An Examination of the Influence of Multi-Scale Processes and Connectivity on the Population and Assemblage Dynamics of Headwater Fishes

The University of Southern Mississippi The Aquila Digital Community

Dissertations

Fall 12-1-2020

An Examination of the Influence of Multi-scale Processes and Connectivity on the Population and Assemblage Dynamics of Headwater Fishes

Josh P. Hubbell University of Southern Mississippi

Follow this and additional works at: https://aquila.usm.edu/dissertations

Part of the Computational Biology Commons, Genetics Commons, Other Ecology and Evolutionary Biology Commons, Population Biology Commons, and the Terrestrial and Aquatic Ecology Commons

Recommended Citation Hubbell, Josh P., "An Examination of the Influence of Multi-scale Processes and Connectivity on the Population and Assemblage Dynamics of Headwater Fishes" (2020). Dissertations. 1845. https://aquila.usm.edu/dissertations/1845

This Dissertation is brought to you for free and open access by The Aquila Digital Community. It has been accepted for inclusion in Dissertations by an authorized administrator of The Aquila Digital Community. For more information, please contact [email protected]. AN EXAMINATION OF THE INFLUENCE OF MULTI-SCALE PROCESSESSES AND CONNECTIVITY ON THE POPULATION AND ASSEMBLAGE DYNAMICS OF HEADWATER FISHES

Joshua Paul Hubbell

A Dissertation Submitted to the Graduate School, the College of Arts and Sciences and the School of Biological, Environmental, and Earth Sciences at The University of Southern Mississippi in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

Approved by:

Dr. Jacob F. Schaefer, Committee Chair Dr. Melvin L. Warren Jr. Dr. Brian R. Kreiser Dr. John B. Alford Dr. Carl P Qualls

December 2020

Joshua Paul Hubbell

2020

Published by the Graduate School

ABSTRACT

Typified by their branching pattern, headwaters are numerically abundant as the density of these habitats increases with increasing distance from the base of a dendritic river system. Connectivity among headwaters is complex, resulting in the spatial isolation of populations. Headwater specialists have evolved a suite of traits that permit these species to permanently reside within these habitats. The spatial configuration and connectivity of headwaters has repercussions for metapopulations and meta-assemblages. I investigated how multi-scale processes and connectivity influenced the patch occupancy, coexistence, movement ecology, population structure, and gene flow of headwater specialists. In chapter two, I used occupancy modeling to assess the patch occupancy and coexistence of three benthic stream fishes. Patch occupancy of each species was influenced by distinct habitat variables. Species co-occurrences were best explained as independent occurrences within a stream-reach according to species specific habitat associations. In chapter three, I used stream fish assemblages to examine how confluence size, as a metric of connectivity, and land cover influenced in-stream habitat, movement behavior, movement rate, and assemblage change at four headwater confluences. Results from hierarchical modeling and multivariate analyses suggested that, generally, habitat change, and movement behavior were related to land cover. There was also indirect evidence that movement rate and assemblage change were more influenced by land cover than confluence size. These findings suggest that land cover may alter the effect of confluence size on movement and assemblage change within headwaters. In chapter four, I used a comparative framework to examine how connectivity influences two benthic headwater fishes ecological response to landscape heterogeneity using presence-absence and genetic data. I further examined

whether river system size altered connectivity and thus influenced the ecological response of both species. Occupancy modeling was used to elucidate species-landscape gradient relationships assuming populations occupied discreet habitat patches. Linear mixed effects models of genetic distance were used to identify whether these relationships changed as a consequence of discrepancies in connectivity between habitat patches. Our results suggested connectivity modified the impact of environmental and anthropogenic gradients on species ecological responses. Further, river system size affected connectivity, thus influencing species-landscape gradient relationships.

iii

ACKNOWLEDGMENTS

I would like to thank the members of my dissertation committee Drs. Jacob F.

Schaefer, Melvin L. Warren Jr., Brian R. Kreiser, John B. Alford, and Carl P. Qualls. Their advice, critique, and support substantially contributed to this dissertation and to my development as a researcher and ecologist. In particular, I want to thank my PhD mentor,

Jacob F. Schaefer, for challenging me to pursue broad, ecologically relevant hypotheses.

He entrusted me with multiple large-scale projects, but also offered advice and creative input that often facilitated the development of solutions to many project-related problems.

I also want to thank Melvin L. Warren Jr. for his exceptional expertise on stream fishes in

Mississippi and for providing me with data that made much of this research possible. I owe

Brian R. Kreiser a huge debt of thanks for helping a budding ecologist develop a strong background in population genetics.

Multiple funding sources made the completion of this dissertation possible. I received funding from the National Fish and Wildlife Foundation, the United States Forest

Service, and the United States Fish and Wildlife Service. I also received travel awards from the School of Biological, Environmental, and Earth Sciences within the University of

Southern Mississippi (USM), and from the Southeastern Fishes Council. I received further funding from the Southeastern Fishes Council for winning third place for best oral presentation at their annual 2018 meeting.

Most importantly, I was supported by my family and friends. My wife, Meagan, endured countless stressful days and nights throughout my tenure at USM but never wavered in her encouragement. My parents and sister gave me unending love and support throughout this journey. I would like to thank my friends in the USM graduate program

Will McFarland, Grover Brown, Peter Flood, Tom Sevick, Scott Clark, Sara Barret,

Andrew Coombes, Loren Stearman, Matt Lodato, Annemarie Fearing, Cybil Huntzinger,

Gabbie Berry, Luke Pearson, and Chris Pellachia for their support or assistance in the field.

Anne Fairly, Austin King, and Sara Rubelowsky also provided assistance in the lab or in the field.

DEDICATION

This dissertation is dedicated to my family for always having faith in my abilities, and to my wife Meagan for supporting me through many arduous trials.

TABLE OF CONTENTS

ABSTRACT ...... ii

ACKNOWLEDGMENTS ...... iv

DEDICATION ...... vi

LIST OF TABLES ...... xii

LIST OF ILLUSTRATIONS ...... xiv

CHAPTER I INTRODUCTION ...... 1

CHAPTER II MODELING PATTERNS OF CO-OCCURRENCE OF THREE

CONGENERIC HEADWATER FISHES ...... 14

Abstract ...... 14

Introduction ...... 15

Methods...... 18

Study System and Species...... 18

Field Methods ...... 20

Data Processing ...... 21

Modeling of Occupancy and Co-occurrence ...... 23

Covariates ...... 25

Model Selection ...... 26

Results ...... 26

Modeling of Detection...... 27

vii

Occupancy Modeling ...... 27

Modeling of Coexistence ...... 29

Discussion ...... 33

Influence of Competition on the Coexistence of Headwater Fishes at the Reach-

Scale ...... 34

Coexistence of Headwater Congeners is Mediated by Niche Differences ...... 35

Groundwater Discharge is an Important Component of Yazoo Darter Occupancy 37

Future Directions ...... 37

Conclusions ...... 38

CHAPTER III CONFLUENCES AND LAND COVER AS AGENTS OF CHANGE:

TEMPORAL HABITAT VARIABILITY MODIFIES THE MOVEMENT AND

ASSEMBLAGE DYNAMICS OF HEADWATER FISHES ...... 51

Abstract ...... 51

Introduction ...... 52

Methods...... 55

Study System...... 55

Study Design ...... 56

Functional Groups ...... 58

Study-Period ...... 58

Mark-Recapture Methods ...... 59

viii

Habitat Sampling ...... 62

Analysis of Habitat Change ...... 63

Analysis of Movement Behavior...... 63

Estimation of Movement Rates and Assemblage Change ...... 65

Analysis of Reach Scale Movement Rates and Assemblage Change ...... 66

Results ...... 67

Recapture Rates ...... 67

Habitat Change ...... 68

Movement Behavior ...... 70

Movement Rates ...... 73

Assemblage Change ...... 74

Analysis of Movement Rates ...... 74

Analysis of Assemblage Change ...... 75

Discussion ...... 75

Effect of Land Cover on Habitat Change ...... 77

Effect of Confluence Size on Habitat Change ...... 78

Movement Behavior of Water-Column Specialists ...... 79

Effect of Woody Structure Availability on the Movement of Water-Column

Specialists ...... 80

Effect of Woody Structure Availability on Assemblage Change ...... 82 ix

Conclusions ...... 84

CHAPTER IV FUNCTIONAL CONNECTIVITY BEGETS CONFLICTING IMPACTS

OF LANDSCAPE GRADIENTS ON THE OCCUPANCY AND GENETIC DISTANCE

OF TWO HEADWATER FISHES ...... 98

Abstract ...... 98

Introduction ...... 99

Methods...... 103

Study System...... 103

Study Species ...... 104

Datasets...... 105

Genetic Analyses ...... 106

Processing of Geospatial Data ...... 109

Hierarchical Modeling...... Error! Bookmark not defined.

Model Selection ...... 114

Results ...... 114

Population Structure and Descriptive Genetic Statistics ...... 114

Migration ...... Error! Bookmark not defined.

Hierarchical Modeling...... 120

Discussion ...... Error! Bookmark not defined.

Isolation By Distance ...... 128

Dendritic Structuring Has Repercussions for Metapopulation Dynamics ...... 128

Habitat Specialization Emphasized When Modeling Genetic Distance ...... 129

Subbasin Size Alters Functional Connectivity ...... 130

Influence of Reservoirs on Genetic Distance ...... 132

Conclusions ...... 133

APPENDIX A ...... 148

APPENDIX B ...... 151

APPENDIX C ...... 158

LIST OF TABLES

Table 2.1 Top single-species occupancy models and intercept-only models for occurrence of three sympatric darter species sampled in the Little Tallahatchie River system, MS, USA. Values are shown for the number of parameters (K), AICC, ∆AICC and model weights (wi). Intercept-only models are designated by periods in place of covariates. Models with parameters θ and θ’ indicate spatial-dependence models...... 28

Table 2.2 Co-occurrence occupancy models used to evaluate the role of interspecific interactions on the habitat use of three sympatric darters sampled in the Little

Tallahatchie River system, MS, USA. Values are shown for the number of parameters

(K), AICC, ∆AICC nd model weights (wi)...... 30

Table 2.3 Occupancy and detection probabilities (p and r) estimated from co-occurrence occupancy models of three sympatric darter species sampled in the Little Tallahatchie

River system, MS, USA...... 32

Table 3.1 Percent recapture rates for trips, years, and between years for three functional groups of fishes (WCS = Water Column Specialists; BS= Benthic Specialists; SS =

Structure Specialists) at each mark-recapture site. (Btw=Between)...... 60

Table 3.2 Numbers of movers by move-type for all four mark-recapture sites across the study period...... 64

Table 3.3 Summary statistics for distances moved by water-column specialists within and among reaches at all four sites. Dist= Distance...... 72

Table 3.4 Top reach scale and multi-scale models of water-column specialists’ proportional daily movement rates (PDMR) and of Bray-Curtis Distances at four

xii

headwater confluences in the Pascagoula River basin, MS, USA. Values are shown for the number of parameters (K), AICC, ∆AICC and model weights (wi)...... 76

Table 4.1 Definitions and a priori hypotheses (+/-) for each covariate for occupancy and genetic distance. Covariates means for ...... 112 genetic distance were estimated using the total pathway distance (km) between pairs of sites...... 112

Table 4.2 Expected heterozygosity, mean allelic richness, and Fis of Yazoo Darter and

Goldstripe Darter STRUCTURE ...... 117 hierarchical groupings...... 117

Table 4.3 Pairwise Fst values for K=5 STRUCTURE, Goldstripe Darter hierarchical groupings...... 118

Table 4.4 Pairwise Fst values for K=6 STRUCTURE, Yazoo Darter hierarchical groupings...... 119

Table 4.5 Yazoo and Goldstripe Darter hierarchical model sets for occupancy and genetic distance (Fst). Two linear mixed effects model sets were constructed for genetic distance to elucidate the effects of subbasin size (Little Tallatchie and Yocona) on functional connectivity...... 122

xiii

LIST OF ILLUSTRATIONS

Figure 2.1 Map of detections, co-detections, and absences for each darter and darter pairing for all 97 localities distributed across the Tippah and upper Little Tallahatchie

River systems included in our analyses. Inset map identifies the locality of the upper

Yazoo River basin. Color (red, orange, yellow, teal, pink) and shape (Δ, О) combinations distinguish which darter or darter pairing was detected at a given locality. All black circles indicate sites where none of the target species were detected...... 19

Figure 3.1 Diagram of site layout. Each site was located at a confluence and consisted of three 100 m reaches. Pools (P) and riffles (R) were delineated as patches of habitat...... 57

Figure 3.2 Principal component analysis (PCA) of in-stream habitat variables for PC1 and PC2 axes. Arrows end at centroids for environmental variables...... 69

Figure 3.3 Frequency distributions in relation to distance moved for each mark-recapture site in the Pascagoula River basin, MS...... 71

Figure 3.4 Plot of interaction between mean difference in woody structure and mean difference in current velocity on the proportional daily movement rate (PDMR) of water- column specialists. Mean difference in woody structure was converted into a factor with two levels (A: small = ≤ 0%; B: large = > 0%) to identify differences in PDMR as a consequence of the interaction between the two habitat variables. The number of plotted points along each regression represents the number of predicted values associated with that level of woody structure. *No predicted estimates of PDMR for Mixon’s Creek were associated with an increase in woody structure...... 81

Figure 3.5 Plot of interaction between mean difference in woody structure and mean difference in depth on reach scale, Bray-Curtis Distances. Mean difference in woody

xiv

structure was converted into a factor with two levels (A: small = ≤ 0%; B: large = > 0%) to identify differences in Bray-Curtis Distance as a consequence of the interaction between the two habitat variables. The number of plotted points along each regression represents the number of predicted values associated with that level of woody structure.

*No predicted estimates of Bray-Curtis Distance for Mixon’s Creek were associated with an increase in woody structure...... 83

Figure 4.1 Map of 39 localities where tissue data for Yazoo Darters, Goldstripe Darters, or both species was collected across the Little Tallahatchie and Yocona River subbasins.

Inset map identifies the locality of the upper Yazoo River basin. Shapes (star, circle, or triangle) identify which species were detected at each site...... 107

Figure 4.2 Plots of probability of occurrence (ψ) for Goldstripe and Yazoo Darters.

Goldstripe Darter occupancy was best predicted as a function of drainage area (A) whereas Yazoo Darter occupancy was best predicted as a function of well depth (B). . 124

Figure 4.3 Plots of Yazoo Darter genetic distance (linearized Fst) regressed against the best predictors. Predictors were chosen based on their inclusion in the best supported model and their effect size (beta estimates). Predictor values represent mean total pathway estimates between pairs of subpopulations. Plot A (YRSB): Genetic distance regressed against drainage area; plot B (LTRSB): genetic distance regressed against river distance...... 125

Figure 4. 4 Plots of Goldstripe Darter genetic distance (linearized Fst) regressed against the best predictors. Predictors were chosen based on their inclusion in the best supported model and their effect size (beta estimates). Predictor values represent mean total pathway estimates between pairs of subpopulations. Plot A (YRSB): Genetic distance

regressed against drainage area; plot B (LTRSB): genetic distance regressed against river distance...... 127

xvi

CHAPTER I INTRODUCTION

Within their geographic range, species can be relatively abundant, being numerous in some habitats but rare or absent in others. The population densities of species are greatest near the center of their range, and decline gradually upon approaching the boundary

(Brown, 1984). Environmental conditions alter patch quality and connectivity at different spatial scales across an ecosystem, which changes patterns in species distributions

(Jackson, Pedro & Olden, 2001).The persistence of a species is a balance between the threat of extinction due to these environmental conditions, and the probability of movement between suitable patches and the boundaries that separate them (Levins, 1966; Wiens,

2002). A patch may best be defined as a spatial subunit that is established by an organism

(Pringle et al., 1988), and connectivity is the extent to which a landscape facilitates or restrains the movement of that organism from one patch to the next (Tischendorf & Fahrig,

2000). Many species are habitat specialists, and exhibit zonation (i.e., the distribution of organisms into specific zones in an ecosystem), because habitat types are distributed nonrandomly throughout an ecosystem dependent on patch connectivity and structure

(Schlosser, 1991; Wiens, 1989; Poole, 2002). Because patch structure and connectivity are scale dependent, an organism’s response will depend on the scale at which it distinguishes distinctions in habitat structure (Kotliar & Wiens, 1990). The hierarchy of patch dynamics

(HPD) provides a framework for viewing ecosystems as ‘nested discontinuous hierarchies of patch mosaics’ (Hollings, 1992; Kotliar & Wiens, 1990), and is significant in its ability to describe the interplay between patch mosaics and their linkage by trans-scale processes

(Poole, 2002). Trans-scale processes regulate the structure of patches throughout an ecosystem. There are two types of trans-scale processes, bottom-up and top-down.

Ecological disturbance is the primary form of top-down trans-scale process, in which a form of disturbance (e.g., fire, flood) is influenced by the ecosystem’s encompassing elemental structure, and influences the component element structure. Contrastingly, in a bottom-up trans-scale process, fine-scale patch structure influences the trans-scale process, which subsequently influences patch structure at coarser scales (Poole, 2002).

In river systems, confluences are origins of habitat construction and augment morphological heterogeneity, forming abrupt transitions in habitat within dendritic river systems (Rice & Church, 2001; Benda, Veldhuisen & Black, 2003). The configuration of confluences in river systems, may be defined by basin size, basin shape, network pattern, drainage and confluent density, and local network geometry (Benda et al., 2004). In regards to basin size, consistent discharge-related morphological changes (i.e., channel width, depth), occur at junctions where the ratio between tributary size and main stem size approaches 0.6 or 0.7 (Rhoads, 1987). In general, as main stem size increases,

“geomorphically significant” confluences are associated with increasingly larger tributaries (Benda et al., 2004). The cumulative effect of confluences should be proportional to the total number of geomorphically significant channel confluences. Local network geometry can be used to describe the kilometer-scale variation of tributary effects in rivers, including the longitudinal sequence of tributary-main stem size ratios, tributary intersection angles, and distance between geomorphically significant confluences (Benda et al., 2004). The ability of confluences to influence longitudinal patterns of riverine heterogeneity is linked to the “watershed disturbance regime” (i.e., the frequency magnitude, and spatial extent of climatic and geomorphic processes). An increase in water supply, sediment load, and the input of woody structure from tributaries results in a higher

frequency and magnitude of morphological changes observed at confluences. Disturbance enhances the role of confluences’ as agents of habitat heterogeneity, and thus it is an important factor that helps shape patch dynamics along across a river system (Benda et al.,

2004; Poole, 2002).

The hierarchical arrangement of patch mosaics across spatial scales influences the distribution and abundance of species across a river system (Diez & Pulliam, 2007; Fagan,

2002; Poole, 2002). In dendritic river systems, individual stream segments are ecologically connected in the longitudinal dimension, and the resulting “branchiness” and hierarchy of habitat arrangements in dendritic river systems affects patch connectivity and the isolation of metapopulations (Fagan, 2002). Also, the longitudinal arrangement of segments within a dendritic river system is unique and dynamic over time (Rice & Church, 2001), further adding to the complexity of patch connectivity across a dendritic river system. Because stream fishes differ in their movement capabilities and habitat preferences, their responses to the heterogeneity of a dendritic river system will be different (Wiens, 2002). At the watershed scale, the biogeographic history of a watershed and the variation of environmental variables encountered provide a diversity of patches from headwaters to the mouth of a lotic system, thus driving the composition of the assemblage of fishes encountered throughout a river basin (Robinson & Matthews, 1998). At the most finite scale, microhabitat studies look at differences in microhabitat availability and species associated with such habitat (Grossman & Freeman, 1987).

Habitat fragmentation limits the dispersal and occurrence of species by further disrupting the connectivity and structure of patches across a river system, and is a symptom of both natural and anthropogenic disturbance (Fagan, 2002; Brown et al., 2001; Walters

et al., 2005). Random fragmentation in river systems increases patch fragments, and the size variability of these patch fragments (Fagan, 2002). Fragmentation affects dispersal in stream fishes differently, and fishes that travel greater distances tend to be more impacted by habitat fragmentation (Pepino, Rodríguez & Magnan, 2012).

Anthropogenic disturbance amplifies the effects of habitat fragmentation across a river system. Changes in land use in a watershed facilitates patch fragmentation in streams, leading to decreased diversity of stream fish assemblages, and limiting species dispersal at multiple spatial scales (Sawyer et al., 2004; Walters et al., 2005). Stream fish diversity is correlated with habitat heterogeneity and water quality, as habitat degradation and water quality decreases, the diversity and abundance of endemics decline as well (Tabit &

Johnson, 2002; Walters et al., 2005). The degradation of streams that flow through non- forested (i.e., urban or agricultural land use types) is associated with poor land management

(e.g., underdeveloped riparian buffers, an increase in impervious surfaces). Streams influenced by anthropogenic disturbance exhibit flashy hydrographs, altered channel morphology, increased concentrations of nutrients and contaminants, decreased species richness, and an increase in the dominance of tolerant species (Meyer et al., 2007; Paul &

Meyer, 2001). Finally, the homogenization of lotic habitats due to incised stream channels and the construction of impoundments has accelerated the dispersal rate of invasive species, furthering the instability of many North American stream fish assemblages (Gido

& Brown, 1999; Gido, Schaefer & Falke, 2009).

Habitat fragmentation limits dispersal by disrupting the connectivity of patches across a landscape, and therefore, influences population structure (Hanski, Peltonen & Kaski,

1991; MacArthur & Wilson, 1967). Reduced gene flow causes reduced intra-population

variation, which affects fitness, population persistence and the adaptability of populations to shifting environmental conditions (Dibattista & Joseph, 2008; Lande, 1987; Reed &

Frankham, 2003). The disturbance of migration pathways can eventually result in the reduction in total genetic variation within a taxon (Allendorf & Luikart, 2007; Epps et al.,

2007). In some cases, genetic distinctions across subpopulations are a factor of within drainage population structuring due to the hierarchical arrangement of stream habitats

(Boizard, Magnan & Angers, 2009). However habitat fragmentation, as an effect of anthropogenic disturbance, is increasingly documented to disrupt gene flow among populations by eliminating dispersal pathways, (Stow, Sunnucks, Briscoe & Gardner,

2001) and isolating populations (Yamamoto, Kojima & Naemura, 2004; Wofford,

Gresswell & Banks, 2005; Bergl & Vigilant, 2007).

Headwater specialists, such as darters and many cyprinids, may best be defined as species which reside permanently in small streams (Meyer et al., 2007). Headwater streams are typified by low species diversity, as they are easily influenced by small changes in abiotic conditions. Therefore, interspecific competition is reduced in these small ecosystems (Meyer et al., 2007) despite many species sharing similar ecological, physiological, and morphological traits (Ross, 2013). Many of these traits are adaptations to the small stream sizes in which they occur (Randinger & Wolter, 2014). Because many headwater systems tend to be highly isolated within a dendritic river system, headwater specialists are characterized by low dispersal rates (Mundahl & Ingersoll, 1983; Petty &

Grossman, 2004; Scalet, 1973). Previous research has shown that the probability of a fish to emigrate is negatively correlated to its distance from the main stem (Albanese,

Angermeier & Doraj-Raj, 2004). Also, morphological adaptations to small stream environs

such as body size, and caudal fin shape, both of which affect swimming performance, may further limit the dispersal of these small stream specialists (Petty & Grossman, 2004; Hudy

& Shiflet, 2009; Ovidio, Detaille, Bontinck & Phillippart, 2009).

Habitat fragmentation, as a function of anthropogenic disturbance, may further limit the dispersal of headwater specialists. Small streams habitats may be fragmented or eliminated due to groundwater extraction (Hubbs, 1995; Cross & Moss, 1987) or land disturbing activities (agriculture, urbanization) (Meyer &Wallace, 2001). Past studies have noted that fish assemblages in low-impacted streams with stable flow regimes were comprised of sensitive species and trophic specialists, whereas channelized streams characterized by highly fragmented habitats possessed species with generalized feeding strategies that were associated with silt and slower current velocities (Poff & Allan, 1995).

By examining the effects of spatial scale, patch quality, and connectivity, inferences can be made to discern factors which influence headwater fish dispersal, gene flow, co- occurrence, and assemblage change within a watershed (Dextrase, Mandrak & Schaefer,

2014; Fagan, 2002; Hudman & Guido, 2012; Wiens, 2002). Many headwater fishes are habitat specialists, and thus are already ecologically limited in their dispersal at multiple spatial scales within a watershed (Ross, 2013; Wiens, 2002). Furthermore, anthropogenic disturbance increases habitat fragmentation, further limiting species dispersal (Fagan,

2002; Sawyer et al., 2004; Walters et al., 2005). The purpose of the following studies is to determine how patch quality and connectivity influence stream fish movement, population structure, gene flow, co-occurrence, and assemblage change in headwaters at multiple spatial scales. Specifically, research objectives of this dissertation are as follows:

1. Which plays a larger role in describing patterns of co-occurrence in three headwater specialists across a dendritic river system: abiotic or biotic factors?

2. Does confluence size or land cover more greatly influence patterns of dispersal and assemblage change in headwater fishes?

3. Does connectivity in a dendritic river system affect how headwater fishes relate to environmental and anthropogenic gradients?

4. How does gene flow of headwater fishes differ in relation to changes in landscape level processes, and connectivity across multiple spatial scales?

REFERENCES

Allendorf, F. W., & Luikart, G. (2009). Conservation and the genetics of populations.

John Wiley & Sons.

Benda, L., Veldhuisen, C., & Black, J. (2003). Debris flows as agents of morphological

heterogeneity at low-order confluences, Olympic Mountains,

Washington. Geological Society of America Bulletin, 115(9), 1110-1121.

Benda, L. E. E., Poff, N. L., Miller, D., Dunne, T., Reeves, G., Pess, G., & Pollock, M.

(2004). The network dynamics hypothesis: how channel networks structure

riverine habitats. BioScience, 54(5), 413-427.

