Macintyre River, Australia

Trophic Ecology and Energy Sources for Fish on the Floodplain of a Regulated Dryland River: Macintyre River, Australia

Author Medeiros, Elvio S. F

Published 2005

Thesis Type Thesis (PhD Doctorate)

School School of Australian Environmental Studies

DOI https://doi.org/10.25904/1912/1443

Downloaded from http://hdl.handle.net/10072/365765

Griffith Research Online https://research-repository.griffith.edu.au Trophic ecology and energy sources for fish on the floodplain of a regulated dryland river: Macintyre River, Australia

Elvio S. F. Medeiros M.Sc.

School of Australian Environmental Studies Faculty of Environmental Sciences Griffith University

Thesis submitted in fulfillment of the requirements of the degree of Doctor of Philosophy

Griffith University, Brisbane, Australia November 2004

Abstract

Drylands occupy about one-third of the world’s land surface area and rivers in these regions have less predictable flow regimes than those in humid tropical and temperate regions. Australia’s dryland river-floodplain systems cycle through recurrent periods of floods and droughts, often resulting in extreme hydrological variability. As a result, these systems have been described as having a “boom and boost” ecology with periods of high productivity associated with flooding. Not surprisingly, flow and its variability have been recognised as major driving forces in the ecological functioning of Australian rivers and responses to flow variability from fish and aquatic invertebrates have been reasonably well described. Furthermore, the reduced amount of water reaching floodplain waterbodies due to river regulation has been held responsible for successional changes in aquatic biota and, consequently, the resources available for both fish and invertebrates. However, information regarding the impacts of water resource development has generally focused on within-channel processes of Australian rivers, not on floodplains, which are arguably more affected by water development.

The following dissertation is concerned with how different types of natural and modified floodplain lagoons are able to trophically support their fish communities in the floodplain of the Macintyre River, Border Rivers catchment (QLD/NSW), a regulated dryland river. This study focuses on the influence of flooding and the implications of an extended dry period, and different levels of flow regulation, on the feeding ecology of selected fish species (Ambassis agassizii, Leiopotherapon unicolor and Nematalosa erebi) between 2001 and 2003. Food resources consumed by fish are hypothesised to vary in response to flooding, when inundation of isolated lagoons and vast floodplain areas can result in a burst of primary and secondary productivity. Given the permanently elevated water levels of some regulated floodplain lagoons, fish diets are hypothesised to be less variable in these floodplain habitats in comparison to diets of fish from floodplain lagoons with natural flow and water regime. Feeding ecology is examined firstly, in terms of diet composition of selected fish species, using stomach content analysis, and secondly, in relation to possible energy sources sustaining fish (using stable isotope analysis) in selected floodplain lagoons and a site in the main channel of the Macintyre River. The information produced should allow managers to take variations in food resources, food ii web structure and dietary ecology into account in management regimes for refugia and dryland systems in general.

Factors such as diel and ontogenetic variations in dietary composition and food intake by fish are shown to considerably affect overall dietary patterns of each study species. Therefore, it is important to understand the contributions of such factors to the variability of fish dietary patterns before performing studies on resource use by fish in floodplain habitats of the Macintyre River. Major food categories consumed by the study species were zooplankton, aquatic invertebrates and detrital material. Zooplankton was of particular importance as this food item was ingested by all three study species at some stage of their life history. Spatial and temporal variation in diet composition of the study species was mostly associated with changes in prey items available across floodplain habitats and between seasons (summer/winter). The low magnitude of flooding events during the study period is arguably the most likely factor influencing the lack of patterns of variation in fish diets in floodplain habitats subject to flooding, whereas in non-flooded lagoons the observed dietary variation was a consequence of successional changes in composition of the aquatic fauna as the dry season progressed. Water regime had an important effect on differences in fish diet composition across lagoons, but further evaluation of the influence of flooding is needed due to overall lack of major flooding events during the study period. Autochthonous resources, namely plankton, were the basis of the food web and phytoplankton in the seston is the most likely ultimate energy source for fish consumers, via planktonic suspension feeders (zooplankton). Nevertheless, organic matter could not be disregarded as an important energy source for invertebrates and higher consumers. In general, the present study does not provide support for the major models predicting the functioning of large rivers, such as the River Continuum Concept and Flood Pulse Concept, which argue that allochthonous organic matter either from upstream or from the floodplain are the most important sources of carbon supporting higher consumers. In contrast, the Riverine Productivity Model would be more appropriate to describe the food web and energy sources for consumers in the Macintyre River floodplain as this model suggests that local productivity, based on autochthonous phytoplankton and organic matter, fuels food webs in large rivers. The results of this study suggest that factors known to affect phytoplankton production in floodplain lagoons (e.g. flow regulation, turbidity and nutrient/herbicide inputs) must be seriously considered in current landscape and water management practices. iii

Statement

I hereby declare that this work has never previously been submitted for a degree or diploma at any University and to the best of my knowledge and belief, this thesis contains no material previously published or written by another person except where due reference is made in the thesis itself.

Elvio S. F. Medeiros

Acknowledgments

I am indebted to several people for their direct or indirect contributions towards the accomplishment of this thesis. Firstly, I would like to express my sincerest gratitude to my principal supervisor Prof. Angela Arthington (CRL, Griffith University) for the many insightful suggestions throughout the project and during the preparation of the thesis, and above all, her guidance, support and patience throughout this process. I am also grateful to my associate supervisor Dr. Glenn Wilson (Northern Basin Laboratory, MDFRC) for his support with most of the fieldwork aspects of this study and for helpful comments on drafts of this thesis. I am also grateful to Prof. Stuart Bunn (CRL, Griffith University) for valuable comments on the stable isotopes chapter and advice on the field design during the earlier stages of this study.

I must also thank my colleagues from the CRL, Mark Kennard, Steve Mackay, Wade Hadwen and Harry Balcombe, for their help and encouragement with assorted issues varying from statistical analyses to fieldwork techniques. I am thankful to Lacey Shaw, Deslie Smith and Maria Barrett (CRL) for their help with everything related to my student account details and paperwork. Thanks also to Rene Diocares and Vanessa Fry from the Stable Isotope Lab. for answering about a million questions. To the landholders in Goondiwindi, Brian Duddy, Peter Campbel, Wally Taylor, David Evans and Rob Newell, I am very grateful for allowing me access through their property to reach the study sites and for providing useful information on some of the many facets of the floodplain lagoons.

I wish to express my gratitude to the Brazilian government agency CAPES (Federal Agency for Post-Graduate Education) for their trust, and for providing financial support in the form of a 4-year scholarship, including University fees and living expenses. I am also grateful to the Faculty of Environmental Sciences and Centre for Riverine Landscapes (Griffith University) for financial support in the form of a grant covering fieldwork expenses, stable isotope analyses and additional University fees.

Finally, I would like to express my deepest gratitude to my family for their understanding and support. I am especially grateful to my dearest Fernanda, for her unconditional love, encouragement and belief in my potential. To her I dedicate this thesis. v

Table of Contents

Abstract...... i

Statement ...... iii

Acknowledgments ...... iv

Table of Contents ...... v

List of Tables ...... ix

List of Figures...... xi

1 General introduction...... 1 1.1 Floodplain rivers...... 1 1.2 Energy sources in large floodplain rivers...... 2 1.3 Feeding ecology of fish in floodplain rivers...... 4 1.3.1 Conceptual model of fish diets for the floodplain of the Macintyre River ...... 6 1.4 Effects of river regulation...... 8 1.4.1 Conceptual model of fish diets for natural and regulated floodplain sites...... 10 1.5 Study species...... 12 1.6 Choice of study sites ...... 13 1.7 Aims ...... 14 1.8 Thesis outline...... 15

2 Study area ...... 18 2.1 Location...... 18 2.2 Climate...... 21 2.3 Hydrology ...... 21 2.4 Geomorphology ...... 24 2.5 Geology...... 26 2.6 Soils...... 26 2.7 Vegetation...... 27 2.8 Land use...... 28 2.9 Fish community and aquatic invertebrate fauna...... 29

3 Spatial and temporal variation in limnological characteristics of the study sites...... 31 3.1 Introduction ...... 31 3.1.1 Aims ...... 32 3.2 Methods...... 33 3.2.1 Hydrology ...... 33 3.2.2 Lagoon morphology...... 33 vi

3.2.3 Habitat characteristics...... 34 3.2.4 Physico-chemical parameters...... 34 3.2.5 Data analysis ...... 35 3.3 Results...... 36 3.3.1 General hydrological patterns ...... 36 3.3.2 General description of the study sites ...... 37 3.3.3 Temporal variation in lagoon morphometry ...... 42 3.3.4 Lagoon microhabitat characteristics ...... 47 3.3.5 Physico-chemical parameters...... 50 3.4 Discussion...... 55 3.4.1 Macintyre River flows...... 55 3.4.2 Morphology and habitat characteristics ...... 55 3.4.3 Water quality characteristics...... 57 3.5 Conclusions...... 61 3.6 Implications for the study of fish diet composition...... 62

4 Diel variation in food intake and diet composition of fish in floodplain lagoons ...... 64 4.1 Introduction ...... 64 4.1.1 Aims ...... 65 4.2 Methods...... 66 4.2.1 Study design...... 66 4.2.2 Collection methods...... 68 4.2.3 Dietary analysis...... 69 4.2.4 Data analysis ...... 70 4.2.4.1 Diel feeding activity ...... 70 4.2.4.2 Diel variation in diet composition...... 71 4.3 Results...... 73 4.3.1 Diel feeding activity...... 73 4.3.2 Diel variation in diet composition...... 84 4.4 Discussion...... 94 4.4.1 Diel feeding activity...... 94 4.4.2 Diel dietary composition...... 98 4.5 Conclusions...... 100 4.6 Implications for the study of fish diet composition...... 101

5 Variation in diet composition of fish among different size classes in floodplain lagoons ...... 102 5.1 Introduction ...... 102 5.1.1 Aims ...... 103 5.2 Methods...... 104 5.2.1 Data analysis ...... 105 5.3 Results...... 106 vii

5.3.1 Ambassis agassizii...... 109 5.3.2 Leiopotherapon unicolor...... 111 5.3.3 Nematalosa erebi ...... 115 5.4 Discussion...... 117 5.5 Conclusions...... 121 5.6 Implications for the study of fish diet composition...... 122

6 Spatial and temporal variation in diets of fish in floodplain lagoons ...... 123 6.1 Introduction ...... 123 6.1.1 Aims ...... 125 6.2 Methods...... 126 6.2.1 Data analysis ...... 128 6.2.1.1 Analysing the significance of patterns ...... 129 6.3 Results...... 131 6.3.1 General dietary composition ...... 131 6.3.2 Spatial and temporal variation in diets of the study species ...... 137 6.3.2.1 Ambassis agassizii...... 137 6.3.2.2 Leiopotherapon unicolor...... 142 6.3.2.3 Nematalosa erebi...... 147 6.4 Discussion...... 157 6.4.1 Overall influences on fish diets...... 157 6.4.2 Influence of spatial variation on the diets of individual species...... 159 6.4.2.1 Ambassis agassizii...... 159 6.4.2.2 Leiopotherapon unicolor...... 162 6.4.2.3 Nematalosa erebi...... 163 6.4.3 Temporal patterns and the effects of flooding on diets of individual species...... 165 6.4.3.1 Ambassis agassizii...... 166 6.4.3.2 Leiopotherapon unicolor...... 167 6.4.3.3 Nematalosa erebi...... 168 6.4.4 Aspects of flow regulation ...... 170 6.5 Conclusions...... 171 6.6 Implications for the study of fish diet composition...... 172

7 Stable isotope analysis of energy sources for fish in floodplain lagoons ...... 174 7.1 Introduction ...... 174 7.1.1 Aims ...... 175 7.2 Methods...... 176 7.2.1 Stable isotope analysis: review of methods ...... 176 7.2.2 Study design...... 177 7.2.3 Collection of primary sources and consumers ...... 178 7.2.4 Sample preparation...... 179 viii

7.2.5 Data analysis ...... 180 7.3 Results...... 181 7.3.1 Primary sources...... 181 7.3.2 Consumers...... 184 7.3.3 Fish...... 186 7.3.4 Spatio-temporal relationships between δ13C values for fish consumers and primary sources...... 187 7.3.5 Contribution of autochthonous versus allochthonous sources of carbon to consumer biomass ...... 190 7.4 Discussion...... 192 7.4.1 Importance of allochthonous versus autochthonous sources of carbon to fish consumers...... 193 7.4.1.1 Leiopotherapon unicolor...... 195 7.4.1.2 Ambassis agassizii and Nematalosa erebi...... 196 7.4.2 Models of energy sources in floodplain rivers...... 197 7.4.2.1 The RCC and FPC...... 197 7.4.2.2 The RPM ...... 198 7.4.2.3 Caveats ...... 199 7.5 Conclusions...... 200

8 General discussion...... 202 8.1 Background...... 202 8.2 Major sources of variation in dietary ecology of fish...... 204 8.3 Feeding ecology of fish in the floodplain of the Macintyre River: dietary data...... 206 8.3.1 Effects of flooding and drought on fish diets: the temporal component...... 207 8.3.2 Effects of flow regime on fish diets: the spatial component...... 208 8.4 Importance of allochthonous and autochthonous energy sources to fish: stable isotope and stomach contents data...... 210 8.5 Management implications...... 212 8.6 Conclusions...... 213

Appendices...... 215 Appendix 1. Photos of the study sites (Chapter 3)...... 215 Appendix 2. Two-way ANOVA tables for diel variation in fish diets (Chapter 4) ...... 219 Appendix 3. Total and standard length relationships for all three species (Chapter 5) ...... 221 Appendix 4. Indicator Species Analysis results (Chapter 6)...... 222

Bibliography ...... 226

List of Tables

Table 3.1 Summary table showing water regulation, flow characteristics and management issues of the study sites during the study period (2002-2003). * Refers to local hydrological characteristics of the study sites. Note that all studied lagoons are affected by flow regulation of the Macintyre River (see Section 2.3). ** See text for details. TSR = Traveling stock reserve...... 38

Table 3.2 List of species recorded in each of the study sites on the floodplain of the Macintyre River, during the study period. * indicates exotic species...... 40

Table 3.3 Morphometry and depth characteristics of the lagoons studied on the Macintyre River floodplain on each sampling occasion. n/a = not applicable. Late dry season sampling occurred early in the summer of 2002 (October-November); Wet season corresponds to later in the summer of 2003 (March) and the Dry season samples correspond to the winter (July) of 2003...... 44

Table 3.4 Percent of variance of lagoon morphometry extracted by PCA for the first 6 axes (showing eigenvalues and broken-stick values) and contribution of the measured variables to the first 6 eigenvectors. Data correspond to the log(x+1) and double square root transformed variables for the three sampling occasions and all sites...... 45

Table 3.5 Coefficients of determination showing the correlations between lagoon habitat characteristics and the first three axes of the ordination space determined by DCA...... 47

Table 3.6 Mean proportions (± SD) of each habitat variable (expressed as % wetted perimeter) for each lagoon and sampling occasion on the floodplain of the Macintyre River. Data correspond to the average of all measurements taken in each individual site. Late dry season sampling occurred early in the summer of 2002 (October-November); Wet season corresponds to later in the summer of 2003 (March) and the Dry season samples correspond to the winter (July) of 2003...... 50

Table 3.7 Physico-chemical characteristics (± SD) of the Macintyre River floodplain lagoons on each sampling occasion. Data correspond to the average of all measurements taken in each individual site. Late dry season sampling occurred early in the summer of 2002 (October-November); Wet season corresponds to later in the summer of 2003 (March) and the Dry season samples correspond to the winter (July) of 2003...... 51

Table 3.8 Percent variance of lagoon water quality extracted by PCA for the first 6 axes (showing eigenvalues and broken-stick values) and contribution of the measured variables to the first 6 eigenvectors...... 53

Table 4.1 List of sites used for the analysis of diel variation in food intake and daily variation in diet composition of fish in dryland river lagoons...... 67

season, standard length (SL) (± SD), body weight (± SD), stomach fullness (± SD), percentage of nearly empty stomachs (fullness < 20 %) (n) and relative content volume (RCV) (± SD)...... 75

Table 4.3 Contribution by volume (% Vol) and frequency of occurrence (% Freq) of different taxa and major dietary categories (in bold) to the overall diets of A. agassizii, L. unicolor and N. erebi collected during two 24 hour periods (summer and winter) from the floodplain of the Macintyre River. Dashes indicate zero values...... 85

Table 5.1 Sequential length classes (TL, and approximate SL based on the regression equation) of the species studied, number of individuals analysed (n), Shannon-Wiener dietary breadth and correspondent season and sampling location in the floodplain of the Macintyre River...... 107

Table 6.1 Summary of results of spatial and temporal variation in diet composition of A. agassizii, L. unicolor and N. erebi from the floodplain of the Macintyre River. Data are sample sizes (N), mean size (TL) of individuals (± SD), stomach fullness (± SD) and mean dietary breadth (± SD) for sampling occasions and sites. (*) See Chapter 3 for site details. (**) Lagoon artificially filled (Serpentine Lagoon). (***) Discrepancies in total N of A. agassizii between this table and Table 6.2 are due to the removal of one outlier sample group of 15 individuals from ordination...... 132

Table 6.2 Contribution by volume (% Vol) and frequency of occurrence (% Freq) of different taxa and major dietary categories (in bold) to the diet of A. agassizii, L. unicolor and N. erebi collected from the floodplain of the Macintyre River throughout the entire study. Dashes indicate zero values...... 135

Table 7.1 Stable carbon isotope ratios (‰) of sources and consumers from floodplain lagoons and one site on the Macintyre River, on each of three sampling occasions, pre-flood (October 2002), post-flood summer (March 2003) and post-flood winter (July 2003). Data correspond to mean values (± SD, for n = 3-7 samples) or individual values (where n < 3)...... 182

Table 7.2 Stable nitrogen isotope ratios (‰) of sources and consumers from floodplain lagoons and one site on the Macintyre River, on each of three sampling occasions, pre-flood (October 2002), post-flood summer (March 2003) and post-flood winter (July 2003). Data correspond to mean values (± SD, for n = 3-7 samples) or individual values (where n < 3)...... 183

Table 7.3 Person’s coefficient of correlation (r2) for relationships between δ13C and δ15N values (‰) for each of the three species of fish and sampling occasions. Only relationships based on n ≥ 3 are shown.... 188

Table 7.4 Percent contribution of major energy sources (zooplankton, algae and organic matter) to the study species of fish on sampling occasions where zooplankton was available. Estimations are based on two- and three-source mixing models using δ13C and δ15N. (-) indicates no data available and (*) indicates solution where sum of proportions was more than 105%. Averages include only the feasible solutions. Note that only Rainbow and South Callandoon lagoons were flooded during the study period...... 192

List of Figures

Figure 1.1 Simple model of how diets of fish are expected to change in an Australian floodplain river in response to flooding and as water levels recede during dry periods...... 7

Figure 1.2 Conceptual model of how diets of fish are expected to vary in relation to different patterns of flow, from natural to regulated. (a) represents major flooding, (b) indicates water flow for regulated floodplain waterbodies, with permanently elevated water levels, (c) indicates semi-permanent natural floodplain lagoons and (d) indicates temporary natural floodplain lagoons...... 11

Figure 2.1 Map showing the location of the Border Rivers catchment within the Murray-Darling River system (inset A) and the study area within the Border Rivers catchment (inset B). Position of the lagoons studied is shown in Figure 2.2...... 19

Figure 2.2 Map showing the location of the lagoons studied in the floodplain of the Macintyre River and some of the major features of the landscape such as effluent creeks and towns...... 20

Figure 2.3 Mean flow volume (megalitres, ML) and runoff (mm) (± SE) for the Macintyre River recorded at the Goondiwindi gauging station (416201A) for the years 1917 to 2003...... 22

Figure 2.4 Hydrograph of highest annual peaks of flow by the Macintyre River recorded for the years 1917 to 2003 at the Goondiwindi gauging station (416201A). Flood classification (minor, moderate and major) are also depicted (source: http: //www.bom.gov.au/ hydro/ flood/ qld/ brochures/ border_rivers/ border_rivers. shtml)...... 24

Figure 3.1 Mean annual discharges (ML/day) for the Macintyre River recorded at the Boggabilla gauging station (416002) for the years 1995 to 2003...... 36

Figure 3.2 Daily discharge (ML/day) in the Macintyre River (recorded at the Boggabilla gauging station - 416002) between 2001 and 2003. Arrows indicate sampling occasions for the study on morphology, water quality and habitat characteristics of the study sites. Dates of flooding events during the study period are also indicated...... 37

Figure 3.3 Position of each study site and sampling occasion within the ordination space defined by the first three factors identified by PCA of the cross-products matrix of the correlation coefficients for lagoon morphometry. (a) axis 1 and 2, and (b) axis 1 and 3. Site codes correspond to the following: late=late dry season, 2002 summer; wet=wet season, 2003 summer; dry=dry season, 2003 winter...... 46

Figure 3.4 Distribution of the study sites within ordination space as defined by DCA of untransformed lagoon habitat characteristics. (a) axis 1 and 2, and (b) axis 1 and 3. Site codes correspond to the following: late=late dry season, 2002 summer; wet=wet season, 2003 summer; dry=dry season, 2003 winter...... 49

following: late=late dry season, 2002 summer; wet=wet season, 2003 summer; dry=dry season, 2003 winter...... 54

Figure 4.1 Plots of stomach fullness and relative content volume against fish standard length for individual species of fish collected in the floodplain of the Macintyre River during the study on diel variation in food intake. Plots show the relationship between RCV and stomach fullness versus fish size for (a) A. agassizii, (b) L. unicolor and (c) N. erebi...... 74

Figure 4.2 Frequency of occurrence histograms of stomach fullness for times of day and seasons of each of the fish species sampled from the floodplain of the Macintyre River. Data represents fullness classes of all stomachs sampled...... 77

Figure 4.3 Plots of mean values of fullness and RCV against seasons and species for a factorial ANOVA to determine the effects of seasonal and inter-specific differences in food intake by fish from the floodplain of the Macintyre River. Plots include the combined results from four diel sampling intervals for each species/season point. Numbers close to each point correspond to the p values of pairwise comparisons from that point to the following (species or season). Note that the level of significance is α=0.0056 after being corrected (9 a-priori comparisons) using the Bonferroni procedure...... 79

Figure 4.4 Plots of mean values of stomach fullness (left) and RCV (right) against time of day and season for each of the three species of fish based on a factorial ANOVA to determine the effects of seasonal and diel differences in food intake by fish from the floodplain of the Macintyre River. Numbers close to each point correspond to the p values of pairwise comparisons between that point and the one with the highest mean value, indicated by an (*) (usually 1200h or 1800h). Note that the level of significance is α=0.0056 after the α=0.05 was corrected (16 a-priori comparisons) using the Bonferroni procedure. 83

Figure 4.5 Relative Sorensen (Bray-Curtis)/Flexible Beta (β=-0.1) dendrogram of dietary composition similarities of A. agassizii based on different times of day (06h, 12h, 18h, 24h) and seasons (S=summer and W=winter) (a); and (b) two-dimensional NMS ordination plot of dietary samples of A. agassizii from different times of day and seasons. Note that, each dietary sample represents the mean volumetric data for groups of 3-6 randomly selected individuals. Dashed lines show secondary groups identified by the classification analysis...... 87

Figure 4.6 Percentage contributions by volume of different dietary items to the diet of A. agassizii collected during two 24-hour periods (summer and winter) from the floodplain of the Macintyre River. () indicates the mean dietary breadth (± SE) for each time of day. See Table 4.3 for full food items names...... 88

Figure 4.7 Relative Sorensen (Bray-Curtis)/Flexible Beta (β=-0.1) dendrogram of dietary composition similarities of L. unicolor based on different times of day (06h, 12h, 18h, 24h) and seasons (S=summer and W=winter) (a); and (b) two-dimensional NMS ordination plot of dietary samples of L. unicolor from different times of day and seasons. Note that, each dietary sample represents the mean volumetric data for groups of 3-6 randomly selected individuals. Dashed lines show secondary groups identified by the classification analysis...... 91 xiii

Figure 4.8 Percentage contributions by volume of different dietary items to the diet of L. unicolor during two 24-hour periods (summer and winter) from the floodplain of the Macintyre River. () indicates the mean dietary breadth (± SE) for each time of day. See Table 4.3 for full food items names...... 92

Figure 4.9 Relative Sorensen (Bray-Curtis)/Flexible Beta (β=-0.1) dendrogram of dietary composition similarities of N. erebi based on different times of day (06h, 12h, 18h, 24h) and seasons (S=summer and W=winter) (a); and (b) two-dimensional NMS ordination plot of dietary samples of N. erebi from different times of day and seasons. Note that, each dietary sample represents the mean volumetric data for groups of 3-6 randomly selected individuals. Dashed lines show secondary groups identified by the classification analysis. The insert box indicates the volumetric contribution of Daphniidae to the ordination. Note that the outliers account for most of the contribution of this food item...... 93

Figure 4.10 Percentage contributions by volume of different dietary items to the diet of N. erebi during two 24-hour periods (summer and winter) from the floodplain of the Macintyre River. () indicates the mean dietary breadth (± SE) for each time of day. See Table 4.3 for full food items names. Note that the percentage volume axis starts at 90% to enhance observation of all items...... 94

Figure 5.1 Two-dimensional NMS ordination plot of dietary samples of sequential size classes for A. agassizii during summer and winter in Rainbow Lagoon. Number of stomach contents examined for each size class collected on each sampling occasion is given in Table 5.1...... 109

Figure 5.2 Percentage contributions by volume of different dietary items to the diets of sequential size classes of A. agassizii collected during different seasons in Rainbow Lagoon. (a) winter of 2002, (b) summer of 2002 and (c) winter of 2003. Number of stomach contents examined for each size class on each sampling occasion is given in Table 5.1. Note that the percentage volume axis in Figure 5.2c is on a different scale to enhance observation of less abundant items. See Table 6.1 for full names of food items...... 111

Figure 5.3 Two-dimensional NMS ordination plot of dietary samples of sequential size classes for L. unicolor collected during different seasons and lagoons. Number of stomach contents examined for each size class on each sampling occasion is given in Table 5.1. The 2003 winter samples were taken from South Callandoon Lg. A and both summer and winter of 2002 samples were taken from South Callandoon Lg. B...... 112

Figure 5.4 Percentage contributions by volume of different dietary items to the diets of sequential size classes of L. unicolor collected during different seasons and lagoons. (a) winter of 2003 at South Callandoon Lg. A, (b) summer of 2002 at South Callandoon Lg. B and (c) winter of 2002 at South Callandoon Lg. B. Number of stomach contents examined for each size class on each sampling occasion is given in Table 5.1. See Table 6.1 for full names of food items...... 113

Figure 5.5 Two-dimensional NMS ordination plot of dietary samples of sequential size classes of N. erebi collected during different years and lagoons. Number of stomach contents examined for each size class on each sampling occasion is given in Table 5.1...... 115 xiv

Figure 5.6 Percentage contributions by volume of different dietary items to the diets of sequential size classes of N. erebi collected during different years and lagoons. (a) summer of 2002 at Maynes Lg., (b) summer of 2002 at Punbougal Lg., (c) summer of 2003 at Rainbow Lg. and (d) summer of 2003 at South Callandoon Lg. A. Number of stomach contents examined for each size class on each sampling occasion is given in Table 5.1. Note that the percentage volume axis in Figure 5.6b is on a different scale to enhance observation of less abundant items. See Table 6.1 for full names of food items...... 116

Figure 6.1 Joint plot showing the two-dimensional NMS ordination of spatial and temporal dietary samples averaged for sampling occasions, and food items correlated (r2>0.2) with sample points in ordination space (denoted by vectors) for A. agassizii, L. unicolor and N. erebi. The direction and length of vectors indicate strength of correlation. Each point represents the mean volumetric diet composition data for each sampling occasion. See Table 6.2 for full names of food items...... 137

Figure 6.2 Two-dimensional NMS ordination plot of spatial and temporal dietary samples for averaged sequential sizes of A. agassizii (a), and (b) joint plot showing the position of food items (+) and food items correlated (r2>0.2) with sample points in ordination space (denoted by vectors). The direction and length of vectors indicate strength of correlation. Each point represents the mean volumetric data for groups of 10-15 individuals. S=summer and W=winter. S1-S5 indicate average size (TL) of individuals within each group, as per Table 5.1...... 139

Figure 6.3 Percentage contribution by volume of different dietary items to the diet of A. agassizii during the entire study period (2002-2003) for different lagoons in the floodplain of the Macintyre River. () indicates the mean dietary breadth (± SE) for each sampling occasion. Arrows indicate flooding up to two months before a sampling occasion. S=summer, W=winter. See Table 6.2 for full names of food items...... 142

Figure 6.4 Two-dimensional NMS ordination plot of spatial and temporal dietary samples for averaged sequential sizes of L. unicolor (a), and (b) joint plot showing the position of food items (+) and food items correlated (r2>0.2) with sample points in ordination space (denoted by vectors). The direction and length of vectors indicate strength of correlation. Each point represents the mean volumetric data for groups of 5-15 individuals. S=summer and W=winter. S1-S6 indicate average size (TL) of individuals within each group, as per Table 5.1...... 144

Figure 6.5 Percentage contribution by volume of different dietary items to the diet of L. unicolor during the entire study period (2001-2003) for different lagoons in the floodplain of the Macintyre River. () indicates the mean dietary breadth (± SE) for each sampling occasion. Arrows indicate flooding up to two months before a sampling occasion. S=summer, W=winter. See Table 6.2 for full names of food items...... 147

Figure 6.6 Two-dimensional NMS ordination plot of spatial and temporal dietary samples for averaged sequential sizes of N. erebi coded for sizes (a), and (b) joint plot showing the position of food items (+) and food items correlated (r2>0.2) with sample points in ordination space (denoted by vectors). Dashed lines indicate outliers for larger individuals (see text). The vector corresponding to ‘detritus’ is 50% its xv

original length. Each point represents the mean volumetric data for groups of 10-15 individuals. S2- S6 indicate average size (TL) of individuals within each group, as per Table 5.1...... 149

Figure 6.7 Two-dimensional NMS ordination plot of spatial and temporal dietary samples for smaller individuals (S2, as per Table 5.1) of N. erebi (a), and (b) joint plot showing the position of food items (+) and food items correlated (r2>0.2) with sample points in ordination space (denoted by vectors). The direction and length of vectors indicate strength of correlation. Each point represents the mean volumetric data for groups of 10-15 individuals...... 151

Figure 6.8 Percentage contribution by volume of different dietary items to the diet of small individuals (S2) of N. erebi during the entire study period (2001-2003) for different lagoons in the floodplain of the Macintyre River. () indicates the mean dietary breadth (± SE) for each sampling occasion. Arrows indicate flooding up to two months before a sampling occasion. S=summer, W=winter. See Table 6.2 for full names of food items...... 152

Figure 6.9 Two-dimensional NMS ordination plot of spatial and temporal dietary samples for larger individuals (>S2, as per Table 5.1) of N. erebi (a), and (b) joint plot showing the position of food items (+) and food items correlated (r2>0.2) with sample points in ordination space (denoted by vectors). The direction and length of vectors indicate strength of correlation. The vector corresponding to ‘detritus’ is 50% its original length. Each point represents the mean volumetric data for groups of 10-15 individuals...... 154

Figure 6.10 Percentage contribution by volume of different dietary items to the diet of large individuals (>S2) of N. erebi during the entire study period (2001-2003) for different lagoons in the floodplain of the Macintyre River. () indicates the mean dietary breadth (± SE) for each sampling occasion. Arrows indicate flooding up to two months before a sampling occasion. S=summer, W=winter. See Table 6.2 for full names of food items...... 156

Figure 7.1 δ13C and δ15N values of primary sources (), primary consumers (S) and fish () for each sampling occasion (from top to bottom), averaged for sites subject to flooding (left) and non-flooded sites (right). Sampling occasions correspond to: before flooding (summer October 2002) and after flooding (summer March 2003) and July 2003 (winter). Major energy sources, consumers and fish species are highlighted...... 185

Figure 7.2 δ13C values of A. agassizii (S), L. unicolor () and N. erebi () versus δ13C of primary sources (zooplankton, algae and organic matter) across study sites and sampling occasions where all three primary sources were available...... 189

Figure 7.3 δ13C and δ15N values of primary sources (CPOM, algae and zooplankton) and fish for the sampling occasions where zooplankton was available. Sampling occasions correspond to: before flooding (summer October 2002) and after flooding (summer March 2003) and July 2003 (winter). Note that only Rainbow and South Callandoon lagoons were flooded during the study period...... 191

Figure 8.1 Simple food web model formulated for the Macintyre River floodplain lagoons from empirical dietary data and stable isotope data collected between late 2001 and 2003, and from information xvi available in the literature (see Chapters 5 and 6). Only strong links are illustrated, as per Table 6.2 (Chapter 6) and Figures 7.1 and 7.3 (Chapter 7). Bold solid lines represent interactions supported by stomach content analyses and stable isotopic signatures. Fine solid lines indicate interactions based on dietary data but lacking stable isotope support. Dashed lines indicate interactions inferred from data from the literature. Box size for energy sources (except for phytoplankton) represent their proportional abundance (as determined by their percent occupation of wetted perimeter, see Chapter 3), averaged for all lagoons and sampling occasions...... 207

Chapter 1. Introduction 1

1 General introduction

1.1 Floodplain rivers

Large floodplain rivers are distinctive landscape features and, in their natural state, are characterised by high biodiversity and productivity (Tockner and Stanford 2002). These environments are disturbance-dominated ecosystems, characterised by a high level of habitat heterogeneity and diverse biota adapted to the high spatio-temporal heterogeneity (Tockner and Stanford 2002). The most important drivers of processes in floodplain rivers and subsequent high biodiversity are the variable flow and the lateral connectivity of the river to the floodplain wetlands (Junk et al. 1989, Walker et al. 1995, Kingsford 2000, Amoros and Bornette 2002, Kingsford et al. 2004).

Defined as ‘areas of low lying land that are subject to inundation by lateral overflow water from rivers with which they are associated’ (Junk and Welcomme 1990), floodplain rivers are distributed throughout a wide range of climatic and geographic regions, from the tropics to temperate and arid regions (Tockner and Stanford 2002, Capon 2004). Drylands occupy about one-third of the world’s land surface area (Thomas 1989) and rivers in these regions have less predictable flow regimes than those in humid tropical and temperate regions (Walker et al. 1995). Rivers in these arid areas, particularly those in Australia, exhibit extreme hydrological variability that has a marked influence on their biological diversity and productivity (Puckridge et al. 1998, Young 1999, Puckridge et al. 2000, Kingsford et al. 2004). Although dryland rivers are renowned for episodic floods extending over vast floodplains, for much of the time they exist as networks of ephemeral channels and turbid waterholes (Bunn and Davies 1999). This fragmentation is a natural feature and contributes to a high degree of spatial and temporal habitat heterogeneity (see Cellot et al. 1994, Usseglio-Polatera 1994, Heiler et al. 1995, Amoros and Bornette 2002). During dry periods, the larger waterbodies represent the only aquatic habitat for biota requiring permanent water and these habitats are likely to function as refugia for a wide variety of aquatic organisms (Morton et al. 1995, Magoulick and Kobza 2003). These refugia can operate as safe habitats (Robertson et al. 1995), serving to reduce mortality associated with flood and drought disturbance (Winterbottom et al. 1997, Lake 2000, 2003). Chapter 1. Introduction 2

1.2 Energy sources in large floodplain rivers

Over the last 10 years an evident trend in river ecosystem research has been the increased emphasis on larger systems and, in particular, on the importance of interconnections between river channels and their floodplains (Thoms and Sheldon 2000, Tockner et al. 2000, Amoros and Bornette 2002, Bunn and Arthington 2002). However, prominent conceptual models such as the River Continuum Concept (RCC, Vannote et al. 1980), nutrient spiralling (Newbold et al. 1981), and the Serial Discontinuity Concept (Ward and Stanford 1983) failed to incorporate interactions between the river channel and the floodplain, in their initial formulations, especially in large rivers, although this was to some extent rectified in subsequent modifications (see Sedell et al. 1989, Walker et al. 1995, Ward and Stanford 1995, Townsend 1996, Fisher et al. 1998). Such models describe upstream/downstream linkages in energetic, nutrient and biological parameters as a longitudinal continuum. The RCC depicts river systems as a continuous gradient of physical conditions which elicit a series of responses from biotic communities, and argues that downstream communities are adapted to capitalize on upstream processing inefficiencies (leakage), therefore, emphasising the influence of nutrients and organic matter from upstream processes on the structure and function of lowland reaches.

According to the RCC, allochthonous inputs contribute comparatively less as energy sources where the riparian zone has been cleared or as the stream broadens downstream. Hence, the reduced importance of terrestrial organic inputs gives way to an enhanced significance of autochthonous primary production as well as organic transport from upstream (Vannote et al. 1980). Heterotrophic streams become autotrophic as the channel becomes wider and riparian regulating factors, such as light and temperature limitation, are reduced, allowing for increased algal or rooted vascular plant production (Vannote et al. 1980, Clapcott 2001). Consequently a change in functional feeding group representation is predicted by the RCC. Filter-feeding and browsing aquatic invertebrates are an important functional group in downstream reaches where they capture fine particles of organic matter and support larger consumers such as fish. In wider more open river reaches direct inputs of leaf litter and other organic matter from adjacent riparian vegetation are therefore considered to be minor. Even though middle-order reaches, where the direct effects of Chapter 1. Introduction 3 riparian shading are diminished, may have higher dependence on in-stream primary production, organic matter derived mostly from upstream processing is claimed to be the major source of nutrients supporting food webs in downstream reaches (Vannote et al. 1980).

One of the most comprehensive models focusing on river-floodplain dynamics is the Flood Pulse Concept (FPC, Junk et al. 1989), which recognised that rivers and their fringing floodplains are integrated components of a single dynamic system, linked by strong interactions between hydrological and ecological processes (Tockner et al. 2000). The major driving force is the pulsing of river discharge that determines the degree of connectivity and the exchange of matter and organisms across river-floodplain gradients (Junk et al. 1989, Tockner et al. 2000, Amoros and Bornette 2002, Tockner and Stanford 2002). In contrast to the RCC, the FPC proposes that riverine food webs are more dependent on production derived laterally from the floodplain and its wetlands than on organic matter transported from upstream. During floods, aquatic organisms migrate to the floodplain and exploit the newly available habitats and their resources, whereas, as floodwaters recede, nutrients and newly produced animal biomass are returned to the main river channel (Johnson et al. 1995, Bunn et al. 2003). This pulse is coupled with a dynamic edge effect, which extends a moving littoral throughout the aquatic/terrestrial transition zone. This moving littoral prevents prolonged stagnation and allows rapid recycling of organic matter and nutrients, thereby resulting in high productivity (Junk et al. 1989). The FPC is, however, derived mainly from research carried out in large tropical floodplain river systems with predictable flood pulses of long duration (Junk and Weber 1995, Tockner et al. 2000), and has received relatively little attention in temperate and arid-zone systems (Walker 1992, Walker et al. 1995, Thoms and Sheldon 2000, Sheldon et al. 2002).

An alternative and supplementary view to the prevailing paradigms of ecosystem function in large rivers (i.e. RCC and FPC) is proposed in the Riverine Productivity Model (RPM, Thorp and Delong 1994). This model emphasizes the importance of local in-stream productivity (mostly algae and other aquatic plants) and, to a lesser extent, direct inputs of organic matter from the adjacent riparian zone. It is argued that the previous models underestimate the role of local autochthonous sources, and over-emphasize the transport of organic matter from both headwater streams (RCC) and floodplains (FPC) (Thorp and Chapter 1. Introduction 4

Delong 1994). Furthermore, Thorp and Delong (1994) point out that nutrients from autochthonous production and the riparian zone are especially important to riverine food webs because they are easily assimilated and readily available for longer periods near the banks, where benthic organisms tend to aggregate.

Although these three major models of large river ecosystem function provide a basis for understanding of potential sources of nutrients and general food web structure, they clearly differ in their emphasis on energy and nutrient inputs from the riparian zone versus upstream and floodplain processes, and their relative importance to ecosystem functioning (Clapcott 2001, Bunn et al. 2003). Moreover, for many rivers, flow regimes may be neither regular nor predictable, as assumed by the above models, and the associated biological processes may also be variable in space and time, as is the case for Australian dryland rivers (Walker et al. 1995, Sheldon et al. 2002).

1.3 Feeding ecology of fish in floodplain rivers

Fish are important consumers in the main channel and floodplain habitats of large tropical rivers and are thought to depend to a great extent directly or indirectly on primary production from the laterally-linked floodplain habitats (Junk et al. 1989). Studies on feeding habits of fish inhabiting floodplains show that despite considerable anatomical specialization (e.g. mouth structure and alimentary tract), there is little exclusiveness of food selected by many species (Lowe-McConnell 1987), the reason being that fish in such environments are opportunistic and must be very flexible in their feeding habits to take advantage of food items as and when they become available (Welcomme 1979). Food supplies in floodplain habitats during the flood phase may be highly abundant (Junk et al. 1989), whereas food resources are usually limited during dry seasons, and the many species stranded in drying waterbodies disconnected from the main river channel have to partition whatever food is available, with fish diets often reliant upon benthic detritus (Lowe-McConnell 1964). In floodplain rivers, periods of fasting coincide with low or falling water levels and are associated with decreases in seasonal fat content in many fish species (Junk 1985). Besides the usual tendency of many species to select a succession of food types as the flood cycle progresses, and as they become available in habitats subject to different degrees of connection with the main channel (Bishop et al. 2001), there is a Chapter 1. Introduction 5 tendency for food preferences to change throughout growth. Thus, the diet of juvenile fish often differs from that of the adults of the same species (Welcomme 1979, Kennard 1995, Bunn et al. 2003).

A variety of species of fish in floodplain river systems make use of the floodplain at times of high flow for feeding (Welcomme 1985) and the role of allochthonous food items to fish diets in many floodplain rivers is well recognised. Commonly, insect and plant foods consumed by fishes originate from overhanging vegetation and several families of fishes in the Amazon rely heavily on fruits and seeds at times of high flows (Goulding 1980). Detritivores and predators are also important consumers in floodplain systems and take advantage of flooding to gain access to accumulations of detritus on the floodplain (Santos 1982, Almeida 1984) and to make use of the large amount of terrestrial insects falling from forest vegetation (Goulding 1980). The importance of allochthonous vegetable matter as a direct food for many species and the role of insects as food for fish have been identified as major trophic features of tropical freshwater fish communities (Lowe-McConnell 1987, Pusey and Arthington 2003). The importance of mud and detritus as food for fishes that become specialised to strain large quantities of these for their contained microorganisms has also been emphasized (Lowe-McConnell 1987). In Australian freshwaters, aquatic insects are the most important food resource for fish species (Kennard et al. 2001), with microcrustaceans (mostly zooplankton), algae and food from terrestrial origin (mainly terrestrial insects) also providing important food sources, whereas macrocrustaceans and fish comprise minor components of Australian fish diets (Pusey et al. 2004).

The periodic connection between main river channels and their floodplains is well recognised as having an extremely important role in riverine trophic processes (Junk et al. 1989, Humphries et al. 1999), but the nature of the relationship between fish feeding and floodplain resources in Australian rivers, e.g. the Murray-Darling system, is not fully understood (Humphries et al. 1999). The majority of Murray-Darling Basin fishes are opportunistic carnivores (Merrick and Schmida 1984, Pusey et al. 2004) with few species known to consume plant material or detritus (Pusey et al. 2000, Balcombe et al. in review).

Quantitative dietary data are available for many fish species in almost all regions of Australia (Kennard et al. 2001, Pusey et al. 2004). However, as recognised by Kennard et al. (2001), there may be a significant regional pattern in fish diets, probably as a result of Chapter 1. Introduction 6 variation in local productivity, food availability and species composition, or a combination of these factors. At a regional level, there is currently a lack of information about spatial and temporal variation in the diets of fishes inhabiting floodplain habitats, especially in Australian dryland rivers. In particular, the role of autochthonous and allochthonous materials in fish dietary ecology is poorly understood in these ecosystems.

The ecological importance of Australian floodplain waterbodies has only recently begun to become appreciated (McCosker 1996, Casanova 1999, Watts 1999, Sheldon et al. 2002), despite the fact that several studies have highlighted the importance of these ecosystems as supporters of a large and diverse biomass of micro- and macroorganisms, e.g. bacteria and invertebrates (Boon 1991, Cranston and Hillman 1992). The idea that these dense populations of billabong invertebrates represent a significant food resource for fish, when rivers and floodplain are connected during floods, may be obvious, but it has yet to be tested (Hillman 1995b). Floodplain waterbodies are believed to serve as feeding and nursery areas for larval and juvenile fish and have been reported to increase recruitment of juvenile fish to main-channel populations (Geddes and Puckridge 1989). However, Humphries and Lake (2000) emphasized that parameters such as food supply and predation must also be taken into account when identifying conditions suitable for fish recruitment. Floodplain waterbodies are regarded as highly variable and productive (mainly with respect to microorganisms), but there is a poor understanding of fish dietary ecology in these systems and little information on how they are able to support their fish communities, especially during periods of drought (cf. Lake 2000, 2003).

1.3.1 Conceptual model of fish diets for the floodplain of the Macintyre River

Based on the literature and information presented above, a simple predictive model of how the diets of fish are expected to vary in floodplain waterbodies in Australian rivers, in relation to dry and wet periods is presented in Figure 1.1 This hypothetical model is applicable to the Macintyre River, Border Rivers catchment, Murray-Darling Basin, and sets the framework for the present study.

Australia’s dryland river-floodplain systems cycle through recurrent periods of floods and droughts often resulting in extreme hydrological variability (Puckridge et al. 1998). As a Chapter 1. Introduction 7 result, these systems have been described as having a “boom and boost” ecology (Kingsford et al. 1999, Kingsford et al. 2004) with periods of high productivity associated with flooding (Bunn and Davies 1999). Not surprisingly, flow and its variability have been recognised as major driving forces in the ecological functioning of Australian rivers (Walker et al. 1995, Bunn and Arthington 2002, Capon 2004) and responses to flow variability from fish (Gehrke et al. 1995) and aquatic invertebrates (Sheldon et al. 2002) have been reasonably well described.

FLOODING

Reduced abundance High floodplain and diversity of food productivity prey types available to fish

Greater contribution of autochthonous material to fish diets Increased biomass Decreased biomass of microorganisms of microorganisms and invertebrates IMPLICATIONS FOR FISH DIETS and invertebrates

Greater contribution of allochthonous material to fish diets

Increased abundance and diversity of food Decreased local prey types available productivity to fish

RECEDING WATERS

Figure 1.1 Simple model of how diets of fish are expected to change in an Australian floodplain river in response to flooding and as water levels recede during dry periods.

The conceptual model for the present study (Figure 1.1) describes predicted changes in diets of fish based on flow variability. It is predicted that, after floodplain inundation, primary productivity will increase in floodplain habitats which will, in turn, increase the abundance and diversity of prey available to fish. It is hypothesised that floodplain inundation will also lead to an increase in food items from allochthonous sources of terrestrial origin, e.g. insects and organic material from riparian vegetation. As flood Chapter 1. Introduction 8 waters recede, and fish are confined to highly turbid (see Bunn and Davies 1999) floodplain waterbodies without connectivity to the main river channel, local productivity on floodplain habitats is hypothesised to decrease (Welcomme 1979). As a result of a decreased biomass of microorganisms and invertebrates, the diversity of food prey types available to fish is reduced. At this stage of the flood-drought cycle, it is thought that autochthonous food sources, e.g. aquatic invertebrates and algae, will make greater contributions to fish diets and allochthonous contributions will decrease (Figure 1.1).

1.4 Effects of river regulation

In Australia, the reduced amount of water reaching floodplain waterbodies due to river regulation has been held responsible for successional changes in aquatic vegetation and for observed declines in population numbers of native fish and invertebrates (Kingsford 2000). However, information regarding the impacts of water resource development on biota in Australia has generally focused on within-channel processes of rivers (Walker 1985, Lake and Marchant 1990, Bren 1993), not on floodplains, which are arguably more affected by water development (Kingsford 2000). Water alienation from floodplains by structures such as levees, block banks and weirs leads to a loss of connectivity between river and floodplain and reduced flows into floodplain waterbodies. Despite the fact that these factors can lead to high turbidity, destroying submerged vegetation (Casanova 1999) and loss of productivity, the ecological impacts of the above-mentioned structures, and floodplain alienation, on floodplain biota are still poorly understood.

The storage and release of water for irrigation purposes is one of the most common and important forms of flow regulation in Australian rivers, especially in the Murray-Darling river system (Humphries and Lake 2000). This type of river regulation may alter the transport of particulate organic and inorganic matter (Petts 1984) and, consequently, local productivity (Walker et al. 1995). Ecological processes concentrated in the rising and falling phases of the flood pulse are disrupted by river regulation, whereas those active during the period of stable waters are reinforced (Walker et al. 1995). River regulation also has detrimental effects on fish populations (Cadwallader 1978, Bunn and Arthington 2002). Natural assemblages of species with flexible and opportunistic life histories may be Chapter 1. Introduction 9 displaced by others adapted to seasonally stable, low-flow environments, and these conditions may favour introduced species (Walker et al. 1995, Bunn and Arthington 2002).

Studies indicate that lacustrine areas in floodplains open to the full flooding amplitude from the main river are more productive (including fish yield) than artificially separated or impounded lakes (Welcomme 1985). Comparative studies of fish yields and biomass between floodplain areas open to the flood pulse of the main river and those areas impounded or separated by artificial levees indicate that the regularly inundated ones are significantly more productive than lakes or backwaters separated from the main channel or impounded reaches on the river (Lelek 1989). Bryan and Sabins (1979) found that fish biomass in floodplain lakes connected to the main river was higher than in lakes separated from the river for most of the year by artificial levees. Fremling et al. (1989) observed that standing-stock estimates from connected lakes were greater following high-water years than during low-water years. The elimination of anabranches or backwater systems by water control systems, such as the construction of levee banks (which reduce the floodplain area), is believed to reduce the extent of waters which previously provided the necessary space and food for the young of many fish species (Cadwallader 1978).

Major effects of river regulation on trophic ecology of fishes are related to changes in food resources available in regulated sections and alterations in contributions from allochthonous and autochthonous organic material as main energy sources supporting fish communities (Goulding 1980, Araujo-Lima et al. 1995, Mérona et al. 2000). However, Mérona et al. (2000) recognised that changes in food resources appear more difficult to foresee in river sections below major impoundments, as they probably depend on a large number of parameters including the amount of nutrients reaching downstream sections and characteristics of the flow regime. The way fish communities adapt to and utilise these new resources is poorly understood and some of the few available results are contradictory (Ferreira 1984, Araujo-Lima et al. 1995, Mérona et al. 2000).

Several water storages constructed along the Macintyre River, Border Rivers catchment, Murray-Darling Basin (see Section 2.3 and Figure 2.1 in Chapter 2), have led to alterations of natural flow and, to some extent, altered patterns of floodplain inundation. In the main channel, flows are more continuous and sustained during summer, and river banks no longer dry out because of the artificially-maintained flow heights. The general model Chapter 1. Introduction 10 presented in Figure 1.1 was based on information describing relatively natural floodplain river systems (see also Section 1.3). Given the nature of land use in the Macintyre River sub-catchment (Section 2.8 in Chapter 2) and the regulated nature of its water flows and floodplain habitats (Section 2.3 in Chapter 2), the model presented in Figure 1.1 needs to be modified to accommodate local conditions of flow regulation by the Macintyre River and many of its floodplain lagoons. Such a model is presented in Figure 1.2 (see next section).

1.4.1 Conceptual model of fish diets for natural and regulated floodplain sites

Two major flow categories can be identified in the floodplain waterbodies of the Macintyre River (see Section 1.6). The first major flow pattern is found in the regulated sites which, in the present study area, are characterised by having permanently elevated water levels. Study sites with such characteristics include the Macintyre River and Rainbow Lagoons (see Sections 3.3.1 and 3.3.2 and Table 3.1 in Chapter 3 for study sites characteristics). An additional level of flow regulation is present in Serpentine Lagoon, which was artificially filled but water levels were not kept artificially elevated during the study period. Therefore, water levels at this lagoon slowly decreased throughout the study period. The second pattern in flow regime is found in floodplain lagoons, which fill through natural overland flows and then dry down through evaporative losses. These include the following study sites: Punbougal, Maynes, South Callandoon (A and B) and Broomfield lagoons (Chapter 3). In such cases, some lagoons can dry out after several months of no flow (temporary lagoons) or they may retain water for years (semi-permanent), drying out much more slowly than temporary lagoons. These floodplain habitats can be classified as a nested hierarchy regarding their water levels and flow patterns (dendrogram on the right of Figure 1.2) and their flow characteristics are depicted as a conceptual hydrograph on the left side of Figure 1.2.

In the conceptual model presented in Figure 1.2, flooding is still hypothesised to be the major factor driving aquatic communities in the Macintyre River as, despite being controlled by upstream water storages, recurrent flooding events are a common feature in the study area (Section 2.3, Chapter 2). One important feature shown in Figure 1.2 is that major flooding events are hypothesised to reset the floodplain habitats of the Macintyre Chapter 1. Introduction 11

River, by flooding extensive areas and inundating all floodplain waterbodies (Figure 1.2a). On the other hand, as flood waters recede and connectivity with the main channel is lost, different floodplain waterbodies will respond in different ways regarding their pattern of variation in water levels (Figure 1.2b-d).

Flood (a)

Low temporal and Floodplain Flow high spatial variability Sites regime in fish diets

Permanently (b) Macintyre River elevated Regulated Rainbow

Water level Artificially filled (c) Serpentine Semi- Punbougal permanent Successional temporal changes in fish diets Maynes and high spatial (d) variability Natural South Callandoon A South Callandoon B Time Broomfield Temporary

Highly variable fish diets, both temporally and spatially

After flooding, regulated sites are expected to dry up with time, but only to some extent, remaining with artificially elevated water levels (Figure 1.2b). In contrast, water levels of natural waterbodies are expected to recede continuously and, in the case of the temporary lagoons, dry up within 6 to 12 months of inundation (Figure 1.2d). Even though water levels of semi-permanent lagoons are expected to decrease, this will generally occur in a less dramatic way in comparison with temporary sites (Figure 1.2c). Note that the Serpentine Lagoon lies in an intermediary position between regulated (this lagoon was Chapter 1. Introduction 12 artificially filled in 2001, see Chapter 3) and natural sites as its water levels steadily decreased throughout the study period.

Diversity in flow characteristics and high variability in water levels within and across floodplain lagoons is thought to contribute to maintaining temporal and spatial diversity of aquatic plant communities and the biota they support (see Boon et al. 1990). It is, therefore, hypothesised that flow variability across floodplain sites will have strong effects on local productivity and, as a result, on resources available for fish. The hypothetical model shown in Figure 1.2 anticipates that diets of fish will vary in response to general flow characteristics of floodplain habitats.

According to Marshall et al. (in review), after disconnection from the main river, floodplain waterbodies act as separate units, with composition of aquatic communities diverging in a manner that reflects the biota present at the time of disconnection. Even though major flooding could reset biological conditions in floodplain habitats to earlier successional stages (see Junk et al. 1989), in the absence of flooding or after major flooding, when biotic processes in individual habitats start to respond individually to the water regime, fish diets are hypothesised to vary in response to patterns in local flow characteristics and food availability (Figure 1.2). Fish from more permanent habitats can be expected to have a less variable dietary composition, reflecting the relatively unchanged flow conditions through time. On the other hand, prey items consumed by fish from temporary and temporally variable habitats, should reflect this variability, whereas fish from semi-permanent habitats should present patterns of dietary composition which reflect slower, successional changes in water regime and food items available (Figure 1.2).

1.5 Study species

In the present study, fish representatives of major functional feeding groups present in the Macintyre River were selected in order to test the hypotheses generated in sections 1.3.1 and 1.4.1 (Figures 1.1 and 1.2). It was important to choose species of fish that feed on different trophic levels of the food web (e.g. detritus/algae, zooplankton or aquatic insects) and, therefore, rely on different energy sources, either allochthonous or autochthonous. Additionally, the ecological importance of the study species also needed to be taken into Chapter 1. Introduction 13 account, e.g. food source for other species of fish, and relative abundance, given that relatively large numbers of individuals are needed to assure proper statistical analysis of diets.

The major trophic groups of fish in the Border Rivers system, based on their expected feeding habits, are detritivores, omnivores and carnivores (McCosker 1996, Moffatt and Voller 2002). During preliminary field trips to prospective study sites, Ambassis agassizii (olive perchlet), Leiopotherapon unicolor (spangled perch) and Nematalosa erebi (bony bream) fulfilled the requirements described above. All of these species are widespread and relatively abundant in the study area. They also present relatively different feeding habits: A. agassizii has been described in the literature as a microcarnivore, L. unicolor as an omnivore and N. erebi as an algal/detritivorous species (Pusey et al. 2004). These species are relatively common in Australian freshwater systems, being important consumers and also sources of food for other species of fish and/or birds (see Section 2.9 in Chapter 2).

1.6 Choice of study sites

The sites sampled were located in the billabong zone of the Macintyre River (Border Rivers catchment) where the river passes through a relatively well-defined floodplain containing numerous intermittent and semi-permanent billabongs on prior river channels (see Section 2.4 in Chapter 2). A number of floodplain lagoons were selected to represent the widest possible range of water regimes available in the study area, from small relatively undisturbed lagoons to large and flow regulated floodplain lagoons (as per Figure 1.2). Seven lagoon sites on the Macintyre River floodplain and one site in the main channel of the river were sampled. Details of the study sites and when they were sampled are given in the appropriate chapters. Here their general characteristics are briefly described.

Study sites consisted of semi-permanent and temporary floodplain lagoons of different sizes located in different areas of the floodplain (within 0.1 to 12 km distant from the main river), and therefore prone to different patterns of flooding from the Macintyre River and subject to different degrees of water management. As the Macintyre River presents evident human influence such as agricultural practices, flow regulation and water Chapter 1. Introduction 14 diversions for irrigation, these factors were taken into account in the present study. The presence of large permanent lagoons and smaller temporary ones, relatively disconnected between themselves, and varying in the influence of flows received from the main river made the present study sites particularly useful for comparisons of how different types of floodplain lagoons are able to trophically support their fish fauna. These lagoons also presented different histories and degrees of water management (e.g. permanently elevated water levels, artificial filling, use by cattle or for recreational activities, etc), which enabled comparisons between modified and relatively natural sites. Further details on individual lagoon sites characteristics are given in Chapter 3 and in subsequent chapters when appropriate.

1.7 Aims

The following dissertation is concerned with how different types of natural and modified floodplain lagoons are able to trophically support their fish communities, and the implications of an extended period of drought on food resources consumed by fish and diet composition. It examines some of the physical and biological processes that occur within selected lagoon sites on the floodplain of the Macintyre River, Border Rivers catchment. More specifically, this study focuses on the influence of flooding and an extended dry period, and different levels of flow regulation, on the feeding ecology of selected fish species. Feeding ecology is examined firstly, in terms of diet composition of selected fish species, using stomach content analysis, and secondly, in relation to possible energy sources sustaining fish (using stable isotope analysis) in selected floodplain lagoons and a site in the main channel of the Macintyre River.

Food resources consumed by fish are hypothesised to vary in response to flooding, when inundation of isolated lagoons and vast floodplain areas can result in a burst of primary and secondary productivity. From the predictive models presented in Figures 1.1 and 1.2, it is anticipated that: (1) after floodplain inundation, increased abundance and diversity of food items available for fish will lead to greater contribution of allochthonous food items to fish diets, and (2) as water levels recede, a decreased biomass of microorganisms and invertebrates will lead to increased reliance of fish on autochthonous food sources. Given the permanently elevated water levels of regulated floodplain lagoons, (3) fish diets are Chapter 1. Introduction 15 hypothesised to be less variable in these floodplain habitats in comparison to diets of fish from floodplain lagoons with natural flow and water regime characteristics.

In particular, the present study aims to answer the following questions:

1. What are the main food resources consumed by A. agassizii, L. unicolor and N. erebi in floodplain lagoons of the Macintyre River?

2. How important are factors such as diel and ontogenetic variations in fish dietary composition and food intake in confounding overall dietary patterns of each fish species?

3. Are there temporal variations in diet composition of each fish species in response to flooding or, in the absence of flooding, in response to extended dry periods?

4. Are there spatial differences in diets of fish in relation to different levels of flow regulation in floodplain habitats? i.e. from natural floodplain lagoons to regulated floodplain lagoons (permanently elevated water levels).

5. Are fish dependent on autochthonous food resources, produced within floodplain lagoons, or on allochthonous food resources of riparian and floodplain origin?

6. What are the ultimate sources of energy supporting the food web and the three species of fish studied in floodplain lagoons of the Macintyre River?

1.8 Thesis outline

In Chapter 2 of the dissertation, physical and biological characteristics of the Border Rivers system and the Macintyre River are described and relevant general features of the study area are presented. Chapter 3 describes the hydrological characteristics of the Macintyre River with respect to seasonal and short-term inter-annual variation in the magnitude of water flows, and provides a broad description of the study sites, characterising their general physico-chemical characteristics and habitat attributes over a summer-winter period, including dry and wet seasons (2002-2003). Chapter 1. Introduction 16

Sampling procedures designed to study food habits of fish need to cover the peak feeding periods of individual species, to maximize the amount of information obtained. Other sources of variation in diet composition, such as diel variation, need to be examined in order to optimize comparisons of diet composition related to spatial and broader temporal patterns. Once diel feeding periodicity is understood, it is advisable to concentrate the sampling effort on periods of active feeding. Chapter 4 examines patterns of feeding activity, i.e. possible changes in the intensity of feeding and dietary composition, on a diel basis to determine the peak feeding times of the study species, olive perchlet (A. agassizii), spangled perch (L. unicolor) and bony bream (N. erebi).

Chapter 5 investigates possible size-related changes in diet composition of the three fish species in floodplain lagoons of the Macintyre River. It addresses the issue raised by Pusey et al. (2000), that ontogenetic changes in fish diets related to increasing body size may overwhelm the detection of any temporal and spatial variation in diet composition associated with fluctuating prey abundance and availability. As ontogenetic patterns in food consumption may be confounded by seasonal and spatial trends in prey availability, this chapter also compares diets of similar sized individuals of the same species in different floodplain lagoons and/or seasons. Information gained in this chapter generates a more precise understanding of size-related differences in diets of the target species, to be used in Chapter 6.

Many species of fish show seasonality in their food uptake related to flood cycles (Wootton 1990). Seasonal changes in food availability may be caused by changes in the habitat available for foraging, changes due to life-history patterns of food organisms and changes caused by the feeding activity of fish themselves (see Marchant 1982, Bishop et al. 2001, Sheldon et al. 2003). Various studies have related spatial and temporal variations in diet composition of fish to variations in abundance of different food items as they became seasonally available across habitats (Welcomme 1979, Griswold et al. 1982, Marchant 1982, Bishop et al. 2001). Chapter 6 investigates spatial and temporal changes in diet composition of the study species. It examines variation in diet composition of each species across a range of floodplain sites with different morphological characteristics and with different levels of flow permanence, from semi-permanent to temporary. The effects of modified water regimes on the trophic ecology of fish are addressed by comparing fish Chapter 1. Introduction 17 diets between natural and regulated floodplain sites (i.e. sites with permanently elevated water levels or subjected to artificial filling). Temporal changes in diet composition of fish are investigated by comparing broad seasonal (summer/winter) patterns of fish diets across different floodplain sites and in relation to patterns of flooding. In the absence of flooding, temporal changes are assessed on the basis of summer/winter variation in water levels and changes in fish diets over an extended dry period.

Studies on aquatic food webs have shown that the use of empirical food consumption data tend to limit the ecological value of food web depictions and cannot provide information on the rate of ingestion and assimilation of food by the species studied (Glasser 1983, Yodzis 1993). On the other hand, an isotopic analysis can provide an estimation of the mean level of organic matter actually assimilated by a given species (Peterson and Fry 1987, Fry 1991). Therefore, when combined with stomach contents analysis, stable isotopes can be used to assess the potential role of unidentifiable or unquantifiable food items (e.g. algal or detrital matter) in the diets of larger consumers such as fish.

Following the previous chapter describing spatial and temporal changes in diet composition based on stomach content data, Chapter 7 investigates the potential ultimate sources of energy supporting the three fish species using stable isotope analysis. It traces the sources of organic carbon and nitrogen through the food web, to assess the contribution of ultimate energy sources, such as algae and detritus, supporting fish species, and the importance of the ingested food items as revealed in the previous chapters, accounting for the availability of those sources over the sampling occasions.

The final chapter of this dissertation synthesizes the overall results presented in the preceding chapters and discusses them within the context of the available literature, the models of predicted patterns in diet composition developed in sections 1.3.1 and 1.4.1 (Figures 1.1 and 1.2) and the general models of energy flow for floodplain rivers, such as the Flood Pulse Concept and the Riverine Productivity Model (see Section 1.2).

Chapter 2. Study area 18

2 Study area

2.1 Location

This study was performed on floodplain lagoons of the Macintyre River in the Border Rivers catchment. This catchment is located along the southern Queensland and northern New South Wales border and comprises a major portion of the headwaters of the Barwon and Darling River systems (McCosker 1996) (Figure 2.1). The catchment encompasses roughly equal areas of the two states and has a total area of about 49500 km2 (DWR 1995).

