Utah State University Utah State University

DigitalCommons@USU DigitalCommons@USU

All Graduate Theses and Dissertations Graduate Studies

12-2018

Assessing the Ecological Implications of the Altered Flow and Assessing the Ecological Implications of the Altered Flow and

Sediment Regimes of the Rio Grande Along the West Texas-Sediment Regimes of the Rio Grande Along the West Texas-

Mexico Border Mexico Border

Demitra E. Blythe Utah State University

Follow this and additional works at: https://digitalcommons.usu.edu/etd

Part of the Terrestrial and Aquatic Ecology Commons

Recommended Citation Recommended Citation Blythe, Demitra E., "Assessing the Ecological Implications of the Altered Flow and Sediment Regimes of the Rio Grande Along the West Texas-Mexico Border" (2018). All Graduate Theses and Dissertations. 7358. https://digitalcommons.usu.edu/etd/7358

This Thesis is brought to you for free and open access by the Graduate Studies at DigitalCommons@USU. It has been accepted for inclusion in All Graduate Theses and Dissertations by an authorized administrator of DigitalCommons@USU. For more information, please contact digitalcommons@usu.edu.

ASSESSING THE ECOLOGICAL IMPLICATIONS OF THE ALTERED FLOW AND

SEDIMENT REGIMES OF THE RIO GRANDE ALONG THE WEST TEXAS-

MEXICO BORDER

by

Demitra E. Blythe

A thesis submitted in partial fulfillment

of the requirements for the degree

of

MASTER OF SCIENCE

in

Aquatic Ecology

Approved:

______________________ ____________________

Phaedra Budy, Ph.D. John C. Schmidt, Ph.D.

Major Professor Committee Member

______________________ ____________________

Janice Brahney, Ph.D. Laurens H. Smith, Ph.D.

Committee Member Interim Vice President for Research and

Dean of the School of Graduate Studies

UTAH STATE UNIVERSITY

Logan, Utah

2018

ii

Copyright © Demitra Blythe 2018

All Rights Reserved

iii

ABSTRACT

Assessing the ecological implications of the altered flow and sediment regimes of the Rio

Grande along the West Texas-Mexico border

by

Demitra E. Blythe, Master of Science

Utah State University, 2018

Major Professor: Dr. Phaedra Budy

Department: Watershed Sciences

Development of water resources alters flow and sediment regimes, often at the

expense of aquatic ecological integrity and diversity. Many riverine species within the

Rio Grande have been reduced or eradicated from the loss, or fragmentation, of viable

habitat, poor water quality, and spread of invasive species. The primary goal of our study

was to determine how the modified hydrologic and sediment regimes of the Rio Grande

have impacted the aquatic ecological integrity, with an emphasis on native fish diversity.

We used a multi-faceted approach to 1) examine the fish community between two

segments of the river with varying levels of degradation (e.g., Forgotten Reach and Big

Bend region) using a novel modeling approach, 2) quantify and describe the native fish

community and aquatic food web structure in Big Bend National Park, and 3) determine

if and how invasive vegetation impacts aquatic food resources. We found native fish

richness to be more influenced by channel narrowing and low flows in the Big Bend

region than in the Forgotten Reach. In contrast, water quality and channel narrowing

were the most influential variables describing native fish richness in the Forgotten Reach.

iv

We found fish abundance (based on catch per unit effort estimates) was greatest and

lowest in reaches we a priori classified as complex and simple, respectively, at the

microhabitat scale. At the macrohabitat scale, where we compared abundance between

alluvial valleys and canyons, fish abundance was greater in canyon reaches than in

alluvial reaches. Food web structure followed a similar pattern as fish abundance,

appearing to be more complex within complex and canyon sites, than simple and alluvial

sites. Contrary to what we expected, native vegetation and periphyton appear to be

contributing to the base of the aquatic food web more than nonnative vegetation.

Collectively, our results suggest the modern flow and sediment regimes may limit the

available habitat and potential food resources suitable for native fishes within the Rio

Grande. Native fish recovery and maintenance may depend largely on the effective

management of stream flow in the Rio Grande, and concordant changes in lower level

productivity and food availability.

(144 pages)

v

PUBLIC ABSTRACT

Assessing the ecological implications of the altered flow and sediment regimes of the Rio

Grande along the West Texas-Mexico border

Demitra E. Blythe

Large, exotic (those whose headwaters are in distant places) rivers are some of the

most unique and diverse ecosystems on earth. Because they often flow through a

multitude of biomes and climates, their waters are a vital resource not only for the

organisms that inhabit these rivers, but for human societies as well. Thus, large rivers,

like the Rio Grande, that flow through arid and agricultural regions are highly regulated

and diverted. Regulation and dewatering upset a river’s natural flow regime (e.g.,

magnitude, duration, timing of large flood events), subsequently impacting the river’s

ability to transport its sediment supply, and eventually perturbing a river into either

sediment surplus or deficit. The combination of altered flow and sediment regimes

influence the availability of habitat essential for the survival and viability of aquatic

organisms, such as fish and invertebrates. In addition, increased deposition of sediment

creates areas suitable for invasive riparian vegetation to establish, likely affecting habitat

complexity and increasing the abundance of leaf litter deposited into the river. The

altered flow and sediment regimes, in combination with invasive riparian vegetation,

culminate and eventually affect the food resources and aquatic communities present in a

river ecosystem. Most often, the links between the physical perturbations to a system

with the biological factors are poorly understood. In this study, we use distinct segments

of the Rio Grande along the US-Mexico border to compare areas with greater and lower

habitat heterogeneity, water quality, and invasive riparian species abundance to better

vi

understand what physical factors can influence aquatic species such as fish and

invertebrate communities. We identify critical limiting factors for the native fish

community present, and link the altered flow and sediment regimes with the aquatic

ecological template of the Rio Grande.

vii

ACKNOWLEDGMENTS

This work was funded by the US Geological Survey South Central Climate

Science Center, the National Park Service, and the Utah State University Ecology Center,

thank you for making this thesis possible. I want to give my husband, Todd Blythe, the

biggest thank you. His levelheadedness and his deep well of patience and encouragement,

gave me what I needed to pursue and persevere my Master’s degree. I am incredibly

thankful to both my advisor, Phaedra Budy, and Jack Schmidt (Todd’s advisor) for

allowing us the opportunity to work on a beautiful, unique river together. I also give my

gratitude to my parents, Gerald and Stephanie, they have always been incredibly

supportive of whatever I do in my life, and they gave me my passion for exploring the

outdoors and pushing my limits. To Phaedra, you have my sincerest respect and gratitude.

You have been a wonderful mentor, role model, friend, and absolute QUEEN of this

fisheries world! You gave me the extra boost of confidence to succeed in the fisheries

realm and to become more than I even expected of myself. To Jack, I am so thankful you

were a part of my project. You helped me think outside the box and pushed me to always

see the bigger picture, you have made me a stronger person because of it. A HUGE thank

you to Jeff Bennett, Dave Larson, Lydia Smith, and the rest of the Science Resource and

Management and Rio Cache staff. Your help and organization with our field trips was

absolutely necessary and would not have been possible without your generosity! To Brian

Laub, Dr. Laub-ster, thank you so much for your guidance, wisdom, and personality

throughout my project and on our field trips, you are awesome! A BIG thank you to my

fellow FEL and GLEL lab mates, your willingness to help with my field work in a remote

place and your wisdom and guidance throughout my project was so beneficial, I feel

viii

more confident in myself because of you guys! Finally, thank you to all the friends,

colleagues, and professors at USU who made being a part of the Natural Resources and

Watershed Department so much fun!

Demitra E. Blythe

ix

CONTENTS

Page

ABSTRACT ...................................................................................................................... iii

PUBLIC ABSTRACT ........................................................................................................v

ACKNOWLEDGMENTS ............................................................................................... vii

LIST OF TABLES ........................................................................................................... xii

LIST OF FIGURES ......................................................................................................... xiii

CHAPTER

1. INTRODUCTION .................................................................................................1

Literature Cited .................................................................................................5

2. REMEBERING A ‘FORGOTTEN’ RIVER: IDENTIFYING

CRITICAL LIMITING FACTORS FOR THE NATIVE FISH

DIVERSITY OF THE RIO GRANDE IN FAR WEST TEXAS ..........................7

Abstract ..............................................................................................................7

Introduction .......................................................................................................8

Contemporary Ecosystem of the Rio Grande ...........................................11

Hydrology ......................................................................................11

Geomorphology .............................................................................14

Associated Changes to the Ecosystem Structure and Function .................15

Methods ...........................................................................................................20

Study Area .................................................................................................20

Data Collection ..........................................................................................21

Fish community and water quality .................................................21

Hydrologic and geomorphic data ...................................................22

Structural Equation Modeling ....................................................................22

Determination of model fit.............................................................23

SEM Conceptual Model .............................................................................24

Defining direct hydrologic and geomorphic variables ...................24

Defining water quality as a latent variable ....................................24

Nonnative fish abundance ..............................................................25

Data Preparation.........................................................................................25

Results ..............................................................................................................26

Comparison of Big Bend Region and Forgotten Reach Models................26

Examining SEM Results Using Standardized Coefficients .......................27

Big Bend SEM model ....................................................................27

Forgotten Reach SEM results ........................................................28

x

Discussion ........................................................................................................29

Mechanisms Influencing Ecosystem Structure and Function ....................29

Identifying Critical Limiting Factors for Each Study Region ...................31

Comparing the Forgotten Reach and Big Bend Regions ...........................34

Model Limitations ......................................................................................35

Implications................................................................................................36

Acknowledgments............................................................................................37

Literature Cited ................................................................................................37

Tables and Figures ...........................................................................................46

3. TOO MUCH SEDIMENT, TOO LITTLE WATER: HOW THE

MODIFIED HYDROLOGIC AND SEDIMENT REGIMES OF THE

RIO GRANDE AFFECT THE NATIVE FISH COMMUNITY AND

FOOD WEB STRUCTURE IN BIG BEND NATIONAL PARK, TEXAS,

UNITED STATES ...............................................................................................53

Abstract ............................................................................................................53

Introduction ......................................................................................................55

Study Area .......................................................................................................60

Methods............................................................................................................63

Field Sampling ...........................................................................................63

Habitat complexity and water quality ............................................63

Fish community .............................................................................64

Primary productivity ......................................................................65

Terrestrial and aquatic invertebrates ..............................................66

Laboratory Methods ...................................................................................68

Chlorophyll a abundance and community analysis .......................68

Terrestrial and aquatic invertebrate community classification ......68

Stable isotope preparation ..............................................................70

Statistical Analyses ....................................................................................71

Random forest classification ..........................................................71

Fish community diversity ..............................................................73

Stable isotope analyses ..................................................................73

Results ..............................................................................................................77

Habitat Complexity ...........................................................................77

Fish Abundance and Diversity ..........................................................77

Abundance of Primary Productivity ..................................................78

Terrestrial and Aquatic Invertebrate Relative Abundance ................79

Stable isotope analyses ......................................................................80

Discussion ........................................................................................................82

Implications................................................................................................92

Acknowledgments............................................................................................93

Literature Cited ................................................................................................94

Tables and Figures .........................................................................................102

4. CONCLUSIONS ................................................................................................116

xi

APPENDIX...............................................................................................................120

xii

LIST OF TABLES

Table Page

2.1 Structural equation model fit statistics calculated

for the Big Bend region and Forgotten Reach ..............................................46

2.2 Structural equation model path coefficients

calculated for the Big Bend region and Forgotten Reach ............................47

3.1 Summary of the measured discharges occurring over

the entire sampling period in the Big Bend region .....................................102

3.2 Catch per unit effort summary statistics .....................................................102

3.3 Layman community fish community metrics .............................................103

3.4 Three-source mixing model summary results

for fish as consumers ..................................................................................103

3.5 Two-source mixing model summary results using

the Hydrogen isotope .................................................................................103

3.6 Three-source mixing model summary results

for aquatic invertebrates as consumers .......................................................103

A1 Average water quality parameters during all

sampling events ..........................................................................................121

A2 Average length and weight results for all fish

captured in complex and simple sites .........................................................123

A3 Average length and weight results for all fish

captured in alluvial and canyon sites ..........................................................124

A4 Total summed abundance of terrestrial invertebrates

captured during the sampling periods ........................................................125

A5 Average relative abundance for aquatic invertebrate

functional feeding groups ...........................................................................126

A6 Shannon’s Diversity Index calculated for terrestrial

invertebrates ...............................................................................................127

xiii

LIST OF FIGURES

Figure Page

2.1 Study area map for the Forgotten Reach

and Big Bend region .....................................................................................48

2.2 Conceptual diagram for the hypothesized

structural equation model .............................................................................49

2.3 Big Bend region structural equation model result ........................................50

2.4 Forgotten Reach structural equation model

using the Big Bend model construct ............................................................51

2.5 Final structural equation model structure for

the Forgotten Reach .....................................................................................52

3.1 Study area map for the Big Bend region sampling

sites .............................................................................................................104

3.2a Conceptual diagram of the mapped mesohabitats ......................................105

3.2b Relative proportion of mapped mesohabitats

across varying levels of habitat heterogeneity ...........................................105

3.3 Average depths and velocities calculated during

each sampling discharge period .................................................................106

3.4 Random forest variable importance plot ....................................................106

3.5a Average catch per unit effort of fish caught

udring sampling periods .............................................................................107

3.5b Shannon’s Diversity Index calculated for the

fish community across varying degrees of habitat heterogeneity ..............107

3.6 Average chlorophyll a abundance across varying

degrees of habitat heterogeneity .................................................................108

3.7 Average relative abundance of aquatic invertebrate

functional feeding groups ..........................................................................109

3.8 Calculated food chain lengths for fish occurring

in complex and simple sampling sites ........................................................110

xiv

3.9 Whole food web structure for varying levels

of habitat heterogeneity ..............................................................................111

3.10 Stable isotope ellipses calculated for the fish community

across varying levels of habitat heterogeneity ...........................................112

3.11 Three-source mixing model summary results

for fish as consumers ..................................................................................113

3.12 Two-source mixing model summary results using

the Hydrogen isotope .................................................................................114

3.13 Three-source mixing model summary results

for aquatic invertebrates as consumers .......................................................115

A1 Total area calculated for each the varying levels of

habitat heterogeneity ..................................................................................127

A2 Average catch per unit effort of fish for each sampling

month ..........................................................................................................128

A3 Fish species evenness calculated across the varying

levels of habitat heterogeneity ....................................................................129

A4 Chlorophyll a abundance calculated by season ..........................................130

CHAPTER 1

INTRODUCTION

Rivers are some of the most ecologically-dynamic and complex systems globally.

However, river ecosystem structure and function are increasingly threatened by the

anthropogenic demand for water (Hauer and Lorang 2004). Thus, the natural flow and

sediment regimes of many rivers have been dramatically altered (e.g., Johnson and others

1995; Poff and others 1997; Allan 2004; Lytle and Poff 2004; Nilsson and others. 2005,

Schmidt and Wilcock 2008, Wohl and others 2015). Water diversion and impoundment

affects the timing and delivery of natural stream flows, ultimately changing

characteristics of the flow regime including magnitude, frequency, and duration of

stream-flow events (Magilligan and Nislow 2005). Diminishing any aspect of a river’s

flood regime can have important geomorphic and ecological consequences.

The natural stream flow variability of a river has been established as the ‘master’

variable for maintaining river ecosystem health and biodiversity (e.g., Walker and others

1995; Poff and others 1997; Hart and Finelli 1999; Poff and others 2010). Large flood

events are important for channel development and maintain habitat complexity (Junk and

others 1989; Ward and Stanford 1989; Ligon and others 1995). Reductions and

simplifications to channel morphology, such as narrowing, aggradation, or incision can

limit instream habitat complexity and create losses in aquatic taxa abundance and

diversity (Ligon and others 1995; Bunn and Arthington 2002). In addition, flows that

maintain longitudinal and lateral connectivity with the active floodplain drive food web

structure and function (Wootton and others 1996; Power and others 1996). Floodplains

that are regularly (or seasonally) inundated increase the area available for aquatic

2

biological activity, and organic material stored in the active floodplain becomes available

for associated consumers. Also, many aquatic species depend on spring flood pulses for

spawning cues (Resh and others 1988; Brouder 2001; King and others 2003; Archdeacon

2016). Finally, the combination of stream flow and channel geometry determines water

level and resource availability, which effectively help or hinder invasive species and the

efficiency of energy dynamics within a river (Bunn and Arthington 2002, Stromberg and

others 2007). A river’s ecosystem health is therefore dependent on stream flows that can

sustain habitat complexity, life history traits, and the endemic taxa. None of this

literature, however, considers that the sediment mass balance is perturbed into sediment

surplus, as is the case in the Big Bend region.

Adequate stream flow and sediment supply are necessary for maintaining

freshwater biodiversity, because each are important drivers of ecosystem processes, drive

species communities and abundances, and maintain habitat heterogeneity and water

quality (e.g., Poff and others 1997; Hart and Finelli 1999; Poff and others 2010). Habitat

heterogeneity can provide adequate areas for growth, safety from predators, food

resources, and reproduction. Varying types of habitat often facilitate the coexistence of

multiple species, and aquatic diversity is often associated with complex habitat

heterogeneity (Nagayama and Nakamura 2018). In addition, water quality, specifically

salinization, is an increasing threat to freshwater ecosystems and water resources globally

(Williams 1999). The ecological effects often include losses in biodiversity, shifts in

community structure, or even trophic cascades. Finally, modified flow and sediment

supply can facilitate nonnative species establishment, potentially displacing native

species and homogenizing both aquatic and terrestrial riparian communities (Thomaz and

3

others 2015). Determining how the hydrological, geomorphic, and ecological variables

interact in large, exotic rivers (e.g., rivers whose flow regime is derived from the climate

of distant headwaters), as is the case for the exotic Rio Grande, is an important question

for effective environmental flow management and conservation efforts.

Linking the physical and ecological components of a river ecosystem has become

a critical aspect of understanding how the aquatic ecosystem interacts with its

hydrological and geomorphic environments (Vaughan and others 2009; Elosegi and

others 2010; Poole 2010). However, there is an added complexity when incorporating

both the physical and ecological environment. Often the questions of interest are better

explained by using multiple direct and indirect variables (Arhonditsis and others 2006).

For the first chapter of my thesis, we used structural equation modeling (SEM) as a

method to address the above ecological complexity, because SEM allows the user to test

both direct and indirect, or intermediary, effects on an overall response variable, or

variables. The overall goals of this study were to provide a synthesis of the state of

knowledge on the existing ecological condition between two hydrologically and

geomorphologically distinct parts of the Rio Grande, the Forgotten Reach and the Big

Bend region. Second, we used that body of knowledge to identify testable hypotheses of

the most important limiting factors for ecosystem integrity within two regions. We also

used SEMs to test those hypotheses, and we compared SEMs between the two river

segments, each with varying degrees of contemporary ecological integrity.

For the second chapter of my thesis, the overarching goal for this study was to use

a habitat availability and food web-based approach to examine how changes to the

physical environment alter the aquatic ecological structure and function of a desert river

4

ecosystem. Within this context, I had two objectives to better understand the altered

physical and ecological environment and associated effects on the native fish community

of the highly regulated Rio Grande along the US-Mexico border. My first objective was

to identify the effects of the altered hydrologic and sediment regimes of the Rio Grande

on habitat heterogeneity and the associated native fish community. Based on prior

geomorphic assessments (Dean and Schmidt and others 2011, 2013; J. Bennett, personal

communication), I chose a suite of sampling sites exhibiting varying levels of channel

morphology and simplification throughout the river. I predicted sampling sites with

greater habitat heterogeneity would be more complex and contain greater fish diversity

than sites with low habitat heterogeneity. My second objective was to determine if and

how these altered hydrologic and sediment regimes have potentially affected energy flow

pathways (e.g., basal resource production) and structure of the aquatic food web. I

predicted food web structure (e.g., food chain length, trophic niche space) would be more

complex within sites with greater habitat heterogeneity than sites with low habitat

heterogeneity because I assumed complex sites to better partition resource-use and

facilitate multiple species. I also predicted nonnative vegetation alters the food web

function by contributing differentially to dietary items of upper trophic-level consumers

than does native vegetation. Overall, I hope to contribute to the ecological knowledge

surrounding desert food web structure and function and provide insight to more effective

and innovative stream flow management solutions for the Rio Grande both in the

Forgotten Reach, and in the Big Bend region, arguably the most ecologically-intact reach

of this river that remains.

5

Literature Cited

Allan JD. 2004. Landscapes and riverscapes: the influence of land use on stream

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

Archdeacon TP. 2016. Reduction in spring flow threatens Rio Grande Silvery Minnow:

trends in abundance during river intermittency. Transactions of the American

Fisheries Society 145: 754-765.

Brouder MJ. 2001. Effects of flooding on recruitment of roundtail chub, Gila robusta, in

a southwestern river. The Southwester Naturalist 46: 302-310.

Bunn SE, Arthington AH. 2002. Basic principles and ecological consequences of altered

flow regimes for aquatic biodiversity. Environmental Management 30: 492-507.

Hart DD, Finelli CM. 1999. Physical-biological coupling in streams: the pervasive effects

of flow on benthic organisms. Annual Review of Ecology and Systematics 30:

363-395.

Hauer FR, Lorang MS. 2004. River regulation, decline of ecological resources, and

potential for restoration in a semi-arid lands river in the western USA. Aquatic

Sciences 66: 388-401.

Johnson BL, Richardson WB, Naimo TJ. 1995. Past, present, and future concepts in large

river ecology. BioScience 45: 134-141.

Junk WJ, Bayley PB, Sparks RE. 1989. The flood pulse concept in river-floodplain

systems. Canadian Special Publication of Fisheries and Aquatic Sciences 106:

110-127.

King AJ, Humphries P, Lake PS. 2003. Fish recruitment on floodplains: the roles of

patterns of flooding and life history characteristics. Canadian Journal of Fisheries

and Aquatic Sciences 60: 773-786.

Ligon FK, Dietrich WE, Trush WJ. 1995. Downstream ecological effects of dams.

Bioscience 45: 183-192.

Lytle, DA, Poff NL. 2004. Adaptation to natural flow regimes. Trends in Ecology and

Evolution 19: 94-100.

Magilligan FJ, Nislow KH. 2005. Changes in hydrologic regime by dams.

Geomorphology 71: 61-78.

Nilsson, C, Reidy CA, Dynesius M, Revenga C. 2005. Fragmentation and flow regulation

of the world’s large river systems. Science 308: 405-408.

6

Poff NL, Allan JD, Bain MB, Karr JR, Prestegaard KL, Richter BD, Sparks RE,

Stromberg JC. 1997. The natural flow regime: a paradigm for river conservation

and restoration. Bioscience 47: 769-784.

Poff NL, Zimmerman JKH. 2010. Ecological responses to altered flow regimes: a

literature review to inform the science and management of environmental flows.

Freshwater Biology 55: 194-205.

Power ME, Dietrich WE, Finlay JC. 1996. Dams and downstream aquatic biodiversity:

potential food web consequences of hydrologic and geomorphic change.

Environmental Management 20: 887-895.

Resh VH, Brown AV, Covich AP, Gurtz ME, Li HW, Minshall GW, Reice SR, Sheldon

AL, Wallace JB, Wissmar RC. 1988. The role of disturbance in stream ecology.

Journal of the North American Benthological Society 7: 433-455.

Schmidt JC, Wilcock PR. 2008. Metrics for assessing the downstream effects of dams.

Water Resources Research 44: 1-19.

Stromberg JC, Beauchamp VB, Dixon MD, Lite SJ, Paradzick C. 2007. Importance of

low-flow and high-flow characteristics to restoration of riparian vegetation along

rivers in arid south-western United States. Freshwater Biology 52: 651-679.

Ward JV, Stanford JA. 1983. The serial discontinuity concept of lotic ecosystems. Pages

29-42 in T.D. Fontaine and S.M. Bartell, editors. Dynamics of Lotic Ecosystems.

Ann Arbor Sciences, Ann Arbor, MI.

Wohl E, Bledsoe BP, Jacobson RB, Poff NL, Rathburn SL, Walters DM, Wilcox AC.

2015. The natural sediment regime in rivers: Broadening the foundation for

ecosystem management. BioScience 65(4): 358-71.

Wootton JT, Parker MS, Power ME. 1996. Effects of disturbance on river food webs.

Science 273: 1558-1561.

7

CHAPTER 2

REMEMBERING A ‘FORGOTTEN’ RIVER: IDENTIFYING LIMITING FACTORS

FOR THE NATIVE FISH DIVERSITY OF THE RIO GRANDE IN FAR WEST

TEXAS

Abstract

Desert river ecosystems are some of the most threatened aquatic systems because

river ecosystems compete with human societies for the most precious resource on earth,

water. Anthropogenic impacts such as stream-flow modification and depletion have

negatively impacted biodiversity and ecosystem integrity at a global scale. Our overall

approach was to identify the most critical limiting factors to the native aquatic diversity

and richness of the Rio Grande within two different segments along the US-Mexico

border: the degraded Forgotten Reach and the more ecologically-intact Big Bend region.