Bergl, R. A., & Vigilant, L. (2007). Genetic analysis reveals population structure and

recent migration within the highly fragmented range of the Cross River gorilla

(Gorilla gorilla diehli). Molecular Ecology, 16(3), 501-516.

Boizard, J., Magnan, P., & Angers, B. (2009). Effects of dynamic landscape elements on

fish dispersal: the example of creek chub (Semotilus atromaculatus). Molecular

Ecology, 18(3), 430-441.

Brown, J. H. (1984). On the relationship between abundance and distribution of

species. The American Naturalist, 124(2), 255-279.

Brown, D. K., Echelle, A. A., Propst, D. L., Brooks, J. E., & Fisher, W. L. (2001).

Catastrophic wildfire and number of populations as factors influencing risk of

extinction for Gila trout (Oncorhynchus gilae). Western North American

Naturalist, 139-148.

Cross, F. B., & Moss, R. E. (1987). Historic changes in fish communities and aquatic

habitats in plains streams of Kansas. Community and evolutionary ecology of

North American stream fishes. University of Oklahoma Press, Norman, 155-165.

Dextrase, A. J., Mandrak, N. E., & Schaefer, J. A. (2014). Modelling occupancy of an

imperilled stream fish at multiple scales while accounting for imperfect detection:

implications for conservation. Freshwater Biology, 59(9), 1799-1815.

DiBattista, J. D. (2008). Patterns of genetic variation in anthropogenically impacted

populations. Conservation Genetics, 9(1), 141-156.

Diez, J. M., & Pulliam, H. R. (2007). Hierarchical analysis of species distributions and

abundance across environmental gradients. Ecology, 88(12), 3144-3152.

Epps, C. W., Wehausen, J. D., Bleich, V. C., Torres, S. G., & Brashares, J. S. (2007).

Optimizing dispersal and corridor models using landscape genetics. Journal of

Applied Ecology, 44(4), 714-724.

Gido, K. B., & Brown, J. H. (1999). Invasion of North American drainages by alien fish

species. Freshwater Biology, 42(2), 387-399.

Gido, K. B., Schaefer, J. F., & Falke, J. A. (2009). Convergence of fish communities

from the littoral zone of reservoirs. Freshwater Biology, 54(6), 1163-1177.

Grossman, G. D., Ratajczak Jr, R. E., Crawford, M., & Freeman, M. C. (1998).

Assemblage organization in stream fishes: effects of environmental variation and

interspecific interactions. Ecological Monographs, 68(3), 395-420.

Hanski, I., Peltonen, A., & Kaski, L. (1991). Natal dispersal and social dominance in the

common shrew Sorex araneus. Oikos, 48-58.

Hubbs, C. (1995). Perspectives: springs and spring runs as unique aquatic

systems. Copeia, 1995(4), 989-991.

Hudy, M., & Shiflet, J. (2009). Movement and recolonization of Potomac sculpin in a

Virginia stream. North American Journal of Fisheries Management, 29(1), 196-

204.

Jackson, D. A., Peres-Neto, P. R., & Olden, J. D. (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.

Kotliar, N. B., & Wiens, J. A. (1990). Multiple scales of patchiness and patch structure: a

hierarchical framework for the study of heterogeneity. Oikos, 253-260.

Lande, R. (1987). Extinction thresholds in demographic models of territorial

populations. The American Naturalist, 130(4), 624-635.

Levins, R. (1966). The strategy of model building in population biology. American

Scientist, 54(4), 421- 431.

MacArthur, R. H., & Wilson, E. O. (2001). The theory of island biogeography (Vol. 1).

Princeton university press.

Matthews, W. J., & Robison, H. W. (1998). Influence of drainage connectivity, drainage

area and regional species richness on fishes of the interior highlands in

Arkansas. The American Midland Naturalist, 139(1), 1-19.

Meyer, J. L., Strayer, D. L., Wallace, J. B., Eggert, S. L., Helfman, G. S., & Leonard, N.

E. (2007). The contribution of headwater streams to biodiversity in river

networks. JAWRA Journal of the American Water Resources Association, 43(1),

86-103.

Meyer, J. L., & Wallace, J. B. (2001). Lost linkages and lotic ecology: rediscovering

small streams. Ecology: Achievement and challenge, 295317.

Mundahl, N. D., & Ingersoll, C. G. (1983). Early autumn movements and densities of

Johnny (Etheostoma nigrum) and fantail (E. flabellare) darters in a southwestern

Ohio stream.

Ovidio, M., Detaille, A., Bontinck, C., & Philippart, J. C. (2009). Movement behaviour of

the small benthic Rhine sculpin Cottus rhenanus (Freyhof, Kottelat & Nolte,

2005) as revealed by radio-telemetry and pit-tagging. Hydrobiologia, 636(1), 119-

128.

Paul, M. J., & Meyer, J. L. (2001). Streams in the urban landscape. Annual review of

Ecology and Systematics, 32(1), 333-365.

Pépino, M., Rodríguez, M. A., & Magnan, P. (2012). Fish dispersal in fragmented

landscapes: a modeling framework for quantifying the permeability of structural

barriers. Ecological Applications, 22(5), 1435-1445.

Petty, J. T., & Grossman, G. D. (2004). Restricted movement by mottled sculpin (Pisces:

Cottidae) in a southern Appalachian stream. Freshwater Biology, 49(5), 631-645.

Poff, N. L., & Allan, J. D. (1995). Functional organization of stream fish assemblages in

relation to hydrological variability. Ecology, 76(2), 606-627.

Poole, G. C. (2002). Fluvial landscape ecology: addressing uniqueness within a river

discontinuum. Freshwater Biology, 47(4), 641-660.

Pringle, C. M., Nainman, R. J., Bretschko, G., Karr, J. R., Oswood, M. W., Webster, J.

R., & Winterbourn, M. J. (1988). Patch dynamics in lotic systems: the stream as a

mosaic. Journal of the North American benthological society, 7(4), 503-524.

Radinger, J., & Wolter, C. (2014). Patterns and predictors of fish dispersal in rivers. Fish

and Fisheries, 15(3), 456-473.

Reed, D. H., & Frankham, R. (2003). Correlation between fitness and genetic

diversity. Conservation Biology, 17(1), 230-237.

Rhoads, B. L. (1987). Changes in stream channel characteristics at tributary

junctions. Physical Geography, 8(4), 346-361.

Rice, S. P., & Church, M. (2001). Longitudinal profiles in simple alluvial systems. Water

Resources Research, 37(2), 417-426.

Ross, S. T. (2013). Ecology of North American freshwater fishes. University of California

Press.

Sawyer, J. A., Stewart, P. M., Mullen, M. M., Simon, T. P., & Bennett, H. H. (2004).

Influence of habitat, water quality, and land use on macro-invertebrate and fish

assemblages of a southeastern coastal plain watershed, USA. Aquatic Ecosystem

Health & Management, 7(1), 85-99.

Scalet, C. G. (1973). Stream movements and population density of the orangebelly darter,

Etheostoma radiosum cyanorum (Osteichthyes: Percidae). The Southwestern

Naturalist, 381-387.

Stow, A. J., Sunnucks, P., Briscoe, D. A., & Gardner, M. G. (2001). The impact of

habitat fragmentation on dispersal of Cunningham’s skink (Egernia cunninghami): evidence from allelic and genotypic analyses of

microsatellites. Molecular Ecology, 10(4), 867-878.

Tabit, C. R., & Johnson, G. M. (2002). Influence of urbanization on the distribution of

fishes in a southeastern upper Piedmont drainage. Southeastern Naturalist, 1(3),

253-268.

Tischendorf, L., & Fahrig, L. (2000). How should we measure landscape, connectivity?

Landscape Ecology, 15(7), 633-641.

Walters, D. M., Leigh, D. S., Freeman, M. C., Freeman, B. J., & Pringle, C. M. (2003).

Geomorphology and fish assemblages in a Piedmont river basin,

USA. Freshwater Biology, 48(11), 1950-1970.

Wiens, J. A. (2002). Riverine landscapes: taking landscape ecology into the

water. Freshwater Biology, 47(4), 501-515.

Wofford, J. E., Gresswell, R. E., & Banks, M. A. (2005). Influence of barriers to

movement on within-watershed genetic variation of Coastal Cutthroat

Trout. Ecological Applications, 15(2), 628-637.

Yamamoto, T., Kojima, M., & Naemura, T. (2004). LIFLET: Light field live with

thousands of lenslets. In ACM SIGGRAPH 2004 Emerging technologies (p. 16).

CHAPTER II MODELING PATTERNS OF CO-OCCURRENCE OF THREE

CONGENERIC HEADWATER FISHES

Abstract

Mechanisms driving patterns of occurrence and co-occurrence among North American freshwater fishes are poorly understood. In particular, the influence of biotic interactions on coexistence among stream reaches and their effects on regional species distribution patterns is not well understood for congeneric headwater fishes. Occupancy models provide a useful framework for examining patterns of co-occurrence while also accounting for imperfect detection. Occupancy models may be extended to test for evidence for evidence that a dominant species influences the occurrence of a subordinate species and thus evaluate support for the hypothesis that species interactions drive patterns of coexistence. We examined patterns of occurrence and co-occurrence at the stream-reach scale among three species of darters (Percidae: Etheostomatinae) that occupy headwater streams within a Gulf Coastal Plain Drainage in the southeastern United States. We assessed species occurrences at 97 sites in 1st to 3rd order streams on one occasion each, and used data from four subreaches sampled with equal effort at each site to estimate species-specific detection probabilities. Following sampling, a suite of habitat variables were collected at three equidistant points along each of the three transects established within a subreach. Coarse (stream-segment, catchment, network) scale variables were also incorporated using geospatial data. Single-species and two-species occupancy models were used to examine patterns of occupancy and coexistence. The occupancy of each species was influenced by distinct habitat variables. Goldstripe Darters (Etheostoma parvipinne) were constrained by a stream size gradient, groundwater input appeared to influence the

occurrence of Yazoo Darters (Etheostoma raneyi), and local habitat heterogeneity (e.g., variation in depth and current velocity) appeared to influence the occupancy of Redspot

Darters (Etheostoma artesiae). We found no evidence that the presence of one species influenced the occurrence of another within a stream-reach based on two-species occupancy models. Rather, species co-occurrences were best explained as independent occurrences within a stream-reach according to species-specific habitat associations.

Occupancy modeling may provide a suitable framework for evaluating the influence of biotic interactions among congeneric stream fishes along species-specific habitat gradients at the stream reach scale. Our study offers insight into how habitat variation can influence coexistence of potential competitors across a large river system.

Introduction

The mechanisms which facilitate the coexistence of sympatric congeners, a group of species within the same genus whose geographic-ranges overlap (Heinrich, Elwen, &

Bräger, 2010), is widely debated among ecologists (Vance, 1972; Schoener 1974;

Hochkirch, Gröning, & Bücker 2007; Sukhikh et al., 2019). Often, the coexistence of these species is regulated by a distinct limiting factor (e.g., the size ratio of one morphological character) (Hutchison, 1959). For example, differences in morphological and behavioral traits may facilitate habitat partitioning, effectively allowing species to exploit different resources and enabling long-term coexistence (Schluter, 2000). Such mechanistic differences limit the ability of ecologists to predict if competitive exclusion is at play

(Davies et al., 2007). Stable coexistence between two potential competitors is predicted when interspecific competition is lessoned (Vergara, Cushman, Urra, & Ruiz-González,

2016). In streams, interspecific competition is reduced by variable abiotic conditions

(Meyer et al., 2007) meaning resident species may share similar ecological, physiological, and morphological traits (Ross, 2013). Thus, abiotic variability may contribute to coarse- scale patterns of co-occurrence in these systems (Giam & Olden, 2016).

Headwater streams (1st – 3rd order streams) are ubiquitous across a stream network, and are structured in hierarchical mosaics of patches, which are ecologically connected longitudinally (Poole, 2002; Wiens, 2002). However, this longitudinal arrangement of patches is unique and dynamic over time (Rice & Church, 2001). The resulting

“branchiness” and hierarchy of these habitat arrangements in stream networks affects patch connectivity and the isolation of metapopulations (Fagan, 2002). Because stream fishes differ in their movement capabilities and habitat preferences, their responses to this diverse, heterogeneous network of patches will be different (Wiens, 2002).

Many headwater fishes exhibit dispersal patterns which adhere to the stream hierarchy model (Meffe & Vrijenhoek, 1988) predicting that hierarchically nested drainages are more likely to exchange organisms. Ecological trait selection within headwater streams is a result of the isolation of unique habitats across a river system (Mundahl & Ingersoll, 1983; Petty

& Grossman, 2004; Schimdt & Schaefer, 2018). Morphological adaptations to these unique habitat patches such as small body size, and caudal fin shape, both of which affect swimming performance, may further limit dispersal of headwater specialists (Petty &

Grossman, 2004; Hudy & Shiflet, 2009; Ovidio, Detaille, Bontinck, & Phillippart, 2009).

As a consequence, these species may have disjunct distributions due to the hierarchical arrangement of headwater habitats, thus influencing assemblage composition.

While heterogeneity of available habitat affects coexistence, it is clear that biological interactions (e.g., predation, interspecific competition, facilitation) are important.

Biological interactions have been shown at micro and mesohabitat scales (Fausch, Nakano

& Ishigaki, 1994; Resetarits, 1997; Grossman et al., 2006); however, other evidence suggests that facilitation and competition may influence coexistence at more coarse scales

(e.g., stream-reach, landscape) (Townsend & Crowl, 1991; Gilliam, Fraser, & Alkins-koo,

1993; Peoples & Frimpong, 2016). Few studies have tested the effects of both biotic and abiotic factors on the coexistence of stream fishes which are congeners or occupy the same trophic guild (Fausk, Nakano, & Ishigaki, 1994; Taylor, 1996; Peres-Neto, 2004; Crow et al., 2010; Peoples & Frimpong, 2016) at these coarser scales; however, we are only aware of one study (Peoples & Frimpong, 2016) that accounted for imperfect detection.

The use of occupancy modeling (MacKenzie et al., 2002) provides a framework to better understand how multi-scale abiotic processes influence the coexistence of congeneric, headwater fishes by testing hypotheses explaining whether co-occurrence of species happens more or less than expected by chance (MacKenzie, Bailey & Nichols,

2004; Miller, Talley, Lips & Grant 2012). To that end, we used this method to evaluate multiple working hypotheses to better understand the coexistence of three congeneric darters within the genus Etheostoma (Percidae: Etheostomatinae, Near et al., 2011) at the stream-reach scale in a Gulf Coastal Plain drainage located within the southeastern United

States. The southeastern United States is the center of diversity for darters; a diverse group of benthic, freshwater fishes which are often headwater specialists (i.e., functional trait databases place 87 of 250 species in springs, or headwater habitats, Frimpong &

Angermeier, 2009). If competition influences occupancy of these headwater fishes, then we predict that a species occupancy or detectability will be lower when another congener is present or detected within a stream reach. If competition does not influence occupancy

of these congeners, then we predict that a species occupancy or detectability will be unaffected when another congener is present or detected within a stream-reach. Assessing the influence of biotic interactions at the stream-reach scale aids in our understanding of how abiotic and biotic processes regulate distributional patterns of aquatic biota at a more coarse spatial extent.

Methods

Study System and Species

We conducted our study in the Little Tallahatchie River system (henceforth LTR), which is positioned within the upper Yazoo River basin in North-Central Mississippi. The

LTR consists of the Little Tallahatchie and Tippah River drainages, and is isolated by the presence of a large (398 km²) reservoir, Sardis reservoir (Figure 2.1). The Yazoo Darter

Etheostoma raneyi, Goldstripe Darter Etheostoma parvipinne, and Redspot Darter

Etheostoma artesiae all occur within the LTR and all three species are most abundant in small to medium sized streams (Ross 2001; Smiley, Dibble, & Schoenholz, 2006; Sterling,

Warren, & Henderson, 2013). These species are members of separate subclades (Yazoo

Darter: Adonia; Goldstripe Darter: Fuscatelum; Redspot Darter: Vexillapinna) within a large genus (Near et al., 2011). While the Goldstripe Darter (Bart & Taylor, 1999) and

Redspot Darter (Piller, Bart, & Walser, 2001) are broadly distributed across the southeastern United States, the Yazoo Darter is endemic to the Upper Yazoo River basin

(Thompson & Muncy, 1986; Suttkus et al., 1994) and listed as vulnerable by the

Southeastern Fishes Council (Warren et al., 2000) and the American Fisheries Society

(Jelks et al., 2008).

Figure 2.1 Map of detections, co-detections, and absences for each darter and darter pairing for all 97 localities distributed across the Tippah and upper Little Tallahatchie River systems included in our analyses. Inset map identifies the locality of the upper Yazoo River basin. Color (red, orange, yellow, teal, pink) and shape (Δ, О) combinations distinguish which darter or darter pairing was detected at a given locality. All black circles indicate sites where none of the target species were detected.

Datasets

Habitat and associated fish assemblage stability in this system (Schaefer, Clark, &

Warren, 2012, Table A1) allowed us to model the occupancy and coexistence of Redspot and Goldstripe Darters with data collected at 53 historic (1999-2003; Sterling, Warren, &

Henderson, 2013) and 44 contemporary (2015-2016) sites (97 sites total) (Figure 2.1). Of the 44 currently surveyed sites, 39 had never been previously sampled. Because Yazoo

Darters are not known to occupy the Eastern portion of the LTR (i.e., the headwaters of the

Little Tallahatchie River, see Hubbell and Schaefer 2017), we used a subset of sites (52 historic, 13 contemporary) to model occupancy and co-occurrence of Yazoo Darters in relation to Redspot and Goldstripe Darters (Figure 2.1). We used 12-digit hydrological unit codes (HUCs, U.S. Geological Survey) to delineate watersheds within the LTR. We randomly selected one to five sites per HUC (mean area ± SD = 82.8 ± 27.4 km²) to reduce historical sampling bias (i.e., site densities were much higher for historically sampled drainages which flow through public land). Due to a lack of accessibility, some HUCs were not able to be sampled.

Field Methods

Subreaches were used as an alternative to re-sampling over time to generate detection histories (Albanese, Peterson, Freeman & Weiler, 2007) for both historic and contemporary sites (sampled May-September). All sites were divided into four evenly sized subreaches.

In this study, we assumed that the occupancy state of a subreach did not change during the survey period, all species were correctly identified, and that all species detections were independent. Reach lengths were calculated by multiplying the average stream wetted width by 30, with minimum and maximum length of 120 m and 300 m, respectively. We

performed backpack electrofishing and seining surveys to sample each subreach with equal effort. Two seine hauls were performed within each subreach, and an attempt was made to sample all available habitats. Electrofishing effort was set at 5 s/m of stream length. We preserved fishes using 10% formalin, and later transferred to 70% ethanol, for cataloging and counting (University of Southern Mississippi Ichthyological Collection). To prevent over-sampling of Yazoo Darter populations, we only vouchered two individuals per site.

Before pooling presence-absence generated by the two gear types for modeling, these data were initially kept separate to assess how detection varied as a result of these two sampling methods; however, because presence-absence data were pooled in the final datasets, we do not present these results here (see Table A2). All sampling protocols were approved by the

University of Southern Mississippi’s Animal Care and Use Committee (IACUC 09-007).

Within each subreach, we established three evenly spaced transects perpendicular to flow.

At three equidistant points (i.e., interval based on wetted width) along each transect we measured depth (cm), current velocity (m·s⁻¹), dominant substrate size (modified

Wentworth scale, Cummins, 1962), and the presence (binary variables) of woody structure or aquatic vegetation. Substrate types were divided into six categories: 1 = silt, 2 = sand,

3 = gravel, 4 = cobble, 5 = boulder, 6 = bedrock. We quantified the coefficient of variation of current velocity (CVCV), substrate size (CVSUB), and depth (CVD) to serve as proxies of habitat heterogeneity. We also calculated means of the following variables (depth = MD; mean current velocity = MCV; mean substrate size = MSUB; WS = woody structure; AV

= aquatic vegetation).

Data Processing

We compiled stream size and network variables (drainage area, confluence link) from

the National Hydrography Dataset Plus (USEPA & USGS, 2006); all data processing was performed in ArcGIS 10.0 (ESRI, 2011). Confluence-link (C-Link) is the number of confluences downstream from each stream segment (Fairchild, Horwitz, Nieman, Boyer &

Knorr, 1998). We defined a stream-segment as an individual segment when using the

NHDPlus flowline vector layer. C-Link values decrease in a given watershed from extreme headwaters to the base of a stream network. We quantified C-Link as the number of confluences downstream from a site to the furthest downstream stream-segment on the main stem of the Little Tallahatchie River. Drainage area (DA) is the total drainage area in km² upstream of each site. We included well depth (WD) (MDWQ, 2015) as a surrogate measure of water table depth to infer stream permanence and relative groundwater input.

Well depth is known to correlate with water table depth across a watershed (Rosenberry,

LaBaugh & Hunt, 2008). We used nearest neighbor interpolation to assign WD vector point data to all sites within the LTR. Nearest neighbor interpolation (implemented using the

Spatial Analysis Toolbox, ESRI, 2011) selects the value of the nearest point and does not consider the values of other points at all. We obtained land cover data (19 classes) for the years 2002 and 2012 from the U.S conterminous wall-to-wall anthropogenic land use dataset (NWALT) (Falcone, 2015). These time periods were used because they represented the closest approximations of land use when fish assemblages were sampled. We reclassified the land cover data into 5 broad land cover types: forested, urban, wetland, open water, and agricultural for both time periods. We conducted a principal components analysis (PCA) on the relative area of each cover type within each site’s DA. The first axis of this PCA explained 45.5% of the variation (with forested land having the highest loading) and was used (PC1) in our occupancy models.

Modeling of Occupancy and Co-occurrence

We used single-species models to characterize important habitat covariates associated with the probability of occurrence (Ψ) and the probability of detection (p) of each headwater darter. Because spatial replicates may not represent truly independent surveys and lead to the inflation of occupancy estimators (Kendall & White, 2009), we also developed spatial dependence models (Hines et al., 2010). Spatial dependence models allow the probability that a spatial segment may or may not be occupied based upon whether the previous segment was occupied (θ’) or not (θ) (Hines et al., 2010) where parameters are modeled as a first-order Markov process. We modeled detection as constant and as a function of sampling covariates to distinguish if p varied among sites. We used untransformed beta estimates to infer relationships (positive or negative) between covariates and parameters. To assess the relative fit of our single-species occupancy models, we used the MacKenzie-Bailey goodness of fit test (MacKenzie & Bailey, 2004), in which overdispersion (ĉ) is estimated by calculating the chi-square goodness of fit statistic for a global model and then dividing it by the mean test statistic of 10,000 bootstrap samples.

Two-species occupancy models using the ΨBa parameterization (Richmond, Hines, &

Beissinger, 2010), an extension of the model described by MacKenzie et al. (2004), were used to test whether species occupancy was influenced by the occupancy or detectability of another congener at the stream-reach scale. This parameterization allows for the estimation of ΨA (probability of occupancy of the dominant species), ΨBA (i.e., probability of co-occurrence), and ΨBa (probability of occupancy of the subordinate species given the dominant species is absent) (Richmond, Hines & Beissinger, 2010). This parameterization

allows for the incorporation of covariates, the direct estimation of the species interaction factor (SIF), and sets ΨB conditional upon ΨA. The SIF represents the probability that the two species co-occur no more or no less than what would be expected if all occurrences of the species were random (MacKenzie et al., 2004). We only calculated SIF if the competition model (i.e., ΨBA ≠ ΨBa) was included among the best models for each darter pair. When using this parameterization, the dominant species and subordinate species must be established a priori (Richmond, Hines & Beissinger, 2010). Because Yazoo Darters are likely more habitat-limited, we assigned Redspot and Goldstripe Darters as ΨA when examining coexistence with Yazoo Darters (designated ΨB). Because of its restriction to smaller streams, we designated Goldstripe Darters as ΨB and Redspot Darters as ΨA when modeling coexistence of these two fishes. We constructed models that assumed occupancy of ΨB would (i.e., ΨBA ≠ ΨBa) or would not (i.e., ΨBA = ΨBa) be influenced by the presence of ΨA. For detection probability, we estimated pA (probability of detecting species A given species B is absent), pB (probability of detecting species B given species A is absent), rA

(probability of detecting species A, given both species are present), rBA (probability of detecting species B, given both species are present), and rBa (probability of detecting species B given species A is present but was not detected). We modeled all detection parameters within our two-species models as both independent (i.e., pA = rA, pB = rBA = rBa), and dependent (i.e., pA ≠ rA, pB ≠ rBA ≠ rBa). Similar to single-species models, we used untransformed beta estimates to infer relationships (positive or negative) between covariates and parameters. We only included the top ranked occupancy covariates and detection covariates (wi > 0.10) from our single-species occupancy models in the two species models to account for habitat preferences and imperfect detection. We constructed

all single and two-species models using the software program PRESENCE (vers. 12.7)

(Hines, 2006).

Covariates

We included four spatial scales of abiotic variables (network, catchment, stream- segment, stream-reach) in our occupancy models using the logit link transformation to model all parameters as a function of covariates (MacKenzie et al., 2002) (Table A3).

Network scale variables captured variation occurring at the largest spatial extent (i.e., the river network), catchment scale variables elucidated variation within distinct drainages

(i.e., 12-digit HUCs), stream-segment variables were indicative of variation among stream- segments, and stream-reach scale variables were indicative of variation among individual sites. C-Link was the only variable included at the network scale, while the PC1 score for land cover was the sole variable included at the catchment scale. At the stream segment scale, we included DA, and WD to estimate the influence of hydrological and geological variables. To assess the effects of stream-reach scale variation on occupancy, we included

MD, MCV, MSUB, CVV CVSUB, CVD, WS, and AV as covariates within our occupancy models. Subreach means for current velocity, substrate size, and depth were included as sampling covariates to estimate p. To evaluate the relative influences of each covariate, we standardized all variables by subtracting the mean and dividing by twice the standard deviation. Prior to modeling, we tested for the correlation between covariates. Any two covariates which had a Pearson correlation greater than the absolute value of 0.5 were not included in the same model. However, correlated variables were used as separate covariates for detection and occupancy within the same model.

