Copyright is owned by the Author of the thesis.  Permission is given for 
a copy to be downloaded by an individual for the purpose of research and 
private study only.  The thesis may not be reproduced elsewhere without 
the permission of the Author. 
 



 

 

 

Multiple scales of biological variability  

in New Zealand streams 
 

 

 

 

 

 

 

A thesis presented in partial fulfilment of the requirements  

for the degree of 

 

Doctor of Philosophy  

in Ecology 

 

at Massey University, Manawatū, New Zealand. 

 

 

 

 

 

 

Withanage Thushantha Sriyan Jayasuriya 
  

2016 
 

 



ii 

 

 

General Abstract 

Stream fish communities in Taranaki, New Zealand, were studied for the patterns 

and drivers of their spatial ecology. The study was focused on three main themes: a) 

complementarity between geography and landuse in driving regional distribution patterns 

of stream fish, b) the impact of agriculture on community composition, structure and 

variability of fish and invertebrates, and c) concordance among environmental distance 

and community dissimilarities of stream fish and invertebrates. 

Stream sampling and data collection for fish was conducted at regional scale using 

96 sites distributed in the protected forest (44 sites) of Egmont National Park in Taranaki, 

and in surrounding farmlands (52 sites). Local scale sampling for fish and invertebrates 

was carried out at 15 stream sites in pasture (8 sites) and in adjacent forest (7 sites). 

Environmental data of geography, landuse and local habitat description were also gathered 

concurrently to biological sampling. The regional scale survey reported fifteen fish 

species, dominated by longfin eels (Anguilla dieffenbachia), redfin bullies (Gobiomorphus 

huttoni) and koaro (Galaxias brevipinnis), while 12 fish species and 69 different 

invertebrate taxa were recorded from the 15 sites at local scale.   

 Regional scale spatial patterns of fish were mainly driven by landuse pattern. 

Catchment landuse (characterised by percentage cover of farming/native forest) effectively 

partitioned the stream fish community structure in Taranaki. Within each level of 

catchment landuse (farming), abundance and richness of fish species were negatively 

correlated with the altitude.  Moreover, the upstream slope in high elevations and intensive 

farming downstream limited the distribution of stream fish across the region. 



iii 

 

 

  Fish community composition differed significantly but weakly between forest and 

pasture in the immediate proximity. The dissimilarity of fish communities between forest 

and pasture increased from regional to local scale, and a similar result was found with 

stream invertebrate dissimilarity at the local scale. Stream communities (fish and 

invertebrates) were equally variable among streams between the two land use classes both 

at regional and local scales. Although the land use difference did not affect within-stream 

variability of fish, invertebrate communities were less variable within a pasture stream. 

Trends in in-stream variability of invertebrates were influenced mainly by altitude, stream 

morphology, pH, and riparian native cover.   

In concordance analysis, Mantel and Procrustes tests were used to compare 

community matrices of fish and invertebrates and the environmental distance between 

stream sites. The spatial patterns of fish and invertebrates were significantly concordant 

with each other among the 15 streams at the local scale. Nevertheless, community 

concordance decreased with lower spatial scales, and the two communities were not 

concordant at local sites within a given stream. Agriculture had a negative impact on the 

concordance between fish and invertebrates among streams, and none of the communities 

correlated with the overall environmental distance between agricultural streams. 

Community concordance between fish and invertebrates was consistently higher than the 

community-environment links, and lower trophic level (invertebrates) linked to their 

environment more closely than the upper trophic level (fish). The overall results suggest a 

bottom-up control of the communities through the stream food web.  



iv 

 

 

Finally, to inform the regional management and conservation decision, stream sites 

were partitioned according to the most important bioenvironmental constraints. The 

ecological similarity was measured by geography, land use pattern and the abundances of 

influential native fish species within the region, and the streams were clustered into seven 

distinct zones, using the method of affinity propagation. Interestingly, the dichotomy in 

proximal land use was not generally represented between zones, and the species diversity 

gradients were not significantly different across the zonal stream clusters. The average 

elevation of a given zone did not influence the community variability, while upstream 

pasture significantly homogenised fish communities between streams within a zone. 

Nonetheless the zones were based on river-system connectivity and geographical 

proximity.    

This study showed separate effects of confounding geography (altitude) and 

landuse on stream fish community structure, which has not explicitly been explored by 

previous studies. Studies with a simultaneous focus on multiple biological (e.g. fish and 

invertebrates) and environmental (e.g. geography, landuse, stream morphology) scales in 

varying spatial scales are not common in freshwater ecology. Therefore, this study has a 

great contribution to the understanding of the spatial ecology of stream communities 

linked with the control of geography, landuse, environment and likely biological 

interactions between fish and invertebrates.         



v 

 

 

Preface 

 

This thesis is based on a research designed to investigate the environmental and 

biological drivers of freshwater community composition, structure and variability, in New 

Zealand streams. Taranaki streams were selected, because of the rich species diversity of 

freshwater fish and invertebrates, reported in previous studies. A special attention was 

paid to separate the effect of geography and land use, which was not explicitly covered by 

the other studies previously conducted in Taranaki. The first part of the study explores the 

important environmental drivers of the fish community in Taranaki, surveyed in a wider 

geographical extent (96 streams), compared to the study area of 15 streams, in the second 

phase of this research.  

Most of the studies on biological variability cover large geographical areas from 

ecosystems to ecoregions. Particularly in New Zealand, previous studies have not mainly 

addressed the inter-site variability change across land-uses and the community 

concordance between fish and invertebrates, within a small geographical extent. 

Therefore, I attempted to address the knowledge gap in biological variability and 

community concordance of stream communities at the local scale, with a special concern 

about the human impacts to stream fish and invertebrates.  

This thesis includes three individual research manuscripts, thus some repetition 

occurs in the introductions, methods and discussions across the chapters.        

 



vi 

 

 

Acknowledgments 

I greatly appreciate my primary supervisor, Dr. Ian Henderson, for his guidance, 

teaching, and encouragement in completing this study. It was relay challenging to remain 

in a doctoral research with a limited previous exposure to the fields of freshwater ecology 

and ecological statistics. Without is excellent supervision, this project would not have 

covered the intellectual discussion provided throughout the thesis. Further, I deeply 

acknowledge his understanding and patience in teaching.  

Many thanks to Prof. Murray Potter for his supervision and constructive 

suggestions, to improve the output quality of this thesis. He was more than helpful during 

the last few years, especially in downturns of my studentship at Massey. His leadership, 

guidence and support were enormous in the completion of this study.  

Special thanks, to Prof. Russell Death for his initiatives and supervision of this 

project, teachings in aquatic ecology and support throughout the study period. Further, I 

am grateful to Dr. Mike Joy for all the inspirations and for sharing his knowledge, ideas 

and experiences, which were helpful in understanding the ecology of New Zealand 

freshwater fish.  

The academic staff of Ecology Group was truly supportive in succeeding this 

research. Prof. Steven Trewick, Associate Prof. Mary Morgan-Richards, Dr. Gillian 

Rapson, Associate Prof. Phil Battley, Dr. Isabel Castro are among the motivating scientists 

who continuously encouraged to be successful in this Ph.D.  



vii 

 

 

 I should extend my deepest gratitude for Prof. Marlena Kruger, the Dean, 

Graduate Research School, for the immense encouragement and guidance given to achieve 

my academic goals at Massey. I am also thankful to Prof. Giselle Byrnes, Assistant Vice-

Chancellor Research, Academic & Enterprise for her support in completion of this study. 

Moreover, the suggestions, motivation and support extended by Prof. Peter Kemp, Head of 

the Institute of Agriculture & Environment are also highly appreciated. Continuous 

support of Dr. Julia Rayner at Graduate Research School was more than helpful in 

improving my studies. Many thanks for the administrative support from Trevor Weir at 

Office of the AVC, Research, Academic and Enterprise.  

International Student Support Team, especially Sylvia Hooker is deeply 

acknowledged for her genuine backing in managing the student life at Massey. 

International office, Library Services, IT Services and Centre for Teaching and Learning 

(Dr. Catherine Stevens) of Massey University are highly valued, because of their support 

to overcome the difficulties in completing this Ph.D. Thanks to Dr. Rajasheker Pullanagari 

at IAE of Massey University for his support in GIS applications and statistics. Many 

thanks to the Scholarship Selection Committee and the Massey University Foundation for 

granting me J.K Skipworth and Doctoral Hardship scholarships and the Baily Bequest 

Bursary.  

  I am further grateful to Rosemary Miller, Christopher Rendall and Natasa Petrova 

of the Department of Conservation in Taranaki for the support given by granting research 

permits and sharing their knowledge about freshwaters in Taranaki. Further, landowners 



viii 

 

 

and people in Taranaki are highly appreciated for their friendly support in long days of 

fieldwork.  

I should thankfully mention Joel Rademaker who always supported my academic 

life in many ways, including fieldwork assistance, proofreading of this thesis, extending 

his brotherly hand especially  in difficult circumstances. 

The support, friendship and inspirations given by Matthew Krna are highly 

appreciated in completion of this study. 

    Acknowledgements to my colleagues, Matthew, Prasad, Rashmi, Lizzy, 

Andrew, Briar, Adam, Josh, Stella, Ishani, Tim, Emily, Charlotte, Alice, Santhi and all the 

office mates at post-grad room 1.41, for their support, encouragement to continue my 

Ph.D. during the last few years. Moreover, I am thankful to Yasalal Wjeweera and Mallika 

Fernando for their funding support to start this doctoral study. 

 Winning this academic goal was extremely cherished by the loving guidance and 

valuable support of my parents and family.   

Finally, I dedicate this thesis to Nipuna Peiris and Shani Fernando, not only for 

their assistance in field work and software handling, but also for all the brotherly support 

and financial backup that strengthened me, from very early stages to the last moment of 

my Ph.D. 

 

 



ix 

 

 

 

Table of Contents Page 

General Abstract ii 

Preface v 

Acknowledgements  vi 

Table of Contents ix 

List of Figures x 

List of Tables xiii 

  

Chapter One: General Introduction  

The effect of land use on the spatial ecology of New Zealand stream communities 1 

Chapter Two  

Spatial patterns of stream fish communities in Taranaki, New Zealand 12 

Chapter Three  

Does land use have an effect on the variability of stream communities? 47 

Chapter Four  

Community concordance of freshwater fish and invertebrates in Taranaki streams 83 

Chapter Five  

Is ecological dissimilarity of fish between streams independent from the proximal land 
use? 107 

Chapter Six: Concluding Remarks 134 

  

Appendices  

Appendix I 144 

Appendix II 147 

Appendix III 149 

Appendix IV 152 



x 

 

 

List of Figures Page 

Fig. 2.1 Stream sampling sites in the Taranaki region of New Zealand used to 
collect fish and habitat data during summer 2012. 18 

Fig 2.2   NMDS ordination on fish presence/absence in 96 Taranaki streams 
fitted with elevation contours (green). Fish species occurring in > 5% of sites 
excluded 

26 

Fig. 2.3 Occurrence of the fish species reported from >  5% of 96 sites, fitted 
with the gradient in upstream average slope (red contours), on Sǿrenson 
similarity ordination between the study sites in Taranaki  27 

Fig. 2.4 Elevation (red) and northing (green) contours fitted onto the surface of 
NMDS ordination of Taranaki sites partitioned on proximal land use pattern 28 

Fig. 2.5 Boxplot diagrams showing the results of Analysis of Similarity 
(ANOSIM) on fish data collected from Taranaki streams, in 2012.  Within and 
between group differences were compared using average ranked Bray-Curtis 
similarity, partitioned between forest (n = 44) and pasture (n = 52) 29 

Fig 2.6 Overall fish abundance of Taranaki streams, constrained by the native 
riparian cover (green), and fitted with elevation contours (purple) and average 
upstream slope  