We synthesized available sampling data conducted within the Rio Grande from El Paso-

Ciudad Juarez to Big Bend National Park (sampling n ~100 events) that describe the fish

community, water quality, and the hydrologic and geomorphic characteristics for each

segment. We then used structural equation modeling (SEM) to explore the hydrologic,

geomorphic, and ecologic components of the Rio Grande in both segments. Both the

Forgotten Reach and the Big Bend region SEM results fit the observed data relatively

well, with both segments having high chi-squared values (χ2; p > 0.05 = 0.400 for the Big

Bend region, p > 0.05 = 0.066 for the Forgotten Reach), where non-significant p-values

indicate better fit. We found native fish richness to be more influenced by channel

narrowing and low flows in the Big Bend region than in the Forgotten Reach. In contrast,

in the Forgotten Reach, water quality and channel narrowing were the most influential

8

variables describing native fish richness. Each region appeared to be unaffected by the

nonnative fish abundance; however, there were higher abundances of generalist, saline-

tolerant native fish species compared to more sensitive species within the Forgotten

Reach. Thus, the community may be shifting more towards generalist native fishes

relative to specialist native fishes or nonnative fishes overall. Our comparative approach

demonstrates how two long river segments can be affected differentially by dewatering

and low flows, and result in different factors that are most limiting to ecological integrity.

In a contemporary world of extreme competition for water and associated physical

impairment to rivers, it is important to understand which factors are most limiting to

ecological health at a local scale, such that management decisions regarding flow can be

effectively prioritized.

Introduction

Reliable water supply is a vital and limited resource for life in arid regions.

Human demand for water has increased due to population growth and expansion of

irrigated agriculture, leading to construction of dams, increased surface-water diversions,

and increased groundwater pumping. The collective increase in river regulation and the

depletion of stream flow threatens aquatic riverine ecosystem structure and function

(Hauer and Lorang 2004; Richter and others 2011). Globally, there have been large

changes to the frequency, magnitude, and duration of flow and sediment inputs associated

with altered water quantity and timing (Poff and others 1997; Brandt 2000; Bunn and

Arthington 2002; Nilsson and others 2005; Schmidt and Wilcock 2008). River ecosystem

integrity is dependent on stream flow and sediment supply that can sustain habitat

complexity, complex life history traits, and diverse endemic taxa (Bunn and Arthington

9

2002; Stromberg and others 2007; Archdeacon 2016).

Freshwater ecosystems are some of the most imperiled in the world and more than

81 percent of freshwater populations have declined since 1970 (WWF 2016). Flow

modification, habitat degradation, water pollution, and species invasions are the primary

threats to freshwater biodiversity worldwide (Dudgeon and others 2006). Losses in

biodiversity can affect ecosystem functions such as carbon and nutrient cycling, and

primary productivity (Hooper and others 2005). For example, disturbances or losses of

benthic functional group diversity influences primary and secondary productivity by

depleting sugars or nutrients necessary for productivity processes such as photosynthesis

(Palmer and others 1997). Thus, upper trophic levels such as secondary and tertiary

consumers may lose critical food resources and subsequently decline in abundance and

diversity, shortening the food chain length and depleting the flow of energy through an

aquatic ecosystem.

Adequate stream flow and sediment supply are necessary for maintaining

freshwater biodiversity, because each are important drivers of ecosystem processes, drive

species communities and abundances, and maintain habitat heterogeneity and water

quality (e.g., Poff and others 1997; Hart and Finelli 1999; Poff and others 2010). Habitat

heterogeneity can provide adequate areas for growth, safety from predators, food

resources, and reproduction. Varying types of habitat often facilitate the coexistence of

multiple species, and aquatic diversity is often associated with complex habitat

heterogeneity (Nagayama and Nakamura 2018). In addition, water quality, specifically

salinization, is an increasing threat to freshwater ecosystems and water resources globally

(Williams 1999). The ecological effects often include losses in biodiversity, shifts in

10

community structure, or even trophic cascades. Finally, modified flow and sediment

supply can facilitate nonnative species establishment, potentially displacing native

species and homogenizing both aquatic and terrestrial riparian communities (Thomaz and

others 2015). Determining how the hydrological, geomorphic, and ecological variables

interact in large, exotic rivers (e.g., rivers whose flow regime is derived from the climate

of distant headwaters), as is the case for the exotic Rio Grande, is an important question

for effective environmental flow management and conservation efforts.

Desert river biodiversity is particularly at risk due to climate-induced drought and

increased regulation and diversion in these water-scarce regions. Receiving less than 500

mm of rainfall a year, desert rivers and their ecosystem communities and processes are

most often limited by water quantity (Kingston and others 2006). Water quantity is

coupled with physical processes such as sediment transport in arid regions. Losses of

stream flow facilitate channel change such as narrowing or incision, which can have

deleterious impacts to habitat heterogeneity and associated biodiversity. Combined with

low water quantity, high evaporation rates in arid systems can degrade water quality

conditions by causing increases in water temperatures, salinity, or nutrients and decreases

in dissolved oxygen, all of which can negatively impact aquatic fauna. Desert rivers pose

a unique challenge to effective management because these ecosystems are partially

dependent on stream flow that is also needed by human society. Therefore, a full

understanding of how the physical and ecological processes are impacted by flow

depletion contributes to innovative solutions for managing these unique and imperiled

ecosystems.

Linking the physical and ecological components of a river ecosystem has become

11

a critical aspect of understanding how the aquatic ecosystem interacts with its

hydrological and geomorphic environments (Vaughan and others 2009; Elosegi and

others 2010; Poole 2010). However, there is an added complexity when incorporating

both the physical and ecological environment. Often the questions of interest are better

explained by using multiple direct and indirect variables (Arhonditsis and others 2006).

We use structural equation modeling (SEM) in this study as a method to address the

above ecological complexity, because SEM allows the user to test both direct and

indirect, or intermediary, effects on an overall response variable, or variables. The overall

goals of this study were to provide a synthesis of the state of knowledge on the existing

ecological condition between two hydrologically and geomorphologically distinct parts

of the Rio Grande, the Forgotten Reach and the Big Bend region. Second, we used that

body of knowledge to identify testable hypotheses of the most important limiting factors

for ecosystem integrity within two regions. We also used SEMs to test those hypotheses,

and we compared SEMs between the two river segments, each with varying degrees of

contemporary ecological integrity.

Contemporary Ecosystem of the Rio Grande

Hydrology

The Rio Grande is the fifth longest river in North America and has had a long

history of human use and modification beginning at the time of Spanish settlement

(Scurlock 1999; Fullerton and Batts 2003). To meet the needs of irrigated agriculture,

municipal, and industrial use, much of the water contributed to the Rio Grande by

snowmelt and monsoon floods is diverted or stored in many reservoirs (Collier and others

1996; Schmidt and others 2003; Porter and others 2009; Sandoval-Solis and others 2010).

12

The contemporary Rio Grande functions hydrologically as two separate rivers, each with

differing headwaters: the northern branch (headwaters to Rio Conchos) and the lower Rio

Grande (Rio Conchos to Gulf of Mexico). The hydrology of the northern branch

headwaters is driven by annual spring snowmelt flows primarily from the San Juan and

Sangre de Cristo mountains in southern Colorado, which peak in early April to June

(Schmidt and others 2003; Blythe and Schmidt 2018). In contrast, the lower Rio Grande

(e.g., the Big Bend region, Rio Conchos to Gulf of Mexico) is predominately sustained

by flows from the Rio Conchos, a large tributary originating in the Sierra Madre

Occidental in northern Mexico, summer thunderstorms over desert tributaries, and

groundwater spring input (Fullerton and Batts 2003; Raines and others 2012). Overall,

the river has declined in total annual volume by as much as 95% (Blythe and Schmidt

2018) and has lost much of its hydrologic variability.

The Forgotten Reach is one of the most impacted segments due to the nearly

complete diversion of upstream flows. This segment is between Fort Quitman, Texas and

Presidio, Texas – Ojinaga, Chihuahua and is approximately 320 river kilometers, its

hydrology is that of the northern branch of the Rio Grande (Figure 1; Collier and others

1996). Prior to extensive irrigation and upstream dam construction, the Forgotten Reach

would have had a snowmelt flood that occurred between April and June, dominating the

annual hydrograph (Blythe and Schmidt 2018). Summer monsoon flash floods also

occurred between July and October, but contributed little runoff compared to the spring

snowmelt. Since closure of Elephant Butte Dam in 1915, snowmelt flows have been

eliminated. All stream flow inputs come from irrigation water return and the remaining

summer monsoon flash floods (Johnson and others 1977; Collier and others 1996; Landis

13

2001; USACE 2008; Schmidt and others 2003). The annual flow regime of the

contemporary Forgotten Reach is largely stable, experiencing relatively minor changes in

flow magnitudes annually. Loss of variability in the flow regime has had compounding

effects on the geomorphology, water quality, and native aquatic communities.

Downstream from the Forgotten Reach, total annual stream flow in the Big Bend

region significantly increases due to inflow from the Rio Conchos (Kelly 2001). Floods

in the Rio Conchos occur during the North American monsoon season and are the

primary contribution to the water quantity of the Rio Grande as it flows through the

Chihuahuan Desert and Big Bend region for approximately 480 km (to Amistad

Reservoir near Del Rio, Texas-Ciudad Acuña, Coahuila; Dahm and others 2005;

Woodhouse and others 2012). Like the northern branch in New Mexico, the Rio Conchos

basin is highly regulated by multiple dams for hydroelectricity, and the river is

extensively diverted for irrigation and municipal uses, and extended periods of low flows

occur (Dean and others 2011; Dean and Schmidt 2011; Dean and others 2016). A

distinctive attribute of the flow regime in the Big Bend region is the occurrence of ‘reset’

floods (Dean and Schmidt 2013). These large (>1000 m3/s), relatively rare flood events

(recurrence of ~10 – 15 years) are ecologically important, because they restore channel

complexity by evacuating fine sediment that narrows the channel during the periods

between the reset floods; abundant exotic vegetation also invades the surface of the fine

sediment deposits that accumulate, and reset floods remove some of the exotic vegetation

(discussed further below; Dean and Schmidt 2011, 2013; Dean and others 2011; Dean

and others 2016).

14

Geomorphology

The Forgotten Reach is significantly perturbed into sediment surplus and has

undergone a series of successive geomorphic changes since the Elephant Butte Dam

closure in 1915 (Everitt 1993). The Forgotten Reach was once a wide, braided, and

shallow river with extensive cottonwood bosques and willow galleries, but has since

narrowed greatly (Everitt 1977, 1993, 1998). Initially, natural tributary flows entering the

study reach deposited substantial amounts of sediment within the channel and on the

floodplain (Everitt 1977, 1993; Collier and others 1996), and the channel shrank and

formed inset floodplains, or sediment berms (Everitt 1993). Shrinkage caused the river to

temporarily disconnect from its former floodplain, and lateral migration of the channel

ceased. As sediment input and low flows continued, the channel began to aggrade high

enough that stream flows once again flowed overbank. The overbank flows cut new

channels to detour around the sediment-laden tributary junctions, causing the river to

migrate away from its tributaries (Everitt 1993). The channel continues to adjust today

with increasing anthropogenic water demand and hydroclimatic variability, but it remains

narrowed and filled with sediment. Restoration of this segment to its historic geomorphic

state would require restoration of the large magnitude duration snowmelt floods last

observed in the 1800s.

Like the Forgotten Reach, reduction in the total annual flow in the Big Bend

region has contributed to channel narrowing and allowed subsequent nonnative

vegetation establishment (Dean and Schmidt 2011). The channel width is estimated to be

50% narrower today than it was more than a century ago, and nonnative saltcedar

(Tamarix spp.) and giant cane (Arundo donax) act as a positive feedback to accelerate

15

channel narrowing between reset floods (Dean and Schmidt 2011, 2013; Dean and others

2011; Dean and others 2016). Historically, the Rio Grande was once a wide, sandy, and

dynamic channel with a low-elevation floodplain in the wide alluvial valleys. The

intervening canyons were historically more confined and may not have had an extensive

floodplain. The present Rio Grande in the Big Bend region has narrowed considerably

and formed high banks in both the alluvial valleys and the canyon-bound reaches.

Subsequently, saltcedar and giant cane have established on parts of the floodplain and

contribute to bank stability and sediment accumulation. In contrast to the Forgotten

Reach, the narrow channel is occasionally rewidened by large, rare flood events (Dean

and Schmidt 2011). These floods are high in magnitude (>1000 m3/s) and duration (>7

days) and erode substantial portions of the inset floodplains that form during low flow

periods (Dean and Schmidt 2011, 2013; Dean and others 2011).

Associated Changes to the Ecosystem Structure and Function

Ecosystem structure is composed of the biota and their interactions with their

physical environment, and ecosystem function is defined as is the physical and chemical

processes influencing structure. For example, the complexity of a food web can drive the

amount of energy transferred/lost within a system (Allan and Castillo, 2007). Similarly, a

floodplain ‘structure’ will influence nutrient deposition and cycling throughout the length

of a river, in turn stimulating aquatic primary and secondary productivity. In fluvial

ecosystems, these feedbacks between structure and function are important for

maintaining ecosystem health. River ecosystem processes are largely driven by the

quantity and timing of stream flow, water quality, and the biological community present.

Alterations to these aspects of flow can affect either structure or function by degrading

16

water quality, changing critical hydrologic cues, reshaping river channel form, or altering

riparian community composition.

The Forgotten Reach is critically altered in terms of its ecological integrity.

Continuous low flows and sediment surplus in the region have contributed to narrowing

the channel, thereby decreasing in-channel and floodplain habitat complexity (USACE

2008). Abundance and diversity of habitat often has a strong correlation with abundance

and diversity of native species (Ricklefs and Schluter 1993). Monitoring surveys on the

lower Rio Grande in west Texas have demonstrated decreasing trends in native species

richness associated with habitat availability (Blythe Chapter 2; Bestgen and Platania

1988; Platania 1991; Edwards and Contreras-Balderas 1991; Heard and others 2012).

Species adapted to particular habitat types have a higher risk of extirpation and are more

sensitive to degraded flow conditions (Laub and Budy 2015). For example, loss of

floodplain habitat associated with high flow events has contributed to the decline in

native fish species, such as the federally endangered Rio Grande silvery minnow

(Hybognathus amarus). These minnows rely on spring flood pulses to cue spawning and

create nutrient-rich floodplain ‘nurseries’ for the larval fish to grow and survive

(Archdeacon 2016), and their endangered-status is directly related to flow and available

habitat for these life strategies.

In addition to decreased habitat heterogeneity, water quality has been a growing

concern in the Forgotten Reach (IBWC 1994, 2008; Miyazono and others 2015). Total

dissolved solids increased substantially after the completion of Elephant Butte Dam, and

increased agricultural diversions in the early 1900s; water quality has continued to

deteriorate in response to recent increased population growth in northern Mexico after

17

signing of the North American Free Trade Agreement. Concentrations of total dissolved

solids (mg/L) are up to five times higher within the Forgotten Reach compared to other

parts of the river (Moyer and others 2013). Consequently, the fish assemblage for the

Forgotten Reach is made up of generalists more tolerant of poor water quality conditions

(Hubbs and others 1977; Bestgen and Platania 1988; USACE 2008; Miyazono and others

2015), reducing food web complexity and subsequent energy transfer for the system by

creating a more homogenous fish community structure. Mollusk and crustacean richness

is also relatively low throughout the region (Metcalf 1978) and is composed primarily of

opportunistic or invasive species, such as the exotic Asian clam (Corbicula fluminea).

Asian clams have high filtration rates and produce substantial amounts of phosphorus,

potentially stimulating primary productivity and increasing the occurrences of

eutrophication in the study region (Sousa and others 2008).

Finally, the riparian community has shifted from native species (e.g., cottonwood

(Populus spp.), willow (Salix spp.), and mesquite (Prosopis spp.)), to almost exclusively

nonnative saltcedar in the Forgotten Reach (Everitt 1998). Compared with native riparian

species, saltcedar decreases the amount of light reaching a narrow river, deposits

substantially more leaf litter and detritus, and creates a system dependent on seasonal

allochthonous sources of energy (Kennedy and others 2004). In addition, two riparian

bird species have become endangered by the loss of native willows, which provide

important nesting habitat for the Southwestern willow flycatcher (Empidonax traillii

extimus) and Western yellow-billed cuckoo (Coccyzus americanus occidentalis; US Fish

and Wildlife 2013). The riparian composition change has likely impacted the Forgotten

Reach ecosystem by altering the sources of energy (e.g., shifting from autochthonous to

18

allochthonous) and altering carbon cycling by changing the types of carbon available for

uptake by secondary consumers. Thus, the removal of nonnative vegetation may increase

ecosystem production and create more available food and habitat resources for threatened

or endangered aquatic native species (Kennedy and others 2005).

Like the Forgotten Reach, the long-term trend of sediment accumulation and

channel narrowing in the Big Bend region has decreased in-channel habitat complexity.

Even after a reset flood, habitat features are only partially recovered, and narrowing and

associated habitat simplification generally resumes relatively quickly (Dean and Schmidt

2011). Because any given species has its own unique habitat requirements, habitat

simplification restricts the amount and types available for a species, often leading to

increased predation and competition (Laub and Budy 2015). Many species of aquatic

organisms in the Big Bend region (e.g., fish, macroinvertebrates, mussels) have become

threatened, endangered, or extirpated due to decreased habitat heterogeneity (Davis 1980;

Heard and others 2012).

Prolonged low flow periods have contributed to the establishment of nonnative

giant cane in the Big Bend region, with multiple ecological consequences. Giant cane

actively displaces native riparian vegetation and uses considerable amounts of water

compared to native species (Yang and others 2011). Giant cane also acts as a positive

feedback to channel narrowing by stabilizing channel banks and trapping sediment,

ultimately contributing to habitat simplification (Dean and Schmidt 2011, 2013).

Terrestrial arthropods have less diverse assemblages in giant cane compared to native

riparian species like willows and cottonwoods (Herrera and Dudley 2003), potentially

shifting the types and abundances of terrestrial arthropod input into the river. Similarly,

19

giant cane produces up to 20 t/ha of leaf litter, likely altering the river’s energy sources

by becoming more reliant on allochthonous inputs (Yang and others 2011) and

influencing the water chemistry through increasing the abundance of organic materials

within the river.

Salinity, nutrients, and organic matter concentrations decrease downstream

because base flows increase downstream from Big Bend National Park where there is

significant groundwater input (URGBBEST 2012). This downstream trend of improving

water quality is reflected within the algal communities as they shift from hypereutrophic-

and eutrophic-affiliated species to mesotrophic and nitrogen-fixing species (Porter and

Longley 2011). Similarly, there have been increased abundances of threatened or

endangered mussel species, such as Texas hornshell (Popenaias popeii), occurring near

higher-quality groundwater in the Lower Canyons part of the Big Bend (Winemiller and

others 2010; Porter and Longley 2011; Burlakova and others 2011a,b). Mussels also have

important roles in river nutrient cycling, often forming the link between benthic and

pelagic energy transfer because they excrete nutrients into the water column and deposit

organic material to river sediments (Vaughn 2008). In addition, diversity and abundance

of the native fish community is tied with the quantity and quality of water, increasing as

base flows increase (Hubbs and others 1977; Bestgen and Platania 1988; Edwards and

Contreras-Balderas 1991; Miyazono and others 2015).

Our goal was to build an SEM for both the Forgotten Reach and Big Bend region

to explore and compare the relationships among the hydrologic and geomorphic

properties, water quality, and nonnative fish species on the native fish richness for each

region. Based on the observed changes to the ecosystem structure and function for the

20

Forgotten Reach and Big Bend region, we hypothesized poor water quality conditions

and degraded habitat heterogeneity have contributed to degraded ecosystem structure and

function for the Forgotten Reach, specifically native fish richness. We also hypothesized

reset floods, elevated base flows, and groundwater spring input likely contribute to higher

ecological integrity in the Big Bend region than within the Forgotten Reach. By

comparing each region, our goal is to describe the mechanisms linking stream flows and

the physical environment with the aquatic ecosystem structure and function. Identifying

these potential mechanisms will contribute to our knowledge of which components of the

flow and sediment regimes are the most important for ecosystem rehabilitation for the

Forgotten Reach relative to the Big Bend region. Our comparative approach allows us to

identify how the two very different regions operate similarly or differently.

Methods

Study Area

The Rio Grande is approximately 550,000 km2 and flows through eight states –

three in the United States and five in Mexico. In addition, the Rio Grande forms the U.S.-

Mexico international border for approximately 1900 km. The Forgotten Reach and Big

Bend region are located within the Basin and Range physiographic province and

Chihuahuan Desert ecosystem (Figure 2.1; USACE 2008). Within each region, the Rio

Grande alternates flowing between wide, alluvial valleys and steep, narrow canyons

(USACE 2008; Dean and Schmidt 2011, 2013). The climate for each region is arid and

dry. Sources of precipitation for the region are primarily from heavy local thunderstorms

associated with the North American monsoon. Total annual precipitation ranges from 10

to 82 cm, with most of the annual precipitation occurring between the months of July and

21

September (Johnson and others 1977; USACE 2008). The average annual temperatures in

the region vary between 5 and 27 C, with the hottest temperatures recorded in the

months of June and July, and the lowest in January (USACE 2008; PRISM Climate

Group 2017).

Data Collection

Fish community and water quality

We compiled fish data for the Forgotten Reach using data collected by Hubbs and

others (1977), Bestgen and Platania (1988), and Miyazono and others (2015) that

extended from El Paso-Ciudad Juarez to Big Bend National Park. For the Big Bend

region, Laub and Budy (2016) compiled fish data from Heard and others (2012), Garrett

and Edwards (2014), Moring and others (2014), and Ken Saunders of the Texas Parks

and Wildlife Department. The fish data included date collected, recorded sampling

locations (given in decimal degrees), total number of fish caught for each species

collected while sampling, and water quality for each sampling location (including water

temperature (°C), dissolved oxygen (mg L-1), specific conductance (µg cm-1), pH). If

water quality was not recorded, we used both the Environmental Protection Agency

(EPA) Water Quality portal tool (https://www.epa.gov/waterdata/water-quality-data-wqx)

and Texas Commission on Environmental Quality database

(https://www.tceq.texas.gov/agency/data) to extract point measurements taken near the

fish sampling locations.

22

Hydrologic and geomorphic data

For our hydrologic parameters, we collected historic mean daily discharge data

from gaging stations operated either by the International Boundary and Water

Commission (IBWC) or U.S. Geological Survey (USGS). Depending on the location of

the sampling site in the Forgotten Reach, we used the following IBWC data from gaging

stations: Rio Grande at Fort Quitman, Texas near Colonia Luis Leon, Chihuahua (08-

3705.00), Rio Grande near Candelaria, Texas and San Antonio Del Bravo, Chihuahua

(08-3712.00), and Rio Grande above Rio Conchos near Presidio, Texas and Ojinaga,

Chihuahua (08-3715.00). For the Big Bend region, Laub and Budy (2016) compiled

instantaneous flow data using both the IBWC Johnson Ranch gaging station (08-3750.00)

and USGS Castolon (08374550) and Rio Grande Village gaging stations (08375300).

For geomorphic data, we remotely measured active channel width (in meters) at

each sampling location using a combination of Google Earth historical aerial imagery and

National Agriculture Imagery Program (NAIP) 1-m aerial photography downloaded from

the Texas Natural Resources Information System (TNRIS). In addition, we used Everitt

(1993) active channel width measurements in the Forgotten Reach to extrapolate

potential channel widths during the Hubbs and others (1977) sampling events. Laub and

Budy (2016) used published average channel width data from Dean and Schmidt (2013)

and Dean and others (2016).

Structural Equation Modeling

Structural equation modeling (SEM) is a useful tool for describing the likely

relationships between the ecology and physical aspects of a river because of its ability to

test multiple direct and indirect effects, and the interactions between each, on a response

23

variable (Grace 2008; Grace and others 2010; Hermoso and others 2011). SEM is a

multivariate statistical technique that uses path analysis and regression to test an a priori

defined pattern of relationships among explanatory and response variables (Arhonditsis

and others 2006). Based on existing knowledge about an ecosystem, users define and

create an initial framework or hypothesis and test their hypothesis against a covariance

matrix of the observed data within an SEM (Sutton-Grier and others 2010). Thus, the

user-defined model can be confirmed or rejected if the data do or do not fit the

hypothetical framework. For example, an SEM was used to understand phytoplankton

dynamics between two different lakes with varying levels of eutrophication - one lake

was eutrophic and the other mesotrophic (Arhonditsis and others 2006). By using two

hypothetical SEMs, phytoplankton dynamics were shown to be dependent on several

ecological processes naturally occurring in these lakes, with more variability explained in

the eutrophic lake.

Determination of model fit

Multiple indices can be used to assess whether the actual data support the

hypothesized model. A chi-square (χ2) test is most often used, and a non-significant p-

value at the α = 0.05 significance level suggests the proposed model structure and the

actual data do not differ significantly (Beaujean 2014). In addition to the chi-square test,

we assessed model fit using a Comparative Fit Index (CFI), Tucker-Lewis Index (TLI),

and Root Mean Square Error of Approximation (RMSEA). For the CFI and TLI, values

closer to 1.0 tend to indicate good fit. For RMSEA, values that are significant (p-value >

0.05) and closer to 0.0 indicate better fit (Beaujean 2014).