Model Selection

We used Akaike’s Information Criterion for small sample sizes (AICC) to assess the quality of competing models (Burnham & Anderson, 2002). Models with small ∆AICC and large Akaike weights (wᵢ) indicate the greater parsimony (Burnham & Anderson,

2002). We only interpreted models with wᵢ > 0.10. To prevent the inclusion of uninformative parameters, models which only differed in ∆AICC by 1-2 units from the best models and possessed similar log-likelihood values were removed (Burnham &

Anderson, 2002). As an alternative to using a single best-supported model, we applied model averaging to quantify unconditional model average estimates of Ψ and p, and associated standard errors (bounded between 0-1.0) for all occupancy and detection parameters within models with wi > 0.001 (Burnham & Anderson, 2002). To infer coexistence patterns, we considered estimated parameters (i.e., ΨA, ΨBA, ΨBa, pA, rA, pB, rBA, rBa) and the relationships among them based on the top ranked model for each species pair.

Results

The three study species were detected in less than half of the surveyed sites. Of the 39 new sites, Goldstripe Darters were detected at 10, Redspot Darters were detected at 13, and

Yazoo Darters were not detected at any of these new localities. Out of the 65 sites sampled occurring within the range of the species, Yazoo Darters were detected at 27 sites (naïve occupancy estimate = 0.42). Out of the 97 total sites sampled, Goldstripe Darters were detected at 30 sites (naïve occupancy estimate = 0.31), and Redspot Darters were detected at 30 sites (naïve estimate = 0.31). Darter species occurred together at rates closer to those expected given independent occurrences. Of the three darter pairings, Yazoo and

Goldstripe Darters were co-detected at 14 of 65 sites (naïve occupancy estimate = 0.22),

Yazoo and Redspot Darters at 8 of 65 sites (naïve estimate = 0.12), and Redspot and

Goldstripe Darters at 11 of 97 sites (naïve estimate = 0.11).

Modeling of Detection

Detection was best modeled as a function of different covariates across the three darter species. Detection of Yazoo Darters was best modeled by depth (wi = 0.96), Goldstripe

Darters by current velocity and depth (wi = 0.84), and Redspot Darters by current velocity and depth (wᵢ = 0.78). Detectability of Yazoo Darters was negatively correlated with depth

(beta estimate, -0.28 ± 0.12). Detection of Goldstripe Darters, was positively related to current velocity (0.26 ± 0.12), and negatively related to depth (-0.74 ± 0.21). Detection of

Redspot Darters was positively related (0.17 ± 0.07) to current velocity, but negatively related to depth (-0.23 ± 0.08). Unconditional, model average estimates for p (i.e., unconditional estimates for dection across all sample sites) for all models with wi > 0.001 for each species were as follows; Yazoo Darter: 0.53 ± 0.09 for p; Goldstripe Darter: 0.33

± 0.18; Redspot Darter: 0.35 ± 0.13.

Occupancy Modeling

Single-season null models were ranked higher than spatial dependence null models for all three darters (Table 2.1), and initial model weights (Redspot Darter, wᵢ = 0.78; Yazoo

Darter, wᵢ = 0.90, Goldstripe Darter, wᵢ = 0.79) suggested that detections were not spatially autocorrelated, justifying the use of spatial replicates for p. All single-species occupancy models converged. Global models for all darters indicated no evidence of a lack of model fit (Yazoo Darter: p = 0.26, ĉ =1.19; Goldstripe Darter: p = 0.62, ĉ = 0.83; Redspot Darter: p = 0.36, ĉ = 1.08). The occupancy of each darter species was associated with distinct

Table 2.1

Top single-species occupancy models and intercept-only models for occurrence of three sympatric darter species sampled in the Little Tallahatchie River system, MS, USA. Values are shown for the number of parameters (K), AICC, ∆AICC and model weights (wi). Intercept-only models are designated by periods in place of covariates. Models with parameters θ and θ’ indicate spatial-dependence models. Species Model K AICC ∆AICC wi

Yazoo p (Depth), Ψ (WD) 5 234.33 0 0.36 Darter p (.), Ψ (.) 2 243.88 11.29 0 p (.), Ψ (.), θ(.) θ’ (.) 5 249.17 16.58 0

Goldstripe p (Depth + Velocity), Ψ (DA) 7 261.98 0 0.68 Darter p (.), Ψ (.) 2 292.09 34.08 0

p (.), Ψ (.), θ(.) θ’ (.) 4 519.63 32.15 0

Redspot p (Depth + Velocity), Ψ (CVD) 5 274.23 0 0.50 Darter p (Depth + Velocity), Ψ (WS) 5 276.59 2.57 0.14

p (Depth + Velocity), Ψ (CVCV) 5 277.21 3.19 0.10

p (.), Ψ (.) 2 282.99 9.07 0

p (.), Ψ (.), θ(.) θ’(.) 5 287.29 13.37 0

habitat variables. Goldstripe and Yazoo Darters were both associated with one model which had wi > 0.10 when modeling occupancy (Ψ), whereas two models had wi > 0.10 when modeling Ψ of Redspot Darters (Table 2.1). Occupancy of Yazoo Darters declined in relation to well depth (beta estimate, -0.26 ± 0.12), and we estimated Ψ at 0.46 ± 0.05 which is a 9.5% increase from the naïve estimate (0.42). Goldstripe Darter occupancy was negatively related to DA (beta estimate, -0.93 ± 0.37), and an average model estimate signified that Ψ for this species was 0.35 ± 0.05 which is a 12.9% increase from the naïve estimate (0.31). Redspot Darter occupancy was positively related to CVD (0.39 ± 0.17), negatively related to WS (-0.26 ± 0.13), and negatively related to CVCV (-0.30 ± 0.16).

Model averaging revealed that Ψ for the Redspot Darter was approximately 0.36 ± 0.06 which is a 16.2% increase from the naïve estimate (0.31).

Modeling of Coexistence

Rankings and parameter estimates for all two-species models suggest that the dominant species did not influence the occupancy of the subordinate species (i.e., ΨBA = ΨBa, pA = rA, pB = rBA = rBa, Table 2.2, Table 2.3). Detection parameters were best modeled as a function of depth for all three species pairs (Table 2.2). For the first species pairing, the best wi = 0.62) assumed that there was a slight, negative effect of stream permanence (i.e.,

WD, beta estimate, -0.21 ± 0.12) on the probability of Yazoo Darter occupancy regardless of presence of Goldstripe Darters (i.e., ΨBA = ΨBa) at the reach-scale, but also indicated that there was a strong negative effect of stream size (i.e., DA, beta estimate, -0.90 ± 0.42) on the probability of Goldstripe Darter occupancy (i.e., ΨA) (Table 2.2). For the second species pairing, the best model (wi = 0.41) only signified that Yazoo Darter occupancy at the reach scale was not influenced by the presence of Redspot Darters (i.e., no habitat covariates

Table 2.2 Co-occurrence occupancy models used to evaluate the role of interspecific interactions on the habitat use of three sympatric darters sampled in the Little Tallahatchie River system, MS, USA. Values are shown for the number of parameters (K), AICC, ∆AICC and model weights (wi). Species Model K AICC ∆AICC wi Yazoo & Goldstripe ΨA (DA), ΨBA = ΨBa (Well), pA = rA (Depth), pB = rBA = rBa (Depth) 8 443.42 0 0.62

ΨA (DA), ΨBA ≠ ΨBa (Well), pA = rA (Depth), pB = rBA = rBa (Depth) 9 445.42 2.50 0.18

ΨA (DA), ΨBA = ΨBa (DA), pA = rA (Depth), pB = rBA = rBa (Depth) 8 446.60 3.18 0.13

ΨA, ΨBA = ΨBa, pA ≠ rA, pB ≠ rBA ≠ rBa 7 461.24 17.79 0

ΨA, ΨBA = ΨBa, pA = rA, pB = rBA = rBa 8 462.19 19.22 0

ΨA, ΨBA ≠ ΨBa, pA = rA, pB = rBA = rBa 9 464.46 21.37 0

ΨA, ΨBA ≠ ΨBa, pA ≠ rA, pB ≠ rBA ≠ rBa 9 466.62 22.74 0

Yazoo & Redspot ΨA, ΨBA = ΨBa, pA = rA (Depth), pB = rBA = rBa (Depth) 6 429.66 0 0.41

ΨA, ΨBA = ΨBa, pA = rA, pB = rBA = rBa 4 431.57 1.91 0.16

ΨA, ΨBA = ΨBa, pA ≠ rA, pB ≠ rBA ≠ rBa 9 432.44 2.78 0.10

ΨA, ΨBA ≠ ΨBa, pA = rA, pB = rBA = rBa 9 433.91 4.25 0.05

Goldstripe & Redspot ΨACVD), ΨBA = ΨBa (DA), pA = rA (Depth), pB = rBA = rBa (Depth) 8 531.93 0 0.46

Table 2.2 (continued) Species Model K AICC ∆AICC wi ΨA(CVD), ΨBA = ΨBa (DA), pA ≠ rA, pB ≠ rBA ≠ rBa (Depth) 14 532.90 1.0 0.28

ΨA(CVD), ΨBA ≠ ΨBa (DA), pA = rA (Depth), pB = rBA = rBa (Depth) 8 534.32 2.39 0.14

ΨA, ΨBA = ΨBa pA = rA, pB = rBA = rBa 4 555.98 18.84 0 ΨA, ΨBA = ΨBa, pA ≠ rA, pB ≠ rBA ≠ rBa 7 556.38 20.13 0 ΨA, ΨBA ≠ ΨBa, pA = rA, pB = rBA = rBa 5 557.95 20.66 0 ΨA, ΨBA ≠ ΨBa, pA ≠ rA, pB ≠ rBA ≠ rBa 8 559.17 22.51 0

Table 2.3

Occupancy and detection probabilities (p and r) estimated from co-occurrence occupancy models of three sympatric darter species sampled in the Little Tallahatchie River system, MS, USA. Species Pair ΨA ΨBA ΨBa rA pA pB rBA rBa Yazoo & Goldstripe 0.39 0.46 0.46 0.39 0.39 0.53 0.53 0.53 Yazoo & Redspot 0.35 0.46 0.46 0.40 0.40 0.53 0.53 0.53 Redspot & Goldstripe 0.36 0.35 0.35 0.35 0.35 0.33 0.33 0.33

appeared in this model, Table 2.2). For the third species pairing, the best model (wi = 0.46) assumed that there was a negative effect of stream size (i.e., DA, beta estimate: -0.93 ±

0.37) on the probability of Goldstripe Darter occupancy regardless of the presence of

Redspot Darters at the reach-scale (i.e., ΨBA = ΨBa), but also implied that there was a slight positive effect of heterogeneity in depth (i.e., CVD, beta estimate, 0.38 ± 0.17) on the probability of Redspot Darter occupancy (i.e., ΨA, Table 2.2).

Discussion

Multi-scale observational studies help ecologists to better understand how fine and large-scale processes influence nonrandom coexistence. However, distinguishing between mechanisms which influence patterns of nonrandom coexistence is difficult because both biotic and abiotic processes shape species distributional patterns. Imperfect species detection may further limit any inferences made. In this study, occupancy modeling was used to integrate habitat covariates and heterogeneous detection probabilities into an investigation of co-occurrence patterns (MacKenzie et al., 2004). Coexistence of headwater fishes is mediated by multi-scale variation, with some studies documenting conflicting results regarding the importance of species interactions within these systems (Townsend &

Crowl, 1991; Taylor, 1996; Grossman et al., 1998; Peoples & Frimpong, 2016). Evidence from our results suggest that (1) competition does not influence the occupancy or detectability of these congeners at the stream-reach scale, (2) and patterns of coexistence at the stream-reach scale may be mediated by habitat preference differences. We therefore suggest that competition likely does not influence the coexistence of these headwater congeners at the stream-reach scale.

Influence of Competition on the Coexistence of Headwater Fishes at the Reach-Scale

The influence of competition on the distributional patterns of small-bodied, stream fishes is inconsistent, and may differ as a consequence of biotic and/or abiotic factors.

Several studies have successfully documented the negative influence of an invasive species on the distribution of native game fish, in which both species are members of the same family (e.g., Salmonidae; Wagner, Deweber, Detar & Sweka, 2013; Hoxomeier &

Dieterman, 2016). However, the extent to which competition affects the distribution of small-bodied fishes is less clear. Because competition among many small-bodied, stream fishes is often restricted to the microhabitat and mesohabitat scales (Resetarits, 1997;

Taylor, 1996; Holomuzki, Feminella, & Power, 2010), phenotypic clustering may allow for the presence of ecologically similar species within a stream-reach (Olden & Kennard,

2010). Many stream fishes have small, restrictive home ranges (Minns, 1995), in particular headwater specialists (Skalski & Gilliam, 2000; Meyer et al., 2007) thus there may be a higher probability of interspecific competition at the microhabitat scale. For example,

Resetarits (1997), using experimental streams, revealed that life stages (juvenile vs. adult) influenced the type of biotic interaction (facilitation or competition) displayed by two benthic strategists at the microhabitat scale. Because headwaters are species depauperate, if competition were important, one might expect ecological release to occur (i.e., a species should exhibit higher abundances in the absence of its competitor, Schoener, 1988). Such a pattern is referred to as density compensation (Crowell, 1962; MacArthur, Diamond &

Karr, 1972) and is associated with intense competition for resources (Angermeier &

Schlosser, 1989). Given our results, we suggest that ecological release, as a biotic process, may only be observed among small-bodied, headwater fishes at the mesohabitat and

microhabitat scales. To date, Taylor (1996) is the only other study we are aware of which has examined the influence of interspecific competition on the coexistence of a headwater benthic fish guild at multiple sample sites across a river system. Although Taylor (1996) found support that small scale interspecific interactions could contribute to the structure of a benthic fish guild (e.g., fish density), his results did not yield significant complementary occurrences. Taylor’s finding of nonsignificant complementary occurrences are supported by another coarse-scale, natural experiment (Peres-Neto, 2004), and the findings of this study. Finally, we also recognize the influence of variation in hydrologic regime on our findings. Extreme hydrologic variability within headwater streams alters demographic patterns, thereby reducing competition for resources within a given stream-reach (Poff &

Allan; 1995; Grossman et al., 1998). Thus, hydrologic variability diminishes the influence of biotic interactions at a fine scale in shaping species distributional patterns at the stream- reach and landscape scales. Therefore, the influence of competition should be assessed across multiple hydrologic regimes.

Coexistence of Headwater Congeners is Mediated by Niche Differences

Coexistence of these headwater congeners may be mediated by distinct habitat preferences. All three of the species assessed in this study are representatives of distinct clades within a large genus (Near et al., 2011), and our results suggest, at the spatial extent examined, habitat use by these three fishes was best explained by distinct habitat parameters (Yazoo Darter: groundwater input; Goldstripe Darter: stream size; Redspot

Darter: variation in depth). One speculation for the distinctive habitat preferences of these fishes is that they may be clade-specific. Streams inhabited by the Redspot Darter and other darters in Vexillapinna are often characterized by strong riffle-pool structure (Scalet, 1973;

Taylor, 2000; Stearman et al., 2015), and may range in size from small headwaters to larger

3rd and 4th order streams (Echelle et al., 1975; Taylor, 2000; Stearman et al., 2015;

Matthews & Turner, 2019). In opposition, in Fuscatelum, the Rush Darter Etheostoma phytophilum and the Goldstripe Darter both appear to be mostly constrained to small headwater tributaries (1st order) within their respective ranges (Robinson & Buchanan,

1988; Metttee et al., 1996; Smiley et al., 2006; Howell, Drennen, & Aarons, 2016). While it seems that the habitat preferences of the Redspot Darter and Goldstripe Darter may be clade-specific, it is difficult to make such a proposition for the Yazoo Darter.

Characterizing commonalities in habitat preference in Adonia is difficult due to the large number of species (19, Near et al., 2011), and widespread distribution (Porter, Cavender,

& Fuerst, 2002) of this group. Our proposition for clade-specific habitat preferences may be indicative of niche conservatism; a classical concept which was recently detected among several clades of stream fishes, including a darter clade (Mcnyset, 2009). Coexistence of two of the three species pairs was best described as a function of stream size and groundwater input (Yazoo and Goldstripe Darter) or habitat heterogeneity and stream size

(Redspot and Goldstripe Darters). Given our results, coexistence among Yazoo and

Goldstripe Darters at the stream-reach scale seems most probable in small, perennial headwaters, whereas the coexistence of Redspot and Goldstripe Darters at this scale appears to be most likely in headwater streams with consistent riffle-pool structure (i.e., high depth heterogeneity). Coexistence of Yazoo and Redspot Darters was best described by a null model in which species occurrences were modeled as independent. Such a result may indicate that at the stream-reach scale, habitat preferences of these two species may be so dissimilar that coexistence at this spatial extent is rare.

Groundwater Discharge is an Important Component of Yazoo Darter Occupancy

Though it is beyond the scope of this study, it is worth noting that the occupancy of

Yazoo Darters was best explained by variability in groundwater input (approximated by the variation in well depths in this study) in relation to all of the other habitat covariates assessed. The influence of groundwater input on shaping stream fish assemblages is well documented (Adams & Warren, 2005; Driver & Hoeinghaus 2015; Mollenhauer, Zhou, &

Brewer, 2019). Spring or groundwater-fed streams are characterized by predictable flow patterns, thus there is limited variation in stream discharge and water temperature (Gordon,

McMahon, Finlayson, & Gippel, 1992); however, the availability, quality, and connectivity of refugia vary within a stream-reach and in relation to drought intensity forming a mosaic of temporally dynamic habitat patches (Magoulick & Kobza, 2003). Thus, aquatic-obligate species occupying ephemeral drainages have often evolved to persist or rapidly recolonize as these habitats fluctuate between flowing and non-flowing states (Dodds et al., 2004).

Adams & Warren (2005) indicate that recolonization probabilities are lower for Yazoo

Darters in relation to Redspot and Goldstripe Darters following a drought within ephemeral drainages. Wider niche breaths may allow eurytopic species (e.g., Redspot and Goldstripe

Darters) to recolonize ephemeral aquatic habitats more rapidly whereas stentopic species

(e.g., Yazoo Darter) may be locally extirpated.

Future Directions

While it appears that coexistence of these congeners at the stream-reach scale may be habitat mediated, other factors cannot be ignored. To further elucidate the extent to which habitat preferences influence the coexistence of a local species pool, Kraft et al. (2015) suggest a multi-step process, focusing on dispersal, persistence, competitive exclusion,

source-sink dynamics, and their contribution to species distributional patterns. Because observational studies are limited in scope, a more thorough test would require manipulative experiments. Such designs assess the coexistence of local guilds through the direct manipulation of inter and intraspecific densities. If species do interact, the extent to which the interaction directly or indirectly affects each species resource usage and acquisition, behavior and other interactions with other species may be assessed (Martin & Martin,

2001).

Conclusions

Our results provide further support that biotic interactions may not be meaningful in describing patterns of co-occurrence of stream fishes; the influence of these interactions on species occupancy may only be detectable at specific spatial extents. There is much support for the regulation of co-occurrence within a stream-reach as a consequence of habitat preferences at the stream-reach and landscape scales (Grossman, Ratajczak, Crawford, &

Freeman, 1998; Jackson, Peres-Neto, & Olden, 2001; Peres-Neto 2004; Giam & Olden

2016); however, there is also limited support that stream fishes may be structured by biotic interactions at the stream-reach scale as well (Peoples & Frimpong 2016; Lamothe,

Dextrase, & Drake 2019). In our study, we sought to examine the influence of biotic and abiotic factors on nonrandom patterns of co-occurrence of three headwater congeners at the stream-reach scale using a method that accounts for imperfect detection. While many studies have detailed the influence of intra and interspecific competition of headwater fishes using experimental streams, these study designs do not provide an appropriate means for extrapolation to the stream-reach scale, due to a lack of variation in physical setting.

Natural study designs which use large datasets and robust analytical tools provide insight

into the significance of how both biotic and abiotic processes structure the coexistence of potential competitors

REFERENCES

Adams, S. B., & Warren Jr, M. L. (2005). Recolonization by warmwater fishes and

crayfishes after severe drought in upper coastal plain hill streams. Transactions of

the American Fisheries Society, 134(5), 1173-1192.

Albanese, B., Peterson, J. T., Freeman, B. J., & Weiler, D. A. (2007). Accounting for

incomplete detection when estimating site occupancy of Bluenose Shiner

(Pteronotropis welaka) in southwest Georgia. Southeastern Naturalist, 6(4), 657-

668.

Angermeier P. L. & Schlosser I. J. (1989). Species–area relationships for stream fishes.

Ecology, 70, 1450–1462.

Bart, H. L., & Taylor, M. S. (1999). Systematic review of subgenus Fuscatelum of

Etheostoma with description of a new species from the upper Black Warrior River

system, Alabama. Tulane Studies in Zoology and Botany, 31(1), 23-50.

Brown, B. L., & Swan, C. M. (2010). Dendritic network structure constrains

metacommunity properties in riverine ecosystems. Journal of Animal

Ecology, 79(3), 571-580.

Burnham, K. P., & Anderson, D. R. (2002). A practical information-theoretic

approach. Model selection and multimodel inference, 2nd ed. Springer, New York.

Crow, S. K., Closs, G. P., Waters, J. M., Booker, D. J., & Wallis, G. P. (2010). Niche

partitioning and the effect of interspecific competition on microhabitat use by two

sympatric galaxiid stream fishes. Freshwater Biology, 55(5), 967-982.

Crowell K.L. (1962). Reduced interspecific competition among the birds of Bermuda.

Ecology, 43, 75–88

Cummins, K. W. (1962). An evaluation of some techniques for the collection and

analysis of benthic samples with special emphasis on lotic waters. The American

Midland Naturalist, 67(2), 477-504.

Dodds, W. K., Gido, K., Whiles, M. R., Fritz, K. M., & Matthews, W. J. (2004). Life on

the edge: the ecology of Great Plains prairie streams. BioScience, 54(3), 205-

216.

Driver, L. J., & Hoeinghaus, D. J. (2016). Spatiotemporal dynamics of intermittent

stream fish metacommunities in response to prolonged drought and

reconnectivity. Marine and Freshwater Research, 67(11), 1667-1679.

Echelle, A. A., Echelle, A. F., Smith, M. H., & Hill, L. G. (1975). Analysis of genic

continuity in a headwater fish, Etheostoma radiosum (Percidae). Copeia, 197-204.

ESRI 2011. ArcGIS Desktop: Release 10. Redlands, CA: Environmental Systems

Research Institute.

Fagan, W. F. (2002) Connectivity, fragmentation, and extinction risk in dendritic

metapopulations. Ecology, 83(12), 3243-3249.

Fairchild, G. W., Horwitz, R. J., Nieman, D. A., Boyer, M. R., & Knorr, D. F. (1998).

Spatial variation and historical change in fish assemblages of the Schuylkill River

drainage, southeast Pennsylvania. American Midland Naturalist 139, 282–295.

Falcone, J. A. (2015). US conterminous wall-to-wall anthropogenic land use trends

(NWALT), 1974–2012 (No. 948). US Geological Survey.

Frimpong, E. A., & Angermeier, P. L. (2010). Trait-based approaches in the analysis of

stream fish communities. In Community ecology of stream fishes: concepts,

approaches, and techniques. American Fisheries Society, Symposium (Vol. 73,

pp. 109-136).

Giam, X., & Olden, J. D. (2016). Environment and predation govern fish community

assembly in temperate streams. Global Ecology and Biogeography, 25(10), 1194-

1205.

Gilliam, J. F., Fraser, D. F., & Alkins-Koo, M. (1993). Structure of a tropical stream fish

community: a role for biotic interactions. Ecology, 74(6), 1856-1870.

Gordon, N. D., McMahon, T. A., Finlayson, B. L., & Gippel, C. J. (2004). Stream

hydrology: an introduction for ecologists. John Wiley and Sons.

Grossman, G. D., Ratajczak Jr, R. E., Crawford, M., & Freeman, M. C. (1998).

Assemblage organization in stream fishes: effects of environmental variation and

interspecific interactions. Ecological Monographs, 68(3), 395-420.

Grossman G. D., Ratajczak R. E., Petty J. T., Hunter M. D., Peterson J. T. & Grenouillet,

G. (2006).Population dynamics of mottled sculpin (Pisces) in a variable

environment: information theoretic approaches. Ecological Monographs, 76,

217–234.

Heinrich, S., Elwen, S., & Bräger, S. (2010). Patterns of sympatry in Lagenorhynchus

and Cephalorhynchus: dolphins in different habitats. In The Dusky Dolphin (pp.

313-332). Academic Press.

Hines, J. E., Nichols, J. D. Royle, J. A., MacKenzie, D. I., Gopalaswamy, A. M., Kumar,

N. S. & Karanth K. U. (2010). Tigers on trails: occupancy modeling for cluster

sampling. Ecological Applications, 20(5), 1456-1466.

Hines, J. E. (2006). PRESENCE v 12.7- Software to estimate patch occupancy and

related parameters.http://www. mbr-pwrc. usgs. gov/software/PRESENCE. html.

Hochkirch, A., Gröning, J., & Bücker, A. (2007). Sympatry with the devil: reproductive

interference could hamper species coexistence. Journal of Animal Ecology, 76(4),

633-642.

Holomuzki, J. R., Feminella, J. W., & Power, M. E. (2010). Biotic interactions in

freshwater benthic habitats. Journal of the North American Benthological

Society, 29(1), 220-244.