33 

Fig. 2.7 Fish abundances in 96 Taranaki, streams constrained by catchment 
farming (A) and native forest cover (B), fitted into altitude (red in (A) & purple 
in (B)) contours and upstream slope 

35 

Fig 2.8 Overall fish abundance of Taranaki streams, constrained by nitrogen 
concentration (ppt), and fitted with elevation contours (purple) and riparian cover  36 

Fig 3.1 Fifteen streams of Taranaki, New Zealand, selected for sampling fish and 
invertebrates in the summer of 2013 52 

Fig. 3.2 Within and between group rank dissimilarities for fish and invertebrate 
communities in 15 Taranaki streams, in the forest (n = 7) or pasture (n = 8) land 
uses 

58 

Fig. 3.3 Difference in in-stream variability between fish and invertebrate 
communities of 15 Taranaki streams, surveyed in 2013 62 

Fig. 3.4 Difference in average multivariate distance to the group centroids (in-
stream variability) for invertebrate communities in 15 Taranaki streams, 
partitioned between contrasting land uses 

62 



xi 

 

 

Fig. 3.5 Boxplot diagrams showing the differences of average values in selected 
environmental factors between forested (n=7) and pasture (n=8) habitats of 
Taranaki 

64 

Fig. 3.6 In-stream variability of invertebrates vs. distance from the forest margin 
(A) and altitude (B), across 15 Taranaki streams, surveyed in 2013 65 

Fig. 3.7 In-stream variability of invertebrates vs. slope (A) and width / depth 
ratio (B), across 15 Taranaki streams, surveyed in 2013 66 

Fig. 3.8 In-stream variability of invertebrates vs. pH value, across 15 Taranaki 
streams, surveyed in 2013   67 

Fig. 3.9 In-stream variability of invertebrates vs. riparian native vegetation 
within 100m, across 15 Taranaki streams, surveyed in 2013   68 

Fig. 3.10 In-stream variability of invertebrates vs. % of bryophytes (moss) in 
streams (A) and % of undercut banks (B), across 15 Taranaki streams, surveyed 
in 2013 

69 

Fig. 4.1 Selected streams in Taranaki, New Zealand, surveyed for fish and 
invertebrate data, during the summer in 2013  87 

Fig. 4.2 Relationship of biological similarities of fish and invertebrates among 15 
streams of Taranaki, surveyed in 2013 92 

Fig. 4.3 Mantel correlation (r) between fish and invertebrate communities across 
different spatial units of Taranaki streams, studied in 2013 92 

Fig. 4.4 Distance comparisons (A & B) and Procrustes superimposition plots 
between fish and invertebrates among forested (n = 7) and agricultural (n = 8) 
streams in Taranaki 

94 

Fig. 4.5 Correlation values of Mantel and Procrustes comparisons between fish 
and invertebrates among forested (n = 7) and agricultural (n = 8) streams in 
Taranaki 

95 

Fig. 4.6 Correlation values of Mantel comparisons between the biological 
distance (Bray-Curtis) and environmental distance (Euclidean) communities , 
with respect to fish and invertebrate communities among  15 Taranaki streams 
and within  the subset of  forested streams (n = 7) 96 

Fig. 5.1 Stream sampling sites of the Taranaki region in New Zealand used to 
collect fish and habitat data during summer 2012 

110 

Fig. 5.2 Classification of bioenvironmental filters used for clustering 96 streams 
considered in this study 

116 



xii 

 

 

Fig. 5.3 Ordination diagram for the first two axes of Constrained 
Correspondence Analysis (CCA) of fish abundances in 96 Taranaki streams, 
fitted with selected bioenvironmental filters.  117 

Fig. 5.4 Two-dimensional similarity ordinations of 96 Taranaki sites (based on 
Euclidean distances between selected vectors) used in affinity propagation 
method to distinguish site clusters   

118 

Fig. 5.5 Heatmap of 96 Taranaki sites, represented by colour coded similarities 
in selected bioenvironmental filters 

119 

Fig. 5.6 Map of  Taranaki streams showing different zones of site clusters 
referring to neighbouring major catchments  

120 

Fig. 5.7 Map of study sites, showing zonal diversity (γ), local diversity (α), 
similarity and variability (β) of Taranaki streams partitioned into seven clusters 
(zones), using affinity propagation method  

122 

Fig. 5.8 Within zone relative abundances of the most abundant fish species in 
each cluster zone of 96 Taranaki streams considered in this study 

123 

Fig. 5.9 Regression plot between biological variability (mean distance to the 
group centroid) and upstream pasture across the seven clusters of streams 
analysed in this study 

125 

Fig. 5.10 Two-dimensional ordinations (Euclidean distance) between the seven 
zones (stress = 0.01), superimposed with within group relative abundances of 
selected fish species in 96 Taranaki streams (sizes of bubbles are proportionate to 
relative abundances shown in the legends) 126 

 

 

 

 

 

 

 



xiii 

 

 

 

List of Tables Page 

Table 2.1:  Frequency of occurrence and relative abundance of   freshwater fish 
species reported from the 96 streams and rivers from Taranaki during summer 
2012  

24 

Table 2.2: Selected fish species important in their abundance to two-dimensional 
NMDS ordination (constructed on Bray-Curtis similarity) of   freshwater fish 
taxa reported from the 96 streams and rivers from Taranaki during summer 2012  24 

Table 2.3: Important environmental vectors of NMDS ordination based on fish 
abundance data and variance inflation factor (VIF) of each vector assessed on 
constrained community ordination (CCA)    

25 

Table 2.4: Results of the Analysis of Similarity (ANOSIM) of Taranaki stream 
fish data grouped on the change in proximal land use pattern between forest and 
pasture 29 

Table 2.5:  Differences in diversity measures of fish between forest (n=44) and 
pasture (n=52) sites in Taranaki 30 

Table 2.6: Ordered contribution by the top six fish species to the Bray-Curtis 
dissimilarity in species abundances (SIMPER test), between the forest (n=44) 
and pasture (n=52), in 96 Taranaki streams 31 

Table 2.7: Axis correlations and variance inflation factor values (VIF) of 
important environmental vectors in constrained correspondence analysis (CCA 
model) of overall fish abundance of 96 Taranaki streams  

32 

Table 2.8: Selected fish species in Taranaki streams partitioned by the native 
riparian width and altitude, according to species abundances correlated in the 
riparian partial model 

33 

Table 2.9: Selected fish species in Taranaki streams partitioned by catchment 
land use and altitude, according to species abundances correlated in the riparian 
partial model 

34 

Table 2.10: Selected fish species in Taranaki streams partitioned by nitrogen 
concentration and altitude, according to species abundances correlated in the 
riparian partial model 

37 

Table 3.1: Relative abundances and frequencies of occurrence of fish species 
reported from 15 Taranaki streams, in 2013 56 



xiv 

 

 

Table 3.2: Species diversity, relative abundances and frequencies of occurrence 
of invertebrate taxa reported from 15 Taranaki streams, in 2013 57 

Table 3.3:Species diversity and in-stream variability (measured in average 
distance to the group centroid in community ordinations) of fish and 
invertebrates in 15 Taranaki streams  57 

Table 3.4: Ordered contribution by the fish species to the Bray-Curtis 
dissimilarity in species abundances (SIMPER test), between the forest (n=7) and 
pasture (n=8), in 15 Taranaki streams 

59 

Table 3.5: Ordered contribution by the top ten invertebrate taxa to the Bray-
Curtis dissimilarity in species abundances (SIMPER test), between the forest 
(n=7) and pasture (n=8), in 15 Taranaki streams 

59 

Table 3.6: Percentage contribution by key taxonomic groups of fish and 
invertebrate to overall the Bray-Curtis dissimilarity in species abundances 
(SIMPER test), differed between the forest (n=7) and pasture (n=8), in 15 
Taranaki streams  60 

Table 3.7: Within group multivariate dispersion (inter-site variability) of fish 
and invertebrate communities between Taranaki streams, (grouped as pasture 
(n=8) vs. forest (n=7)), sampled in January- February 2013, by electrofishing and 

2

61 

Table 3.8 Within stream multivariate dispersion (in-stream variability) of fish 
and invertebrate communities between Taranaki streams, (grouped as pasture 
(n=8) vs. forest (n=7)) 61 

Table 3.9: Differences in average values of selected environmental factors 
between forested (n=7) and pasture (n=8) habitats of Taranaki 63 

Table 3.10: Results of regression analyses between community (in-stream) 
variability and selected environmental factors (significant values (P < 0.05) are 
in bold letters) 

70 

Table 4.1: Environmental factors grouped into relevant categories for measuring 
the Euclidean distance between 15 streams considered in this study      90 

Table 4.2: Results of Mantel and Procrustes tests between fish and invertebrates 
of Taranaki streams grouped into contrasting land use classes (forest (n=7), 
pasture (n=8)) 

93 

Table 4.3: Mantel Correlations between the biological distance (Bray-Curtis) 
and environmental distance (Euclidean) of fish and invertebrate communities 
among 15 Taranaki streams, within the subsets of forested (n = 7) and 
agricultural (n = 8) streams 96 



xv 

 

 

Table  4.4: Mantel correlations between the biological distance (Bray-Curtis) 
and  categorical environmental distances (Euclidean) communities , relating to 
fish and invertebrate communities among  15 Taranaki streams  and within  
forested (n = 7)  streams 

97 

Table 5.1: Fitted correlation (CCA) and co-linearity (VIF) values of selected 
environmental and biological filters of the fish community structure between 96 
streams in Taranaki 

115 

Table 5.2: Major catchments represented different zones of site clusters in 
Taranaki 120 

Table 5.3: Zonal diversity (γ), local diversity (α), similarity and variability (β) of 
Taranaki streams partitioned into 7 clusters (zones), using affinity propagation 
method  

121 

Table 5.4: Explainable variations of the trends (r2) between freshwater fish 
diversity measures across seven clusters of 96 sites in Taranaki 123 

Table 5.5: Global R-values of ANOSIM test for the pairwise comparisons of the 
bioenvironmental similarity, between the seven zones of Taranaki streams 124 

 



1 

 

Chapter One 

The effect of land use on the spatial ecology of New Zealand stream 

communities 

 

Biological communities and environmental changes 

 The global environment has been drastically changed over the last three centuries 

(Goldewijk & Ramankutty, 2004; Ramankutty & Foley, 1999; Turner, 1990). Land use 

change (e.g. forest clearance), over-exploitation of species (e.g. animal farming, commercial 

forestry and floriculture), invasive species, climate change (increase in global temperature 

and precipitation) and high nutrient applications in agriculture intensification are among key 

direct anthropogenic drivers of the contemporary status of natural ecosystems worldwide 

(McDowall, 1990; Morris, 2010; Nelson et al., 2006; Turner, 1990). Human impacts have 

become increasingly critical determinants of biological diversity, within the framework of 

biogeographical history, spatial heterogeneity (e.g. topography), and temporal changes such 

as seasonal variations in precipitation and temperature (Chapin III et al., 2000; Chapman & 

Reiss, 1998; 2010; Olden, 2006). Biotic communities fluctuate in response to their changing 

environment, and the particular responses may be informative of the level of environmental 

disturbances (Conrad, 1977).   

 Biomonitoring has become a popular tool among resource managers for assessing 

ecological health and sustainability of natural resources in response to anthropogenic 

activities such as farming, urbanization and introduced exotic species (Allan, 2004; Death & 

Joy, 2004; Joy & Death, 2002).  It may be useful to identify the most sensitive communities 

to a particular human impact, in order to focus mitigation strategies. For instance, stream 

fauna such as fish and invertebrates are widely used to monitor the impacts of land 



2 

 

conversions, particularly for agricultural developments. Hence, in countries with economies 

largely based on agriculture, such as New Zealand, fluvial habitats are extensively studied to 

compare the stream communities between upstream natural forests and downstream pasture, 

to find out the effects of farming on the natural environment and essentially the biological 

diversity (Cowie, 1980; Death & Joy, 2004; McDowall, 2010). 

