CONSERVING STREAMS WITH CHANGING CLIMATE: A MULTI-SCALED RESEARCH
FRAMEWORK TO CONSIDER CURRENT AND FUTURE CONDITION OF HAWAIIAN
STREAM HABITATS
By
Ralph William Tingley III

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Fisheries and Wildlife —Doctor of Philosophy
2017

ABSTRACT
CONSERVING STREAMS WITH CHANGING CLIMATE: A MULTI-SCALED RESEARCH
FRAMEWORK TO CONSIDER CURRENT AND FUTURE CONDITION OF HAWAIIAN
STREAM HABITATS
By
Ralph William Tingley III
Human land use in stream catchments and direct alterations to stream channels have
degraded stream habitats across the world, which in turn has resulted in declines in biodiversity.
In addition, climate change has and will continue to alter stream habitat and lead to additional
changes in species distributions and characteristics of existing populations, which in some cases,
may lead to species extinction. Given the current degraded condition of stream habitats and
potential effects of climate change, effective conservation of stream species requires that
decision-makers anticipate climate change and incorporate knowledge on its effects into
conservation strategies (i.e., proactive conservation). Proactive conservation therefore requires
that current influences of climate on streams and the organisms they support are understood.
However, observed effects of climate on organisms depend on the scale over which they are
examined. Studies at multiple spatial extents can increase our understanding of climate change
effects on streams in different ways and can provide effective insights when they are
complementary. The goal of my dissertation is to implement a research framework that
increases understanding of climate influences on streams over multiple scales in support of
proactive conservation. I applied this framework across the five largest Hawaiian Islands, where
little research has been conducted on effects of climate on stream organisms. In these chapters, I
examined the influence of climate on native stream organisms over two spatial scales, the
entirety of study area, which has a wide range in rainfall and other natural landscape features,
and within a region of Hawaii Island with little variability in most landscape features but a

gradient in mean annual rainfall. I then used my results to inform a spatial prioritization approach
that identified areas of high conservation value given potential changes in rainfall. I found that
across the entire study area, influences of climate and other natural landscape features on species
distributions could be used to characterize differences in stream habitats across Hawaii, while
over smaller spatial extents, influences of rainfall on native stream species population
characteristics (i.e., individual size, disease occurrence) were observed indirectly through
influences of flow magnitude, variability, and the occurrence of low flows. The results of the
spatial prioritization of stream habitats indicated that many areas capable of supporting unique
stream habitats and taxa assemblages will continue to do so as rainfall magnitude changes
through the 21st century, but some streams will experience annual or seasonal drying that may
result in a shift in their ability to support taxa. My study generated useful information on
potential effects of climate on species in Hawaii and demonstrated the value of considering
changes in population characteristics of stream organisms to understand underlying mechanisms
that may drive species loss with climate change. In addition, the multi-scale framework
implemented in my study can be applied in other understudied regions that require information
on influences of climate on stream ecosystems to inform proactive conservation decisionmaking.

Copyright by
RALPH WILLIAM TINGLEY III
2017

To Mom, Dad, Bea, Laura, Rob, George, and Samwise

v

ACKNOWLEDGEMENTS

I would like to thank the National Fish Habitat Partnership, the U.S. Fish and Wildlife
Service, the Hawaii Fish Habitat Partnership, the Association of Fish and Wildlife Agencies, and
the USDA Forest Service for providing financial support for the research described in this
dissertation. I would also like to thank the Department of Fisheries and Wildlife at Michigan
State University and the USGS National Climate Change and Wildlife Science Center for
awarding me with travel grants and fellowships that allowed me to complete and share this
research with a wide audience of researchers and stakeholders.
No single individual has had more influence on my development as a scientist and
researcher than my advisor, Dr. Dana Infante. Dana, thank you for taking a chance on the kid
who showed up for his graduate school interview in a Dr. Pepper t-shirt. It’s impossible to
imagine where I would be without your guidance, patience, and friendship. I’ll never forget the
times and experiences we have shared, but you are never getting those 40 dollars.
I would also like to thank my dissertation research committee members, Dr. Doug Beard,
Dr. Steve Hamilton, Dr. Brian Roth, and Dr. Rich MacKenzie for all of their advice and
guidance. I would especially like to thank Dr. MacKenzie for dreaming up the North Hilo coast
research and allowing me to take part in so many research projects. Thank you for all of the
professional and personal support over the last 6 years, Rich.
Thank you to the members of the Hawaii Fish Habitat Partnership for their input and
ideas that influenced the research described in this dissertation. I would especially like to thank
Dr. Gordon Smith for his advice and assistance in developing much of this work. Thank you
also to the individuals at the USGS National Climate Change and Wildlife Science Center,

vi

especially Dr. Shawn Carter and Dr. Abigail Lynch, for your support and guidance during my
time as a Science to Action Fellow. Also, thank you to Dr. Chuck Elzinga for helping me
discover how rewarding teaching can be. You are truly an inspiration.
Thank you to all the Aquatic Landscape Ecology lab members, current and former, who
have influenced my work and been such a huge part of my life (even Jared Ross and Dr. Darren
Thornbrugh). A special thanks to Arthur Cooper, Dr. Yin-Phan Tsang, Dr. Wesley Daniel, and
Kyle Herreman, who supplied technical and intellectual support for much of the research
described in this dissertation. Thank you to all the individuals at the Institute of Pacific Islands
Forestry for being so welcoming and kind during my time in Hawaii. I would especially like to
thank Dr. Ayron Strauch, Mike Riney, Therese Frauendorf, Zach Johnson, James Akau, Kristina
Black and Patra Foulk for their support in the field and for just being great people in general.
Hilo would have been a very lonely place without all of you.
I simply don’t have the space to name of all of the wonderful friends I have made at MSU
who mean so much to me, but thank you all for the experiences we have shared. A special thank
you Tracy Kolb, Kendra Kozlauskos, and Shikha Singh, who have been by my side through all
the ups and downs of grad school. To all of my family, especially Mom, Dad, Bea, Rob, George
and Sam; thank you for always being there for me and making life worth living. You all mean
the world to me. Finally, thank you to my best friend and fiancĂŠ Laura Claus, who supported me
through one of the toughest times of my life. I cannot wait to see where life takes us.

vii

PREFACE

The major research chapters in this dissertation have been prepared and formatted for
publication. Therefore, there is some repetition in concept, study site descriptions and methods
among chapters.

viii

TABLE OF CONTENTS

LIST OF TABLES ..................................................................................................................... xiii
LIST OF FIGURES ................................................................................................................... xvi
INTRODUCTION......................................................................................................................... 1
LITERATURE CITED .......................................................................................................... 5
CHAPTER 1 .................................................................................................................................. 9
INFLUENCES OF NATURAL LANDSCAPE FACTORS ON TROPICAL STREAM
ORGANISMS: AN ECOLOGICAL CLASSIFICATION OF HAWAIIAN ISLAND STREAMS
......................................................................................................................................................... 9
Abstract .................................................................................................................................... 9
Introduction ........................................................................................................................... 10
Methods .................................................................................................................................. 14
Study region ..................................................................................................................... 14
Spatial framework and natural and biological datasets .............................................. 15
Spatial units and framework ....................................................................................... 15
Natural landscape variables ....................................................................................... 16
Stream size and channel slope ......................................................................... 17
Influences on migration ................................................................................... 17
Soil characteristics .......................................................................................... 17
Rainfall ............................................................................................................ 18
Stream flow metrics..................................................................................................... 19
Biological data ............................................................................................................ 19
Identifying variables associated with distributions of stream organisms .................. 20
Identifying rainfall variables correlated with stream flow metrics ............................ 20
Identifying landscape variables important to distributions of stream organisms ...... 20
Ecological classification of stream reaches ................................................................... 21
Conditional inference tree analysis ............................................................................ 21
Validation of conditional inference tree results.......................................................... 22
Finalizing reach classes and examining taxa associations ........................................ 22
Results .................................................................................................................................... 23
Study site description ...................................................................................................... 23
Identifying variables associated with distributions of stream organisms .................. 24
Ecological classification of stream reaches ................................................................... 25
Conditional inference tree .......................................................................................... 25
Validation of conditional inference tree results.......................................................... 25
Final stream reach classes and associated taxa ......................................................... 26
Representation of reach classes across islands .......................................................... 27
ix

Discussion............................................................................................................................... 27
Landscape variables strongly associated with stream taxa distributions .................. 28
Ecological classification of stream reaches ................................................................... 29
Influences of landscape variables on stream taxa distributions ................................. 29
Inventory of classes across the Hawaiian Islands ...................................................... 31
Utility of the analytical approach ............................................................................... 32
Utility for conservation ................................................................................................... 33
Conclusion ....................................................................................................................... 34
APPENDICES ....................................................................................................................... 36
APPENDIX 1.A: Tables ................................................................................................. 37
APPENDIX 1.B: Figures ................................................................................................ 43
APPENDIX 1.C: Supplemental tables .......................................................................... 47
LITERATURE CITED ........................................................................................................ 53
CHAPTER 2 ................................................................................................................................ 61
ANTICIPATING EFFECTS OF CHANGING CLIMATE ON TROPICAL ISLAND
STREAMS: INFLUENCES OF STREAM FLOW ON ATYID MOUNTAIN SHRIMP ........... 61
Abstract .................................................................................................................................. 61
Introduction ........................................................................................................................... 62
Methods .................................................................................................................................. 66
Study area ........................................................................................................................ 66
Data collection and processing ....................................................................................... 67
Mean annual rainfall .................................................................................................. 67
Stream flow ................................................................................................................. 67
Daily stream flow data .................................................................................... 67
Stream flow metrics ......................................................................................... 67
Habitat ........................................................................................................................ 69
Stream habitat data sampling .......................................................................... 69
Stream habitat factors ..................................................................................... 69
A. bisulcata ................................................................................................................. 70
A. bisulcata sampling ...................................................................................... 70
Laboratory processing of A. bisulcata ............................................................ 71
A. bisulcata population characteristics ........................................................... 71
Data analysis .................................................................................................................... 73
Variation in flow metrics across the study region and relationships to rainfall ........ 73
Association between habitat factors and flow metrics ................................................ 73
Association between A. bisulcata population characteristics, flow metrics, and habitat
factors ......................................................................................................................... 73
Results .................................................................................................................................... 74
Variation in flow metrics across the study region and relationships with rainfall ... 74
Variation in habitat factors across the study region and relationships with rainfall 75

x

Associations among A. bisulcata population characteristics, flow metrics, and
habitat factors.................................................................................................................. 76
Discussion............................................................................................................................... 78
Variation in flow metrics and their relationship to rainfall ........................................ 78
Variation in habitat factors across the study region and relationships with rainfall 79
Association between A. bisulcata population characteristics and flow metrics ........ 80
Baseflow and frequency of low flows .......................................................................... 80
Temporary habitat desiccation ........................................................................ 81
Food availability.............................................................................................. 81
Stream temperature ......................................................................................... 82
Flow variability and brown spot disease .................................................................... 83
Implications for assessing climate change effects and prioritizing conservation action
..................................................................................................................................... 83
APPENDICES ....................................................................................................................... 85
APPENDIX 2.A: Tables ................................................................................................. 86
APPENDIX 2.B: Figures ................................................................................................ 96
APPENDIX 2.C: Supplemental tables ........................................................................ 102
LITERATURE CITED ...................................................................................................... 109
CHAPTER 3 .............................................................................................................................. 117
SPATIAL PRIORITIZATION OF HAWAII’S STREAM ECOSYSTEMS FOR
CONSERVATION IN THE CONTEXT OF CHANGING CLIMATE..................................... 117
Abstract ................................................................................................................................ 117
Introduction ......................................................................................................................... 118
Methods ................................................................................................................................ 121
Study area ...................................................................................................................... 121
Study units ..................................................................................................................... 122
Ranking reaches based on current and future conservation value .......................... 122
Defining reach conservation value based on uniqueness of ecological potential .... 123
Determining current ecological potential ..................................................... 123
Determining future ecological potential........................................................ 123
Creating and assigning the similarity matrix values to reaches ................... 124
Creating and assigning taxa association weights to reaches ........................ 124
Modifying reach conservation value based on current and future disturbance ....... 125
Applying current habitat condition score to estimates of conservation value
...................................................................................................................... .125
Applying climate change exposure scores to estimates of future conservation
value............................................................................................................... 125
Allowing reach conservation value to vary based on downstream connectivity ...... 126
Creating and applying downstream habitat loss penalty .............................. 126
Ranking stream reaches based on conservation value ............................................. 127
Assessing differences in prioritization when accounting for climate change .......... 127
xi

Results .................................................................................................................................. 128
Changes in reach class and reductions in rainfall with climate changes ................. 128
Ranking reaches based on future conservation value ............................................... 129
Mid-century time period ........................................................................................... 129
Late-century time period ........................................................................................... 130
Differences in priority reaches in current vs. future rankings ................................. 130
Discussion............................................................................................................................. 131
Ranking reaches based on current and future conservation value .......................... 131
Differences in priority reaches in current vs. future rankings ................................. 134
Implications for the proactive conservation of Hawaiian streams ........................... 134
Consideration of approach and future directions ...................................................... 135
APPENDICES ..................................................................................................................... 138
APPENDIX 3.A: Tables ............................................................................................... 139
APPENDIX 3.B: Figures .............................................................................................. 144
APPENDIX 3.C: Supplemental tables ........................................................................ 152
APPENDIX 3.D: Supplemental figures ...................................................................... 154
LITERATURE CITED ...................................................................................................... 167
CHAPTER FOUR ..................................................................................................................... 173
CONCLUSIONS: SUMMARY OF FINDINGS AND IMPLICATIONS FOR MANAGEMENT
..................................................................................................................................................... 173
Principal findings ................................................................................................................ 173
Chapter One: Influences of natural landscape factors on tropical stream organisms:
An ecological classification of Hawaiian Island streams ........................................... 173
Chapter Two: Anticipating the effects of changing climate on tropical island
streams: Influences of stream flow on atyid mountain shrimp ................................ 174
Chapter Three: Spatial prioritization of Hawaii’s stream ecosystems for
conservation in the context of changing climate ........................................................ 175
Management implications .................................................................................................. 176
Chapter One .................................................................................................................. 176
Chapter Two .................................................................................................................. 177
Chapter Three ............................................................................................................... 178
LITERATURE CITED ...................................................................................................... 180

xii

LIST OF TABLES

Table 1.1. Landscape variables considered for use in the Hawaiian stream classification.
Variables were summarized at three spatial scales: local (L), upstream (U) and downstream main
channel (D) catchments. Individual variables were summarized at multiple spatial extents,
resulting in 82 natural landscape variables. Asterisks (*) indicate the eleven natural landscape
variables included in the CCA. Two asterisks (**) indicate the variables that explained a
significant amount of variation within the biological dataset. .......................................................38
Table 1.2. Stream taxa used in the ecological classification of Hawaiian streams ........................39
Table 1.3. Descriptive statistics for all natural landscape variables summarized for perennial
stream catchments across the study region (n=4,732) and for those with biological samples
(n=403)...........................................................................................................................................40
Table 1.4. Descriptions, area under the curve (AUC) for validation datasets, and associated taxa
for each stream reach class. ...........................................................................................................41
Table 1.5. Percent of the total stream length across and within individual islands of each reach
class. ...............................................................................................................................................42
Table C1.1. Hydrological metrics calculated from Hydrological Index Tool and seasonal metrics
calculated for the wet and dry seasons (denoted with *). Hydrological Index Tool metric
descriptions follow Olden & Poff (2003). .....................................................................................48
Table C1.2. The eight principal components (PCs) generated from rainfall variables with Îť>1
explained 93.8% of the total variance. PC loadings ≥ ±0.60 are denoted by an asterisk. ............50
Table C1.3. Correlations between selected stream flow metrics and rainfall variables. Highly
correlated variables (r ≥ ±0.60) are denoted by an asterisk (*). ....................................................52
Table 2.1. Network catchment area, mean annual rainfall in catchments, and elevation of the
eleven study reaches. Streams are listed geographically from south to north across the study area
(Figure 2.1). ...................................................................................................................................87
Table 2.2. Stream flow metrics generated for each stream reach using daily flow data. ..............88

xiii

Table 2.3. Descriptive statistics for selected stream flow metrics across the 11 study reaches for
the entire period of record. Also shown are the mean values of each stream flow metric for each
7-month period prior to the sampling season and the coefficient of variation of these mean
values. ............................................................................................................................................89
Table 2.4. Pearson’s correlation coefficients describing relationships between mean annual
rainfall (MAR), drainage area (DA) and stream flow metrics. Correlation coefficients greater
than an absolute value of 0.6 and significant (Îą=0.05) are shown in bold. ...................................90
Table 2.5. Descriptive statistics for habitat factors across study reaches averaged across years. .91
Table 2.6. Pearson’s correlation coefficients between stream flow metrics and habitat factors.
Correlation coefficients greater than an absolute value of 0.60 and significant (Îą=0.05) are shown
in bold. ...........................................................................................................................................92
Table 2.7. Descriptive statistics for A. bisulcata population characteristics across sites for each
sample year. ...................................................................................................................................93
Table 2.8. Pearson’s correlation coefficients between A. bisulcata population characteristics and
stream flow metrics. Correlation coefficients greater than 0.6 and statistically significant
(Îą=0.05) are shown in bold. ..........................................................................................................94
Table 2.9. Pearson’s correlation coefficients between A. bisulcata population characteristics and
habitat factors. Correlation coefficients greater than 0.6 and statistically significant (Îą=0.05) are
shown in bold. ...............................................................................................................................95
Table C2.1. Stream flow metric values for each stream over the entire period of record and
each 7-month flow period (October to May) of each year (2012,2013 and 2014). .....................103
Table C2.2. Substrate size categories and references following Higashi & Nishimoto (2007) used
to estimate percent substrate composition and mean substrate size in stream reaches. ...............105
Table C2.3. Length-mass regression coefficients for each reach each year. Asterisks indicate
regressions not used to calculate geometric mean coefficient values due to a dominance by
juveniles, limited sample size (n<30) or R2 < 0.8. .......................................................................106
Table C2.4. Averages values of habitat factors from transect surveys. .......................................107
Table C2.5. Atyoida bisulcata population characteristics summarized by each reach each year108

xiv

Table 3.1. Stream classes, taxa association weightings and taxa associated with each class
resulting from the Hawaii stream reach classification. Associated taxa are defined by prevalence
within a given class compared to prevalence across all samples reaches (Tingley this volume a).
.................................................................................................................................................... 140
Table 3.2. Changes in stream reach length for each class under each climate scenario and time
period. Also shown are total lengths of stream reaches projected to be greater than one standard
deviation from current mean annual or dry season rainfall under each scenario and time period.
.................................................................................................................................................... 141
Table 3.3. Total length of stream reaches that overlap in the top 20% of ranked reaches at each
time period for the two climate scenarios. The percent of total reach length that is within the top
20% within a region and across islands as well as the mean, standard deviation and range of
UHCI scores within a region are also shown ............................................................................. 142
Table C3.1. Similarity matrix applied to stream reach classes. ................................................. 153

xv

LIST OF FIGURES

Figure 1.1. Study area and locations with biological data. ............................................................44
Figure 1.2. Conditional inference tree and resulting characteristics of reach classes (A-I). The
relative species occurrence is represented within each histogram by its taxa value (1-9). Nodes
(I-VIII) indicate landscape variables that create significant splits in species distributions. ..........45
Figure 1.3. Final classification of Hawaiian stream reaches. ........................................................46
Figure 2.1. North Hilo coast of Hawaii Island and locations of study reaches along a gradient in
mean annual rainfall. ......................................................................................................................97
Figure 2.2. Average post orbital carapace length (mm) of males versus the natural log (x+1) of
Q70 yield. Panels show 2012 (A), 2013 (B) and 2014 (D) data. ....................................................98
Figure 2.3. Average relative mass of individuals collected during each sampling event versus the
natural log (x+1) of Q70 yield. Significant correaltions are denoted by a solid (vs. dashed) trend
line. Panels show 2012 (A), 2013 (B) and 2014 (D) data. .............................................................99
Figure 2.4. The percent of male A. bisulcata with brown spot disease versus the natual log (x+1)
transformed coefficient of variation in daily flows (CV). Panels show 2012 (A) and 2013 (B)
data. ..............................................................................................................................................100
Figure 2.5. Average biomass (g/m2 ) versus the natural log (x+1) Q70 yield. Panels show 2012
(A), 2013 (B) and 2014 (D) data. .................................................................................................101
Figure 3.1. The five Hawaiian Islands considered for spatial prioritization of stream reaches ...145
Figure 3.2. Spatial prioritization approach taken for the ranking of stream reaches given current
ecological potential and habitat condition and with projected changes in rainfall. .....................146
Figure 3.3. Reach classes used to estimate ecological potential (from Tingley this volume a).
Intermittent streams not shown. ...................................................................................................147
Figure 3.4. Spatial prioritization ranking of stream reaches under mid-century projected climate
scenarios. Results were applied to local catchments for ease of interpretation. Rankings reflect
those reaches that were above a given percentage in both scenarios. ..........................................148
Figure 3.5. Spatial prioritization ranking of stream reaches under late-century projected climate
scenarios. Results were applied to local catchments for ease of interpretation. Rankings reflect
those reaches that were above a given percentage in both scenarios. ..........................................149

xvi

Figure 3.6. Similarities and differences in the top 5 and 10% ranked sites under current condition
and mid-century scenarios. ..........................................................................................................150
Figure 3.7. Similarities and differences in the top 5 and 10% ranked sites under current condition
and late-century scenarios. ...........................................................................................................151
Figure D3.1. Changes in stream class under RCP 4.5 climate scenario for the mid-century time
period.. “Negative” refers to a change in class resulting from a decline in rainfall, while
“Positive” indicates a change in class resulting from an increase in rainfall...............................155
Figure D3.2. Changes in stream class under RCP 4.5 climate scenario for the late-century time
period.. “Negative” refers to a change in class resulting from a decline in rainfall, while
“Positive” indicates a change in class resulting from an increase in rainfall...............................156
Figure D3.3. Changes in stream class under RCP 8.5 climate scenario for the mid-century time
period.. “Negative” refers to a change in class resulting from a decline in rainfall, while
“Positive” indicates a change in class resulting from an increase in rainfall...............................157
Figure D3.4. Changes in stream class under RCP 8.5 climate scenario for the late-century time
period.. “Negative” refers to a change in class resulting from a decline in rainfall, while
“Positive” indicates a change in class resulting from an increase in rainfall. .............................158
Figure D3.5. Stream reaches that increased or decreased by greater than one standard deviation
from current conditions. during the dry season under the RCP 4.5 climate scenario for the midcentury time period. .....................................................................................................................159
Figure D3.6. Stream reaches that increased or decreased by greater than one standard deviation
from current conditions. during the dry season under the RCP 4.5 climate scenario for the latecentury time period. .....................................................................................................................160
Figure D3.7. Stream reaches that increased or decreased by greater than one standard deviation
from current conditions. during the dry season under the RCP 8.5 climate scenario for the midcentury time period. .....................................................................................................................161
Figure D3.8. Stream reaches that increased or decreased by greater than one standard deviation
from current conditions. during the dry season under the RCP 8.5 climate scenario for the latecentury time period. .....................................................................................................................162
Figure D3.9. Stream reaches that increased or decreased by greater than one standard deviation
from current annual mean annual rainfall under the RCP 4.5 climate scenario for the late-century
time period. .................................................................................................................................163

xvii

Figure D3.10. Stream reaches that increased or decreased by greater than one standard deviation
from current annual mean annual rainfall under the RCP 8.5 climate scenario for the midcentury time period. .....................................................................................................................164
Figure D3.11. Stream reaches that increased or decreased by greater than one standard deviation
from current annual mean annual rainfall under the RCP 8.5 climate scenario for the late-century
time period. ..................................................................................................................................165
Figure D3.12. Stream reaches that increased or decreased by greater than one standard deviation
from current annual mean annual rainfall under the RCP 4.5 climate scenario for the midcentury time period. .....................................................................................................................166

xviii

INTRODUCTION
Current rates of species extinction are estimated to be several orders of magnitude greater
than historical baselines (Pimm et al., 2013), and approximately one in three freshwater
vertebrate species are threatened with extinction across the world (Collen et al., 2014). In
streams, habitat degradation resulting from indirect effects of anthropogenic land use and
channel and flow alterations has contributed substantially to declines in abundance and species
extinctions (Allan et al., 2004; Dudgeon et al., 2006; Helfman, 2007). Further degradation of
freshwater habitats is likely to occur as anthropogenic disturbance to the landscape continues to
increase over the 21st century (Martinuzzi et al., 2014). The identification of areas that support
stream species of conservation interest is therefore a priority of management agencies and
conservation partnerships alike (e.g., USFWS, 2006; American Rivers, 2013). However, future
effects of changing climate on stream habitats and the species they support may confound
conservation actions if only current condition is taken into account.
Over the last half century, effects of increasing air temperature and changes in
precipitation resulting from climate change have been observed in streams in multiple regions.
Hari et al. (2006) linked increases in air temperature to reductions in available habitat and
increases in disease occurrence for brown trout, Salmo trutta, in streams of Switzerland. In the
Pacific Northwest, U.S.A., Ward et al. (2015) associated declines in the natural reproduction of
Chinook salmon, Oncorhynchus tshawytscha, with increases in winter stream flow variability
likely resulting from a greater number of winter rainfall events. In addition, changes in the
distribution of stream fishes resulting from changes in climate have been documented across the
globe (as reviewed by Comte et al., 2013). As temperatures continue to rise and regional
precipitation changes, gains or losses in species distributions (e.g., Mohseni et al. 2003; Comte et

1

al., 2013; Domisch et al., 2013), changes in population characteristics (i.e., abundance, growth,
occurrence of disease; e.g., Pease and Paukert, 2014), and higher rates of extinction of stream
species (e.g., Xenopoulos et al., 2005; Tedesco et al., 2013) are all projected to occur.
The current state of stream systems and potential effects of climate change suggest that to
effectively conserve stream species and their habitats we must incorporate knowledge on
anticipated changes in air temperature and precipitation into conservation decision-making
(Palmer et al., 2009; Zeigler et al., 2012). Therefore, understanding habitat factors most
important to stream organisms as well as their response to changes in these conditions (i.e., water
temperature, stream flow) are necessary to develop effective conservation strategies. These
responses include both changes in species distributions as well as underlying changes to
characteristics of populations, such as disease prevalence and fecundity and growth of
individuals, which may ultimately threaten population persistence. However, understanding
organism response to climate change can be difficult, given that influences of precipitation and
rainfall are context-dependent, and biotic response to stream habitat will vary depending on the
spatial scale over which it is examined (Heino et al., 2007; Gornish & Tylianakis, 2013).
Across large spatial extents, regional differences in climate and interactions of climate
with geology, topography and other natural landscape features within stream catchments control
stream habitat and in turn influence distributions of stream organisms (e.g., Hynes, 1975;
Frissell, 1986; Allan, 2004). By first understanding influences of landscape factors on species
distributions over large spatial extents, we can more effectively examine the response of
population characteristics to climate-mediated habitat changes within smaller regions with less
variability in landscape features. These responses can then provide richer insight into underlying
mechanisms that may eventually lead to population loss (Whitney et al., 2016) while also

2

providing information on the value of a given stream system for species conservation.
Therefore, studies that increase our understanding of effects of climate change on stream species
over several spatial scales can complement one another to improve proactive conservation
decision making.
Tropical island streams often support endemic species of fish and shrimp that require
hydrologic connections to marine habitats to complete their amphidromous life history
(Fitzsimons, 2002; Kikkert et al., 2009; Jenkins et al., 2010; Bauer, 2013). Variation in habitat
requirements and migratory abilities results in complex patterns of species distributions within
stream catchments and across islands. Tropical island streams and the species they support have
been historically understudied, but the need for research grows as climate change progresses and
anthropogenic disturbance, largely from land use and channel alteration, increases (Smith et al.,
2003). Because many tropical-island 1st and 2nd order streams are dependent on persistent
rainfall for perennial stream flows due to low groundwater input (Craig, 2003) and because they
provide important habitat for many endemic species, alterations to precipitation magnitude or
frequency resulting from climate change may have severe effects on stream organisms (Leong et
al., 2004). Given the severity of current and future threats, greater understanding of differences
in stream habitats within and among island systems and their relationships to natural landscape
features, including rainfall, can be useful in understanding differences in habitat sensitivity to
change and ultimately to prioritizing conservation actions.
The goal of my dissertation was implement a research framework that increases
understanding of climate influences on streams over multiple scales in support of proactive
conservation. I focus my research on the sub-tropical Hawaiian Islands where effects of climate
on Hawaiian stream systems have historically received little attention, yet declines in annual

3

rainfall over the last century have been associated with declines in baseflow (Bassiouni & Oki,
2013; Frazier & Giambelluca, 2016) and regional declines in rainfall are projected to continue
(Timm et al., 2015). I begin by identifying influences of the natural landscape on distributions of
stream organisms across the five largest Hawaiian Islands with perennial streams, which I then
use as the basis for development of an ecological classification of streams at the reach scale
across the islands. Then, over a set of reaches with similar landscape characteristics, I assess
influences of rainfall on the stream flow regime and in turn, influences of the stream flow regime
on population characteristics of an endemic atyid shrimp, Atyoida bisulcata. Finally, using the
ecological classification of reaches and projected climate data, I identify areas across the five
islands that are projected to retain high conservation value under both current and future
conditions. My research will increase our understanding of climate change effects on tropical
island streams by examining biological response to climate over multiple spatial scales,
providing information that can help inform and implement proactive conservation of the unique
stream organisms of the Hawaiian Islands.

4

LITERATURE CITED

5

LITERATURE CITED

American Rivers Board of Directors, 2013. American Rivers Strategic Plan 2014 -2018
http://americanrivers.org/wpcontent/uploads/2016/03/AmericanRiversStrategicPlan20142018.pdf Accessed 17.03.04.
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.
Bassiouni, M. & D. S. Oki, 2013. Trends and shifts in streamflow in Hawai'i, 1913-2008.
Hydrological Processes 27: 1484-1500.
Bauer, R. T., 2013. Amphidromy in shrimps: a life cycle between rivers and the sea. Latin
American Journal of Aquatic Research 41: 633-650.
Collen, B., F. Whitton, E.E. Dyer, J.E.M. Baillie, N. Cumberlidge, W.R.T. Darwall, C. Pollock,
N.I. Richman, A.M. Soulsby & M. Bohm, 2014. Global patterns of freshwater species
diversity, threat and endemism. Global Ecology and Biogeography 23: 40-51.
Comte, L., L. Buisson, M. Daufresne & G. Grenouillet, 2013. Climate-induced changes in the
distribution of freshwater fish: observed and predicted trends. Freshwater Biology 58:
625-639.
Craig, D. A., 2003. Geomorphology, development of running water habitats, and evolution of
black flies on Polynesian islands. Bioscience 53: 1079-1093
Domisch, S., M. B. Araujo, N. Bonada, S. U. Pauls, S. C. Jahnig & P. Haase, 2013. Modelling
distribution in European stream macroinvertebrates under future climates. Global Change
Biology 19: 752-762.
Dudgeon, D., A. H. Arthington, M. O. Gessner, Z. I. Kawabata, D. J. Knowler, C. Leveque, R. J.
Naiman, A. H. Prieur-Richard, D. Soto, M. L. J. Stiassny & C. A. Sullivan, 2006.
Freshwater biodiversity: importance, threats, status and conservation challenges.
Biological Reviews 81: 163-182.
Fitzsimons, J. M., J. E. Parham & R. T. Nishimoto, 2002. Similarities in behavioral ecology
among amphidromous and catadromous fishes on the oceanic islands of Hawai'i and
Guam. Environmental Biology of Fishes 65: 123-129.
Frazier, A. G. & T. W. Giambelluca, 2016. Spatial trend analysis of Hawaiian rainfall from
1920-2012. International Journal of Climatology

6

Frissell, C. A., W. J. Liss, C. E. Warren & M. D. Hurley, 1986. A hierarchical framework for
stream habitat classification - Viewing streams in a watershed context. Environmental
Management 10: 199-214.
Gornish, E. S. & Tylianakis, J. M., 2013. Community shifts under climate change: mechanisms
at multiple scales. American Journal of Botany 100: 1422-1434.
Hari, R. E., D. M. Livingstone, R. Siber, P. Burkhardt-Holm & H. Guttinger, 2006.
Consequences of climatic change for water temperature and brown trout populations in
Alpine rivers and streams. Global Change Biology 12: 10-26.
Heino, J., H. Mykra, J. Kotanen & T. Muotka, 2007. Ecological filters and variability in stream
macroinvertebrate communities: do taxonomic and functional structure follow the same
path? Ecography 30: 217-230.
Helfman, G. S., 2007. Fish conservation: a guide to understanding and restoring global aquatic
biodiversity and fishery resources. Island Press, Washington, D.C.
Hynes, N. B., 1975. Edgardo Baldi Memorial Lecture: The stream and its valley. Verhandlungen
Internationale Vereinigung fĂźr Theoretische und Angewandte Limnologie 19: 1-15.
Jenkins, A. P., S. D. Jupiter, I. Qauqau & J. Atherton, 2010. The importance of ecosystem-based
management for conserving aquatic migratory pathways on tropical high islands: a case
study from Fiji. Aquatic Conservation: Marine and Freshwater Ecosystems 20: 224-238.
Kikkert, D. A., T. A. Crowl & A. P. Covich, 2009. Upstream migration of amphidromous
shrimps in the Luquillo Experimental Forest, Puerto Rico: temporal patterns and
environmental cues. Journal of the North American Benthological Society 28: 233-246.
Leong, J.A., J. J. Marra, M. L. Finucane, T. Giambelluca, M. Merrifield, S. E. Miller, J.
Polovina, E. Shea, M. Burkett, J. Campbell, P. Lefale, F. Lipschultz, L. Loope, D.
Spooner & B. Wang, 2014. Ch. 23: Hawai‘i and U.S. Affiliated Pacific Islands In J. M.
Melillo, T.C. Richmond & G. W. Yohe (eds.), Climate Change Impacts in the United
States: The Third National Climate Assessment U.S. Global Change Research Program
537-556. U.S. Government Printing Office, Washington, DC.
Martinuzzi, S., S. R. Januchowski-Hartley, B. M. Pracheil, P. B. McIntyre, A. J. Plantinga, D. J.
Lewis & V. C. Radeloff, 2014. Threats and opportunities for freshwater conservation
under future land use change scenarios in the United States. Global Change Biology 20:
113-124.
Mohseni, O., H. G. Stefan & J. G. Eaton, 2003. Global warming and potential changes in fish
habitat in US streams. Climatic Change 59: 389-409.

