratios

and the peaks of primary production (Figure 4.6b). Both of the correlation between the

iC:iP ratios of Aphanizomenon sp. and the peaks of their primary production (Rs=0.657)

and the correlation between the iN:iP ratios of Aphanizomenon sp. and their primary

production (Rs=0.784) were higher than the other two phytoplankton groups because

the peaks of iN:iP ratios and iC:iP ratios matched the peaks of primary production well

(Figure 4.6c).

88

0

0.01

0.02

0.03

0.04

0

100

200

300

0 10 20 30 40 50

primary production

iN:iP

(a)

iN:iP primary production

0

0.01

0.02

0.03

0.04

0

1000

2000

3000

4000

0 10 20 30 40 50

primary production

iC:iP

month from 1997

IC:IP primary production

0

0.02

0.04

0.06

0.08

0

5

10

15

0 10 20 30 40 50 primary production

iN:iP

(b)


0

0.02

0.04

0.06

0.08

0

100

200

300

0 10 20 30 40 50

primary production

iC:iP

month from 1997

iC:iP primary production

89

Figure 4.6 Simulated primary production of different phytoplankton groups: (a) Peridinium, (b) Microcystis, (c)

Aphanizomenon.

Both of the correlation between the iC:iP ratios of Peridinium sp. and Microcystis sp.

and their respiration were higher than the correlation between their iN:iP ratios and their

respiration (Table 4.2). Although the major peaks of iC:iP ratios and iN:iP ratios

matched the major peaks of their respiration; however, some minor peaks of their iN:iP

ratios could not match the peaks of their respiration (Figure 4.7a & b). The correlation

between the iC:iP ratios of Aphanizomenon sp. and the peaks of their respiration (Rs=0.

406) was almost the same as the correlation between the iN:iP ratios of Aphanizomenon

sp. and their respiration (Rs=0.413). There was a time lag between both the peaks of

iC:iP ratios and iN:iP ratios and the peaks of their respiration (Figure 4.7c).

From the above analysis, the double peak features of iN:iP ratios and iC:iP features of

Peridinium sp. and Microcystis sp. caused their lake metabolism Rs values lower than

Aphanizomenon sp.

0

0.05

0.1

0.15

0

2

4

6

8

0 10 20 30 40 50

primary production

iN:iP

(c)


0

0.05

0.1

0.15

0

20

40

60

80

0 10 20 30 40 50

primary production

iC:iP

month from 1997

iC:iP primary production

90

0

0.01

0.02

0.03

0.04

0.05

0

50

100

150

200

250

0 10 20 30 40 50

respiration

iN:iP

(a)

iN:iP respiration

0

0.01

0.02

0.03

0.04

0.05

0

1000

2000

3000

4000

0 10 20 30 40 50

respiration

iC:iP

month from 1997

iC:iP respiration

0

0.02

0.04

0.06

0.08

0

5

10

15

0 10 20 30 40 50

respiration

iN:iP

(b)

iN:iP respiration

0

0.02

0.04

0.06

0.08

0

100

200

300

0 10 20 30 40 50

respiration

iC:iP

iC:iP respiration

91

Figure 4.7 Simulated respiration of different phytoplankton groups: (a) Peridinium, (b) Microcystis, (c)

Aphanizomenon.

Table 4.2 The Spearman rank correlation coefficients (Rs) between simulated iC:iN:iP ratios of phytoplankton and

lake metabolism processes.

Phytoplankton groups Primary production Respiration

iC:iP iN:iP iC:iP iN:iP

Peridinium 0.129 0.042 0.480 0.358

Microcystis 0.340 0.384 0.507 0.106

Aphanizomenon 0.657 0.784 0.406 0.413

0

0.05

0.1

0.15

0

2

4

6

8

0 10 20 30 40 50

respiration

iN:iP

(c)

iN:iP respiration

0

0.05

0.1

0.15

0

20

40

60

80

0 10 20 30 40 50

respiration

iC:iP

month from 1997

iC:iP respiration

92

4.4.6 Environmental factors

Different phytoplankton groups had different iC:iN:iP dynamics features: The iN:iP

ratios of Peridinium sp. and Microsystis sp. had double peaks per year: major peak and

minor peak. Their iC:iP ratios had one peak per year (Figure 4.7a & b). Moreover, the

timing of the peaks of iC:iP ratios matched the timing of their minor peaks of iN:iP

ratios. As for Aphanizomenon sp., the iC:iP ratios matched the iN:iP ratios well no

matter timing and magnitude (Figure 4.7c). The double peak features or single peak

features of iN:iP ratios and iC:iP ratios were in response to environmental factors (Table

4.3) in details as follows:

(1) For Peridinium sp., their iC:iP ratios were highly correlated to light (Rs=0.775) and

N (Rs=0.498) but their iN:iP ratios were most correlate to P (Rs=0.580) and equally

correlated to the rest environmental factors (Rs around 0.30-0.35). The timing of the

minor peaks of iN:iP ratios of Peridinium sp. matched the timing of the peaks of light

(Figure 4.7a).

(2) For Microcystis sp., their iC:iP ratios were most correlated to light (Rs=0.846) but

also highly correlated to the rest environmental factors (Rs around 0.50-0.65).Their

iN:iP ratios were only poorly correlated to P (Rs=0.261) and light (Rs=0.233). The

small time lags existed between the minor peaks of iN:iP ratios of Microcystis sp. and

the peaks of light (Figure 4.7b).

(3) For Aphanizomenon sp., their iC:iP ratios were highly correlated to light (Rs=0.744)

and temperature (Rs=0.625) and their iN:iP ratios were also highly correlated to light

(Rs=0.866) and temperature (Rs=0.659). The peaks of iN:iP ratios of Aphanizomenon

sp. matched the peaks of light well (Figure 4.7c).

Table 4.3 The Spearman rank coefficients (Rs) between simulated iC:iN:iP ratios of phytoplankton and

environmental factors.

Phytoplankton groups Light N P T

iC:iP iN:iP iC:iP iN:iP iC:iP iN:iP iC:iP iN:iP

Peridinium 0.775 0.320 0.498 0.312 0.278 0.580 0.089 0.350

Microcystis 0.846 0.233 0.635 0.065 0.496 0.261 0.595 0.079

Aphanizomenon 0.744 0.866 Null Null 0.194 0.232 0.625 0.659

93

0

50

100

150

200

250

0

500

1000

1500

2000

2500

3000

3500

0 10 20 30 40 50

iN:iP

iC:iP

month from 1997

(a)

iC:iP iN:iP

0

0.1

0.2

0

0.1

0.2

0 10 20 30 40 50

II

Light limitation

0

0.5

1

1.5

0

0.5

1

1.5

0 10 20 30 40 50

NN

N limitation

0

0.5

1

1.5

0

0.5

1

1.5

0 10 20 30 40 50

PP

P limitation

94

0

0.5

1

1.5

0

0.5

1

1.5

0 10 20 30 40 50

TTT limitation

0

5

10

15

0

50

100

150

200

250

300

0 10 20 30 40 50

iN:iP

iC:iP

month from 1997

(b)

iC:iP iN:iP

0

0.1

0.2

0

0.1

0.2

0 10 20 30 40 50

II

Light limitation

0

0.5

1

1.5

0

0.5

1

1.5

0 10 20 30 40 50

NN

N limitation

0

0.5

1

0

0.5

1

0 10 20 30 40 50

PP

P limitation

95

0

0.5

1

1.5

0

0.5

1

1.5

0 10 20 30 40 50

TT

T limitation

0

2

4

6

8

0

20

40

60

80

0 10 20 30 40 50

iN:iP

iC:iP

month from 1997

(c)

iC:iP iN:iP

0

0.1

0.2

0.3

0

0.1

0.2

0.3

0 10 20 30 40 50

II

Light limitation

0

1

2

0

1

2

0 10 20 30 40 50

NN

N limitation

0

0.5

1

1.5

0

0.5

1

1.5

0 10 20 30 40 50

PP

P limitation

96

Figure 4.8 The emergent property of the simulated iC:iN:iP ratios of three phytoplankton groups:(a) Peridinium, (b)

Microcystis, (c) Aphanizomenon, in response to environmental factors (I, T, N and P).

4.5 Discussion

Bacterial competition with phytoplankton for inorganic nutrients has a positive impact

on the primary production of phytoplankton via the microbial loop in Lake Kinneret.

This differs from the traditional view of primary production in oceanography, where the

competition between bacteria and phytoplankton is thought to have a negative effect on

the primary production of phytoplankton (Joint and Morris, 1982; Bratbak and

Thingstad, 1985; Joint et al., 2002). This paradox has been noted repeatedly (Bratbak

and Thingstad, 1985; Stone, 1990; Kirchman, 1994) but lacks a generally accepted

mechanistic explanation.

In Lake Kinneret, because of the ‘cancelling negative effects’ of the microbial food

web, the primary production of the phytoplankton community was positively impacted

by bacterial competition for inorganic nutrients. According to the indirect mutualism

theory (Boucher et al., 1982; Stone, 1990), the paradox results from the following

conditions: (1) microzooplankton graze on bacteria, which has a negative effect to

bacteria; (2) phytoplankton excrete organic matter that may stimulate bacterial growth,

which has a positive effect to bacteria; (3) bacteria compete with phytoplankton for

inorganic nutrients, which has a negative effect to phytoplankton. Through the 1 and 3

double negative interactions the microbial loop resulted in a net positive impact on

phytoplankton primary production. Moreover, there is widespread empirical evidence

that microzooplankton grazing on bacteria reduces bacterial abundances but increases

bacteria-mediated decomposition of organic matter (Sherr et al., 1988; Ratsak et al.,

1996; Wang et al., 2007). When bacteria compete with phytoplankton for inorganic

nutrients, bacteria become ‘enemies’ of phytoplankton; however, microzooplankton are

0

0.5

1

1.5

2

0

0.5

1

1.5

2

0 10 20 30 40 50

TTT limitation

97

‘enemies’ of bacteria in the microbial loop. Therefore, microzooplankton become

indirect ‘friends’ of phytoplankton and increase their primary production.

The nutrient recycling processes via the microbial loop explain why the bacterial

competition with phytoplankton has a positive effect on primary production only

suitable for non-N-fixation phytoplankton species. As Lake Kinnneret is P limited (Hart

et al., 2000), bacteria compete with phytoplankton for the same inorganic nutrients

(especially PO4), which enhances primary production of the non-N-fixation

phytoplankton species. Specifically, bacteria store more P in their cells and recycle N

faster, especially DON (Berman and Bronk, 2003). For the non-N-fixing phytoplankton

species, they need to take up more inorganic N for enhancing their primary production.

However, for the N-fixing species, Aphanizomenon sp. can fix N from the atmosphere.

Therefore bacteria take up inorganic N via the microbial loop, which does not influence

their primary production. As a result, the biomass of Aphanizomenon sp. decreased

(Figure 4.3a) but the iN:iP ratio was constant (Table 4.1), even if bacteria competed

with them for inorganic nutrients.

Our analysis of iC:iN:iP stoichiometric variations highlighted the significant impact of

bacterial competition with phytoplankton for inorganic nutrients on the iC:iN:iP ratios

of the simulated phytoplankton species under nutrient limiting conditions. Moreover,

the ability for bacteria to regulate their stoichiometry through the uptake of inorganic

nutrients can significantly impact the availability of inorganic nutrients and the overall

rate of organic matter mineralization (Li et al., 2012). Therefore, bacterial competition

with phytoplankton for inorganic nutrients in the microbial loop has a significant

influence on the functioning of aquatic ecosystems by recycling nutrients. When

bacteria compete with phytoplankton for inorganic nutrients and phytoplankton growth

increases, the ecosystem becomes unstable for a period. One reason is that bacteria are

better competitors for nutrients over phytoplankton; the other is that phytoplankton

excrete more dissolved organic carbon (DOC) for bacteria, when phytoplankton suffer

nutrient stress. However, the competition via the microbial loop indirectly supports

primary production of phytoplankton with definite advantages and restores the stability

of the ecosystem over time (Stone, 1990). Moreover, high iC:iN and iC:iP ratios of

DOM reduce growth and reproduction and alter related nutrient release (Frost et al.,

2002). Bacteria switch the nutrient ratios that limit algal growth and change the algal

composition (Li et al., in preparation; Danger et al., 2007). For these above reasons, the

historically stable phytoplankton assemblage in Lake Kinneret was observed to be

98

disrupted during 1997-2001, although the lake was well known for the once regular

occurrences of the dinoflagellate Peridinium gatunese (Zohary et al., 1998). Nowadays

the frequent occurrences of nuisance cyanobacterial species have become a concern

(Zohary et al., 2011), which suggests the dominant phytoplankton species in Lake

Kinneret ecosystem change from Peridinium dominant community to cyanobacterial

dominant community.

Our study expanded Stone (1990)’s simple mathematical modelling work at the

community-level to the whole food web by incorporating the microbial loop into the

coupled mechanistic model (DYRESM-CAEDYM). In Lake Kinneret, we further

assessed the role of bacterial competition with phytoplankton for inorganic nutrients via

the microbial loop. We found bacterial competition with phytoplankton for inorganic

nutrients has a positive impact on the phytoplankton primary production in Lake

Kinneret only suitable for non-N-fixation species.

For different phytoplankton groups, there are different correlation levels between their

dynamic iC:iN:iP ratios and lake matebolism processes (i.e. primary production and

respiration). Moreover, the iC:iN:iP ratios of phytoplankton have highly correlate to

some key environmental factors, especially light and nutrients. In the future, to further

examine how the abundance of iC:iN:iP ratios of phytoplankton at a particular time is

dependent on the abundance of different process variables and in response to

environmental factors, the convergent cross mapping (CCM) will be used to explore the

emergent property of the iC:iN:iP ratios of phytoplankton, and unravel the mechanism

of how bacterial competition with phytoplankton for inorganic nutrients influences the

primary production.

99

5 The role of the microbial loop in regulating nutrient availability and phytoplankton dynamics

5.1 Abstract

The recycling of organic material through bacteria and microzooplankton to higher

trophic levels, known as the ‘microbial loop’, is an important process in aquatic

ecosystems. In this study, the significance of the microbial loop in controlling nutrient

supply to phytoplankton is investigated in Lake Kinneret (Israel). Microbial interactions

were quantified using a coupled hydrodynamic-ecosystem model that accounted for the

physical and ecological processes controlling the cycling of carbon, nitrogen and

phosphorus through bacteria, phytoplankton and zooplankton. The model was validated

to a comprehensive field dataset and three microbial loop sub-model configurations

were used to understand the mechanisms by which it could influence phytoplankton: (a)

static bacterial biomass and mineralization rates, (b) dynamic bacterial biomass and

mineralization rates, but without ability for dissolved inorganic nutrient uptake by

bacteria, and (c) dynamic bacteria biomass with mineralization rates with the ability for

dissolved inorganic nutrient uptake when organic matter was poor in nutrient content.

The results were analyzed in terms of nutrient flux pathways and the patterns of nutrient

limitation on phytoplankton growth. Considerable variation in phytoplankton biomass

and dissolved organic matter demonstrated that the predictions were highly sensitive to

assumptions of the three configurations and the mechanisms by which phytoplankton

growth rates were affected. Organic matter phosphorus content was a critical factor

driving microbial loop processes and when bacterial growth became P-limited, bacterial

100

competition with phytoplankton switched phytoplankton internal nutrient limitation

from nitrogen or phosphorus, and led to changes in phytoplankton community

composition. We conclude that the microbial loop plays an important role in nutrient

recycling by regulating not only the quantity but also the stoichiometry of available

nutrients. It is therefore an important model component that should be carefully

parameterized when simulating phytoplankton succession and water quality dynamics in

freshwater ecosystems.