 

What is β-diversity? 

  Biological diversity encompasses the diversity of genes, species, and communities, 

within life-systems, ranging from a single organism to complex ecosystems or ecoregions 

(Magurran, 2002). Conventional biological diversity measures (e.g. species richness) have a 

limited capability to capture variability within and between ecosystems (Leprieur et al., 2011; 

Munoz et al., 2008; Soininen et al., 2007). Community ecologists have therefore focused on 

diversity measures that cover greater scales to help with this (Magurran, 1988). The concept 

of β diversity was introduced by Whittaker (Magurran, 2004; Whittaker, 1960), and has 

developed extensively in community ecology  (Anderson et al., 2011; Koleff et al., 2003; 

Soininen et al., 2007). β diversity measures the change of diversity between sampling units, 

across  time and/or space, and it  has a great utility in exploring community patterns, related 

to the functionality of an ecosystem (Magurran, 2002).  

Since the introduction of Whittaker’s differentiation index in 1960, around 24 

different measures have been described to measure β diversity for presence/absence data, and 

several quantitative coefficients  such as the Bray-Curtis index (Sǿrenson quantitative) have 

also been introduced to measure β diversity with density data in biological communities 

(Koleff et al., 2003; Oksanen, 2012). All of these β diversity measures fall into three basic 

categories: 1). Measures of differentiation (difference in α diversity (e.g. number of species) 



3 

 

between two plots), 2). Measures of complementarity (similarity / dissimilarity) and 3). 

Measures of species-area relationship or average species turnover per area (Magurran, 2002).  

 

New Zealand stream communities and their determinants  

 Fish and benthic macroinvertebrates are  the most extensively studied organisms  in 

New Zealand lotic habitats (Collier & Winterbourn, 2000; McDowall, 2010).  New Zealand 

freshwater fish communities comprise a high proportion of migratory (amphidromous and 

catadromous) species, 21 exotic species and seven marine species (McIntosh & McDowall, 

2004). Generally, nocturnal and benthic native fish fauna include representatives in seven 

families: Geotridae, Anguillidae, Retropinnidae, Galaxiidae, Pinuipedidae, Gobiidae, and 

Pleuronectidae, with the majority being galaxiids (Galaxias spp.) or bullies (Gobiomorphus 

spp.). Almost the entire fish community feeds on freshwater macroinvertebrates, while larger 

species such as eels and trout prey on other fish and semi-aquatic birds or mammals as well 

as invertebrates (McDowall, 2000).  

  Apart from being a major component of the diet of fish, freshwater 

macroinvertebrates play an important role in many key ecosystem functions (e.g. breaking 

down allochthonous organic matter and transferring photosynthetic energy to higher trophic 

levels) of New Zealand running waters (Closs et al., 2009; Collier & Winterbourn, 2000). 

Freshwater macroinvertebrates are dominated by insects but include other taxonomic groups 

ranging from Porifera to Mollusca (Collier & Winterbourn, 2000). New Zealand freshwater 

insects belong to 58 families in nine orders; Odonata, Ephemeroptera, Plecoptera, 

Megaloptera, Mecoptera, Hemiptera, Trichoptera, Coleoptera and Diptera (Winterbourn et 

al., 1989). The ecological role of freshwater insects is mainly related to their functional 

feeding groups: collector-gatherers, browsers, scrapers, shredders, filter feeders, predators 

and piercer-suckers (Cowie, 1980; Cowley, 1978; Cummins, 1973; Winterbourn, 2000). 



4 

 

Insects disperse between freshwater habitats by flying, drifting,  moving between substrates 

or banks by walking or swimming (Delucchi, 1989).   

 New Zealand freshwater communities are affected by several environmental factors, 

including altitude, distance inland, land use, and migratory barriers (Collier & Winterbourn, 

2000; Jowett & Richardson, 2003; McDowall, 1990). However, the impact of  each factor  

varies between habitats, regions and at the national scale (McIntosh & McDowall, 2004). For 

instance, predator-prey interactions are important habitat scale drivers, while geographical 

factors such as latitude play a major role in regional or national scale community structure 

and composition of freshwater communities (Geist, 2011; McIntosh & McDowall, 2004).  

Conversion of forest to pastoral land results in increased deposited sediments, nutrient 

enrichment, removal of riparian vegetation and declines in water quality (e.g. high primary 

productivity, deoxygenation, and ammonia toxicity), and may negatively influence the 

ecological balance of stream communities (Quinn, 2000).  

 

Importance of studying variability in New Zealand stream communities  

 Multivariate community assessments are commonly used to explore the 

environmental determinants of New Zealand freshwater community structure and 

composition (Death & Winterbourn, 1994; Jowett & Richardson, 2003; Leathwick et al., 

2005). Although several studies have looked at the β diversity patterns of freshwater fish and 

invertebrates, particularly along the geographical extent of the country, there has been no 

investigation of how community variability is partitioned between land uses (e.g. forest vs. 

pasture) (Astorga et al., 2014; McDowall, 2010). For example, are the fish faunas more or 

less variable between streams (β diversity) in pasture than in native forest? In addition, 

landscape ecology of New Zealand stream communities is yet to be explained in terms of 

links between the physical environment and trophic levels of the stream food web. Some of 



5 

 

the aspects that have not been explicitly addressed by previous exploratory studies include 

community concordance, factors of spatial stratification, relative importance of geography, 

and land use for freshwater communities. Further, the conventional studies are heavily 

weighted towards assessing the land use impacts and/or national scale biogeography of New 

Zealand freshwater communities (Collier & Winterbourn, 2000; McDowall, 2010; 

Winterbourn, 1991). In-depth investigations on the spatial ecology of regional stream 

communities would provide scientific insights for interest groups such as resource managers, 

conservationists, and local decision makers. Hence, this study would potentially contribute to 

the future sustainability of freshwater ecosystems.   

 

Study design and the research goals 

 This research program was designed in three spatial levels within Taranaki: a) 

regional, b) local and c) in-stream, to explore the relative effects of bioenvironmental factors 

and cross-community links for fish and invertebrates between specific spatial levels. Stream 

fish and invertebrates were selected because of their: a) popularity/applicability in ecological 

health assessments (Joy & Death, 2002; Lewis et al., 2007), b) high contribution to New 

Zealand stream food webs (Winterbourn, 1991), and c) biological diversity value in 

conservation of New Zealand freshwater ecosystems (Geist, 2011). Further, the analyses of 

this study are based on β diversity of the particular communities, to enhance the predictability 

of community models alongside multiple bioenvironmental scales. Further, the particular  β 

diversity-based statistical analyses (described in the method sections of this thesis) were more 

effective and informative in achieving the key research goals of this study. For instance, 

partitioning the community patterns against land use/ geographical gradients , comparing the 

community ordinations between fish and invertebrates would have not been straightforward 

with conventional α diversity measures such as species richness (Magurran, 1988). 



6 

 

Agriculture was considered as the predominant human impact indicator of the Taranaki 

streams, and the study attempted to gauge the relative importance between land use and 

elevation as portioning factors (Joy & Death, 2001; Townsend, 1996). The key research goals 

of the research project include: 

 Investigating the relative importance of geography and land use control of the stream 

communities in Taranaki 

 Comparison of the effect of farming for community similarity and variability 

 Understanding the likelihood of top-down or bottom-up control of the stream 

communities through trophic levels 

 Measuring the differences in community-environment links between stream fish and 

invertebrates 

 Comparing the importance in community concordance for stream community 

structure  between different spatial levels and land use classes 

  To suggest a pragmatic conservation/management approach for the fish community 

in Taranaki.  

  

References  

Allan, J. D. (2004). Landscapes and riverscapes: the influence of land use on stream 

ecosystems. Annual Review of Ecology, Evolution, and Systematics, 35, 257-284.  

Anderson, M. J., Crist, T. O., Chase, J. M., Vellend, M., Inouye, B. D., Freestone, A. L., 

Sanders, N. J., Cornell, H. V., Comita, L. S., & Davies, K. F. (2011). Navigating the 

multiple meanings of β diversity: a roadmap for the practicing ecologist. Ecology 

Letters, 14(1), 19-28.  



7 

 

Astorga, A., Death, R., Death, F., Paavola, R., Chakraborty, M., & Muotka, T. (2014). 

Habitat heterogeneity drives the geographical distribution of beta diversity: the case 

of New Zealand stream invertebrates. Ecology and Evolution, 4(13), 2693-2702.  

Chapin III, F. S., Zavaleta, E. S., Eviner, V. T., Naylor, R. L., Vitousek, P. M., Reynolds, H. 

L., Hooper, D. U., Lavorel, S., Sala, O. E., & Hobbie, S. E. (2000). Consequences of 

changing biodiversity. Nature, 405(6783), 234-242.  

Chapman, J. L., & Reiss, M. J. (1998). Ecology: Principles and Applications: Cambridge 

University Press, Cambridge, UK. 

Closs, G., Downes, B., & Boulton, A. (2004). Freshwater Ecology: a Scientific Introduction: 

Blackwell Science Ltd, Oxford, UK . 

Collier, K. J., & Winterbourn, M. J. (2000). New Zealand Stream Invertebrates: Ecology and 

Implications for Management: New Zealand Limnological Society; with the 

assistance of the National Institute of Water and Atmospheric Research. Hamilton, 

New Zealand. 

Conrad, M. (1977). Functional significance of biological variability. Bulletin of Mathematical 

Biology, 39(2), 139-156.  

Cowie, B. (1980). Community dynamics of the bethnic fauna in a West Coast stream 

ecosystem. (PhD  thesis), University of Canterbury, Christchurch, New Zealand. 

Cowley, D. (1978). Studies on the larvae of New Zealand Trichoptera. New Zealand Journal 

of Zoology, 5(4), 639-750.  

Cummins, K. W. (1973). Trophic relations of aquatic insects. Annual Review of Entomology, 

18(1), 183-206.  

Death, R. G., & Joy, M. K. (2004). Invertebrate community structure in streams of the 

Manawatu–Wanganui region, New Zealand: the roles of catchment versus reach scale 

influences. Freshwater Biology, 49(8), 982-997.  



8 

 

Death, R. G., & Winterbourn, M. J. (1994). Environmental stability and community 

persistence: a multivariate perspective. Journal of the North American Benthological 

Society, 13 (2), 125-139.  

Delucchi, C. (1989). Movement patterns of invertebrates in temporary and permanent 

streams. Oecologia, 78(2), 199-207.  

Geist, J. (2011). Integrative freshwater ecology and biodiversity conservation. Ecological 

Indicators, 11(6), 1507-1516.  

Goldewijk, K. K., & Ramankutty, N. (2004). Land cover change over the last three centuries 

due to human activities: the availability of new global data sets. GeoJournal, 61(4), 

335-344.  

Jowett, I. G., & Richardson, J. (2003). Fish communities in New Zealand rivers and their 

relationship to environmental variables. New Zealand Journal of Marine and 

Freshwater Research, 37(2), 347-366.  

Joy, M., & Death, R. (2002). Predictive modelling of freshwater fish as a biomonitoring tool 

in New Zealand. Freshwater Biology, 47(11), 2261-2275.  

Joy, M. K., & Death, R. G. (2001). Control of freshwater fish and crayfish community 

structure in Taranaki, New Zealand: dams, diadromy or habitat structure? Freshwater 

Biology, 46(3), 417-429.  

Koleff, P., Gaston, K. J., & Lennon, J. J. (2003). Measuring beta diversity for presence–

absence data. Journal of Animal Ecology, 72(3), 367-382.  

Leathwick, J. R., Rowe, D., Richardson, J., Elith, J., & Hastie, T. (2005). Using multivariate 

adaptive regression splines to predict the distributions of New Zealand's freshwater 

diadromous fish. Freshwater Biology, 50(12), 2034-2052.  