7

Palmer, M. A., D. P. Lettenmaier, N. L. Poff, S. L. Postel, B. Richter & R. Warner, 2009.
Climate change and river ecosystems: Protection and adaptation options. Environmental
Management 44: 1053-1068.
Pease, A. A. & C. P. Paukert, 2014. Potential impacts of climate change on growth and prey
consumption of stream-dwelling smallmouth bass in the central United States. Ecology of
Freshwater Fish 23: 336-346.
Pimm, S. L., C. N. Jenkins, R. Abell, T. M. Brooks, J. L. Gittleman, L. N. Joppa, P. H. Raven, C.
M. Roberts, & J. O. Sexton, 2014. The biodiversity of species and their rates of
extinction, distribution, and protection. Science 344: 987-.
Smith, G. C., A. R. Covich & A. M. D. Brasher, 2003. An ecological perspective on the
biodiversity of tropical island streams. Bioscience 53: 1048-1051.
Tedesco, P. A., T. Oberdorff, J.-F. Cornu, O. Beauchard, S. Brosse, H.H. Durr, G. Grenouillet, F.
Leprieur, C. Tisseuil, R. Zaiss & B. Hugueny, 2013. A scenario for impacts of water
availability loss due to climate change on riverine fish extinction rates. Journal of
Applied Ecology 50: 1105-1115.
Timm, O. E., T. W. Giambelluca & H. F. Diaz, 2015. Statistical downscaling of rainfall changes
in Hawaiˈi based on the CMIP5 global model projections. Journal of Geophysical
Research-Atmospheres 120: 92-112.
United States Fish & Wildlife Service, 2016. Strategic plan for the U.S. Fish and Wildlife
Service Fish and Aquatic Conservation Program FY 2016-2020, U. S. Fish and Wildlife
Service, Falls Church, VA.
Ward, E. J., J. H. Anderson, T. J. Beechie, G. R. Pess & J. M. Ford, 2015. Increasing hydrologic
variability threatens depleted anadromous fish populations. Global Change Biology 21:
2500-2509.
Whitney, J. E., R. Al-Chokhachy, D.B. Bunnell, C.A. Caldwell, S.J. Cooke, E.J. Eliason, M.
Rogers, A.J. Lynch & C.P. Paukert, 2016. Physiological basis of climate change impacts
on North American inland fishes. Fisheries 41:332-345.
Xenopoulos, M. A., D. M. Lodge, J. Alcamo, M. Marker, K. Schulze & D. P. Van Vuuren, 2005.
Scenarios of freshwater fish extinctions from climate change and water
withdrawal. Global Change Biology 11: 1557-1564.
Zeigler, M. P., A. S. Todd & C. A. Caldwell, 2012. Evidence of recent climate change within the
historic range of Rio Grande Cutthroat Trout: Implications for management and future
persistence. Transactions of the American Fisheries Society 141: 1045-1059.

8

CHAPTER 1
INFLUENCES OF NATURAL LANDSCAPE FACTORS ON TROPICAL STREAM
ORGANISMS: AN ECOLOGICAL CLASSIFICATION OF HAWAIIAN ISLAND STREAMS
Abstract
Natural landscape features influence stream biology through controls on stream habitat. While
this hierarchical relationship is widely acknowledged as important to the understanding of stream
systems and has been supported by several studies at small spatial extents, regional differences in
natural landscape controls on habitat and complexity due to variation in natural landscape
features over large spatial extents are less well understood. Ecological stream classifications are
useful tools that can describe some of this complexity over large study areas and are especially
valuable when habitat can be inferred from influences of natural landscape data on stream
organisms. In this study, we develop a classification of Hawaiian stream reaches based on
influences of natural landscape features on stream habitat and biota. Our objectives include 1)
identifying a parsimonious set of natural landscape variables that are anticipated to influence
stream habitat and are associated with distributions of stream taxa; and 2) classifying Hawaiian
stream reaches by identifying influences of specific natural landscape variables on distributions
of stream taxa. An initial data reduction process that included use of a forward selection
canonical correspondence analysis (CCA) resulted in a set of 7 natural landscape variables that
have strong associations with distributions of stream taxa in Hawaii, some of which are
commonly identified to influence habitat in many systems and some that are regionally specific.
A conditional inference tree was then used to identify significant influences of these 7 natural
landscape variables on taxa distributions. Results indicate that stream taxa distributions in
Hawaii are well described by differences in elevation, channel slope, groundwater delivery and
rainfall. These results were used to develop 12 reach classes that, when extrapolated using
9

available spatial datasets across the five largest Hawaiian Islands, describe differences in stream
habitat and associated taxa. This research adds to our current understanding of landscape
controls on the aquatic biota of tropical island streams and provides a useful tool for decision
makers who require system-wide understanding of stream habitats at fine spatial scales.
Introduction
Streams are comprised of diverse habitats that vary throughout networks and influence
local biodiversity (Robinson et al., 2002; Ward et al., 2002). Patterns in habitat are in part the
result of natural landscape features, such as geology, topography, and climate, operating across
landscapes drained by streams (e.g., Hynes, 1975; Frissell et al., 1986). Principles of landscape
ecology support this understanding of interrelationships, including the emphasis on habitat
heterogeneity being driven in part by landscape factors as well as hierarchical influences of
landscape factors on stream organisms through effects on habitat (Wiens, 2002; Allan, 2004).
Accounting for such relationships is now considered essential when conducting stream studies
(Fausch et al., 2002), but influences of natural landscape features on stream biota are often
complex and difficult to characterize over large spatial extents. Further, most studies that have
documented influences of landscape features on biota via habitat are limited in spatial extent,
focusing on a few catchments (e.g., Roth et al., 1996; Infante & Allan, 2010) or relatively small
regions (e.g., Richards et al., 1997; Wang et al., 2003; Infante et al., 2006; Dala-Corte et al.,
2016; Paller et al., 2016). Additional research is therefore required to identify regional
differences in landscape effects on stream ecology over large spatial extents.
Stream classifications are analytical tools that can aid in clarifying complex patterns in
stream characteristics, including landscape influences on habitat, across large regions (Melles et
al., 2012). A broad goal of many stream classifications is to describe and categorize physical

10

characteristics of stream habitats or their catchments. Early stream classifications were
developed to describe changes in stream size based on drainage patterns throughout upstream
networks (e.g., Horton et al., 1945; Strahler, 1952). Other more complex classifications have
summarized differences in segments of streams based on characteristics including sinuosity,
geology, and channel slope (e.g., Rosgen, 1994; Montgomery & Buffington, 1997; Brierley &
Fryris, 2000).
Recent increases in availability of landscape-scale datasets as well as improved
geospatial data processing capabilities have supported classifications of stream systems derived
from physical characteristics of stream channels and catchments across large spatial extents.
Snelder & Biggs (2002) applied the River Environment Classification (REC) to all streams of
New Zealand, classifying stream segments based on differences in climate, geology, topography,
and land cover known to have important influences on stream habitat. Parham et al. (2002) used
similarities in size and shape of stream catchments, network stream channel slope, and extent of
bay development at catchment pour points to classify Hawaiian streams and estimate habitat
availability for stream fishes. The above classifications are based on physical characteristics of
stream channels or catchments and effectively summarize potential differences in stream habitats
over large regions, but they lack explicit links between landscape factors, habitat features they
are attempting to characterize, and stream organisms.
Using biological data for stream classifications can yield greater ecological insights by
linking physical characteristics of the channel and catchment to biological patterns across a study
area. In the Pacific Northwest region of the U.S, Higgins et al. (2005) grouped adjacent
catchments into large spatial units (Ecological Drainage Units) based on similarities in natural
landscape features and historical fish distributions within catchments. Seelbach et al. (2006)

11

identified segments of Michigan streams with similar types of catchment surficial geology and
land cover, amounts of groundwater delivery to segments, and presence of fish species, resulting
in identification of ecologically homogenous stream units across the Lower Peninsula. The
aforementioned studies developed classifications by assessing patterns in natural landscape
features and patterns in biology independently, then combining this information to group units
into ecologically similar classes.
An alternative to this approach involves using statistical analyses to group a subset of
stream units into homogenous groupings based on influences of natural landscape variables on
species then using identified relationships to classify all streams within the study area that may
lack species data. Brenden et al. (2008) implemented such an approach in the Lower Peninsula
of Michigan by identifying influences of stream size, channel gradient and modeled water
temperature on stream fishes, then extrapolating this information to reaches lacking fish data to
develop an ecological stream classification (Brenden et al., 2008). This type of classification
approach allows for greater understanding of the natural ability of a stream to support aquatic
biota, or its ecological potential. In addition, it allows for identification of influences of natural
landscape features on stream biota via habitat independent of a stream unit’s position within the
study area, thereby identifying influences that may have been masked if streams were first
clustered into groups (Leathwick et al. 2011). Stream classifications that describe ecological
potential using identified influences of natural landscape features on stream organisms are
grounded in current ecological theory, can provide insight into the relative importance of
individual natural landscape features in shaping local biodiversity, and are particularly useful for
classifying streams across large areas of interest where biological data are not available within all
stream units.

12

Tropical island streams are historically understudied systems that support unique
biological assemblages. Assemblages often include endemic species that require surface
hydrological connections between freshwater and marine habitats to complete amphidromous life
histories (Fitzsimons, 2002; Kikkert et al., 2009; Jenkins et al., 2010; Bauer, 2013). Differing
habitat requirements and migratory abilities for each species combined with variation in natural
landscape features of catchments and stream channels control differences in distributions of
species throughout individual river networks. In many tropical island systems, species are
threatened by habitat degradation from agricultural and urban land use, which results in
pollution, alterations to stream temperature regimes, and sedimentation, as well as habitat loss
due to extensive diversions or alterations of stream flow (Brasher et al., 2003; Smith et al., 2003;
Jenkins et al., 2010). In addition, current climate change projections for the tropics indicate
regionally specific shifts in rainfall, with some regions experiencing wetter conditions and others
experiencing reductions in rainfall (IPCC, 2013). Therefore, many tropical island streams may
be subject to reductions in rainfall, in some cases resulting in reductions in stream flow. Given
the severity of current and future threats, greater understanding of differences in stream habitats
across island systems and their relationships to natural landscape features, including rainfall, can
be useful in prioritizing conservation action and understanding differences in habitat sensitivity
to change.
The goal of this study is to develop a classification of Hawaiian stream reaches based on
influences of natural landscape features on stream organisms. To meet this goal, our first
objective is to identify a parsimonious set of natural landscape variables that are anticipated to
influence stream habitat, are available across the entire study area, and that are strongly
associated with distributions of stream organisms. Our second objective is to classify Hawaiian

13

stream reaches by identifying influences of specific natural landscape variables on distributions
of stream taxa, and then extrapolate results across the study area using available spatial datasets.
The resulting stream classes will highlight the ecological potential of stream reaches across
Hawaii while allowing for the identification of rare and common habitats associated with
different taxa, which in turn can help prioritize conservation actions.
Methods
Study region
This study evaluated perennial streams of the five largest Hawaiian Islands: Hawaii,
Molokai, Maui, Oahu, and Kauai (Figure 1.1). The islands increase in age from east to west.
Volcanic activity ceased on Kauai approximately 6 million years ago, while portions of Hawaii
Island currently experience new lava flows. Hydraulic conductivity of underlying geology varies
due to differences in age and direction of lava flows (e.g., vertical dikes resulting from flows can
impound groundwater), leading to substantial differences in groundwater contributions to stream
baseflow, especially during periods of low rainfall (Lau & Mink, 2006, Izuka et al., 2015). Due
to prevailing trade winds, leeward sides of each island are generally drier than windward sides
(Giambelluca et al., 2011), and the majority of perennial streams occur on windward sides of
islands. Together, these landscape factors contribute to diverse stream habitat conditions across
islands, including flow regimes ranging from intermittent streams that only flow during major
precipitation events to perennial streams that have continuous flow throughout the year due to
groundwater inputs.

14

Spatial framework and natural and biological datasets
Spatial units and framework
As the basis for the spatial framework for this study, we modified the 1:24,000 National
Hydrography Dataset (NHD 2008; http://nhd.usgs.gov/) to create a consistently-defined and
ecologically-meaningful set of spatial units for analysis. We began by evaluating all arcs (i.e.,
spatial representations of surface water channels) in the NHD for the five main Hawaiian Islands,
and we excluded those that were classified as ditches or canals; only arcs classified as perennial
or intermittent were used for development of the spatial framework. Second, we modified breaks
in some stream arcs in two ways. In many cases, arcs in the NHD represent ecologically-defined
sections of streams extending between stream origins and confluences; confluences and
confluences; or confluences and lotic water bodies, river mouths, or in some cases, waterfalls.
However, arcs in the NHD could also be defined by non-ecological criteria, including stream arc
intersections with boundaries of USGS Topographic Quadrangle Maps. To eliminate this artifact
of the NHD, we identified arcs defined by topographic line intersections with streams and then
grouped sets of adjacent arcs that would otherwise comprise a single, ecologically-defined unit
(following Wieferich et al., 2015). In addition, waterfalls were underrepresented within the
NHD dataset and were supplemented using the Hawaii Division of Aquatic Resources waterfall
layer (http://hawaii.gov/dbedt/gis/waterfalls.htm) to create additional arc breaks. The third step
in modifying the NHD was to divide arcs where they intersected ecological zone boundaries
associated with changes in elevation known to influence stream organisms throughout the
Hawaiian Islands (Parham & Lapp, 2006). This step was taken to ensure that analyses could
most effectively discriminate between groups of stream arcs based on differences in their
ecological potential. Through these modifications, we developed the Hawaii Fish Habitat

15

Partnership (HFHP) stream layer (http://www.fishhabitat.org/). We refer to individual arcs
within this modified hydrography layer as stream reaches (following Wang et al., 2011).
We created catchment boundaries to encompass landscapes that drain directly to stream
reaches in the HFHP stream layer, referred to as local catchments (Wang et al., 2011). We
generated local catchments using the ArcMap extension ArcHydro 9.0. We reconditioned a 10
m DEM from the National Elevation Dataset (NED; http://ned.usgs.gov/) to the HFHP stream
layer while filling sinks across the landscape, then generated a flow direction grid that was
subsequently used to create local catchments for each reach within the spatial framework.
We generated two additional spatial units as part of the spatial framework for our study
region. First, we defined upstream catchments, or the entire upstream area draining to a given
stream reach, by converting the local catchment grid to a polygon layer, then generating a nested
polygon layer that included an individual polygon for each reach’s upstream catchment. Next,
we used the ArcGIS extension Network Analyst
(http://www.esri.com/software/arcgis/extensions/networkanalyst) to generate a matrix that
connected single stream reaches with all upstream reaches to establish directionality and to
aggregate landscape data within the upstream catchment. Finally, we used Network Analyst to
establish downstream connectivity and develop the downstream main channel catchment, which
represents the portion of the stream connecting a given reach to the marine environment.
Natural landscape variables
To characterize important natural landscape feature influences on distributions of stream
organisms in Hawaiian streams, we began our analysis with 82 variables summarized in multiple
spatial scales (Table 1.1). All landscape datasets used to generate variables were publically
available or created for this study. Variables were considered for analysis based on expert

16

opinion and relationships established in previous research (Kido, 2008; Parham et al., 2009).
Categories of natural landscape variables included stream size and channel slope, influences on
migration, soil characteristics, and rainfall.
Stream size and channel slope
We first calculated catchment area for each upstream catchment, a representation of
stream size. Next, we calculated stream reach slope from the change in elevation between the
uppermost point of a stream reach and its terminal end divided by total reach length. Finally, we
calculated downstream main channel slope using the elevation change between the terminal point
of a reach and the point of the downstream main channel where the stream enters the marine
environment.
Influences on migration
We determined the minimum elevation of each stream reach using the HFHP stream
layer and the 10 m DEM to estimate elevation at the terminal point of a given reach. We also
measured the distance inland from the terminal point of stream reaches to the marine
environment along the channel. We estimated maximum waterfall height using information
available in the World Waterfall Database (www.worldwaterfalldatabase.com), available
elevation data, and Google Earth. For each stream reach, we then calculated maximum waterfall
height in the downstream main channel.
Soil characteristics
Characteristics of soil and geology across the study region were represented using
attribute data within the Soil Survey Geographic Database (SSURGO; USDA, 1995). SSURGO
hydrologic soil groupings are classed into four types (Groups A, B, C, D) and are estimates of
relative soil infiltration rates. Group A soils have the highest soil infiltration rates, while Group

17

D soils have the lowest. Group D also may indicate soils that are a short distance to an
impermeable geology layer or that have high groundwater tables. For this study, we reclassified
the hydrologic groupings as 4 (A), 3 (B), 2 (C) and 1 (D). We summarized the area-weighted
value within the local catchment and also found the minimum area-weighted value of any local
catchment upstream of a given reach. The upstream minimum hydrologic soil grouping was
calculated to identify reaches that may drain high elevation groundwater resources impounded by
dikes that are often spatially correlated with soil grouping D in perennial streams (i.e., rocky
outcrops; Lau & Mink, 2006). We summarized average SSURGO soil erodibility, classed as
highly erodible (3), potentially highly erodible (2), and not highly erodible (1), for both local and
upstream catchments. We also calculated percent of wetlands within a local catchment using
data from the Coastal Change Analysis Program (CCAP; NOAA, 2005).
Rainfall
Average monthly rainfall values were available as raster data sets for the years 1992 to
2007 and were developed using a network of rainfall gages across Hawaii at a resolution of 250
m (Frazier et al., 2015). We first attributed monthly and annual rainfall values to local
catchments and aggregated rainfall values within the upstream catchment for each reach. For
local catchments that extended above 1830 m in elevation, we calculated area weighted averages
only within the area of local catchments below 1830 m and also excluded any local catchments
located completely above 1830 m (Strauch et al., 2015). This step was taken to account for the
presence of the tradewind inversion layer (TWI), which results in extremely dry conditions at
high elevations on Hawaii Island and Maui where potential evapotranspiration exceeds rainfall
(Ehlmann et al, 2005; Erasmus, 1986). Seventy rainfall metrics (Table 1.1) were then calculated
within local and upstream catchments that described monthly, wet season (May through

18

October), dry season (November through April) and annual rainfall average amounts; variability
in rainfall across months and years; and ratios between wet and dry season rainfall.
Stream flow metrics
We downloaded daily stream flow data from the USGS National Water Information
System (http://waterdata.usgs.gov/nwis) for 24 stream gages in the study region that had at least
10 years of continuous flow data from 1992-2007, with no more than five consecutive days of
missing data and that occurred along perennial stream reaches represented in the HFHP stream
layer. We then used the Hydrologic Index Tool (HIT; Henriksen et al., 2006) to generate 26
stream flow metrics selected based on their ability to describe different characteristics of the flow
regime in flashy perennial or runoff streams (Table C1.1.; Olden & Poff, 2003). We also
calculated eight additional metrics summarizing average values and variability in flow during
Hawaii’s wet and dry seasons, resulting in a final set of 34 stream flow metrics describing
magnitude, frequency, timing, duration and rate of change of streamflow.
Biological data
Data characterizing distributions of native stream taxa collected by state researchers from
1992 to 2010 were provided by the Hawaii Division of Aquatic Resources. Presence/absence
data were collected using standardized visual surveys (Higashi & Nishimoto, 2007) from streams
located across the five main Hawaiian Islands, and sampling locations were attributed to reaches
of the HFHP stream layer for analysis. A total of 403 perennial stream reaches were attributed
with data, and in some cases, a single reach was attributed with multiple samples collected at
different times during the study period. In these cases, we characterized taxa presence within a
reach based on a taxa representation in at least one sample (following Steen et al., 2008). The
final dataset used to classify stream reaches characterizes current distributions of nine native

19

stream taxa including five amphidromous fish species, two freshwater shrimp species, a neritid
gastropod species, and two Kuhliidae fish species that periodically access streams from the
nearshore marine environment (Table 1.2). The two Kuhliidae species were included as a single
grouping in our analysis because they were only recently reclassified as two unique species
(McRae et al., 2011).
Identifying variables associated with distributions of stream organisms
We reduced the initial set of 82 natural landscape variables using a series of data
reduction steps specific to variable types to identify a parsimonious set of landscape variables for
use in the classification. Prior to data reduction, all natural landscape variables were log(x+1)
transformed (following Wang et al., 2001).
Identifying rainfall variables correlated with stream flow metrics
We used principal components analysis (PCA) to identify sets of variables that were
similar in the rainfall dataset and to select a parsimonious set of variables that were minimally
redundant. We used the statistical program SPSS to generate PCs from the rainfall variable
correlation matrix and implemented a Varimax rotation to increase interpretability of the PCs.
We selected one highly-loaded rainfall variable from each PC with λ>1 and examined Pearson’s
correlation coefficients with stream flow metrics to identify variables that had strong
relationships (r≥|0.6|) with stream flow characteristics. Those rainfall variables that had the
strongest relationships with flow were used in the reach classification.
Identifying landscape variables important to distributions of stream organisms
To identify a subset of natural landscape variables showing strong associations with
stream taxa, we ran a forward selection canonical correspondence analysis (CCA) with the
program CANOCO. Forward selection CCA is an ordination technique commonly used in

20

ecological studies to identify a subset of factors explaining variation in a set of dependent
variables. In our study, we used forward selection CCA to identify those variables that explained
a significant (p < 0.10) amount of variation in the biological dataset (Wang et al., 2003;
Esselman & Allan, 2010).
Ecological classification of stream reaches
Conditional inference tree analysis
We used conditional inference (CI) trees to develop groupings of stream reaches based on
relationships between natural landscape variables and native stream taxa distributions. CI trees
are a recursive partitioning technique that aims to maximize differences among groups based on
differences in a set of predictor variables. We chose to use CI trees for two reasons. First, CI
trees allow for generating groupings based on relationships between multiple predictor variables
and multiple binary responses (Hothorn et al., 2006). Second, CI trees address two common
criticisms of tree-based analyses, overfitting of trees and the bias towards predictor variables
with many possible breaks, by incorporating a two-step splitting and stopping procedure that
tests for statistical significance at each split (Hothorn et al., 2006).
We used the function “ctree” within the R package “party” to generate the initial CI tree
describing relationships between the natural stream catchment landscape variables and the taxa
distributions within study reaches. The tree was generated using data from 323 sites, while the
remaining 80 sites were withheld for validation. The alpha value was set to 0.05 with a
Bonferroni adjustment, and the minimum bucket size (the number of reaches that must be
present in a terminal node for a split to occur) was set to 20, equal to 5% of the original dataset.
Following the creation of the initial tree, we identified competitor splits at each node to
identify whether additional significant relationships between natural landscape variables and taxa

21

distributions were present. Examining competitor splits in recursive partitioning analysis can
allow researchers to create more simplistic or meaningful trees and gain a greater understanding
of observed relationships (De'ath & Fabricius, 2000). Competitor splits can suggest the
influence of alternative predictor variables that also have significant relationships with the
response dataset at a given node. In some instances, multiple variables could be identified as
predictors at a single break; ctree defaults to selecting the variable with the lowest p-value. In
our study, examining competitors in the context of current information on Hawaii streams
systems allowed for the potential selection of those that are most ecologically relevant.
Validation of conditional inference tree results
To evaluate relationships between natural landscape variables and the biological dataset
identified by the CI tree, we calculated area under the curve (AUC) estimates for each reach
class. AUC can be used with binary data to select among a set of competing models or to
validate the effectiveness of the model using a withheld dataset and has been used to examine the
predictive capabilities of conditional inference trees and other recursive partitioning techniques
(e.g., Zipkin et al., 2012; Blank & Blaustein, 2014). Individual AUC scores were calculated for
each reach class using the R program ROCR (Sing et al., 2005).
Finalizing reach classes and examining taxa associations
We used the output of the CI tree as the basis for development of an ecological stream
classification of the 4,732 perennial stream reaches in Hawaii. In addition to the CI tree results,
three additional reach classes were created to account for established ecological relationships
between natural landscape variables and distributions of native stream taxa. Stream taxa
associated with each reach class were identified by comparing the percent taxa occurrence within
a given class with their percent occurrence across all sampled reaches (Liu et al., 2005). Taxa

22

were considered associated with a particular reach class when percent occurrence was greater in
the reach class than across all sampled reaches.
Results
Study site description
Many of the natural landscape features were similar in catchments of perennial stream
reaches with biological samples and catchments of all perennial streams of the five main
Hawaiian Islands (Table 1.3). Mean upstream catchment area was similar in reaches with
samples vs. all reaches (15.9 vs. 18.7 km2, respectively), as was downstream main channel slope
(9.81% in reaches with samples vs. 9.09% in all reaches). Average upstream mean annual
rainfall was similar among reaches with and without biological samples (3298 and 3416
mm/year, respectively). The only factors that were substantially lower in sampled reaches vs.
the larger set of perennial reaches included distance inland (3797 vs. 8816 m) and minimum
stream elevation (154 vs. 353 m). These differences reflect the fact that biological samples were
not typically collected in difficult-to-access headwater reaches.
The goby Awaous stamineus was the most common taxa found in Hawaiian streams,
occurring in nearly half of all sampled stream reaches (46%; Table 1.2). Two gobies (Lentipes
concolor and Sicyopterus stimpsoni) and one species of atyid (Atyoida bisulcata) were found in
just over one third of all sampled stream reaches (38, 35 and 36%, respectively). The prawn
Macrobrachium grandimanus and goby Stenogobius hawaiiensis were least common and were
found in only 5 and 6% of sampled stream reaches, respectively. The amphidromous neritid
snail, Neritina granosa, was found in 24% of sampled reaches. All taxa were observed in
streams on Maui, Hawaii, Oahu, and Kauai, while only five of the nine were sampled in Molokai
streams (Table 1.2).
23

Identifying variables associated with distributions of stream organisms
The first eight principal components (PCs) of the rainfall dataset had Îť>1 and accounted
for 93.8% of the total variance (Table C1.2). Variables describing magnitude and variability of
annual, seasonal, and monthly rainfall summarized in the local catchment had the highest
loadings on PC1. Rainfall variables with high loadings on PC2 included total annual rainfall in
the upstream catchment along with other upstream catchment summaries of annual, seasonal, and
monthly rainfall. Variables describing variability in rainfall in the wet and the dry season across
years had high loadings on PC3 and PC4, respectively. Variables describing annual variability in
rainfall during February in the local and upstream catchment were strongly weighted on PC5.
Variables most strongly weighting PCs 6 through 8 included annual variability in rainfall during
August, April and March, respectively. We selected one rainfall variable from each of the eight
PCs with a high individual loading for use in the correlation analysis with stream flow metrics
(Table C1.3.).
Five flow metrics (annual runoff, wet season runoff, maximum monthly flow for August,
maximum monthly runoff during the dry season, and maximum monthly runoff during the wet
season) had strong correlations (r ≥ |0.60|) with at least one of the rainfall variables identified
through the PCA (Table C1.3.). The variable selected from PC2, upstream mean annual rainfall,
had strong positive correlations with all five metrics. Maximum monthly dry season runoff also
had a strong negative correlation with yearly variability in August rainfall. We included only
upstream mean annual rainfall in the set of natural landscape variables for further consideration
given its strong relationship with multiple metrics describing stream flow magnitude and a lack
of additional unique relationships between stream flow metrics and other rainfall variables.