The main rivers in the catchment include the Macintyre and New South Wales Severn rivers in the south-east, the Dumaresq River in the north-east, Macintyre Brook in the north and the Weir River in the north-west areas of the catchment. The eastern boundary of the catchment is formed by the Great Dividing Range, while the western extremity occurs at the junction of the Barwon and Moonie rivers, at about 45 km downstream of Mungindi (McCosker 1996).

The Macintyre River rises in the Waterloo Range, in the Great Dividing Range east of Inverell, and flows west then north-west through Wallangra (McCosker 1996). After the junction with the New South Wales Severn River, one of the major tributaries, the Macintyre, continues north-westerly and is joined by the Dumaresq River about 16 km upstream of Boggabilla (DWR 1995). The only main tributary downstream of Boggabilla is the Weir River, which enters the Macintyre River about 23 km upstream of Mungindi.

A number of effluent streams diverge from the Macintyre River in the vicinity of Boggabilla and Goondiwindi (DWR 1995) to form some of the streams that feed the many waterholes and floodplain lagoons in the region (Figure 2.2). The sites selected for this study were located within the billabong zone (see Section 2.4), to the west and east of the towns of Goondiwindi and Boggabilla on both the Queensland and New South Wales sides of the Macintyre River floodplain (Figure 2.2). Further details of the study lagoons are provided in Section 1.6 and Chapter 3. Chapter 2. Study area 19

N (A)

Brisbane Murray-Darling River system

Sydney r

catchment

e Border Rivers Border v

i R

NSW

r r

e e

i Mole v R i

R VIC

QLD

e S n r e

S SA

a a NT

d m d

a k n a n

r i

e D

u o e v D i

l R P o

a r m

l G

l B D o l

C e -Darling River system (inset A) and the studysystemthe -Darling River (inset and A)

v y

I q s e r a m u

a M er iv R n

Goondiwindi

e e

r k

e C

e r

(B)

v i

R e

y t

a M

l e

e G

R v

h i

R W

km 50

w r a MapthelocationMurra showing the the Border within of catchment Rivers

e i

n o

o M arealagoonsis(insetcatchment the in the Rivers within B). Position shown 2.2. studied Figure Border of Figure 2.1

Chapter 2. Study area 20

Flow Rainbow Lg.

r k

C e e

r C

c r

t Weir

a Boggabilla W y igg P gy ig

P Serpentine Lg.

l l

r Mo Boggabilla Maynes Lg.

e n

a iv l

R a h W

Weir Goondiwindi Goondiwindi

e r

Punbougal Lg.

i c

a M

k e e r C

Boobera Lg.

a r

r r

e o

a o

i D

Broomfield Lg.

River site k

e n r

i e

e r a

b r C o d

o B thefeaturesthethe Mapthethelocation showingMacintyre studied of major the of lagoons in of floodplain of and some River SouthCallandoon B Lg.

r o

B landscape such as effluent creeks and towns. towns. and creeks as effluent such landscape Figure 2.2

o 5 km

Figure 2.2 Mapd showing the location of the lagoons studied in the floodplain of the Macintyre River and

l some of the majorl features of the landscape such as effluent creeks and towns.

South Callandoon Lg. A QLD NSW Chapter 2. Study area 21

2.2 Climate

The Border Rivers system lies in Australia’s subtropical zone and is located on a broad transitional zone between the predominantly summer active rainfall mechanism of northern Australia and the winter active mechanisms of southern Australia (Hobbs and Jackson 1977). Rainfall in this catchment shows a uniform decrease from a maximum along the eastern boundary to a minimum in the west and occurs predominantly during summer, with 55% of precipitation falling between November and March (DWR 1995). The summer rains result from convective thunderstorm activity and tropical low pressure systems originating in northern Australia, and rainfall during this time is often of high intensity, whereas winter rains result from the intrusion of low pressure systems and associated cold fronts that are the major rain producing systems of the southern part of the continent (Hobbs and Jackson 1977). Annual average rainfall is in excess of 800 mm in the headwaters along the Great Dividing Range, whereas Mungindi receives an annual average of around 500 mm (DWR 1995). Evaporation exceeds precipitation throughout the whole catchment and pan evaporation ranges from around 1200 mm a year in the east to 2000 mm in the west (Wylie 1995).

There is an east-west gradient in temperatures, associated with altitude (Johnson 1999). At lower elevations on western areas of the catchment, average daily maximum and minimum temperatures in summer are around 35°C and 20°C, respectively, whereas corresponding winter temperatures are about 20°C and 5°C. In higher altitude areas to the east, average maximum and minimum summer temperatures are 27°C and 14°C respectively, while average winter maximum and minimum temperatures are 13°C and 1°C, respectively. Extreme temperatures have been recorded up to 45°C and down to -10°C (DWR 1995).

2.3 Hydrology

The Border Rivers system has an average annual discharge of one million megalitres (ML) (SCMCC 1990). The Macintyre River is a relatively large river with an average annual flow volume of 1115913 megalitres and a monthly mean flow volume of 85784 ML (measures at the Goondiwindi gauging station - 416201A). Even though this is a regulated Chapter 2. Study area 22 river, the flow regime of the Macintyre River can be characterised by a wet summer season, with the peak flow volumes reaching an average of 158251 ML between October and March, and a dry winter season from April to September (Figure 2.3). An average of 61.1% of the monthly runoff occurs during the summer (Figure 2.3).

160000 14 140000 12 120000 10 100000 8 80000

60000 6

40000 4 Mean monthly runoff (mm) Mean monthly flow volume (ML) 20000 2

0 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Month

Discharge Runoff

Figure 2.3 Mean flow volume (megalitres, ML) and runoff (mm) (± SE) for the Macintyre River recorded at the Goondiwindi gauging station (416201A) for the years 1917 to 2003.

Major water storages have been constructed in the Macintyre River valley, including Glenlyon Dam, on Pike Creek and Coolmunda Dam, on Macintyre Brook, both in Queensland, for water supply, and Pindari Dam on the Severn River, New South Wales, which was constructed for flood mitigation and supply of water (Arthington 1995) (Figure 2.1). This scheme regulates three river systems in the Border Rivers catchment: the Severn, Macintyre and Barwon Rivers are jointly regulated by Glenlyon and Pindari dams (as far downstream as the town of Mungindi); the Dumaresq and Macintyre/Barwon Rivers to Mungindi are regulated by Glenlyon and Coolmunda dams; and Macintyre Brook is regulated by Coolmunda Dam (DNR and DLWC 2000) (see Figure 2.1).

Chapter 2. Study area 23

Regulated river systems in this region are those in which the natural stream flow is supplemented by releases from Government storages. In the Border Rivers system, such regulated releases are monitored using a network of stream gauges (DNR and DLWC 2000). In the study area, the towns of Goondiwindi and Boggabilla are reliant on water supply from nearby weirs (Figure 2.2) and in low flow periods the release of water from Glenlyon and Pindari dams is necessary to meet stock and domestic requirements (Arthington 1995). Both Goondiwindi and Boggabilla weirs regulate the lower reaches of the Macintyre River, upstream of Boomi (Figure 2.1). Boggabilla weir is a re-regulating storage, which improves water distribution efficiencies by allowing the capture and use of small flows, as well as the capture, holding and re-regulation of releases from upstream dams (DNR and DLWC 2000). During periods of regulated flow, the minimum passing flow target at the end of the system, at Mungindi, is 20 ML per day (DNR and DLWC 2000). The numerous remaining weirs in the catchment (see DNR and DLWC 2000) are mostly between one and five metres high, and are used mainly as pumping pools for irrigation, stock and domestic users, as well as town water supply.

Before river regulation, periods of low flows or droughts were a common natural occurrence in the Macintyre River, downstream of Goondiwindi, and native fauna and flora had adapted to cope with such characteristics by surviving in deeper pools (Hillman 1995a, McCosker 1996). Presently, approximately 40% of the long term average flows in the Border Rivers are diverted for irrigation purposes (Gutteridge et al. 1992). River regulation has resulted in major changes in low-flow periods, where summer flows are more continuous and sustained, and river banks no longer dry out during summer because of artificially-maintained high flows (Hillman 1995a, McCosker 1996). Changes in river flows attributable to flow regulation and water diversion have reduced these dryland rivers to systems which now experience low flows for much longer periods of time than before regulation and high flow events have been much reduced, or fail to occur in some years (Thomson 1993, Arthington 1995, DNR and DLWC 1999, 2000).

Despite flow regulation, the pattern of flooding in the Macintyre River valley can vary considerably over time and major floods have occurred at an average frequency of once every 20 years over the past 80 years (Arthington 1995), with the major recorded flooding events occurring in 1890, 1976 and 1996 (Laurie et al. 1983, G. Wilson, pers. comm.). Additionally, moderate and minor floods have been recorded during the past several years Chapter 2. Study area 24

(Figure 2.4). Smaller streams in this catchment present highly variable discharges, with periods of no flow during long dry periods (Arthington 1995). Ring et al. (1984) reported that water flow in tributary streams in the upper Macintyre valley show a marked contrast with rainfall records as, despite a relatively abundant rainfall, flows exhibit a marked lack of persistence with periods of no flow being common. In the upland zone, as the rivers flow through well defined valleys, flooding occurs in a narrow band on either side of the main channel. The width of this band can vary from a narrow strip to several hundred metres. In the billabong and plains zones, the floodplains widen dramatically during flooding as the river enters an area of lower slope (Laurie et al. 1983).

2.4 Geomorphology

Based on the occurrence of distinct landforms, geomorphology, soils and vegetation, the Border Rivers system has been recognised as containing three distinct zones, termed uplands, billabong zone and riverine plains (Ring et al. 1984, McCosker 1996).

Chapter 2. Study area 25

The uplands include the upper reaches of the Weir River east of Moonie, Macintyre Brook above Yelarbon, the Dumaresq River upstream of Beebo and the Macintyre River upstream from Yetman, including the Severn River. A fine dendritic drainage pattern with short steep slopes in a hilly to rugged terrain on sedimentary, granite and basalt parent materials characterises this area. The rivers flow swiftly through well-defined valleys and riverbeds consisting mainly of coarse gravels derived from the parent granite, basalt, slate and greywacke (Ring et al. 1984, McCosker 1996).

The billabong zone, where the present study was carried out (Figure 2.2), extends downstream from Yetman to approximately 20 km west of Goondiwindi on the Macintyre River. Topography in this zone is undulating on slightly metamorphosed and sedimentary parent material, with plains comprising recent alluvia adjacent to the rivers. Throughout the billabong zone the rivers pass through a relatively well-defined floodplain which contains numerous intermittent and semi-permanent billabongs on prior river channels (McCosker 1996). The floodplain becomes broader in the lower reaches of this zone, and effluent watercourses associated with prior river channels are more pronounced (McCosker 1996).

Even though many small and larger billabongs exist within the riparian zone of the Macintyre River, according to McCosker (1996) there has been no comprehensive assessment of the biological status of floodplain wetlands in the billabong zone, with the exception of Boobera Lagoon, on the Morella Watercourse, which departs the Macintyre River upstream of Boggabilla and has considerable cultural significance to the local aboriginal community and is a popular recreation site. McCosker (1996) points out that other lagoons and effluents of the Macintyre River have been identified in various reports (e.g. DBBRC 1988, DWR 1990), including some of the study sites, such as Rainbow Lagoon, Maynes Lagoon, Serpentine Lagoon, and the Morella Watercourse, where lies the Punbougal Lagoon, and the Callandoon Branch and Dingo Creek, which are located nearby the Broomfield Lagoon (Figure 2.2), but there is little information regarding their current condition.

The riverine plains zone consists of an extensive alluvial floodplain and extends from approximately 20 km west of Goondiwindi to the end of the basin. Effluents and Chapter 2. Study area 26 anabranches are common and virtually no semi-permanent water bodies exist away from the main river channels (McCosker 1996).

2.5 Geology

Two distinct geological provinces divide the Borders River catchment, namely the western province, which comprises alluvial plains of the Weir, Macintyre and Dumaresq River systems, and the eastern province, which is characterised by a more elevated undulating hills and rugged mountain ranges (DWR 1995). In the western province, the land surface is flat and featureless, and rock outcrops are virtually non-existent (Johnson 1999). Surface geology is limited to the alluvial and riverine plain deposits which form a thin layer over a considerable thickness of sediment that was laid down on the edge of the Great Artesian Basin during the Jurassic-Cretaceous (Johnson 1999).

Extensive basalt extrusions occurred over southern parts of the uplands zone during the Tertiary period and in some places are still over 300 m thick, despite subsequent erosion which has exposed underlying rocks elsewhere (McCosker 1996). This dissection of basalts and granites in the uplands zone has resulted in extensive westward transportation of sediments (Ring et al. 1984). During the late Pliocene the eastern highlands were uplifted and subsequent extensive erosion and fluvial deposits in the western part of the basin formed the extensive alluvial plains that characterise the present day landscape of the riverine plains (Taylor 1978).

2.6 Soils

McCosker (1996) stated for soils in the Border Rivers basin that “soils developed on the granite and sedimentary parent materials are typically duplex, with a strong texture contrast between A and B horizons, often with a bleached A2 horizon. The exceptions are some yellow earths and lithosols. In the uplands zone, yellow and gray brown podsolics and red brown earths are common. These grade into podsolised red brown earths further west. Yellow earths typically occur on the Jurassic sandstones, while solodised red brown earths are common on sediments in the west of the basin (Ring et al. 1984). Chocolate and Chapter 2. Study area 27 black earths occur on basaltic parent material. These are often stony in nature in the higher uplands. Redder examples are found near the top of the catena, while chocolate, prairie, meadow and black earths are found lower down (Ring et al. 1984). Uniform deep cracking grey and brown clays have developed on the extensive alluvial deposits of the riverine plains. The darker soils occur in the lowest situations. Freely drained areas which may represent older surfaces or terraces are often brown or red-brown (McGarity 1977)”.

2.7 Vegetation

The upper areas of the Border Rivers catchment suffered widespread clearing in the past and nowadays only remnants of the original eucalypt woodland remain (Johnson 1999). However, some areas of steep hill country in the east still support a relatively rich diversity of plants (Johnson 1999). In the middle area of the catchment, original stands of Mountain gum (Eucalyptus pulverulenta), Silver-leaf ironbark (E. paniculata) and Smooth-barked apple (Angophora costata) woodland still stand, although the composition of the understorey has been considerably altered due to grazing (Boey et al. 1995). Many of the plant species found in the upper and mid sections of the catchment have been replaced by dryer floodplain species in the lower reaches of the basin, between Boggabilla and Mungindi, such as River red gums (Eucalyptus camaldulensis) and Coolabah trees (E. microtheca) (Johnson 1999). The composition of the plant communities in this area has been significantly affected by agricultural activities (Boey et al. 1995). According to Johnson (1999), large areas of natural grasslands have been lost due to grazing, cropping and pasture improvement pressures and in many areas introduced species are now predominant (Boey et al. 1995).

Species composition of the riparian vegetation in the Border Rivers catchment shows clear correlation with geographical position, as found by Boddy and Bales (1996), who show that at higher altitude to the east of the catchment, Casuarina cunninghamiana (river she- oak), Callistemon viminalis (bottlebrush) and Leptospermum brachyandrum (tea tree) are the most common species. Eucalyptus camaldulensis (River red gum) is present at the lower reaches of the basin and, downstream of Goondiwindi, E. microtheca (coolabah) becomes the dominant tree species. River red gum stands occur where there is a reliable water source, provided by ground water on active floodplains or other areas with a ground Chapter 2. Study area 28 water source or from surface water when rivers are in flood (Arthington 1995). The understorey of River red gum forests includes plants that can withstand inundation, such as Acacia spp. (cooba), Lignum spp. and various grasses and rushes, such as Eleocharis spp. (Arthington 1995).

A number of non-woody riparian species are also found, with the greatest diversity along the Mole and Dumaresq Rivers, and fewer species along the Macintyre River (Boddy and Bales 1996). The extent of non-woody vegetation is also determined by the intensity of grazing pressure, as grazing can alter the species composition by favouring the growth of unpalatable species (Boddy and Bales 1996).

2.8 Land use

Cropping and grazing comprise the major land uses in the Border Rivers basin, with approximately 10000 km2 used for crop production and 54000 km2 of grazing land (Wylie and Greggery 1995). Grazing of native and improved pastures dominates land use in the Border Rivers catchment, with livestock numbers around 850000 head of cattle and 5500000 sheep (Wylie and Greggery 1995). A substantial portion of the catchment has been sown to improved pastures over the last 40 years, with an estimated 330000 hectares now supporting sown pastures (Wylie and Greggery 1995, McCosker 1996). Poultry production, horticulture, cereal cropping (e.g. wheat, barley and sorghum) are also established in the region (Wylie 1995).

In recent years, cotton has become an important crop in the Boggabilla, Goondiwindi and Mungindi areas with around 50000 hectares of cotton grown annually in this part of the basin, and the use of irrigation has been increasing, with large scale irrigation occurring below the junction of the Dumaresq and Macintyre Rivers (DWR 1995). Other major crops in these areas are soybeans and wheat (Johnson 1999). In the upper region of the catchment, irrigation farming has been restricted to the alluvial flats and areas next to the Dumaresq River (DWR 1995).

Several land degradation issues related to agricultural practices and land use in the Border Rivers catchment have been identified by Peasley (1993) and McCosker (1996), and Chapter 2. Study area 29 include soil and stream bank erosion, dryland salinity, tree decline and deteriorating water quality. Wylie (1995) recognised four main water issues in the Border Rivers basin, which are, enhanced nutrient content of stream and river waters, run-off of pesticides from irrigated land, sediments suspended by erosion from stream banks and agricultural land earthworks, and reduced but more stable river flows resulting from diversion of water for irrigation (see Section 2.3.).

2.9 Fish community and aquatic invertebrate fauna

The Border Rivers system supports 15 species of native fish and up to six species of exotic fish (McCosker 1996, Moffatt and Voller 2002). Recreational fishing species such as golden perch (Macquaria ambigua), silver perch (Bidyanus bidyanus), Murray cod (Maccullochella peeli) and catfish (Tandanus tandanus) occur throughout the system (Swales and Curran 1995), although populations, except for catfish, have experienced significant decline in recent years (Faragher and Harris 1993, Mallen-Cooper 1993).

The spangled perch (Leiopotherapon unicolor) is relatively abundant in the Darling River (Pusey et al. 2004). Found mostly in the billabong and riverine plains zones, this species is known to inhabit potentially any body of water in these zones (Moffatt and Voller 2002). Purple-spotted gudgeon (Mogurnda adspersa), Darling River hardyhead (Craterocephalus amniculus), the latter only found in tributaries of the upper Darling River, and olive perchlet (Ambassis agassizii) occur sporadically in the catchment and were listed as rare by Wager and Jackson (1993), even though Allen (1996) stated that the olive perchlet is relatively common throughout much of its range. This author also points out that A. agassizii can have an important role in controlling mosquito populations and is an important food for larger predators.

Bony bream (Nematalosa erebi) is abundant throughout the Border Rivers and is known to form huge populations (Briggs and McDowall 1996). The bony bream is found in turbid waters and can tolerate a wide range of temperatures, salinities and other habitat conditions (Briggs and McDowall, 1996). It has been described as a key link in the food chain, being an important source of food for piscivores, including golden perch and Murray cod (Briggs and McDowall 1996, Bunn et al. 2003). Exotic species, such as mosquitofish (Gambusia Chapter 2. Study area 30 holbrooki) and European carp (Cyprinus carpio), also occur in the Border Rivers system and are thought to have had a significant impact on native species by competing for habitat and food (Lloyd et al. 1986, McDowall 1996) or altering the natural habitat (Roberts et al. 1995).

The macroinvertebrate community in the Border Rivers catchment is relatively high in diversity of taxa but moderate to low in number of individuals, with the orders Coleoptera, Diptera, Trichoptera, Hemiptera, Odonata and the phylum Mollusca accounting for the majority of known taxa in the catchment (Boddy and Bales 1996). The confluence of the Macintyre and Dumaresq Rivers marks a transition in the nature of the macroinvertebrate community, as upstream of the confluence, a large number of individuals belonging to many taxa are found, whereas a decrease in the number of taxa is apparent below the confluence (Boddy and Bales 1996). This may be attributed to reduced habitat diversity, there being no riffles and few macrophytes downstream of Goondiwindi (Boddy and Bales 1996). In autumn and winter, the communities are dominated by Ephemeroptera (mayflies), Hemiptera (true bugs) and Trichoptera (caddisflies), whereas a shift in taxa composition occurs during summer, when Crustacea (e.g. shrimps and yabbies) and Hemiptera predominate (Boddy and Bales 1996).

Chapter 3. Limnological characteristics of the study sites 31

3 Spatial and temporal variation in limnological characteristics of the study sites

3.1 Introduction

The Murray-Darling system presents a wide range of wetlands and many of these are formed by meanderings of lowland rivers, therefore associated with the rivers during flooding periods (Mitchell 1994). The importance of floodplain lagoons in these dynamic systems is well recognised (Hamilton et al. 1990, Sippel et al. 1992, Castillo 2000). Because such waterbodies are strongly influenced by periodic fluctuations in water level, they function as both lentic systems, during the dry season, when they are relatively isolated from the main river channel, and as lotic environments, during the wet season, when their interaction with the river is at its maximum (Rai and Hill 1984).

Many ecological issues regarding floodplain inundation in large tropical rivers have been identified (Junk et al. 1989) and among these is the potential for higher productivity on the floodplain in comparison to the main river channel (Junk et al. 1989, Bayley 1995, Tockner et al. 2000). In Australian rivers, floodplain lagoons are regarded as highly productive systems, supporting a large and diverse biomass of microorganisms and invertebrates (Bunn and Boon 1993, Butcher 1997, Hillman 1998) that are hypothesised to represent a significant food resource for fish (Geddes and Puckridge 1989, Hillman 1995b). It is believed that a positive relationship exists between floodplain productivity and water flow (Gosselink and Turner 1978), the latter boosting production and allowing exchange of nutrients and biota between river and floodplain (Boon et al. 1990). Many studies have shown that the flood cycle in floodplain lagoons is followed by fluctuations in physical, chemical and biological characteristics of these water bodies, such as depth, turbidity, dissolved oxygen levels, nutrient and organic carbon concentrations and biological production (Rai and Hill 1984, Hamilton and Lewis 1987, Junk and Weber 1995).

Resulting from variation in flooding regime, biological communities within such environments vary geographically and seasonally (Hillman 1986, Nielsen et al. 2002), as they have to survive changes in chemical and physical characteristics of the environment, Chapter 3. Limnological characteristics of the study sites 32 such as extremes of temperature, fluctuations in salinity and oxygen levels and, in some instances, complete drying (Williams 1985, Williams and Allen 1987). With the decrease in river flow after the wet season, the aquatic floodplain environment suffers a substantial contraction in size as water levels recede and lagoons become isolated and start to dry up due to evaporation (Rodriguez and Lewis 1994). The aquatic biota restricted to these habitats can experience substantial stress, as they are subjected to variable and/or harsh physico-chemical conditions resulting from deterioration in water quality with the progression of the dry season. For example, contraction in floodplain lagoon size may cause an increase in solute concentrations (Hart and McGregor 1982). Variations in water level and reduction in water quality can affect the germination, growth and survival of aquatic vegetation (Bruton 1985, Gopal 1986), as well as the seasonal composition and abundance of phytoplankton (Osborne et al. 1987), bacteria (Boon 1991), microcrustaceans (Shiel et al. 1982), macroinvertebrates (Marchant 1982, Boddy and Bales 1996) and fish (Kennard 1995). Therefore, because the hydrological characteristics of these environments are so strongly influenced by spatial end temporal patterns of water flow, quite proximal floodplain lagoons can have very different physical, chemical and biological characteristics based on their morphology and position on the floodplain (Walker and Thoms 1993).

Even though some of the floodplain lagoons of the Macintyre River have been identified in various reports (e.g. DBBRC 1988, DWR 1990), there is little information regarding their current condition and overall morphological and limnological characteristics.

3.1.1 Aims

It is essential to seek some understanding of the limnology of the floodplain lagoons to be studied, in order that the flow regime and limnological characteristics existing in these sites at the time of study are taken into consideration. The variety of sites chosen and the fact that some of them are subject to unnatural flow manipulations makes such a study particularly important. This chapter provides a broad description of the study sites and characterises their general limnological attributes over a summer-winter period, including dry and wet seasons. Specific objectives are:

Chapter 3. Limnological characteristics of the study sites 33

1. to describe the flow regime of the Macintyre River with respect to seasonal and short-term inter-annual variation in the magnitude of water flows, as the flow regime directly influences the morphological, water quality and habitat characteristics of the nearby floodplain lagoons (see next), 2. to understand spatial and temporal variation in some of the morphological and physico-chemical characteristics of the lagoons studied, 3. to describe the habitat characteristics within the study lagoons and nearby reaches of the Macintyre River.

3.2 Methods

3.2.1 Hydrology

Water flow data were supplied by the Department of Infrastructure, Planning and Natural Resources - NSW Water Information and Queensland Department of Natural Resources, Mines and Energy for two gauging stations on the Macintyre River (Goondiwindi Town and Boggabilla gauging stations - 416201A and 416002, respectively). Both stations are between 5 to 30 km from the study sites (Figure 2.2).

3.2.2 Lagoon morphology

A range of floodplain lagoons was selected to represent the widest possible range of waterbody types available in the study area (see Chapter 2), from small relatively undisturbed to large and flow-regulated floodplain lagoons. Six lagoon sites on the Macintyre River floodplain and one site in the main channel of the river were sampled on three occasions: summer, prior to and after flooding and winter, several months after flooding. The first samplings were undertaken late in the dry season of 2002 (20-31 October), the second sampling occasion occurred near the end of the 2002-2003 summer, soon after the wet season (10-20 March 2003), and the third sampling occasion occurred in the winter of 2003 (15-25 July), during the dry season.

Chapter 3. Limnological characteristics of the study sites 34

On all sampling occasions, lagoon length and width were measured by tape for distances of up to 50 m in smaller lagoons and by GPS satellite receiver (Magellan Platinum) for distances greater than 50 m in the larger lagoons. Two or three transects were established in each lagoon (depending on lagoon length and accessibility) to measure their width. Eleven water column depths were taken at approximately equivalent distances along each transect. These measures were used to estimate lagoon surface area and volume. The elevation, overall shape and distance to the river for each site were estimated from a combination of GPS receiver readings, aerial photos and 1:100000 topographic maps. It is important to note that, given the resolution of these methods, the measurements were made mainly to compare general features of lagoon morphology and should not be taken as absolute measures.

3.2.3 Habitat characteristics

The following habitat variables were quantified: presence of macrophytes, submerged grass, floating vegetation, emergent vegetation, submerged vegetation (these were later collapsed to one single element - aquatic vegetation, during the statistical analysis), filamentous and attached algae, overhanging vegetation, leaf litter, root masses, woody debris, mud, sand, fine and coarse gravel, rocks and cobbles (these substrate composition elements were also collapsed into mud, sand and gravel for the statistical analysis). These habitat elements were estimated by eye and represent their proportional contribution to lagoon wetted perimeter, since most habitat elements were concentrated mostly along the shallow lagoon margins and visual estimation was impracticable in deeper and turbid waters. Measurements were taken from a 10 x 1 m transect parallel to the lagoon shore. This sampling procedure was performed on the bank adjacent to the transects, and data were expressed as the mean proportional (%) cover of the habitat variables for each lagoon on each sampling occasion.

3.2.4 Physico-chemical parameters

The following water quality parameters were taken: water temperature, pH, conductivity, turbidity, salinity (Horiba U-10 water checker) and dissolved oxygen (TPS Model WP-82). Chapter 3. Limnological characteristics of the study sites 35

These measurements were taken at the lagoon margins (depths up to 1 m) at the water surface. Replicates were made in a similar fashion to the habitat characteristics, at the ends of each transect and when possible at the extremities of the lagoons. Data were generally collected between 9:30 am and 12:00 pm.

3.2.5 Data analysis

Variations in lagoon morphology and water quality characteristics were summarised in two different sets of analyses using Principal Component Analysis (PCA). This analysis used the correlation coefficients of the cross-products matrix, with the variables scores calculated by weighted averaging (McCune and Mefford 1999). After testing for univariate skewness, the variables were transformed as appropriate (see Sokal and Rohlf 1969). Volume and depth variables were fourth root and log(x+1) transformed, respectively, as they were skewed to the right, and the area and length were not transformed as they were not significantly skewed. With regard to the water quality variables, conductivity and salinity were fourth root transformed as they had positive skewness, and the remaining water quality measures were not transformed because their skewness was not significantly different from zero. According to McCune and Grace (2002), PCA is an ideal technique for summarising data with approximately linear correlations among variables, as is the case for lagoon morphological characteristics and water quality parameters. Further investigation of variations in lagoon morphology and water quality parameters was performed by two-way ANOVA of log(x+1) transformed data.

Variation in lagoon habitat characteristics between sites and sampling occasions was investigated using Detrended Correspondence Analysis (DCA) of the untransformed data (McCune and Mefford 1999). The coefficients of determination for the correlations between the DCA axis scores and the distances of the habitat variables in the original n- dimensional space were also used to assist the interpretation of the DCA results. Further investigation of specific habitat variables was performed by two-way ANOVA of arcsine- transformed data.

Chapter 3. Limnological characteristics of the study sites 36

3.3 Results

3.3.1 General hydrological patterns

The average annual discharges in the Macintyre River show wide variation from year to year since 1995 (Figure 3.1), with approximately alternating wet and dry years. The 2002- 2003 period was relatively dry compared to previous years, with average discharges of about 780 to 845 ML/day.

8000

7000

6000

5000

4000

3000

2000

Average annual discharge (ML/day) discharge annual Average 1000

0 1995 1996 1997 1998 1999 2000 2001 2002 2003 Year

Figure 3.1 Mean annual discharges (ML/day) for the Macintyre River recorded at the Boggabilla gauging station (416002) for the years 1995 to 2003.

Major to moderate floods occurred early and late in 2001, however only minor to moderate floods were recorded for the study period (2002-2003) with mean discharges of 14487 ML/day on 31 March 2002 and 17412 ML/day on 26 February 2003 (Figure 3.2). Mean water levels rose up to 4 m during the 2002 flood and 4.6 m during the 2003 flood (data from the Boggabilla gauging station - 416002). Even though such floods may cause inundation of low lying areas adjacent to the main river channel, they were not sufficient to inundate all study sites, as shown by the water level changes measured in each of the lagoons studied. A detailed summary of hydrological characteristics of the study sites is provided in Table 3.1 and in Section 3.3.2.

Chapter 3. Limnological characteristics of the study sites 37

90000 2001 2002 2003 80000

70000

60000

30 Nov 01 50000

40000

Discharge (ML/day) 30000

20000 31 Mar 02 26 Feb 03

10000 late dry wet dry

0 Jul Jul Jul Jun Oct Jun Oct Jun Oct Sep Feb Sep Feb Sep Feb Dec Apr Dec Apr Dec Apr Mar Mar Mar Nov Nov Nov Aug Aug Aug May May May Jan/03 Jan/02 Jan/01 Month

3.3.2 General description of the study sites

The morphological characteristics of the lagoons studied were diverse. On average, their surface area ranged from 83155 to 767380 m2. Average lagoon length varied between 643 to 5137 m, whereas maximum water depth ranged from 36 cm to 6 m. Apart from the Macintyre River site, the lagoons studied can be classified into four groups, based on their size, shape and relation to the river: the first group corresponds to the small and relatively long South Callandoon Lagoon A (and B, see below), which were located within 100 m of the main channel of the Macintyre River. The second group corresponds to the large and meandering lagoons, including Rainbow and Serpentine lagoons, which lay within 1000 m of the Macintyre River. Punbougal and Maynes lagoons comprise the third group. These are large, long lagoons lying more than 4000 m from the Macintyre River. The fourth group includes solely the Broomfield Lagoon, which is a relatively short but wide lagoon, located at a moderate distance from the river (approximately 3000 m) (refer to Figure 2.2 for site locations and Appendix 1 for photos of the study sites).

Chapter 3. Limnological characteristics of the study sites 38

Overall hydrological Flow history Lagoon Wate r regime Management issues status * (2000-2003) South Callandoon Natural Temporary (dries out after Flooded in November 2001 May receive groundwater Lg. A a few months of no flow) and February 2003 from nearby ring tank. Provides drinking water for cattle South Callandoon Natural Temporary (dries out after Flooded in November Provides drinking water Lg. B several months of no flow) 2001. Dried out before for cattle August 2002

Rainbow Lg. Regulated Semi-permanent (has not Flooded in November 2001 Pumped into and out for (permanently elevated fully dried out in the past and February 2003. May cotton irrigation. water level) 20-30 years) have received water in TSR for cattle March 2002 **

Serpentine Regulated Semi-permanent (was dry Filled out after flooding of Pumped into from (artificially filled) in 2000). Probably dries the Macintyre River late Macintyre River. out after several years of 2000. Not flooded since Recreational activities. no flow then Turned into a water park in 2004 Punbougal Lg. Natural Semi-permanent (has not Not flooded since at least Used for recreational dried out in several years) November 2001 activities

Maynes Lg. Natural ** Semi-permanent (has not Not flooded since at least TSR for cattle. Provides dried out in several years) November 2001 drinking water for sheep and horses Broomfield Lg. Natural ** Semi-permanent (dries out Probably flooded in Provides drinking water Artificial lake (?) after a few years of no November 2001. Dried out for cattle flow) before March 2003

Macintyre River Regulated Permanent/Semi- Major flooding in early Pumped out for permanent months, and November, of irrigational purposes 2001 and minor to moderate floods in March 2002 and February 2003

Although only some morphological and physico-chemical data were collected for South Callandoon Lagoon B, both the collected data and field observations indicated that South Callandoon lagoons A and B were very similar in their morphological characteristics. These were small anabranch billabongs (mean area =83155 m2 and mean length =643 m) positioned very close to the Macintyre River (about 100 m), at an elevation of about 203 m, and separated from the main channel by a low natural levee. At both sites, a shallow channel can be observed crossing the river bank and leading to the lagoons. During the wet season these lagoons are likely to be filled by runoff from their own catchment first and subsequently, water from the river may enter the lagoon by flowing over the natural levee, through the anabranch channels. The South Callandoon lagoons were consistently muddy and shallow, with a maximum depth of 1.22 m, elongated and channel-shaped water bodies (L:W ratio =31.8), with gently sloping banks on the river side of the lagoons and steep banks on the opposite side. Although the surroundings of these lagoons are subject to grazing by cattle, there is still a considerable amount of overhanging Eucalyptus Chapter 3. Limnological characteristics of the study sites 39 and Melaleuca (45.8% riparian cover). Partially exposed root masses and submerged logs and branches were also common, but aquatic vegetation was abundant only when the lagoons were fullest. When present, aquatic vegetation consisted mostly of the floating species, water primrose (Ludwigia spp.) and red azolla (Azolla spp.) (Sainty 1973, Stephens and Dowling 2002), which were found over the shallower and marginal sections of the lagoon. Filamentous and attached (to small and large debris) algae were also present at these sites and consisted mostly of Rhizoclonium spp. and Cladophora spp. (Prescott 1970, Sainty 1973, Entwisle et al. 1997). These were the only species of aquatic plants and algae recorded for the other study sites as well. The fish fauna in both South Callandoon lagoons was amongst the most diverse when compared to the other study sites, with 6 to 9 different species recorded (Table 3.2). Although capable of retaining water for several months after flooding, these lagoons are temporary, and dry up if not flooded by water from the Macintyre River. During the study period, both lagoons were flooded in the summer (late November) of 2001, but even though South Callandoon Lagoon A retained water throughout the whole study period, South Callandoon Lagoon B dried up during the winter (August) of 2002. It is possible that South Callandoon Lagoon A receives groundwater inflow from a nearby ring tank but the extent of this supply is unknown. Although the hydrograph for the Macintyre River shows a sharp rise in water flow in late March 2002 (Figure 3.2), observations on the study sites in May 2002 indicate that this event did not cause flooding of the South Callandoon lagoons.

Rainbow and Serpentine lagoons are large (approx. area ranging from 278437 to 431750 m2), long (approx. length ranging from 2950 to 5300 m) and meandering (L:W ratio from 38.8 to 186.9) lagoons, located about 500 m from the main channel of the Macintyre River (see Figure 2.2). These are deeper lagoons, with recorded maximum depths of 2.7 to 2.8 m (respectively), situated at an elevation of about 232 m. Bank slopes varied from steep on the river side to gentle on the banks opposite to the river for Rainbow Lagoon, and mostly gentle to slightly steeper in Serpentine Lagoon. Filamentous and attached algae were present at both sites, though in greater amounts at Rainbow Lagoon. Although subject to flooding in a similar fashion to the South Callandoon lagoons, both of these lagoons are fairly permanent and have their natural water level regulated for irrigation and recreational purposes. Rainbow Lagoon is pumped into from the river and out for cotton irrigation and retained water during all of the study period. As this site is also a traveling stock reserve, cattle are frequently brought to its shores to drink and people also camp along its Chapter 3. Limnological characteristics of the study sites 40 foreshores (Appendix 1, Photos 9 and 10). This lagoon had a muddy substrate, although sand could be observed in some areas. Aquatic vegetation was not abundant and was present mostly as submerged vegetation. The riparian overhanging vegetation was very abundant with an average of 71.6% of cover, as well as partially exposed root masses and submerged logs and branches. The fish fauna was one of the most diverse of the studied sites with 9 species recorded (Table 3.2).

Table 3.2 List of species recorded in each of the study sites on the floodplain of the Macintyre River, during the study period. * indicates exotic species.

South South Macintyre Species Callandoon Rainbow Serpentine Callandoon Punbougal Maynes Broomfield River A B Clupeidae Nematalosa erebi (Günther 1868). 3 3 3 3 3 3 3 3 Bony Bream Terapontidae Leiopotherapon unicolor (Günther 3 3 3 3 3 3 1859). Spangled perch Retropinnidae Retropinna semoni (Weber 1866). 3 3 3 3 3 3 Australian smelt Eleotrididae

Hypseleotris spp. Carp gudgeon 3 3 3 3 3 3 Ambassidae Ambassis agassizii Steindachner 1867. 3 3 3 3 3 Olive perchlet Melanotaeniidae Melanotaenia fluviatilis (Castelnau 3 3 3 1878). Crimson-spotted rainbowfish Percichthyidae Macquaria ambigua (Richardson 3 3 3 1845). Golden perch or yellowbelly Plotosidae Tandanus tandanus Mitchell 1828. 3 Ell-tailed catfish * Cyprinidae Cyprinus carpio Linnaeus 1758. 3 3 3 3 3 3 3 3 Carp * Cyprinidae Carassius auratus Linnaeus 1758. 3 Goldfish * Poecilidae Gambusia holbrooki Girard 1859. 3 Mosquito fish Total 9 9 8 7 6 3 3 3

Serpentine Lagoon is also pumped into and presented relatively high water levels throughout the whole study period. Local information indicates that this lagoon was dry in 2000 and was artificially filled up later in that year. There is no indication that the water levels were artificially changed throughout the study period, but in the summer of 2004, after this study was completed, Serpentine Lagoon was artificially filled again and turned into a water park for recreational activities. This lagoon had a muddy bottom with very little submerged debris and no exposed logs or root masses. Overhanging riparian Chapter 3. Limnological characteristics of the study sites 41 vegetation was almost absent (only 4.6% cover) and aquatic vegetation was relatively scarce, being present mostly as floating species, i.e. Azolla spp. and water primrose (Ludwigia spp.). The fish fauna was relatively diverse when compared to the other study sites, with 7 species recorded (Table 3.2).

Punbougal and Maynes lagoons are large (approx. area ranging from 242763 to 463125 m2) and long lagoons (approx. length from 1440 to 3360 m), lying relatively distant from the Macintyre River, at a distance greater than 4000 m from the main channel, and an elevation of approximately 227 m. These are less meandering lagoons than Rainbow and Serpentine, with gentle bank slopes. Although both lagoons are fairly deep, Punbougal Lagoon was considerably deeper than any other study site with a maximum depth of 6 m, whereas Maynes Lagoon was only up to 2 m in depth. There was no indication of water management practices affecting these lagoons, even though Punbougal Lagoon is used for fishing and water skiing, and both lagoons are permanent, not known to have dried up in many years. Punbougal and Maynes lagoons are located on the floodplain of the Macintyre River and are subject to flooding both from the river itself and from Morella Watercourse (Figure 2.2). The western end of Maynes Lagoon is a traveling stock reserve used by cattle as a drinking site. The remaining surroundings of this lagoon are also used by horses and sheep for drinking water. The substrate of Maynes Lagoon was very muddy with hardly any debris and little aquatic vegetation. Some filamentous and attached algae were present but in very small quantities. Overhanging riparian vegetation, although present in the area, was not recorded near the shores of this lagoon. The fish fauna was one of the poorest, with only 3 species recorded (Table 3.2). Punbougal Lagoon presented a sandy to muddy substrate, with gravel, cobbles and small rocks also recorded. Given its depth and relative distance from riparian vegetation, very little debris was observed, and overhanging riparian vegetation was scarce (3.3% cover). Aquatic vegetation was present as the submerged water milfoil (Myriophyllum spp.) and the floating water primrose (Ludwigia spp.) and red azolla (Azolla spp.) (Sainty 1973, Stephens and Dowling 2002). Filamentous algae, mostly Rhizoclonium spp. and Cladophora spp. (Prescott 1970, Sainty 1973, Entwisle et al. 1997), were relatively abundant in shallower areas. The fish fauna was very poor, with only 3 species recorded (Table 3.2).

Broomfield Lagoon was the largest of the study sites, with an approximate area of 767380 m2, located at an approximate distance of 3000 m from the river, at an elevation of 223 m. Chapter 3. Limnological characteristics of the study sites 42

This is a relatively short (490 m), wide (L:W ratio =1.1; mean width =440 m) and shallow lagoon (maximum depth =37 cm), with very gently sloping banks. Broomfield Lagoon is located on the floodplain of the Macintyre River and is flooded via the adjacent Dingo and Callandoon Creeks (Figure 2.2). Local information indicated that this is an artificial lake created by the damming of a channel off Callandoon Creek, which runs parallel to the Macintyre River. Even though the water flow into this lagoon can be artificially controlled, there was no evidence of water management during the study period, as the lagoon slowly dried up throughout 2000-2002. Broomfield Lagoon presented a consistently muddy substrate and filamentous algae were relatively abundant. No aquatic vegetation was recorded during the study period, although preliminary field trips during 2001 indicated the presence of macrophytes. Overhanging riparian vegetation, although present in the area, was not recorded near the shores of this lagoon and only small amounts of debris were observed. The fish fauna was one of the poorest, with only 3 species recorded (Table 3.2).

Finally, the river site constituted a stretch of about 200 m length in the main channel of the Macintyre River, at an elevation of about 222 m (Figure 2.2). The maximum depth recorded was 1.37 m and the mean width was about 24.1 m. The Macintyre River banks were relatively steep at this site, and the substrate was mostly muddy with some sand. Overhanging vegetation was present at an approximate proportion of 26.6% of cover. The studied stretch of the river contained very few submerged root masses but small and large debris were present in relatively large amounts. Aquatic vegetation was very scarce, comprised mostly of submerged vegetation and filamentous algae, which were relatively abundant in the shallow slowly flowing margins. The fish fauna in the river site was relatively diverse, with a total of 8 species recorded (Table 3.2).

3.3.3 Temporal variation in lagoon morphometry

During the late dry season (October-November 2002) South Callandoon Lagoon A was generally very shallow with a maximum depth of 40 cm and a mean depth of 27 cm. Even though the overall shape of this lagoon was similar throughout the study period, South Callandoon Lagoon A was shorter and narrower during the late dry, with a length of 520 m and a mean width of 17.7 m. Soon after the wet season (March 2003), this lagoon had Chapter 3. Limnological characteristics of the study sites 43 increased in size and depth, to a length of 750 m, a mean width of 22.6 m and an average depth of 89 cm (the maximum depth was 1.22 m) (Table 3.3). Rainbow Lagoon was also deeper (mean depth =1.56 m; maximum depth =2.82 m) and larger (length =3028 m and mean width =78 m) during the wet season than during the preceding dry season. By the next dry season (July 2003), receding water levels due to evaporation, and possibly water extraction (in Rainbow Lagoon’s case), caused both lagoons to contract in size once again, resulting in an increased representation of shallow waters, and consequently lower mean and maximum depths (68 and 89 cm for South Callandoon, and 1.24 and 1.89 m for Rainbow Lagoon). Overall size was also reduced in both lagoons. Even though the Macintyre River was not sampled during the flooding period, data collected during the wet season, after the flood, show that the river was deeper during this season (mean depth =90 cm and maximum depth =1.37 m), but the overall width was similar on all occasions (Table 3.3).

In contrast, the remaining lagoons had their water levels steadily reduced throughout the study period, due to evaporation and because flood waters from the Macintyre River were unable to reach these sites during the wet season. The most extreme cases were Broomfield Lagoon, which dried up completely before the wet season (2003) and has not flooded since, and South Callandoon B, which dried up before this study commenced, but was wet when sampled in May 2002. Even though Broomfield Lagoon can be artificially filled from the nearby creeks, it is likely that the rainfall during the wet season and the water flow in the Macintyre River were not sufficient to fill this lagoon. Similarly, Serpentine Lagoon had its depths reduced from 1.13 m (average depth) and 2.7 m (maximum depth) in the beginning of the study period (late dry season of 2002), to 87 cm (average depth) and 1.74 m (maximum depth) by the end of the study period (dry season of 2003) (Table 3.3). Overall lagoon size was also reduced from approximately 5300 m in length and 39 m of mean width to 5000 m length and a mean width of 26 m during the dry season of 2003 (Table 3.3).

Maynes and Punbougal lagoons had their overall sizes and depths steadily reduced from the late dry season in 2002 to the dry season of 2003. For instance, Punbougal Lagoon had its maximum depth reduced from 6 to 5.1 m and overall width decreased from 137 m on average to 121 m (Table 3.3). By the dry season of 2003, the surface area of Punbougal Lagoon had decreased considerably and consequently, the water receded to the deeper Chapter 3. Limnological characteristics of the study sites 44 channel, which resulted in the increased mean depths recorded for this waterhole. Likewise, Maynes lagoon had its overall surface area reduced from approximately 430000 to 242000 m2 by the dry season of 2003. The overall lagoon size was reduced from 2170 m in length and 113 m of mean width, during the late dry season of 2002, to 1440 m length and a mean width of 68 m during the dry season of 2003 (Table 3.3).

Distance Length: Area Volume Length Mean Elevation Maximum Mean Lagoon Season to the Width Slope (m2) (m3) (m) width (m) (m) depth (m) depth (m) river (m) ratio South Late dry 80850 21830 520 17.7 100 29 205 0.40 0.27 2.8 Callandoon A Wet 86470 77247 750 22.7 33 207 1.22 0.89 3.0

Dry 82147 55860 660 20.0 33 198 0.89 0.68 3.3

Rainbow Late dry 291104 436656 3000 64.5 500 47 225 2.40 1.50 3.0

Wet 431750 673530 3028 78.0 39 239 2.82 1.56 3.3

Dry 340166 423507 2950 62.3 47 236 1.89 1.25 3.5

River Late dry n/a n/a n/a 27.3 n/a n/a 225 0.81 0.58 1.5

Wet n/a n/a n/a 25.0 n/a 223 1.37 0.90 2.5

Dry n/a n/a n/a 25.0 n/a 218 0.86 0.61 2.0

Punbougal Late dry 463125 896919 3460 137.5 11500 25 226 6.00 1.94 3.3

Wet 415625 879740 3400 125.3 27 223 5.20 2.12 3.6

Dry 403750 1030908 3360 121.2 28 225 5.10 2.55 3.9

Maynes Late dry 431525 371112 2170 113.3 4000 19 232 2.00 0.86 4.0

Wet 296710 186927 1590 81.3 20 232 1.35 0.63 4.0

Dry 242763 135947 1440 68.5 21 225 1.15 0.56 4.0

Serpentine Late dry 278437 314634 5300 39.3 500 135 233 2.70 1.13 3.2

Wet 176343 149010 5112 29.0 176 237 1.85 0.85 3.3

Dry 157781 137269 5000 26.8 187 226 1.74 0.87 3.2

Broomfield Late dry 767380 188008 490 440.0 3000 1 223 0.37 0.25 4.0

Principal Component Analysis of the lagoon morphometry and depth characteristics during the three sampling occasions demonstrated some spatial and temporal differences between the study sites (Figure 3.3). The overall variation in sampling sites can be summarized on a two-dimensional plot explaining 88% of the total variance, indicating a high level of correlation (or collinearity) between the measured variables. A further third axis, giving a cumulative total of 97% of the variance explained, is also interpretable (Table 3.4), but the Chapter 3. Limnological characteristics of the study sites 45 rest of the variance is noise, as indicated by the broken-stick eigenvalue. The broken-stick model is a guide for determining the number of significant axes to be interpreted in the PCA. If the broken-stick eigenvalue for a given axis is less than the actual eigenvalue for that axis, then that axis contains more information than expected by chance, and should be considered for interpretation (McCune and Grace 2002).

% of Cum.% of Broken-stick Axis Eigenvalue Variance Var. Eigenvalue 1 3.827 63.781 63.781 2.450 2 1.438 23.960 87.741 1.450 3 0.538 8.974 96.715 0.950 4 0.148 2.472 99.187 0.617 5 0.034 0.566 99.753 0.367 6 0.015 0.247 100.000 0.167

Characteristic First 6 Eigenvectors 1 2 3 4 5 6 Area -0.3879 -0.5328 -0.0980 0.0647 0.3363 0.6624 Volume -0.4758 -0.062 -0.2437 -0.7976 -0.1998 -0.1852 Length -0.3636 0.3526 -0.7326 0.4154 0.1212 -0.1398 Mean width -0.3587 -0.5649 0.2023 0.3743 -0.1686 -0.5855 Maximum depth -0.4422 0.3450 0.3296 0.2091 -0.6310 0.3673 Mean depth -0.4083 0.3871 0.4948 -0.0573 0.6369 -0.1723

As observed in Table 3.4, the first principal component (64% of the total variation) represents a negative linear function of the correlated depth (mean and maximum depths) and volume measures, reflecting the overall volumetric changes associated with each lagoon throughout the study period. From the plots of sites scores observed in Figure 3.3, deeper lagoons such as Punbougal and Rainbow, which also had greater water volume, were aligned on the left of the first factor, while shallower lagoons were arrayed to the right. In general, sites subject to flooding during the wet season (South Callandoon, Rainbow and the river) had their wet season samples located on the left of this axis when compared to samples from the dry seasons. On the other hand, sites that dried continuously throughout the study period had the first dry season (late dry) located on the left of the first axis, when compared to samples from other seasons for same lagoons (Figure 3.3). Chapter 3. Limnological characteristics of the study sites 46

Area late Width South Callandoon Rainbow Macintyre River Punbougal Maynes Serpentine Broomfield

xis 2 A

late wet dry

wet dry late late wet late dry late (a) dry wet late wet dry dry wet

Axis 1 Length wet dry South Callandoon Rainbow late Macintyre River Punbougal Maynes Serpentine Broomfield

late

dry dry xis 3 wet A late late wet late dry

wet

late wet late dry dry (b) wet

Axis 1 Depth Vol ume Figure 3.3 Position of each study site and sampling occasion within the ordination space defined by the first three factors identified by PCA of the cross-products matrix of the correlation coefficients for lagoon morphometry. (a) axis 1 and 2, and (b) axis 1 and 3. Site codes correspond to the following: late=late dry season, 2002 summer; wet=wet season, 2003 summer; dry=dry season, 2003 winter. Chapter 3. Limnological characteristics of the study sites 47

The second principal component (24% of the variation) explains the variables more related to lagoon size and is negatively correlated to the area and width measures. In this regard, Broomfield Lagoon clearly stands out from the other sites studied (Figure 3.3a), as this lagoon was the larger in area and width. Finally, the third principal component, which explains only 9% of the total observed variation, is also a function of lagoon size, reflecting a positive correlation with lagoon length (Figure 3.3b). Consequently, longer lagoons such as Serpentine and Rainbow lagoons are aligned upwards on this factor. Note that, although long, Punbougal Lagoon is not located upwards on the third factor. This was caused by other variables, mostly from factor 1, having a stronger effect on the position of this lagoon in ordination space.

3.3.4 Lagoon microhabitat characteristics

Ordination by DCA of the habitat characteristics of the study sites for each sampling occasion also demonstrated spatial and temporal variation between lagoons (Figure 3.4). The first three axes accounted for 94% of the total variation, with 82% of the variation explained by the first axis alone. The correlations between DCA scores of the first three axes and the habitat characteristics are shown in Table 3.5.

Table 3.5 Coefficients of determination showing the correlations between lagoon habitat characteristics and the first three axes of the ordination space determined by DCA.

Axis 1 2 3

Aquatic vegetation 0.682 0.397 -0.704 Algae 0.218 -0.076 0.550 Overhanging vegetation -0.350 0.760 -0.042 Leaf litter -0.236 0.547 -0.080 Root masses -0.215 0.650 0.023 Debris -0.332 0.705 -0.038 Mud -0.963 -0.383 -0.066 Sand 0.961 0.385 0.068 Gravel 0.931 0.277 -0.002

The first axis was positively related to sand and gravel and negatively correlated with mud, representing the substrate composition variables. Consequently, the lagoons with higher proportion of sand and gravel will be arrayed to the right of the axis, as was the case for Chapter 3. Limnological characteristics of the study sites 48

Punbougal Lagoon. The remaining sites, which presented a predominantly muddy substrate, were located on the left of this axis. The second axis was positively correlated with riparian vegetation characteristics such as overhanging vegetation, leaf litter, root masses and woody debris (Figure 3.4a). As a result, sites where the riparian vegetation was distant from the lagoon edges, and consequently had smaller representation of root masses, debris and leaf litter, were arrayed to the bottom of the second axis. The third axis showed a negative correlation with aquatic vegetation and a positive correlation with algae. These two variables were also positively correlated with the first axis.

Therefore, the sites with a higher proportion of aquatic vegetation were arrayed to the right of the first axis and downwards on axis 3, as was the case for Punbougal Lagoon, and Rainbow and South Callandoon lagoons, during the wet and second dry seasons (respectively), whereas sites with smaller proportions of aquatic vegetation were aligned to the top left of the ordination space. On the other hand, sites with higher proportions of algae were arrayed on the right of the first axis and upwards on the third axis, as was the case for Punbougal and Broomfield lagoons (Figure 3.4b).

Spatial and temporal variation in the mean proportions of habitat characteristics of particular interest was further investigated using two-way ANOVA. The differences in mean proportions of each habitat variable between lagoons and sampling occasions are shown in Table 3.6. The lagoons differed significantly in their mean proportion of aquatic vegetation (F=11.5; df=4,65; p<0.0001), algae (F=12.5; df=6,85; p<0.0001), leaf litter (F=4.7; df=6,85; p=0.0004) and woody debris (F=17.2; df=6,85; p<0.0001). On the other hand, the observed variation in proportion of aquatic vegetation between sampling occasions was significant only in the Serpentine (F=8.6; df=2,85; p=0.0005) and South Callandoon (F=19.7; df=2,85; p<0.0001) lagoons. The South Callandoon Lagoon also had a significant difference in mean proportion of leaf litter (F=5.1; df=2,85; p=0.0083) between the sampling occasions, whereas Rainbow Lagoon had a significant variation in the mean proportion of both algae (F=3.6; df=2,85; p=0.0321) and leaf litter (F=8.1; df=2,85; p=0.0006). The mean proportion of woody debris between sampling occasions did not differ significantly for any of the study sites (F=0.2; df=2,85; p=0.8780).

Chapter 3. Limnological characteristics of the study sites 49

Overh. V. wet Leaf L. South Callandoon Root M. late Rainbow dry Macintyre River Debris dry Punbougal Maynes wet late Serpentine Broomfield wet wet dry

late late

2 xis xis A

late dry

dry (a) wet late wet dry late

Axis 1

late Algae South Callandoon Rainbow Macintyre River Punbougal dry dry Maynes Serpentine dry Broomfield late late wet late dry late late wet

3 late late

xis xis wet A wet late

Aq. Veg. dry (b)

dry

Axis 1

Mud Sand Aq. Veg. Gravel Algae Figure 3.4 Distribution of the study sites within ordination space as defined by DCA of untransformed lagoon habitat characteristics. (a) axis 1 and 2, and (b) axis 1 and 3. Site codes correspond to the following: late=late dry season, 2002 summer; wet=wet season, 2003 summer; dry=dry season, 2003 winter.

Chapter 3. Limnological characteristics of the study sites 50

Aquatic Overhanging Root Woody Lagoon Season Algae Leaf litter Mud Sand Gravel Vegetation vegetation masses debris South Late dry 0.0 0.1 31.3 3.3 1.9 13.6 100.0 0.0 0.0 Callandoon A (±0.2) (±25.3) (±3.1) (±3.0) (±12.7) (±0) Wet 12.2 1.0 51.3 10.8 0.6 11.7 100.0 0.0 0.0 (±6.6) (±1.5) (±29.0) (±14.2) (±0.9) (±6.6) (±0) Dry 38.8 0.8 55.0 4.8 2.2 12.0 100.0 0.0 0.0 (±13.7) (±1.5) (±34.6) (±3.6) (±3.2) (±6.1) (±0) Rainbow Late dry 0.0 0.5 82.5 3.0 2.4 9.2 100.0 0.0 0.0 (±0.9) (±35.0) (±2.7) (±2.8) (±6.5) (±0) Wet 4.9 6.3 92.5 15.2 3.4 14.8 100.0 0.0 0.0 (±5.4) (±3.3) (±9.6) (±11.1) (±4.5) (±6.4) (±0) Dry 0.0 1.7 40.0 2.4 3.3 14.1 58.8 41.3 0.0 (±1.5) (±46.2) (±1.2) (±6.2) (±9.7) (±48.0) (±48.0) River Late dry 0.0 0.6 30.0 1.9 0.1 5.1 91.8 8.3 0.0 (±0.1) (±42.4) (±2.1) (±0.1) (±5.5) (±2.5) (±2.5) Wet 0.1 0.7 50.0 2.0 0.0 7.7 92.5 7.5 0.0 (±0.1) (±0.5) (±70.7) (±0.4) (±1.4) (±3.5) (±3.5) Dry 0.0 6.3 0.0 3.0 0.0 3.4 92.5 7.5 0.0 (±5.2) (±2.8) (±4.1) (±3.5) (±3.5) Punbougal Late dry 30.0 1.3 10.0 0.8 0.0 1.2 31.4 67.1 1.4 (±26.4) (±2.8) (±26.5) (±0.5) (±2.3) (±47.1) (±46.1) (±3.8) Wet 31.3 4.1 0.0 1.7 0.0 0.5 28.6 70.1 1.4 (±9.6) (±4.3) (±2.0) (±0.8) (±45.6) (±44.8) (±3.6) Dry 20.7 4.9 0.0 1.0 0.0 1.1 29.3 68.4 2.3 (±6.1) (±4.4) (±0.6) (±1.2) (±48.3) (±47.0) (±5.6) Maynes Late dry 0.0 0.4 0.0 0.5 0.0 1.4 99.2 0.8 0.0 (±0.6) (±0.4) (±1.1) (±2.0) (±2.0) Wet 0.0 0.0 0.0 0.0 0.0 0.0 100.0 0.0 0.0 (±0) Dry 0.0 1.4 0.0 0.0 0.0 0.0 100.0 0.0 0.0 (±2.3) (±0) Serpentine Late dry 0.2 0.1 14.0 1.1 0.0 2.8 100.0 0.0 0.0 (±0.4) (±0.1) (±31.3) (±1.0) (±4.0) (±0) Wet 0.0 0.0 0.0 2.6 0.0 2.0 100.0 0.0 0.0 (±4.1) (±1.1) (±0) Dry 26.9 0.0 0.0 0.3 0.0 1.7 100.0 0.0 0.0 (±21.6) (±0.4) (±1.4) (±0) Broomfield Late dry 0.0 11.4 0.0 0.3 0.0 1.1 100.0 0.0 0.0 (±9.8) (±0.1) (±1.0) (±0)

3.3.5 Physico-chemical parameters

Spatial and temporal patterns in water quality parameters averaged across lagoons and sampling occasions are shown in Table 3.7. Physico-chemical parameters varied considerably across different seasons (sampling occasions) and lagoons. There was a significant difference in mean water temperatures between all sites (two-way ANOVA, F=5.2; df=6,85; p=0.0001), with the highest mean levels observed in Broomfield Lagoon (during the late dry summer season). Mean temperatures also varied across seasons in all Chapter 3. Limnological characteristics of the study sites 51 studied sites (p values of two-way ANOVA for each site <0.0001), with lower water temperatures observed during the winter (dry season). Maximum water temperatures reached 34°C late in the afternoon during the 2002 summer in Rainbow Lagoon, whereas minimum water temperatures reached 12°C early in the morning in South Callandoon Lagoon A during the winter (2003).

Wate r Dissolved Turbidity Conductivity Salinity Lagoon Season temperature pH oxygen (ppm) (NTU) (mS/cm) (%) (°C) South Late dry 23.9 5.86 7.09 940 0.377 0.01 Callandoon A (±1.5) (±0.52) (±0.06) (±36) (±0.004) (±0) Wet 25.3 4.57 6.92 54 0.247 0.00 (±1.2) (±1.13) (±0.12) (±27) (±0.006) (0) Dry 13.4 7.54 7.10 70 0.270 0.01 (±0.8) (±0.23) (±0.11) (±13) (±0.005) (0) Rainbow Late dry 22.9 7.26 7.40 472 0.280 0.01 (±2.1) (±0.59) (±0.15) (±42) (±0.001) (0) Wet 24.2 3.25 7.12 335 0.223 0.00 (±1.0) (±0.45) (±0.14) (±33) (±0.002) (0) Dry 14.4 8.69 7.22 346 0.252 0.00 (±0.7) (±0.30) (±0.23) (±6) (±0.012) (±0.01) River Late dry 25.7 7.69 7.64 61 0.217 0.00 (±0.4) (0) (±0.04) (±9) (±0.000) (0) Wet 24.5 5.99 7.16 210 0.206 0.00 (±0.0) (±0.04) (±0.15) (±5) (±0.001) (0) Dry 13.1 10.14 7.77 38 0.335 0.01 (±0.1) (±0.14) (±0.01) (±5) (±0.001) (0) Punbougal Late dry 24.2 8.80 8.24 100 0.365 0.01 (±1.6) (±1.90) (±0.16) (±27) (±0.002) (0) Wet 25.9 8.26 8.86 68 0.400 0.01 (±1.1) (±0.78) (±0.09) (±6) (±0.001) (0) Dry 15.1 11.73 8.71 43 0.419 0.01 (±0.4) (±0.49) (±0.08) (±13) (±0.008) (0) Maynes Late dry 25.6 8.03 8.34 581 0.775 0.03 (±1.9) (±0.70) (±0.07) (±96) (±0.044) (0) Wet 24.1 7.86 9.26 385 1.342 0.06 (±1.0) (±0.41) (±0.04) (±176) (±0.012) (0) Dry 13.7 10.80 8.64 461 1.533 0.07 (±0.3) (±0.90) (±0.14) (±93) (±0.005) (0) Serpentine Late dry 28.5 11.21 8.19 180 0.734 0.03 (±1.7) (±1.78) (±0.09) (±37) (±0.023) (0) Wet 24.5 4.72 8.61 483 1.273 0.05 (±0.8) (±1.51) (±0.11) (±100) (±0.014) (0) Dry 13.3 7.74 7.87 636 0.857 0.03 (±0.7) (±0.75) (±0.17) (±145) (±0.093) (±0.01) Broomfield Late dry 27.5 9.73 8.76 244 0.839 0.03 (±2.2) (±0.78) (±0.15) (±126) (±0.010) (0)

Chapter 3. Limnological characteristics of the study sites 52

Dissolved oxygen was generally high at all sites, with averages between 6.0 ppm (South Callandoon Lagoon A) and 9.7 ppm (Broomfield Lagoon). Significant differences were observed between lagoons (F=32.7; df=6,85; p<0.0001), with higher concentrations in Punbougal, Maynes, Broomfield and Serpentine lagoons. Similarly, significant differences were also observed between sampling occasions in all sites (p values of two-way ANOVA for each site ≤ 0.002), with lower mean values during the wet season (summer). Dissolved oxygen concentrations were higher in the dry season (during the winter) with mean values of up to 11.7 ppm in Punbougal Lagoon.

The study sites contained considerably turbid waters, with mean values of up to 940 NTU, even though the river and Punbougal Lagoon had relatively low turbidity values reaching a minimum of 37 and 43 NTU (respectively), during the winter dry season. Mean turbidity values were highly variable between sites (F=136.9; df=6,85; p<0.0001) and sampling occasions, with the exception of Rainbow Lagoon where the variation was not significant (F=2.1; df=2,85; p=0.1366). The sites with higher turbidity were Maynes, Serpentine and Rainbow lagoons, with turbidity higher than 180 NTU, but South Callandoon Lagoon A presented a mean turbidity of 940 NTU during the summer, prior to flooding.