Leprieur, F., Tedesco, P. A., Hugueny, B., Beauchard, O., Dürr, H. H., Brosse, S., & 

Oberdorff, T. (2011). Partitioning global patterns of freshwater fish beta diversity 



9 

 

reveals contrasting signatures of past climate changes. Ecology Letters, 14(4), 325-

334.  

Lewis, B. R., Jüttner, I., Reynolds, B., & Ormerod, S. J. (2007). Comparative assessment of 

stream acidity using diatoms and macroinvertebrates: implications for river 

management and conservation. Aquatic Conservation: Marine and Freshwater 

Ecosystems, 17(5), 502-519.  

Magurran, A. E. (1988). Ecological diversity and its measurement. Princeton University 

Press, Kent, England, U.K. 

Magurran, A. E. (2002). Measuring biological diversity. Blackwell Publishing, London, 

England, UK. 

McDowall, R. M. (1990). New Zealand Freshwater Fishes: a Natural History and Guide. 

Heinemann Reed. Auckland, New Zealand. 

McDowall, R. M. (2000). The Reed Field Guide to New Zealand Freshwater Fishes. 

Heinemann Reed. Auckland, New Zealand. 

McDowall, R. M. (2010). New Zealand Freshwater Fishes: an Historical and Ecological 

Biogeography (Vol. 32): Springer Science & Business Media, New York, USA. 

McIntosh, A., & McDowall, R. (2004). Fish communities in rivers and streams: Freshwaters 

of New Zealand (ed. by J.Harding, P.Mosley, C.Pearson and B.Sorrell). New Zealand 

Hydrological and Limnological Societies, Christchurch. (pp. 17.11-17.19). 

Morris, R. J. (2010). Anthropogenic impacts on tropical forest biodiversity: a network 

structure and ecosystem functioning perspective. Philosophical Transactions of the 

Royal Society B: Biological Sciences, 365(1558), 3709-3718.  

Munoz, F., Couteron, P., & Ramesh, B. R. (2008). Beta diversity in spatially implicit neutral 

models: A new way to assess species migration. American Naturalist, 172(1), 116-

127.  



10 

 

Nelson, G. C., Bennett, E., Berhe, A. A., Cassman, K., DeFries, R. S., Dietz, T., Dobermann, 

A., Dobson, A., Janetos, A., & Levy, M. A. (2006). Anthropogenic drivers of 

ecosystem change: an overview. Ecology and Society, 11(2), article 29.  

Oksanen, J. (2012). Ordination and analysis of dissimilarities: tutorial with R and Vegan. 

URL:http://www.pelagicos.net/MARS6910_spring2015/manuals/R_vegan_ordination

.pdf 

Olden, J. D. (2006). Biotic homogenization: a new research agenda for conservation 

biogeography. Journal of Biogeography, 33(12), 2027-2039.  

Quinn, J. (2000). Effects of pastoral development: New Zealand Stream Invertebrates: 

Ecology and Implications for Management. New Zealand Limnological Society, 

Christchurch. (pp. 208-229). 

Ramankutty, N., & Foley, J. A. (1999). Estimating historical changes in global land cover: 

Croplands from 1700 to 1992. Global Biogeochemical Cycles, 13(4), 997-1027.  

Soininen, J., Lennon, J. J., & Hillebrand, H. (2007). A multivariate analysis of beta diversity 

across organisms and environments. Ecology, 88(11), 2830-2838.  

Townsend, C. R. (1996). Invasion biology and ecological impacts of brown trout Salmo trutta 

in New Zealand. Biological Conservation, 78(1), 13-22.  

Turner, B. L. (1990). The earth as transformed by human action: global and regional changes 

in the biosphere over the past 300 years. Cambridge University Press, Cambridge, 

UK. 

Whittaker, R. H. (1960). Vegetation of the Siskiyou mountains, Oregon and California. 

Ecological Monographs, 30(3), 279-338.  

Winterbourn, M. (2000). Feeding ecology of New Zealand Stream Invertebrates: Ecology and 

Implications for Management. New Zealand Limnological Society, Christchurch, 

New Zealand. (pp. 100-124). 



11 

 

Winterbourn, M. J. (1991). Coping with current: Research on running freshwaters in New 

Zealand, 1967–91. New Zealand Journal of Marine and Freshwater Research, 25(4), 

381-391.  

Winterbourn, M. J., Gregson, K. L., & Dolphin, C. H. (1989). Guide to the Aquatic Insects of 

New Zealand (Vol. 9): Entomological Society of New Zealand, Auckland. 

 

 



12 

 

Chapter Two 

Spatial patterns of stream fish communities in Taranaki, 

New Zealand 

 

Abstract 

 The environmental and biological drivers of the similarity structure of fish 

communities were studied in 96 streams of Taranaki, New Zealand. Eight fish species 

occurred in > 5% of the sites. Long fin eels (Anguilla dieffenbachii) followed by brown trout 

(Salmo trutta) showed the widest distribution, occurring at 75% and 41% of the sites 

respectively.  Fish community similarity differed significantly between forest and pasture 

within region. Catchment land use and native riparian cover effectively partitioned fish 

abundances, and fish species consistently correlated with the altitudinal gradient within 

different land use and riparian classes. Land use vectors were more important than the habitat 

factors (e.g. habitat type and substrate type) in structuring fish communities. Upstream slope 

and farming limited species distribution in the extreme ends of upper and lower catchments, 

respectively. 

 

Introduction 

Stream fauna often reflect the quality of their physical and biological environment via  

community structure, composition, diversity and variability (Closs et al., 2004). During the 

past three decades, freshwater fish community structure has become increasingly popular as 

an indicator of the ecological quality of New Zealand lotic habitats (Jowett & Richardson, 

2003; Joy & Death, 2002; Leathwick et al., 2005), because of their well-known biology and 



13 

 

life history information (Karr, 1981; 1990). New Zealand native freshwater fish are 

characteristic for their migratory patterns, nocturnal behaviour and for generally occupying 

benthic water layers of streams and rivers (McDowall, 1990). Most of the freshwater fish in 

New Zealand are primarily riverine, and 17 native species migrate between sea and 

freshwaters to complete their life cycles.  More than 80% of New Zealand migratory fish 

species are either amphidromous or catadromous and spend most of their life in freshwater. 

In addition to the 38 native species, 21 exotic fish species have been introduced to freshwater 

habitats in New Zealand since the 19th century (McDowall, 1990; 2000; 2010; McIntosh & 

McDowall, 2004).  

Determinates of freshwater fish community structure and composition vary with the 

magnitude of spatial units such as region, reach or microhabitat (Closs et al., 2004; McIntosh 

& McDowall, 2004). In addition to evolutionary and geological history of the country, 

altitude and/or inland distance have a strong impact on regional scale distribution of 

freshwater fish in New Zealand (McIntosh & McDowall, 2004). It is not surprising to observe 

altitude playing an important role in the distribution of a fish community largely consisting of 

migratory species, because only the good climbers such as eels (Anguilla spp.) and koaro 

(Galaxias brevipinnis) are able to penetrate the steep hills to reach the headwaters. Therefore, 

the occurrence of migratory species generally declines with increasing altitude in inland 

freshwater habitats (Joy & Death, 2000; McDowall, 1990). Although the latitudinal diversity 

patterns of freshwater fish in the country have not been completely explained, all diadromous 

species occur across almost the entire latitudinal range of New Zealand, while the distribution 

of non-diadromous species is not consistent throughout the country’s geographic extent along 

latitudes  (McDowall, 2010).  

In addition to dispersal patterns at large spatial scales (e.g. island-wide), geography 

may influence species distribution also within a small region or sub-region (Heino, 2001). 



14 

 

However particular within region, patterns are often modified by land use constraints (Ricotta 

et al., 2014; Turner, 1989). It is therefore difficult to distinguish the effects of land use from 

regional scale fish community patterns often confounded by geographical factors such as 

altitude, mainly because of the confusion caused by co-linearity between multiple 

environmental drivers of community structure in relatively limited spatial scales (Graham, 

2003; Olden & Jackson, 2002). Hence, it is important to separate the geographical effects 

from land use impacts on communities, to understand the ecological processes, which 

underlie community distribution patterns, particularly in ecosystems impacted by agriculture. 

Nevertheless, in ecological studies of small regions (or limited areas), there has not been an 

explicit concern about geographical influence for community distribution patterns, compared 

to the research interest in prominent land use impacts caused by farming in particular (Heino, 

2001; Turner, 1989). In this study, I argue that the influence of geography may exist as 

‘signatures’ in species’ distribution patterns, even across different land uses, because of the 

strong connection between altitude and life cycle strategies of migratory fish species in New 

Zealand streams (McDowall, 2010). The likely fish distribution patterns influenced by 

geography (across natural and modified landuse classes) are introduced in this study, as “geo-

signatures” of the community structure. The term “geo-signature” was used to reflect the 

persistence of geographical influence on community patterns, at the presence of a given 

human modification such as conversion of natural forest to pasture.   

At relatively smaller spatial scales such as catchment or reach, vegetation, land use 

and water quality are strong environmental constraints for the composition of riverine fish 

communities (McIntosh & McDowall, 2004; Winterbourn, 1991). Farming is the 

predominant land use practice in New Zealand, and makes a major contribution to changes in 

freshwater habitats, including nutrient enrichment in water, removal of riparian native canopy 

and sediment deposition on streambeds (Ling, 2010; Quinn & Hickey, 1990; Zimmermann & 



15 

 

Death, 2002). Pasture streams receive less allochthonous organic matter such as leaf litter 

compared to the forested streams, but show high levels of algal growth with increased 

nitrogen levels and reduced riparian shade. Therefore, pastoral land use contributes to a 

dramatic change in freshwater fish habitats from their natural condition, and hence has a 

strong impact on reach scale structure and composition of freshwater communities in New 

Zealand (McDowall, 2001; McIntosh & McDowall, 2004; Winterbourn, 1991). 

 Despite well-known bottom-up effects of agriculture such as increased in-stream 

primary production, some researchers argue that top-down control by introduced fish has 

community-wide effects on stream fish in New Zealand (Flecker & Townsend, 1994; 

Schlosser, 1995; Simon & Townsend, 2003). Exotic species, for instance brown trout, have 

been studied for their predation, competition and distribution in relation to population 

dynamics and diversity changes in native fish, invertebrates, and algal growth in streams. For 

example, invertebrate densities have decreased and the algal standing crop has increased in 

streams invaded by brown trout compared to the streams occupied by native galaxiids 

(Flecker & Townsend, 1994). The particular effects of exotic species have, however, 

generally been assessed at lower spatial levels (e.g. reach), compared to the regional scale 

analyses of the effect of geography or land use change on stream communities (Heino, 2001; 

Jowett & Richardson, 2003). Thus, there is limited information on ecosystem-wide effects of 

introduced species, in comparison to regional scale impacts of geography and land use on 

freshwater communities. 

Studies largely based on inventory measures (e.g. population density, biomass or 

species richness) have a limited capacity to capture the full range of changes in spatial 

distribution patterns of natural communities influenced by multiple anthropogenic stressors 

(Soininen et al., 2007). Contrary to alpha diversity based on the number of species, beta 

diversity compares communities between given samples, and quantifies the level of 



16 

 

compositional similarity among the sampling sites. Moreover, community ecologists have 

developed comprehensive spatial analyses (e.g. homogeneity test of beta diversity), to 

investigate biological variability and its determinants (Anderson & Walsh, 2013; Clarke & 

Warwick, 2001; Dale & Fortin, 2014). Comparative analyses based on beta diversity are 

statistically useful to overcome the limited capability of conventional inventory studies for 

understanding the complexity of ecological patterns influenced by multiple geographical, 

anthropogenic and biological constraints. For instance, beta diversity is commonly used to 

analyse the compositional similarities of communities within and/or between different 

treatment groups (community partitioning) (Anderson et al., 2011; Magurran, 2002). In this 

study, I used beta diversity to question: a) Does agriculture partition the regional-scale 

freshwater fish community similarity? b) Are land-use factors more predictive than the 

habitat (scale) determinants of fish in freshwater ecosystems affected by agriculture? c) Do 

geo-signatures exist in the spatial structure of fish communities constrained by land use 

patterns?  