24

SEM Conceptual Model

Defining direct hydrologic and geomorphic variables

Given the knowledge of the hydrologic and geomorphic properties of the Rio

Grande for both the Forgotten Reach and Big Bend region, we chose variables to capture

the effects of stream flow variability on fish habitat and communities. Thus, our

hydrologic variables included mean daily discharge on the day of fish sampling, the

median and minimum mean daily discharge during the previous 30 days, and the number

of flow spikes (defined as a flow rise 1.5 times the previous day’s flow based on mean

daily discharge) during the previous 90 days. Base flows were represented by median

flow, extreme low flow or drying events were captured within minimum flows, and

tributary-derived flash flood events were associated with flow spikes. In addition, we

chose active channel width (m) to represent the geomorphic and physical habitat

conditions for each region. While we understand there are other physical/geomorphic

variables that also affect habitat conditions (e.g., depth, velocity, area of in-channel

habitat features, sediment loads), active channel width was the most consistent variable

reported and we assumed it was an adequate measure of habitat complexity because

wider channels tend to support multiple, varying types of in-channel habitat.

Defining water quality as a latent variable

SEMs often include the use of latent variables, or variables not directly measured

or observed. Latent variables are thus defined using indicator variables, or variables

directly measured. There are two types of latent variables: reflective and formative. For

reflective latent variables, the latent variables cause the indicator variables (e.g.,

intelligence level (latent variable, not directly measured) predicts how well one does on a

25

test (test score, indicator variable)). In contrast, formative latent variables are caused by

the indicator variables (e.g., socioeconomic status (latent variable) is determined by

income or occupation (indicator variables; Beaujean 2012). For our model, salinity and

sulfates are some of the greatest threats to primary productivity and fish health for the

Rio Grande (Porter and Longley 2011). Therefore, we defined water quality as a

reflective latent variable, because we assumed water quality causes the chosen indicator

variables to covary. Our indicator variables for water quality were specific conductance

(µS cm-1), sulfate (mg L-1), and water temperature (°C). We chose specific conductance

as a measurement of salinity, because it was the most consistent parameter measured, and

we considered water temperature an important variable to consider because it controls

multiple physiological functions for aquatic species.

Nonnative fish abundance

Finally, invasive species tend to be higher in abundance and can influence and

restructure aquatic communities. Therefore, we included the summed nonnative fish

relative abundance in the model, because there were at least five different nonnative

species present in the fish sampling data for each study region, including: common carp

(Cyprinus carpio), inland silverside (Menidia beryllina), plains killifish (Fundulus

zebrinus), white bass (Morone chrysops), and blue tilapia (Oreochromis aurea).

Data Preparation

Before we tested the conceptualized SEM model for the Forgotten Reach, we

scaled all variables using z-scores, because some parameters had much higher values than

others (e.g., measurements of specific conductance, channel width, and sulfates were at

26

times almost 1000 times greater than native fish richness estimates). We completed all

data exploration using R 3.4.3 (R Development Core Team, 2017), and modeled all

SEMs using the lavaan package in R (Yves Rosseel 2012).

Results

Comparison of Big Bend Region and Forgotten Reach SEM Models

Because the Big Bend region and the Forgotten Reach models were developed

from independent datasets, we began by testing the Big Bend region against our

hypothesized conceptual model and considered the Big Bend SEM model the baseline

model (Figure 2.2). We then compared the Forgotten Reach to the baseline Big Bend

model using the original hypothesized model structure. Following the baseline

comparison, we adjusted our Forgotten Reach model accordingly, using a combination of

correlation and parameter estimates from the baseline model structure. Finally, we

compared the final, better-fit Forgotten Reach model with the baseline Big Bend model.

The Big Bend data supported the hypothesized model, and the SEM fit measures

indicated a good fit (χ2 p-value = 0.400, CFI= 0.994, TLI = 0.985, RMSEA = 0.03; Table

2.1). The hypothesized model included the direct effects of active channel width, all the

hydrologic variables (e.g., flow median and minimum during the previous 30 days, and

spikes during the previous 90 days), the water quality latent variable (defined using

specific conductance and sulfate), and nonnative fish abundance on native fish richness

(Figure 2.3). In addition, the hypothesized model included the indirect effects of the flow

variables on the water quality latent variable and nonnative fish abundance. While the

Big Bend model performed relatively well overall, the native fish richness variance was

only partially explained by the model structure (R2 = 0.269).

27

Next, the Forgotten Reach dataset was compared using the baseline Big Bend

SEM structure (hypothesized model, Figure 2.4). The Forgotten Reach data did not fit the

baseline structure well, and the overall fit measures were relatively low (χ2 p-value =

0.006, CFI = 0.905, TLI = 0.752, RMSEA = 0.178; Table 2.1). However, native fish

richness was explained by the model relatively well (R2 = 0.664). Because model fit was

relatively low for the Forgotten Reach using the baseline structure, we altered the

Forgotten Reach structure based on observing the significant parameter estimates within

the Forgotten Reach baseline model results. The final alternate model structure included

the direct effects of flow spikes, active channel width, nonnative fish abundance, and the

water quality latent variable on native fish richness (Figure 2.5). we also included the

indirect effects of flow spikes on active channel width and water quality, and water

quality on nonnative fish abundance. In addition to specific conductance and sulfate, we

added water temperature to the water quality latent variable. The data fit the altered

Forgotten Reach model structure relatively well (χ2 p-value = 0.066, CFI = 0.946, TLI =

0.915, RMSEA = 0.104; Table 2.1). Native fish richness continued to be explained well

even with the altered model structure (R2 = 0.630).

Examining SEM Results Using Standardized Coefficients

Big Bend SEM model

We present the standardized coefficients (unitless parameter estimates) to further

understand the similarities and differences between each of the above presented models.

By comparing the standardized coefficients, we are essentially comparing the strengths of

each pathway on overall native fish richness (Grace 2006). Thus, within the Big Bend

SEM structure, the most influential, or statistically significant, paths for native fish

28

richness were the direct effects of median and minimum flow during the past 30 days and

active channel width (Table 2.2). Native fish richness tended to increase with increased

active channel width and median flows and decrease with increases in minimum flow

events. Indirect effects of the hydrologic variables on water quality and nonnative

abundance were relatively negligible.

Forgotten Reach SEM results

For the Forgotten Reach under the baseline structure (hypothesized), native fish

richness was most strongly influenced by the direct effects of active channel width, water

quality, flow on the day of sampling, and flow spikes during the previous 90 days (Table

2.2). In contrast to the Big Bend model, the indirect effect of the median flows during the

previous 30 days had a statistically significant influence on water quality. For the altered

Forgotten Reach model, the direct effects were similar to the baseline model results

where water quality, active channel width, flow on the day of sampling, and flow spikes

during the previous 90 days had the most influence on overall native fish richness (Table

2.2). For each Forgotten Reach model structure, native fish richness tended to increase

with increases in flow spikes and active channel width and decrease with decreases in

water quality and flow on day of sampling. Overall, both the Big Bend and Forgotten

Reach model had significant direct paths between active channel width and native fish

richness. However, each model was influenced by different hydrologic variables (e.g.,

median and minimum mean daily discharge for the Big Bend region, median mean daily

discharge and flow spikes for the Forgotten Reach), and the Forgotten Reach was most

affected by water quality.

29

Discussion

The goal of this study was to understand and identify the linkages between the

physical and ecological components of a regulated riverine ecosystem and to identify the

most important limiting factors to native fish richness for the Rio Grande in the

Chihuahuan Desert. Using a novel modeling approach to accomplish this goal, we were

able to provide a better understanding of how ecosystem structure and function is

affected by both the modified hydrologic and sediment regimes of an extensively diverted

arid river. Comparing two regions similar in physical structure (e.g., geology, sediment

input, gradient and elevation) but differing in the magnitude of change of the natural flow

regime, allowed us to identify the most critical limiting factors for ecosystem integrity

and biodiversity in the Rio Grande in these two regions. In addition, comparing the Big

Bend region, a segment of greater ecological integrity, with the degraded Forgotten

Reach allowed us to explore the ways in which each region functioned similarly or

differently, and affected the native fish richness for the Rio Grande within each segment.

Mechanisms Influencing Ecosystem Structure and Function

Water quality and habitat heterogeneity appear to be the most critical limiting

factors for native fish richness in the Forgotten Reach. Across the entire Rio Grande,

salinity has been increasing since the 1950s; however, the highest concentrations occur

between Fort Quitman and Presidio-Ojinaga (2000 to 5000 mg L-1; Miyamoto and others

1995). An estimated 1.84 million MT of salt flow is input annually into the Forgotten

Reach and salt concentration is exacerbated by high evaporation rates in the region

(Miyamoto and others 1995). Trace elements in the Forgotten Reach often exceed the

U.S. Fish and Wildlife Service standards for aquatic species health. Fish samples near

30

Presidio contained elevated levels of mercury, lead, and copper in their tissues which

were attributed to mining sediments input from surrounding tributaries (Miyamoto and

others 1995; Gray and others 2006; Miyazono and others 2015). Fish and mussel species

are particularly sensitive to changes in water quality, and populations are often affected

by limited growth and abundance associated with poor water quality (e.g., Lorenz and

others 2016; Morcillo and others 2016; Hintz and Relyea 2017; Hintz and others 2017).

In addition, the channel in the Forgotten Reach has shrunk dramatically, with an almost

90% decrease in width (approximately 300 m wide pre-Elephant Butte Dam in 1902 to

30-50 m wide by 1970; Everitt 1993; Dean and Schmidt 2011). Habitat heterogeneity has

likely decreased as a result, having gone from a broad, shallow channel with multiple

oxbows to a narrow, aggraded and sediment-filled river (Everitt 1993).

In contrast, the Big Bend region water quality improves with increased stream

flows supplied by the Rio Conchos and groundwater aquifer complexes within the region

(Hubbs and others 1977; Miyamoto and others 1995; Porter and Longley 2011; Raines

and others 2012; Miyazono and others 2015). Thus, while still a key factor influencing

ecosystem integrity in the Big Bend region, water quality may not be a critical limiting

factor. Rather, sources of ecosystem degradation associated with low flows appear to be

more strongly impacted by losses in available in-channel habitat due to channel

narrowing and extreme low flow events in the region (Laub and Budy 2016).

Historically, the Rio Grande in the Big Bend region likely maintained multiple

types of shallow, slow-velocity habitats including floodplain pools, backwaters, and

embayments (Mueller 1975; Stotz 2000; Dean and Schmidt 2011). Presently, the river

becomes more complex immediately following a reset flood but narrows within a decade

31

after the reset event and has a small width-to-depth ratio (Everitt 1993; Dean and Schmidt

2011). Thus, there are many sections of the Big Bend region consisting almost entirely of

large, high-velocity runs and/or deep pools, and are considered ‘simple’ or ‘low’ in terms

of habitat heterogeneity (Moring and others 2014; Blythe, Chapter 2). In addition, the

alluvial valleys are narrowing at a faster rate compared to canyon-bound regions

following a recent 2008 reset flood (Dean et al. 2016; T. Blythe, in progress). The Big

Bend region is just one example of a desert river that has become simplified. For

example, the Green River in Utah, the longest tributary to the Colorado, experienced

decreases in average channel width by nearly 20% within the twentieth century (Allred

and Schmidt 1999). China’s desert Yellow River has also experienced channel change

and aggradation associated with regulated stream flow (Ta and others 2008), illustrating

that patterns of channel narrowing and habitat simplification are ubiquitous in desert

rivers.

Identifying Critical Limiting Factors for Each Study Region

As we hypothesized, our model results suggest low habitat heterogeneity may

have important consequences for the native fish richness currently present in the Big

Bend region. Channel width and low flows had the strongest influence on native fish

richness for the Big Bend region, likely because the channel is still dynamic and rewidens

after a reset flood event. The influence of channel width and low flows are unsurprising,

given that wider channels tend to support multiple types of habitat, and low flows, or

drying events, decrease the availability of habitat. For example, saltcedar removal on the

San Rafael River, Utah, a smaller but similar desert river, caused the channel to become

wider after a large flood event and increased the abundance of complex habitat areas

32

(e.g., riffles, backwaters, pools; Keller and others 2014). Subsequently, state-sensitive

(e.g., flannelmouth sucker (Catostomus latipinnis), bluehead sucker (Catostomus

discobolus)), and roundtail chub (Gila robusta)) native fish species were higher in

abundance associated with increased abundance of their preferred habitats within this

upper reach. In addition, the relatively small impact of water quality on the native fish

richness was expected with the increased groundwater input in the downstream direction

through the Big Bend region (Raines and others 2012).

Nonnative fish species often deleteriously affect the native fish communities in

many aquatic ecosystems. However, our results indicate the nonnative fish abundance did

not significantly affect the native fish richness. These non-significant results are contrary

to what we predicted, given nonnative fish species generally negatively impact the native

fish community by either predation or outcompeting for food and habitat resources.

Similarly, in the lower San Rafael River, native fish species were affected by a

synergistic combination of degraded habitat and inter- and intra-specific competition and

had to compete for habitat and food resources among native and non-native species (e.g.,

common carp (Cyprinus carpio), red shiner (Cyprinella lutrensis), and fathead minnow

(Pimephales promelas); Walsworth and others 2013). Overall our SEM for the Big Bend

region suggests a wider channel facilitates the maintenance of native fish diversity, and it

appears the current habitat simplifications have likely contributed to the decline and local

extirpations of native species such as the endangered Rio Grande silvery minnow.

Consistent with our observations in the available data, our model results for the

Forgotten Reach suggested water quality and channel width as the primary influences on

native fish richness. In addition, our results are consistent with previous monitoring

33

surveys in the Forgotten Reach (e.g., Hubbs and others 1977; Bestgen and Platania 1988;

Miyazono and others 2015), where higher abundances of generalist, saline-tolerant fish

species (e.g., red shiner, common carp, gizzard shad (Dorosoma cepedianum), bullhead

minnow (Pimephales vigilax)) occurred more frequently at sampling sites than other,

more sensitive species (e.g., Mexican tetra (Astyanax mexicanus), Chihuahua shiner

(Notropis chihuahua), Rio Grande shiner (Notropis jemezanus)). Low flows in the

Forgotten Reach, combined with increased maquiladores (manufacturing plants) in the El

Paso valley and irrigation return, are the primary sources of high salinization in this

region (Miyamoto et al. 1995).

Poor water quality has important ecological effects which can be felt at multiple

levels of biological organization. For example, fish predator response in a high salinity

environment via mesocosm experiments significantly reduced the zooplankton richness

and biomass, which induced a phytoplankton bloom due to the absence of zooplankton

grazing pressure (Hintz and others 2017). Similarly, salinity can influence fish growth by

increasing osmoregulation, subsequently increasing energy costs and limiting growth of

an individual fish (Bœuf and Payan 2001). Limited growth impacts reproduction

capabilities, occurrences of predation, competitive ability, and associated recruitment into

a population (Fogarty and others 1991; Cargnelli and Gross 1996). In addition, flow

spikes during the previous 90 days had a more direct impact on native fish richness,

which suggests any increases in quantity of stream flows may benefit native fish species.

Similarly, flow spikes indirectly affected native fish richness through negatively

influencing water quality and positively influencing channel width, suggesting flow

spikes may increase water quality (e.g., flow spikes may decrease the concentrations of

34

specific conductance and sulfates) and channel width.

Comparing the Forgotten Reach and Big Bend Regions

When we compared model results for the Forgotten Reach to the Big Bend region,

the native fish richness for each region appeared to be influenced by active channel

width. Compared with the Forgotten Reach, the Big Bend region experiences a history of

reset and rewidening of channel width over a decade and Forgotten Reach does not. As

wider channels are associated with increased habitat heterogeneity, native fish richness

may be subsequently increased in areas containing greater channel width due to greater

potential for multiple habitat types. In contrast, each region is affected directly by

different aspects of the flow regime. For the Big Bend region, minimum flows during the

previous 30 days (or low flow/drying events) had a more significant negative impact on

the native fish richness, where native fish richness tended to decrease with increases in

low flow/drying events. The only direct hydrologic effect observed with our SEM

modeling for the native fish richness within the Forgotten Reach was flow spikes or

assumed tributary-derived flash flood events, where native fish richness tended to

increase with increases in flow spikes. Poor water quality appeared to be a more

influential factor in determining the integrity of the native fish community for the

Forgotten Reach. Each region also appeared to be unaffected by the nonnative fish

abundance; however, there were higher abundances of generalist, saline-tolerant native

species compared with more sensitive species in the Forgotten Reach. Thus, generalist

native fishes may be affecting other native fishes more than nonnative species within the

region. Overall, we were able to observe how two regions, similar in their historical

physical template but different in ecological integrity, are affected similarly (e.g., habitat

35

simplification) and differently (e.g., poor water quality) by the modern flow and sediment

regimes of the Rio Grande.

Model Limitations

SEMs have become increasingly popular in ecological studies due to their

multivariate nature and ability to include latent variables (Grace 2008; Grace and others

2010; Tomarken and Waller 2004). They are especially intriguing for use in ecology

because of their ability to test multiple different structures for a given system. However,

as with all statistical models, they are only an estimation of reality, and SEMs routinely

omit important variables to increase model fit (Tomarken and Waller 2004). These

omitted variables can result in biased model constructs and final inferences for a given

system. For our models, we had to work largely with sporadic monitoring data and thus

had limited sample sizes for both the Forgotten Reach and the Big Bend region. We were

limited to using mean daily data that may completely miss most flows spikes as some can

occur over a matter of hours. Furthermore, many of the sampling sites within the

Forgotten Reach are in anthropogenically altered reaches of Forgotten Reach segment

(e.g., levees, agriculture), which may not be representative of the entire Forgotten Reach

segment as there are reaches not directly affected by human impact. Discussed

previously, there are multiple (potentially infinite) ways to specify an SEM, and for our

models, we assumed geomorphic, hydrologic, water quality, and nonnative fish species to

be the most critical limiting factors impacting the native fish diversity for the Rio Grande

in the Forgotten Reach and Big Bend region. Nonetheless, our models presented here

illustrate their potential use for identifying limiting factors on large, regulated riverine

ecosystems. Perhaps more importantly, the SEM results presented here also provide the

36

impetus for future studies to directly measure and test both the physical and ecological

components of an ecosystem.

Implications

Using SEMs have allowed us to better understand changes to ecosystem structure

and function for a degraded desert river where few comprehensive studies have

previously been conducted. Habitat heterogeneity and adequate stream flows may be

critical mechanisms driving alterations to the aquatic ecosystem and associated

biodiversity for the Big Bend region. In combination with adequate stream flows, water

quality and habitat heterogeneity may drive the ecosystem structure and associated

biodiversity for the Forgotten Reach. Having identified the most critical limiting factors

affecting the Rio Grande’s ecosystem integrity may contribute to the identification of

management policies aimed at innovative solutions to increase the amount, magnitude, or

timing of flows. Specifically, the Big Bend region will benefit most from the restoration

of flows that can reestablish and maintain critical habitat. The Forgotten Reach is most

affected by water quality and habitat degradation associated with the substantial channel

narrowing and simplification in this segment. Thus, there would need to be a substantial

change in the quantity and timing of flows to re-widen and reestablish complex habitat;

however, flows that can maintain better water quality conditions should be the primary

consideration in terms of restoration. Prescribing adequate stream flows for maintaining

sufficient habitat complexity and water quality conditions may be key for restoring the

Rio Grande back to a suitable level of ecosystem integrity and resilience that allows the

persistence of imperiled native fishes. We now have a better understanding of how the

physical template of a river can structure the ecological factors, and in turn, how the

37

ecological factors are used to inform the importance of the physical structure.

Acknowledgments

Funding was provided by the U.S. Geological Survey, South Central Climate

Science Center, the National Park Service, USU award by the Ecology Center at Utah

State University, and by the U.S. Geological Survey, Utah Cooperative Fish and Wildlife

Research Unit (in-kind). Data were generously shared by S. Miyazono, R. Patiño, C.M.

Taylor, and K. Saunders for both the Forgotten Reach and the Big Bend region. I would

like to thank B. Laub for his generous help with both the modeling process and reviews

of earlier drafts. In addition, I would like to thank J. Brahney for their guidance and

reviews throughout my project. Any use of trade, firm, or product names is for

descriptive purposes only and does not imply endorsement by the United States

Government.

Literature Cited

Allan JD, Castillo MM, editors. 2007. Stream ecology: structure and function of running

waters, second edition. The Netherlands: Springer-Dordrecht.

Allred TM, Schmidt JC. 1997. Channel narrowing by vertical accretion along the Green

River near Green River, Utah. Geological Society of America Bulletin 111: 1757-

1772.

Archdeacon TP. 2016. Reduction in spring flow threatens Rio Grande Silvery Minnow:

trends in abundance during river intermittency. Transactions of the American

Fisheries Society 145: 754-765.

Arhonditsis GB, Stow CA, Steinberg LJ, Kenney MA, Lathrop RC, McBride SJ,

Reckhow KH. 2006. Exploring ecological patterns with structure equation

modeling and Bayesian analysis. Ecological Modelling 192: 385-409.

Beaujean AA. 2014. Latent variable modeling using R: a step-by-step guide. Taylor and

Francis: New York, New York.

38

Bestgen KR, Platania SP. 1988. The ichthyofauna and aquatic habitats of the Rio Grande

from the New Mexico-Texas border to Big Bend National Park. U.S. Fish and

Wildlife Service. 59 p.

Blythe TL, Schmidt JC. 2018. Estimating the natural flow regime of rivers with long-

standing development: The Northern branch of the Rio Grande. Water Resources

Research 54: 1212-1236.

Bœuf G, Payan P. 2001. How should salinity influence fish growth? Comparative

Biochemistry and Physiology 130: 411-423.

Brandt SA. 2000. Classification of geomorphological effects downstream of dams.

Catena 40: 375-401.

Bunn SE, Arthington AH. 2002. Basic principles and ecological consequences of altered

flow regimes for aquatic biodiversity. Environmental Management 30: 492-507.

Burlakova LE, Karatayev AY, Karatayev VA, May ME, Bennett DL, Cook MJ. 2011a.

Biogeography and conservation of freshwater mussels (Bivalvia: Unionidae) in

Texas: patterns of diversity and threats. Diversity and Distributions 17: 393-407.

Burlakova LE, Karatayev AY, Karatayev VA, May ME, Bennett DL, Cook MJ. 2011a.

Endemic species: contribution to community uniqueness, effect of habitat

alteration, and conservation priorities. Biological Conservation 144: 155-165.

Cargnelli LM, Gross MR. 1996. The temporal dimension in fish recruitment: birth date,

body size, and size-dependent survival in a sunfish (bluegill: Lepomis

macrochirus). Canadian Journal of Fisheries and Aquatic Sciences 53: 360-367.

Collier M, Webb RH, Schmidt JC. 1996. Dams and rivers: primer on the downstream

effects of dams. U.S. Geological Society Circular 1126.

Dahm CN, Edwards RJ, Gelwick FP. 2005. Gulf Coast rivers of the Southwestern United

States. Benke AC, Cushing CE, editors. Rivers of North America. El Sevier

Academic Press. 180-228.

Davis JR. 1980. Species composition and diversity of benthic macroinvertebrates in the

upper Rio Grande, Texas. Southwestern Naturalist 25: 137-150.

Dean DJ, Schmidt JC. 2011. The role of feedback mechanisms in historic channel

changes of the lower Rio Grande in the Big Bend region. Geomorphology 126:

333-349.

Dean DJ, Scott ML, Shafroth PB, Schmidt JC. 2011. Stratigraphic, sedimentologic, and

dendrogeomorphic analyses of rapid floodplain formation along the Rio Grande in

Big Bend National Park, Texas. Geological Society of America Bulletin 123:

39

1908-1925.

Dean DJ, Schmidt JC. 2013. The geomorphic effectiveness of a large flood on the Rio

Grande in the Big Bend region: insights on geomorphic controls and post-flood

geomorphic response. Geomorphology 201: 183-198.

Dean DJ, Topping DJ, Schimdt JC, Griffiths RE, Sabol TA. 2016. Sediment supply

versus local hydraulic controls on sediment transport and storage in a river with

large sediment loads. Journal of Geophysical Research: Earth Surface 121: 1-29.

Dudgeon D, Arthington AH, Gessner MO, Kawabata Z, Knowler DJ, Lévêque C,

Naiman RJ, Prieur-Richard A, Soto D, Stiassny MLJ, Sullivan CA. 2006.

Freshwater biodiversity: importance, threats, status, and conservation challenges.

Biological Reviews 81: 163-182.

Edwards RJ, Contreras-Balderas S. 1991. Historical changes in the ichthyofauna of the

lower Rio Grande (Río Bravo del Norte), Texas and Mexico. Southwestern

Naturalist 36: 201-212.

Elosegi A, Díez J, Mutz M. 2010. Effects of hydromorphological integrity on biodiversity

and functioning of river ecosystems. Hydrobiologia 657: 199-215.

Everitt BL. 1977. Historical Background, Chapter V. In A preliminary appraisal of the

cultural and historical resources found along the Rio Grande between Fort

Quitman and Haciendita, Texas. Prepared for the International Boundary and

Water Commission. Anthropology 5: 64 p.

Everitt BL.1993. Channel responses to declining flow on the Rio Grande between Ft.

Quitman and Presidio, Texas. Geomorphology 6: 225-242.

Everitt BL. 1998. Chronology of the spread of tamarisk in the central Rio Grande.

Wetlands 18: 658-668.

Fogarty MJ, Sissenwine MP, Cohen EB. 1991. Recruitment variability in the dynamics of

exploited marine populations. Trends in Ecology and Evolution 8: 241-246.

Fullerton W, Batts D. 2003. Hope for a living river: a framework for a restoration vision

for the Rio Grande. The Alliance for Rio Grande Heritage, World Wildlife Fund.