5.2 Introduction

One of the principal objectives for water quality management of freshwater bodies is to

reduce the magnitude and frequency of nuisance algal blooms. Excess nutrients are

generally implicated in the production of nuisance blooms since they fuel primary

production and organic matter accumulation. However, a simple ‘bottom-up’ (i.e.,

nutrient driven) view of algal blooms does not account for the complicated nature of

energy and nutrient pathways within aquatic food webs and the non-linearity of

ecosystem response to environmental changes (Jeppesen et al., 2005; Roelke et al.,

2007). A beneficial approach to decipher the driving processes of nuisance blooms is to

study aquatic ecosystems that have experienced considerable changes in patterns of

nuisance algal blooms. Lake Kinneret (Sea of Galilee, Israel) is a highly studied

freshwater ecosystem in this regard. The well-documented stable phytoplankton

succession pattern prior to the early 1990s (Pollingher, 1986; Berman et al., 1992;

Berman et al., 1995) has notably deteriorated and resulted in more frequent occurrences

of cyanobacterial blooms (Banker et al., 1997; Zohary, 2004; Schatz et al., 2005; Ballot

et al., 2011). In order to unravel the complexities that have led to this change, it has

become critically important to understand the manner in which nutrients move among

planktonic communities. In this study we employ a dynamic ecosystem model to

explore the pathways of nutrient transfers between the pelagic ecosystem components of

Lake Kinneret.

Much work in limnology is based on the classic ‘N-P-Z-D’ (nutrients-phytoplankton-

101

zooplankton-detritus) paradigm which assumes a relatively simple flow of nutrients

between autotrophic and heterotrophic pools. However, it is now well-documented both

in oceanographic and, to a lesser extent, in limnological applications, that higher order

predators such as crustacean zooplankton or fish can be supported by two paths: the so-

called ‘green’ (algal-based) and ‘brown’ (detrital-based) food web components (Moore

et al., 2004). The latter refers to the dynamics of the heterotrophic bacteria and the

microzooplankton grazers (defined here as size less than 125µm to account for rotifers,

ciliates, and juvenile macrograzers, Thatcher et al., 1993) – often termed the ‘microbial

loop’. This has been shown to play an important role in shaping carbon fluxes in lakes,

including Lake Kinneret (Stone et al., 1993; Berman et al., 2010), and in enhancing

nutrient cycling at the base of food webs (Hart et al., 2000; Hambright et al., 2007).

Less studied is how the microbial loop can affect patterns of phytoplankton growth and

its potential for shaping phytoplankton succession. There are four main mechanisms by

which microbial loop processes are thought to influence phytoplankton dynamics: (1)

the provision of bacterially mineralised nutrients for phytoplankton growth, (2) the

provision of an alternative food source since micrograzers prey on bacteria instead of

small phytoplankton (e.g., as evidenced by Hambright et al., 2007); (3) the excretion of

readily available nutrients by micrograzers that directly support primary production

(Johannes, 1965; Wang et al., 2009); (4) the competition of bacteria with phytoplankton

for inorganic nutrients when organic detritus becomes nutrient depleted (Barsdate et al.,

1974; Bratbak & Thingstad, 1985; Stone, 1990; Kirchman, 1994; Caron, 1994; Joint et

al., 2002; Danger et al., 2007). The relative significance of each of these mechanisms

remains unclear, and in particular how they interact in a dynamic environment. For

simplicity, it is often argued that the biomass of heterotrophic bacteria is fairly stable

and that the majority of bacterial production is lost to respiration (Cole, 1999). As a

result, most quantitative models of carbon and nutrient fluxes in freshwater ecosystems

essentially ‘lump’ microbial loop processes by assuming a static mineralisation rate of

organic material and simulating direct zooplankton consumption of detritus as a proxy

for the consumption of microzooplankton on bacteria (e.g., Janse et al., 1992; Saito et

al., 2001; Bruce et al., 2006; Mooij et al., 2010). These simplifications however do not

capture the range of nutrient ‘adjustments’ that occur as a result of microbial loop

processes, since stoichiometric composition of organisms and the range of the fluxes

between them in reality are not uniform or static (Elser & Urabe, 1999; Sterner &

Elser, 2002).

102

Ecosystem models of varying complexity are becoming increasingly common in the

management and study of water quality problems, such as those identified above, as the

benefits are well established. Models provide scientists and resource managers alike

with a platform to integrate their understanding of lake processes in a quantitative

manner, and provide a ‘virtual’ laboratory for exploring ecosystem processes (Van Nes

& Scheffer, 2005; Mooij et al., 2010). Following our increased understanding of the

importance of the microbial loop in recycling nutrients, representation of microbial loop

processes have been developed in marine ecosystem models (e.g., Faure et al., 2010),

however, there are fewer reports of freshwater ecosystem models that include explicit

incorporation of key microbial loop processes and no reported accounts that

simultaneously resolve the stoichiometry of the main ecosystem pools of carbon (C),

nitrogen (N) and phosphorus (P).

Specifically for Lake Kinneret, a steady-state C flux model was developed to examine C

cycling through the planktonic biota, including consideration of the microbial loop

(Stone et al., 1993; Hart et al., 2000). Recently, a one dimensional (1D) coupled

hydrodynamic-ecosystem model (DYRESM-CAEDYM) was presented by Bruce et al.

(2006), which focused specifically on the zooplankton dynamics and their contribution

to nutrient recycling within the lake. However, the model presented by Bruce et al.

(2006) had a simplistic representation of the microbial loop dynamics, and two

cyanobacterial species, Microcystis sp. and Aphanizonmimen sp., were also not included

within the simulation, but continue to remain a concern to the overall health of the

ecosystem (Zohary, 2004). Gal et al. (2009) expanded this model to include a dynamic

microbial loop parameterization and accounted for the two cyanobacterial species listed

above in order to study the impact of changes in nutrient loading on water quality trends

within Lake Kinneret. This study aims to further the analysis of Gal et al. (2009) to

specifically explore the effects of the microbial loop on the patterns of phytoplankton

growth and succession within the lake.

Since the microbial loop can regulate both the quantity and stoichiometry of nutrient

transfers (e.g. organic matter recycling), we hypothesize that inclusion of the microbial

loop in a numerical model not only impacts our ability to directly model the role of

zooplankton and bacteria in lake ecosystems, but also impacts our ability to simulate the

ratios of inorganic nutrients available to primary producers and predict algal succession

patterns. It is therefore the aim of this study to examine how microbial loop processes

regulate the nutrient fluxes between different groups of bacteria, phytoplankton and

103

zooplankton via nutrient recycling pathways and how these processes shape the

phytoplankton growth patterns in a freshwater ecosystem. To achieve this aim, three

microbial loop configurations have been analysed using the DYRESM-CAEDYM

model of Gal et al. (2009), as applied to Lake Kinneret.

5.3 Methods

5.3.1 Site description

Lake Kinneret (Sea of Galilee) is a large monomictic lake located in the Syrian-African Rift

Valley in north-eastern Israel. It covers an area of 170 km2, is 21 km long and 16 km wide and

has a maximum depth of 43 m, and has been the focus of considerable limnological research

over the past few decades. Major phytoplankton groups present in the lake include Peridinium

sp., Aulacoseira sp., Aphanizomenon sp., Microcystis sp., and nanophytoplankton. A number of

zooplankton species occur in the lake and can be grouped as rotifers, ciliates, metazooplankton

and larger predators. The maximum ciliate abundance is observed in autumn, generally

preceding a metazooplankton peak. Heterotrophic nanoflagellates are most abundant in winter

and spring, and least abundant in autumn. Bacteria numbers are highest during the decline of the

Peridinium gatunense (hereafter referred to as Peridinium) bloom and are the lowest during the

winter (Hadas et al., 1998). Lake Kinneret was once well known for seasonal blooms of

Peridinium that regularly occurred until the late 1990s (Zohary et al., 1998; Zohary, 2004;

Roelke et al., 2007). However, observations over the last decade have seen a remarkable decline

in Peridinium due to fungal epidemics and a disruption in the historically stable phytoplankton

(Zohary, 2004) the biomass of Aulacoseira blooms has increased in winter; the contribution of

cyanobacteria and nanophytoplankton to the total phytoplankton biomass has increased in

summer. Due to increased anthropogenic stresses, the frequent occurrence of nuisance

cyanobacterial blooms has become a concern (Ballot et al., 2011).

5.3.2 Model overview and approach

In order to examine how the microbial loop can influence patterns of phytoplankton

growth within Lake Kinneret, a one dimensional hydrodynamic-ecological model

(DYRESM-CAEDYM) was simulated and validated against field data from 1997-2001.

The results of three alternative microbial loop sub-model configurations were then

compared to evaluate the relative importance of the four key mechanisms by which the

microbial loop can affect phytoplankton succession patterns.

5.3.2.1 Hydrodynamic-ecological model platform

104

The Dynamic Reservoir Simulation Model (DYRESM) has previously been applied to

Lake Kinneret (Gal et al., 2003; Yeates & Imberger, 2003). The Computational Aquatic

Ecosystem Dynamics Model (CAEDYM) has been linked to DYRESM and utilized for

numerous freshwater lakes (Romero et al., 2004; Burger et al., 2007; Trolle et al.,

2008), including Lake Kinneret (Bruce et al., 2006; Gal et al., 2009; Makler-Pick et al.,

2011a,b).

The model setup and parameterization has been based on Gal et al. (2009) and used to

simulate variables including phytoplankton dynamics, bacterial production, C and

nutrient recycling, sediment-water interactions, and the relevant inflows, outflows and

mixing processes. The inflow data (daily volume, temperature, salinity, nutrient), the

outflow data (the total daily outflow volume from released outflow and local pumping),

and meteorological data (hourly short- and long-wave radiation, air temperature, vapour

pressure, wind speed and precipitation) were forced in the model (Gal et al., 2003; Gal

et al., 2009). The model uses dynamic intracellular stores that are able to regulate

phytoplankton growth based on Droop’s model. This model allows for phytoplankton to

have variable internal nutrient concentrations with dynamic nutrient uptake bounded by

minimum and maximum limits. The model can therefore capture the dynamic response

of phytoplankton stoichiometry to environmental conditions and food web structure. In

each configuration, five phytoplankton species are included, each with three state

variables (internal C, N, and P, denoted as A, AIN, and AIP, respectively): Peridinium

(A1), Microcystis (A2), Aphanizomenon (A3), nanophytoplankton (A4), Aulocaseira (A5).

Three zooplankton functional groups, Z, each with fixed internal nutrient ratios, were

also simulated: predatory copepods (Z1), macrograzers (Z2), microzooplankton (Z3).

Bacteria (B) were modelled as a separate state variable for two of the microbial loop

configurations. An additional ten nutrient variables (FRP, NO3, NH4, DIC, DOC, DON,

DOP, POC, PON, POP), dissolved oxygen (DO) and temperature (T) were also

modelled, giving a total of 40 key state variables (Table 5.1). In addition to the base

simulation presented in Gal et al. (2009), we implemented different microbial loop sub-

model configurations (Figure 5.1), as described below.

105

Table 5.1 Overview of the variables configured with DYRESM-CAEDYM for Lake Kinneret.

Notation CAEDYM Name Description Units

BIOGEOCHEMICAL VARIABLES

DOC DOCL Dissolved organic carbon concentration mg C L-1

POC POCL Detrital particulate organic carbon concentration mg C L-1

TN Total nitrogen concentration mg N L-1

PON PONL Detrital particulate organic nitrogen concentration mg N L-1

DON DONL Dissolved organic nitrogen concentration mg N L-1

NH4 NH4 Ammonium concentration mg N L-1

NO3 NO3 Nitrate concentration mg N L-1

TP Total phosphorus concentration mg P L-1

POP POPL Detrital particulate organic phosphorus concentration mg P L-1

DOP DOPL Dissolved organic phosphorus concentration mg P L-1

FRP PO4 Filterable reactive phosphorus mg P L-1

DO DO Dissolved oxygen concentration mg O L-1

BIOLOGICAL VARIABLES

NA Number of algal groups being simulated (=5) -

A Algal group index (1… NA) -

A1 DINOF Algae #1 (Dinoflagellate: Peridinium gatunense the main, bloom-forming species) C biomass concentration

mg C L-1

A2 CYANO Algae #2 (Cyanobacteria: Non N2 fixing group represented by Microcystis, toxin-producing species) C biomass concentration

mg C L-1

A3 NODUL Algae #3 (Cyanobacteria: Filamentous N2 fixing group represented mostly by Aphanizomenon ovalisporum and Cylindrospermopsis cuspis) C biomass concentration

mg C L-1

A4 CHLOR Algae #4 (Nanophytoplankton: A large suite of species that are nanoplanktonic in size and are readily grazed by zooplankton) C biomass concentration

mg C L-1

A5 FDIAT Algae #5 (Diatom: Aulacoseira granulata, a winter bloom forming filamentous diatom) C biomass concentration

mg C L-1

AIN 1 IN_DIN Algae #1 (Dinoflagellate: Peridinium) internal N concentration mg N L-1

106

AIN 2 IN_CYA Algae #2 (Cyanobacteria: Microcystis) internal N concentration

mg N L-1

AIN 3 IN_NOD Algae #3 (Cyanobacteria: Aphanizomenon) internal N concentration

mg N L-1

AIN 4 IN_CHL Algae #4 (Nanophytoplankton) internal N concentration mg N L-1

AIN 5 IN_FDI Algae #5 (Diatom: Aulacoseira) internal N concentration mg N L-1

AIP 1 IP_DIN Algae #1 (Dinoflagellate: Peridinium) internal P concentration mg P L-1

AIP 2 IP_CYA Algae #2 (Cyanobacteria: Microcystis) internal P concentration mg P L-1

AIP 3 IP_NOD Algae #3 (Cyanobacteria: Aphanizomenon) internal P concentration

mg P L-1

AIP 4 IP_CHL Algae #4 (Nanophytoplankton) internal P concentration mg P L-1

AIP 5 IP_FDI Algae #5 (Diatom: Aulacoseira) internal P concentration mg P L-1

NZ Number of zooplankton groups being simulated (=3) -

Z Zooplankton group index (1… NZ) -

Z1 ZOOP1 Zooplankton #1 (Predators: adult copepods, predatory rotifers) C biomass concentration

mg C L-1

Z2 ZOOP2 Zooplankton #2 (Large herbivores/macrozooplankton: cladocerans, copepodites) C biomass concentration

mg C L-1

Z3 ZOOP3 Zooplankton #3 (Microzooplankton: copepod nauplii, most rotifers, ciliates, heterotrophic flagellates) C biomass concentration

mg C L-1

ZIN 1 Zooplankton #1 (Predators: Copepods) internal N concentration

mg N L-1

ZIN 2 Zooplankton #2 (Macro-grazers: Cladocerans) internal N concentration

mg N L-1

ZIN 3 Zooplankton #3 (Micro-grazers: Rotifers/Ciliates) internal N concentration

mg N L-1

ZIP 1 Zooplankton #1 (Predators: Copepods) internal P concentration mg P L-1

ZIP 2 Zooplankton #2 (Macro-grazers: Cladocerans) internal P concentration

mg P L-1

ZIP 3 Zooplankton #3 (Micro-grazers: Rotifers/Ciliates) internal P concentration

mg P L-1

B BAC Heterotrophic bacterial C biomass concentration mg C L-1

BIN Heterotrophic bacterial internal nitrogen concentration mg N L-1

BIP Heterotrophic bacterial internal phosphorus concentration mg P L-1

107

Figure 5.1 Conceptual diagram highlighting the general ecosystem model configuration for Lake Kinneret (top) and processes and feedbacks for the three microbial loop models (bottom) explored in this study: (1) NOBAC (mineralization is not dependent on the bacterial biomass), (2) BAC-DIM (bacteria only take up DOM), and (3) BAC+DIM (bacteria not only take up DOM but also DIM) in with the aquatic ecological model CAEDYM (refer to Table 5.1 and Table 5.2 for notation).