 

Methods 

Study sites 

The study sites include 96 streams (Appendix 1) at Egmont National Park, on Mount 

Taranaki and adjacent area, in the west of North Island of New Zealand (Fig. 2.1). Mount 

Taranaki peaks at 2,518 m above sea level is at the hub of the Egmont National Park. 

Protected area is predominately surrounded by pastoral land, dominated by dairy and beef 

farming, while natural forest covers the most of the area in upper catchments  (Joy & Death, 

2000; Winterbourn, 1991). Pasture covers more than half of the proximal land use along the 

total length of Taranaki streams (Taranaki Regional Council, 2010).  



17 

 

About 140 streams and rivers drain in a radial pattern through the Taranaki ring plain. 

Typically, Taranaki running waters are relatively short (compared to major New Zealand 

river systems: Waikato, Manawatu) first or second order streams (Joy & Death, 2000; 

Taranaki Regional Council, 2013).  Downstream dams are widespread across the Taranaki 

region, and likely to obstruct the upstream migration of fish, particularly above 100 m of the 

sea level, but all of the dams/ structures are not obligatory barriers for migration, and fish-

passes have been established by some of the downstream dams (Joy, 1999).  

Although the specific history of brown trout introduction to Taranaki streams is 

unclear, this particular species had been introduced to most North Island rivers since 1872 

(McDowall, 1990; Townsend, 1996). Further, studies show scientific records of brown trout 

in Taranaki streams since the late 1940s (Alien & Cunningham, 1957; Jowett, 1990). In 

Taranaki streams, the native fish community is mainly characterised by diadromous species 

(except for two species: Gobiomorphus basalis and G. breviceps) (Joy & Death, 2001). The 

sites used in this study sampled 96 different reaches of Taranaki streams in both protected 

forest areas (44 sites) and pasture (52 sites).  



18 

 

  

Fig. 2.1 Stream sampling sites in the Taranaki region of New Zealand used to collect fish and               

habitat data during summer 2012. 

Fish sampling and data collection 

Night spotlighting for fish species presence and abundance was carried out (once) at 

46 sites in Taranaki streams from January to June 2012.  These data were supplemented with 

recent (within 10 years) records from the New Zealand Freshwater Fish dataset (NZFFD).  

These historical data provided additional species records for 12 of the 46 sampling sites and 

added another 50 sites; making a total of 96 sites.  

  A stretch of nearly 150 m was sampled in each stream, using spotlighting and bank 

side observations, by moving in an upstream direction. Spotlighting was commenced 

approximately one hour after sunset. Each site was scanned between banks from the 

downstream end, by using a 30 W spotlight. Smaller fish species (e.g. bullies) were caught 



19 

 

alive using a dip net, and were collected into a container, while large fish were observed 

closely without catching. Fish species were identified to species level, and the collected fish 

were released back into their habitats after species identification was completed (McDowall, 

2000). 

Environmental data were collected in 50 different vectors at habitat, reach and 

regional scales for each sampling site. Visually observed habitat scale measures of this study 

include percentage cover of habitat type, substrate type, riparian vegetation type, in-stream 

vegetation, and leaf litter. Stream size (wetted width and depth) was measured by measuring 

tapes and rulers. Proximal land use pattern (within 500 m from the stream bank) was recorded 

by visual observations, and was further confirmed using 1:50,000 maps. Catchment land use 

data were recorded using 1:50,000 maps and Freshwater Ecosystems in New Zealand (FENZ) 

geo-based spatial data layers (Leathwick et al., 2010). Water quality parameters including 

pH, temperature and conductivity were measured by a ‘EuTech cyber scan PC-10 pH-

conductivity-temperature meter’. Further, I extracted habitat data, including climatic factors 

(rainfall and air temperature), geological composition (% composition of Ca and P of the 

surface rocks), proportional upstream land use, total Nitrogen concentration in water, for each 

sampling site, from the records in geo-based spatial data layers (Leathwick et al., 2010). In 

addition, NZFFD also provided geomorphological and land use data at habitat, riparian and 

catchment scales (Appendix II).   

Statistical Analysis  

Fish and environmental data were compiled into two major categories; a) fish 

(biological) database of 96 sites and b) environmental data. For fish data set, beta diversity 

was computed between streams, using Bray-Curtis and Sǿrenson indices which are popular 

(dis)similarity measures among ecologists, because they are suitable for communities with 



20 

 

limited numbers of species (Clarke, 1993; Magurran, 1988; 2002). Abundance data were 

standardised using Wisconsin double standardisation method to increase the gradient 

detection capability of beta diversity indices (Oksanen, 2008). Bray-Curtis and  Sǿrenson 

dissimilarities were organised into a triangular matrices (Clarke & Warwick, 2001; Oksanen, 

2012).   

  The non-metric multidimensional scaling (NMDS) method was used to construct a 

two dimensional ordination between sampling sites, on the biological similarity of fish 

between streams (Clarke, 1993; Clarke & Warwick, 2001; Ramette, 2007; Wickelmaier, 

2003).  NMDS models were generated for abundance data matrix. Biological similarity 

between sites is represented by the distance between points in an NMDS ordination, and the 

particular distance in a biological ordination is referred to as the biological distance.   

Euclidean distance between sites was computed, using normalised environmental data, and 

the environmental distance was compared with the biological distance, by using BIO-ENV/ 

BVSTEP procedures (Clarke & Warwick, 2001). I used the BIO-ENV/ BVSTEP test as an 

exploratory tool, to find out the most important environmental variables of the fish 

community structure in Taranaki streams. Because of the large number of environmental 

variables (46), stepwise matching (BVSTEP) was selected to explore important 

environmental variables (Appendix II).  In addition, I also compared raw density data matrix 

of fish with the biological similarity matrix, to identify important fish species for the 

biological distance between streams in Taranaki (BIO-BIO procedure). Significances of the 

most influential environmental variables and fish taxa of the fish community structure were 

identified by the Spearman correlation values (Clarke, 1993; Clarke & Warwick, 2001; Zar, 

1972).   

Fish similarity matrices were used to assess the partitioning dissimilarity of fish 

between forest and pasture (Oksanen et al., 2013). ANOSIM (Analysis of Similarity) test was 



21 

 

used to find out the effect of factor predictors (forest vs. pasture) on the similarity of fish data 

between and within groups (Clarke & Warwick, 2001; Oksanen, 2012). ANOSIM detects the 

importance of factor predictors, in partitioning beta diversity (Anderson & Walsh, 2013). 

Global R-values measure the degree of separation between selected groups of sites (forest vs. 

pasture), using average rank dissimilarities and the number of samples in a data matrix 

(Clarke & Warwick, 2001; Oksanen, 2011). Top contributors for between-group Bray-Curtis 

dissimilarity were selected by SIMPER (Similarity Percentage) analysis (Clarke & Warwick, 

2001; Shepherd et al., 1992). 

Further, I compared the average number of species in a stream (α diversity) and total 

number of species (γ diversity) between the groups of sites significantly different in their 

group similarities. Within-group variability was compared between impacted and non-

impacted sites, using the difference in group homogeneities (Oksanen, 2011). Group 

homogeneities were computed by partitioning similarities between sites for each factor 

predictor. In an ordination, distance to the centroid in a partitioned distance matrix was used 

to quantify the degree of variability within a group of samples. The distance between the 

group centroid and a sampling site (within a group of sites) was calculated by dividing the 

sum of squared inter-point (sites are represented as points in an ordination) distances by the 

number of points (Anderson, 2001). The difference in variability was analysed between 

groups by using the multivariate analysis of variance. Permutation tests (999) assessed the 

significance of variation between selected groups of sites. 

I used constrained correspondence analysis (CCA) to observe the biological variation, 

which is explainable by the important community drivers. Compared to previous NMDS 

models, CCA is rather linear mapping approach that uses Chi-squared distances between the 

objects (sites) (Ramette, 2007). In the CCA of this study, multiple co-linearity between all the 

gathered environmental constraints was gauged by variance inflation factor (VIF) of 



22 

 

individual factors among full environmental data set, and factors were considered to be fully 

independent of other variables, when their VIF is < 10. To minimise the co-linearity effect 

among factors, a reduced model was constructed by removing the least important variables 

found by BIO-ENV analysis  (Graham, 2003; Oksanen, 2011).  

Important environmental constraints were selected by CCA factor fitting approach, 

which is similar to BIO-ENV procedure. r2 value of each variable was used to assess the 

strength of fitting, and significances were derived by 999 permutation tests (Oksanen, 2008). 

Important vectors of CCA factor fitting results were selected to develop a reduced model, for 

the removal of co-linearity effect among multiple vectors.  Moreover, individual partial 

models were developed by constraining the biological ordination with selected most 

important vectors, and the significance of conditioned partial models was assessed by 

permutation tests. Each partial model  was constrained by only one selected  factor, but one to 

three of other explanatory vectors were fitted into each model, using vector fitting and/or 

surface fitting approaches (Oksanen, 2011). Further, within-strata variation in community 

composition was analysed from species fitted into their correlating sites in each partial model. 

Vegan package in R software (version 3.0.2) and PRIMER (6.0) were used for statistical 

analysis of this study (Clarke & Warwick, 2001; Oksanen, 2011).   



23 

 

Results 

Fish community composition  

Fifteen species of freshwater fish in five families: Anguillidae, Galaxiidae, Gobiidae 

(Eleotridae), Pinguipedidae and Salmonidae, were observed in the survey. Eight fish species 

occurred in > 5% of the 96 sites. Long fin eels (Anguilla dieffenbachii) occurred at 75% of 

the sites while koaro (Galaxias brevipinnis), shortjaw kokopu (Galaxias postvectis) and 

redfin bullies (Gobiomorphus huttoni) were found in more than 27% of the sites. Brown trout 

(Salmo trutta) were found from over 41% of the sites (Table 2.1 and Fig. 2.2).  

Fish community structure 

Abundance of eight fish species had a significant effect on the community structure.  

The impact of galaxiids was prominent in their abundance, but their axis correlations varied 

among species. For example, koaro and banded kokopu showed negative correlations while 

giant kokopu had positive correlations to both NMDS axes.  Additionally, redfin and 

common bullies had strong positive correlations to NMDS axis 2. Axis correlations of brown 

trout were positive in both NMDS 1 and 2 (Table 2.2).  

 



24 

 

Table 2.1:  Frequency of occurrence and relative abundances of freshwater fish species 

reported from the 96 streams and rivers from Taranaki during summer 2012.  

 
Common Name Scientific Name % Frequency 

of occurrence 
Relative 

Abundance (%) 

Longfin eel  Anguilla dieffenbachia 75.00 27.58 
Brown trout           Salmo trutta 41.67 09.23 
Koaro           Galaxias brevipinnis 37.50 17.03 
Short jaw kokopu  Galaxias postvectis 31.25 10.44 
Red fin bully     Gobiomorphus huttoni 27.08 26.26 
Shortifn eel     Anguilla australis 16.67 2.86 
Banded kokopu   Galaxias fasciatus 09.38 1.87 
Common bully     Gobiomorphus cotidianus 05.21 2.31 
Torrentfish      Cheimarrichthys fosteri 04.17 0.55 
Giant kokopu Galaxias argenteus 04.17 0.44 
Inanga          Galaxias maculatus 01.04 0.55 
Smelt           Retropinna retropinna 01.04 0.22 
Cran’s bully      Gobiomorphus basalis 01.04 0.22 
Upland bully     Gobiomorphus breviceps 01.04 0.22 
Bluegill bully   Gobiomorphus hubbsi 01.04 0.22 
 

Table 2.2: Selected fish species important in their abundance to two dimensional NMDS 

ordination (constructed on Bray-Curtis similarity) of   freshwater fish taxa reported from the 

96 streams and rivers from Taranaki during summer 2012.  