142 p.

Garrett GP, Edwards RJ. 2014. Changes in fish populations in the Lower Canyons of the

Rio Grande. Pages 396-408 in C. A. Hoyt and J. Karges, eds. Proceedings of the

Sixth Symposium on the Natural Resources of the Chihuahuan Desert Region.

October 14-17. Chihuahuan Desert Research Institute, Fort Davis, Texas, USA.

40

Grace JB. 2008. Structural equation modeling for observational studies. Journal of

Wildlife Management 72: 14-22.

Grace JB, Anderson TM, Olff H, Scheiner SM. 2010. On the specification of structural

equation models for ecological systems. Ecological Monographs 80: 67-87.

Graf WL. 2006. Downstream hydrologic and geomorphic effects of large dams on

American rivers. Geomorphology 79: 336-360.

Gray JE, Hines ME, Biester H. 2006. Mercury methylation influenced by areas of past

mercury mining in the Terlingua district, Southwest Texas, USA. Applied

Geochemistry 21: 1940-1954.

Hannah DM, Wood PJ, Sadler JP. 2004. Ecohydrology and hydroecology: a ‘new

paradigm’? Hydrological Processes 18: 3439 – 3445.

Hart DD, Finelli CM. 1999. Physical-biological coupling in streams: the pervasive effects

of flow on benthic organisms. Annual Review of Ecology and Systematics 30:

363-395.

Hauer FR, Lorang MS. 2004. River regulation, decline of ecological resources, and

potential for restoration in a semi-arid lands river in the western USA. Aquatic

Sciences 66: 388-401.

Heard TC, Perkin JS, Bonner TH. 2012. Intra-annual variation in fish communities and

habitat associations in a Chihuahuan Desert reach of the Rio Grande/Rio Bravo

del Norte. Western North American Naturalist 72: 1-15.

Hermoso V, Clavero M, Blanco-Garrido F, Prenda J. 2011. Invasive species and habitat

degradation in Iberian streams: an analysis of their role in freshwater fish

diversity loss. Ecological Applications 21: 175-188.

Herrera AM, Dudley TL. 2003. Reduction of riparian arthropod abundance and diversity

as a consequence of giant reed (Arundo donax) invasion. Biological Invasions 5:

167-177.

Hintz WD, Mattes BM, Schuler MS, Jones DK, Stoler AB, Lind L, Relyea RA. 2017.

Salinization triggers a trophic cascade in experimental freshwater communities

with varying food chain length. Ecological Applications 27: 833-844.

Hintz WD, Relyea RA. 2017. A salty landscape of fear: responses of fish and

zooplankton to freshwater salinization and predatory stress. Oecologia 185: 147-

156.

Hooper DU, Chapin III FS, Ewel JJ, Hector A, Inchausti P, Lavorel S, Lawton JH, Lodge

DM, Loreau M, Naeem S, Schmid B, Setälä H, Symstad AJ, Vandermeer J,

41

Wardle DA. 2005. Effects of biodiversity on ecosystem functioning: a consensus

of current knowledge. Ecological Monographs 75: 3:35.

Hubbs C, Miller RR, Edwards RJ, Thompson KW, Marsh E, Garrett GP, Powell GL,

Morris DJ, Zerr RW. 1977. Fishes inhabiting the Rio Grande, Texas and Mexico,

between El Paso and the Pecos confluence. Pages 91-97 in R. R. Johnson and D.

A. Jones, eds. Symposium on the Importance, Preservation, and Management of

the Riparian Habitat. USDA Forest Service General Technical Report, RM-43.

July 9, 1977. Tucson, Arizona, USA.

International Boundary and Water Commission (IBWC). 1994. Binational study

regarding the presence of toxic substances in the Rio Grande/Rio Bravo and its

tributaries along the boundary portion between the United States and Mexico.

Final Report. 315 p.

Johnson CA, Everitt BL, Gerald RE. 1977. A preliminary appraisal of the cultural and

historical resources found along the Rio Grande between Fort Quitman and

Haciendita, Texas. Prepared for the International Boundary and Water

Commission. Anthropology 5: 64 p.

Keller DL, Laub BG, Birdsey P, Dean DJ. 2014. Effects of flooding and tamarisk

removal on habitat for sensitive fish species in the San Rafael River, Utah:

implications for fish habitat enhancement and future restoration efforts.

Environmental Management 54: 465-478.

Kelly ME. 2001. The Río Conchos: A preliminary overview. Texas Center for Policy

Studies, Austin, Texas, USA. 27 p.

Kennedy TA, Hobbie SE. 2004. Saltcedar (Tamarix ramosissima) invasion alters organic

matter dynamics in a desert stream. Freshwater Biology 49: 65-76.

Kennedy TA, Finlay JC, Hobbie SE. 2005. Eradication of invasive Tamarix ramosissima

along a desert stream increases native fish density. Ecological Applications 15:

2072-2083.

Kingsford RT, editor. 2006. Ecology of desert rivers. Cambridge University Press,

Cambridge.

Landis ME. 2001. The “Forgotten River” of the Rio Grande/Rio Bravo: investigation into

the reclamation of an arid riparian ecosystem. Bureau of Reclamation, El Paso

Field Division, El Paso, Texas. 34 p.

Laub BG, Budy P. 2016. Ecological changes in aquatic communities in the Big Bend

reach of the Rio Grande: synthesis and future monitoring recommendations.

Desert Landscape Conservation Cooperative: 1-43.

42

Laub BG, Budy P. 2015. Assessing the likely effectiveness of multispecies management

for imperiled desert fishes with niche overlap analysis. Conservation Biology 00:

1-11.

Lorenz OT, Riccobono SE, Smith P. 2016. Effects of salinity on the survival and

aggression of the invasive Rio Grande cichlid (Herichthys cyanoguttatus). Marine

and Freshwater Behaviour and Physiology 49: 1-8.

Metcalf AL. 1978. Some Quaternary molluscan faunas from the northern Chihuahuan

desert and their paleoecological implications. Symposium on Biological

Resources for the Chihuahuan Desert Region, National Park Service. Transactions

and Proceedings Series 3:53:66.

Miyamoto S, Fenn LB, Swietlik D. 1995. Flow, salts, and trace elements in the Rio

Grande: A review. Texas A&M University Agricultural Research Center, El Paso,

Texas, USA. 30 p.

Miyazono S, Patiño R, Taylor CM. 2015. Desertification, salinization, and biotic

homogenization in a dryland river ecosystem. Science of the Total Environment

511: 444-453.

Morcillo P, Esteban MA, Cuesta A. 2016. Heavy metals produce toxicity, oxidative stress

and apoptosis in the marine teleost fish SAF-1 cell line. Chemosphere 144: 225-

233.

Morcillo P, Esteban MA, Cuesta A. 2016. Heavy metals produce toxicity, oxidative stress

apoptosis in the marine teleost SAF-1 cell line. Chemosphere 144: 225-233.

Moring JB, Braun CL, Pearson DK. 2014. Mesohabitats, fish assemblage composition,

and mesohabitat use of the Rio Grande silvery minnow over a range of seasonal

flow regimes in the Rio Grande/Rio Bravo del Norte, in and near Big Bend

National Park, Texas, 2010-11. U.S. Geological Suvery, Scientific Investigations

Report 2013-5210. 89 p.

Moyer DL, Anderholm SK, Hogan JF, Philips FM, Hibbs BJ, Witcher JC, Matherne AM,

Falk SE. 2013. Knowledge and understanding of dissolved solids in the Rio

Grande—San Acacia, New Mexico to Fort Quitman, Texas, and plan for future

studies and monitoring. U.S. Geological Survey Open-File Report 2013-1190: 67

p.

Mueller JE. 1975. Restless river. El Paso, TX: Texas Western Press, University of Texas.

155 p.

Nagayama S, Nakamura F. 2018. The significance of meandering channel to habitat

diversity and fish assemblage: a case study in the Shibetsu River, northern Japan.

Limnology 19: 7-20.

43

Palmer MA, Ambrose RF, Poff NL. 1997. Ecological theory and community restoration

ecology. Restoration Ecology 5: 291-300.

Petts GE, Gurnell AM. 2005. Dams and geomorphology: research progress and future

directions. Geomorphology 71: 27-47.

Platania SP. 1991. Fishes of the Rio Chama and upper Rio Grande, New Mexico, with

preliminary comments on their longitudinal distribution. Southwestern Naturalist

36: 186-193.

Poff NL, Allan JD, Bain MB, Karr JR, Prestegaard KL, Richter BD, Sparks RE,

Stromberg JC. 1997. The natural flow regime: a paradigm for river conservation

and restoration. Bioscience 47: 769-784.

Poff NL, Zimmerman JKH. 2010. Ecological responses to altered flow regimes: a

literature review to inform the science and management of environmental flows.

Freshwater Biology 55: 194-205.

Poole GC. 2010. Stream hydrogeomorphology as a physical science basis for advances in

stream ecology. Journal of the North American Benthological Society 29: 12-25.

Porter SD, Barker RA, Slade Jr. RM, Longley G. 2009. Historical perspective of surface

water and groundwater resources in the Chihuahuan Desert Network, National

Park Service. Edwards Aquifer Research & Data Center, R1-09. 112 p.

Porter SD, Longley G. 2011. Influence of ground and surface water relations on algal

communities in the Rio Grande wild & scenic river. Edwards Aquifer Research

and Data Center, R1-11. 47 p.

PRISM Climate Group. 2004. Oregon State University, http://prism.oregonstate.edu.

Date accessed: October 20, 2017.

R Development Core Team. 2008. R: A language and environment for statistical

computing. Vienna, Austria: R Foundation for Statistical Computing. ISBN: 3–

900051-07-0. Retrieved from http://www.R-project.org.

Raines TH, Turco MJ, Connor PJ, Bennett JB. 2012. Streamflow gains and losses and

selected water-quality observations in five subreaches of the Rio Grande/Rio

Bravo del Norte from near Presidio to Langtry, Texas, Big Bend area, United

States and Mexico, 2006. U.S. Geological Survey, Scientific Investigations

Report 2012-5125. 30 p.

Richter BD, Davis MM, Apse C, Konrad C. 2011. A presumptive standard for

environmental flow protection. River Research Applications 28: 1312-1321.

44

Ricklefs RE, Schluter D. 1993. Species diversity in ecological communities: historical

and geographical perspectives. University of Chicago Press, Chicago, USA.

Sandoval-Solis S, Reith B, McKinney DC. 2010. Hydrologic analysis before and after

reservoir alteration at the Big Bend reach, Rio Grande/Rio Bravo. Center for

Research in Water Resources, Online Report 10-06. 72 p.

http://www.crwr.utexas.edu/reports/2010/rpt10-6.shtml

Schmidt JC, Everitt BL, Richard GA. 2003. Hydrology and geomorphology of the Rio

Grande and implications for river rehabilitation. Pages 25-45 in G. P. Garrett and

N. L. Allan, editors. Aquatic Fauna of the Northern Chihuahuan Desert,

Contributed Papers from a Special Session within the Thirty-Third Annual

Symposium of the Desert Fishes Council. Special Publications of the Museum of

Texas Tech University, Number 46.

Schmidt JC, Wilcock PR. 2008. Metrics for assessing the downstream effects of dams.

Water Resources Research 44: 1-19

Stotz NG. 2000. Historic reconstruction of the ecology of the Rio Grande/Rio Bravo

channel and floodplain in the Chihuahuan desert. Chihuahuan Desert Program,

World Wildlife Fund. 152 p.

Sousa R, Antunes C, Guilhermino L. 2008. Ecology of the invasive Asian clam

Corbicula fluminea (Müller, 1774) in aquatic ecosystems: an overview. Annales

de Limnologie – International Journal of Limnology 44: 85-94.

Stromberg JC, Beauchamp VB, Dixon MD, Lite SJ, Paradzick C. 2007. Importance of

low-flow and high-flow characteristics to restoration of riparian vegetation along

rivers in arid south-western United States. Freshwater Biology 52: 651-679.

Sutton-Grier AE, Kenney MA, Richardson CJ. 2010. Examining the relationship between

ecosystem structure and function using structural equation modelling: a case study

examining denitrification potential in restored wetland soils. Ecological

Modelling 221: 761-768.

Swanson BJ, Meyer GA, Coonrod JE. 2011. Historical channel narrowing along the Rio

Grande near Albuquerque, New Mexico in response to peak discharge reductions

and engineering: magnitude and uncertainty of change from air photo

measurements. Earth Surface Processes and Landforms 36: 885-900.

Ta W, Xiao H, Dong Z. 2008. Long-term morphodynamic changes of a desert reach of

the Yellow River following upstream large reservoirs’ operation. Geomorphology

97: 249-259.

45

Thomaz SM, Kovalenko KE, Havel JE, Kats LB. 2015. Aquatic invasive species: general

trends in the literature and introduction to the special issue. Hydrobiologia 746: 1-

12.

Tomarken AJ, Waller NG. 2005. Structural equation modeling: strengths, limitations, and

misconceptions. Annual Review in Clinical Psychology 1:31-65.

Upper Rio Grande Basin and Bay Expert Science Team (URGBBEST). 2012.

Environmental flows recommendations report. Final Submission to the

Environmental Flows Advisory Group, Rio Grande Basin and Bay Stakeholders

Committee and Texas Commission on Environmental Quality. 197 p.

U.S. Army Corps of Engineers (USACE). 2008. Forgotten Reach of the Rio Grande, Fort

Quitman to Presidio, Texas. Texas Commission on Environmental Quality

Section 729. 169 p.

U.S. Fish and Wildlife Service (USFWS). 2013. Federal Register. Volume 78 (2): 192 p.

Vaughan IP, Diamond M, Gurnell AM, Hall KA, Jenkins A, Milner NJ, Naylor LA, Sear

DA, Woodward G, Ormerod SJ. 2009. Integrating ecology with

hydromorphology: a priority for river science and management. Aquatic

Conservation: Marine and Freshwater Ecosystems 19: 113-125.

Vaughn CC, Nichols SJ, Spooner DE. 2008. Community and food web ecology of

freshwater mussels. Journal of the North American Benthological Society 27:

409-423.

Walsworth TE, Budy P, Thiede GP. 2013. Longer food chains and crowded niche space:

effects of multiple invaders on desert stream food web structure. Ecology of

Freshwater Fish 22: 439-452.

Williams WD. 1999. Salinisation: a major threat to water resources in the arid and semi-

arid regions of the world. Lakes and Reservoirs Research and Management 4: 85-

91.

Winemiller K, Lujan NK, Wilkins RN, Snelgrove RT, Dube AM, Skow KL, Snelgrove

AG. 2010. Status of freshwater mussels in Texas. Texas A&M Institute of

Renewable Natural Resources, College Station, Texas, USA. 62 p.

Woodhouse CA, Stahle DW, Díaz JV. 2012. Rio Grande and Rio Conchos water supply

variability over the past 500 years. Climate Research 51: 125-136.

World Wildlife Fund (WWF). 2016. Living Plant Report 2016. Risk and resilience in a

new era. WWF International, Gland, Switzerland.

46

Yang C, Everitt JH, Goolsby JA. 2011. Mapping Giant Reed (Arundo donax) infestations

along the Texas-Mexico portion of the Rio Grande with aerial photography.

Invasive Plant Science and Management 4: 402-410.

Tables and Figures

Table 2.1. Structural equation model fit statistics for the explained variation in the native

fish richness in the Big Bend and Forgotten Reach regions. The alternate models for the

Forgotten Reach are compared with the Big Bend model (Chi-squared value (χ2 ),

Comparative Fit Index (CFI), Tucker Lewis Index (TLI), Root Mean Square Error of

Approximation (RMSEA), proportion of explained variance in native fish richness (R2)).

Model

χ2 (p-value)

CFI

TLI

RMSEA

R2

Big Bend Region

0.400 0.994 0.985 0.030* 0.269

Forgotten Reach (compared to baseline Big Bend)

0.006

0.905

0.752

0.178*

0.664

Forgotten Reach (altered model structure)

0.066

0.946

0.915

0.104

0.630

47

Table 2.2. Standardized path coefficients for each SEM structure (=~ symbolizes a latent

variable, → symbolizes a regression; estimates in bold were statistically significant at p-

value < 0.05).

Path coefficients

Big Bend

Estimates

Forgotten

Reach

(baseline)

Estimates

Forgotten

Reach

(altered)

Estimates

Water quality =~ specific conductance

0.954

1.00

1.00

Water quality =~ sulfate 0.721 0.966 0.954

Water quality =~ water temperature -0.762

Water quality → flow median -0.054 -0.238 -0.356

Water quality → flow minimum 0.222 -0.343

Water quality → flow spikes -0.687 -0.214

Water quality → flow day of sampling -0.423

Nonnative → water quality 0.244 0.147 0.258

Nonnative → channel width

Nonnative → flow median 0.452 -0.009

Nonnative → flow minimum 0.006 -0.105

Nonnative → flow spikes 0.037 -0.204

Nonnative → flow day of sampling

Native richness → water quality 0.027 -0.437 -0.535

Native richness → channel width 0.483 0.378 0.376

Native richness → nonnative 0.147 0.009

Native richness → flow median 0.396 0.012

Native richness → flow minimum -0.544 0.209

Native richness → flow spikes -0.078 0.408 0.374

Native richness → flow day of sampling -0.137 -0.219 -0.213

Channel width → flow spikes 0.375

48

Figure 2.1. Map of the study areas: Forgotten Reach through Big Bend region along the

Rio Grande in West Texas, United States. Outline represents park boundary for Big Bend

National Park.

49

Figure 2.2. Hypothesized model for the overall native fish diversity of the Rio Grande

for both the Forgotten Reach and the Big Bend region.

50

Figure 2.3. SEM result for the Big Bend region using the hypothesized model structure

(Figure 2). Width of lines are scaled to the strength of the path coefficient, which is also

given over the path line.

51

Figure 2.4. SEM result for Forgotten Reach using the baseline Big Bend model structure.

Width of lines are scaled to the strength of the path coefficient, which is also given over

the path line.

52

Figure 2.5. SEM result for Forgotten Reach using an altered model structure where the

direct effects of minimum flows, median flows and nonnative relative abundance have

been removed from native richness. In addition, water temperature has been added to the

water quality latent variable and flow spikes now influence channel width. Width of lines

are scaled to the strength of the path coefficient, which is also given over the path line.

53

CHAPTER 3

TOO MUCH SEDIMENT, TOO LITTLE WATER: HOW THE MODIFIED

HYDROLOGIC AND SEDIMENT REGIMES OF THE RIO GRANDE

AFFECT THE NATIVE FISH COMMUNITY AND FOOD WEB

STRUCTURE IN BIG BEND NATIONAL PARK, TEXAS,

UNITED STATES

Abstract

Desert rivers are complex and dynamic and are also some of the most degraded

ecosystems globally, in part because they are competing with man for water resources.

Many native riverine species endemic to the arid and highly regulated Rio Grande have

been reduced or eradicated due to the loss of viable and connected habitat, poor water

quality, and nonnative plants and animals, all often associated with modified flow and

sediment regimes. The overall goal of my study was to determine if and how the

modified hydrologic and sediment regimes of the Rio Grande in the Big Bend region

have impacted the aquatic ecosystem structure and function with an emphasis on the

native fish community and food web structure. I used a multi-faceted approach involving

habitat availability analysis, fish community and abundance comparisons, and stable

isotope food web techniques to compare 12 sampling sites of differing habitat

heterogeneity, a priori classified as either complex or simple, and/or canyon and alluvial.

I found complex and canyon sites to contain a greater diversity of habitat types than

simple or alluvial sites. In addition, using random forest classification, the categorization

of complex or simple was best predicted based on the proportion of backwater

mesohabitats. Though fish diversity was not significantly different between sites of

54

varying habitat heterogeneity, I observed significantly greater catch per unit effort

(CPUE) in complex and canyon sites (~19% greater) relative to simple or alluvial sites,

indicating higher abundance. Chlorophyll a concentrations were relatively low and not

variable between sites of varying habitat complexity; however, the ‘scraper’

macroinvertebrate functional feeding group was significantly greater in all sites of greater

habitat complexity suggesting there is an abundance of attached algae, or periphyton in

those areas. Food web structure also appeared to be influenced by habitat heterogeneity,

and complex and canyons sites tended to be more trophically-diverse (nitrogen range

(NR) = 10.6 and 9.25, respectively) and contained more types of basal resources (carbon

range (CR) = 1.83 for both complex and canyon) than simple or alluvial sites (NR = 5.49

and 2.57; CR = 0.586 and 0.710, respectively). In addition, the niche breadth of fish and

aquatic and terrestrial invertebrates was wider and more complex in complex and canyon

sites. There was very little variability in isotopic niche space for fish and invertebrates

across vegetation types, and nonnative vegetation did not appear to trophically influence

the fish community. Finally, my mixing models demonstrated that aquatic

macroinvertebrates provided a greater contribution to the proportion of dietary items

consumed by fish across most sites except for alluvial sites. Collectively, my results

contribute to a greater understanding of desert river ecosystems by linking the physical

components (e.g., hydrology and sediment supply) to the habitat for biota and the

ecological components (e.g., fish community and food web structure). Ultimately these

results have important implications for native fish conservation by providing an

ecological framework to inform effective water management solutions.

55

Introduction

Rivers are some of the most ecologically-dynamic and complex ecosystems on

Earth. Stream ecologists have long studied the multi-dimensional interactions of a river

ecosystem between both the terrestrial and aquatic environment, such that multiple

theories have been proposed to explain the structure and function of these aquatic

systems (e.g., Vannote et al. 1980; Junk et al. 1989; Thorp and Delong 1994).

Understanding how biotic assemblages and nutrient pathways (e.g., aquatic and

terrestrial) interact to influence food web structure in large riverine ecosystems remains a

relevant question in stream ecology. Vannote et al. (1980) suggested a longitudinal

gradient where downstream biotic assemblages of a river are increasingly dependent on

upstream sources of nutrients. As a result, the biotic assemblages are structured

accordingly as the abundance of upstream basal resources and local processes change

along the river gradient. Junk et al. (1989) stressed the importance of flooding and

floodplain interactions to the biotic community. Thorp and Delong (1994) described

autochthonous primary production and direct, local sources of riparian organic matter as

important basal resources of food for upper aquatic trophic levels (e.g., primary

consumers). While all these theories have proven fundamental to our growing

understanding of large river ecosystem dynamics, a question remaining is how and if

these theories can be applied to large, dewatered desert rivers with extremely altered

hydrologic and sediment regimes, a now common phenomenon.

Desert river ecosystem structure and function are increasingly threatened by both

human society’s demand for water and climate-induced drought (Hauer and Lorang

2004). Long-term water regulation and diversions on many North American arid land

56

rivers have altered the natural variability and magnitude of stream flows and sediment

delivery (Ward and Stanford 1983; Poff et al. 1997; Schmidt and Wilcock 2008). For

example, reductions in stream flow that occur while the local sediment supply remains

unchanged, have led to channel simplification of many desert rivers in the southwest US,

including the Green River in Utah, the Rio Grande in far west Texas, and the Little

Colorado River in Arizona (e.g., Hereford 1984; Allred and Schmidt 1997; Dean and

Schmidt 2011). Dewatering often reduces flood magnitudes to the point where the river

can no longer maintain the former floodplain, disconnecting it from the main channel. As

a result, the channel banks often become more stable and can facilitate native and

nonnative riparian vegetation encroachment (Brandt 2000; Graf 2006).

Ecosystems can be fundamentally changed in cases where their physical structure

is altered by changes to the flow and sediment regimes. Habitat for biota can become

simplified and reduced in availability, and biotic assemblages typically change in

response. Endemic taxa are often lost or forced into sub-optimal habitat due to decreased

habitat availability or suitability (Townsend and Winfield 1985; McDonald et al. 1992;

Kalb et al. 2018). For example, a simulated spring flood pulse within the northern branch

of the Rio Grande in New Mexico created access to adequate floodplain spawning

grounds for the Rio Grande silvery minnow, a federally-listed endangered species, and

cued spawning behaviors (Archdeacon 2016). Access to these floodplain-spawning

grounds created habitat and productivity necessary for the development of larval

minnows, which appeared to increase larval survival (Dudley and Platania 1997, 2007;

Medley and Shirey 2013). However, the timing of this simulated flood pulse may have

also resulted in the early evacuation of developing fry (D. Strohm, personal

57

communication), with potentially negative consequences at the next life stage.

Nonetheless, this example of flood management demonstrated how hydrologic timing

and magnitude of a river are inextricably coupled with the life history of organisms, and

even small alterations to flood regimes can have important consequences for a sensitive

species.

Changing the hydrologic and sediment regimes can also impact riparian

communities and their distributions. Associated modifications to riparian composition

influences the trophic structure and associated food web energy pathways within a river

(Mineau et al. 2011, 2012). For example, nonnative saltcedar (Tamarix spp) altered a

small desert river once seasonally-dependent on autochthonous sources of energy to an

ecosystem dependent on continuously-available allochthonous sources of energy,

resulting in nonnative fish species out-competing native fish species (Kennedy et al.

2004). Consequently, the removal of saltcedar in this system contributed to the increased

abundance of native fish species (Kennedy et al. 2005). Anthropogenic changes or

impacts to either the physical (e.g., hydrology or sediment supply) or ecological integrity

(e.g., shifted riparian composition) of desert rivers compounds the already sensitive

nature of these harsh, drought-prone environments (Ward and Stanford 1983; Cross et al.