Howell, W. M., Drennen, D. J., & Aarons, D. J. (2016). A new population of two

endangered Alabama fishes: watercress (Etheostoma nuchale) and rush

(Etheostoma phytophilum) darters. Journal of the Alabama Academy of

Science, 87(3-4), 156-159.

Hoxmeier, R. J. H., & Dieterman, D. J. (2016). Long-term population demographics of

native Brook Trout following manipulative reduction of an invader. Biological

Invasions, 18(10), 2911-2922.

Hudy, M. & Shiflet, J. 2009. Movement and recolonization of Potomac sculpin in a

Virginia stream. North American Journal of Fisheries Management 29, 196-204.

Hutchinson, G. E. 1959. Homage to Santa Rosalia or why are there so many kinds of

animals? American Naturalist 93, 145–159.

Jackson, D. A., Pedro, R. P., & Olden, J. D. 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, 157-170.

Jelks, H. L., Walsh, S. J., Burkhead, N. M., Contreras-Balderas, S., Diaz-Pardo, E.,

Hendrickson, D. A., & Platania, S. P. (2008). Conservation status of imperiled

North American freshwater and diadromous fishes. Fisheries, 33(8), 372-407.

Kendall, W. L., & White, G. C. (2009). A cautionary note on substituting spatial subunits

for repeated temporal sampling in studies of site occupancy. Journal of Applied

Ecology, 46(6), 1182-1188.

Knouft, J. H. (2003). Convergence, divergence, and the effect of congeners on body size

ratios in stream fishes. Evolution, 57(10), 2374-2382.

Kraft, N. J. B., Adler, P. B., Godoy, O., James, E. C., Fuller, S., and Levine, J. M. (2015).

Community assembly, coexistence and the environmental filtering metaphor.

Functional Ecology 29, 592-599.

Lamothe, K. A., Dextrase, A. J., & Drake, D. A. R. (2019). Characterizing species co‐

occurrence patterns of imperfectly detected stream fishes to inform species

reintroduction efforts. Conservation Biology, 0, 1-12.

MacArthur R.H., Diamond, J.M., & Karr, J.R. (1972) Density compensation in island

faunas. Ecology, 53, 330–342.

MacKenzie, D. I., & Bailey, L. L. (2004). Assessing the fit of site-occupancy

models. Journal of Agricultural, Biological, and Environmental Statistics, 9(3),

300-318.

MacKenzie, D. I., Bailey, L. L., & Nichols, J. D. (2004). Investigating species co‐

occurrence patterns when species are detected imperfectly. Journal of Animal

Ecology, 73(3), 546-555.

MacKenzie, D. I., Nichols, J. D., Lachman, G. B., Droege, S., Andrew Royle, J., &

Langtimm, C. A. (2002). Estimating site occupancy rates when detection

probabilities are less than one. Ecology, 83(8), 2248-2255.

Magoulick, D. D., & Lynch, D. T. (2015). Occupancy and abundance modeling of the

endangered Yellowcheek Darter in Arkansas. Copeia, 103(2), 433-439.

Martin, P. R., & Martin, T. E. (2001). Behavioral interactions between coexisting species:

song playback experiments with wood warblers. Ecology, 82(1), 207-218.

Matthews, W. J., & Turner, T. F. (2019). Redescription and Recognition of Etheostoma

cyanorum from Blue River, Oklahoma. Copeia, 107(2), 208-218.

Meffe G.K., & Vrijenhoek, R. C. 1988. Conservation genetics in the management of

desert fishes. Conservation Biology, 2,157−169.

Mettee, M. F., O'Neil, P. E., & Pierson, J. M. (1996). Fishes of Alabama and the Mobile

basin. Oxmoor House.

Meyer, J. L., Strayer, D. L., Wallace, J. B., Eggert, S. L., Helfman, G. S., & Leonard, N.

E. (2007). The contribution of headwater streams to biodiversity in river

networks. American Water Resources Association 1, 86-103.

Miller, D. A., Talley, B. L., Lips, K. R., & Grant, E. H. C. (2012). Estimating patterns

and drivers of infection prevalence and intensity when detection is imperfect and

sampling error occurs. Methods in Ecology and Evolution, 3(5), 850-859.

McNyset, K. M. (2009). Ecological niche conservatism in North American freshwater

fishes. Biological Journal of the Linnean Society, 96(2), 282-295.

Mississippi Department of Water Quality (MDWQ). 2015. State Well Depth Data.

Unpublished raw data.

Near, T. J., Bossu, C. M., Bradburd, G. S., Carlson, R. L., Harrington, R. C.,

Hollingsworth Jr, P. R., & Etnier, D. A. (2011). Phylogeny and temporal

diversification of darters (Percidae: Etheostomatinae). Systematic Biology, 60(5),

565-595.

Ovidio, M., Detaille, A., Bontinck, C., & Philippart, J. C. (2009). Movement behaviour of

the small benthic Rhine sculpin Cottus rhenanus (Freyhof, Kottelat & Nolte,

2005) as revealed by radio-telemetry and pit-tagging. Hydrobiologia, 636(1), 119-

128.

Peoples, B. K., & Frimpong, E. A. (2016). Biotic interactions and habitat drive positive

co‐occurrence between facilitating and beneficiary stream fishes. Journal of

Biogeography, 43(5), 923-931.

Petty, J. T., & Grossman, G. D. (2004). Restricted movement by mottled sculpin (Pisces:

Cottidae) in a southern Appalachian stream. Freshwater Biology, 49(5), 631-645.

Peres-Neto, P. R. (2004). Patterns in the co-occurrence of fish species in streams: the role

of site suitability, morphology and phylogeny versus species

interactions. Oecologia, 140(2), 352-360.

Piller, K. R., Bart Jr, H. L., & Walser, C. A. (2001). Morphological variation of the

Redfin Darter, Etheostoma whipplei, with comments on the status of the

subspecific populations. Copeia, 2001(3), 802-807.

Poff, N. L., & Allan, J. D. (1995). Functional organization of stream fish assemblages in

relation to hydrological variability. Ecology, 76(2), 606-627.

Poole, G. C. (2002). Fluvial landscape ecology: addressing uniqueness within the river

discontinuum. Freshwater Biology, 47(4), 641-660.

Porter, B. A., Cavender, T. M., & Fuerst, P. A. (2002). Molecular phylogeny of the

snubnose darters, subgenus Ulocentra (genus Etheostoma, family

Percidae). Molecular Phylogenetics and Evolution, 22(3), 364-374.

Resetarits Jr, W. J. (1997). Interspecific competition and qualitative competitive

asymmetry between two benthic stream fish. Oikos, 429-439.

Rice, S. P., & Church, M. (2001). Longitudinal profiles in simple alluvial systems. Water

Resources Research, 37(2), 417-426.

Richmond, O. M., Hines, J. E., & Beissinger, S. R. (2010). Two‐species occupancy

models: a new parameterization applied to co‐occurrence of secretive

rails. Ecological Applications, 20(7), 2036-2046.

Robinson, H. W., & Buchanan, T. M. (1988). Fishes of Arkansas. University of Arkansas

Press, Fayetteville.

Rosenberry, D. O., LaBaugh, J. W., & Hunt, R. J. (2008). Use of monitoring wells,

portable piezometers, and seepage meters to quantify flow between surface water

and ground water. Field Techniques for Estimating Water Fluxes Between Surface

Water and Ground Water, 4, 39-70.

Ross, S. T. (2013). Ecology of North American freshwater fishes. Univ of California

Press.

Scalet, C. G. (1973). Stream movements and population density of the Orangebelly

Darter, Etheostoma radiosum cyanorum (Osteichthyes: Percidae). The

Southwestern Naturalist, 381-387.

Schaefer, J. F., Clark, S. R., & Warren Jr, M. L. (2012). Diversity and stability in

Mississippi stream fish assemblages. Freshwater Science, 31(3), 882-894.

Schmidt, B. V., & Schaefer, J. (2018). Ecological and landscape effects on genetic

distance in an assemblage of headwater fishes. Ecology of Freshwater

Fish, 27(2), 617-631.

Schoener, T. W. (1974). Resource partitioning in ecological

communities. Science, 185(4145), 27-39.

Schoener, T.W. (1988). Ecological interactions. In A. A. Myers & P.S. Giller (Eds.)

Analytical Biogeography, an Integrated Approach to the Study of Animal and

Plant Distributions (pp. 255–297). New York: Chapman and Hall.

Schluter, D. (2000). The ecology of adaptive radiation. OUP Oxford.

Skalski, G. T., & Gilliam, J. F. (2000). Modeling diffusive spread in a heterogeneous

population: a movement study with stream fish. Ecology, 81(6), 1685-1700.

Smiley Jr, P. C., Dibble, E. D., & Schoenholtz, S. H. (2006). Spatial and temporal

variation of Goldstripe Darter abundance in first-order streams in north-central

Mississippi. The American Midland Naturalist, 156(1), 23-36.

Sterling, K. A., Warren Jr, M. L., & Henderson, L. G. (2013). Conservation assessment

of the Yazoo Darter (Etheostoma raneyi). Southeastern naturalist, 816-842.

Sterling, K. A., & Warren Jr, M. L. (2017). Microhabitat estimation of an imperiled

headwater fish, the Yazoo Darter (Etheostoma raneyi), in Coastal Plain

streams. Environmental Biology of Fishes, 100(10), 1223-1235.

Sukhikh, N., Souissi, A., Souissi, S., Holl, A. C., Schizas, N. V., & Alekseev, V. (2019).

Life in sympatry: coexistence of native Eurytemora affinis and invasive

Eurytemora carolleeae in the Gulf of Finland (Baltic Sea). Oceanologia, 61(2),

227-238.

Suttkus, R. D., Bailey, R. M., Bart, H. L., Warren, M. A., Cashner, R. C., Knight, C. L.,

& Hastings, R. W. (1994). Three New Species of Etheostoma, Subgenus

Ulocentra, from the Gulf Coastal Plain of Southeastern United States: Fishes of

the Buffola River System, Wilkinson County, Southwestern Mississippi. Tulane

University.

Taylor, C. M. (1996). Abundance and distribution within a guild of benthic stream fishes:

local processes and regional patterns. Freshwater Biology, 36(2), 385-396.

Taylor, C. M. (1997). Fish species richness and incidence patterns in isolated and

connected stream pools: effects of pool volume and spatial

position. Oecologia, 110(4), 560-566.

Thompson, K. W., & Muncy, R. J. (1986). Darters of the Little Tallahatchie watershed in

northern Mississippi. Journal of Mississippi Academy of Sciences, 31, 63-77.

Townsend C.R. & Crowl T.A. (1991). Fragmented population structure in a native New

Zealand fish: an effect of introduced Brown Trout? Oikos, 61, 347–354.

Troia, M. J., & Gido, K. B. (2014). Towards a mechanistic understanding of fish species

niche divergence along a river continuum. Ecosphere, 5(4), 1-18.

United States Environmental Protection Agency, United States Geological Survey.

(2006). The national hydrography dataset plus. http://www.horizon

systems.com/NHDPlus/index.php. Accessed 1 May 2015.

Vance, R. R. (1972). Competition and mechanism of coexistence in three sympatric of

intertidal Hermit Crabs. Ecology, 53(6), 1062-1074.

Vergara, M., Cushman, S. A., Urra, F., & Ruiz-González, A. (2016). Shaken but not

stirred: multiscale habitat suitability modeling of sympatric marten species

(Martes martes and Martes foina) in the northern Iberian Peninsula. Landscape

Ecology, 31(6), 1241-1260.

Wagner, T., Deweber, J. T., Detar, J., & Sweka, J. A. (2013). Landscape-scale evaluation

of asymmetric interactions between brown trout and brook trout using two-species

occupancy models. Transactions of the American Fisheries Society, 142(2), 353-

361.

Wiens, J. A. 2002. Riverine landscapes: taking landscape ecology into the

water. Freshwater Biology 47(4): 501-515.

CHAPTER III CONFLUENCES AND LAND COVER AS AGENTS OF CHANGE:

TEMPORAL HABITAT VARIABILITY MODIFIES THE MOVEMENT AND

ASSEMBLAGE DYNAMICS OF HEADWATER FISHES

Abstract

Discerning the environmental and anthropogenic factors that alter species dispersal rate, thus influencing metacommunity dynamics, is key to improving our understanding of how species turnover is regulated. In this study, we use headwater fish assemblages as a model system to assess the influences of confluence size and land cover on changes to in- stream habitat, movement behavior, movement rate, and assemblage structure. Sites were paired based on confluence size ratio (> 0.6 <) and upstream land cover (urban or forested).

We established three reaches at each confluence and then used mark-recapture methods to estimate changes in movement behavior and movement rate within each reach. Three functional groups were used to assess changes to movement rate and behavior associated with confluence size or land cover. Further, we used count data collected during mark- recapture events to calculate Bray-Curtis Distances which served as a proxy of assemblage change within our study reaches. To accomplish these objectives, we used multivariate analyses and non-linear, mixed effects models to examine temporal changes to habitat, movement behavior, movement rate, and assemblage dynamics. Generally, changes to habitat, movement behavior, movement rate, and assemblage structure at headwater confluences are chiefly a consequence of land cover. Patch stability, associated with land cover, altered the movement behavior and distances travelled by water-column specialists.

Changes in movement rate and assemblage structure were a consequence of differences in woody structure availability. Our understanding of the extent to which land cover alters the

geomorphic and ecological gradients associated with confluences in dendritic networks will be critical to ensure the conservation of sensitive species whose fitness is dependent on the integrity of these habitats.

Introduction

Dispersal is a fundamental principle integral to species survival because it influences colonization, gene flow, and resource use. The dispersal ability of a species is a function of its probability of movement between suitable habitat patches and the barriers that separate them (Levins, 1966; Wiens, 2002). Habitat fragmentation limits the dispersal of species by disrupting the connectivity and structure of habitat patches, and is a symptom of natural and anthropogenic disturbance (Fagan, 2002; Brown et al., 2001). Because organisms move to complete their life histories, even a small degree of fragmentation across an ecosystem can be detrimental to their survival. In dendritic stream networks

(Campbell-Grant, Lowe & Fagan, 2007), random habitat fragmentation increases the number of patch fragments, variance in patch size (Fagan, 2002) and distance between patches, thus diminishing dispersal of resident species (Roberts & Angermeier, 2007).

Confluences are areas of active habitat change and geomorphic activity, amplifying the effects of local disturbance (Rice, Greenwood & Joyce, 2001; Benda et al., 2004). The magnitude of geomorphic change at confluences in small headwater drainages is characterized by high flow events, which transport sediment and woody structure (Johnson

& Rodine, 1984; Hogan, Bird & Hassan, 1998). High flow events scour sediment and larger substrates and transports wood downstream, forming depositional areas such as gravel bars and alluvial fans (Benda & Cundy, 1990; Benda, Veldhuisen & Black, 2003). Sediment deposition at confluences induces predictable, localized, geomorphic responses in

mainstem channels (Benda et al., 2004). For example, decreasing sediment transport at confluences should facilitate reductions in upstream channel gradient and substrate size, while increasing channel meandering and floodplain width in the mainstem. Such changes are counterbalanced in the downstream reach of the mainstem with increases to channel gradient, channel width, substrate size, pool depth, and bar occurrence. Confluence density in low order streams influences the regularity of discontinuities in channel morphology

(Rice 1998; Benda et al., 2004). Thus, confluences may serve as biological hotspots (see

McClain et al., 2003) through increases in water and sediment input that will ultimately facilitate downstream habitat heterogeneity (Rice & Macklin, 2008).

Confluence size (i.e., the ratio between tributary size and mainstem size) and network geometry contribute substantially to the extent of geomorphology at confluences. Across all drainage sizes, discharge-related morphological changes (i.e., channel width, depth) occur at confluences where the ratio between tributary size and main stem size approaches

0.6 to 0.7 (Rhoads, 1987), and confluences that approach this ratio are referred to as being

“geomorphically significant” (Benda et al., 2004). Local network geometry describes the angle at which tributaries intersect the mainstem. Generally, as a confluence angle approaches 90º, the probability of certain geomorphic effects increases (Mosley 1976; Best

1986); however, in headwaters, such effects may still be promoted when a confluence angle surpasses 70º (Benda & Cundy, 1990).

Urban and agricultural land cover fragment stream habitats (Leitäo et al., 2018), and have well documented deleterious effects on aquatic ecosystems (Wang, Lyons, Kanehl &

Bannerman, 2001; Allan, 2004; Leal et al., 2016). Within a watershed, habitat fragmentation at small spatiotemporal scales is intimately tied to land use at coarser

spatiotemporal scales (Wang, Lyons, Kanehl & Bannerman, 2001). For example, increases in agricultural and urban land cover degrades water quality and alters channel hydrology of stream segments (Allan, 2004). Thus, broad anthropogenic impacts reduce the connectivity, stability, and diversity of habitat patches at smaller scales (Padgham & Webb,

2010). Because of such fragmentation, anthropogenic disturbance disrupts gene flow

(Stow, Sunucks, Briscoe & Gardner, 2001) while also degrading regional emergent properties such as species richness among local aquatic communities (Perkins & Gido,

2012) by the elimination of dispersal pathways. However, the extent to which land cover modifies the role of confluences as local agents of habitat change, and therefore, altering the dispersal ability of aquatic biota, is poorly understood.

Stream fish assemblages provide an excellent system for studying dispersal as a function of spatial scale. Understanding patterns of individual stream fish movement has proven to be difficult due to tag loss and low rates of recapture among highly mobile species (Albanese, Angermeier & Doraj-Raj, 2004; Gowan & Fausch, 1996; Rodriguez,

2002). Thus, multiple competing conceptual frameworks such as the restricted movement paradigm (Gerking, 1959), long distance dispersal (Rodriguez, 2010) and confluence exchange hypothesis (Thornbrugh & Gido, 2010) were developed to model movement ecology of stream fishes. Although much work has been done to better understand the influences of habitat variability at the patch (Smithson & Johnston, 1999; Belica & Rahel,

2008; Clark & Schaefer, 2016) and reach scales (Albanese, Angermeier & Doraj-Raj, 2004,

Walker & Adams, 2016), our understanding of how confluences regulate rates of dispersal as a consequence of the branching connectivity of streams remains relatively unclear. Much indirect evidence (Gorman, 1986; Hitt & Angermeir, 2008; Thornbrugh & Gido, 2010)

suggests that confluences perform an important role in regulating stream fish movement, but few studies (Cathcart, Gido & McKinstry, 2015; Cathcart, Gido, McKinstry &

MacKinnon, 2018) have directly investigated their influence. Overall, our understanding of the interactive effects of land cover and confluences on movement of resident species is unknown.

In this study, we examine habitat structure, dispersal, and stream fish assemblages at four headwater confluences that differed in size and surrounding land cover. We used mark-recapture methods to assess movement and count data to examine assemblage dynamics. We applied a multi-scale approach to test which habitat characteristics best described movement and assemblage change across confluence size or land cover factors.

We also assessed how differences in movement behavior among patches (i.e., riffles and pools) may change because of these two factors. We predicted that 1) habitat change across the confluence should be most extensive within urban reaches characterized by a confluence size > 0.6, 2) differences in movement behavior among patches are related to both land cover and confluence size, 3) variations in rate of movement and assemblage change are associated with distinct changes to habitat characteristics mediated by land cover and confluence size, 4) urban sites will be distinguished by an increased rate of movement and homogeneous habitat and assemblages, and 5) rate of movement should increase at sites characterized by a larger confluence size as a consequence of increasing geomorphic change.

Methods

Study System

We conducted our study in the Pascagoula River drainage, which is positioned within

the Gulf Coastal Plain province, and is the largest unimpounded river system in the contiguous United States (Dynesius & Nilsson, 1994). Our study sites were located within the Southern Pine Plains and Hills ecoregion within the province. Streams that drain this ecoregion tend to be either alluvial or blackwater systems and are characterized by low to moderate gradients and gravel, sand and clay substrates (Hupp, 2000).

Study Design

Each of our four sites was located at the confluence of two headwater streams. Of the two upstream reaches, we identified the reach with the larger upstream drainage area as the mainstem (Figure 3.1). We then delineated three contiguous 100 m reaches containing alternating riffle and pool patches (hereafter habitat) so that our mark-recapture data would capture movement at the reach and patch scales. We established two factorial levels for land cover (urban, forested) and confluence size (>0.61, < 0.60) for each of the four sites.

Confluence size ratios are calculated by dividing the drainage area of the smaller stream by the drainage area of the larger stream, and a threshold of > 0.6 represents the lowest value at which geomorphic change may be observed because of confluence size (Benda et al., 2004). We estimated confluence size for each site using the National Hydrography Data

Plus (NHDPlus) (USEPA &USGS 2016) set in ArcGIS (version 10.0, ESRI). We also measured stream gradient and sinuosity using the NHDPlus dataset and USGS national elevation dataset (https://www.nrcs.usda.gov/wps/portal/nrcs/detail/national/). Sites were only chosen if their confluence angle was < 60º to diminish local geomorphic effects that were a symptom of network geometry. The 2016 national land cover dataset (NLCD, https://www.mrlc.gov/data) was used to estimate the proportions of urban and forested land cover within each site’s upstream catchment. Our urban sites (Priest Creek and Mixon’s

Figure 3.1 Diagram of site layout. Each site was located at a confluence and consisted of three 100 m reaches. Pools (P) and riffles (R) were delineated as patches of habitat.

Creek, > 50% urban land cover) were located within the city limits of Hattiesburg, MS, and our forested sites (Garraway Creek and Sweetwater Creek, 100% forested land cover) were located within De Soto National Forest (Figure B1).

Functional Groups

We established three functional groups from fishes encountered at the study sites. The benthic specialists included the Brighteye Darter Etheostoma lynceum, the Gulf Darter E. swaini, and the Blackbanded Darter Percina nigrofasciata, water-column specialists included the Blacktail Shiner Cyprinella venusta, the Cherryfin Shiner Lythrurus. roseipinnis, the Rough Shiner Notropis baileyi, and the Flagfin Shiner Pteronotropis signippinis, and structure specialists included the Bluegill Lepomis macrochirus the

Longear Sunfish L.megalotis, the Green Sunfish L. cyanellus and the Shadow Bass

Ambloplites arriomus.

Study-Period

We treated repeated visits to reaches within a site as replicates. Initially, we marked fish in August of 2017 and May of 2018; subsequent mark-recapture events occurred at 4 to 6 week intervals (7 marking events total). We conducted six recapture events at all four sites across two field seasons; two in the fall of 2017 and four in the spring-summer of

2018. However, flooding in the fall of 2017, resulted in a bridge failure that restricted access to one site. We replaced this site with a new site prior to beginning sampling in the summer of 2018. Thus, we conducted six recapture events at three sites, but only four recapture events were performed at the fourth site. We used non-parametric bootstraps

(n=10,000) with 95% confidences intervals (CIs) to determine whether there was a significant, effect of site on our response variables (e.g., movement rate, assemblage

change) dependent on whether the sample sizes for three (6 recapture events * 3 reaches *

3 sites = 54 replicate reaches) or four sites (63 replicates) were used to estimate the random- intercept. Because 95% CIs (Table B1) for both parameter estimates overlapped, we assumed that there was not a significant effect of site, despite the unbalanced design.

Mark-Recapture Methods

At each riffle or pool we made three passes using a 4.8 mm mesh seine and a ETS

Badger 1 backpack electrofishing unit using pulsed DC (range: 100 to 250 V). All darters

> 40 mm standard length (SL), all sunfish > 80 mm in SL, and all shiners > 40 mm in SL were marked. We gave all fish habitat unit, batch-specific marks using visual implant elastomer (VIE; Northwest Marine Technology, Inc., Shaw Island, WA, USA). Small bodied minnows (Bangs et al., 2013; Neufeld, Blair & Poesch, 2015), darters (Roberts &

Angermeier, 2004), and centrarchids (Laux, Koupal & Hoback, 2007) often exhibit high retention of VIE tags over extended periods. We used three colors and six body locations to distinguish habitat units (range of codes =7 to 13). Prior to marking, we anesthetized fish with tricaine methanesulfonate. Initially, we marked fish to identify their initial habitat unit of capture. During consecutive mark-recapture events, we gave fish which displayed movement an extra mark unique to the new habitat unit from which they were captured.

We released all marked fish that displayed no movement back into their original habitat unit of capture. After marking, we placed fish into live jars for 30 minutes to allow for recovery from anesthesia. Following recovery, we released fish in low velocity microhabitats near the midpoint of the habitat unit from which they were sampled. We used the midpoint distance of the original habitat unit of capture and habitat unit of recapture to establish the beginning and end points of the total distance moved by a fish for all pairs of

59 Table 3.1

Percent recapture rates for trips, years, and between years for three functional groups of fishes (WCS = Water Column Specialists; BS= Benthic Specialists; SS = Structure Specialists) at each mark-recapture site. (Btw=Between). Site Land Confluence Trip Year Days Btw Total WCS % BS % SS% cover Ratio Size Trips Recapture % Garraway Forested >0.6 1 2017 60 15.3% 15.3% 0% 0% Creek 2 2017 30 14.8% 18.4% 0% 0% Mean 2017 45 15.0% 16.9% 0% 0%

Btw 192 1.6% 1.4% 4.5% 0% Yrs.

4 2018 21 10.6% 12.0% 10.0% 0% 5 2018 21 17.5% 19.0% 11.1% 0% 6 2018 30 14.2% 16.8% 0% 0% Mean 2018 24 14.1% 15.9% 7.0% 0%

Priest Urban <0.6 1 2017 92 9.8% 6.8% 22.2% 27.8% Creek 2 2017 42 16.5% 17.2% 20.0% 7.7% Mean 2017 67 13.2% 12.0% 21.2% 17.8%

Btw 174 3.0% 7.5% 11.7% 15.4% Yrs.