Common Name Scientific Name 
Axis Correlation 

 
  r2

  Significance 

NMDS1 NMDS2 
Brown trout            Salmo trutta 0.97 0.26 0.29 0.001 *** 
Banded kokopu   Galaxias fasciatus -0.99 -0.08 0.27 0.001 *** 
Koaro            Galaxias brevipinnis -0.34 -0.94 0.16 0.001 *** 
Redfin bully    Gobiomorphus huttoni -0.03 0.99 0.16 0.001 *** 
Shortjaw kokopu  Galaxias postvectis -0.71 0.71 0.12 0.002 **  
Giant kokopu     Galaxias argenteus 0.35 0.94 0.12 0.005 **  
Torrentfish Cheimarrichthys fosteri 0.05 0.99 0.08 0.022 *   
Common bully    Gobiomorphus cotidianus 0.21 0.98 0.08 0.027 *   

 ‘*’ P < 0.05, ‘**’ P < 0.01, ‘***’ P < 0.001 



25 

 

Environmental drivers of the fish community structure 

Among geographical vectors, a strong negative correlation occurred between NMDS 

axis 1 and northing (rs = -0.97), and upstream average slope was important for similarity on 

abundance of fish. Further, the occurrence of fish species was influenced by the gradient of 

upstream average slope, which is collinear with altitude (Fig. 2.3 and Table 2.7). Strong 

positive links occurred between NMDS axis 1 and riparian native vegetation in both 

percentage cover (rs = -0.90) and the proportionate width (rs = -0.90), and in contrast, farming 

in catchment positively correlated with the same axis of abundance ordination (rs = 0.87).  

Upstream geology characterised by calcium and phosphorus concentrations in surface rocks 

had significant but contrasting effects on the fish community ordination. However, in NMDS 

ordination of fish, the biological variation was poorly explained by environmental factors 

(very low r2 values), and multiple co-linearity was common among explanatory variables 

(Table 2.3). 

Table 2.3: Important environmental vectors of NMDS ordination based on fish abundance 

data and variance inflation factor (VIF) of each vector assessed on constrained community 

ordination (CCA). 

   

          Vector          Axis correlation    r2 Significance VIF 
NMDS1 NMDS2 

Northing         -0.97 0.23 0.13 0.003 ** 11.70 
% Farming©        0.87 -0.49 0.07 0.029 * 1034.07 
% Native ®        -0.94 0.34 0.06 0.044 * 793.12 
Native riparian cover within 100m      -0.90 -0.43 0.07 0.033 * 10.85 
Average slope (US)     -0.93 -0.36 0.07 0.041 * 19.53 
Calcium conc. in surface rocks (US)       0.60 0.80 0.07 0.039 * 21.73 
Phosphorus conc. in surface rocks (US)      -0.81 -0.59 0.08 0.027 * 9.86 

   US= upstream, © = catchment, ® = riparian                ‘*’ P < 0.05, ‘**’ P < 0.01, ‘***’ P < 0.001 

 



26 

 

 

 

Fig. 2.2   NMDS ordination on fish presence/absence in 96 Taranaki streams fitted with 

elevation contours (green). Fish species occurring in > 5% of sites excluded (random noise 

has been applied to remove the convergence among multiple objects (sites) with identical 

species composition). 

 

Surface fitting methods revealed that both altitude and northing had a linear 

relationship. Although the geographic extent of the study area is within a single region, the 

influence of northing was strong along the NMDS axis 1. Further, the number of forested 

sites was generally high in higher northings, alternatively to the higher proportion of 

-1.0 -0.5 0.0 0.5 1.0 

-1.0 

-0.5 

0.0 

0.5 

1.0 

NMDS1 

N
M

D
S2

 

Shortjaw kokopu 
Brown trout 

Koaro 

Banded kokopu 

Common bully 
Shortifn eel 

Redfin bully 
Longfin eel 

Forest 
Pasture 

Altitude Stress = 0.16 
Sǿrenson 



27 

 

agricultural streams clustered on the positive side of NMDS axis 1, showing southern sites of 

this study (Fig. 2.4).  

 

  

Fig. 2.3 Occurrence of the fish species reported from > 5% of 96 sites, fitted with the gradient 

in upstream average slope (red contours), on Sǿrenson similarity ordination between the 

study sites in Taranaki (random noise has been applied to remove the convergence among 

multiple objects (sites) with identical species composition). 

Co-linearity of fish community drivers  

Co-linearity of multiple environmental variables was very prominent among the 

important drivers (e.g. northing and altitude) of similarity structure considered in this study. 

However, phosphorus concentrations of the upstream surface rocks remained independent of 

other environmental vectors (VIF < 10) (Table 2.3 and Fig. 2.4).  

-1.0 -0.5 0.0 0.5 1.0 

-1.0 

-0.5 

0.0 

0.5 

1.0 

NMDS1 

N
M

D
S2

 

Forest 
Pasture 

Longfin eel
Shortfin eel 

Banded kokopu 

Koaro 

Shortjaw kokopu 

Redfin bullyCommon bully 

Trout 

 4  5 

 6 

 7  8 

 9 
 10  11 

 12 

 13 



28 

 

 

Fig. 2.4 Elevation (red) and northing (green) contours fitted onto the surface of NMDS 

ordination of Taranaki sites partitioned on proximal land use pattern (random noise has been 

applied to remove the convergence among multiple objects (sites) with identical species 

composition). 

Partitioning of fish similarity between forest and pasture 

 Proximal land use had a significant effect on the compositional similarity in fish 

communities between forested and pasture streams. However, the effect of land use was not 

very strong on the community similarities among 96 streams in Taranaki (Table 2.4 and Fig. 

2.5).           

-1.5 -1.0 -0.5 0.0 0.5 1.0 

-1.0 

-0.5 

0.0 

0.5 

1.0 

NMDS1 

N
M

D
S2

 

Forest 
Pasture 

Altitude 

Northing 

Stress = 0.16 
Sǿrenson Similarity 



29 

 

Table 2.4: Results of the Analysis of Similarity (ANOSIM) of Taranaki stream fish data 

grouped on the change in proximal land use pattern between forest and pasture. 

Community Index Factor Global R Significance 

Sǿrenson Land use 0.059 0.006** 

Bray-Curtis Land use 0.082 0.002** 

          ‘*’ P < 0.05, ‘**’ P < 0.01, ‘***’ P < 0.001 

   

 

   
   

 D
is

si
m

ila
rit

y 
R

an
k 

 

 

Fig. 2.5 Boxplot diagrams showing the results of Analysis of Similarity (ANOSIM) on fish 

data collected from Taranaki streams, in 2012.  Within and between group differences were 

compared using average ranked Bray-Curtis similarity, partitioned between forest (n = 44) 

and pasture (n = 52). 

Between Forest Pasture 

0 

1000 

2000 

3000 

4000 

R = 0.082,  P = 0.002



30 

 

 

Table 2.5:  Differences in diversity measures of fish between forest (n=44) and pasture 

(n=52) sites in Taranaki. 

Type of 
Diversity 

Measure Forest Pasture F 
value 

P 
value 

Degrees of 
freedom (d.f.) 

α Diversity Average number of species 2.50 2.63 0.27 0.60 1,94 

β Diversity Group multivariate dispersion 0.46 0.46 0.00 0.99 1,94 

γ Diversity Total number of species 9 15    

 

Agricultural streams generally had a higher number fish species (15) than the forested 

sites (9). However, the average species diversity (α) remained similar in streams between two 

land uses. Moreover, biological variability (within group multivariate dispersion) was not 

affected by pasture (Table 2.5). 

 

Community changes between forested and agricultural streams  

Longfin eels were the most abundant species in both pasture (39.25%) and forest 

(34.31%), while contributing to 25.86% of the compositional dissimilarity between the two 

site groups.  Koaro, shortjaw kokopu and banded kokopu were clearly more abundant in 

forest than within the agricultural streams. Brown trout and redfin bullies increased nearly by 

two fold in their relative abundance within streams from forest to pasture (Table 2.6).  



31 

 

Table 2.6: Ordered contribution by the top six fish species to the Bray-Curtis dissimilarity in 

species abundances (SIMPER test), between the forest (n=44) and pasture (n=52), in 96 

Taranaki streams. 

Species 

Average 
abundance (%) 

 

Average 

dissimilarity 

 

% 

contribution 

 

Cumulative% 

contribution Pasture Forest 

Long fin eel 39.25 34.31 18.83 25.86 25.86 

Koaro 7.15 23.88 13.00 17.85 43.71 

Trout 20.30 8.37 11.63 15.97 59.68 

Shortjaw kokopu 6.56 16.88 9.94 13.65 73.33 

Redfin bully 15.57 7.22 9.65 13.25 86.58 

Banded kokopu 1.49 6.33 3.74 5.14 91.72 

 

Fish community constraints 

In the results of constrained correspondence analysis (CCA) of this study, the full 

range of environmental data explained 70% of the variation in abundance (P < 0.05) of the 

fish community in Taranaki.  Amongst the important environmental constraints of fish 

abundance, land use indicators generally explained a greater biological variation compared to 

the particular measure explained by geographical vectors. Impact of reach riparian native 

cover (within 100m), % farming in the catchment, % native cover in the catchment, proximal 

land use pattern and total nitrogen concentration were significant in the fish CCA  model of 

abundance data. Among the geographical vectors, distance inland and downstream dams 

remained collinear (VIF > 10), even in the reduced model (Table 2.7).  



32 

 

Table 2.7: Axis correlations and variance inflation factor values (VIF) of important 

environmental vectors in constrained correspondence analysis (CCA model) of overall fish 

abundance of 96 Taranaki streams.  

Category Factor CA1 CA2 r2 Significance 
VIF 

(overall 
model) 

VIF 
(reduced 
model) 

Geography Easting          -0.22 -0.98 0.20 0.018 *   22.11 6.57 
Geography Altitude          -1.00 -0.09 0.23 0.008 **  14.68 7.55 
Geography Inland distance (km)    0.03 -1.00 0.35 0.001 *** 47.05 23.95 
Habitat Width (m)           1.00 0.04 0.19 0.029 *   7.84 1.80 
Land use % Native forest© -0.72 0.70 0.31 0.001 *** 1528.83 5.90 
Land use % Farming©        0.40 -0.92 0.41 0.001 *** 1037.07 4.97 
Land use Native Forest (US)      -0.75 0.66 0.30 0.001 *** 9052.63 Removed 
Land use Pasture (US)       0.75 -0.66 0.29 0.002 **  8883.47 14.41 
Land use Native riparian cover 

within 100m      -0.78 0.62 0.41 0.001 *** 10.85 4.01 

Land use Total Nitrogen 
concentration (ppt)     0.74 -0.68 0.23 0.011 * 34.39 9.88 

Land use Proximal Land use# 0.82 -0.57 0.35 0.001*** Not 
included 3.82 

Land use Dams (DS)           0.18 -0.98 0.28 0.006 **  20.79 11.42 
Climate Summer temperature  1.00 0.00 0.29 0.001 *** 37.42 9.52 
Geography Average Slope (US)      -0.81 0.59 0.25 0.007 **  19.53 2.11 

US= upstream, DS= Downstream © = catchment       ‘*’ P < 0.05, ‘**’ P < 0.01, ‘***’ P < 0.001 
   ‘#’ = forest or pasture 

 

Partial models  

Riparian cover 

The proportion of native riparian cover (within 100 m) significantly constrained fish 

abundance (P < 0.001), fish species decomposed mainly into three main strata across the 

gradient of reach riparian native buffer width (F = 7.31, d.f. = 3, 92). In each stratum of the 

riparian gradient, fish abundances differed in their attachment to the altitudinal range (Fig. 

2.6, Table 2.8).  