24

Overall, 26% of the total variance in the biological dataset was explained by the 11
landscape variables included in the forward selection CCA. Of the explained variance, 94% was
attributed to 7 landscape variables: upstream catchment area, local reach slope, downstream
main channel slope, reach elevation, maximum waterfall height in the downstream main channel,
minimum hydrological soil grouping, and mean annual rainfall (Table 1.1). Local and upstream
soil erodibility, local hydrologic soil grouping and distance inland did not explain a significant
amount of variance and were not used in the generation of the stream reach classification.
Ecological classification of stream reaches
Conditional inference tree
The CI tree resulted in nine classes of stream reaches (A-I) defined by relationships
between natural landscape variables and the biological dataset (Figure 1.2; Table 1.4). Minimum
elevation had the strongest association with taxa distributions at three nodes (values at which the
CI tree splits into two subsequent groups of reaches) within the CI tree, generating splits within
the dataset at 22, 76, and 231 m (nodes I, II, and VII, respectively). Downstream main channel
slope had the strongest relationship with the biological dataset in low elevation reaches (22 > X ≤
76 m). Stream reaches above 76 m in elevation and with minimum hydrologic soil grouping
values greater than 1 (i.e., excluding soil grouping D) were grouped into reach classes delineated
by differences in upstream mean annual rainfall. In a single instance, a competitor variable
(upstream mean annual rainfall; Node VIII) was chosen to replace the original split variable
(local slope).
Validation of conditional inference tree results
Overall, AUC values suggest the analysis was effective at delineating reaches into classes
based on relationships between taxa distributions and landscape variables (Table 1.4). AUC

25

values at or below 0.5 would suggest the analysis was ineffective in delineating reach classes,
poor if less than 0.69, fair to good if between 0.70 and 0.89, and excellent if greater than 0.90
(Swets, 1988; Blank & Blaustein, 2014). The AUC values generated from the validation dataset
ranged from fair (0.71) to very good (0.94) and had an average value of 0.84.
Final stream reach classes and associated taxa
Stream reach classes J, L and K were based on known ecological information of
Hawaiian streams. First, reach class J was created for reaches with terminal waterfalls (falling
directly to the ocean). Reaches with terminal falls are known to exclude all non-climbing
organisms and to limit moderate climbers (Nishimoto & Fitzsimons, 2006) but were not well
represented by sites sampled for biology. Two hundred and sixty reaches with terminal falls
from reach classes A-E were grouped into reach class J. We also created an additional reach
class (L) for reaches with a minimum elevation greater than 750 m. At very high elevations,
stream taxa considered in this study are not often observed, likely due to limitations on upstream
migration (Polhemus et al., 1992). Finally, from a subset of reaches classed as L, we created a
final reach class (K) to represent high elevation bogs, a unique habitat of interest to local
managers. High elevation bogs are found in several mountainous regions of Hawaii with high
rainfall (Polhemus, personal communication). Class L reaches with annual rainfall values
greater than 2500 mm and with more than 5% of a local catchment classified as wetlands were
changed to reach class K. Six hundred and forty-four reaches from classes G, H, I and F were
grouped into reach class L, while 168 from reach were grouped into reach class K.
The number of taxa associated with each reach class varied from 7 (reach class A) to 0
(G; Table 1.4). The most common taxa associated with reach classes were the atyid A. bisulcata
(E, F, H, I, J) and the goby Awaous stamineus (A, B, C, D, E). Two goby species with limited

26

climbing abilities, S. hawaiiensis and E. sandwicensis, were associated with only a single reach
class (A). The goby species L. concolor and S. stimpsoni were associated with reaches with high
downstream main channel slopes. In general, number of taxa associated with a given reach class
declined with increasing elevation and with lower mean annual rainfall.
Representation of reach classes across islands
Our results suggest that a diversity of stream habitats exists across the Hawaiian Islands
(Figure 1.3). On the island of Oahu, low-gradient, low-elevation stream habitat is common, and
18.7% of total stream perennial stream length is classified as class A (Table 1.5). Reach class C
is more prominent on the islands of Hawaii, Maui, and Molokai, while reach class B is more
common on Kauai and Oahu. On the island of Maui, the majority of streams present on the
eastern half of the island above 76 m in elevation are classes H and I, while the western half is
dominated by reach classes D, E, and F. Reach classes G and L were most common across the
entire study area (17.0% of total stream length). Kauai is dominated by reach class F in
catchments that extend to high elevations near the center of the island. Headwater streams (reach
class L) are found across all islands and make up the highest proportion of perennial stream
reaches on Maui (23.7% of total stream length). Together, these results present a holistic view of
habitat diversity across Hawaiian Islands at a fine spatial scale.
Discussion
In this study, we developed a classification of Hawaiian perennial stream reaches based
on identified influences of natural landscape features on distributions of stream taxa. Natural
landscape variables identified as having significant associations with stream taxa distributions
included several known to influence aquatic biota in many stream systems. These included
rainfall, groundwater delivery and channel slope. Further, others are more commonly associated

27

with changes in taxa distributions in regions similar to Hawaii. Elevation was one such variable,
and results of our classification support current understanding of Hawaiian stream ecology and
indicate that reach elevation strongly controls taxa distributions in Hawaii. Besides
identification of important influences on organisms, the specific landscape variable split values
identified within the CI tree allowed for generation of unique reach classes. Because these
classes reflected influences of landscape on biology, they are suggestive of not only potential
differences in habitat across reaches but they also provide insight into unique taxa assemblages.
When applied across the entire study area, the classes provide a representation of ecological
potential for all stream reaches in Hawaii and allow for identification of common and rare
habitats at the reach scale, which has particular relevance for conservation prioritization
decisions in Hawaii.
Landscape variables strongly associated with stream taxa distributions
Several of the natural landscape variables identified through the CCA that have strong
associations with the biological dataset, including local channel slope (e.g., Walters et al., 2003;
Infante et al., 2006; Maret et al., 2007), mean annual rainfall (e.g., Derolph et al., 2014), and
hydrologic soil grouping (Stauffer et al., 2000; Brewer et al., 2007), have been shown to
influence biology in streams in other regions. However, some natural landscape variables that
had strong associations with the biological dataset are more regionally specific. Elevation and
downstream maximum waterfall height are associated with differences in species composition in
tropical island streams and have been used to broadly describe patterns in species assemblages
(Polhemus et al., 1992; Parham & Lapp, 2006). In addition, in streams of New Zealand where
migratory species are also common, differences in fish and macroinvertebrate distributions were
linked to differences in average downstream main channel slope (Leathwick et al., 2011).

28

Together, these results suggest that commonalities in the hierarchical effects of landscape on
biology exist across many stream systems, but regionally-specific studies can provide further
insight into understanding of a streams ability to support aquatic biota.
Ecological classification of stream reaches
Influences of landscape variables on stream taxa distributions
The results of the CI tree yielded richer insights into the influence of landscape features
on stream taxa distributions. Among the natural landscape variables examined in this study, the
elevation at the terminal point of a given reach had the strongest influence on taxa distributions
across all sampled reaches. In addition, the individual break values associated with two of the
three elevation splits (22 m and 231 m) are similar to those used to broadly classify Hawaiian
streams into biological zones (20 m and 200 m; Parham & Lapp, 2006), linking the results of the
classification to current understanding of taxa distributions in Hawaii. The presence of an
additional elevation split (76 m) suggests that our results further define the influence of elevation
on taxa distributions in Hawaii, separating climbing taxa from non-climbing. The importance of
elevation in defining stream reach classes is likely the result of differing abilities of individual
taxa to ascend step reaches (Polhemus et al., 1992; Nishimoto & Fitzsimons, 2006) and is
reflected by the taxa associated with each reach class. For instance, the association of L.
concolor and A. bisulcata with highest elevation reach classes is likely due to their abilities to
traverse most natural barriers in Hawaii. Similarly, the association of nearshore and nonclimbing taxa with low elevation classes suggests that with increasing elevation the location of a
single migratory barrier along the stream network prevents further upstream migration.
We also found that reach classes with higher channel slopes were associated with a
greater number of taxa than reach classes with lower slopes. Channel slope is known to be a

29

primary factor in determining stream power, or a stream’s ability to move material, and relatively
higher gradient reaches can therefore have larger substrate than lower gradient reaches
(Knighton, 1998). In streams of Michigan, higher channel slope was also associated with greater
habitat complexity independent of its effect on stream power, which in turn had positive
influences on species richness (Infante & Allan, 2010). In our study, the association of certain
stream taxa with higher sloped stream classes at low elevations may be related to an increase in
foraging opportunities. Large substrate composition in reaches with higher slopes may benefit
herbivorous stream taxa (N. granosa, S. stimpsoni) that feed by scraping diatoms and algae from
surface of cobble and boulder (Fitzsimons et al., 2007). In addition, reaches with higher slopes
may also have increased stream velocity, resulting in increased delivery of fine particulate
organic matter (FPOM) to filter feeders (A. bisulcata; Couret, 1976). Preferential habitat
selection may also be reflected in taxa associations in low gradient reach classes. For instance,
despite moderate climbing abilities, A. stamineus is associated with low elevation and low
gradient reaches. This may be due to a habitat preference for deep and slow pools where A.
stamineus forages for macroinvertebrates by burrowing within fine sediments, a unique feeding
strategy not practiced by other Hawaiian gobies (Kinzie & Ford, 1982; Kinzie, 1988).
Our results also indicate that the amount of rainfall that occurs within a catchment
influences the habitat of Hawaiian stream reaches. This is not surprising, as mean annual rainfall
was positively correlated with measures of stream flow magnitude in this study and likely
reflects broad differences in total available habitat and the likelihood that a stream experiences
very low flow or drying events. However, the CI tree results also indicate that effects of rainfall
on taxa distributions are most important in systems with less potential groundwater delivery (as
indicated by upstream minimum hydrologic soil grouping). This result has particular relevance

30

for understanding differences in the sensitivity of Hawaiian stream reaches to climate change and
also highlights the benefits of implementing approaches that allow for the consideration of how
landscape variables predict ecological potential.
Inventory of classes across the Hawaiian Islands
Differences in the prevalence and distribution of reach classes within and across islands
emphasize the importance of considering reach to reach variation in ecological potential when
assessing freshwater resources in Hawaii. On each island, changes in reach class and subsequent
declines in associated taxa occur with increasing elevation and are represented by the
classification. A higher percentage of lower elevation, lower gradient reach classes on Oahu and
Kauai indicates that the majority of habitat associated with nonclimbing and nearshore taxa
occur on these islands. Reaches at moderate to high elevation on the Hilo/Hamakua coast of
Hawaii Island as well as eastern Maui consist predominantly of classes more dependent on
rainfall to support stream flow, suggesting that stream habitat in these regions may be more
vulnerable to declines in mean annual rainfall that may occur with climate change. The overall
change in dominant reach classes at moderate to high elevations from east to west along the
island chain is indicative of changes in groundwater delivery. Streams on older islands with
greater channel incision are more likely to be influenced by dike impounded groundwater that
contributes to baseflow (Craig, 2003; Lau & Mink, 2006; Izuka et al., 2015).
The extrapolation of the classification to all reaches across the study area allows for the
consideration of how additional biological surveys or spatial data layer development can help
further delineate unique classes and increase our understanding of Hawaiian stream ecology. For
instance, additional breaks within reach class D associated with waterfall height or increases in
elevation may distinguish additional differences in taxa distributions, but high elevation reaches

31

are underrepresented in the biological dataset due to limited access and higher cost of sampling.
Similarly, increased biological sampling in low elevation streams with limited groundwater input
across a gradient in rainfall may also indicate whether differences in taxa associations observed
in classes at moderate to high elevations (I, H, and G) are mirrored in some lower elevation
reaches (C and J). In addition, while the relationships between upstream minimum hydrologic
soil grouping and taxa distributions are related to differences in the influence of groundwater on
baseflow and may be better represented with additional data that represents groundwater
resources not captured by minimum hydrologic soil grouping (i.e., perched groundwater; Izuka
et al., 2015).
Utility of the analytical approach
We conducted the Hawaii stream reach classification using an approach that considered
the natural landscape characteristics of an individual reach regardless of its location within the
study area. This place-independent approach can delineate differences among reaches that may
not be represented by classification approaches that begin by clustering rivers in large or multiple
catchments (Leathwick et al., 2011). While a potential limitation of place-independent
approaches is the lack of consideration of directional connectivity of stream systems, our study
accounted for this directional relationship by including natural landscape variables summarized
in upstream catchments.
We acknowledge that human disturbance is likely to have some influence on the
distribution of stream organisms in Hawaii. However, in similar ecological classifications, the
influence of disturbance is acknowledged, but the potential of streams to support taxa is
considered in the context of just natural landscape factors (i.e., Brenden et al., 2008). In
addition, we assert that the potential bias from landscape disturbance is limited within the

32

biological distributions represented by our data. This results from two factors. First, a recently
completed Habitat Condition Index of Hawaii indicates that 80% of sampled reaches used to
generate the classification have very low to moderate risk of disturbance from current
anthropogenic sources (Tsang et al., in preparation), suggesting that the influence of disturbance
on the presence of native taxa is limited within the dataset. In addition, the amphidromous life
history and the presumed lack of natal stream homing (Chubb et al., 1998; McDowall, 2003;
Bebler & Foltz, 2004) of native stream taxa results in an open population in which individuals
are likely to migrate into both “source” and “sink” stream reaches (McRae, 2007). This
continuous influx of new individuals suggests that even in degraded stream reaches, native taxa
may be represented in presence/absence datasets.
Taxa associated with each reach class offer a useful biological characterization of a given
reach. For instance, the lack of a taxa association with a reach class, such as the absence of nonclimbing taxa in reaches above 76 m in elevation, is a strong indication that taxa are not capable
of accessing areas above this elevation and can be accepted with confidence. However, in other
instances the lack of an association may be the result of low occurrence of the taxa overall in the
biological dataset. For example, S. hawaiiensis and E. sandwicensis are not associated with
reach class B, despite substantial overlap in habitat preference of these two taxa with A.
stamineus (McRae et al., 2013). Therefore, taxa association for a specific reach class should be
considered in the context of available ecological knowledge.
Utility for conservation
Classifications are most useful when they incorporate and add to current understanding of
ecology and are built to address specific objectives. In Hawaii, endemic shrimp, fish and snails
are threatened by increasing human landscape disturbance, invasive species and climate change

33

(Brasher, 2003; Smith et al., 2003; Walter et al., 2012). The stream reach classification
developed in this study can be used to account for natural variation in stream reaches to increase
understanding of ecological patterns and responses to threats, which can in turn aid in
conservation decision making. For example, the results of the stream classification allow for the
visualization of differences in habitat across the islands, which can be used in spatial
prioritization analyses that identify areas of conservation importance based on unique habitats
and associated biotic assemblages. The classification can also be paired with human disturbance
datasets and habitat condition indices to assess which habitats and taxa are at particular risk from
human disturbance. A similar approach can be used to assess Hawaiian stream vulnerability to
climate change by utilizing downscaled projected climate change data to assess how streams may
experience shifts in reach class under the influence of climate change and variability. Beyond
assessment of disturbance or change within and among reach classes, finer-resolution biological
data (i.e., abundance data, disease prevalence, growth rates of individuals in different habitats)
can be examined within or across classes associated with particular taxa to assess how
disturbances may affect stream populations. This information can then be used to improve
understanding of the potential effects of climate change on stream organisms, for example in the
case of reduced or more variable streamflow.
Conclusion
The results of our study suggest that natural landscape features of Hawaiian catchments
and stream channels are influential to the distribution of stream taxa, and that these relationships
can be used to understand the ecological potential of Hawaiian streams. The selection of some
variables that are known to influence species distributions in many regions as well as the
selection of some variables that are more regionally specific suggest that continued research on

34

the relationships between the landscape, habitat and biota in understudied regions is required to
fully characterize differences in habitat across large spatial extents.
Our application of a quantitative analytical approach that identifies relationships between
the landscape and stream taxa supported the development of a study area-wide classification of
streams that includes reaches where habitat or biological data are not available but a spatial
understanding of stream resources is needed to inform conservation. We chose to conduct our
study in Hawaii to demonstrate the value of such an approach in understudied regions like
tropical islands of the Pacific, which support many endemic organisms that are threatened by
disturbance from anthropogenic land use and climate change. However, we believe the
development of stream classifications using such an approach is beneficial for informed
conservation decision making in any system where variation in and interactions among natural
landscape features have inherently complex and poorly understood relationships with stream
habitat and biota.

35

APPENDICES

36

APPENDIX 1.A

Tables

37

Table 1.1. Landscape variables considered for use in the Hawaiian stream classification.
Variables were summarized at three spatial scales: local (L), upstream (U) and downstream main
channel (D) catchments. Individual variables were summarized at multiple spatial extents,
resulting in 82 natural landscape variables. Asterisks (*) indicate the eleven natural landscape
variables included in the CCA. Two asterisks (**) indicate the variables that explained a
significant amount of variation within the biological dataset.
Category Code
Variable description
Stream size and channel slope

Units

Spatial
scales

AREAKM
SLOPE
MC_SL
Influences on migration
MIN_ELE
DIST_IN
DROP_WF
EROD
Soil characteristics
HY
MIN_HY

Upstream catchment area
Reach slope
Main channel slope

km2
%
%

U**
L**
D**

Reach elevation
Reach distance inland
Maximum waterfall height
Soil erodibility

m
m
m
.

L**
L*
D**
U*,L*

Hydrologic soil grouping
Minimum soil grouping

.
.

L*
U**

WETLAND

Wetland land cover

km2

U,L

MAR
D
W
(1-12)

Mean annual rainfall
Mean dry season rainfall
Mean wet season rainfall
Mean monthly rainfall

mm/yr
mm/yr
mm/yr
mm/yr

U**,L
U,L
U,L
U,L

MAR_VAR

C.V.1 of mean annual rainfall

mm/yr U,L

Rainfall

1

D_VAR

C.V. of mean dry season rainfall

mm/yr U,L

W_VAR

C.V.1 of mean wet season rainfall

mm/yr U,L

MM_VAR
(1-12)_VAR
WD_RAT
1
Coefficient of variation

1

Mean C.V. of monthly means
1

C.V. of rainfall for a given month
Ratio of total wet to dry season rainfall

38

mm/yr U,L
mm/yr U,L
mm/yr U,L

Table 1.2. Stream taxa used in the ecological classification of Hawaiian streams
Sampled
Present in sampled reaches
Taxa reaches where
Taxa
Common name code taxa occur (%) Hawaii Maui Molokai Oahu Kauai
Aytoida bisulcata
'opae kala'ole
1
38
Y
Y
Y
Y
Y
Lentipes concolor
o' opu alamo'o
2
35
Y
Y
Y
Y
Y
Sicyopterus stimpsoni
o' opu nopili
3
36
Y
Y
Y
Y
Y
Awaous stamineus
o' opu nakea
4
46
Y
Y
Y
Y
Y
Neritina granosa
hihiwai
5
24
Y
Y
Y
Y
Y
Eleotris sandwicensis
o' opu 'akupa
6
12
Y
Y
N
Y
Y
Stenogobius hawaiiensis
o 'opu nahina
7
6
Y
Y
N
Y
Y
Macrobrachium grandimanus
opae oeha'a
8
5
Y
Y
N
Y
Y
Kuhlia sp.
āholehole
9
13
Y
Y
N
Y
Y

39

Table 1.3. Descriptive statistics for all natural landscape variables summarized for perennial
stream catchments across the study region (n=4,732) and for those with biological samples
(n=403).
90th
10th
Category Variable
Mean Median percentile percentile
All perennial stream reaches
Upstream catchment area (km2)
15.86
2.39
0.20
31.44
Local reach slope (%)
19.01
9.80
1.80
48.50
Downstream main channel slope (%)
9.09
4.40
0.80
15.60
Minimum reach elevation (m)
352.91
236.50
16.00
773.00
Distance inland (m)
Upstream soil erodibility (1-3)
Local soil erodibility (1-3)
Local hydrologic soil grouping (1-4)
Upstream minimum hydrologic soil
grouping (1-4)
Upstream mean annual rainfall
(mm/yr)
Perennial stream reaches with biological samples
Upstream catchment area (km2)
Local reach slope (%)
Downstream main channel slope (%)
Minimum elevation (m)
Distance inland (m)
Upstream soil erodibility (1-3)
Local soil erodibility (1-3)

8816.55
2.37
2.40
1.80
1.50

6505.10
2.38
2.49
2.00
1.10

161.20
1.97
1.65
1.00
1.00

21473.47
2.99
3.00
2.80
2.30

3298.18

3150.00

1600.00

5232.00

18.74
11.05
9.81
154.48
3797.36
2.41
2.40

8.00
7.70
6.10
95.00
2535.15
2.33
2.47

3.00
3.70
2.60
21.00
724.74
2.11
2.03

43.00
22.16
17.36
370.00
8827.81
2.94
3.00

Local hydrologic soil grouping (1-4)

1.92

2.00

1.23

2.72

Upstream minimum hydrologic soil
grouping (1-4)
Upstream mean annual rainfall
(mm/yr)

1.46

1.20

1.00

2.00

3416.39

3429.00

2108.00

5100.00

40

Table 1.4. Descriptions, area under the curve (AUC) for validation datasets, and associated taxa for each stream reach class.
Reach
Validation
class
Description
AUC
Associated taxa
A
Coastal
0.73
S. stimpsoni, A. stamineus, N. granosa,
E. sandwicensis, S. hawaiiensis,
M. grandimanus, Kuhlia sp.
B*

Low gradient downstream channel
at low elevation

N/A

A. stamineus, M. grandimanus

C

High gradient downstream channel
at low elevation

0.78

L. concolor, S. stimpsoni, A. stamineus,
N. granosa, E. sandwicensis

D

Low gradient, moderate to high
elevation, potential high water table
High gradient, moderate elevation,
potential high water table

0.90

A. stamineus, S. stimpsoni

0.91

A. bisulcata, L. concolor, S. stimpsoni,
A. stamineus, N. granosa

High gradient, high elevation,
potential high water table
Moderate to high elevation, low
rainfall
Moderate to high elevation,
moderate rainfall

0.88

A. bisulcata, L. concolor

0.87

A. bisulcata

0.94

A. bisulcata, L. concolor

J

Moderate to high elevation, high
rainfall
Terminal falls at low elevation

N/A

A. bisulcata, L. concolor

K

High elevation bogs

N/A

E
F
G
H
I

0.71

L
Headwater streams
N/A
AUC was not calculated for the validation dataset of reach class B due to low sample size

*

41

Table 1.5. Percent of the total stream length across and within individual islands of each reach
class.
Reach
All
class
islands Hawaii Maui Molokai Oahu Kauai
A
7.4
2.5
3.4
4.5
18.7
9.0
B
5.0
0.6
0.2
3.7
11.5
8.5
C
4.5
6.5
8.3
6.2
1.6
2.0
D
12.5
0.3
2.4
3.7
39.7
17.3
E
3.8
0.7
4.5
12.4
4.7
5.5
F
11.4
3.7
14.4
36.6
8.9
17.0
G
17.0
24.2
8.1
5.2
8.3
18.2
H
6.1
9.2
13.5
2.3
2.7
2.0
I
8.7
17.6
16.2
1.1
2.5
0.0
J
3.3
7.6
2.8
0.9
0.0
0.5
K
3.3
0.0
2.5
19.9
0.0
6.8
L
17.0
27.0
23.7
3.5
1.4
13.3

42

APPENDIX 1.B

Figures

43

Figure 1.1. Study area and locations with biological data.
44

Figure 1.2. Conditional inference tree and resulting characteristics of reach classes (A-I). The relative species occurrence is
represented within each histogram by its taxa value (1-9). Nodes (I-VIII) indicate landscape variables that create significant
splits in species distributions.
45

Figure 1.3. Final classification of Hawaiian stream reaches.
46

APPENDIX 1.C

Supplemental tables

47

Table C1.1. Hydrological metrics calculated from Hydrological Index Tool and seasonal metrics
calculated for the wet and dry seasons (denoted with *). Hydrological Index Tool metric
descriptions follow Olden & Poff (2003).
Category Code
Description
Magnitude
MA10 Range in daily flow (20th/80th) divided by median daily flows
MA26 Coefficient of variation of March flows
MA41 Mean annual runoff
MA46* Average of dry season runoff (May-September)
MA47* Average wet season runoff (November-March)
MA48* Coefficient of variation of flows during dry season
MA49* Coefficient of variation of flows during dry season
ML14
Mean of annual minimum flows
ML16
Median of annual minimum flows
ML17
Baseflow index
ML23* Mean minimum yield in dry season
ML24* Mean minimum yield in wet season
MH8
Mean maximum monthly flow for August
MH14 Median of annual maximum flows
MH28* Mean maximum yield in dry season
MH29* Mean maximum yield in wet season
MH23 High flow volume
Frequency
FL2
Variability in low flood pulse count
FL3
Frequency of low flow spells
FH4
High flood pulse count
FH6
Flood frequency (threshold 3x median flow)
FH7
Flood frequency (threshold 7x median flow)
Duration
DL6
Variability in annual minima of daily discharge (one day)
DL10
Variability in annual minima of daily discharge (90 day)
DL17
Variability in low flow pulse duration
DH13
Means of 30 day maxima of daily discharge
DH16
Variability in high flow pulse duration
DH24
Flood free days
Timing
TA1
Constancy
TA2
Predictability of flow

48

Table C1.1 (cont’d).
Category Code Description
Timing
TH3 Seasonal predictability of non-flooding
Rate of change
RA6 Change of flow (increasing)
RA7 Change of flow (decreasing)
RA9 Coefficient of variation in reversals

49

Table C1.2. The eight principal components (PCs) generated from rainfall variables with Îť>1
explained 93.8% of the total variance. PC loadings ≥ ±0.60 are denoted by an asterisk.
Rainfall variable
PC1
PC2 PC3 PC4 PC5 PC6 PC7
PC8
Variance (%)
53.47 17.35 7.37 6.33 3.60 2.41 1.83
1.45
L_mar
L_m_d
L_mwr
L_m_1
L_m_2
L_m_3
L_m_4
L_m_5
L_m_6
L_m_7
L_m_8
L_m_9
L_m_10
L_m_11
L_m_12
L_mar_var
L_m_d_var
L_m_w_var
L_m_1_var
L_m_2_var
L_m_3_var
L_m_4_var
L_m_5_var
L_m_6_var
L_m_7_var
L_m_8_var
L_m_9_var
L_m_10_var
L_m_11_var
L_m_12_var
L_mm_var
L_d_mm_var
L_mm_w_var

0.92*
0.92*
0.92*
0.86*
0.90*
0.91*
0.92*
0.92*
0.91*
0.91*
0.91*
0.89*
0.88*
0.90*
0.90*
-0.71*
-0.26
-0.59
-0.26
-0.46
-0.42
-0.57
-0.24
-0.11
-0.23
-0.70*
-0.45
-0.51
-0.22
-0.78*
-0.73*
-0.71*
-0.62*

0.37
0.35
0.37
0.41
0.36
0.34
0.30
0.35
0.30
0.33
0.32
0.36
0.40
0.39
0.39
-0.30
-0.18
-0.33
0.39
0.07
-0.42
-0.30
-0.25
-0.07
-0.05
-0.23
-0.16
-0.08
-0.13
-0.24
-0.24
-0.20
-0.20

-0.02 -0.05 -0.04
-0.05 -0.07 -0.03
0.01 -0.04 -0.04
-0.08 -0.08 0.04
-0.06 -0.02 -0.09
0.07 -0.10 -0.05
0.07 -0.12 -0.08
-0.01 -0.13 0.02
-0.09 -0.06 -0.06
0.04 -0.12 -0.02
0.01 -0.09 0.00
-0.09 -0.04 -0.02
-0.19 0.00 -0.07
-0.02 0.07 -0.02
-0.02 -0.02 -0.02
-0.05 0.32 0.31
0.91* 0.06 0.16
0.02 0.39 0.37
0.37 0.68* 0.21
0.35 0.13 0.72*
-0.31 0.22 -0.15
0.35 0.14 0.00
0.74* 0.29 -0.18
0.89* 0.06 0.12
0.81* 0.01 0.36
0.33 0.12 0.19
0.66* -0.04 0.10
0.76* 0.10 0.29
0.15 0.91* 0.05
-0.14 0.03 0.26
0.35 0.32 0.05
0.52 0.19 0.00
0.28 0.56 0.19

50

-0.01
-0.08
0.05
0.15
0.07
0.08
-0.10
-0.06
-0.08
-0.15
-0.16
-0.08
0.06
0.08
0.04
0.27
-0.07
0.22
0.04
-0.18
0.00
-0.21
0.00
0.11
-0.04
0.33
-0.27
-0.04
0.07
0.11
0.16
0.18
-0.05

-0.02
-0.12
0.06
0.18
0.06
0.11
0.03
-0.05
-0.14
-0.11
-0.17
-0.18
-0.07
-0.05
0.06
0.14
0.05
-0.09
0.00
-0.04
0.09
0.51
-0.01
0.24
-0.28
0.00
0.23
0.07
0.01
0.28
0.33
0.26
0.22

-0.03
-0.04
-0.01
0.03
-0.01
0.00
-0.02
-0.04
-0.05
-0.11
-0.04
-0.03
0.00
-0.04
-0.01
0.22
0.09
0.32
-0.19
-0.04
0.66*
0.19
0.23
-0.03
0.14
0.23
-0.25
-0.02
0.12
0.29
0.16
0.12
0.26

Table C1.2. (cont’d)
Rainfall variable
PC1
PC2 PC3 PC4 PC5 PC6 PC7
L_wd_rat
-0.72* -0.24 0.33 0.10 0.02 0.26 0.38
L_wd_rat_std
-0.47 -0.17 0.72* 0.11 0.07 0.12 0.20
U_mar
0.52 0.83* -0.15 -0.05 -0.02 -0.06 -0.08
U_d
0.50 0.81* -0.17 -0.08 0.04 -0.15 -0.16
U_w
0.53 0.83* -0.13 -0.04 -0.06 0.01 0.01
U_1
0.34 0.85* -0.26 -0.08 0.04 0.15 0.14
U_2
0.56 0.78* -0.18 -0.06 -0.10 0.04 0.03
U_3
0.51 0.83* 0.02 -0.10 -0.04 -0.01 0.06
U_4
0.55 0.76* -0.05 -0.15 -0.11 -0.22 -0.02
U_5
0.54 0.78* -0.15 -0.12 0.03 -0.15 -0.13
U_6
0.42 0.81* -0.23 -0.09 0.03 -0.16 -0.13
U_7
0.48 0.78* 0.00 -0.14 0.07 -0.28 -0.15
U_8
0.52 0.77* -0.06 -0.08 0.11 -0.21 -0.24
U_9
0.49 0.80* -0.23 -0.04 0.05 -0.10 -0.19
U_10
0.46 0.79* -0.38 0.00 -0.04 0.04 -0.11
U_11
0.57 0.77* -0.16 0.11 -0.04 0.06 -0.11
U_12
0.47 0.84* -0.19 -0.01 -0.06 0.04 0.01
U_mar_var
-0.28 -0.53 -0.00 0.42 0.28 0.49 0.21
U_d_var
0.24 -0.15 0.92* 0.11 0.03 0.02 -0.03
U_w_var
-0.22 -0.52 0.16 0.47 0.40 0.34 -0.12
U_1_var
0.06
0.38 0.37 0.73* 0.15 0.14 -0.01
U_2_var
-0.08
0.06 0.42 0.23 0.77* 0.10 -0.08
U_3_var
-0.28 -0.63* -0.25 0.19 -0.12 0.00 0.04
U_4_var
-0.26 -0.44 0.42 0.09 -0.21 -0.03 0.63*
U_5_var
0.07 -0.30 *0.74 0.30 -0.32 0.11 -0.02
U_6_var
0.29
0.06 *0.83 0.05 0.02 0.19 0.18
U_7_var
0.31
0.04 *0.78 0.04 0.28 0.05 -0.31
U_8_var
-0.10 -0.17 0.30 0.30 -0.01 0.78* -0.08
U_9_var
0.01 -0.11 *0.78 0.11 -0.11 -0.03 0.34
U_10_var
-0.09 -0.12 *0.87 0.14 0.20 0.21 0.06
U_11_var
0.05 -0.22 0.15 0.93 -0.03 0.08 -0.01
U_12_var
-0.51 -0.47 -0.15 0.12 0.32 0.27 0.27
U_mm_var
-0.21 -0.49 0.52 0.46 -0.12 0.33 0.30
U_mm_d_var
-0.13 -0.34 0.67* 0.25 -0.18 0.44 0.26
U_mm_w_var
-0.16 -0.34 0.38 0.74* 0.09 0.11 0.18
U_wd_rat
-0.23 -0.58 0.43 0.13 -0.21 0.35 0.36
U_wd_rat_std
0.28 -0.01 0.86* 0.20 -0.02 -0.08 -0.25

51

PC8
0.10
0.14
-0.06
-0.06
-0.06
-0.02
-0.03
-0.09
-0.05
-0.05
-0.06
-0.14
-0.04
-0.03
-0.03
-0.08
-0.04
0.12
-0.07
0.23
-0.22
-0.11
0.59
0.01
0.03
-0.24
0.00
-0.03
-0.31
-0.08
0.07
0.35
0.05
-0.09
0.16
0.05
0.00