108

5.3.2.2 Bacteria and microbial loop sub-models

Three alternative microbial loop sub-model configurations are tested to explore their

impacts on phytoplankton growth, and include: (1) NOBAC: bacteria state variable

replaced with constant organic matter mineralization rates and zooplankton grazing

directly on POM; (2) BAC-DIM: bacteria simulated with dynamic biomass and

hence mineralization rates, but unable to take up dissolved inorganic N and P; (3)

BAC+DIM: dynamic bacteria (as per 2) with an additional ability for supplementing

their internal nutrient requirement with dissolved inorganic N and P (PO4 and

NO3/NH4) if the available organic matter becomes nutrient depleted. The general

mathematical description of the mass balance for each of the variables relevant to

this analysis and the relevant notations are shown in Table 5.2. Parameterizations of

the common microbial loop process pathways for all configurations are described in

detail within the specific configurations. The microbial loop parameters used in each

configuration are summarized in Table 5.3. For other variable descriptions, process

representations and parameter values and justifications, readers are referred to Gal et

al. (2009).

109

Table 5.2 Equations for C, N and P within nutrients, organic matter, bacteria and zooplankton pools. Note that the pools and processes related to phytoplankton are not included here

for brevity since they are not different between the three configurations.

NOBAC BAC-DIM BAC+DIM

CA

RB

ON

POC

t Mz

z Ma

a DPOC SPOC GZ 3 POC

DOC

t DPOC UDOC EA

a EDOCZ

z DSF

DOC

t DPOC UDOC EA

a EDOCZ

z EB DSF

NIT

RO

GE

N

BIN

tUDON B UNH4

B UNO3B ENH 4 GZ 3(B) SB

Z3

tGZ 3(POC) EZ RZ 3 PZ1

POC

t Mz

z Ma

a DPOC SPOC

DOC

t DPOC UDOC EA

a EDOCZ

z EB DSF

B

tUDOC B EB RB GZ 3(B) SB

Z3

tGZ 3(B) EDOC RZ 3 PZ1

POC

t Mz

z Ma

a DPOC SPOC

B

tUDOC EB RB GZ 3(B) SB

Z3

tGZ 3(B) EDOC RZ 3 PZ1

PON

t Mz

z Ma

a DPON SPON GZ 3 PON

DON

t DPON UDON EA

a EDONZ

z DSF

ZIN3

t GZ 3(PON ) EDON PZ1

NH4

tUNH 4 A DSF NIT

NO3

tUDON UNO3 A DSF NIT DEN

PON

t Mz

z Ma

a DPON SPON

DON

t DPON UDON B EA

a EDONZ

z DSF

BIN

tUDON B ENH 4 GZ 3(B) SB

ZIN3

t GZ 3(B) EDON PZ1

NH4

t ENH 4 UNH 4 A DSF NIT

NO3

t ENH 4 UNO3 A DSF NIT DEN

PON

t Mz

z Ma

a DPON SPON

DON

t DPON UDON B EA

a EDONZ

z DSF

ZIN3

t GZ 3(B) EDON PZ1

NH4

t ENH 4 UNH 4 A, B DSF NIT

NO3

t ENH 4 UNO3 A, B DSF NIT DEN

110

PH

OS

PH

OR

US

D is particulate decomposition,

S is sedimentation (SPOM is particulate organic matter sedimentation, SB is bacterial sedimentation),

GZ3 is grazing by microzooplankton,

Mz is zooplankton mortality and messy feeding,

Ma is mortality of phytoplankton

RB is bacterial respiration, RZ3 is respiration of microzooplankton

Pz1 is predation by Z1,

EA is phytoplankton excretion of DOM,

EPO4 & ENH4 refer to bacterial mineralization of nutrients

EDOM is DOM excretion from zooplankton

DSF is dissolved sediment flux, NIT is nitrification, DEN is denitrification,

UDOM is dissolved organic matter uptake, either independent or linked to B biomass in the case of NOBAC and the other simulations, respectively.

UNH4, UNO3 and UPO4 refer to inorganic nutrient uptake, and the functions are designed to account for phytoplankton uptake only in the case of NOBAC and BAC-DIM, U(A), and phytoplankton and bacteria in the case of BAC+DIM, U(A,B).

POP

t Mz

z Ma

a DPOP SPOP GZ 3 POP

DOP

t DPOP UDOP EA

a EDONz

z DSF

ZIP3

t GZ 3(POP) EDOP PZ1

PO4

tUDOP DSF UPO4 A

POP

t Mz

z Ma

a DPOP SPOP

DOP

t DPOP UDOP B EA

a EDONz

z DSF

BIP

tUDOP B EPO4 GZ 3(B) SB

ZIP3

t GZ 3(B) EDOP PZ1 SZ 3

PO4

t EPO4 UPO4 A DSF

POP

t Mz

z Ma

a DPOP SPOP

DOP

t DPOP UDOP B EA

a EDOPz

z DSF

BIP

tUDOP B UPO4 B EPO4 GZ 3(B) SB

ZIP3

t GZ 3(B) EDOP PZ1 SZ 3

PO4

t EPO4 UPO4 A, B DSF

111

Table 5.3 Microbial loop related parameters used in the three model simulations (refer to Gal et al., 2009 for other parameter values). Note that the shaded parameters are those selected as key parameters for the sensitivity analysis.

Parameter Units Description NOBAC BAC-DIM BAC+DIM Comments / Other Literature / Justification

POM parameters

µPOCmax d-1 Maximum transfer of POCDOC 0.07 0.07 0.07 Gal et al., (2009) values adopted. 0.001[1]

µPONmax d-1 Maximum transfer of PONDON 0.01 0.01 0.01 0.02[1] ; 0.01-0.03[2]

µPOPmax d-1 Maximum transfer of POPDOP 0.1 0.1 0.1 0.01[1] ; 0.01-0.1[2]

POMda m Diameter of POM particles 5.5010-6 5.5010-6 5.5010-6 Gal et al., (2009) values adopted; 1.5010-5 [1]

POMDensity kg m-3 Density of POM particles 1040 1040 1040 Gal et al., (2009) values adopted; 1.08103 [1]

DOM parameters

µDOCmax d-1 Max mineralisation of DOCDIC 0.0008 N/A N/A Estimated from average output from BAC+DIM

µDOPmax d-1 Max mineralisation of DOPPO4 0.1 N/A N/A 0.01 [1] ; 0.01-0.1[2]

µDONmax d-1 Max mineralisation of DONNH4 0.008 N/A N/A calibrated values adopted; 0.02 [1]; 0.01-0.03[2]

Bacteria parameters

vB Arrhenius temperature scaling factor 1.08 1.08 1.08 Gal et al. (2009) values adopted.

Tstd oC Standard temperature 20 20 20 Gal et al. (2009) values adopted.

TOPTB oC Optimum temperature 30 30 30 Gal et al. (2009) values adopted.

TMAXB oC Maximum temperature 38 38 38 Gal et al. (2009) values adopted.

KDOB mg O2 L−1 Half saturation constant for dependence of

POM/DOM d i i DO1.5 1.5 1.5 Gal et al. (2009) values adopted.

fAnB - Aerobic/anaerobic factor 0.8 0.8 0.8 Gal et al. (2009) values adopted.

kBr d-1 Bacterial respiration rate at 20◦C N 0.12 0.12 Gal et al. (2009) values adopted.

µDECDOC d-1 Maximum bacterial DOC uptake rate N 0.05 0.05 Gal et al. (2009) values adopted.

KB mg C L−1 Half saturation constant for bacteria function N 0.01 0.01 Gal et al. (2009) values adopted.

KBIN mg N (mg C)-1 Internal C:N ratio of bacteria N 0.13 0.13 Gal et al. (2009) values adopted.

KBIP mg P (mg C)-1 Internal C:P ratio of bacteria N 0.0575 0.0575 Gal et al. (2009) values adopted.

KBe - DOC excretion N 0.7 0.7 Gal et al. (2009) values adopted.

µDIMupt DIM uptake N N Y

112

Micrograzer (Z3) parameters

KZIN mg N (mg C)-1 Internal ratio of nitrogen to carbon 0.2 0.2 0.2 0.2 [1] ; 0.24-0.27[3]

KZIP mg P (mg C)-1 Internal ratio of phosphorus to carbon 0.016 0.016 0.016 0.01 [1] ; 0.016-0.43[3]

Pzp - Preference of zooplankton for POC 1 0 0 Pzp = 1 in NOBAC as no bacteria present; 1[1] ; 0.75[4]

Pzb - Preference of zooplankton for bacteria 0 1 1 Z3 assumed to only graze on bacteria

gMAX mg C L−1(mg Z L−1)−1 d−1 Grazing rate 9 9 9 Gal et al. (2009) values adopted;

Kmf - Messy feeding(Grazing efficiency) 0.75 0.75 0.75 Gal et al. (2009) values adopted; 1 [1]

KZe d-1 Excretion fraction of grazing 0.25 0.25 0.25 Gal et al. (2009) values adopted; 0.2 [1]

KZ mg C L−1 Half saturation constant for grazing 0.4 1.5 1.5 0.5[1] [5] ; 0.1[5] ; 1.64[6]

MINPOC mg C L−1 Minimum grazing limit for POC 0.075 N/A N/A Assumed

MINBAC mg C L−1 Minimum grazing limit for bacteria N/A 0.05 0.05 Gal et al. (2009) values adopted.

TOPTZ oC Optimum temperature 24 24 24 Gal et al. (2009) values adopted.

TMAXZ oC Maximum temperature 30 30 30 Gal et al. (2009) values adopted.

[1] Bruce et al. (2006)

[2] Jorgensen & Bendoricchio (2001)

[3] Martin et al. (2005)

[4] Gophen & Azoulay (2002)

[5] Makler-Pick et al. (2011b)

[6] Stemberger & Gilbert (1985)

113

Common processes in all configurations:

POM hydrolysis: This process considers the enzymatic hydrolysis and

decomposition (DPOM) of particulate detrital materials depending on bacterial

biomass (B), if bacteria are simulated. It is similar in all configurations:

(1)

where µPOMmax is the maximum transfer of POM to DOM, and refers to one of

µPOCmax, µPONmax, or µPOPmax (Table 5.3).

DOM mineralization: The mineralization of DOM to DIM occurs in all

configurations. However, in configurations including bacteria the process adopts a

two-stage breakdown pathway as shown in the subsequent details of configuration 2

and 3. The general uptake rate of DOM by bacteria (U) is simulated as:

(2)

where µDECDOM is the maximum bacterial DOM uptake rate, and refers to one of

µDECDOC, µDECDON, or µDECDOP (Table 5.3).

Micrograzer grazing: All simulations include microzooplankton (Z3), which

graze either on a lumped detrital pool (configuration 1) or on bacteria, if the latter are

explicitly simulated (configuration 2 and 3). For simplicity in this study, we mainly

investigate how the microbial loop regulates the nutrient fluxes to influence

phytoplankton nutrient limitation. We have set up microzooplankton to only graze

on bacteria in configuration 2 and 3, even though it has been reported that

microzooplankton can also graze on small size phytoplankton (Hambright et al.,

2007).

Micrograzer excretion and respiration: In all configurations micrograzers

respire (R) and excrete (E) labile organic matter:

(3)

EDOC 1 kmf kZeGZ 3

(4)

DPOM POM max fBT1(T ) min fB

DOB (DO) fB B POM

UDOM DECDOM fBT1(T ) min fB

DOB (DO) fB B DOM

RZ 3 kZr fZ 3T 2 T Z3

114

where kZr is the respiration rate, kZe is the DOC excretion rate and GZ3 is the grazing

rate. Since the micrograzers are configured to have a stable C:N:P requirement, their

excretion of N and P is calculated to be dynamic in order to balance the other input

and output nutrient fluxes. This is numerically achieved by performing the excretion

at the end of the time step after other terms have been accounted for, according to:

EDON ZIN3

* Z3t1kZIN3

t where

ZIN3* ZIN3

t GZ 3 BIN EDON MZ 3 PZ1 (5)

EDOP ZIP3

* Z3t1kZIP3

twhere ZIP3

* ZIP3t GZ 3 BIP EDOP MZ 3 PZ1

(6)

where kZIN is the internal ratio of N to C, kZIP is the internal ratio of P to C of the

particular zooplankton class and PZ1 is the amount grazed by ZOOP1.

Configuration 1 – NOBAC:

This configuration assumes organic matter is mineralized at a rate that is not

dependent on the bacterial biomass (i.e., the bacterial biomass is constant and fB(B)

in POM hydrolysis Eq. (1) and DOM mineralization Eq. (2) are fixed at 1). This

approach moves C, N and P fluxes between DOM and DIM proportionally. Since

there are no bacteria simulated for micrograzers to graze, all the grazing preferences

were adjusted to consume POM in place of bacteria; thereby the bacterial biomass is

lumped within the detrital pool. Therefore the grazing rate of microzooplankton

simplifies to:

(7)

where POM is one of POC, PON, or POP. The grazing rate parameter (gMAX) was

adjusted to make GZ3(POM) in NOBAC approximately equal to GZ3(B) in

BAC+DIM (Table 5.3), to keep the general C flow and biomass patterns comparable

between these simulations.

Configuration 2 – BAC-DIM:

POMPOMK

POMgPOMG MAXZ

3

115

This configuration includes the heterotrophic bacteria state variable, B (as in Gal et

al., 2009). However, bacteria are configured to only consume DOM during the

mineralization process. Under this scenario, the bacterial biomass and their

mineralisation rate increase and decrease depending on temperature and organic

matter availability, but they must get the necessary amount of C, N and P from the

DOM pool. The basic equations for BAC-DIM are similar to NOBAC except the

inclusion of the bacterial equation and their associated growth and loss processes

(Table 5.2). Bacterial uptake of DOC is similarly defined using Eq. (2) with fB(B)

defined as:

(8)

Bacterial uptake of DON and DOP is based on the C mineralization rate. They are

converted according to the bacterial stoichiometric requirement of N and P (kBIN and

kBIP), but limited to the available pool to enforce mass conservation:

(9)

(10)

Note that if they cannot support the stoichiometric requirement in line with the

growth rate estimate, their growth rates (U) are also limited until both N and P

requirements can be satisfied with the available pool. In this configuration, POM

decomposition is also dependent on the changing bacterial biomass through fB(B) and

micrograzers solely graze on bacteria (B) rather than POM. Therefore GZ3(B) is set

as:

(11)

Configuration 3 – BAC+DIM:

Bacteria can supplement their internal nutrient requirement for their growth by

BK

BBf

BB

)(

UDON B U kBIN DON > U kBINt

DON DON < U kBINt

UDOP B U kBIP DOP > U kBIPt

DOP DOP < U kBIPt

BBK

BgBG MAXZ

3

116

taking inorganic nutrients. In doing so, they compete with phytoplankton for nutrient

resources. To investigate the role of this competition, this configuration extends

BAC-DIM by allowing the bacterial state variable to support their growth by taking

up NH4, NO3 and/or PO4, if there is insufficient N and P in the DOM pool to support

the prescribed growth rate. Therefore the bacterial uptake of N and P requires the

following additional terms (Table 5.2):

(12)

tkUUU

tkUUUUUkU

tNONHDONkUNO

U

BINNHDON

BINNHDONNHDONBIN

BIN

NO

0

< + --

4

443

343

(13)

tkUU

tkUUUkUU

BIPDOP

BIPDOPDOPBIPFRP 0

< - (14)

In configuration 2, if there is insufficient organic and inorganic N or P, the growth

rate, UDOC, is similarly limited to enforce mass balance.

5.3.3 Analysis procedure

5. 3.3.1 Model sensitivity

To determine the significance of the three microbial loop configurations on a number

of water quality variables in Lake Kinneret, the seasonal averages of these variables

from the upper 10m of the water column were computed over the period (1997-

2001) for winter–spring (January–June) and summer–autumn (July–December), as

per Gal et al. (2009). The simulated outputs of NOBAC, BAC-DIM, and BAC+DIM

on physical and chemical variables (T, DO, TN, TP, NO3, NH4, PO4) and biological

variables (A1-5, Z1-3) were statistically compared by One Way ANOVA (5%

significance level, SPSS software version 18.0) and Multiple Comparisons (POST

HOC, SPSS software version 18.0) to determine significant differences between the

simulated outputs of the three alternative microbial loop sub-models.