33 

 

 

Fig. 2.6 Overall fish abundance of Taranaki streams, constrained by the native riparian cover 

(green), and fitted with elevation contours (purple) and average upstream slope.  

 

Table 2.8: Selected fish species in Taranaki streams partitioned by the native riparian width 

and altitude, according to species abundances correlated in the riparian partial model. 

Primary constraint 
Reach riparian native cover (%) 

Fitted gradient 
Altitude (m) 

Fish species 

20-40 250-300 Torrent fish 
300-350 Common bully, Shortfin eel 

40-60 200-250 Redfin bully 
300-350 Longfin eel, Brown trout 
350-400 Banded kokopu 

60-70 300-350 Shortjaw kokopu 
400-450 Koaro 

 

-4 -2 0 2 4 

-1 

0 

1 

2 

CCA1 

C
A

1 
Forest 
Pasture 

Koaro 

Shortjaw kokopu 

Redfin bully 

Common bully Banded kokopu 

Torrentfish 

Shortifn eel 

Trout ully Band
fn eel

BTrout
Longfin eel 

Average slope 
(upstream) 

Riparian native cover 
(within 100m) 

Altitude



34 

 

Catchment land use 

Farming of the catchment partitioned fish community into three main strata (F= 6.75, 

P <0.001, d.f. = 3, 92). Most of the species were abundant when farmlands occurred less than 

20% of the catchment. In addition, native forest in the catchment had more or less similar 

partitioning effect on the overall fish community (F = 5.32, P < 0.01, d.f. = 3, 92). In the 

reduced CCA model (Table 2.7), two particular vectors varied independently from each other 

(VIF= 4.97 and 5.90, respectively), after the removal of other catchment land use vectors 

such as percentage of exotic forest, scrub, swamp land and alpine (Appendix II and Table 

2.7). In both of the partial models constrained by native forest and farming in the catchment, 

species abundances consistently fitted into altitudinal gradients (Fig.2.7 & Table 2.9). 

Table 2.9: Selected fish species in Taranaki streams partitioned by catchment land use and 

altitude, according to species abundances correlated in the riparian partial model. 

Primary Constraint 
Catchment farming (%) 

Fitted gradients 
Altitude (m) 

Fish species 

10-20 200-220 Redfin bully 
220-240 Torrent fish 
280-300 Shortjaw kokopu, Brown trout 
320-340 Longfin eel 
380-400 Banded kokopu 
400-420 Koaro 

20-30 300-320 Shortfin eel 
30-40 280-300 Common bully 
   
Catchment native forest (%)   
80-90 300-350 Shortjaw kokopu 

350-400 Banded kokopu 
400-450 Koaro 

70-80 200-250 Redfin bully 
250-300 Torrent fish 
300-350 Longfin eel, Shortfin eel, Brown 

trout 
60-70 250-300 Common bully 



35 

 

 

 

Fig. 2.7 Fish abundances in 96 Taranaki streams constrained by catchment farming (A) and 
native forest cover (B), fitted into altitude (red (A) & purple (B)) contours and upstream 
slope. 

-2 0 2 4 6 

-2 

-1 

0 

1 

2 

CCA1 

C
A

1 

Farming (catchment) 

Forest 
Pasture 

Koaro 

Banded kokopu 

Shortjaw kokopu 
Longfin eel 

Trout 
n eel
tttttShortifn eel 

 

Altitude 

Common bully 

Redfin bully 

Torrentfish 

A 

-2 0 2 4 6 

-2 

-1 

0 

1 

2 

CCA1 

C
A

1 

Native forest (catchment) 

Forest 
Pasture 

Common bully 
Koaro 

Redfin bully 

Banded kokopu 

Shortjaw kokopu Torrentfish 

Longfin eel gfin eel
Shortifn eel Sh

Trout 

Slope 
(upstream) 

Altitude 

B 



36 

 

Nitrogen concentration  

The total Nitrogen concentration of reaches significantly stratified the fish abundances 

(F=4.21, P < 0.05, d.f. = 4, 91). Most of the species were abundant in streams, when the total 

nitrogen concentration ranged between 0.6 and 1.4 ppb (parts per billion). Shortjaw kokopu, 

banded kokopu and brown trout abundances correlated with sites having less than 0.8 ppb of 

total nitrogen in streams. Species abundances further stratified across elevation gradients 

within each stratum of nitrogen concentration values in streams (Fig.2.8 & Table 2.10). 

 

 

Fig. 2.8 Overall fish abundance of Taranaki streams, constrained by nitrogen concentration 

(ppb), and fitted with elevation contours (purple) and riparian cover   

 

-2 0 2 4 6 

-2 

-1 

0 

1 

2 

3 

CCA1 

C
A

1 

Nitrogen concentration (ppb) 

Forest 
Pasture 

Koaro Common bully 

Redfin bully 

Shortjaw kokopu 

Banded kokopu 

Trout 

d k k
Longfin eel 

okopup
t
Torrentfish 

k
in eelShortifn eel 

Riparian native cover 
(within 100m) 

Altitude 



37 

 

Table 2.10: Selected fish species in Taranaki streams partitioned by nitrogen concentration 

and altitude, according to species abundances correlated in the riparian partial model. 

Primary constraint 
Total Nitrogen concentration (ppb) 

Fitted gradient 
Altitude (m) 

Fish species 

0.6-0.8 280-300 Shortjaw kokopu, Brown trout 
360-380 Banded kokopu 
380-400 Koaro 

0.8-1.0 220-240 Torrent fish 
300-320 Shortfin eel 
320-340 Longfin eel 

1.2-1.4 260-280 Common bully 
 

 

Discussion 

 The magnitude of the spatial scale is one of the most important factors in 

understanding the structural and functional diversity of a biological community (Ingels & 

Vanreusel, 2013). In this study, I attempted to investigate the key regional scale drivers of the 

fish community over Taranaki freshwater ecosystem from geographical, biological and 

anthropic perspectives. Regional-scale studies contribute to link our informative 

understanding of communities in their natural habitats, to the state of continental and global 

ecology (Cheruvelil et al., 2013), and therefore play a vital role in conservation and 

management measures.  

 The first hypothesis of this study questions not only the importance of agriculture for 

the fish community, but also the most effective mechanism of regional scale control of the 

top consumers of stream community food web. Studies have shown that New Zealand stream 

fish are feeding generalists, thus differences in fish community composition do not affect the 

diversity of their prey communities (Flecker & Townsend, 1994; McDowall, 2000). Results 

of this study show that brown trout is an important component of the fish communities of 



38 

 

Taranaki streams, because of their wide distribution and high relative abundance. However, 

the presence of brown trout has shown to be less significant in ecosystem scale fish 

community changes, compared to dairy farming, which shares a similar history with the trout 

in the Taranaki region (Jowett, 1990; McDowall, 2010; Townsend, 1996). Moreover, from 

ecosystem ecology point of view, brown trout is unlikely to qualify as a ‘keystone species’  

of New Zealand freshwaters, as shown in previous studies (Davic, 2003; Payton et al., 2002). 

Nonetheless, this study was mainly focused on the effects of land use and geography on fish 

communities in Taranaki. 

 In contrast to having lack of evidence for top-down effects from brown trout, land use 

constraints have led the (regional scale) structural and compositional changes in the fish 

community considered in this study.  Land use change from natural forest to pasture has 

consistently partitioned both distribution and abundance of fish community. Thus, the fish 

community considered in this study is clearly predicted by the catchment land use and, in 

particular, by variables such as percentage forest cover, percentage farming or total nitrogen 

concentration.  However, most of the habitat quality measures and land use vectors showed a 

high degree of co-linearity, in terms of their variance inflation factor (VIF). Multiple co-

linearity among environmental factors can confuse the statistical interpretation of community 

structure matched with large environmental data matrices (Graham, 2003). Although co-

linearity effect reduces the statistical importance of a particular variable in a community 

model, priori assumptions are required to assess the nature of co-linearity between interested 

environmental vectors, since some vectors are co-linear in nature, for instance temperature 

and elevation, while regional vectors such as geographical distance may confound with land 

use practices because of the human interference. Thus, co-linearity requires careful 

discussion, in both statistical and ecological senses. In this study, three types of co-linearity 

occurred between the important environmental vectors of fish community: a). Natural co-



39 

 

linearity (e.g. altitude and slope), b). Impact related co-linearity (e.g. forest cover and 

pasture) and c). Complex multiple co-linearity associated with habitat scale vectors 

potentially influenced by several other factors (e.g. % mud and sand). 

 To resolve the confusion of co-linearity, I reduced the number of variables to 

constrain biological ordinations considered in this study. Reduced models were advantageous 

in understanding how geography and land use pattern interactively, but independently, affect 

the fish community within the ecosystem.  Further, the partial models of this study revealed 

limiting factors of fish dispersal within the ecosystem. For instance, upstream slope and 

effect of farming have restricted the abundance of fish species at extreme ends of each 

variable; hence, optimal space for the fish abundance has been reduced by both land use and 

geography within the ecosystem. Particular independent effects of geo-land use factors for 

the  dispersal of fish are unlikely to be explored in simple bio-environmental data matching 

methods (Jowett & Richardson, 2003), as well as inventory models, based on alpha diversity 

(Dudley & Platania, 2007; Joy, 1999; Schlosser, 1995).  Therefore, from a meta-community 

point of reference, the results of this study contribute to explaining the validity of 

geographical attachment of the species, even in an ecosystem highly affected by agriculture 

(Leibold et al., 2004; Planque et al., 2011).  

Besides having the effect of co-linearity in biological-environmental data matching, 

the NMDS model explained a very limited variation of the fish community across any given 

environmental factor (Table 2.3). Even though the NMDS algorithm is popular in exploratory 

studies (because of its efficiency at identifying the relationships between communities and 

their environment), linear models such as CCA remain more suitable in explaining the 

community variations (Ramette, 2007). Therefore, I used both NMDS (non-linear) and CCA 

(linear) ordination methods to construct the fish community structure. Further, CCA models 

were very effective in constructing partial models presented in this study. 



40 

 

   Although proximal land use effectively partitioned the community composition, most 

of the important land uses vectors of community structure described catchment or regional 

scale changes in forest cover and pasture. Therefore, large-scale drivers (e.g. catchment forest 

cover) are likely to be more effective than medium or small-scale vectors (e.g. bank cover, 

habitat type), for the inter-site biological distance within the ecosystem. The impact of 

regional scale vectors consistently occurred from land use, geography, and climate, to the 

within ecosystem community structure analysed in this study. Hence, it is important 

considering a zonal clustering approach among particular important drivers, for the regional 

scale management and conservation plans, to improve the current conservation approach 

based on the coarse land use dichotomy in Taranaki (Chantepie et al., 2011; Januchowski-

Hartley et al., 2011; Roset et al., 2007). 

  

References 

 Alien, K., & Cunningham, B. (1957). New Zealand angling 1947-1952: results of the diary 

scheme. New Zealand Marine Department of Fisheries. Bulletin, 12.  

Anderson, M. J. (2001). A new method for non parametric multivariate analysis of variance. 

Austral Ecology, 26(1), 32-46.  

Anderson, M. J., Crist, T. O., Chase, J. M., Vellend, M., Inouye, B. D., Freestone, A. L., 

Sanders, N. J., Cornell, H. V., Comita, L. S., & Davies, K. F. (2011). Navigating the 

multiple meanings of β diversity: a roadmap for the practicing ecologist. Ecology 

Letters, 14(1), 19-28.  

Anderson, M. J., & Walsh, D. C. I. (2013). PERMANOVA, ANOSIM, and the Mantel test in 

the face of heterogeneous dispersions: What null hypothesis are you testing? 

Ecological Monographs, 83(4), 557-574.  



41 

 

Chantepie, S., Lasne, E., & Laffaille, P. (2011). Assessing the conservation value of 

waterbodies: the example of the Loire floodplain (France). Biodiversity and 

Conservation, 20(11), 2427-2444.  