Conductivity also varied significantly between lagoons (F=4119.5; df=6,85; p<0.0001) and sampling occasions within each lagoon (p values of two-way ANOVA for each site always <0.0001). The non-flooded sites presented higher overall conductivity as opposed to the flooded ones. In general, conductivity was lower during the wet season in the sites subjected to flooding (South Callandoon A, Rainbow Lagoon and the river), whereas lower conductivity in the non-flooded sites (Punbougal, Maynes and Serpentine lagoons) was observed mostly during the first sampling occasion, i.e. later in the summer. Significant differences between lagoons were also observed for salinity (F=1129.3; df=6,85; p<0.0001) and pH (F=592.8; df=6,85; p<0.0001).

Temporal variation in mean pH and salinity within lagoons were also significant (p values of two-way ANOVA for each lagoon ≤ 0.0026), except for Punbougal Lagoon where the mean salinity values were virtually unchanged across sampling occasions (F=0.0; df=2,85; p=1.0000). All sites were neutral to alkaline and, similarly to the conductivity results, the flooded sites presented more neutral pH, with mean values ranging from 6.9 to 7.7 and with lower values during the wet season, whereas the remaining sites were consistently Chapter 3. Limnological characteristics of the study sites 53 alkaline, with mean values between 7.8 and 9.2. Salinity was also particularly high in the non-flooded lagoons, when compared to the sites subjected to flooding.

Temporal and spatial variations in mean water quality parameters are further summarized by PCA (Figure 3.5). The overall spread of the lagoons in ordination space confirms the previous finding of a high degree of heterogeneity in water quality characteristics among the study lagoons. Even so, the first three axes of PCA explained 90% of the total variation in the data, with 47% attributed to the first axis alone (Table 3.8). This axis reflects variation in the chemical water qualities, i.e. pH, conductivity and salinity, being negatively correlated with these variables. The second axis explains 24% of the variation and is negatively correlated to dissolved oxygen and positively correlated with turbidity. Therefore, the highly turbid samples are arrayed on the top of this axis, while the water samples with higher oxygen concentrations are aligned to the bottom of this axis. The third axis, accounting for 18% of the variation, is negatively correlated to water temperate. Therefore, lower temperatures, as observed during the winter dry season, are arrayed towards the top of this axis.

% of Cum.% of Broken-stick Axis Eigenvalue Variance Var. Eigenvalue 1 2.854 47.568 47.568 2.45 2 1.431 23.846 71.414 1.45 3 1.101 18.351 89.765 0.95 4 0.333 5.543 95.309 0.617 5 0.197 3.281 98.59 0.367 6 0.085 1.41 100 0.167

Characteristic First 6 Eigenvectors 1 2 3 4 5 6 Water temperature 0.0770 0.3772 -0.8110 0.3639 0.2322 0.0880 pH -0.5064 -0.0919 -0.3871 -0.1958 -0.5342 -0.5115 Conductivity -0.5432 0.2378 -0.0219 -0.2851 -0.1175 0.7435 Dissolved oxygen -0.3557 -0.5700 0.0325 0.7138 -0.0337 0.1918 Turbidity -0.1422 0.6770 0.4147 0.4826 -0.2962 -0.1699 Salinity -0.5439 0.0972 0.1377 -0.0745 0.7470 -0.3349

Chapter 3. Limnological characteristics of the study sites 54

dry South Callandoon Rainbow dry Macintyre River Punbougal dry dry Maynes late Serpentine Broomfield dry

late

dry xis 3 A

wet

wet late wet wet

wet Water late (b) Temperature late late late wet

Axis 2

late Turbidity South Callandoon Rainbow wet Macintyre River Punbougal Maynes Serpentine late Broomfield wet

wet dry late

wet wet xis 2 xis

A late

dry late wet late dry late

dry Dissolved (a) oxygen

dry

Axis 1 pH Salinity Conductivity

Figure 3.5 Position of each study site and each sampling occasion within the ordination space defined by the first three factors identified by PCA of the cross-products matrix of the correlation coefficients for lagoon water quality data. (a) axis 1 and 2, and (b) axis 2 and 3. Site codes correspond to the following: late=late dry season, 2002 summer; wet=wet season, 2003 summer; dry=dry season, 2003 winter. Chapter 3. Limnological characteristics of the study sites 55

3.4 Discussion

3.4.1 Macintyre River flows

Reduced river flows resulting from water diversions for irrigation have become a major issue in the Border Rivers basin (Wylie 1995), and McCosker (1996) recognised reduced water quantities as one of the main issues related to agricultural practices and land use in this system. Approximately 40% of the long term average flows in the Border Rivers are diverted for irrigation purposes (Gutteridge et al. 1992).

The results of this study show that the majority of the study period experienced very low water flows along the stretch of the Macintyre River adjacent to the study sites. Although 2001 was a relatively wet year with several major to moderate floods, 2002 and 2003 were very dry, with only one minor to moderate flood occurring in each of these years. Most of the water quality parameters and morphological characteristics of the lagoons studied should be strongly related to water flow and the incidence of floods in the Macintyre River, as these sites are located on its floodplain and consequently subject to flooding by the Macintyre River. It is clear from the results presented that only a few of the study sites, i.e. South Callandoon A and Rainbow lagoons, were flooded during the period of this study. These lagoons are closer to the main channel and are, consequently, more likely to be filled by minor floods like the ones recorded during the 2002-2003 period.

3.4.2 Morphology and habitat characteristics

Seasonal changes in morphological and biological characteristics of floodplain lagoons have been documented for several river systems in Australia (Hart and McGregor 1982, Mackey 1991, Kennard 1995) and overseas (Rodriguez and Lewis 1994, Vyverman 1994, Silva and Thomaz 1997). In the present study, fluctuations in the main river water levels were found to influence the morphology of the study sites by increasing overall size (area and volume) and depths of the lagoons subjected to flooding during the wet season. Water level fluctuations were not sufficient to affect all of the study sites, and therefore, the non- Chapter 3. Limnological characteristics of the study sites 56 flooded lagoons also presented temporal changes in their morphology, mostly as a result of receding water levels due to evaporation throughout the study period.

Despite the observed temporal and spatial variations in the measured morphological characteristics of the sites studied, the general habitat characteristics varied mostly between lagoons rather than sampling occasions. Johnson (1999) described aquatic habitat in the Macintyre River as very poor to good, with very poor and poor ratings accounting for 73% of the river length and only 5% of the river length classified as good. Although Johnson’s (1999) data describes the main channel of the Macintyre River, the results of the present study also indicate that, on the whole, the aquatic habitat was similarly poor for most of the lagoons studied. As shown in Figure 3.4a, sites subject to flooding during the study period (cluster of sampling occasions on the top left of the plot) presented a higher diversity and greater overall proportion of aquatic structures, such as debris, root masses and leaf litter. Other elements such as aquatic and overhanging vegetation were also relatively abundant in these lagoons. Of the non-flooded lagoons, only Serpentine and Punbougal presented significant contributions of aquatic vegetation. Although the results indicate that macrophyte growth was higher after flooding, Serpentine Lagoon showed an increase in the contribution by aquatic vegetation at the end of the study period. The fact that this lagoon dried consistently throughout the study period represented an increase in the proportion of shallow water available for colonisation by aquatic vegetation. Moreover, shading produced by riparian overhanging vegetation is considered a primary factor limiting growth (Friday 1987, Mackay et al. 2003), distribution and abundance (Brookes 1994, Bunn et al. 1998) of aquatic plants, and may have influenced vegetation growth in lagoons such as Serpentine and Punbougal, where riparian cover was low. Another factor that may have affected aquatic vegetation and algal growth in the study sites is water turbidity, as only the sites with lower turbidity values presented higher proportions of macrophytes and algae. While higher light intensity is known to favour germination (Gopal 1986), suspended sediments can shade out aquatic macrophytes, leading to a reduced plant biomass (Bruton 1985). It is important to note that even though Punbougal Lagoon presented higher levels of algae and aquatic vegetation, this lagoon was also much larger and deeper. Because these habitat elements were present only in the very narrow stretches of the margins where depths were lower, it is likely that both of these habitat elements were considerably overestimated in the analyses.

Chapter 3. Limnological characteristics of the study sites 57

Much of the temporal variation in habitat characteristics was observed within the flooded lagoons (South Callandoon A and Rainbow). This can be attributed to the influence of the wet season, when increased proportions of aquatic vegetation and algae were recorded for both lagoons. Increased growth in aquatic macrophytes after the wet season is reported in other studies (Marchant 1982, Osborne et al. 1987, Finlayson et al. 1989) and seems to be the result of larger areas of shallow water around the waterbody margins, leading to higher availability of habitat suitable for colonisation. The increase in area of the flooded lagoons during the wet season is also an important factor explaining the higher contribution of leaf litter, as these lagoons’ margins were drawn closer to the riparian vegetation.

3.4.3 Water quality characteristics

Water quality parameters of floodplain lagoons are expected to show temporal changes, mostly related to wet/dry seasons and water flow (Hart and McGregor 1982, Osborne et al. 1987, Kennard 1995). Even though the oxygen dissolved in freshwaters is derived mostly from the atmosphere and photosynthesis in algae and aquatic plants (Hart and McGregor 1982), ultimately, the total amount of dissolved oxygen will be dependent on water temperature and, to a lesser extent, salinity (Cole 1994).

The study lagoons presented temporal and spatial variations in dissolved oxygen concentrations. Temporal patterns were more pronounced in the flooded lagoons, where dissolved oxygen levels were lower during the wet season soon after flooding, whereas spatial patterns were evident between flooded and non-flooded lagoons, where the latter had overall higher concentrations of dissolved oxygen in the water. Houldsworth (1995) found that dissolved oxygen in sites along the Border Rivers catchment was depressed in the summer as a result of increased degradation of organic matter, stimulated by the resuspension of sediments during preceding high flows. This could have been a major factor explaining the overall higher dissolved oxygen levels in the flooded lagoons as they presented higher proportions of organic matter (leaf litter), aquatic vegetation and algae. These elements were also abundant in Punbougal Lagoon (non-flooded) but only in the shallow margins and, given the size and depth of this lagoon, it is unlikely that they had a significant effect on dissolved oxygen levels in this lagoon. Hart and McGregor (1982) and Kennard (1995) also found that oxygen levels in floodplain lagoons were lowest Chapter 3. Limnological characteristics of the study sites 58 during the wet/early dry season, and that this could be a result of decay of aquatic vegetation and/or bacterial decomposition of allochthonous organic matter washed into the lagoons during the wet season. The lower dissolved oxygen levels during the late dry and wet seasons in the present study can also be interpreted as the result of the warming up of the epilimnion during the summer, when overall temperatures were higher. Increased proportions of aquatic vegetation and algae, and consequently respiration by aquatic vegetation, mostly during the wet season, in the flooded lagoons, may also have contributed to lowering the amount of dissolved oxygen in the water of these sites, although there is no strong evidence from the data to support such argument, as aquatic vegetation and algae contributions were widely variable between seasons in the flooded lagoons.

The concentration of dissolved salts in the water can also affect the volume of dissolved oxygen in water by lowering the solubility of this gas (Cole 1994). Interestingly, this does not accord with the results of this study. Despite the fact that no clear relation was evident between conductivity, or salinity, and dissolved oxygen in the study lagoons, conductivity and salinity were generally higher in the non-flooded lagoons than in the flooded ones, where the overall dissolved oxygen levels were higher. In general, the conductivity values in this study are comparable with those documented by DIPNR (2002) for the Border Rivers catchment, although some of the study sites (Maynes and Serpentine lagoons) had much higher values, up to 1.533 mS/cm (see Table 3.7).

Conductivity is a measure of the total concentration of dissolved inorganic salts in the water, and is expected to increase as the dry season progresses due to evaporation and the consequent concentration of dissolved salts (Hart and McGregor 1982, Mackey 1991). In the flooded lagoons, conductivity was higher during the first sampling occasion, late dry season, and lower after flooding in the wet season. This is in accordance with Osborne et al. (1987) who reported an increase in conductivity accompanied by the fall in water levels in Lake Murray (Papua New Guinea), and in the present study is obviously a result of higher concentrations of salts at the end of the dry season, which were subsequently diluted by flooding after the wet season. In addition, Houldsworth (1995) found that the increase in water dilution during the summer in the Border Rivers catchment, in this case by rainfall, led to a trough in salinity values and also that higher salinities during the spring (October) were the result of a higher proportion of groundwater flow. On the other hand, Chapter 3. Limnological characteristics of the study sites 59 non-flooded sites examined during this study tended to have higher conductivities during the last two sampling occasions, during the wet and dry seasons. As these sites were not flooded and decreased continuously in volume, salt concentrations increased as time progressed.

Mean values of pH in the study sites are in accordance with those reported by Houldsworth (1995) for the Border Rivers catchment and, although less variable, overall pH levels were also comparable with those of other studies in floodplain lagoons (Hart and McGregor 1982, Mackey 1991, Kennard 1995). An increase in pH should be expected with the decrease in water levels (Osborne et al. 1987) which was the case in almost all of the study sites, except Serpentine Lagoon where pH decreased with water level. In general, changes in pH were very small, even though sites subject to flooding presented lower overall pH levels than non-flooded ones. Additionally, flooded lagoons showed their smaller pH values during the wet season, whereas non-flooded lagoons presented lower pH usually at the first sampling occasion (late dry season). McLachlan et al. (1972) found that a fall in water levels in an African lake was accompanied by an elevation in pH, and argued that this was a result of an increase in the concentration of major ions due to evaporation.

There is strong evidence that the non-flooded lagoons had not received water since the end of 2001 when the last major floods occurred in the Macintyre River. Therefore, these lagoons have been drying continuously during the 2002-2003 period, as the next moderate to major floods only occurred early in 2004 after the end of this study. On the other hand, the remaining lagoons (South Callandoon and Rainbow) and the river received water intermittently throughout this time. Therefore, it is possible that the flooding of some of the study lagoons resulted in their overall lower pH, whereas non-flooded lagoons retained higher pH levels. Kennard (1995) reported lower pH levels in recently-flooded floodplain lagoons, and suggested that such results could have been a consequence of leaching of organic matter from submerged leaf litter and other terrestrial vegetation. Previous work has also suggested that raised pH in water can result from the photosynthetic activity of algae (Houldsworth 1995) and high phytoplankton production (Osborne et al. 1987). Even though the leaching of submerged organic matter may have resulted in the lower pH levels recorded in the flooded lagoons in the present study, as these lagoons had a relatively higher leaf litter component then the others, it is unlikely that algal or phytoplankton production had any significant effect on the pH levels in the non-flooded lagoons, given Chapter 3. Limnological characteristics of the study sites 60 the low contributions of algae to the habitat characteristics and the high turbidity recorded for these lagoons.

Turbidity is a measure of the scattering of light by suspended particulate and colloidal matter, such as clay, silt, phytoplankton and detritus (DIPNR 2002). In the Border Rivers catchment, although a natural feature of these systems, turbidity is among the main water quality issues identified by McCosker (1996), and Boddy and Bales (1996) suggest that lower diversity and abundance of macroinvertebrates and macrophytes in the Macintyre River may be related to high turbidity.

Even though the overall turbidity on the study sites was considerably higher than reported by DIPNR (2002) for the Macintyre River, and higher than the basin-wide median of 180 NTU reported by Houldsworth (1995) for the Border Rivers system, Punbougal Lagoon and the Macintyre River site had turbidity levels comparable to those found by these authors. Water bodies in inland Australia are commonly very turbid as a result of high concentrations of fine suspended particles (Kirk 1985) from turbulent re-suspension of bottom sediments caused by wave action (Davis et al. 2002). The higher turbidities observed in the floodplain lagoons in the present study are probably the result of the low to nil flow conditions experienced by most of the lagoons, where water was confined to the channel and erosion of bed and banks and sediment re-suspension is likely to have occurred. Additionally, the presence of cattle and sheep (see Bowling and Jones 2003) associated with the lack of protection from wind by riparian vegetation, are very likely to have contributed to the high turbidity in many of the study lagoons. The fact that Punbougal Lagoon retained relatively lower turbidity throughout the study period is possibly the result of its high depths and substrate composition. The re-suspension of bottom sediments is prevalent in shallow waterbodies (Davis et al. 2002), and the high proportions of sand and gravel in the substrate of Punbougal Lagoon would have limited the amount of fine sediment available for re-suspension in the water.

Temporal changes in turbidity were also evident from the results of this study even though they were only partially in agreement with results from other studies. A positive correlation between water flow and turbidity in the Border Rivers catchment is reported in several studies (e.g. Houldsworth 1995, Johnson 1999) and turbidity is usually higher during or shortly after rainfall as these events cause runoff into the waterways (DIPNR Chapter 3. Limnological characteristics of the study sites 61

2002). Although this was the case in the Macintyre River site, all other study sites showed the highest turbidity during the sampling occasion before the wet season, except for Serpentine Lagoon.

McLachlan (1972) found that the drop in water levels in an African lake was accompanied by an increase in turbidity and Hart and McGregor (1982) reported that by the end of the dry season, floodplain lagoons of the Magela Creek in the Northern Territory were very turbid, but after initial flushing the water clarity was markedly improved. They explained these patterns in terms of high concentration of particles in the water due to evaporation in the African lake and low levels of suspended solids in the water flowing from the Magela Creek into the floodplain. Although it is unclear which factors led to the seasonal variations in the non-flooded lagoons of the Macintyre River, the flooded ones had high turbidity during the late dry season when their sizes were smaller and, due to the low magnitude of floods during the study period, it is likely that these lagoons actually had their suspended particles diluted after flooding by the Macintyre River, resulting in a decrease in turbidity after flooding.

3.5 Conclusions

The study period was characterised by relatively low flows in the main channel of the Macintyre River, with only two minor to moderate floods occurring between 2002 and 2003, usually in the early months of each year (late summer). This directly influenced the hydrology of the study sites as, apart from the river site, only two of the 7 lagoons studied were flooded (South Callandoon A and Rainbow lagoons) and two dried up completely (South Callandoon B and Broomfield lagoons). The remaining sites, Serpentine, Punbougal and Maynes lagoons, decreased in size and volume continuously throughout the study period but did not dry up completely.

Even though the habitat characteristics varied on a spatial (between lagoons) and temporal (between sampling occasions) basis, most variation occurred between flooded and non- flooded lagoons, as the latter sites presented significantly lower proportions of habitat elements, such as aquatic and overhanging vegetation, algae, debris, leaf litter and root masses. On the other hand, the flooded lagoons showed greater indication of temporal Chapter 3. Limnological characteristics of the study sites 62 changes, given that many of these habitat elements increased in proportion during the wet season, after flooding.

In a similar fashion, the water quality parameters presented significant variation between lagoons and sampling occasions. Most variation occurred between flooded and non- flooded lagoons, and even though patterns were not clear for the non-flooded sites, the flooded lagoons showed clear patters in water quality parameters in relation to flooding, when pH, conductivity and dissolved oxygen were lower and water temperatures were higher. Flooding also decreased turbidity in the lagoon sites, whereas flooding at the river site increased turbidity.

3.6 Implications for the study of fish diet composition

The Macintyre River is regulated in the vicinity of the study sites with water flows controlled by two artificial water storages, the Goondiwindi and Boggabilla weirs, constructed in the main channel adjacent to both towns (see Figure 2.2), as wells as other major dams constructed further upstream (see Section 2.3 of Chapter 2).

Accordingly, the hydrology and flooding history of the study lagoons and the river site, represent non-natural conditions. In addition, some lagoons are subject to pumping into from river water and out for irrigation. Thus, an additional level of modification of water regime has been superimposed on these lagoons. Therefore, it is important to stress that there are at least two levels of water regulation which affect the study sites. On a local scale, some of the study lagoons are subject to water modifications by artificially changed water levels, e.g. Rainbow and Serpentine lagoons. On a larger scale, all study sites, as well as the floodplain of the Macintyre river, are subject to un-natural patterns of inundation resulting from the artificial flow regime of the river, due to flow regulation by major and minor water storages (Section 2.3 of Chapter 2). These hydrological modifications coupled with the variable distances from the river and elevations of the study lagoons, and their flow history during the study period, have produced wide spatial and seasonal variations in lagoon water regime, morphology, habitat structure and water quality.

Chapter 3. Limnological characteristics of the study sites 63

The study lagoons, therefore, present a wide range of spatial and seasonal contrasts that can be expected to influence the food resources available for fish and fish dietary composition. These circumstances of wide ranging spatial and seasonal variations among lagoons are ideal for addressing the question of how floodplain lagoons trophically support their fish assemblages in a regulated river system.

Chapter 4. Diel variation in food intake and diet composition 64

4 Diel variation in food intake and diet composition of fish in floodplain lagoons

4.1 Introduction

Several studies of feeding rhythms of fish have reported diel patterns of feeding activity, and that these patterns may even change seasonally (Gray et al. 1998, Jacobsen and Berg 1998, Mookerji et al. 1998, Cardona 1999, Horppila et al. 2000). Keast and Welsh (1968) suggested that diel variation in diet of fishes probably reflects changes in the activity and, consequently, vulnerability of their prey. In contrast, Boujard and Leatherland (1992) argued that these circadian rhythms may also be under endogenous control and synchronized by light. A more direct effect of light on feeding activity of fish is related to the natural variations in light levels throughout the day. As light intensity declines near sunset, the contrast between prey and background also declines, therefore visually hunting fish should show a strong relationship between light intensity and feeding intensity. In contrast, as light levels increase, feeding activity shows little change until a threshold of light intensity is achieved, and only then will feeding intensity increase with further increases in light intensity (Wootton 1990).

In natural populations, changes in the rate of food consumption can also be correlated with water temperature (Worobec 1984), though the causal effects of temperature may be confounded or correlated with those of other abiotic factors and changes in the availability of food (Wootton 1990). Keast (1968) reported that several species of fish significantly reduced their feeding activity rates at low temperatures, and several Australian species present similar behaviour, for example, Macquaria novemaculeata (Harris 1985) and Arius graeffei (Rimmer 1985) have been found to cease feeding or reduce food intake in the winter as a result of low water temperatures. Leiopotherapon unicolor was also found to decrease its food intake and metabolic rates associated with lower water temperatures (Gehrke 1988a). Other studies on feeding rhythms of fish that report strong diel patterns of feeding suggest that control of feeding time is not regulated necessarily by natural variations in food availability (Boujard 1995) or light intensity or wavelength (Gibson and Keenleyside 1966). Alternatively, the choice of feeding time of fish may be influenced by temporal variation in the intensity of competition (Kadri et al. 1997). Chapter 4. Diel variation in food intake and diet composition 65

Likewise, variations with time of day can also be found in the composition of food items ingested by fish (Bowen 1983), and can be related to the availability and/or accessibility of specific food types at certain times of day (Keast and Welsh 1968). According to Keast and Welsh (1968), the main advantages of such behaviour are that a higher number of food items can be cropped by a given fish species and that different feeding times also serve to reduce interspecific contact and competition. Other studies (e.g. O'Brien 1979, Gliwicz 1986) have emphasized the effects of diel vertical migrations in zooplankton on the dietary composition of fish that feed on this resource. Even though the contribution of zooplankton can be expected to increase at low light times when these organisms migrate up in the water column, other factors such as light availability should be considered, as low light conditions can reduce visibility of predators (Mookerji et al. 1998, Horppila 1999).

In this sense, sampling procedures designed to study food habits of fish (e.g. food preference, rates of food consumption, ontogenetic shifts in diet composition or predation) need to cover the peak feeding periods of individual species, in order to maximize the amount of information obtained per fish collected. Such procedures should reveal the time of day when individual fish would be most likely to contain food in their stomachs, and should also determine if individuals collected at different times of day could be treated as samples from the same feeding population with respect to the composition of their diet.

4.1.1 Aims

This chapter is part of a broader study on diet composition of fish that addresses variations in diets as a consequence of seasonal and spatial patterns of flow history and flow regulation (Chapters 5 and 6). It is important that other sources of variation in diet composition, such as diel variation, are understood in order to optimize comparisons of diet composition related to spatial and seasonal patterns. Therefore, once diel feeding periodicity is understood, it is advisable to concentrate the sampling effort on the periods of active feeding (as per Boisclair and Leggett 1988, Hayward et al. 1991).

This chapter represents the first of several steps in the description of the feeding ecology of Ambassis agassizii, Leiopotherapon unicolor and Nematalosa erebi. It aims to uncover the Chapter 4. Diel variation in food intake and diet composition 66 peak feeding periods of these species in floodplain lagoons of the Macintyre River, by investigating their diel patterns of feeding activity on a short-term scale, i.e. possible changes in the intensity of feeding and dietary composition on a diel basis. Specific research questions addressed in this chapter are:

1. Is there any variation in food intake by fish throughout a 24 hour period on a seasonal (summer/winter) basis? If so, when is the peak feeding time for each of the species studied during each time of the year? 2. Does dietary composition change as a result of diel or diel/seasonal patterns of variation in food intake? 3. As a consequence of 1 and 2, what is the optimal time of day to sample fish in summer and winter?

4.2 Methods

4.2.1 Study design

To evaluate diel patterns in feeding activity and dietary composition, fish samples were taken on two occasions (summer and winter) from sites where each species was most abundant at the time of sampling, as shown in Table 4.1. Three lagoons on the Macintyre River floodplain (see Chapter 3 for site information) were sampled during the summer of 2001/2002 and the winters of 2002 and 2003. Nematalosa erebi was sampled during the summer of 2001 (27-28 November) and the winter of 2002 (09-10 August) in Rainbow Lagoon. Ambassis agassizii was sampled from South Callandoon Lagoon B during the summer of 2002 (20-21 February) and from Rainbow Lagoon during the 2002 winter (9-10 August). Leiopotherapon unicolor was sampled from South Callandoon Lagoon B during the 2002 summer (20-21 February) and from South Callandoon Lagoon A during the 2003 winter (4-5 August) (Table 4.1). The reason for samples of A. agassizii and L. unicolor being taken from two different sites from summer to winter was that South Callandoon Lagoon B dried up soon after the summer of 2002 and both species became unavailable at this site.

Chapter 4. Diel variation in food intake and diet composition 67

Table 4.1 List of sites used for the analysis of diel variation in food intake and daily variation in diet composition of fish in dryland river lagoons.

Time Species Location Sunrise / Sunset Diel Feeding Activity Summer Nov/01 N. erebi Rainbow 0450 / 1840 h Feb/02 A. agassizii South Callandoon B 0550 / 1830 h Feb/02 L. unicolor South Callandoon B 0550 / 1830 h Winter Aug/02 N. erebi Rainbow 0630 / 1730 h Aug/02 A. agassizii Rainbow 0630 / 1730 h Aug/03 L. unicolor South Callandoon A 0630 / 1730 h

Fish feeding patterns are expected to fit into three general categories, namely nocturnal (night feeding), diurnal (day feeding) and crepuscular (twilight feeding) (e.g. Boujard and Leatherland 1992), although food intake may present some variation within these categories (Eggers 1977). Therefore, times of sampling were chosen based on these three main categories, in order to take into account both diurnal (1200h) and nocturnal (2400h), and also twilight (1800h and 0600h) feeding activity. Each collection consisted of four diel samples taken at 6-hour intervals (0600h, 1200h, 1800h, and 2400h) during a 24-hour period.

Data replication was in part driven by the availability of enough individuals for meaningful statistical analysis, given that the effect size sought in this kind of survey to attain sufficient statistical power requires a minimum number of individuals being used per sample to account for variation among individual fishes (Platell and Potter 2001). Another factor affecting replication was that, although the sampling gear (Section 4.2.2) used presented a relatively low selectivity for species and size-classes (unlike most fish sampling gear) (Lyons 1986, Borgstrom and Plahte 1992), its efficiency and adequacy can vary with different habitat types (Vigg 1981, Parsley et al. 1989), thus limiting accessibility in some areas and, therefore, sampling effectiveness. These difficulties were compounded by the overall low abundance of fishes in many of the study lagoons and high variability in catches between and within lagoons through time.

Strong evidence supports the idea that circadian rhythms in fish are synchronized by light/dark alternation (Boujard and Leatherland 1992) and can be shifted seasonally (Landless 1976, Boujard et al. 1990) when daylight duration changes. Because the main goal at this point was to understand the peak feeding times of selected species rather than characterize diets between lagoon sites, spatial replication was considered unnecessary, Chapter 4. Diel variation in food intake and diet composition 68 although spatial replication was achieved for two of the three species studied. Instead, data sampling for each target species was repeated on a seasonal basis (summer and winter), to test for seasonal differences in the diel feeding times of fish.

4.2.2 Collection methods

Fish collections were performed as per Table 4.1 during summer and winter periods. Each lagoon was sampled using a seine net (25 m length x 2 m height x 1 cm mesh). Given the size of most sites, samples were taken at random points along the margins (maximum sampled depth of 1.5 m), as these areas generally contain the majority of fish and because seining becomes impractical at water depths greater than 1.5 metres. Individual seine- transect lengths were approximately 10 m and from 5 to 15 m wide depending on the lagoon depth. Duration of each seine was about 5 minutes with 5 to 10 minutes between seines to allow time to sort the fish caught and remove woody debris from the net. This was repeated a minimum of three times along the lagoon margins until a maximum of 30 individuals of each fish species was caught for each time-sample or one hour had elapsed from the beginning of the first seine. Thus, total sampling duration never exceeded one hour after the sampling had started. In some sites (e.g. Rainbow Lagoon) the presence of snags, logs and tree roots and high depths in the margins limited the total seined distance to about 40 to 50 m in the same location, but in most cases the overall sampled distance consisted of about 100 m along the shores in one to four locations within each lagoon.

The number of individuals for each fish diet sample was chosen on the basis of previous studies (Kennard 1995, Ebner et al. in press). Once the maximum number of individuals (n=30) of a given species was achieved, any additional individuals of that species were returned unharmed to the point of capture, as well as any individuals of unwanted species. To avoid size-related variations in relative food intake and diet composition between samples, only individuals within similar size classes were kept for analysis. Given that the overall size of the study species was relatively small (less than approximately 150 mm standard length), the individuals of the target species were preserved whole in 4% buffered formalin immediately after capture. Later in the laboratory, fish were transferred to 70% ethanol and the necessary measurements were performed. Chapter 4. Diel variation in food intake and diet composition 69

4.2.3 Dietary analysis

In the laboratory, each individual fish was washed in tap water to remove excess ethanol and blotted with absorbent paper before being measured to the nearest mm with Vernier calipers and weighed using an analytical balance (Mettler AE240) to 0.0001 g. The standard (SL) and total lengths (TL), total weight (TW) and body weight (BW) (the weight of the fish after all organs were removed from the body cavity), were recorded.

For the dietary analysis, the stomach of each fish was removed, blotted and then placed in a petri dish where the degree of stomach fullness was estimated by eye. Stomach fullness was estimated by (1) assigning a score between 0 (empty) and 10 (fully distended with food), reflecting a percentage value of food that occupied the stomach, and (2) the relation between the indirect volume estimate (see below) of the stomach contents and the fish’s standard length. Volume was used as opposed to weight because fish size was usually small (less than 150 mm SL), which made weighing impractical or unreliable on most occasions. It is also important to note that N. erebi has a two-part stomach (Atkins 1984, Gerking 1994), comprising a muscular and distensible cardiac portion, also know as the pharynx or foregut, and a gizzard-like pyloric portion which is also muscular but rigid, henceforth called the gizzard. In this species, both foregut and gizzard were considered in the estimation of stomach fullness and diet composition.

Direct volumetric measurement of stomach fullness and prey items could not be determined due to the small size of most of the stomach items examined (often smaller than 10 mm3). Instead, the proportional contribution of each class of items to the diet was estimated using Hyslop’s (1980) indirect volumetric method, where stomach items within each prey category are lumped together and squashed to a uniform depth of 1 mm and the area covered taken to represent the volumetric contribution of that food category to the total stomach contents in mm3 (Arthington 1992, Pusey et al. 2000).

Prey items were identified to the lowest level of taxonomic resolution, resulting in 29 dietary items and 6 broader food categories recognised for analysis (adapted from Kennard et al. 2001) (see Results). Microcrustaceans were identified to family level, based on Ingram et al. (1997) and Shiel (1995), although copepods were resolved only to order. Other invertebrates, mostly insects, were identified to the level of family according to Chapter 4. Diel variation in food intake and diet composition 70

Williams (1980), Hawking (1986), Hawking and Smith (1997) and Ingram et al. (1997). Algal items where identified, when possible, using Prescott (1970) and Entwisle et al. (1997).

4.2.4 Data analysis

4.2.4.1 Diel feeding activity

Patterns of diel variation in food intake were derived from the variation in frequency of occurrence and mean stomach fullness (%), and differences in mean relative content volume (RCV). The RCV was calculated for each fish as the stomach contents volume divided by the fish standard length. To test for possible bias associated with fish size, the stomach fullness and RCV were plotted against fish standard length and the R2 taken as an indicator of the strength of the correlation. Notice that the term ‘feeding activity’ in this study refers to the amount of food ingested by fish, as measured by gut contents fullness and relative stomach contents volume. To describe the general diel variation in food intake for each species, plots of stomach fullness and RCV over times of day were constructed which incorporated mean values from each of the times of day/season/species samples. Exploratory analysis was performed by constructing the marginal means table and using preliminary runs of ANOVA for times of day per species and seasons. Examination of residuals and probability plots were performed to verify the assumptions of the ANOVA and, when necessary, log(x+1) transformation was applied to the RCV and Arcsine transformation was applied to stomach fullness, as is appropriate for percentage or proportion data (Sokal and Rohlf 1969).

For comparisons of stomach fullness and RCV between species and seasons, each measurement of the response variables (stomach fullness and RCV) was classified according to two factors (species and season), where species has 3 factor levels (= N. erebi, A. agassizii and L. unicolor), and season has 2 factor levels (= summer and winter). To test for differences in stomach fullness and RCV for each species between times of day and seasons, each measurement of the response variables was classified according to two factors (time of day and season), where time of day has 4 factor levels (0600h, 1200h, 1800, 2400h), and season has 2 factor levels (summer and winter). Chapter 4. Diel variation in food intake and diet composition 71

This design is crossed, with data for each combination of factor levels, and unbalanced, with unequal sample sizes. Because of the nature of the data, more than one factor could be affecting the response variables, so a factorial (two-way) ANOVA with fixed factors was applied to the data using the GLM procedure (SAS Institute 1988), which can handle unbalanced designs. In the case of a strong interaction between species and season, a sub- analysis was conducted on the effect of season (which has only two levels) on either species or times of day (SLICE option of the LSMEANS statement, (SAS Institute 1988)). Restricted a-priori pairwise comparisons were performed using the Bonferroni method, to test for differences in mean fullness and RCV of fish species or times of day within and between seasons. This method adjusts the critical value of significance (α=0.05) to control experimentwise error rates in multiple testing situations. Despite being conservative when there are too many comparisons, it provides great control over type I error (Byrkit 1987).

4.2.4.2 Diel variation in diet composition

The contribution of each dietary item to the diet was expressed as the percentage frequency of occurrence of each dietary item in the stomachs of all individuals (n is shown in Table 4.2) of each fish species and as the mean of the percentage volumetric contribution made by each dietary item to the stomach contents of each individual of each fish species.

The unidentified material category was included only in the description of the total mean diet of the selected species to give an indication of the average proportional contribution to the total mean diet of the unidentifiable fraction (see Table 4.3). Since the inclusion of values of unidentified material might bias the results of multivariate analysis (Platell and Potter 2001) and their overall contribution to the diets of all species was very small (less than 11%) (see Table 4.3), their proportional contribution to the total was excluded in all subsequent calculations, and the data was re-normalised to sum 100% (as in Kennard et al. 2001).

For all subsequent statistical analysis on diel diet composition between seasons, only stomachs with fullness equal or greater than 2 (or 20%) were considered. Stomachs with less than 20% content fullness were excluded in order to prevent bias in the calculation of Chapter 4. Diel variation in food intake and diet composition 72 proportional contribution of prey or diversity indices. According to Pusey et al. (1995), nearly empty stomachs may contain only one item, for example, which can erroneously indicate that the diet of the fish was not diverse. If that item was infrequently encountered in the diets of other individuals with fuller stomachs, inclusion of that individual would place undue importance on that one unusual item.

To understand the relative extent to which the dietary compositions of fish were influenced by overall differences in species among times of day and seasons, the proportional volumetric dietary data for individual representatives of each species from each time of day and each season were randomly allocated into groups of 3 to 6 individuals and the means calculated for each of those groups (as in Schafer et al. 2002). The allocation of individuals of each species into groups, prior to analysis, overcomes the problem posed by the fact that the stomachs of individual fish often contained only a few of the total number of dietary items recorded. These data were also used to calculate the mean Shannon- Wiener diversity index of prey in the stomachs of each species at each time of day and season, this index corresponding to dietary breadth (B) (Marshall and Elliott 1997).

Overall similarities in dietary composition of each species between seasons and times of day were evaluated using classification (sequential agglomerative hierarchical non- overlapping clustering, SAHN), to find discontinuities between samples. Ordination (Non- metric Multidimensional Scaling, NMS) was used for further evaluation and interpretation of how discrete seasonal and diel samples were. As described by McCune and Mefford (1999), NMS is a robust technique that does not require that the original distances are maintained in the ordination space, but only their rank order.

Prior to clustering and NMS, the mean proportional contribution by volume of dietary items for the different groups of fish individuals were arcsine square root transformed, as is appropriate for proportional multivariate data (McCune and Grace 2002), and an outlier analysis was performed to check for outliers and skewness. The Relative Sorensen (Kulcynski) distance was used for the cluster analysis and ordination. This is a relativised Manhattan distance by sample unit totals. All graph outputs were Varimax-rotated. The fusion strategy for clustering was the Flexible clustering with beta =-0.1, which results in a conservative effect on space and better separates the groups. Chapter 4. Diel variation in food intake and diet composition 73

4.3 Results

The plots of the RCV and stomach fullness against fish size (SL) for each of the three species showed that the RCV and fullness displayed no significant relationship with fish length (Figure 4.1a-c). The standard length of A. agassizii (Figure 4.1a) could explain only 0.6% of the variation in stomach fullness and 0.1% of the variation in RCV. Leiopotherapon unicolor also showed no significant relationship between size and the RCV or fullness (Figure 4.2b), with no more than 0.41% of the variation in both variables being explained by the standard length of fish. In a similar fashion, N. erebi had only 2.3 and 1.3% of the variation in stomach fullness and RCV (respectively) being explained by fish size (Figure 4.1c). Therefore, there was no need for further adjustment for bias associated with size in the RCV and stomach fullness data in any of the species studied. This absence of a relationship was most likely a result of the relatively restricted size range of the individuals sampled.

4.3.1 Diel feeding activity

Relative stomach content volume (RCV) and percent fullness varied widely among individuals both within and between times of day for each species, and standard deviations were in many cases almost equal to or greater than the mean values (Table 4.2). Frequency of occurrence of nearly empty stomachs (fullness ≤20%) was higher mostly in the early morning (0600h) and midnight (2400h) samples, although some variation was observed (Figure 4.2). For instance, during the winter, frequency of occurrence of nearly empty stomachs seemed to be higher across times of day than during the summer. In contrast, full or nearly full stomachs (fullness ≥60%) seemed to be more frequent at midday (1200h) and in the late afternoon (1800h), mostly the former (Figure 4.2).

Mean stomach fullness varied significantly across the studied combinations of seasons and species (two-way ANOVA, F=12.31; df=5,635; p<0.0001), although the percentage of variation that could be attributed to the effect of these two factors, individually and in combination, was very low (r2=0.0884).

Chapter 4. Diel variation in food intake and diet composition 74

(a) A. agassizii

100 5 2 2 90 r = 00058. r = 00008. 80 4 70 60 3 50 40 2 30

Stomach fullness (%) fullness Stomach 20 1 10 content Relative volume 0 0 0 10203040 0 10203040 Standard length (mm) Standard length (mm)

(b) L. unicolor 100 60 2 2 90 r = 0. 0041 r < 00001. 80 50 70 40 60 50 30 40 30 20

Stomach fullness (%) 20 10 10 content Relative volume 0 0 0 102030405060 0 102030405060 Standard length (mm) Standard length (mm)

Stomach fullness (%) fullness Stomach 20 10 volume content Relative 5 0 0 0 20406080100 0 20406080100 Standard length (mm) Standard length (mm)

Table 4.2 Summary of results from two (summer and winter) diel feeding studies of three fish species from the floodplain of the Macintyre River. Data are sample sizes (N), mean sample sizes (± SD) per season, standard length (SL) (± SD), body weight (± SD), stomach fullness (± SD), percentage of nearly empty stomachs (fullness < 20 %) (n) and relative content volume (RCV) (± SD).

Species and Mean time- Mean SL Mean body Mean % near Total N Mean RCV Season sample N (mm) weight (g) fullness (%) empty

A. agassizii

Summer 109 27.3 21 0.2229 33.9 33.9 0.9101 (± 3.77) (± 1.5) (± 0.04) (± 26.05) (n=37) (± 0.91) 0600h 27 20 0.2105 31.1 33.3 0.6893 (± 1.5) (± 0.04) (± 23.1) (n=9) (± 0.70) 1200h 22 19 0.1912 59.5 0 1.6249 (± 1.6) (± 0.05) (± 13.6) (± 1.00) 1800h 30 22 0.2381 40.0 16.7 0.9963 (± 1.1) (± 0.03) (± 26.4) (n=5) (± 0.75) 2400h 30 21 0.2420 11.3 76.7 0.4984 (± 0.9) (± 0.03) (± 12.2) (n=23) (± 0.85) Winter 107 26.8 33 0.8156 31.2 44.9 0.8754 (± 2.50) (± 2.2) (± 0.19) (± 30.1) (n=48) (± 0.99) 0600h 28 34 0.9391 5.7 85.7 0.1536 (± 2.3) (± 0.22) (± 7.3) (n=24) (± 0.21) 1200h 23 32 0.7642 50.0 4.3 1.3324 (± 2.5) (± 0.20) (± 23.9) (n=1) (± 0.96) 1800h 28 32 0.7634 58.2 7.1 1.7650 (± 1.9) (± 0.14) (± 25.1) (n=2) (± 0.99) 2400h 28 32 0.7866 14.3 75.0 0.3323 (± 2.0) (± 0.14) (± 19.4) (n=21) (± 0.51) Total 216 27.0 27 0.5165 32.5 39.4 0.8929 (± 2.98) (± 6.4) (± 0.33) (± 28.1) (n=85) (± 0.95)

L. unicolor

Summer 113 28.3 36 1.3247 46.2 20.4 12.5468 (± 2.06) (± 3.7) (± 0.39) (± 30.9) (n=23) (± 12.04) 0600h 30.0 37 1.3624 42.0 6.7 7.0407 (± 2.4) (± 0.25) (± 22.3) (n=2) (± 3.95) 1200h 30.0 39 1.5119 79.7 0 27.2234 (± 4.6) (± 0.52) (± 20.4) (± 12.19) 1800h 27.0 36 1.2999 43.0 11.1 12.0934 (± 3.7) (± 0.37) (± 21.3) (n=3) (± 6.79) 2400h 26.0 34 1.0908 15.6 69.2 2.4364 (± 2.0) (± 0.18) (± 19.8) (n=18) (± 3.12) Winter 100 25.0 41 1.6372 35.2 17.0 4.1592 (± 4.08) (± 6.6) (± 0.76) (± 22.3) (n=17) (± 3.65) 0600h 25.0 40 1.5272 24.6 24.0 2.6154 (± 8.0) (± 0.86) (± 15.5) (n=6) (± 2.21) 1200h 30.0 43 1.8813 46.2 6.7 5.5865 (± 5.0) (± 0.70) (± 24.2) (n=2) (± 4.19) 1800h 25.0 42 1.7010 36.8 24.0 4.9385 (± 6.9) (± 0.81) (± 22.4) (n=6) (± 4.12) 2400h 20.0 38 1.3291 30.0 15.0 2.9742 (± 5.2) (± 0.51) (± 19.9) (n=3) (± 2.53) Total 213 26.6 39 1.4714 41.0 18.8 8.6090 (± 3.46) (± 5.7) (± 0.61) (± 27.7) (n=40) (± 10.02) Chapter 4. Diel variation in food intake and diet composition 76

Table 4.2 Continued.

Species and Mean time- Mean SL Mean body Mean % near Season Total N sample N (mm) weight (g) fullness (%) empty Mean RCV

N. erebi

Summer 110 27.5 64 4.0793 55.6 21.8 9.7009 (± 1.73) (± 7.2) (± 1.51) (± 31.2) (n=24) (± 9.57) 0600h 27.0 62 3.5442 64.1 0 9.0599 (± 7.7) (± 1.47) (± 14.5) (± 7.85) 1200h 27.0 68 4.7446 80.0 0 19.1549 (± 6.3) (± 1.43) (± 17.8) (± 7.91) 1800h 30.0 66 4.3185 67.0 0 9.8405 (± 6.9) (± 1.60) (± 19.7) (± 8.47) 2400h 26.0 62 3.6679 8.3 92.3 0.3879 (± 6.3) (± 1.25) (± 6.9) (n=24) (± 0.40) Winter 102 25.5 73 6.4076 34.8 32.4 2.0213 (± 3.11) (± 12.7) (± 2.99) (± 25.9) (n=33) (± 2.57) 0600h 21.0 60 3.7007 12.6 71.4 0.3269 (± 14.3) (± 2.98) (± 14.7) (n=15) (± 0.63) 1200h 27.0 76 6.7217 59.1 0 4.3075 (± 9.9) (± 2.67) (± 18.5) (± 3.35) 1800h 26.0 78 7.6495 47.5 3.8 2.6552 (± 9.6) (± 2.57) (± 18.1) (n=1) (± 1.87) 2400h 28.0 76 6.9818 16.1 60.7 0.4987 (± 9.5) (± 2.50) (± 14.9) (n=17 (± 0.39) Total 212 26.5 69 5.1995 45.6 26.9 6.0060 (± 2.56) (± 11.1) (± 2.61) (± 30.5) (n=57) (± 8.08)

There was a significant interaction between fish species and season for stomach fullness (F=5.64; df=2,635; p=0.0037). Although fish species had no effect on stomach fullness during the winter samples (F=0.64; df=2,635; p=0.5302), a pronounced effect of fish species on stomach fullness was evident during the summer (F=16.54; df=2,635; p<0.0001). In particular, stomach fullness during the summer for A. agassizii (fullness=33.9%) was significantly lower than for L. unicolor (fullness=46.2%) (p=0.0011,<0.0056 after Bonferroni correction) and N. erebi (fullness=55.6%) (p<0.0001, <0.0056 after Bonferroni correction) during the same season. However, no significant differences in stomach fullness were observed for any of the fish species during the winter (see Figure 4.3a for p values and Bonferroni correction).

Chapter 4. Diel variation in food intake and diet composition 77

(a) A. agassizii 100 100 90 90 80 80 70 70 60 60 50 50 40 40

Frequency (%) 30 Frequency (%) 30 20 20 10 10 0 0 0600h 1200h 1800h 2400h 0600h 1200h 1800h 2400h summer winter (b) L. unicolor 100 100 90 90 80 80 70 70 60 60 50 50 40 40

Frequency (%) 30 Frequency (%) 30 20 20 10 10 0 0 0600h 1200h 1800h 2400h 0600h 1200h 1800h 2400h summer winter (c) N. erebi 100 100 90 90 80 80 70 70 60 60 50 50 40 40

Frequency (%) 30 Frequency (%) 30 20 20 10 10 0 0 0600h 1200h 1800h 2400h 0600h 1200h 1800h 2400h summer winter Time of day

Stomach fullness classes

0 - 20% 20 - 40% 40 - 60% 60 - 80% 80 - 100%

The interaction was also evident when considering seasonal variation, although the mean stomach fullness was similar across seasons only for A. agassizii (summer =33.9%, winter =31.2%), while for L. unicolor and N. erebi the difference between the two seasons was substantial. Mean stomach fullness for L. unicolor was significantly greater during the summer (=46.2%) than during the winter (=35.2%) (p=0.0046, <0.0056 after Bonferroni correction). Similarly, N. erebi had significantly greater fullness during the summer (=55.6%) than during the winter (=34.8%) (p<0.0001, <0.0056 after Bonferroni correction) (Figure 4.3b).

Similarly to the fullness values, the mean relative stomach content volume also varied significantly between seasons and species (two-way ANOVA, F=82.72; df=5,635; p<0.0001), and the percentage of variation that could be attributed to the effect of the two factors, individually and in combination, was low (r2=0.3944). There was a significant interaction between fish species and seasons in their influence on RCV (F=21.02; df=2,635; p<0.0001). As opposed to the stomach fullness, fish species clearly had a pronounced effect on RCV during both seasons (F=124.96; df=2,635; p<0.0001 and F=32.96; df=2,635; p<0.0001, summer and winter respectively).

During the summer the RCV for A. agassizii (RCV=0.91) was significantly lower than the RCV for L. unicolor (RCV=12.54) (p<0.0001, <0.0056 after Bonferroni correction) and N. erebi (RCV=9.70) (p<0.0001, <0.0056 after Bonferroni correction). However, no significant difference in RCV was observed between L. unicolor and N. erebi (p=0.0063, >0.0056 after Bonferroni correction). Winter RCV values were significantly different across all three species, with L. unicolor having the greatest mean RCV (=4.15) and A. agassizii (RCV=0.87) having the lowest (see Figure 4.3c for p values and Bonferroni correction). Ambassis agassizii had similar mean RCV levels across seasons (RCV=0.91 and 0.87) (p=0.6526, >0.0056 after Bonferroni correction), while L. unicolor and N. erebi had significantly different mean RCV across seasons. Both species presented greater values of RCV during the summer (RCV=12.54 and 9.70, respectively) than during the winter (RCV=4.15 and 2.02, respectively) (see Figure 4.3d for p values and Bonferroni correction).

Chapter 4. Diel variation in food intake and diet composition 79

60 60 <0.0001 (a)<0.0001 (b) B H 55 55

50 50 0.0122 0.0046 B J 45 45

40 40

Stomach fullness (%) 0.0011 0.9102 0.3620 Stomach fullness (%) J J 0.4896 HJ 35 B 35 B 0.3073J B 30 30 A. agassizii L. unicolor N. erebi summer winter Species Season 14 14 (c) <0.0001 (d) 0.0063B J 12 12 <0.0001 <0.0001 10 B 10 H

8 8

6 6 <0.0001 4 J 4 J

Relative content volume 0.0024 Relative content volume 2 <0.0001 J 2 0.6526 H BJ BB <0.0001 0 0 A. agassizii L. unicolor N. erebi summer winter Species Season

B summer J winter B A. agassizii H N. erebi J L. unicolor

Factorial ANOVA performed on stomach fullness of A. agassizii at different times of day per season revealed significant differences across seasons and times of day combinations (two-way ANOVA, F=30.70; df=7,208; p<0.0001) and that the percentage of variation explained by the effect of the two factors (season and time), individually and in combination, was high (r2=0.5082). Significant interaction between time of day and season in their influence on stomach fullness was evident (F=11.97; df=3,208; p<0.0001), so that time of day had a significant effect on stomach fullness in both studied seasons, summer (F=25.82; df=3,208; p<0.0001) and winter (F=45.50; df=3,208; p<0.0001).

Chapter 4. Diel variation in food intake and diet composition 80

The results also show that the stomach fullness for A. agassizii was higher during the midday (1200h) summer sample (=59.5%) and during midday (=50.0%) and late afternoon (1800h) (58.2%) samples in winter, with no significant difference between the last two (p<0.1465, <0.0032 after Bonferroni correction) (see Figure 4.4a for additional pairwise comparisons and p values). Differences in time of day fullness between seasons were observed only during the 0600h and 1800h sampling occasions (Bonferroni correction p<0.0001 and p=0.0007 <0.0032, respectively).

The same analysis performed on RCV produced similar results, showing that during the summer, RCV was higher during the 1200h sample (RCV=1.62) whereas during the winter the RCV was higher at the 1200h (RCV=1.33) and 1800h (RCV=1.76) samples, although there was no significant difference in mean RCV between the 1200h and 1800h time samples (p=0.0520, <0.0032 after Bonferroni correction) (Figure 4.4a). Seasonal variation in RCV also showed results similar to stomach fullness, with significant variation between summer and winter observed only at the 0600h and 1800h time samples (in both cases p=0.0001, <0.0032 after Bonferroni correction).

For L. unicolor, factorial ANOVA showed that stomach fullness also varied significantly across the different combinations of times of day and season (two-way ANOVA, F=23.37; df=7,205; p<0.0001) and that the percentage of variation explained by the effect of seasons and times of day, individually and in combination, was high (r2=0.4438). A significant interaction between time of day and season was observed in their influence on stomach fullness (F=11.83; df=3,205; p<0.0001), and time of day had a pronounced effect on mean stomach fullness in both studied seasons, summer (F=44.43; df=3,205; p<0.0001) and winter (F=5.31; df=3,205; p=0.0015).

During the summer, mean fullness of L. unicolor stomachs was significantly greater (=79.7%) at the 1200h time, with the lowest fullness being observed at the midnight sample (=15.6%) (see Figure 4.4b for p values). Dusk and dawn samples presented intermediary mean values (=43.0 and 42.0%, respectively), not being significantly different from each other (p=0.8630, >0.0032 after Bonferroni correction). As for the winter samples, only 0600h (=24.6%) and 1200h (=46.2%) were significantly different (lower) from their counterparts during the summer (Bonferroni correction <0.0032, p<0.0001 and p=0.0025, respectively). During the winter, 1200h presented the highest mean stomach Chapter 4. Diel variation in food intake and diet composition 81 fullness (=46.2%), although this was not significantly different from the 1800h and 2400h samples (Bonferroni correction <0.0032, p=0.1011 and 0.0083, respectively). Similar results were observed from the analysis performed on RCV. There was a significant difference across the combinations of times of day and season (F=35.01; df=7,205; p<0.0001), with high percentage of the variation (r2=0.5445) explained by the effect of the two factors, individually and in combination, and with significant interaction between times of day and season on RCV (F=15.51; df=3,205; p<0.0001). Time of day also had a pronounced effect on mean RCV in both seasons, summer (F=55.98; df=3,205; p<0.0001) and winter (F=4.33; df=3,205; p=0.0055).

The summer and winter results on mean RCV of L. unicolor strengthened patterns observed for stomach fullness, showing midday as the peak time of feeding (RCV=27.22), followed by dusk and dawn (RCV=12.09 and 7.04, respectively), the latter two not differing significantly (p=0.0143, >0.0032 Bonferroni correction). The winter results of RCV presented consistently lower values in comparison with their correspondents in summer for all time periods (p<0.0001, <0.0032 after Bonferroni correction), except for 2400h, which did not vary significantly from summer to winter (p=0.1590, >0.0032 Bonferroni correction). During the winter, mean RCV levels were higher at the 1200h sample (RCV=5.58), although not significantly different from the 1800h (RCV=4.93) and 2400h (RCV=2.97) time periods (Figure 4.4b).

Factorial ANOVA on N. erebi stomach fullness also showed that mean fullness varied considerably across the different combinations of times of day and seasons (two-way ANOVA, F=78.12; df=7,204; p<0.0001) and that the percentage of variation explained by the effect of the two factors, individually and in combination, was high (r2=0.7283). Interaction between time of day and season also had a significant influence on stomach fullness (F=28.12; df=3,204; p<0.0001), and time of day had a pronounced effect on mean stomach fullness in both seasons, summer (F=101.90; df=3,204; p<0.0001) and winter (F=51.17; df=3,204; p<0.0001) (Table 4.6a).

Mean stomach fullness of N. erebi during the summer was significantly higher at the 1200h time period (=80.0%), whereas the lowest fullness was observed at midnight (=8.3%) (see Figure 4.4c for pairwise comparison p values). Dusk and dawn mean fullness levels were similar (p=0.4966, >0.0032 Bonferroni correction), and relatively high (=67.0 Chapter 4. Diel variation in food intake and diet composition 82 and 64.1%, respectively), although they were significantly lower than the midday sample (p<0.0001, <0.0032 Bonferroni correction). Mean stomach fullness for the winter samples was significantly lower then during the summer (Bonferroni correction, p<0.001 <0.0032) for N. erebi, except from the midnight sample which was not significantly different from its summer correspondent (p=0.0784, >0.0032 after Bonferroni correction). During the winter, 1200h presented the highest mean fullness (=59.1%), although it was not significantly higher than the 1800h sample (=47.5%) (p=0.0100, >0.0032 Bonferroni correction). Dawn and midnight fullness were lowest during the winter with a mean fullness of 12.6% and 16.1%, respectively (Figure 4.4c).

Analysis on RCV of N. erebi showed that variation across the combinations of times of day and season was significant (F=92.52; df=7,205; p<0.0001), with a high percentage of the variation (r2=0.7604) explained by the effect of the two factors, individually and in combination, and with significant interaction between time of day and season on RCV (F=31.04; df=3,205; p<0.0001). Time of day also had a pronounced effect on mean RCV in both seasons, summer (F=115.99; df=3,205; p<0.0001) and winter (F=35.00; df=3,205; p=0.0055).

The results for RCV analysis of N. erebi also show that, during the summer, the highest mean RCV was observed during the midday sample (RCV=19.15), with midnight presenting the lowest mean RCV (=0.38) (see Figure 4.4c for p values of pairwise comparisons). Similar to the stomach fullness analysis, the winter mean values of RCV per time of day were significantly lower than during the summer (p<0.0001, <0.0032 after Bonferroni correction), except at 2400h which did not vary between seasons (p=0.5842, >0.0032 Bonferroni correction). The winter samples of mean RCV also showed a similar pattern to the one found for stomach fullness, where midday had the highest mean value (=4.30), followed by 1800h (=2.65), although this difference was not significant (p=0.0295, >0.0032 Bonferroni correction).

Chapter 4. Diel variation in food intake and diet composition 83

(a) A. agassizii 100 2.0 90 1.8 * 80 1.6 * 70 1.4 * 60 * 1.2 0.0520 50 1.0 <0.0001 <0.0001 40 0.1465 0.8 <0.0030 <0.0001 30 0.0006 0.6 <0.0001 Stomach fullness (%) 20 0.4 Relative content volume 10 0.2 <0.0001 <0.0001 <0.0001 <0.0001 0 0.0 0600h 1200h 1800h 2400h 0600h 1200h 1800h 2400h Time of day Time of day (b) L. unicolor 100 30 90 * 80 * 25 70 20 60 <0.0001 <0.0001 50 * 15 <0.0001 40 0.0083 10 30 0.1011 <0.0001 Stomach fullness (%) *

20 Relative content volume 0.0002 5 <0.0001 10 0.2677 <0.0001 0.0012 0.0122 0 0 0600h 1200h 1800h 2400h 0600h 1200h 1800h 2400h Time of day Time of day (c) N. erebi 100 20 * 90 18 * 80 16 0.0004 70 0.0028 14 60 * 12 <0.0001 <0.0001 50 10 40 0.0100 8 30 6 <0.0001 Stomach fullness (%) 20 <0.0001 4 * Relative contentvolume 0.0295 10 2 <0.0001 <0.0001 <0.0001 <0.0001 0 0 0600h 1200h 1800h 2400h 0600h 1200h 1800h 2400h Time of day Time of day

summer winter

Chapter 4. Diel variation in food intake and diet composition 84

4.3.2 Diel variation in diet composition

The overall contribution by volume or frequency of occurrence of dietary items varied markedly between the three species studied (Table 4.3). Detritus and microcrustaceans were the only food categories ingested by all three species, with detritus being ingested by 6.4% (A. agassizii) to 89.1% (N. erebi) of individuals, and microcrustaceans being ingested by 11.3% (N. erebi) to 84.5% (L. unicolor) of all individuals. Although not used in any further analysis, unidentified matter made consistent contributions by frequency of occurrence to the stomach contents of most of the species studied. Items of this category were present in 26.4% of the stomachs of A. agassizii and in 54.9% of the stomachs of L. unicolor (Table 4.3). On the other hand, overall volumetric contributions of unidentified matter varied from insignificant (0.27% in N. erebi) to small (10.3% in L. unicolor).

From Table 4.3, it can be observed that microcrustaceans were the main food category ingested by A. agassizii, having significant contributions both in volume (=90.7%) and frequency of occurrence in stomachs (=81.9%). Of this category, nearly 90% of the volume can be attributed to calanoids and cladocerans (Moinidae), which were also present in about 50% of the stomachs analysed. The next most important identifiable category for A. agassizii was detritus, with merely 1.9% of the overall volume and being present in 6.4% of the stomachs analysed. Microcrustaceans were also the most important prey item for L. unicolor, comprising 53% of the total volume and being ingested by 84.5% of the individuals analysed, although in this species calanoids were less abundant (volume =4.5%). Aquatic insects were also important items consumed by L. unicolor (volume =30.3%), occurring in stomachs of 65.7% of studied individuals. Of the aquatic insects, main contributions to the overall diet were by chironomids (pupae and larvae), which jointly contributed to about 73% of the total of aquatic insects ingested. Detritus and vegetable matter, although not being important contributors by volume to the diets of L. unicolor were present in 23.4% and 12.2% (respectively) of the stomachs. In contrast, these two categories were amongst the most frequently ingested food items by N. erebi. Detritus corresponded to 97.1% of the total volume of items ingested by this species, and was present in 89.1% of the stomachs. Despite having contributed insignificantly to the overall volume, algae and vegetable matter were ingested by 52.3% and 33.0% (respectively) of individuals of this species. Chapter 4. Diel variation in food intake and diet composition 85

Major categories A. agassizii L. unicolor N. erebi and dietary ite ms % Vol % Freq % Vol % Freq % Vol % Freq Detritus 1.96 6.48 5.79 23.47 97.19 89.15 Algae - - - - 0.87 52.36 Volvox colony - - - - 0.30 47.17 Filamentous algae - - - - 0.01 1.89 Algae matter - - - - 0.56 13.68 Vegetable matter - - 0.39 12.21 0.93 33.02 Microcrustaceans 90.76 81.94 53.09 84.51 0.74 11.32 Calanoida (Copepoda) 43.27 46.30 4.55 42.25 0.23 5.19 Cyclopoida (Copepoda) 0.23 0.93 0.32 6.10 - - Moinidae (Cladocera) 37.52 56.02 39.62 46.48 0.30 5.66 Daphniidae (Cladocera) 9.34 19.44 5.51 29.11 0.20 4.72 Bosminidae (Cladocera) 0.02 0.46 - - - - Chydoridae (Cladocera) - - 0.07 1.88 - - Sididae (Cladocera) 0.37 1.39 - - - - Ostracoda - - 2.86 21.13 0.01 0.47 Conchostraca - - 0.18 4.69 - - Aquatic insects 0.67 1.85 30.31 65.73 - - Leptophlebiidae (Ephemeroptera) - - 0.43 3.76 - - Coenagrionidae (Odonata) - - 1.65 4.23 - - Notonectidae (Hemiptera) - - 0.08 0.47 - - Corixidae (Hemiptera) - - 4.34 21.13 - - Chironomidae (Larvae) 0.29 1.39 16.15 52.58 - - Chironomidae (Pupae) 0.37 0.46 6.49 21.60 - - Ecnomidae (Trichoptera) - - 0.98 6.10 - - Leptoceridae (Trichoptera) - - 0.17 0.94 - - Hydroptilidae (Trichoptera) - - 0.01 0.47 - - Molluscs - - 0.06 0.94 - - Planorbidae - - 0.06 0.94 - - Other invertebrates 0.04 - 0.001 0.47 - - Nematoda 0.04 0.46 0.001 0.47 - - Unidentified 6.58 26.39 10.36 54.93 0.27 4.25 Insect fragment 0.25 0.46 1.57 14.55 0.07 3.30 Digested zooplankton 1.64 4.63 0.06 0.47 0.18 0.47 Digested matter 4.68 22.22 5.18 39.91 0.02 0.47 Unidentifiable matter - - 3.55 13.62 - - Total number of stomachs 216 213 212 Percentage of stomachs containing food 84.72 94.37 90.09

Chapter 4. Diel variation in food intake and diet composition 86

When the volumetric dietary data for A. agassizii recorded for each of the four times of day in each season was subjected to cluster analysis, the classification dendrogram (Figure 4.5a) showed two distinct clusters in diet composition, corresponding to the summer and winter samples. Although no clustering indicated differences between times of day for A. agassizii during the summer, the winter presented a readily identifiable cluster of samples representing the 1200h dietary composition. The discreteness of these groups can be observed in the NMS ordination plot (Figure 4.5b). The two-dimensional ordination explained 98% of the variation in the original, unreduced space, resulting in a very low stress value of 5.4%. Of the explained variation, 89% can be assigned to the first axis alone. Nevertheless, Figure 4.5b indicates a seasonal gradient along the axis 1, from summer (at the left) to winter (middle and right). Ordination, thus, revealed that there was considerable overlap in dietary composition during the summer, whereas for the winter samples, a consistent segregation of the midday samples from the other times of day was evident, and some segregation can also be observed for the 1800h samples.