Table C1.3. Correlations between selected stream flow metrics and rainfall variables. Highly
correlated variables (r ≥ ±0.60) are denoted by an asterisk (*).
Flow
metric U_mar L_mar U_d_var L_11_var L_2_var U_8_var U_4_var L_3_var
MA10
0.17
0.30
0.50
0.16
-0.35
-0.46
0.33
-0.06
MA26
0.36
0.32
0.36
0.06
-0.38
-0.44
0.24
-0.29
MA41
0.66*
0.53
0.20
-0.08
-0.14
-0.56
0.46
-0.49
MA46
0.57
0.43
0.26
-0.02
0.00
-0.47
0.47
-0.46
MA47
0.68*
0.51
0.21
0.02
-0.05
-0.41
0.47
-0.47
MA48
-0.12
0.09
0.55
0.21
-0.11
-0.26
0.54
0.16
MA49
0.03
0.15
0.57
0.10
-0.22
-0.24
0.20
-0.18
ML14
-0.04
-0.17
-0.34
-0.23
0.26
0.32
-0.17
-0.11
ML16
-0.04
-0.18
-0.35
-0.22
0.27
0.31
-0.17
-0.11
ML17
-0.16
-0.21
-0.39
-0.21
0.32
0.38
-0.24
0.03
ML23
0.33
0.09
-0.18
-0.16
0.29
-0.07
0.15
-0.27
ML24
0.27
0.02
-0.27
-0.15
0.31
0.05
0.09
-0.20
MH8
0.68*
0.36
0.21
0.27
0.05
-0.27
0.18
-0.50
MH14
0.18
0.25
0.26
0.12
-0.35
-0.25
0.04
0.01
MH23
0.01
0.16
0.10
0.19
-0.21
-0.08
-0.05
0.25
MH28
0.71*
0.55
0.25
-0.03
-0.20
-0.61*
0.39
-0.50
MH29
0.69*
0.57
0.34
0.07
-0.24
-0.48
0.49
-0.47
FL2
-0.35
-0.31
-0.28
-0.23
0.20
0.34
-0.18
0.15
FL3
0.04
0.23
0.17
-0.13
-0.22
-0.43
0.16
0.17
FH4
0.19
0.27
0.45
0.16
-0.39
-0.44
0.31
-0.09
FH6
0.30
0.20
0.37
0.06
-0.37
-0.47
0.30
-0.30
FH7
0.19
0.25
0.43
0.12
-0.42
-0.45
0.29
-0.12
DL6
-0.33
-0.16
0.00
0.07
-0.15
-0.11
-0.09
0.40
DL10
-0.08
0.13
0.33
0.28
-0.14
-0.16
0.07
0.20
DL17
-0.24
-0.13
-0.36
-0.09
0.29
0.37
-0.17
0.32
DH13
0.24
0.30
0.32
0.18
-0.33
-0.40
0.21
0.00
DH16
-0.44
-0.29
-0.37
-0.05
0.32
0.39
-0.23
0.45
DH24
0.16
0.07
-0.23
-0.22
0.06
-0.09
-0.10
-0.21
TA1
0.34
0.02
-0.13
-0.06
0.23
-0.18
0.15
-0.25
TA3
-0.04
0.08
0.04
0.22
-0.18
0.13
-0.34
0.14
TH3
-0.16
0.01
0.10
0.48
0.02
0.48
0.00
0.26
RA6
0.29
0.25
0.30
-0.01
-0.40
-0.50
0.23
-0.23
RA7
0.15
0.13
0.34
0.07
-0.25
-0.40
0.16
-0.10
RA9
-0.24
-0.01
-0.31
-0.14
-0.01
-0.09
-0.20
0.21

52

LITERATURE CITED

53

LITERATURE CITED

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.
Bauer, R. T., 2013. Amphidromy in shrimps: a life cycle between rivers and the sea. Latin
American Journal of Aquatic Research 41: 633-650.
Bebler, M. H. & D. W. Foltz, 2004. Genetic diversity in Hawaiian stream macroinvertebrates.
Micronesica 37: 119-128.
Blank, L. & L. Blaustein, 2014. A multi-scale analysis of breeding site characteristics of the
endangered fire salamander (Salamandra infraimmaculata) at its extreme southern range
limit. Hydrobiologia 726: 229-244.
Brasher, A. M. D., 2003. Impacts of human disturbances on biotic communities in Hawaiian
streams. Bioscience 53: 1052-1060.
Brenden, T. O., L. Z. Wang & P.W. Seelbach, 2008. A river valley segment classification of
Michigan streams based on fish and physical attributes. Transactions of the American
Fisheries Society 137: 1621-1636.
Brewer, S. K., C.F. Rabeni, S.P. Sowa & G. Annis. 2007. Natural landscape and stream segment
attributes influencing the distribution and relative abundance of riverine smallmouth bass
in Missouri. North American Journal of Fisheries Management 27: 326-341.
Brierley, G. J. & K. Fryirs, 2000. River styles, a geomorphic approach to catchment
characterization: Implications for river rehabilitation in Bega catchment, New South
Wales, Australia. Environmental Management 25: 661-679.
Chubb, A. L., R. M. Zink & J. M. Fitzsimons, 1998. Patterns of mtDNA variation in Hawaiian
freshwater fishes: The phylogeographic consequences of amphidromy. Journal of
Heredity 89: 8-16.
Couret Jr., C. L., 1976. The biology and taxonomy of a freshwater shrimp, Atyoida bisulcata
randall, endemic to the Hawaiian islands, MS Thesis. University of Hawai'i at Hilo, Hilo,
HI.
Craig, D. A., 2003. Geomorphology, development of running water habitats, and evolution of
black flies on Polynesian islands. Bioscience 53: 1079-1093.
Dala-Corte, R. B., X. Giam, J. D. Olden, F. G. Becker, T. D. F. Guimaraes & A.S. Melo, 2016.
Revealing the pathways by which agricultural land-use affects stream fish communities
in South Brazilian grasslands. Freshwater Biology 61: 1921-1934.
54

De'ath, G. & K. E. Fabricius, 2000. Classification and regression trees: A powerful yet simple
technique for ecological data analysis. Ecology 81: 3178-3192.
DeRolph, C. R., S. A. C. Nelson, T. J. Kwak & E. F. Hain, 2015. Predicting fine-scale
distributions of peripheral aquatic species in headwater streams. Ecology and Evolution
5: 152-163.
Ehlmann, B. L., R. E. Arvidson, B. L. Jolliff, S. S. Johnson, B. Ebel, N. Lovenduski, J. D.
Morris, J. A. Byers, N. O. Sinder & R. E. Criss, 2005. Hydrologic and isotopic modeling
of alpine Lake Waiau, Mauna Kea, Hawai'i. Pacific Science 59: 1-15.
Erasmus, D.A., 1986. Meteorological conditions affecting observing quality on Mauna Kea.
Publications of the Astronomical Society of the Pacific 98: 254–259.
Esselman, P. C. & J. D. Allan, 2010. Relative influences of catchment- and reach-scale abiotic
factors on freshwater fish communities in rivers of northeastern Mesoamerica. Ecology of
Freshwater Fish 19: 439-454.
Fausch, K. D., C. E. Torgersen, C. V. Baxter & H. W. Li, 2002. Landscapes to riverscapes:
Bridging the gap between research and conservation of stream fishes. Bioscience 52:
483-498.
Fitzsimons, J. M., J. E. Parham & R. T. Nishimoto, 2002. Similarities in behavioral ecology
among amphidromous and catadromous fishes on the oceanic islands of Hawai'i and
Guam. Environmental Biology of Fishes 65: 123-129.
Fitzsimons, J. M., M. G. McRae & R. T. Nishimoto, 2007. Behavioral ecology of indigenous
stream fishes in Hawai'i. In Evenhuis, N. L. & J. M. Fitzsimons (ed.), Biology of
Hawaiian Stream and Estuaries, Bishop Museum Bulletin in Cultural and Environmental
Studies 3: 11-21.
Frazier, A. G., T. W. Giambelluca, H. F. Diaz & H. L. Needham, 2015. Comparison of
geostatistical approaches to spatially interpolate month-year rainfall for the Hawaiian
Islands. International Journal of Climatology 12: 1459-1470.
Frissell, C. A., W. J. Liss, C. E. Warren & M. D. Hurley, 1986. A hierarchical framework for
stream habitat classification - Viewing streams in a watershed context. Environmental
Management 10: 199-214.
Giambelluca, T. W., J. K. Delay, J. K. Nullet, M. A. Scholl & S. B. Gingerich, 2011. Canopy
water balance of windward and leeward Hawaiian cloud forests on Haleakala, Maui,
Hawai'i. Hydrological Processes 25: 438-447.
Henriksen, J. A., J. Heasley, J. G. Kennen & S. Nieswand, 2006. Users’ Manual for the
Hydroecological Integrity Assessment Process Software (including the New Jersey

55

Assessment Tools), U.S. Geological Survey, Biological Resource Discipline, Open File
Report 2006-1093. U.S. Department of the Interior, Washington, DC.
Higashi, G. R. & R. T. Nishimoto, 2007. The point quadrant method: rapid assessment of
hawaiian streams. In Evenhuis, N. L. & J. M. Fitzsimons (eds.), Biology of Hawaiian
Stream and Estuaries, Bishop Museum Bulletin in Cultural and Environmental Studies 3:
305-312.
Higgins, J. V., M. T. Bryer, M. L. Khoury & T. W. Fitzhugh, 2005. A freshwater classification
approach for biodiversity conservation planning. Conservation Biology 19: 432-445.
Horton, R. E., 1945. Erosional development of streams and their drainage basins - hydrophysical
approach to quantitative morphology. Geological Society of America Bulletin, 56: 275370.
Hothorn, T., K. Hornik & A. Zeileis, 2006. Unbiased recursive partitioning: A conditional
inference framework. Journal of Computational and Graphical Statistics 15: 1-21.
Hynes, N. B., 1975. Edgardo Baldi Memorial Lecture: The stream and its valley. Verhandlungen
Internationale Vereinigung fĂźr Theoretische und Angewandte Limnologie 19: 1-15.
Infante, M. D., M. J. Wiley & P. W. Seelbach, 2006. Relationships among channel shape,
catchment characteristics, and fish in Lower Michigan streams. American Fisheries
Society Symposium 48: 339–357.
Infante, D. M. & J. D. Allan, 2010. Response of stream fish assemblages to local-scale habitat as
influenced by landscape: a mechanistic investigation of stream fish assemblages.
American Fisheries Society Symposium 73: 71–397.
IPCC, 2013. Climate Change 2013: The Physical Science Basis. Contribution of Working Group
I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change.
Stocker T. F., D. Qin, G. K. Plattner, M. Tignor, S. K. Allen, J. Boschung, A. Nauels, Y.
Xia, V. Bex & P. M. Midgley (ed.). Cambridge, United Kingdom and New York, NY.
Izuka, S. K., J. A. Engott, M. Bassiouni, A. G. Johnson, L. D. Miller, K. Rotzoll & A. Mair,
2015. Volcanic aquifers of Hawai’i – Hydrogeology, water budgets and conceptual
models, Scientific Investigations Report 2015-5164. U.S. Geological Survey,
Washington, DC.Jackson, J. K. & B. W. Sweeney, 1995. Present status and futuredirections of tropical stream research. Journal of the North American Benthological
Society 14: 5-11.
Jenkins, A. P., S. D. Jupiter, I. Qauqau & J. Atherton, 2010. The importance of ecosystem-based
management for conserving aquatic migratory pathways on tropical high islands: a case
study from Fiji. Aquatic Conservation: Marine and Freshwater Ecosystems 20: 224-238.

56

Kido, M. H., 2008. A persistent species assemblage structure along a Hawaiian stream from
catchment-to-sea. Environmental Biology of Fishes 82: 223-235.
Kikkert, D. A., T. A. Crowl & A. P. Covich, 2009. Upstream migration of amphidromous
shrimps in the Luquillo Experimental Forest, Puerto Rico: temporal patterns and
environmental cues. Journal of the North American Benthological Society 28: 233-246.
Kinzie, R. A., III & J. I. Ford, 1982. Population biology in small Hawaiian streams, Technical
Report 147. Water Resources Research Center, Honolulu, HI.
Kinzie, R. A., 1988. Habitat utilization by Hawaiian stream fishes with reference to community
structure in oceanic island streams. Environmental Biology of Fishes 22: 179-192.
Knighton, D., 1998. Fluvial forms and processes: A new perspective. Arnold, London, UK.
Lau, L. S. & J. F. Mink, 2006. Hydrology of the Hawaiian Islands. University of Hawaiˈi Press,
Honolulu, HI.
Leathwick, J. R., T. Snelder, W. L. Chadderton, J. Elith, K. Julian & S. Ferrier, 2011. Use of
generalised dissimilarity modelling to improve the biological discrimination of river and
stream classifications. Freshwater Biology 56: 21-38.
Liu, C. R., P. M. Berry, T. P. Dawson & R. G. Pearson, 2005. Selecting thresholds of occurrence
in the prediction of species distributions. Ecography 28: 385-393.
Maret, T. R., C. T. Robinson & G. W. Minshall, 1997. Fish assemblages and environmental
correlates in least-disturbed streams of the upper Snake River basin. Transactions of the
American Fisheries Society 126: 200-216.
McDowall, R. M., 2003. Hawaiian biogeography and the islands' freshwater fish fauna. Journal
of Biogeography 30:703-710.
McRae, M. G., 2007. The potential for source-sink population dynamics in Hawaii’s
amphidromous fishes. In Evenhuis, N. L. & J. M. Fitzsimons (ed.), Biology of Hawaiian
Stream and Estuaries, Bishop Museum Bulletin in Cultural and Environmental Studies 3:
87-98.
McRae, M. G., L. B. McRae & J. M. Fitzsimons, 2011. Habitats Used by Juvenile Flagtails
(Kuhlia spp.; Perciformes: Kuhliidae) on the Island of Hawai'i. Pacific Science 65: 441450.
McRae, M. G., L. B. McRae & J. M. Fitzsimons, 2013. Microhabitat use in an assemblage of
native and introduced fishes in a Hawaiian stream. Pacific Science 67: 205-217.

57

Melles, S. J., N. E. Jones & B. J. Schmidt, 2012. Review of theoretical developments in stream
ecology and their influence on stream classification and conservation
planning. Freshwater Biology 57: 15-434.
Montgomery, D. R. & J. M. Buffington, 1997. Channel-reach morphology in mountain drainage
basins. Geological Society of America Bulletin 109: 596-611.
Nishimoto, R. T. & J. M. Ftizsimons, 2006. Status of native Hawaiian stream fishes, a unique
amphidromous biota. American Fisheries Society Symposium 53: 21-27.
National Oceanic and Atmospheric Administration, Office for Coastal Management, 2005. CCAP high resolution land cover state of Hawaii. U.S. Department of Commerce,
Washington, DC. http://maps.coast.noaa.gov/arcgis/rest/services/CCAP_HR
/Hawaii_HR_2005/MapServer/ Accessed 10.05.01.
Olden, J. D. & N. L. Poff, 2003. Redundancy and the choice of hydrologic indices for
characterizing streamflow regimes. River Research and Applications 19: 101-121.
Paller, M. H., B.A. Prusha, D.E. Fletcher, E. Kosnicki, S. A. Sefick, M. S. Jarrell, S. C. Sterrett,
A. M. Grosse, T. D. Tubervile & J. W. Feminella, 2016. Factors influencing stream fish
species composition and functional properties at multiple spatial scales in the sand hills
of the southeastern United States. Transactions of the American Fisheries Society 145:
545-562.
Parham, J. E., G. R. Higashi, R. T. Nishimoto, S. Hau, D. G. K. Kuamo'o, L. K. Nishiura, T. S.
Sakihara, T. E. Shimoda & T. T. Shindo, 2009. The use of Hawaiian Stream Habitat
Evaluation Procedure to provide biological resource assessment in support of instream
flow standards for East Maui streams. Bishop Museum and the Division of Aquatic
Resources, Honolulu, HI.
Parham, J. E. & E. K. Lapp, 2006. Reach classification for hawaiian streams: Standardized GIS
methodology for determining reaches. Digital product.
Parham, J. E., 2002. Spatial models of Hawaiian streams and stream fish habitats. Doctoral
dissertation. Louisiana State University, Baton Rouge, LA.
Polhemus, D. A., J. Maciolek & J. Ford, 1992. An ecosystem classification of inland waters for
the tropical Pacific islands. Micronesica 25: 155-173.
Richards, C., R. J. Haro, L. B. Johnson & G. E. Host, 1997. Catchment and reach-scale
properties as indicators of macroinvertebrate species traits. Freshwater Biology 37: 219230.
Robinson, C. T., K. Tockner & J. V. Ward, 2002. The fauna of dynamic riverine
landscapes. Freshwater Biology 47: 661-677.
58

Rosgen, D. L., 1994. A classification of natural rivers. Catena 22: 169-199.
Roth, N. E., J.D. Allan & D. L. Erickson, 1996. Landscape influences on stream biotic integrity
assessed at multiple spatial scales. Landscape Ecology 11: 141-156.
Seelbach, P. W., M. J. Wiley, M. E. Baker & K. E. Wehrly, 2006. Initial classification of river
valley segments across Michigan's Lower Peninsula. American Fisheries Society
Symposium 48: 25-48.
Sing, T., O. Sander, N. Beerenwinkel & T. Lengauer, 2005. ROCR: visualizing classifier
performance in R. Bioinformatics 21: 3940-3941.
Smith, G. C., A. R. Covich & A. M. D. Brasher, 2003. An ecological perspective on the
biodiversity of tropical island streams. Bioscience 53: 1048-1051.
Snelder, T. H. & B. J. F. Biggs, 2002. Multiscale River Environment Classification for water
resources management. Journal of the American Water Resources Association 38: 12251239.
Steen, P. J., T. G. Zorn, P. W. Seelbach & J. S. Schaeffer, 2008. Classification tree models for
predicting distributions of Michigan stream fish from landscape variables. Transactions
of the American Fisheries Society 137: 976-996.
Strahler, A. N., 1952. Dynamic basis of geomorphology. Geological Society of America Bulletin
63: 923-938.
Strauch, A. M., R. A. MacKenzie, C. P. Giardina & G. L. Bruland, 2015. Climate driven changes
to rainfall and streamflow patterns in a model tropical island hydrological system. Journal
of Hydrology 523: 160-169.
Stauffer, J. C., R. M. Goldstein & R. M. Newman, 2000. Relationship of wooded riparian zones
and runoff potential to fish community composition in agricultural streams. Canadian
Journal of Fisheries and Aquatic Sciences 57: 307-316.
Swets, J. A., 1988. Measuring the accuracy of diagnostic systems. Science 240: 1285-1293.
Tsang, Y-P., R. W. Tingley III, J. Hsaio & D. M. Infante, In preparation. Identifying high value
areas for conservation: Accounting for connections among terrestrial, freshwater, and
marine habitats in a tropical island system.
U. S. Department of Agriculture, Natural Resources Conservation Service, 1995. Soil Survey
Geographic (SSURGO) Data Base, Report 1527. U.S. Department of Agriculture,
Washington, DC.

59

Walter, R. P., J. D. Hogan, M. J. Blum, R. B. Gagne, E. F. Hain, J. F. Gilliam & P. B. McIntyre,
2012. Climate change and conservation of endemic amphidromous fishes in Hawaiian
streams. Endangered Species Research 16: 261-272.
Walters, D. M., D. S. Leigh, M. C. Freeman, B. J. Freeman & C. M. Pringle, 2003.
Geomorphology and fish assemblages in a Piedmont River basin, USA. Freshwater
Biology 48: 1950-1970.
Wang, L. Z., J. Lyons & P. Kanehl, 2001. Impacts of urbanization on stream habitat and fish
across multiple spatial scales. Environmental Management 28: 255-266.
Wang, L. Z., J. Lyons, P. Rasmussen, P. Seelbach, T. Simon, M. Wiley, P. Kanehl, E. Baker, S.
Niemela & P. M. Stewart, 2003. Watershed, reach, and riparian influences on stream fish
assemblages in the Northern Lakes and Forest Ecoregion, USA. Canadian Journal of
Fisheries and Aquatic Sciences 60: 491-505.
Wang, L. Z., D. Infante, P. Esselman, A. Cooper, D. Y. Wu, W. Taylor, D. Beard, G. Whelan &
A. Ostroff, 2011. A hierarchical spatial framework and database for the National River
Fish Habitat Condition Assessment. Fisheries 36: 436-449.
Ward, J. V., K. Tockner, D. B. Arscott & C. Claret, 2002. Riverine landscape
diversity. Freshwater Biology 47: 517-539.
Wieferich, D. J., W. M. Daniel & D. M. Infante, 2015. Enhancing the utility of the NHDplus
river coverage: characterizing ecological river reaches for improved management and
summary of information. Fisheries 40: 562-564.
Wiens, J. A., 2002. Riverine landscapes: taking landscape ecology into the water. Freshwater
Biology 47: 501-515.
Zipkin, E. F., E. H. C. Grant & W. F. Fagan, 2012. Evaluating the predictive abilities of
community occupancy models using AUC while accounting for imperfect
detection. Ecological Applications 22: 1962-1972.

60

CHAPTER 2
ANTICIPATING EFFECTS OF CHANGING CLIMATE ON TROPICAL ISLAND
STREAMS: INFLUENCES OF STREAM FLOW ON ATYID MOUNTAIN SHRIMP
Abstract
Climate change will alter stream flow regimes and may have negative effects on
populations of stream organisms. This ultimately threatens population persistence in regions that
experience changes in climate, including tropical islands, which harbor many endemic species.
This study examined associations between flow regimes and population characteristics of the
endemic mountain shrimp, Atyoida bisulcata, found in high elevation streams of Hawaii. We
first identified aspects of the flow regime strongly associated with rainfall across a set of study
streams. Then, we identified relationships between rainfall-sensitive flow metrics and stream
habitat factors. Finally, we characterized relationships between rainfall-sensitive flow metrics
and population characteristics of A. bisulcata. Our results showed that multiple aspects of the
flow regime were strongly associated with mean annual rainfall, including baseflow, flow
variability, and flow event duration and frequency, and these rainfall-sensitive flow metrics were
associated with changes in available habitat type and moss and filamentous algae cover. A.
bisulcata in streams with lower baseflows were smaller and in poorer condition than in streams
with higher baseflows, and individuals in streams with higher flow variability had a greater
prevalence of brown spot disease. Together, our results indicate that declines in annual rainfall
leading to lower baseflow and greater flow variability may negatively affect A. bisulcata.
Further, these findings demonstrate the value of considering relationships between population
characteristics and stream flow metrics as opposed to presence/absence or abundance data as a
way to understand how climate change could affect stream organisms, offering insights that
could be used in support of proactive conservation strategies.
61

Introduction
The stream flow regime is a dominant factor influencing stream organisms (Poff et al.,
1997). Flow regimes include flow magnitude, frequency, duration, timing and rate of change of
events (Poff & Ward, 1989), all of which can differentially affect distributions and abundances
of stream organisms. Reductions in stream flow magnitude and increases in duration or
frequency of low flow events can change physical characteristics of the stream environment
including reducing available habitat, increasing temperature and/or reducing dissolved oxygen.
Such changes can increase stress to and reduce growth of organisms (as reviewed by Rolls et al.,
2012), and in more extreme cases, exclude organisms entirely from streams (e.g., Canton et al.,
1984). Increased frequency of high flow events may also be problematic, resulting in a
temporary loss of species unable to persist during high flows (e.g., Meffe & Minckley, 1987).
For some stream fishes, timing of seasonal high flows controls timing of spawning events (Lytle
& Poff, 2004) and larval hatching (e.g., Naesje et al., 1995). The rate of change of stream flows
also affects organisms, stranding individuals in suboptimal habitats following quick flow
recessions (e.g., Bradford, 1997; Davey et al., 2006) or flushing them downstream after rapid
increases in flow (Cushman, 1985).
Physical characteristics of the catchment (such as geology, topography, and land cover)
as well as regional climate influence stream organisms through indirect controls on the flow
regime (Poff et al., 1997). Examples of climatic controls include the total amount and timing of
precipitation events as well as effects of air temperature on snowmelt and evapotranspiration.
Given the influences of climate on stream flow, changes in precipitation and air temperature
affect the physical habitat of streams and species they support through alterations to the flow
regime. Recent studies have documented such changes in multiple regions. In snowmelt-driven

62

streams of the western United States over the past half century, warmer air temperatures have
caused earlier snowmelts, resulting in increased winter and spring discharges and declines in
summer flows (Leppi et al., 2012; Zeigler et al., 2012), reducing available habitat for native
cutthroat trout (Isaak et al., 2012). In Pacific Northwest streams, more fall and winter rainstorms
over the past 30 years have likely caused greater flow variability during winter months, reducing
natural reproduction of Chinook salmon (Ward et al., 2015). In the Rio Grande, shorter and
smaller snowmelt-driven spring discharges resulting from warmer air temperatures have been
shown to reduce differences in spawning dates of eight stream fish species, potentially increasing
interspecific competition among fish larvae for food (Krabbenhoft et al., 2014). Together, these
studies offer important insights into diverse flow-induced ecological changes resulting from
documented changes in climate.
While the studies above describe climate influences within temperate regions, we are
currently aware of no studies that have documented climate effects on flow regimes and
organisms in tropical island streams. This lack of information is particularly problematic for
several reasons. First, changes in the magnitude and timing of rainfall are likely to occur with
climate change throughout the tropics (IPCC, 2013), and reduced rainfall may alter stream flow
regimes and threaten endemic species that inhabit tropical island streams (Leong et al., 2014).
This is especially concerning given the fact that many tropical island streams require persistent
rainfall for maintenance of baseflow (Craig, 2003). Reductions in rainfall that reduce flows
could change stream habitat by increasing temperatures or lowering dissolved oxygen; in
extreme cases, reductions in flow could lead to complete loss of stream habitat. An additional
cause for concern is that many freshwater shrimp, fish and snails inhabiting tropical islands
require hydrological connectivity to marine environments to complete amphidromous life

63

histories where larvae hatch in stream systems, passively drift downstream to the ocean, spend
weeks to months developing in the marine environment, then use freshwater cues to migrate
back upstream (Keith, 2003; Nishimoto & Fitzsimons, 2006; McDowall, 2007; Bauer, 2013).
Reductions in streamflow may limit drift of larvae to the ocean and may also slow or stop
juvenile migration upstream after their marine stage (Gorbach et al., 2012; Walter et al., 2012;
Leong et al., 2014). Finally, stream flow reductions may change food availability by reducing
available foraging space and by limiting delivery of fine particulate organic matter (FPOM) and
availability of basal food resources (James et al., 2007), reducing individual growth and
condition. Ultimately, changes in stream flow within tropical island systems represent a
significant threat that could lead to loss of species from altered stream habitats.
In response to changes in stream flow and prior to species loss, changes are likely to
occur to individual populations such as size or age structure, average body condition, disease
prevalence, or fecundity. For example, reductions in growth may result from lower habitat and
food availability associated with lower stream flows, which may be reflected in the average size
of individual organisms. Increases in flow variability may increase stress on aquatic organisms,
potentially increasing their vulnerability to disease and infection (Jones et al., 2012). In
temperate streams, researchers have begun to study relationships between flow regimes and
population characteristics (e.g., age structure, fecundity) to understand effects of changing
climate on stream organisms operating through aspects of the flow regime (Krabbenhoft et al.,
2014; Ward et al., 2015). Studies such as these remain uncommon (as reviewed by Lynch et al.,
2016), yet understanding specifically how climate influences populations can provide insight into
the mechanisms that may eventually result in species loss as climate change progresses.

64

The goal of our study is to understand how the flow regime influences population
characteristics of tropical stream organisms. We focus our research on high elevation streams of
Hawaii inhabited by the endemic Atyoida bisulcata, an amphidromous atyid mountain shrimp
found in streams throughout the islands. A. bisulcata are listed as “near threatened” by the
International Union for Conservation of Nature (IUCN; De Grave and Cai, 2013) due to their
extremely limited distribution (restricted to the Hawaiian Islands) as well as their declining
populations resulting from habitat loss and degradation from anthropogenic land use and water
diversions (Brasher, 2003). Despite their conservation status, ecological information on A.
bisulcata is limited to a single thesis (Couret, 1976) and several published reports (e.g.,
Nishimoto & Kauamo’o, 1996; Brasher, 1997; Gingerich & Wolff, 2005), although these studies
provide little information on influences of stream flow regimes on populations. This represents a
critical knowledge gap, as changes in rainfall over the last century have been linked to reductions
in baseflow across Hawaiian streams (Bassiouni & Oki, 2013; Frazier & Giambelluca, 2016) and
projected climate data indicate that regionally specific declines in rainfall will continue as
climate change progresses over the next century (Timm et al., 2015).
Our first objective is to assess characteristics of flow regimes strongly influenced by
rainfall across a set of study streams. Second, we identify relationships between those rainfallsensitive flow metrics and select stream habitat factors to consider how changes in rainfall could
affect habitat. Finally, we characterize relationships between population characteristics of A.
bisulcata and those rainfall-sensitive flow metrics and habitat factors. By comparing streams
with similar landscape characteristics across a wide gradient in annual rainfall, we can make
inferences about how climate may affect stream biota based on identified relationships between
rainfall-sensitive flow metrics and select stream habitat factors. Our results will provide insight

65

into how climate operates through the flow regime to influence populations of tropical island
stream organisms, thereby increasing our understanding of the potential effects of climate change
on these systems.
Methods
Study area
The study area is located along the North Hilo coast of Hawaii Island, where a gradient in
mean annual rainfall ranging from 3000 to 7000 mm/year exists over a linear distance of less
than 30 km (Giambelluca et al., 2013; Figure 2.1). We evaluated 11 first- and second-order
stream reaches lying between 470 and 510 m in elevation and bounded by plunge pools created
by waterfalls at their inflows and outflows (Figure 2.1; Table 2.1). Catchments draining these
reaches have minimal urban and agricultural land use and are composed primarily of mixed
native and non-native forests (Price et al., 2012). The streams drain land on porous volcanic
geology ranging in age from 5,000 to 64,000 years (Sherrod et al., 2008) with soils that are
classified as erodible and are primarily Akaka and Honokaa soil types (USDA, 1995). Recent
research across our study reaches indicates that baseflow magnitude is lower and flow variability
is higher in streams with lower mean annual rainfall, suggesting that flow regimes in these
systems are heavily dependent on rainfall (Strauch et al., 2015). Similarities in landscape
characteristics of stream catchments in our study area and the strong relationships between
stream flow and mean annual rainfall make the North Hilo Coast a useful system to study
linkages between climate, the stream flow regime, and stream organisms.