UNH4B

NH4 U kBIN DON NH4 t

U kBIN -UDON UDON < U kBINt

0 UDON = U kBINt

117

A sensitivity analysis for the impact of the microbial loop parameters on the

simulated outputs of the base configuration of BAC+DIM was conducted since this

configuration had the most complete process representation. The limited selection of

these parameters (the shaded parameters in Table 5.3) were chosen based on the

detailed sensitivity analysis of the complete set of ecological parameters by Makler-

Pick et al. (2011a) and the relevance to the key microbial loop processes investigated

here. These parameters were scaled one at a time by +20% and -20% to determine

the degree of sensitivity of both the state variable concentrations, and also the key

process pathways to the relevant parameters.

5.3.3.2 Quantification of pools, fluxes and limitation

To determine the influence of the microbial loop on the key nutrient pathways of

nutrient recycling processes, the pools and fluxes of C, N, and P cycles were

averaged over the simulation period of 56 months (Jan. 1997 – Aug. 2001). Nutrient

and biological state variables and fluxes were vertically integrated to calculate lake-

wide averages.

For each of the phytoplankton groups, the nutrient limitation functions, fa(N) and

fa(P), at a depth of 1m below the water surface were assessed to explore the impact

of the microbial loop on phytoplankton nutrient limitation. The functions were

calculated by the model based on the internal nutrient concentrations:

(15)

(16)

Which range from 0 (extreme limitation) to 1 (no limitation). A comparison was

made between results from the three microbial loop sub-models configurations,

NOBAC, BAC-DIM and BAC+DIM for each phytoplankton group as a function of

time.

a

MIN

MINMAX

MAXa AIN

IN

ININ

INNf a

aa

a 1)(

a

MIN

MINMAX

MAXa AIP

IP

IPIP

IPPf a

aa

a 1)(

118

5.4 Results

5.4.1 Comparison of model outputs

The physical, chemical, and biological simulated results using the explicitly

modelled bacteria configuration (BAC+DIM) have been validated by Gal et al.

(2009), to which we refer the reader for a detailed description of the model

performance against the range of available field data. Here we further adjusted the

microbial loop parameters as listed in Table 5.3 and produced another two

alternative microbial loop configurations (NOBAC and BAC-DIM). For all

configurations, the simulated water level, thermal structure, and dissolved oxygen

patterns were almost identical to the earlier version and matched the field data well.

The simulated major nutrient results (TN, TP, NO3, NH4, and PO4) for the three

configurations were noticeably different in the surface waters although similar in the

bottom water where sediment fluxes dominate (Figure 5.2a). Most noticeable was

the reduced surface water concentrations of NH4 and NO3 in the simulated output of

BAC-DIM. In contrast the BAC-DIM configuration simulated higher PO4

concentrations. Increase levels of TN were also simulated in both the NOBAC and

BAC-DIM configurations. There were some differences in the simulated

concentrations of the biological variables with BAC+DIM, however they were

almost identical to the earlier study and so not reproduced here. All three

configurations followed the general seasonal trends, with the most noticeable

differences being reduced bacteria and Peridinium and increased Aphanizomenon

concentrations in the BAC-DIM configuration output (Figure 5.2b), and these

simulations had increase discrepancy compared to the field data.

119

Figure 5.2 Comparison of model simulations for a) nutrient variables in the surface 10m (left) and bottom 10m

(right) of the water column, and b) for the nine microbial groups (mg C L-1 for A1-5 and B, and mg C m-2 for Z1-3).

Specifically, the impact of the three alternative microbial loop configurations on the

15 physical, chemical and biological variables was statistically analyzed by One

Way ANOVA and Multiple Comparisons (Table 5.4). Although the simulated

results for T and DO were not significantly different in the three configurations (p-

value>0.05), the simulated results for nutrients were significantly different: NH4,

TN, and TP of BAC+DIM were significantly different from BAC-DIM and NOBAC

6 12 18 24 30 36 42 48 540

0.1

0.2

NH 4 (

mg

/L)

TOP 10m

6 12 18 24 30 36 42 48 540

0.1

0.2

NO

3 (m

g /L

)

6 12 18 24 30 36 42 48 540

0.5

1

1.5

TN

(m

g /L

)

6 12 18 24 30 36 42 48 540

0.005

0.01

PO

4 (m

g /L

)

6 12 18 24 30 36 42 48 540

0.02

0.04

0.06

0.08

TP (m

g /L

)

month from 1997

6 12 18 24 30 36 42 48 540

1

2

NH 4 (

mg

/L)

BOT 10m

6 12 18 24 30 36 42 48 540

0.2

0.4

NO

3 (m

g /L

)

6 12 18 24 30 36 42 48 540

1

2

TN

(m

g /L

)

6 12 18 24 30 36 42 48 54

0

0.05

0.1

PO

4 (m

g /L

)

6 12 18 24 30 36 42 48 540

0.05

0.1

0.15

TP (m

g /L

)

month from 1997

6 12 18 24 30 36 42 48 540

0.5

1

1.5

2

mg

C/L

6 12 18 24 30 36 42 48 540

0.1

0.2

0.3

0.4

mg

C/L

6 12 18 24 30 36 42 48 540

0.1

0.2

0.3

0.4

mg

C/L

6 12 18 24 30 36 42 48 540

0.1

0.2

0.3

0.4

mg

C/L

6 12 18 24 30 36 42 48 540

0.1

0.2

0.3

0.4

mg

C/L

6 12 18 24 30 36 42 48 540

0.1

0.2

0.3

0.4

mg

C/L

6 12 18 24 30 36 42 48 540

1

2

3

4

mg

C/m

2

6 12 18 24 30 36 42 48 540

1

2

3

4

mg

C/m

2

6 12 18 24 30 36 42 48 540

1

2

3

4

mg

C/m

2

Peridinium

nanophytoplankton

Aphanizomenon Aulacoseira

Mycrocystis Bacteria

Predators Macrozooplankton Microzooplankton

month from 1997 month from 1997 month from 1997

NOBAC BAC‐DIM BAC+DIM

NOBAC BAC‐DIM BAC+DIM

(a)

(b)

120

(p-value<0.05); NO3 and PO4 of BAC+DIM were significantly different from BAC-

DIM (p-value<0.05), but similar to NOBAC. Biological variables were also

significantly different between these microbial loop configurations: Peridinium,

Aphanizomenon, and microzooplankton of BAC+DIM were significantly different

from NOBAC and BAC-DIM; copepods of BAC+DIM were significantly different

from NOBAC but similar to BAC-DIM; cladocerans of BAC+DIM and BAC-DIM

were significantly different from NOBAC but similar to each other; Microcystis of

BAC+DIM was also significantly different from BAC-DIM.

121

Table 5.4 Statistical analysis of water quality variables comparing the three microbial loop configurations by

ANOVA and Multiple Comparisons.

Dependent Variable

(I) Group

(J) Group

Mean Difference (I-J)

Std. Error

P-value

(pairwise)

P-value

(between groups)

T NOBAC BAC-DIM -.135 1.051 .898

.989 NOBAC BAC+DIM -.140 1.051 .894 BAC-DIM BAC+DIM -.005 1.051 .996

DO NOBAC BAC-DIM .052 .236 .826

.237 NOBAC BAC+DIM .372 .236 .117 BAC-DIM BAC+DIM .320 .236 .178

NH4 NOBAC BAC-DIM .048* .006 .000

.000 NOBAC BAC+DIM .023* .006 .000 BAC-DIM BAC+DIM -.025* .006 .000

NO3 NOBAC BAC-DIM .015* .005 .002

.003 NOBAC BAC+DIM .001 .005 .794 BAC-DIM BAC+DIM -.014* .005 .005

PO4 NOBAC BAC-DIM -.000* .000 .000

.000 NOBAC BAC+DIM .000 .000 .958 BAC-DIM BAC+DIM .000* .000 .000

TN NOBAC BAC-DIM -.072* .015 .000

.000 NOBAC BAC+DIM .141* .015 .000 BAC-DIM BAC+DIM .213* .015 .000

TP NOBAC BAC-DIM -.006* .001 .000

.000 NOBAC BAC+DIM -.008* .001 .000 BAC-DIM BAC+DIM -.003* .001 .005

Nanophytoplankton (A4) NOBAC BAC-DIM -.004 .007 .555

.126 NOBAC BAC+DIM -.013* .007 .048 BAC-DIM BAC+DIM -.009 .007 .163

Microcystis (A2) NOBAC BAC-DIM .004 .009 .669

.100 NOBAC BAC+DIM -.014 .009 .107 BAC-DIM BAC+DIM -.018* .009 .042

Peridinium (A1) NOBAC BAC-DIM .321* .057 .000

.000 NOBAC BAC+DIM .161* .057 .005 BAC-DIM BAC+DIM -.161* .057 .005

Aulacoseria (A5) NOBAC BAC-DIM .033 .020 .105

.125 NOBAC BAC+DIM -.005 .020 .795 BAC-DIM BAC+DIM -.0385 .020 .060

Aphanizomenon (A3) NOBAC BAC-DIM -.028* .005 .000 .000 NOBAC BAC+DIM -.012* .005 .009 BAC-DIM BAC+DIM .0156* .005 .001

Predators (Z1) NOBAC BAC-DIM .345* .168 .041 .001 NOBAC BAC+DIM .617* .168 .000 BAC-DIM BAC+DIM .272 .168 .107

Macrograzers (Z2) NOBAC BAC-DIM .585* .171 .001 .001 NOBAC BAC+DIM .552* .171 .001 BAC-DIM BAC+DIM -.033 .171 .848

Microzooplankton (Z3) NOBAC BAC-DIM .241* .068 .000 .001 NOBAC BAC+DIM .027 .068 .691 BAC-DIM BAC+DIM -.214* .068 .002

* The mean difference is significant at the 0.05 level.

122

5.4.2 Model parameter sensitivity analysis

Changes in the key parameters relevant to microbial loop processes had different

degrees of impact on various state variables and process pathways. Several

phytoplankton state variables, microzooplankton and the various process pathways

that connected them, were particularly sensitive to a number of key microbial loop

parameters (above the 20% sensitivity level) (Figure 5.3). In particular, Peridinium

was sensitive to the diameter of POM particles (POMda) and the bacterial optimum

temperature (TOPTB). In addition to POMda and TOPTB, Microcystis was sensitive to

the zooplankton internal N:C ratio (kZIN), and Aphanizomenon was also highly

sensitive to TOPTB (>50%). Microzooplankton biomass, bacterial grazing rates and

zooplankton excretion rates were strongly sensitive to KZe (> 30%), with mild

sensitivity to POMda, TOPTB, KZIN, and the half saturation constant for bacterial

function (KB). The DOM concentration was sensitive to POMda, particularly for N

(>50%), and the maximum bacterial DOC uptake rate (µDECDOC), and KB and kZIN (>

30%). Looking specifically at the process pathways, rates of algal excretion and algal

uptake were sensitive to TOPTB, particularly in the P cycle (>30%). To summarize, the

model output was most sensitive to changes in the microbial loop parameters

POMda, TOPTB, and KZe, which had a significant effect on DOM, the biomass of

Peridinium, cyanobacteria, heterotrophic bacteria, and microzooplankton.

5.4.3 Nutrient pools

The multi-annual and lake-wide volumetric nutrient pools were compared between

the three microbial loop configurations to understand how the microbial loop shifts

the partitioning of nutrients between different ecosystem compartments (Table 5.5).

In each configuration, the stoichiometry of the organic and inorganic matter pools

was free to change, whereas the stoichiometry of zooplankton and bacteria were

fixed, and the stoichiometry of phytoplankton was allowed to vary only within the

range prescribed by the minimum and maximum parameters of internal nutrient

ratios (unchanged from those in Gal et al., 2009). In each configuration, the DIC

pools were similar, but the DOC pool in BAC+DIM was significantly lower (1.79

mg C L-1) than in NOBAC (9.56 mg C L-1) and BAC-DIM (7.81 mg C L-1).

Similarly the DON and DOP pools in BAC+DIM were also lower than the

corresponding pools in NOBAC and BAC-DIM, even though bacteria were able to

take up DIN and DIP to meet their nutrient needs in this configuration. The N:P ratio

123

of DOM in NOBAC was 307:1, and with bacteria included (both BAC-DIM and

BAC+DIM), the N:P ratios increased significantly to 47215:1 and 3479:1

respectively. For configurations with dynamically simulated bacteria, the DIP pools

in BAC-DIM (6.410-3 mg P L-1) and BAC+DIM (5.210-3 mg P L-1) were higher

than that in NOBAC (3.610-3 mg P L-1), suggesting enhanced P availability for

phytoplankton uptake when bacteria are present. The POM pools in BAC-DIM and

BAC+DIM were also higher than those in NOBAC.

124

Figure 5.3 Sensitivity analysis of state variables and process rates for the C, N and P cycles presented as the lake average absolute change after a +/-20% parameter shift.

125

Table 5.5 Summary of average values (1997-2001) for C, N, and P contents (mg L-1) and N:P molar ratios of the various food web components in different microbial loop

configurations.

Configurations:

NOBAC

BAC-DIM

BAC+DIM

Variables C N P N:P C N P N:P C N P N:P

DIM 24.63 0.176 0.0036 109:1 24.61 0.068 0.0064 23:1 25.00 0.157 0.0052 67:1

DOM 9.56 0.319 0.0023 307:1 7.81 0.421 3.310-5 28475:1 1.79 0.050 3.110-5 3543:1

POM 0.09 0.028 0.0011 57:1 0.17 0.137 0.0035 86:1 0.26 0.214 0.0040 119:1

BAC (B) N/A N/A N/A N/A 0.06 0.007 0.0032 5:1 0.16 0.021 0.0091 5:1

Microcystis (A2) 0.02 0.004 0.0009 9:1 0.02 0.002 0.0010 4:1 0.03 0.004 0.0011 8:1

Peridinium (A1) 0.19 0.038 0.0006 150:1 0.04 0.005 0.0002 59:1 0.11 0.018 0.0004 107:1

Aphanizomenon (A3) 210-3 3.510-4 0.0002 3:1 0.02 0.004 0.0018 4:1 0.01 0.002 0.0008 4:1

Nanophytoplankton (A4) 0.07 0.022 0.0009 55:1 0.08 0.013 0.0016 18:1 0.08 0.021 0.0010 47:1

Aulacoseria (A5) 0.06 0.004 0.0006 15:1 0.03 0.002 0.0004 10:1 0.08 0.005 0.0006 16:1

Predators (Z1) 0.03 0.004 0.0004 27:1 0.02 0.003 0.0002 27:1 0.01 0.002 0.0001 27:1

Macrograzers (Z2) 0.06 0.012 0.0013 20:1 0.03 0.008 0.0008 20:1 0.04 0.008 0.0009 20:1

Microzooplankton (Z3) 0.01 0.002 0.0001 28:1 210-3 410-4 310-5 28:1 0.01 0.001 0.0001 28:1

Total dissolved 34.20 0.496 0.0059 186:1 32.43 0.489 0.0064 168:1 26.79 0.207 0.0052 88:1

Total particulate 0.53 0.115 0.0061 42:1 0.46 0.180 0.0126 31:1 0.79 0.295 0.0182 36:1

126

The biomass of bacteria and zooplankton varied in the different microbial loop

configurations, although the stoichiometry of zooplankton and bacteria were fixed at

5:1 (bacteria), 27:1 (copepods), 20:1 (cladocerans), and 28:1 (microzooplankton).