Figure 4.6 shows the percentage contribution by volume of all dietary items per season and time of day, and the dietary breadth for A. agassizii. Once again the dietary composition varied more markedly between seasons than between times of day. The diet of A. agassizii during the summer consisted mostly of moinid cladocerans (volumetric range =52-92%) and, secondly, calanoids (volumetric range =5-17%), regardless of the time of day, though during the midnight samples Moinidae abundance was lower (volume =52%) and pupae of chironomids were more abundant (volume =25%). During the winter A. agassizii fed primarily on calanoids (volumetric range =49-84%) and secondly on copepods (mainly Daphniidae, volumetric range =7-50%), though the 1200h sample had a high volume of Moinidae (volume =37%). More importantly, the dietary breadth (B), measured by the Shannon-Wiener diversity index (Figure 4.6) showed that during the 1200h period, dietary composition was more diverse for both seasons (summer =0.6 and winter =0.9), although the winter samples showed an overall higher diversity of items ingested by A. agassizii. Interestingly, the 2400h period for the summer also showed a high diversity (B =0.8), with a total 4 dietary items ingested throughout this period of time. Although valid, the consistency of this result is obscured by a high standard error. Chapter 4. Diel variation in food intake and diet composition 87

Distance (Objective Function) 5E-05 2.3E-01 4.6E-01 7E-01 9.3E-01 Information Remaining (%) 100 75 50 25 0 S12h S24h S12h (a) S18h S06h S06h S06h S18h S12h S18h S06h S18h S18h S12h S24h W06h W24h W18h W18h W18h W18h W18h W24h W12h W12h W12h W12h

W Time of day W 06h W 12h W 18h S W 24h W W S W S S S W S S S W S S S W S S W

Axis 2 S

S Stress=5.4% (b) Axis 1

100 1.0

90 0.9

80 0.8

70 0.7

60 0.6

50 0.5

40 0.4 breadth Dietary Percentage volume

30 0.3

20 0.2

10 0.1

0 0.0 0600h 1200h 1800h 2400h 0600h 1200h 1800h 2400h

summer winter Time of day

detrit cyclop bosm sidid chiron-p

calan moin daph chiron-l Dietary breadth

Cluster analysis of the volumetric dietary data for L. unicolor for times of day and season resulted in the dendrogram shown in Figure 4.7a. Similarly to the results for A. agassizii, the classification dendrogram shows two clear clusters representing summer and winter samples. The discreteness of these groups can be observed in the NMS ordination plot (Figure 4.7b). The two-dimensional ordination explained 93% of the variation observed in the original, unreduced space, thus the low ordination stress (=9.8%), and the first axis alone accounts for 60% of the explained variation. Although no clustering can be observed for times of day in either of the two seasons studied, the classification dendrogram (Figure Chapter 4. Diel variation in food intake and diet composition 89

4.7a) shows one minor cluster for the summer data, corresponding to samples from 1200 and 1800 hours, and three minor clusters of samples for the winter data, one corresponding to the diets of L. unicolor during the 1800h and 2400h periods and two corresponding to the diets during the 0600h and 1200h periods. From Figure 4.7b, an overall gradient is observable along the 1800-2400h and the 0600-1200h winter groups, and some discreteness is observed for the three clusters, where the segregation is between the 1800- 2400h cluster and the rest of the samples is identifiable.

The percentage contribution by volume of all dietary items of L. unicolor per season and time of day confirms previous results that the variation between seasons is greater than between times of day, within seasons (Figure 4.8). During the summer, the diets of L. unicolor consisted mostly of Moinidae cladocerans (volumetric range =35-74%), larvae of Chironomidae (volumetric range =11-30%) and detritus (volumetric range =5-20%). Early in the morning (0600h) Corixidae and Ostracoda were also abundant (volume =17 and 8%, respectively), but their abundances decreased after the 1200h and 1800h samples, when chironomid pupae were more abundantly consumed (volume =6 and 9%, respectively). Interestingly, the diet breadth of L. unicolor during the summer was greater at the 0600h (B =1.3) and 1200h (B =1.2) samples. As for the winter samples, Figure 4.8 indicates a markedly different composition of diet, while the dietary breadth values were relatively high when compared to those for the summer. Although 0600h, 1200h and 1800h had equally high dietary breadth values (=1.5, 1.7 and 1.6, respectively), the composition of diets was quite different, mostly between early morning-midday and late afternoon- midnight, as indicated by ordination and classification. Larval stages of chironomids were the basis of the diet of L. unicolor during the first two samples of the day (0600h and 1200h) (volume =40 and 29%), but Odonata (Coenagrionidae) (volume =27 and 8%), Daphniidae cladocerans (volume =9 and 15%), calanoids (volume =7 and 13%) and trichopterans (Ecnomidae and Hydroptilidae) (volume <7%) were also abundant at these times, the later being observed mostly during the 0600h and 1200h sampling occasions. A considerable shift can then be observed to the 1800h and 2400h dietary composition. The main difference is that Daphnia became considerably more abundant (volumetric range =31-59%), and so did the calanoids (at 1800h) (volume =35%), as the chironomids (larvae) (volumetric range =7-10%) and Odonata (volume <3%) became less important to the dietary composition. Other relatively abundant items consumed throughout all the winter Chapter 4. Diel variation in food intake and diet composition 90 samples were pupae of chironomids (volumetric range =3-12%) and hemipterans (Corixidae) (volumetric range =3-9%).

The results of classification analysis on volumetric dietary data for N. erebi for times of day and seasons were characterised by the presence of three outliers, corresponding to two 1800h and one 1200h winter samples (Figure 4.9). Average distances standard deviation for these points was higher than the cut off value (=2.0). The classification dendrogram (Figure 4.9a) also shows three other major clusters generally identifying one group comprised of 0600h samples, and two others with mostly 1200h, 1800h and 2400h samples intermingled. The discreteness of the groups formed by classification analysis can be observed in the NMS ordination plot (Figure 4.9b). Even though the two-dimensional NMS plot explained 93% of the variation found in the unreduced space, each axis individually accounted for 59% (axis 1) and 34% (axis 2) of the explained variation, resulting in a relatively high stress value of 11.8%. Therefore, the resultant NMS plot shows that the dietary composition of N. erebi was remarkably homogeneous across time of day samples and seasons, and no clear segregation in the cloud of points can be observed, apart from the outliers. Nevertheless, two main groups can be drawn from the classification dendrogram, broadly representing the 0600h samples and, the second, representing a combination of the dietary composition during the 1800h and 2400h samples. Interestingly, neither classification nor ordination revealed readily identifiable summer/winter groups for dietary composition of N. erebi.

The percentage contribution by volume of all dietary items per season and time of day (Figure 4.10) shows the clear dominance of detritus material in the diet of N. erebi (volumetric range =94-99%) (note that the percentage volume axis begins at 90%, to enable observation of the contribution of the remaining dietary items). Interestingly, the contribution of detritus increased slightly during the 2400h times, when consumption of other items decreased. The 2400h periods also had lower dietary breadth values (=0.2 and 0.3). Comparison of the summer and winter composition of diet reveals that contribution of detritus was higher during the summer (minimum volume =97%, against up to 97% during the winter), and as a result the overall values of dietary breadth were smaller compared with those of the winter (B range =0.2-0.7 in summer and 0.3-0.8 in winter). Chapter 4. Diel variation in food intake and diet composition 91

Distance (Objective Function) 5.4E-03 4.7E-01 9.4E-01 1.4E+00 1.9E+00 Information Remaining (%) 100 75 50 25 0 S12h S12h S18h (a) S12h S18h S18h S12h S12h S12h S06h S06h S06h S24h S06h S06h S18h S24h W06h W06h W12h W12h W18h W18h W18h W18h W24h W24h W24h W06h W06h W12h W12h W12h

W Time of day W W W W 06h 12h W 18h W W 24h W W W W W W W S W

Axis 2 S S

S S S S S S S S S S S S S (b)

Stress=9.8% S

Axis 1 Figure 4.7 Relative Sorensen (Bray-Curtis)/Flexible Beta (β=-0.1) dendrogram of dietary composition similarities of L. unicolor based on different times of day (06h, 12h, 18h, 24h) and seasons (S=summer and W=winter) (a); and (b) two-dimensional NMS ordination plot of dietary samples of L. unicolor from different times of day and seasons. Note that, each dietary sample represents the mean volumetric data for groups of 3-6 randomly selected individuals. Dashed lines show secondary groups identified by the classification analysis. Chapter 4. Diel variation in food intake and diet composition 92

100 1.8

90 1.6

80 1.4

70 1.2 60 1.0 50 0.8

40 Dietary breadth Percentage volume 0.6 30

0.4 20

10 0.2

0 0.0 0600h 1200h 1800h 2400h 0600h 1200h 1800h 2400h

summer winter Time of day

detrit cyclop chydo nema noton chiron-p hydrop

veg-m moin ostrac leptop corix ecnom planor

calan daphn conch coenag chiron-l leptoc Dietary breadth

However, on the whole, dietary breadth was low for N. erebi. Observation of the contribution of the remaining dietary items indicates a significant variation in their consumption. Although Volvox colonies and vegetable matter were consistently ingested throughout times of day and seasons, their overall contribution by volume was minimal (volume <3%) and it is likely that they were ingested incidentally rather than being actively sought by N. erebi. The same can be applied to the algal matter (including filamentous algae). The contribution of microcrustaceans (Moinidae, Daphniidae, Calanoida and Ostracoda) was similarly low (volume <2.4%), although they were ingested more abundantly at dusk and dawn. In the case of the Daphniidae cladocerans, the outlier samples (1800h) accounted for most of their volumetric contribution to the dietary composition of N. erebi (see insert box, Figure 4.9c). Chapter 4. Diel variation in food intake and diet composition 93

Distance (Objective Function) 2.3E-05 3.6E-02 7.2E-02 1.1E-01 1.4E-01

Information Remaining (%) 100 75 50 25 0

S12h S12h S12h S12h (a) S18h W18h S18h S06h W12h W12h W12h W12h W18h S12h S18h S18h S18h S24h W24h W24h S06h S06h S06h S06h W06h W06h W18h W12h W18h W18h

S Stress=11.8 Time of day S S S 06h S 12h S 18h S S S 24h S W W W W W W S W W W W

S S Axis 2

W S W

(b) W S

S Daphniidae

Axis 1 Figure 4.9 Relative Sorensen (Bray-Curtis)/Flexible Beta (β=-0.1) dendrogram of dietary composition similarities of N. erebi based on different times of day (06h, 12h, 18h, 24h) and seasons (S=summer and W=winter) (a); and (b) two-dimensional NMS ordination plot of dietary samples of N. erebi from different times of day and seasons. Note that, each dietary sample represents the mean volumetric data for groups of 3-6 randomly selected individuals. Dashed lines show secondary groups identified by the classification analysis. The insert box indicates the volumetric contribution of Daphniidae to the ordination. Note that the outliers account for most of the contribution of this food item. Chapter 4. Diel variation in food intake and diet composition 94

100 1.0

99 0.9

98 0.8

97 0.7

96 0.6

95 0.5

94 0.4 Dietary breadth Percentage volume

93 0.3

92 0.2

91 0.1

90 0.0 0600h 1200h 1800h 2400h 0600h 1200h 1800h 2400h

summer winter Time of day

detrit alg-m veg-m moin ostrac

volvox f-alg calan daphn Dietary breadth

4.4 Discussion

4.4.1 Diel feeding activity

Overall stomach fullness was higher for L. unicolor and N. erebi than for A. agassizii, as shown in Figure 4.3. Omnivorous and herbivorous species such as L. unicolor and N. erebi are known to have food consumption rates higher than species belonging to other trophic levels, e.g. carnivores (De Silva et al. 1996), due to the fact that plant-derived food items (in N. erebi’s case) are of low nutritive quality (De Silva 1985) therefore having to Chapter 4. Diel variation in food intake and diet composition 95 be ingested in larger quantities, and that the omnivorous nature of L. unicolor enables this species to feed on a higher number of food items.

The results of this study also demonstrate that there are differences in food intake by all study species throughout the day. On the whole, these patterns changed on a seasonal basis, but individual species presented slightly different patterns of diel and diel/seasonal variation. All of the species studied were most active during daylight, with higher feeding activity during midday and/or late in the afternoon (Table 4.5), depending on which season is being considered (see below). Overall feeding activity was higher during the summer for L. unicolor and N. erebi, but A. agassizii presented similar levels between summer and winter (Table 4.4). Despite the fact that summer days were longer (Table 4.1), feeding activity of all species during this season was mostly concentrated at midday. In contrast, feeding activity during the winter seemed to be equally high from midday until late in the afternoon (Figure 4.4). Interestingly, because winter days were shorter, the late afternoon samples were taken when sunlight was very low or absent, but still all species presented a level of feeding activity similar (but lower) to the levels observed at midday. The fact that the species studied fed intensely even shortly after sunset (in winter), indicates that food intake was extended until very low light was available during this season, even though samples may represent feeding activity well before time of capture.

Even though the sunrise varied between 0450h in summer to 0630h in winter, the small feeding activity observed early in the morning is likely to be a result of the combination of low light intensity and lower water temperatures. According to Wootton (1990) the increase in light levels associated with sunrise will only lead to increased feeding activity when a threshold of light intensity is reached, and only then will feeding intensity increase with further increases in light intensity. The low activity at midnight is obviously the result of the total absence of sunlight as none of the species studied is thought to be nocturnal (Merrick and Schmida 1984, Pusey et al. 2004).

Despite the fact that photoperiod can influence the endocrine system and indirectly cause seasonal changes in fish hunger or appetite (Keast and Welsh 1968), sunlight can have a more direct effect on their food consumption, by inducing water temperature changes, and consequently leading to alternating periods of activity and inactivity, or by reducing the ability of fish to perceive food by visual means. Because most zooplankton feeders, such Chapter 4. Diel variation in food intake and diet composition 96 as A. agassizii, locate their prey by visual means (Gerking 1994), it is not surprising that this species presented strong correlations with daylight feeding. The fact that the results show that this species fed intensely even shortly after sunset during the winter (the sunset was approximately at 1730h and the collections started only at 1800h), may simply be a result of sampling fish that had ingested food before being caught. Likewise, L. unicolor was also considered a daylight feeder, with peak feeding time at noon, although feeding activity decreased significantly from summer to winter. During the latter, midday feeding activity, though higher, was less conspicuous, remaining on a level similar to the late afternoon activity. Visual predators like L. unicolor (Gehrke 1988b) are expected to feed mostly during daylight hours as they experience some difficulty in locating their prey at night (Keenleyside 1979). According to Gehrke (1988a), L. unicolor may also be inactive at night, as captive individuals became quiescent in darkened aquaria.

The rate of food consumption in fish can also be affected by temperature, through its effects on the rate of gastric evacuation and consequently systemic demand (Wootton 1990). At low temperatures fish may cease to feed but, as temperature increases, consumption rates will also increase up to a maximum level (Elliott 1972). Despite the fact that A. agassizii showed similar feeding activity during summer and winter, this assumption was particularly accurate for the remaining species studied, given that they were less active during the winter when overall temperatures were lower (water quality data presented in Chapter 3). Despite that, A. agassizii did not show similar patterns of diel feeding activity from summer to winter, as feeding activity during the latter was mostly concentrated at noon. Even though there are no specific data on temperature tolerances for A. agassizii, this species has been collected from river sites with temperatures up to 32°C (Pusey et al. 2004). According to Elliott (1981), increases in temperature beyond an optimum level may lead to a rapid decrease in food consumption by fishes. Maximum surface water temperatures during the summer at Rainbow Lagoon, where A. agassizii was collected, reached values of 34°C at 1830h, as opposed to a maximum of 29°C at noon. This is an indication that higher temperatures during the summer may have inhibited feeding activity of A. agassizii at dusk during this season.

Results for L. unicolor, on the other hand, are in accordance with Gehrke (1988a) who reported an abrupt reduction in catches of this species during the winter and that this Chapter 4. Diel variation in food intake and diet composition 97 species may reduce food intake at low temperatures. This is likely to be the main factor explaining the diel patterns in food consumption found in this study, given that, as reported in Chapter 3, diel variation in water temperature (mostly during the summer) was observed in one of the lagoons studied (Rainbow Lagoon) and water temperatures were lower early in the morning. In contrast, previous work by Gehrke (1988a) reported a bimodality in diel catch rates of this species, with higher catch rates during early morning (around 0630h) and late afternoon (around 1630h). He suggested that feeding activity should be concentrated into similar time periods. The results of this study, on the other hand, show that catches of L. unicolor were similar throughout the day (Table 4.2) and that during the summer, feeding activity was mostly concentrated around midday, and midday to late afternoon during the winter. On the other hand, it is important to note that Gehrke (1988a) used angling techniques as his sampling method, thereby targeting much larger fishes (usually more than 100 mm in length) than the ones analysed in the present study. Given the fact that Pusey et al. (2004) described significant ontogenetic differences in diet of L. unicolor in the Burdekin River, it is likely that the differences between the present study and Gehrke’s are due to size differences in the fish sampled. The similar catch rates throughout the day observed in the present study are more likely to be a result of the combination of small depths and area of the sites sampled, rather than a reflection of fish activity.

Like the other species studied, seasonal reduction in abundances of N. erebi has been reported in previous studies (Puckridge and Walker 1990, Arthington et al. 1992a). Although these changes correspond to the period of lower temperatures (Arthington et al. 1992a), reduced abundances can result from increased infection by pathogens associated with lower temperatures during the winter (Puckridge 1989) as well as cumulative mortalities following summer recruitment (Ruello 1976). Data on stomach fullness of N. erebi showed that, despite the fact that feeding activity was higher at noon, early morning and late afternoon activity were also relatively high compared to the other species. Although water temperature may have influenced the rates of food consumption of this species, it is important to note that N. erebi fed almost exclusively on detritus (Table 4.3). Therefore, the fact that the availability of this resource would not be expected to change throughout the day may have enabled this species to forage at very low light levels.

Chapter 4. Diel variation in food intake and diet composition 98

4.4.2 Diel dietary composition

The results of dietary composition of the study species suggested that the increased rates of food consumption during midday to late afternoon did not necessarily imply an increase in the number of dietary items consumed at these times of day. Overall results (Table 4.3) showed that the study species had a relatively restricted dietary range, which is likely to be a consequence of low diversity of prey items available. Two of the species (A. agassizii and L. unicolor) are considered to be opportunistic predators. Ambassis agassizii consumed mostly microcrustaceans (calanoid copepods and cladocerans, Moinidae and Daphniidae), L. unicolor fed on microcrustaceans (mostly Moinidae and Daphniidae cladocerans and calanoids) and aquatic insects (larvae and pupae of Chironomidae, corixid hemipterans and Odonata) and that N. erebi’s diet consisted almost exclusively of detritus material. Even though algae were very frequent in this species stomach contents. It is also important to note that comparisons of the dietary composition of fish in this study with those of other investigators are limited by differences in physical characteristics of the habitats sampled, and consequently food types available, and fish sizes. A detailed study of spatial and temporal variation in dietary composition of the target species is presented in Chapter 6.

As opposed to the other species studied, the results for N. erebi did not indicate diel or seasonal variation in diet composition. This is likely to be a result of the significant specialization of this species on a readily available dietary item, i.e. detritus material. Even though individuals of this species are known to be zooplanktivorous or omnivorous when young (Atkins 1984, Bunn et al. 2003, see also Chapter 5), adults feed almost exclusively on detritus and algal material. Although some similarities between specific times of day were found across seasons, these are of low importance as they were based on a few prey items that contributed insignificantly to the overall dietary composition of this species. As observed from Figure 4.9, the dietary data were weakly structured due mostly to the presence of outliers, which were the result of a high volumetric contribution by a single unusual prey item (Daphniidae) to a few individuals, and because one single item (detritus) contributed a much greater proportion of the diet than other food types.

The composition of A. agassizii diet in the present study (Table 4.3) is only partially in accordance with previous work for the species (e.g. Arthington 1992, Pusey et al. 2004). Chapter 4. Diel variation in food intake and diet composition 99

These authors report a high contribution of aquatic insects to the diets of A. agassizii, as opposed to the overwhelming contribution of microcrustaceans in this study. This may have resulted from differences in the habitat sampled and fish size between studies. The results presented here suggest that the dietary composition of A. agassizii was influenced mostly by seasonal changes in prey composition rather than prey availability within a diel cycle, in particular during the summer, when the richness of prey items was lower and dominated by moinid cladocerans. During winter, A. agassizii tended to feed more on calanoids and daphniid cladocerans. Despite some variation in dietary composition throughout the 24h period studied during the winter, no clear trend can be observed during the summer. During the winter, the volume of Daphniidae in stomach contents was higher at midnight and early morning. Although not being very active at this time, it is likely that A. agassizii benefited from the vertical migration of zooplankton.

Differences in dietary composition were also more evident between summer and winter than between times of day for L. unicolor. Additionally, a wider breadth in dietary composition characterised winter samples. Gehrke (1988a) suggested that L. unicolor may change dietary preferences in winter, as found for other Australian species of fish (Harris 1985), although whether this is due to changes in prey composition/availability or preferences for specific prey types is unclear. The observed dietary differences between summer and winter in the present study are likely to be a result of changes in prey composition and availability as a result of recent flooding of the lagoon studied (South Callandoon A) prior to the winter sampling occasion. Although dietary composition remained similar between times of day during the summer, some differentiation was observed during the winter, as daphniid cladocerans were more abundant in the stomachs of L. unicolor during the late afternoon and midnight samples.

Vertical migration of zooplankton may well be one factor affecting the diel patterns of dietary composition for A. agassizii and L. unicolor, as the results indicate that both species presented a switch in diet composition, mainly in the consumption of daphniid cladocerans, after the midday samples. It is well known that zooplankton may undertake diel vertical migrations within water bodies (Gliwicz 1986), such as deep and clear lakes (Hutchinson 1967), where stratification is evident. Studies on shallow lakes in river floodplains (Hart and McGregor 1982, Mackey 1991, Vyverman 1994) also suggest that stratification may be present at least during part of the year, usually during the dry season Chapter 4. Diel variation in food intake and diet composition 100

(Kennard 1995) and that the vertical migration of zooplankton may be associated with this. There are no data to support that this would be the case in the sites studied here, but the low overall depths observed in some of the sites sampled indicate that vertical migration of zooplankton would be unlikely at least in the more shallow lagoons surveyed. More importantly, cladocerans can conduct diurnal lateral migrations in the littoral zone, where they tend to hide among macrophyte beds or other aquatic structures, and move into the nearby open water at night (Timms and Moss 1984).

4.5 Conclusions

The results of this exploratory study, based on stomach fullness and relative stomach content volume, showed that food intake by A. agassizii, L. unicolor and N. erebi varied within a 24-h cycle, and that these patterns changed from summer to winter. Data from this and other studies indicate that these patterns are mostly driven by a combination of changes in water temperature and light availability. Overall, food consumption was significantly higher during the summer than during the winter. As supported by general literature, this is likely to be a result of lower water temperatures leading to lower metabolic rates during this season (Harris 1985, Gehrke 1988a).

During the summer, peak feeding activity of the three species was concentrated mostly around midday, although data on stomach fullness of N. erebi indicate that early morning (0600h) and late afternoon (1800h) feeding activity was also high. During the winter, food consumption was also high during midday hours, but there was some indication that feeding activity of fish was extended until late in the afternoon. This is particularly the case for A. agassizii, which presented slightly higher food consumption in the late afternoon.

The results of dietary composition suggest that, although overall feeding activity was higher during the summer, dietary breadth was higher during the winter. This may have resulted from either higher contribution of unusual items, due to smaller overall fullness at winter, or lower availability of food items during this season. Similarly, the increased rates of food consumption during midday to late afternoon did not necessarily imply an increase in the number of dietary items consumed at these times of day. Chapter 4. Diel variation in food intake and diet composition 101

Diet composition of A. agassizii and L. unicolor varied significantly between summer and winter, probably as a consequence of changes in prey availability, even though N. erebi presented very similar composition of diets between the two seasons, due to its highly specific diet based on readily available food (detritus material). Even though there was no indication of diel variation in diet composition of N. erebi, A. agassizii and L. unicolor showed an increased consumption of daphniid cladocerans during darker times of the day, which may be a consequence of diel migration of zooplankton and their greater availability as prey.

4.6 Implications for the study of fish diet composition

The results of this study indicate that sampling of A. agassizii, L. unicolor and N. erebi to determine their exploitation of available food resources should be concentrated at midday hours (1130 to 1300 hours) during the summer and midday (1130 to 1300 hours) to late afternoon (1730 to 1900 hours) during the winter. These are the times of day when food consumption was greatest, as demonstrated by stomach fullness and relative stomach content volume. Given the resolution of time intervals chosen in this study, it is unknown if any of the species feed continuously between the midday and late afternoon periods.

With regard to the dietary breadth and composition, although midday samples presented a relatively high number of food items ingested, it is not clear if this was the result of higher number of prey items being consumed at midday hours or simply higher contribution of unusual items during other times, as stomach fullness was higher during midday to late afternoon hours.

Chapter 5. Ontogenetic shifts in diet composition 102

5 Variation in diet composition of fish among different size classes in floodplain lagoons

5.1 Introduction

Differences in resource use are not only observed between species, but within species between ontogenetic intervals (Werner and Gilliam 1984) as almost all fish species change trophically during ontogeny (Miller 1979b). Several studies on fish dietary ecology have observed that ontogeny influences diet composition (Ross and Baker 1983, Gerking 1994, Pusey et al. 1995, 2000) and stressed the importance of ascertaining ontogenetic changes in studies on the feeding ecology of fish (Persson and Greenberg 1990, Piet 1998). Many Australian freshwater fish also present pronounced ontogenetic variation in dietary composition. Kennard et al. (2001) showed that a progression from juvenile to adult was usually accompanied by a general increase in the diversity of prey items, as well as in the consumption of larger items, particularly for carnivorous species. Van Valen (1965) argued that such variations in diet reflect the adaptation of individuals to different micro- environments. This hypothesis was further advanced by Polis (1984) who attributed the ontogenetic (within-individual) component of diet variation to niche changes.

Ontogenetic shifts in taxonomic composition and size range of prey exhibited by fish have been attributed to increasing body size, changes in morphology of the feeding apparatus, habitat or behavioural shifts, or to age of the predator (Werner and Hall 1977, Grossman 1980, Schmitt and Holbrook 1984). Other changes associated with increased body size, such as increased swimming speed and greater mouth gape, can also substantially increase the array of prey items suitable to a predator as it grows (Werner 1974, Keast 1985). Visual acuity and reaction distance to prey also increase with increasing predator size (Breck and Gitter 1983, Dunbrack and Dill 1983). Diets will also vary with developmental stage (larvae, juvenile and adult) as fish leaving the embryonic phase and reliance upon the egg yolk as their main energy source will switch to exogenous feeding (Schlosser 1991). From then onward, trophic interactions (resource use and predator-prey interactions) and food availability become particularly important (Wilbur 1980).

Chapter 5. Ontogenetic shifts in diet composition 103

Ontogenetic variation has also been shown experimentally to depend on interspecific competition, as the increase in density of one species may induce a shift in diet composition in a competing species (Persson and Greenberg 1990), and on spatial heterogeneity as a shift in diet was found to depend on contrasting availability of food resources in different habitats (Garcia-Berthou 1999). The effect of spatio-temporal habitat variation in food availability on size-related diet shifts is further emphasized by Winemiller (1989), who argued that switching from invertebrate to fish prey coincided with changing food availability related to seasonal variation in abundance of prey items in the Venezuelan ilanos.

5.1.1 Aims

Given that ontogenetic changes in diet related to increasing body size may overwhelm the detection of any potential temporal and spatial variation associated with fluctuating prey abundance and availability (Pusey et al. 2000), this chapter will examine possible size- related changes in diet composition of Ambassis agassizii, Leiopotherapon unicolor and Nematalosa erebi in floodplain lagoons of the Macintyre River.

As ontogenetic patterns in food consumption may be confounded by seasonal and spatial trends in prey availability, this study also aims to compare diets of similar size individuals of the same species in different floodplain lagoons and/or seasons. Information gained in this chapter is expected to generate a more precise understanding of the influence of size- related differences in diets of the target species, to be used in the next chapter on temporal and spatial changes in diet composition. Specific research questions addressed are:

1. Is there ontogenetic (size-related) variation in the dietary composition of A. agassizii, L. unicolor and N. erebi?

2. Do similar size classes of individuals of each fish species present similar composition of food items across different lagoons and/or seasons?

Chapter 5. Ontogenetic shifts in diet composition 104

5.2 Methods

Evaluation of ontogenetic variation in diet composition of the species studied was performed in lagoon sites and/or seasons where enough individuals were available for meaningful statistical analysis of individual size classes (see Table 5.1). Even though there is some evidence from the literature that L. unicolor and N. erebi undergo ontogenetic changes in diet (Atkins 1984, Bunn et al. 2003, Pusey et al. 2004), size classes of each species were chosen based on the range of sizes available in this study (based on total length, TL) and the number of individuals in each size class. Results from the previous chapter (Chapter 4) indicated that there is some seasonal variation in diets of all species studied (see also Chapter 6) and, to a degree, spatial variation. Therefore, whenever possible, samples from different seasons and/or sites were kept separate for analysis.

Fish sampling aimed at a wide range of sizes was performed throughout the entire study period. Each lagoon was sampled using a seine net (25 m length x 2 m height x 1 cm mesh) from approximately 11:00 am to 2:00 pm. Based on results from Chapter 4, sampling of A. agassizii, L. unicolor and N. erebi to determine their exploitation of available food resources should be concentrated at midday hours (1130 to 1300 hours), as this is the time of day when food consumption was greatest, and the dietary breadth was wider.

Samples were taken at random points along the margins (maximum depth of 1.5 m), as these areas generally contained the majority of fish and because seining becomes impractical at water depths greater than 1.5 metres. Individual seine-transects were approximately 10 m long and 5 to 15 m wide depending on the lagoon depth. The duration of each seine haul was about 5 minutes with 5 to 10 minutes between seines to allow time to sort the fish caught and remove woody debris from the net. This procedure was repeated a minimum of three times along the lagoon margins until a maximum of 30 individuals of different size classes of each species was caught for each sampling occasion. Once the maximum number of individuals (n=30) of a given size class of a given species was achieved, any additional individuals of that species were returned unharmed to the point of capture, as well as any individuals of unwanted species.

Chapter 5. Ontogenetic shifts in diet composition 105

In the field, the number of individuals for each size class was taken based on the overall abundance of each species. Due to discrepancies in size classes of all three studied species between sampling occasions, at each collection a range of individuals with different sizes was sought and size classes were determined later in the laboratory (Table 5.2). All individuals of the target species were preserved whole in 4% buffered formalin immediately after capture. Later in the laboratory, fish were transferred to 70% ethanol and the necessary measurements were performed.

The dietary analysis of the stomach contents was performed as described in Section 4.2.3 of Chapter 4. The proportional contribution of each class of items to the diet of individual fish was estimated using Hyslop’s (1980) indirect volumetric method, where stomach items within each prey category are lumped together and squashed to a uniform depth of 1 mm and the area covered taken to represent the volumetric contribution of that food category to the total stomach contents in mm3 (cf. Arthington 1992, Pusey et al. 2000). Prey items were identified to the lowest level of taxonomic resolution (see Section 4.2.3) to mach items and food categories listed in Table 4.3.

5.2.1 Data analysis

The contribution of each dietary item to the diet of any given size class of each species was expressed as the mean of the percentage volumetric contribution made by each dietary item to the stomach contents of each individual of each size class and each species. The unidentified material category was excluded from the analysis, since the inclusion of values of unidentified material might bias the results of multivariate analysis (Platell and Potter 2001) and, from the results of Chapter 4, the overall contribution to the diets of each species was expected to be small. For all subsequent statistical analysis on size class diet composition, only stomachs with fullness equal or greater than 20% were considered, in order to prevent bias in the calculation of the proportional contribution of prey or diversity indices (Pusey et al. 1995) (see also Section 4.2.4.2 in Chapter 4).

To understand whether variations in the dietary composition of fish were influenced by differences in size among different sampling occasions, the proportional volumetric dietary data for individual representatives of each species from each sampling occasion were Chapter 5. Ontogenetic shifts in diet composition 106 allocated into size classes based on the range of sizes available (TL) and the number of individuals of each size class, and the mean diet composition calculated for each of these groups (Schafer et al. 2002). These data were also used to calculate the mean Shannon- Wiener diversity index of prey for each size class of each species on each sampling occasion, this index corresponding to dietary breadth (B) (Marshall and Elliott 1997).

Overall similarities in dietary composition of size classes within and between sampling occasions for each species were analysed using ordination (McCune and Mefford 1999). Non-metric Multidimensional Scaling (NMS) was used for the evaluation and interpretation of how discrete size classes varied in diet composition across sampling occasions and species. Prior to ordination, the mean proportional contribution by volume of food types to the diet of the size classes on different sampling occasions was arcsine square root transformed, as is appropriate for proportional multivariate data (McCune and Grace 2002) and an outlier analysis was performed to check for outliers and skewness (McCune and Mefford 1999). The Relative Sorensen (Kulcynski) distance was used for the analysis and all graph outputs were Varimax-rotated. The coefficient of determination (r2) between Euclidean distances in the ordination space and distances in the original space was used as an after-the-fact method for calculating proportion of variance represented by ordination axes (McCune and Grace 2002). This method is useful to evaluate the quality of data reduction and, therefore, how well the distances in the ordination space represent the distances in the original, unreduced space. One-way ANOVA of arcsine transformed data was used to further investigate the variance in proportional contribution of specific dietary items between size classes.

5.3 Results

Of the several floodplain lagoons studied, ontogenetic data were available for two to four locations and at least two seasons, depending on the species (Table 5.1), and some degree of ontogenetic shift in diet was evident for all three species in the relative position of sequential size class samples in ordination space.

Chapter 5. Ontogenetic shifts in diet composition 107

Size-class (mm) TL Species Site (approx. SL) n Dietary Breadth A. agassizii Rainbow Lg. S1 35-40 0 - (26-29) Winter / 2002 S2 41-45 9 0.4927 (30-33) S3 46-50 13 0.5059 (34-37) S4 51-55 14 0.4824 (38-41) S5 56-60 8 0.3476 (42-45) Rainbow Lg. S1 35-40 4 0.3457 (26-29) Summer / 2002 S2 41-45 56 0.2714 (30-33) S3 46-50 47 0.1259 (34-37) S4 51-55 4 0.0564 (38-41) S5 56-60 0 - (42-45) Rainbow Lg. S1 35-40 9 0.0201 (26-29) Winter / 2003 S2 41-45 18 0.0210 (30-33) S3 46-50 10 0 (34-37) S4 51-55 3 0.0102 (38-41) S5 56-60 0 - (42-45) L. unicolor South Callandoon Lg. A S1 30-39 11 0.7675 (24-31) Winter / 2003 S2 40-49 68 0.8420 (32-39) S3 50-59 46 0.8202 (40-48) S4 60-69 13 0.8292 (49-56) S5 70-79 0 - (57-64) S6 80-89 0 - (65-72) South Callandoon Lg. B S1 30-39 4 0.6585 (24-31) Summer / 2002 S2 40-49 111 0.6755 (32-39) S3 50-59 35 0.7301 (40-48) S4 60-69 0 - (49-56) S5 70-79 0 - (57-64) S6 80-89 0 - (65-72)

Chapter 5. Ontogenetic shifts in diet composition 108

Table 5.1 Continued.

Size-class (mm) TL Species Site (approx. SL) n Dietary Breadth L. unicolor South Callandoon Lg. B S1 30-39 0 - (24-31) Winter / 2002 S2 40-49 0 - (32-39) S3 50-59 18 0.7465 (40-48) S4 60-69 48 0.6439 (49-56) S5 70-79 14 0.6903 (57-64) S6 80-89 6 0.6801 (65-72) N. erebi Punbougal Lg. S1 40-49 0 - (32-38) Summer / 2002 S2 50-69 29 0.0427 (39-53) S3 70-89 23 0.0227 (54-69) S4 90-109 3 0.0132 (70-84) S5 110-129 22 0.0383 (85-99) S6 130-139 7 0.0340 (100-107) Maynes Lg. S1 40-49 0 - (32-38) Summer / 2002 S2 50-69 30 0.4131 (39-53) S3 70-89 9 0.0144 (54-69) S4 90-109 8 0.0125 (70-84) S5 110-129 11 0.0216 (85-99) S6 130-139 0 - (100-107) Rainbow Lg. S1 40-49 3 0.4971 (32-38) Summer / 2003 S2 50-69 23 0.5891 (39-53) S3 70-89 9 0.0381 (54-69) S4 90-109 3 0.0266 (70-84) S5 110-129 0 - (85-99) S6 130-139 0 - (100-107) South Callandoon Lg. A S1 40-49 0 - (32-38) Summer / 2003 S2 50-69 20 0.5979 (39-53) S3 70-89 16 0.1570 (54-69) S4 90-109 4 0.1410 (70-84) S5 110-129 0 - (85-99) S6 130-139 0 - (100-107)

Chapter 5. Ontogenetic shifts in diet composition 109

5.3.1 Ambassis agassizii

The discreteness in similarities of proportional volumetric dietary data for A. agassizii recorded for sequential size classes in different seasons is shown in the NMS ordination plot (Figure 5.1). The two-dimensional ordination explained 99% of the variation in the original, unreduced, space, resulting in a very low stress of 0.27%. Of the explained variation, 93% can be assigned to the first axis alone.

Observation of Figure 5.1 indicates an already expected (from Chapter 4) temporal gradient where samples from the winter of 2002 are arrayed to the top left, from the summer of 2003 are approximately in the middle of the ordination plot and samples from the winter of 2003 are arrayed on the bottom right.

S2 Sampling occasion Winter/02 S3 Summer/02 Winter/03 S4

Size classes (mm)

S2 S1 = 35 - 40 S2 = 41 - 45 S5 S3 = 46 - 50 S4 = 51 - 55 S5 = 56 - 60

S1 xis 2 A S3

S2 S1

Stress=0.27% S3

Axis 1

Chapter 5. Ontogenetic shifts in diet composition 110

More importantly at this stage, an ontogenetic shift in dietary composition of A. agassizii is observable for all sampling occasions, where the points for dietary samples of sequential size classes at each site tended to progress from left to right and/or bottom of the ordination plot. Even though such patterns are not as clear for the 2003 winter, in general, size classes S1 (35-40 mm TL) and S2 (41-45 mm TL) tended to show higher segregation compared to the other size classes. Ordination did not indicate similarities in diet composition between similar size classes across different sampling occasions.

Segregation amongst size classes of A. agassizii is also evident from the percentage contribution by volume of each dietary item per sampling occasion, shown in Figure 5.2. Even though the main dietary items remain similar for all sampling occasions, their proportional contribution changed across seasons and size classes. On all sampling occasions, calanoid copepods were the main food items consumed by all size classes of A. agassizii with a minimum contribution of 41% (S2, 2002 winter), followed by moinid cladocerans with contributions of up to 59% of the ingested items for any given size class. During the 2003 winter, contributions of dietary items other than calanoids were nil or insignificant, which led to the lack of clear segregation in the ordination space among size classes of A. agassizii on this sampling occasion. Apart from that, the proportional contribution of Moinidae tended to decrease from smaller to larger individuals, although the differences in mean proportion of Moinidae and Calanoida between size classes were significant only during the 2002 winter (one-way ANOVA, F=5.86; df=3,40; p=0.0020 and F=5.4; df=3,40; p=0.0032, respectively). Nevertheless, calanoids became increasingly more important to the diet of A. agassizii as individuals of this species grew.

Dietary breadth also showed temporal variation, with higher values being observed during the 2002 winter (Table 5.1). Even though during the summer of 2002 diet breadth tended to decrease with size class, on the whole, this parameter added little to the overall patterns of diet composition for A. agassizii given the low range of dietary items consumed by this species.

Chapter 5. Ontogenetic shifts in diet composition 111

(a) Rainbow Lg. - Winter /02 100 90 80 70

60 detrit moin chydo 50

40 calan daphn chiron-p

Percentage volume 30 20 10 0 35 - 40 41 - 45 46 - 50 51 - 55 56 - 60 Size class (mm)

(b) Rainbow Lg. - Summer /02 (c) Rainbow Lg. - Winter /03 100 100 90 80 90 70 60 80 50 40 70

Percentage volume 30 Percentage volume 20 60 10 0 50 35 - 40 41 - 45 46 - 50 51 - 55 56 - 60 35 - 40 41 - 45 46 - 50 51 - 55 56 - 60 Size class (mm) Size class (mm)

5.3.2 Leiopotherapon unicolor

Similarities of proportional volumetric dietary data for sequential size classes of L. unicolor are shown in Figure 5.3. The two-dimensional NMS ordination explained 64% of the overall variation in the original space, with 44% of the explained variation attributed to the first axis. The stress value for the two-dimensional solution was 9.1%. The NMS ordination plot shows some indication of a spatio-temporal gradient within the dietary composition of L. unicolor, as samples from South Callandoon A are arrayed to the bottom of the ordination space and samples from South Callandoon B are spread on the middle and top of the plot.

Chapter 5. Ontogenetic shifts in diet composition 112

S3 Sampling occasion Winter/03 Summer/02 S2 Winter/02

Size classes (mm) S1 S1 = 30 - 39 South Callandoon B S2 = 40 - 49 S3 S4 S3 = 50 - 59 S4 = 60 - 69 S5 = 70 - 79 S6 S6 = 80 - 89

xis 2 A

S1 S5 S2

S3 South Callandoon A

Stress=9.1% S4

Axis 1

Some discreteness between the two South Callandoon B seasonal groups of samples (summer and winter of 2002) is also evident. The ordination plot also shows a clear shift in diet composition across sequential size classes for L. unicolor on all three sampling occasions, as each dietary sample of sequential size classes progressed from left to right on the ordination plot. The points corresponding to smaller individuals (size S1, 30-39 mm TL) were clearly segregated from the more intermingled remaining corresponding size classes during the 2002 summer and 2003 winter, whereas for the winter of 2002 all sizes classes were considerably segregated from each other. Interestingly, no similarity in diet composition is observable between correspondent size classes across different sampling occasions.

Chapter 5. Ontogenetic shifts in diet composition 113

Figure 5.4 shows the percentage contribution by volume of dietary items to the diet of L. unicolor for each sequential size class and sampling occasion. Even though this species presented considerably different diets on each sampling occasion, a dietary shift between size classes was also observed (Figure 5.4). Size class S1 (30-39 mm TL) in both 2003 winter and 2002 summer presented a markedly different dietary composition when compared to the remaining size classes. In general, during the 2003 winter at South Callandoon Lg. A, smaller (S1) individuals of L. unicolor showed higher consumption of pupae of chironomids (13.5%), and the contribution of this item to the diet tended to decrease with size (Figure 5.4a).

(a) South Callandoon Lg. A - Winter /03 100 90 detrit cyclop conch chiron-l 80 volvox moin nema chiron-p 70

60 alg-m daphn leptop ecnom 50 40 f-alg bosm coenag leptoc

Percentage volume Percentage 30 veg-m chydo noton hydrop 20 10 rotif sidid corix planor 0 30 - 39 40 - 49 50 - 59 60 - 69 70 - 79 80 - 89 calan ostrac Size class (mm)

(b) South Callandoon Lg. B - Summer /02 (c) South Callandoon Lg. B - Winter /02 100 100 90 90 80 80 70 70 60 60 50 50 40 40

Percentage volume Percentage 30 volume Percentage 30 20 20 10 10 0 0 30 - 39 40 - 49 50 - 59 60 - 69 70 - 79 80 - 89 30 - 39 40 - 49 50 - 59 60 - 69 70 - 79 80 - 89 Size class (mm) Size class (mm)

Chapter 5. Ontogenetic shifts in diet composition 114

In contrast, other aquatic insects, such as Ecnomidae (Trichoptera) and Coenagrionidae (Odonata), which were not consumed by S1 individuals, became increasingly important to the diet of L. unicolor as fish size increased. For example, consumption of Ecnomidae increased from nil at S1 to 8.4% at S4 (60-69 mm TL) and Coenagrionidae consumption increased from 2.7% at S1 to 22.6% at S4. Other important dietary items that made significant contributions to the diet of L. unicolor and were similarly consumed by all size classes were microcrustaceans, Calanoida (8.1 to 23.8%) and Daphniidae (14.9 to 31.4%), and larval stages of chironomids (19.2 to 30.8%). Corixidae, Leptophlebiidae and Conchostraca, although also being consumed by all size classes, made small contributions to overall diets.

Similarly, contributions of calanoids (37.9%) and chironomid larvae (42.5%) and pupae (11.5%) to the diet of small (S1) L. unicolor were also higher in samples from South Callandoon Lg. B during the 2002 winter (Figure 5.4b), although these items were progressively less consumed by larger individuals (S2, 40-49 and S3, 50-59 mm TL). In contrast, larger individuals showed an increase in consumption of microcrustaceans during the 2002 winter, such as Moinidae (from 1.1% at S1 to 46.3% at S3) and Ostracoda (from nil at S1 to 15.5% at S3). Corixid hemipterans also showed an increase in consumption from nil at S1 to 6.9% at S3. Contribution of detrital matter to the diets of L. unicolor was similar across size classes, ranging from 5.7 to 9.5%. At the same lagoon (South Callandoon B), but during the winter (2002) (Figure 5.4c), size classes S1 and S2 were not present. Even so, a decrease in consumption of chironomids (larvae and pupae) and Moinidae was observed with the remaining size classes, where chironomid larvae progressively decreased from 19.8% of contribution to S3 to only 2.9% at S6 (80-89 mm TL), chironomid pupae decreased from 34.6% at S3 to 0.4% at S6 and consumption of Moinidae progressively decreased from 12.1% at S3 to 1.2% at S6. Interestingly, during the 2002 winter, larger size classes (S4 to S6) showed an increase in consumption of aquatic invertebrates (except for chironomids, see above), with high consumption of Coenagrionidae (34% at S5, 70-79 mm TL), Hemiptera (18.2% of Notonectidae at S6 and up to 14.8% of Corixidae at S6) and Ecnomidae with up to 4.2% at S6.

Dietary breadth data from Table 5.1 indicated an increase in the range of food items consumed by individuals of L. unicolor after the S1 size class, but further growth did not necessarily lead to higher breadth in dietary composition. Chapter 5. Ontogenetic shifts in diet composition 115

5.3.3 Nematalosa erebi

The NMS ordination plot of proportional volumetric dietary data for sequential size classes of N. erebi is shown in Figure 5.5. The two-dimensional ordination explained 95% of the variation in the original data, with 67% of that attributed to the first axis. The resulting stress was only 0.01%. The NMS plot shows a gradient from bottom right to top left, with size classes of larger individuals being arrayed to the left side of the ordination space. Furthermore, a clear segregation between smaller individuals of size classes S1 (40-49 mm TL) and S2 (50-69 mm TL) was observed, as well as between these size classes and the remaining ones. In contrast to the other two species studied, with the exception of S2 at Punbougal Lagoon, smaller (S2) size classes of N. erebi were clustered together, regardless of season or location.

S4 S3 Sampling occasion S2 S4 S3 Summer/02 S4 S6 S5 S5 Summer/02 S3 Summer/03 S3 Summer/03

S4 Lagoon Pungbougal Maynes Rainbow S. Callandoon A

Size classes (mm)

xis 2 xis S1 = 40 - 49 A S2 = 50 - 69 S3 = 70 - 89 S4 = 90 - 109 S2 S5 = 110 - 129 S2 S2 S6 = 130 - 139

Stress=0.01% S1

Axis 1

The proportional contribution by volume of each dietary item to each sequential size class and sampling occasion, shown in Figure 5.6, also indicated a clear ontogenetic shift in the dietary composition of N. erebi.

(a) Maynes Lg. - Summer /02 (b) Pungbougal Lg. - Summer /02 100 100 90 95 80 90 70 85 60 80 50 75 40 70

Percentage volume 30 Percentage volume 65 20 60 10 55 0 50 40 - 49 50 - 69 70 - 89 90 - 109 110 - 129 130 - 139 40 - 49 50 - 69 70 - 89 90 - 109 110 - 129 130 - 139 Size class (mm) Size class (mm)

Percentage volume 30 Percentage volume 30 20 20 10 10 0 0 40 - 49 50 - 69 70 - 89 90 - 109 110 - 129 130 - 139 40 - 49 50 - 69 70 - 89 90 - 109 110 - 129 130 - 139 Size class (mm) Size class (mm)

detrit moin f-alg alg-m chiron-l veg-m

ostrac calan volvox cyclop nema

Although detritus material contributed at least 52.9% of the diet of this species in all size classes, smaller individuals (size classes S1 and S2) showed consistently higher consumption of microcrustaceans (mostly Moinidae and Calanoida) than larger individuals (> 69 mm TL, S3-S6) in all lagoons except Punbougal. In this lagoon, diet of all size Chapter 5. Ontogenetic shifts in diet composition 117 classes consisted mostly of detritus and, to a much smaller extent, vegetable matter. Interestingly, individuals from 40 to 49 mm TL at Rainbow Lagoon presented a diet consisting exclusively of Moinidae (53.8%) and Calanoida (46.2%). Consequently, dietary breadth was much narrower in larger individuals (> 6.9, S3-S6) when detritus material dominated the food items ingested by N. erebi (Table 5.1).

5.4 Discussion

Several species of fish are known to exhibit substantial ontogenetic changes in both taxonomic composition and size range of prey consumed (Brodeur 1991). These changes have been attributed to increasing body size associated with changes in morphology of the feeding apparatus, habitat or behavioural shifts, or to age of the predator (Werner and Hall 1977, Grossman 1980, Schmitt and Holbrook 1984). In the present study, all three species of fish showed distinct shifts in diet composition across size classes and these trends generally varied across seasons and/or locations. Winemiller (1989) recognised that food availability is a primary factor involved in producing size-related patterns of feeding among fish. According to this author, an ontogenetic switch from invertebrate (zooplankton and benthic invertebrates) to fish prey in piscivorous fish coincides with changing food availabilities associated with seasonal changes. It is, therefore, probable that differences in food item availability, related to spatio-temporal differences in the sites, studied have influenced the size-related patterns in the fish species studied.

Ambassis agassizii tended to feed more on calanoids and, to a lesser extent on daphniid cladocerans, as it grew, whereas smaller individuals tended to consume more moinid cladocerans than larger ones. There was no evidence from the literature indicating ontogenetic shifts in diet composition of A. agassizii. Kennard (1995) described size- related changes in the diet of a similar species, Ambassis macleayi, in floodplain lagoons of the Normanby River (Cape York Peninsula, north Queensland). Results from Kennard (1995) show that smaller individuals (<18 mm SL) presented a less diverse diet and relied mostly on cladocerans (69.1% mean proportional contribution to their diet) and chironomids (18% mean proportional contribution), whereas larger individuals (>18 mm SL) presented a more diverse diet with an increase in consumption of aquatic invertebrates Chapter 5. Ontogenetic shifts in diet composition 118

(mostly chironomids, corixids, notonectids, ephemeropterans and trichopterans) and a decreased consumption of cladocerans (from 69.1 to 44.1%).

Changes associated with increasing body size, such as higher swimming speed and mouth gape size, visual acuity and reaction distance to prey, may increase the range of prey available to a predator as it grows (Werner 1974, Breck and Gitter 1983, Dunbrack and Dill 1983, Keast 1985). Kennard (1995) demonstrated that small microphagic planktivore/insectivore species were limited in prey choice by constraints in body and mouth sizes. On the whole, the fact that A. agassizii fed mostly on small microcrustaceans is an indication that this species is restricted to such dietary items by mouth size and locomotory ability. Additionally, even though diets were found to differ between sampling occasions, the main dietary items were similar with only the overall proportional contribution of items the diet changing. Therefore, another important factor that must be considered is the fact that resource availability probably had a strong influence on the diet composition of different size classes of this species. Arthington (1992) found that individuals of A. agassizii in a size range similar to the ones in this study (22-45 mm SL) consumed exclusively aquatic insects and terrestrial material in tributary streams of the Brisbane River. Furthermore, O’Brien (1979) demonstrated in a laboratory experiment, that fish consistently selected prey that appeared to be larger, either because of their absolute size or because the prey item was closer to the fish. It is possible that the progressive increase in consumption of calanoids by A. agassizii with growth relates to this tendency to consume larger prey, as calanoids were usually larger than moinid cladocerans.

Ordination of dietary items of L. unicolor for sequential size classes showed a higher segregation of smaller individuals (approx. 24-31 mm SL) against the remaining size classes on the same sampling occasion. These smaller individuals tended to rely mostly on microcrustaceans (calanoids and daphniid cladocerans) and chironomid larvae and pupae. Even though these items were also important to the diet of larger fish (>31 mm SL), an increase in the consumption of aquatic invertebrates (Ecnomidae, Coenagrionidae, Corixidae, and Notonectidae) was observed in larger individuals. These results are partially in accordance with Pusey et al. (2004) who found that smaller individuals of L. unicolor (<40 mm SL) consumed mostly chironomid larvae whereas larger fish had a higher consumption of larger prey such as Odonata and Trichoptera. Interestingly, even though the size range reported by Pusey et al. (2004) for L. unicolor roughly matches the Chapter 5. Ontogenetic shifts in diet composition 119 size ranges examined in this study (note that individuals were fixed in formalin before being measured, so some shrinkage should be expected), the contribution of microcrustaceans to the mean diet of this species reported by Pusey et al. (2004) was very low, suggesting that geographic or habitat variation must also be accounted for when explaining ontogenetic variation in diets of this species. Bluhdorn and Arthington (1994) observed that larger individuals of L. unicolor tended to consume mostly shrimp and occasionally fish, whereas smaller individuals (≤ 70 mm SL) consumed greater amounts of insects (larvae and ephemeropterans) and microcrustaceans (copepods and ostracods). Even though fish and shrimp were not recorded in the present study, additional results presented in Chapter 6 agree with Bluhdorn and Arthington (1994) by showing greater contributions of these two food items to the diets of relatively large individuals of L. unicolor in Rainbow and Serpentine lagoons. One feature of particular interest is that a spatio-temporal gradient existed in the ontogenetic shifts observed for L. unicolor in the present study, as different sampling occasions showed particularly different patterns in the composition of dietary items across size classes. During the 2002 summer sampling occasion at South Callandoon Lg. B, for example, larger individuals switched from calanoid copepods as the main microcrustacean item to ostracods and moinid cladocerans, and to a lesser extent, corixids, and the contribution of calanoids at the same location was much lower during the 2002 winter.

Although not being a frequently collected species, L. unicolor was very abundant when present in a site sampled. This species is known to show considerable variation in abundance levels in space and time (Woodland and Ward 1992, Pusey et al. 2004). Pusey et al. (2004) suggested that the ontogenetic dietary shifts in L. unicolor is an important factor allowing this species to maintain high densities by minimizing intraspecific competition. Furthermore, Gehrke (1988b) suggested that there is an important effect of gape limitation on prey choice in smaller individuals of this species. Interestingly, in the present study, smaller individuals (size class 3.0-3.9 TL) reflected a narrower niche breadth when compared to the remaining correspondent size classes.

Nematalosa erebi also showed size-related changes in diet composition, with a clear shift from a mostly planktivorous diet (Moinidae and Calanoida) in smaller individuals (≤69 mm TL) to a predominantly detritivorous diet in larger fish (>69 mm TL). A clear segregation of smaller individuals was demonstrated by ordination, whereas larger Chapter 5. Ontogenetic shifts in diet composition 120 individuals presented similar dietary composition. Interestingly, this species was the only one to show similarities between correspondent size classes across different sampling occasions. Pronounced size-related patterns in the diet of N. erebi are a known feature in the dietary ecology of this species recognised by Kennard et al. (2001) and Pusey et al. (2004) among others. Bunn et al. (2003), noted a size-related shift in stable isotope signatures of this species, reflecting a decrease in dependence on zooplankton and a greater dependence on benthic algae and terrestrial detritus as individuals of N. erebi grew (see also Chapter 7). Even though the cut off value for the size when this species switches from a diet based on invertebrates to a more herbivorous/detrital diet varies, in general, individuals smaller than 70 mm TL can be expected to have higher contribution of invertebrates to their diet. This is in accordance with results presented in this study, which shows that individuals larger than 70 mm TL were mostly detritivorous. It is important to bear in mind that individuals examined in this study were actually larger than the sizes given, as they were fixed in formalin prior to being measured. It is likely that the morphological changes that occur with the ontogeny of N. erebi played a major role in the observed dietary shifts in this species. Kennard et al. (2001) suggested that variations in diet between juveniles and adults of N. erebi probably reflect ontogenetic changes in morphology, such as increasing mouth size, as well as shifts in foraging habitat. Moreover, Atkins (1984) found that some of the morphological correlates of dietary shifts in this species included a change in mouth positioning, from dorso-terminal to terminal in fry and juveniles (related to planktivory), to ventro-terminal in adults (related to omnivory and detritivory), and ontogenetic changes in gut morphology, where smaller individuals have a relatively simple, straight and short alimentary tract and only later in life the gizzard (used to help digest detrital/plant materials) becomes muscular and functional.

Even though ontogenetic niche shifts can be influenced by changes in the morphology of the feeding apparatus, factors such as composition and availability of food items are also an important element that must be taken into account. Even though being primarily a detritivore/algivore species, studies indicate that the composition or abundance of dietary items consumed by N. erebi may change on a seasonal basis, as pointed out by Balcombe et al. (in review) for the Cooper Creek (western Queensland), or on a spatial scale, reflecting the effect of habitat structure on trophic style (Kennard 1995). Even though the diet of N. erebi in the present study remained similar across size classes and sampling occasions, during the 2002 summer at Punbougal Lg., the diet of this species did not show Chapter 5. Ontogenetic shifts in diet composition 121 clear changes in composition between size classes. Therefore, similar to the other species studied, size-related changes in dietary composition of N. erebi are also dependent on spatial or temporal changes in food availability, as zooplankton densities in Punbougal Lg. were very low, as demonstrated in Chapter 7.

5.5 Conclusions

On the whole, A. agassizii, L. unicolor and N. erebi experienced size-related changes in diet, either in the relative abundance of prey consumed, as was the case for A. agassizii, or in the specific composition of the food items ingested. In general, dietary breadth of A. agassizii was very narrow, with ontogenetic diet changes mostly related to an increase in the abundance of calanoids and a decrease in abundance of moinid cladocerans consumed with age/size. Leiopotherapon unicolor shifted from a narrower diet breadth in smaller individuals, based on microcrustaceans and chironomids, to a wider range of food items in larger individuals, with an increase in consumption of aquatic insects with age/size. Nematalosa erebi displayed the clearest dietary shift, with a change from microcrustaceans in smaller individuals, to detritus material in lager fish, as the main food category.

Although ontogenetic changes in diet composition were observed in all species and on most sampling occasions, it is evident from these results that the degree of ontogenetic shift in prey items consumed by the species studied is also dependent on the season and/or site where the fish were collected. This is supported by previous work that indicates that, with exception of N. erebi, the species studied are expected to present strong spatial patterns in composition of dietary items. This finding is further supported by the fact that, with exception of N. erebi, the species studied did not show similarities in dietary composition between similar size classes across different locations and/or seasons.

In the studied floodplain lagoons, a combination of prey availability, most probably related to spatial and temporal variation in prey abundance and composition, and ontogenetic morphological changes in the feeding apparatus and gape limitation of the study species of fish, are suggested as the main factors influencing the size-related changes in their diet composition. This also raises the issue of temporal and spatial trends in diets of fish from floodplain being confounded by ontogenetic shifts in prey consumption. Chapter 5. Ontogenetic shifts in diet composition 122

Even though N. erebi had a predominantly detritivorous diet, juveniles of this species showed a diet similar to the other species studied, based on microcrustaceans. These results confirm the importance of microcrustaceans, as a food resource for small and/or juvenile individuals of A. agassizii, L. unicolor and N. erebi in floodplain lagoons of the Macintyre River.

5.6 Implications for the study of fish diet composition

Given the results presented in this chapter, it is clear that the species of fish studied experienced size-related changes in diet, on both relative abundance of prey consumed and/or specific composition of the food items ingested. In general, ontogenetic changes can influence the detection of temporal and spatial changes in diet composition of fish, and therefore it is advisable that, for further analysis, average sizes or size ranges of groups of individuals of the study species are known and taken into consideration when explaining patterns in diet composition. This is more important in the case of N. erebi, which presented the strongest changes in diets with in relation to fish size.

On the other hand, whether these changes in diets with size are likely to confound the detection of potential temporal and spatial patterns in food items composition of the study species is subject to debate. From the results presented in this study, the degree of shift in the items consumed by the species studied is also dependent on the season and/or site where the fish were collected, as similar sized fish presented different diets in different lagoons or seasons (mostly A. agassizii and L. unicolor). Therefore, the results so far demonstrate that overall differences in dietary composition between sites seem to overwhelm size-related shifts in diets within sites or sampling occasions (see also results from Chapter 6). An exception to this argument is N. erebi, which presented similar diet between similar size classes across sampling occasions. Despite that, the approach for further chapters (6 and 7) will be to keep some control over size of fish within samples and therefore account for any possible variation in size across sites and/or sampling occasions, as this will give further insight in possible patterns of diet changes.

Chapter 6. Spatial and temporal variation in diet composition 123

6 Spatial and temporal variation in diets of fish in floodplain lagoons

6.1 Introduction

Tropical floodplain ecosystems are biologically important aquatic systems supporting diverse fish assemblages (Junk et al. 1989, Junk and Weber 1995). The richness and variability of floodplain habitats is known to provide a wide range of possible food organisms and substrates for fish (Welcomme 1979), which use the inundated areas directly as sources of food (Ross and Baker 1983, Geddes and Puckridge 1989, Bayley and Li 1992, O’Connell 2003). It has been suggested that without the intermittent connection to terrestrial habitats, river and stream systems may not have sufficient production to support the local aquatic organisms (Junk et al. 1989, Junk and Weber 1995, Sheldon et al. 2003).

The energy pool available in the biomass of aquatic primary producers (microorganisms, algae, aquatic plants) is ultimately the only food available for fish production, unless supplemented by a supply of allochthonous material (Bishop et al. 2001). Studies of freshwater fish diets emphasize the importance of riparian resources, e.g. terrestrial arthropods and fruit seeds (Goulding 1980) and detritus, e.g. fine and coarse particulate organic matter as sources of food for river fishes (Welcomme 1985). Alternatively, the importance of algae as the primary energy source fueling floodplain food webs has been emphasized in a number of recent studies on arid and tropical rivers (Hamilton et al. 1992, Thorp et al. 1998, Bunn et al. 2003). Welcomme (1979) argued that the feeding cycle of fish in floodplains is linked to two main factors, i.e. food supply and population densities. During the flood, a rapid increase in food organisms, together with wide dispersal of fish over the inundated area favours intensive feeding, whereas at low water when the aquatic environment is contracted, fish are concentrated in a few permanent waterbodies disconnected from the main river, and food resources may become limiting. This pressure on food resources is apparently greatest when water levels fall after the wet season, and the number of fish species and individuals rises as they leave the contracting wet season environments (Zaret and Rand 1971, Pusey et al. 1995). Several authors found that during Chapter 6. Spatial and temporal variation in diet composition 124 the dry season fish diets overlap least due to the fact that food is more scarce and competition stronger (Lowe-McConnell 1964, Goulding 1980).

Many species of fish show seasonality in their food uptake related to flood cycles (Wootton 1990). Seasonal changes in food availability may be caused by changes in the habitat available for foraging, changes due to life-history patterns of food organisms and changes caused by the feeding activity of fish themselves (see Marchant 1982, Bishop et al. 2001, Sheldon et al. 2003). Many studies have related spatial and temporal variations in diet composition of fish with variations in abundance of different food items across habitats as they became seasonally available (e.g. Arthington et al. 1994, Bishop et al. 2001). Bishop et al. (2001) reported that changes in composition of food items consumed by several species of fish were associated with fluctuations of food items available across seasons and habitats in a floodplain system. On the other hand, workers who measured food availability for fish in floodplain lagoons found that maximum spatial variations in abundance of major food items did not necessarily coincide with variations in dietary breadth of fish consumers (e.g. Kennard 1995). Furthermore, O’Connell (2003) reported that, for the floodplain of a low-order, blackwater stream, there was no evidence that fish diets became more specialised on the floodplain in response to increases in food availability due to flooding.

The reduced amount of water reaching floodplain waterbodies of Australian rivers due to river regulation has been held responsible for successional changes in aquatic vegetation and for observed declines in population numbers of native fish and invertebrates (Kingsford 2000). However, information regarding the impacts of water resource development on biota in Australia has generally focused on within-channel processes of rivers (Walker 1985, Lake and Marchant 1990, Bren 1993), and less so on floodplains which are arguably more affected by water development (Kingsford 2000). Water alienation from floodplains by structures such as levees, block banks and weirs may lead to loss of connectivity between river and floodplain, by reducing or stopping flows to floodplain waterbodies. Despite the fact that these factors can lead to high turbidity and destroy submerged vegetation (Casanova 1999) and result in loss of productivity and habitat, the ecological impacts of the above-mentioned structures, and floodplain alienation, on food resources for fish in floodplain habitats are still poorly understood.

Chapter 6. Spatial and temporal variation in diet composition 125

In Australian floodplain rivers, billabongs and floodplain lagoons are regarded as highly productive systems, supporting a large and diverse biomass of microorganisms and invertebrates (Bunn and Boon 1993, Butcher 1997, Hillman 1998) that are hypothesised to represent a significant food resource for fish (Geddes and Puckridge 1989, Hillman 1995b). Even though quantitative dietary data are available for many species of fish in almost all regions of Australia (Kennard et al. 2001, Pusey et al. 2004), there is currently a lack of information about spatial and temporal variations in the diets of fishes inhabiting floodplain lagoons in Australian dryland rivers, and the role of autochthonous and allochthonous materials in fish dietary ecology in these ecosystems (but see Balcombe et al. in review). Moreover, Kennard et al. (2001) recognised that there may be a significant regional pattern in fish diets, probably as a result of variation in local productivity, food availability and species composition, or a combination of these factors.