66

Data collection and processing
Mean annual rainfall
We generated estimates of mean annual rainfall from 1992-2007 for each study reach
using 250-m monthly rainfall raster datasets developed by Frazier et al. (2015). We summed
monthly raster datasets to create annual rainfall estimates for each of 16 years, which were then
attributed to local catchments, or the area draining directly to a stream reach (Tingley, this
volume a). We then summarized these yearly data in the entire landscape in the upstream stream
network draining to the reach (i.e., the upstream catchment; following Wang et al., 2011) using
the aggregation method described in Tsang et al. (2014). Finally, we averaged annual values to
obtain a long-term estimate of mean annual rainfall at the catchment scale for each study reach.
Stream flow
Daily stream flow data
We obtained daily stream flow records for the 2012, 2013, and 2014 water years from
multiple sources. For a single study reach, Honolii, we downloaded daily flow data from the
USGS National Water Information System (https://waterdata.usgs.gov/usa/nwis/16717000). We
averaged daily stream flows from 15-min interval instantaneous flow data collected at gaging
stations installed between spring of 2011 and fall of 2012 in all other stream reaches (see Strauch
et al. (2015) for additional detail). In three study reaches (Manoloa, Waikaumalo and Pahoehoe),
we supplemented gaps in daily stream flow using a Distributed Hydrology Soil Vegetation
Model for the North Hilo coast (DHSVM; Strauch et al., 2016).
Stream flow metrics
We generated nine flow metrics from daily stream flow data describing magnitude,
frequency, duration, timing, and rate of change of stream flow for study reaches (Table 2.2). In

67

addition to representing different aspects of the flow regime, we chose these metrics based on
their suggested relevance or observed relationships to stream biology within published literature.
Two of the metrics describe stream flow magnitude. Oki et al. (2010) found that the 70th
percent exceedance value (Q70) is broadly representative of baseflow conditions in Hawaiian
streams, and therefore we included Q70 normalized by catchment area (hereafter Q70 yield) to
characterize potential differences in baseflow across study reaches. April runoff normalized by
catchment area (AR yield) quantified stream flow magnitude prior to the sampling season, and
coefficient of variation (CV) of daily stream flows characterized flow variability. We
represented the frequency of low flows as the total number of events where daily flow fell below
5% of mean annual flow (FL) and differences in the duration of low flow events using the
minimum of 7-day mean flow normalized by catchment area (DL yield). We considered high
flow periods when daily flows were above 25th percentile exceedance values; the frequency and
duration of these events were also included as metrics for analysis. We represented aspects of
stream flow timing and rate of change using constancy (CON) and rise rate yield (RR yield),
respectively. Constancy is a measure of overall day to day stability in stream flow throughout
the year, meaning a stream with seasonal high flows would have a lower constancy than one with
numerous but regular high flow events throughout the year (Colwell, 1974; Gordon et al., 2004).
We included average rise rate of daily flow normalized by catchment area (RR yield) to
represent differences in the relative rate of change in days with increased flow over the previous
day.
We calculated stream flow metrics for four time periods: the full period of flow record
(2012, 2013 and 2014 water years) and for three seven-month time periods prior to each
sampling season (October-April; Table C2.1.). The seven-month time period was chosen for

68

calculation of flow metrics to allow for a comparative period in which flow data were available
across all years leading up to the beginning of the sampling season. We calculated Q70 yield and
AR yield manually and the other seven flow metrics using the USGS Hydrologic Index Tool
(Henriksen et al., 2006; Table 2.2).
Habitat
Stream habitat data sampling
A transect approach was used to collect habitat data from stream reaches during MayAugust of 2012, 2013, and 2014. Study reach lengths equal to 20 times mean stream width
(MSW) at the time of sampling were established to capture a representative sample of habitat in
each of the 11 reaches (Kido, 2008; Kido, 2013). Transects lines were established once every
MSW over study reaches. Moving upstream, two researchers visually recorded the percent
(within Âą5%) of substrate categories comprising the stream reach including silt, sand, gravel,
cobble, boulder and bedrock (following Higashi & Nishimoto, 2007; Table C2.2). Similarly,
geomorphic units (riffle, run, pool, cascade) within 1 m of the transect line were also recorded
(Higashi & Nishimoto, 2007). Wetted width was measured to the nearest 0.1 m, and maximum
water depth along the transect line was measured to the nearest 0.01 m. In 2013 and 2014, we
also recorded the total percent (Âą 5%) substrate covered by aquatic moss and filamentous algae
within 1 m of the transect line. Values were averaged across transects to assess relative habitat
conditions within each reach.
Stream habitat factors
We calculated six habitat factors from sampled habitat data in each reach each year to
assess potential changes in habitat which may influence food resources and available habitat for
A. bisulcata. Average wetted transect area (Knighton, 1998) was calculated by multiplying

69

transect wetted width by one-half of maximum depth, then averaging these products across
transects. Mean substrate size was calculated by multiplying the average percent substrate
within a reach by geometric mean of particle size range, with bedrock set to 5656 mm (following
Kaufmann & Robison, 1998; Table C2.2). Percent riffles and pools were averaged across all
wetted transects within a study reach to capture differences in geomorphic habitat composition.
In 2013 and 2014, we averaged percent algae and percent moss cover across transects to assess
differences in available food resources including both filamentous algae (Couret, 1976) and
FPOM held within moss (Maurer & Brusven, 1983).
A. bisulcata
A. bisulcata sampling
We sampled Atyoida bisulcata each year when habitats were sampled, approaching each
transect from downstream with a hand seine net (2 mm mesh; 0.91 m wide and 0.45 m deep )
and completing a single pass to capture A. bisulcata in habitats within 1 m of each transect line.
The total numbers of A. bisulcata collected from each transect were recorded, and a subsample
of all individuals (typically 75-125 individuals) was taken for further analysis. In some
instances, transects were excluded from the study if their locations were not accessible (e.g., due
to debris) in any sample year. In 2013 and 2014, we also sampled individuals from the upstream
plunge pool in reaches with few individuals found within transects. At 2 m intervals around the
perimeter of the plunge pool, we seined all A. bisulcata within a 1 m2 area and a subsample of all
captured individuals was considered representative of the population within the stream reach and
taken to the lab for further analysis.

70

Laboratory processing of A. bisulcata
Within 24 hours of collection, we measured individual specimens in the lab. We
measured post-orbital carapace (POC) length, the linear distance from the posterior of the ocular
cavity to the posterior of the carapace, to the nearest 0.1 mm using an ocular micrometer. Sex
was determined based on shape of the endopod of the first pleopod (sinusoidal in males; long and
tapering in females) and the presence of the appendix masculina on the second pleopod of males
(Chace, 1983; Couret, 1976). We recorded small specimens as juveniles if their first pleopod did
not show characteristics associated with either sex. In 2013 and 2014, we inspected individuals
for brown spot disease, which is an infection of the exoskeleton by chitinoclastic bacteria that is
theorized to be more common in A. bisulcata living in extreme or degraded environmental
conditions (Chan, 1978).
On a subset of measured individuals (typically 75-100) from each subsample, we
recorded wet weights for each specimen to the nearest 0.001 g. We dried specimens at 60-105°
C for at least 24 hours, reweighed, then combusted them at 500°C in a muffle furnace for at least
2.5 hours to obtain AFDM (Benke et al., 1999). In Kaawalii Stream in 2013, AFDM
measurements were not available and were derived from dry weights.
A. bisulcata population characteristics
We calculated four population characteristics for each reach each year. We calculated the
first population characteristic, average POC (mm) length, by averaging individual POC length
measurements from males within a subsample. Females were excluded in calculation of POC
(mm) length due to their low representation in samples and evidence of size stratification by sex.
We calculated the second characteristic, the prevalence of brown spot disease, as the percent
occurrence of the disease for subsamples with greater than 30 adult males present. Females and

71

juveniles were excluded from brown spot disease prevalence calculations to avoid disease
detection biases resulting from post-egg release molting in females (Chan, 1978; Nishimoto &
Kuamo’o, 1996) and high molting rates in juveniles (Couret, 1976). The third population
characteristic, average relative mass, was obtained by first generating length-mass relationships
for each reach each year using the linear form of the basic length-mass model:
log10 M = log 𝑎 + 𝑏 log L

[1]

where M is mass (AFDM; g), L is length (POC length; mm) and a and b are constants (Benke et
al., 1999). Geometric means of a and b (am and bm, respectively) were calculated from lengthmass regressions from each reach each year (Froese, 2006; Table C2.3). We calculated the
relative mass (Mrm) of each individual using the formula:
𝑀rm = 100 𝑎

𝑀

𝑏
𝑚𝐿 𝑚

[2]

and average relative mass was then calculated for all males for each reach each year. Froese
(2006) suggests that when distinguishing differences in individual condition across multiple
sample areas, comparing observed relative mass to predicted relative mass from a mean lengthweight relationship is appropriate. Individuals with weights greater than 100% would potentially
indicate those in relatively “better” condition than those with weights that fall well under 100%.
We excluded females and juveniles from averages of relative mass to avoid biases resulting from
differences in body shape (Froese, 2006) and post-egg release declines in relative mass among
females. The final population characteristic, average biomass per m2 of wetted width (g/m2),
was calculated by first finding the proportions of individuals in 0.5 mm POC length-classes in
each reach each year. Abundance counts were multiplied by the size-class proportions to
estimate the number of individuals within each 0.5 mm size-class at each transect (following
Cross et al., 2008). Length-mass regressions were used to predict an average AFDM for each
72

size-class, which was multiplied by the size-class counts and summed to estimate biomass at
each transect. Biomass at each transect was then divided by wetted width and averaged across
transects. If an individual transect was dry at the time of sampling, it was excluded from average
biomass estimates.
Data analysis
Variation in flow metrics across the study region and relationships to rainfall
We calculated descriptive statistics (average, minimum, maximum, standard deviation,
and coefficient of variation) of the nine stream flow metrics across stream reaches for the entire
period of record. In addition to descriptive statistics over the entire period of record, means for
the seven-month flow period each year were generated along with the coefficient of variation of
these means across years to assess relative differences in annual variation of stream flow metrics
within the study region. We assessed Pearson’s correlations between stream flow metrics and
drainage area (as calculated in Tingley, this volume a) to determine if flow metrics co-varied
with stream size, and we also examined correlations between mean annual rainfall and the stream
flow metrics. Prior to analysis, flow metrics were transformed using natural log (x+1).
Association between habitat factors and flow metrics
We generated descriptive statistics for the six habitat factors averaged across years for
each stream reach. We examined Pearson correlations between the six habitat factors and nine
stream flow metrics calculated over the entire period of record.
Association between A. bisulcata population characteristics, flow metrics, and habitat factors
We calculated descriptive statistics each year to assess the range and variation in A.
bisulcata population characteristics across the study reaches and across years. When then used a
multi-step approach to assess strength and consistency of relationships between A. bisulcata

73

population characteristics and stream flow metrics across years. First, because we were
interested in the effects of differences in rainfall on A. bisulcata operating through the flow
regime, we selected stream flow metrics that had been shown to have strong (r >|0.60|) and
significant (p < 0.05) correlations with mean annual rainfall for use in the analysis with
population characteristics. In some cases, if selected stream flow metrics were highly redundant
(r > |0.9|), we included a single flow metric. Then, we calculated Pearson’s correlation
coefficients between the selected stream flow metrics summarized over each seven-month flow
period and the A. bisulcata population characteristics independently for each year that detailed
biological data were collected (2012, 2013, 2014).
Results
Variation in flow metrics across the study region and relationships with rainfall
Most stream flow metrics varied across the study region and some had high measures of
variability (as indicated by coefficients of variation greater than 100) across study years (Table
2.3). Q70 yield and minimum 7-day mean flow yield had the most variability (coefficient of
variations of 132.2 and 175.2, respectively) across study reaches over the entire period, and both
metrics had minimum values below measurable levels. Coefficients of variation of daily flows
were above 100% in all sites with a mean of 472.8, suggesting that stream organisms often
experience high and in some cases extreme variability in stream flows. The frequency of low
flow events and rise rate yield were also relatively variable across sites, with coefficients of
variation of 66.9 and 75.8, respectively. The duration of high flow events was relatively short
across all stream reaches, ranging from 2.1 to 4.2 days with a mean of 3.4. For most stream flow
metrics, mean values of the 7-month period prior to each sampling season were similar across
years, but coefficients of variation were highest for April runoff yield (112.1) and Q70 yield

74

(50.3). Mean of Q70 yield summarized over the 7-month period across all sites in 2012 was
nearly double that of the 2013 value (13.3 vs. 7.8 L/s/km2, respectively) and one third of the
mean in 2014 (4.7 L/s/km2), suggesting that inter-annual variation in total rainfall likely resulted
in substantial differences in baseflow over our study years. Flow conditions prior to the
sampling season differed greatly across years; April runoff was approximately 25 times higher in
2014 than 2013 (23.6 vs 588.9 ML/km2).
Mean annual rainfall had strong and significant correlations with six of the nine stream
flow metrics calculated over the entire period of record, indicating that rainfall along the North
Hilo Coast has strong controls on stream flow magnitude and variability (Table 2.4). Of the nine
stream flow metrics, Q70 yield had the strongest positive correlation with mean annual rainfall (r
= 0.91). We selected five stream flow metrics with strong and significant correlations with mean
annual rainfall for further comparison with population characteristics. We excluded minimum 7day mean flow yield given its redundancy (r ≥ 0.9) with multiple metrics and its limited
ecological interpretability when compared to Q70 yield. April runoff yield was included despite
its redundancy with the frequency of high flow events (r = 0.94), given that it is a short-term
representation of flow conditions prior to the sampling season that has greater year-to-year
variability than other stream flow metrics.
Variation in habitat factors across the study region and relationships with rainfall
Similar to stream flow metrics, stream habitat factors also varied across the study reaches
(Table 2.5). Mean wetted area was 2.1 m2, with values ranging from 0.1 to 9.1 m2. The
minimum substrate size across stream reaches was 1071 mm and the mean was 2353 mm,
indicating that study reaches were predominately composed of large substrate (i.e., bedrock,
boulders). Reaches ranged from those predominantly composed of pools (90.7%) to those in

75

which pools were rare (0.8%). Riffles were less common across sites, with a maximum of
29.5%. Mean percent moss and percent algae cover were both close to 10% (9.9 and 9.6%,
respectively).
Percent pool habitat, riffle habitat, moss cover, and algal cover all had strong and
significant correlations with at least one of the six stream flow metrics correlated with mean
annual rainfall (Table 2.6). This suggests that rainfall likely has indirect relationships with other
characteristics of habitat through controls on the stream flow regime. Percent pools and riffles
had directionally opposite correlations with all stream flow metrics. For instance, pools were
negatively correlated with metrics of flow magnitude, the frequency of high flow events and
minimum 7-day mean flow yield, while riffles were positively correlated with these metrics.
Both percent pools and percent riffles and were both most strongly correlated with Q70 yield (0.77 and 0.91, respectively). While multiple stream flow metrics were correlated with percent
moss cover, it was most strongly correlated with frequency of low-flow events (r = -0.90). The
frequency of low flow events was also the only stream flow metric with strong and significant
correlations with percent algae (r = 0.67).
Associations among A. bisulcata population characteristics, flow metrics, and habitat
factors
All four population characteristics showed a range in values across study reaches, but
descriptive statistics were similar across years (Table 2.7). Average and maximum POC length
was relatively constant across years and range in POC length across sites was less than 2.5 mm
in all years. The range in average relative mass across sites was similar across years, with ranges
between 26% and 32% in all years of the study. The mean of average relative mass across sites
was below 100% in 2012 (97.9%) and 2013 (96.2%), but above 100% in 2014 (108.0%). The

76

minimum value of percent relative mass for stream reaches was also much higher in 2014 than
2013 and 2012 (95.2 vs 84.1 and 80.3%, respectively), suggesting that overall individuals were
in better condition across sites in 2014. Brown spot disease was never found on more than
20.5% of males within a reach in a given year, and the disease was not observed in at least one
stream in both years it was monitored. Average biomass had minimum values of zero in all
years and mean values ranging from 0.24 to 0.33 g/m2. In two streams in 2012 (Pahale and
Manowaiopae) and one stream in 2014 (Kaawalii), estimates of relative mass, average POC
length and the prevalence of brown spot disease were not calculated due a lack of males within
samples (Table C2.5).
All population characteristics had strong correlations with at least one of the five stream
flow metrics considered in at least two years, suggesting that multiple aspects of stream flow
may have effects on populations of A. bisulcata (Table 2.8). Average POC length had a strong
positive correlation with median baseflow yield in all three years (r = 0.82, 0.87, and 0.80;
Figure 2.2). Average relative mass had a strong positive correlation with median baseflow yield
in 2012 and 2013 (r = 0.72 and 0.78) but not in 2014 (Figure 2.3). The prevalence of brown spot
disease had strong positive correlations with the coefficient of variation in daily flows in both
2013 and 2014 (r = 0.62 and 0.77, respectively; Figure 2.4). Biomass had strong positive
correlation with the coefficient of variation in daily flows in two years but was most strongly
correlated with median baseflow yield (r = 0.75, 0.82, and 0.87; Figure 2.4) and the frequency of
low flow events (r = -0.79, -0.83, and -0.77). Associations between population characteristics of
A. bisulcata and habitat factors largely reflect changes in habitat associated with differences in
the stream flow regime (Table 2.9) which suggests that changes in the habitat factors and
population characteristics occur with changes to the stream flow regime.

77

Discussion
Our results highlight the potential effects of climate change on tropical island streams,
which can be used to inform the proactive conservation of stream species as climate change
progresses. Multiple aspects of the stream flow regimes across our study reaches were sensitive
to mean annual rainfall, and several habitat factors also varied with differences in the rainfallsensitive flow metrics. Relationships between rainfall-sensitive flow metrics and population
characteristics of A. bisulcata (Figure 2.2, Figure 2.3, and Figure 2.4) suggests that populations
across the North Hilo coast are sensitive to changes in the flow regime controlled by rainfall. In
streams with lower baseflows at the drier end of the rainfall gradient, we observed populations of
A. bisulcata composed of smaller individuals in poorer condition, which,along with an increase
in the frequency of low flows, contributed to lower total biomass within streams. Higher flow
variability in drier streams was also associated with increased prevalence of brown spot disease,
suggesting that populations exposed to more extreme flow conditions are more prone to
infection. Overall, our study offers important insights into influences of stream flow
characteristics on populations of A. bisulcata and demonstrates the value of using population
characteristics to study effects of stream flow on tropical stream organisms.
Variation in flow metrics and their relationship to rainfall
Our results indicate strong relationships between magnitude of annual rainfall and
multiple metrics of stream flow, emphasizing that annual rainfall is the dominant factor
influencing stream flow magnitude, variability, and duration and frequency of ecologically
important flow events in our study streams. While these relationships have been described
previously and have been attributed to relatively small drainage areas, high slopes, and limited
groundwater inputs of these streams (Strauch et al., 2015), our results underscore the strength of

78

these relationships. The consistency of values of Q70 yield below measurable levels in several
drier reaches across years and the high inter-annual variability in average Q70 yield across
reaches shows the sensitivity of baseflow within these streams to differences in mean annual
rainfall. This supports current understanding that baseflows in Hawaiian streams are extremely
sensitive to changes in rainfall (Bassiouni and Oki, 2013; Frazier and Giambelluca, 2016), which
in turn suggests these systems are highly vulnerable to projected declines in rainfall associated
with climate change (Timm et al, 2015).
Variation in habitat factors across the study region and relationships with rainfall
Our results also show that the stream flow regime influences the type of habitat and
availability of moss and filamentous algae within reaches, outcomes that may have implications
for habitat and food availability for A. bisulcata. This is not surprising, as declining baseflow
results in changes to stream habitat types and reduces hydraulic habitat complexity (Rolls et al.,
2012). In headwater streams of West Virginia, abnormally low baseflow due to drought
temporarily reduced riffle habitat within seven stream reaches by half but had little to no effect
on total amount of pool habitat (Hakala & Hartman, 2004). In our study region, this
homogenization of habitat types is apparent in streams with lower baseflow at the drier end of
the gradient, with pools more prevalent in drier reaches and riffles largely absent. These
differences in habitat in drier reaches may have negative consequences for A. bisulcata, given
that the species may preferentially select higher velocity habitats (Couret, 1976; Brasher et al.,
1997).
Greater aquatic moss cover within a reach is often associated with more stable
environments and high baseflow (Bowden, 1999), and strong correlations between stream flow
magnitude, frequency of low flow events, and flow variability with percent moss cover in our

79

study reaches supports this understanding. While A. bisulcata is not suspected to feed on aquatic
moss (Riney, 2015), it can potentially act as high-quality habitat for macroinvertebrate species
that traps FPOM and contains high levels of epiphytic microalgae (Maurer & Brusven, 1983;
Brusven et al., 1990), increasing foraging opportunities indirectly for grazers (Bowden, 1999).
Percent filamentous algae cover is only significantly correlated with increasing frequency of low
flow events, but similar insignificant correlations with Q70 yield and the minimum 7-day mean
flow further indicate that filamentous algae are more abundant in streams with lower baseflow
and longer and more frequent low flow periods. Streams with low baseflow and long low-flow
periods have fewer high flow events, higher temperatures, and more light associated with
shallower habitat, all which can encourage the growth of filamentous algae (Cummins, 1974).
While not a primary component of their diet, a recent stable isotope analysis indicated that
filamentous algae may provide a secondary food source for A. bisulcata (Riney, 2015). While
this may suggest that A. bisulcata have ample food resources in streams with lower rainfall,
Riney (2015) also found that the percent contribution of algae to their diets did not increase even
in streams with very low baseflow, suggesting that reductions in rainfall with climate change will
not result in increased food availability.
Association between A. bisulcata population characteristics and flow metrics
Baseflow and frequency of low flows
Across the North Hilo Coast of Hawaii Island, stream reaches with lower baseflows
support populations of A. bisulcata composed of smaller individuals in relatively poorer
condition than populations in streams with higher baseflows. In addition, streams with the
highest baseflow had the highest average biomass, while those with the lowest baseflow and the
highest frequency of low flows had the lowest average biomass. These results suggest that the

80

stream flow regime influences populations of A. bisulcata both directly through controls on
available habitat and potentially through indirect controls on food availability and temperature.
Temporary habitat desiccation
Very low baseflow and frequent low flows can lead to temporary habitat desiccation,
resulting in substantial loss of organism biomass within streams (Rolls et al., 2012). We
observed very low biomass estimates in reaches with higher frequencies of low flows and lower
baseflow, suggesting that temporary reductions in habitat in these reaches may have been
responsible. In all years of this study, habitat surveys showed that reaches with low mean annual
rainfall were comprised mostly of isolated pools throughout the channel. However, A. bisulcata
were rarely captured within these habitats and instead were found in upstream plunge pools
(depth >2 m). Use of plunge pool habitats by A. bisulcata in rivers with low baseflow supports
the understanding that these habitats, while comprising a low proportion of the stream under
normal flows, are important to the persistence of populations during periods of reduced flow
(McRae, 2007). Our results also indicate that reaches with larger upstream catchments may
provide refugia for A. bisulcata during times of low flow, as reaches with large upstream
catchments (>30 km2) with low baseflow (Waikaumalo and Umauma) had moderate measures of
biomass in all years (Figure 2.5; Table C2.4).
Food availability
In our study, limited food availability in streams with lower baseflow likely resulted in
lower average relative mass in 2012 and 2013 and may be responsible for observed differences
in average POC length. A. bisulcata are opportunistic filterers/grazers assumed to filter FPOM
in high velocity areas of the reach as well as graze on benthic resources (Couret, 1976). This
suggests that declines in overall foraging space and habitats for filter feeders associated with

81

lower baseflow may limit food availability. Riney (2015) found through stable isotope analysis
that biofilm was a predominant component of adult A. bisulcata diets in our study region, even in
reaches with lower stream flow and limited concentrations of biofilm. The diet of a similar
filterer/grazer atyid species in Australian streams, Paratya australiensis, consists largely of
cyanobacteria selectively grazed from biofilm (Burns & Walker, 2000). Given that biofilm shifts
from predominantly autotrophic to heterotrophic communities as baseflow declines (Timoner et
al., 2012), a reduction in both the quality and quantity of food may occur with declining stream
flow, with likely negative effects on both the growth and condition of A. bisulcata.
The lack of a relationship between Q70 yield and average relative mass in 2014 may have
resulted from seasonal anomalies in baseflow. April runoff yield was substantially higher in
2014 than 2013 and 2012 across our stream reaches, suggesting that average relative mass may
be sensitive to temporary increases in food availability resulting from higher baseflow. Many
studies have documented variability in organism condition with short-term or seasonal changes
in food availability (e.g., Hakala & Hartman, 2004; Currinder et al., 2014; Miller et al., 2015),
which may indicate differences in individual growth rates (Bentley & Schindler, 2013).
Stream temperature
In addition to limited food availability, higher stream temperatures may contribute to
smaller A. bisulcata in streams with lower baseflow. Increases in rearing temperature of
ectotherms results in lower maximum adult size (as reviewed by Atkinson et al., 1994). Across
several of our study reaches, Strauch et al. (2017) demonstrated that mean dry season water
temperature increased as baseflow decreased. Given that we observed declines in adult POC
length with declines in steam baseflow, it is possible that higher average stream temperature

82

results in lower maximum adult size. This relationship requires further study, as climate change
will result in increased air temperatures and in turn alter stream temperatures.
Flow variability and brown spot disease
Increases in the coefficient of variation in daily flows associated with declines in mean
annual rainfall may result in higher rates of carapace infection by chitinoclastic bacteria in
populations of A. bisulcata. While the exact mechanism causing increased prevalence of brown
spot disease in A. bisulcata is not clear, previous research suggests that carapace surface
abrasions increase susceptibility to infection (Chan, 1978). Abrasive forces acting upon
macroinvertebrates in streams are increased during periods of increased suspended sediment
loads (Jones et al., 2012). Preliminary analysis of measures of turbidity from 2012-2014 in our
study area does show higher sediment loads during flood events in streams receiving less annual
precipitation, but these data are only available from three reaches (unpublished data). Further
study of differences in sediment loading during high flow events may identify additional factors
influencing prevalence of brown spot disease.
Implications for assessing climate change effects and prioritizing conservation action
Our results demonstrate the value of considering population characteristics when
assessing effects of climate change through changes in flow regimes on stream organisms.
Understanding how climate-driven changes in flow may affect population characteristics such as
individual size, condition and susceptibility to disease is essential when considering effects of
climate change, as these measures may reflect the mechanisms which drive species loss. This
approach holds promise for the proactive conservation of other organisms as well, as changes in
the flow regime are likely to occur in many systems with future changes in climate. In addition,
Hawaiian streams support other endemic fish, shrimp and snails of conservation concern.

83

Differences in stream flow likely influence these species, and further research that identifies
relationships between population characteristics of these organisms and the stream flow regime
can improve proactive conservation decision-making in Hawaii.
Current climate projections indicate that changes in total annual rainfall will vary
regionally across the five main Hawaiian Islands in the coming century, with rainfall increasing
in some areas and decreasing in others (Timm et al., 2015). Our results suggest that in areas
where rainfall declines, baseflows may decrease but flow variability may increase. This will
likely have negative effects on stream organisms, as evidenced in our study by declines in A.
bisulcata individual size and relative mass with reductions in baseflow as well as increases in
disease prevalence with higher flow variability. This information suggests that proactive
conservation strategies that consider prioritization in the context of climate change are necessary,
but specific actions may differ depending on projected changes in rainfall and current habitat
condition. For instance, in areas that currently support populations of native stream organisms
but are likely to experience declines in rainfall, it is important to maintain and if possible
increase stream baseflow by minimizing and removing water diversions. In areas not likely to
experience declines in rainfall, conservation strategies may include the protection of streams that
currently have high baseflow and minimal anthropogenic disturbance in the catchment, given
these areas will likely provide the best habitat for stream species in the future (Tingley this
volume c). In addition, areas that are moderately disturbed but are not projected to decline in
annual rainfall may be optimal for restoration projects that improve habitat and increase
baseflow.

84

APPENDICES

85

APPENDIX 2.A

Tables

86

Table 2.1. Network catchment area, mean annual rainfall in catchments, and elevation of the
eleven study reaches. Streams are listed geographically from south to north across the study area
(Figure 2.1).
Network
Mean annual
catchment
rainfall
Elevation
2
Stream
area (km )
(mm/year)
(m)
Honolii
30.4
5252
470
Pahoehoe
1.3
5487
430
Kolekole
17.2
5982
500
Kapue
20.7
5257
460
Umauma
46.0
4441
490
Waikaumalo
36.0
3410
450
Manoloa
2.6
4759
420
Makahiloa
9.6
4263
400
Pahale
10.1
3676
510
Manowaiopae
3.8
4271
470
Kaawalii
33.1
2724
475

87

Table 2.2. Stream flow metrics generated for each stream reach using daily flow data.
Aspect of
flow regime Flow metric
Magnitude
Q70 yield
AR yield
CV*
Frequency
FL*

Metric name

Units

Description

Q70 yield
April runoff yield
Coefficient of variation

L/s/km2
ML/km2
-

Seventieth percentile exceedance
Total April runoff in million liters divided by catchment area
Standard deviation of daily flow divided by the mean

Low flow frequency

Count

Total number of low flow events below 5% of mean annual
daily flow

FH*

High flow frequency

Count

Total number of flow events in a year greater than the 25th
percentile exceedance

DL yield*

Minimum 7-day mean flow L/s/km2

Minimum of 7-day mean flow divided by catchment area

DH*

Duration of high flow

Days

Duration of high flow events as represented by FH

Constancy

-

Stability in day to day stream flows over the study period

Rise rate yield

L/s/km2

Duration

Timing
CON*
Rate of change
RR yield*

Average rise rate of days with increased flow over previous
day divided by catchment area
*Metrics calculated using the USGS Hydrologic Index Tool (HIT; Henriksen et al., 2006) with descriptions modified from Olden &
Poff (2003).

88

Table 2.3. Descriptive statistics for selected stream flow metrics across the 11 study reaches for the entire period of record. Also
shown are the mean values of each stream flow metric for each 7-month period prior to the sampling season and the coefficient of
variation of these mean values.
Aspect of
CV of
flow
Flow
Std.
2012
2013 2014 yearly
regime
metric
Units
Mean
Min Max deviation CV mean
mean mean means
Magnitude
Q70 yield
L/s/km2
8.2
0.0
34.5
10.9 132.2
13.3
7.8
4.7
50.3
2
April runoff yield
ML/km 261.4
16.3 497.7 135.4
55.6 172.0
23.6 588.9
112.1
Coefficient of variation
- 472.8 178.3 840.1 214.1
45.2 198.8 340.7 425.2
35.6
Frequency
Frequency of low flows
Count
13.9
0.7
23.3
9.3
66.9
7.8
7.5
8.6
7.6
Frequency of high flows
Count
22.8
10.3
27.7
5.3
23.3
15.6
11.7
13.4
14.2
Duration
Minimum 7-day mean flow L/s/km2
3.8
0.0
22.2
6.7 175.2
5.2
4.5
3.3
22.5
Duration of high flows
Days
3.4
2.2
4.2
0.5
16.3
3.9
4.7
4.4
9.7
Timing
Constancy
0.47
0.32
0.61
0.11
22.6
0.45
0.48
0.46
3.3
Rate of change
Rise rate yield
L/s/km2 215.8
10.3 565.4 163.7
75.8 282.3 129.2 301.5
39.7

89

Table 2.4. Pearson’s correlation coefficients describing relationships between mean annual rainfall (MAR), drainage area (DA) and
stream flow metrics. Correlation coefficients greater than an absolute value of 0.6 and significant (Îą=0.05) are shown in bold.

MAR
DA
Q70 yield
CV
AR yield
FL

MAR

DA

Q70 .
yield

CV

AR.
yield

FL

1.00

-0.31

0.91

-0.72

0.62

-0.68

0.64

0.82

0.19

0.31

0.32

1.00

-0.13

0.15

-0.14

0.12

-0.05

-0.17

0.12

0.24

-0.10

1.00

-0.82

0.50

-0.78

0.46

0.95

0.30

0.53

0.12

1.00

-0.27

0.83

-0.18

-0.89

0.15

-0.61

0.11

1.00

-0.03

0.94

0.33

0.02

-0.22

0.90

1.00

-0.05

-0.90

-0.17

-0.64

0.34

1.00

0.27

0.14

-0.18

0.87

1.00

0.26

0.60

-0.09

1.00

0.10

-0.08

1.00

-0.56

FH
DL yield
DH
CON
RR yield

FH

DL
yield

DH

CON

RR
yield

1.00

90

Table 2.5. Descriptive statistics for habitat factors across study reaches averaged across years.