When bacteria were able to uptake dissolved inorganic nutrients in BAC+DIM, the

total bacterial biomass increased by 2.7 times that simulated in BAC-DIM. For

zooplankton, biomass of microzooplankton (Z3) was similar in NOBAC and

BAC+DIM and effectively absent in BAC-DIM. For predatory zooplankton (Z1)

simulated biomass was greatest in NOBAC and lowest in BAC+DIM and for

macrograzers (Z2) greatest in NOBAC and lowest in BAC-DIM.

The N:P stoichiometry of the five simulated phytoplankton groups varied

individually in response to the presence of bacteria in BAC-DIM and bacterial

uptake of inorganic nutrients in BAC+DIM. The molar N:P ratios of phytoplankton

in BAC+DIM (Peridinium 107:1; Microcystis 8:1; nanophytoplankton 47:1;

Aulacoseira 16:1) were also higher than their N:P ratios in BAC-DIM (Peridinium

59:1; Microcystis 4:1; nanophytoplankton 18:1; Aulacoseira 10:1). Conversely, for

Aphanizomenon, simulated biomass in BAC-DIM was higher than in BAC+DIM,

but no change was observed in their molar N:P ratios (4:1). The total phytoplankton

biomass in BAC+DIM was higher than that in BAC-DIM.

5.4.4 Nutrient fluxes

Simulated fluxes of C, N and P from the three microbial loop configurations

representing the dominant C, N and P recycling pathways demonstrate significant

differences in the relative magnitude of bacterial mineralization, zooplankton

excretion, zooplankton grazing, and bacterial competition with phytoplankton for

inorganic nutrients (Figure 5.4). For C recycling processes, in NOBAC, we

investigated the magnitude of mineralization and zooplankton excretion. Relative to

the algal CO2 fixation rate (defined as 100% for each simulation), the mineralization

rate returned 32.7% of the total DIC assimilated by phytoplankton, and zooplankton

excretion returned 29.7%. In BAC-DIM, we investigated the magnitude of bacterial

mineralization, zooplankton excretion, and zooplankton grazing. Bacterial

mineralization returned 43.3%, zooplankton grazing took up 0.9%, and zooplankton

excretion returned 10.8%. In BAC+DIM, we investigated the magnitude of bacterial

mineralization, zooplankton excretion, zooplankton grazing, and bacterial

competition with phytoplankton for inorganic nutrients. Bacterial mineralization

127

returned 77.3% of the total photosynthesized C, zooplankton grazing took up 17.5%,

and zooplankton excretion recycled 10.2% back to the water column.

For N recycling processes, in NOBAC, bacterial mineralization recycled 77.4% of

the total DIN taken up by phytoplankton, with zooplankton excretion being the

primary source of organic N with a similar relative magnitude (68.4%). In BAC-

DIM, bacterial mineralization recycled 47.2% of N, however only 17.6% was

supplied through zooplankton excretion. In BAC+DIM, the bacterial mineralization

returned 74.3%, with zooplankton excretion supplying 21.5%. When bacteria were

simulated in BAC-DIM and BAC+DIM, hydrolysis of particulate detritus was a

larger source of labile organic nitrogen than from zooplankton excretion (>50%).

For P recycling processes, in NOBAC, bacterial mineralization recycled 84.2% of

total DIP assimilated by phytoplankton, and zooplankton excretion provided 29.3%

of this P to bacteria. In BAC-DIM, however, bacterial mineralization recycled

94.0%, with zooplankton excretion contributing just 12.8%. When uptake of DIM by

bacteria was simulated in BAC+DIM, DIP uptake shifted significantly to 27.8% by

algae and 72.2% by bacteria. Of this total consumed PO4, bacterial mineralization

was responsible for 95.9%, with DOM supplied by zooplankton excretion

contributing just 10.9%.

In order to better quantify the relative contribution of the microbial loop to the

supply of inorganic nutrients for primary production, we computed how much

dissolved inorganic N and P come from recycling processes compared to the inflows

and sediment fluxes in BAC+DIM. The model predicted that 95.9% of dissolved

inorganic P was sourced from recycling within the water column, only 4.4% from

the sediments, and less from the inflows. For N, the model predicted a reduced

dependence on recycling (74.3%), higher sediment flux (22.3%) and a similar low

contribution (0.7%) from the inflows.

128

Figure 5.4 Summary of C, N, and P fluxes in a) NOBAC, b) BAC-DIM, and c) BAC+DIM (C pathways-black values; N pathways-red values; P pathways-blue values), presented as the lakewide average flux rates in brackets (×10-5mg L-1d-1). Values are also presented as % of total DIM taken up by phytoplankton and bacteria where relevant (The notations for variables and fluxes refer to Table 5.1 and Table 5.2).

129

5.4.5 Phytoplankton succession patterns

In conjunction with changes in temperature, light and vertical mixing, changes in

nutrient availability resulting from the variation in nutrient recycling processes leads

to variation in phytoplankton nutrient uptake and their nutrient limitation functions,

fa (N) and fa (P). The different patterns of seasonal variation in the nutrient limitation

of the five simulated phytoplankton groups within the three model configurations

highlight the potential for microbial loop processes to influence phytoplankton

patterns (Figure 5.5). For Peridinium, in NOBAC and BAC+DIM, the model

predicted N and P co-limitation. However, in BAC-DIM, it was evident that N

limitation was predicted to dominate most of the year. For Aulacoseira, in BAC-

DIM, N and P co-limitation was experienced most of year, but in NOBAC and

BAC+DIM, it switched from P limitation to N and P co-limitation. For Microcystis,

in NOBAC, P was the limiting factor for the algal growth, however, in BAC-DIM, it

was predicted to switch from P limitation to significant N limitation, and in

BAC+DIM it experienced significant P limitation with an annual occurrence of N

and P co-limitation in spring. For Aphanizomenon, in all three configurations, P

limitation dominated growth, since it is an N2-fixing species. For the

nanophytoplankton, in NOBAC and BAC+DIM, its growth was P limited with

annual N and P co-limitation but in BAC-DIM, growth switched between N

limitation and P limitation annually.

130

Figure 5.5 Comparison of nutrient limitation functions fa(N) and fa(P) defined in equations (15) and (16) respectively for the five simulated phytoplankton groups in a) NOBAC, b) BAC-DIM and c) BAC+DIM.

131

5.5 Discussion

5.5.1 Model performance and sensitivity

Given the complexity of the interactions affecting phytoplankton succession and

bloom dynamics, our ability to accurately predict all species accurately remains a

challenge. To date, there are limited modelling examples for a complete lake

ecosystem that confidently simulate the successional dynamics of phytoplankton and

zooplankton at the level of multiple trophic complexity. This is due to nonlinearity

of these complex models and a large number of parameters relative to poor data

availability (Arhonditsis & Brett, 2004; Rigosi et al., 2010; Mooij et al., 2010).

Nonetheless, our models were successful in capturing the seasonal dynamics and the

inter-annual variation of the key plankton functional groups in Lake Kinneret,

though their absolute concentrations tended to be under predicted. This is not

unexpected given we have adopted a laterally averaged one-dimensional approach

which is being compared to inherently patchy field data, particularly inherent in

Peridinium blooms (e.g., Ng et al., 2011; Hillmer et al., 2008). However, the models

were able to match the annual sequence and timing of the predicted peaks of these

blooms, particularly in the BAC+DIM configuration. Within this simulation the

time-scales of growth or decay of the biomass of biological variables generally

matched the observed data, and seasonal trends were accurately captured for physical

and chemical variables since the model responds significantly to the strong seasonal

forcing of the lake (Makler-Pick et al., 2011a). While we acknowledge further

improvements could be made, the focus of our study is to use the dynamic model to

help us gain insights into the significance of microbial loop processes on

phytoplankton growth in accordance with the approach suggested by van Nes &

Scheffer (2005) for application of complex models to explore ecological theory. For

this purpose, the model captures the variability of key physical, chemical and

biological processes to a suitable level to allow us investigate the mechanisms

governing the microbial interactions between the configurations.

Accordingly, different microbial loop configurations were found to have a

significant impact on the sensitivity of most state variables based on One Way

ANOVA and Multiple Comparison analysis. The predicted surface water nutrient

concentrations appeared to be the most sensitive variables to microbial configuration

132

(Figure 5.2a and Table 5.4), with particular sensitivity noted in the concentrations of

inorganic nutrients available for phytoplankton growth. Generally, it was noted that

in BAC-DIM inorganic nutrients were lower on average even though the total

nutrients were higher, suggesting accumulation of organic matter since it was not

being processed as efficiently. However, in the bottom water, nutrient variables were

not sensitive to the microbial loop configurations since the high concentrations of

nutrients result from sediment release, and biological activity is limited during the

long stratified period due to anoxic conditions.

Differences in predicted surface nutrient concentrations had considerable impact on

predicted plankton biomass and growth rates. The structure of the microbial loop

model had a significant impact on the total phytoplankton biomass as has similarly

been reported by Faure et al. (2010) for a coastal ecosystem. They demonstrated that

DIN (both NH4 and NO3) and phytoplankton biomass were strongly impacted by

microbial loop processes, such as bacterial remineralization and inorganic nutrient

uptake. Here, by extending the analysis to include phosphorus and several different

functional groups of phytoplankton and zooplankton, we have further identified

nutrients, Peridinium, Aphanizomenon and zooplankton as variables that are highly

sensitive to assumptions related to microbial loop configuration.

The parameter sensitivity analysis focused on several bacteria and microbial loop

parameters that were hypothesized to have the greatest impact on microbial

interactions relevant to the aims of this study, and based on the extensive sensitivity

analysis performed by Makler-Pick et al. (2011a). The dominant algal species

(Peridinium) and the nuisance algal species (Microcystis) in the lake were both

found to be highly sensitive to two microbial loop parameters: the optimum

temperature for bacteria growth (TOPTB) and the diameter of detrital particles

(POMda). According to Stoke's law, the diameter of POM particles influences the

settling rate of these particles. When the settling velocity is small, the residence time

becomes long, so that POM particles persist in the water column for a prolonged

period, allowing bacteria to more completely transform and mineralize organic

matter. Conversely, higher settling rates increases the loss of TN and TP to the

sediment from the photic zone. Thus identification of POMda as a significant

parameter signifies the importance of POM in the nutrient budget and contribution to

133

recycled nutrients. A similar finding was demonstrated by Makler-Pick et al.

(2011a) who identified that correct parameterization of POMda was important to

ensure a stable balance of TN and TP.

The sensitivity analysis indicated that both microzooplankton biomass and bacterial

growth were sensitive to the excretion fraction of the ingested material (KZe) grazed

by microzooplankton. Adjusting this excretion fraction parameter not only impacted

their own biomass and grazing rates, but also impacted the biomass of other

zooplankton groups and the phytoplankton community more broadly, including

Peridinium. Although Peridinium is not grazed directly by zooplankton, any

reduction in nutrient supply from micrograzers leads to reduced P availability and

ultimately reduced growth. These results suggest that the interaction between

phytoplankton and zooplankton is non linear and that there is a strong potential both

for top-down (i.e., grazing-mediated) and bottom-up (i.e., microbial loop nutrient

supply) control of phytoplankton. Interestingly, the smaller microzooplankton have a

significant overall impact shaping the food web structure in the model simulations

despite having the lowest biomass. These findings are in line with the conclusions of

Hart et al. (2000) and Hambright et al. (2007), who highlighted the critical role of

small micrograzers in the microbial loop processes. Since there exists a range of

uncertainty surrounding the parameterization of microzooplankton excretion with

large ranges being reported (Fasham et al., 1999; Faure et al., 2010), it remains an

important challenge for modellers to correctly parameterize.

5.5.2 Role of the microbial loop in regulating nutrient flows

In this study we investigated four mechanisms by which bacterial and microbial

loop processes influence primary production: 1) bacterial mineralization of organic

nutrients, 2) zooplankton excretion, 3) zooplankton grazing pressure, and 4) bacterial

competition with phytoplankton for inorganic nutrients when organic matter quality

is poor (i.e., nutrient depleted). By comparing fluxes between pools of C, N and P

we were able to gain insights into the role of the microbial loop in the recycling of

nutrients. Results of this study identified the different relative importance this role

has on the C, N and P nutrient flux pathways.

The complex assemblage of bacteria and zooplankton simulated in the Lake Kinneret

model allows us to study the relative affect of microzooplankton grazing and nutrient

134

excretion simultaneously. Due to their small size and high mass-specific grazing

rates, microzooplankton can transfer energy and nutrients via bacterial grazing to

higher trophic levels (Hart et al., 2000; Loladze et al., 2000) and therefore play an

important role in carbon and nutrient recycling (Stone et al., 1993; Dolan 1997;

Hambright et al., 2007). In turn, larger zooplankton grazing on microzooplankton

further provide organic matter for bacterial growth through excretion of nutrient rich

organic compounds (DOM) and fecal pellet production (POM) (Peduzzi & Herndl,

1992). From this point view, the recycling of organic nutrients is facilitated by

bacterial consumers rather than bacteria themselves, known as consumer-driven

nutrient recycling (CNR) (Elser & Urabe, 1999). In this study we have been able to

estimate the significance of this pathway and characterize the relative contributions

of upward and downward nutrient fluxes and the stoichiometry of these pathways.

For example, in BAC+DIM, as a fraction of algal uptake, microzooplankton

excretion was predicted to account for 10% of C, 22% of N and 11% of P returned

for mineralization, which was significantly larger than that supplied from algal

excretion for N and P (but not for C), and not matching the relative proportions

consumed through bacterial grazing (18% for C, 15% for N and 20% for P). This

therefore highlights the dissimilarity in the C, N and P cycles, and the importance of

nutrient adjustments that occur during these microbial interactions.

Bacterial mineralization also had a strongly regulatory effect on nutrient recycling,

and the model predicted more than 70% of N and around 95% of P available for

primary production was from bacterial mineralization of organic matter. These

figures are based on a five year average and relative contributions were found to vary

seasonally in response to temperature and organic matter availability. However, in

terms of carbon biomass, the bacterial population was found to be relatively stable.

A key result emerging from the simulations is that the lowest concentration of DOC

occurred in BAC+DIM, suggesting bacterial metabolism is enhanced when nutrient

supplementation is considered. Although bacterial growth is C limited in many lakes

(Coveney et al., 1992), bacteria in our simulations were mainly limited by P and also

occasionally co-limited by N and P, as indicated by the relative use of inorganic

nutrients. In the model, the DOM is assumed to be relatively labile; however in

reality different bioavailability of the various organic matter constituents may mean

that limitation due to a lack of suitably bioavailable carbon may also occur. There is

135

therefore scope for further extension of the model to understand how processes of

mineralization compare when multiple liability fractions of organic matter are

considered.

In freshwater ecosystems, the concentration of DON can often be higher than that of

DIN, and the DON pool plays an important role in providing N to both bacteria and

algae (Berman, 2001; Berman & Bronk, 2003), though the latter is not considered in

our model conceptualisation. In the present study, concentrations of DON were

higher than those of DIN in NOBAC and BAC-DIM, which fits with observations by

Berman & Bronk (2003). However, DON was lower than DIN in BAC+DIM where

bacteria biomass and mineralization rates were higher. As a result of increased DIN,

DOP became the limiting factor when competition by bacteria for inorganic nutrients

was included in the model configuration. Therefore, the variable stoichiometry of

organic matter, and different stoichiometric requirements of various process

pathways, leads to a complex interplay between the groups (Gaedke et al., 2002) and

future studies should further consider the significance of organic matter

stoichiometry, microzooplankton excretion rates and rates of nutrient immobilization

by bacteria when modelling planktonic food webs (Hessen, 1997; Muller et al.,

2001).

5.5.3 Impact of the microbial loop on phytoplankton growth

Bacterial competition for inorganic nutrients has a two-fold effect on phytoplankton

growth by limiting nutrient supply and regulating the N:P ratio of available nutrients.