In dryland floodplain rivers, food resources for fish can be expected to vary in response to flooding, when inundation of isolated lagoons and surrounding floodplain areas can result in a burst of primary and secondary productivity (Puckridge 1999). Fish utilising this production whilst feeding on the floodplain are expected to differ in diet composition from fish living in unflooded areas where they have access only to food resources available within floodplain waterbodies, and, possibly, riparian food resources. Fish diets may also be expected to vary when floodplain lagoons are modified by water resource developments, such as maintenance of artificially high water levels or, alternatively, abstraction of river water, such that lagoons are deprived of natural flood flows from adjacent rivers or flooded less frequently. Floodplain lagoons that do not experience flooding, and those that are flooded less frequently than under natural conditions, may experience increasing water drawdown over dry periods. Fish confined to such lagoons must, therefore, depend on autochthonous food resources.

6.1.1 Aims

This chapter will ascertain whether diets of Ambassis agassizii, Leiopotherapon unicolor and Nematalosa erebi varied spatially, that is, between floodplain lagoons and a river site with different flow histories, and if fish diets at these sites are influenced by broad seasonal patterns (summer/winter). This chapter also addresses the implications of an extended Chapter 6. Spatial and temporal variation in diet composition 126 period of drought and only minor flooding of some lagoons on fish diets as well as the effect of water resource development, as indicators of the impact of modified water regimes on the trophic ecology of fish living in an arid zone floodplain-river system. Therefore, the present study aims to describe (1) spatial variation in prey items consumed by fish from floodplain lagoons with different flow histories, and in relation to flow regulation, and (2) temporal variations in fish diets in relation to minor flooding events and, in the absence of flooding, broad seasonal (summer/winter) changes in diet, and diet composition as the dry season progressed in the Macintyre River.

Specific questions addressed in the present chapter are:

1. Given the natural differences in morphology, water quality and flow history of floodplain lagoons, is there variation in food items consumed by three species of fish with contrasting feeding ecology across different sites on the floodplain of the Macintyre River? 2. Are there temporal changes in diets of the study species as a consequence of floodplain and lagoon rejuvenation following flooding? If so, does the contribution of allochthonous versus autochthonous food resources change significantly as the dry season progresses, and before and following a flood? 3. In the absence of a flooding event when fish are confined to permanent or semi- permanent waterbodies with receding water levels, will fish be dependent on food resources produced within floodplain lagoons and/or food resources of riparian origin, and will their diet change as the dry season progresses? 4. Do fish confined in floodplain lagoons subjected to particular forms of flow modification, such as artificial filling and permanently elevated water levels, present different patterns in diet composition compared to fishes in lagoons subject only to regulation of river flows?

6.2 Methods

Seven lagoon sites and one river channel site on the floodplain of the Macintyre River (see Figure 2.2) were sampled to evaluate variations in dietary composition of the study species. A total of 6 field trips was made between late 2001 and 2003 (see Table 6.1) and the study sites sampled roughly during spring-summer and autumn-winter months, Chapter 6. Spatial and temporal variation in diet composition 127 henceforth summer and winter (summary data are given in Table 6.1). Given that the study period experienced low water flows along the Macintyre River, pre- and post- flooding samples were collected only during the summer of 2002 (October) prior to flooding of the South Callandoon A and Rainbow lagoons and later in March (summer) and July (winter) of 2003 following flooding of the above mentioned lagoons. Note that, although not flooded, all study sites were sampled on all occasions. Summary data on overall characteristics of the study sites in relation to lagoon type, flow history and water management are given in Table 3.1.

All sampling procedures were performed as described in Chapters 4 and 5. Each site was sampled using a seine net (25 m length x 2 m height, 1 cm mesh) from approximately 11:00 am to 2:00 pm. Results of Chapter 4 suggested that, sampling of A. agassizii, L. unicolor and N. erebi for dietary analysis, should be concentrated at midday (1130 to 1300 hours), as this is the time of day when food consumption was greatest and dietary breadth was wider. Samples were taken at random points along the margins of lagoons and the river site (maximum depth of 1.5 m), as these areas generally contain the majority of fish and because seining becomes impracticable at water depths greater than 1.5 m. Individual seine lengths were approximately 10 m and from 5 to 15 m wide depending on lagoon depth. Duration of each seine was about 5 minutes with 5 to 10 minutes between seines to allow time to sort the fish caught and remove any woody debris from the net (Section 4.2.2, Chapter 4). This procedure was repeated a minimum of three times along the margins of each study site until enough individuals of a range of different sizes of each target species had been caught. Size classes were later determined in the laboratory after fish lengths were measured (as per Section 5.2). All individuals of the target species were preserved whole in 4% buffered formalin immediately after capture. Later in the laboratory, fish were transferred to 70% ethanol and the necessary measurements and gut content analyses were performed.

Dietary analysis of the stomach contents was performed as described in Section 4.2.3 of Chapter 4. The proportional contribution of each category of food item to the diet of individual fish was estimated using Hyslop’s (1980) indirect volumetric method, where stomach items within each prey category are lumped together and squashed to a uniform depth of 1 mm and the area covered taken to represent the volumetric contribution of that food category to the total stomach contents in mm3 (cf. Arthington 1992, Pusey et al. Chapter 6. Spatial and temporal variation in diet composition 128

2000). Prey items were identified to the lowest level of taxonomic resolution as described in Section 4.2.3 (Chapter 4).

6.2.1 Data analysis

The contribution of each dietary item to the overall diet of the species studied was expressed as the percentage frequency of occurrence of each dietary item in the stomachs of all individuals of each fish species collected on each sampling occasion and as the mean of the percentage volumetric contribution made by each dietary item to the stomach contents of each individual of each species (see next paragraphs). The unidentified material category was included only in the description of the total diet of the selected species to give an indication of the proportional contribution to the total mean diet of the unidentifiable fraction. This category was excluded from any further analysis, since the inclusion of values of unidentified material might bias the results of multivariate analysis (Platell and Potter 2001) and, the overall contribution of unidentified material was found to be small (less than 10%) as demonstrated on Table 4.2 and in the results from Chapter 5. For all analyses performed, only stomachs with fullness equal to or greater than 20% were considered, in order to prevent bias in the calculation of the proportional contribution of prey items or diversity indices (Pusey et al. 1995) (see also Section 4.2.4.2 in Chapter 4).

Non-metric Multidimensional Scaling (NMS) (McCune and Mefford 1999) was used for the evaluation and interpretation of overall similarities in dietary composition among species, sites and sampling occasions. Based on results from Chapter 5, the species studied showed ontogenetic changes in composition and/or abundance of prey items consumed, therefore, to account for the influence of size on spatial and temporal variations in dietary composition, the proportional volumetric dietary data for individuals of each species from each sampling occasion were ordered by fish body size (TL).

The proportional volumetric dietary data for single individuals of each species from each sampling occasion were then allocated into groups of 5 to 15 individuals, according to their size (TL) and the mean diet composition calculated for each of those groups. Size classes were the same as those examined in Chapter 5 and are shown in Table 5.1. Group size depended on the number of individuals of each species collected on each sampling Chapter 6. Spatial and temporal variation in diet composition 129 occasion. Note that each of these groups of 5-15 individual fish included sequentially sized individuals and correspond to a point or dietary sample in the NMS ordination space. Point samples from a-priori groups, e.g. sampling occasions, sites or seasons were used in the analysis of the significance of differences in diet for each species (Section 6.2.1.1). The allocation of individuals of each species collected on each sampling occasion into size groups, prior to analysis, overcomes the problem posed by the fact that the stomachs of individual fish often contained only a few of the total number of dietary items recorded for that species (Schafer et al. 2002) and also allows size variation within and between sampling occasions to be taken into account. These data were also used to calculate the mean Shannon-Wiener diversity index of prey in the stomachs of each species collected on each sampling occasion and at each site, this index corresponding to dietary breadth (B) (Pusey et al. 1995, Marshall and Elliott 1997, Pusey et al. 2000).

Prior to ordination, the data were arcsine square root transformed, as is appropriate for proportional multivariate data (McCune and Grace 2002) and an outlier analysis was performed to check for outliers and skewness (McCune and Mefford 1999). The Relative Sorensen (Kulcynski) distance was used for the analysis and all graph outputs were Varimax-rotated. The coefficient of determination (r2) between Euclidean distances in the ordination space and distances in the original space was used as an after-the-fact method for calculating proportion of variance represented by ordination axes (McCune and Grace 2002). This method is useful to evaluate the quality of data reduction and, therefore, how well the distances in the ordination space represent the distances in the original, unreduced space.

6.2.1.1 Analysing the significance of patterns

Further analysis of temporal and spatial patterns of associations in fish diets was performed using a combination of multivariate techniques as suggested by McCune and Grace (2002), as the design of this study involves a relatively complex array of multivariate data. Initially, to test for differences in food items consumed by each species between two or more groups of entities corresponding to dietary groups of sample points for seasons, sites or sampling occasions, the Multi-Response Permutation Procedure (MRPP) was used (Biondini et al. 1985, McCune and Mefford 1999). This is a nonparametric method for Chapter 6. Spatial and temporal variation in diet composition 130 testing multivariate differences among pre-defined groups, by calculating the probability (p) that group differences detected are due to chance alone (Zimmerman et al. 1985). MRPP has the advantage of not requiring distributional assumptions, such as multivariate normality and homogeneity of variances (McCune and Grace 2002), which are unlikely to be met by the data collected in the present study. Rank-transformation of the similarity matrix (based on same distance measure used in ordination, Relative Sorensen) was performed to correct for loss of sensitivity of the distance measure as heterogeneity increases (McCune et al. 2000). Consequently, the chance-corrected within-group agreement (A) (see below) is considerably higher after rank transformation of the distance matrix (McCune and Grace 2002).

For all MRPP analyses, the chance-corrected within-group agreement (A) is presented. The value of the A statistic represents the degree of within group homogeneity, compared to random expectation, and is analogous to the ‘effect size’. When all sample points in a group (e.g. sampling occasions or sites) have identical composition of food items, A=1; when within group heterogeneity equals that expected by random chance, A=0. On the other hand, if there is less agreement within groups than expected by chance, A<0 (McCune and Grace 2002). It has been reported that values of A below 0.3 are often meaningful in community data, although A values are commonly below 0.1 even when within-group heterogeneity differs significantly from the expected (McCune and Grace 2002, Wolf et al. 2003). Even though MRPP provides a measure of ‘effect size’ (A) and a p-value, the description of any differences between the a-priori groups (e.g. sampling occasions, sites, seasons) should be in association with other related methods (McCune and Grace 2002).

Therefore, where MRPP detected significant differences in composition of food items between groups, further analysis was performed to reveal which particular food items contributed significantly as the source of difference in diet composition. Significant group associations of food items were determined using McCune and Mefford’s (1999) Indicator Species Analysis (ISA). This method combines information on food item abundance (or proportional volume, in the present study) in a particular group and the faithfulness of occurrence of a food item in a particular group. It also produces indicator values (IV), for each food item in each group, which describe the separation of groups. IVs were Chapter 6. Spatial and temporal variation in diet composition 131 calculated using the method of Dufrene and Legendre (1997). These were tested for statistical significance (p<0.05) using a Monte Carlo technique with 1000 runs, where:

1+ n(observed runs) p = 1+ n(randomised runs)

Finally, the food items that defined the group membership, based on ISA, and/or were correlated with groups of sample points, were overlaid onto the NMS ordination plot. Vectors of the corresponding food items show the nature of the association between each food item and the ordination points. The vector lines radiate from the centroid of the ordination scores and the angle and length of the line indicates the direction and strength of the relationship. When a vector length was greater than the plot area (due to high correlation with a given food item), such vector was expressed as a proportion (%) of its original length. In such cases, the proportion of the plotted vector in relation to its original length is stated where appropriate. To eliminate ‘weak’ variables only food items with a correlation coefficient (r2) (with scores on either ordination axis) larger than 0.2 are plotted. Note that this is a joint plot of food items and sample points, not a correspondence analysis. Such an approach is particularly useful when there are only a few strong variables (in this case food items) defining the main groups, and to ease the visualisation of the underlying patterns of variation. Frequency histograms based on the proportional contribution of food items averaged across all individuals of each species and size groups collected on each sampling occasion were constructed for further visual comparisons of major dietary patterns of each species.

6.3 Results

6.3.1 General dietary composition

The overall abundance and mean size of the species studied showed considerable variation across sampling sites and occasions (Table 6.1). Average sizes of A. agassizii varied from 28 mm (TL) on February 2002 (South Callandoon Lagoon B) to 46 mm (TL) on May of the same year at Rainbow Lagoon, whereas L. unicolor mean sizes ranged from 46 mm (TL) in February 2002 (South Callandoon Lagoon B) to an average size of 137 mm (TL) in Chapter 6. Spatial and temporal variation in diet composition 132

Sampling occasions Time of last Mean TL Mean fullness Mean dietary Mean dietary Season N (and site codes) flooding * (mm) (%) breadth breadth per site

A. agassizii ***

Rainbow Lagoon 0.4734 (RBW) (± 0.28)

May 2002 Winter November 96 46 58 0.3051 2001 (± 3) (± 4.9) (± 0.14) August 2002 Winter November 59 45 53 0.8293 2001 (± 3) (± 5.7) (± 0.14) October 2002 Summer November 60 50 47 0.6923 2001 (± 5.7) (± 5.1) (± 0.06) March 2003 Summer February 30 39 54 0.4867 2003 (± 1.8) (± 2.3) (± 0.11) July 2003 Winter February 30 42 63 0.0867 2003 (± 1.8) (± 7.8) (± 0.01) South Callandoon 0.6637 Lagoon A (SCA) (± 0.09)

October 2002 Summer November 30 46 41 0.6637 2001 (± 2.9) (± 4.8) (± 0.09) South Callandoon 0.9877 Lagoon B (SCB) (± 0.37)

February 2002 Summer November 72 28 48 0.5502 2001 (± 2.3) (± 5.4) (± 0.23) May 2002 Winter November 127 35 39 1.2308 2001 (± 2.4) (± 3.6) (± 0.13)

L. unicolor

South Callandoon 1.8267 Lagoon B (± 0.17)

February 2002 Summer November 90 46 57 1.7578 2001 (± 4.7) (± 10.1) (± 0.17) May 2002 Winter November 86 66 39 1.8955 2001 (± 8.2) (± 7.2) (± 0.16) South Callandoon Lagoon A

July 2003 Winter February 138 49 42 2.3130 2003 (± 7.3) (± 9.1) (± 0.13) Rainbow Lagoon 1.1385 (± 1.23) November 2001 Summer November 30 90 53 1.1080 2001 (± 12.7) (± 11.5) (± 0.28) October 2002 Summer November 8 96 80 1.2300 2001 (± 15.9) (± 21.4) Serpentine Lagoon 1.2605 (SPN) (± 0.34)

October 2002 Summer Late 2000 ** 5 121 93 1.5030 (± 6.3) (± 9.7) March 2003 Summer Late 2000 ** 5 137 82 1.0180 (± 14.8) (± 8.4) Broomfield Lagoon (BFD)

February 2002 Summer November 8 64 69 1.6550 2001 (± 5.1) (± 21.7)

Chapter 6. Spatial and temporal variation in diet composition 133

Table 6.1 Continued.

Sampling occasions Time of last Mean TL Mean fullness Mean dietary Mean dietary (and site codes) Season flooding * N (mm) (%) breadth breadth per site

N. erebi

Rainbow Lagoon 0.6595 (± 0.32) November 2001 Summer November 86 87 70 0.6680 2001 (± 10.1) (± 9.2) (± 0.18) May 2002 Winter November 97 72 64 0.2609 2001 (± 5.8) (± 3.0) (± 0.12) August 2002 Winter November 69 100 49 0.8418 2001 (± 15.8) (± 6.4) (± 0.17) October 2002 Summer November 30 92 83 0.7893 2001 (± 7.6) (± 8.7) (± 0.31) March 2003 Summer February 60 67 62 0.8285 2003 (± 16.2) (± 8.5) (± 0.41) July 2003 Winter February 30 66 55 0.7397 2003 (± 6.7) (± 1.6) (± 0.28) South Callandoon 0.9816 Lagoon A (± 0.31)

May 2002 Winter November 106 78 60 0.7486 2001 (± 6.2) (± 3.9) (± 0.19) March 2003 Summer February 60 73 72 1.0458 2003 (± 13.2) (± 5.4) (± 0.23) July 2003 Winter February 30 57 49 1.3967 2003 (± 3.2) (± 1.2) (± 0.08) Maynes Lagoon 0.7071 (MNS) (± 0.40)

October 2002 Summer November 60 78 80 0.6565 2001 (± 28.4) (± 14.4) (± 0.33) March 2003 Summer November 60 82 68 0.7577 2001 (± 22.7) (± 7.6) (± 0.48) Punbougal Lagoon 0.3470 (PBL) (± 0.11)

October 2002 Summer November 90 88 67 0.3809 2001 (± 27.2) (± 11.1) (± 0.12) March 2003 Summer November 60 74 61 0.2962 2001 (± 13.9) (± 5.1) (± 0.10) Macintyre River 0.2234 (RIV) (± 0.04)

March 2003 Summer February 30 108 80 0.1917 2003 (± 11.3) (± 2.5) (± 0.00) July 2003 Winter February 50 99 79 0.2424 2003 (± 10.8) (± 3.7) (± 0.04) Serpentine Lagoon

March 2003 Summer Late 2000 ** 30 124 70 0.9860 (± 12.1) (± 3.5) (± 0.15)

Chapter 6. Spatial and temporal variation in diet composition 134

March 2003 (Serpentine Lagoon). Similarly, N. erebi also showed considerable variation in size, ranging from 55 mm (TL) on average in July 2003 (South Callandoon A) to 124 mm (TL) in March 2003 (Serpentine Lagoon) (Table 6.1).

The range of food items consumed varied between species. Ambassis agassizii presented a generally lower dietary breadth per site (B =0.4734-0.9877), N. erebi had a slightly wider, though more variable, dietary breadth (B =0.2234-0.9860), whereas L. unicolor presented the greatest range in food items consumed (B =1.2605-1.8267) (Tables 6.1 and 6.2). Similarly, the contribution by volume and frequency of occurrence of dietary items varied between the three species studied (Table 6.2). Detritus, microcrustaceans and aquatic insects were the main food items consumed, making significant contributions to the diets of all species. Even so, contributions of detritus were much greater for N. erebi, corresponding to 93.7% of the diet of this species, and having been present in the stomachs of 85.9% of the individuals analysed. This species also consumed microcrustaceans, although in a much smaller amount (volume =4.9%). These were mostly Calanoida and Moinidae, which were present in approximately 22% of the stomachs. Even though algae and vegetable matter contributed very little, by volume, to the diet of N. erebi, these items were consumed by a relatively large number of individuals, being present in 31.2% and 47% (respectively) of the stomachs analysed (Table 6.2).

Microcrustaceans were the main food category in the diet of A. agassizii, being consumed by about 99% of the individuals of this species and with a 97.8% contribution by volume. Of this, 62.2% were calanoids and 25.5% were moinid cladocerans. Other food items contributed insignificantly to the diet of this species. Leiopotherapon unicolor also showed a relatively high consumption of microcrustaceans (24.6% by volume), with moinid (13.6% by volume) and daphniid (5.8% by volume) cladocerans as the main taxa consumed. However, the main food category in the diet of L. unicolor was aquatic insects, which contributed 49% of the diet of this species and were present in 80.5% of the stomachs analysed. Interestingly, even though volumetric contributions were greater for Corixidae (17.7%) and Notonectidae (15.8%), the latter were present in relatively few individuals, representing 6.2% of the total number of stomachs analysed. In contrast, larvae of chironomids, which contributed only 7% by volume, were present in 51.9% of the stomachs analysed. Other food items that made significant contributions by volume to Chapter 6. Spatial and temporal variation in diet composition 135

Major categories A. agassizii L. unicolor N. erebi and dietary ite ms % Vol % Freq % Vol % Freq % Vol % Freq Detritus 0.28 1.93 1.98 14.05 93.73 85.86 Algae - - - - 0.36 31.22 Volvox colony - - - - 0.23 25.27 Filamentous algae - - - - 0.02 4.43 Algae matter - - - - 0.11 5.38 Vegetable matter - - 0.23 11.89 0.86 47.07 Microcrustaceans 97.77 99.42 24.60 80.27 4.93 35.23 Calanoida (Copepoda) 65.19 77.26 3.45 36.22 1.34 22.57 Cyclopoida (Copepoda) 0.55 7.90 0.26 15.41 0.04 2.64 Moinidae (Cladocera) 25.46 67.24 13.56 34.05 3.38 22.78 Daphniidae (Cladocera) 6.44 25.82 5.81 38.65 0.09 2.64 Bosminidae (Cladocera) 0.01 0.19 - - 0.01 1.05 Chydoridae (Cladocera) 0.06 0.96 0.28 13.24 0.001 0.11 Sididae (Cladocera) 0.07 0.58 - - - - Ostracoda - - 0.97 14.05 0.06 6.43 Conchostraca - - 0.27 9.19 - - Aquatic insects 0.85 3.08 49.03 80.54 0.001 0.32 Leptophlebiidae (Ephemeroptera) - - 0.47 4.59 - - Coenagrionidae (Odonata) - - 4.00 7.57 - - Zigoptera (Odonata) 0.04 0.39 - - - - Notonectidae (Hemiptera) - - 15.78 6.22 - - Corixidae (Hemiptera) - - 17.73 26.76 - - Chironomidae (Larvae) 0.16 1.35 7.08 51.89 0.001 0.32 Chironomidae (Pupae) 0.66 1.35 3.13 25.95 - - Ecnomidae (Trichoptera) - - 0.76 8.38 - - Leptoceridae (Trichoptera) - - 0.07 0.81 - - Hydroptilidae (Trichoptera) - - 0.01 0.54 - - Molluscs - - 0.02 0.54 - - Planorbidae - - 0.02 0.54 - - Other invertebrates - - 15.94 6.76 0.002 1.58 Rotifera 0.001 0.95 Nematoda - - 0.001 0.27 0.001 0.63 Shrimp (Macrobrachium spp) - - 15.94 6.49 - - Fish - - 2.59 1.35 - - Unidentified 1.10 7.32 5.62 53.24 0.12 4.64 Insect fragment - - 0.54 9.19 0.02 2.64 Digested zooplankton 0.30 1.73 0.02 0.27 0.03 0.11 Digested matter 0.80 5.78 3.70 44.05 0.06 1.79 Unidentifiable matter - - 1.36 10.81 0.01 0.11

Total number of stomachs 519 370 948

Chapter 6. Spatial and temporal variation in diet composition 136 the diet of L. unicolor were shrimps (volume =15.9%) and fish (volume =2.6%), although these items were present only in 6.5 and 1.3% (respectively) of the stomachs analysed.

Unidentified matter made relatively little volumetric contribution to the diets of all three species, comprising up to 5.6% by volume for L. unicolor, even though 53.2% of the stomachs of this species presented some unidentifiable matter. This category contributed only 1.1% by volume to the diet of A. agassizii and 0.12%, by volume, to the diet of N. erebi. For both species unidentifiable matter was present in the stomachs of less than 10 individuals. This food category was characterised by a high level of digestion and/or fragmentation and was excluded from any further analysis.

The overall similarities in volumetric dietary composition of each species for each sampling occasion, coded for species and shown in Figure 6.1, further illustrate these dietary patterns. The two-dimensional NMS ordination explained 85% of the variation in the original unreduced space, with 43.3% being attributed to the first axis alone (the stress for the ordination was 12.4%). The ordination plot shows a relatively tighter cluster of points corresponding to the volumetric dietary samples of A. agassizii arrayed to the left of the ordination space. A band of points from the top right to the middle left, with two points intermingled with the A. agassizii samples, corresponds to N. erebi, whereas another band from the middle to the low right corresponds to L. unicolor samples.

Correlations between volumetric dietary sample points and main food items (r2>0.2), denoted by the vectors, corroborate the patterns of diet composition for each species, with N. erebi diets dominated by detritus and to a lesser extent vegetable matter, L. unicolor consuming mostly aquatic invertebrates, i.e. Notonectidae and Corixidae, shrimps and fish, and A. agassizii diets dominated by Calanoida and Moinidae. Interestingly, some points for L. unicolor and N. erebi are arrayed to the left, forming bands instead of clouds. This is a result of the contribution of microcrustaceans (Calanoida and Moinidae) to the diet of these two species on some of the occasions and/or sites sampled.

Chapter 6. Spatial and temporal variation in diet composition 137

Species A. agassizii detrit L. unicolor N. erebi

veg-m

calan moin

fish xis 2 xis A shrimp noton corix

Stress=12.4%

Axis 1

6.3.2 Spatial and temporal variation in diets of the study species

6.3.2.1 Ambassis agassizii

As observed in Table 6.1, the overall abundance and mean size of the study species varied considerably across lagoons and sampling occasions. Therefore, comparisons of spatial and temporal changes in diet composition were constrained since, according to results from Chapter 5, changes in fish size also lead to changes in proportion of food items ingested and in diet composition. Constraints in interpretation of spatial and temporal patterns are further constrained by a lack of individuals on some sampling occasions and/or sites. In Chapter 6. Spatial and temporal variation in diet composition 138 the case of A. agassizii, individuals from Rainbow and South Callandoon Lagoon A were usually larger than individuals from South Callandoon Lagoon B (Table 6.1 and Figure 6.2a). These were the only sites where this species was abundant.

Similarities of the proportional volumetric dietary data of A. agassizii, averaged for sequential sizes of fish collected on each sampling occasion, are shown in the NMS ordination plot (Figure 6.2). The two-dimensional solution explained 97% of the variation in the original, unreduced space, resulting in a stress of 6.4%. Of the explained variation, 71% is attributed to the first axis. Interestingly, the first axis represents the spatial variation in diets of A. agassizii with Rainbow Lagoon samples arrayed on the left and both South Callandoon lagoons arrayed on the right, whereas the second axis represents a temporal gradient with samples from warmer months, mostly summer, arrayed on the bottom of this axis and samples from cooler months, mostly winter, arrayed on the top of the second axis (Figure 6.2).

Observation of Figure 6.2a shows clear segregation between summer and winter dietary samples, where the winter samples formed a band that extended diagonally from bottom left to top right in the ordination plot and the summer samples formed another band extending approximately horizontally along the bottom of the plot. Interestingly, two of the three points representing samples from Rainbow Lagoon during the late summer (March 2003) are intermingled with the winter samples. These two points correspond to groups of individuals with relatively lower proportional contributions of Moinidae and higher contributions of Calanoida, therefore, being arrayed to the left of the ordination plot (see Figure 6.2b).

When individual lagoons are considered, the ordination plot (Figure 6.2a) shows that there was virtually no overlap in the diet of A. agassizii between lagoons across seasons and/or sampling occasions. Nevertheless, the summer/winter groups corresponding to diets of A. agassizii collected from Rainbow Lagoon were much closer together than the summer/winter groups corresponding to the two South Callandoon lagoons.

Chapter 6. Spatial and temporal variation in diet composition 139

W- S1 Sampling occasions SCB-Feb/02 W- S1 SCB-May/02 W-

W- S2 W- S2 W- S2 Rainbow Lg. S- S2

Axis 2 S- S2 W- S2 W- S2 W- S3 W- S3 South Callandoon B W- S3 S- S5 W- S2 W- S3 S- S4 W- S2 S- S1 S-

S- S1 S- S2

S-

Axis 1

chydo chiron-l chiron-p daphn sidid daphn

cyclop

cyclop chydo xis 2

A calan calan

zi gop moi n moin

bosm

detrit

(b)

Axis 1

The Multi-Response Permutation Procedure (MRPP) corroborated the observed summer/winter differences in dietary composition of A. agassizii by showing a significant difference between the two seasons (A=0.17, p<0.0001). Differences between the three lagoons (A=0.37, p<0.0001) and all eight sampling occasions were also significant (A=0.72, p<0.0001). Note that the chance-corrected within-group agreement, A statistic, which is a descriptor of within-group homogeneity, was lower for sites and sampling occasions then for seasons. Pairwise comparisons showed that the dietary compositions of fish from any one sampling occasion were significantly different from each of the other sampling occasions (p<0.025 in each case), except from South Callandoon Lagoon A (October 2002) and South Callandoon Lagoon B during the summer (p=0.0448), and the March 2003 and May 2002 samples from Rainbow Lagoon (p=0.0945). MRPP between summer/winter samples from each site also revealed that seasonal differences in dietary composition were significant in both Rainbow Lagoon (A=0.10, p=0.009), despite some intermingling of summer and winter groups, and South Callandoon Lagoon B (A=0.42, p=0.001).

Even though A. agassizii consumed relatively large amounts of calanoids and moinid cladocerans (Table 6.2), Indicator Species Analysis (ISA) based on dietary items showed that of the 11 food items consumed by this species, Moinidae (IV=96.6%), Daphniidae (IV=76.6%), Calanoida (IV=58.9%) and, to a lesser degree, detritus (IV=24.1%), were significant contributors to the seasonal (summer/winter) pattern found (p<0.03 in each case) (Appendix 4). ISA revealed that these seasonal (summer/winter) changes in the diet of A. agassizii were attributable to the consumption of relatively greater volumes of Moinidae and detritus during the summer, and Daphniidae and Calanoida during the winter.

The NMS correlations with sample positions in the ordination space (Figure 6.2b) indicate strong associations between seasonal samples and specific dietary items based on the length of vectors corresponding to the main food items), and corroborates the results from ISA by showing a higher correlation (r2 >0.2) between Moinidae and the summer samples, and between Calanoida and Daphniidae and the winter diet samples. Note the absence of a detritus vector in the plot, resulting from a lack of association (r2 <0.2) between this food item and the sample points.

Chapter 6. Spatial and temporal variation in diet composition 141

NMS correlations presented in Figure 6.2b also show associations between the lagoons studied and dietary items in A. agassizii stomach contents, where Rainbow Lagoon samples show clear correlation with calanoids and the two South Callandoon lagoons are more strongly correlated with moinid cladocerans during the summer and with daphniids during the winter. ISA on the three lagoons studied, showed that the only site with a significant indicator food item was Rainbow Lagoon (p=0.001), with an IV of 52.6% for Calanoida. On the other hand, a similar analysis on the eight sampling occasions showed that only three of the dietary items consumed by A. agassizii (Bosminidae, Chydoridae and pupae of Chironomidae) were not indicator species on a given sampling occasion, which indicates a correspondence between individual sampling occasions and specific dietary items.

Figure 6.3 shows the volumetric contributions of food items to the diet of A. agassizii averaged for each of the sampling occasions. As expected from ISA and Figure 6.2a, calanoids were consistently consumed by A. agassizii at Rainbow Lagoon (volumetric range =57.5-98.3%) with smaller contributions by Moinidae (volumetric range =2-42%) and Daphniidae (up to 12% by volume), whereas both South Callandoon lagoons presented relatively lower contributions of calanoids (up to 31%) to the diet of this species and increased consumption of Moinidae and Daphnia (from 26 to 83% and up to 34.2%, respectively). In the case of South Callandoon Lagoon B, summer samples show a higher contribution of Moinidae (volume =82.9%) to the diet of A. agassizii, whereas an increase in consumption of Daphniidae (volume =34.2%) and Calanoida (volume =30.7%) is observed during the winter.

As a result of the high dominance of calanoids in the diets of A. agassizii at Rainbow Lagoon, this site presented the narrowest diet breadth of the lagoons studied (B =0.4734), whereas South Callandoon B showed a higher breadth in dietary items consumed by A. agassizii (B =0.9877), mostly due to the lower dominance of a single food item in the diet of this species. Samples from South Callandoon Lagoon A also showed a relatively shorter dietary breadth of A. agassizii (B =0.6637). Differences in food items ingested between early and late flood diet samples were significantly different (A=0.06, p=0.0150) and on average, A. agassizii presented a lower dietary breadth (B =0.4231 ±0.20) during early flood then during drier periods later after flooding (B =0.8441 ±0.40) (Figure 6.3).

Chapter 6. Spatial and temporal variation in diet composition 142

100 1.4 90

80 1.2

70 1.0 60 0.8 50

40 0.6 Dietary breadth Dietary Percentage volume

30 0.4 20 0.2 10

0 0.0 May/02 Aug/02 Oct/02 Mar/03 Jul/03 Oct/02 Feb/02 May/02 W WSS W W S W Rainbow South South Callandoon A Callandoon B

zigop sidid daphn calan

chiron-l chydo moin detrit

chiron-p bosm cyclop Dietary breadth

6.3.2.2 Leiopotherapon unicolor

Similar to the results for A. agassizii, average sizes and abundances of L. unicolor varied considerably across sampling occasions and/or sites (Table 6.1 and Figure 6.4a), with mostly smaller individuals having been collected from South Callandoon B during the summer, medium-sized individuals during the winter and larger individuals from the remaining sites (Rainbow, Serpentine and Broomfield lagoons).

NMS ordination performed on similarities of proportional volumetric dietary data for averaged sequential sizes of L. unicolor collected on each sampling occasion is shown in Chapter 6. Spatial and temporal variation in diet composition 143

Figure 6.4. The two-dimensional solution explained 91.4% of the total variation, with 73.7% attributed to the first axis and a stress of 11%. Similar to the results for A. agassizii, Figure 6.4a shows a seasonal (summer/winter) segregation for diet samples of L. unicolor. Winter samples formed a relatively tighter group arrayed on the top left of the ordination plot, whereas summer samples were more spread, scattered along the bottom and bottom right of the ordination space.

Groups corresponding to individual lagoons and sampling occasions are clearly identified for the South Callandoon lagoons, although South Callandoon B during the winter had a point considerably distant (to the right) from the main group. This point corresponds to a group of larger individuals of L. unicolor which presented higher proportions of larger aquatic insects, mostly Notonectidae and Corixidae, and Coenagrionidae, as food items, therefore being arrayed left and upwards in relation to remaining points from the same sampling occasion (see Figure 6.4b). Interestingly, summer and winter samples from South Callandoon B are clearly segregated with relatively tight clusters for each occasion, but this is likely to be an artifact of different mean sizes of fish collected on the two sampling occasions. Regarding the sites and sampling occasions with three or less points available in the ordination space, Figure 6.4a shows that for Rainbow Lagoon, L. unicolor had a relatively similar diet on different sampling occasions (November 2001 and October 2002), whereas at Serpentine Lagoon the diet of this species showed some variation between the two occasions sampled (October 2002 and March 2003).

MRPP revealed that the dietary compositions between summer and winter samples were significantly different (A=0.27, p<0.0001), and that differences between sites with more than two points (that is South Callandoon lagoons A and B, and Rainbow Lagoon) were also significantly different (A=0.48, p<0.0001). Similar results were found for sampling occasions with more than two points (in this case data for South Callandoon B was split for summer and winter), with an A=0.66 and p<0.0001.

Note that the smaller the spatio-temporal scale, the greater the A statistic, which indicates tighter groups for the spatial variation (that is sites and/or occasions), than for the temporal scale (summer and winter samples). Chapter 6. Spatial and temporal variation in diet composition 144

W- S2 Sampling occasions W- S2 South Callandoon A SCB-Feb/02 W- S4 SCB-May/02 W- S3 W- S3 Winter Summer RBW-Oct/02 W- S3 SPN-Oct/02 W- S2 W- S3 SPN-Mar/03 SCA-Jul/03 W- S2 RBW-Nov/01 W- S1 BFD-Feb/02 W- S4

W- S3 W- S6 S- > S6

W- S5 W- S4 S- S4

W- S4 xis 2 xis A South Callandoon B

S- S1 South Callandoon B S- S5

S- S1 S- S1 S- >S6 S- S1 S- >S6 S- S3 S- >S6 S- S1 (a) Stress=11% S- >S6

Axis 1

hydrop bayarlept oc leptop conchcoenag ecnom cyclop coenag conch daphn cal an veg-m cyclop leptop ecnom chi-l calan chydochi-p daphn chi-l chi-p

noton ostrac

Axis 2 fish corix corix de trit noton shrimp detrit moin moin

nema fish

shrimp

(b)

Axis 1 Figure 6.4 Two-dimensional NMS ordination plot of spatial and temporal dietary samples for averaged sequential sizes of L. unicolor (a), and (b) joint plot showing the position of food items (+) and food items correlated (r2>0.2) with sample points in ordination space (denoted by vectors). The direction and length of vectors indicate strength of correlation. Each point represents the mean volumetric data for groups of 5-15 individuals. S=summer and W=winter. S1-S6 indicate average size (TL) of individuals within each group, as per Table 5.1. Chapter 6. Spatial and temporal variation in diet composition 145

Pairwise comparisons by MRPP revealed that dietary composition between any sampling occasion (with groups comprising more than two points) was significantly different from each other sampling occasion (p<0.003 in each case). Such comparisons also include the summer and winter samples of South Callandoon B (A=0.43, p<0.0001).

Further analysis revealed clear trends in consumption of food items by L. unicolor between summer and winter. ISA (Appendix 4) revealed that, of the 22 food items ingested by this species, Corixidae (IV=67.9%), shrimp (IV=53.8%), Notonectidae (IV=49.4%) and fish (IV=30.8%) were all significant indicators of summer samples (p<0.03 for each case), whereas Chironomids (Pupae IV=72.5% and larvae IV=68.5%), Cyclopoida (IV=97.6%), Daphniidae (IV=100%), Chydoridae (IV=62.6%), Ecnomidae (IV=62.4%), Conchostraca (IV=56.2%), Coenagrionidae (IV=60.6%), Leptophlebiidae (IV=46.1%) and Calanoida (IV=66%) were significant indicators of winter samples (p≤ 0.02 in each case).

Similar analysis performed on sampling occasions with groups comprising more than two points, revealed that most of the dietary items identified as indicators for the winter samples (plus vegetable matter, IV=50.5%), were also indicators for the winter samples of South Callandoon Lagoon A (p<0.02 in each case), except for Chydoridae (IV=84.4%), Daphniidae (IV=63.4%) and Pupae of chironomids (IV=42.4%) which were indicators for South Callandoon B during the winter (p<0.015 in each case). Interestingly, detritus (IV=73%), Moinidae (IV=72.3%) and Ostracods (IV=69.7%) were significant indicators for the South Callandoon samples during the summer (p=0.001 in each case), whereas shrimp (IV=100%) and Notonectidae (IV=56.5%) were significant indicators for Rainbow Lagoon during the summer (p<0.015 in each case).

These patterns are more easily observed in Figure 6.4b, which shows the correlations between food items and sample points in the NMS ordination space (r2>0.2). The relatively short length of vectors indicate that correlations between individual food items and points in ordination space were not as strong as those observed, for example, for A. agassizii, which is a result of several food items making contributions to the dietary patterns of L. unicolor. Even so, Figure 6.4b illustrates the results described in the previous paragraph, with vectors corresponding to most of the food items ingested by L. unicolor being relatively correlated to the winter samples, and Daphniidae and pupae of chironomid vectors associated with South Callandoon B winter samples. Moinidae and Chapter 6. Spatial and temporal variation in diet composition 146 detritus are relatively correlated with South Callandoon B during the summer, and larger items such as fish, shrimp, Corixidae and Notonectidae are clearly correlated to Rainbow and remaining sites, i.e. Serpentine and Broomfield lagoons.

The volumetric proportion of food items consumed by L. unicolor averaged for each sampling occasion (Figure 6.5) shows the major trends in dietary composition. Firstly, in both South Callandoon lagoons, L. unicolor showed greater consumption of zooplankton, mostly calanoids (volumetric range =0.2-15%), moinid (volumetric range =0.7-58.7%) and daphniid cladocerans (up to 50.1% by volume) and smaller aquatic insects, such as Coenagrionidae (up to 16.4%), chironomids (larvae =9.6-23.6% and pupae =6.5%-14%) and to a lesser extent corixids (up to 4.8%) and notonectids (up to 2%). On the other hand, Moinidae were ingested mostly at South Callandoon Lagoon B (summer), whereas Daphniidae were mostly ingested during the winter in both South Callandoon lagoons.

The remaining study lagoons (Rainbow, Serpentine and Broomfield) showed little to no contribution by microcrustaceans to the diet of L. unicolor (11.4% of calanoids in Rainbow Lagoon and 1.3% by volume of conchostracans on Broomfield Lagoon), whereas larger aquatic insects, mostly Corixidae (volumetric range =10.7-65.2%) and Notonectidae (volumetric range =2.0-56.1%), made consistent contributions to the diet of this species at these sites. Furthermore, these lagoons were the only study sites where L. unicolor fed on unidentified shrimps (probably Macrobrachium, which was abundant in these lagoons) with contributions of this food item to the diet between 3.1 (Serpentine, October 2002) and 59.6% (Rainbow, October 2002) and fish remains, with contributions to the diet of up to 20.1% at Rainbow Lagoon (October 2002).

Because the range of food items ingested by L. unicolor from both South Callandoon lagoons was relatively wide, including microcrustaceans and a variety of aquatic insects, the average dietary breadth for these lagoons was considerably greater (B =2.0477 ±0.29) than for the remaining sites where the diet of this species was more specialised (B =1.2471 ±0.29). Dietary breadth for sampling occasions soon after flooding events showed little variation in comparison to other sampling occasions at the same lagoons (Figure 6.5), as was observed for A. agassizii.

Chapter 6. Spatial and temporal variation in diet composition 147

100 2.5

80 2.0

60 1.5

40 1.0 breadth Dietary Percentage volume

20 0.5

0 0.0 Feb/02 May/02 Jul/03 Nov/01 Oct/02 Oct/02 Mar/03 Feb/02 S WWSSSS S South South RainbowSerpentine Broomfield Calandoon B Calandoon A

fish leptoc corix nema daphn calan

shrimp ecnom noton conch moin veg-m

planor chiron-p coenag ostrac cyclop detrit

hydrop chiron-l leptop chydo Dietary breadth

6.3.2.3 Nematalosa erebi

This species was abundant in most of the study sites and on most sampling occasions and, therefore individuals belonging to a range of sizes were also available (Table 6.1). Similarities of the proportional contributions of food items to diets for fish of averaged sequential sizes are shown in Figure 6.6. The two-dimensional NMS ordination explained 96.4% of the total variance, with 74.6% being attributed to the first axis alone. The resulting stress was 10.2%. With the points on the ordination plot coded for fish size (Figure 6.6a), it becomes evident that the ordination procedure separated the samples of Chapter 6. Spatial and temporal variation in diet composition 148 smaller individuals from medium and larger ones (see Table 5.1 for size class ranges). With the exception of a few points corresponding to sizes S2 (50-69 mm TL), medium to larger individuals (S3-S5, 70-129 mm TL) were arrayed to the bottom right of the plot, while smaller individuals (S2) were spread on the middle and top left of the ordination space. This outcome was expected, given the strong size-related changes in diets of N. erebi recorded in Chapter 5.

Interestingly, a group of three points corresponding to large individuals (S5-S6, 110-139 mm TL) from Serpentine Lagoon was arrayed closer to smaller individuals than to larger ones, and a fourth point corresponding to S3 size from Rainbow Lagoon (July 2003) is closer to individuals with average sizes corresponding to S2 (the mentioned points are indicated in Figure 6.6a). Further analysis on larger individuals separately (see next paragraphs) revealed such points to be outliers, with more than two standard deviations away from the mean and were, therefore, excluded from ordination analysis. Figure 6.6b further strengthens the results from Chapter 5, by showing higher correlations (r2>0.2) of microcrustaceans with smaller individuals of N. erebi, whereas larger individuals are more associated with detritus. Because smaller individuals fed on a variety of food items, points corresponding to sizes S2 and, to a lesser extent S3, are relatively more spread in the ordination plot than points corresponding to larger individuals of N. erebi which primarily consumed detritus (Figure 6.6b).

Ordination on smaller individuals (S2, as per Table 5.1) yielded a lower stress (8.6%) on a two-dimensional solution (Figure 6.7), explaining 93.6% of the variation in the original unreduced space. Of this, 81.2% is attributed to the second axis. From Figure 6.7a, even though some segregation between summer and winter samples is evident, with summer points arrayed in the middle of the plot and winter samples towards the edges, summer and winter samples for Punbougal Lagoon are considerably close together (at the far top of the ordination plot) and segregated from the remaining samples. This relatively tight group of points formed at the top of the ordination plot corresponds to dietary samples from Punbougal and Rainbow (May 2002) lagoons. This cluster corresponds to individuals with greater contributions of detritus and vegetable matter to their diet (Figure 6.7b). Furthermore, although diets of fish from most lagoons and sampling occasions are relatively segregated in ordination space, variation within some groups is considerable.

Chapter 6. Spatial and temporal variation in diet composition 149

Average size S2 S3 S4 S5 S6

Axis 2

(a) Stress=10.4%

Axis 1

bosm rotif

da phn

chi-l calan

moin calan xis 2 xis A cyclop

moin

daphn rotif nema bosm

ostracvolvox

chydo veg-m detrit (b) detrit alg f-alg

Axis 1 Figure 6.6 Two-dimensional NMS ordination plot of spatial and temporal dietary samples for averaged sequential sizes of N. erebi coded for sizes (a), and (b) joint plot showing the position of food items (+) and food items correlated (r2>0.2) with sample points in ordination space (denoted by vectors). Dashed lines indicate outliers for larger individuals (see text). The vector corresponding to ‘detritus’ is 50% its original length. Each point represents the mean volumetric data for groups of 10-15 individuals. S2-S6 indicate average size (TL) of individuals within each group, as per Table 5.1. Chapter 6. Spatial and temporal variation in diet composition 150

MRPP on smaller individuals revealed that dietary composition of N. erebi between summer and winter was not significantly different (p=0.4535), and the A statistic indicates that heterogeneity within summer and winter groups was greater than expected by chance alone (A=-0.006). On the other hand, pairwise comparisons between summer and winter points for each site revealed that the dietary composition of N. erebi was significantly different for Rainbow Lagoon (A=0.58, p=0.0074) and South Callandoon Lagoon A (A=0.42, p=0.0221), whereas differences between summer samples (even though they were taken on different occasions) for Maynes Lagoon were not significant (A=0.17, p=0.0546). Despite the fact that both summer samples from Punbougal Lagoon were considerably closer together in the ordination plot, MRPP revealed that these differences were significant (A=0.24, p=0.0265). This result is a consequence of different proportional contributions of vegetable matter to the diet of N. erebi at this site and the very low number of food items ingested (see further analysis below). Differences between the four sampling sites (A=0.29, p=0.0002) and sampling occasions (A=0.64, p<0.0001) with more than one sample point were also significant.

Even though dietary data are available for four sites, ISA revealed that only two sites had significant indicator food items (Appendix 4). South Callandoon Lagoon A had Cyclopoida (IV =42.9%), Rotifera (IV=42.9%), Daphniidae (IV=42.9%) and Bosminidae (IV=42.9%) (p<0.045 in each case) as significant indicator food items, whereas the indicator food items for Punbougal Lagoon were detritus (IV=44.3%) and vegetable matter (IV=59.7%) (p<0.003 in each case). The relatively low IVs and high p values at South Callandoon Lagoon result from the fact that these items were consumed on only one of the three occasions when this lagoon was sampled. Similarly, the relatively low IV for detritus at Punbougal Lagoon reflects the generally high proportion of this food item on almost all sampling occasions. ISA for each sampling occasion resulted in generally low or non- significant (p>0.05) IVs for the main food items consumed by smaller individuals of N. erebi (Moinidae IV=31.6%, p=0.011, Calanoida IV=26.5%, p=0.011 and detritus IV=20.4%, p=0.1), which is due to the fact that these items were consumed in relatively large amounts on most of sampling occasions.

Chapter 6. Spatial and temporal variation in diet composition 151

S S W Sampling occasion S S W S S RBW-May/02 SCA-May/02 MNS-Oct/02 PBL-Oct/02 SCA-Mar/03 RBW-Mar/03 W MNS-Mar/03 S PBL-Mar/03 RBW-Jul/03 SCA-Jul/03

S S S xis 2 xis S

A S S

S W

S W S

S S

W W (a) W Stress = 8.6%

Axis 1

f-alg alg

detrit

veg-m

detrit

veg-m

vol vox cyclop Axis 2 rotif daphn

moin calan moin calan

l-chi

rotif bosm (b) daphn

Axis 1

Figure 6.8 shows the volumetric proportion of food items of smaller N. erebi averaged for sampling occasions. Even though microcrustaceans were preferred by smaller individuals of N. erebi, contributions of detritus to the diet of this size group were still large on most of the sampled occasions, reaching up to 99.4% by volume.

100 1.4 90

80 1.2

70 1.0

60 0.8 50

40 0.6 Percentage volume Percentage breadth Dietary

30 0.4 20 0.2 10

0 0.0 May/02 Mar/03 Jul/03 May/02 Mar/03 Jul/03 Oct/02 Mar/03 Oct.02 Mar/03 W SWW S W SSSS Rainbow South Maynes Pungbougal Calandoon A

chiron-l moin rotif f-alg volvox

bosm cyclop veg-m alg detrit

daphn calan Dietary breadth

Even so, microcrustaceans represented an important source of food, mostly in the form of Calanoida (up to 81.1%), Moinidae (up to 56.6%) and Daphniidae (up to 22.8%), with the remaining food items contributing less than 3.3% ot any given occasion. As a consequence, dietary breadth was generally low, with lower average breadth for Rainbow (B =0.7760 ±0.42) and Punbougal (B =0.2848 ±0.11) lagoons, where detritus or calanoids Chapter 6. Spatial and temporal variation in diet composition 153 dominated as food items. In South Callandoon and Maynes lagoons the average dietary breadth was wider (B =1.2896 ±0.13 and B =1.0493 ±0.24, respectively) due to a lower proportion of detritus associated with higher contributions from other food items, such as Moinidae and Calanoids. It is important to note that variation in dietary composition in Rainbow Lagoon was relatively high, with proportions of detritus and calanoids varying from 98.7 and 0% (respectively) in May 2002 to 18.9% of detritus and 81.1% of calanoids in July 2003 (Figure 6.8). Recent flooding did not appear to have any marked effects on diet composition.

Ordination on larger individuals (>S2, as per Table 5.1) resulted in a stress of 9.6% on a two-dimensional solution (Figure 6.9), explaining 96.4% of the total variance in the original space. Of the explained variance, 54.6% is attributed to the first axis. The results from Figure 6.9a are somewhat similar to those for smaller individuals, with no clear segregation between summer and winter dietary samples of N. erebi, and a tight cluster of points on the bottom of the plot, with points becoming sparser on the top of the ordination space.

Observation of Figure 6.9b indicate that contributions of fewer dietary items, mostly detritus and vegetable matter, characterise the tighter points on the bottom of the plot, whereas points arrayed to the top of the ordination plot present a relatively less restricted diet also including microcrustaceans (Calanoida and Moinidae) and Volvox colonies. Interestingly, sample points arrayed to the top of the ordination plot showed some segregation between summer and winter samples, with points corresponding to summer being arrayed to the top left of the ordination space.

Nevertheless, MRPP revealed that dietary composition of larger individuals of N. erebi was significantly different between summer and winter (p=0.0298), although the chance- corrected within-group agreement, A statistic, indicates that the heterogeneity within groups is very close to expectation by chance (A=0.02). MRPP on sites (excluding samples from outlier sampling occasions, i.e. Serpentine Lagoon and Rainbow Lagoon on July 2003) indicates that differences in proportional diet composition among the five sites are significant (A=0.30, p<0.0001).

Chapter 6. Spatial and temporal variation in diet composition 154

S Sampling occasion RBW-Nov/01 S RBW-May/02 SCA-May/02 RBW-Aug/02 W S W RBW-Oct/02 MNS-Oct/02 PBL-Oct/02 SCA-Mar/03 RBW-Mar/03 S W RIV-Mar/03 MNS- Mar/03 S PBL-Mar/03 W RIV-Jul/03

W W W

Axis 2 W

S W W W S- S S S W S S S S S S S S W S S S W S W S S WS W W (a) W W S S S W Stress=9.6% W S S S S S Axis 1

daphn

calan moin

volvox xis 2 xis

A moin chydo

calan volvox

alg

detrit veg-m veg-m nema ostrac detrit (b) f-alg l-chi

Axis 1

Pairwise comparisons between sampling occasions or seasons for each site revealed that diets of larger N. erebi between sampling occasions for Rainbow Lagoon were significantly different (p<0.03 in each case) except for comparisons between October 2002 (summer) and November 2001 (summer) (A=-0.01, p=0.5234), October 2002 (summer) and August 2002 (winter) (A=0.14, p=0.0537) and October 2002 (summer) and March 2003 (summer) (A=0.06, p=0.1947). For the remaining study sites, differences between sampling occasions or seasons were significantly different for all sites (p<0.042 in each case) except for Maynes Lagoon (A=-0.08, p=0.8902).

ISA (Appendix 4) revealed that Moinidae (IV=52.7%), Calanoida (IV=76.4%) and Volvox colonies (IV=74.8%) were significant food item indicators for South Callandoon Lagoon A (p<0.002 in each case), Ostracods were significant indicators for Maynes Lagoon (IV=74.9%, p=0.001), algae for Rainbow Lagoon (IV=34.3%, p=0.02), vegetable matter for Punbougal Lagoon (IV=29.3%, p=0.003) and detritus for the River site (IV=20.8%, p=0.006). Because the last three items were fairly common among sites, their overall IVs were relatively low. Similar analysis performed on sampling occasions revealed that indicator food items were found for only five of the 13 sampling occasions (Appendix 4).

The volumetric proportion of food items consumed by larger individuals of N. erebi averaged for sampling occasions (Figure 6.10) clearly indicates the importance of detrital matter to the diet of larger individuals of this species. This food item contributed to at least 91.6% of the diet of larger N. erebi, on most sampling occasions. Interestingly, on two occasions which correspond to the outliers identified in the ordination analysis, contributions by detritus were very low, 64.2% (Rainbow Lagoon on July 2003) and only 0.9% for Serpentine Lagoon. In the latter, despite the relatively large size of individuals (≥ S5), moinid cladocerans was the most consumed food item with 93.9% of contribution by volume of the overall diet. At Rainbow Lagoon (July 2003) Cyclopoida were often consumed despite the relatively large size of N. erebi individuals (S3), contributing 31.2% of the diet by volume on this sampling occasion.

The remaining food items contributed very little, by volume, to the diet of larger N. erebi, with proportions ranging from nil to 5.8% (Volvox colonies, on South Callandoon A, March 2003). As a result, dietary breadth was generally very low, with lower averages for Chapter 6. Spatial and temporal variation in diet composition 156 the River site (B =0.0354 ±0.01), Maynes (B =0.0522 ±0.01) and Punbougal (B =0.0756 ±0.02) lagoons, where detritus was the dominant food item, whereas Rainbow (B =0.1861 ±0.18), South Callandoon A (B =0.2412 ±0.11) and Serpentine (B =0.2893 ±0.14) lagoons presented relatively higher breadths in diet due to higher proportions of other food items, such as microcrustaceans. As with smaller individuals, recent flooding did not appear to influence patterns in dietary composition or breadth, although in lagoon sites subject to flooding (Rainbow and South Callandoon A), N. erebi showed relatively higher breadth in diet.

100 1.0

90 0.9

80 0.8

70 0.7

60 0.6

50 0.5

40 0.4 breadth Dietary Percentage volume Percentage

30 0.3

20 0.2

10 0.1

0 0.0 Jul/03 Jul/03 Oct/02 Oct/02 Oct/o2 Mar/03 Mar/03 Mar/03 Mar/03 Mar/03 Mar/03 Nov/01 Aug/02 May/02 W May/02WWS S S S W S S SS S SWRainbowSouth Maynes Pungbougal Macintyre Serpentine Calandoon A River

chiron-l chydo cyclop f-alg volvox

nema daphn calan alg detrit

ostrac moin veg-m Dietary breadth

Chapter 6. Spatial and temporal variation in diet composition 157

6.4 Discussion

6.4.1 Overall influences on fish diets

The diets of A. agassizii, L. unicolor and N. erebi from isolated lagoons on the floodplain of the Macintyre River were temporally and spatially variable, both between seasons and sites. Even though closely comparable studies are not available, the general diets of the study species conform to those described in the literature (Merrick and Schmida 1984, McDowall 1996, Pusey et al. 2004). On the whole, results suggest that during the study period, food resources available for fish in floodplain lagoons of the Macintyre River were relatively scarce since, despite the range of food items recorded in fish stomachs (Table 6.2), individual species fed mostly on relatively few food categories and few food items within each category, and most of the variation observed was associated with different proportional contributions of a few food items.

The results of the present study provide evidence that the dietary composition of the study species on the floodplain of the Macintyre River is influenced by both spatial variation in the availability and composition of prey items, and the time of year during which lagoons were sampled, as well as by the feeding modes of the fish species themselves. Despite the relatively low range of food items consumed by the species studied, the results suggest that lagoons on the floodplain of the Macintyre River are highly variable regarding the array of food resources available to fish. As expected from dietary data for A. agassizii, L. unicolor and N. erebi available from other studies (see Pusey et al. 2004), the three species show distinctly different diet composition, reflecting differences in species specific morphological and behavioural characteristics. Nevertheless, some overlap in diet composition was observed, as all three species consumed zooplankton at some stage of their life history, as demonstrated in Chapter 5 and in the results of the present chapter.

In Chapter 5, it is indicated that the study species undergo size related changes in the abundance and/or composition of prey items consumed, which raises the issue of possible effects of size on spatial and temporal variations in diet composition. Furthermore, it is suggested that the degree of ontogenetic shift in food items consumed by the study species was dependent on the season and/or site where the individuals were collected. Indeed, the Chapter 6. Spatial and temporal variation in diet composition 158 overall results from the present chapter for A. agassizii and L. unicolor (see ordination plots, Figures 6.2 and 6.4) show that, even though size was an important factor affecting the diet of these two species on a local scale (i.e. within sites), most of the variation in dietary composition was observed at a larger scale, between sites, and/or on different sampling occasions. The range of size classes of L. unicolor examined in the present study confirms that this species underwent size related changes in diet, with larger individuals feeding on larger prey, such as fish, shrimps and larger aquatic invertebrates (Corixidae and Notonectidae). Even so, similarly sized individuals presented quite different diets across different sites and/or sampling occasions. Nematalosa erebi also showed great variation in composition of food items ingested by smaller and larger individuals. When smaller and larger individuals were analysed separately, ordination showed that variation in diet between sites and/or sampling occasions was considerable. However, although smaller individuals of N. erebi had relatively variable diets between sampling occasions and/or sites, based on a range of microcrustaceans and detritus, larger individuals of this species presented a much more consistent diet, based almost exclusively on detrital material.

Diet breadth of the study species reflected their intrinsic morphological and behavioural characteristics. Leiopotherapon unicolor showed a generally higher range of food items consumed based on microcrustaceans, aquatic insects, shrimp and fish. Fish with opportunistic feeding habits, such as L. unicolor, display little morphological or behavioural specialisation to capture and process specific prey (Gehrke 1988b) and, whereas large prey can be disabled by repeated grasping of less protected regions or appendages, smaller prey just need to be large enough to be handed individually. According to Gehrke (1988b), this behavioural characteristic, associated with the use of visual, olfactory and auditory stimuli to locate prey, broadens the potential diet of L. unicolor by overcoming morphological restrictions such as mouth gape, and enabling this species to feed on prey of a wider variety of types and range of sizes. On the other hand, A. agassizii presented a narrower breadth due to its diet being based mostly on microcrustaceans. As discussed in Chapter 5 and demonstrated by Kennard (1995) small microphagic species, like A. agassizii, are often limited in prey choice by constraints in body size, and consequently mouth gape and locomotory ability. Regarding N. erebi, even though this species fed mostly on detrital matter, the presence of microcrustaceans and vegetable and algal material in its diet is reflected in a relatively higher (in comparison Chapter 6. Spatial and temporal variation in diet composition 159 with A. agassizii), although more variable, dietary breadth. Although morphological characteristics are thought to play an important role in diet composition of N. erebi, as discussed in Chapter 5, the relatively wider range of food items consumed by this species can be explained by the wide range of sizes of individuals collected and is, therefore, the result of the combination of smaller individuals (which consumed relatively large amounts of microcrustaceans) and larger individuals (which are mostly algivorous/detritivorous) in the calculations of the index of dietary breadth of this species.

6.4.2 Influence of spatial variation on the diets of individual species

The three study species showed differences in diet composition among sites, and even though the major food categories were similar within species, composition and abundance of prey items consumed changed substantially for each species across the study sites. Tropical rivers have large spatial heterogeneity across a continuum of spatial scales that ranges from microhabitats to landscape scale (Winemiller and Jepsen 1998). Habitats on the floodplain can be associated with very different hydrologies, soils, water quality and vegetation (Wissmar et al. 1981, Day and Davies 1986) that ultimately affects the food resources available for fish. Such variability in natural characteristics of floodplain lagoons, which was demonstrated for the study sites in Chapter 3, is likely to have affected the food resources available for fish among sites on the floodplain of the Macintyre River. Interestingly, Marshall et al. (in review) found that temporal variation in faunal composition of macroinvertebrates within floodplain waterholes (across four sampling times over two years) in Cooper Creek was greater than spatial variation within a single time period at any scale investigated. This suggests that temporal variation in composition of potential food items (at least for A. agassizii and L. unicolor) may have been greater than spatial variation in the present study area.

6.4.2.1 Ambassis agassizii

The main components in the diets of the study species were microcrustaceans, aquatic insects and detritus. Ambassis agassizii consumed mostly microcrustaceans (Calanoida, Moinidae and Daphniidae) and can, therefore, be classified as a microphagous carnivore, Chapter 6. Spatial and temporal variation in diet composition 160 feeding mostly in mid-waters and only occasionally in benthic areas. This is supported by indications that the closely related A. nigrippinnis, collected from south-east Queensland rivers, tended to feed within littoral vegetation and in the mid-water (Milton and Arthington 1985). There is indication that in Rainbow Lagoon, where A. agassizii was most abundant, individuals tended to concentrate on the relatively steeper (0.6 to 1.0 m deep) margins and nearby aquatic structures, such as large debris and root masses (author’s personal observation). This is consistent with habitat preferences (see Arthington and Milton 1983, Pusey et al. 2004) and feeding habits of this species and its relative susceptibility to predation. According to Miller (1979a), small open-water foragers, like A. agassizii, have been shown to be more susceptible to predation than species that feed in more sheltered areas like aquatic vegetation patches.

Spatial variation in diets of A. agassizii is apparent from studies in two different areas in south-east and central Queensland. Arthington (1992) showed that A. nigrippinnis, a junior synonym for A. agassizii (Allen and Burgess 1990), consumed mostly ephemeropterans (37%), chironomids (13%) and terrestrial items (17%) (including Formicidae and Diptera) in tributary streams of the Brisbane River, whereas individuals from the Burdekin and Burnett Rivers presented higher proportions of microcrustaceans in their diet (Pusey et al. 2004). Pusey et al. (2004) argue that such variations reflect differences in physical characteristics and food availability of the rivers studied, since individuals from the Brisbane River tributaries were obtained from shallow lotic stream reaches, whereas individuals from the Burdekin and Burnett Rivers were collected from deeper and slow flowing waters. In the present study, the greater abundance of microcrustaceans in the diet of A. agassizii is likely to be due to the fact that the study sites where A. agassizii was collected presented low to nil water flow, therefore favouring the consumption of microcrustaceans rather than other small invertebrates such as aquatic insects. Comparisons of diets of a similar species, A. agrammus, from different rivers in Cape York Peninsula revealed that, although invertebrate larvae dominated the diet of this species, composition of microplankton varied across rivers (Pusey et al. 2000). Changes in diets are also reported between sandy creekbeds and a range of floodplain sites (e.g. backflow, corridor and floodplain billabongs) in the Magela and Nourlangie catchments, Alligator Rivers region, Northern Territory, for A. agrammus and A. macleayi (Bishop et al. 2001). Ambassis agrammus fed primarily on chironomid larvae in creekbeds, whereas the microcrustacean component, particularly cladocerans, copepods and ostracods, was Chapter 6. Spatial and temporal variation in diet composition 161 much stronger in the floodplain sites. Similar results were found for Ambassis macleayi, although chironomids were an important food item in the floodplain sites as well (Bishop et al. 2001).

Differences in diets of fish sampled from different waterbodies may reflect several factors, including developmental stages and body sizes, sampling methods, time of day and season, and the influence of habitat structure on food resources (Arthington 1992). Even though sampling methods and time of day for sampling were standardised in the present study (seasonal factors being discussed in the next section), individuals of any given species from different sampling occasions or sites often presented different body sizes and, therefore, some comparisons were constrained. Ambassis agassizii presented different diets between Rainbow and both South Callandoon lagoons resulting from higher consumption of calanoids in the former and Moinidae and Daphniidae in the latter, but differences in average sizes between fish individuals from these two lagoons hampers further interpretation. From results presented in Chapter 5, consumption of Calanoids tended to increase with increasing size of A. agassizii, whereas proportions of Moinidae tended to decrease in larger individuals. Given that individuals from South Callandoon were generally smaller than individuals from Rainbow Lagoon, the lower proportions of Calanoids would be expected. Mouth size and locomotory ability may also have restricted the diet of A. agassizii to small microcrustaceans, since many of the food items available in the study sites and consumed by other species were larger prey items such as corixids and notonectids. Even though Arthington (1992) showed high contributions of larger prey such as ephemeropterans, trichopterans and coleopterans to the diet of A. agassizii, she analysed a relatively wider range of body sizes (22-45 mm SL) in flowing water river sites, where mid-water microcrustaceans are less likely to be consumed by small fish like A. agassizii. The relationship between prey and fish sizes is relatively well documented (Werner and Gilliam 1984, Sheldon and Meffe 1993) and Kennard (1995) showed that larger prey items (i.e. corixids and notonectids) gave relatively greater proportional contributions to the diet of larger individuals of A. macleayi, whereas cladocerans were less important in their diet (see also Chapter 5, Section 5.4).