2011).

Examining fish communities in modified river systems can be a useful tool to

assess the ecological integrity of these aquatic systems. The structure of communities and

how they change through time and space is important for determining both the physical

components of these environments and investigating the potential timing and magnitude

of human-induced impacts (Sá-Oliveira et al. 2015). In the Murray-Darling river basin,

58

there was an observed trend of reduced species diversity in highly regulated catchments

of the basin presumably due to modified stream flows (Gehrke et al. 1995). Fish

community assessment is also beneficial for determining the effectiveness of restoration.

For example, flow restoration to reaches with previous modified flow changes increased

the abundance of fish guilds present in the Rhone River, France (Daufresne et al. 2015).

In general, understanding the response of fish communities to modified hydrologic and

sediment regimes will increase our knowledge of degraded desert river systems and help

prioritize management actions.

Riverine food web structure is also influenced by degraded habitat conditions

associated with anthropogenic water use and drought. Habitat size and structure

determine productivity, food types and abundances, resource partitioning, and

interactions among organisms (Finlay 2011; Takimoto and Post 2013). Thus, changes to

habitat availability can influence the types and abundances of food resources for aquatic

consumers, impacting the food web structure by changing the overall communities

present (Bain et al. 1988; McHugh et al. 2015). Low complexity habitat may lead to

increased competition because of the limited access and availability of resources. Those

that outcompete other organisms within these simplified habitats may homogenize the

native community if they are dominant competitors (Menge and Sutherland 1987;

Jackson et al. 2001; Hayden et al. 2015). Food web structure is subsequently affected by

these changes to the communities, and food chain length and lower trophic diversity (e.g.,

increased omnivory) are often the result of increased community homogenization (Post

2002a).

Given the well documented impacts and associated changes of desert river

59

ecosystems based on altered flow and sediment regimes, it is sufficient to say the

fundamental theories of river ecology described previously all likely apply at some level.

There are important hydrologic and sediment dynamics that propagate downstream,

critical flood stages/pulses allow access to essential floodplain habitats and nutrients, and

direct riparian input alters energy pathways (Kennedy et al. 2004, 2005; Schmidt and

Wilcock 2008; Archdeacon 2016). Thus, the overarching goal for this study was to use a

habitat availability and food web-based approach to examine how changes to the physical

environment alter the aquatic ecological structure and function of a desert river

ecosystem.

Within this context, I had two objectives to better understand the altered physical

and ecological environment and associated effects on the native fish community of the

highly regulated Rio Grande along the US-Mexico border. My first objective was to

identify the effects of the altered hydrologic and sediment regimes of the Rio Grande on

habitat heterogeneity and the associated native fish community. Based on prior

geomorphic assessments (Dean and Schmidt et al. 2011, 2013; J. Bennett, personal

communication), I chose a suite of sampling sites exhibiting varying levels of channel

morphology and simplification throughout the river. I predicted sampling sites with

greater habitat heterogeneity would be more complex and contain greater fish diversity

than sites with low habitat heterogeneity. My second objective was to determine if and

how these altered hydrologic and sediment regimes have potentially affected energy flow

pathways (e.g., basal resource production) and structure of the aquatic food web. I

predicted food web structure (e.g., food chain length, trophic niche space) would be more

complex within sites with greater habitat heterogeneity than sites with low habitat

60

heterogeneity because I assumed complex sites to better partition resource-use and

facilitate multiple species. I also predicted nonnative vegetation alters the food web

function by contributing differentially to dietary items of upper trophic-level consumers

than does native vegetation. Overall, I hope to contribute to the ecological knowledge

surrounding desert food web structure and function, and provide insight to more effective

and innovative stream flow management solutions for the Rio Grande in the Big Bend

region, arguably the most ecologically-intact reach of this river that remains.

Study area

The Rio Grande, known as the Río Bravo in Mexico, is the fifth longest river in

North America, extending approximately 3,000 km-long across eight states and two

countries and draining a basin of approximately 558,000 km2 (Patiño-Gomez et al. 2007).

My study was conducted in the Big Bend region of the Lower Rio Grande Basin,

designated as the segment extending from the confluence of the Rio Grande and Rio

Conchos, a large tributary originating in the Sierra Madre Occidental of northern Mexico,

to Amistad Reservoir for approximately 490 km (Figure 3.1; Dean and Schmidt 2011).

The Big Bend region is located in the Chihuahuan Desert, receiving as little as 200 mm

annually in total precipitation and having maximum air temperatures reaching 33°C

(PRISM Climate Group, 2017).

Historically, floods within the Big Bend region were supplied by both spring

snowmelt in the southern Rocky Mountains and rains associated with the North American

monsoon or dissipating tropical cyclones. Thus, the high flow period in the Big Bend

region of the Rio Grande likely extended from early April to early fall (Blythe and

Schmidt 2018, T. Blythe 2018 thesis). Construction of Elephant Butte Dam in New

61

Mexico in 1915 and irrigation diversions eliminated the spring snowmelt stream flows

such that approximately 95% of the total annual flow upstream of the Rio Conchos is

consumed before it reaches the confluence. Contemporary stream flows in the Big Bend

region are now reliant upon flows provided by natural floods of the Rio Conchos,

occasional dam releases from Luis L. Leon Dam on the Rio Conchos in Mexico, and

floods on Chihuahuan Desert tributaries (Dean and others 2016). Consumptive water use

in the Rio Conchos watershed is not as complete as it is upstream of the Rio Conchos-Rio

Grande confluence, and the Rio Conchos continues to supply most of the flow to the Big

Bend region. In other words, most of the Rio Grande’s stream flow in the Big Bend

region has come from Mexico. Today, the Rio Grande in the Big Bend is one of the most

ecologically-intact portions of the river (CEC 2014).

Though stream flows are higher in magnitude through the Big Bend region, the

loss of spring snowmelt flows is likely to have impacted the river’s ability to effectively

transport the continued sediment supply from tributary-sourced flash flood events (T.

Blythe 2018 thesis). Thus, sediment is not evacuated and contributes to channel

narrowing and vertical accretion in the Big Bend region (Dean and Schmidt 2011; Dean

and other 2011). In addition, nonnative riparian plant species, primarily giant reed

(Arundo donax), have become well established on the inset floodplains associated with

sediment accumulation. These nonnative species, in combination with the modified

timing and magnitude of large flood events, likely prevent the re-establishment of native

riparian vegetation such as cottonwood (Populus spp.) and willow (Salix spp.), and act as

a positive feedback to channel narrowing as they trap sediment and actively stabilize the

channel banks (Dean and Schmidt 2011, 2013). Occasionally the channel is re-widened

62

by large (>1000 m3 s-1), rare (10- to 15-year frequency) flood events, termed ‘reset’

floods (Dean and Schmidt 2013). These reset events are ecologically-important, because

they have the capacity to restore channel complexity by actively rewidening the channel

and reducing nonnative giant cane (Dean and Schmidt 2011, 2013; Dean et al. 2011;

Dean et al. 2016).

The native fish community of the Rio Grande in the Big Bend region has

responded in kind to these hydrologic, geomorphic, and riparian community composition

changes. Many native species have become threatened (e.g., state-threatened Mexican

stoneroller (Campostoma ornatum), Chihuahua shiner (Notropis chihuahua), blue sucker

(Cycleptus elongatus)) or listed as federally endangered (e.g., Rio Grande silvery

minnow) and extirpated from the Big Bend region (Miyazono et al. 2013, 2015). The

relative abundances of many sensitive native species within this region have decreased in

the last 20th century associated with altered hydrologic conditions, poor water quality,

and loss of necessary habitat associated with increased sedimentation. In contrast, native

generalist, saline-tolerant species such as red shiner (Cyprinella lutrensis), Tamaulipas

shiner (Notropis braytoni), and western mosquitofish (Gambusia affinis) have increased

in their relative abundances in recent years (Miyazono et al. 2015). Nonnative fish

species such as plains killifish (Fundulus zebrinus) and common carp (Cyprinus carpio)

are also more abundant in the present-day Big Bend region of the Rio Grande.

Restoration of stream flows that would increase water quality and habitat heterogeneity

will be necessary to prevent species homogenization and rehabilitate and restore the

native fish community for the Big Bend region.

63

Methods

Field Sampling Habitat complexity and water quality

Sufficient habitat complexity, to the same extent that was the case in an

unregulated river, is often associated with diverse species and aquatic communities, and

thus the degree of habitat complexity can provide one measure of the ecological integrity

of an aquatic ecosystem. I sampled 12 1-km reaches with different degrees of habitat

complexity, a priori classified as either simple or complex based on previous geomorphic

surveys (Dean and Schmidt 2011, 2013; J. Bennett and D. Dean, personal

communication), and classified as either in canyon-bound portions of the river or in wide

alluvial valleys (hereafter called alluvial). I sampled these sites five times in 2016 and

2017. I collected samples in different seasons: late winter/early spring 2016, spring 2016,

fall 2016, winter 2017, and summer 2017 (Table 3.1). I measured discharge during each

sampling event to determine if and how the magnitude of stream flow influences

available in-channel habitat complexity and thus, native aquatic diversity (e.g., do higher

magnitude flows provide more or less diversity for in-channel habitat types and perhaps

increase/decrease native fish diversity?). Discharge differed during each sampling trip by

approximately one order of magnitude: 1.9 m3/s in May 2016 and 30 m3/s in October

2016 (Table 3.1). I considered the May 2016 discharge measurement to be considered

low stream flow and the October 2016 discharge is considered a moderate stream flow

magnitude.

I sampled eight sites within Big Bend National Park, two of which were in the

alluvial valleys and six in Boquillas Canyon. I also selected four sample sites located

within the Lower Canyons of the Rio Grande Wild and Scenic River. Stream flow

64

increases in the Lower Canyons due to numerous springs from the Edwards-Trinity

aquifer (Raines et al. 2012). As a result, baseflow progressively increases downstream

and water quality improves.

Using methods similar to Moring et al. (2014), I mapped complexity at the

mesohabitat scale for each 1-km site at baseflow conditions. and delineated runs, pools,

riffles, backwaters, forewaters, embayments, boulder bars, and side pools (Figure 3.2a). I

measured depth (m), velocity (m/s), and substrate composition of each mesohabitat type

in proportion to their overall distribution in each sampling site (Figure 3.2b, 3.3). Within

a geographic information system (GIS), I digitized habitat maps using aerial imagery

flown at similar discharges in ESRI ArcMap software (version 10.3.1) to quantify total

area and calculate the proportions of mesohabitat area for each sampling site. I used a

one-way analysis of variance to identify differences in total area between simple and

complex sampling sites and between alluvial and canyon reaches. I recorded water

quality measurements using a YSI 556 Multiprobe meter for each mesohabitat type to

identify potential patterns in water quality associated with each mesohabitat and overall

habitat complexity. I measured water temperature (°C), specific conductance (µS/cm),

total dissolved solids (mg/L), dissolved oxygen concentration (mg/L) and saturation (%),

and pH. I calculated the mean of each water quality parameter for each mesohabitat type

and between seasons (Appendix, Table A1).

Fish community

Freshwater fish community assemblages are determined in large part by their

interactions with their surrounding physical environment (Jackson et al. 2001). Thus,

modifications to fishes preferred and required habitat can impact and alter the fish

65

community present (Fausch et al. 1990). I compared fish diversity and abundance

between simple and complex sampling sites and between canyon and alluvial reaches to

assess how the fish community has responded to the modified flow and sediment regimes

of the Rio Grande. I used a combination of active and passive field sampling techniques

and multiple types of gear to quantify abundance of all size classes of fishes present and

minimize potential gear bias within my results for describing the native fish community.

I actively seined (seine size: 6 m x 1 m, 6 mm mesh) and used angling techniques

in all mesohabitat types for each sampling site and recorded length of the seine haul (m)

and time (hrs) spent angling to calculate catch per unit effort (CPUE; fish m-2, fish hr-1).

In addition, I set passive fyke nets, hoop nets, and minnow traps in side pool, pool, and

boulder bar mesohabitats that are deep and have slow velocity, and recorded the time set

for each sampling site. For all catch, I identified, counted, and measured total length and

weight for each individual fish. In addition, I fin-clipped at least five individual fish for

each species per length class and air-dried fin-clips in a 1.5 mL micro-centrifuge tube for

stable isotope analysis.

Primary productivity

Primary producers are important basal food resources for both fish that directly

consume primary producers and fish that feed on secondary consumers feeding on

primary productivity. To better understand the contemporary patterns in primary

productivity abundance and identify the importance of primary producers supporting

secondary production, I measured the concentrations of chlorophyll a (µg/L) for seston,

soft sediment algae, and periphyton and analyzed each using stable isotopes. Within each

sampling site, I sampled periphyton by randomly selecting five different stones within a

66

riffle mesohabitat and scraping a known area for each stone into a sampling pan. I then

measured total volume scraped (mL) and filtered a known volume through a 2.4 cm

Whatman glass fiber filter (GF/C), collecting multiple filters for both chlorophyll a and

stable isotope analysis. For my soft sediment algae samples, I targeted slow, shallow

mesohabitat types (e.g., backwaters, forewaters, and embayments) with either a silt or

sand substrate at the time of sampling. I then gently placed a petri-dish with a known area

of 31.7 cm2 within the thin, flocculent layer of algae on top of the silt/sand substrate,

placed a flattened piece of plastic between the flocculent layer and the substrate, and

deposited sample into a sample pan. I used the same method of filtering for the soft

sediment samples as the periphyton. I collected seston samples by directly sampling from

the mid-point of the river’s depth within a run mesohabitat and filtering the known

volume through a glass fiber filter. For all algal samples collected, I placed filters in

aluminum foil, labeled by type and volume filtered, and kept frozen until further analysis.

Terrestrial and aquatic invertebrates

In many riverine ecosystems throughout the world, nonnative vegetation has

changed river ecosystem efficiency by increasing the amount of leaf litter deposited in

the river and decreasing autochthonous primary productivity (Mineau et al. 2011, 2012).

Along much of the Rio Grande, nonnative giant cane has become well established and

altered the contemporary riparian community (Everitt 1998; Dean et al. 2011). As giant

cane can produce up to 20 t ha-1 of leaf litter, a substantial portion of organic matter from

giant cane is likely input into the river (Yang et al. 2011). It is important to note that Big

Bend National Park has undergone substantial effort to eradicate giant cane throughout

the park, particularly within Boquillas Canyon. Thus, the sites I sampled were a

67

combination of treated and untreated sites in terms of nonnative vegetation eradication

efforts. To quantify the impacts of nonnative vegetation to the food web structure of the

Rio Grande and identify if the system has shifted in terms of domination by allochthony

or autochthony, I collected terrestrial invertebrates and leaf litter from native and

nonnative vegetation, and open sites, or sites where there were exposed gravel and sand

bars near water’s edge. I used small plastic tubs equipped with UV blacklights and filled

each with 250 mL of distilled water and one to two drops of dish detergent (to break

surface tension) in each of the three sites. This method was pioneered by Kennedy et al.

(2016) in a Citizen Science protocol used along the Colorado River in Grand Canyon. I

turned lights on for one hour after approximate nautical sunset and preserved the entire

contents collected during that hour in a 250-mL Nalgene bottle filled with a 95% ethanol

solution in the field. Upon return to the laboratory, I changed the ethanol solution to 70%

to prevent invertebrate breakdown.

Aquatic invertebrates are important primary consumers and food resources for

native fishes, as well as excellent indicators of aquatic ecosystem health (Hawkins et al.

2000). Thus, aquatic invertebrates are an important aspect for understanding the structure

of the Rio Grande’s aquatic food web and ecological integrity. I sampled invertebrates in

both slow velocity (e.g., pools, backwaters, forewaters, embayments) and high velocity

(e.g., riffles, runs) areas. I used a combination of 3-min kicks and 1-m sweeps in all

mesohabitat types using a 0.30-m diameter, heavy duty D-frame dip net with 500-µm

mesh. Because I was interested in community differences between complex and simple

sites, and between alluvial and canyon sites, I combined all samples collected in 18.9-L

plastic bucket and picked all invertebrates in the field. I preserved invertebrates in a 250-

68

mL Nalgene bottle filled with a 95% ethanol solution. Like my terrestrial invertebrate

samples, I changed the ethanol solution to 70% upon return to the laboratory.

Laboratory Methods

Chlorophyll a abundance and community analysis

In order to estimate primary productivity abundance, I analyzed my algal samples

using fluorometric analysis. Fluorometric analysis is a highly sensitive technique to

extract chlorophyll a (Welschmeyer 1994). I began by extracting all algal filter samples

in a 10-mL 95% ethanol solution for 24 hours. After extraction, I thoroughly mixed the

chlorophyll a from the filter into the solution and measured 6-mL of solution into a glass

cuvette. I then measured the extracted chlorophyll a concentrations with a Turner 10 AU

Fluorometer. To calculate the final concentrations from the original samples, I used two

different equations based on the types of algae being analyzed:

For seston samples:

𝜇𝑔 𝐿−1 =𝑓𝑙𝑢𝑜𝑟𝑜𝑚𝑒𝑡𝑒𝑟 𝑟𝑒𝑎𝑑𝑖𝑛𝑔 (𝜇𝑔 𝐿−1)∗𝑒𝑥𝑡𝑟𝑎𝑐𝑡𝑖𝑜𝑛 𝑣𝑜𝑙𝑢𝑚𝑒 (𝑚𝐿)

𝑣𝑜𝑙𝑢𝑚𝑒 𝑓𝑖𝑙𝑡𝑒𝑟𝑒𝑑 (𝑚𝐿)

For periphyton and soft sediment samples:

𝜇𝑔 𝑐𝑚−2 =𝑓𝑙𝑢𝑜𝑟𝑜𝑚𝑒𝑡𝑒𝑟 𝑟𝑒𝑎𝑑𝑖𝑛𝑔 (𝜇𝑔 𝐿−1)∗𝑒𝑥𝑡𝑟𝑎𝑐𝑡𝑖𝑜𝑛 𝑣𝑜𝑙𝑢𝑚𝑒 (𝑚𝐿)∗(𝑖𝑓 𝑢𝑠𝑒𝑑) 𝑑𝑖𝑙𝑢𝑡𝑖𝑜𝑛 𝑓𝑎𝑐𝑡𝑜𝑟

𝑣𝑜𝑙𝑢𝑚𝑒 𝑓𝑖𝑙𝑡𝑒𝑟𝑒𝑑 (𝑚𝐿)∗𝑎𝑟𝑒𝑎 𝑠𝑐𝑟𝑎𝑝𝑒𝑑 (𝑐𝑚2)

Terrestrial and aquatic invertebrate community classification

Though allochthonous input can be a key basal resource for many rivers (Vannote

et al. 1980), shifts in the types of allochthonous input can change the types and quality of

carbon input into upper levels of the aquatic food web (Kennedy et al. 2004, 2005). For

69

example, nonnative riparian vegetation encroachment may alter the communities and

abundances of terrestrial invertebrates input into river ecosystems compared to native

riparian vegetation. I compared terrestrial invertebrate communities between native and

nonnative vegetation, and open sites. For all sample types, I counted and identified the

entire contents of each native, nonnative, and open site within each sampling site. I

identified aquatic invertebrate emergent adults to family level and all other invertebrates

to order. My goal was to identify the functional importance of aquatic emergent adults

within the terrestrial environment (e.g., riparian vegetation) and link that importance with

the potential impacts of nonnative riparian vegetation to the aquatic food web structure

for the Rio Grande.

Functional groups of aquatic invertebrates can serve as important indicators of

ecological integrity, primarily because of the multiple ways they can consume food

resources (Wallace and Webster 1996). In addition, aquatic invertebrates can influence

primary productivity, decomposition, and pathways of nutrient cycles. I sorted, counted,

and identified aquatic invertebrate samples to functional feeding group for each sampling

site. I classified aquatic invertebrate samples into five different categories including:

scrapers, shredders, filtering collectors, gathering collectors, and predators. These

functional feeding groups follow the classifications set forth by Cummins et al. (1979)

and provide an important linkage with the physical template of the river and the

ecological integrity. For example, by comparing functional groups between different

types of habitat complexity (e.g., simple versus complex), I identified mesohabitats

critical for maintaining diversity at the secondary productivity level, and subsequently

food resources important for the native fish community.

70

Stable isotope preparation

Stable isotopes have the power to infer trophic position, trace the sources of

carbon and nitrogen, and provide insight on the integrity of an ecosystem (Vander

Zanden et al. 1997, Grey 2006). Carbon and nitrogen isotopes are useful indictors of

trophic position and the source of carbon throughout the system (Vander Zanden et al.

1996, Post 2002, Grey 2006). The amount of the heavy isotope of nitrogen in an

organism’s tissue is related to the diet items each organism consumes, and thus, their

trophic level. Carbon isotopes stay constant as they are transferred from an ultimate

source to the aquatic food web and can be used to identify sources of allochthonous of

energy to the river. In addition, hydrogen isotopes are useful for both deriving

allochthonous sources of energy and providing distinctions among local food webs.

I used a subset of all the different samples taken as described previously (fish fin-

clips n = 311, algae n = 47, terrestrial and aquatic invertebrate n = 140, and leaf litter n =

13) and analyzed each for carbon (δ13C) and nitrogen (δ15N) content. In addition, I

analyzed both fish fin-clips (n = 92) and invertebrates (n = 140) for hydrogen (δ2H)

content. To analyze sufficient mass of fish fin-clips, I consolidated fish samples of the

same species and similar length class; however, this reduced my sample size of total fish

analyzed. I dried all samples for 48-hours at 70°C and ground entire sample (e.g., fin-

clips, terrestrial and aquatic invertebrates, or leaf litter) into a homogenous powder using

a mortar and pestle. I placed each sample into either an Ultra-pure tin capsule for δ13C

and δ15N isotopes, or a silver capsule for the δ2H isotope (Costech Analytical, Valencia,

CA, USA), and weighed samples to the nearest 0.0008 mg.

I sent samples to both Washington State University Stable Isotope Laboratory and

71

the Dennis Newell Stable Isotope Laboratory at Utah State University. Methods at each

laboratory were similar, and technicians converted samples for δ13C and δ15N analysis to

N2 and CO2 with an elemental analyzer (ECS 4010 Costech Analytical, Valencia, CA,

USA). N2 and CO2 were separated with a 3-meter Gas Chromatography (GC) column and

analyzed with a continuous flow isotope ratio mass spectrometer (Delta PlusXP, Thermo

Finnigan, Bremen). The standards for the δ13C and δ15N samples were PeeDee belemnite

(δ13C) and atmospheric nitrogen (δ15N). For δ2H, samples were converted to CO and H2

with a pyrolysis elemental analyzer (TC/EA, Thermo Finnigan, Bremen), then each gas

was separated with a GC column and analyzed using a continuous flow isotope ratio mass

spectrometer (Delta PlusXP, Thermo Finnigan, Bremen). The standards for δ2H were

potassium nitrate, caribou hoof, and kudu horn and were normalized relative to the

Vienna standard Mean Ocean Water content. Isotopic signatures are reported in -

notation:

δ13C, δ15N, or δ2H = [(Rsample

Rstandard) -1] x1000

Where R is the ratio of 13C/12C, 15N/14N, or 2H/1H.

Statistical Analyses

Random forest classification

Classification regression methods are widely used in ecology, because they are

simple to interpret, accurate, and characterize complex interactions (Cutler et al. 2007).

Classification trees can predict complex patterns using either numerical or categorical

response data. These types of models use an optimization pattern to select a ‘split’ in the

data, a predictor variable, and a ‘cut-off’ that homogenizes the data into subgroups. The

72

‘splitting’ of a classification tree continues until the homogenized groups of data no

longer decrease the accuracy of the model (Cutler et al. 2007). Random forests are a

particular class of classification tree that fit many trees to a given data set and combine

each prediction from every tree, essentially bootstrapping the data. Those data that are

not randomly selected during the bootstrapping process are used for later prediction once

a tree is ‘fully grown’, and are known as ‘out-of-bag’ (OOB) estimates. The OOB data

are then used to assess the accuracy and error estimates of each predictor analyzed within

the model. Validation of the model fit is assessed using five different metrics: overall

percentage of models correctly classified (PCC), sensitivity, specificity, kappa, which is a

measure of agreement between predicted presences and absences, and area under the

curve (AUC) estimates. Model fit is ‘good’ when all metrics are high in value, and

specifically, if the AUC estimates are closer to one.

To predict the level of habitat complexity for each of my sampling sites, and to

verify my a priori classifications of simple or complex sites, I used random forest (RF)

regression analysis to identify which types of mesohabitat are important for predicting

overall habitat complexity for each sampling reach. As described previously, I used my

digitized habitat maps to calculate area (m2) of each mesohabitat type for every sampling

reach. I then calculated the relative proportions of each mesohabitat for every simple (n =

3) and complex (n = 9) sampling site. Thus, my predictors for the RF model were the

relative proportions of each mesohabitat including: runs, pools, riffles, backwaters,

forewaters, embayments, boulder bars, and side pools. Within the model, I ran 5000

bootstrap replicates (number of trees drawn) with replacement and the number of splits

was equal to two. I assessed model fit using the OOB metrics described previously.