In this study we compared the time series of nutrient limitation functions for the five

simulated phytoplankton groups for each of the three alternative microbial loop

configurations to decipher the effect of bacterial competition on phytoplankton

growth. Whilst most freshwaters are considered to be P-limited (Schindler et al.,

2008), Elser et al. (2007) asserts that N and P co-limitation is also prevalent. During

the simulation period in this study, Lake Kinneret had an average TN:TP ratio ~50:1

suggesting strong P limitation, as suggested by other authors. However, Gophen

(2011) argues N limitation is also occurring, potentially due to large fractions of

unavailable organic nitrogen distorting nutrient ratios (Ptanick et al., 2010). In this

study, growth of the five simulated phytoplankton groups in Lake Kinneret were

predominantly P limited or periodically N and P co-limited depending on the

microbial loop configuration. When organic matter became P depleted, it could not

136

support bacterial growth and therefore bacteria were supplementing with PO4 as

evident in the increased uptake rates (Fig. 4c). In addition, bacteria generally have

faster P uptake rates in comparison with phytoplankton (Berman, 1985). In our

model, while phytoplankton growth is nutrient limited we have not included a

specific mechanism for limiting the rate of uptake of PO4 by bacteria and they can

essentially outcompete the phytoplankton to meet their stoichiometric requirement.

Several phytoplankton groups experienced differences in the degree of N and P

limitation when bacteria were configured not to take up inorganic nutrients (BAC-

DIM), as opposed to when bacteria were also consuming inorganic nutrients

(BAC+DIM). For Peridinium growth, the BAC-DIM simulation was dominated by

N limitation, but in BAC+DIM, periods of phosphorus limitation also emerged

generally following periods of accelerated growth. For Aulacoseira, when not

competing with bacteria for nutrients (BAC-DIM) severe N limitation was simulated

this switched to predominant P limitation in BAC+DIM coinciding with Peridinium

blooms. Similarly for Microcystis and the mixed nanophytoplankton community

stronger N limitation simulated in BAC-DIM switched to predominant P limitation

in BAC+DIM. The model results indicate that stoichiometric regulation of bacteria

through DIM supplementation shifted patterns of phytoplankton nutrient limitation.

It therefore follows that bacteria-induced shifts in nutrient limitation can ultimately

influence the overall biomass and composition of the phytoplankton community

(Andersen et al., 2004). Indeed here we noted relative differences in the simulated

phytoplankton biomass, and in particular, when competition with bacteria for

inorganic nutrients was simulated (BAC+DIM), Peridinium dominated and

Aulacoseira also occurred in significant numbers. When this competition is

switched off (BAC-DIM), the model simulates reduced Peridinium and Aulacoseira

biomass, with a corresponding significant increase in Aphanizomenon. Microcystis

were also slightly reduced and the nanoplankton appeared to exhibit greater

seasonality. Based on these observations, our study indicates that phytoplankton

community composition is affected by changes in bacterial nutrient uptake

conditions (as triggered by microbial loop adjustment of C, N and P stoichiometry)

and the relative nutrient stoichiometric requirements of bacteria versus

phytoplankton. This further illustrates the role that the microbial loop plays in

shaping algal biomass and patterns of community composition.

137

6 Conclusion

This study improves the current understanding of eutrophication and algal blooms by

studying microbial interactions and their roles in regulating the pathways of C and

nutrient cycling processes in aquatic ecosystems with ecological modelling tools. It

provides insights into the nutrient flux pathways between viruses, bacteria,

phytoplankton and zooplankton, and the key biogeochemical processes that

ultimately shape phytoplankton succession patterns.

6.1 Summary of research findings

This study has developed several ecological models based on the classic ‘Nutrient-

Phytoplankton-Zooplankton-Detritus’ (NPZD) model for unravelling microbial

interactions in aquatic ecosystems. The ‘Nutrient-Phytoplankton-Viruses-Detritus’

(NPVD) model was firstly developed to compare the influence of zooplankton

mediated mortality and virus mediated mortality on phytoplankton. The ‘Nutrient-

138

Phytoplankton-Zooplankton-Detritus+Viruses’ (NPZD+V) model, the ‘Nutrient-

Phytoplankton-Zooplankton-Detritus+Bacteria’ (NPZD+B) model, and the

‘Nutrient-Phytoplankton-Zooplankton-Detritus+Viruses+Bacteria’ (NPZD+VB)

model were built for describing the viral shunt, the microbial loop, and the effect of

the viral shunt short circuit the microbial loop in aquatic ecosystems.

Using Lake Kinneret (Israel) as a study site, the Microbial Loop Absent Scenario

(MLAS) and the Microbial Loop Present Scenario (MLPS) sub-models were

incorporated into a one dimensional coupled hydrodynamic-ecosystem model

(DYRESM-CAEDYM) for exploring the impact of the microbial loop on the N:P

ratios of phytoplankton. The DYRESM-CAEDYM model was validated to a

comprehensive dataset in the lake over a five year period (1997-2001). Two bacterial

nutrient uptake sub-models were further designed for examining the impact of

bacterial uptake of inorganic nutrients on the internal C: N: P stoichiometry of the

phytoplankton community and detrital and dissolved nutrient pools. Finally, three

microbial loop sub-model configurations were compared to understand the

mechanisms by which it could influence phytoplankton succession patterns.

The main research findings addressed in this thesis are:

Compared to the NPZD model, the developed NPVD model indicates that

virus mediated mortality on phytoplankton via infection and lysis is as

important as zooplankton mediated mortality on phytoplankton via grazing.

When the microbial interactions are included (e.g., the microbial loop, the

viral shunt), the NPZD+B model and the NPZD+V model capture the

positive impact of the microbial loop on phytoplankton growth in aquatic

ecosystems and the movement of nutrients catalysed by the viral shunt from

phytoplankton to detritus.

The NPZD+VB model indicates that the viral shunt short circuits the

microbial loop via viral infection and lysis, and thereby increases the transfer

of C and nutrients from phytoplankton and bacteria to detritus.

Based on the numerical stoichiometric analysis, the seasonal patterns of

phytoplankton nutrient ratios reflect patterns of water column nutrient ratios,

in particular the DIN:TP ratios.

Bacterial competition with phytoplankton for inorganic nutrients has a

139

positive effect on the primary production of the phytoplankton community.

The microbial loop regulates the nutrient flux pathways between different

groups of bacteria, phytoplankton and zooplankton to shape phytoplankton

succession patterns within freshwater ecosystems. Especially, the phosphorus

content of the dissolved organic matter pool is a critical factor driving

microbial loop processes.

As few model studies have directly simulated the role of the microbial loop in

nutrient recycling in lakes, the importance of understanding the ‘bottom-up’

processes has been recognised. In particular, the microbial loop influences nutrient

recycling processes in aquatic ecosystems. These results help provide an improved

mechanistic understanding of ‘bottom-up’ control of algal blooms via microbial

interactions, and ecological stoichiometry for viral-bacterial-phytoplankton-

zooplankton interactions to protect water quality in an aquatic environment.

6.2 Implications for water quality management

The improved ecological models for microbial interactions in aquatic ecosystems

help guide the design of the modelling structure and the monitoring program, which

support real-time decision-making about the plankton dynamics and ensure that the

sampling locations and frequency are focused on the areas that present the largest

risk of algal blooms. Moreover, the microbial loop plays an important role in nutrient

recycling by regulating not only the quantity but also the stoichiometry of available

nutrients. It is therefore an important water quality model component that should be

carefully parameterized when simulating phytoplankton succession and water quality

dynamics in freshwater ecosystems.

Although the microbial loop processes affect the elemental composition of

phytoplankton communities, the internal N:P (iN:iP) ratio patterns of the

phytoplankton community reflect their C biomass patterns. This can help build our

understanding of the relationship between different forms of N:P ratios in the water

column and algal internal N:P ratios and C biomass in the real world. This in turn

can provide an effective means to resolve water quality problems in lake ecosystems.

140

The stoichiometry of the total phytoplankton biomass in Lake Kinneret is closely

correlated to the nutrient status of the water column. Based on the numerical

stoichiometric analysis, the seasonal patterns of phytoplankton nutrient ratios reflect

the seasonal patterns of water column nutrient ratios. However, the iN:iP patterns of

individual phytoplankton groups did not necessarily relate to nutrient ratios in the

water column, which highlight that simply inferring the limitation of particular

phytoplankton groups based on the water column nutrient stoichiometry may be

misleading.

Currently, many types of N:P ratios have been used for discriminating nutrient

limitation of phytoplankton growth. Because of water depth and hydrodynamics,

different nutrient ratios have different usages for describing the relationship between

nutrient supply in the water column and phytoplankton nutrient limitation. In this

study, DIN:TP ratios in the water column has been demonstrated to be a useful

indicator for reflecting the N:P stoichiometry of the phytoplankton community as a

whole in the surface water of Lake Kinneret. The seasonality has a significant impact

on the correlation between the iN:iP ratios of the combined phytoplankton

community, Microcystis, and nanophytoplankton, and the DIN:TP ratios of the water

column.

The findings of the hydrodynamic-ecological model are not only suitable for Lake

Kinneret but also are potentially suitable for other aquatic ecosystems, in which

microbial interactions play an important role in biogeochemical cycles. Ultimately,

this improved understanding of the relationship between the iN:iP ratios of

phytoplankton and the N:P ratios of the water column can help develop more

accurate nutrient limitation metrics and ecological models for predicting algal

blooms in aquatic ecosystems.

6.3 Recommendations for future work

The serial NPVD, NPZD+V, NPZD+B, NPZD+VB models based on the classic

NPZD model can be configured in a way that facilitates the user to develop nutrient

mass balance equations in specific aquatic ecosystem modelling configurations.

These simple models from this study cannot be directly extrapolated to natural

141

aquatic ecosystems, because they do not consider allochthonous nutrients and

detritus, which highly depend on the situation of a study site. However, they clearly

demonstrate some important microbial interactions influencing plankton dynamics.

These improved models help understand and describe the dynamics of natural viral-

bacterial-phytoplankton-zooplankton community and their interactions in aquatic

ecosystems. In particular, the viral shunt short circuits the microbial loop, which may

constitute new important areas about the indirect impact of microbial interactions on

algal blooms for protecting water quality by ecological modelling in future research.

The model application to Lake Kinneret performed well. However, it does suffer

from a few of shortcomings. While the model successfully simulated many

components of the food web and provided valuable insights into the processes

difficult to study in the site or laboratory, there were some discrepancies between

time serial simulations and field observations. Only in the presence of additional

process-based data will it be possible to illuminate some of those discrepancies. For

example, there was still a discrepancy between the simulated C:N:P stoichiometry,

the field C:N:P stoichiometry, and the literature values of some phytoplankton

groups. In the future, the dynamic relationship between the growth rate and nutrient

quota should be considered based on variants of Droop models to resolve the

discrepancy between simulations and in-situ observations on the C:N:P


The microbial loop plays a crucial role in nutrient recycling by regulating the

quantity and stoichiometry of available nutrients, which has a significant effect on

phytoplankton growth. Future research should be conducted to understand the impact

of the microbial loop in more detail, by including processes such as, the viral shunt

short circuits the microbial loop. These processes should be considered with

seasonality and heterogeneity, which influence the stoichiometry of aquatic food

webs. Moreover, they should also be supported with targeted empirical studies, such

as Lake Kinneret. Therefore the ecological model can provide more accurate

prediction for algal blooms in aquatic ecosystems.

142

References

Andersen，T., Elser, J. & Hessen, D. (2004) Stoichiometry and population dynamics.

Ecology Letters, 7, 884-900.

Anderson, D. M., Glibert, P. M. & Burkholder, J. M. (2002) Harmful Algal Blooms and

Eutrophication: Nutrient Sources, Composition, and Consequences. Estuaries, 25, 704–726.

Arhonditsis, G.B. & Brett, M.T. (2004) Evaluation of the current state of mechanistic aquatic

biogeochemical modelling. Marine Ecology-Progress Series, 271, 13–26.

Azam, F., Fenchel, T., Field, J., Gray, J.S., Meyer-Reil, L.A. & Thingstad, F. (1983) The

ecological role of water-column microbes in the sea. Marine Ecology-Progress Series, 10,

257-263.

Ballantyne IV, F., Menge, D.N.L., Ostling, A. & Hosseini, P. (2008) Nutrient recycling affects

autotroph and ecosystem stoichiometry. American Naturist, 171, 511-523.

Ballantyne IV, F., Menge, D.N.L. & Weitz, J.S. (2010) A discrepancy between predictions of

saturating nutrient uptake models and nitrogen-to-phosphorus stoichiometry in the surface

ocean. Limnology and Oceanography, 55, 997–1008.

Ballot, A., Ramm, J., Rundberget, T., Kaplan-Levy, R.N., Hadas, O., Sukenik, A. & Wiedner,

C. (2011) Occurrence of non-cylindrospermopsin-producing Aphanizomenon ovalisporum

and Anabaena bergii in Lake Kinneret (Israel). Journal of Plankton Research, 33, 1736-

1746.

Banker, R., Carmeli, S., Hadas, O., Teltsch, B., Porat, R., & Sukenik, A. (1997) Identification

of cylindrospermopsin in Aphanizomenon ovalisporum (Cyanophyceae) isolated from Lake

Kinneret, Israel. Journal of Phycology, 33, 613-616.

Barsdate, R.J., Prentki, R.T. & Fenchel, T. (1974) Phosphorus cycle of model ecosystems:

Significance for decomposer food chains and effect of bacterial grazers. Oikos, 25, 239-251.

Beardall, J., Berman, T. & Heraud, P. (2001a) Comparison of methods for detection of

phosphate limitation in microalgae. Aquatic Sciences, 63, 107–121.

Beardall, J., Young, E. & Roberts, S. (2001b) Approaches for determining phytoplankton

nutrient limitation. Aquatic Sciences, 63, 44–69.

143

Bergström, A.K. (2010) The use of TN:TP and DIN:TP ratios as indicators for phytoplankton

nutrient limitation in oligotrophic lakes affected by N deposition. Aquatic Sciences, 72, 277-

281.

Berman, T. (1985) Uptake of [32P] orthophosphate by algae and bacteria in Lake Kinneret.

Journal of Plankton Research, 7, 71-84.

Berman, T. (2001) The role of DON and the effect of N:P ratios on occurrence of

cyanobacterial blooms: implications from the outgrowth of Aphanizomenon in Lake

Kinneret. Limnology and Oceanography, 46, 443–447.

Berman, T., Yacobi, Y.Z. & Pollinger, U. (1992) Lake Kinneret phytoplankton - stability and

variability during 20 years. Aquatic Sciences, 54, 104-127.

Berman, T., Stone, L., Yacobi, Y.Z., Kaplan, B., Schlichter, M., Nishri, A. & Pollingher, U.

(1995) Primary production and phytoplankton in Lake Kinneret - A long-term record (1972-

1993). Limnology and Oceanography, 40, 1064-1076.

Berman, T. (2001) The role of DON and the effect of N: P ratios on occurrence of

cyanobacterial blooms: Implications from the outgrowth of Aphanizomenon in Lake

Kinneret. Limnology and Oceanography; 46, 443-447.

Berman, T. & Bronk, D.A. (2003) Dissolved organic nitrogen: a dynamic participant in

aquatic ecosystems. Aquatic Microbial Ecology, 31, 279–305.

Berman, T., Parparov, A. & Yacobi, Y.Z. (2004) Planktonic community production and

respiration and the impact of bacteria on carbon cycling in the photic zone of Lake Kinneret.

Aquatic Microbial Ecology, 34, 43–55.

Berman, T., Yacobi, Y.Z., Parparov, A. & Gal, G. (2010) Estimation of long-term bacterial

respiration and growth efficiency in Lake Kinneret. FEMS Microbiology Ecology, 71, 351–

363.

Berman-Frank, I., Zohary, T., Erez, J. & Dubinsky, Z. (1994) CO2 availability, carbonic

anhydrase and the annual dinoflagellate bloom in Lake Kinneret. Limnology and

Oceanography, 39, 1822-1834.