Habitat characteristic
Wetted area
Mean substrate size
Pool habitat
Riffle habitat
Moss cover
Algal cover

Units
m2
mm
%
%
%
%

Mean
2.1
2353
35.6
12.8
9.9
9.6

Minimum
0.1
1071
0.8
0.0
0.0
0.3

91

Maximum
9.1
4356
90.7
29.5
33.6
20.8

Std.
deviation
2.66
1120.3
31.98
10.99
14.39
7.08

Table 2.6. Pearson’s correlation coefficients between stream flow metrics and habitat factors.
Correlation coefficients greater than an absolute value of 0.60 and significant (Îą=0.05) are shown
in bold.
Aspect of
Mean
flow
Wetted substrate Pool
regime
Metric
area
size
habitat
Magnitude
Q70 yield
0.57
-0.40
-0.77
April runoff yield
0.27
-0.16
-0.76
Coefficient of variation
-0.51
0.34
0.46
Frequency
Frequency of low flows
-0.59
0.41
0.40
Frequency of high flows
0.44
-0.16
-0.72
Duration
Minimum 7-day mean flow 0.52
-0.37
-0.65
Duration of high flows
0.25
0.00
-0.34
Timing
Constancy
0.53
-0.06
-0.14
Rate of change
Rise rate yield
0.06
-0.04
-0.50

92

Riffle
habitat

Moss Algae
cover cover

0.91
0.49
-0.73

0.76 -0.52
0.24 -0.06
-0.71 0.44

-0.68
0.46

-0.90 0.67
0.23 -0.13

0.86
0.26

0.82 -0.57
0.15 -0.40

0.43

0.36 -0.30

0.13

-0.06

0.16

Table 2.7. Descriptive statistics for A. bisulcata population characteristics across sites for each
sample year.
Population
characteristic Statistic
2012
2013
2014
POC length (mm)
Mean
6.1
6.1
6.0
Minimum
5.3
5.2
4.8
Maximum
7.3
7.3
7.1
Range
2.0
2.1
2.3
Std. Deviation
0.78
0.76
0.77
Relative mass (%)
Mean
97.9
96.2
108.0
Minimum
84.1
80.3
95.2
Maximum
110.2
112.1
120.1
Range
26.1
31.8
24.9
Std. Deviation
8.21
9.78
6.74
Brown spot (%)
Mean
.
6.7
6.5
Minimum
.
0.0
0.0
Maximum
.
20.5
13.5
Range
20.5
13.5
Std. Deviation
.
7.13
5.23
2
Biomass (g/m )
Mean
0.33
0.27
0.24
Minimum
0.00
0.00
0.00
Maximum
0.82
0.80
0.67
Range
0.82
0.80
0.67
Std. Deviation
0.300
0.290
0.240

93

Table 2.8. Pearson’s correlation coefficients between A. bisulcata population characteristics and
stream flow metrics. Correlation coefficients greater than 0.6 and statistically significant
(Îą=0.05) are shown in bold.
Population
2012 2013 2014
characteristic Flow metric
POC length
Q70 yield
0.82 0.87 0.80
Coefficient of variation
-0.60 -0.88 -0.47
April runoff yield
0.54 0.83 0.68
Frequency of low flows
-0.34 -0.73 -0.55
Frequency of high flows
-0.19 -0.24 0.33
Relative mass
Q70 yield
0.72 0.78 0.36
Coefficient of variation
-0.74 -0.59 0.11
April runoff yield
0.46 0.78 0.51
Frequency of low flows
-0.58 -0.49 -0.22
Frequency of high flows
-0.59 -0.59 -0.39
Biomass
Q70 yield
0.75 0.82 0.87
Coefficient of variation
-0.76 -0.85 -0.45
April runoff yield
0.16 0.78 0.23
Frequency of low flows
-0.79 -0.83 -0.77
Frequency of high flows
-0.55 -0.35 -0.20
Brown spot
Q70 yield
. -0.56 -0.63
Coefficient of variation
. 0.62 0.77
April runoff yield
. -0.53 0.51
Frequency of low flows
. 0.36 0.63
Frequency of high flows
. 0.55 0.55

94

Table 2.9. Pearson’s correlation coefficients between A. bisulcata population characteristics and
habitat factors. Correlation coefficients greater than 0.6 and statistically significant (Îą=0.05) are
shown in bold.
Population
characteristic
Habitat factor
2012 2013 2014
POC length
Wetted area
0.48
0.46
0.48
Mean substrate size
-0.59 -0.38 -0.29
Pool habitat
-0.24 -0.76 -0.72
Riffle habitat
0.58
0.77
0.84
Moss cover
0.75
0.79
.
Algal cover
. -0.49 -0.49
Relative mass
Wetted area
-0.08
0.38 -0.26
Mean substrate size
-0.53 -0.27 -0.01
Pool habitat
-0.64 -0.91 -0.49
Riffle habitat
0.53
0.41
0.68
Moss cover
0.31
0.65
.
Algal cover
. -0.52 -0.10
Biomass
Wetted area
0.57
0.67
0.63
Mean substrate size
-0.04 -0.18 -0.21
Pool habitat
-0.53 -0.64 -0.79
Riffle habitat
0.30
0.58
0.62
Moss cover
0.80
0.86
.
Algal cover
. -0.62 -0.72
Brown spot
Wetted area
. -0.40 -0.52
Mean substrate size
0.13
0.36
.
Pool habitat
0.52
0.20
.
Riffle habitat
. -0.71 -0.45
Moss cover
. -0.25 -0.55
Algal cover
0.66
. -0.08

95

APPENDIX 2.B

Figures

96

Figure 2.1. North Hilo coast of Hawaii Island and locations of study reaches along a gradient in mean annual rainfall.

97

Figure 2.2. Average post orbital carapace length (mm) of males versus the natural log (x+1) of
Q70 yield. Panels show 2012 (A), 2013 (B) and 2014 (D) data.
98

Figure 2.3. Average relative mass of individuals collected during each sampling event versus the
natural log (x+1) of Q70 yield. Significant correaltions are denoted by a solid (vs. dashed) trend
line. Panels show 2012 (A), 2013 (B) and 2014 (D) data.
99

Figure 2.4. The percent of male A. bisulcata with brown spot disease versus the natural log (x+1)
transformed coefficient of variation in daily flows (CV). Panels show 2012 (A) and 2013 (B)
data.

100

Figure 2.5. Average biomass (g/m2 ) of A. bisulcata versus the natural log (x+1) Q70 yield.
Panels show 2012 (A), 2013 (B) and 2014 (D) data.
101

APPENDIX 2.C

Supplemental tables

102

Table C2.1. Stream flow metric values for each stream over the entire period of record and each
7-month flow period (October to May) of each year (2012, 2013 and 2014)
Time
Q70
AR
DL
RR
Stream period yield
CV
yield
FL
FH
yield
DH CON yield
Honolii
Full
record 19.30 240.88 253.21
0.67 25.67
8.49
3.77 0.61 113.50
2012
27.20 198.83 228.74
1.00 16.00
7.86
3.31 0.56 169.39
2013
18.09 223.64
58.43
0.00
8.00 11.71
6.63 0.68 91.05
2014
11.45 262.69 472.46
0.00 12.00
7.82
4.42 0.57 121.75
Pahoehoe
Full
record 34.53 178.33 262.75
0.67 19.00 22.25
3.50 0.52 80.00
2012
63.80 126.21 278.68
0.00
6.00 34.90
8.67 0.52 93.15
2013
36.58 179.71
88.10
0.00 14.00 23.80
3.79 0.59 75.65
2014
21.44 285.74 421.46
0.00
7.00 16.55
7.57 0.53 99.18
Kolekole
Full
record 14.01 274.30 283.25
1.00 27.67
6.02
3.10 0.54 150.45
2012
20.20 220.12 258.01
5.00 18.00
6.51
2.94 0.53 218.96
2013
13.09 262.27
40.31
0.00
9.00
6.86
5.89 0.52 126.09
2014
7.67 294.55 551.44
0.00 15.00
5.90
3.53 0.55 183.67
Kapue
Full
record 10.99 304.93 418.80 13.33 26.67
2.54
3.47 0.51 239.97
2012
17.90 224.11 345.32
5.00 17.00
2.84
3.12 0.48 360.59
2013
9.41 270.42
40.44 11.00 11.00
4.19
4.82 0.56 185.62
2014
5.28 326.53 870.63 13.00 16.00
2.15
3.31 0.51 273.38
Umauma
Full
record
4.85 455.47 515.32 23.33 26.67
1.12
3.08 0.46 454.56
2012
5.40 371.68 179.20 11.00 19.00
0.96
2.79 0.43 385.51
2013
2.58 372.88
11.71 13.00 12.00
0.87
4.42 0.53 229.55
2014
1.24 416.68 764.92 10.00 16.00
0.53
3.31 0.43 406.46
Waikaumalo
Full
record
0.18 840.13 160.97 22.33 22.00
0.04
4.24 0.35 166.73
2012
1.70 164.25
7.66
0.00 11.00
0.92
4.82 0.54
3.41
2013
0.14 435.65
0.64 12.00 11.00
0.08
4.82 0.36 50.46
2014
0.10 806.19 474.62 13.00 13.00
0.07
4.08 0.40 354.75

103

Table C2.1. (cont’d)
Time
Q70
Stream period yield
Manoloa
Full
record
3.90
2012
5.30
2013
3.54
2014
3.25
Makahiloa
Full
record
2.17
2012
3.90
2013
2.32
2014
1.01
Pahale
Full
record
0.00
2012
0.00
2013
0.00
2014
0.00
Manowaiopae
Full
record
0.52
2012
0.90
2013
0.41
2014
0.57
Kaawalii
Full
record
0.00
2012
0.00
2013
0.00
2014
0.00

CV

AR
yield

CON

RR
Yield

FL

FH

611.42
136.30
434.65
644.87

290.87
21.41
14.30
836.90

16.00
0.00
8.00
18.00

25.33
10.00
16.00
19.00

1.58
2.76
1.40
2.39

3.93
5.30
3.31
2.79

0.51
0.48
0.45
0.56

188.74
6.15
121.48
652.37

710.15 497.68
760.32 228.26
385.54
3.87
453.01 1260.90

22.67
18.00
12.00
9.00

27.33
19.00
13.00
14.00

0.13
0.01
0.00
0.04

3.30
2.79
4.08
3.64

0.32 565.44
0.33 1103.13
0.32 323.59
0.27 496.79

388.68
275.64
378.24
364.24

278.81
181.28
0.00
655.13

22.33
15.00
14.00
13.00

22.00
18.00
12.00
16.00

0.00
0.00
0.00
0.00

2.15
2.94
4.42
3.31

0.40
0.28
0.47
0.43

269.65
289.79
182.11
630.64

618.93
535.26
389.39
395.58

95.12
162.22
1.28
121.87

19.00
22.00
7.00
13.00

18.00
22.00
10.00
13.00

0.01
0.04
0.00
0.29

3.25
2.41
5.30
4.08

0.32
0.26
0.33
0.36

134.69
457.82
28.55
82.84

578.11
543.59
415.56
427.38

16.33
1.43
0.05
47.50

11.33
9.00
5.00
6.00

10.33
15.00
13.00
6.00

0.00
0.00
0.00
0.00

3.10
3.40
4.00
8.83

0.60
0.66
0.56
0.47

10.28
17.36
7.18
15.03

104

DL
yield

DH

Table C2.2. Substrate size categories and references following Higashi & Nishimoto (2007) used
to estimate percent substrate composition and mean substrate size in stream reaches.
Particle diameter
Geometric mean
Size category
range (mm)
(mm)
5656.854
Bedrock
4000.00 – 8000.00
1011.929
Boulder
256.00 – 4000.00
128.00
Cobble
64.00 – 256.00
11.314
Gravel
2.00 – 64.00
0.256
Sand
0.06 - 2.00
0.024
Silt
0.01 - 0.06

105

Table C2.3. Length-mass regression coefficients for A. bisulcata in each reach each year.
Asterisks indicate regressions not used to calculate geometric mean coefficient values due to a
dominance by juveniles, limited sample size (n<30) or R2 < 0.8.
Stream
Year
a'
B
R2
Honolii
2012
-3.59 Âą 0.083 3.03 Âą 0.099 0.94
2013
-3.75 Âą 0.078 3.26 Âą 0.093 0.94
2014
-3.37 Âą 0.087 2.80 Âą 0.104 0.89
Kaawalii
2012*
-3.63 Âą 0.234 3.02 Âą 0.305 0.85
2013*
-3.48 Âą 0.178 2.84 Âą 0.236 0.91
2014*
-3.52 Âą 0.115 2.83 Âą 0.221 0.65
Kapue
2012
-3.68 Âą 0.095 3.18 Âą 0.114 0.92
2013
-3.77 Âą 0.091 3.31 Âą 0.107 0.93
2014
-3.46 Âą 0.083 2.97 Âą 0.100 0.91
Kolekole
2012
-3.50 Âą 0.109 2.99 Âą 0.125 0.90
2013
-3.34 Âą 0.124 2.83 Âą 0.139 0.86
2014
-3.51 Âą 0.090 3.01 Âą 0.104 0.90
Makahiloa
2012*
-3.98 Âą 0.129 3.61 Âą 0.185 0.95
2013*
-3.36 Âą 0.122 2.73 Âą 0.175 0.76
2014*
-2.98 Âą 0.135 2.27 Âą 0.191 0.61
Manoloa
2012
-3.81 Âą 0.080 3.39 Âą 0.097 0.95
2013
-3.54 Âą 0.081 3.00 Âą 0.101 0.93
2014
-3.33 Âą 0.088 2.78 Âą 0.112 0.86
Manowaiopae
2012*
-3.93 Âą 0.222 3.28 Âą 0.272 0.86
2013
-3.76 Âą 0.079 3.17 Âą 0.107 0.92
Pahale
2014
-3.80 Âą 0.092 3.34 Âą 0.134 0.87
2013
-3.48 Âą 0.164 2.90 Âą 0.207 0.87
2014
-3.68 Âą 0.165 3.24 Âą 0.213 0.93
Pahoehoe
2012
-3.61 Âą 0.113 3.15 Âą 0.134 0.90
2013
-3.47 Âą 0.132 2.95 Âą 0.152 0.84
2014
-3.14 Âą 0.084 2.63 Âą 0.098 0.88
Umauma
2012
-3.93 Âą 0.063 3.47 Âą 0.085 0.96
2013
-3.71 Âą 0.101 3.19 Âą 0.127 0.90
2014
-3.55 Âą 0.060 3.03 Âą 0.077 0.94
Waikaumalo
2012
-3.75 Âą 0.076 3.25 Âą 0.101 0.94
2013
-3.66 Âą 0.076 3.15 Âą 0.099 0.93
2014
-3.65 Âą 0.057 3.21 Âą 0.077 0.95
Geometric mean

-3.57

3.08

106

Table C2.4. Averages values of habitat factors from transect surveys

Year Stream
2012
Honolii
Pahoehoe
Kolekole
Kapue
Umauma
Waikaumalo
Manoloa
Makahiloa
Pahale
Manowaiopae
Kaawalii
2013
Honolii
Pahoehoe
Kolekole
Kapue
Umauma
Waikaumalo
Manoloa
Makahiloa
Pahale
Manowaiopae
Kaawalii
2014
Honolii
Pahoehoe
Kolekole
Kapue
Umauma
Waikaumalo
Manoloa
Makahiloa
Pahale
Manowaiopae
Kaawalii

Area

Mean
Pool
Riffle substrate
habitat habitat
size

Algae
cover

Moss
cover

9.97
0.73
4.73
4.41
3.18
1.46
0.72
1.91
0.67
0.14
0.37

31
1
6
7
7
0
17
5
46
88
88

16
26
31
14
7
2
6
28
1
0
0

1321.70
1730.94
1902.42
1830.22
4103.02
4643.37
995.48
1855.32
1910.30
1042.86
3232.99

8.24
0.60
3.46
2.85
2.09
1.67
0.59
0.28
0.66
0.06
0.28

30
0
9
16
18
46
27
44
66
85
100

15
34
19
28
8
6
10
13
0
12
0

1292.07
1315.66
2076.11
1907.81
4251.17
4435.84
958.90
1555.68
2684.30
1600.13
2878.56

0.0
1.8
0.3
13.3
5.1
1.3
0.4
17.3
5.6
20.5
9.2

32.8
32.5
31.8
0.0
0.1
0.0
22.5
0.0
0.0
0.5
0.0

9.05
0.68
4.29
3.37
2.37
1.69
0.46
0.30
0.53
0.13
0.26

26
1
7
9
15
38
11
70
88
89
84

12
29
37
33
17
1
12
6
0
0
0

1434.31
1901.82
2269.36
2561.48
4534.51
3988.10
1257.19
2043.65
3071.37
1990.08
3065.82

0.7
4.6
3.8
7.0
15.6
2.3
21.1
24.4
30.9
13.0
12.8

30.0
34.8
24.0
0.3
0.0
0.0
10.9
0.0
0.0
0.0
0.0

107

Table C2.5. Atyoida bisulcata population characteristics summarized for each reach each year.
Pahoe
Kole
Uma Waikau Maka
ManoHonolii
-hoe
-kole Kapue Manoloa -uma
-malo -hiloa Pahale waiopae
Average POCL (mm)
2012
6.7
6.8
7.3
6.6
6.1
5.5
5.5
5.3
.
.
2013
6.5
7.3
7.3
6.8
5.9
5.7
5.5
5.5
6.0
5.2
2014
6.6
7.1
7.1
6.5
5.8
5.6
5.4
5.4
5.8
4.8
Average relative mass (%)
2012
94.5 110.2 105.5
97.7
107.0
93.4
95.1
93.7
.
.
2013
99.0 105.2 112.1 102.8
100.4
93.0
96.6
98.6
89.5
80.3
2014
101.5 120.1 108.8 109.7
109.9
104.1 112.2 111.8 106.6
95.2
Brown spot disease (%)
2013
1.6
0.0
2.9
0.0
17.1
8.9
2.4
5.2
12.9
2.6
2014
0.0
1.2
3.8
1.2
13.5
11.4
10.2
12.0
9.2
2.5
2
Average biomass (g/m )
2012
0.82
0.66
0.36
0.27
0.57
0.49
0.40 <0.01
0.00 <0.01
2013
0.80
0.78
0.43
0.20
0.12
0.33
0.23
0.00 <0.01 <0.01
2014
0.61
0.67
0.36
0.27
0.19
0.29
0.26
0.01
0.00
0.00

108

Kaawalii
5.5
5.4
.
84.1
80.7
.
20.5
.
0.01
0.09
<0.01

LITERATURE CITED

109

LITERATURE CITED

Atkinson, D., 1994. Temperature and organism size: A biological law for ectotherms? Advances
in Ecological Research 25: 1-58.
Bassiouni, M. & D. S. Oki, 2013. Trends and shifts in streamflow in Hawai'i, 1913-2008.
Hydrological Processes 27: 1484-1500.
Bauer, R. T., 2013. Amphidromy in shrimps: a life cycle between rivers and the sea. Latin
American Journal of Aquatic Research 41: 633-650.
Benke, A. C., A. D. Huryn, L. A. Smock & J. B. Wallace, 1999. Length-mass relationships for
freshwater macroinvertebrates in North America with particular reference to the
southeastern United States. Journal of the North American Benthological Society 18:
308-343.
Bentley, K. T. & D. E. Schindler, 2013. Body condition correlates with instantaneous growth in
stream-dwelling Rainbow Trout and Arctic Grayling. Transactions of the American
Fisheries Society 142: 747-755.
Bowden, W. B. & Stream Bryophyte Group, 1999. Roles of bryophytes in stream ecosystems.
Journal of the North American Benthological Society 18: 151-184.
Bradford, M. J., 1997. An experimental study of stranding of juvenile salmonids on gravel ears
and in side channels during rapid flow decreases. Regulated Rivers-Research &
Management 13: 395-401.
Brasher, A. M., 1997. Habitat use by fish (o'opu), snails (hihiwai), shrimp ('opae) and prawns in
two streams on the island of Moloka'i, University of Hawaii Cooperative National Park
Resources Studies Unit Technical Report, 116. Department of Botany, Honolulu, HI.
Brasher, A. M. D., 2003. Impacts of human disturbances on biotic communities in Hawaiian
streams. Bioscience 53: 1052-1060.
Brusven, M. A., W. R. Meehan & R. C. Biggam, 1990. The role of aquatic moss on community
composition and drift of fish-food organisms. Hydrobiologia 196: 39-50.
Burns, A. & K. F. Walker, 2000. Biofilms as food for decapods (Atyidae, Palaemonidae) in the
River Murray, South Australia. Hydrobiologia 437: 83-90.
Canton, S. P., L. D. Cline, R. A. Short & J. V. Ward, 1984. The macroinvertebrates and fish of a
Colorado stream during a period of fluctuating discharge. Freshwater Biology 14: 311316.

110

Chace, F.A. Jr., 1983. The Atya-like shrimps of the Indo-Pacific Region (Decapoda: Atyidae).
Smithsonian Contributions to Zoology 384: 1-54.
Chan, J. G., 1978. Some aspects of a shell disease in the Hawaiian freshwater shrimp, Atya
bisulcata (Randall). In Smith, C.W. (ed), Proceedings of the Second Conference in
Natural Sciences Hawaii Volcanoes National Park 42-50.
Colwell, R. K., 1974. Predictability, constancy, and contingency of periodic
phenomena. Ecology 55: 1148-1153.
Couret Jr., C. L., 1976. The biology and taxonomy of a freshwater shrimp, Atyoida bisulcata
randall, endemic to the Hawaiian islands, MS Thesis. University of Hawai'i at Hilo, Hilo,
HI.
Craig, D. A., 2003. Geomorphology, development of running water habitats, and evolution of
black flies on Polynesian islands. Bioscience 53: 1079-1093.
Cross, W. F., A. P. Covich, T. A. Crowl, J. P. Benstead & A. Ramirez, 2008. Secondary
production, longevity and resource consumption rates of freshwater shrimps in two
tropical streams with contrasting geomorphology and food web structure. Freshwater
Biology 53: 2504-2519.
Cummins, K. W., 1974. Structure and function of stream ecosystems. Bioscience 24: 631-641.
Currinder, B., K. K. Cecala, R. M. Northington & M. E. Dorcas, 2014. Response of stream
salamanders to experimental drought in the southern Appalachian Mountains, USA.
Journal of Freshwater Ecology 29: 579-587.
Cushman, R. M., 1985. Review of ecological effects of rapidly varying flows downstream from
hydroelectric facilities. North American Journal of Fisheries Management 5: 330-339.
Davey, A. J. H., D. J. Kelly & B. J. F. Biggs, 2006. Refuge-use strategies of stream fishes in
response to extreme low flows. Journal of Fish Biology 69: 1047-1059.
De Grave, S. & X. Cai, 2013. Atyoida bisulcata. IUCN 2013. IUCN Red List of Threatened
Species, Version 2013.2. http://www.iucnredlist.org Accessed 13.01.24.
Frazier, A. G., T. W. Giambelluca, H. F. Diaz & H. L. Needham, 2015. Comparison of
geostatistical approaches to spatially interpolate month-year rainfall for the Hawaiian
Islands. International Journal of Climatology 12: 1459-1470.
Frazier, A. G. & T. W. Giambelluca, 2016. Spatial trend analysis of Hawaiian rainfall from
1920-2012. International Journal of Climatology
Froese, R., 2006. Cube law, condition factor and weight-length relationships: history, metaanalysis and recommendations. Journal of Applied Ichthyology 22: 241-253.
111

Giambelluca, T. W., Q. Chen, A. G. Frazier, J. P. Price, Y. L. Chen, P. S. Chu, J. K. Eischeid &
D.M. Delparte, 2013. Online Rainfall Atlas of Hawai‘i. Bulletin of the American
Meteorological Society 94: 313-316.
Gingerich, S. B. & R. H. Wolff, 2005. Effects of surface-water diversions on habitat availability
for native macrofauna, Northeast Maui, Hawaii. United States Geological Survey
Scientific Inverstigations Report, 2005-5213. U. S. Geological Survey, Reston, VA.
Gorbach, K. R., M. E. Benbow, M. D. McIntosh & A. J. Burky, 2012. Dispersal and upstream
migration of an amphidromous neritid snail: implications for restoring migratory
pathways in tropical streams. Freshwater Biology 57: 1643-1657.
Gordon, N. A., T. A. McMahon, B.L. Finlayson, C. J. Gippel, & R. J. Nathan, 2004. Stream
hydrology-an introduction for ecologists, 2nd edition. John Wiley and Sons, New York,
NY.
Hakala, J. P. & K. J. Hartman, 2004. Drought effect on stream morphology and brook trout
(Salvelinus fontinalis) populations in forested headwater streams. Hydrobiologia 515:
203-213.
Henriksen, J. A., J. Heasley, J. G. Kennen & S. Nieswand, 2006. Users’ Manual for the
Hydroecological Integrity Assessment Process Software (including the New Jersey
Assessment Tools), U.S. Geological Survey, Biological Resource Discipline, Open File
Report 2006-1093. U.S. Department of the Interior, Washington, DC.
Higashi, G. R. & R. T. Nishimoto, 2007. The point quadrant method: rapid assessment of
hawaiian streams. In Evenhuis, N. L. & J. M. Fitzsimons (ed.), Biology of Hawaiian
Stream and Estuaries, Bishop Museum Bulletin in Cultural and Environmental Studies 3:
305-312.
IPCC, 2013. Climate Change 2013: The Physical Science Basis. Contribution of Working Group
I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change.
Stocker T. F., D. Qin, G. K. Plattner, M. Tignor, S. K. Allen, J. Boschung, A. Nauels, Y.
Xia, V. Bex & P. M. Midgley (ed.). Cambridge, United Kingdom and New York, NY.
Isaak, D. J., S. Wollrab, D. Horan & G. Chandler, 2012. Climate change effects on stream and
river temperatures across the northwest US from 1980-2009 and implications for
salmonid fishes. Climatic Change 113: 499-524.
James, A. B. W., Z. S. Dewson & R. G. Death, 2007. The effect of experimental flow reductions
on macroinvertebrate drift in natural and streamside channels. River Research and
Applications 24: 22-35.
Jones, J. I., J. F. Murphy, A. L. Collins, D. A. Sear, P. S. Naden & P. D. Armitage, 2012. The
impact of fine sediment on macro-invertebrates. River Research and Applications 28:
1055-1071.
112

Kaufmann, P. R. & E. G. Robison, 1998. Physical Habitat Assessment In Lazorchak, J. L., D. J.
Klemm & D. V. Peck (eds.), Environmental Monitoring and Assessment Program Surface Waters: Field Operations and Methods for Measuring the Ecological Condition
of Wadeable Streams, EPA/ 620/R-94/004F. U.S. Environmental Protection Agency,
Washington, D.C.
Keith, P., 2003. Biology and ecology of amphidromous Gobiidae of the Indo-Pacific and the
Caribbean regions. Journal of Fish Biology 63: 831-847.
Kido, M. H., 2008. A persistent species assemblage structure along a Hawaiian stream from
catchment-to-sea. Environmental Biology of Fishes 82: 223-235.
Kido, M. H., 2013. A native species-based index of biological integrity for Hawaiian stream
environments. Environmental Monitoring and Assessment 185: 4063-4075.
Knighton, D., 1998. Fluvial forms and processes: A new perspective. Arnold, London, UK.
Krabbenhoft, T. J., S. P. Platania & T. F. Turner, 2014. Interannual variation in reproductive
phenology in a riverine fish assemblage: implications for predicting the effects of climate
change and altered flow regimes. Freshwater Biology 59: 1744-1754.
Leong, J.A., J. J. Marra, M. L. Finucane, T. Giambelluca, M. Merrifield, S. E. Miller, J.
Polovina, E. Shea, M. Burkett, J. Campbell, P. Lefale, F. Lipschultz, L. Loope, D.
Spooner & B. Wang, 2014. Ch. 23: Hawai‘i and U.S. Affiliated Pacific Islands In J. M.
Melillo, T.C. Richmond & G. W. Yohe (eds.), Climate Change Impacts in the United
States: The Third National Climate Assessment U.S. Global Change Research Program
537-556. U.S. Government Printing Office, Washington, DC.
Leppi, J. C., T. H. DeLuca, S. W. Harrar & S. W. Running, 2012. Impacts of climate change on
August stream discharge in the Central-Rocky Mountains. Climatic Change 112: 9971014.
Lynch, A. J., B. J. E. Myers, C. Chu, L. A. Eby, J. A. Falke, R. P. Kovach, T. J. Krabbenhoft, T.
J. Kwak, J. Lyons, C. P. Paukert & J. E. Whitney, 2016. Climate Change Effects on
North American Inland Fish Populations and Assemblages. Fisheries 41: 346-361.
Lytle, D. A. & N. L. Poff, 2004. Adaptation to natural flow regimes. Trends in Ecology &
Evolution 19: 94-100.
Maurer, M. A. & M. A. Brusven, 1983. Insect abundance and colonization rate in fontinalis-neomexicana (bryophyta) in an Idaho batholith stream, USA. Hydrobiologia 98: 9-15.
McDowall, R. M., 2007. On amphidromy, a distinct form of diadromy in aquatic organisms. Fish
and Fisheries 8: 1-13.

113

McRae, M. G., 2007. The Potential for source-sink population dynamics in Hawaii’s
amphidromous fishes. In Evenhuis, N. L. & J. M. Fitzsimons (ed.), Biology of Hawaiian
Stream and Estuaries, Bishop Museum Bulletin in Cultural and Environmental Studies 3:
87-98.
Meffe, G. K. & W. L. Minckley, 1987. Persistence and stability of fish and invertebrate
assemblages in a repeatedly disturbed sonoran desert stream. American Midland
Naturalist 117: 177-191.
Miller, S. J., D. T. VanGenechten & C. E. Cichra, 2015. Length weight relationships and an
evaluation of fish size and seasonal effects on relative condition (K-n) of fishes from the
Wekiva River, Florida. Florida Scientist 78: 1-19.
Naesje, T., B. Jonsson & J. Skurdal, 1995. Spring flood: A primary cue for hatching of river
spawning Coregoninae. Canadian Journal of Fisheries and Aquatic Sciences 52: 21902196.
Nishimoto, R. T. & J. M. Fitzsimons, 2006. Status of Native Hawaiian Stream Fishes, A Unique
Amphidromous Biota. American Fisheries Society Symposium 53.
Nishimoto, R. T. & D. G. K. Kuamo'o, 1996. Monitoring study of the endemic mountain shrimp,
Atyoida bisulcata Randall, in Waikaumalo Stream, Hawaii Island. Proceedings of the
Western Association of Fish and Wildlife Agencies 76: 262-273.
Oki, D. S., R. H. Wolff & J. A. Perreault, 2010. Effects of surface-water diversion on
streamflow, recharge, physical habitat, and temperature, Na Wai Eha, Maui, Hawaii, U.S.
Geological Survey Scientific Investigations Report 2010-5011. U. S. Geological Survey,
Reston, VA.
Poff, N. L. & J. V. Ward, 1989. Implications of streamflow variability and predictability for lotic
community structure: a regional-analysis of streamflow patterns. Canadian Journal of
Fisheries and Aquatic Sciences 46: 1805-1818.
Poff, N. L., J. D. Allan, M. B. Bain, J. R. Karr, K. L. Prestegaard, B. D. Richter, R. E. Sparks &
J. C. Stromberg, 1997. The natural flow regime. Bioscience 47: 769-784.
Price, J. P., J. D. Jacobi, S. M. Gon, III, D. Matsuwaki, L. Mehrhoff, W. Wagner, M. Lucas & B.
Rowe, 2012. Mapping plant species ranges in the Hawaiian Islands—Developing a
methodology and associated GIS layers, U.S. Geological Survey Open-File Report 2012–
1192. U. S. Geological Survey, Reston, VA.
Riney, M. H., 2015. The impacts of stream flow rate on food resource quality and diets of native
endemic atyid shrimps (Atyoida bisulcata) in Hawaiian streams, MS Thesis. University
of Hawaii at Hilo, Hilo, HI.

114

Rolls, R. J., C. Leigh & F. Sheldon, 2012. Mechanistic effects of low-flow hydrology on riverine
ecosystems: ecological principles and consequences of alteration. Freshwater Science 31:
1163-1186.
Sherrod, D. R., 2008. Database to accompany geologic map of the state of Hawaii.
http://pubs.usgs.gov/imap/ Accessed 10.05.01.
Strauch, A. M., R. A. MacKenzie, C. P. Giardina & G. L. Bruland, 2015. Climate driven changes
to rainfall and streamflow patterns in a model tropical island hydrological system. Journal
of Hydrology 523: 160-169.
Strauch, A. M., C. P. Giardina, R. A. Mackenzie, C. Heider, T. W. Giambelluca, E. Salminen, &
G. L. Bruland, 2016. Modeled effects of climate change and plant invasion on watershed
function across a steep tropical rainfall gradient. Ecosystem 19: 1-18.
Strauch, A. M., R. A. Mackenzie & R. W. Tingley III, 2017. Baseflow driven shifts in tropical
stream temperature regimes across a mean annual rainfall gradient. Hydrological
Processes
Timm, O. E., T. W. Giambelluca & H. F. Diaz, 2015. Statistical downscaling of rainfall changes
in Hawaiˈi based on the CMIP5 global model projections. Journal of Geophysical
Research-Atmospheres 120: 92-112.
Timoner, X., V. Acuna, D. Von Schiller & S. Sabater, 2012. Functional responses of stream
biofilms to flow cessation, desiccation and rewetting. Freshwater Biology 57: 1565-1578.
Tingley, R. W. III, This volume a. Influences of natural landscape factors on tropical stream
organisms: An ecological classification of Hawaiian Island streams in A multi-scaled
research framework for conserving streams with changing climate, Doctoral
Dissertation. Michigan State University, East Lansing, MI.
Tingley, R. W. III, This volume c. Spatial prioritization of Hawaii’s stream ecosystems for
conservation in the context of changing climate: An ecological classification of
Hawaiian Island streams in A multi-scaled research framework for conserving streams
with changing climate, Doctoral Dissertation. Michigan State University, East Lansing,
MI.
Tsang, Y-P., D. Wieferich, K. Fang, D. M. Infante & A. R. Cooper, 2014. An efficient and
integrated approach of aggregating upstream catchment information. SpringerPlus 3: 589.
U. S. Department of Agriculture, Natural Resources Conservation Service, 1995. Soil Survey
Geographic (SSURGO) Data Base, Report 1527. U.S. Department of Agriculture,
Washington, D.C.