73

Fish community diversity

Diversity indices provide information on species richness and associated evenness

of diversity for a given ecosystem (Kwak and Peterson 2007). In addition, diversity

indices are useful for assessing assemblage structure and the associated interactions with

the physical environment. For example, diversity can be compared between sites of

varying ecosystem integrity, such as habitat heterogeneity, water quality, or nonnative

species. Shannon’s diversity index (H’; Shannon and Weaver 1949) is one of many

common indices applied to examining freshwater fish assemblages. I used Shannon’s

index, because it is sensitive to the abundance of rare species in a community, and rare

species are of overall interest herein. When I compared diversity between simple and

complex sites, there were species that only occurred in complex sites and were typically

low in abundance. I calculated Shannon’s index for each sampling reach as:

𝐻′ = ∑ (𝑝𝑖)(𝑙𝑜𝑔𝑒𝑝𝑖)𝑠𝑖=1

Where, s is the number of species, pi is the proportion of the total sample represented by

the ith species. I used a one-way analysis of variance (ANOVA) to detect any significant

differences of diversity between the different levels of habitat complexity.

Stable isotope analysis

Food web structure of an ecosystem provides useful information on the different

interactions between species, trophic niches, and the ecosystem integrity of the

surrounding physical environment. I began by correcting the carbon and nitrogen isotopic

signatures for upper trophic consumers using a common herbivorous invertebrate (Order:

Ephemeroptera, functional feeding group: scraper) as a baseline to compare between my

74

sampling sites classified as either simple or complex, and alluvial or canyon. Thus, I used

the following equations to correct for basal resource variation using carbon and nitrogen

isotopes for each sample:

𝛿13𝐶𝑐𝑜𝑟 =𝛿13𝐶𝑖− 𝛿13𝐶𝑚𝑒𝑎𝑛𝑐ℎ

𝐶𝑅𝑐ℎ

Where, 𝛿13𝐶𝑖 is the carbon isotope signature of consumer i, 𝛿13𝐶𝑚𝑒𝑎𝑛𝑐ℎ is the mean

common herbivore carbon isotopic signature from the site where consumer i was

collected, and 𝐶𝑅𝑐ℎ is the carbon range of the common herbivores. I corrected nitrogen

values using:

𝛿15𝑁𝑐𝑜𝑟𝑗,𝑖 = 𝛿15𝑁𝑗,𝑖 − (𝛿15𝑁𝑐ℎ,𝑖 − 𝛿15𝑁𝑐ℎ,𝑚𝑖𝑛

)

Where, 𝛿15𝑁𝑗,𝑖 is the nitrogen signature of consumer j at site i, 𝛿15𝑁𝑐ℎ,𝑖 is the mean 15N

of common herbivores at site i, and 𝛿15𝑁𝑐ℎ,𝑚𝑖𝑛 is the minimum mean common

herbivore of all sites (from Cabana and Rasmussen 1996 and Walsworth et al. 2013). I

also calculated the carbon and nitrogen isotope ranges to supplement the corrected carbon

and nitrogen values and to provide information on the trophic length of the fish

community (NR) and the range of basal resources used by consumers (CR) using the

following equations:

𝑁𝑅 = 𝛿15𝑁𝑐𝑜𝑟𝑚𝑎𝑥 − 𝛿15𝑁𝑐𝑜𝑟𝑚𝑖𝑛

𝐶𝑅 = 𝛿13𝐶𝑐𝑜𝑟𝑚𝑎𝑥 − 𝛿13𝐶𝑐𝑜𝑟𝑚𝑖𝑛

Using my corrected carbon and nitrogen values, I then calculated the trophic

position for each consumer using the equation adapted from Post (2002):

75

𝑇𝑃𝑖 =𝛿15𝑁𝑐𝑜𝑟𝑖−𝛿15𝑁𝑐𝑜𝑟𝑐ℎ

3.4+ 2

Where, 𝑇𝑃𝑖 is the trophic position for each consumer i, 𝛿15𝑁𝑐𝑜𝑟𝑖 is the corrected nitrogen

signature of consumer i, and 𝛿15𝑁𝑐𝑜𝑟𝑐ℎ is the corrected nitrogen signature of common

herbivorous invertebrates. I assumed trophic fractionation of 15N to be 3.4‰ with every

increase in trophic position (Minagawa and Wada 1984) and the common herbivorous

invertebrates I assumed to have a trophic position of two. I used the calculated trophic

positions to determine food chain length of both simple and complex, and alluvial and

canyon habitat sampling sites.

I analyzed carbon and nitrogen isotopes of upper level consumers (all fish species

caught) using community multivariate ellipses, or SIBER (Stable Isotope Bayesian

Ellipses in R). SIBER metrics use a Bayesian framework to calculate convex hulls, or the

area of all species plotted in a C-N bivariate space (Layman et al. 2007; Jackson et al.

2011). Using SIBER, I estimated the 15N range (NR) and 13C range (CR), total area

(TA) of convex hull, and mean and standard deviation nearest neighbor distance (NND,

SDNND) for the entire fish communities within simple and complex, and alluvial and

canyon sampling sites. The NR represents the vertical structure of a food web and a

larger range suggests more trophic diversity. Estimating the CR range provides a measure

of the diversity for basal resources and an increased range indicates multiple types of

basal resources. The TA measures the total isotopic niche space occupied and provides a

measure of trophic diversity. Finally, the NND and the SDNND are measures of the

‘evenness’ of species within a food web structure, where smaller values for each metric

suggests increased trophic redundancy and a more even distribution of ecological feeding

76

niches.

I also used Bayesian mixing models of the carbon, nitrogen, and hydrogen stable

isotope data to predict patterns in resource partitioning or dietary preferences (Parnell et

al. 2010). Mixing models estimate the proportion of dietary items contributed to upper

trophic levels of interest. I used the Stable Isotope Mixing Models in R (simmr) package

(Parnell et al. 2010) to estimate the basal resources contributed to upper trophic

consumers (fish and aquatic invertebrates). For basal resources, I used aquatic (using the

common herbivore isotope signals), the combined three levels of terrestrial input: native

and nonnative vegetation, and periphyton. I first analyzed stable isotope food webs for

each type of habitat complexity: complex and simple, and alluvial or canyon. I then used

mixing models to observe any broad differences in the proportion of dietary items (pk)

between vegetation types across all habitat complexities. In addition, I analyzed separate

mixing models of each habitat complexity. For each model, I ran 500,000 iterations,

discarded 100,000 and retained one of every 10 iterations for analysis. I chose not to

include the residual error term for each simmr mixing model as there were sampling

periods where I only caught one individual (Parnell et al. 2010); however, I report the

95% confidence intervals for each pk mean estimate and assumed non-overlapping

intervals to indicate a more robust estimate of each pk. I performed and analyzed all

models using the free statistical software R (R Core Team 2017) and assumed models to

be significant at the P < 0.05 level.

77

Results

Habitat Complexity

The relative proportions of mesohabitat types between simple and complex sites

appeared to differ, where complex sampling sites tended to contain all types of

mesohabitat present in some proportion (e.g., backwaters, forewaters, embayments,

riffles, runs, pools, side pools, side channels, and boulder bars). In contrast, simple sites

contained only embayments, riffles, runs, pools, and side channels. Run mesohabitats

were 22% greater in simple sites compared to complex (Figure 3.2b).

Having determined the total area of simple and complex sites and the relative

proportions of mesohabitat for each habitat classification, I next used a random forest

classification using the relative proportion of mesohabitat types to predict habitat

heterogeneity for each sampling reach (complex n = 9, simple n = 3). Overall, the

validation of fit of the model was high for the PCC, specificity, sensitivity, kappa, and

AUC (PCC = 91.7%, specificity = 100%, sensitivity = 66.7%, kappa = 0.75, AUC =

0.89). I then examined the variable importance plot for the mesohabitat predictors, and it

appeared backwater mesohabitats were the most important variable for predicting the

level of complexity for a sampling site (Figure 3.4).

Fish Abundance and Diversity

I captured 16 species (total n for complex sites = 1464 individual fish; total n for

simple sites = 247) across all sampling sites (Appendix, Tables A2 and A3). I collected

only two nonnative species in either simple or complex sampling sites: channel catfish

(Ictalurus punctatus; total n = 57) and common carp (Cyprinus carpio; total n = 55), each

of which were in relatively low abundance compared to the rest of the catch. I captured

78

two generalist, saline-tolerant species in high abundance in each simple and complex site,

red shiner (Cyprinella lutrensis; total n = 529) and Tamaulipas shiner (Notropis braytoni;

total n = 542). In addition, I caught three Rio Grande silvery minnow (Hybognathus

amarus) in one complex sampling site.

Total catch per unit effort (CPUE; fish m-2) across all seasons was significantly

greater in complex habitats (CPUE = 0.52 ± 0.59 fish m2 [mean ± standard deviation];

repeated measures ANOVA, p < 0.001) and canyon (CPUE = 0.51 ± 0.59 fish m2;

repeated measures ANOVA, p < 0.001) sampling sites than in simple and alluvial sites

(Table 3.2, Figure 3.5a). March and May sampling events contained the greatest CPUE

(CPUEMarch = 0.74 ± 0.34 fish m2; CPUEMay = 1.69 ± 0.41 fish m2; p < 0.001 Two-way

ANOVA; Appendix, Figure A2). January, October, and June all demonstrated similar

CPUE and did not differ significantly.

Fish diversity appeared slightly higher in simple sites than complex sites (Figure

3.5b); however, the mean Shannon’s Index did not differ significantly between each type

of habitat complexity (H’simple = 1.42 ± 0.31, H’complex = 1.35 ± 0.52; repeated measures

ANOVA, p = 0.53). In contrast, canyon sites were significantly greater in fish diversity

than alluvial sites (H’canyon = 1.47 ± 0.38, H’alluvial = 0.88 ± 0.68; repeated measures

ANOVA, p < 0.05).

Abundance of Primary Productivity

As an indicator of primary productivity, I analyzed the chlorophyll a abundance

for all sampling sites across seasons and for three different types of algae: seston, soft

sediment, and periphyton, or attached algae (Figure 3.6). The concentration of Chl a

appeared to be greater in complex sampling sites compared to simple, but these

79

differences were not significant (repeated measures ANOVA p > 0.05). Similarly,

alluvial sampling sites appeared to contain greater concentrations of Chl a for both seston

and soft sediment algae abundance than canyon-bound sites. Periphyton abundance in

contrast, appeared to be greater in the canyon-bound sites. I observed no significant

differences in Chl a concentrations across seasons; however, summer sampling events

appeared to have greater concentrations of Chl a than either fall or winter sampling

events (Appendix, Figure A4).

Terrestrial and Aquatic Invertebrate Relative Abundance

I analyzed the relative abundances of terrestrial and aquatic invertebrates between

complex and/or simple sites, alluvial and/or canyon sites, vegetation types, and by season

(e.g., sampling months; Appendix, Tables A4 and A5). There were no significant

differences in relative abundances of terrestrial invertebrates between either simple,

complex, alluvial, or canyon-bound sampling sites, or between seasons and vegetation

types (e.g, native, nonnative, or open). However, the relative abundances of terrestrial

invertebrates between simple or complex sampling sites were significantly greater at the

Family classification-level (p-value < 0.01, repeated measures one-way ANOVA) than at

the Order classification-level. Terrestrial abundance at the family level, however, did not

appear to differ between nonnative and native vegetation types (p-value = 0.9868,

repeated measures one-way ANOVA). When I tested for seasonality (e.g., sampling

months), there appeared to be significant differences between the sampling months of

March and October (total relative abundanceMarch = 3.78, total relative abundanceOctober =

2.03; p-value < 0.01), but not May, June, or January. The terrestrial invertebrate relative

abundance was highest in March and June (total relative abundanceJune = 4.00), and

80

lowest in January (total relative abundanceJanuary = 0.036).

I analyzed functional feeding groups of the aquatic invertebrate catch between

complex and simple sites, alluvial and canyon sites, and season (sampling reach ncomplex =

9, aquatic invertebrate total ncomplex = 8604; sampling reach nsimple = 3, aquatic

invertebrate total nsimple = 947). The ‘scraper’ functional feeding group was 45.1% and

42.5% and significantly greater in relative abundance (p-value = 0.05) than all other

feeding groups in the complex and canyon sampling sites (mean scraper relative

abundancecomplex = 0.65; mean scraper relative abundancecanyon= 0.62; Figure 3.7a,b). The

relative abundance of the remaining four functional feeding groups was variable but

similar between simple and complex sites, and canyon and alluvial sites. I observed no

effect of season on the relative abundances of functional feeding groups (p–value = 0.23).

Stable isotope analyses

Overall, complex sampling sites appeared to contain a shorter food chain length

than simple sites (FCLcomplex = 3.99, FLCsimple = 4.81). Canyon-bound sites appeared to

contain a longer food chain length than alluvial sites (FCLcanyon = 4.81, FCLalluvial = 2.45).

I also compared the FCLs of species occurring in both complex and simple sampling

sites, and found species demonstrating a longer FCL in simple sites compared to complex

sites (Figure 3.8).

The trophic diversity (NR) appeared to be greater in complex (NRcomplex = 10.6)

and canyon sampling sites (NRcanyon = 9.25), and lower in simple (NRsimple = 5.49) and

alluvial sites (NRalluvial = 2.57; Table 3.3). Similarly, basal resources appeared to be more

diverse in complex and canyon sites (CRcomplex = 1.83, CRcanyon = 1.83) than in simple

and/or alluvial sites (CRsimple = 0.506, CRalluvial = 0.710). Total area (TA) of community

81

isotopic hulls appeared to be greater in complex and canyon sites compared to simple and

alluvial sites (TAcomplex = 9.32, TAcanyon = 8.49, TAsimple = 1.87, TAalluvial = 0.986). The

trophic redundancy (NDD) appeared to be lower in simple and alluvial sites than

complex and canyon sites (NDDsimple = 0.416, NDDalluvial = 0.446, NDDcomplex = 0.844,

NDDcanyon = 0.715). Finally, I compared the evenness of trophic niche diversity

(SDNDD), which is less influenced by sample size, and observed higher SDNDD values

for complex and canyon sites than simple and alluvial sites (SDNDDcomplex = 0.749,

SDNDDcanyon = 0.611, SDNDDsimple = 0.322, SDNDDalluvial = 0.0954).

Overall, the trophic niche breadth of fish, aquatic and terrestrial invertebrates was

wide and more complex in complex and canyon sites relative to simple and alluvial sites

as demonstrated by stable isotope bi-plots (Figure 3.9). Periphyton in particular was

much less in depleted in 13C in the complex and canyon sites relative to the simple and

alluvial. In addition, fish tended to hold a greater diversity of trophic positions and higher

trophic positions in the complex and canyon sites, but there was considerable overlap and

variability. There was only one fish, the Rio Grande silvery minnow, that appeared to

rely more on periphyton as a carbon source. All other fishes appeared to rely on a

combination of terrestrial and aquatic macroinvertebrates.

When I compared isotopic trophic niche space of fish and invertebrates across

vegetation type (native, nonnative, open), the fish community overall appeared to be

more dependent on terrestrial-based food resources than aquatic macroinvertebrates

(Figure 10). Aquatic macroinvertebrates (e.g., common herbivorous consumers),

appeared to be more depleted in the carbon isotope and may also be partially reliant upon

terrestrially-based sources of food. Fishes, however, demonstrated very little variability

82

in isotopic niche space across vegetation type, and did not appear to be trophically

influenced by nonnative vegetation.

Mixing models provide a more detailed and accurate description of the dietary

contribution of fishes as compared to general trophic positions. The mixing model results

demonstrated aquatic food resources potentially contributing to a higher proportion of

dietary (pk) items consumed by fish between complex, simple, and canyon sites (mean

complex paquatic = 0.453 ± 0.098, mean simple paquatic = 0.617 ± 0.099, mean canyon

paquatic = 0.569 ± 0.080) than terrestrial food resources (mean complex pterrestrial = 0.398 ±

0.116, mean simple pterrestrial = 0.211 ± 0.115, mean canyon pterrestrial = 0.237 ± 0.094;

Table 3.4, Figure 3.11). Terrestrial food resources appeared to contribute to a higher

proportion of dietary items consumed by fish in alluvial sites than aquatic food resources

(mean alluvial paquatic = 0.335 ± 0.077, mean alluvial pterrestrial = 0.636 ± 0.082; Figure

3.11). Periphyton did contribute to fish diets, but the proportion was small, albeit not

variable (Table 3.4).

A similar pattern appeared when I ran the mixing model using the hydrogen

isotope data, where complex, simple, and canyon sites (mean complex paquatic = 0.641 ±

0.046, mean simple paquatic = 0.780 ± 0.102, canyon paquatic = 0.711 ± 0.045) appeared to

contribute to a higher proportion of dietary items for fish than terrestrial food resources

(mean complex pterrestrial = 0.359 ± 0.046, mean simple pterrestrial = 0.220 ± 0.102, mean

canyon pterrestrial = 0.289 ± 0.045; Table 3.5, Figure 3.12). When I tested the mixing

model using aquatic invertebrates as the top consumers and terrestrial vegetation (native

and nonnative) and periphyton as the basal resources, native vegetation appeared to

contribute to a higher proportion of dietary items for aquatic invertebrates than nonnative

83

vegetation or periphyton for complex, simple, and canyon sites (Table 3.6; Figure 3.13).

Periphyton appeared to contribute to a higher proportion of dietary items for aquatic

invertebrates in alluvial sites compared to canyon, complex, or simple sampling sites.

Discussion

The highly regulated and dewatered Rio Grande along the US-Mexico border has

had a century’s long history of human use and modification (Scurlock 1998; Schmidt et

al. 2003). As a result, the modern flow regime is greatly reduced hydrologically, and the

river channel has become narrowed and simplified associated with the modified flow

regime and continued sediment supply (Dean and Schmidt 2011). The physical

degradation (e.g., low flows, increased sediment accumulation) of this region has likely

contributed to the reduction of many native riverine species through the loss of viable and

connected habitat, poor water quantity and quality, and nonnative plants and animals.

Though the physical changes (e.g., hydrology and geomorphology) have been well

documented (e.g., Schmidt et al. 2003; Dean and Schmidt 2011, 2013; Dean et al. 2016,

Blythe and Schmidt 2018), little is known or understood about the aquatic ecological

structure and function of the Rio Grande in the Big Bend region. In this study, I examined

how the modified flow and sediment regimes of the Rio Grande in the Big Bend region

have influenced the native fish community and aquatic food web structure.

Understanding if and how the native fish community and food resources have changed in

response to this physical degradation may provide unique insight on how to better

manage water resource use for the conservation and rehabilitation of native fish species

endemic to the Rio Grande in the Big Bend region.

I used a multi-faceted approach involving habitat availability analysis, fish and

84

macroinvertebrate community and abundance comparisons, and stable isotope techniques

to compare 12 sampling sites with different levels of habitat complexity a priori

classified as either complex or simple, and/or canyon and alluvial. Using this type of

comparison created a study design that used space for time substitution, wherein complex

habitats represent large scale conditions immediately after a reset flood and simple

habitats represent conditions long after a reset flood. I observed complex sites containing

multiple types of habitat in some proportion, generally having both deep and shallow

slow habitats (e.g., backwaters, pools), and deep and shallow fast habitats (e.g., runs,

riffles). In contrast, simple sites contained only runs, pools, riffles, and relatively few

embayments. The levels of habitat complexity I classified also appear to be best predicted

by shallow mesohabitats (e.g., backwaters). In addition, complex sites contained greater

abundances of fish, an important functional feeding group of macroinvertebrates (e.g.,

scrapers), and greater trophic diversity. CPUE for fish and the relative abundance of

scrapers were greater in complex and canyon sites than alluvial and/or simple sites.

Complex and canyon sites also appeared to support multiple ecological feeding niches

(e.g. trophic diversity) compared alluvial and/or simple sites. Similarly, fish consumers

may be more dependent on aquatic-based food resources than terrestrial-based food

resources. Collectively, these results indicate that the hydrologic and geomorphic

degradation that has occurred within the Rio Grande extend to fish habitat and food web

structure, confirming that complex and heterogeneous habitat is likely one of the factors

limiting the viability and persistence of the native fish community in relation between

simple and complex habitats changing with time.

The once wide, meandering, multi-threaded alluvial valleys of the Rio Grande in

85

the Big Bend region likely have been affected the most by the altered hydrologic and

sediment regimes, and have presently become narrowed, single-threaded channels. In

addition, the alluvial valleys are narrowing at a faster rate compared to canyon-bound

regions following a recent 2008 reset flood (Dean et al. 2016; T. Blythe, in progress).

These large scale geomorphic changes also appear to manifest in my study. Notably,

most of the complex sampling sites were located within the canyon-bound portions of the

river. Thus, the geomorphic processes of channel narrowing occurring in the alluvial

valleys may not be having as great of an impact on habitat heterogeneity in the canyon-

bound sites.

Habitat heterogeneity is important for sustaining diverse fish communities,

because it provides protection from predators, reduces competition, and provides

necessary resources or refugia for critical life stages (Gorman and Karr 1978; Nagayama

and Nakamura 2018). Habitat heterogeneity may also influence growth patterns for

fishes. The synergistic effects of adequate food resources and vegetated habitats allowed

for increased growth and subsequent survival from predators in an estuarine fish species

compared to habitat sites with no vegetation (Levin et al. 1997). Similarly, increased

turbidity and varying habitat substrates decreased the reaction time and subsequent

consumption rates for roach (Rutilus rutilus L.), a freshwater fish native to many

European waters, potentially affecting growth for roach within more turbid habitat

environments (Murray et al. 2016). Based purely on these results, I would expect there to

be lower abundance and diversity of fish and macroinvertebrates in sites of lower habitat

heterogeneity associated with the decreased availability of food, limited growth, and

predation.

86

Somewhat surprisingly, fish diversity was not significantly different between

complex and simple sampling sites. Diversity serves as an abundance-weighted measure

of presence or absence, and though the individuals present may be similar between

different levels of habitat heterogeneity, their abundances can be dramatically lower yet

still demonstrate a similar diversity score. In addition, observed abundance was much

greater in complex and canyon sites relative to simple and alluvial sites, and these sites

may serve as a potential source of native fishes to other habitat types. It would follow that

the same set of species may be colonizing sub-optimal habitats but in much lower

abundance, and indeed, CPUE was lower in simple and alluvial sites. This pattern has

also been shown to serve as a form of population persistence. When optimal spring pool

habitat was unavailable or disconnected, the Arkansas darter (Etheostoma cragini)

remained in sub-optimal pools until large spring floods re-scoured new pool habitat

allowing the darter to recolonize and increase in population once again (Labbe and

Fausch 2000). The combined effects of climate change and flow alterations were also

predicted to alter viable spawning and rearing habitat for West Balkan trout (Salmo

fariodes) in the southwestern Balkan river of Greece, ultimately decreasing population

abundances and distributions of this endemic trout in the Balkan river (Papadaki et al.

2016). Proximity to viable habitat therefore likely controls the persistence of native fish

populations and communities, thus demonstrating how fishes and other aquatic biota

require complimentary habitat types (pools, riffles, backwaters) that are accessible to

complete their life cycle and persist (White and Rahel 2008).

Riverine food web structure is inherently influenced by the physical template,

and decreases in the amount or availability of habitat associated with alterations to the

87

physical template ultimately affects the biotic community composition, food resources,

and other ecosystem functions (Ledger et al. 2013; Datry et al. 2014; McHugh et al.

2015). Thus, I would expect food web structure to become more simplified with

increased habitat loss and degraded habitat complexity. Streams within the Waimakariri

and Rakaia river basins in Canterbury, New Zealand experienced significantly low

minimum flows during many parts of the year, ultimately creating smaller habitat size

and availability in many regions within these basins (McHugh et al. 2015). The food web

structure responded in kind to decreased habitat size and availability by becoming shorter

in food chain length, smaller in trophic niche area, and smaller in species richness. In this

study, food chain length and community trophic structure were linked to habitat

heterogeneity and the physical template of the Rio Grande. For example, the higher

trophic range (TA and NR) for complex and canyon sites suggests these sites have

increased trophic diversity, or multiple species of upper-level consumers (e.g., primary-,

secondary-, tertiary-, or quaternary-consumers). Thus, reaches more complex in habitat

may allow for more variability in ecological feeding niches, because of the increased area

available to better partition necessary resources (Walsworth et al. 2013). It has also been

shown that larger ecosystems may contain longer food chain lengths and increased

trophic diversity because of their ability to contain larger areas available for resources,

and they may also facilitate more ecosystem processes because of this increased area

(Takimoto and Post 2013). Basal resource use also appeared to be more diverse within

complex and canyon sites, which follows that a ‘larger’ or more ‘complex’ ecosystems

may facilitate niche specialization (trophic diversity) and provide abundant and diverse

food resources.

88

Given that trophic and basal resource diversity were greater in complex and

canyon sites, it was surprising to observe a shorter food chain length in these types of

habitat compared to simple and alluvial sites. However, river food web structure and

function is incredibly heterogeneous and dynamic in nature because food webs in any

given location are affected by upstream processes, terrestrial influences, and seasons (in

temperate regions; Baxter et al. 2005). In addition, while their abundance appears to be

relatively low in comparison to other desert rivers similarly affected (e.g., Colorado

River, San Rafael River, and many others), nonnative fish species are present within the

Rio Grande. Nonnative species have been demonstrated to significantly alter the food

web structure by narrowing the isotopic niche space of native species and increasing the

food chain length, often with added predation and competition, in many freshwater

systems (Walworth et al. 2013; Sagouis et al. 2015). A nonnative, top-level predator

altered a previous complex food web structure by increasing food chain length and

omnivory in a small stream located in Sussex, England (Woodward and Hildrew 2001,

2002). Though this invasion did not necessarily reduce the native top-level predators

present immediately, competition ultimately narrowed the isotopic niche space. Thus, my

results suggest food web chain length may be more influenced by other biotic factors,

potentially including upstream processes and nonnative species, than by abiotic factors

such as habitat complexity at a reach scale.