Interestingly, individuals of A. agassizii from South Callandoon lagoons A and B showed quite similar diets (within the same season), even though their body sizes were different. Therefore, this may suggest that at least in some circumstances, dietary differences among Chapter 6. Spatial and temporal variation in diet composition 162 sites may have been spatial rather than ontogenetic. Similar results for L. unicolor (see next Section) also indicate that diets of this species were relatively similar between South Callandoon A and B (in the same season), but further comparisons are not possible due to different sizes of individuals across sites. The close proximity of these two lagoons and their overall similarities (as indicated in Chapter 3) are likely to be the main reasons for the relatively similar diets of fish feeding in these lagoons. Such an argument is supported by results from Marshall et al. (in review) who found that macroinvertebrate assemblages tended to be more similar the closer the samples were collected within the physical landscape in floodplain waterholes of Cooper Creek, central Queensland.

6.4.2.2 Leiopotherapon unicolor

Leiopotherapon unicolor consumed a relatively wide range of prey items across the study sites, from a variety of aquatic insects and microcrustaceans in the South Callandoon lagoons (including ephemeropterans, chironomids and trichopterans, and moinids, daphniids and cladocerans) to corixids, notonectids, shrimp and fish in Serpentine and Rainbow lagoons. Therefore, this species can be classified as an opportunistic meiophagic omnivore that feeds from mid-water as well as benthic areas. Morphological and behavioural characteristics further supporting this argument are discussed in Chapter 5 and described by Gehrke (1988b). Even though the herbivorous component was an important part of the diet of L. unicolor in the Alligator River Region (Bishop et al. 2001), contributions made by material from vegetative origin in the present study were very small and probably incidental (Merrick 1974).

The diet of L. unicolor, in the study sites, was similar to findings for this species from elsewhere (Pusey et al. 2004). Dietary data on L. unicolor are available from several studies undertaken mostly in Queensland, on a range of natural and modified habitats, such as artificial reservoirs, riverine habitats and floodplain billabongs and lagoons (Pusey et al. 2004). However, descriptions of spatial variation in dietary composition of this species from habitats similar to those in the present study are scarce. Comparisons of diets of L. unicolor from rivers in Cape York Peninsula revealed little differences in diet (Pusey et al. 2000). These authors reported that, despite differences in consumption of trichopterans larvae for L. unicolor across the Normanby, Pacoe and Stewart rivers, aquatic insect larvae Chapter 6. Spatial and temporal variation in diet composition 163 were the dominant prey across all three rivers. On the other hand, Bishop et al. (2001) reported significant changes in composition of food items consumed by L. unicolor from different habitats in the Magela catchment in the Northern Territory. For example, in escarpment mainchannel waterbodies, L. unicolor ate terrestrial plant material and equal proportions of filamentous algae, aquatic insects (chironomid larvae), fish and terrestrial insects. However, in lowland sandy creekbeds this species ate aquatic insects (baetid and chironomid larvae), but little terrestrial insects and little plant material, though micro- and macrocrustaceans were also consumed. Further, in backflow billabongs, the diet was based primarily on aquatic insects, but with a greater variety of fish present (including plotosids, M. splendida inornata, Ambassis spp., G. giuris and M. mogurnda). In corridor waterbodies L. unicolor consumed equal proportions of microcrustaceans (mainly cladocerans), aquatic insects (baetid and chironomid larvae and corixids) and fish. The diet in floodplain habitats had a large organic component (presumably digested fish) and fairly equal proportions of aquatic insects, algae, and fish (Bishop et al. 2001). Bishop et al. (2001) explained such variations on the basis of local invertebrate availability and on overall differences in substrate composition. In the present study the main changes in diets of L. unicolor observed between lagoon sites are clearly associated with the size of individuals sampled. Generally larger fish from Serpentine and Rainbow lagoons fed mostly on shrimps, fish, corixids and notonectids, whereas smaller individuals from the South Callandoon lagoons consumed greater amounts of microcrustaceans and smaller aquatic insects (ephemeropterans, chironomids and trichopterans). These results are in accordance with Bluhdorn and Arthington (1994) who observed that larger individuals of L. unicolor tended to consume mostly shrimp and occasionally fish, whereas smaller individuals (≤ 70 mm SL) consumed greater amounts of insects (larvae and ephemeropterans) and microcrustaceans (copepods and ostracods) (see also Arthington et al. 1994).

6.4.2.3 Nematalosa erebi

The main food categories in the diet of N. erebi were detritus, and to a much lesser extent, microcrustaceans, mostly moinid cladocerans. Microcrustaceans were mainly eaten by juvenile individuals, although large individuals from Serpentine and Rainbow lagoons also fed on these prey items, whereas larger individuals consumed mostly detritus and small Chapter 6. Spatial and temporal variation in diet composition 164 amounts of vegetable matter and algae. On the whole, this species can be classified as a microphagic omnivore, with smaller fish being generally omnivorous and larger fish being mainly detritivorous. Hydro-acoustic sampling has demonstrated that similar species (N. papuensis and N. flyensis) are present throughout the water column (Bishop et al. 2001). Although N. erebi may also be observed throughout the water column, they are most commonly found in the lower one third (Kennard 1995). This further emphasizes the preference of this species for substrate detritus from bottom waters. This species was reported to be frequently observed ‘pecking’ at the substrate and spitting out the sand it accidentally ingested (Bishop et al. 2001).

Summary dietary data from Pusey et al. (2004) showed that N. erebi feeds mostly on detritus and algae material. Given the considerable dominance of detritus over the other food items consumed by N. erebi on the study sites, relatively little spatial variation in diet composition of this species was observed. Bishop et al. (2001) described differences in composition of food items for N. erebi between sandy creekbeds and a range of floodplain sites (e.g. backflow, corridor and floodplain billabongs) in the Magela and Nourlangie catchments, Alligator Rivers region, Northern Territory, but these are mostly related to changes in either algal or detrital matter, and to a less extent incidental organic material, across sites. Spatial patterns related to the consumption of other prey items, such as microcrustaceans, reported by those authors reflected a higher proportion of juvenile fish, which frequently ate microcrustaceans. They also found that N. erebi fed more extensively on phytoplankton in downstream habitats with muddy substrata, and on diatomaceous periphyton in escarpment mainchannel waterbodies. In the present study, algal material represented a minor part in the diet of N. erebi and substrate composition across study sites was very similar, comprising mostly of mud (Chapter 3). Alternatively, Pusey et al. (2004) emphasized that contribution of organic detritus to the diet of N. erebi must also be considered as indicating the effect of habitat structure on trophic style. The low abundance of algae in the floodplain lagoons of the Macintyre River, during most of the period studied, and the relative abundance of fine and coarse organic detritus (see Chapter 7), may have played a major role in the dietary composition of N. erebi. On the other hand, major dietary differences were observed between particular sites, that is Punbougal Lagoon, and the remaining sites (regardless of the size of individuals) as a diet based predominantly on detritus was observed in the former lagoon. Samples of zooplankton from Punbougal Lagoon for stable isotope analysis (see Chapter 7) showed very low abundances and it is Chapter 6. Spatial and temporal variation in diet composition 165 likely that this was the main reason for the absence of microcrustaceans in the diets of small N. erebi from Punbougal Lagoon. This indicates that prey availability was an important factor influencing diets composition between sites as fish were merely feeding on food resources available at time of capture.

Regarding the remaining sites, diets of smaller individuals of N. erebi were quite similar, based mostly on detritus and microcrustaceans (mostly Moinidae, Daphniidae and Calanoida) with most of the variation being accounted by temporal changes in proportional volume of a few dietary items, namely detritus, Moinidae and Calanoida. Larger individuals of this species showed consistent consumption of detritus, and with the exception of Serpentine Lagoon, very little variation in diet composition was observed between sites. It is important to bear in mind that ordination scores for large individuals of N. erebi are based on extremely low proportions of all food items except for detritus. Therefore, the importance of contributions of food items other than detritus to the diet of this species is significantly overestimated in the ordination plots and joint plots shown in Figure 6.9. This is clearly observed in Figure 6.10. Note also that the length of the detritus vector is actually twice the length shown in Figure 6.9b. The reasons for the consumption of almost exclusively microcrustaceans (mostly Moinidae) by larger individuals in Serpentine Lagoon are unknown as detritus was fairly abundant in this location (as well as in the other study sites). Given the fact that plankton samplings for stable isotopic analysis in Serpentine Lagoon have consistently produced large numbers of zooplankters (mostly Moinidae), it is possible that N. erebi was merely taking advantage of an abundant resource available at the time of sampling. This can also explain similar results for one sampling occasion (July 2003) in Rainbow Lagoon, where Cyclopoida were relatively abundant in the diet of larger N. erebi.

6.4.3 Temporal patterns and the effects of flooding on diets of individual species

The annual cycle of flooding and drying of floodplain waterbodies is known to affect aquatic food webs dramatically as aquatic plants and invertebrates show a significant burst in production during the early stages of wet-season flooding (Winemiller 1990, Walker et al. 1995, Bunn et al. in press). Temporal changes in diets of the three fish species studied Chapter 6. Spatial and temporal variation in diet composition 166 were clearly demonstrated in the results of the present study, although the extent of such changes depended on the sites. In general, most changes in diet composition were associated with summer/winter variations in composition of food items ingested. Chapter 3 demonstrated that the study period was relatively dry in relation to previous years and, furthermore, only minor flooding of the Macintyre River was recorded. As a result many of the study sites dried out or remained isolated and only South Callandoon A and Rainbow lagoons were flooded during the study period (detailed information on flow history of the study sites is given in Tables 3.1 and 6.1). Given the absence of floods of greater magnitude during the study period, the observed summer/winter changes in diets of the study species are likely to be the result of successional changes in composition of prey items as the dry season progressed.

6.4.3.1 Ambassis agassizii

Most fishes display relative dietary plasticity, feeding on different foods as they become seasonally available (Lowe-McConnell 1987). This is particularly true for riverine fishes in more seasonal rivers in Australia where flooding and scouring, as well as periods of low or absent flow may destroy aquatic habitats and disrupt the production of aquatic food resources (Arthington 1992, Bishop et al. 2001). In the present study, A. agassizii from South Callandoon Lagoon B showed strong temporal changes, mostly associated with fluctuations in the consumption of Moinidae, as higher abundances of this item were observed during the summer and lower abundances of Moinidae, associated with increased consumption of Daphniidae, were observed during the winter. Bishop et al. (2001) found similar results, for two species of Ambassis (A. agrammus and A. macleayi) collected from floodplain sites at the Alligator River Region, where the microcrustacean component of the diets was large throughout the study period with seasonal fluctuations in copepods and cladocerans across seasons. It is argued that such fluctuations are associated with dry-wet seasonal variations in the availability of such prey items in the studied billabongs (Marchant 1982, Bishop et al. 2001). Interestingly, in the present study, temporal changes in diets of A. agassizii from Rainbow Lagoon were less conspicuous, with a diet based mostly on calanoids (and to a lesser extent Moinidae) throughout the period of study (see Section 6.4.3.1).

Chapter 6. Spatial and temporal variation in diet composition 167

6.4.3.2 Leiopotherapon unicolor

Similar to the results for A. agassizii, L. unicolor showed relatively similar diets across different sampling occasions (both during summer) in Rainbow Lagoon, based mostly on shrimp. In contrast, Serpentine Lagoon showed marked temporal changes in diets of this species, not necessarily related to summer/winter patterns, but associated with different sampling occasions. As for the differences in diets across sampling occasions observed in South Callandoon Lagoon B, these are likely to be a result of the considerable differences in average sizes of fish collected between the two occasions sampled. Interestingly, the diets of L. unicolor in South Callandoon lagoons A and B during the same season (summer) were relatively similar, with higher contributions of Daphniidae and Chironomidae (both larvae and pupae). Even though ordination shows overall seasonal differences in diet of L. unicolor, Figure 6.4a indicates that these results may have been confounded by a further variable, which is fish size. As shown in this ordination, the major patterns in the data seem to be related to the average size of the individuals sampled, therefore some of the temporal changes across lagoons are not readily interpreted. Significant changes in diet of L. unicolor from floodplain sites at the Alligator River Region mostly associated with wet and dry seasons are described by Bishop et al. (2001). They found differences regarding the abundance of different aquatic insects, microcrustaceans, and to a less degree algae, and claimed that measured resurgences of microcrustaceans and peak abundances of aquatic insect larvae in the study billabongs generally during the wet season were reflected in the diet of L. unicolor (Bishop et al. 2001). Additionally, more fish appeared in the stomachs of this species during the wet to late-wet seasons when the number of fish species captured in the billabongs had arguably peaked (Marchant 1982, Bishop et al. 2001).

In a similar fashion, Arthington et al. (1994) found that, despite little variation with season, the diet of L. unicolor from North Pine Dam (south-east Queensland) changed markedly after a rise in water level as this species switched from a balanced diet of animal and plant sources to feeding on resources which became abundant as a consequence of extensive shoreline inundation, namely fish and insects. They argued that this was a response to either increased abundance or higher vulnerability of prey after the loss of cover that dense aquatic macrophytes provided at low water levels. There is no strong evidence that the observed differences in diet of L. unicolor in the present study are the result of flooding of Chapter 6. Spatial and temporal variation in diet composition 168 the study sites and the larger contribution of fish and shrimp to the diet of this species in Serpentine and Rainbow lagoons is clearly associated with larger fish sampled from these sites in comparison with those from both South Callandoon lagoons.

Sheldon et al. (2003) and Marshall et al. (in review) suggested that the connection and disconnection of floodplain waterbodies in dryland rivers can influence the invertebrate community in a sequential fashion associated with fluctuations in hydrology and that, after disconnection each waterbody will behave as a separate unit with assemblage compositions diverging in a manner that reflects those species present at the time of disconnection. The absence of a significant flood at Rainbow Lagoon since November 2001 and the fact that water levels were relatively unchanged throughout most of the study period may have had a significant influence on the composition of prey items available in this site, as variations in diet of A. agassizii and L. unicolor at Rainbow Lagoon were less conspicuous than, for instance, at the South Callandoon lagoons. Therefore, this lack of variation in food items consumed is likely to be a result of the relatively less variable water regime in Rainbow Lagoon. Because A. agassizii and L. unicolor feed on resources that are more subject to seasonal and spatial changes (i.e. zooplankton and aquatic invertebrates) (see Marchant 1982, Bass et al. 1997, Bishop et al. 2001), both of these species presented more evident variation in composition of diets in more variable sites such as the South Callandoon lagoons. This is further evidence that the composition of available prey items plays an important role in the diet of these species in the floodplain lagoons studied. Arthington et al. (1992b) found that diet of L. unicolor from Barker-Barambah Creek showed no major seasonal trends and was dominated by crustaceans (shrimps). They argued that this may simply indicate the relative abundance and/or ease of capture of these freshwater crustaceans in relation to other food sources available such as small fish and aquatic insects.

6.4.3.3 Nematalosa erebi

Very little temporal variation in dietary composition of N. erebi was observed in the present study and most of the differences in diet composition are accounted for by ontogenetic changes associated with morphological changes on feeding apparatus as fish grow (see Chapter 5, Section 5.4). When individuals were separated by sizes, the temporal Chapter 6. Spatial and temporal variation in diet composition 169 component of the diet of larger individuals was still subject to little change, with detritus predominating as the main food item. Smaller individuals, on the other hand, showed differences in diet with time due to variations in consumption of different microcrustacean food items (mostly Calanoida and Moinidae) and in the amount of detritus eaten. Similar results are reported by Bishop et al. (2001) for floodplain habitats in the Alligator River Region. Even though seasonal changes in diets of N. erebi were observed by Bishop et al. (2001), detritus was the main dietary component of this species with temporal variations in the consumption of algae and the presence of unidentified organic material. Changes in the consumption of detrital matter, though mentioned by Bishop et al. (2001), are argued to be the result of anoxic conditions at the bottom of the study billabongs, therefore, N. erebi filter-fed on phytoplankton from the mid-water zone rather than entering anoxic bottom waters to feed on the substrate. Interestingly, Arthington et al. (1992b) observed a marked increase in consumption of algae by N. erebi from the Barker-Barambah Creek during the summer, associated with an increase in gut fullness, whereas the detritus component decreased during the same period. Since algal blooms were more noticeable during summer months, these authors argued that the dominance of algae in the diet of N. erebi may have been a function of its availability. In the study sites, food availability seems to be the major factor influencing the diets of the study species as, in the case of N. erebi, the main food item ingested, namely detritus, was considerably abundant in all study sites as revealed by organic matter samples for the stable isotopes study (Chapter 7) and also author’s personal observations. When detritus was not abundant in the diets of this species, this trend was associated with increases in consumption of microcrustaceans, in both small and large individuals. Interestingly this happened only at sites where the zooplankton were also relatively abundant (see Chapter 7).

In Cooper Creek, Balcombe et al. (in review) argued that fish diet changes reflected successional changes of invertebrate species composition resulting from rapid changes in resources available due to flooding. In the Macintyre River system, the results for A. agassizii and L. unicolor show that despite the absence of major flooding, diet composition of these species varied considerably with time. The overall lack of correspondence between changes in diet composition and sampling occasions associated with minor flooding events does not necessarily downplay the role of floods in the dietary ecology of fish from the Macintyre River floodplain. From results presented in Chapter 3, the study period was relatively dry with only minor flooding of some of the study sites, therefore Chapter 6. Spatial and temporal variation in diet composition 170 limiting meaningful assessment of the effects of flooding on fish diets. Even so, the fact that fish collected from the floodplain lagoons more subject to flooding presented more variable diets is an indication that flooding influences fish diets in such environments. The results from the present study indicate that without flooding of natural floodplains, diets of fish became constrained by the resources available within floodplain lagoons. Because direct exploitation of floodplain invertebrates by fishes is one of the most important benefits of seasonal inundation, if river-floodplain connectivity is eliminated or disturbed through anthropogenic activities, fish dietary ecology and population processes are likely to be affected. Therefore aquatic floodplain habitats and food resources need to be maintained to ensure the conservation of fishes in floodplain ecosystems.

6.4.4 Aspects of flow regulation

According to Kingsford (2000) the substitution of a variable flooding pattern with a permanent one and the consequent loss of wet-dry cycles may have a significant impact on floodplain waterbodies. Invertebrate communities may become less diverse or abundant than those in more variable floodplain habitats (Boulton and Lloyd 1991) or may become dominated by species adapted to standing waters (Shiel 1990). In the present study, the analysis of whether fish diets differ in response to water regulation is subject to the same constraints related to the analysis of temporal variations, due to the confounding effect of individual fish size, as previously discussed in terms of diet variations between sites and sampling occasions. Even so, it is possible to claim that at least one type of water regulation (permanently elevated water levels), present in Rainbow Lagoon, had some effect on fish diets. Diets of A. agassizii and L. unicolor were to a great extent similar through time in this site, whereas diets of these species from the smaller and more temporally variable South Callandoon Lagoon B showed significant changes through time, mainly between summer and winter. Despite the unusual diet of N. erebi from Serpentine Lagoon, there was no evidence from the results that water regulation had any effect on the diet of this species. On the contrary, most of the variation in diets of all three species studied seemed to be related to temporal changes in abundances of food items consumed and/or size related patterns. The former could be a function of flow regulation, such as in Rainbow Lagoon, but further work is necessary to determine the effect of permanent lagoon flooding on fish diets. Chapter 6. Spatial and temporal variation in diet composition 171

6.5 Conclusions

Results from the present study show that the characterisation of fish diets can be significantly affected by variability in the spatial and temporal patterns of food availability and ontogenetic patterns in food consumption, being further supported by the results from Chapter 5. Although comparable to studies elsewhere, dietary composition of the study species was relatively narrow, based on three major food categories (aquatic insects, microcrustaceans and detritus) and few food items within each category.

Spatial variation in diet composition, i.e. across the study sites, was observed for all three study species and was mostly associated with changes in contribution of specific food items to fish diets at each study site. Although masked by differences in average size of individuals across sites, the results indicate that differences in prey item availability at each study site played a major role in the observed patterns. As indicated from temporal data, flow history may also have had an important effect on differences in fish diets composition across lagoons, but further evaluation is needed due to overall lack of major flooding events during the study period and, as previously mentioned, differences in sizes of fish between sites.

Temporal changes in dietary composition of the study species seemed to reflect changes in availability and abundance of specific food items associated with major seasonal patterns (summer/winter differences). Sampling occasions associated with relatively recent minor flooding in some of the study sites did not reflect clear patterns of dietary changes in comparison with sampling occasions prior to or long after such floods. Given the low magnitude or total absence of flooding events and flood influences on lagoons during the study period, it is likely that the observed dietary variation was a consequence of successional changes in composition of the aquatic fauna as the dry season progressed (see Bishop et al. 2001, Sheldon et al. 2003, Balcombe et al. in review). Spatial and temporal variations in diets of N. erebi were not as significant as those observed for A. agassizii and L. unicolor, as larger and, to a lesser extent, smaller N. erebi fed mostly on detritus material, a relatively abundant and widespread resource, throughout most sites and sampling occasions. Chapter 6. Spatial and temporal variation in diet composition 172

The total absence of food items of terrestrial origin in the stomach contents of A. agassizii and L. unicolor indicates that the study species were reliant on resources produced within the floodplain lagoons, even after the occurrence of small floods. This is likely to be the result of the absence of major flooding during the study period, therefore preventing allochthonous sources of food from entering floodplain lagoons. However, detrital material of terrestrial origin made high contributions to the diet of N. erebi, suggesting the importance of riparian sources to this species in the study sites.

Of the sites subject to water regulation (Rainbow and Serpentine lagoons) there was sufficient data for meaningful comparisons only for Rainbow Lagoon. Overall temporal changes in diets of A. agassizii and L. unicolor in Rainbow Lagoon were not as significant as, for instance, variations in the diets of these species from the smaller and more temporally variable lagoons, such as South Callandoon Lagoon B. This is probably the result of a combination of the absence of major flooding of this site throughout most of the study period and permanently elevated water levels which may have inhibited significant variation in composition of prey items available, since the overall habitat characteristics remained relatively similar during the study period. Furthermore, despite the diet of N. erebi from Serpentine Lagoon being based mostly on zooplankton, there was not sufficient evidence from the results to argue that this was a result of water regulation in Serpentine Lagoon.

6.6 Implications for the study of fish diet composition

Results presented in this chapter suggest that a variety of factors contribute to the variation in diet composition of fish in floodplain lagoons. Even though flow variability has been recognised as playing a central role in the structure and functioning of river-floodplain systems (Junk et al. 1989, Walker et al. 1995), in the absence of such events local attributes such as spatio-temporal variations in general physical (morphology, habitat characteristics and, possibly, degree of water management) and biological (fish species and sizes being considered and differences in available food resources) characteristics of floodplain lagoons become increasingly important contributors in determining dietary composition of fish. Even though some lagoons were flooded, diets changed temporally Chapter 6. Spatial and temporal variation in diet composition 173 mainly in relation to seasons and the drying phase of isolated lagoons, rather than due to the influx of riparian food types sourced from the floodplain.

The present study has found little evidence of allochthonous food types in diet of A. agassizii, L. unicolor and, to a lesser extent small, N. erebi. Based on stomach contents, autochthonous resources seemed to be sustaining these fish species in floodplain lagoons of the Macintyre River during the period of study. Alternatively, indication of the importance of local riparian sources to the diet of larger N. erebi is also presented, as this species fed almost exclusively at detrital material, most likely from terrestrial origin. However, instantaneous diet composition information, obtained from stomach content analysis, may not reveal the true dependencies of fish on particular food resources. This is particularly the case for detritivorous species such as N. erebi, which present highly unidentifiable stomach contents.

Alternatively, stable isotope analysis can provide an estimation of mean levels of organic carbon and nitrogen assimilated by higher consumes such as fish, in relation to several dietary sources with distinguishable isotopic signatures (e.g. aquatic invertebrates, zooplankton) and trace the ultimate sources of carbon actually assimilated by fish consumers. Therefore, when combined with stomach contents analysis stable isotopes can be used to assess the potential role of unidentifiable or unquantifiable food items (e.g. algal or detrital matter) in the diets of larger consumers such as fish. Chapter 7 attempts to determine the ultimate sources of carbon (allochthonous versus autochthonous) assimilated by the study species of fish and relate the results with dietary data from stomach contents analysis presented in the present chapter.

Chapter 7. Energy sources for fish using stable isotope analysis 174

7 Stable isotope analysis of energy sources for fish in floodplain lagoons

7.1 Introduction

Food webs in tropical waters are often complex and based on relatively few energy sources (Lowe-McConnell 1987). In freshwaters, food webs are thought to run from either bottom detritus, through microorganisms to detritus-feeding invertebrates or fish, to several levels of piscivore; or in the pelagic zone, from phytoplankton to zooplankton, to zooplankton feeders, then to one or more levels of piscivore (Lowe-McConnell 1987). In river systems, the detrital chain is argued to be more important, based largely on allochthonous materials from headwaters or decomposition of aquatic macrophytes (the River Continuum Concept, Vannote et al. 1980) or, in the case of large floodplain rivers, from terrestrial inputs and dissolved organic matter originating laterally from the floodplain (the Flood Pulse Concept, Junk et al. 1989). In both cases, strong reliance on allochthonous inputs has been emphasized.

However, these views have been challenged in recent studies and a number of researchers have been suggesting that autochthonous primary production is an important, often major, contributor to metazoan production in rivers (see reviews by Thorp and Delong 2002, Bunn et al. in press). Studies on the Orinoco River and its floodplain showed that, even though macrophytes and terrestrial litter composed about 98% of the potentially available carbon, phytoplankton and periphyton were the ultimate carbon sources for invertebrates and fish (Lewis et al. 2000, 2001). Similarly, Araujo-Lima et al. (1986) showed that detritivorous fish in the Amazon River floodplain derived most of their carbon from phytoplankton production. Even though the Flood Pulse Concept has been thought to be an appropriate model for large floodplain rivers in Australia (Walker et al. 1995, Pusey and Arthington 2003), studies on dryland floodplain rivers have shown that despite the large amounts of terrestrial carbon available, there is no evidence that terrestrial carbon is a significant contributor to the aquatic food web (Bunn and Davies 1999, Bunn et al. 2003). Instead, these studies show that algae growing along the shallow littoral zone of waterholes and floodplain lagoons are the major sources of energy for aquatic consumers. An alternative perspective to the RCC and FPC, the Riverine Productivity Model (Thorp and Chapter 7. Energy sources for fish using stable isotope analysis 175

Delong 1994), highlights the importance of local in-stream production by phytoplankton, benthic algae and/or other aquatic plants as opposed to direct inputs of organic matter from the adjacent riparian zone, therefore indicating that the role of autochthonous sources has been underestimated in previous models for large river ecosystems.

Studies on food webs have shown that the use of empirical dietary data tend to limit the ecological value of food web depictions and cannot provide information on the rate of ingestion and assimilation of food by the species studied (Glasser 1983, Yodzis 1993). On the other hand, an isotopic analysis can provide an estimation of the mean level of organic matter actually assimilated by a given species (when signatures are distinct among food sources, see Section 7.2.1). Therefore, when combined with stomach contents analysis, stable isotopes can be used to assess the potential role of unidentifiable or unquantifiable food items (e.g. algal or detrital matter) in the diets of larger consumers such as fish.

7.1.1 Aims

Despite the fact that dryland rivers in Australia feature extensive floodplains and a network of anastomosing tributaries providing great terrestrial-water interface (Walker et al. 1995), for most of the time these systems exist as disconnected and highly turbid waterholes and floodplain lagoons that act as refugia for aquatic organisms (Morton et al. 1995, Bunn and Davies 1999). Given the high turbidity of the study sites (Chapter 3, see also Houldsworth 1995), and the corollary assumption that the lower diversity and abundance of macroinvertebrates and macrophytes presented in these systems may relate to high turbidity (Boddy and Bales 1996), aquatic plant and algal production should be limited by light and by the absence of water flow during dry periods. Given these features it should be expected that fish consumers would be dependent on energy and nutrients derived ultimately from floodplain exchange during and soon after flooding, and by direct inputs from the riparian vegetation during dry periods.

Therefore as a follow up to the previous chapter on spatial and temporal changes in diets of target fish species and by combining information presented from stomach contents analysis, this present chapter aims to investigate the ultimate sources of energy supporting the study species using stable isotope analysis to trace sources of organic carbon and Chapter 7. Energy sources for fish using stable isotope analysis 176 nitrogen through the food web, in order to assess (1) the relative contribution of ultimate energy sources, for instance, algal versus detrital sources, and (2) potential major sources of food for the study species, based on results from Chapter 6 (e.g. zooplankton, aquatic insects and invertebrates), taking into account the availability of those sources over the study period. Specific objectives of the present chapter are:

1. to understand the types and origin of the ultimate sources of energy supporting the food web and fish consumers in floodplain lagoons of the Macintyre River, given that the current models for large river ecosystem function vary considerably in their predictions of the importance of terrestrial versus in-stream production to the aquatic food web, 2. to identify possible seasonal changes in energy flow pathways from sources to consumers in relation to minor flooding of specific floodplain lagoons and/or as the dry season progressed, 3. to identify spatial variation in the ultimate sources of energy for fish consumers in different floodplain lagoons with different flow regimes, and in relation to flow regulation.

7.2 Methods

7.2.1 Stable isotope analysis: review of methods

Stable isotope ratios offer an effective natural tracer for following energy and nutrient flows in ecosystems. Stable isotopes are atoms of a particular element that share a common number of protons but vary in number of neutrons and do not decay with time. Because the chemical properties of elements are largely determined by the number of electrons, which do not vary between isotopes, they have common chemical properties, while being analytically distinguishable (Schimel 1993). Stable isotope ratios are measured using an isotope ratio mass spectrometer, which measures the ratio of the heavy and light isotopes in a sample and compares this to a standard. Differences in absolute abundances of isotopes within a sample are usually very small and subject to sample heterogeneity, fluctuation in the mass spectrometer and sample preparation (Hayes 1982). Therefore, the stable isotopic ratio of a sample is compared to a standard, so that fluctuations will be reflected equally in both standard and sample. The differences in the Chapter 7. Energy sources for fish using stable isotope analysis 177 ratio of the heavy to light isotope are calculated in δ notation and are expressed as parts per million (‰) (Lajtha and Michener 1994).

Stable isotopes can be used to identify the sources of energy to consumers and to trace the flow of energy through ecosystems. In one of the earliest studies, Fry et al. (1978) used analysis of δ13C from vegetation, grasshopper guts and grasshoppers to show that herbivores have the same isotopic composition as their food. Furthermore, they concluded that specialist species had carbon isotope ratios similar to those of the plants on which they fed, whereas generalist species had ratios reflecting the mixture of species they consumed. The stable isotope method is thus based on the fact that certain isotopes fractionate in predictable ways as elements travel through the food web (Fry 1991, Michener and Schell 1994). The nitrogen isotope value (δ15N) is used to determine feeding and other trophic relationships among animals and plants, because there is significant enrichment of 15N between organism and diet (about 2-4‰). Since they change consistently through the food web, 15N isotopes can be used to evaluate a consumer’s trophic level (DeNiro and Epstein 1981). The δ13C value is used to determine the sources of energy, because the 13C content of an organism reflects the 13C contents of its diet with little or no change. In numerous studies on food web relationships, the dual relative ratios of stable isotopes of carbon (13C and 12C) and nitrogen (15N and 14N) in biota have been used as indicators of food source to resolve ambiguities in sources of energy and energy flows (Fry 1991, Creach et al. 1997, Johannsson et al. 2001, Leite et al. 2002). Furthermore, when combined with gut contents data, stable isotopes are an important tool in assessing the potential role of unidentifiable, undetectable or unquantifiable prey items in the diets of consumers (Johannsson et al. 2001), as well as the origins of organic matter found in digestive tracts and the role played by bacteria in the diet of macro consumers (Creach et al. 1997).

7.2.2 Study design

To understand the variations in contribution of allochthonous (organic matter from riparian origin) versus autochthonous (algae and zooplankton) organic carbon and nitrogen to the food web, particularly to fish, predicted sources, e.g. algae, organic matter and aquatic plants, were sampled from lagoons on the floodplain of the Macintyre River. Prey items known to be important to the diet of the study species, i.e. zooplankton, aquatic insects and Chapter 7. Energy sources for fish using stable isotope analysis 178 shrimp (Chapter 6), were also sampled, as well as the study species (Ambassis agassizii, Leiopotherapon unicolor and Nematalosa erebi). Seven of the study sites, six floodplain lagoons and one site in the main channel of the Macintyre River, were sampled on three occasions: summer (prior to and after flooding) and winter (several months after flooding) (see Chapter 3 for site information). The first samplings were undertaken late in the dry season of 2002, early in the summer of that year (20-31 October), the second sampling occasion occurred near the end of the 2002-2003 summer, soon after the wet season (10-20 March 2003) when some of the study sites experienced minor flooding, and the third sampling occasion occurred in the winter of 2003 (15-25 July), during the dry season.

7.2.3 Collection of primary sources and consumers

Major primary sources of terrestrial and aquatic organic carbon and nitrogen were collected from each study site. Fallen leaves from major riparian trees (mostly Eucalyptus spp.) (RIP.VEG) were collected by hand from the margins of each lagoon and the river site. Benthic detritus was collected with dip and hand nets and wet-sieved in the field into fine (250 µm to 1 mm) and coarse (>1 mm to 1 cm) particulate organic matter fractions (CPOM and FPOM, respectively).

Algae samples were taken from the shallow littoral margins both directly off the mud surface and from submerged wood or rocks, and washed in the field to remove any associated organic debris. These samples consisted mostly of Rhizoclonium spp. and Cladophora spp. (Prescott 1970, Sainty 1973, Entwisle et al. 1997). Because of the high levels of suspended sediment (Chapter 3) and the presence of particulate organic matter in the water, it was not possible to take useful samples of phytoplankton. Aquatic plants, where present, were collected by hand from the lagoon margins. The aquatic plants collected included the submerged water milfoil (Myriophyllum spp.) (SUB.PLANT) and the floating water primrose (Ludwigia spp.) and red azolla (Azolla spp.) (FLOAT.PLANT) (Sainty 1973, Stephens and Dowling 2002) (see Chapter 3). Zooplankton (calanoids and cladocerans - moinid and daphniid) (ZOOPL) was sampled at dusk and dawn by towing a 250 µm plankton net just below the surface of the water.

Chapter 7. Energy sources for fish using stable isotope analysis 179

Aquatic insects were abundant in many of the study sites though not very diverse. Insects sampled included the hemipterans Corixidae (CORIX), which feed mostly on detritus and particulate plant and animal matter (Ingram et al. 1997), Notonectidae (NOTON), which are carnivores, and will prey on zooplankton, other aquatic insects and small fish (Williams 1980), and odonate larvae (ODON) (mostly Gomphidae and Coenagrionidae), which are ferocious predators, commonly consuming rotifers, molluscs, small crustaceans, other aquatic insects and small fish (Ingram et al. 1997). Other insects sampled included larvae of chironomids (CHIRON), which are mostly herbivores and detritivores (Ingram et al. 1997), but these were abundant only on few occasions. Aquatic invertebrates were collected by hand net and a small seine net (2 m long with 1 mm mesh) from the bottom of the lagoons and from aquatic vegetation (when present). Other invertebrates, mostly shrimps (Macrobrachium spp.) were sampled by seining.

The three species of fish studied were sampled by seining as described in Chapters 4 and 5 and with fyke nets set overnight. Note that fish from fyke nets were not used in stomach contents analysis due to possible changes in food items ingested while these fish remained trapped for several hours. Where possible three replicate samples of each of the potential food sources and consumers, and five replicates of fish samples, were collected from different areas in each lagoon. All animal and plant samples were immediately refrigerated, frozen within 4 hours of collection and stored frozen for stable isotope analysis.

7.2.4 Sample preparation

In the laboratory, primary sources (tree and aquatic plant leaves, organic matter and algae) were rinsed in distilled water and in the case of algae samples, any remaining organic debris was removed. All samples were oven-dried at 60°C for 36 to 48 hours and then ground to a powder-like consistency in a ring grinder. For each of the zooplankton samples collected, half was treated in 10% HCl for approximately two hours to remove carbonates from exoskeletons for δ13C analysis. The remaining half of the sample was not treated in acid and was used for δ15N analysis (Bunn et al. 1995). Aquatic insects were prepared whole and pooled for each site. Similar to the zooplankton samples, shrimp exoskeletons were removed to prevent contamination from carbonates of non-dietary Chapter 7. Energy sources for fish using stable isotope analysis 180 origin. Their digestive tracts were also removed as they could represent a significant source of contamination from unassimilated material (Bunn et al. 2003). Samples of muscle tissue were then taken from the tail. Samples of muscle tissue were also taken from individuals of each fish species, from the region above the lateral line and adjacent to the dorsal fin (after the fish was scaled and skinned), in each case recording total and standard body length and weight. Fish samples analysed were from single individuals of the three study species (A. agassizii, L. unicolor and N. erebi). All animal samples were oven-dried at 60°C for 24 to 48 hours and then ground by hand with a mortar and pestle.

Dried and ground samples were oxidized at high temperature and the resultant CO2 and N2 were analysed for percentage C, N and stable isotope ratios with a continuous-flow isotope-ratio mass spectrometer (Europa Tracermass and Roboprep, Crewe, England) at the Stable Isotope Analysis laboratory at Griffith University. Ratios of 13C/12C and 15N/14N were expressed as the relative per million (‰) difference between the sample and the conventional standards (PeeDee Belemnite carbonate and N2 in air), where:

RR− δX (‰) = sample stan dard ×1000 Rstan dard where X =13C or 15N and R =13C/12C or 15N/14N.

Measurement precision was approximately 0.1‰ for 13C/12C and 0.3‰ for 15N/14N.

7.2.5 Data analysis

Data were first analysed using the stable isotope values of sources and consumers for each sampling occasion and site, and then averaged for flooded and non-flooded sites across each sampling occasion. Samples were considered enriched or depleted if their ratios departed from standards in a positive (i.e. more heavy isotopes) or negative direction, respectively. Field studies generally find a carbon enrichment of 0 to 2‰ and 3 to 5‰ for nitrogen, for each increase in trophic level (see Peterson and Fry 1987). The relationships between δ13C and δ15N signatures of the studied species of fish, and between the stable isotope signatures and fish size, were examined using simple Pearson correlations. Chapter 7. Energy sources for fish using stable isotope analysis 181

To evaluate the contribution of the more abundant primary sources to the biomass of consumers, a two-source linear mixing model based on δ13C signatures of zooplankton and either organic matter (CPOM) or algae was used. Where all three major sources were available, a three-source linear mixing model based on δ13C and δ15N signatures of zooplankton, algae and organic matter (CPOM) was used (Phillips 2001, Phillips and Gregg 2001). To account for fractionation the assumed 3 to 3.5‰ per trophic level increase from the nitrogen isotopic signature of the fish was added to δ15N values of food sources. δ13C isotopic fractionation is close to zero, so an adjustment of 0.1-0.2‰ per trophic level was made (DeNiro and Epstein, France 1996, Vander Zanden and Rasmussen 2001, Bunn et al. 2003). Mixing models with an unfeasible solution, summing more than 105% of contribution of sources to a given fish consumer, usually as a result of fish consumers being either too 13C depleted or 15N enriched in relation to the primary sources, were expressed as 100% contribution of the closest source and mentioned when appropriate. For the mixing models only, algal δ13C and δ15N values for Rainbow Lagoon on October 2002 were estimated from algae values from South Callandoon Lagoon on the same sampling occasion, as both sites had similar algal compositions.

7.3 Results

7.3.1 Primary sources

Algal δ13C values were generally more enriched than any other primary source, even though there was considerable variability in algae δ13C values across sites and sampling occasions (from −10.9 to –31.9‰) (see Table 7.1). In the sites subject to flooding, lower δ13C values were observed on the sampling occasions after flooding, whereas non-flooded sites presented generally more enriched algal δ13C values. Submerged aquatic plants (Myriophyllum spp.) were also 13C enriched (−9.0 to −12.4‰) in comparison with other primary sources and other aquatic plants. The floating aquatic plants, water primrose (Ludwigia spp.) and red azolla (Azolla spp.), had similar average δ13C values generally consistent between sites, thus they were pooled together. These species were consistently 13C depleted with δ13C values ranging from −27.4 and −29.6‰ (Table 7.1). Chapter 7. Energy sources for fish using stable isotope analysis 182

------Jul. 2003

------Mar. 2003

) ) ) ) ) ) )

------0.2 0.1 0.1 0.4 1.3 0.1 0.8 Broomfield Lagoon Broomfield 10.9 24.7 25.7 29.1 20.5 24.8 19.1 19.2 23.4 Oct. Oct. 2002 ± ± ± ± ± ± ± − − − − − − − − − ( ( ( ( ( ( (

) ) ) ) ) ) ) )

- - - - - 0.5 0.1 0.2 0.2 0.3 0.3 0.2 0.2 28.1 27.1 28.0 27.9 27.4 29.3 24.7 25.7 21.6 21.8 20.7 22.1 16.7 17.9 Jul. 2003 ± ± ± ± ± ± ± ± − − − − − − − − − − − − − − ( ( ( ( ( ( ( (

) ) ) ) ) ) ) ) )

------2.0 0.1 0.1 0.1 0.8 1.0 1.1 0.6 0.9 26.1 28.4 29.6 19.1 19.6 22.7 16.7 19.6 21.3 17.1 Mar. 2003 ± ± ± ± ± ± ± ± ± − − − − − − − − − − of three of occasions, pre-flood sampling ( ( ( ( ( ( ( ( (

) ) ) ) ) ) ) ) )

- - - - 1.1 1.9 1.1 0.2 2.8 0.2 1.0 0.7 0.4 Serpentine Lagoon Serpentine 17.0 24.9 26.7 29.2 25.9 22.6 24.5 22.0 22.2 23.8 22.3 22.2 Oct. Oct. 2002 ± ± ± ± ± ± ± ± ± − − − − − − − − − − − − ( ( ( ( ( ( ( ( (

) ) ) ) )

------0.8 0.1 0.6 0.5 0.2 23.4 24.1 27.4 29.6 27.3 21.5 23.2 25.1 26.0 21.2 Jul. 2003 ± ± ± ± ± − − − − − − − − − − ( ( ( ( (

) ) ) ) ) ) ) SD, for n = 3-7 n = samples)(where values individual 3). for n < SD, or

------0.3 0.3 0.2 0.2 2.7 0.6 0.1 23.3 23.8 27.7 22.9 16.6 17.3 20.1 22.0 25.7 15.8 19.8 16.8 ± Mar. 2003 ± ± ± ± ± ± ± − − − − − − − − − − − − ( ( ( ( ( ( (

) ) ) ) ) ) ) Maynes LagoonMaynes

------0.5 0.4 2.5 1.2 1.6 1.6 1.3 11.6 24.5 27.6 28.4 21.4 26.0 20.5 23.3 22.0 Oct. Oct. 2002 ± ± ± ± ± ± ± − − − − − − − − − ( ( ( ( ( ( (

) ) ) ) ) )

- - - - - 9.0 0.2 0.1 0.4 0.5 0.4 0.1 13.9 15.2 23.6 24.4 28.6 28.3 19.7 20.0 23.9 24.4 23.8 25.0 24.2 Jul. − 2003 ± ± ± ± ± ± − − − − − − − − − − − − − ( ( ( ( ( (

) ) ) ) ) ) ) ) ) )

------0.3 0.2 0.3 0.6 0.8 0.3 0.6 1.1 0.1 1.0 23.9 26.8 28.5 28.4 11.4 19.7 21.0 19.2 24.4 24.1 Mar. 2003 ± ± ± ± ± ± ± ± ± ± − − − − − − − − − − ( ( ( ( ( ( ( ( ( (

) ) ) ) ) ) ) )

------0.3 0.6 0.3 0.5 4.2 0.7 0.2 1.0 Punbougal Lagoon Punbougal 25.5 25.3 28.6 12.4 16.8 20.6 26.5 26.9 Oct. Oct. 2002 ± ± ± ± ± ± ± ± − − − − − − − − ( ( ( ( ( ( ( (

) ) ) ) ) )

------1.4 0.3 0.3 0.2 3.1 1.6 27.9 27.6 28.6 28.9 27.3 25.8 26.3 26.4 Jul. 2003 ± ± ± ± ± ± − − − − − − − − ( ( ( ( ( (

) ) ) ) ) ) )

------0.4 0.1 0.3 0.6 0.4 1.0 0.7 31.9 27.6 28.8 29.6 26.9 23.9 24.8 27.5 26.2 25.3 Mar. 2003 ± ± ± ± ± ± ± − − − − − − − − − − ( ( ( ( ( ( (

) ) ) ) ) ) om floodplain lagoons and one site on the Macintyre River, on each River, the Macintyre floodplain and one site on lagoons om Macintyre RiverMacintyre ------0.2 0.7 1.2 1.4 1.7 1.0 15.5 27.4 27.8 28.0 26.6 30.8 29.5 29.7 27.1 Oct. Oct. 2002 ± ± ± ± ± ± − − − − − − − − − ( ( ( ( ( (

) ) ) ) ) ) ) )

- - - - - 1.3 0.2 0.2 0.5 0.3 1.2 0.3 0.8 17.6 27.0 27.8 29.0 32.2 22.3 29.6 29.1 25.5 26.4 28.8 28.7 32.7 Jul. 2003 ± ± ± ± ± ± ± ± − − − − − − − − − − − − − ( ( ( ( ( ( ( (

) ) ) ) ) ) ) ) ) ) )

- - - 2.8 0.2 0.3 0.3 0.1 0.4 2.1 0.8 0.3 1.0 1.2 26.1 24.4 27.0 29.2 31.7 26.7 25.2 27.4 25.8 28.2 26.9 27.4 27.5 28.4 Mar. 2003 ± ± ± ± ± ± ± ± ± ± ± − − − − − − − − − − − − − − ( ( ( ( ( ( ( ( ( ( (

) ) ) ) ) ) ) ) ) ) )

- - - - - Rainbow Lagoon 0.1 0.1 0.3 0.3 4.3 2.3 0.7 0.3 0.5 0.3 1.2 27.3 27.5 28.7 32.9 28.2 30.2 27.3 30.0 27.0 31.0 27.9 Oct. Oct. ± ± ± ± ± ± ± ± ± ± ± 2002 − − − − − − − − − − − ( ( ( ( ( ( ( ( ( ( (

) ) ) ) ) ) ) ) ) ) ) )

- - 0.7 0.1 0.7 0.3 0.9 0.1 2.3 0.7 0.2 1.1 0.2 0.1 18.9 27.5 28.4 28.7 29.0 37.3 28.2 30.0 31.2 35.6 30.9 34.3 32.5 34.3 33.8 Jul. 2003 ± ± ± ± ± ± ± ± ± ± ± ± − − − − − − − − − − − − − − − ( ( ( ( ( ( ( ( ( ( ( (

) ) ) ) ) ) ) ) ) ) )

- - - - 4.1 0.3 0.5 0.2 0.5 0.6 0.5 3.8 0.5 0.3 0.9 23.4 27.6 28.1 29.2 28.6 36.5 28.1 29.3 27.3 29.3 30.0 30.5 29.2 Mar. 2003 ± ± ± ± ± ± ± ± ± ± ± − − − − − − − − − − − − − ( ( ( ( ( ( ( ( ( ( ( Lagoon A ) ) ) ) ) ) ) )

------South Callandoon Callandoon South 0.5 0.2 0.3 0.5 0.2 1.6 0.3 0.5 13.7 25.9 28.1 29.7 32.6 26.0 29.1 28.7 29.0 Oct. Oct. 28.4 2002 ± ± ± ± ± ± ± ± − − − − − − − − − ( ( ( ( ( ( ( (

(small) (large) Stable carbon isotope ratios (‰) of sources and consumers fr sources (‰) of Stable ratios isotope carbon Sample type A. agassizii L. unicolor N. erebi N. erebi rimary sources rimary consumersrimary ish econdary consumers Floating aquatic Floating plants aquatic Submerged plants Chironomidae (larvae) P Algae FPOM CPOM vegetation Riparian P Zooplankton Corixidae S Notonectidae Odonata Shrimp F

(Octobermeancorrespond (March to ( 2003) values (July 2003). 2002), winter summer and post-flood Data post-flood Table 7.1 Chapter 7. Energy sources for fish using stable isotope analysis 183

------Jul. 2003

------Mar. 2003

) )

------0.1) 0.1) 1.2) 1.1) 0.1) 0.4 0.3) 0.1 Broomfield Lagoon Broomfield 8.2 6.4 4.0 6.3 9.6 9.8 Oct. Oct. 10.4 10.0 13.8 2002 ± ± ± ± ± ± ± ( ( ( ( ( ( (

) )

- - - - - 1.4) 1.4) 0.3) 0.3) 0.5) 7.8 0.4) 0.7) 0.3 7.7 6.4 5.3 9.1 Jul. 19.8 21.9 10.6 12.0 16.0 16.5 19.5 21.6 15.6 15.0 2003 ± ± ± ± ± ± ± ± ( ( ( ( ( ( ( (

) )

------0.2) 0.2) 0.1) 0.7) 0.1) 0.5 0.6) 0.9) 0.4) 0.6 5.0 5.2 5.5 9.8 9.4 10.5 12.1 16.8 17.9 12.4 Mar. 2003 ± ± ± ± ± ± ± ± ± ( ( ( ( ( ( ( ( (

) ch of threesampling ch of occasions, pre-flood

- - - - 2.7) 2.7) 0.2) 0.6) 0.2) 2.6 0.2) 1.1) 0.3) 0.6) Serpentine Lagoon Serpentine 5.2 6.1 5.3 Oct. Oct. 14.6 14.3 12.6 15.1 15.4 16.5 16.6 18.2 16.8 2002 ± ± ± ± ± ± ± ± ± ( ( ( ( ( ( ( ( (

)

------0.3) 0.3) 0.2) 0.4) 0.6 0.4) 8.1 8.2 8.3 6.1 8.2 7.9 9.1 Jul. 11.0 11.7 11.2 2003 ± ± ± ± ± ( ( ( ( (

) )

------0.1) 0.1) 0.4) 0.6) 0.1) 1.0 0.1) 0.2 7.4 7.1 7.6 7.8 6.5 6.7 7.1 8.5 6.2 7.8 12.3 10.8 Mar. 2003 ± ± ± ± ± ± ± SD, for n = 3-7 n = samples)(where values individual 3). for n < SD, or ( ( ( ( ( ( ( ±

) ) Maynes LagoonMaynes

------0.4) 0.4) 1.0) 1.5) 1.1) 1.7) 1.6 1.0 6.6 6.5 6.5 8.1 6.6 Oct. Oct. 10.7 13.3 14.6 13.4 2002 ± ± ± ± ± ± ± ( ( ( ( ( ( (

)

- - - - - 0.1) 0.1) 0.2) 0.2) 0.5 0.1) 0.8) 7.4 7.6 7.0 6.8 8.1 8.0 9.6 5.7 5.9 9.0 8.9 Jul. 11.1 12.5 11.2 2003 ± ± ± ± ± ± ( ( ( ( ( (

) ) )

------0.4) 0.4) 0.1) 0.5) 0.9 1.5) 0.3 0.7) 0.3) 0.4) 1.0 5.9 6.4 5.4 5.9 6.3 7.7 7.9 10.6 11.9 11.3 Mar. 2003 ± ± ± ± ± ± ± ± ± ± ( ( ( ( ( ( ( ( ( (

) )

------0.3) 0.3) 0.5) 0.5) 1.1) 2.9 0.5) 0.2) 1.2 Punbougal Lagoon Punbougal 6.4 6.5 5.5 8.1 8.9 8.8 Oct. Oct. 14.4 11.1 2002 ± ± ± ± ± ± ± ± ( ( ( ( ( ( ( (

)

------3.6) 3.6) 0.5) 0.1) 0.1) 1.8 1.3) 4.4 5.3 4.5 6.4 7.7 9.6 9.7 Jul. 10.4 2003 ± ± ± ± ± ± ( ( ( ( ( (

) )

------0.4) 0.4) 0.4) 0.3) 1.5) 0.4 1.0) 0.4 6.4 2.6 4.4 5.5 8.5 9.6 8.7 14.8 12.4 14.8 Mar. 2003 ± ± ± ± ± ± ± ( ( ( ( ( ( (

) ) Macintyre RiverMacintyre ------1.0) 1.0) 0.6) 0.7) 0.1) 1.5 1.1 7.5 4.5 4.9 4.7 6.3 9.6 9.7 Oct. Oct. 12.7 13.3 2002 ± ± ± ± ± ± ( ( ( ( ( (

- - - - - 1.1) 1.1) 0.1) 0.1) 0.4) 0.3) 0.6) 0.5) 1.1) 6.2 5.1 6.3 Jul. 11.0 14.5 11.9 13.1 15.2 12.6 13.0 13.8 11.4 11.9 2003 ± ± ± ± ± ± ± ± ( ( ( ( ( ( ( (

) )

- - - 2.4) 2.4) 0.6) 0.4) 1.4) 0.1) 0.2 0.8) 0.1) 0.2) 0.8) 1.1 8.5 3.7 2.8 4.8 7.9 8.6 12.8 11.3 12.6 15.2 16.1 16.5 12.8 12.3 Mar. 2003 ± ± ± ± ± ± ± ± ± ± ± ( ( ( ( ( ( ( ( ( ( (

) )

- - - - - Rainbow Lagoon 1.7 0.7 0.4) 0.4) 0.1) 0.8) 0.2) 3.2) 0.9) 0.6) 0.7) 0.4) 3.8 3.8 5.9 Oct. Oct. 16.2 11.3 11.9 12.1 16.5 15.4 16.8 13.0 ± ± 2002 ± ± ± ± ± ± ± ± ± ( ( ( ( ( ( ( ( ( ( (

) ) )

- - 1.9) 1.9) 0.2) 0.8) 0.5) 5.3 0.1) 0.7 0.5) 0.2) 0.3) 0.1) 0.2 8.0 9.5 6.5 5.6 7.0 7.8 9.8 Jul. 12.7 10.8 12.0 11.4 16.0 15.5 15.8 14.9 2003 ± ± ± ± ± ± ± ± ± ± ± ± ( ( ( ( ( ( ( ( ( ( ( (

) ) )

- - - - 1.5) 1.5) 0.3) 0.4) 4.3) 1.6 0.1) 0.6 1.8) 0.6) 0.2) 1.3 3.5 9.0 7.8 7.0 9.4 14.0 12.3 11.7 12.0 10.9 12.4 14.2 12.4 Mar. 2003 ± ± ± ± ± ± ± ± ± ± ± ( ( ( ( ( ( ( ( ( ( ( Lagoon A ) )

------South Callandoon Callandoon South 0.5) 0.5) 0.0) 0.8) 0.2) 0.7 0.8) 0.9) 1.4 7.6 4.2 6.5 7.8 3.8 9.8 Oct. Oct. 22.3 13.6 14.5 13.2 2002 ± ± ± ± ± ± ± ± ( ( ( ( ( ( ( (

(small) (large) Sample type A. agassizii L. unicolor N. erebi N. erebi Stable(‰)consumersRiver,isotope and Macintyre ratios one lagoons site from of floodplain on ea on the nitrogen sources and rimary sources rimary consumersrimary ish econdary consumers Floating aquatic Floating plants aquatic Submerged plants Chironomidae (larvae) P Algae FPOM CPOM vegetation Riparian P Zooplankton Corixidae S Notonectidae Odonata Shrimp F

(Octobermeancorrespond (March to ( 2003) values (July 2003). 2002), winter summer and post-flood Data post-flood Table 7.2 Chapter 7. Energy sources for fish using stable isotope analysis 184

Primary sources, except for floating vegetation collected in March 2003, were relatively 15N depleted in comparison with consumers (Table 7.2 and Figure 7.1). Spatial and temporal variation in algal δ15N values was relatively low, with algae from most sites being relatively 15N depleted (3.5 to 11.0‰). Interestingly, Serpentine Lagoon had highly enriched algal 15N, with values of 14.6 and 19.8‰. Submerged aquatic plants were relatively 15N depleted (5.9 to 9.6‰) in comparison with the floating species (6.1 to 14.0‰) (Table 7.2). The δ13C values of organic matter and riparian tree leaves were consistent across sites and sampling occasions, highlighting similar riparian vegetation characteristics of the study sites. Riparian tree leaves were 13C-depleated (−27.4 to −29.7‰) in comparison with coarse (−23.8 to −28.8‰) and fine (−23.3 to −27.6‰) organic matter, suggesting a potential contribution of algal carbon to the latter (Table 7.1). Variation in δ15N of organic matter and riparian tree leaves across sites and sampling occasions was relatively low, with values ranging from 2.6 to 9.5‰ (Table 7.2 and Figure 7.1).

Zooplankton samples were not available for all sites and sampling occasions. Nevertheless, it is clear that zooplankton δ13C values were both temporally and spatially variable, even though they were consistently more 13C depleted than the primary sources and most primary and secondary consumers (Table 7.1 and Figure 7.1). Lagoons subject to flooding during the study period (South Callandoon A and Rainbow) presented generally lower δ13C zooplankton values (−31.7 to −37.3‰), whereas zooplankton samples from non-flooded lagoons were consistently more 13C enriched (−19.1 to −29.3‰) (Table 7.1). Similarly, δ15N values for zooplankton were relatively higher than most of the primary sources and consumers (Table 7.2). Although more variable than δ13C values, zooplankton were slightly more 15N enriched in flooded lagoons (7.8 to 22.3‰) than in non-flooded ones (7.8 to 21.9‰) (see Table 7.2).

7.3.2 Consumers

Values of δ13C and δ15N for primary and secondary consumers, i.e. aquatic invertebrates, presented considerable spatial and temporal variation (Tables 7.1 and 7.2). In general, δ13C and δ15N values for aquatic invertebrates (excluding zooplankton), were intermediate Chapter 7. Energy sources for fish using stable isotope analysis 185

within the bounds of major sources and higher consumers, namely fish (Figure 7.1). Larvae of chironomids were available only for three sites and were considerably 13C depleted (−26.7 to −30.0‰) in South Callandoon A and Rainbow lagoons (subject to flooding) than in Maynes Lagoon (non-flooded) (−16.6 and −17.3‰). Similarly, δ15N values of chironomids were higher for the flooded study sites (7.9 to 12.7‰) than for the non-flooded ones (6.5 and 6.7‰).

Flooded sites Non-flooded sites

19.0 ZOOPL LEIOP NEMA-S Before flooding AMBAS SHRIMP LEIOP (summer) NEMA-S 14.0 AMBAS ZOOPL NEMA-L SHRIMP NOTON NOTON NEMA-L CORIX ODON ALGAE

N (‰) N (‰) 9.0 CORIX

15 ALGAE δ FPOM SUB.PLANT RIP.VEG RIP.VEG CPOM 4.0 FPOM CPOM October 2002 -1.0 19.0 After flooding LEIOP AMBAS AMBAS LEIOP (summer)

14.0 ZOOPL FLOAT.PLANT NEMA-S SHRIMP SHRIMP NEMA-S ODON NOTON FLOAT.PLANT NEMA-L NEMA-L

N (‰) N (‰) CHIRON NOTON 9.0 CORIX ZOOPL 15

δ OD ON CPOM CORIX ALGAE RIP.VEG CHIRON RIP.VEG FPOM SUB.PLANT 4.0 CPOM FPOM

March 2003 -1.0 19.0 AMBAS After flooding SHRIMP LEIOP (winter) ZOOPL OD ON 14.0 NOTON NEMA - S ALGAE NOTON NEMA - L ODON NEMA-L ZOOPL CHIRON CORIX NEMA-S 9.0 N (‰) FLOAT.PLANT CORIX FLOAT.PLANT SUB.PLANT 15 ALGAE FPOM δ FPOM RIP.VEG CPOM RIP.VEG 4.0 CPOM

July 2003 - 1.0 - 35.0 - 30.0 - 25.0 - 20.0 - 15.0 - 10.0 -35.0- 5.0 -30.0 -25.0 -20.0 -15.0 -10.0 -5.0 δ13C (‰) δ13C (‰)

13 15 Figure 7.1 δ C and δ N values of primary sources (), primary consumers (S) and fish () for each sampling occasion (from top to bottom), averaged for sites subject to flooding (left) and non-flooded sites (right). Sampling occasions correspond to: before flooding (summer October 2002) and after flooding (summer March 2003) and July 2003 (winter). Major energy sources, consumers and fish species are highlighted. Chapter 7. Energy sources for fish using stable isotope analysis 186

The remaining aquatic insects (Corixidae, Notonectidae and Odonata) also presented generally more depleted δ13C values from study sites subject to flooding compared to non- flooded sites (Table 7.1). Corixids were generally 5.5 to 6.5‰ more 13C depleted in flooded sites (−22.3 to −31.2‰) than in non-flooded ones (−16.8 to −24.7‰). Notonectids were 3.3 to 9.6‰ 13C depleted in flooded sites in comparison with non-flooded ones, whereas Odonata were at least 6.2‰ 13C depleted in flooded sites compared to non-flooded ones. δ15N values for corixids and notonectids were less variable across sites and sampling occasions, being only slightly 15N enriched in flooded sites, on average 0.3 and 0.8‰, respectively. On the other hand, Odonata were consistently more 15N enriched in flooded sites, 11.4 and 12.4‰, against 6.2 to 9.8‰ in non-flooded sites (Table 7.2). Note that Serpentine Lagoon had considerably 15N enriched Odonata with δ15N values of 15.4 and 19.5‰.

Shrimp (Macrobrachium spp.) were available only from Rainbow and Serpentine lagoons. These consumers had relatively more depleted δ13C values from Rainbow (−25.8 and −27.3‰) than from Serpentine Lagoon, even though the latter showed some temporal variation in δ13C values (−16.7 to −22.2‰). Although variable between these two lagoons, δ15N values for shrimp were less temporally variable as shown in Table 7.2, with values of 12.1 and 12.6‰ from Rainbow and 12.1 to 16.5 from Serpentine Lagoon.

7.3.3 Fish

In general, the three study species of fish had relatively 13C depleted mean carbon isotope signatures, particularly so in flooded sites, whereas δ15N values were enriched at all sites (Figure 7.1), consistent with an ultimate dietary source of zooplankton. Interestingly, the study species of fish were relatively less 13C depleted in the non-flooded sites. Trophic positions of the study species of fish consumers were temporally similar (Figure 7.1). With the exception of Odonata, the more predaceous taxa of invertebrates and fish were more enriched in 15N. The most 15N enriched species of fish were the omnivorous and microcarnivorous, L. unicolor and A. agassizii respectively, whereas N. erebi was Chapter 7. Energy sources for fish using stable isotope analysis 187 relatively depleted in 15N, though smaller individuals were relatively less 15N depleted than larger ones, thus indicating the trophic level of the study species.

Despite their contrasting diets (Chapter 6), the study species presented relatively similar δ13C and δ15N values, even though A. agassizii presented slightly more 13C depleted signatures than L. unicolor and N. erebi (Table 7.1). On the other hand, N. erebi were consistently more 15N depleted than the other two species of fish (Table 7.2). Values of δ13C for fish presented considerable spatial variation, whereas δ15N were less variable across sites (Tables 7.1 and 7.2). Ambassis agassizii showed more 13C depleted values at the sites subject to flooding during the study period (−28.2 to −34.3‰) whereas, at the non-flooded sites, δ13C values were slightly enriched (−19.6 and 23.8‰). In a similar fashion, L. unicolor was more 13C depleted in flooded sites (−24.8 to −32.5‰) than in non- flooded ones (−21.3 and −22.3‰). δ15N values for fish were relatively consistent through sampling occasions (Table 7.2), even though flooded sites were on average up to 3.4‰ 15N depleted compared with non-flooded ones.

7.3.4 Spatio-temporal relationships between δ13C values for fish consumers and primary sources

The three study species of fish, particularly A. agassizii and L. unicolor, showed significantly enriched values of 15N and depleted 13C values in comparison with most of the sources and consumers. In general, A. agassizii, L. unicolor and N. erebi showed a negative relationship between δ13C and δ15N values (Table 7.3). That is, for individual sampling occasions, more 13C depleted individuals were also 15N enriched. Therefore, the three study species (though to a lesser extent, larger individuals of N. erebi) had stable isotope signatures more similar to zooplankton, than to algae or organic matter and riparian vegetation. On average, smaller individuals of N. erebi (<7.0 mm SL) were slightly more 13C depleted and 15N enriched than the larger ones (>7.0 mm SL). However, regression analysis showed no correlation between standard length of N. erebi and δ13C or δ15N values (r2 =0.0038 and 0.0662, respectively).

Chapter 7. Energy sources for fish using stable isotope analysis 188

Table 7.3 Pearson’s coefficient of correlation (r2) for relationships between δ13C and δ15N values (‰) for each of the three species of fish and sampling occasions. Only relationships based on n ≥ 3 are shown.

Ambassis agassizii Leiopotherapon unicolor Nematalosa erebi

Sampling October March July October March July October March July occasion 2002 2003 2003 2002 2003 2003 2002 2003 2003 Rainbow 0.9120 0.4793 0.0017 0.1172 - - 0.7869 0.0638 0.0351

South 0.5748 - 0.0089 - - 0.3093 0.9314 0.7445 0.6356 Callandoon Macintyre ------0.4862 0.6448 0.5761 River Maynes ------0.7391 0.9409 -

Punbougal ------0.0052 0.0046 0.7462

Serpentine - - - 0.8829 0.9169 - - 0.4871 0.5739

Broomfield ------0.5702 - -

It is clear from the data presented, that algae are generally too 13C enriched and 15N depleted in relation to fish isotopic signatures, indicating that this autochthonous source contributed little, if anything, to the fish consumers. Furthermore, 15N values for riparian vegetation and organic matter were too 15N depleted (approximately 3.1 to 10.0‰ lower than any of the invertebrate consumers) for these allochthonous sources to be supporting the food web (see also Figure 7.3).