4 2018 30 15.4% 10.0% 22.2% 23.3% 5 2018 30 39.4% 26.8% 45.7% 63.4% 6 2018 30 38.0% 56.2% 11.5% 35.7% Mean 2018 30 30.9% 31.0% 26.5% 40.8%

Table 3.1 (continued) Site Land Confluence Trip Year Days Btw Total WCS % BS % SS% cover Ratio Size Trips Recapture % Mixon’s Urban >0.6 1 2017 57 7.2% 0% 0% 9.6% Creek 2 2017 56 23.3% 22.2% 4.3% 30.0% Mean 2017 56.5 15.3% 11.1% 2.2% 19.8%

Btw 2018 183 3.9% 9.0% 0% 17.3% Yrs.

4 2018 30 19.1% 15.0% 0% 25.3% 5 2018 30 9.3% 4.2% 12.5% 13.0% 6 2018 30 19.2% 9.1% 0% 22.0% Mean 2018 30 15.9% 9.4% 4.2% 20.1%

Sweetwater Forested <0.6 1 2018 40 9.1% 10.1% 5.3% 0% Creek 2 2018 21 45.3% 51.0% 37.5% 0% 3 2018 31 24.6% 24.6% 0% 0% Mean 2018 31 26.3% 28.6% 14.3% 0%

patches. Because we used batch markings, our data only captured the distance moved by a marked fish between sampling events, and not for the duration of the study period; thus our ability to deduce patterns in fish movement was restrained to these shorter time intervals.

Because recapture rates of benthic specialists and structure specialists were extremely variable among sites (Table 3.1), we did not analyze differences in movement behavior or movement rate for these two groups. Our first recapture event of the 2018 field season allowed us to estimate the percentage of fishes recaptured between our two field seasons.

At the end of the 2018 field season, we performed a long-distance recapture event (1 – 2 river km) for each site to identify any individuals that may have moved outside of the study- reaches.

Habitat Sampling

Following each sampling event, we collected habitat data along 10 evenly spaced transects perpendicular to flow within each of the three stream reaches at a site. At one meter intervals along each transect we measured depth (m), current velocity (m·s⁻¹), substrate size (modified Wentworth scale, six categories; Cummins, 1962), and the presence (binary variable) of small wood (<10 cm diameter or <1.5 m in length), and large wood (>10 cm diameter or>1.5 m in length). We also measured the wetted and bankfull widths (m) for each transect. Substrate size represented the mean size of alluvial material at a site. We condensed the percentage occurrence of small and large wood into one variable (woody structure) because the two variables were highly correlated (r > 0.5). We calculated mean, reach-scale differences to quantify the degree of change in each habitat variable between sampling events. We then divided habitat variable mean differences by the number of days between sampling events to account for differences in time between

mark-recapture events.

Analysis of Habitat Change

We used principal components analysis (PCA) to reduce the dimensionality of habitat variables using a correlation matrix. Using the broken stick method (Legendre & Legendre,

1983), we retained PCA scores for the first two axes (PC1 and PC2). We used non- parametric multivariate analysis of variance (NP-MANOVA) with Euclidean distance matrices to test for effects of confluence size and land cover among reach PC1 and PC2 scores. NP-MANOVA is considered to be more robust than traditional MANOVA because it is permuted and partitions variance among groups on the basis of any symmetric distance or dissimilarity measure. P-values were estimated based on 1000 permutations. To test our prediction that urban reaches characterized by a confluence size > 0.6 (i.e., Mixon’s Creek reaches) would exhibit the most habitat change, we performed one-way ANOVAs on the reach PC1 and PC2 scores with site as a main effect. We then corrected for multiple comparisons using pairwise t-tests and the Bonferroni method.

Analysis of Movement Behavior

We examined differences in movement behavior by assessing the influence of land cover or confluence size on the frequency distributions of distances travelled and move type (i.e., movement between discreet patches). We summed the number of recaptures for each species that exhibited a specific move type (e.g., pool A to riffle B) at each site for all mark-recapture events (N=51). To assess the effect of confluence size and land cover on movement behavior, we then assigned move types using two criteria, 1) whether a move was or was not a confluence move, and 2) the sequential habitat unit history of the fish

(e.g., riffle to pool). In total, we defined eight levels for move type (Table 3.2). We also

Table 3.2

Numbers of movers by move-type for all four mark-recapture sites across the study period. Site Land cover Confluence Move-Type N Ratio Size Garraway Forested > 0.6 Pool to Pool 6 Creek Pool to Riffle 18 Pool to Riffle via Confluence 1 Riffle to Pool 14 Riffle to Pool via Confluence 2 Pool to Pool via Confluence 0 Riffle to Riffle 0 Riffle to Riffle via Confluence 0 Mixon’s Urban > 0.6 Pool to Pool 1 Creek Pool to Riffle 0 Pool to Riffle via Confluence 1 Riffle to Pool 0 Riffle to Pool via Confluence 1 Pool to Pool via Confluence 7 Riffle to Riffle 0 Riffle to Riffle via Confluence 0 Priest Urban < 0.6 Pool to Pool 9 Creek Pool to Riffle 12 Pool to Riffle via Confluence 2 Riffle to Pool 10 Riffle to Pool via Confluence 3 Pool to Pool via Confluence 7 Riffle to Riffle 4 Riffle to Riffle via Confluence 6 Sweetwater Forested < 0.6 Pool to Pool 2 Creek Pool to Riffle 9 Pool to Riffle via Confluence 1 Riffle to Pool 7 Riffle to Pool via Confluence 4 Pool to Pool via Confluence 1 Riffle to Riffle 7 Riffle to Riffle via Confluence 2

64 examined the influence of distance moved and land cover or confluence size on the frequency distribution of move types. We used Poisson regression (r-package “lmer”) to test our hypotheses that differences in the distributions of move-type and distance moved would differ as a result of either land cover or confluence size. In our models, we included species as a nested random effect within site.

Estimation of Movement Rates and Assemblage Change

We estimated rates of movement and assemblage change for reaches between sampling events. Sample sizes for estimates of movement (N =54) and assemblage change (N=63) differed because we did not recapture any individuals in some reaches during a sampling event. Thus, for a given sampling event, reaches in which we did not recapture any individuals were subsequently dropped from our movement dataset prior to any analyses.

Because the number of water-column specialists marked within these reaches was usually small (14 ± 19 individuals), we are confident that any inability to recapture individuals from these reaches was not a result of poor sampling efficiency but rather an effect of low fish abundance. We estimated proportional daily movement rates (PDMR) at the reach scale using the formula M *R-¹ D-1, where M is the number of fish that moved, R is the total number of recaptures within a reach, and D is the number of days since the first marking period of the field season (Warren & Pardew, 1998). We estimated rates of reach movement separately for 2017 and 2018 movement data because low recapture rates (1.6%. see Table 3.1) suggested that mark loss or mortality (electrofishing or natural) was high between years. From count data, we calculated the relative abundance of each functional group within all reaches during each sampling event and then estimated reach scale,

assemblage change between sampling events and across all functional groups using the

Bray-Curtis distance of dissimilarity.

Analysis of Reach Scale Movement Rates and Assemblage Change

We modeled movement and assemblage change using Beta-regression and Zero-

Inflated Beta regression models (r-package, “gamlss”) because, preliminary analyses suggested that movement rates and Bray-Curtis distances were not normally distributed.

As an expansion of generalized linear models, Beta regression models can evaluate the effects of explanatory variables for proportional and probability data falling within the interval (0, 1). In addition, Zero-Inflated Beta regression models allow the user to account for overdispersion in the response variable by modeling zeros in a dataset as a function of a two-state process (Liu & Eugenio, 2018). One of the two states, the zero state, may be defined as the probability of an event being so low that it cannot be readily differentiated from zero. The second state, the normal state, includes both zeros and continuous values falling within the interval (0, 1) (Liu & Eugenio, 2018). Prior to the inclusion of any habitat covariates, we assessed whether the addition of “site” as a random intercept term increased model fit. For both response variables, random-intercept, null models were ranked higher than null models (PDMR: wᵢ = 0.76; Bray-Curtis Distance: wᵢ = 0.74) suggesting that reach movement rates varied as a consequence of site effects, justifying the inclusion of a random intercept term.

We constructed both reach and multi-scale models (Table B2). We used reach scale models to identify habitat variables at this spatial extent that best explained variation in movement rates and Bray-Curtis distances. We standardized all habitat variables by subtracting the mean and dividing by twice the standard deviation. Prior to modeling, we

tested for the correlation between covariates. For pairs with a Pearson correlation greater than the absolute value of 0.5, we dropped one of the two covariates. We included a null model and global model in our reach scale model sets. In total, we constructed 23 models to test our hypotheses on movement rate and assemblage change at the reach scale. Multi- scale model sets were configured so that both coarse scale factors, confluence ratio size and land cover, could be integrated into our best reach scale models as both additive and interactive effects. Our multi-scale model sets, also featured the best reach scale model without land cover or confluence size included as additive or interactive effects. For both reach scale and multi-scale model sets, we only incorporated two-way interactions. We excluded three-way and higher interaction terms because our capacity to derive inferences from such model structures would be problematic. We used Akaike’s Information Criterion for small sample sizes (AICC) to assess the quality of competing models (Burnham &

Anderson, 2002). Models with small ∆AICC and large Akaike weights (wᵢ) indicate the greater parsimony (Burnham & Anderson, 2002). We only used reach scale models with wᵢ > 0.10 to construct multi-scale models.

Results

Recapture Rates

Over the duration of the study period, we marked 2459 fish (average of 614 at each site) which varied between sites (Sweetwater Creek: 334; Garraway Creek: 675; Mixon’s

Creek: 717, Priest Creek: 733). Recapture rates varied by functional group, site, trip, year, and between years (Table 3.1). The total number of recaptures across all sites were higher in 2018 than in 2017 (Mean + 95% CI; 2017: 28.67 + 4.19, 51.14; 2018: 75.33 + 0, 163.98), but there was extensive overlap between 95% confidence intervals. The mean percentage

of individuals recaptured between years was lower than within years at all sites (Table 3.1).

Our greatest yearly recapture rates for each functional group were all achieved during our

2018 field season at Priest Creek. We recaptured 34.5% of marked structure specialists,

30.5% of marked water-column specialists, and 26.2% of marked benthic specialists at this site (Table 3.1). Proportionally, across all sites, 61% (152 individuals), of water-column specialists were recaptured in pools, 63% (37 individuals) of benthic specialists were recaptured in riffles, and 93.6% (103 individuals) of structure specialists were recaptured in pools.

Habitat Change

The first two components from the PCA explained 50% of the variance in habitat variables among reaches across time (Figure. 3.2, Table B3). The PC1 axis described a gradient driven primarily by mean difference in depth (1.32) and mean difference in woody structure (1.15). The PC2 axis described a gradient driven primarily by mean difference in flow (1.04) and mean difference in wetted width (0.96). There was a significant main effect of land cover (pseudo R² = 15.9%, pseudo F1,62,= 10.4, p = 0.0001) on habitat change.

There was not a significant main effect of confluence size on habitat change. The interaction between confluence size and land cover was not significant (p > 0.05). Results for one-way ANOVAs indicated that there was not a significant effect of site on PC1 scores

(p > 0.05); however, there was a significant effect of site on PC2 scores (F = 7.283,59, p =

0.0003). In association with this result, pairwise comparisons indicated that Mixon’s Creek reaches were significantly different from Garraway Creek reaches (p = 0.001), and

Sweetwater Creek reaches (p = 0.01). Also, Priest Creek reaches were significantly different from Garraway Creek reaches (p = 0.01), but not Sweetwater Creek reaches

Figure 3.2 Principal component analysis (PCA) of in-stream habitat variables for PC1 and PC2 axes. Arrows end at centroids for environmental variables.

(p > 0.05). There was no evidence that Mixon’s Creek reaches were significantly different from Priest Creek reaches (p > 0.05) or that Garraway Creek reaches were significantly different from Sweetwater Creek reaches (p > 0.05).

Movement Behavior

The distribution of distances travelled by water-column specialists varied among sites

(Mean ± SD; Garraway Creek: 53.7 m ± 22.3; Mixon’s Creek: 84.8 m ± 23.5; Priest Creek:

39.9 m ± 14.9; Sweetwater Creek: 53.5 m ± 17.6, Figure 3.3). Generally, water-column specialists appeared to frequently move greater distances at urban sites (e.g., Mixon’s

Creek and Priest Creek), shorter distance moves were more common at forested sites (e.g.,

Garraway Creek and Sweetwater Creek) (Figure 3.3). The minimum (18.25 m) and maximum observed distances (113.5 m) moved by water-column specialists were recorded at urban sites (Table 3.3). The mean, minimum, and maximum observed distances moved by water-column specialists at urban sites were highly dissimilar. (Table 3.3). Conversely, mean, minimum, and maximum observed distances moved by water-column specialists at forested sites were similar (Table 3.3). No long distance moves outside of our study reaches were detected. There was a significant interaction between distance moved and urban land cover (beta = 0.03, p = 0.001) on the number of movers. Distance (-0.03, p = 0.006), and urban land cover (-1.67, p = 0.0004) both exhibited negative, significant main effects on the number of movers. There was a weak significant, interaction between confluence size

< 0.6 and distance moved on the number of movers (0.02, p = 0.02). There were also significant, negative main effects of confluence size (-1.55, p = 0.0005) and distance (-0.02

, p = 0.0001) on the number of movers.

N=41 N=10

N=53 N=33

Figure 3.3 Frequency distributions in relation to distance moved for each mark-recapture site in the Pascagoula River basin, MS.

Table 3.3

Summary statistics for distances moved by water-column specialists within and among reaches at all four sites. Dist= Distance. Site Land Confluence Mean ± SD Min. Max. Dist. Cover Ratio Size Distance (m) Dist. (m) (m) Garraway Creek Forested > 0.6 53.68 ± 22.30 26.3 88 Mixon’s Creek Urban > 0.6 84.8 ± 23.54 56.5 113.5 Priest Creek Urban < 0.6 34.94 ± 14.93 18.25 70.5 Sweetwater Forested < 0.6 53.53 ± 17.62 23.71 82.0 Creek

Pool to riffle (n=39) and riffle to pool (n=31) were the most common move types observed throughout the study period (Table 3.2). The frequency of all confluence move- types was greater at urban sites (Mixon’s Creek: n= 9; Priest Creek: n= 18) than at forested sites (Garraway Creek: n=3, Sweetwater Creek: n=8). Furthermore, at urban sites, the number of pool to pool moves, independent of whether an individual crossed a confluence, were much greater (n=24) at urban sites than at forested sites (n=9). Conversely, the number of riffle to pool or pool to riffle moves, independent of whether an individual crossed a confluence, were much greater at urban sites (n=56) than at forested sites (n=23).

There was a significant effect of land cover on the number of water-column specialists that displayed riffle to riffle moves (beta = -1.63; p = 0.04). There was also a significant, main effect of move type; the number of pool to riffle (1.33, p = 0.001), riffle to pool (1.08, p =

0.01), and riffle to riffle (1.18; p = 0.02) moves displayed by water-column specialists were significantly different from all other move types. There was not a significant interaction between confluence size and move-type or significant main effect of confluence size on the frequency distribution of move types displayed by water-column specialists.

Movement Rates

Movement rates were similar among sites (Mean ± SD; Garraway Creek: 0.009 ±

0.012; Mixon’s Creek: 0.007 ± 0.009; Priest Creek: 0.006 ± 0.008; Sweetwater Creek:

0.007 ± 0.009). Mean reach movement rates were similar for forested (0.009 ± 0.011) and urban (0.007 ± 0.008) reaches. Mean reach movement rates were also similar independent of confluence size (>0.6: 0.008 ± 0.011; <0.6: 0.007 ± 0.007). Cumulative movement rates were similar for forested reaches (20.2%) and urban reaches (18.8%), and cumulative movement rates for reaches typified by a confluence ratio size >0.6 were similar (20.2%)

to the cumulative movement rates of reaches characterized by a confluence ratio size <0.6

(19.4%).

Assemblage Change

Throughout the study-period, we sampled 776 individuals at Garraway Creek, 458 individuals at Sweetwater Creek, 823 individuals at Priest Creek, and 802 individuals at

Mixon’s Creek. Mean species richness was highest at our urban sites (Mean + 95% CI;

Priest = 6.86; 6.24, 7.47; Mixon’s Creek = 4.57; 2.8, 6.35) and lowest at our forested site

(Sweetwater Creek = 4.08; 0.66, 7.5; Garraway Creek = 4.24; 3.16, 7.89) but confidence intervals broadly overlapped. Mean relative abundance of water-column specialists was greatest at Garraway Creek (0.77; 0.68, 0.85) and Sweetwater Creek (0.67; 0.59, 0.76) and lowest at Priest Creek (0.49; 0.40, 0.58) and Mixon’s Creek (0.27; 0.18, 0.35). Mean relative abundance of benthic specialists was greatest at Priest Creek (0.26; 0.20, 0.32) and

Sweetwater Creek (0.25; 0.16, 0.35), and lowest at Mixon’s Creek (0.06; 0.02, 0.10) and

Garraway Creek (0.13; 0.08, 0.19). Finally, mean relative abundance of structure specialists was greatest at Mixon’s Creek (0.69; 0.58, 0.76) and Priest Creek (0.25; 0.19,

0.32), and lowest at Garraway Creek (0.10; 0.03, 0.17) and Sweetwater Creek (0.08; 0.02,

0.13). Means and SDs in Bray-Curtis distances were similar among sites (Mixon’s Creek:

0.008 ± 0.004; Garraway Creek: 0.008 ± 0.006; Sweetwater Creek: 0.006 ± 0.003; Priest

Creek: 0.006 ± 0.003).

Analysis of Movement Rates

For our reach scale model set, one interaction model (wᵢ = 0.42) best explained the variation in reach movement rates of water-column specialists (Table 3.4). The model

(pseudo R² = 0.17) characterized reach movement rates as the outcome of a significant,

negative interaction (beta = -3.58, p = 0.003) between mean difference in current velocity and mean difference in woody structure. Our multi-scale modeling results suggested the addition of land cover or confluence size as additive effects to the model did not increase model fit.

Analysis of Assemblage Change

A well, supported model (wᵢ = 0.99, pseudo R² =0.28) suggested that reach scale, Bray-

Curtis distances were best explained by a significant, large positive interaction (beta =

249.95, p = 0.0005) between mean difference in depth and mean difference in woody structure (Table 3.4). There were also significant, negative effects of mean difference in woody structure (-63.92, p = 0.009) and depth (-2.21, p = 0.004) on reach scale, Bray distances. Multi-scale modeling results suggested the addition of land cover or confluence size as additive effects to the model did not increase model fit

Discussion

In this study, we assessed the influences of confluence size and land cover on habitat change, the movement ecology of a functional guild, and on temporal assemblage change at four headwater confluences in a Gulf Coastal Plain dendritic river system. Hydrologic variability (e.g., decreases in wetted width) altered habitat patches, influenced the rate of movement of water-column specialists, and promoted assemblage change at these four headwater confluences; however, no direct effects of land cover or confluence size were apparent. Our analyses yielded six primary findings: (1) a hydrogeomorphic gradient (PC2) associated with differences in land cover facilitated habitat change (2) along this gradient, urban reaches with a confluence size > 0.6 (i.e., Mixon’s Creek reaches) exhibited the.

Table 3.4

Top reach scale and multi-scale models of water-column specialists’ proportional daily movement rates (PDMR) and of Bray- Curtis Distances at four headwater confluences in the Pascagoula River basin, MS, USA. Values are shown for the number of parameters (K), AICC, ∆AICC and model weights (wi). Response Scale Model K AICC ∆AICC wi

PDMR Reach Current Velocity *Woody Debris + 1| Stream 5 -216.3 0.0 0.42

Multi Flow * Woody Debris + 1| Stream 5 -215.4 0.0 0.71

Flow * Woody Debris + Land cover + 1| Stream 6 -214.1 2.3 0.22

Bray-Curtis Reach Depth * Woody Debris + 1| Stream 5 -544.9 0.0 0.99

Multi Depth * Woody Debris + 1| Stream 5 -546.3 0.0 0.54

Depth * Woody Debris + CR + 1| Stream 6 -543.8 1.6 0.25

Depth * Woody Debris + Land cover + 1| Stream 6 -542.9 1.9 0.21

greatest habitat change, (3) patterns in patch selection of water-column specialists were interrelated to land cover, (4) the availability of woody structure strongly influenced the magnitudes of variation in assemblage change and movement rate of water-column specialists, (5) rate of movement of water-column specialists predictably increased within increasing change in stream flow due to decreased habitat complexity, and (6) Bray-Curtis distance predictably declined with increasing change in depth due to decreased habitat complexity. Our results highlight the influence of confluence size and land cover on differences in habitat variation, assemblage change, and the rate of dispersal of stream fishes in headwater streams. Because the movement and assemblage datasets used for this study were small (N=51, N=63); replication was low (9-18), reducing the power of analyses.

Effect of Land Cover on Habitat Change

Our results indicated that land cover facilitated habitat change among all stream reaches more so than confluence size. Differences in habitat change between sites was associated with differences in the magnitude of geomorphic variation (e.g., changes in bankfull width, current velocity, Fig. 3.2, Table B3). Urban reaches were generally characterized by greater bankfull widths, greater variation in wetted width, diminished availability of woody structure, shallower depths, and greater variance in current velocity (Table B4). In small, coastal plain streams, limited impervious cover (<11%) may increase median water temperatures, flashiness, and generally limit the influence of base flow on total stream discharge (Schneid, Anderson & Feminella, 2017). However, the extent to which land cover alters active channel width (Utz & Hilderbrand, 2011; Schneid, Anderson &

Feminella, 2017), and sediment transport (Kang & Marston, 2006; Riley, 2009; Utz &

Hilderbrand, 2011; Hardison et al., 2009) in Coastal Plain streams remains unclear. For example, low gradient drainages in Oklahoma and Georgia were characterized by low alterations to sediment transport size and channel enlargement despite increases in watershed urbanization (Kang & Marston, 2006; Riley, 2009). However, increases in active channel width are reported for other small, urban Coastal Plain drainages (Hardison et al.,

2009; Schneid, Anderson & Feminella, 2017). Generally, our results indicate that there was little variation in mean sediment size during the study period across all sites, but urban drainages were typified by greater variability in active channel width (Table B4). Utz &

Hildebrand (2011) suggested that the stability of sediment structure in small, urban Coastal

Plain streams may be related to bed material. Coarse substrate particles, which are often in great abundance in Piedmont and montane drainages, are rarely present in Coastal Plain streams.

Effect of Confluence Size on Habitat Change

Our prediction that urban reaches characterized by a confluence size > 0.6 would exhibit the greatest amount of habitat change was supported. Differences in habitat change between sites was associated with differences in the magnitude of geomorphic variation

(e.g., changes in bankfull width, current velocity, Fig. 3.2, Table B3). We did observe trends (Table B4) in habitat variation that support the predictions made by Benda et al.

(2004). For example, mean depth was deeper for all mainstem, downstream reaches, measures of sinuosity were greater for all mainstem, upstream reaches, and generally, stream gradient was greater in mainstem, downstream reaches (Table A3.4, Table B5).

Conversely, excluding 2018 data for Mixon’s Creek, our findings that mean wetted width, and mean bankfull width were greater in mainstem, downstream reaches, but mean

substrate size was larger in mainstem upstream reaches (see Table B4) are in disagreement with the predictions of Benda et al. (2004).

Movement Behavior of Water-Column Specialists

Distance between patches influenced the inter-patch movements of water-column specialists. The number of fish movements between patches at Sweetwater Creek and

Garraway Creek were negatively related to inter-patch distance (Figure 3.3). Belica &

Rahel (2008) and Lonzarich, Lonzarich & Warren (2000) suggested that movement rates of fishes out of habitat patches (pools) declined as the distance between patches increased.

Schaefer (2001) also found that the probability of movement by minnows receded as riffle lengths between pools increased. Interestingly, the distribution of fish movements between patches at Priest Creek and Mixon’s Creek appeared to be either bimodal, or right skewed

(Figure 3.3). While we found significant interactions between distance travelled and both categorical factors on the number of movers, the observational data suggests that land cover had a greater influence on the movement behavior of water-column specialists.

Patch instability, specifically, the dewatering of riffle habitats, likely contributed to differences in patch selection and distances traveled by water-column specialists as a consequence of land cover. Riffle dewatering is a consequence of the alteration of the infiltration-recharge process and increases the likelihood that drainages characterized by a reduction in percent forested cover will be subjected to increased periods of low base flow

(Price et al., 2011). There was a significant reduction in the number of observed riffle moves at urban sites. Variation in riffle wetted width at Garraway Creek and Sweetwater

Creek was approximately equivalent to a 25% change in mean wetted width (Garraway

Creek: 4.43 m ± 1.08 m; Sweetwater Creek: 2.43 m ± 0.64 m), but variation in riffle wetted

width at Priest Creek and Mixon’s Creek was a nearly 40% change in mean wetted width

(Priest Creek: 3.27 m ± 1.21 m; Mixon’s Creek: 3.0 m ± 1.08 m). Thus, we observed a greater number of riffle to riffle (n=7) and pool to riffle (n=29) moves at forested sites, than at urban sites (riffle to riffle: n= 4, pool to riffle: n =11). Riffle dewatering as a result of periods of low flow strongly alters the movement behavior of stream fishes (Labbe &

Fausch 2000; Schaefer, 2001), often inducing a reduction in fitness (Davey & Kelly, 2007;

Roberts & Angermeier, 2007; Rosenfeld, Beecher & Ptolemy, 2016).

Effect of Woody Structure Availability on the Movement of Water-Column Specialists

Differences in the percentage occurrence of in-stream wood altered the size of the effect of hydrologic (e.g., current velocity) variation on movement rates of water-column specialists. For example, when change in woody structure availability was small (< 0.002 mean difference), in relation to growing change in current velocity, movement rates of water-column specialists seemed to increase (Figure 3.4A). This trend was more common

(i.e., more data points) among Mixon’s Creek and Priest Creek reaches. Conversely, when change in woody structure availability was large (> 0.002 mean difference), movement rates seemed to decline in response to cumulative changes in current velocity. Movement rates seemed to decline steeply within and among Garraway Creek and Sweetwater Creek reaches, but decreased moderately within and among Priest Creek reaches (Figure 3.4B).