Cheruvelil, K., Soranno, P., Webster, K., & Bremigan, M. (2013). Multi scaled drivers of 

ecosystem state: quantifying the importance of the regional spatial scale. Ecological 

Applications, 23(7), 1603-1618.  

Clarke, K. R. (1993). Non parametric multivariate analyses of changes in community 

structure. Australian Journal of Ecology, 18(1), 117-143.  

Clarke, K. R., & Warwick, R. M. (2001). An approach to statistical analysis and 

interpretation: Change in Marine Communities, 2nd edition. PRIMER-E Ltd, 

Plymouth, UK.  

Closs, G., Downes, B. J. B., & Andrew, J. (2004).  Freshwater Ecology: a Scientific 

Introduction: Blackwell Science Ltd, Oxford, UK . 

Dale, M. R., & Fortin, M. J. (2014). Spatial analysis: a guide for ecologists. Cambridge 

University Press. Cambridge, UK. 

Davic, R. D. (2003). Linking keystone species and functional groups: a new operational 

definition of the keystone species concept. Conservation Ecology, 7(1), r11.  

Dudley, R. K., & Platania, S. P. (2007). Flow regulation and fragmentation imperil pelagic-

spawning riverine fishes. Ecological Applications, 17(7), 2074-2086.  

Flecker, A. S., & Townsend, C. R. (1994). Community-wide consequences of trout 

introduction in New Zealand streams: Ecosystem Management. Springer, New York, 

USA. (pp. 203-215). 

Graham, M. H. (2003). Confronting multicollinearity in ecological multiple regression. 

Ecology, 84(11), 2809-2815.  



42 

 

Heino, J. (2001). Regional gradient analysis of freshwater biota: do similar biogeographic 

patterns exist among multiple taxonomic groups? Journal of Biogeography, 28(1), 69-

76.  

Ingels, J., & Vanreusel, A. (2013). The importance of different spatial scales in determining 

structural and functional characteristics of deep-sea infauna communities. 

Biogeosciences, 10(7), 4547-4563.  

Januchowski-Hartley, S., Hermoso, V., Pressey, R., Linke, S., Kool, J., Pearson, R., Pusey, B. 

J., & VanDerWal, J. (2011). Coarse-filter surrogates do not represent freshwater fish 

diversity at a regional scale in Queensland, Australia. Biological Conservation, 

144(10), 2499-2511.  

Jowett, I. G. (1990). Factors related to the distribution and abundance of brown and rainbow 

trout in New Zealand clear water rivers. New Zealand Journal of Marine and 

Freshwater Research, 24(3), 429-440.  

Jowett, I. G., & Richardson, J. (2003). Fish communities in New Zealand rivers and their 

relationship to environmental variables. New Zealand Journal of Marine and 

Freshwater Research, 37(2), 347-366.  

Joy, M., & Death, R. (2002). Predictive modelling of freshwater fish as a biomonitoring tool 

in New Zealand. Freshwater Biology, 47(11), 2261-2275.  

Joy, M. K. (1999). Freshwater fish community structure in Taranaki: dams, diadromy or 

habitat quality?.M.Sc. Thesis. Massey University, New Zealand.    

Joy, M. K., & Death, R. G. (2000). Development and application of a predictive model of 

riverine fish community assemblages in the Taranaki region of the North Island, New 

Zealand. New Zealand Journal of Marine and Freshwater Research, 34(2), 241-252.  



43 

 

Joy, M. K., & Death, R. G. (2001). Control of freshwater fish and crayfish community 

structure in Taranaki, New Zealand: dams, diadromy or habitat structure? Freshwater 

Biology, 46(3), 417-429.  

Karr, J. R. (1981). Assessment of biotic integrity using fish communities. Fisheries, 6(6), 21-

27.  

Karr, J. R. (1990). Fish communities as indicators of environmental degradation. American 

Fisheries Society Symposium. 8, 123-144. 

Leathwick, J. R., Rowe, D., Richardson, J., Elith, J., & Hastie, T. (2005). Using multivariate 

adaptive regression splines to predict the distributions of New Zealand's freshwater 

diadromous fish. Freshwater Biology, 50(12), 2034-2052.  

Leathwick, J. R., West, D., Gerbeaux, P., Kelly, D., Robertson, H., Brown, D., Chadderton, 

W., & Ausseil, A. G. (2010). Freshwater Ecosystems of New Zealand (FENZ) 

Geodatabase: User Guide. Department of Conservation Wellington, New Zealand. 

51(1).  

Leibold, M. A., Holyoak, M., Mouquet, N., Amarasekare, P., Chase, J., Hoopes, M., Holt, R., 

Shurin, J., Law, R., & Tilman, D. (2004). The metacommunity concept: a framework 

for multi scale community ecology. Ecology letters, 7(7), 601-613.  

Ling, N. (2010). Socio economic drivers of freshwater fish declines in a changing climate: a 

New Zealand perspective. Journal of Fish Biology, 77(8), 1983-1992.  

Magurran, A. E. (1988). Ecological diversity and its measurement. Princeton University 

Press, Kent, England, U.K. 

Magurran, A. E. (2002). Measuring biological diversity. Blackwell Publishing, London, 

England,  UK.  

McDowall, R. (2001). Getting the measure of freshwater fish habitat in New Zealand. 

Aquatic Ecosystem Health & Management, 4(4), 343-355.  



44 

 

McDowall, R. M. (1990). New Zealand Freshwater Fishes: a Natural History and Guide. 

Heinemann Reed, Auckland, New Zealand.  

McDowall, R. M. (2000). The Reed Field Guide to New Zealand Freshwater Fishes. 

Heinemann Reed, Auckland, New Zealand. 

McDowall, R. M. (2010). New Zealand Freshwater Fishes: an Historical and Ecological 

Biogeography (Vol. 32). Springer Science & Business Media. New York, USA. 

McIntosh, A., & McDowall, R. (2004). Fish communities in rivers and streams: Freshwaters 

of New Zealand (ed. by J.Harding, P.Mosley, C.Pearson and B.Sorrell). New Zealand 

Hydrological and Limnological Societies, Christchurch. (pp. 17.11-17.19). 

Oksanen, J. (2008). Multivariate analysis of ecological communities in R: Vegan tutor. URL: 

http://cc.oulu.fi/~jarioksa/opetus/metodi/vegantutor.pdf 

Oksanen, J. (2011). Multivariate analysis of ecological communities in R: vegan tutorial. 

URL: http://cc.oulu.fi/~jarioksa/opetus/metodi/vegantutor.pdf 

Oksanen, J. (2012). Ordination and analysis of dissimilarities: tutorial with R and Vegan. 

URL:http://www.pelagicos.net/MARS6910_spring2015/manuals/R_vegan_ordination

.pdf  

Oksanen, J., Blanchet, F. G., Kindt, R., Legendre, P., Minchin, P. R., O'Hara, R., Simpson, G. 

L., Solymos, P., Stevens, M. H. H., & Wagner, H. (2013). Package ‘vegan’. 

Community ecology package, 2(9). URL: https://cran.ism.acjp/web/packages/vegan.pdf  

Olden, J. D., & Jackson, D. A. (2002). A comparison of statistical approaches for modelling 

fish species distributions. Freshwater Biology, 47(10), 1976-1995.  

Payton, I. J., Fenner, M., & Lee, W. G. (2002). Keystone species: the concept and its 

relevance for conservation management in New Zealand. Department of 

Conservation, Wellington, New Zealand. 



45 

 

Planque, B., Loots, C., Petitgas, P., LindstrøM, U., & Vaz, S. (2011). Understanding what 

controls the spatial distribution of fish populations using a multi model approach. 

Fisheries Oceanography, 20(1), 1-17.  

Quinn, J. M., & Hickey, C. W. (1990). Magnitude of effects of substrate particle size, recent 

flooding, and catchment development on benthic invertebrates in 88 New Zealand 

rivers. New Zealand Journal of Marine and Freshwater Research, 24(3), 411-427.  

Ramette, A. (2007). Multivariate analyses in microbial ecology. FEMS Microbiology 

Ecology, 62(2), 142-160.  

Ricotta, C., Celesti-Grapow, L., Kühn, I., Rapson, G., Pyšek, P., La Sorte, F. A., & 

Thompson, K. (2014). Geographical constraints are stronger than invasion patterns for 

European urban floras. PloS one, 9(1), e85661.  

Roset, N., Grenouillet, G., Goffaux, D., Pont, D., & Kestemont, P. (2007). A review of 

existing fish assemblage indicators and methodologies. Fisheries Management and 

Ecology, 14(6), 393-405.  

Schlosser, I. J. (1995). Critical landscape attributes that influence fish population dynamics in 

headwater streams. Hydrobiologia, 303, 71–81. 

Shepherd, A. R. D., Warwick, R. M., Clarke, K. R., & Brown, B. E. (1992). An analysis of 

fish community responses to coral mining in the Maldives. Environmental Biology of 

Fishes, 33(4), 367-380.  

Simon, K. S., & Townsend, C. R. (2003). Impacts of freshwater invaders at different levels of 

ecological organisation, with emphasis on salmonids and ecosystem consequences. 

Freshwater Biology, 48(6), 982-994.  

Soininen, J., Lennon, J. J., & Hillebrand, H. (2007). A multivariate analysis of beta diversity 

across organisms and environments. Ecology, 88(11), 2830-2838.  



46 

 

Taranaki Regional Council. (2010). Small Stream Modification in Taranaki–An Assessment 

of the Ecological and Hydrological Values of Small Streams, the Cumulative Extent 

and Ecological Effects of Modification, and Implications for Policy and practice. 

Taranaki Regional Council, Stratford, New Zealand. 

Taranaki Regional Council. (2013). Maintaining indigenous freshwater biodiversity in the 

Taranaki region. Taranaki Regional Council, Stratford, New Zealand.  

Townsend, C. R. (1996). Invasion biology and ecological impacts of brown trout Salmo trutta 

in New Zealand. Biological Conservation, 78(1), 13-22.  

Turner, M. G. (1989). Landscape ecology: the effect of pattern on process. Annual Review of 

Ecology and Systematics, 20, 171-197.  

Wickelmaier, F. (2003). An introduction to MDS. Sound Quality Research Unit, Aalborg 

University, Denmark.  

Winterbourn, M. J. (1991). Coping with current: Research on running freshwaters in New 

Zealand, 1967–91. New Zealand Journal of Marine and Freshwater Research, 25(4), 

381-391.  

Zar, J. H. (1972). Significance testing of the Spearman rank correlation coefficient. Journal 

of the American Statistical Association, 67(339), 578-580.  

Zimmermann, E. M., & Death, R. G. (2002). Effect of substrate stability and canopy cover on 

stream invertebrate communities. New Zealand Journal of Marine and Freshwater 

Research, 36(3), 537-545.  



47 

 

Chapter Three 

Does land use have an effect on the variability of stream communities? 

 

Abstract 

 Fish and invertebrates were sampled in 15 streams in Taranaki, New Zealand, to 

investigate the impact of land use on the variability of stream communities. Fish and 

invertebrate community compositions differed significantly between forested and agricultural 

streams. Longfin eels (Anguilla dieffenbanchii), redfin bullies (Gobiomorphus huttoni) and 

brown trout (Salmo trutta) contributed for nearly 78% of the Bray-Curtis dissimilarity of fish, 

between forest and pasture. Deleatidium spp., Pycnocentrodes spp., Elmidae, Orthcladiinae, 

Aoteapsyche spp.  and Polypedilum spp.  influenced more than 60% of compositional 

changes of invertebrate communities from forest to pasture.  The compositional difference 

was stronger in invertebrates than in the fish communities between two land use classes.   

 Despite compositional differences in fish and invertebrate communities between 

native forest and pasture streams, the stream communities were equally variable between 

streams, within the two contrasting groups. Fish and invertebrates had similar average in-

stream variability across the study sites, regardless of their differenc