115

Walter, R. P., J. D. Hogan, M. J. Blum, R. B. Gagne, E. F. Hain, J. F. Gilliam & P. B. McIntyre,
2012. Climate change and conservation of endemic amphidromous fishes in Hawaiian
streams. Endangered Species Research 16: 261-272.
Wang, L. Z., D. Infante, P. Esselman, A. Cooper, D. Y. Wu, W. Taylor, D. Beard, G. Whelan &
A. Ostroff, 2011. A hierarchical spatial framework and database for the National River
Fish Habitat Condition Assessment. Fisheries 36: 436-449.
Ward, E. J., J. H. Anderson, T. J. Beechie, G. R. Pess & J. M. Ford, 2015. Increasing hydrologic
variability threatens depleted anadromous fish populations. Global Change Biology 21:
2500-2509.
Zeigler, M. P., A. S. Todd & C. A. Caldwell, 2012. Evidence of recent climate change within the
historic range of Rio Grande Cutthroat Trout: Implications for management and future
persistence. Transactions of the American Fisheries Society 141: 1045-1059.

116

CHAPTER 3
SPATIAL PRIORITIZATION OF HAWAII’S STREAM ECOSYSTEMS FOR
CONSERVATION IN THE CONTEXT OF CHANGING CLIMATE
Abstract
Declines in biodiversity resulting from anthropogenic disturbances to stream catchments
and habitats have occurred at a global scale. Climate change threatens to exacerbate these
effects, and in some cases, alter systems to the point where they can no longer support current
assemblages of organisms. Spatial prioritization of streams that identifies habitats for protection
or enhancement can aid in efforts to conserve stream habitats and their organisms with current
and future threats. The goal of this study was to address this need for streams of the five largest
Hawaiian Islands. Our first objective was to identify areas of high conservation priority by
considering both current and future characteristics of stream habitats, with future characteristics
determined by projected changes in rainfall at multiple time steps under multiple climate
scenarios. To provide additional insight for prioritizing conservation action, we also assessed
how future priority rankings compared to rankings based on current conditions. Our results
indicate that many regions that currently have stream reaches of high conservation value will
retain these characteristics through the end of the century under two different climate scenarios.
In addition to ranking streams by their within-reach habitat value, we established the importance
of connectivity to marine habitats in our ranking procedure, thus ensuring that complete
migratory pathways for Hawaii’s native amphidromous organisms were prioritized. Our results
suggest that streams that drain leeward slopes of the islands will likely experience reductions in
rainfall and may be negatively influenced by climate change. This study offers useful insight into
potential effects of climate change on tropical island streams and helps inform the location and

117

type of conservation action that may be most effective to ensure the persistence of native
Hawaiian stream organisms.
Introduction
Anthropogenic land use and alterations to stream channels including water diversions and
dams have degraded stream habitats worldwide (Allan, 2004; Helfman, 2007). Disturbances to
habitats have contributed to global declines in stream species biodiversity and increasing
numbers of species at risk of extinction (Strayer et al., 2006; Helfman, 2007; Jelks et al., 2008).
In addition, recent climate change has also had various impacts on stream organisms documented
in multiple regions. Increases in air temperature and changes in precipitation have reduced
population size and recruitment of salmonid fishes in North America (e.g., Ward et al., 2015;
Bassar et al., 2016). Increases in water temperature have also increased disease prevalence and
loss of habitat for Salmo trutta in alpine streams of Switzerland, contributing to declines in
abundance (Hari et al., 2006). Climate-induced reductions in distributions of cold and cool water
fish species have also been reported in multiple regions across the globe (as reviewed by Comte
et al., 2013). Given that increases in air temperature and changes in precipitation are projected to
continue over the coming century (IPCC, 2013), climate change effects on stream organisms will
likely be magnified and exacerbate global reductions in biodiversity and increases in species at
risk of extinction (Xenopoulos et al., 2005; Tedesco et al., 2013).
The current state of the world’s streams combined with the likelihood of continued
climate change underscore the need to consider both current and future disturbances when
prioritizing stream habitats for conservation. Conservation prioritization efforts that only
consider current conditions may not be fully effective at identifying priority habitats to protect
at-risk species, as even pristine systems may change in their ability to support assemblages with

118

changing climate (e.g., Battin et al., 2007). It is therefore imperative to anticipate and utilize
information on potential effects of climate change on stream habitats and organisms when
prioritizing areas for conservation (Palmer et al., 2009; Zeigler et al., 2012). However, a recent
review shows that a majority of spatial prioritization studies have not incorporated climate
change as a potential source of disturbance, and of those that do, very few focus on streams
(Jones et al., 2016). This underscores the need for additional regional efforts that support
proactive conservation of stream habitats and organisms in the context of climate change.
Streams of the Hawaiian Islands are home to an assemblage of native shrimp, fish and
snails, the majority of which are endemic to the island chain. Some of these organisms are also a
traditional food source of Hawaiian people and are still harvested by island residents. Currently,
anthropogenic land use and channel diversions have degraded many stream habitats, which has
in turn reduced species abundances in select areas of the islands (Brasher, 2003; Walter et al.,
2012). Due to their limited distributions and recent population declines, the International Union
for Conservation of Nature (IUCN) lists several of these species as near threatened or vulnerable
including Atyoida bisulcata, Neritina granosa, and Sicyopterus stimpsoni (World Conservation
Monitoring Centre, 1996; Cordeiro & Perez, 2012; De Grave & Cai, 2013).
In recent years, conservation decision makers have prioritized Hawaiian streams for
native species conservation based on estimates of current habitat condition and available species
data (e.g., Hawaii Fish Habitat Partnership, 2010). However, a lack of downscaled projected
climate data for the Hawaiian Islands has prevented past efforts from incorporating climate
change into spatial prioritizations, therefore limiting the ability to implement proactive
conservation strategies. This represents a critical knowledge gap for decision makers interested
in prioritizing areas for conservation action, as declines in baseflow over the past century across

119

the island system have been attributed to reductions in mean annual rainfall (Bassiouni & Oki,
2013; Frazier & Giambelluca, 2016), suggesting that further declines in rainfall will have
negative consequences for Hawaiian stream organisms. Recently, downscaled rainfall
projections across the Hawaiian Islands have become available (Timm et al. 2015), and these
projections indicate that climate change may result in substantial and regionally-specific
reductions in total rainfall across Hawaii. Changes in the ability of streams to support a given
taxa assemblage may occur as mean annual rainfall changes (Tingley this volume a), and where
populations persist within streams, recent research suggests that moderate flow reductions and
increased flow variability associated with lower rainfall can result in smaller individuals in
poorer condition (Tingley this volume b).
The goal of this study is to incorporate projected climate data into a prioritization of
stream habitats of the Hawaiian Islands for conservation of native species. Our first objective is
to identify areas of conservation value by considering both current and future characteristics of
stream habitats, with future characteristics determined by projected changes in rainfall.
Completing this objective will allow us to identify streams that are likely to retain high
conservation value given differences in climate change scenarios at different time periods. Our
second objective is to assess how outcomes of future rankings of conservation value compare to
a ranking based only on current condition to assess how prioritization strategies differ when
accounting for projected climate change. The results of this study will highlight the importance
of using proactive approaches when identifying high-value conservation areas while also
providing information to decision-makers in Hawaii.

120

Methods
Study area
The study area included all perennial and intermittent stream reaches and their
catchments from the five largest Hawaiian Islands (Hawaii, Maui, Molokai, Oahu and Kauai;
Figure 3.1). Though regional variation in rainfall does occur throughout the islands, interactions
between eastern trade-winds and topographic characteristics of each island control dominant
rainfall patterns and result in generally higher precipitation at higher elevations on the windward
(eastern) sides of islands and lower precipitation on leeward (western) facing slopes
(Giambelluca et al., 2011). Groundwater contributions to stream flow are substantial in some
catchments and near zero in others (Lau & Mink, 2006; Izuka et al., 2015). Disturbance from
agricultural and urban land use in stream catchments tends to be more severe along coastlines
and at lower elevations, while many high elevation areas are relatively undisturbed (Crawford et
al., 2016).
Native Hawaiian stream organisms include five species of fish (four in the family
gobiidae and one eleotridae), two shrimp (A. bisulcata and Macrobrachium grandimanus) and a
neritid snail (N. granosa; Kinzie, 1990). All of these species are amphidromous; they hatch in
streams, spend a short time (weeks to months) developing in the marine environment, and then
migrate upstream to mature and reproduce (McDowall, 1988). Current research suggests that
these species do not exhibit natal stream homing, and there is little distinguishable genetic
structure across the study region (Chubb et al., 1998; McDowall, 2003; Bebler & Foltz, 2004).
Coastal species, most notably striped mullet (Mugil cephalus) and flagtails (Kuhlia sp.), move to
and from freshwater stream habitats (Nishimoto et al., 2007; McRae et al., 2011).

121

Study units
The spatial unit assessed through our rankings was the stream reach as defined in the
Hawaii Fish Habitat Partnership (HFHP) stream layer (Tingley this volume a). The HFHP
stream layer is a modified version of the 1:24,000 National Hydrography Dataset (NHD 2008;
http://nhd.usgs.gov/) and includes 11,437 intermittent and perennial stream reaches delineated by
hydrologic breaks within the river network (e.g., confluences with other stream reaches,
waterfalls, and/or reservoirs). Each stream reach has an associated local catchment, defined as
the portion of the landscape where surface runoff enters the reach directly, as well as an
upstream catchment, defined as the entire area of the landscape draining to a given reach by
other tributaries and including the local catchment (following Wang et al., 2011).
Ranking reaches based on current and future conservation value
We used the spatial prioritization software program Zonation version 4.0 (Moilanen et
al., 2014) to generate multiple rankings of current and future conservation value of Hawaiian
stream reaches. Zonation prioritizes spatial units based on their ecological uniqueness by
accounting for presence of a habitat or species of interest (i.e., conservation features),
disturbances to study units, and their connectivity to other study units. We generated five
rankings with Zonation. We conducted one ranking of conservation value using only current
conditions, and four additional rankings that characterized conservation value based on current
conditions and projected changes in rainfall from two different downscaled climate change
scenarios (RCP 4.5 and 8.5) at two time periods (mid- and late-century; Timm et al., 2015).
Prior to ranking, we took multiple steps to define conservation value of individual reaches
(Figure 3.2).

122

Defining reach conservation value based on uniqueness of ecological potential
Determining current ecological potential
Across the five main Hawaiian Islands, variation in natural landscape characteristics
including average annual rainfall, reach elevation, channel slope, and catchment soil
permeability results in a diversity of stream habitats within and across catchments (Tingley this
volume a). The current ecological potential of stream reaches was defined based on results of a
stream reach classification that associated the landscape factors listed above with distributions of
stream taxa (Tingley this volume a). Thirteen classes of stream reaches were identified through
the classification based on differences in types and specific values of landscape factors
associated with differences in assemblages. For ranking, we considered classes that reaches
occurred within to be ecologically unique habitats, and we used these classes as the conservation
features of interest selected for within Zonation (Figure 3.3).
Determining future ecological potential
We estimated future ecological potential for the four future stream rankings using stream
reaches reclassified based on summarized projected annual rainfall data. Downscaled projected
rainfall data for Hawaii are currently available at two time periods (mid-century and latecentury) for both medium-low (RCP 4.5) and high emission scenarios (RCP 8.5; Timm et al.,
2015). We summarized projected annual rainfall data in the upstream catchment of stream
reaches for each of the two emission scenarios at each time period using the aggregation tool
described in Tsang et al. (2014). We then generated four representations of future classes by
reclassifying reaches that had current rainfall associated with taxa distributions. Four of 13
classes met this criteria, and we considered these susceptible to climate change and likely to
change in their ability to support current taxa assemblages (classes G, H, I, K; Table 3.1).

123

Creating and assigning the similarity matrix values to reaches
To ensure that rare habitats are not underrepresented in rankings of conservation value
generated with Zonation, it is important to account for similarity among habitats (in our case,
reach classes). We developed a measure of biological similarity for reach classes following
Leathwick et al. (2010) based on prevalent native stream taxa comprising assemblages of each of
the 13 stream reach classes (Table 3.1). Similarity values assigned to pairs of stream reach
classes ranged from 0 (prevalent taxa for a stream reach class are fully distinct from taxa of
another class) to 1 (all prevalent taxa are shared between two classes; Table C3.1). We then
incorporated the resulting similarity matrix using the community analysis option in Zonation for
our rankings (following Moilanen et al., 2014).
Creating and assigning taxa association weights to reaches
Because efforts to conserve streams in Hawaii have conservation of native stream
organisms as a primary focus, we incorporated a biodiversity measure into our rankings. We
developed a relative weighting of numbers of prevalent taxa within each reach class to account
for differences in reach class biodiversity (Table 3.1). We assigned reach classes with five or
more prevalent taxa a value of 4, classes with two to four taxa a value of 3, and classes with only
one prevalent taxon a value of 2. Classes that did not support amphidromous species but that
represent unique habitats still important for conservation (e.g., high elevation bogs) were
assigned a value of 1. Intermittent streams were assigned a value of 0 because we assumed that
these streams were least suitable for native Hawaiian stream organisms.

124

Modifying reach conservation value based on current and future disturbance
Applying current habitat condition score to estimates of conservation value
For the current conservation ranking using Zonation, we characterized disturbance to
stream reaches using a landscape-scale assessment of stream habitats conducted across the five
main Hawaiian Islands (Crawford et al., 2016). The assessment is based on the assumption that
greater amounts of disturbance in stream catchments will result in greater disturbance to stream
habitat (following Danz et al., 2007; Esselman et al., 2011). Assessment scores were derived
from multiple measures of anthropogenic landscape and stream channel disturbances likely to
have negative influences on habitat in stream reaches (Crawford et al., 2016), and results include
disturbance scores as indices available in multiple spatial extents. To create our score, we
summed an index characterizing fish habitat condition in local catchments of our stream reaches
with a second index characterizing fish habitat condition in their upstream catchments. We then
standardized this summed index to develop our cumulative upstream habitat condition index
(UHCI) from 0 (worst habitat condition) to 1 (best habitat condition) for analysis. We
incorporated the UHCI into the analysis for the current condition ranking using the “condition
layer” option in Zonation, which modifies the value of each reach given its measure of habitat
condition, up-weighting the conservation value of reaches with high UHCI scores in a given
reach class. In the four future rankings, we used both the UHCI scores with the climate exposure
score, described below, to modify reach conservation value.
Applying climate change exposure scores to estimates of future conservation value
Stream reach classes were defined based solely on the range of conditions characterized
by current climate data, however, strong deviations from current climate could still yield novel
conditions that could negatively affect stream organisms. To account for possible climate

125

changes beyond the range expressed by current conditions in our rankings, we generated a
“climate exposure score” by identifying projected annual precipitation amounts in stream
catchments that were substantially different from current conditions (Bush et al., 2014). We first
summarized current (1992-2007) mean annual and dry season rainfall in the upstream catchment
of stream reaches using monthly rainfall raster datasets developed by Frazier et al. (2015). We
then found the average and standard deviation of current annual and dry season rainfalls for each
stream reach. Next, using projected rainfall data, we determined whether mean annual and dry
season rainfalls for a given time step and climate change scenario were greater than one standard
deviation from current means. For those reaches where projected annual or dry season rainfalls
were reduced by more than one standard deviation from current means, we assigned a value of
0.5, meaning their conservation value would be lowered given the potential exposure to climate
change . Reaches that changed class with changes in climate were also assigned a value of 0.5,
and all other reaches were assigned a value of 1.
Allowing reach conservation value to vary based on downstream connectivity
Creating and applying downstream habitat loss penalty
Given the amphidromous life histories of many Hawaiian stream taxa, the conservation
value of a reach depends on its hydrologic connectivity to the marine environment. For each of
the five Zonation reach rankings, we used information on stream network connectivity available
in the HFHP stream layer to identify the portion of a stream network connecting the reach to the
marine environment. We then used the neighborhood quality penalty (NQP; Moilanen & Kujala,
2008) to prioritize migratory pathways in each stream ranking. Following Bond et al. (2014), the
value of a reach was set to be dependent on the conservation value of downstream reaches
connecting the reach to the marine environment.

126

Ranking stream reaches based on conservation value
We used the Core-Area Selection (CAZ) procedure for all five conservation value
rankings, given that this procedure is appropriate for studies in which priority units are those
with the greatest amount of conservation features in the best condition (Leathwick et al., 2010).
Priority ranking in Zonation begins with a representation of the entire landscape of interest and
proceeds by the iterative removal of single study units that contribute the least conservation
value until all study units have been removed (Di Minin et al., 2014). This results in the ranking
of each individual study unit, in this case the stream reach, on a scale from 0 to 100%, with the
reach ranking 100% having the highest conservation value, and the reach with 0% having the
lowest conservation value.
When designing proactive conservation strategies based on projected conservation value,
managers should consider results in the context of uncertainty associated with projected climate
changes. Therefore, we examined overlap in the ranks of individual reaches across the study
area for each of the projected climate scenarios at both mid and late century time periods. In
some instances, we interpreted results by region (Figure 3.1) and examined regional differences
with an emphasis on the top 5, 10 and 20% of overlapping ranked reaches. Reach ranking values
were attributed to local catchments and mapped across the study region.
Assessing differences in prioritization when accounting for climate change
We compared stream reach rankings at each future time-period with current conservation
value rankings to assess how proactive approaches to prioritization differ from approaches
focused only on current condition by considering overlap in the top 5 and top 10% of reaches in
each ranking. In addition, we identified those reaches that were only in the top 10% of the

127

current conservation value ranking or the future conservation value ranking for each time period.
Results were attributed to local catchments and mapped across the study region.
Results
Changes in reach class and reductions in rainfall with climate changes
Changes in total length of stream reaches found within each reach class were projected to
occur under both climate scenarios at each time period (Table 3.2). These changes suggest a
shift in the ability of a stream reach to support its current assemblage of stream organisms.
Overall, changes to classes were projected to be greater in the latter part of the 21st century, and
most changes were projected to occur in the Hilo/Hamakua and East Maui regions (Figures D3.1
– D3.4). Several high elevation reaches on the leeward sides of Kauai and Molokai were also
projected to change classes. Across time periods and for each climate scenario, the total length
of stream reaches classified as G, H and K were projected to decrease with changes in mean
annual rainfall, while classes L and I were projected to increase. The greatest absolute change in
total reach class length occurred to class I under the RCP 4.5 climate scenario (118 km gained),
but the greatest change relative to current reach length occurred in class K under the RCP 8.5
climate scenario (73 of 233 km, or a reduction of ~31%).
Changes in average annual or dry season rainfall were projected to exceed one standard
deviation from current estimates of mean annual rainfall for many stream reaches across the
Hawaiian Islands (Table 3.2). Most changes in annual and dry season rainfall occurred under the
late century RCP 8.5 scenario, with 2,419 km of streams projected to have mean annual rainfall
values lower than one standard deviation from current values, and 3,174 km of streams projected
to have dry season rainfall declines. Projected negative changes in annual and dry season rainfall

128

were more common the leeward side of each island where many intermittent reaches occur
(Figures D3.5 – D3.12).
Ranking reaches based on future conservation value
Mid-century time period
Overall, rankings of stream reaches based on mid-century projections suggest that each of
the five islands will continue to support areas of high conservation value (Figure 3.4). The
Kohala Mountains, northeastern Molokai, and the windward side of Kauai (denoted as
Windward Kauai) had many reaches ranked in the top 5 and top 10%. In addition, these highlyranked reaches tend to extend up to high elevations and include complete stream networks. Less
disturbed areas along the windward slopes of east and west Maui also contain several stream
networks that have many reaches ranked in the top 5 and 10%. Windward Kauai had the largest
percentage of total reach length within the top 20% of all ranked reaches across the islands
(26.2%), followed by leeward Oahu (15.6%) and the Kohala Mountains (13.4%). Windward
Kauai, Windward Oahu, and the Kohala Mountains had the highest percentages of total reaches
ranked within the top 20% (33.8, 30.3 and 28.8%). This is likely a reflection of multiple factors,
including a high percentage of perennial reaches within these regions, presence of less common
reach classes, and relatively low UHCI and climate exposure scores. In addition, high ranking of
low elevation reaches occurred in many areas across the islands, in some cases, even in stream
networks where upstream reaches had low rankings (i.e., Windward Kauai). Considering the
large number of perennial reaches along the north Hilo coast, a comparatively small percentage
of reaches within the region were ranked in the top 20% (6.6%).

129

Late-century time period
Overall, the rankings of stream reaches using late-century climate scenarios were similar
to those of the mid-century time period, but some regions were projected to have declines in
conservation value (Figure 3.5). Leeward Kauai, for example, had no stream reaches ranked in
the top 5%, and only 6 km of stream reaches ranked in the top 10% in the late-century ranking,
compared to 7.8 and 29.3 km in the mid-century ranking (Table 3.3). Similar to the mid-century
conservation value rankings, Windward Kauai, northeastern Molokai, and the Kohala Mountains
had a high a proportion of stream reaches ranking in the top 5 and 10%, but windward Kauai had
a greater length of reaches ranked in the top 10% compared to the mid-century rankings (239 vs.
277 km). However, several reaches along the southeastern Na Pali Coast show some declines in
stream ranking. In addition, Molokai had 11.4 km fewer of stream reaches ranked in the top 5%
in the late-century compared to the mid-century scenario (63.7 versus 52.3 km).
Differences in priority reaches in current vs. future rankings
Many of the reaches in the top 5 and 10% of the current conservation value ranking were
also in the top 5 and 10% in both the mid and late-century rankings (Figures 3.6 and 3.7), but
some notable differences occurred. For example, high elevation reaches of northeast Molokai
were in the top 10% of reaches in the current conservation ranking, but were not in the top 10%
in the mid and late-century rankings. In addition, reaches that drain leeward facing slopes on
Kauai and West Maui were less commonly selected in the top 10% in future conservation value
rankings than in the current conservation value ranking. Along the Hilo/Hamakua Coast of
Hawaii, high elevation reaches were less commonly ranked in the top 10% in future rankings
than in the current reach ranking. Along the windward coast of Oahu, however, several stream

130

reaches were ranked in the top 10% only in future conservation value rankings while few were
only selected in the top 10% in the current conservation value ranking.
Discussion
Given anticipated changes in climate, proactive strategies are needed to effectively
conserve stream organisms (Palmer et al., 2009; Zeigler et al., 2012). In this study, we used best
available information on the ecological potential of stream reaches, estimates of current habitat
condition, and projected changes in rainfall to identify Hawaiian stream reaches of high
conservation value under two future climate scenarios for two time periods. Summaries of
projected average annual and dry season rainfall within upstream catchments of reaches indicate
that many intermittent and perennial streams may be influenced by lower precipitation in the
future. The results of our future ranking scenarios indicate that areas of high conservation value
occur across all five of the largest Hawaiian Islands, and that several regions known to support
important habitat for native stream organisms will likely continue to be of high value in the
future, regardless of the climate scenario or time period considered. Comparison of future
conservation value rankings with the current conservation value ranking indicates that leeward
perennial streams and some high elevation reaches may lose value due to reductions in average
dry season or annual rainfall and, in some cases, due to changes in the ability to support the
assemblage of stream organisms that they currently support. Our results are particularly valuable
for developing proactive stream conservation strategies in Hawaii because they can be used to
consider the type and location of most effective conservation measures.
Ranking reaches based on current and future conservation value
The high ranking of many low elevation reaches across the study area results from
migratory limitations of Hawaii’s stream species as well as patterns in anthropogenic disturbance

131

to the catchment and channel. In this study, we defined conservation value as the ecological
uniqueness of a stream reach and prioritized those reaches that supported more taxa. Because
many low elevation reaches are associated with high numbers of unique taxa that are not likely to
migrate far upstream, such as Stenogobius hawaiiensis and Macrobrachium grandimanus (Kido,
2008; Parham et al, 2009; Tingley this volume a), the relatively high ranking of low-elevation
stream reaches is not surprising. In addition, land use and stream channel disturbance to
Hawaiian streams tends to be most severe in low-elevation areas near the coast (Crawford et al.,
2016). This limits the conservation value of many reaches in these areas, therefore magnifying
the importance of those low-elevation reaches that are in moderate to good condition in the
Zonation ranking procedure (De Minin et al., 2014). Application of the climate exposure score
also lowered the conservation value of reaches where changes in ecological potential or declines
in annual or dry season rainfall were projected to occur. This contributed to higher rankings of
perennial reaches draining windward slopes, given the lower number of reaches in these regions
projected to experience declines in rainfall as compared to leeward draining streams. This is
perhaps most apparent in the high ranking of many low elevation reaches along Windward Oahu,
which have lower UHCI scores than reaches in other regions but are unlikely to experience
reductions in rainfall (Table 3.3).
Highly-ranked stream reaches at higher elevations were not common in the future
rankings, but when they did occur, they were located in stream networks that were connected to
the marine environment by other highly-ranked reaches. For instance, streams located on the
north and northwest side of the windward coast of Kauai as well as in the Kohala Mountains
region of Hawaii Island were highly ranked through the majority of their stream networks. In
general, these regions contain low elevation reaches with low anthropogenic disturbance relative

132

to many other regions of the Hawaiian Islands (Crawford et al., 2016). The selection of reaches
throughout these stream networks indicates that accounting for the condition of reaches in the
downstream network in the ranking procedure effectively prioritizes streams that are likely to
support complete amphidromous migratory pathways in good condition. In some regions, such
as east Maui and on the island of Molokai, the prioritization of high elevation bogs may have
also resulted in the selection of entire watersheds, given their relative uniqueness compared to
other reach classes.
The future magnitude of greenhouse gas emissions is uncertain, but significant climate
change is now considered inevitable, and the implementation of proactive conservation strategies
requires that decisions be made despite uncertainty. Examination of overlap in reaches with high
conservation value from both climate scenario rankings at a given time period allowed for
identification of streams most likely to have high conservation value with changes in climate.
Reaches that were highly ranked in both scenarios are those that are most likely to retain
conservation value regardless of projected future climate change, suggesting that decision
makers can prioritize protection of these reaches under the assumption that they represent the
best opportunities to continue to support habitat for native stream organisms.
The comparison of rankings at the mid and late-century time periods offers two benefits.
First, it demonstrates that changes in rainfall may have greater effects on the conservation value
of specific reaches or regions as climate change progresses. For instance, a decline in
conservation value in reaches that drain the leeward slopes of Kauai and West Maui is indicative
of the “wet getting wetter, dry getting drier” outcome anticipated to be a general pattern in
precipitation change at a global scale (IPCC, 2013). Second, because levels of human induced
radiative forcing are relatively similar at the mid-century time period under all four of the RCP

133

scenarios (IPCC, 2013), use of the late-century projection demonstrates how differences in the
magnitude of anthropogenic forcing may differentially affect the long-term conservation value of
Hawaiian streams.
Differences in priority reaches in current vs. future rankings
Comparison of the current ranking with future rankings of conservation value
demonstrates how accounting for changes in ecological potential and exposure to changing
climate can affect conservation prioritization. Differences are most pronounced on the leeward
sides of Kauai and West Maui, where many reaches are projected to experience declines in either
annual or dry season rainfall, and sometimes both. Declines in the conservation value of these
reaches likely contributed to observed increases in the ranking of reaches along windward sides
of islands in areas that have ecologically similar reach classes to those that drain leeward slopes.
Several reaches at high elevations on Molokai and in the Hilo/Hamakua region of Hawaii Island
are ranked in the top 10% in the current condition ranking but not in future rankings. This is
likely the result of a shift in the ecological potential of stream reaches resulting from projected
changes in average annual rainfall, which will potentially alter the current ability of these
streams to support current assemblages.
Implications for the proactive conservation of Hawaiian streams
Our results can be used to assess what conservation actions may be most effective in
different regions of the Hawaiian Islands. Proactive conservation approaches that include
preservation of habitats or establishment of protected areas may be most appropriate in regions
where entire stream networks consist of highly ranked reaches under multiple climate scenarios.
In addition, comparison of current and future rankings can aid in conservation decision making.
For example, in streams projected to have reduced conservation value due to declines in average

134

dry season or annual rainfall, conservation approaches that support maintenance of current
baseflow (i.e., prevention of additional water diversions) can increase the likelihood that these
reaches retain habitat for native stream species in the future. In addition, if flow in these streams
is currently diverted, as it is along much of west Maui (Cheng, 2014), the removal or reduction
of water diversions may be an effective management action that buffers streams from future
changes in climate. The identification of regions where many reaches were highly ranked only
in future rankings can also help guide conservation actions. Streams in these regions may have
lower habitat condition scores due to anthropogenic disturbance to the landscape and channel,
but are unlikely to experience reductions in rainfall. This suggests that in regions like Windward
Oahu, the restoration or rehabilitation of stream habitats may be an effective proactive
conservation strategy.
Consideration of approach and future directions
Using multiple emission scenarios when conducting spatial prioritizations based on future
conditions is necessary given the uncertainty associated with future climate change. While we
addressed this in our approach by ranking reaches using two greenhouse gas emission scenarios,
two additional scenarios (RCP 2.6 and 6.0) exist that are not currently downscaled for the
Hawaiian Islands. We therefore recommend that when downscaled versions of these data
become available, they be used in our spatial prioritization approach to offer decision makers a
broader depiction of potential future conditions when developing proactive conservation
strategies.
In this study, we used the finest resolution, statistically downscaled climate data available
for the Hawaiian Islands. Statistical downscaling requires the assumption of “stationarity,”
meaning that current factors that control climate in a region remain the same through time

135

(Timm and Diaz, 2009). If regional controls on climate are altered as a result of climate change,
biases in both the direction (i.e., increases vs. decreases in rainfall) and magnitude of projections
developed using statistical downscaling will occur (Timm et al., 2015). Other approaches, such
as dynamic downscaling, do not assume stationarity and will likely be available for the Hawaiian
Islands soon (e.g., Zhang et al., 2016). Therefore, we suggest that as downscaled climate data
are made available, additional rankings are developed using our approach and compared with the
results of our study.
The climate exposure score was developed based on two factors: 1) whether a stream
reach changed in class, therefore representing a shift in ecological potential and its ability to
support native assemblages of organisms, and 2) whether it was projected to experience a
reduction in average annual or dry season rainfall greater than one standard deviation from the
current mean value. While the responses of stream flow and populations of stream organisms to
differences in annual or dry season rainfall have not been studied explicitly in all regions of
Hawaii, current research suggests that declines in rainfall will lead to declines in groundwater
recharge and subsequent reductions in baseflow across the islands (Bassiouni and Oki, 2013).
The assumption that any substantial reduction in rainfall will have negative consequences for
stream organisms in Hawaii is well-founded, but increased study of specific effects of changes in
the flow regime on stream organisms or stream habitat can be useful to increase our
understanding of potential mechanisms by which climate change can impact streams.
The effective conservation of stream ecosystems requires that we continue to develop
approaches that aid in prioritization of conservation actions in the face of changing climate. In
this study, we incorporated the best available information on climate change effects on stream
systems to identify areas for conservation and suggest specific conservation actions. Our results

136

draw attention to potential consequences of climate change for Hawaiian streams and provide a
template for similar studies in other tropical island systems. In addition, this approach can be
easily extended to include new information on ecological potential, current habitat condition and
projected changes in climate. We therefore view our work as a building block for future studies
that develop a dynamic picture of climate change effects in Hawaii, which can then be used to
direct conservation actions to support the unique organisms of these stream ecosystems.

137

APPENDICES

138

APPENDIX 3.A

Tables

139

Table 3.1. Stream classes, taxa association weightings and taxa associated with each class
resulting from the Hawaii stream reach classification. Associated taxa are defined by prevalence
within a given class compared to prevalence across all samples reaches (Tingley this volume a).
Taxa
Reach
association
type
Description
weighting Associated taxa
A
Coastal
4
S. stimpsoni, A. stamineus, N. granosa,
E. sandwicensis, S. hawaiiensis, M.
grandimanus, Kuhlia sp.
B

Low gradient downstream
channel at low elevation

3

A. stamineus, M. grandimanus

C

High gradient downstream
channel at low elevation

4

L. concolor, S. stimpsoni, A. stamineus,
N. granosa, E. sandwicensis

D

3

S. stimpsoni, A. stamineus

4

A. bisulcata, L. concolor, S. stimpsoni,
A. stamineus, N. granosa

3

A. bisulcata, L. concolor

0

No taxa

2

A. bisulcata

3

A. bisulcata, L. concolor

3

A. bisulcata, L. concolor

K

Low gradient, moderate
elevation, potential high
water table
High gradient, moderate
elevation, potential high
water table
High gradient, high
elevation, potential high
water table
Moderate to high
elevation, low rainfall
Moderate to high
elevation, moderate
rainfall
Moderate to high
elevation, high rainfall
Terminal falls at low
elevation
High elevation bogs

1

No taxa

L

Headwater reaches

1

No taxa

M

Intermittent reaches

0

No taxa

E

F

G
H

I
J

140

Table 3.2. Changes in stream reach length for each class under each climate scenario and time period. Also shown are total lengths of
stream reaches projected to be greater than one standard deviation from current mean annual or dry season rainfall under each scenario
and time period.