Thus our results indicate that small perturbations in geomorphic change engendered greater movement at urban confluences due to their reduced capacity to retain in-stream wood.

Density and retention of in-stream wood is much higher in forested streams as opposed to drainages dominated by urban or agricultural land cover (Angradi et al., 2004; Blauch &

Jefferson, 2019). Further, in drainages dominated by urban land cover, in-stream wood has

B A

Figure 3.4 Plot of interaction between mean difference in woody structure and mean difference in current velocity on the proportional daily movement rate (PDMR) of water-column specialists. Mean difference in woody structure was converted into a factor with two levels (A: small = ≤ 0%; B: large = > 0%) to identify differences in PDMR as a consequence of the interaction between the two habitat variables. The number of plotted points along each regression represents the number of predicted values associated with that level of woody structure. *No predicted estimates of PDMR for Mixon’s Creek were associated with an increase in woody structure.

little influence on reach scale, channel morphology (Blauch & Jefferson, 2019). Because large wood accumulation in small streams may last for hundreds of years (Dollof &

Warren, 2003), in-stream wood regulates hydrological and sediment transfer processes which facilitates habitat heterogeneity and stabilization (Angermeier & Karr, 1984;

Gurnell, Tockner, Edwards & Petts, 2005; Dolloff & Warren, 2003). Consequently, the maintenance of deepwater or pool habitats is greatest in small streams typified by a large volume of in-stream wood. Pool and deepwater habitats provide stream fishes refuge from droughts and streams typified by small shallow pools (Lisle & Hilton, 1992) are often species depauperate and exhibit greater variation in species abundance (Schlosser, 1987).

Thus, stream fish are less prone to move in physically, complex habitats (Roni & Quinn,

2001; Belica & Rahel, 2008; Clark & Schaefer, 2016).

Effect of Woody Structure Availability on Assemblage Change

Our finding that assemblage change was a corollary of the interaction between woody structure availability and channel morphology (e.g. depth) strengthens the argument that the stability of stream fish assemblages is strongly related to physical habitat complexity

(Pearsons, Li & Lamberti, 1992; Paller, 2002; Shea & Petersen, 2007) linked to land cover

(Wang, Lyons, Kanehl & Bannerman 2001; Hitt & Angermeier, 2008; Hubbell et al.,

2020). The abiotic and biotic benefits of in-stream wood to macroinvertebrate (Drury &

Kelso, 2000) and fish assemblages in small streams is well documented (Karr &

Angermeier, 1984; Shields, Knight, Morin & Blank, 2003; Scott & Roberts, 2017). When change in woody structure availability was small (< 0.002 mean difference assemblage distances appeared to decrease as a consequence of increasing change in depth (Figure

3.5A). However, when change in woody structure availability was large (> 0.002 mean

A B

Figure 3.5 Plot of interaction between mean difference in woody structure and mean difference in depth on reach scale, Bray-Curtis Distances. Mean difference in woody structure was converted into a factor with two levels (A: small = ≤ 0%; B: large = > 0%) to identify differences in Bray-Curtis Distance as a consequence of the interaction between the two habitat variables. The number of plotted points along each regression represents the number of predicted values associated with that level of woody structure. *No predicted estimates of Bray-Curtis Distance for Mixon’s Creek were associated with an increase in woody structure.

difference), assemblage distances generally declined within Priest Creek reaches, but was highly variable within Sweetwater Creek and Garraway Creek reaches in response to increasing change in depth (Figure 3.5B). Thus, changes in functional relative abundances among forested reaches were likely a consequence of an increase in habitat complexity.

Scarce availability of woody structure within Mixon’s Creek reaches reduced habitat complexity as a consequence of hydrologic variability, thus decreasing the likelihood of assemblage change (Figure 3.5A). Hydrologic variability reduces habitat complexity, shifting changes in stream fish assemblages to favor generalists as opposed to more specialized functional groups (Poff & Allan, 1995; Bunn & Arthington, 2002). The effect of hydrologic variability on assemblage dynamics is dampened in stream reaches characterized by more complex habitat (Pearsons, Li & Lamberti, 1992), but the influence of land cover alters the hydrologic regime, prompting magnified rates of erosion to the streambed (Walsh et al., 2005), thus disturbing critical, habitat refugia of many sensitive fishes (Walters et al., 2005; Hubbell et al., 2020).

Conclusions

We assumed that there would be an interaction between confluence size and land cover, and thus, urban reaches with a larger confluence size would exhibit the greatest variability in habitat, movement, and assemblage change. Geomorphic characteristics were most variable at Mixon’s Creek, subjecting water-column specialists to greater geomorphic change. Thus, water-column specialists in these reaches were more likely to move greater distances. However, although these individuals moved greater distances within and among reaches at this site, there was no statistical evidence confluence size contributed to substantial variation in rate of movement or assemblage change among all of our study

reaches.

We used a 2 X 2 design to assess the effects of confluence size and land cover on stream fish movement and assemblage change. Most stream fish movement studies have sought to understand what proportion of populations are mobile (Skalski & Gilliam, 2000; Fraser et al., 2001; Rodriguez, 2002), what ecological factors regulate movement at the patch and reach scales (Albanese, Angermeier & Doraj-Raj, 2004; Belica & Rahel, 2008; Walker &

Adams, 2016); and how physical barriers may prevent movement (Warren & Pardew,

1998; Knott, Mueller, Pander & Geist, 2020; Williams et al., 2020). Additionally, studies assessing the role of confluences on metapopulation and meta-assemblage dynamics have mostly examined their effects at the landscape scale (see Smith & Kraft, 2005; Kiffney et al., 2006; Angermeir & Hitt, 2008; Hubbell et al., 2020), and few studies have been conducted to investigate their influence on dispersal (but see Cathcart, Gido, McKinstry &

MacKinnon, 2018). No observational studies, to our knowledge, have attempted to connect local geomorphic change at confluences to biotic change in a natural setting. Thus, our study provides baseline data for future studies further investigating the relationship between confluences and land cover on dispersal and assemblage change at localized

(reach, mesohabitat unit) scales. Limited replication diminished our statistical power and the use of batch markings constrained the temporal scale at which movement was measured in this study. Given our results, an influence of confluence size on movement and assemblage change in headwaters would more likely be detected if upstream land cover between sites was similar and the method of replication yielded larger sample sizes. Thus, future studies should consider the use of other tagging methods (e.g., passive integrated transponders; acoustic tags) to increase the number of replicates (e.g., marking of

individuals), but also allowing for long-term monitoring to better understand how confluences may influence movement and assemblage change in headwaters.

REFERENCES

Albanese, B., Angermeier, P. L., & Dorai-Raj, S. (2004). Ecological correlates of fish

movement in a network of Virginia streams. Canadian Journal of Fisheries and

Aquatic Sciences, 61(6), 857-869.

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

ecosystems. Annu. Rev. Ecol. Evol. Syst., 35, 257-284.

Hitt, N. P., & Angermeier, P. L. (2008). Evidence for fish dispersal from spatial analysis

of stream network topology. Journal of the North American Benthological

Society, 27(2), 304-320.

Angradi, T. R., Schweiger, E. W., Bolgrien, D. W., Ismert, P., & Selle, T. (2004). Bank

stabilization, riparian land use and the distribution of large woody debris in a

regulated reach of the upper Missouri River, North Dakota, USA. River Research

and Applications, 20(7), 829-846.

Belica, L. A. T., & Rahel, F. J. (2008). Movements of creek chubs, Semotilus

atromaculatus, among habitat patches in a plains stream. Ecology of Freshwater

Fish, 17(2), 258-272.

Benda, L. E., & Cundy, T. W. (1990). Predicting deposition of debris flows in mountain

channels. Canadian Geotechnical Journal, 27(4), 409-417.

Benda, L. E. E., Poff, N. L., Miller, D., Dunne, T., Reeves, G., Pess, G., & Pollock, M.

(2004). The network dynamics hypothesis: how channel networks structure

riverine habitats. BioScience, 54(5), 413-427.

Benda, L., Veldhuisen, C., & Black, J. (2003). Debris flows as agents of morphological

heterogeneity at low-order confluences, Olympic Mountains,

Washington. Geological Society of America Bulletin, 115(9), 1110-1121.

Best, J. L. (1986). The morphology of river channel confluences. Progress in

Physical Geography, 10(2), 157-174.

Bangs, B. L., Falcy, M. R., Scheerer, P. D., & Clements, S. (2013). Comparison of three

methods for marking a small floodplain minnow. Animal Biotelemetry, 1(1), 18.

Blauch, G. A., & Jefferson, A. J. (2019). If a tree falls in an urban stream, does it stick

around? Mobility, characteristics, and geomorphic influence of large wood in

urban streams in northeastern Ohio, USA. Geomorphology, 337, 1-14.

Brown, D. K., Echelle, A. A., Propst, D. L., Brooks, J. E., & Fisher, W. L. (2001).

Catastrophic wildfire and number of populations as factors influencing risk of

extinction for Gila trout (Oncorhynchus gilae). Western North American

Naturalist, 139-148.

Bunn, S. E., & Arthington, A. H. (2002). Basic principles and ecological consequences of

altered flow regimes for aquatic biodiversity. Environmental Management, 30(4),

492-507.

Burnham, K. P., & Anderson, D. R. (2002). A practical information-theoretic

approach. Model selection and multimodel inference, 2nd ed. Springer, New York.

Campbell Grant, E. H., Lowe, W. H., & Fagan, W. F. (2007). Living in the branches:

population dynamics and ecological processes in dendritic networks. Ecology

Letters, 10(2), 165-175.

Cathcart, C. N., Gido, K. B., & McKinstry, M. C. (2015). Fish community distributions

and movements in two tributaries of the San Juan River, USA. Transactions of the

American Fisheries Society, 144(5), 1013-1028.

Cathcart, C. N., Gido, K. B., McKinstry, M. C., & MacKinnon, P. D. (2018). Patterns of

fish movement at a desert river confluence. Ecology of Freshwater Fish, 27(1),

492-505.

Clark, S. R., & Schaefer, J. F. (2016). Ecological influences on the local movement

dynamics of the blackspotted topminnow, Fundulus olivaceus. Behavioral

Ecology and Sociobiology, 70(4), 557-567.

Cummins, K. W. (1962). An evaluation of some techniques for the collection and

analysis of benthic samples with special emphasis on lotic waters. The American

Midland Naturalist, 67(2), 477-504.

Davey, A. J., & Kelly, D. J. (2007). Fish community responses to drying disturbances in

an intermittent stream: a landscape perspective. Freshwater Biology, 52(9), 1719-

1733.

Dolloff, C. A., & Warren, M. L. (2003). Fish relationships with large wood in small

streams. In American Fisheries Society Symposium 37: 179-193, 2003.

Drury, D. M., & Kelso, W. E. (2000). Invertebrate colonization of woody debris in

coastal plain streams. Hydrobiologia, 434(1-3), 63-72.

Dynesius, M., & Nilsson, C. (1994). Fragmentation and flow regulation of river systems

in the northern third of the world. Science, 266(5186), 753-762.

Fagan, W. F. (2002). Connectivity, fragmentation, and extinction risk in dendritic

metapopulations. Ecology, 83(12), 3243-3249.

Fraser, D. F., Gilliam, J. F., Daley, M. J., Le, A. N., & Skalski, G. T. (2001). Explaining

leptokurtic movement distributions: intrapopulation variation in boldness and

exploration. The American Naturalist, 158(2), 124-135.

Gerking, S. D. (1959). The restricted movement of fish populations. Biological

Reviews, 34(2), 221-242.

Gorman, O. T. (1986). Assemblage organization of stream fishes: the effect of rivers on

adventitious streams. The American Naturalist, 128(4), 611-616.

Gurnell, A., Tockner, K., Edwards, P., & Petts, G. (2005). Effects of deposited wood on

biocomplexity of river corridors. Frontiers in Ecology and the Environment, 3(7),

377-382.

Hardison, E. C., O’Driscoll, M. A., DeLoatch, J. P., Howard, R. J., & Brinson, M. M.

(2009). Urban Land Use, Channel Incision, and Water Table Decline Along

Coastal Plain Streams, North Carolina 1. JAWRA Journal of the American Water

Resources Association, 45(4), 1032-1046.

Hitt, N. P., & Angermeier, P. L. (2008). Evidence for fish dispersal from spatial analysis

of stream network topology. Journal of the North American Benthological

Society, 27(2), 304-320.

Hogan, D. L., Bird, S. A., & Hassan, M. A. (1998). Spatial and temporal evolution of

small coastal gravel-bed streams: influence of forest management on channel

morphology and fish habitats. Gravel-bed Rivers in the Environment, 365-392.

Hubbell, J., Warren, M. L., Schaefer, J. F., Sterling, K., & Flood, P. J. (2020).

Fragmentation alters ecological gradients and headwater fish assemblage

composition relative to land use in a dendritic river system. Canadian Journal of

Fisheries and Aquatic Sciences, 77, 1281-1291.

Hupp, C. R. (2000). Hydrology, geomorphology and vegetation of Coastal Plain rivers in

the southeastern USA. Hydrological Processes, 14(16‐17), 2991-3010.

Johnson, A. M., & Rodine, J. R. (1984). Debris flow (Chapter 8). Slope Instability, eds.

D. Brunsden & DB Prior.

Kang, R. S., & Marston, R. A. (2006). Geomorphic effects of rural-to-urban land use

conversion on three streams in the Central Redbed Plains of

Oklahoma. Geomorphology, 79(3-4), 488-506.

Kiffney, P. M., Greene, C. M., Hall, J. E., & Davies, J. R. (2006). Tributary streams

create spatial discontinuities in habitat, biological productivity, and diversity in

mainstem rivers. Canadian Journal of Fisheries and Aquatic Sciences, 63(11),

2518-2530.

Knott, J., Mueller, M., Pander, J., & Geist, J. (2020). Seasonal and diurnal variation of

downstream fish movement at four small‐scale hydropower plants. Ecology of

Freshwater Fish, 29(1), 74-88.

Koupal, K., & Hoback, W. (2007). Successful application of visible implant

elastomer tags on crappies, Pomoxis spp., without the use of

anaesthetic. Fisheries Management and Ecology, 14, 235-238.

Labbe, T. R., & Fausch, K. D. (2000). Dynamics of intermittent stream habitat regulate

persistence of a threatened fish at multiple scales. Ecological Applications, 10(6),

1774-1791.

Leal, C. G., Pompeu, P. S., Gardner, T. A., Leitão, R. P., Hughes, R. M., Kaufmann, P.

R., & Mac Nally, R. (2016). Multi-scale assessment of human-induced changes to

Amazonian instream habitats. Landscape Ecology, 31(8), 1725-1745.

Legendre, L., & Legendre, P. (1983). Partitioning ordered variables into discrete states

for discriminant analysis of ecological classifications. Canadian Journal of

Zoology, 61(5), 1002-1010.

Leitão, R. P., Zuanon, J., Mouillot, D., Leal, C. G., Hughes, R. M., Kaufmann, P. R., &

Ferraz, S. F. (2018). Disentangling the pathways of land use impacts on the

functional structure of fish assemblages in Amazon streams. Ecography, 41(1),

219-232.

Lisle, T. E., & Hilton, S. (1992). The volume of fine sediment in pools: An index of the

supply of mobile sediment in stream channels. Water Resources Bulletin, 28(2),

371-383.

Liu, F., & Eugenio, E. C. (2018). A review and comparison of Bayesian and likelihood-

based inferences in beta regression and zero-or-one-inflated beta

regression. Statistical methods in Medical Research, 27(4), 1024-1044.

Lonzarich, D. G., Lonzarich, M. R., & Warren Jr, M. L. (2000). Effects of riffle length on

the short-term movement of fishes among stream pools. Canadian Journal of

Fisheries and Aquatic Sciences, 57(7), 1508-1514.

McClain, M. E., Boyer, E. W., Dent, C. L., Gergel, S. E., Grimm, N. B., Groffman, P.

M., & McDowell, W. H. (2003). Biogeochemical hot spots and hot moments at

the interface of terrestrial and aquatic ecosystems. Ecosystems, 301-312.

Mosley, M. P. (1976). An experimental study of channel confluences. The Journal of

Geology, 84(5), 535-562.

Neufeld, K., Blair, S., & Poesch, M. (2015). Retention and stress effects of visible

implant tags when marking Western Silvery Minnow and its application to other

cyprinids (Family Cyprinidae). North American Journal of Fisheries

Management, 35(5), 1070-1076.

Padgham, M., & Webb, J. A. (2010). Multiple structural modifications to dendritic

ecological networks produce simple responses. Ecological Modelling, 221(21),

2537-2545.

Paller, M. H. (2002). Temporal variability in fish assemblages from disturbed and

undisturbed streams. Journal of Aquatic Ecosystem Stress and Recovery, 9(3),

149-158.

Pearsons, T. N., Li, H. W., & Lamberti, G. A. (1992). Influence of habitat complexity on

resistance to flooding and resilience of stream fish assemblages. Transactions of

the American Fisheries Society, 121(4), 427-436.

Perkin, J. S., & Gido, K. B. (2012). Fragmentation alters stream fish community structure

in dendritic ecological networks. Ecological Applications, 22(8), 2176-2187.

Poff, N. L., & Allan, J. D. (1995). Functional organization of stream fish assemblages in

relation to hydrological variability. Ecology, 76(2), 606-627.

Price, K., Jackson, C. R., Parker, A. J., Reitan, T., Dowd, J., & Cyterski, M. (2011).

Effects of watershed land use and geomorphology on stream low flows during

severe drought conditions in the southern Blue Ridge Mountains, Georgia and

North Carolina, United States. Water Resources Research, 47(2).

Rhoads, B. L. (1987). Changes in stream channel characteristics at tributary

junctions. Physical Geography, 8(4), 346-361.

Rice, S. (1998). Which tributaries disrupt downstream fining along gravel-bed rivers?

Geomorphology, 22(1), 39-56.

Rice, S. P., Greenwood, M. T., & Joyce, C. B. (2001). Tributaries, sediment sources, and

the longitudinal organisation of macroinvertebrate fauna along river

systems. Canadian Journal of Fisheries and Aquatic Sciences, 58(4), 824-840.

Rice, S. P., & Macklin, M. G. (2008). Geomorphology and Earth system science: a

reflection. ESPL, 33(13), 2118-2120.

Riley, J. W. (2009). Assessing morphological adjustments of stream channels in the

Piedmont area of Georgia, USA (Doctoral dissertation, MSc Thesis, University of

Georgia, Athens, Georgia).

Roberts, J. H., & Angermeier, P. L. (2004). A comparison of injectable fluorescent marks

in two genera of darters:effects on survival and retention rates. North American

Journal of Fisheries Management, 24(3), 1017-1024.

Roberts, J. H., & Angermeier, P. L. (2007). Spatiotemporal variability of stream habitat

and movement of three species of fish. Oecologia, 151(3), 417-430.

Rodríguez, M. A. (2002). Restricted movement in stream fish: the paradigm is

incomplete, not lost. Ecology, 83(1), 1-13.

Rodríguez, M. A. (2010). A modeling framework for assessing long-distance dispersal

and loss of connectivity in stream fish. In Community ecology of stream fishes:

concepts, approaches, and techniques. American Fisheries Society,

Symposium (Vol. 73, pp. 263-279).

Roni, P., & Quinn, T. P. (2001). Effects of wood placement on movements of trout and

juvenile Coho Salmon in natural and artificial stream channels. Transactions of

the American Fisheries Society, 130(4), 675-685.

Rosenfeld, J., Beecher, H., & Ptolemy, R. (2016). Developing bioenergetic-based habitat

suitability curves for instream flow models. North American Journal of Fisheries

Management, 36(5), 1205-1219.

Schaefer, J. (2001). Riffles as barriers to interpool movement by three cyprinids

(Notropis boops, Campostoma anomalum and Cyprinella venusta). Freshwater

Biology, 46(3), 379-388.

Schlosser, I. J. (1987). A conceptual framework for fish communities in small warmwater

streams. Community, and Evolutionary Ecology of North American Stream

Fishes. 17-24.

Schneid, B. P., Anderson, C. J., & Feminella, J. W. (2017). The influence of low-

intensity watershed development on the hydrology, geomorphology,

physicochemistry and macroinvertebrate diversity of small coastal plains

streams. Ecological Engineering, 108, 380- 390.

Scott, R. L., & Roberts, J. H. (2017). Influences of Local and Regional Environmental

Factors on Atlantic Coastal Plain Stream Fish Communities.

Shea, C. P., & Peterson, J. T. (2007). An evaluation of the relative influence of habitat

complexity and habitat stability on fish assemblage structure in unregulated and

regulated reaches of a large southeastern warmwater stream. Transactions of the

American Fisheries Society, 136(4), 943-958.

Shields, F. D., Knight, S. S., Morin, N., & Blank, J. (2003). Response of fishes and

aquatic habitats to sand-bed stream restoration using large woody

debris. Hydrobiologia, 494(1-3), 251-257.

Skalski, G. T., & Gilliam, J. F. (2000). Modeling diffusive spread in a heterogeneous

population: a movement study with stream fish. Ecology, 81(6), 1685-1700.

Smith, T. A., & Kraft, C. E. (2005). Stream fish assemblages in relation to landscape

position and local habitat variables. Transactions of the American Fisheries

Society, 134(2), 430- 440.

Smithson, E. B., & Johnston, C. E. (1999). Movement patterns of stream fishes in a

Ouachita Highlands stream: an examination of the restricted movement

paradigm. Transactions of the American Fisheries Society, 128(5), 847-853.

Stow, A. J., Sunnucks, P., Briscoe, D. A., & Gardner, M. G. (2001). The impact of

habitat fragmentation on dispersal of Cunningham’s skink (Egernia

cunninghami): evidence from allelic and genotypic analyses of

microsatellites. Molecular Ecology, 10(4), 867-878.

Utz, R. M., Hilderbrand, R. H., & Raesly, R. L. (2010). Regional differences in patterns

of fish species loss with changing land use. Biological Conservation, 143(3), 688-

699.

Walker, R. H., & Adams, G. L. (2016). Ecological factors influencing movement of creek

chub in an intermittent stream of the Ozark Mountains, Arkansas. Ecology of

Freshwater Fish, 25(2), 190-202.

Walsh, C. J., Roy, A. H., Feminella, J. W., Cottingham, P. D., Groffman, P. M., &

Morgan, R. P. (2005). The urban stream syndrome: current knowledge and the

search for a cure. Journal of the North American Benthological Society, 24(3),

706-723.

Walters, D. M., Freeman, M. C., Leigh, D. S., Freeman, B. J., & Pringle, C. M. (2005).

Urbanization effects on fishes and habitat quality in a southern Piedmont river

basin. In American Fisheries Society Symposium (Vol. 47, pp. 69-85).

Wang, L., Lyons, J., Kanehl, P., & Bannerman, R. (2001). Impacts of urbanization on

stream habitat and fish across multiple spatial scales. Environmental

Management, 28(2), 255-266.

Warren Jr, M. L., & Pardew, M. G. (1998). Road crossings as barriers to small-stream

fish movement. Transactions of the American Fisheries Society, 127(4), 637-644.

Williams, J. M., Dodd, H. R., & Finn, D. S. (2020). A low-water crossing impacts

Northern Hog Sucker Hypentelium nigricans movement in an Ozark

stream. Journal of Freshwater Ecology, 35(1), 157-171.

CHAPTER IV FUNCTIONAL CONNECTIVITY BEGETS CONFLICTING IMPACTS

OF LANDSCAPE GRADIENTS ON THE OCCUPANCY AND GENETIC DISTANCE

OF TWO HEADWATER FISHES

Abstract

Patch structure and connectivity in dendritic river systems is a consequence of their bifurcating structure; however, the influence of this branching pattern on functional connectivity is not present in patch occupancy models. Thus, the weight of influence of environmental and anthropogenic gradients on the patch occupancy of many aquatic obligates in dendritic river systems may be under or overestimated when using this modeling framework. In our study, we modeled occupancy and genetic distance using an information theoretic approach to examine whether functional connectivity in a dendritic river system affects how two co-occurring, headwater darters relate to environmental

(stream size, groundwater input) and anthropogenic (road crossing density, presence/distance of reservoir, and land cover) gradients in a dendritic river system. We also tested whether the influence of these resistance gradients on genetic distance may vary dependent on subbasin size and species. We conducted our study in two parallel subbasins that are hierarchically nested within a larger Gulf Coastal Plain drainage in the southeastern

United States. For one subbasin, we examined patterns of site-occupancy using detection data collected at the reach-scale. We used Genotype by Sequencing to identify single nucleotide polymorphisms to approximate genetic distance between pairs of sites, but also to evaluate of population structure, population demographics, and dispersal within both subbasins. In both subbasins, Yazoo and Goldstripe Darters exhibited hierarchical structure, and estimates of the number of migrants per generation suggested that historical

migrant exchange was limited to a few watersheds. Rankings of site-occupancy and models of genetic distance suggested that functional connectivity strongly affected how either species related to environmental and anthropogenic gradients. Subbasin size altered the influence of anthropogenic gradients on Yazoo and Goldstripe Darter genetic distance.

Relative to site occupancy models, differences in habitat specialization were emphasized when relating Yazoo Darter and Goldstripe Darter genetic distances to environmental gradients. The comparative, modeling approach used in this study suggests that patch occupancy models may over or underestimate the influence of environmental and anthropogenic gradients on stream fishes. Thus, the development of new models that feature dendritic structure should provide better insight into how landscape heterogeneity influences stream fish metapopulations.