Bjoerkman, K. & Karl, D.M. (1994) Bioavailability of inorganic and organic phosphorus

compounds to natural assemblages of microorganisms in Hawaiian coastal waters. Marine

Ecology-Progress Series, 111, 265– 273.

Bratbak, G. & Thingstad, T.F. (1985) Phytoplankton-bacteria interactions-an apparent

paradox-analysis of a model system with both competition and commensalism. Marine

Ecology-Progress Series, 25, 23-30.

144

Bruce, L.C., Hamilton, D.P., Imberger, J., Gal, G., Gophen, M., Zohary, T. & Hambright, D.

(2006) A numerical simulation of the role of zooplankton in C, N and P cycling in Lake

Kinneret, Israel. Ecological Modelling, 193, 412-436.

Brussaard, C. & Riegman R. (1998) Influence of bacteria on phytoplankton cell mortality with

phosphorus or nitrogen as the algal-growth-limiting nutrient, Aquatic Microbial Ecology, 14,

271–280.

Boucher, D.H., James, S. & Keeler, K.H. (1982) The ecology of mutualism. Ann. Rev. Ecol. Syst. 13, 315-347.

Burchard, H., Deleersnijder, E. & Meister, A. (2005) Application of modified Patankar

schemes to stiff biogeochemical models for the water column. Ocean Dynamics, 55, 326–

337.

Burger, D.F., Hamilton, D.P. & Pilditch, C.A. (2007) Modelling the relative importance of

internal and external nutrient loads on water column nutrient concentrations and

phytoplankton biomass in a shallow polymictic lake. Ecological Modelling, 211, 411–423.

Caron, D.A. (1994) Inorganic nutrients, bacteria, and the microbial loop. Microbial ecology,

28, 295-298.

Clasen, J.L., Brigden, S.M., Payet, J.P. & Suttle, C.A. (2008) Evidence that viral abundance

across oceans and lakes is driven by different biological factors. Freshwater Biology, 53,

1090–1100.

Cole, J.J. (1999) Aquatic microbiology for ecosystem scientists: new and recycled paradigms

in ecological microbiology. Ecosystems, 2, 215–225.

Cotner, J.B., Jr. & Wetzel, R.G. (1992) Uptake of dissolved inorganic and organic phosphorus

compounds by phytoplankton and bacterioplankton. Limnology and Oceanography, 37, 232–

243.

Coveney, M.F. & Wetzel, R.G. (1992) Effects of Nutrients on Specific Growth Rate of

Bacterioplankton in Oligotrophic Lake Water Cultures. Applied and Environmental

Microbiology, 58, 150-156.

Currie, D.J. & Kalff, J. (1984) The relative importance of bacterioplankton and phytoplankton

in phosphorus uptake in freshwater. Limnology and Oceanography, 29, 311–321.

Danger, M., Oumarou, C., Benest, D. & Lacroix, G. (2007) Bacteria can control stoichiometry

and nutrient limitation of phytoplankton. Functional Ecology, 21, 202–210.

Danger, M., Lacroix, G., Oumarou, C., Benest, D.& Meriguet, J. (2008) Effects of food-web

structure on periphyton stoichiometry in eutrophic lakes: a mesocosm study. Freshwater

Biology, 53, 2089–2100.

145

Dolan, J.R. (1997) Phosphorus and ammonia excretion by planktonic protists. Marine

Geology, 139, 109-122.

Edwards, A.M. (2001) Adding Detritus to a Nutrient–Phytoplankton–Zooplankton Model: A

dynamical-Systems Approach. Journal of Plankton Research, 23, 389-413.

Elser, J.J. (1999) The pathway to noxious cyanobacterial blooms in lakes: the food web as the

final turn. Freshwater Biology, 42, 537–543.

Elser, J.J. & Urabe, J. (1999) The stoichiometry of consumer-driven nutrient recycling:

Theory, observations, and consiquences. Ecology, 80, 735–751.

Elser, J.J., Loladze, I., Peace, A. L. & Kuang, Y. (2012) Lotka re-loaded: Modeling trophic

interactions under stoichiometric constraints. Ecological Modeling,

doi:10.1016/j.ecolmodel.2012.02.006.

Faure, V., Pinazo, C., Torréton, J.P. & Jacquet, S. (2010) Modelling the spatial and temporal

variability of the SW lagoon of New Caledonia I: A new biogeochemical model based on

microbial loop recycling. Marine Pollution Bulletin, 61, 465-79.

Fasham, M.J.R., Boyd, P.W. & Savidge, G. (1999) Modeling the relative contributions of

autotrophs and heterotrophs to carbon flow at a Lagrangian JGOFS station in the Northeast

Atlantic: the importance of DOC. Limnology and Oceanography, 44, 80–94.

Ferrier, P. C. & Rassoulzadegan, F. (1994) N-remineralization in planktonic protozoa.

Limnology and Oceanography, 39, 411–418.

Franks, P.J.S. (2002) NPZ models of Plankton Dynamics: Their Construction, Coupling to

Physics, and Application. Journal of Oceanography, 58, 379-387.

Frost, P.C., Stelzer, R.S., Lamberti, G.A. & Elser, J.J. (2002) Ecological Stoichiometry of

Trophic Interactions in the Benthos: Understanding the Role of C:N:P Ratios in Lentic and

Lotic Habitats. Journal of the North American Benthological Society, 21, 515-528.

Frost, P.C., White, M., Finkel, Z., Jensen, T. & Matzek, V. (2005) Are you what you eat?

Physiological constraints on organismal stoichiometry in an elementally imbalanced world.

Oikos, 109, 18-28.

Gaedke, U., Hochstadter, S., Straile, D. (2002) Interplay between energy limitation and

nutritional deficiency: empirical data and food web models. Ecological Monographs, 72,

251-270.

Gal, G., Imberger, J., Zohary, T., Antenucci, J.P., Anis, A. & Rosenberg, T. (2003) Simulating

the thermal dynamics of Lake Kinneret. Ecological Modelling, 162, 69-86.

Gal, G., Hipsey, M.R., Parparov, A., Wagner, U., Makler, V. & Zohary, T. (2009)

146

Implementation of ecological modeling as an effective management and investigation tool:

Lake Kinneret as a case study. Ecological Modelling, 220, 1697–1718.

Gal, G. & Anderson, W. (2010) A novel approach to detecting a regime shift in a lake

ecosystem. Methods in Ecology and Evolution, 1, 45–52.

Gillor, O., Hadas, O., Post, A.F. & Belkin, S. (2010) Phosphorus and nitrogen in a

monomictic freshwater lake: employing cyanobacterial bioreporters to gain new insights into

nutrient bioavailability. Freshwater Biology, 55, 1182–1190.

Gobler, C.J., Hutchins, D.A., Fisher, N.S., Cosper, E.M. & Sanudo-Wilhelmy, S.A. (1997)

Release and bioavailability of C, N, P, Se and Fe following viral lysis of a marine

chrysophyte. Limnology and Oceanography, 42, 1492–1504.

Goldman, J.C., McCarthy, J.J. & Peavey, D.G. (1979) Growth rate influence on the chemical

composition of phytoplankton in oceanic waters. Nature, 279, 210–215.

Gophen, M. (2011) The cladoceran trophic status in the nitrogen limited ecosystem of lake

Kinneret (Israel). Journal of Environmental Biology, 32, 455–462.

Gophen, M. & Azoulay, B. (2002) The trophic status of zooplankton communities in lake

Kinneret (Israel). Verhandlungen des Internationalen Verein Limnologie, 28, 836–839.

Hadas, O. & Berman, T. (1998) Seasonal abundance and vertical distribution of Protozoa

(flagellates, ciliates) and bacteria in Lake Kinneret, Israel. Aquatic Microbial Ecology, 14,

161-170.

Hambright, K. D., Zohary T. & Gude, H. (2007) Microzooplankton dominate carbon flow and

nutrient cycling in a warm subtropical freshwater lake. Limnology and Oceanography, 52,

1018–1025.

Hambright, K. D., Zohary, T., Eckert, W., Eckert, S., Schwartz, S. S., Schelske, C. L., Laird,

K. & Leavitt, P. R. (2008) Exploitation and destabilization of a warm, freshwater ecosystem

through engineered hydrological change. Ecological Applications, 18, 1591-1603.

Hessen, D. (1997) Stoichiometry in food webs - Lotka revisited. Oikos, 79,195-200.

Hart, D. R., Stone, L. & Berman, T. (2000) Seasonal dynamics of the Lake Kinneret food

web: The importance of the microbial loop. Limnology and Oceanography, 45, 350-361.

Hellweger, F. L., Kravchuk, E. S. & Novotny, V. (2008) Agent-based modeling of the

complex life cycle of a cyanobacterium (Anabaena) in a shallow reservoir. Limnology and

Oceanography, 53, 1227–1241.

Hillmer, I., van Reenen, P., Imberger, J. & Zohary, T. (2008). Phytoplankton patchiness and

their role in the modelled productivity of a large, seasonally stratified lake. Ecological

147

Modelling, 218, 49-59.

Hipsey, M.R. & Hamilton, D.P. (2008) Computational Aquatic Eco-system Dynamics Model.

v3.3 Science Manual. Centre for Water Research Report, University of Western Australia.

Howarth, R.W. (2008) Coastal nitrogen pollution: A review of sources and trends globally and

regionally. Harmful Algae, 8, 14-20.

Howarth, R.W. & Marino, R. (2006) Nitrogen as the limiting nutrient for eutrophication in

coastal marine ecosystems: Evolving views over three decades. Limnology and


Hudnell, H. K. (2010) The state of U.S. freshwater harmful algal blooms assessments, policy

and legislation. Toxicon, 55, 1024–1034.

Janse, J. H., Aldenberg, T. & Kramer, P. R. G. (1992) A mathematical model of the

phosphorus cycle in Lake Loosdrecht and simulation of additional measures. Hydrobiologia,

233,119–136.

Jellison, R.&Melack, J.M. (1993). Meromixis and vertical diffusivities in hypersaline Mono

Lake California. Limnology and Oceanography, 38, 1008-1019.

Jeppesen, E., Sondergaard, M., Jensen, J. P., Havens, K. E., Anneville, O., Carvalho, L.,

Coveney, M. F., Deneke, R., Dokulil, M. T., Foy, B., Gerdeaux, D., Hampton, S. E., Hilt,

S., Kangur, K., Kohler, J., Lammens, E. H. H. R., Lauridsen, T. L., Manca, M., Miracle, M.

R., Moss, B., Noges, P., Persson, G., Phillips, G., Portielje, R., Romo, S., Schelske, C. L.,

Straile, D., Tatrai, I., Willen, E. & Winder, M. (2005) Lake responses to reduced nutrient

loading - an analysis of contemporary long-term data from 35 case studies. Freshwater

Biology, 50, 1747-1771.

Johannes, R. E. (1965) Influence of marine protozoa on nutrient regeneration. Limnology and


Joint I. & Morris, R. (1982) The role of bacteria in the turnover of organic matter in the sea,

Oceanogrraphy and Marine Biology Annual Review, 20, 65–118.

Joint, I., Henriksen, P., Fonnes, G. A., Bourne, D., Thingstad, T. F. & Riemann, B. (2002)

Competition for inorganic nutrients between phytoplankton and bacterioplankton in nutrient

manipulated mesocosms. Aquatic Microbial Ecology, 29, 145-159.

Jorgensen, S.E. & Bendoricchio, G. (2001) Fundamentals of Ecological Modeling, third ed.

Elsevier.

Karl, D.M., Bjorkman, K.M., Dore, J.E., Fujieki, L., Hebel, D.V., Houlihan, T., Letelier, R.M.

& Tupas, L.M. (2001) Ecological nitrogen-to-phosphorus stoichiometry at station ALOHA.

148

Deep Sea Research II, 48, 1529–1566.

Klausmeiser, C.A., Litchman, E., Daufresne, T. & Levin, S.A. (2004) Optimal nitrogen-to-

phosphorus stoichiometry of phytoplankton. Nature, 429, 171-174.

Kirchman, D.L. (1994) The uptake of inorganic nutrients by heterotrophic bacteria. Microbial

Ecology, 28, 255-271.

Kooijman, S.A.L.M., Andersen, T. & Kooi, B.W. (2004) Dynamic energy budget

representations of stoichiometric constraints on population dynamics. Ecology, 85, 1230–

1243.

Kromkamp, J. & Walsby, A.E. (1990) A computer model of buoyancy and vertical migration

in cyanobacteria. Journal of Plankton Research, 12, 191–183.

Laybourn-Parry, J., HÖfer, J. & Sommaruga, R. (2001) Viruses in Antarctic freshwater and

saline lakes. Freshwater Biology, 46, 1279–1287.

Lewis, D. M., Brookes, J. D. & Lambert, M. F. (2004) Numerical models for management of

Anabaena circinalis. Journal of Applied Phycology, 16, 457–468.

Li, Y., Gal, G., Waite, A.M. & Hipsey, M.R. (2011) Microbial loop processes shape the food

web stoichiometry in Lake Kinneret. In: Chan, F., Marinova, D. and Anderssen, R.S. (eds)

MODSIM2011, 19th International Congress on Modelling and Simulation. Modelling and

Simulation Society of Australia and New Zealand, December 2011 ISBN: 978-0-9872143-1-

7, 3726-3732.

Li, Y., Gal, G., Waite, A. & Hipsey, M. (2013) An analysis of the relationship between

phytoplankton internal stoichiometry and water column N:P ratios in a dynamic lake

environment. Ecological Modelling, 252, 196-213.

Loladze, I., Kuang, Y. & Elser, J. (2000) Stoichiometry in producer-grazer systems: Linking

energy flow with element cycling. Bulletin Mathematics Biology, 62,1137-1162.

Likens, G.E. (2010) Biogeochemistry of Inland Waters. Elsevier Science, Burlington.

Madan, N. J., Marshall, W.A. & Laybourn-Parry, J. Virus and microbial loop dynamics over

an annual cycle in three contrasting Antarctic lakes. Freshwater Biology, 50, 1291–1300.

Makler-Pick, V., Gal, G., Gorfine, M., Hipsey, M.R. & Carmel, Y. (2011a) Sensitivity

analysis for complex ecological models - A new approach. Environmental Modelling &

Software, 26, 124-134.

Makler-Pick, V., Gal, G., Shapiro, J. & Hipsey. M.R. (2011b) Exploring the role of fish in a

lake ecosystem (Lake Kinneret, Israel) by coupling an individual-based fish population

model to a dynamic ecosystem model. Canadian Journal of Fisheries and Aquatic Sciences,

149

68, 1265-1284.

Makino, W., Cotner, J.B., Sterner, R.W. & Elser, J.J. (2003) Are bacteria more like plants or

animals? Growth rate and resource dependence of bacterial C: N: P stoichiometry.

Functional Ecology, 17, 121-130.

Martin-Creuzburg, D., Bec, A., & von Elert, E. (2005) Trophic upgrading of

picocyanobacterial carbon by ciliates for nutrition of Daphnia magna. Aquatic Microbial

Ecology, 41, 271-280.

McGillicuddy, D.J., Robinson A.R. & McCarthy, J.J. (1995) Coupled physical and biological

modelling of the spring bloom in the North-Atlantic.1.3-dimensional bloom and post-bloom

processes. Deep-Sea Reserach I, 42, 1359-1398.

Michaels, A.F., Karl, D.M. & Capone, D.G. (2001) Elemental stoichiometry, new production,

and nitrogen fixation. Oceanography, 14, 68–77.

Morris, D.P. & Lewis, W.M. (1988) Phytoplankton nutrient limitation in Colorado mountain

lakes. Freshwater Biology, 20, 315–27.