Also contrary to what I expected, my results suggest trophic redundancy was

higher in simple and alluvial sites than complex and canyon sites. Thus, complex and

canyon sites appear to be more divergent in terms of trophic niche and perhaps contain a

less stable food web structure. However, complex and canyon sites still demonstrated a

89

more diverse, preferred food resource base. In addition, nonnative species have also been

shown to simplify food web structure, and increased trophic redundancy may be a result

of elevated competition or predation on native species by nonnative individuals (Sagouis

et al. 2015; Busst and Britton 2017). The food web structure of the San Rafael River, a

small desert river in southeastern Utah was impacted by the synergistic effects nonnative

species and habitat loss (Walsworth et al. 2013). The food chain length was shorter and

isotopic niche narrower in regions of the San Rafael with increased abundance of

nonnative fish species and degraded habitat. It is important to note, however, that

synthetic food web metrics such as trophic redundancy may not vary in association with

local study sites as used herein, but rather vary over a larger area of river and again are

affected by both upstream and downstream processes (Fausch et al. 2002).

River ecosystems strongly interact with their riparian zones through the

movement and exchange of both organic and inorganic materials (e.g., invertebrates, leaf

litter; Baxter et al. 2005). Thus, the establishment of nonnative riparian species may alter

the types of organic and inorganic materials input to the river, potentially changing the

forms of carbon and food resources available for uptake by aquatic consumers (Mineau et

al. 2011). Nonnative giant reed has become well established in the riparian zone of the

Rio Grande within the Big Bend region. I therefore examined potential alterations to the

food resources available associated with the increased nonnative riparian vegetation

abundance and habitat simplification. Overall, nearly all fish species analyzed were

remarkably similar and had low variability in isotopic niche space when compared across

vegetation type (native, nonnative, and open).

While upper trophic consumers (e.g., fish) did not appear to be trophically

90

influenced by nonnative vegetation, aquatic macroinvertebrates appeared more depleted

in the 13C element and could be partially reliant upon terrestrially-based sources of food

in canyon-bound reaches of the Rio Grande in the Big Bend region. Surprisingly, native

vegetation appeared to contribute a higher proportion to the dietary items consumed by

aquatic invertebrates than either nonnative vegetation or periphyton. This result is

contrary to my prediction of nonnative riparian vegetation having a greater influence on

the food web function of the Rio Grande. However, the likely greater abundance of both

native and nonnative riparian species due to availability of habitat to become well

established (e.g., inset floodplains) can still affect the food web structure. For example, if

the riparian leaf litter is poor in terms of food quality for macroinvertebrates (e.g.,

necessary nutrients), the aquatic macroinvertebrate community could experience shifts in

abundance and diversity. The nonnative Rhododendron ponticum plant was shown to

decrease relative abundance and diversity of the aquatic macroinvertebrate communities

within multiple streams in Western Ireland, ultimately limiting ecosystem functioning by

decreasing decomposition rates and suppressing algal production (Hladyz et al. 2011).

Similarly, nonnative Russian olive (Elaeagnus angustifolia) impaired ecosystem

functioning by creating a large influx of relatively recalcitrant nonnative leaf litter into a

small stream, ultimately decreasing the ability of consumers to process this large influx of

leaf litter (Mineau et al. 2012). Thus, the food web structure of the Rio Grande may be

more influenced by bottom-up effects, where the potentially threatened macroinvertebrate

community may become a limited resource for fish consumers.

Contrary to what I appeared to observe based on trophic position and isotopic

niche space, the mixing model demonstrated that aquatic macroinvertebrates provided a

91

greater contribution to the proportion of dietary items consumed by fish across most sites

except for alluvial sites. Similarly, the mixing model results comparing terrestrial

vegetation and periphyton illustrated periphyton appearing to contribute to a greater

proportion of dietary items to aquatic invertebrate consumers than either native or

nonnative vegetation. These results are concurrent with the functional feeding group

results, where scrapers were significantly greater in abundance than other functional

feeding groups present. Scraper abundance has been shown to be dependent on increased

concentrations of chlorophyll a on cobble substrates (e.g., periphyton; Hawkins and

Sedell 1981). In addition, scraper abundance can also be negatively impacted by

increased sedimentation rates. For example, in four low-gradient, alluvial streams located

in the Missouri Ozark Highlands, scraper abundance was decreased in the alluvial

streams with elevated sedimentation rates (Rabeni et al. 2005). Thus, a high density of

scrapers in complex, simple, and canyon sites may indicate increased abundance of

periphyton, or attached algae, and lower turbidity due to sedimentation within these sites.

Interestingly, based on my stable isotope bi-plots, the federally endangered Rio Grande

silvery minnow appeared to be more dependent on periphyton as a source of food and

was only captured in a complex site. Periphyton forms by integrating inorganic materials

from the water, and thus, its nutrient quality can be influenced by the overall water

quality (Sekar et al. 2002; Azim et al. 2005). Because of its high nutrient quality,

periphyton is often a preferred source of food for many fish species. Thus, having

observed a specialized fish species potentially consuming a greater abundance of

nutrient-rich periphyton in complex and canyon sites further contributes to the knowledge

of these sites containing higher quality food resources.

92

As the Rio Grande through the Big Bend region is hydrologically dynamic (e.g.,

tributary flash flood events, dam releases associated with monsoonal events), the River

Wave Concept (Humphries et al. 2014) may also explain, in part, the patterns I observed

regarding the dietary proportions I modeled. The River Wave Concept suggests the

timing and sources of either autochthonous or allochthonous production to a river’s food

web structure and function is largely dependent on the ‘river wave’, or the flood pulses of

a river. This concept was applied to three major rivers in Texas of varying stream flow

and sediment regimes: the Brazos, Guadalupe, and Neches rivers (Roach and Winemiller

2015). The physical characteristics (e.g., sediment supply) and the timing and magnitude

of flood pulses predicted when each river was more dependent on either allochthonous or

autochthonous sources of energy. These patterns may also hold for the Rio Grande as the

hydrologic and associated sediment environments during my sampling events may have

allowed the fish and macroinvertebrates communities to consume more autochthonous

sources of energy. In addition, I assumed the isotopic ratios of consumer tissues was

reflective of the resource availability during the time of sampling, and isotopes can reflect

a longer time period of consumption (e.g., up to one month for fishes; Vander Zanden et

al. 2001; Post 2002).

Implications My results have important implications for native fish conservation and

understanding food web structure in exotic rivers that cross desert lands, especially those

exotic rivers that have been significantly dewatered. Specifically, I have provided an

example of how food web structure and native fish communities in such rivers are

inextricably linked with their local physical environment, which is driven by both

93

hydrologic and sediment supply processes. By linking the physical processes with the

ecology of this system, I provide additional impetus to find innovative solutions for water

management that can maintain channel complexity and perhaps rehabilitate and restore

the native aquatic biodiversity for this unique river. I have demonstrated that availability

of complex habitat is likely a limiting factor for native fishes, and the availability of

complex habitat also affects the structure of the food web, potentially including food

availability for fishes. Increasing flood magnitude and duration to levels that evacuate

sediment and rewiden the channel, reconnecting the river to its floodplain, and improving

water quality (not addressed here) may allow populations of sensitive native fish species,

such as the federally endangered Rio Grande silvery minnow, to be restored and

maintained within this region.

Acknowledgments

Funding for this study was provided by the U.S. Geological Survey, South Central

Climate Science Center, the National Park Service, the Ecology Center at Utah State

University, and by the U.S. Geological Survey, Utah Cooperative Fish and Wildlife

Research Unit (in-kind). I would like to thank G. Thiede, B. Laub, and members of both

the Fish Ecology Lab and Lake Ecology lab their generous help in the field and overall

guidance with my project. In addition, I would like to thank J. Brahney for their guidance

and reviews throughout my project. Any use of trade, firm, or product names is for

descriptive purposes only and does not imply endorsement by the United States

Government.

94

Literature Cited

Allred TM, Schmidt JC. 1999. Channel narrowing by vertical accretion along the Green

River near Green River, Utah. Geological Society of America Bulletin, 111(12),

1757-1772.

Archdeacon TP. 2016. Reduction in spring flow threatens Rio Grande Silvery Minnow:

trends in abundance during river intermittency. Transactions of the American

Fisheries Society, 145(4), 754-765.

Azim MA, Verdegem MCJ, van Dam AA, Beveridge MCM. 2005. Periphyton. Ecology,

exploitation, and management. CABI Publishing, Oxfordshire, UK. 319 p.

Bain MB, Finn JT, Booke HE. 1988. Streamflow regulation and fish community

structure. Ecology, 69(2), 382-392.

Bennett JB. 2016. Big Bend National Park. Personal communication.

Blythe TL, Schmidt JC. 2018. Estimating the natural flow regime of rivers with long‐

standing development: the northern branch of the Rio Grande. Water Resources

Research, 54(2), 1212-1236.

Brandt SA. 2000. Classification of geomorphological effects downstream of dams.

Catena, 40(4), 375-401.

Bunn SE, Arthington AH. 2002. Basic principles and ecological consequences of altered

flow regimes for aquatic biodiversity. Environmental management, 30(4), 492-

507.

Busst G, Britton JR. 2017. Comparative trophic impacts of two globally invasive cyprinid

fishes reveal species‐specific invasion consequences for a threatened native fish.

Freshwater Biology, 62(9), 1587-1595.

Cabana G, Rasmussen JB. 1996. Comparison of aquatic food chains using nitrogen

isotopes. Proceedings of the National Academy of Sciences, 93(20), 10844-10847.

Commission for Environmental Cooperation (CEC). 2014. Conservation assessment for

the Big Bend-Rio Bravo region: a binational collaborative approach to

conservation. Montreal 889 (QC), Commission for Environmental Cooperation,

106 p.

Cross WF, Baxter CV, Donner KC, Rosi-Marshall EJ, Kennedy TA, Hall RO, Wellard

Kelly HA, Rogers RS. 2011. Ecosystem ecology meets adaptive management:

food web response to a controlled flood on the Colorado River, Glen Canyon.

Ecological Applications, 21(6), 2016-2033.

95

Cross WF, Baxter CV, Rosi-Marshall EJ, Hall RO, Kennedy TA, Donner KC, Wellard

Kelly HA, Seegert SEZ, Behn KE, Yard MD. 2013. Food web dynamics in a

large river discontinuum. Ecological Monographs, 83(3), 311-337.

Cummins KW, Klug MJ. 1979. Feeding ecology of stream invertebrates. Annual review

of ecology and systematics, 10(1), 147-172.

Cutler DR, Edwards TC, Beard KH, Cutler A, Hess KT, Gibson J, Lawler JJ. 2007.

Random forests for classification in ecology. Ecology, 88(11), 2783-2792.

Datry T, Larned ST, Tockner K. 2014. Intermittent rivers: a challenge for freshwater

ecology. BioScience, 64(3), 229-235.

Daufresne M, Boët P. 2007. Climate change impacts on structure and diversity of fish

communities in rivers. Global Change Biology, 13(12), 2467-2478.

Dean DJ, Scott M, Shafroth P, Schmidt JC. 2011. Stratigraphic, sedimentologic, and

dendrogeomorphic analyses of rapid floodplain formation along the Rio Grande in

Big Bend National Park, Texas. GSA Bulletin, 123(9-10), 1908-1925.

Dean DJ, Schmidt JC. 2011. The role of feedback mechanisms in historic channel

changes of the lower Rio Grande in the Big Bend region. Geomorphology, 126(3-

4), 333-349.

Dean DJ, Schmidt JC. 2013. The geomorphic effectiveness of a large flood on the Rio

Grande in the Big Bend region: Insights on geomorphic controls and post-flood

geomorphic response. Geomorphology, 201, 183-198

.

Dean DJ, Topping DJ, Schmidt JC, Griffiths RE, Sabol TA. 2016. Sediment supply

versus local hydraulic controls on sediment transport and storage in a river with

large sediment loads. Journal of Geophysical Research: Earth Surface, 121(1),

82-110.

Dean DJ. 2016. Grand Canyon Monitoring and Research Center. Personal

communication.

Dudley RK, Platania SP. 1997. Habitat use of Rio Grande silvery minnow: Division of

Fishes, Museum of Southwestern Biology, Department of Biology, University of

New Mexico.

Dudley RK, Platania SP. 2007. Flow regulation and fragmentation imperil pelagic‐

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

Everitt BL. 1998. Chronology of the spread of tamarisk in the central Rio Grande.

Wetlands, 18(4), 658-668.

96

Fausch KD, Torgersen CE, Baxter CV, Li HW. 2002. Landscapes to riverscapes:

bridging the gap between research and conservation of stream fishes: A

continuous view of the river is needed to understand how processes interacting

among scales set the context for stream fishes and their habitat. AIBS Bulletin,

52(6), 483-498.

Finlay JC. 2001. Stable isotope carbon ratios of river biota: implications for energy flow

in lotic food webs. Ecology, 82(4), 1052-1064.

Gehrke P, Brown P, Schiller C, Moffatt D, Bruce A. 1995. River regulation and fish

communities in the Murray Darling river system, Australia. River Research and

Applications, 11(3‐4), 363-375.

Graf WL. 2006. Downstream hydrologic and geomorphic effects of large dams on

American rivers. Geomorphology, 79(3-4), 336-360.

Grey J. 2006. The use of stable isotope analyses in freshwater ecology: current

awareness. Polish Journal of Ecology 54: 563-584.

Gorman OT, Karr JR. 1978. Habitat structure and stream fish communities. Ecology,

59(3), 507-515.

Hauer FR, Lorang MS. 2004. River regulation, decline of ecological resources, and

potential for restoration in a semi-arid lands river in the western USA. Aquatic

Sciences, 66(4), 388-401.

Hawkins CP, Norris RH, Hogue JN, Feminella JW. 2000. Development and evaluation of

predictive models for measuring the biological integrity of streams. Ecological

Applications, 10(5), 1456-1477.

Hawkins CP, Sedell JR. 1981. Longitudinal and seasonal changes in functional

organization of macroinvertebrate communities in four Oregon streams. Ecology,

62(2), 387-397.

Hayden B, Harrod C, Sonninen E, Kahilainen KK. 2015. Seasonal depletion of resources

intensifies trophic interactions in subarctic freshwater fish communities.

Freshwater Biology, 60(5), 1000-1015.

Hereford R. 1984. Climate and ephemeral-stream processes: Twentieth-century

geomorphology and alluvial stratigraphy of the Little Colorado River, Arizona.

Geological Society of America Bulletin, 95(6), 654-668.

Hladyz S, Åbjörnsson K, Giller PS, Woodward G. 2011. Impacts of an aggressive

riparian invader on community structure and ecosystem functioning in stream

food webs. Journal of Applied Ecology, 48(2), 443-452.

97

Humphries P, Keckeis H, Finlayson B. 2014. The river wave concept: integrating river

ecosystem models. BioScience, 64(10), 870-882.

Jackson AL, Inger R, Parnell AC, Bearhop S. 2011. Comparing isotopic niche widths

among and within communities: SIBER–Stable Isotope Bayesian Ellipses in R.

Journal of Animal Ecology, 80(3), 595-602.

Jackson DA, Peres-Neto PR, Olden JD. 2001. What controls who is where in freshwater

fish communities the roles of biotic, abiotic, and spatial factors. Canadian journal

of fisheries and aquatic sciences, 58(1), 157-170

Junk WJ, Bayley PB, Sparks RE. 1989. The flood pulse concept in river-floodplain

systems. Canadian Special Publication of Fisheries and Aquatic Sciences 106:

110-127.

.

Kalb BW, Huntsman BM, Caldwell CA, Bozek MA. 2018. A mechanistic assessment of

seasonal microhabitat selection by drift-feeding rainbow trout Oncorhynchus

mykiss in a Southwestern headwater stream. Environmental Biology of Fishes,

101(2), 257-273.

Kennedy TA, Finlay JC, Hobbie SE. 2005. Eradication of invasive Tamarix ramosissima

along a desert stream increases native fish density. Ecological Applications,

15(6), 2072-2083.

Kennedy TA, Hobbie SE. 2004. Saltcedar (Tamarix ramosissima) invasion alters organic

matter dynamics in a desert stream. Freshwater Biology, 49(1), 65-76.

Kennedy TA, Muehlbauer JD, Yackulic CB, Lytle DA, Miller SW, Dibble KL,

Kortenhoeven EW, Metcalfe AN, Baxter CV. 2016. Flow management for

hydropower extirpates aquatic insects, undermining river food webs. BioScience,

66(7), 561-575.

Kwak TJ, Peterson JT. 2007. Community indices, parameters, and comparisons. Analysis

and interpretation of freshwater fisheries data. American Fisheries Society,

Bethesda, Maryland, 677-763.

Labbe TR, Fausch KD. 2000. Dynamics of intermittent stream habitat regulate

persistence of a threatened fish at multiple scales. Ecological Applications, 10(6),

1774-1791.

Layman CA, Arrington DA, Montaña CG, Post DM. 2007. Can stable isotope ratios

provide for community‐wide measures of trophic structure? Ecology, 88(1), 42-

48.

98

Layman CA, Arrington DA, Montaña CG, Post DM. 2007. Can stable isotope ratios

provide for community‐wide measures of trophic structure? Ecology, 88(1), 42-

48.

Ledger ME, Brown LE, Edwards FK, Milner AM, Woodward G. 2013. Drought alters

the structure and functioning of complex food webs. Nature Climate Change,

3(3), 223.

Levin P, Petrik R, Malone J. 1997. Interactive effects of habitat selection, food supply

and predation on recruitment of an estuarine fish. Oecologia, 112(1), 55-63.

Marcarelli AM, Baxter CV, Mineau MM, Hall RO. 2011. Quantity and quality: unifying

food web and ecosystem perspectives on the role of resource subsidies in

freshwaters. Ecology, 92(6), 1215-1225.

McDonald ME, Hershey AE, O'Brien WJ. 1992. Cost of predation avoidance in young-

of-year lake trout (Salvelinus namaycush): Growth differential in sub-optimal

habitats. Hydrobiologia, 240(1-3), 213-218.

McHugh PA, Thompson RM, Greig HS, Warburton HJ, McIntosh AR. 2015. Habitat size

influences food web structure in drying streams. Ecography, 38(7), 700-712.

Menge BA, Sutherland JP. 1987. Community regulation: variation in disturbance,

competition, and predation in relation to environmental stress and recruitment.

The American Naturalist, 130(5), 730-757.

Minagawa M, Wada E. 1984. Stepwise enrichment of 15N along food chains: further

evidence and the relation between δ15N and animal age. Geochimica et

cosmochimica acta, 48(5), 1135-1140.

Mineau MM, Baxter CV, Marcarelli AM. 2011. A non-native riparian tree (Elaeagnus

angustifolia) changes nutrient dynamics in streams. Ecosystems, 14(3), 353-365

.

Mineau MM, Baxter CV, Marcarelli AM, Minshall GW. 2012. An invasive riparian tree

reduces stream ecosystem efficiency via a recalcitrant organic matter subsidy.

Ecology, 93(7), 1501-1508.

Miyazono S, Patiño R, Taylor CM. 2015. Desertification, salinization, and biotic

homogenization in a dryland river ecosystem. Science of the Total Environment,

511, 444-453.

Miyazono S, Taylor CM. 2013. Effects of habitat size and isolation on species

immigration–extinction dynamics and community nestedness in a desert river

system. Freshwater Biology, 58(7), 1303-1312.

99

Murray GP, Stillman RA, Britton JR. 2016. Habitat complexity and food item size

modify the foraging behaviour of a freshwater fish. Hydrobiologia, 766(1), 321-

332.

Nagayama S, Nakamura F. 2018. The significance of meandering channel to habitat

diversity and fish assemblage: a case study in the Shibetsu River, northern Japan.

Limnology, 19(1), 7-20.

Medley C, Shirey PD. 2013. Review and reinterpretation of Rio Grande silvery minnow

reproductive ecology using egg biology, life history, hydrology, and

geomorphology information. Ecohydrology, 6(3), 491-505.

Papadaki C, Soulis K, Muñoz-Mas R, Martinez-Capel F, Zogaris S, Ntoanidis L,

Dimitriou E. 2016. Potential impacts of climate change on flow regime and fish

habitat in mountain rivers of the south-western Balkans. Science of the Total

Environment, 540, 418-428.

Parnell, AC, Inger R, Bearhop S, Jackson AL. 2010. Source partitioning using stable

isotopes: coping with too much variation. PloS one, 5(3), e9672.

Patiño-Gomez C, McKinney DC, Maidment DR. 2007. Sharing water resources data in

the binational Rio Grande/Bravo Basin. Journal of Water Resources Planning and

Management, 133(5), 416-426.

Phillips DL, Inger R, Bearhop S, Jackson AL, Moore JW, Parnell AC, Semmens BX,

Ward EJ. 2014. Best practices for use of stable isotope mixing models in food-

web studies. Canadian Journal of Zoology, 92(10), 823-835.

Poff NL, Allan JD, Bain MB, Karr JR, Prestegaard KL, Richter BD, Sparks RE,

Stromberg JC. 1997. The natural flow regime. BioScience, 47(11), 769-784.

Post DM. 2002. Using stable isotopes to estimate trophic position: models, methods, and

assumptions. Ecology, 83(3), 703-718.

PRISM Climate Group, Oregon State University, http://prism.oregonstate.edu, created 4

Feb 2004. Date accessed: October 20, 2017.

R Development Core Team. 2008. R: A language and environment for statistical

computing. Vienna, Austria: R Foundation for Statistical Computing. ISBN: 3–

900051-07-0. Retrieved from http://www.R-project.org.

Rabení CF, Doisy KE, Zweig LD. 2005. Stream invertebrate community functional

responses to deposited sediment. Aquatic Sciences, 67(4), 395-402.

Raines TH, Turco MJ, Connor PJ, Bennett JB. 2012. Streamflow gains and losses and

selected water-quality observations in five subreaches of the Rio Grande/Rio

100

Bravo del Norte from near Presidio to Langtry, Texas, Big Bend area, United

States and Mexico, 2006. U.S. Geological Survey, Scientific Investigations

Report 2012-5125. 30 p.

Roach KA, Winemiller KO. 2015. Hydrologic regime and turbidity influence entrance of

terrestrial material into river food webs. Canadian journal of fisheries and aquatic

sciences, 72(7), 1099-1112.

Sagouis A, Cucherousset J, Villéger S, Santoul F, Boulêtreau S. 2015. Non‐native species

modify the isotopic structure of freshwater fish communities across the globe.

Ecography, 38(10), 979-985.

Sá-Oliveira JC, Isaac VJ, Ferrari SF. 2015. Fish community structure as an indicator of

the long-term effects of the damming of an Amazonian river. Environmental

Biology of Fishes, 98(1), 273-286.

Schmidt JC, Wilcock PR. 2008. Metrics for assessing the downstream effects of dams.

Water Resources Research, 44(4).

Schmidt JC, Everitt BL, Richards GA. 2003. Hydrology and geomorphology of the Rio

Grande and implications for river rehabilitation. Pages 25-45 in G. P. Garrett and

N. L. Allan, editors. Aquatic Fauna of the Northern Chihuahuan Desert,

Contributed Papers from a Special Session within the Thirty-Third Annual

Symposium of the Desert Fishes Council. Special Publications of the Museum of

Texas Tech University, Number 46.

Scurlock D. 1998. An environmental history of the middle Rio Grande basin. Rocky

Mountain Forest and Range Experiment Station, United States Department of

Agriculture Forest Service, Fort Collins, Colorado.

Sekar R, Nair K, Rao V, Venugopalan V. 2002. Nutrient dynamics and successional

changes in a lentic freshwater biofilm. Freshwater Biology, 47(10), 1893-1907.

Shannon C, Weaver W. 1949. The mathematical study of communication. Urbana, IL:

University of Illinois Press.

Strohm D. 2017. Utah Division of Wildlife Resources, Fisheries Experiment Station.

Personal communication.

Takimoto G, Post DM. 2013. Environmental determinants of food-chain length: a meta-

analysis. Ecological research, 28(5), 675-681.

Thorp JH, Delong MD. 1994. The riverine productivity model: an heuristic view of

carbon sources and organic processing in large river ecosystems. Oikos, 305-308.

101

Townsend CR, Winfield IJ. 1985. The application of optimal foraging theory to feeding

behaviour in fish. In Fish energetics (pp. 67-98): Springer.

Vander Zanden MJ, Cabana G, Rasmussen JB. 1997. Comparing trophic position of

freshwater fish calculated using stable nitrogen isotope ratios (δ15N) and

literature dietary data. Canadian journal of fisheries and aquatic sciences, 54(5),

1142-1158.

Vander Zanden MJ, Shuter BJ, Lester N, Rasmussen JB. 1999. Patterns of food chain

length in lakes: a stable isotope study. The American Naturalist, 154(4), 406-416.

Vannote RL, Minshall GW, Cummins KW, Sedell JR, Cushing CE. 1980. The river

continuum concept. Canadian journal of fisheries and aquatic sciences, 37(1),

130-137.

Wallace JB, Webster JR. 1996. The role of macroinvertebrates in stream ecosystem

function. Annual review of entomology, 41(1), 115-139.