Introduction

Patchiness is a central concept in spatial ecology, occurring across multiple spatial and temporal extents. A patch may best be defined as a spatial subunit established by an organism (Pringle et al., 1988) where connectivity defines the extent that a landscape facilitates or restrains movement between or among patches (Tischendorf & Fahrig, 2000).

Connectivity characterizes a species persistence as a tradeoff between resource limitation at occupied patches and the risk of migration to unoccupied patches (Levins, 1966; Wiens,

2002). Thus, connectivity is a product of the arrangement and structure of patches within the broader context of the landscape. Structural connectivity describes the proximity of patches, irrespective of the individual patch quality or ecology of resident species (Taylor,

Henein & Merriam, 1993). Alternatively, functional connectivity describes the extent that a landscape inhibits or enables dispersal among patches (Bélisle, 2005). Because patch

structure and connectivity are scale dependent, an organism’s response will depend on the scale at which it distinguishes differences habitat structure (Kotliar & Wiens, 1990).

Understanding how fragmentation influences population dynamics is a fundamental objective of metapopulation theory. This lead to the formulation of patch occupancy models, with stochastic patch occupancy models (Hanksi, 1994; Ovaskainen & Hanski,

2004) and site-occupancy models (MacKenzie et al., 2002; 2003) being the most frequently used. The integration of crucial mechanisms (dispersal and connectivity) of spatially realistic metapopulation theory and site-occupancy models (imperfect detection) has allowed for the simultaneous modeling of phase-specific colonization and extinction dynamics, population size, while also adjusting for imperfect detection (Sutherland, Elston

& Lambin 2012). However, an explicit weakness of spatially realistic metapopulation theory is the connectivity metric, which uses Euclidean distance to assess the degree that a patch is isolated from all other patches (Hanski, 1999). Because Euclidean distance is a measure of the straight-line distance between two points, this metric is not representative of patterns of connectivity in the natural world, particularly in rivers and streams.

Patch structure and connectivity in many river systems is a consequence of their bifurcating structure, thus they are characterized as dendritic ecological networks. Stream branches increase in number exponentially as drainage size decreases (Campbell-Grant et al., 2007). In river systems, stream segments represent the branches of the dendritic ecological network and are ecologically connected in the longitudinal dimension. This resulting “branchiness” and hierarchy of habitat arrangements affects the connectivity of stream segments and thus promotes the isolation of metapopulations (Poole, 2002; Fagan,

2002; Diez & Pulliam, 2007). The spatial and hierarchical arrangement of habitats within

100

a dendritic river system influences metapopulation dynamics and induces genetic processes such as local adaptation and drift to act on subpopulations bringing about drainage substructuring (Boizard, Magnan & Angers, 2009). Thus, not accounting for such alterations to the functional connectivity of patches in a modeling framework may engender spurious species-environmental relationships.

Headwaters account for 80% of the surface water within a river system (Colvin et al.,

2019), and provide habitat to unique assemblages of fishes (Paller, 1994). Ecological trait selection among headwater specialists may result from the unique habitats occupied

(Mundahl & Ingersoll, 1983; Petty & Grossman, 2004; Schimdt & Schaefer, 2018).

Because headwaters are small, isolated, and often unstable (Ostrand & Wilde, 2002;

Grenouillet & Hérissé 2004), these fishes are much more susceptible to local extinction due to the resulting small population sizes (Fagan, 2002; Richardson, 2019). The spatial isolation of these habitats encourages headwater specialists (i.e, species which permanently reside in headwaters) to exchange migrants between neighboring drainages as described by the stream hierarchy model (Meffe & Vrijenhoek, 1988).

Isolation by resistance threatens species persistence in streams by fragmenting habitats

(Fagan, 2002; Perkin & Gido, 2011). Isolation by resistance is a product of reduced functional connectivity, thus inhibiting gene flow among populations (McRae, 2006).

Spatial patterns of natural fragmentation within a dendritic river system are influenced by drainage size. As the density and size of streams within a dendritic river system multiply, the number of confluences capable of producing consistent geomorphic change increases

(Benda et al., 2004). Streamflow permanence is dependent on relative groundwater input, which may vary among small to medium sized streams. As groundwater discharge

101

declines, headwaters become increasingly vulnerable to surface dewatering (Gordon,

McMahon, Finlayson, & Gippel, 1992; Richardson 2019). Thus, headwater specialists’ occurrence (Mollenhauuer, Zhou, & Brewer, 2019) and these organisms’ propensity to colonize (Adams & Warren, 2005) new habitats is closely associated with relative groundwater input.

Two key sources of anthropogenic fragmentation within dendritic river systems are road crossings and reservoirs (Warren & Pardew, 1998; Hudman & Gido, 2012; Perkin &

Gido, 2012; Fluker, Kuhajda & Harris, 2014). Connectivity across a dendritic river system is reduced upon the establishment of these barriers in contrast to most non-dendritic river systems which are resilient to fragmentation induced by a small number of barriers (Cote,

Kehler, Bourne & Wiersma, 2009). Dispersal, and gene flow may be halted by impoundments and road crossings in two key ways. First, migration between neighboring streams may simply be blocked as a dam or road crossing acts as a physical barrier. Second, the formation of unsuitable habitat by a reservoir or road crossing may also reduce dispersal and gene flow (Skalski, Landis, Grose & Hudman, 2008; Lamphere & Blum, 2012; Perkin,

Gido, Al-Ta’ani, O., & Scoglio, 2013). Road crossings with rapid drops in streambed elevation at the outflow end (perching) represent semi-permeable barriers to fish dispersal dependent on streamflow (Norman, Hagler, Freeman & Freeman, 2009; Nislow, 2011).

Road crossings are more numerable than dams, and 53 to 97% form semi-permeable barriers to fish passage (Gibson, Haedrich & Wernerheim, 2005; Poplar-Jeffers et al.,

2009). Therefore, it is well established that barriers such as road crossings disrupt functional connectivity in river systems, thus affecting metapopulation (Perkin, Gido, Al-

102

Ta’ani, O., & Scoglio, 2013; Gido, Whitney, Perkin & Turner, 2016) and metacommunity

(Perkin & Gido, 2012) dynamics.

In our study, we use a comparative framework to examine whether functional connectivity in a dendritic river system affects how two co-occurring, headwater darters relate to environmental (stream size, groundwater input) and anthropogenic (road crossing density, presence/distance of reservoir, and land cover) gradients in a dendritic river system. We modeled occupancy and genetic distance using an information theoretic approach to explore whether functional connectivity modified how either species related to these gradients. Genetic distance was used as a surrogate of gene flow, to identify how a resistance gradient may affect patterns of colonization and extinction. We also tested whether the influence of these resistance gradients on genetic distance may vary dependent on subbasin size and species. The southeastern United States is the center of diversity for darters; a speciose group of benthic, freshwater fishes which are often headwater specialists

(i.e., functional trait databases place 87 of 250 species in springs, or headwater habitats,

Frimpong & Angermeier, 2009). We predicted (1) as headwater specialists, patterns of genetic distance should support the stream hierarchy model, (2) functional connectivity should alter the influence of environmental and anthropogenic gradients on genetic distance in relation to occupancy, (3) as darter occupancy increases along an environmental gradient, genetic distance among darter subpopulations should decline; and (4) subbasin size should alter functional connectivity thus influencing the magnitude of environmental and anthropogenic gradients on genetic distance.

Methods

Study System

103

We conducted our study in the Upper Yazoo River basin which is positioned in North-

Central Mississippi in the southeastern United States. The upper Yazoo River basin consists of two subbasins, the Little Tallahatchie River subbasin (hencehforth, LTRSB) and the Yocona River subbasin (henceforth, YRSB). The Little Tallahatchie and Tippah

River systems drain the LTRSB, and Otoucalofa Creek and the Yocona River system drain the YRSB. Both subbasins are characterized by the presence of large reservoirs (LTRSB:

Sardis Reservoir, 117 km²; YRSB: Enid Reservoir, 59 km²). In the LTRSB and YRSB, much of the surrounding land cover is forested (LTRSB: 50.8%; YRSB: 56.6%); however, there is a strong agricultural and pastoral influence (LTRSB: 32.1%; YRSB: 27.9%) and little land area is urbanized (LTRSB: 5.4%; YRSB: 5.8%). Land cover in this river basin is rapidly changing. Population growth in Lafayette County (county seat, Oxford) increased by 22.76% from 2010 to 2017 (U.S. Census Bureau, 2010-2017), thus promoting an increase in the number of culverts and extensive channel incision throughout this river basin as road densities multiplied.

Study Species

Two darters, the Yazoo Darter Etheostoma raneyi and Goldstripe Darter Etheostoma parvipinne were chosen for this study. The Goldstripe Darter (Bart & Taylor, 1999) is broadly distributed across the southeastern United States, and the Yazoo Darter is endemic to the Upper Yazoo River basin (Thompson & Muncy, 1986; Suttkus, Bailey & Bart, 1994) and listed as vulnerable by the Southeastern Fishes Council (Warren et al., 2000) and the

American Fisheries Society (Jelks et al., 2008). Recently, the Yazoo Darter was formally split into two species (Sterling & Warren, 2020) in accordance with the stated criteria of the unified species concept (de Queiroz, 2007). Much historical evidence has suggested

104

that the LTRSB and YRSB Yazoo Darter populations represent two distinct lineages

(Powers & Warren, 2009; Sterling, Reed, Noonan & Warren, 2012; Sterling et al., 2020).

The newly described species, the Yoknapatawpha Darter, Etheostoma faulkneri is endemic to the YRSB. However, given their recent divergence (0.4 – 0.8 my, Sterling et al., 2020), and almost identical habitat use of the two lineages, we chose two treat them as a singular species for ease of interpretation. Goldstripe and Yazoo Darters are constrained to small and medium sized streams (Smiley, Dibble, & Schoenholz, 2006; Sterling, Warren, &

Henderson, 2013), and co-occur throughout this study area, presumably due to niche differences (Hubbell, Schaefer, Warren & Sterling, 2020). As headwater specialists, populations across drainages exhibit genetic substructure (Sterling, Reed, Noonan &

Warren, 2012; Bjorn & Schaefer, 2018). However, the Yazoo Darter is less tolerant of variability in streamflow and has a restricted range throughout the basin (Adams & Warren,

2005; Hubbell, Schaefer, Warren & Sterling, 2020).

Datasets

Habitat and associated fish assemblage stability in this system (see Hubbell, Schaefer,

Warren & Sterling, 2020) allowed us to model the occupancy of Goldstripe Darters with presence-absence data collected at 61 historic (1999-2003; Sterling, Warren, & Henderson,

2013) and 43 contemporary (2015-2016) sites within the LTRSB. Because the number of sites sampled within the YRSB was low, we did not include these sites when modeling the occupancy of either species. Because Yazoo Darters are not known to occupy the easternmost portion of the LTRS (i.e., the uppermost Little Tallahatchie River, see Hubbell

& Schaefer 2017), we used a subset of sites (59 historic, 21 contemporary) to model occupancy of Yazoo Darters. Subreaches were used as an alternative to re-sampling over

105

time to generate detection histories (Albanese, Peterson, Freeman & Weiler, 2007) for historic and contemporary sites (sampled May-September; see Hubbell et al., 2020).

Yazoo Darter tissue (106 individuals) samples were collected at seven sites within the

YRSB and 17 sites within the LTRSB, and Goldstripe Darter tissue (84 individuals) samples were collected at 12 sites within the YRSB and 16 sites within the LTRSB

(sampled May-August, 2016 & 2017, Figure. 4.1). Fin clips were preserved in a tissue preservation buffer (Seutin et al., 1991) and returned to the laboratory where they were stored at -20ºC.

Genetic Analyses

Single Nucleotide Polymorphisms (SNPs) were identified through Genotype by

Sequencing (GBS) (Elshire et al., 2011; Narum et al., 2013) for the evaluation of population structure, population demographics and dispersal. Following the GBS strategy a single restriction enzyme (Eco T22I) is used for the digestion of genomic DNA, adapters are then ligated to the DNA fragments, allowing for PCR amplification and sequencing.

We extracted genomic DNA using the Qiagen DNeasy Blood and Tissue kit (Qiagen)

Paired end sequencing data, consisting of 100 bar coded reads, was then analyzed with the

TASSEL GBS pipeline (Bradbury et al., 2007; Glaubitz et al., 2014). First, TASSEL pipelines identify all unique reads, eliminate any that do not meet the 100 bp length requirement, and then assemble the remaining reads into a list. Goldstripe Darter reads were then aligned using the Orangethroat Darter Etheostoma spectabile genome as a reference (Moran, Catchen & Fuller, 2020), and Yazoo Darter reads were aligned using the

Tallapoosa Darter genome as a reference (dartergenomics.org). We used bowtie2 alignment which requires an entire read to align without clipping bps from sequence ends

106

Goldstripe Darter

Yazoo Darter

Both

Figure 4.1 Map of 39 localities where tissue data for Yazoo Darters, Goldstripe Darters, or both species was collected across the Little Tallahatchie and Yocona River subbasins. Inset map identifies the locality of the upper Yazoo River basin. Shapes (star, circle, or triangle) identify which species were detected at each site.

TASSEL was used to process the raw sequence data and calculate the binomial likelihood of heterozygotes using read counts and an expected error rate of 0.01 (Bradbury et al.,

2007). A hapmap file was then exported from TASSEL and then further filtering of loci and individuals was performed using R (version 3.2.3, R Development Core Team, Vienna,

Austria). SNP genotypes were filtered based on the following criteria: the elimination of any loci that contained ≥ 25% missing data; elimination of any loci with extreme (> 50%) heterozygosity, and finally removal of loci within 100 bp to reduce linkage. Our filtering processes yielded a final Yazoo Darter dataset consisting of 7541 loci and a final Goldstripe

Darter dataset consisting of 9541 loci.

We assessed hierarchical population genetic structure in both species using

STRUCTURE (Pritchard, Stephens & Donnelly, 2000) v. 2.3.4. For both species, we ran

10 replicates of K ranging from 1-10 (750,000 Markov Chains with a burnin of 350,000), followed by the Evanno method (Evanno, Regnaut & Goudet, 2005) (Structure Harvestor,

Earl, 2012) to identify the best value of K. Additional STRUCTURE analyses were performed on these genetic groups using the parameters described above to determine if additional genetic structure was present. For the best value of K in each analysis, we calculated the average ancestry coefficient (q) for all individuals from the 10 runs with

CLUMPP v. 1.1.2 (Jakobsson & Rosenberg, 2007). We used the program Structure Plot

(vs. 2.0) (Ramasamy, Ramasamy, Bindroo & Naik, 2014) to visualize CLUMPP results.

Descriptive population genetics statistics (mean allelic richness, observed and expected heterozygosity, mean inbreeding coefficient) and pairwise FST values using the R packages

“pegas”, “adegenet”, and “hierfstat” on the groups identified by the STRUCTURE analyses.

108

Following STRUCTURE analysis, we considered population clusters as unique populations and estimated migration rates using the program Migrate 3.6.4 (Beerli &

Felsenstein, 2001). A coalescent method is implemented by this program to estimate mutation-scaled migration rates over the last 4Ne generations (approximately 100 to 500 years, dependent on a species generation time) for each population (Beerli, 2010). For both species, we specified a full migration model, and estimated parameters using Bayesian inference with uniform priors. For both species, we performed ten replicate Migrate runs.

Each replicate consisted of 4 heated Markov-chains, 1,600,000 in length sampled over 20 steps after discarding an initial 75,000 (constant mutation rate for all loci). To assess model convergence, we inspected acceptance ratios, ESS values, and posterior distributions for all model parameters.

For both darters, a second set of pairwise FST values were calculated to represent estimates of genetic distance between pairs of sites in the YRSB and LTRSB where we sampled ≥5 individuals. Genetic differentiation between subpopulations may be successfully measured with low sample sizes when SNP datasets consist of a large (>

1,000) number of bi-allelic sites (Willing, Dreyer & Oosterhout, 2012). These pairwise values were used as response variables in hierarchical models. Pairwise Fst values for sites

(3) sampled downstream of Sardis Reservoir were not quantified, because specific covariate effects could not be estimated for these localities.

Processing of Geospatial Data

We extracted environmental (drainage area, waterbodies, groundwater input) and anthropogenic (land cover, road crossing density, reservoir distance) data for inclusion as covariates in our hierarchical models; all data processing was performed in QGIS 3.4.14

109

(2019). Hubbell, Schaefer, Warren & Sterling (2020) identified drainage area and groundwater input as the two most important environmental covariates on Goldstripe and

Yazoo Darter occupancy across four spatial scales. We retrieved drainage area and water bodies from the National Hydrography Dataset Plus (USEPA & USGS, 2006), and the road layer data was retrieved from the Mississippi Department of Transportation (MDOT,

2006). We included well depth (MDWQ, 2015) as a surrogate measure of water table depth to infer stream permanence and relative groundwater input. Well depth is correlated with water table depth across a watershed (Rosenberry, LaBaugh & Hunt, 2008). We used nearest neighbor interpolation to assign well depth, vector point data to all stream segments within the LTRSB and YRSB. We defined a stream-segment as an individual segment when using the NHDPlus flowline vector layer. We used the line intersections and polygon-line intersections tools to convert all stream-road intersections to vector point data. We extracted land cover data for 2001 and 2016 from the national land cover dataset

(NLCD, https://www.mrlc.gov/data). These years were used because they were the closest to when sites were sampled. NLCD data includes 20 classes of land cover that we reclassified into 5 broad land cover variables: forested, urban, wetland, open water, and agricultural.

Two versions of each covariate were quantified to represent the presence or absence of functional connectivity effects (Table 4.1). We extracted reclassified land cover data using two different approaches: 1) a 200 m stream buffer was used to calculate the percentage of anthropogenic land cover associated with each stream-segment to represent functional connectivity effects, and 2) the percentage of anthropogenic land cover associated with each site’s upstream drainage area was used to represent the absence of functional

110

connectivity effects. The percentage of anthropogenic land cover was calculated as the sum of the percentages of urban, and agricultural land cover. We calculated mean stream- segment estimates of well depth, anthropogenic land cover, and road crossing density using the join by location tool. For each of these variables and drainage area, we calculated mean total pathway estimates to represent functional connectivity effects on genetic distance. We extracted independent values of drainage area and mean stream-segment estimates of road crossing density and well depth for all site localities to represent the absence of functional connectivity effects on occupancy. For occupancy modeling, reservoir effects were measured as a site’s river distance to reservoir (i.e., river distance (km) between a site and the dam forming the reservoir). When modeling genetic distance, the influence of large reservoir habitat was measured as the fraction of the total pathway distance that crossed either large impoundment. In addition to these other variables, we also quantified pathway distance (i.e., the total river distance between pairs of subpopulation account for patterns of isolation by distance exhibited by either species in the YRSB and LTRSB. Prior to model construction, we tested for the correlation between covariates. Any two covariates that had a Pearson correlation greater than the absolute value of 0.5 were not included in the same model.

Hierarchical Modeling

We used single-species models (MacKenzie et al., 2002) to characterize environmental and anthropogenic gradients associated with the probability of occurrence (Ψ) of both darters within the LTRSB. Because spatial replicates may not represent truly independent surveys and lead to the inflation of occupancy estimators (Kendall & White, 2009), we also developed spatial dependence models (Hines et al., 2010). Spatial dependence models

111 Table 4.1

Definitions and a priori hypotheses (+/-) for each covariate for occupancy and genetic distance. Covariates means for genetic distance were estimated using the total pathway distance (km) between pairs of sites. Covariate Definition Effect (+/-)

Drainage Area (km²) Occupancy: drainage area of stream segment associated with site - locality Genetic Distance: mean drainage area of total pathway between pairs - of sites

Well Depth (ft.) Occupancy: mean well depth of stream-segment associated with site - locality Genetic Distance: mean well depth between pairs of sites -

Anthropogenic Land Cover (%) Occupancy: percent anthropogenic land cover upstream of site -

Genetic Distance: mean percentage of anthropogenic land cover - between pairs of sites

Road crossing Density Occupancy: mean road crossing density of stream-segment associated - with site locality

Genetic Distance: mean percentage of anthropogenic land cover - between pairs of sites

Reservoir Occupancy: river distance (km) between site and reservoir -

Genetic Distance: fraction of distance (km) traversing reservoir habitat - between pairs of sites

allow the probability that a spatial segment may or may not be occupied based upon whether the previous segment was occupied (θ’) or not (θ) (Hines et al., 2010) where parameters are modeled as a first-order Markov process. Preliminary analyses suggested that Yazoo Darter detections were spatially independent, but Goldstripe Darter detections were not. Thus, Goldstripe Darter detection data was modeled using the spatial dependence model. We modeled the probability of detection (p) as constant and as a function of sampling covariates (see Hubbell, Schaefer, Warren & Sterling, 2020) to distinguish if p varied among sites. We used untransformed beta estimates to infer relationships (positive or negative) between covariates and parameters. We used the MacKenzie-Bailey goodness of fit test (MacKenzie & Bailey, 2004) to test model fit and preliminary analyses indicated no evidence of a lack of fit (Yazoo Darter: p = 0.26, ĉ =1.19; Goldstripe Darter: p = 0.62).

We constructed all occupancy models using the software program PRESENCE (vers. 12.7)

(Hines, 2006).

We used linear mixed effects modeling (r-package “lme4”) to examine how functional connectivity effects influenced the strength of environmental and anthropogenic gradients on genetic distance within the LTRSB. Linear mixed effects models were also used to assess whether the influence of these gradients on genetic distance may vary dependent on differences in subbasin size between the LTRSB and YRSB. Within-drainage pairwise Fst values were linearized with the transformation of FST/(1- FST). (Rousset, 1997). We constructed four model sets (two species X two subbasins). The LTRSB model sets were constructed to assess whether a species response to an environmental or anthropogenic gradient shifted in the presence of functional connectivity effects, and the YRSB model sets were constructed to assess how subbasin size may influence the strength of any

113

environmental or anthropogenic gradient on genetic distance. For all linear mixed effects model sets, population (i.e., site ID) was included as a random-intercept to account for the non-independence of the pairwise data (Clarke, Rothery, & Raybould, 2002). Variance explained by fixed effects was determined by calculating the marginal coefficient of determination (mR²) and variance explained by random effects was resolved by quantifying the conditional coefficient of determination (cR²).

Model Selection

For all hierarchical model sets, we used Akaike’s Information Criterion for small sample sizes (AICC) to assess the quality of competing models (Burnham & Anderson,

2002). Models with small ∆AICC and large Akaike weights (wᵢ) indicate greater parsimony

(Burnham & Anderson, 2002). We only interpreted models with wᵢ > 0.10. To prevent the inclusion of uninformative parameters, models which only differed in ∆AICC by 1-2 units from the best models and possessed similar log-likelihood values were removed (Burnham

& Anderson, 2002).

Results

Population Structure and Descriptive Genetic Statistics

Our initial STRUCTURE runs suggested that most of the genetic variation in both species was a result of hierarchical structuring between the two subbasins. Thus, for our first STRUCTURE run for both darters, we obtained a K of 2 (LTRSB, YRSB). Standalone

STRUCTURE runs of these K=2 groupings when consulting q scores elucidated that

Yazoo Darters sampled from drainages within the LTRSB could be divided into four groups (Hotopha; Little Tallahatchie, Cypress; Tippah) and Yazoo Darters sampled from the YRSB could be partitioned into two groups (Yocona, Otoucalofa), thus producing a K

114

of 6 (Figure C1, Figure C2). The Hotopha grouping represents the lone Yazoo Darter population that is known downstream of Sardis Reservoir within the LTRSB. When referencing q scores, our results suggested that admixture was present in Yazoo Darters sampled from Hurricane Creek (Little Tallahatchie grouping); individuals sampled from this drainage shared similar lineages with the Tippah and Hotopha groupings. Therefore, our results suggest there was recent gene flow between the Hotopha and Little Tallahatchie groupings and between the Little Tallahatchie and Tippah groupings. In the YRSB, there was no evidence of recent gene flow between the Yazoo Darter Yocona and Otoucalofa groupings.

Independent STRUCTURE runs of our K=2 Goldstripe Darter groupings (LTRSB;

YRSB) exposed that the LTRSB hierarchical grouping could only be divided into two groups (Little Tallahatchie, Tippah). Our results from this initial run also indicated that

Goldstripe Darters sampled from streams downstream of Sardis Reservoir were more genetically similar to individuals sampled from the YRSB. Therefore, we performed an independent STRUCTURE run to assess further hierarchical structure between Goldstripe

Darters sampled from downstream of Sardis Reservoir and Goldstripe Darters sampled from the YRSB. Our results for this run suggested that samples could be partitioned into three groups (Downstream of Sardis Reservoir, Yocona, Otoucalofa), thus yielding an overall K of 5 (Figure C3, Figure C4). Bar plots and q scores indicated the existence of admixture in Goldstripe Darters sampled from Big Spring Creek (Little Tallahatchie grouping); individuals from this drainage appear to share a common ancestry with the

Tippah grouping. Thus our results suggests there was recent gene flow between the

Goldstripe Darter Little Tallahatchie and Tippah groupings. We found no evidence of

115

recent gene flow between Goldstripe Darter groupings in the Downstream of Sardis

Reservoir, Yocona, and Otoucalofa groupings.

Estimates of expected heterozygosity and mean allelic richness were similar across all five Goldstripe Darter groupings (Table 4.2), but values for both of these parameters were highly variable among the six Yazoo Darter groupings. Generally, estimates of FIS were greater for Goldstripe Darter groupings than Yazoo Darter groupings (Table 4.2). In both darters, genetic differentiation was greatest between any pair of LTRSB and YRSB groupings; however, pairwise FST values were more variable for Yazoo Darter groupings in relation to Goldstripe Darter groupings (Tables 4.3 & 4.4).