Therefore, if zooplankton, or to a lesser extent algae or organic matter, were an important source of organic carbon for fish consumers, it would be expected that the variability in isotopic signatures of zooplankton would track the variability in signatures of fish in space and time. That is, the isotopic signatures of each of the three species of fish should be positively correlated with the isotopic signatures of their food source across sampling occasions. Figure 7.2 shows that this was clearly the case for zooplankton and all of the study species, where zooplankton explained 98% of the variation in δ13C across study sites for A. agassizii, 85% for L. unicolor and 81% for N. erebi. Organic matter (CPOM) had relatively consistent δ13C values across sites and explained only 13% of the variation in carbon isotopic signatures for N. erebi, whereas for A. agassizii and L. unicolor less than 6% of δ13C variation across sampling occasions was explained by this primary source. Algal δ13C was also weakly correlated with δ13C of fish consumers, explaining only 6% of the variation in δ13C for N. erebi and less than 1% of A. agassizii and L. unicolor variation in δ13C values (Figure 7.2). Chapter 7. Energy sources for fish using stable isotope analysis 189

0 Zooplankton -5 N. erebi y = 0.9557x + 3.3136 R2 = 0.8076 -10 A. agassizii y = 0.7836x - 3.9719 -15 R2 = 0.9777

-20 L. unicolor y = 0.5785x - 8.774 C (‰) (‰) C R2 = 0.855 13 -25 δ Consumers Consumers

-30

-35

-40 0-40-35-30-25-20-15-10-50 Algae -5

-10

-15 N. erebi y = 0.1288x - 23.159 2 -20 R = 0.061 C (‰) 13 δ -25 Consumers Consumers

-30 L. unicolor y = 0.0209x - 26.338 A. agassizii R2 = 0.0014 -35 y = 0.033x - 28.348 R2 = 0.0018

-40 0 -35 -30 -25 -20 -15 -10 -5 0 CPOM -5

-10 L. unicolor y = 1.1418x + 5.6994 2 -15 R = 0.0572 N. erebi y = 1.0378x + 2.9111 R2 = 0.1324 -20 A. agassizii C (‰) (‰) C y = 1.0293x + 0.6453 13 -25 2

δ R = 0.0215 Consumers Consumers

-30

-35

-40 -40-35-30-25-20-15-10-50 δ13C (‰) Sources Figure 7.2 δ13C values of A. agassizii (S), L. unicolor () and N. erebi () versus δ13C of primary sources (zooplankton, algae and organic matter) across study sites and sampling occasions where all three primary sources were available. Chapter 7. Energy sources for fish using stable isotope analysis 190

7.3.5 Contribution of autochthonous versus allochthonous sources of carbon to consumer biomass

Isotopic signatures (δ13C and δ15N) of fish consumers and predicted sources were plotted for sampling occasions where all three predicted sources (zooplankton, CPOM and algae) where available (Figure 7.3). Data presented in Figure 7.3 corroborate the previous finding that zooplankton is a major contributor to fish consumers’ carbon, and clearly indicate that algae are unlikely to be making a significant contribution to the consumer food web in Macintyre River floodplain sites. Organic matter (CPOM) explained little of the variability in δ13C through sites and sampling occasions, as demonstrated in Figure 7.2. However, even though organic matter was considerably 15N depleted (approximately 3.1 to 10.0‰ lower than any of the invertebrate consumers), δ13C isotopic signatures of these allochthonous sources indicate that there may be an important contribution of organic matter to higher consumers in some of the study sites (Figure 7.3).

Estimates derived from mixing models for study sites where zooplankton was present also suggest an important contribution of zooplankton carbon to the biomass of fish consumers (Table 7.4). Data from Table 7.4 also show that contributions of organic matter and algae were variable across sampling occasions. Fish consumers derived, on average, 64.7% of their biomass from zooplankton and another 19.5% and 19.7% from algae and organic matter, respectively. Ambassis agassizii derived on average 74.3% of its biomass carbon from zooplankton and only 8.3 to 40.4% from algae or organic matter carbon (Table 7.4). Similarly, L. unicolor derived most of its carbon from zooplankton (61.8% on average), although contributions from algae were comparatively higher than for the other species (from 25.5 to 38.9%). Nematalosa erebi also derived most of its carbon from zooplankton, although it was the species least reliant on this ultimate primary source. Even so, an average of 56.2% of N. erebi biomass was derived from zooplankton. Despite being relatively variable across sampling occasions, organic matter contributed another 27.4%, on average, to the biomass of this species (Table 7.4).

Chapter 7. Energy sources for fish using stable isotope analysis 191 ALGAE NEMA NEMA inter). Note that that Note inter). C (‰) (‰) C NEMA 13 CPOM CPOM ZOOPL δ ZOOPL CPOM ZOOPL -35.0 -30.0 -25.0 -20.0 -15.0 -10.0 -5.0 0.0 -35.0 -30.0 -25.0 -20.0 -15.0 -10.0 -5.0 0.0 -35.0 -30.0 -25.0 -20.0 -15.0 -10.0 -5.0 0.0 ALGAE NEMA NEMA ZOOPL AMBAS NEMA LEIOP C (‰) (‰) C 13 LEIOP ZOOPL δ CPOM CPOM ALGAE CPOM Serpentine Lg. Lg. Serpentine ZOOPL AMBAS -35.0 -30.0 -25.0 -20.0 -15.0 -10.0 -5.0 -35.0 -30.0 -25.0 -20.0 -15.0 -10.0 -5.0 -35.0 -30.0 -25.0 -20.0 -15.0 -10.0 -5.0 ALGAE ALGAE ALGAE C (‰) (‰) C 13 δ CPOM CPOM CPOM AMBAS NEMA NEMA LEIOP ZOOPL South Callandoon A Lg. Maynes Lg. NEMA ZOOPL -35.0 -30.0 -25.0 -20.0 -15.0 -10.0 -5.0 -35.0 -30.0 -25.0 -20.0 -15.0 -10.0 -5.0 -35.0 -30.0 -25.0 -20.0 -15.0 -10.0 -5.0 ZOOPL AMBAS N values of primary sources (CPOM, algae and zooplankton) and fish for the sampling zooplanktonzooplankton) algae the of sources (CPOM, values and occasions fish for primary where andN available. was ALGAE 15 δ C (‰) (‰) C LEIOP ALGAE 13 LEIOP CPOM CPOM and C Rainbow Lg. Lg. Rainbow CPOM AMBAS NEMA δ AMBAS 13 AMBAS δ NEMA

LEIOP NEMA ZOOPL ZOOPL ZOOPL October 2002 March 2003 July 2003 -40.0 -35.0 -30.0 -25.0 -20.0 -15.0 -10.0 -5.0 -40.0 -35.0 -30.0 -25.0 -20.0 -15.0 -10.0 -5.0 -40.0 -35.0 -30.0 -25.0 -20.0 -15.0 -10.0 -5.0 Figure 7.3 onlythe and Southflooded Rainbow lagoons study Callandoon were during period Sampling2003 (w(summerJuly flooding correspond and before October to: after (summer 2003) flooding March 2002) occasions and 5.0 0.0 5.0 0.0 5.0 0.0 25.0 20.0 15.0 10.0 25.0 20.0 15.0 10.0 25.0 20.0 15.0 10.0 N N N 15 15 15 (‰) (‰) (‰) (‰) δ δ δ

Figure 7.3 δ13C and δ15N values of primary sources (CPOM , algae and zooplankton) and fish for the sampling occasions where zooplankton was available. Sampling occasions correspond to: before flooding (summer October 2002) and after flooding (summer March 2003) and July 2003 (winter). Note that only Rainbow and South Callandoon lagoons were flooded during the study period. Chapter 7. Energy sources for fish using stable isotope analysis 192

A. agassizii L. unicolor N. erebi Average Algae CPOM Zoopl. Algae CPOM Zoopl. Algae CPOM Zoopl. Algae CPOM Zoopl. October 2002 (prior to flooding) Rainbow 10.1 9.0 80.9 25.5 4.0 70.5 10.2 19.7 70.2 15.2 10.9 73.9 South Callandoon 11.3 40.4 48.3 - - - 6.9 53.1 40.0 9.1 46.7 44.1 Serpentine 21.1 1.4 77.5 38.9 3.1 58.1 40.6 4.2 55.3 33.5 2.9 63.6 Maynes ------15.6 0.0 87.9 - - - March 2003 (after flooding, summer) Rainbow 0.0 0.0 100.0 * 0.0 0.0 100.0 * 71.5 * 0.0 35.6 * - - - South Callandoon ------0.0 0.0 100.0 * - - - Serpentine - 8.3 91.7 - 26.3 73.7 - 0.0 100.0 * - 17.3 82.7 Maynes ------0.0 0.0 100.0 * - - - July 2003 (after flooding, winter) Rainbow 16.2 10.5 73.2 36.3 18.9 44.8 8.2 32.5 59.3 20.2 20.7 59.1 South Callandoon 0.0 0.0 100.0 * 0.0 0.0 100.0 * 0.0 0.0 100.0 * - - - Serpentine ------100.0 * 0.0 0.0 - - - Maynes ------0.0 100.0 * 0.0 - - - Average 14.7 13.9 74.3 33.5 13.1 61.8 16.5 27.4 56.2 19.5 19.7 64.7

7.4 Discussion

In general, stable isotope signatures of carbon for consumers in floodplain lagoons of the Macintyre River are similar to those reported for other Australian rivers, though considerably enriched in nitrogen, whereas primary sources, are generally similar in carbon and nitrogen values to those reported elsewhere in Australia (Bunn and Boon 1993, Boon and Bunn 1994, Bunn and Davies 1999, Bunn et al. 2003). Even though some variability in δ13C and δ15N of consumers and sources has been detected across individual sites, the results of the present study indicate that the major energy sources for consumers were generally similar between sampling occasions for flooded and non-flooded lagoons. Seasonal variations in isotopic signatures of carbon and nitrogen have been previously reported for floodplain billabongs (Bunn and Boon 1993) and seem to be related to inundation periods when terrestrial biomass contributes extensively to aquatic food webs (Fisher et al. 2001). Spatial differences were also observed in the present study and were mostly related to δ13C and δ15N values of zooplankton and algae, and overall isotopic signatures between flooded and non-flooded sites. Such variations are likely to be the result of differences in sources of carbon and nitrogen to primary producers across the Chapter 7. Energy sources for fish using stable isotope analysis 193 study sites. On the other hand, relatively unchanged isotopic signatures of organic matter reflect the similarities in riparian vegetation between study sites (see Chapter 3).

Of the major primary sources likely to be supporting fish consumers, algae were clearly too 13C enriched to be contributing significantly to fish biomass. Even though N. erebi was generally 15N depleted in comparison with the other two study species, in general, fish had relatively similar δ13C and δ15N isotopic signatures, indicating similar sources of organic carbon, despite their general differences in feeding habits (Chapter 6). The nitrogen isotopic signatures of organic matter and zooplankton, the two assumed major sources supporting fish consumers, were markedly different. The fact that fish had strong negative relationships between δ13C and δ15N is also consistent with a 15N enriched zooplankton endpoint. Even though the estimates derived from mixing models generally support this view, fish presented generally intermediate δ13C and δ15N values between the more 15N enriched zooplankton and the consistently 15N depleted organic matter (see next section).

7.4.1 Importance of allochthonous versus autochthonous sources of carbon to fish consumers

It has been suggested that allochthonous resources and detritus are important energy sources in floodplain rivers during periods of inundation (Fisher et al. 2001). Nevertheless, no major changes in isotopic signatures of consumers were observed between sampling occasions in flooded Macintyre sites. Interestingly, Delong et al. (2001) found that trophic structure did not change in response to flooding in the Mississippi River, suggesting that consumers continued to rely on sources of organic matter that were being used in the absence of flooding.

Fisher et al. (2001) suggested that allochthonous inputs of coarse organic matter play an important role in shaping the dynamics of primary consumers and influencing the overall community. However, in the present study, although organic matter in the form of detritus material was a substantial portion of food consumed by one of the study species of fish (Chapter 6), isotopically, it appeared to contribute little to the food web. δ15N values of organic matter seemed too depleted in relation to most consumers for this allochthonous source to be supporting detritivorous species, such as N. erebi, or some invertebrate Chapter 7. Energy sources for fish using stable isotope analysis 194 consumers. However, based on δ13C values, organic matter cannot be eliminated as a potential energy source to the food web. In the Cooper Creek (central Queensland), although algae was found to be the major energy source supporting the food web, terrestrial organic carbon (CPOM) was reported to be a major contributor to isotopic signatures of some consumers, such as shrimp (Macrobrachium spp.) and catfish (Neosilurus spp.), whereas isotopic signatures of N. erebi reflected a more even contribution of organic matter, algae and zooplankton (Bunn et al. 2003).

Autochthonous algae have been recognised as being an important source of organic carbon sustaining food webs in both temperate and tropical rivers (e.g. Hamilton et al. 1992, Forsberg et al. 1993, Thorp and Delong 1994, Bunn et al. 2003). However, data presented in the present study indicate that algal carbon contributed little to overall biomass of invertebrate and fish consumers. This is further supported by results presented in Chapter 3, which shows that algae were not abundant in most of the study sites. Despite the fact that the mixing models showed algae as an important energy resource to fish on some of the sampling occasions, this was the result of relatively 13C depleted algal values rather than changes in consumer isotopic signatures. It is, therefore, likely that fish consumers relied on sources other than algae throughout the study period.

Overall results indicate that zooplankton predominated as a dietary energy source for fish consumers in the Macintyre River floodplain. Fisher et al. (2001) recognised zooplankton as an important energy transfer taxa to small body-sized fish and, in turn, to higher piscivorous fish consumers. It has also been shown by Fisher et al. (2001), based on stable isotopic data, that the major source supporting zooplankton communities in floodplain habitats was an autochthonous primary producer, namely phytoplankton. In the present study, given the high concentrations of suspended sediment in the study sites it was impossible to sample this important autochthonous primary source, therefore preventing the determination of phytoplankton as the ultimate carbon source for zooplankton. Nevertheless, it can be assumed that phytoplankton was the most likely carbon source. Despite the high turbidity in floodplain waterholes of Cooper Creek, Queensland, Bunn et al. (2003) suggested that marked diel variation in dissolved oxygen saturation in the open surface water indicates high rates of phytoplankton production. Though not measured, there were indications of phytoplankton production in at least some of the study sites, in Chapter 7. Energy sources for fish using stable isotope analysis 195 the present study, based on diel variation in dissolved oxygen saturation in surface water and field observations.

The seston also provides further evidence that phytoplankton is the major source of energy for zooplankton. Huryn et al. (2001) found significant temporal changes in isotopic signatures in several size classes of seston and argued that such changes were the result of changes in the ratio of terrestrial to aquatic sources of organic carbon related to seasonal changes in water flow. This could explain the variations in isotopic signatures of zooplankton in the present study. Assuming that zooplankton consumed fine and ultra-fine sestonic organic matter and phytoplankton, changes in the proportion of terrestrial and aquatic sources of organic carbon from the seston across lagoons and sampling occasions may have led to the observed variations in isotopic signatures of zooplankton recorded in the present study. Furthermore, the variability observed for zooplankton isotopic signatures across sampling occasions is an indication that bacterially processed organic carbon is unlikely to be a major source of carbon to zooplankton (see Hadwen 2003). Because organic matter was consistent across sites in the Macintyre floodplain, a significant contribution of allochthonous carbon sources, that is organic matter, would be more likely to generate relatively constant isotopic signatures of zooplankton across sampling occasions, which clearly was not the case (see Table 7.1) (Hadwen 2003).

7.4.1.1 Leiopotherapon unicolor

Aquatic invertebrates were an abundant food item in the diet of the L. unicolor and as a result, organic matter should be an important source for this species, most likely via corixids and chironomids, which are known to feed on detritus and particulate organic matter (Ingram et al. 1997) and were amongst the food items most frequently consumed by L. unicolor (Chapter 6). However, the mixing models indicate that zooplankton and, to a lesser extent, algae were important contributors to the biomass of L. unicolor. Bunn et al. (2003) reported isotopic signatures of L. unicolor in Cooper Creek to be consistent with an ultimate dietary source of benthic algae, whereas other species of fish showed isotopic values intermediate between zooplankton and benthic algal signatures. This species has been reported to feed mostly on terrestrial prey, aquatic dipterans, zooplankton and other aquatic fauna in Cooper Creek (Balcombe et al. in review). In the present study, algae Chapter 7. Energy sources for fish using stable isotope analysis 196 were clearly not important to any of the consumers, whereas zooplankton, besides being a substantial portion of L. unicolor diet (Chapter 6), seemed to be the most important primary source supporting the other study species, though to a lesser extent for N. erebi (see next).

7.4.1.2 Ambassis agassizii and Nematalosa erebi

Overall results showed that A. agassizii was highly dependent on zooplankton as an ultimate dietary source of carbon, whereas N. erebi seemed to rely on zooplankton and to a lesser extent, organic matter. Bunn et al. (2003) reported size-related shifts in δ13C and δ15N signatures of N. erebi, and argued that this reflected a decreased dependence in zooplankton and a greater dependence on benthic algae and terrestrial detritus as individuals grow. Despite the fact that no correlation was found between body size of N. erebi and δ13C or δ15N values in the present study, on average, smaller individuals of this species tended to be slightly more 15N enriched than larger individuals, probably reflecting decreased consumption of zooplankton with size. This is supported by dietary data based on stomach contents presented in Chapter 5, which shows higher consumption of zooplankton in smaller individuals.

Smaller individuals of N. erebi and A. agassizii were also planktivorous, the latter feeding exclusively on zooplankton. It is, therefore, likely that the 13C depleted and 15N enriched values for N. erebi relative to the expected source (organic matter), from dietary data presented in Chapter 6, resulted from phytoplankton which is expected to be slightly 15N depleted in relation to zooplankton (≈2‰) (see Hadwen and Bunn in press). Furthermore, the phytoplanktonic component of fine organic matter from seston is known to represent an important form of autochthonous carbon input into the food web (Thorp et al. 1998, Delong et al. 2001).

In other studies, it has been found that despite having stomach contents dominated by detritus, fish showed little contribution of organic carbon derived from tree leaves, macrophytes or periphyton. Instead, detritivorous species received a large fraction of their carbon from phytoplankton (Araujo-Lima et al. 1986, Forsberg et al. 1993). In the floodplain lagoons of the Macintyre River, seston-derived phytoplankton probably Chapter 7. Energy sources for fish using stable isotope analysis 197 provides the main energy base for the study species, most likely via zooplankton in the case of A. agassizii and L. unicolor, or either via zooplankton or sestonic detritus matter for N. erebi. The large detrital component in the stomachs of N. erebi is most likely to be material that has to be ingested in order to extract the nutritious phytoplankton from the seston. There is evidence supporting the idea that organic matter consumed by detritivorous fishes is composed of a mixture of phytoplankton and other detrital matter (see Forsberg et al. 1993), with the phytoplankton component being either physically separated and ingested (Bowen 1984) or simply selectively digested and assimilated. Thorp et al. (1998) suggested that the more labile nature of the autochthonous component of detritus material would make it a more suitable and readily digestible source of nutrition for consumers than the more refractory allochthonous organic matter. Thus, it is reasonable to conclude that, in the present study, phytoplankton is generally easier to assimilate than detritus, and therefore an important source of energy for the fish food chain, especially in the case of N. erebi.

7.4.2 Models of energy sources in floodplain rivers

7.4.2.1 The RCC and FPC

Though not conclusive, the findings in the present study do not support the predictions of the RCC and FPC, as most of the evidence indicates a strong dependence on autochthonous sources by consumers in floodplain lagoons of the Macintyre River, rather than allochthonous organic matter. Water depths and high turbidity were amongst the main reasons that the RCC and FPC discounted the importance of autochthonous primary production in large floodplain rivers (Thorp et al. 1998, but see Bunn et al. 2003). In the present study, high turbidity and consequently low abundances of algae are thought to be the main reasons that algal carbon contributed little to the food web in the Macintyre River. Nevertheless, another autochthonous primary source, i.e. phyplankton, is considered to be the major source of carbon for fish consumers.

The RCC and FPC differ greatly in the sources of energy identified as primarily supporting food webs in large rivers when compared to the RPM, which emphasizes the importance of autochthonous carbon as a primary energy source for consumers. As stated in the RCC, Chapter 7. Energy sources for fish using stable isotope analysis 198 large rivers receive quantities of particulate organic matter from upstream processing of dead leaves and woody debris, and the contribution of local riparian vegetation to the food web is insignificant (Vannote et al. 1980). In the present study, given the low magnitude or total absence of flooding, it was assumed that riparian vegetation along the shores of floodplain lagoons may have been an important source of organic matter for invertebrates and fish, as suggested by the relatively similar carbon isotopic signatures between organic matter and fish consumers. Many studies have highlighted that tree leaves contribute significantly to the total terrestrial input into streams (e.g. Campbell et al. 1992, Pusey and Arthington 2003). Furthermore, results presented in Chapter 3 show that, despite a lack of overhanging vegetation, patches of leaf litter and to a lesser extent woody debris, were present in all of the study sites.

However, organic matter may be an important source for the study species that feed mostly on aquatic invertebrates, that is L. unicolor, and larger N. erebi, which feed on organic detritus (Chapter 6). Zooplanktivorous species, such as A. agassizii and smaller individuals of N. erebi, did not feed on macroinvertebrates or detritus (in the case only for A. agassizii) and therefore are more likely to be depending on organic carbon from an autochthonous source, that is phytoplankton. Although known to be more abundant in large rivers than smaller streams, according to the RCC phytoplankton is thought to contribute little to food webs because of light limitations related to turbidity and depth (Vannote et al. 1980). The FPC also minimizes the role of phytoplankton as a primary source for consumers, given its argued low contribution to the total carbon production in the floodplain and ease of decomposition (Junk et al. 1989). Nevertheless, other studies have indicated that the contribution of autochthonous phytoplankton has been underestimated (Thorp and Delong 1994, Thorp et al. 1998, Bunn et al. 2003).

7.4.2.2 The RPM

Thorp and Delong (1994) proposed in the RPM that a combination of local autochthonous production (by phytoplankton, benthic algae, etc) and direct inputs from the riparian zone (by tree leaves and particulate organic matter) during periods not limited to flood pulses are the major sources of carbon assimilated by consumers in large rivers. Furthermore, the RPM also emphasizes that, for floodplain rivers, phytoplankton is a significant contributor Chapter 7. Energy sources for fish using stable isotope analysis 199 to secondary productivity. Data presented in the present study provide support to the above concept, assuming that the main energy source for zooplankton was phytoplankton. Even though the study species presented relatively similar δ13C isotopic values, there is some indication from δ15N and the mixing models that the study species (particularly N. erebi) were relying on a combination of zooplankton (and, therefore, ultimately phytoplankton) and organic matter of riparian origin as their main energy sources.

The RPM predicts that organisms will vary greatly in their dependence on energy sources from in-stream primary production or lateral/downstream transport of organic matter (Thorp and Delong 1994). Furthermore, Thorp et al. (1998) and Fisher et al. (2001) reported that stable isotopic signatures of different species of fish reflected their level of omnivory, with more predacious species being relatively more 15N enriched in relation to other species, reflecting the known selective excretion of lighter isotopes with increasing trophic level (Peterson and Fry 1987). However, the present study provides evidence indicating that different species of fish, displaying different feeding habits, relied on fairly similar sources of carbon, regardless of their trophic level.

7.4.2.3 Caveats

This stable isotope analysis provided important insight into the carbon flow through the food web of the Macintyre River floodplain. Despite the fact that a growing body of evidence emphasizes the importance of algal contribution to consumer food webs in floodplain rivers, data from the present study show that plankton and, less importantly, organic matter were the major sources of energy supporting fish in floodplain lagoons of the Macintyre River. Nevertheless, some additional factors could alter the outcomes and interpretation of the results presented. For instance, it is important to bear in mind that the study period was relatively dry with no major flooding of any of the study sites. Inundation of the floodplain has been suggested to redistribute, and provide increased availability, of terrestrial litter (Soares et al. 1986, Junk et al. 1989) and autochthonous organic matter (Wissmar et al. 1981, Araujo-Lima et al. 1986, Hamilton et al. 1992). Given the importance of flooding to floodplain ecosystems it is possible that the patterns observed are the result of an absence of flow. In the case of major flooding, it is possible that algal growth would have been increased, as was the case for aquatic macrophytes (see Chapter 7. Energy sources for fish using stable isotope analysis 200

Chapter 3) and, therefore, this source could have contributed more significantly to the food web. Additionally, it has been suggested that riverine food webs, including those found in floodplain habitats, are seasonally dynamic with massive and swift shifts between microbial- and photosynthetic-based energy sources, in response to flow variability (Winemiller 1996).

Furthermore, the increased macrophyte growth in flooded lagoons, after flooding, is an indication that this potential energy source may have been overlooked during the study period as macrophytes were not abundant in most of the study sites. Another potentially important carbon source, which was identified in the present study (i.e. phytoplankton), was not physically sampled and its importance is inferred from estimates based on relatively variable zooplankton isotopic signatures. The considerable spatio-temporal variability in zooplankton signatures can be explained based on the influence of seston, and possibly phytoplankton, on isotopic signatures of this primary consumer. On the whole, further work needs to be done on the contribution of phytoplankton to the food web, and on the contribution of algae and macrophytes to consumers during a wetter period, when moderate to major flooding of the Macintyre River inundates the floodplain.

7.5 Conclusions

Analysis of stable isotope signatures of major sources known (from stomach content analysis in Chapter 6) and/or expected to contribute to the food web and particularly fish, indicate that zooplankton is the most important source of organic carbon supporting the study species of fish in the floodplain of the Macintyre River. It is proposed that phytoplankton, and possibly organic matter, of an autochthonous origin in the seston are the most likely energy sources for the planktonic suspension feeders (zooplankton) and consequently higher consumers, namely fish.

Even though organic matter was less important as a food source for fish, despite being consumed in quantity by N. erebi, there is some indication, mostly from δ13C values, that organic matter is also important to the food web, most likely via aquatic invertebrates, especially corixids and chironomids, which are known to feed on detritus and particulate organic matter, and are important food items for one of the study species (L. unicolor). Chapter 7. Energy sources for fish using stable isotope analysis 201

Benthic and littoral algae, on the other hand, were very low in abundance in all study sites, probably due to high turbidity, and seemed to contribute little to the aquatic food web.

The results presented in this study do not support some of the models for the functioning of floodplain rivers (RCC and FPC), which argue that allochthonous organic matter either from upstream or from the floodplain are the most important sources of carbon supporting higher consumers. In contrast, the data presented are partially in agreement with the RPM, which suggests that local productivity, based on autochthonous phytoplankton and organic matter, fuels the food web in large rivers.

It is important to emphasize that, given the variability of the results obtained in the present study and the impossibility of estimating isotopic signatures for all possible sources, such as phytoplankton and to a lesser extent algae, a more thorough investigation of the factors associated with phytoplankton production and the movement of phytoplankton carbon through the food web, particularly for fish, is needed.

Chapter 8. General discussion 202

8 General discussion

8.1 Background

Floodplain river systems are a conspicuous feature of dry inland Australian landscapes (Thoms and Sheldon 2000). Water flows of rivers in such regions are less regular than those in more humid tropical and temperate regions (Walker et al. 1995). Nevertheless, dry region floods are no less significant as drivers of riverine processes, even though flow variability is also a dominant feature (Poff and Ward 1989, 1990). In such systems, flow variability underpins the rates of many ecosystem processes and the transport of organisms, nutrients, organic carbon, and other materials within rivers and on their floodplain (Baldwin and Mitchell 2000). Responses of fish (Gehrke et al. 1995, Puckridge et al. 2000) and their food resources (Boulton 1999, Sheldon et al. 2002) are, therefore, closely related to water flow.

Lotic systems are thought to be organised as a nested hierarchy, both temporally and spatially (Johnson et al. 1995, Poff 1997). In a nested hierarchy, large-scale processes, such as climate and geomorphology, define higher levels of organisation of the physical and biological aspects (e.g. river morphology, flow regime, species pool) of the river- floodplain ecosystem (Johnson et al. 1995). These higher levels of organisation influence physical and biological factors at a variety of lower spatial and temporal scales (Poff 1997), which may be particularly relevant to floodplain river ecosystems (Kennard 1995). Therefore, over relatively short temporal and spatial scales (i.e. over dry-wet seasons and across floodplain lagoons) several factors may be influencing variations in resource use and the trophic ecology of fish communities.

In the context of the present study, several small-scale physical and biological factors were identified as being potentially important in understanding the trophic ecology of fish in a dryland floodplain river and in evaluating (1) how different types of natural and modified floodplain lagoons are able to trophically support their fish communities and (2) the ultimate sources of energy supporting fish consumers in floodplain lagoons.

Chapter 8. General discussion 203

Intrinsic, morphologic and behavioural factors identified as being of particular importance in influencing the diet of fish were (1) diel variation in food intake and in diet composition and (2) ontogenetic variation in composition of food items consumed. External, spatial and temporal factors thought to be influencing dietary ecology of fish in floodplain lagoons were (1) the temporal variability of the Macintyre River flow regime and, therefore, frequency, magnitude, timing and duration of floodplain inundation, (2) the spatial arrangement of lagoons on the floodplain and, more importantly, the morphological characteristics of floodplain lagoons, i.e. size, habitat characteristics and water quality, and (3) the degree of water management (e.g. artificially elevated water levels) of the floodplain lagoons. The main factors expected to be affecting contributions of ultimate energy sources to fish through the food web were (1) the availability of primary sources, such as algae and organic matter, and the consumption by fish of primary consumers relying on these primary sources, and (2) seasonal changes in energy flow pathways from sources to consumers in relation to flooding. The factors presented above will henceforth be addressed in the context of the results presented in this thesis and the predictive models presented in sections 1.3.1 and 1.4.1 of the introductory chapter (Figures 1.1 and 1.2).

To test the hypotheses from sections 1.3.1 and 1.4.1, fish that feed on different trophic levels of the food web (e.g. detritus/algae, zooplankton or aquatic insects) were selected for study in the Macintyre River floodplain. Ambassis agassizii is as a microphagous carnivore, feeding mostly in mid-waters and only occasionally in benthic areas. During the study period this species consumed mostly microcrustaceans (Calanoida, Moinidae and Daphniidae) (Chapter 6). Leiopotherapon unicolor consumed a wider range of prey items, from a variety of aquatic insects and microcrustaceans to shrimps and fish. Therefore, this species can be classified as an opportunistic meiophagic omnivore that feeds from mid- water as well as benthic areas (Chapter 6). Nematalosa erebi is a microphagic omnivore, with smaller fish being generally omnivorous, feeding mostly on microcrustaceans and detritus, and larger fish being mainly detritivorous, but also consuming small amounts of vegetable matter and algae (Chapter 6).

As shown in Chapter 3, lagoons in the floodplain of the Macintyre River vary considerably in morphology, flow regime and, to some degree, physico-chemical characteristics of the water, and these are some of the major factors hypothesised to be affecting food resources used by fish in these floodplain habitats (Figures 1.1 and 1.2). Floodplain waterbodies are Chapter 8. General discussion 204 strongly influenced by periodic fluctuations in water level, as they function as both lentic systems during the dry season, when they are relatively isolated from the main river channel, and as lotic environments, during the wet season, when their interaction with the river is at its maximum (Rai and Hill 1984). During the study period the Macintyre River presented only two minor to moderate floods occurring between 2002 and 2003, which led to a prolonged dry period and, consequently, affected the hydrology of the study sites.

Many studies have shown that the flood cycle in floodplain lagoons is followed by fluctuations in physical, chemical and biological characteristics of these waterbodies, such as depth, turbidity, dissolved oxygen levels, nutrient and organic carbon concentrations and biological production (Rai and Hill 1984, Hamilton and Lewis 1987, Junk and Weber 1995). Therefore, as a result of the low magnitude of flooding events during the study period, variation in habitat characteristics in the study sites occurred mostly between flooded and non-flooded lagoons (Chapter 3), as the latter presented significantly lower proportions of habitat elements, such as aquatic and overhanging vegetation, algae, debris, leaf litter and root masses. On the other hand, the flooded lagoons showed greater indication of temporal changes, given that many of these habitat elements increased in proportion during the wet season, after flooding. The water quality parameters also varied mostly between flooded and non-flooded lagoons (Chapter 3), and even though patterns were not clear for the non-flooded sites, the flooded lagoons showed lower pH, conductivity and dissolved oxygen and higher water temperatures. Flooding also decreased turbidity in the lagoon sites, whereas flooding increased turbidity at the river site.

8.2 Major sources of variation in dietary ecology of fish

It has been recognised that a major drawback in using stomach content analysis in studies of fish diets and/or food web structure is the possibility of bias arising from too few individuals in one sample and/or nearly empty stomachs in fish samples. Although the problem posed by small samples is easier to overcome (Schafer et al. 2002), nearly empty stomachs may be an important issue as, in such cases, the stomach may contain only a few food items, often one or two, which can erroneously indicate that the diet of that fish was not diverse, or place undue emphasis on that single unusual item in the overall diet Chapter 8. General discussion 205 composition of the species being studied (Pusey et al. 1995). Therefore, it is important to know the time of day when fish individuals would be most likely to contain food in their stomachs to ensure that interpretation of diet patterns is not biased.

Chapters 4 and 5 were aimed at determining possible sources of variation in diet composition of fish in order to optimize comparisons of fish diets in relation to spatial and temporal patterns. Besides the spatio-temporal variability in dietary composition (discussed in the next section), the study species presented considerable variation in food intake throughout the day. Furthermore, it has been shown in Chapter 4 that these patterns can change from summer to winter. Diel variation in food intake is widely reported in the literature and thought to be related to changes in activity of prey (Keast and Welsh 1968) and to be under endogenous control, synchronized by light (Wootton 1990, Boujard and Leatherland 1992). It is argued in Chapter 4 that the diel patterns in food intake by the study species are mostly driven by a combination of changes in water temperature and light availability. In general, food consumption was significantly higher during the summer than during the winter. During the summer, feeding activity of the study species was concentrated around midday, and during the winter, food consumption was high both at midday hours and possibly extended until late in the afternoon. These feeding patterns indicated that sampling of A. agassizii, L. unicolor and N. erebi should be concentrated at midday hours during the summer and midday to late afternoon during the winter. These are the times of day when food consumption was greatest, as demonstrated by stomach fullness and relative stomach content volume. Given the resolution of time intervals chosen in this study, it is unknown whether any of the species feed continuously between the midday and late afternoon periods.

Another important source of variation influencing dietary composition of the study species was related to changes in the relative abundance of prey consumed and/or specific composition of the food items ingested, as fish grew. Differences in resource use by fish species between ontogenetic intervals are common, as almost all species of fish change trophically during ontogeny (Miller 1979b, Werner and Gilliam 1984). Such ontogenetic changes in diet related to increasing body size have been claimed to overwhelm the detection of any potential temporal and spatial variation associated with fluctuating prey abundance and availability (Pusey et al. 2000, Schafer et al. 2002).

Chapter 8. General discussion 206

Chapter 5 explored size-related changes in diet composition of the three study species and possible variations in ontogenetic patterns across sites and seasons. Size-related shifts in dietary ecology of the study species were clearly shown in Chapter 5. Nevertheless, the degree of ontogenetic changes varied between lagoons and seasons and across species. Whereas A. agassizii experienced size-related changes mostly related to the relative abundance of prey items consumed, smaller L. unicolor shifted from a narrow diet of microcrustaceans and chironomids, to a wider range of food items in larger individuals, with an increase in consumption of aquatic insects, shrimps and fish with age/size (Chapters 5 and 6). Similar to results from other studies (e.g. Kennard et al. 2001, Bunn et al. 2003), N. erebi displayed the strongest dietary shift, with a change from microcrustaceans in smaller individuals, to a largely detrital diet in larger fish in most of the study lagoons. Therefore, it was important for average sizes or size ranges of groups of individuals of the study species to be known and taken into account when explaining patterns in diet composition. This is particularly the case for N. erebi, which presented the strongest changes in diets with size.

8.3 Feeding ecology of fish in the floodplain of the Macintyre River: dietary data

In Australian freshwaters, aquatic insects, microcrustaceans, algae and terrestrial material have been recognised as the most important food resources for fish (Kennard et al. 2001, Pusey et al. 2004). Even though all these food items were consumed by A. agassizii, L. unicolor and N. erebi, the results presented in Chapters 6 and 7 clearly indicate that zooplankton is the most important food resource consumed by these species. As summarised in the conceptual food web presented in Figure 8.1, zooplankton is part of the diet of all three species of fish and contributes significantly as an energy source to the study species. The importance of zooplankton/phytoplankton as energy sources has been clearly demonstrated in the present study (see Section 8.4). Even though aquatic insects and detritus were consumed by some of the study species, stable isotopic signatures indicate that these food items were less important than zooplankton (Figure 8.1). Chapter 8. General discussion 207

Fish L. unicolor A. agsssizii N. erebi consumers

Invertebrate Aquatic insects Zooplankton consumers Shrimps

Energy sources (producers) Aquatic plants Detritus Algae Phytoplankton

Figure 8.1 Simple food web model formulated for the Macintyre River floodplain lagoons from empirical dietary data and stable isotope data collected between late 2001 and 2003, and from information available in the literature (see Chapters 5 and 6). Only strong links are illustrated, as per Table 6.2 (Chapter 6) and Figures 7.1 and 7.3 (Chapter 7). Bold solid lines represent interactions supported by stomach content analyses and stable isotopic signatures. Fine solid lines indicate interactions based on dietary data but lacking stable isotope support. Dashed lines indicate interactions inferred from data from the literature. Box size for energy sources (except for phytoplankton) represent their proportional abundance (as determined by their percent occupation of wetted perimeter, see Chapter 3), averaged for all lagoons and sampling occasions.

8.3.1 Effects of flooding and drought on fish diets: the temporal component

Aquatic food webs, particularly those in floodplain rivers, can be significantly affected by spatial and temporal variability in patterns of waterbody connectivity, succession and inundation (Winemiller 1996, Jepsen and Winemiller 2002). Fish assemblages in tropical river food webs are characterised by high taxonomic diversity, diverse foraging modes, omnivory and an abundance of detritivores. In the present study, the dietary composition of the study species of fish was narrow, based on relatively few food items within three major food categories (microcrustaceans, aquatic insects and detritus) (Figure 8.1).

Despite predictions that floodplain inundation would lead to higher abundance and diversity of prey available for fish (conceptual model depicted in Figure 1.1), observed dietary changes in A. agassizii and L. unicolor seemed to reflect changes in availability and Chapter 8. General discussion 208 abundance of specific food items associated with major seasonal patterns (summer/winter), rather than flooding. Flooding of some of the study sites was not reflected in clear patterns of dietary changes in comparison with sampling occasions prior to or long after floods. Given the low magnitude or total absence of flooding events and, consequently, flood influences on lagoons during the study period, it is likely that the observed dietary variation was largely a consequence of successional changes in composition of the aquatic fauna as the dry season progressed (see Bishop et al. 2001, Sheldon et al. 2003, Balcombe et al. in review). Therefore, the hypothesis that floodplain inundation leads to increased food availability for fish remains to be tested as flooding during the study period was mild (Chapter 3), and apparently had little influence on feeding patterns of the study species.

Nevertheless, the hypothesis that in the absence of flooding, fish confined to floodplain lagoons without connectivity to the main river would be dependent on scarce food resources seem to be supported by the data presented in Chapters 6 and 7. This is demonstrated by the overall low diversity in food items, within the major categories, consumed by the study species and the fact that food items were overwhelmingly of autochthonous origin (Figure 8.1). It is important to keep in mind, though, that spatial and temporal variations in diets of N. erebi were not as significant as those observed for A. agassizii and L. unicolor, as larger N. erebi and, to a lesser extent, smaller individuals fed mostly on detritus material, a relatively abundant and widespread resource, throughout most sites and sampling occasions.

8.3.2 Effects of flow regime on fish diets: the spatial component

The most important drivers of processes in floodplain rivers are the variable flow and the lateral connectivity of the river to its floodplain wetlands (Junk et al. 1989, Walker et al. 1995, Kingsford 2000, Amoros and Bornette 2002, Kingsford et al. 2004). In Australian rivers, flow regime and its variability have been recognised as major driving forces in ecological structure and functioning (Walker et al. 1995, Bunn and Arthington 2002, Capon 2004), and responses to flow variability of fish (Gehrke et al. 1995) and aquatic invertebrates (Sheldon et al. 2002) have been reasonably well described.

Chapter 8. General discussion 209

As indicated from temporal data (Chapter 6), flow regime had an important effect on differences in fish diet composition across lagoons as the three study species showed differences in diet composition among sites. Even though the major food categories were similar within species, the composition and relative abundance of prey items consumed changed substantially for each species across the study sites. Even though flooding has been hypothesised to be the major driving force affecting feeding ecology in floodplain lagoons of the Macintyre River (conceptual models in sections 1.3.1 and 1.4.1), flooding events were minor during the study period and, therefore, contributed little to the dietary patterns of the study species (Chapter 6). Nevertheless, it is assumed (based on field observations) that the major flooding of November 2001 reached all study sites (Chapter 3) and, therefore, reset floodplain lagoons as hypothesised in the conceptual model presented in Figure 1.2(a) (Section 1.4.1, Chapter 1). From then onward, resulting from the absence of major flooding and evaporative water loss, the study sites dried out continuously, except for the sites with permanently elevated water levels (Macintyre River and Rainbow Lagoon) (Figure 1.2).

It has been suggested that the connection and disconnection of floodplain waterbodies in dryland rivers can influence the biota in a sequential fashion associated with fluctuations in hydrology and that, after disconnection each waterbody will behave as a separate unit with assemblage composition diverging in a manner that reflects those species present at the time of disconnection (Sheldon et al. 2003, Marshall et al. in review). The effects of different hydrological history on fish diets were clear but results were limited due to the fact that all study species were not necessarily present in all study sites. Nevertheless, as predicted by the model in Section 1.4.1, fish diets were variable across sites with different flow histories. As hypothesised in Section 1.4.1, diets of fish from Rainbow Lagoon, which has permanently elevated water levels, were consistently less variable through time than diet of fish from lagoons with more variable water regimes. The lack of major flooding since November 2001, and the fact that water levels in Rainbow Lagoon were relatively unchanged throughout most of the study period, may have had a significant influence on the composition of prey items available in this site, as variations in diet of A. agassizii and L. unicolor from Rainbow Lagoon were less conspicuous than, for instance, at the South Callandoon lagoons. It appears that this lack of variation in food items consumed is likely to be a result of the relatively less variable water regime in Rainbow Chapter 8. General discussion 210

Lagoon and the low diversity of prey items available (see Table 6.2 in Chapter 6 and Chapter 7).

On the other hand, fish from more hydrologically variable lagoons (i.e. South Callandoon A) showed considerable variation in diet composition for summer and winter periods (Chapter 6). Because species like A. agassizii and L. unicolor feed mostly on resources that are more subject to seasonal and spatial changes (i.e. zooplankton and aquatic invertebrates) (see Marchant 1982, Bass et al. 1997, Bishop et al. 2001), seasonal variation in dietary composition was more evident, showing that the composition of available prey items associated with contrasting flow regime characteristics plays an important role in the diet of these species in the floodplain lagoons studied. The hypothesis that fish from semi- permanent lagoons should present slower successional changes in patterns of diet composition remains to be tested as only N. erebi was collected from these habitats.

The results from the present study indicate that a variety of factors contribute to the observed variation in diet composition of fish in floodplain lagoons. Flow variability has been recognised as playing a central role in the structure and functioning of river- floodplain systems (Junk et al. 1989, Walker et al. 1995). In the absence of floods and significant floodplain inundation during the study period, local attributes of lagoons became increasingly important factors in determining dietary composition of fish. Such attributes are the spatio-temporal variations in general physical conditions (morphology, habitat characteristics and, possibly, degree of water management) and biological characteristics (fish species and sizes being considered and differences in available food resources) of floodplain lagoons.

8.4 Importance of allochthonous and autochthonous energy sources to fish: stable isotope and stomach contents data

Major conceptual models predict that allochthonous organic matter either from upstream (RCC) or from the floodplain (FPC) are the most important sources of carbon supporting higher consumers in large river ecosystems (Vannote et al. 1980, Junk et al. 1989). However, the present study has found little evidence of allochthonous food types in the diet of A. agassizii, L. unicolor and, to a lesser extent, N. erebi. On the contrary, results of Chapter 8. General discussion 211 stable isotope and stomach content analysis show that these species were reliant on resources produced within the floodplain lagoons, particularly zooplankton, even after the occurrence of small floods (Chapter 6). In floodplain rivers, the flood-pulse is known to inundate a substantial area of terrestrial biomass that contributes extensively to increased inputs of terrestrial detritus and other allochthonous resources (Junk et al. 1989, Fisher et al. 2001). Because major flooding was not recorded during the study period, allochthonous materials may have been prevented from entering the floodplain lagoons, resulting in the relatively low contributions of allochthonous food items to the diet of these species.

However, detrital material of terrestrial origin made high contributions to the diet of N. erebi, suggesting the importance of riparian sources to this species in the study sites. Despite the importance of terrestrial organic matter as food for fish in some river systems, as emphasized by Pusey and Arthington (2003), detrital matter was less important as a food source for fish, despite being consumed by both small and large N. erebi. Nevertheless, there was some indication, mostly from carbon isotopic values (Chapter 7), that organic matter is also important to the food web, most likely via aquatic invertebrates, especially corixids and chironomids, which are known to feed on detritus and particulate organic matter, and are important food items for one of the study species (L. unicolor).

Results of the present study are partially in agreement with the RPM, which suggests that local productivity, based on autochthonous phytoplankton and organic matter, fuels the food web in large rivers (Thorp and Delong 1994). Analysis of stable isotope signatures indicates that zooplankton is the most important source of organic carbon supporting the study species of fish in floodplain lagoons of the Macintyre River (Chapter 7). This food item was consistently consumed by A. agassizii, L. unicolor and smaller individuals of N. erebi (Chapter 6). Furthermore, it is proposed in Chapter 7 that phytoplankton (an autochthonous energy source) in the seston is the most likely energy source for the planktonic suspension feeders (zooplankton) and consequently fish consumers, including N. erebi. Other autochthonous energy sources, such as benthic algae and aquatic plants, were generally low in abundance probably due to high turbidity and lack of flooding, and were regarded as making little contribution to the aquatic food web. Despite high turbidity in floodplain habitats in Cooper Creek, Bunn et al. (2003), found that benthic algae were the most important energy source for fish consumers. However, extensive algal growth Chapter 8. General discussion 212 was exclusive to the shallow littoral zone of larger floodplain waterholes (Bunn et al. 2003), which was not the case in the present study.

8.5 Management implications

The findings of the present study have important implications for the way that floodplain river ecosystems are managed. Several studies have recognised that factors such as flow regulation, nutrient dynamics, turbidity, herbicide inputs and stock trampling (McCosker 1996, Robertson 1997, Bunn et al. 2003) are likely to have important influences on the food web of large rivers because they affect the distribution, composition and production of important primary sources for consumers, such as microalgae (Bunn et al. 2003). Results presented in this thesis strengthen evidence of the importance of energy resources produced within floodplain lagoons (i.e. phyto- and zooplankton), which can affected by alterations in turbidity levels due to stock trampling and by herbicide inputs (Bowling and Jones 2003), and other management practices such as flow regulation, by maintaining permanently elevated water levels (Chapter 6).

The substitution of a variable flooding pattern with a stable one and the consequent loss of wet-dry cycles is thought to have a significant impact on floodplain waterbodies, as invertebrate communities may become less diverse or abundant than those in more variable floodplain habitats or may become dominated by species adapted to standing waters (Shiel 1990, Boulton and Lloyd 1991, Kingsford 2000). In the present study, permanently elevated water levels are thought to have had direct effects on fish diets by leading to a less variable dietary composition of species like A. agassizii and L. unicolor, which feed on resources that are more subject to seasonal changes, such as zooplankton and aquatic invertebrates (Marchant 1982, Bass et al. 1997, Bishop et al. 2001). In lagoons offering diminished trophic diversity for fish consumers, fish are potentially vulnerable to sudden or extended losses of important dietary components. For example, an acute herbicide spill could have a cascading effect on primary and secondary consumers (e.g. zooplankton), further diminishing the energy basis of the food web. This would be important where few other food types are available. Additionally, fish that are constrained to feed on a narrow range of diet items may experience loss of body condition, growth potential and even fecundity (see Welcomme 1979, Junk 1985). Chapter 8. General discussion 213

8.6 Conclusions

Factors such as diel and ontogenetic variations in fish dietary composition and food intake have been shown to considerably affect overall dietary patterns of the three fish species. It is, therefore, important to understand the contributions of such factors to the variability of dietary patterns before performing studies on resource use by fish using stomach contents analysis. The extent of the influence of diel and ontogenetic variability in fish diets depends on temporal patterns of food intake and spatial variability in prey items available.

The major food categories consumed by A. agassizii, L. unicolor and N. erebi on the floodplain of the Macintyre River were zooplankton, aquatic invertebrates (mostly insects - corixids, notonectids and chironomids) and detrital material. Zooplankton was of particular importance as this food item was ingested by all three study species at some stage of their life history. As a consequence, factors that influence the distribution and abundance of phytoplankton and, consequently, zooplankton, such as flow regulation, nutrients, light, herbicides, stock trampling, etc (see Robertson 1997, Bunn et al. 2003), are likely to have an important impact on the food web of the Macintyre River.

Spatial and temporal variation in diet composition of the study species was mostly associated with changes in prey items available across floodplain habitats and between major time periods (summer/winter). The low magnitude of flooding events during the study period is the most likely factor affecting the lack of variation in patterns of fish diets in flooded sites, whereas in non-flooded lagoons the observed dietary variation was a consequence of successional changes in composition of the aquatic fauna as the dry season progressed. Nematalosa erebi showed less dietary variability, when compared to that of A. agassizii and L. unicolor, as larger N. erebi and, to a lesser extent, smaller individuals fed mostly on detritus material, which is a relatively abundant and widespread resource in these floodplain lagoons. Flow regime had an important effect on differences in fish diets composition across lagoons, but further evaluation of the role of flooding is needed due to overall lack of major flooding events during the study period.

Chapter 8. General discussion 214

During the study period, autochthonous resources formed the basis of the aquatic food web. Phytoplankton in the seston is the most likely energy source for fish consumers via the pathway of planktonic suspension feeders (zooplankton). Nevertheless, organic matter could not be discarded as an important energy source for invertebrates and higher consumers, and further study is necessary to address this issue. Therefore, the present study does not provide support for the RCC and FPC, which argue that allochthonous organic matter either from upstream or from the floodplain are the most important sources of carbon supporting higher consumers. In contrast, the RPM would be a more appropriate model to describe the food web and energy sources for consumers in lagoons on the Macintyre River floodplain, as this model suggests that local productivity, based on autochthonous phytoplankton and organic matter, fuels food webs in large rivers.

These results of the present study show that factors known to affect microalgae production in floodplain lagoons, such as flow regulation, turbidity and nutrient/herbicide inputs must be seriously considered in current management practices, because microalgae (namely phytoplankton) was recognised as the major energy source for primary and secondary consumers in floodplain lagoons of the Macintyre River. Additionally, flow regulation of floodplain habitats by maintaining permanently elevated water levels is likely to affect the dietary composition of fish consumers that feed on resources subject to seasonal changes, such as zooplankton and aquatic invertebrates. Data from the present study showed that fish species in a regulated lagoon, where a variable flooding pattern has been altered to a more permanent one, presented much less temporal variability in their diet than species from floodplain habitats with a more variable water regime.

215

Appendices

Appendix 1. Photos of the study sites (Chapter 3)

Photo 1. South Callandoon Lagoon A soon after minor flooding by the Macintyre River (March 2003). This lagoon is characterised by abundant overhanging vegetation and the presence of floating and submerged aquatic vegetation usually associated with flooding.

Photo 2. South Callandoon Lagoon B soon after flooding by the Macintyre River (February 2002). This lagoon is relatively similar to the South Callandoon A, with abundant overhanging vegetation and aquatic vegetation associated with flooding. South Callandoon B dried out before August 2002. 216

Photo 3. Study site on the Macintyre River, adjacent to the South Callandoon lagoons (Summer 2003). Note the relatively steep banks and the presence of large woody debris in the margins. Maximum depth on this site was approximately 1.4 m.

Photo 4. Punbougal Lagoon during the summer of 2003. The low levels of water turbidity and high depths (up to 6 m) were major characteristics of this lagoon. Note the overall absence of overhanging vegetation and floating aquatic vegetation.

Photo 5. Maynes Lagoon was a very muddy and relatively sterile lagoon with few aquatic plants and fish catches dominated by bony bream (Nematalosa erebi). Note the gentle banks and the absence of overhanging vegetation. Cattle were often seen drinking on its shores.

217

Photo 6. Rainbow Lagoon is pumped into and out for irrigation. Photo taken late in the dry season (July of 2003). Main features in this lagoon include a relatively diverse fish fauna and the presence of several habitats, such as aquatic vegetation and woody debris.

Photo 7. Serpentine Lagoon, depicted here during the summer of 2003, is located adjacent to the town of Goondiwindi. Artificially filled late in 2000, this lagoon contains relatively few habitat elements, despite a relatively diverse fish fauna. Aquatic vegetation was relatively scarce in this lagoon.

Photo 8. Broomfield Lagoon dried out soon after the summer of 2002/03. This is a large and shallow lagoon with relatively few habitat elements and low amounts of aquatic vegetation. 218

Photo 9. Water regulation is a common feature in many floodplain lagoons in the Border Rivers system. This picture shows a typical irrigation channel and control gates which drain water from Rainbow Lagoon, in the background.

Photo 10. Cattle grazing is one of the main forms of land use in the Border Rivers basin. This picture shows cattle drinking from the shores of Rainbow Lagoon. Trampling usually contributes to bank and soil erosion.

Photo 11. Water diversion for irrigation is another major issue in the Border Rivers catchment. This picture shows the pipes used for pumping water from the Macintyre River, adjacent to the South Callandoon lagoons, to a nearby ring tank used for water storage.

219

Appendix 2. Two-way ANOVA tables for diel variation in fish diets (Chapter 4)

Results table of factorial ANOVA to determine the effects of different species and seasons on mean stomach fullness (a) and RCV (b) of fish. Note that in the presence of a significant interaction, the significance of tests for the main effects are unreliable, thus the necessity to split the analysis.

Source DF SS MS F value Pr > F (a) Stomach fullness Among cells 5 484.27 96.85 12.31 <.0001 A|B 2 175.82 87.91 11.18 <.0001 B|A 1 210.62 210.62 26.78 <.0001 AxB 2 88.68 44.34 5.64 0.0037 Within 635 4994.26 7.86 Total 640 5478.53 AxB effect sliced by season and species Summer 2 260.12 130.06 16.54 <.0001 Winter 2 9.99 5.00 0.64 0.5302 (b) RCV Among cells 5 259.81 51.96 82.72 <.0001 A|B 2 169.52 84.76 134.93 <.0001 B|A 1 57.83 57.83 92.06 <.0001 AxB 2 26.41 13.20 21.02 <.0001 Within 635 398.88 0.63 Total 640 658.68 AxB effect sliced by season and species Summer 2 156.98 78.49 124.96 <.0001 Winter 2 41.40 20.70 32.96 <.0001 Factor A: Species Factor B: Season

Results table of factorial ANOVA to determine the effects of different times of day and seasons on mean stomach fullness (a) and RCV (b) of A. agassizii. Note that in the presence of a significant interaction, the significance of tests for the main effects are unreliable, thus the necessity to split the analysis.

Source DF SS MS F value Pr > F (a) Stomach fullness Among cells 7 862.09 123.16 30.70 <.0001 A|B 3 716.51 238.84 59.54 <.0001 B|A 1 6.33 6.33 1.58 0.2105 AxB 3 144.05 48.02 11.97 <.0001 Within 208 834.41 4.01 Total 215 1696.50 AxB effect sliced by season and species Summer 3 310.73 103.58 25.82 <.0001 Winter 3 547.60 182.53 45.50 <.0001 (b) RCV Among cells 7 17.96 2.57 24.86 <.0001 A|B 3 14.71 4.90 47.50 <.0001 B|A 1 0.17 0.17 1.68 0.1957 AxB 3 3.27 1.09 10.57 <.0001 Within 208 21.46 0.10 Total 215 39.42 AxB effect sliced by season and species Summer 3 4.46 1.49 14.41 <.0001 Winter 3 13.37 4.46 43.18 <.0001 Factor A: Time Factor B: Season

220

Results table of factorial ANOVA to determine the effects of different species and seasons on mean stomach fullness (a) and RCV (b) of L. unicolor. Note that in the presence of a significant interaction, the significance of tests for the main effects are unreliable, thus the necessity to split the analysis.

Source DF SS MS F value Pr > F (a) Stomach fullness Among cells 7 721.78 103.11 23.37 <.0001 A|B 3 471.15 157.05 35.6 <.0001 B|A 1 59.51 59.51 13.49 0.0003 AxB 3 156.57 52.19 11.83 <.0001 Within 205 904.30 4.41 Total 212 1626.08 AxB effect sliced by season Summer 3 587.94 195.98 44.43 <.0001 Winter 3 70.22 23.41 5.31 0.0015 (b) RCV Among cells 7 112.82 16.12 35.01 <.0001 A|B 3 56.70 18.90 41.05 <.0001 B|A 1 28.32 28.32 61.52 <.0001 AxB 3 21.43 7.14 15.51 <.0001 Within 205 94.38 0.46 Total 212 207.20 AxB effect sliced by season and species Summer 3 77.32 25.77 55.98 <.0001 Winter 3 5.99 2.00 4.33 0.0055 Factor A: Time Factor B: Season

Results table of factorial ANOVA to determine the effects of different species and seasons on mean stomach fullness (a) and RCV (b) of N. erebi. Note that in the presence of a significant interaction, the significance of tests for the main effects are unreliable, thus the necessity to split the analysis.

Source DF SS MS F value Pr > F (a) Stomach fullness Among cells 7 1433.90 204.84 78.12 <.0001 A|B 3 1010.56 336.85 128.46 <.0001 B|A 1 232.01 232.01 88.48 <.0001 AxB 3 221.23 73.74 28.12 <.0001 Within 204 534.92 2.62 Total 211 1968.82 AxB effect sliced by season Summer 3 801.58 267.19 101.90 <.0001 Winter 3 402.56 134.19 51.17 <.0001 (b) RCV Among cells 7 179.30 25.61 92.52 <.0001 A|B 3 102.68 34.23 123.63 <.0001 B|A 1 54.11 54.11 195.44 <.0001 AxB 3 25.78 8.59 31.04 <.0001 Within 204 56.48 0.28 Total 211 235.78 AxB effect sliced by season and species Summer 3 96.34 32.11 115.99 <.0001 Winter 3 29.06 9.69 35.00 <.0001 Factor A: Time Factor B: Season

221

Appendix 3. Total and standard length relationships for all three species (Chapter 5)

Ambassis agassizii (n=519) 50

45 y = 0.7555x + 0.9344 40 r2 = 0.9892 35

20 Standard length (mm) length Standard 15

0 0 102030405060 Total length (mm) Leiopotherapon unicolor (n=378) 140

y = 0.8165x - 0.5269 120 r2 = 0.9976

100

60 Standard length (mm) length Standard 40

0 0 20 40 60 80 100 120 140 160 Total length (mm) Nematalosa erebi (n=948) 120

y = 0.7598x + 0.8679 100 r2 = 0.9937

Standard length (mm) length Standard 40

0 0 20406080100120140 Total length (mm) 222

Appendix 4. Indicator Species Analysis results (Chapter 6)

Ambassis agassizii ISA results output from PC-ORD:

MONTE CARLO test of significance of observed maximum indicator value for food items 1000 permutations. Random number seed: 1614 Maxgrp 1=summer, 2=winter ------IV from Observed randomized Indicator groups Column Maxgrp Value (IV) Mean S.Dev p * ------1 detrit 1 24.1 12.7 4.84 0.0250 2 calan 2 58.9 52.9 2.71 0.0170 3 cyclop 2 18.8 25.9 6.38 0.9420 4 moin 1 69.6 54.2 3.59 0.0010 5 daphn 2 76.6 42.2 6.46 0.0010 6 bosm 1 6.7 5.3 1.14 0.3930 7 chydo 2 21.7 12.4 5.18 0.1360 8 sidid 2 8.7 7.2 3.53 0.5160 9 chiron-p 2 12.9 12.8 5.19 0.4830 10 chiron-l 2 21.7 12.5 4.99 0.1290 11 zigop 1 13.3 7.1 3.36 0.1480 ------* proportion of randomized trials with indicator value equal to or exceeding the observed indicator value. p = (1 + number of runs >= observed)/(1 + number of randomized runs) Maxgrp = Group identifier for group with maximum observed IV

MONTE CARLO test of significance of observed maximum indicator value for food items 1000 permutations. Random number seed: 1119 Maxgrp 1=SCB-Feb/02, 2=SCB-May-02, 4=RBW-Aug/02, 7=RBW-Mar/03, 8=RBW-Jul/03 ------IV from Observed randomized Indicator groups Column Maxgrp Value (IV) Mean S.Dev p * ------1 detrit 1 66.3 18.3 10.24 0.0040 2 calan 8 19.8 16.4 1.38 0.0060 3 cyclop 2 36.9 18.9 6.89 0.0260 4 moin 1 25.9 18.4 2.19 0.0020 5 daphn 2 48.1 21.2 5.09 0.0010 6 bosm 1 20.0 21.0 8.51 0.5780 7 chydo 2 23.2 17.3 9.73 0.2150 8 sidid 4 50.0 19.8 10.11 0.0410 9 chiron-p 2 21.8 20.0 10.69 0.3120 10 chiron-l 4 45.5 17.6 10.30 0.0270 11 zigop 7 66.7 19.6 9.05 0.0120 ------* proportion of randomized trials with indicator value equal to or exceeding the observed indicator value. p = (1 + number of runs >= observed)/(1 + number of randomized runs) Maxgrp = Group identifier for group with maximum observed IV

223

Leiopotherapon unicolor ISA results output from PC-ORD:

MONTE CARLO test of significance of observed maximum indicator value for food items 1000 permutations. Random number seed: 386 Maxgrp 1=summer, 2=winter ------IV from Observed randomized Indicator groups Column Maxgrp Value (IV) Mean S.Dev p * ------1 detrit 1 34.3 29.4 7.57 0.2100 2 veg-m 2 50.9 36.6 7.36 0.0530 3 calan 2 66.0 45.6 7.52 0.0200 4 cyclop 2 97.6 38.2 7.31 0.0010 5 moin 1 34.5 34.9 7.52 0.4210 6 daphn 2 100.0 36.3 6.85 0.0010 7 chydo 2 62.6 31.7 7.68 0.0020 8 ostrac 1 29.0 35.3 7.71 0.7640 9 conch 2 56.2 27.5 7.38 0.0050 10 nema 1 7.7 6.9 0.75 0.4530 11 leptop 2 46.1 24.3 7.26 0.0110 12 coenag 2 60.6 31.4 7.37 0.0060 13 noton 1 49.4 22.4 7.35 0.0070 14 corix 1 67.9 52.0 5.82 0.0100 15 chi-l 2 68.5 45.3 5.87 0.0020 16 chi-p 2 72.5 44.1 6.52 0.0010 17 ecnom 2 62.4 31.5 7.61 0.0040 18 leptoc 2 18.7 11.8 5.07 0.2330 19 hydrop 2 12.5 9.1 4.71 0.5080 20 planor 2 6.2 6.9 0.74 1.0000 21 shrimp 1 53.8 20.2 6.77 0.0010 22 fish 1 30.8 13.7 5.68 0.0250 ------

Random number seed: 4036 Maxgrp 1=S. Callandoon B summer, 2= S. Callandoon B winter, 6= S. Callandoon A winter, 7=Rainbow summer. ------IV from Observed randomized Indicator groups Column Maxgrp Value (IV) Mean S.Dev p * ------1 detrit 1 73.0 27.8 9.42 0.0010 2 veg-m 6 50.5 30.7 8.13 0.0180 3 calan 6 54.6 33.5 6.43 0.0040 4 cyclop 6 63.2 30.6 7.20 0.0010 5 moin 1 72.3 30.1 8.77 0.0010 6 daphn 2 63.4 29.8 7.44 0.0010 7 chydo 2 84.4 29.6 9.89 0.0010 8 ostrac 1 69.7 30.5 8.86 0.0010 9 conch 6 100.0 25.6 9.18 0.0010 10 nema 1 16.7 15.8 7.03 0.5830 11 leptop 6 71.7 25.1 10.07 0.0020 12 coenag 6 82.7 26.4 9.02 0.0010 13 noton 7 56.5 20.6 11.63 0.0120 14 corix 7 34.9 31.6 4.26 0.2020 15 chi-l 6 40.8 30.3 3.17 0.0010 16 chi-p 2 42.4 31.1 4.59 0.0140 17 ecnom 6 63.6 27.6 9.40 0.0040 18 leptoc 6 30.0 18.2 10.25 0.1000 19 hydrop 6 20.0 17.2 9.54 0.3990 20 planor 6 10.0 15.9 6.88 1.0000 21 shrimp 7 100.0 18.6 10.18 0.0010 22 fish 7 33.3 16.2 7.27 0.1290 ------* proportion of randomized trials with indicator value equal to or exceeding the observed indicator value. p = (1 + number of runs >= observed)/(1 + number of randomized runs) Maxgrp = Group identifier for group with maximum observed IV

224

Small Nematalosa erebi ISA results output from PC-ORD:

MONTE CARLO test of significance of observed maximum indicator value for food items 1000 permutations. Random number seed: 2535 Maxgrp 2=South Callandoon A, 3=Maynes, 4=Punbougal ------IV from Observed randomized Indicator groups Column Maxgrp Value (IV) Mean S.Dev p * ------1 detrit 4 44.3 30.9 3.88 0.0010 2 volvox 3 33.3 22.9 8.81 0.1080 3 alg 4 16.7 15.4 1.28 0.4550 4 f-alg 4 16.7 15.4 1.29 0.4840 5 veg-m 4 59.7 25.1 7.80 0.0030 6 rotif 2 42.9 16.3 9.35 0.0450 7 calan 3 35.9 29.1 6.17 0.1550 8 cyclop 2 42.9 15.3 9.38 0.0390 9 moin 3 42.7 29.1 6.91 0.0520 10 daphn 2 42.9 15.5 9.58 0.0450 11 bosm 2 42.9 16.4 9.33 0.0450 12 l-chi 3 10.8 14.3 9.00 0.4720 ------* proportion of randomized trials with indicator value equal to or exceeding the observed indicator value. p = (1 + number of runs >= observed)/(1 + number of randomized runs) Maxgrp = Group identifier for group with maximum observed IV

MONTE CARLO test of significance of observed maximum indicator value for food items 1000 permutations. Random number seed: 2299 Maxgrp 7=PBL-Oct/02, 8=SCA-Mar/03, 12=MNS-Mar/03, 13=PBL-Mar/03, 14=RBW-Jul/03, SCA-Jul/03 ------IV from Observed randomized Indicator groups Column Maxgrp Value (IV) Mean S.Dev p * ------1 detrit 13 20.4 17.9 2.05 0.1000 2 volvox 12 47.0 24.0 10.53 0.0510 3 alg 13 33.3 36.5 6.68 1.0000 4 f-alg 7 33.3 35.9 6.11 1.0000 5 veg-m 7 34.9 21.6 6.11 0.0030 6 rotif 15 100.0 26.0 13.79 0.0050 7 calan 14 26.5 19.7 2.91 0.0110 8 cyclop 8 100.0 23.2 14.89 0.0060 9 moin 12 31.6 21.2 4.35 0.0110 10 daphn 15 100.0 23.8 13.89 0.0050 11 bosm 15 100.0 26.1 13.97 0.0050 12 l-chi 12 20.4 26.8 13.67 1.0000 ------* proportion of randomized trials with indicator value equal to or exceeding the observed indicator value. p = (1 + number of runs >= observed)/(1 + number of randomized runs) Maxgrp = Group identifier for group with maximum observed IV

225

Large Nematalosa erebi ISA results output from PC-ORD:

MONTE CARLO test of significance of observed maximum indicator value for food items 1000 permutations. Random number seed: 2311 Maxgrp 1=Rainbow, 2=South Callandoon A, 3=Maynes, 4=Punbougal, 5=Macintyre River ------IV from Observed randomized Indicator groups Column Maxgrp Value (IV) Mean S.Dev p * ------1 detrit 5 20.8 20.4 0.66 0.0060 2 volvox 2 74.8 22.4 6.50 0.0010 3 alg 1 34.3 16.8 6.47 0.0200 4 f-alg 4 25.6 15.7 6.40 0.0810 5 veg-m 4 29.3 23.4 2.03 0.0030 6 calan 2 76.4 17.0 6.90 0.0010 7 moin 2 52.7 18.7 6.86 0.0020 8 daphn 1 15.5 11.9 6.69 0.2280 9 chydo 2 11.1 9.3 4.42 0.5970 10 ostrac 3 74.9 15.1 6.79 0.0010 11 nema 1 4.3 8.8 4.29 1.0000 12 l-chi 3 16.7 9.1 4.32 0.1090 ------* proportion of randomized trials with indicator value equal to or exceeding the observed indicator value. p = (1 + number of runs >= observed)/(1 + number of randomized runs) Maxgrp = Group identifier for group with maximum observed IV

MONTE CARLO test of significance of observed maximum indicator value for food items 1000 permutations. Random number seed: 1205 Maxgrp 1=RBW-Nov/01, 3=SCA-May/02, 4=RBW-Aug/02, 6=MNS-Oct/02, 8=SCA-Mar/03, 9=RBW-Mar/03, 10=RIV-Mar/03, 12=MNS-Mar/03, 13=PBL-Mar/03 ------IV from Observed randomized Indicator groups Column Maxgrp Value (IV) Mean S.Dev p * ------1 detrit 10 8.0 8.0 0.26 0.1140 2 volvox 8 40.2 16.3 4.74 0.0010 3 alg 1 33.1 15.5 6.29 0.0280 4 f-alg 13 43.5 15.1 6.27 0.0010 5 veg-m 4 13.4 10.9 0.98 0.0150 6 calan 3 30.4 15.9 6.80 0.0530 7 moin 8 33.9 16.3 6.04 0.0200 8 daphn 4 71.9 16.1 8.91 0.0030 9 chydo 3 16.7 23.4 7.64 1.0000 10 ostrac 12 39.3 14.7 6.40 0.0090 11 nema 9 33.3 24.0 7.85 0.4060 12 l-chi 6 33.3 23.4 7.68 0.3700 ------* proportion of randomized trials with indicator value equal to or exceeding the observed indicator value. p = (1 + number of runs >= observed)/(1 + number of randomized runs) Maxgrp = Group identifier for group with maximum observed IV

226

Bibliography

Allen, G. R. 1996. Freshwater fishes of south-eastern Australia. Glassfishes, chanda perches. Pages 146-149 in R. M. McDowall, editor. Freshwater fishes of south- eastern Australia. Reed, Chatswood. 247 pp.