Walsworth TE, Budy P, Thiede GP. 2013. Longer food chains and crowded niche space:

effects of multiple invaders on desert stream food web structure. Ecology of

Freshwater Fish, 22(3), 439-452.

Ward J, Stanford J. 1983. Intermediate-disturbance hypothesis: an explanation for biotic

diversity patterns in lotic ecosystems. Dynamics of Lotic Systems, Ann Arbor

Science, Ann Arbor MI. 1983. 347-356 p.

Welschmeyer NA. 1994. Fluorometric analysis of chlorophyll a in the presence of

chlorophyll b and pheopigments. Limnology and Oceanography, 39(8), 1985-

1992.

White SM, Rahel FJ. 2008. Complementation of habitats for Bonneville cutthroat trout in

watersheds influenced by beavers, livestock, and drought. Transactions of the

American Fisheries Society, 137(3), 881-894.

Woodward G, Hildrew AG. 2001. Invasion of a stream food web by a new top predator.

Journal of Animal Ecology, 70(2), 273-288.

Woodward G, Hildrew AG. 2002. Food web structure in riverine landscapes. Freshwater

Biology, 47(4), 777-798.

Zanden M, Rasmussen JB. 2001. Variation in δ15N and δ13C trophic fractionation:

implications for aquatic food web studies. Limnology and Oceanography, 46(8),

2061-2066.

102

Tables and Figures

Table 3.1. Mean, minimum, and maximum instantaneous discharge (in cubic meters per

second, m3 s-1) of the Rio Grande at the US Geological Survey (USGS) Castolon

(08374550) and Rio Grande Village (08375300) gaging stations and the International

Boundary and Water Commission (IBWC) Dryden Crossing (08376400) gaging station

for each field sampling date over the years 2016 to 2017 within the Big Bend region.

Sampling Dates

Mean Discharge

(m3 s -1)

Minimum

Discharge (m3 s -1)

Maximum

Discharge (m3 s -1)

Castolon Gage

March 1-11, 2016 1.61 1.37 2.11

May 26-30, 2016 2.38 2.02 2.71

October 17-21, 2016 23.4 22.4 23.8

January 11- 20, 2017 17.9 15.7 20.3

Rio Grande Village

March 1-11, 2016 2.97 2.40 3.46

May 26-30, 2016 4.73 0.977 28.1

October 17-21, 2016 23.4 22.4 23.8

January 11- 20, 2017 22.1 16.1 61.2

Dryden Crossing

June 24-30, 2017 8.11 7.06 10.9

Table 3.2. Summary statistics (average and standard deviation) of the catch per unit

effort (fish m-1) for the four different sites of habitat heterogeneity sampled along the Rio

Grande in the Big Bend region.

Complexity

Average

CPUE (fish m-1)

Standard Deviation

CPUE (fish m-1)

complex 0.519 0.576

simple 0.400 0.504

alluvial 0.392 0.056

canyon 0.510 0.590

103

Table 3.3. Layman community metrics (Layman et al. 2007) from the stable isotopic

analysis of the fish community for the four different sites of habitat heterogeneity within

the Rio Grande in the Big Bend region (NR = nitrogen isotopic range, CR = carbon

isotopic range, TA = total area of community hull calculated, MNND = mean nearest

neighbor distance, SDNND = standard deviation nearest neighbor distance).

Complexity NR CR TA MNND SDNND

complex 10.6 1.83 9.32 0.845 0.749

simple 5.49 0.586 1.87 0.416 0.322

alluvial 2.57 0.710 0.986 0.446 0.095

canyon 9.25 1.83 8.49 0.715 0.611

Table 3.4. Mean and standard deviation (SD) for the proportion of dietary items (aquatic,

terrestrial, or periphyton) consumed by fish calculated for each of the four different sites

of habitat heterogeneity using a three-source stable isotope mixing model (simmr).

Complexity Mean Paquatic ± SD Mean Pterrestrial ± SD Mean Pperiphyton ± SD

Complex 0.453 ± 0.098 0.398 ± 0.116 0.149 ± 0.024

Simple 0.617 ± 0.099 0.211 ± 0.115 0.172 ± 0.026

Alluvial 0.335 ± 0.077 0.636 ± 0.082 0.029 ± 0.014

Canyon 0.569 ± 0.080 0.237 ± 0.094 0.194 ± 0.020

Table 3.5. Mean and standard deviation (SD) for the proportion of dietary items (aquatic

or terrestrial) consumed by fish calculated for each of the four different sites of habitat

heterogeneity using the hydrogen isotope and a two-source mixing model (simmr).

Complexity Mean Paquatic ± SD Mean Pterrestrial ± SD

Complex 0.641 ± 0.046 0.359 ± 0.046

Simple 0.780 ± 0.102 0.220 ± 0.102

Alluvial 0.423 ± 0.068 0.577 ± 0.068

Canyon 0.711 ± 0.045 0.289 ± 0.045

Table 3.6. Mean and standard deviation (SD) for the proportion of dietary items (aquatic,

terrestrial, or periphyton) consumed by aquatic invertebrates calculated for each of the

four different sites of habitat heterogeneity using a three-source mixing model (simmr).

Complexity Mean Pnative ± SD Mean Pnonnative ± SD Mean Pperiphyton ± SD

Complex 0.304 ± 0.257 0.292 ± 0.114 0.404 ± 0.170

Simple 0.301 ± 0.244 0.299 ± 0.161 0.399 ± 0.206

Alluvial 0.309 ± 0.259 0.291 ± 0.128 0.400 ± 0.182

Canyon 0.445 ± 0.292 0.243 ± 0.122 0.312 ± 0.197

104

Figure 3.1. Study area map of the Rio Grande through the Big Bend region. Sampling

sites within Big Bend National Park are designated by the park boundary, those outside

the boundary is the Lower Canyons portion of the Big Bend region.

105

Figure 3.2. (a) Conceptual diagram illustrating the differences between simple and

complex sampling sites and the mesohabitats used to infer the level of complexity. (b)

Relative proportion of the total area (m2) summed for each mesohabitat type between all

simple and complex sampling sites.

106

Figure 3.3. Average depth and velocity for all mesohabitats mapped within two different

discharge ranges that occurred over the entire period of sampling. Box represents the

interquartile range, line indicates the median, whiskers are 1.5 the interquartile range, and

black circles indicate outliers.

Figure 3.4. Variable importance plot for the random forest classification model. The

increased ‘mean decreased accuracy’ values indicate the mesohabitats most important for

predicting the level of complexity.

107

Figure 3.5. (a) The average catch per unit effort for the four different types of habitat

sites sampled in the Rio Grande within the Big Bend region. The solid line denotes the

median, box indicates the 25th and 75th percentiles, and ‘whiskers’ represent 1.5 of the

interquartile range. (b) Shannon’s diversity index for the four different types of habitat

sites sampled in the Rio Grande within the Big Bend region. Box and whisker data

structure described previously in other figures.

108

Figure 3.6. Chlorophyll a abundance estimates for seston, soft sediment, and periphyton

alga across the four different levels of habitat heterogeneity sampled along the Rio

Grande in the Big Bend region. Top panels represent Chlorophyll a concentrations

between complex and simple sampling sites, and bottom panels represent Chlorophyll a

concentrations between alluvial and canyon sampling sites. Box and whisker data

structure described previously in other figures.

109

Figure 3.7. (a) Average relative abundance estimates for each functional feeding group

across complex and simple sampling sites. (b) Average relative abundance estimates for

each functional feeding group across alluvial and canyon sampling sites. The mean is

represented by the symbol ‘x’, the median is the solid line within the box, the box

represents the 25th and 75th interquartile range, and the ‘whiskers’ denote 1.5 of the

interquartile range.

110

Figure 3.8. Calculated trophic positions (food chain lengths) of fish species occurring in

both simple and complex sampling sites.

111

Figure 3.9. Food web structure for the four different habitat sites in the Rio Grande in the

Big Bend region. The food web structure was determined by calculating the mean stable

isotope signatures of individual fish species, common herbivorous macroinvertebrates

(aquatic), terrestrial invertebrates (terrestrial), and attached algae (periphyton). Bars for

each point represent the 95% confidence intervals for each mean organism calculated

(e.g., fish species, aquatic invertebrates, terrestrial invertebrates, and periphyton).

112

Figure 3.10. Stable isotope ellipses grouped and calculated for the entire fish community,

terrestrial invertebrates, and aquatic invertebrates across vegetation types (e.g., native,

nonnative, and open) in the Rio Grande in the Big Bend region. Border of ellipses (area)

are portrayed as the 95% confidence interval for each organism type (e.g., fish, aquatic

macroinvertebrates, and terrestrial invertebrates).

113

Figure 3.11. Carbon and Nitrogen stable isotope mixing model results (using the simmr

package in R) for the four different types of habitat heterogeneity sites, where the mean

proportion of dietary items (Proportion) contributed by either aquatic (common

herbivorous invertebrates), terrestrial (terrestrial invertebrates), or attached algae

(periphyton) to fish consumers was estimated using a Bayesian framework. Box and

whisker data structure described previously in other figures.

114

Figure 3.12. Hydrogen stable isotope mixing model results (using the simmr package in

R) for the four different types of habitat heterogeneity sites, where the mean proportion

of dietary items (Proportion) contributed by either aquatic (common herbivorous

invertebrates) or terrestrial (terrestrial invertebrates) to fish consumers was estimated

using a Bayesian framework. Box and whisker data structure described previously in

other figures.

115

Figure 3.13. Carbon and Nitrogen stable isotope mixing model results (using the simmr

package in R) for the four different types of habitat heterogeneity sites, where the mean

proportion of dietary items (Proportion) contributed by either native (native terrestrial

leaf litter), nonnative (nonnative terrestrial leaf litter), or attached algae (periphyton) to

aquatic invertebrate consumers was estimated using a Bayesian framework. Box and

whisker data structure described previously in other figures.

116

CHAPTER 4

CONCLUSIONS

The Rio Grande varies along its longitudinal length both in its physical and

ecological structure, as well as the anthropogenic impacts that affect the Rio Grande.

Thus, the Rio Grande is a unique river in which to understand how the interactions

between the physical environment and human impacts affect the aquatic ecological

integrity and associated native fish diversity of a large river system. Collectively, the

results of my project indicate that the hydrologic and geomorphic degradation that has

occurred within the Rio Grande extend to fish habitat and food web structure, confirming

complex and heterogeneous habitat is likely one of the factors limiting the viability and

persistence of the native fish community of the Rio Grande in far west Texas.

For Chapter II, we compared two different segments with different problems,

allowing us to explore the ways in which each region functioned and affected the native

fish diversity for the Rio Grande in each segment. Both the Big Bend and Forgotten

Reach native fish richness was significantly influenced by active channel width; however,

each model was influenced by differing hydrologic variables. For the Big Bend,

baseflows and low flow events may affect native fish richness to a greater degree than

within the Forgotten Reach. The Forgotten Reach native fish richness appeared to be

largely influenced by poor water quality and flash flood events, likely because any

change in flow magnitude has a greater impact within the Forgotten Reach due to its

significantly lower base flows. Given the Big Bend region is a gaining reach due to the

Rio Conchos and the large Edwards-Trinity groundwater aquifer, it follows that the Big

Bend region native fish richness is not as affected by poor water quality conditions

117

compared to the Forgotten Reach. Overall, our modeling results from this chapter

allowed us to test the links between the physical components of a river ecosystem with

the ecological components, and identify important limiting factors for native fish richness

associated with the degree of degradation within each segment of the Rio Grande.

The Big Bend region is diverse in terms of river landscapes containing simple

reaches, whether they are naturally simple or simple because of channel narrowing, and

complex reaches. For Chapter III, I found physical habitat differs between these types of

reaches such that complex reaches have a greater diversity of mesohabitats compared to

simple (specifically backwaters). Subsequently, the fauna appears to be influenced by the

degree of habitat heterogeneity within the Rio Grande in the Big Bend region. Fish

abundance and diversity differs between canyon and alluvial sites, but not between

simple and complex sites. Aquatic food resources (e.g., aquatic invertebrates and

periphyton) also appeared to be influenced by the level of habitat heterogeneity. The

‘scraper’ functional feeding group for aquatic invertebrates appeared to be greater in

abundance in complex sites than simple, and between canyon and alluvial sites. In

contrast, terrestrial sources of food (e.g., terrestrial invertebrates) appeared to be more

influenced by seasonality than complexity.

Food web structure in the Big Bend region appeared to be linked with the physical

template as well, where food web structure appeared to be more complex within complex

sites than simple sites. Interestingly, food web function was more associated with native

or nonnative terrestrial vegetation than habitat complexity. Native vegetation appeared to

contribute to a greater proportion of dietary items for aquatic invertebrates, and

subsequently, aquatic invertebrates appeared to contribute to a greater proportion of

118

dietary items for fish consumers. It is important to note, however, that nonnative

vegetation tended to be greater in abundance within simple and alluvial sites than

complex or canyon reaches, and nonnative vegetation has been treated for eradication in

Boquillas Canyon, one of my main sampling canyons within the Big Bend region.

The Rio Grande is a complex river having undergone significant hydrologic and

associated geomorphic changes. My thesis explored this complexity from an ecological

perspective, and illustrates the importance of considering both the physical and ecological

components of a riverine ecosystem. Having used a modeling approach to link the

physical template to ecosystem structure and function in Chapter II, we were then able to

test these modeling results in Chapter three using field observations from the Big Bend

region. My thesis confirms stream flows are a critical variable influencing ecosystem

structure (e.g., faunal assemblages and abundances) and function (e.g., food resource

partitioning and pathways), and habitat heterogeneity is an important predictor of native

fish richness and abundance. We now have a better understanding for native fish

conservation through understanding food web structure and function in large, dewatered

rivers. Overall, my thesis provides an impetus for finding innovative water management

solutions to rehabilitate and restore native aquatic abundance, distribution, and

biodiversity through restoration to physical habitat and improved water quality.

The innovative water management solutions I would suggest are specific for each

region studied in my thesis, the Forgotten Reach and the Big Bend region. For the

Forgotten Reach, improving water quality would likely be the best solution to restoring

the aquatic faunal assemblages and abundances for this degraded segment. In a perfect,

yet unrealistic world, I would suggest we remove Elephant Butte Dam entirely, and

119

demand that the agriculture within Colorado and New Mexico be changed to cultivate

crops less dependent on large water infrastructure and more suited to the biomes they are

cultivated in (e.g., Hatch chilis in New Mexico). In a more realistic world, I would

suggest a significant proportion of the water in Elephant Butte be designated for

environmental flows and emulate a more natural flow regime conducive to improving

water quality, increasing flood magnitudes for native fish to cue in to, and providing

greater habitat heterogeneity in the Forgotten Reach. In addition, I would also suggest

redesigning municipal and agricultural flow infrastructure so that discharges into the river

are treated extensively to improve water quality (e.g., eliminate point sources of salinity

or sulfates, for example).

For the management recommendations in the Big Bend, we need stream flows

that can restore base flow magnitudes and increase habitat heterogeneity. Thus, I would

suggest the United States and Mexico work on an agreement to increase the minimum

release magnitudes from both Luis L. León Dam and Elephant Butte Dam. If we could

maintain greater levels of flow following a reset flood, when habitat is at its most

complex, we would likely observe a reduction in the degree of channel narrowing

throughout the Big Bend region. Like the Forgotten Reach, I would also suggest a

management solution to emulate the natural flow regime (e.g., Blythe and Schmidt 2018).

A simulated spring flood pulse may allow access to the Rio Grande’s floodplain and

subsequently, sensitive fish species such as the Rio Grande silvery minnow, may be able

to re-establish and thrive in the Big Bend region once again.

120

APPENDIX

121

121

Table A1. Average water quality parameters for each mesohabitat mapped in every one-kilometer sampling site and month of

sampling. Month of Sampling Mesohabitat Water Temperature (C) Specific Conductance (µS/cm) Total Dissolved Solids (mg L-1) pH Dissolved Oxygen (mg L-1) Dissolved Oxygen (%)

March backwater 20.3 2310.1 1.5 8.5 8.6 92.5

embayment 20.3 2065.8 1.3 8.2 8.1 89.9

forewater 17.1 2030.0 1.3 8.3 9.1 94.8

pool 20.4 2424.1 1.7 8.3 8.5 92.3

riffle 20.6 2403.0 1.6 8.5 8.2 91.4

run 20.1 2388.5 1.6 8.5 8.6 94.2

side pool 17.3 3068.0 2.0 9.6 9.4 98.7

May backwater 25.7 1308.7 0.9 8.0 8.2 103.6

embayment 26.3 1254.8 0.8 8.2 8.0 99.2

forewater 27.3 1262.3 0.8 8.1 7.9 100.6

pool 27.2 1267.6 0.8 8.1 7.6 95.9

riffle 27.7 1269.0 0.8 8.1 7.7 98.2

run 26.3 1160.9 0.8 8.1 7.5 93.9

side channel 26.6 1263.3 0.8 8.1 7.8 97.7

side pool 30.5 1276.0 0.8 8.2 7.4 100.1

October backwater 26.6 1308.0 0.9 NA 6.5 81.6

embayment 26.5 1385.0 0.9 NA 7.5 93.5

forewater 25.5 1404.5 0.9 NA 4.8 59.0

riffle 25.9 1392.8 0.9 NA 6.1 75.9

run 25.1 1422.7 0.9 NA 6.1 74.8

side channel 26.0 1440.0 0.9 NA 7.3 90.9

122

Table A1. (cont.) Month of Sampling Mesohabitat Water Temperature (C) Specific Conductance (µS/cm) Total Dissolved Solids (mg L-1) pH Dissolved Oxygen (mg L-1) Dissolved Oxygen (%)

January backwater 18.3 1607.0 1.0 7.7 7.2 77.1

embayment 15.7 1640.0 1.1 7.9 7.4 74.8

forewater 16.8 1608.0 1.0 NA 8.1 83.3

pool 16.7 1611.5 1.0 NA 7.3 75.1

riffle 15.9 1626.2 1.1 NA 8.3 83.7

run 16.0 1625.7 1.1 NA 8.5 86.4

side channel 15.0 1643.8 1.1 NA 8.5 84.4

June backwater 30.2 1069.8 NA 8.1 7.6 101.3

embayment 29.9 1240.5 NA 8.3 7.7 101.7

forewater 31.2 1196.3 NA 8.3 8.0 109.7

pool 29.3 1279.4 NA 8.2 7.1 93.1

riffle 29.1 1266.7 NA 8.3 7.5 97.6

run 30.3 1177.6 NA 8.1 7.6 101.6

side channel 30.3 1086.0 NA 8.5 7.9 105.2

side pool 31.2 969.0 NA 7.8 7.3 98.0

122

123

Table A2. Average total length for all fish species caught between complex and simple sampling sites (Min is minimum length, Max

is maximum length). In addition, those denoted with I are nonnative species.

Complex

Simple

Species

n

Mean

length

(mm)

Min

length

(mm)

Max

length

(mm)

n

Mean

length

(mm)

Min

length

(mm)

Max

length

(mm)

Blue sucker 175 37.3 16 80 17 33.5 22 105

Channel catfish (I)* 31 157.7 24 441 23 70.7 27 623

Common carp (I) 49 29.3 18 65 6 34.2 25 57

Flathead catfish 3 201.3 75 360 - - - -

Freshwater drum - - - - 1 188 188 188

Gizzard shad 8 42.9 35 50 - - - -

Longear sunfish 3 44.7 39 52 4 34 29 37

Longnose dace 7 51.0 42 90 1 75 75 75

Longnose gar 7 222.7 42 656 2 509 450 568

Mexican tetra 53 42.6 21 68 12 43.2 29 51

Red shiner 489 39.4 22 63 41 44.2 33 61

Rio Grande silvery minnow 3 62.7 61 66 - - - -

River carpsucker 124 44.6 18 211 34 37.4 22 95

Speckled chub 46 40.2 25 52 18 48.7 39 63

Tamaulipas shiner 472 36.1 16 67 85 42 22 66

Western mosquitofish 18 33.6 21 51 2 41 38 44

*Channel catfish may not be a nonnative species but its own Rio Grande catfish species

123

124

Table A3. Average total length for all fish species caught between alluvial and canyon sampling sites (Min is minimum length, Max is

maximum length). In addition, those denoted with I are nonnative species.

Alluvial

Canyon

Species

n Mean

Length

(mm)

Min

length

(mm)

Max length

(mm)

n Mean length (mm) Min length (mm) Max length (mm)

Blue sucker - - - - 192 36.9 16 105

Channel catfish (I)* 1 134 134 134 53 120.4 24 623

Common carp (I) - - - - 55 29.8 18 65

Flathead catfish - - - - 3 201.3 75 360

Freshwater drum - - - - 1 188 188 188

Gizzard shad - - - - 8 42.9 35 50

Longear sunfish - - - - 7 38.6 29 52

Longnose dace 1 75 75 75 7 51 42 90

Longnose gar - - - - 9 286.3 42 656

Mexican tetra 8 46.4 42 51 57 42.2 21 68

Red shiner 194 40.2 22 63 336 39.6 22 62

Rio Grande silvery

minnow

- - - - 3 62.7 61 66

River carpsucker 3 57.7 33 95 155 42.8 18 211

Speckled chub 26 44.1 26 63 38 41.6 25 51

Tamaulipas shiner 47 42.8 24 57 510 36.5 16 67

Western mosquitofish - - - - 20 34.3 21 51

*Channel catfish may not be a nonnative species but its own Rio Grande catfish species

124

125

Table A4. Total sum for all orders and families of terrestrial invertebrates caught in both

complex and simple sites as well as native and nonnative vegetation, and open sites.

complex

Simple

Coleoptera native nonnative open native nonnative open

N Order 819 785 1119 27 50 192

Curculionidae

2

Dryopidae 1 5 4

Dytiscidae 5

2

Elmidae 13 19 29

Hydrophilidae 1

8

Hydroscaphidae 18 21

Noteridae

1

Sphaeriusidae 9 4 17

Staphylinidae 41 234 52

Diptera

N Order

6960 6384 9470 478 1013 1878

Canacidae

1

Cecidomyidae 2

Ceratopogonidae 1447 984 1912

154 260

Chironomidae 5182 5032 6902

113 320

Culicidae 3

1

Empididae 1

Ephydridae

9 5

Muscidae 20 28 24

1 1

Phoridae 47 36 27 2

Sciomyzidae 5 1

1

Simullidae 344 207 470

3

Tipulidae 14 13 52

1

Ephemeroptera

N Order 723 820 1292 2 132 132

Baetidae 9 2

Behningiidae 611 917 1085

Caenidae 103 22 135

4

Ephemerellidae 2 4 11

Hemiptera

N Order 600 375 299 40 97 54

Corixidae

61 1

Reduviidae 1

Hymenoptera

N Order

243 247 81 34 44 10

Braconidae

1

Trichogrammatidae 12 2 6 2

126

Table A4. (cont.) Lepidoptera

N Order 567 769 287 177 103 38

Crambidae

9 2

Noctuidae 1 3

Mecoptera

N Order

1

Bittacidae 1

Neuroptera

N Order

30 23 10 2 1 2

Myrmeleontidae

2 1

Myrmeliodae 3 6

Odonata

N Order

1 0 1

Lestidae 1

1

Orthoptera

N Order

1 2 7 0 0 1

Tridactylidae

1 4

1

Trichoptera

N Order

5437 4121 4207 1029 654 687

Hydropsychidae 186 258 105 24 18 5

Hydroptilidae 4846 3726 3707 312 136 178

Leptoceridae 63 37 33 15 8 3

Thysanoptera

N Order 28 6 16 6 11 6

Isoptera

N Order 26 9 4 0 1 2

Collembola

N Order 22 16 1 1 5 1

Arachnida

N Order 7 3 3 0 3 0

Table A5. Average relative abundance of functional feeding groups for the four different

sites of habitat heterogeneity sampled along the Rio Grande in the Big Bend region.

Complexity

Filtering

Collectors

Gathering

Collectors

Predators

Scrapers

Shredders

complex 0.141 0.139 0.062 0.649 0.009

simple 0.235 0.202 0.260 0.293 0.010

alluvial 0.350 0.185 0.190 0.263 0.011

canyon 0.128 0.149 0.096 0.619 0.008

127

Table A6. Shannon’s diversity index for terrestrial invertebrates sampled over the

varying levels of habitat heterogeneity and across vegetation type for the Rio Grande

within the Big Bend region.

Habitat Type

All Complex

Native

Nonnative

Open

Shannon's

Diversity Index

1.47

1.52

1.55

Simple

Shannon’s

Diversity Index

1.56

1.57

1.67

Lower Canyons

Shannon's

Diversity Index

1.06

1.17

0.96

Figure A1. Total area calculated for each type of habitat heterogeneity. I show the mean

(x), median, interquartile range (25th and 75th percentiles), and outliers calculated for each

habitat classification.

128

Figure A2. The average catch per unit effort for each sampling month between the years

2016 and 2017. March and May contained the most significant (p-value < 0.05) CPUE

compared to January, June, and October sampling months.

129

Figure A3. Species evenness calculated for the fish communities sampled among the

varying levels of habitat heterogeneity. Evenness measure is based on the calculated

relative abundance for each complex, simple, alluvial, and canyons (Boquillas Lower

Canyons) sampling sites.

130

Figure A4. Chlorophyll a abundance estimates for seston, soft sediment, and periphyton

alga across sampling seasons.