Rainfall
Current
Mid century
RCP 4.5
RCP 8.5
Late century
RCP 4.5
RCP 8.5

Class G Class H
(km)
(km)
849
321

Class I
(km)
418

Class K Class L
(km)
(km)
233
783

Annual rainfall
(-/+ 1 standard
deviation; km)

Dry season rainfall
(-/+ 1 standard
deviation; km)

-24
-24

-71
-77

96
102

-25
-36

25
36

238/334
477/316

521/559
1,334/547

-40
-23

-77
-87

118
110

-41
-73

41
73

573/570
2,419/415

909/589
3,174/754

141

Table 3.3. Total length of stream reaches that overlap in the top 20% of ranked reaches at each time period for the two climate
scenarios. The percent of total reach length that is within the top 20% within a region and across islands as well as the mean, standard
deviation and range of UHCI scores within a region are also shown.
Top Top 20%
Total Perennial Top
90.0- 80.0Top 20% in across
Mean
reach length length >95.0% 94.9% 89.9% 20.0% region islands UHCI
(km)
(km)
(km)
(km) (km) (km) (%)
(%)
score

Period Region
Mid-century rankings
Island of Kauai
Leeward Kauai
Windward Kauai
Island of Oahu
Leeward Oahu
Windward Oahu
Molokai
Island of Maui
West Maui
East Maui
Island of Hawaii
Hilo/Hamakua
Kohala
Kona
Late-century rankings
Island of Kauai
Leeward Kauai
Windward Kauai
Island of Oahu

Standard
deviation
of UHCI
score

Range of
UHCI
score

2215
856
1359
437
1717
437
964
1870
684
1187
3950
1884
814
1252

1544
582
962
214
627
214
230
574
186
388
1819
1243
569
7

129
8
121
22
17
22
66
34
10
24
72
12
60
0

147
29
118
36
113
36
57
74
19
55
88
30
58
0

343
122
221
74
145
74
48
109
47
62
184
67
117
0

619
159
460
132
274
132
168
218
76
142
342
108
234
0

27.9
18.6
33.8
18.9
16.0
30.3
17.4
11.7
11.2
11.9
8.7
5.7
28.8
0.0

35.3
9.1
26.2
23.2
15.6
7.5
9.6
12.4
4.4
8.1
19.5
6.2
13.4
0.0

0.88
0.87
0.89
0.83
0.83
0.85
0.97
0.91
0.90
0.93
0.86
0.85
0.85
0.89

0.101
0.072
0.114
0.146
0.151
0.134
0.053
0.101
0.099
0.101
0.117
0.108
0.126
0.123

0.37-1.00
0.44-1.00
0.37-1.00
0.00-1.00
0.00-1.00
0.26-1.00
0.62-1.00
0.53-1.00
0.53-1.00
0.56-1.00
0.35-1.00
0.35-1.00
0.47-1.00
0.37-1.00

2215
856
1359
437

1544
582
962
214

133
0
133
24

150
6
144
35

331
154
177
66

614
160
454
125

27.7
18.8
33.4
18.6

33.8
8.8
25.0
22.1

0.88
0.87
0.89
0.83

0.101
0.072
0.114
0.146

0.37-1.00
0.44-1.00
0.37-1.00
0.00-1.00

142

Table 3.3. (cont’d)
Top Top 20%
Total Perennial Top
90.0- 80.0Top 20% in across
Mean
reach length length >95.0% 94.9% 89.9% 20.0% region islands UHCI
(km)
(km)
(km)
(km) (km) (km) (%)
(%)
score

Period Region
Late-century rankings
Leeward Oahu
Windward Oahu
Molokai
Island of Maui
West Maui
East Maui
Island of Hawaii
Hilo/Hamakua
Kohala
Kona

1717
437
964
1870
684
1187
3950
1884
814
1252

627
214
230
574
186
388
1819
1243
569
7

17
24
52
30
10
20
64
6
58
0

113
35
78
67
25
42
126
35
91
0

145
66
29
129
54
75
228
138
90
0

143

275
125
159
225
89
136
418
178
240
0

16.0
28.5
16.5
12.0
13.0
11.5
10.6
9.4
29.5
0.0

15.2
6.9
8.7
12.4
4.9
7.5
23.0
9.8
13.2
0.0

0.83
0.85
0.97
0.91
0.90
0.93
0.86
0.85
0.85
0.89

Standard
deviation
of UHCI
score

Range of
UHCI
Score

0.151
0.134
0.053
0.101
0.099
0.101
0.117
0.108
0.126
0.123

0.00-1.00
0.26-1.00
0.62-1.00
0.53-1.00
0.53-1.00
0.56-1.00
0.35-1.00
0.35-1.00
0.47-1.00
0.37-1.00

APPENDIX 3.B

Figures

144

Figure 3.1. The five Hawaiian Islands considered for spatial prioritization of stream reaches
145

Figure 3.2. Spatial prioritization approach taken for the ranking of stream reaches given current
ecological potential and habitat condition and with projected changes in rainfall.

146

Figure 3.3. Reach classes used to estimate ecological potential (From Tingley this volume a). Intermittent streams not shown.

147

Figure 3.4. Spatial prioritization ranking of stream reaches under mid-century projected climate scenarios. Results were applied to
local catchments for ease of interpretation. Rankings reflect those reaches that were above a given percentage in both scenarios.
148

Figure 3.5. Spatial prioritization ranking of stream reaches under late-century projected climate scenarios. Results were applied to
local catchments for ease of interpretation. Rankings reflect those reaches that were above a given percentage in both scenarios.
149

Figure 3.6. Similarities and differences in the top 5 and 10% ranked sites under current condition and mid-century scenarios
150

Figure 3.7. Similarities and differences in the top 5 and 10% ranked sites under current condition and late-century scenarios

151

APPENDIX 3.C

Supplemental tables

152

Table C3.1. Similarity matrix applied to stream reach classes
Stream
type
A
B
C
D
E
F
G
H
A
1.00 0.29 0.50 0.29 0.33 0.00 0.00 0.00
B
1.00 0.29 0.33 0.20 0.00 0.00 0.00
C
1.00 0.40 0.67 0.17 0.00 0.00
D
1.00 0.40 0.00 0.00 0.00
E
1.00 0.40 0.00 0.20
F
1.00 0.00 0.50
G
1.00 0.00
H
1.00
I
J
K
L
M

153

I
0.00
0.00
0.11
0.00
0.40
1.00
0.00
0.50
1.00

J
0.00
0.00
0.11
0.00
0.40
1.00
0.00
0.50
1.00
1.00

K
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
1.00

L
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
1.00

M
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
1.00

APPENDIX 3.D

Supplemental figures

154

Figure D3.1. Changes in stream class under RCP 4.5 climate scenario for the mid-century time period.. “Negative” refers to a change
in class resulting from a decline in rainfall, while “Positive” indicates a change in class resulting from an increase in rainfall.
155

Figure D3.2. Changes in stream class under RCP 4.5 climate scenario for the late-century time period.. “Negative” refers to a change
in class resulting from a decline in rainfall, while “Positive” indicates a change in class resulting from an increase in rainfall.
156

Figure D3.3. Changes in stream class under RCP 8.5 climate scenario for the mid-century time period.. “Negative” refers to a change
in class resulting from a decline in rainfall, while “Positive” indicates a change in class resulting from an increase in rainfall.
157

Figure D3.4. Changes in stream class under RCP 8.5 climate scenario for the late-century time period.. “Negative” refers to a change
in class resulting from a decline in rainfall, while “Positive” indicates a change in class resulting from an increase in rainfall.
158

Figure D3.5. Stream reaches that increased or decreased by greater than one standard deviation from current conditions. during the dry
season under the RCP 4.5 climate scenario for the mid-century time period.
159

Figure D3.6. Stream reaches that increased or decreased by greater than one standard deviation from current conditions. during the dry
season under the RCP 4.5 climate scenario for the late-century time period.
160

Figure D3.7. Stream reaches that increased or decreased by greater than one standard deviation from current conditions. during the
dry season under the RCP 8.5 climate scenario for the mid-century time period.
161

Figure D3.8. Stream reaches that increased or decreased by greater than one standard deviation from current conditions. during the dry
season under the RCP 8.5 climate scenario for the late-century time period.
162

Figure D3.9. Stream reaches that increased or decreased by greater than one standard deviation from current annual mean annual
rainfall under the RCP 4.5 climate scenario for the late-century time period.
163

Figure D3.10. Stream reaches that increased or decreased by greater than one standard deviation from current annual mean annual
rainfall under the RCP 8.5 climate scenario for the mid-century time period.
164

Figure D3.11. Stream reaches that increased or decreased by greater than one standard deviation from current annual mean annual
rainfall under the RCP 8.5 climate scenario for the late-century time period.
165

Figure D3.12. Stream reaches that increased or decreased by greater than one standard deviation from current annual mean annual
rainfall under the RCP 4.5 climate scenario for the mid-century time period.
166

LITERATURE CITED

167

LITERATURE CITED

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.
Bassar, R. D., B. H. Letcher, K. H. Nislow & A. R. Whiteley, 2016. Changes in seasonal climate
outpace compensatory density-dependence in eastern brook trout. Global Change Biology
22: 577-593.
Bassiouni, M. & D. S. Oki, 2013. Trends and shifts in streamflow in Hawai'i, 1913-2008.
Hydrological Processes 27: 1484-1500.
Battin, J., M.W. Wiley, M. H. Ruckelshaus, R. N. Palmer, E. Korb, K. K. Bartz & H. Imaki,
2007. Projected impacts of climate change on salmon habitat restoration. Proceedings of
the National Academy of Sciences of the United States of America 104: 6720-6725.
Bebler, M. H. & D. W. Foltz, 2004. Genetic diversity in Hawaiian stream
macroinvertebrates. Micronesica 37: 119-128.
Bond, N. R., J. R. Thomson & P. Reich, 2014. Incorporating climate change in conservation
planning for freshwater fishes. Diversity and Distributions 20: 931-942.
Brasher, A. M. D., 2003. Impacts of human disturbances on biotic communities in Hawaiian
streams. Bioscience 53: 1052-1060.
Bush, A. A., D. A. Nipperess, D. E. Duursma, G. Theischinger, E. Turak & L. Hughes, L., 2014.
Continental-scale assessment of risk to the Australian odonata from climate change. Plos
One: 9.
Cheng, C.L., 2014. Low-flow characteristics of streams in the Lahaina District, West Maui,
Hawai‘i: U.S. Geological Survey Scientific Investigations Report 2014–5087. U.S.
Geological Survey, Washington, DC.
Chubb, A. L., R. M. Zink & J. M. Fitzsimons, 1998. Patterns of mtDNA variation in Hawaiian
freshwater fishes: The phylogeographic consequences of amphidromy. Journal of
Heredity 89: 8-16.
Comte, L., L. Buisson, M. Daufresne & G. Grenouillet, 2013. Climate-induced changes in the
distribution of freshwater fish: observed and predicted trends. Freshwater Biology 58:
625-639.
Cordeiro, J. & K. Perez, 2012. Neritina granosa. The IUCN Red List of Threatened Species
2012. http://www.iucnredlist.org Accessed 17.02.21.

168

Crawford, S., G. Whelan, D. M. Infante, K. Blackhart, W. M. Daniel, P. L. Fuller, T. Birdsong,
D. J. Wieferich, R. McClees-Funinan, S. M. Stedman, K. Herreman & P. Ruhl, 2016.
Through a Fish's Eye: The Status of Fish Habitats in the United States 2015. National
Fish Habitat Partnership. http://assessment.fishhabitat.org/ Accessed 17.02.15.
Danz, N. P., G. J. Niemi, R. R. Regal, T. Hollenhorst, L. B. Johnson, J. M. Hanowski & G. E.
Host, 2007. Integrated measures of anthropogenic stress in the US Great Lakes
basin. Environmental Management 39: 631-647.
De Grave, S. & X. Cai, 2013. Atyoida bisulcata. IUCN 2013. IUCN Red List of Threatened
Species, Version 2013.2. http://www.iucnredlist.org Accessed 13.01.24.
Di Minin, E., V. Veach, J. Lehtomäki, F. Montesino-Pouzols & A. Moilanen, 2014. A quick
introduction to Zonation.
http://cbig.it.helsinki.fi/files/zonation/Z_quick_intro_manual_B5_final_3.pdf> Accessed
17.01.01.
Esselman, P. C., D. M. Infante, L. Z. Wang, D. Y. Wu, A. R. Cooper & W. W. Taylor, 2011. An
index of cumulative disturbance to river fish habitats of the conterminous United States
from landscape anthropogenic activities. Ecological Restoration: 29: 133-151.
Frazier, A. G., T. W. Giambelluca, H. F. Diaz & H. L. Needham, 2015. Comparison of
geostatistical approaches to spatially interpolate month-year rainfall for the Hawaiian
Islands. International Journal of Climatology 12: 1459-1470.
Frazier, A. G. & T. W. Giambelluca, 2016. Spatial trend analysis of Hawaiian rainfall from
1920-2012. International Journal of Climatology
Giambelluca, T. W., J. K. Delay, M. A. Nullet, M. A. Scholl & S. B. Gingerich, 2011. Canopy
water balance of windward and leeward Hawaiian cloud forests on Haleakala, Maui,
Hawai'i. Hydrological Processes 25: 438-447
Hari, R. E., D. M. Livingstone, R. Siber, P. Burkhardt-Holm & H. Guttinger, 2006.
Consequences of climatic change for water temperature and brown trout populations in
Alpine rivers and streams. Global Change Biology 12: 10-26.
Hawaii Fish Habitat Partnership, 2010. Hawaii Fish Habitat Partnership Strategic Plan.
https://www.fws.gov/pacificislands/Publications/Hawaii%20FHP%20Strategic%20Plan.p
df Accessed 17.01.01.
Helfman, G. S., 2007. Fish conservation: a guide to understanding and restoring global aquatic
biodiversity and fishery resources. Island Press, Washington, D.C.
IPCC, 2013. Climate Change 2013: The Physical Science Basis. Contribution of Working Group
I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change.
169

Stocker T. F., D. Qin, G. K. Plattner, M. Tignor, S. K. Allen, J. Boschung, A. Nauels, Y.
Xia, V. Bex & P. M. Midgley (ed.). Cambridge, United Kingdom and New York, NY.
Izuka, S. K., J. A. Engott, M. Bassiouni, A. G. Johnson, L. D. Miller, K. Rotzoll & A. Mair,
2015. Volcanic aquifers of Hawai’i – Hydrogeology, water budgets and conceptual
models, Scientific Investigations Report 2015-5164. U.S. Geological Survey,
Washington, DC.
Jelks, H. L., S. J. Walsh, N. M. Burkhead, S. Contreras-Balderas, E. Diaz-Pardo, D. A.
Hendrickson, J. Lyons, N. E. Mandrak, F. McCormick, J. S. Nelson, S. P. Platania, B. A.
Porter, C. B. Renuad, J. J. Schmitter-Soto, E. B. Taylor & M. L. Warren, 2008.
Conservation status of imperiled North American freshwater and diadromous fishes.
Fisheries 33: 372-407.
Jones, K. R., J. E. M. Watson, H. P. Possingham & C. J. Klein, 2016. Incorporating climate
change into spatial conservation prioritisation: A review. Biological Conservation 194:
121-130.
Kido, M. H., 2008. A persistent species assemblage structure along a Hawaiian stream from
catchment-to-sea. Environmental Biology of Fishes 82: 223-235.
Kinzie, R. A., 1990. Amphidromous macrofauna of Hawaiian Island streams species profiles:
Life histories and environmental requirements of coastal vertebrates and invertebrates:
Pacific Oceans region, Enivronmental Impact Research Program Technical Report EL89-10. U. S. Army Corps of Engineers, Washington, D.C.
Lau, L. S. & J. F. Mink, 2006. Hydrology of the Hawaiian Islands. University of Hawaiˈi Press,
Honolulu, HI.
Leathwick, J. R., A. Moilanen, S. Ferrier & K. Julian, 2010. Complementarity-based
conservation prioritization using a community classification, and its application to
riverine ecosystems. Biological Conservation 143: 984-991.
McDowall, R. M., 1988. Diadromy in shes: migrations between marine and freshwater
environments. Timber Press, Portland, OR.
McDowall, R. M., 2003. Hawaiian biogeography and the islands' freshwater fish fauna. Journal
of Biogeography 30:703-710.
McRae, M. G., L. B. McRae & J. M. Fitzsimons, 2011. Habitats used by juvenile flagtails
(Kuhlia spp.; Perciformes: Kuhliidae) on the Island of Hawai'i. Pacific Science 65: 441450.
Moilanen, A. & H. Kujala, 2008. Zonation spatial conservation planning framework and
software v. 2.0, User manual http:/www.helsinki.fi/bioscience/consplan/ Accessed
17.01.01.
170

Moilanen, A, L. Meller, J. Leppanen, F. M. Pouzols, A. Arponen, H. Kujala, 2014. Zonation
spatial conservation planning framework and software v. 4.0, User Manual. University of
Helsinki, Helsinki, Finland.
Nishimoto, R. T., T. E. Shimoda & L. K. Nishiura, 2007. Mugilids in the Muliwai: a Tale of Two
Mullets. Bishop Museum Bulletin in Cultural and Environmental Studies 3.
Palmer, M. A., D. P. Lettenmaier, N. L. Poff, S. L. Postel, B. Richter & R. Warner, 2009.
Climate change and river ecosystems: Protection and adaptation options. Environmental
Management 44: 1053-1068.
Parham, J. E., G. R. Higashi, R. T. Nishimoto, S. Hau, D. G. K. Kuamo'o, L. K. Nishiura, T. S.
Sakihara, T. E. Shimoda & T. T. Shindo, 2009. The use of Hawaiian Stream Habitat
Evaluation Procedure to provide biological resource assessment in support of instream
flow standards for East Maui streams. Bishop Museum and the Division of Aquatic
Resources, Honolulu, HI.
Strayer, D. L., 2006. Challenges for freshwater invertebrate conservation. Journal of the North
American Benthological Society 25: 271-287.
Tedesco, P. A., T. Oberdorff, J.-F. Cornu, O. Beauchard, S. Brosse, H.H. Durr, G. Grenouillet, F.
Leprieur, C. Tisseuil, R. Zaiss & B. Hugueny, 2013. A scenario for impacts of water
availability loss due to climate change on riverine fish extinction rates. Journal of
Applied Ecology 50: 1105-1115.
Timm, O., & H. F. Diaz, 2009. Synoptic-statistical approach to regional downscaling of IPCC
twenty-first-century climate projections: Seasonal rainfall over the Hawaiian Islands.
Journal of Climate 22: 4261-4280.
Timm, O. E., T. W. Giambelluca & H. F. Diaz, 2015. Statistical downscaling of rainfall changes
in Hawaiˈi based on the CMIP5 global model projections. Journal of Geophysical
Research-Atmospheres 120: 92-112.
Tingley, R. W., III, This volume a. Influences of natural landscape factors on tropical stream
organisms: An ecological classification of Hawaiian Island streams in A multi-scaled
research framework for conserving streams with changing climate, Doctoral
Dissertation. Michigan State University, East Lansing, MI.
Tingley, R. W., III, This volume b. Anticipating effects of changing climate on tropical island
streams: influences of stream flow on atyid mountain shrimp in A multi-scaled research
framework for conserving streams with changing climate, Doctoral Dissertation.
Michigan State University, East Lansing, MI.
Tsang, Y-P., D. Wieferich, K. Fang, D. M. Infante & A. R. Cooper, 2014. An efficient and
integrated approach of aggregating upstream catchment information. SpringerPlus 3: 589.
171

Wang, L. Z., D. Infante, P. Esselman, A. Cooper, D. Y. Wu, W. Taylor, D. Beard, G. Whelan &
A. Ostroff, 2011. A hierarchical spatial framework and database for the National River
Fish Habitat Condition Assessment. Fisheries 36: 436-449.
Walter, R. P., J. D. Hogan, M. J. Blum, R. B. Gagne, E. F. Hain, J. F. Gilliam & P. B. McIntyre,
2012. Climate change and conservation of endemic amphidromous fishes in Hawaiian
streams. Endangered Species Research 16: 261-272.
Ward, E. J., J. H. Anderson, T. J. Beechie, G. R. Pess & J. M. Ford, 2015. Increasing hydrologic
variability threatens depleted anadromous fish populations. Global Change Biology 21:
2500-2509.
World Conservation Monitoring Centre, 1996. Sicyopterus stimpsoni. IUCN Red List of
Threatened Species 1996. http://www.iucnredlist.org Accessed 17.02.21.
Xenopoulos, M. A., D. M. Lodge, J. Alcamo, M. Marker, K. Schulze & D. P. Van Vuuren, 2005.
Scenarios of freshwater fish extinctions from climate change and water
withdrawal. Global Change Biology 11: 1557-1564.
Zeigler, M. P., A. S. Todd & C. A. Caldwell, 2012. Evidence of recent climate change within the
historic range of Rio Grande Cutthroat Trout: Implications for management and future
persistence. Transactions of the American Fisheries Society 141: 1045-1059.
Zhang, C., Y. Wang, K. Hamilton & A. Lauer, 2016. Dynamical downscaling of the climate for
the Hawaiian Islands Part I: Present-day. Journal of Climate.

172

CHAPTER FOUR
CONCLUSIONS: SUMMARY OF FINDINGS AND IMPLICATIONS FOR MANAGEMENT
In this chapter, I synthesize findings described in Chapters 1, 2 and 3. In addition, I
describe potential management implications for each chapter, with an emphasis on the
conservation of Hawaiian streams and their native organisms.
Principal findings
Chapter One: Influences of natural landscape factors on tropical stream organisms: An
ecological classification of Hawaiian Island streams
Using currently available and newly developed spatial datasets, I explored influences of
natural landscape factors on distributions of native stream taxa, which provided insight into the
ecological potential of stream reaches across the Hawaiian Islands for supporting those taxa. My
results indicate that stream size, channel slope, elevation, soil permeability, annual rainfall, and
waterfall height all explained a substantial amount of variance in distributions of stream taxa in
Hawaii. I then used a conditional inference tree to group stream reaches based on influences of
individual natural landscape factors on taxa distributions, which resulted in groupings defined by
differences in elevation, rainfall, channel slope and soil permeability. I used these results to
develop a classification of 13 stream classes that, when extrapolated across the study region,
spatially identify differences in the ecological potential of stream reaches. My study culminated
in the first stream reach-scale ecological classification for the Hawaiian Islands. My approach
effectively integrated existing knowledge on influences of natural landscape characteristics on
distributions of stream organisms with new insight into important factors influencing taxa
distributions at the reach scale. This study demonstrates the utility of implementing a landscape
approach when assessing effects of natural landscape factors on stream organisms.

173

The results of this study also suggest that sensitivity of stream reach ecological potential
to changing climate in Hawaii will depend on both relative exposure to change (i.e., the
magnitude of change in rainfall) and the underlying characteristics of the natural landscape.
Three stream classes associated with differences in organism distributions were all grouped
based on statistically significant relationships with mean annual rainfall. These climate-sensitive
classes occur at moderate to high elevations and likely lack substantial contributions of
groundwater to baseflow, as suggested by comparatively lower soil permeability in their
catchments. This suggests that stream reaches within these classes may change in their
ecological potential if mean annual rainfall changes with climate change. Overall, my results
add new insight into the sensitivity of stream habitat in Hawaii to changes in climate.
Chapter Two: Anticipating the effects of changing climate on tropical island streams:
Influences of stream flow on atyid mountain shrimp
I assessed differences in population characteristics of an endemic mountain shrimp,
Atyoida bisulcata¸ across 11 study streams with similar geology, topography, land cover,
elevation and soil characteristics but that differed substantially in mean annual rainfall. I
summarized daily stream flow data from each of the streams to assess differences in the flow
regime across sites and years. My results indicate that mean annual rainfall has strong
associations with multiple aspects of the flow regime including baseflow, flow variability, and
frequency and duration of flow events. Several of these flow metrics were also strongly
associated with habitat types available to A. bisulcata. Based on relationships between baseflow
and measures of average size and relative mass, I determined that individual A. bisulcata in
streams with lower baseflow are generally smaller and in poorer condition than in streams with
higher baseflow. Baseflow and the frequency of low flows were also associated with differences

174

in biomass, which was highest in streams with the highest baseflow and approached zero in
several reaches where low flows were frequent. Observed differences may be related to
temporary habitat desiccation, less food availability or higher water temperature stream resulting
from lower baseflow. I also documented increasing prevalence of the carapace infection brown
spot disease in streams with greater flow variability. My results suggest that even moderate
reductions in rainfall may affect populations of native stream organisms through changes to
multiple aspects of the flow regime. This highlights the importance of considering measures
beyond the presence or absence of species when assessing potential effects of changing climate
on stream organisms.
Chapter Three: Spatial prioritization of Hawaii’s stream ecosystems for conservation in
the context of changing climate
Using information on the ecological potential of stream reaches, current habitat condition
and projected changes in rainfall, I implemented a spatial prioritization approach that culminated
in the ranking of stream reaches based on estimates of current and future conservation value for
two time periods. Overall, my results indicate that areas of high conservation value will occur on
all five of the largest Hawaiian Islands under both moderate and high emission scenarios.
Despite projected changes in rainfall, many regions that are currently considered to be of high
conservation value will likely continue to be of high value through the 21st century when
considering climate change as the only driver. A comparison of rankings at mid to late-century
time periods indicates that streams that drain leeward slopes may have limited conservation
value due to substantial reductions in rainfall, and the severity of this effect will be dependent on
future greenhouse gas concentrations. A comparison of reach rankings generated with future
conditions to rankings generated only with current conditions indicated that leeward perennial

175

streams may lose value in the future due to reductions in average dry-season and annual rainfall.
My results highlight the importance of considering future effects of climate change when
developing strategies for stream conservation. Further, these collective findings can provide
guidance for conservation strategies across the Hawaiian Islands.
Management implications
Chapter One
My results show influences of natural landscape factors on taxa distributions and allow
for a representation of ecological potential of reaches across the Hawaiian Islands. The
development of an ecological classification from these results allowed for a visual representation
of locations of rare and common habitats across all streams in Hawaii. When paired with data
describing habitat condition of stream reaches resulting from anthropogenic disturbances, the
classification can be used to assess the relative condition of different habitats, which has utility
when considering the value of particular conservation actions. In addition, by broadly
controlling for natural variation across the study region, the reach classification can be valuable
in selecting sites for examining effects of landscape or channel disturbance on existing
populations of stream organisms.
The classification also highlights the sensitivity of particular regions of Hawaii to
changes in climate. For instance, streams of northeast Maui and the Hilo/Hamakua coast of
Hawaii Island include many reaches that are classified based on differences in mean annual
rainfall, suggesting that streams in these areas may be sensitive to reductions in rainfall.
Therefore, conservation actions should be taken to maintain and restore as much flow as possible
to these reaches to retain their ecological potential and buffer them against future climate change.

176

The classification is an effective way to represent differences in habitat across Hawaii,
but we also recognize that certain stream habitats (i.e., high elevation reaches, hard to access
reaches deep within stream valleys) may be underrepresented in the biological dataset.
Therefore, targeted sampling efforts in these regions may help identify additional landscape
influences on stream taxa not captured in the presence/absence database used in this study which
can further improve the value of the classification to management. In addition, my study
identified relationships between multiple natural landscape factors that have significant
correlations with the distribution and abundance of stream organisms, which in some cases draws
attention to important data needs across Hawaii. For instance, upstream minimum hydrologic
soil grouping effectively captured upstream contributions of groundwater to baseflow from high
elevation dikes, but geologic anomalies in Hawaii may result in site-specific differences in
groundwater resources (e.g., perched groundwater; Izuka et al., 2015). The development of an
island-wide spatial data layer describing groundwater delivery or identifying unique geologic
features comprehensively across the islands could be beneficial to improving understanding of
site-specific ecological potential.
Chapter Two
Observed declines in the condition and size of individual A. bisulcata with diminishing
flow are indicative of the vulnerability of Hawaiian stream organisms to reduced baseflow.
Protecting streams in Hawaii from water diversions or other disturbances that may reduce
baseflow is a major focal point for management in Hawaii (Nishimoto & Fitzsimons, 2006). My
results suggest that populations of stream organisms will be in the best condition when baseflow
is highest, implying that efforts should be taken to maintain and restore flows to streams and not
simply manage for minimum flows. Given that populations of A. bisulcata were largest and

177

biomass was greatest in streams with highest flows, I infer that reductions in baseflow may have
negative effects on populations of A. bisulcata. This information can help direct management
action (e.g., preservation vs. restoration) when paired with projected rainfall data.
In this study, I focus on a single species endemic to the Hawaiian Islands, but my results
indicate that effective conservation requires continued research on effects of changes in the
stream flow regime on populations of other Hawaiian stream organisms. While I broadly
generalize that reductions in stream flow may influence other species, understanding specific
effects of the flow regime on other species will improve conservation efforts. Using the
classification of stream reaches developed in Chapter 1 to account for patterns in natural
variability can be valuable in determining areas where observational studies like the one
described in this chapter can be used to assess finer resolution changes to populations of other
Hawaiian stream organisms. In addition, many tropical island systems support similar species
assemblages to that of Hawaii, suggesting that my results may be broadly applicable for
understanding effects of declining baseflow and increasing flow variability on other species
found in tropical island streams.
Chapter Three
The results of this study demonstrate the importance of considering changes in climate
when assessing the conservation value of streams in Hawaii. Streams in Hawaii and the
organisms they support are sensitive to reductions in rainfall. Therefore, considering future
changes in rainfall is an essential component when developing proactive conservation strategies.
The high ranking of many stream reaches that drain windward slopes under multiple climate
scenarios at different time periods indicates that these areas are likely to retain high conservation
value regardless of projected changes in rainfall. This result suggests that establishing protected

178

areas or preserving catchments with many high ranking reaches may be an effective form of
conservation. Areas that are ranked lower in future rankings when compared to the ranking
based on current conservation value may be more suitable for conservation actions that restore or
retain current flow, thereby buffering streams from anticipated effects of reduced rainfall.
Finally, those reaches that increase in conservation value and are currently moderately degraded
may be ideal locations for restoration actions that improve habitat, as they will likely be
minimally affected by future changes in rainfall.
While I focus on protection of total native stream biodiversity in my prioritization
analysis, the flexibility of my approach does allow for more targeted rankings that consider
specific reach classes or species of interest for conservation. For instance, researchers interested
in protection of Lentipes concolor, an endemic stream goby found in high elevation reaches,
could alter my prevalent taxa association weights and adjust climate exposure scores to favor
habitats that may support this species in the future. Furthermore, additional climate data from
dynamic downscaling approaches will become available in the near future (as described in Zhang
et al., 2016), thus offering further information to decision makers considering conservation
measures in Hawaii.

179

LITERATURE CITED

180

LITERATURE CITED

Izuka, S. K., J. A. Engott, M. Bassiouni, A. G. Johnson, L. D. Miller, K. Rotzoll & A. Mair,
2015. Volcanic aquifers of Hawai’i – Hydrogeology, water budgets and conceptual
models, Scientific Investigations Report 2015-5164. U.S. Geological Survey,
Washington, DC.
Nishimoto, R. T. & J. M. Fitzsimons, 2006. Status of Native Hawaiian Stream Fishes, A Unique
Amphidromous Biota. American Fisheries Society Symposium 53.
Zhang, C., Y. Wang, K. Hamilton & A. Lauer, 2016. Dynamical downscaling of the climate for
the Hawaiian Islands Part I: Present-day. Journal of Climate.

181