Mooij, W.M., Trolle, D., Jeppesen, E., Arhonditsis, G., Belolipetsky, P., Chitamwebwa,

D.B.R., Degermendzhy, A.G., DeAngelis, D. L., Senerpont Domis de, L.N., Downing, A.S.,

Elliott, J.A., Fragoso Jr., C.R., Gaedke, U., Genova, S.N., Gulati, R.D., Håkanson, L.,

Hamilton, D.P., Hipsey, M.R., Hoen, P.J., Hülsmann, S., Los, F.J., Makler-Pick, V.,

Petzoldt, T., Prokopkin, I., Rinke, K., Schep, S.A., Tominaga, K., Dam Van, A.A., Nes van,

E.H., Wells, S.A. & Janse, J.H. (2010) Challenges and opportunities for integrating lake

ecosystem modelling approaches. Aquatic Ecology, 44, 633-667.

Moore, J.C., Berlow, E.L., Coleman, D.C., Ruiter, P.C., Dong, Q., Hastings, A., Johnson,

N.C., McCann, K.S., Melville, K., Morin, P.J., Nadelhoffer, K., Rosemond, A.D., Post,

D.M., Sabo, J.L., Scow, K.M., Vanni, M.J. & Wall, D.H. (2004). Detritus, trophic dynamics

and biodiversity. Ecological Letters, 7, 584–600.

Muller, E., Nisbet, R., Koojman, S, Elser, J. & McCauley, E. (2001) Stoichiometric food

quality and herbivore dynamics. Ecology Letters, 4, 519-529.

Ng, S., Antenucci, J.P., Hipsey, M.R., Tibor, G. & Zohary, T. (2011) Physical controls on the

spatial evolution of a dinoflagellate bloom in a large lake. Limnology and Oceanography,

56, 2265–2281.

Özkundakci, D., Hamilton, D.P. & Trolle, D. (2011) Modelling the response of a highly

eutrophic lake to reductions in external and internal nutrient loading. New Zealand Journal

of Marine and Freshwater Research, 45, 165-185.

Parparov, A. & Gal, G. (2012) Assessment and implementation of a methodological

150

framework for sustainable management: Lake Kinneret as a case study. Journal of

Environment and Management, 101, 111-117.

Peduzzi, P. & Herndl, G.J. (1992) Zooplankton activity fueling the microbial loop:

Differential growth response of bacteria from oligotrophic and eutrophic waters. Limnology

and Oceanography, 37, 1087-1092.

Pollingher, U. (1986) Phytoplankton periodicity in a subtropical lake (Lake Kinneret, Israel).

Hydrobiologia, 138, 127-138.

Pomeroy, L.R., Leb, P.J., Azam, W.F. & Hobbie, J.E. (2007) The Microbial Loop.

Oceanography, 20, 28-33.

Popova, E.E., Ryabchenko, V.A. & Fasham, M.J.R. (2000) Biological pump and vertical

mixing in the southern ocean: their impact on atmospheric CO2. Global Biogeochemical

Cycles, 14, 477–498.

Ptacnik, R., Andersen, T. & Tamminen, T. (2010) Performance of the Redfield ratio and a

family of nutrient limitation indicators as thresholds for phytoplankton N vs. P limitation.

Ecosystems, 13, 1201–1214.

Ratsak, C.H., Maarsen, K, A. & Kooijman, S.A.L.M. (1996) Effects of protozoa on carbon

mineralization in activated sludge. Water Resources, 30, 1–12.

Redfield, A.C. (1958) The biological control of chemical factors in the environment.

American Scientist, 46, 205-221.

Regel, R.H., Brookes, J.D. & Ganf, G.G. (2004) Vertical migration, entrainment and

photosynthesis of the freshwater dinoflagellate Peridinium cinctum in a shallow urban lake.

Journal of Plankton Research, 26, 143-157.

Reynolds, C.S. (1984) The ecology of freshwater phytoplankton. London: Cambridge

University Press.

Rhee, G.Y. (1978) Effects of N:P atomic ratios and nitrate limitation on algal growth, cell

composition, and nitrate uptake. Limnology and Oceanography, 23, 10–25.

Rigosi, A., Fleenor, W. & Rueda, F. (2010) State-of-the-art and recent progress in

phytoplankton succession modelling. Environmental Reviews, 18, 423-440.

Rigosi, A., Marcé, R., Escot, C. & Rueda, F.J. (2011) A calibration strategy for dynamic

succession models including several phytoplankton groups. Environmental Modelling &

Software, 26, 697-710.

Roelke, D.L., Zohary, T., Hambright, K.D. & Montoya, J.V. (2007) Alternative states in the

phytoplankton of Lake Kinneret, Israel (Sea of Galilee). Freshwater Biology, 52, 399-411.

151

Romero, J.R., Antenucci, J.P. & Imberger, J. (2004) One- and three- dimensional

biogeochemical simulations of two differing reservoirs. Ecological Modelling, 174, 143-160.

Rose, C. & Axler, R.P. (1998) Uses of alkaline phosphatase activity in evaluating

phytoplankton community phosphorus deficiency. Hydrobiologia, 361, 145–156.

Saito, L., Johnson, B.M., Bartholow, J. & Hanna, R.B. (2001) Assessing ecosystem effects of

reservoir operations using food web-energy transfer and water quality models. Ecosystems,

4, 105–125.

Sanches, L.F., Guariento, R.D., Caliman, A., Bozelli, R.L. & Esteves, F.A. (2011) Effects of

nutrients and light on periphytic biomass and nutrient stoichiometry in a tropical black-water

aquatic ecosystem. Hydrobiologia, 669, 35–44.

Sanudo-Wilhelmy, S.A., Tovar-Sanchez, A., Fu, F.X., Capone, D.G., Carpenter, E.J.&

Hutchins, D.A. (2004) The impact of surface-adsorbed phosphorus on phytoplankton

Redfield stoichiometry. Nature, 432, 897–901.

Saxton, M.A., Arnold, R.J., Bourbonniere, R.A., McKay, R.M.L. & Wilhelm, S.W. (2012)

Plasticity of total and intracellular phosphorus quotas in Microcystis aeruginosa cultures and

Lake Erie algal assemblages. Frontiers in Microbiology, 3, 1-9.

Schatz, D., Keren, Y., Hadas, O., Carmeli, S., Sukenik, A. & Kaplan, A. (2005) Ecological

implications of the emergence of non-toxic subcultures from toxic Microcystis strains.

Environmental Microbiology, 7, 798-805.

Schindler, D.W., Hecky, R.E., Findlay, D.L., Stainton, M.P., Parker, B.R., Paterson, M.J.,

Beaty, K.G., Lyng, M. & Kasian, S.E.M. (2008) Eutrophication of lakes cannot be

controlled by reducing nitrogen input: Results of a 37-year whole-ecosystem experiment.

Proceedings of the National Academy of Sciences, 105,11254–11258.

Smith, V.H. (1983) Low nitrogen to phosphorus ratios favor dominance by blue-green algae in

lake Phytoplankton. Science, 221, 669-671.

Smith, V.H. (2003) Eutrophication of freshwater and coastal marine ecosystems: a global

problem. Environmental Science and Pollution Research, 10, 126-139.

Sherr, B.F., Sherr, E.B. & Hopkinson, C.S. (1988) Trophic interactions within pelagic

microbial communities: indications of feedback regulation of carbon flow.

Hydrobiologia, 159, 19–26.

Suttle, C. A. (2005) Viruses in the sea. Nature, 437, 356-361.

Sterner, R.W. & Elser, J.J. (2002) Ecological stoichiometry: The Biology of Elements from

Molecules to the Biosphere. Princeton University Press, New Jersey.

152

Stemberger, R.S. & Gilbert, J.J. (1985) Body size, food concentration, and population growth

in plantonic rotifers. Ecology, 66, 1151–1159.

Stone, L. (1990) Phytoplankton-bacteria-protozoa interactions - a qualitative model portraying

indirect effects. Marine Ecology-Progress Series, 64, 137-145.

Stone, L., Berman, T., Bonner, R., Barry, S. &Weeks, S.W. (1993) Lake Kinneret: A seasonal

model for carbon flux through the planktonic biota. Limnology and Oceanography, 38,

1680-1695.

Thatcher, S.J., Davis, C.C. & Gardner, G.A. (1993) Physical and chemical effects of

macrograzers and micrograzers on enclosed,in situ phytoplankton in a Newfoundland lake.

Hydrobiologia, 250, 127-141.

Thingstad T. & Lignell R. (1997) Theoretical models for the control of bacterial growth rate,

abundance, diversity and carbon demand, Aquatic Microbial Ecology, 13, 19–27.

Thingstad, T.F., Krom, M.D., Mantoura, R.F.C., Flaten, G.A.F., Groom, S., Herut, B., Kress,

N., Law, C.S., Pasternak, A., Pitta, P., Psarra, S., Rassoulzadegan, F., Tanaka, T.,

Tselepides, A., Wassmann, P., Woodward, E.M.S., Riser, C.W., Zodiatis, G. & Zohary, T.

(2005) Nature of Phosphorus Limitation in the Ultraoligotrophic Eastern Mediterranean.

Science, 309, 1068-1071.

Thingstad, T.F., Bellerby, R.G.J., Bratbak, G., Børsheim, K.Y., Egge1, J. K., Heldal, M.,

Larsen, A., Neil, C., Nejstgaard, J., Norland, S., Sandaa, R. A., Skjoldal, E. F., Tanaka, T.,

Thyrhaug, R. & Töpper, B. (2008) Counterintuitive carbon-to-nutrient coupling in an Arctic

pelagic ecosystem. Nature, 455, 387-390.

Trolle, D., Jørgensen, T.B. & Jeppesen, E. (2008) Predicting the effects of reduced external

nitrogen loading on the nitrogen dynamics and ecological state of deep Lake Ravn,

Denmark, using the DYRESM–CAEDYM model. Limnologica, 38, 220-232.

Trolle, D., Hamilton, D.P., Hipsey, M.R., Bolding, K., Bruggeman, J., Mooij, W.M., Janse,

J.H., Nielsen, A., Jeppesen, E., Elliott, J., Makler-Pick, V., Petzoldt, T., Rinke, K., Flindt,

M., Arhonditsis, G., Gal, G., Bjerring, R., Tominaga, K., Hoen, J., Downing, A. (2012) A

community-based framework for aquatic ecosystem models. Hydrobiologia, 683, 25-34.

Tyrrell, T. (1999) The relative influences of nitrogen and phosphorus on oceanic primary

production. Nature, 400, 525–531.

Vadstein, O. (2000) Heterotrophic, planktonic bacteria and cycling of phosphorus: phosphorus

requirements, competitive ability, and food web interactions. Advanced Microbial Ecology,

16,115–167.

Van Nes, E.H. & Scheffer, M. (2005) A strategy to improve the contribution of complex

153

simulation models to ecological theory. Ecological Modelling, 185,153-164.

Vrede, T., Dobberfuhl, D.R., Kooijman, S.A.L.M. & Elser, J.J. (2004) Fundamental

connections among organism C:N:P stoichiometry, macromolecular composition, and

growth. Ecology, 85, 1217–1229.

Wang, H., Jiang, L. & Weitz, J.S. (2009) Bacterivorous grazers facilitate organicmatter

decomposition: a stoichiometric modeling approach. FEMS Microbiology Ecology, 69, 170–

179.

Webb, W.L., Newton, M. & Starr, D. (1974) Carbon dioxide exchange of Alnus rubra: a

mathematical model. Oecologia, 17, 281–291.

Wilhelm, S.W. & Suttle, C.A. (1999) Viruses and nutrient cycles in the sea. BioScience, 49,

781–788.

Wohlers-Zollner, J., Breithaupt, P., Walther, K., Jurgens, K. & Riebesell, U. (2011)

Temperature and nutrient stoichiometry interactively modulate organic matter cycling in a

pelagic algal–bacterial community. Limnology and Oceanography, 56, 599–610.

Wommack, K.E. & Colwell R.R. (2000) Virioplankton: viruses in aquatic ecosystems.

Microbiology and Molecular Biology Reviews, 64, 69–114.

Yeates, P.S. & Imberger, J. (2003) Pseudo two-dimensional simulations of internal and

boundary fluxes in stratified lakes and reservoirs. International Journal of River Basin

Management, 1, 297-319.

Zohary, T., Pollingher, U., Hadas, O., & Hambright, K.D. (1998) Bloom dynamics and

sedimentation of Peridinium gatunense in Lake Kinneret. Limnology and Oceanography, 43,

175-186.

Zohary, T. (2004) Changes to the phytoplankton assemblage of Lake Kinneret after decades of

a predictable, repetitive pattern. Freshwater Biology, 49, 1355-1371.

Zohary, T., Gal, G. & Antenucci, J. (2006) Lake Kinneret Water Quality Management and

Optimisation Support System - Phase 2 Final Report., IOLR-KLL Report T8/2004. Kinneret

Limnological Laboratory, Israel.

Zohary, T. & Ostrovsky, I. (2011) Ecological impacts of excessive water level fluctuations in

stratified freshwater lakes. Inland Waters, 1, 47-59.

154

Appendix

Sim1(NPZD):

Sim2(NPZD):

155

Sim1(NPVD):

Sim2(NPVD):

156

Sim1(NPZD+V):

157

Sim2(NPZD+V):

158

Sim1(NPZD+B):

Sim2(NPZD+B):

0 100 200 3000

2

4

6

mm

ol/m

3

Nutrients

0 100 200 3000

1

2

3

mm

ol/m

3

Phytoplankton

0 100 200 3000

1

2

3

4

mm

ol/m

3

Zooplankton1

0 100 200 3000

0.5

1

mm

ol/m

3

Zooplankton2

0 100 200 3001

2

3

4

5

Time (d)

mm

ol/m

3

Detritus

0 100 200 3000

0.2

0.4

0.6

0.8

Time (d)

mm

ol/m

3

bacteria

0 100 200 3000

2

4

6

mm

ol/m

3

Nutrients

0 100 200 3000

1

2

3

mm

ol/m

3

Phytoplankton

0 100 200 3000

1

2

3

mm

ol/m

3

Zooplankton1

0 100 200 3000

0.5

1

mm

ol/m

3

Zooplankton2

0 100 200 3000

2

4

6

Time (d)

mm

ol/m

3

Detritus

0 100 200 3000

0.5

1

Time (d)

mm

ol/m

3

bacteria

159

Sim1(NPZD+VB):

0 100 200 3004.5

5

5.5

mm

ol/m

3

Nutrients

0 100 200 3000

0.05

0.1

mm

ol/m

3

Phytoplankton

0 100 200 3000

0.5

1x 10

-3

mm

ol/m

3

Zooplankton1

0 100 200 3000

0.01

0.02

mm

ol/m

3

Zooplankton2

0 100 200 3003

4

5

mm

ol/m

3

Detritus

0 100 200 3000

0.1

0.2

mm

ol/m

3

Bacteria

0 100 200 3000

0.005

0.01

Time (d)

mm

ol/m

3

Viruses1

0 100 200 3000

0.1

0.2

Time (d)

mm

ol/m

3

Viruses2

160

Sim2(NPZD+VB):

0 100 200 3004

5

6m

mol

/m3

Nutrients

0 100 200 3000

0.1

0.2

mm

ol/m

3

Phytoplankton

0 100 200 3000

0.005

0.01

mm

ol/m

3

Zooplankton1

0 100 200 3000

0.02

0.04

mm

ol/m

3

Zooplankton2

0 100 200 3003

4

5

mm

ol/m

3

Detritus

0 100 200 3000

0.1

0.2

mm

ol/m

3

Bacteria

0 100 200 3000

0.02

0.04

Time (d)

mm

ol/m

3

Viruses1

0 100 200 3000

0.2

0.4

Time (d)

mm

ol/m

3

Viruses2

Unravelling microbial interactions in aquatic ecosystems ... · Unravelling microbial interactions...

Documents

Transcript of Unravelling microbial interactions in aquatic ecosystems ... · Unravelling microbial interactions...