Allen, G. R., and W. E. Burgess. 1990. A review of the glassfishes (Chandidae) of Australia and New Guinea. Records of the Western Australian Museum Suppl. 34:139-206.

Almeida, R. G. 1984. Biologia alimentar de três espécies de Triportheus (Pisces: Characoidei, Characidae) do lago do Castanho, Amazonas. Acta Amazônica 14:48- 76.

Amoros, C., and G. Bornette. 2002. Connectivity and biocomplexity in waterbodies of riverine floodplains. Freshwater Biology 47:517-539.

Araujo-Lima, C. A. R. M., A. A. Agostinho, and N. N. Fabre. 1995. Trophic aspects of fish communities in Brazilian rivers and reservoirs. Pages 105-136 in J. G. Tundisi, C. E. M. Bicudo, and T. M. Tundisi, editors. Limnology in Brazil. Brazilian Academy of Sciences, Rio de Janeiro.

Araujo-Lima, C. A. R. M., B. R. Forsberg, R. Victoria, and L. Martinelli. 1986. Energy sources for detritivorous fishes in the Amazon. Science 234:1256-1258.

Arthington, A., D. R. Bluhdorn, and M. J. Kennard. 1994. Food resource partitioning by the introduced cichlid, Oreochromis mossambicus, and two native fishes in a subtropical Australian impoundment. Pages 425-428 in L. M. Chou, A. D. Munro, T. J. Lam, T. W. Chen, L. K. K. Cheong, J. K. Ding, K. K. Hooi, H. W. Khoo, V. P. E. Phang, K. F. Shim, and C. H. Tan, editors. Proceeding of the Third Asian Fisheries Forum, Singapore. 26-30 October, 1992. Asian Fisheries Society, Manila, Philippines. 1135 pp.

Arthington, A. H. 1992. Diets and trophic guild structure of freshwater fishes in Brisbane streams. Proceedings of the Royal Society of Queensland 102:31-47.

Arthington, A. H. 1995. State of the rivers in cotton growing areas. Northern NSW and the border rivers with Queensland. Occasional Paper Series No. 95/02, Land and Water Resources, Research and Development Corporation, Canberra. 99 pp.

Arthington, A. H., B. M. Bycroft, and D. L. Conrick. 1992a. The fish of Barker-Barambah Creek: flow and passage requirements. in A. H. Arthington, editor. Environmental Study Barker Barambah Creek. Water quality, ecology and water allocation strategy. Queensland Government. Scientific Report, Brisbane. 456 pp.

Arthington, A. H., D. L. Conrick, and B. M. Bycroft. 1992b. Environmental study, Barker- Barambah Creek. Volume 2. Scientific report: Water quality, ecology and water allocation strategy, Water Resources Commission, Department of Primary 227

Industries and Centre for Catchment and In-Stream Research, Griffith University. 457 pp.

Arthington, A. H., and D. A. Milton. 1983. Effects of urban development and habitat alterations on the distribution and abundance of native and exotic freshwater fish in the Brisbane region, Queensland. Australian Journal of Ecology 8:87-101.

Atkins, B. 1984. Feeding ecology of Nematolosa erebi in the lower River Murray. Honors Thesis. University of Adelaide, Adelaide. 82 pp.

Balcombe, S. R., S. E. Bunn, P. M. Davies, and F. J. McKenzie-Smith. in review. Temporal variability in fish diets of an arid zone river, Cooper Creek, Western Queensland. Journal of Fish Biology:18.

Baldwin, D. S., and A. M. Mitchell. 2000. The effects of drying and re-flooding on the sediment or soil nutrient dynamics of lowland river floodplain systems: a synthesis. Regulated Rivers: Research & Management 16:457-467.

Bass, J. A. B., L. C. V. Pinder, and D. V. Leach. 1997. Temporal and spatial variation in zooplankton populations in the river Great Ouse: an ephemeral food resource for larval and juvenile fish. Regulated Rivers: Research & Management 13:245-258.

Bayley, P. B. 1995. Understanding large river-floodplain ecosystems. BioScience 45:153- 158.

Bayley, P. B., and H. W. Li. 1992. Riverine fishes. Pages 251-281 in P. Calow and G. E. Petts, editors. The rivers handbook: hydrological and ecological principles. Blackwell Scientific Publications, Oxford ; Melbourne. 526 pp.

Biondini, M. E., C. D. Bonham, and E. F. Redente. 1985. Secondary successional patterns in a sagebrush (Artemisia tridentata) community as they relate to soil disturbance and soil biological activity. Vegetatio 60:25-36.

Bishop, K. A., S. A. Allen, D. A. Pollard, and M. G. Cook. 2001. Ecological studies on the freshwater fishes of the Alligator Rivers Region, Northern Territory: Autecology. Supervising Scientist Report 145, Environment Australia, Department of the Environment and Heritage, Darwin. 570 pp.

Bluhdorn, D. R., and A. H. Arthington. 1994. The effects of flow regulation in the Barker- Barambah catchment. Volume 2 - Biotic studies and synthesis, Griffith University, Centre for Catchment and In-Stream Research, Brisbane. 421 pp.

Boddy, J., and M. Bales. 1996. Border Rivers instream biological resources study. Report on biological surveys of macroinvertebrates, aquatic macrophytes and riparian vegetation of the borders rivers. TS96.068, Department of Land and Water Conservation Technical Services Division, Parramatta, NSW. 103 pp.

Boey, A., H. Daly, and B. Cooper. 1995. Water quality in the Border Rivers basin: 1960- 1995. Department of Land and Water Conservation. 228

Boisclair, D., and W. C. Leggett. 1988. An in situ experimental evaluation of the Elliot and Persson and the Eggers models for estimating fish daily rations. Canadian Journal of Fisheries and Aquatic Sciences 46:457-467.

Boon, P. I. 1991. Bacterial assemblages in rivers and billabongs of southeastern Australia. Microbial Ecology 22:27-52.

Boon, P. I., and S. E. Bunn. 1994. Variations in the stable isotope composition of aquatic plants and their implications for food web analysis. Aquatic Botany 48:99-108.

Boon, P. I., J. Frankenburg, T. J. Hillman, R. L. Oliver, and R. J. Sheil. 1990. Billabongs. Pages 183-198 in N. Mackay and D. Eastburn, editors. The Murray. Murray- Darling Basin Commission, Canberra. 363 pp.

Borgstrom, R., and E. Plahte. 1992. Gillnet selectivity and a model for capture probabilities for a stunted brown trout (Salmo trutta) population. Canadian Journal of Fisheries and Aquatic Sciences 49:1546-1554.

Boujard, T. 1995. Diel rhythms of feeding activity in the European catfish, Silurus glains. Physiology & Behaviour 58:641-645.

Boujard, T., P. Keith, and P. Luquet. 1990. Diel cycle in Hoplosternum littorale (Teleostei): evidence for synchronization of locomotor, air breathing and feeding activity by circadian alternation of light and dark. Journal of Fish Biology 36:133- 140.

Boujard, T., and J. F. Leatherland. 1992. Circadian rhythms and feeding time in fishes. Environmental Biology of Fishes 35:109-131.

Boulton, A. 1999. Why variable flows are needed for invertebrates of semi-arid rivers. Pages 113-128 in R. T. Kingsford, editor. A Free-flowing River: The Ecology of the Paroo River. National Parks and Wildlife Service, Hurstville, N.S.W. 320 pp.

Boulton, A. J., and L. N. Lloyd. 1991. Aquatic macroinvertebrate assemblages in floodplain habitats of the lower River Murray. Regulated Rivers: Research & Management 6:183-201.

Bowen, S. H. 1983. Quantitative description of the diet. Pages 325-336 in L. A. Nielsen and D. L. Johnson, editors. Fisheries Techniques. American Fisheries Society, Bethesda, Maryland. 468 pp.

Bowen, S. H. 1984. Detritivory in neotropical fish communities. Pages 59-66 in T. M. Zaret, editor. Evolutionary ecology of neotropical freshwater fishes. Dr. W. Junk Publishers, The Hague, Netherlands. 173 pp.

Bowling, L., and H. Jones. 2003. Impacts of cattle grazing on water quality at Glennies Creek water storage. Centre for Natural Resources. NSW Department of Land and Water Conservation, Parramatta. 45 pp.

Breck, J. E., and M. J. Gitter. 1983. Effect of fish size on the reactive distance of bluegill (Lepomis macrochirus) sunfish. Canadian Journal of Fisheries and Aquatic Sciences 40:162-167. 229

Bren, L. J. 1993. Riparian zone, stream, and floodplain issues - a review. Journal of Hydrology 150:277-299.

Briggs, I. C., and R. M. McDowall. 1996. Freshwater fishes of south-eastern Australia. Herrings. Pages 44-47 in R. M. McDowall, editor. Freshwater fishes of south- eastern Australia. Reed, Chatswood. 247 pp.

Brodeur, R. D. 1991. Ontogenetic variations in the type and size of prey consumed by juvenile coho, Oncorhynchus kisutch, and chinook, Oncorhynchus tshawytscha, salmon. Environmental Biology of Fishes 30:303-315.

Brookes, A. 1994. River channel change. Pages 55-75 in P. Calow and G. E. Petts, editors. The rivers handbook. Hydrological and ecological principles. Blackwell Scientific Publications, Oxford.

Bruton, M. N. 1985. Effects of suspensoids on fish. Pages 221-241 in B. R. Davies and R. D. Walmsley, editors. Perspectives in Southern Hemisphere Limnology. Dr. W. Junk Publishers, Dordrecht.

Bryan, C. F., and D. S. Sabins. 1979. Management implications in water quality and fish standing stock information in the Atchafalaya river basin, Louisiana. Pages 239- 316 in J. W. Day, Jr., D. D. Culley, Jr., R. E. Turner, and A. J. Mumphrey, Jr., editors. Proc. Third Coastal Marsh and Estuary Management Symposium, Louisiana State University Division of Continuing Education, Baton Rouge, LA.

Bunn, S. E., and A. H. Arthington. 2002. Basic principles and ecological consequences of altered flow regimes for aquatic biodiversity. Environmental Management 30:492- 507.

Bunn, S. E., S. R. Balcombe, P. M. Davies, C. S. Fellows, and F. J. McKenzie-Smith. in press. Productivity and aquatic food webs of desert river ecosystems.

Bunn, S. E., and P. I. Boon. 1993. What sources of organic carbon drive food webs in billabongs? A study based on stable isotope analysis. Oecologia 96:85-94.

Bunn, S. E., and P. M. Davies. 1999. Aquatic foodwebs in turbid, arid-zone rivers: preliminary data from Cooper Creek, Western Queensland. Pages 67-76 in R. T. Kingsford, editor. A Free-flowing River: The Ecology of the Paroo River. National Parks and Wildlife Service, Hurstville, N.S.W. 320 pp.

Bunn, S. E., P. M. Davies, D. M. Kellaway, and I. P. Prosser. 1998. Influence of invasive macrophytes on channel morphology and hydrology in an open tropical lowland stream, and potential control by riparian shading. Freshwater Biology 39:171-178.

Bunn, S. E., P. M. Davies, and M. Winning. 2003. Sources of organic carbon supporting the food web of an arid zone floodplain river. Freshwater Biology 48:619-635.

Bunn, S. E., N. R. Loneragan, and M. A. Kempster. 1995. Effects of acid washing on stable isotope ratios of C and N in panaeid shrimp and seagrass: implications for food-web studies using multiple stable isotopes. Limnology and Oceanography 40:622-625. 230

Butcher, R. 1997. Marshes, swamps and billabongs - bugs love 'em. Watershed October:3- 5.

Byrkit, D. R. 1987. Statistics today: a comprehensive introduction. Benjamin/Cummings Pub. Co., Menlo Park, California. 850 pp.

Cadwallader, P. L. 1978. Some causes of the decline in range and abundance of native fish in the Murray-Darling river system. Proceedings of the Royal Society of Victoria 90:211-224.

Campbell, I. C., C. R. James, B. T. Hart, and A. Deveraux. 1992. Allochthonous coarse particulate organic material in forest and pasture reaches of two south-eastern Australian stream. I. Litter accession. Freshwater Biology 27:341-352.

Capon, S. J. 2004. Flow variability and vegetation dynamics in a large arid floodplain: Cooper Creek, Australia. PhD Thesis. Griffith University, Brisbane. 226 pp.

Cardona, L. 1999. Seasonal changes in the food quality, diel feeding rhythm and growth rate of juvenile leaping grey mullet Liza saliens. Aquatic Living Resources 12:263- 270.

Casanova, M. T. 1999. Plant establishment in Paroo wetlands: the importance of water regime. Pages 138-148 in R. T. Kingsford, editor. A free-flowing river: the ecology of the Paroo River. National Parks and Wildlife Service, Hurstville, N.S.W. 320 pp.

Castillo, M. M. 2000. Influence of hydrological seasonality on bacterioplankton in two neotropical floodplain lakes. Hydrobiologia 437:57-69.

Cellot, B., M. J. Dole-Olivier, G. Bornette, and G. Pautou. 1994. Temporal and spatial environmental variability in the Upper Rhone River and its floodplain. Freshwater Biology 31:311-325.

Clapcott, J. 2001. Can C4 plants contribute to the aquatic food webs of subtropical Queensland streams? M.Sc. Thesis. Griffith University, Brisbane. 51 pp.

Cole, G. A. 1994. Textbook of limnology, 4th edition. Waveland Press, Prospect Heights, Illinois. 412 pp.

Cranston, P. C., and T. J. Hillman. 1992. Rapid assessment of biodiversity using 'Biological Diversity Techniques'. Australian Biologist 5:144-154.

Creach, V., M. T. Schricke, G. Bertru, and A. Mariotti. 1997. Stable isotopes and gut analysis to determine feeding relationships in saltmarsh macroconsumers. Estuarine, Coastal and Shelf Science 44:599-611.

Davis, L., M. C. Thoms, C. Fellows, and S. E. Bunn. 2002. Physical and ecological associations in dryland refugia: waterholes of the Cooper Creek, Australia. Pages 77-84 in The Structure, Function and Management Implications of Fluvial Sedimentary Systems. IAHS Publ. no. 276, Proceeding of an international symposium held at Alice Springs, Australia, September, 2002. 231

Day, J. A., and B. R. Davies. 1986. The Amazon River system. Pages 289-318 in B. R. Davies and K. F. Walker, editors. The ecology of river systems. Dr. W. Junk Publishers, Dordrecht. 793 pp.

DBBRC. 1988. Boggabilla Weir Environmental Impact Statement. Dumaresq-Barwon Border Rivers Commission, Brisbane.

De Silva, S. S. 1985. Body condition and nutritional ecology of Oreochromis mossambicus (Pisces, Cichlidae) populations of man-made lakes in Sri Lanka. Journal of Fish Biology 27:621-633.

De Silva, S. S., U. S. Amarasinghe, and N. D. N. S. Wijegoonawardena. 1996. Diel feeding patterns and daily ration of cyprinid species in the wild determined using an iterative method, MAXIMS. Journal of Fish Biology 49:1153-1162.

Delong, M. D., J. H. Thorp, K. S. Greenwood, and M. C. Miller. 2001. Responses of consumers and food resources to a high magnitude, unpredicted flood in the upper Mississippi river basin. Regulated Rivers: Research & Management 17:217-234.

DeNiro, M. J., and S. Epstein. 1981. Influence of diet on the distribution of nitrogen isotopes in animals. Geochimica et Cosmochimica Acta 45:341-351.

DIPNR. 2002. Water quality in the Border Rivers catchment 2001-2002. Department of Infrastructure, Planning and Natural Resources, Tamworth. 13 pp.

DNR, and DLWC. 1999. Current ecological condition of streams in the Border Rivers catchment. Border Rivers flow management planning. Information Paper, DNRQ00020, Department of Natural Resources (Queensland) and Department of Land and Water Conservation (New South Wales). 58 pp.

DNR, and DLWC. 2000. Border Rivers flow management planning. Stage 1. Information Paper, DNRQ00115, Department of Natural Resources (Queensland) and Department of Land and Water Conservation (New South Wales). 40 pp.

Dufrene, M., and P. Legendre. 1997. Species assemblages and indicator species: the need for a flexible asymmetrical approach. Ecological Monographs 67:345-366.

Dunbrack, R. L., and L. M. Dill. 1983. A model of size dependent surface feeding in a stream dwelling salmonid. Environmental Biology of Fishes 8:203-216.

DWR. 1990. Pindari Dam enlargement proposal environmental impact statement. Volume 1, New South Wales Department of Water Resources.

DWR. 1995. Water resources of the Border Rivers system in northern New South Wales. New South Wales Department of Water Resources.

Ebner, B., R. McAllister, and P. Suter. in press. Effects of sample size on numerical estimates of prey consumption in a fish population. Journal of Fish Biology.

Eggers, D. M. 1977. Factors in interpreting data obtained by diel sampling of fish stomachs. Journal of the Fisheries Research Board of Canada 34:290-294. 232

Elliott, J. M. 1972. Rates of gastric evacuation in brown trout, Salmo trutta L. Freshwater Biology 2:1-18.

Elliott, J. M. 1981. Some aspects of thermal stress on freshwater teleosts. Pages 209-245 in A. D. Pickering, editor. Stress and Fish. Academic Press, London.

Entwisle, T. J., J. A. Sonneman, and S. H. Lewis. 1997. Freshwater algae in Australia: a guide to Conspicuous Genera. Sainty & Associates, Sydney. 242 pp.

Faragher, R. A., and J. H. Harris. 1993. The historical and current status of freshwater fish in New South Wales. Australian Zoologist 29:166-176.

Ferreira, E. J. G. 1984. A ictiofauna da represa hidrelétrica de Curuá-Una, Santarém, Pará. II - Alimentação e hábitos alimentares das principais espécies. Amazoniana 9:1-16.

Finlayson, C. M., B. J. Bailey, and I. D. Cowie. 1989. Macrophyte vegetation of the Magela Creek flood plain, Alligator Rivers region, Northern Territory. Research report 5, Australian Government Publishing Service, Canberra. 38 pp.

Fisher, S. G., N. B. Grimm, E. Marti, R. M. Holmes, and J. B. Jones. 1998. Material spiraling in stream corridors: a telescoping ecosystem model. Ecosystems 1:19-34.

Fisher, S. J., M. L. Brown, and D. W. Willis. 2001. Temporal food web variability in an upper Missouri River backwater: energy origination points and transfer mechanisms. Ecology of Freshwater Fish 10:154-167.

Forsberg, B. R., C. A. R. M. Araujo-Lima, L. A. Martinelli, R. L. Victoria, and J. A. Bonassi. 1993. Autotrophic carbon sources for fish of the central Amazon. Ecology 74:643-652.

France, R. L. 1996. Carbon-13 conundrums: limitations and cautions in the use of stable isotope analysis in stream ecotonal research. Canadian Journal of Fisheries and Aquatic Sciences 53:1916-1919.

Fremling, C. R., J. L. Rasmussen, R. E. Sparks, S. P. Cobb, C. F. Bryan, and T. O. Caflin. 1989. The Mississippi River fisheries: a case history. Special Publication of the Canadian Journal of Fisheries and Aquatic Sciences 106:309-351.

Friday, L. E. 1987. The diversity of macroinvertebrate and macrophyte communities in ponds. Freshwater Biology 18:87-104.

Fry, B. 1991. Stable isotope diagrams of freshwater food webs. Ecology 72:2293-2297.

Fry, B., A. Joern, and P. L. Parker. 1978. Grasshopper food web analysis: use of carbon isotope ratios to examine feeding relationships among terrestrial herbivores. Ecology 59:498-506.

Garcia-Berthou, E. 1999. Spatial heterogeneity in roach (Rutilus rutilus) diet among contrasting basins within a lake. Archive fur Hydrobiologie 146:239-256.

Geddes, M. C., and J. T. Puckridge. 1989. Survival and growth of larval and juvenile native fish. The importance of the floodplain. Pages 101-115 in Murray-Darling 233

Basin Commission, editor. Proceedings of the Workshop on Native Fish Management, Canberra, 16-17 June, 1988. Murray-Darling Basin Commission, Canberra. 174 pp.

Gehrke, P. C. 1988a. Feeding energetics and angling catches of spangled perch, Leiopotherapon unicolor (Gunther 1859), (Percoidei: Teraponidae). Australian Journal of Marine and Freshwater Research 39:569-577.

Gehrke, P. C. 1988b. Influence of gut morphology, sensory cues and hunger on feeding behaviour of spangled perch, Leiopotherapon unicolor (Gunther, 1859), (Percoidei, Terapontidae). Journal of Fish Biology 33:189-201.

Gehrke, P. C., P. Brown, C. B. Schiller, D. B. Moffatt, and A. M. Bruce. 1995. River regulation and fish communities in the Murray-Darling river system, Australia. Regulated Rivers: Research & Management 11:363-375.

Gerking, S. D. 1994. Feeding ecology of fish. Academic Press, San Diego. 416 pp.

Gibson, R. J., and M. H. A. Keenleyside. 1966. Responses to light of young Atlantic Salmon (Salmo salar) and Brook Trout (Salvelinus fontinalis). Journal of Fisheries Research Board of Canada 23:1007-1024.

Glasser, J. W. 1983. Variation in niche breadth with trophic position: on the disparity between expected and observed species packing. The American Naturalist 122:542- 548.

Gliwicz, M. Z. 1986. Predation and the evolution of vertical migration in zooplankton. Nature 320:746-748.

Gopal, B. 1986. Vegetation dynamics in temporary and shallow freshwater habitats. Aquatic Botany 23:391-396.

Gosselink, J. G., and R. E. Turner. 1978. The role of hydrology in freshwater wetland ecosystems. Pages 63-78 in R. E. Good, D. F. Whigham, R. L. Simpson, and C. G. Jackson, Jr., editors. Freshwater wetlands : ecological processes and management potential. Academic Press, New York. 378 pp.

Goulding, M. 1980. The fishes and the forest: explorations in Amazonian natural history. California Univ. Press, Berkeley. 280 pp.

Gray, C. A., R. C. Chick, and D. J. McElligott. 1998. Diel changes in assemblages of fishes associated with shallow seagrass and bare sand. Estuarine, Coastal and Shelf Science 46:849-859.

Griswold, B. L., C. J. Edwards, and L. C. Woods. 1982. Recolonization of macroinvertebrates and fish in a channelized stream after a drought. Ohio Journal of Science 82:96-102.

Grossman, G. D. 1980. Ecological aspects of ontogenetic shifts in prey size utilization in the bay goby (Pisces: Gobiidae). Oecologia 47:233-238. 234

Gutteridge, Haskins, and Davey. 1992. An investigation of nutrient pollution in the Murray-Darling River system. Murray-Darling Basin Commission, Camberra.

Hadwen, W. L. 2003. Effects of nutrient additions on dune lakes on Fraser Island, Australia. PhD Thesis. Griffith University, Brisbane. 150 pp.

Hadwen, W. L., and S. E. Bunn. in press. Tourists increase the contribution of autochthonous carbon to littoral zone food webs in oligotrophic dune lakes. Marine and Freshwater Research 55.

Hamilton, S. K., and W. M. Lewis, Jr. 1987. Causes of seasonality in the chemistry of a lake on the Orinoco River floodplain, Venezuela. Limnology and Oceanography 32:1277-1290.

Hamilton, S. K., S. J. Sippel, W. M. Lewis, Jr., and I. J.F. Saunders. 1990. Zooplankton abundance and evidence for its reduction by macrophyte mats in two Orinoco floodplain lakes. Journal of Plankton Research 12:345-363.

Hamilton, S. K., J. W.M. Lewis, and S. J. Sippel. 1992. Energy sources for aquatic animals in the Orinoco River floodplain: evidence from stable isotopes. Oecologia 89:324- 330.

Harris, J. H. 1985. Diet of the Australian bass, Macquaria novemaculeata (Perciformes: Percichthyidae), in the Sydney basin. Australian Journal of Marine and Freshwater Research 36:219-234.

Hart, B. T., and R. J. McGregor. 1982. Water quality characteristics of eight billabongs in the Magela Creek catchment. Research report 2, Australian Government Publishing Service, Canberra. 54 pp.

Hawking, J. H. 1986. Dragonfly larvae of the River Murray system: a preliminary guide to the identification of known final instar odonate larvae of south-eastern Australia. Technical report No. 6, Albury-Wodonga Development Corporation, Wodonga, Vic. 65 pp.

Hawking, J. H., and F. J. Smith. 1997. Colour guide to invertebrates of Australian inland waters. Identification guide No. 8, Co-operative Research Centre for Freshwater Ecology, Albury, N.S.W. 213 pp.

Hayes, J. M. 1982. Fractionation et al.: an introduction to isotopic measurements and terminology. Spectra 8:3-8.

Hayward, R. S., F. J. Margraf, Jr., D. L. Parrish, and B. Vondracek. 1991. Low-cost field estimation of yellow perch daily ration in young perch daily ration. Transactions of the American Fisheries Society 120:589-604.

Heiler, G., T. Hein, F. Schiemer, and G. Bornette. 1995. Hydrological connectivity and flood pulses as the central aspects for the integrity of a river-floodplain system. Regulated Rivers: Research & Management 11:351-361.

Hillman, T. 1995a. Flow Regimes of Rivers in the Murray-Darling Basin: Ecological Characteristics and Ecological Consequences. Pages 11-18 in Proceedings of the 235

Water Use and Environmental Flows Workshop, 22-23 August. Murray-Darling Basin Commission.

Hillman, T. 1998. River-billabong interactions: the links that keep our rivers healthy. Watershed March:4-5.

Hillman, T. J. 1986. Billabongs. Pages 457-470 in P. De Deckker and W. D. Williams, editors. Limnology in Australia. Dr W. Junk Publishers, Dordrecht.

Hillman, T. J. 1995b. Billabongs, floodplains and the health of rivers. Water May/June:16-19.

Hobbs, J. E., and I. J. Jackson. 1977. Climate. Pages 75-99 in D. A. M. Lea, J. J. J. Pigram, and L. Greenwood, editors. An Atlas of New England - The Commentaries. Department of Geography, University of New England, Armidale.

Horppila, J. 1999. Diet changes in diet composition of an omnivorous cyprinid - a possible source of error in estimating food consumption. Hydrobiologia 400:33-39.

Horppila, J. J. R., M. Rask, C. Karppinen, K. Nyberg, and M. Olin. 2000. Seasonal changes in the diets and relative abundances of perch and roach in the littoral and pelagic zones of a large lake. Journal of Fish Biology 56:51-72.

Houldsworth, B. 1995. Central and north west regions water quality program. 1994/95 report on nutrients and general water quality monitoring. TS95.088, Department of Land and Water Conservation Technical Services Division, Parramatta, NSW. 95 pp.

Humphries, P., A. J. King, and J. D. Koehn. 1999. Fish, flows and flood plains: links between freshwater fishes and their environment in the Murray-Darling river system, Australia. Environmental Biology of Fishes 56:129-151.

Humphries, P., and P. S. Lake. 2000. Fish larvae and the management of regulated rivers. Regulated Rivers: Research & Management 6:421-432.

Huryn, A. D., R. H. Riley, R. G. Young, C. J. Arbuckle, K. Peacock, and G. Lyon. 2001. Temporal shift in contribution of terrestrial organic matter to consumer production in a grassland river. Freshwater Biology 46:213-226.

Hutchinson, G. E. 1967. A treatise on limnology. John Wiley & Sons, New York. 1115 pp.

Hyslop, E. J. 1980. Stomach contents analysis - a review of methods and their application. Journal of Fish Biology 17:411-429.

Ingram, B. A., J. H. Hawking, and R. J. Shiel. 1997. Aquatic life in freshwater ponds: a guide to the identification and ecology of life in aquaculture ponds and farm dams in south eastern Australia. Identification guide No. 9, Co-operative Research Centre for Freshwater Ecology, Albury, N.S.W. 105 pp.

Jacobsen, L., and S. Berg. 1998. Diel variation in habitat use by planktivores in field enclosure experiments: the effect of submerged macrophytes and predation. Journal of Fish Biology 53:1207-1219. 236

Jepsen, D. B., and K. O. Winemiller. 2002. Structure of tropical river food webs revealed by stable isotope ratios. Oikos 96:46-55.

Johannsson, O. E., M. F. Leggett, L. G. Rudstam, M. R. Servos, M. A. Mohammadian, G. Gal, R. M. Dermott, and R. H. Hesslein. 2001. Diet of Mysis relicta in Lake Ontario as revealed by stable isotope and gut content analysis. Canadian Journal of Fisheries and Aquatic Sciences 58:1975-1986.

Johnson, B. L., W. B. Richardson, and T. J. Naimo. 1995. Past, present, and future concepts in large river ecology. BioScience 45:134-141.

Johnson, D. P. 1999. Border Rivers and Moonie River catchment. An ecological and physical assessment of the condition of streams in the Border Rivers and Moonie River catchments. Queensland Government. Department of Natural Resources. 77 pp.

Junk, W. J. 1985. Temporary fat storage, and adaptation of some fish species to the waterlevel fluctuations and related environmental changes of the Amazon River. Amazoniana IX:315-351.

Junk, W. J., P. B. Bayley, and R. E. Sparks. 1989. The flood pulse concept in river- floodplain systems. Pages 110-127 in D. P. Dodge, editor. Proceedings of the International Large Rivers Symposium. Canadian Special Publication of Fisheries and Aquatic Sciences, 106 pp.

Junk, W. J., and G. E. Weber. 1995. Amazonian floodplains: a limnological perspective. Verhandlungen Internationale Vereinigung für theoretische und angewandte Limnologie 26:149-157.

Junk, W. J., and R. L. Welcomme. 1990. Floodplains. Pages 491-524 in B. C. Patten, editor. Wetlands and Shallow Continental Water Bodies. SPB Academic Publishers, The Hague.

Kadri, S., N. B. Metcalfe, F. A. Huntingford, and J. E. Thorpe. 1997. Daily feeding rhythms in Atlantic salmon II: size-related variation in feeding patterns of post- smolts under constant environmental conditions. Journal of Fish Biology 50:273- 279.

Keast, A. 1968. Feeding of some Great Lakes fishes at low temperatures. Journal of Fisheries Research Board of Canada 25:1199-1218.

Keast, A. 1985. The piscivore feeding guild of fishes in small freshwater ecosystems. Environmental Biology of Fishes 12:119-129.

Keast, A., and L. Welsh. 1968. Daily feeding periodicities, food uptake rates and dietary changes with hour of day in some lake fishes. Journal of the Fisheries Research Board of Canada 25:1133-1144.

Keenleyside, M. H. A. 1979. Diversity and adaptation in fish behavior. Springer-Verlag, Berlin. 208 pp. 237

Kennard, M. J. 1995. Factors influencing freshwater fish assemblage in floodplain lagoons of the Normanby River, Cape York Peninsula: a large tropical Australian River. M.Sc. Thesis. Griffith University, Brisbane. 225 pp.

Kennard, M. J., B. Pusey, and A. Arthington. 2001. Trophic ecology of freshwater fishes in Australia. CCISR - CRCFE, Brisbane. 42 pp.

Kingsford, R. T. 2000. Ecological impacts of dams, water diversions and river management on floodplain wetlands in Australia. Austral Ecology 25:109-127.

Kingsford, R. T., A. L. Curtin, and J. Porter. 1999. Water flows on Cooper Creek in arid Australia determine "boom" and "bust" periods for waterbirds. Biological Conservation 88:231-248.

Kingsford, R. T., K. M. Jenkins, and P. J.L. 2004. Imposed hydrological stability on lakes in arid Australia and effects on waterbirds. Ecology 85:2478-2492.

Kirk, J. T. O. 1985. Effects of suspensoids (turbidity) on penetration of solar radiation in aquatic ecosystems. Hydrobiologia 125:195-209.

Lajtha, K., and R. H. Michener. 1994. Stable isotopes in ecology and environmental science. Blackwell Scientific Publications, Oxford ; Boston. 316 pp.

Lake, P. S. 2000. Disturbance, patchiness, and diversity in streams. Journal of the North American Benthological Society 19:573-592.

Lake, P. S. 2003. Ecological effects of perturbation by drought in flowing waters. Freshwater Biology 48:1161-1172.

Lake, P. S., and R. Marchant. 1990. Australian upland streams: ecological degradation and possible restoration. Proceedings of the Ecological Society of Australia 16:79-91.

Landless, P. J. 1976. Demand-feeding behaviour of rainbow trout. Aquaculture 7:11-25.

Laurie, M., Pettit PTY. LTD. Consulting Engineers and Surveyors, and New South Wales Water Resources Commission. 1983. Macintyre Valley: summary report. New South Wales inland rivers flood plain management studies Water Resources Commission, Sydney. 27 pp.

Leite, R. G., C. A. R. M. Araujo-Lima, R. L. Victoria, and L. A. Martinelli. 2002. Stable isotope analysis of energy sources for larvae of eight fish species from the Amazon floodplain. Ecology of Freshwater Fish 11:56-63.

Lelek, A. 1989. The Rhine River and some of its tributaries under human impact in the last two centuries. Special Publication of the Canadian Journal of Fisheries and Aquatic Sciences 106:469-487.

Lewis, W. M., Jr., S. K. Hamilton, M. A. Lasi, M. Rodriguez, and J. F. Saunders, III. 2000. Ecological determinism on the Orinoco floodplain. BioScience 50:681-692.

Lewis, W. M., Jr., S. K. Hamilton, M. A. Rodriguez, I. J.F. Saunders, and M. A. Lasi. 2001. Foodweb analysis of the Orinoco floodplain based on production estimates 238

and stable isotope data. Journal of the North American Benthological Society 20:241-254.

Lloyd, L. N., A. H. Arthington, and D. A. Milton. 1986. The mosquitofish - a valuable mosquito control agent or a pest? Pages 7-25 in R. L. Kitching, editor. The ecology of exotic animals and plants: some Australian case histories. John Wiley and Sons, Brisbane. 276 pp.

Lowe-McConnell, R. H. 1964. The fishes of the Rupununi savanna district of British Guiana, South America. Part 1. Ecological groupings of fish species and effects of the seasonal cycle on the fish. Journal of the Linnean Society Zoology 45:1-103.

Lowe-McConnell, R. H. 1987. Ecological Studies in Tropical Fish Communities. Cambridge University Press, London. 382 pp.

Lyons, J. 1986. Capture efficiency of a beach seine for seven freshwater fishes in a north temperate lake. North American Journal of Fisheries Management 6:288-289.

Mackay, S. J., A. H. Arthington, M. J. Kennard, and B. J. Pusey. 2003. Spatial variation in the distribution and abundance of submersed macrophytes in an Australian subtropical river. Aquatic Botany 77:169-186.

Mackey, A. P. 1991. Aspects of the limnology of Yeppen Yeppen lagoon, Central Queensland. Australian Journal of Marine and Freshwater Research 42:309-325.

Magoulick, D. D., and R. M. Kobza. 2003. The role of refugia for fishes during drought: a review and synthesis. Freshwater Biology 48:1186-1198.

Mallen-Cooper, M. 1993. Habitat changes and declines of freshwater fish in Australia: what is the evidence and do we need more? Pages 118-123 in D. A. Hancock, editor. Proceedings of Sustainable Fisheries Through Sustaining Fish Habitat,. Australian Society for Fish Biology Workshop, Bureau of Resource Sciences Proceedings, Australian Government Publishing Service, Canberra.

Marchant, R. 1982. Seasonal variation in the macroinvertebrate fauna of billabongs along Magela Creek, Northern Territory. Australian Journal of Marine and Freshwater Research 33:329-342.

Marshall, J. C., F. Sheldon, M. Thoms, and S. Choy. in review. The macroinvertebrate fauna of an Australian dryland river system: spatial and temporal patterns and environmental relationships.

Marshall, S., and M. Elliott. 1997. A comparison of univariate and multivariate numerical and graphical techniques for determining inter- and intraspecific feeding relationships in estuarine fish. Journal of Fish Biology 51:526-545.

McCosker, R. O. 1996. Environmental scan of the border rivers catchment. New South Wales / Queensland. Department of Natural Resources & Department of Land and Water Conservation. 41 pp.

McCune, B., and J. B. Grace. 2002. Analysis of Ecological Communities. MjM Software Design, Gleneden Beach, Oregon, U.S.A. 300 pp. 239

McCune, B., and M. J. Mefford. 1999. PC-ORD. Multivariate Analysis of Ecological Data. MjM Software Design, Gleneden Beach, Oregon, U.S.A.

McCune, B., R. Rosentreter, J. M. Ponzetti, and D. C. Shaw. 2000. Epiphyte habitats in an old conifer forest in western Washington, USA. Bryologist 102:417-427.

McDowall, R. M., editor. 1996. Freshwater fishes of south-eastern Australia. Reed, Chatswood. 247 pp.

McGarity, J. W. 1977. Soils. Pages 47-70 in D. A. M. Lea, J. J. J. Pigram, and L. Greenwood, editors. An Atlas of New England - The Commentaries. Department of Geography, University of New England, Armidale.

McLachlan, A. J., P. R. Morgan, C. Howard-Williams, S. M. McLachlan, and D. Bourn. 1972. Aspects of the recovery of a saline African lake following a dry period. Archive fur Hydrobiologie 70:325-340.

Mérona, B., G. M. Santos, and R. G. Almeida. 2000. Short term effects of Tucuruí dam (Amazonia, Brazil) on the trophic organization of fish communities. Environmental Biology of Fishes 60:375-392.

Merrick, J. 1974. Preliminary systematic studies of the fishes included in the genera Bidyanus and Madigania (Teleostei: Theraponidae). Report for Hydrobiology Course, University of Sydney.

Merrick, J. R., and G. E. Schmida. 1984. Australian freshwater fishes: biology and management. North Ryde, N.S.W. 409 pp.

Michener, R. H., and D. M. Schell. 1994. Stable isotope ratios as tracers in marine aquatic food webs. Pages 139-157 in R. H. Michener, editor. Stable isotopes in ecology and environmental science. Blackwell Scientific Publications, Oxford ; Boston. 316 p.

Miller, P. J. 1979a. Adaptativeness and implications of small size in teleosts. Symposium of the Zoological Society, London 44:263-306.

Miller, R. J. 1979b. Relatioships between habitat and feeding mechanisms in fishes. in H. Clepper, editor. Predator-Prey systems in Fisheries Management. Sport Fishing Institute, Washington, DC.

Milton, D. A., and A. H. Arthington. 1985. Reproductive strategy and growth of the Australian smelt, Retropinna semoni (Weber) (Pisces: Retropinnidae), and the olive perchlet, Ambassis nigripinnis (De Vis) (Pisces: Ambassidae), in Brisbane, south- eastern Queensland. Australian Journal of Marine and Freshwater Research 36:329- 341.

Mitchell, D. 1994. Floodplain wetlands of the Murray-Darling Basin: Management issues and challenges. Pages 1-5 in T. Sharley and C. Huggan, editors. Proceedings of the Floodplain Wetlands Management Workshop. Murray–Darling Basin Commission, Canberra.

Moffatt, D., and J. Voller. 2002. Fish and fish habitat of the Queensland Murray-Darling basin. Queensland Government. Department of Primary Industries. 98 pp. 240

Mookerji, N., C. Heller, H. J. Meng, H. R. Bürgi, and R. Müller. 1998. Diel and seasonal patterns of food intake and prey selection by Coregonus sp. in re-oligotrophicated Lake Lucerne, Switzerland. Journal of Fish Biology 52:443-457.

Morton, S. R., J. Short, and R. D. Barker. 1995. Refugia for biological diversity in arid and semi-arid Australia. Paper no. 4, Report to the Biodiversity Unit of the Department of Environment, Sport and Territories, Canberra, A.C.T. 171 pp.

Newbold, J. D., J. W. Elwood, R. V. O'Neill, and W. Van Winkle. 1981. Measuring nutrient spiralling in streams. Canadian Journal of Fisheries and Aquatic Sciences 38:860-863.

Nielsen, D. L., T. J. Hillman, F. J. Smith, and R. J. Shiel. 2002. The influence of seasonality and duration of flooding on zooplankton in experimental billabongs. River Research and Applications 18:227-237.

O’Connell, M. T. 2003. Direct exploitation of prey on an inundated floodplain by cherryfin shiners (Lythrurus roseipinnis) in a low order, blackwater stream. Copeia 2003:635-645.

O'Brien, W. J. 1979. The predator-prey interaction of planktivorous fish and zooplankton. American Scientist 67:572-581.

Osborne, P. L., J. H. Kyle, and M. S. Abramski. 1987. Effects of seasonal water level changes on the chemical and biological limnology of Lake Murray, Papua New Guinea. Australian Journal of Marine and Freshwater Research 38:397-408.

Parsley, J. J., D. E. Palmer, and R. W. Burkhardt. 1989. Variation in capture efficiency of a beach seine for small fishes. North American Journal of Fisheries Management 9:239-244.

Peasley, B. 1993. Macintyre River catchment. Land management proposals for the integrated treatment and prevention of land degradation. Department of Conservation and Land Management, Inverell. 130 pp.

Persson, L., and L. A. Greenberg. 1990. Juvenile competitive bottlenecks: the perch (Perca fluviatilis) - roach (Rutilus rutilus) interaction. Ecology 71:44-56.

Peterson, B. J., and B. Fry. 1987. Stable isotopes in ecosystem studies. Annual Review of Ecology and Systematics 18:293-320.

Petts, G. E. 1984. Impounded rivers: perspectives for ecological management. John Wiley and Sons, Chichester. 326 pp.

Phillips, D. L. 2001. Mixing models in analysis of diet using multiple stable isotopes: a critique. Oecologia 127:166-170.

Phillips, D. L., and J. W. Gregg. 2001. Uncertainty in source partitioning using stable isotopes. Oecologia 127:171-179.

Piet, G. J. 1998. Ecomorphology of a size-structured tropical freshwater fish community. Environmental Biology of Fishes 51:67-86. 241

Platell, M. E., and I. C. Potter. 2001. Partitioning of food resources amongst 18 abundant benthic carnivorous fish species in marine waters on the lower wets coast of Australia. Journal of Experimental Marine Biology and Ecology 261:31-54.

Poff, N. L. 1997. Landscape filters and species traits: towards mechanistic understanding and prediction in stream ecology. Journal of the North American Benthological Society 16:391-409.

Poff, N. L., and J. V. Ward. 1989. Implications of streamflow variability and predictability for lotic community structure: a regional analysis of streamflow patterns. Canadian Journal of Fisheries and Aquatic Sciences 46:1805-1818.

Poff, N. L., and J. V. Ward. 1990. Physical habitat template of lotic systems: recovery in the context of historical pattern of spatiotemporal heterogeneity. Environmental Management 14:629-645.

Polis, G. A. 1984. Age structure component of niche width and intraspecific resource partitioning: can age groups function as ecological species? American Naturalist 123:541-564.

Prescott, G. W. 1970. How to know the freshwater algae. W. C. Brown Co., Dubuque, Iowa. 348 pp.

Puckridge, J. T. 1999. The role of hydrology in the biology of dryland rivers. Pages 97-112 in R. T. Kingsford, editor. A Free-flowing River: The Ecology of the Paroo River. National Parks and Wildlife Service, Hurstville, N.S.W. 320 pp.

Puckridge, J. T., K. F. Walker. 1989. Myotic dermatitis in a freshwater gizzard shad, the bony bream, Nematalosa erebi (Gunther), in the River Murray, South Australia. Journal of Fish Diseases 12:205-221.

Puckridge, J. T., F. Sheldon, K. F. Walker, and A. J. Boulton. 1998. Flow variability and the ecology of large rivers. Marine and Freshwater Research 49:55-72.

Puckridge, J. T., and K. F. Walker. 1990. Reproductive biology and larval development of a gizzard shad, Nematalosa erebi (Günther) (Dorosomatinae: Teleostei), in the river Murray, South Australia. Australian Journal of Marine and Freshwater Research 41:695-712.

Puckridge, J. T., K. F. Walker, and J. F. Costelloe. 2000. Hydrological persistence and the ecology of dryland rivers. Regulated Rivers: Research & Management 16:385-402.

Pusey, B., M. Kennard, and A. Arthington. 2004. Freshwater Fish of North-eastern Australia. CSIRO Publishing, Melbourne, Australia. 682 pp.

Pusey, B. J., and A. H. Arthington. 2003. Importance of the riparian zone to the conservation and management of freshwater fish: a review. Marine and Freshwater Research 54:1-16.

Pusey, B. J., A. H. Arthington, and M. G. Read. 2000. The dry-season diet of freshwater fishes in monsoonal tropical rivers of Cape York Peninsula, Australia. Ecology of Freshwater Fish 9:177-190. 242

Pusey, B. J., M. G. Read, and A. H. Arthington. 1995. The feeding ecology of freshwater fishes in two rivers of the Australian wet tropics. Environmental Biology of Fishes 43:85-103.

Rai, H., and G. Hill. 1984. Primary production in the Amazonian aquatic ecosystems. Pages 311-335 in H. Sioli, editor. The Amazon: Limnology and Landscape Ecology of a Mighty Tropical River and its Basin. Dr W. Junk Publishers, Dordrecht.

Rimmer, M. A. 1985. Growth, feeding and condition of the fork-tailed catfish Arius graeffei Kner & Steindachner (Pisces: Ariidae) from the Clarence River. Australian Journal of Marine and Freshwater Research 36:33-39.

Ring, P. J., R. J. Banens, and S. J. Perrens. 1984. Changes in land use and hydrology in the Macintyre Valley. Report No WWF 2, Water for the Future Committee. 86 pp.

Roberts, J., A. Chick, L. Oswald, and P. Thompson. 1995. Effect of carp (Cyprinus carpio) L., an exotic benthivorous fish, on aquatic plants and water quality in experimental ponds. Australian Journal of Marine and Freshwater Research 46:1171-1180.

Robertson, A. I. 1997. Land-water linkages in floodplain river systems: the influence of domestic stock. Pages 207-218 in N. Klomp and I. Lunt, editors. Frontiers in ecology: building the links. Elesevier Science, Oxford. 344 pp.

Robertson, A. L., J. Lancaster, and A. G. Hildrew. 1995. Stream hydraulics and the distribution of microcrustacea: a role for refugia? Freshwater Biology 33:469-484.

Rodriguez, M. A., and W. M. Lewis, Jr. 1994. Regulation and stability in fish assemblages of neotropical floodplain lakes. Oecologia 99:166-180.

Ross, S. T., and J. A. Baker. 1983. The response of fishes to periodic spring floods in a southeastern stream. The American Midland Naturalist 109:1-14.

Ruello, N. V. 1976. Observations on some massive fish kills in Lake Eyre. Australian Journal of Marine and Freshwater Research 27:667-672.

Sainty, G. R. 1973. Aquatic plants: identification guide. Water Conservation and Irrigation Commission, New South Wales, Sydney. 111 pp.

Santos, G. M. 1982. Caracterização, habitos alimentares e reprodutivos de quatro espécies de "aracus" e considerações ecológicas sobre o grupo no lago Janauacá-AM. (Osteichthyes, Characoidei, Anostomidae). Acta Amazonica 12:713-739.

SAS Institute. 1988. SAS procedures guide: release 6.03 edition. SAS Institute, Cary, N.C. 441 pp.

Schafer, L. N., M. E. Platell, F. J. Valesini, and I. C. Potter. 2002. Comparisons between the influence of habitat type, season and body size on dietary compositions of fish species in nearshore marine waters. Journal of Experimental Marine Biology and Ecology 278:67-92.

Schimel, D. S. 1993. Theory and application of tracers. Academic Press, San Diego. 119 pp. 243

Schlosser, I. J. 1991. Stream fish ecology: a landscape perspective. BioScience 41:704- 712.

Schmitt, R. J., and S. J. Holbrook. 1984. Gape-limitation, foraging tactics and prey size selectivity of two microcarnivorous species of fish. Oecologia 63:6-12.

SCMCC. 1990. Catchment management issues in the Macintyre River catchment. State Catchment Management Coordinating Committee, New Sotu Wales. 52 pp.

Sedell, J. R., J. E. Richey, and F. J. Swanson. 1989. The river continuum concept: a basis for the expected ecosystem behaviour of very large rivers? Pages 49-55 in D. P. Dodge, editor. Proceedings of the International Large Rivers Symposium. Canadian Special Publication of Fisheries and Aquatic Sciences, 106 pp.

Sheldon, A. L., and G. K. Meffe. 1993. Multivariate analysis of feeding relationships of fishes in blackwater streams. Environmental Biology of Fishes 37:161-171.

Sheldon, F., A. J. Boulton, and J. T. Puckridge. 2002. Conservation value of variable connectivity: aquatic invertebrate assemblages of channel and floodplain habitats of a central Australian arid-zone river, Cooper Creek. Biological Conservation 103:13-31.

Sheldon, F., A. J. Boulton, and J. T. Puckridge. 2003. Variable hydrological connection structures aquatic invertebrate composition in dryland rivers: data from Cooper Creek and Diamantina River. Records of the South Australian Museum 7:119-130.

Shiel, R. J. 1990. Zooplankton. Pages 275-284 in N. Mackay and D. Eastburn, editors. The Murray. Murray-Darling Basin Commission, Canberra. 363 pp.

Shiel, R. J. 1995. A guide to identification of rotifers, cladocerans and copepods from Australian inland waters. Identification guide No. 3, Co-operative Research Centre for Freshwater Ecology, Murray-Darling Freshwater Research Centre, Albury, NSW. 144 pp.

Shiel, R. J., K. F. Walker, and W. D. Williams. 1982. Plankton of the lower River Murray, South Australia. Australian Journal of Marine and Freshwater Research 33:301- 327.

Silva, L. P., and S. M. Thomaz. 1997. Diel variation of some limnological parameters and metabolism of a lagoon of the high Parana river floodplain, MS. Pages 169-189 in J. E. Santos, editor. Anais do VIII Seminário Regional de Ecologia. UFSCar, PPG- ERN.

Sippel, S. J., S. K. Hamilton, and J. M. Melack. 1992. Inundation area and morphometry of lakes on the Amazon River floodplain, Brazil. Archiv fur Hydrobiologie 123:385- 400.

Soares, M. G. M., R. G. Almeida, and W. J. Junk. 1986. The trophic status of the fish fauna in Lago Camaleao, a macrophyte dominated floodplain lake in the middle Amazon. Amazonia 9:511-526. 244

Sokal, R. R., and F. J. Rohlf. 1969. Biometry: the principles and practice of statistics in biological research. W.H. Freeman, San Francisco. 776 pp.

Stephens, K. M., and R. M. Dowling. 2002. Wetland plants of Queensland: a field guide. CSIRO Publishing, Collingwood, Vic. 146 pp.

Swales, S., and S. J. Curran. 1995. Pindari Dam enlargement study fish population investigations. New South Wales Department of Land and Water Conservation. Cooperative Research Centre for Freshwater Ecology.

Taylor, G. 1978. A brief Cainozoic history of the upper Darling Basin. Proceedings of the Royal Society of Victoria 90:53-60.

Thomas, D. S. G. 1989. The nature of arid environments. Pages 1-10 in D. S. G. Thomas, editor. Arid zone geomorphology. Belhaven Press, London and Halsted Press, New York.

Thoms, M. C., and F. Sheldon. 2000. Lowland rivers: an Australian introduction. Regulated Rivers: Research & Management 16:375-383.

Thomson, C. 1993. The impact of river regulation on the natural flows of the Murray Darling Basin. Technical Report 92/5.3, Murray-Darling Basin Commission. 14 pp.

Thorp, J. H., and M. D. Delong. 1994. The riverine productivity model: an heuristic view of carbon sources and organic processing in large river ecosystems. Oikos 70:305- 308.

Thorp, J. H., and M. D. Delong. 2002. Dominance of autochthonous autotrophic carbon in food webs of heterotrophic rivers. Oikos 96:543-550.

Thorp, J. H., M. D. Delong, K. S. Greenwood, and A. F. Casper. 1998. Isotopic analysis of three food web theories in constricted and floodplain regions of a large river. Oecologia 117:551-563.

Timms, R. M., and B. Moss. 1984. Prevention of growth of potentially dense phytoplankton populations by zooplankton grazing, in the presence of zooplanktivorous fish in a shallow wetland ecosystem. Limnology and Oceanography 29:472-486.

Tockner, K., F. Malard, and J. V. Ward. 2000. An extension of the flood pulse concept. Hydrological Processes 14:2861-2883.

Tockner, K., and J. A. Stanford. 2002. Riverine flood plains: present state and future trends. Environmental Conservation 29:308-330.

Townsend, C. R. 1996. Concepts in river ecology: pattern and process in the catchment hierarchy. Archiv fur Hydrobiologie 113:3-21.

Usseglio-Polatera, P. 1994. Theoretical habitat templets, species traits, and species richness: aquatic insects in the Upper Rhone River and its floodplain. Freshwater Biology 31:417-437. 245

Van Valen, L. 1965. Morphological variation and width of ecological niche. American Naturalist 99:377-388.

Vander Zanden, M. J., and J. B. Rasmussen. 2001. Variation in δ15N and δ13C trophic fractionation: implications for aquatic food web studies. Limnology and Oceanography 46:2061-2066.

Vannote, R. L., G. W. Minshall, K. W. Cummins, J. R. Sedell, and C. E. Cushing. 1980. The river continuum concept. Canadian Journal of Fisheries and Aquatic Sciences 37:130-137.

Vigg, S. 1981. Species composition and relative abundance of adult fish in Pyramid Lake Nevada. Great Basin Naturalist 41:395-408.

Vyverman, W. 1994. Limnological features of lakes on the Sepik-Ramu floodplain, Papua New Guinea. Australian Journal of Marine and Freshwater Research 45:1209-1224.

Wager, R., and P. Jackson. 1993. The action plan for Australian freshwater fishes. Project Number 147, Australian Nature Conservation Agency Endangered Species Program, Canberra.

Walker, K. F. 1985. A review of ecological effects of river regulation in Australia. Hydrobiologia 125:111-129.

Walker, K. F. 1992. The River Murray, Australia: a semiarid lowland river. Pages 472-492 in P. Calow and G. E. Petts, editors. The rivers handbook: hydrological and ecological principles. Blackwell Scientific Publications, Oxford ; Melbourne. 526 pp.

Walker, K. F., F. Sheldon, and J. T. Puckridge. 1995. A perspective on dryland river ecosystems. Regulated Rivers: Research & Management 11:85-104.

Walker, K. F., and M. C. Thoms. 1993. Environmental effects of flow regulation on the lower Murray River, Australia. Regulated Rivers: Research & Management 8:103- 119.

Ward, J. V., and J. A. Stanford. 1983. The serial discontinuity concept of lotic ecosystems. Pages 29-42 in T. D. Fontaine, III and S. M. Bartell, editors. Dynamics of lotic ecosystems. Ann Arbor Science, Ann Arbor, Mich. 494 pp.

Ward, J. V., and J. A. Stanford. 1995. The serial discontinuity concept: extending the model to floodplain rivers. Regulated Rivers: Research & Management 10:159- 168.

Watts, R. J. 1999. Biodiversity in the Paroo river and its wetlands. Pages 13-22 in R. T. Kingsford, editor. A Free-flowing River: The Ecology of the Paroo River. National Parks and Wildlife Service, Hurstville, N.S.W. 320 pp.

Welcomme, R. L. 1979. Fisheries ecology of floodplain rivers. Longman, London; New York. 317 pp.

Welcomme, R. L. 1985. River Fisheries. FAO Fish. Tech. Pap. 262. 330 pp. 246

Werner, E. E. 1974. The fish size, prey size, handling time relation in several sunfishes and some implications. Journal of the Fisheries Research Board of Canada 31:1531- 1536.

Werner, E. E., and J. F. Gilliam. 1984. The ontogenetic niche and species interactions in size-structured populations. Annual Review of Ecology and Systematics 15:393- 425.

Werner, E. E., and D. J. Hall. 1977. Competition and habitat shift in two sunfishes (Centrarchidae). Ecology 58:869-876.

Wilbur, H. M. 1980. Complex life cycles. Annual Review of Ecology and Systematics 11:67-93.

Williams, W. D. 1980. Australian freshwater life: the invertebrates of Australian inland waters, 2nd ed. edition. Macmillan, Melbourne. 321 pp.

Williams, W. D. 1985. Biotic adaptations in temporary lentic waters, with special reference to those in semi-arid and arid regions. Hydrobiologia 125:85-110.

Williams, W. D., and G. R. Allen. 1987. Origins and adaptations of the fauna of inland waters. Pages 184-201 in G. R. Dyne and D. W. Walton, editors. Fauna of Australia. General Articles. Australian Government Publishing Service, Canberra. 339 pp.

Winemiller, K. O. 1989. Ontogenetic diet shifts and resource partitioning among piscivorous fishes in the Venezuelan ilanos. Environmental Biology of Fishes 26:177-199.

Winemiller, K. O. 1990. Spatial and temporal variation in tropical fish trophic networks. Ecological Monographs 60:331-367.

Winemiller, K. O. 1996. Factors driving temporal and spatial variation in aquatic floodplain food webs. Pages 298-312 in G. A. Polis and K. O. Winemiller, editors. Food webs: integration of patterns & dynamics. Chapman & Hall, New York. 472 pp.

Winemiller, K. O., and D. B. Jepsen. 1998. Effects of seasonality and fish movement on tropical river food webs. Journal of Fish Biology 53:267-296.

Winterbottom, J. H., S. E. Orton, A. G. Hildrew, and J. Lancaster. 1997. Field experiments on flow refugia in streams. Freshwater Biology 37:569-580.

Wissmar, R. C., J. E. Richey, R. F. Stallard, and J. M. Edmond. 1981. Plankton metabolism and carbon processes in the Amazon River, its tributaries, and floodplain waters, Peru-Brazil, May-June 1977. Ecology 62:1622-1633.

Wolf, J., N. C. Johnson, D. L. Rowland, and P. B. Reich. 2003. Elevated CO2 and plant species richness impact arbuscular mycorrhizal fungal spore communities. New Phytologist 157:579-588. 247

Woodland, D. J., and P. J. Ward. 1992. Fish communities in sandy pools of Magela Creek, Alligator Rivers region. Research report 9, Australian Government Publishing Service, Canberra. 81 pp.

Wootton, R. J. 1990. Ecology of Teleost Fishes. Chapman & Hall, London. 404 pp.

Worobec, M. N. 1984. Field estimates of the daily ration of winter flounder, Pseudopleuronectes americanus (Walbaum), in a southern New England salt pond. Journal of Experimental Marine Biology and Ecology 77:183-196.

Wylie, P. 1995. Land use in the Border Rivers catchment. Implications for water quality. Border Rivers Catchment Coordinating Committee, Dalby.

Wylie, P., and I. Greggery. 1995. Pesticide audit Border Rivers catchment - Implications for water quality. A report prepared for the Border Rivers Catchment Coordinating Committee.

Yodzis, P. 1993. Environmental and trophodiversity. Pages 26-38 in R. E. Ricklefs and D. Schluter, editors. Species Diversity in Ecological Communities. Historical and geographical perspectives. The University of Chicago Press, Chicago. 416 pp.

Young, W. J. 1999. Hydrologic descriptions of semi-arid rivers: an ecological perspective. Pages 77-96 in R. T. Kingsford, editor. A Free-flowing River: The Ecology of the Paroo River. National Parks and Wildlife Service, Hurstville, N.S.W. 320 pp.

Zaret, T. M., and A. S. Rand. 1971. Competition in tropical stream fishes: support for the competitive exclusion principle. Ecology 52:336-342.

Zimmerman, G. M., H. Goetz, and P. W. Mielke, Jr. 1985. Use of an improved statistical method for group comparisons to study effects of prairie fire. Ecology 66:606-611.