Migration

Estimates of effective population size and Nm suggests that migration between hierarchical groupings of both species has historically been asymmetric (Tables C1, C2,

C3, & C4). The Little Tallahatchie (theta = 0.0222, 0.021 – 0.024 95% credible interval) and Tippah (0.0156, 0.0139 – 0.017) Yazoo Darter groupings were typified by the largest mutation-scaled effective population sizes for this species. Effective population sizes for

Yazoo Darter groupings in the (0.004, 0.003 – 0.006), Yocona (0.003, 0.001 – 0.004), and

Otoucalofa (0.006, 0.005 – 0.008) were approximately a third as large as the estimates for the Little Tallahatchie and Tippah groupings. We found no evidence that gene flow occurred between the greater YRSB and LTRSB Yazoo Darter populations; estimates of

Nm and their credible intervals indicated that the number of migrants exchanged per generation between all possible combinations of YRSB and LTRSB groupings were consistently < 1. However, our results indicate that there was historic asymmetrical gene flow between the Below Sardis Reservoir and YRSB Goldstripe Darter groupings (i.e.,

116 Table 4.2

Expected heterozygosity, mean allelic richness, and Fis of Yazoo Darter and Goldstripe Darter STRUCTURE hierarchical groupings.

Species Grouping Expected Observed Mean Allelic Mean FIS Heterozygosity Heterozygosity Richness Yazoo Darter Little Tallahatchie 0.24 0.19 1.53 0.22 Tippah 0.21 0.18 1.48 0.14 Yocona 0.06 0.05 1.13 0.12

Hotopha 0.10 0.10 1.23 0.04 Cypress 0.15 0.13 1.34 0.15 Otoucalofa 0.08 0.07 1.17 0.20

Goldstripe Little Tallahatchie 0.25 0.20 1.25 0.19 Darter Tippah 0.22 0.16 1.23 0.24 Yocona 0.21 0.16 1.22 0.26 Otoucalofa 0.21 0.17 1.22 0.20 Below Sardis 0.26 0.22 1.28 0.18

Table 4.3

Pairwise Fst values for K=5 STRUCTURE, Goldstripe Darter hierarchical groupings. Outocalofa Yocona Little Tallahatchie Tippah Below Sardis Otoucalofa 0 0.03 0.51 0.54 0.14 Yocona 0.03 0 0.51 0.54 0.13 Little Tallahatchie 0.51 0.51 0 0.02 0.36 Tippah 0.54 0.54 0.02 0 0.41 Below Sardis 0.14 0.13 0.36 0.41 0

118

Table 4.4

Pairwise Fst values for K=6 STRUCTURE, Yazoo Darter hierarchical groupings. Little Tippah Hotopha Cypress Yocona Otoucalofa Tallahatchie Little 0 0.15 0.35 0.24 0.66 0.65 Tallahatchie Tippah 0.15 0 0.42 0.33 0.69 0.68 Hotopha 0.35 0.41 0 0.56 0.81 0.80 Cypress 0.24 0.33 0.56 0 0.76 0.75 Yocona 0.66 0.69 0.81 0.76 0 0.34 Otoucalofa 0.65 0.68 0.80 0.75 0.34 0

119

Yocona, Otoucalofa), thus validating our STRUCTURE results. Individuals from the

Goldstripe Otoucalofa and Yocona groupings migrated into streams in the lower Little

Tallahatchie River system. In the LTRSB, a pattern of asymmetric gene flow is present among Goldstripe Darter groupings, individuals migrated downstream from the Tippah river system into streams draining directly into the Little Tallahatchie River. An opposite pattern appears to be present among Yazoo Darter groupings in the LTRSB; Yazoo Darters migrated upstream from the Little Tallahatchie grouping into streams within the Tippah

River system. In the YRSB, gene flow was apparently negligible between the Otoucalofa and Yocona groupings of both species. An opposite pattern appears to be present among

Yazoo Darter groupings in the LTRSB; Yazoo Darters migrated upstream from the Little

Tallahatchie grouping into streams within the Tippah River system. In the YRSB, gene flow was apparently negligible between the Otoucalofa and Yocona groupings of both species.

Hierarchical Modeling

Rankings of occupancy and linear mixed effects models suggested that functional connectivity strongly affected how either species related to environmental and anthropogenic gradients (Table 4.5). In the absence of functional connectivity, Goldstripe and Yazoo Darter occupancy in the LTRSB were both best modeled (wi > 0.10) as the function of a single environmental covariate (Table 4.5, Figure 4.2). Yazoo Darter occupancy was negatively related to well depth (beta estimate = -0.34 ± 0.12), and

Goldstripe Darter occupancy was negatively related to drainage area (beta estimate = -1.33

± 0.45).

120

Anthropogenic gradients were better predictors of Goldstripe and Yazoo Darter genetic distance in the smaller YRSB subbasin (Table 4.5). When modeling Yazoo Darter genetic distance in the LTRSB, the best model (mR² = 0.53; cR² = 0.68) explained variation in genetic distance as a consequence of the additive effects of well depth (0.04 ± 0.01) and pathway distance (0.10 ± 0.01) (Figure 4.3), but Yazoo Darter genetic distance in the YRSB was best modeled (mR² = 0.82; cR² = 0.84) as a consequence of the additive effects of drainage area (0.69 ± 0.005) (Figure 4.3) and road crossing density (0.09 ± 0.05). Genetic distance of the Goldstripe Darter in the YRSB was bested modeled (mR² = 0.38; cR² =

0.53) as the summation of the additive effects of anthropogenic land cover (0.006 ± 0.0002) and pathway distance (0.01 ± 0.002) (Figure 4.4). When modeling Goldstripe Darter genetic in the LTRSB, three models had similar support (Table 4.5). The best model indicated variation in Goldstripe Darter genetic distance was best explained (mR² = 0.15; cR² = 0.42) as the summation of the additive effects function of road crossing density

(0.003 ± 0.0003) (Figure 4.4) and well depth (0.002 ± 0.0003), whereas the second best model (mR² = 0.14; cR² = 0.42) described this variance as the summation of the additive effects of pathway distance (0.02 ± 0.007) and well depth (0.01 ± 0.007).

Discussion

Functional connectivity regulates the degree of influence an environmental or anthropogenic gradient imposes on stream fish dispersal, thus modifying metapopulation and meta-assemblage dynamics (Perkin & Gido, 2012; Perkin, Gido, Al-Ta’ani, O., &

Scoglio, 2013; Schimdt & Schaefer, 2018). Therefore, the exclusion of landscape heterogeneity between patches within occupancy models suggests we may have a restricted

121 Table 4.5

Yazoo and Goldstripe Darter hierarchical model sets for occupancy and genetic distance (Fst). Two linear mixed effects model sets were constructed for genetic distance to elucidate the effects of subbasin size (Little Tallatchie and Yocona) on functional connectivity. Species Response Subbasin Model K AICC ∆AICC wi

Yazoo Occupancy Little Tallahatchie p (.), Ψ (Well Depth) 3 327.30 0 0.95 Darter Fst Little Tallahatchie Well Depth + Pathway Distance + 1|Pop 5 -155.1 0 0.80

Fst Little Tallahatchie Pathway Distance + Reservoir Distance + 5 -151.8 3.2 0.16 1|Pop

Fst Yocona Drainage Area + Road crossing Density + 4 31.8 0.0 0.40 1|Pop

Fst Yocona Drainage Area + 1|Pop 4 32.0 0.2 0.36

Fst Yocona Reservoir Distance + 1|Pop 4 34.1 2.4 0.12

Goldstripe Occupancy Little Tallahatchie p (Depth + Velocity), Ψ (Drainage Area) θ(.) 5 473.22 0 0.97 Darter θ’(.)

Fst Little Tallahatchie Well Depth + Road crossing Density + 1|Pop 5 -160.6 0.0 0.27

Fst Little Tallahatchie Well Depth + Pathway Distance + 1|Pop 5 -160.1 0.5 0.21

Table 4.5 (continued)

Species Response Subbasin Model K AICC ∆AICC wi Fst Little Tallahatchie Pathway Distance + 1|Pop 4 -159.2 1.3 0.14

Fst Yocona Anthropogenic Land Cover + Pathway 5 -376.6 0 0.69 Distance + 1|Pop Fst Yocona Reservoir Distance + Anthropogenic Land 5 -373.1 3.5 0.12 Cover + 1|Pop

A B

Figure 4.2 Plots of probability of occurrence (ψ) for Goldstripe and Yazoo Darters. Goldstripe Darter occupancy was best predicted as a function of drainage area (A) whereas Yazoo Darter occupancy was best predicted as a function of well depth (B).

A B

understanding of how environmental and anthropogenic gradients influence the patch occupancy and colonization and extinction probabilities of many stream fishes. The interplay between dendritic structuring and landscape heterogeneity influences functional connectivity, limiting gene flow in stream fishes (Brauer et al., 2018; Schimdt & Schaefer

2018; Nathan, Welsh & Vokoun, 2019). However, to our knowledge, no empirical studies have investigated how the exclusion of dendritic structuring in models of stream fish metapopulation parameters may influence model inferences. We juxtaposed how the occupancy and genetic distance of two co-occurring, headwater darters related to environmental and anthropogenic gradients to evaluate the influence of functional connectivity on these fishes’ ecological response in a dendritic river system. Our results suggest that, (1) patterns of genetic distance in both species support the predictions of the stream hierarchy model (2) functional connectivity modified the impact of environmental and anthropogenic gradients on genetic distance relative to site-occupancy, (3) while

Goldstripe and Yazoo Darter occupancy was best explained by environmental gradients, habitat specialization accentuated the degree of influence an environmental gradient exerted on genetic distance; and (4) given the influence of functional connectivity, the strength of anthropogenic gradients were a consequence of subbasin size. Given our results, it is evident that site-occupancy models may underestimate the influence of environmental or anthropogenic gradients on the patch occupancy of aquatic organisms in dendritic river systems unless functional connectivity is an integral component of the modeling framework. Thus, we support the development of novel modeling approaches which incorporate dendritic structure to examine how landscape heterogeneity influences stream fish metapopulation dynamics.

126

B A

Isolation By Distance

Our results suggest that patterns of migrant exchange in Yazoo and Goldstripe Darters follow the stream hierarchy model. Pathway distance was included among the best models of Yazoo Darter genetic distance in the LTRSB. Similarly, pathway distance was included among the best models for both subbasins when modeling the genetic distance of the

Goldstripe Darter. Thus our results conflict with the findings of Schimdt & Schaefer (2018) who documented no discernible patterns in Goldstripe Darter genetic distance as an effect of isolation by distance. Differences in data type (SNP versus microsatellite) are a likely explanation for these conflicting findings.

Dendritic Structuring Has Repercussions for Metapopulation Dynamics

Dendritic structuring modifies population connectivity, thus influencing stream fish metapopulations and meta-assemblages, and many theoretical models (Campbell-Grant,

Lowe & Fagan, 2007; Campbell-Grant, 2011; Brown et al., 2011) have been developed to describe these ecological processes. Our prediction that functional connectivity should modify the relationship between a landscape gradient and genetic distance relative to site- occupancy was supported. Thus, our findings suggest that the interplay between dendritic structure and landscape gradients influences stream fish metapopulations. Because patch occupancy models omit dendritic structure effects, it is apparent that any inferences made about stream fish metapopulation dynamics may be tenuous. Recent advances (Howell et al., 2018) in wildlife ecology have unified the fields of metapopulation ecology and landscape ecology, developing a modeling framework which considers a species occupancy, extinction, and colonization probabilities to be a consequence of differences in functional connectivity across a landscape. Howell et al. (2018) integrates least cost path

128

analysis (i.e., identification of the path with the lowest cost-weighted distance between two localities, Dijkstra, 1959) to expand the biological realism of the model; however, the use of this approach to explain landscape heterogeneity as a derivative of dendritic structuring would be inappropriate. While patch occupancy models of colonization and extinction

(e.g., MacKenzie et al., 2003), are well adapted to enhance our understanding of how patch dynamics influences stream fish metapopulations, the exclusion of dendritic structure from such models inhibits ecologists from formulating predictions about how connectivity influences stream fish populations. The interaction between a dendritic river system’s size and topological complexity have repercussions for stream fish metapopulations (Campbell-

Grant et al., 2011). The size of a dendritic river system determines the number of suitable patches available to a species while the arrangement of habitat patches modifies functional connectivity (Labbe & Fuasch 2000, Fausch, Torgersen, Baxter & Li, 2002; Wiens, 2002).

Habitat Specialization Emphasized When Modeling Genetic Distance

Relative to site occupancy models, differences in habitat specialization were emphasized when relating Yazoo Darter and Goldstripe Darter genetic distances to environmental gradients. Variability in the degree of habitat specialization influences how the genetic distances of stream fishes relate to environmental gradients (Heithaus &

Laushman, 1997; Turner & Robinson, 2006). As habitat specialization increases, connectivity declines, decreasing gene flow among subpopulations (Whiteley, Spruell &

Allendorf 2004; Schimdt & Schaefer, 2018). Although both species are identified as headwater specialists (Smiley, Dibble, & Schoenholz, 2006; Sterling, Warren, &

Henderson, 2013), the range of the Yazoo Darter is more restricted due to this species dependency on groundwater-fed streams (Sterling, Warren, & Henderson, 2013; Hubbell,

129

Schaefer, Warren & Sterling, 2020). Thus, Yazoo Darter occupancy declined with decreasing groundwater input (approximated by the variation in well depths in this study).

Similarly, Yazoo Darter genetic distance was best modeled by the additive effects of well depth and pathway distance, whereby both covariates were positively related to genetic distance. Many rheophilic specialists are more sensitive to fragmentation relative to eurytopic species (Musil et al., 2012). Goldstripe Darter occupancy declined with increasing stream size. However, we found no clear evidence that Goldstripe Darter genetic distance was strongly related to stream size in the LTRSB or YRSB. In the LTRSB, fixed effects in the two best supported models explained no more than 15% of the variance in

Goldstripe Darter genetic distance, and no environmental gradient was included within the best YRSB models for this species. Goldstripe Darters regularly occupy, colonize, and persist in small ephemeral drainages (Adams & Warren, 2005; Smiley, Dibble, &

Schoenholz, 2006). Thus, environmental gradients such as groundwater input and stream size may act as limited impediments on gene flow in this species.

Subbasin Size Alters Functional Connectivity

Subbasin size altered the influence of anthropogenic gradients on Yazoo and Goldstripe

Darter genetic distance. Drainage size influences resource availability, connectivity, and patterns of disturbance, thus influencing metapopulation responses (Dunham & Rieman

1999; Huntsman, Petty, Sharma & Merriam, 2016). Alterations to gene flow and genetic drift can be safely inferred as a consequence of human-induced fragmentation because alterations to genotype frequencies by mutation often takes thousands of years (Gido,

Whitney, Perkin & Turner, 2016). In our occupancy models and models of genetic distance in the LTRSB, Yazoo Darters were not strongly influenced by any anthropogenic gradient.

130

Similarly, Goldstripe Darter occupancy was not strongly related to any anthropogenic gradient; however, there was weak support (wi = 0.27, mR² = 0.15) that Goldstripe Darter genetic distance was positively related to road crossing density in the LTRSB.

In the YRSB, there were positive relationships between Yazoo Darter genetic distance and road crossing density, and Goldstripe Darter genetic distance and anthropogenic land cover. Thus, our results provide further support that physical barriers and modifications to land cover along dispersal corridors negatively impact the genetic integrity of stream fishes

(Prunier et al., 2017; Nicol, Stevens & Jobling, 2017). Mean road crossing density between

Yazoo Darter subpopulations in the YRSB (31.9 ± 17.49) and LTRSB (37.7 ± 13.77) were comparable; however, pathway distances between subpopulations were not (LTRSB: 52.5 km ± 21.1 km; YRSB: 26.3 km ± 17.2 km). Similarly, the percentage of anthropogenic land cover in the YRSB (27.2% ± 6.2%) and the LTRSB (25.5% ± 17.6%) between

Goldstripe Darter subpopulations was also comparable; but pathway distances between these subpopulations were greater in the LTRSB (54.5 km ± 25.5 km) relative to the YRSB

(42.2 km ± 21.2 km). Two non-exclusive possibilities can explain these results. Because the LTRSB is approximately 3X larger than the YRSB, any effect of anthropogenic disturbance may be overshadowed by groundwater input and isolation by distance gradients. The contemporary distributions of both darters within the LTRSB may also contribute to these results. Within the LTRSB, many of the stream reaches occupied by

Yazoo Darters occur on or are within 2 km of land managed by state or federal agencies

(Sterling et al., 2011). Similarly, within the LTRSB, Goldstripe Darters have historically been detected within highly forested catchments (78.5% ± 14.8%). Within the upper

LTRSB, there is approximately 632.29 km² of public land, but in the YRSB public land

131

only comprises an area of approximately 227.61 km². These discrepancies in land cover between the two subbasins likely affected the magnitude of anthropogenic effects on Yazoo and Goldstripe Darter genetic distance. Limited sample size prevented us from modeling the occupancy of either darter in the YRSB. Therefore, it is difficult to ascertain whether or not Goldstripe or Yazoo Darter occupancy would be strongly related to these anthropogenic gradients.

Influence of Reservoirs on Genetic Distance

Though beyond the scope of this study, our findings that Enid and Sardis Reservoirs had a limited effect on the genetic distance in upstream subpopulations of either species is noteworthy. There was slight evidence that Enid Reservoir influenced genetic distances between upstream subpopulations of both species in the YRSB. However, there was no indication that Sardis Reservoir has served as a substantial barrier to gene flow among upstream subpopulations of either species in the LTRSB. The study of the influence of reservoirs on the genetic distance of headwater fishes has yielded conflicted findings

(Skalski et al., 2008; Fluker, Kuhajda & Harris, 2014; Schimdt & Schaefer, 2018). Fluker,

Kuhajda & Harris (2014) suggested two explanations given similar findings for another benthic stream fish. The limited effect of the reservoirs on genetic distance may be a consequence of reservoir age; both reservoirs are relatively young (Sardis: 1936; Enid:

1952). Thus, despite changes to either subbasin, detectable genetic differentiation may be slow to accumulate through time (Epps & Keyghobadi, 2015). Previous research on other closely related species with short generation times (1 – 3 years; Page 1983) suggests that both darters have been subjected to this fragmented riverscape for 27 to 40 generations. A second possibility is that gene flow between subpopulations of both species has been

132

historically limited. Therefore, both darters may be capable of maintaining relatively large effective population sizes despite structuring on small spatial scales. Similar findings have been proposed for other close relatives of the Yazoo Darter (Porter, Cavender & Fuerst,

2002; Powers & Mayden, 2007; Fluker, Kuhajda & Harris, 2014). Given the large degree of hierarchical substructure in both species, we suggest that the second option is more likely.

Conclusions

Our study highlights the influence of functional connectivity on two headwater fishes’ ecological responses to environmental and anthropogenic gradients across a riverscape.

The comparative, analytical approach used in this study suggests that the use of traditional patch occupancy models may underestimate the influence of environmental and anthropogenic gradients on stream fishes. Thus, novel modeling frameworks which integrate a dendritic structure component would likely provide better insight into how landscape heterogeneity influences stream fish metapopulation dynamics. Our finding that habitat specialization influenced the relatedness between environmental gradients and genetic distance suggests that subtle niche differences may influence of patterns of gene flow, and thus patterns of colonization and extinction, and a species’ distribution. We also noted that subbasin size altered the influence of anthropogenic gradients. Thus, metapopulations may respond more rapidly to sources of human-induced fragmentation in smaller drainages. The lack of a reservoir-effect on upstream subpopulations in both species is interesting, and adds to a body of evidence suggesting that small benthic fishes can maintain their genetic integrity in spite of limited gene flow. Because our study made use of SNPs, our results help clarify findings of previous landscape genetics studies on

133

benthic headwater fishes which were limited due to the use of microsatellites (e.g., Sterling,

Reed, Noonan & Warren, 2012; Fluker, Kuhajda & Harris, 2014; Schimdt & Schaefer,

2018).

134

REFERENCES

Adams, S. B., & Warren Jr, M. L. (2005). Recolonization by warmwater fishes and

crayfishes after severe drought in upper coastal plain hill streams. Transactions of

the American Fisheries Society, 134(5), 1173-1192.

Albanese, B., Peterson, J. T., Freeman, B. J., & Weiler, D. A. (2007). Accounting for

incomplete detection when estimating site occupancy of Bluenose Shiner

(Pteronotropis welaka) in southwest Georgia. Southeastern Naturalist, 6(4), 657-

668.

Bart, H. L., & Taylor, M. S. (1999). Systematic review of subgenus Fuscatelum of

Etheostoma with description of a new species from the upper Black Warrior River

system, Alabama. Tulane Stud. Zool. Bot, 31(1), 23-50.

Bélisle, M. (2005). Measuring landscape connectivity: the challenge of behavioral

landscape ecology. Ecology, 86(8), 1988-1995.

Benda, L. E. E., Poff, N. L., Miller, D., Dunne, T., Reeves, G., Pess, G., & Pollock, M.

(2004). The network dynamics hypothesis: how channel networks structure

riverine habitats. BioScience, 54(5), 413-427.

Beerli, P., & Felsenstein, J. (2001). Maximum likelihood estimation of a migration matrix

and effective population sizes in n subpopulations by using a coalescent

approach. Proceedings of the National Academy of Sciences, 98(8), 4563-4568.

Boizard, J., Magnan, P., & Angers, B. (2009). Effects of dynamic landscape elements on

fish dispersal: the example of Creek Chub (Semotilus atromaculatus). Molecular

Ecology, 18(3), 430-441.

Bradbury, P. J., Zhang, Z., Kroon, D. E., Casstevens, T. M., Ramdoss, Y., & Buckler, E.

135

S. (2007). TASSEL: software for association mapping of complex traits in diverse

samples. Bioinformatics, 23(19), 2633-2635.

Brauer, C. J., Unmack, P. J., Smith, S., Bernatchez, L., & Beheregaray, L. B. (2018). On

the roles of landscape heterogeneity and environmental variation in determining

population genomic structure in a dendritic system. Molecular Ecology, 27(17),

3484-3497.

Brown, B. L., Swan, C. M., Auerbach, D. A., Campbell Grant, E. H., Hitt, N. P.,

Maloney, K. O., & Patrick, C. (2011). Metacommunity theory as a multispecies,

multiscale framework for studying the influence of river network structure on

riverine communities and ecosystems. Journal of the North American

Benthological Society, 30(1), 310-327.

Burnham, K. P., & Anderson, D. R. (2002). A practical information-theoretic

approach. Model selection and multimodel inference, 2nd ed. Springer, New York.

Campbell Grant, E. H. (2011). Structural complexity, movement bias, and

metapopulation extinction risk in dendritic ecological networks. Journal of the

North American Benthological Society, 30(1), 252-258.

Campbell Grant, E. H., Lowe, W. H., & Fagan, W. F. (2007). Living in the branches:

population dynamics and ecological processes in dendritic networks. Ecology

Letters, 10(2), 165-175.

Clarke, R. T., Rothery, P., & Raybould, A. F. (2002). Confidence limits for regression

relationships between distance matrices: estimating gene flow with

distance. Journal of Agricultural, Biological, and Environmental Statistics, 7(3),

361.

136

Colvin, S. A., Sullivan, S. M. P., Shirey, P. D., Colvin, R. W., Winemiller, K. O.,

Hughes, R. M., & Danehy, R. J. (2019). Headwater streams and wetlands are

critical for sustaining fish, fisheries, and ecosystem services. Fisheries, 44(2), 73-

91.

Cote, D., Kehler, D. G., Bourne, C., & Wiersma, Y. F. (2009). A new measure of

longitudinal connectivity for stream networks. Landscape Ecology, 24(1), 101-

113.

De Queiroz, K. (2007). Species concepts and species delimitation. Systematic

Biology, 56(6), 879-886.

Diez, J. M., & Pulliam, H. R. (2007). Hierarchical analysis of species distributions and

abundance across environmental gradients. Ecology, 88(12), 3144-3152.

Dijkstra, E. W. (1959). A note on two problems in connexion with graphs. Numerische

Mathematik, 1(1), 269-271.

Dunham, J. B., & Rieman, B. E. (1999). Metapopulation structure of bull trout:

influences of physical, biotic, and geometrical landscape

characteristics. Ecological Applications, 9(2), 642-655.

Earl, D. A. (2012). STRUCTURE HARVESTER: a website and program for visualizing

STRUCTURE output and implementing the Evanno method. Conservation

Genetics Resources, 4(2), 359-361.

Elshire, R. J., Glaubitz, J. C., Sun, Q., Poland, J. A., Kawamoto, K., Buckler, E. S., &

Mitchell, S. E. (2011). A robust, simple genotyping-by-sequencing (GBS)

approach for high diversity species. PloS One, 6(5), e19379.

Epps, C. W., & Keyghobadi, N. (2015). Landscape genetics in a changing world:

137

disentangling historical and contemporary influences and inferring

change. Molecular Ecology, 24(24), 6021-6040.

Evanno, G., Regnaut, S., & Goudet, J. (2005). Detecting the number of clusters of

individuals using the software STRUCTURE: a simulation study. Molecular

Ecology, 14(8), 2611- 2620.

Fagan, W. F. (2002). Connectivity, fragmentation, and extinction risk in dendritic

metapopulations. Ecology, 83(12), 3243-3249.

Fausch, K. D., Torgersen, C. E., Baxter, C. V., & Li, H. W. (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. BioScience, 52(6),

483-498.

Fluker, B. L., Kuhajda, B. R., & Harris, P. M. (2014). The effects of riverine

impoundment on genetic structure and gene flow in two stream fishes in the

Mobile River basin. Freshwater Biology, 59(3), 526-543.

Frimpong, E. A., & Angermeier, P. L. (2009). Fish traits: a database of ecological and

life-history traits of freshwater fishes of the United States. Fisheries, 34(10), 487-

495.

Gibson, R. J., Haedrich, R. L., & Wernerheim, C. M. (2005). Loss of fish habitat as a

consequence of inappropriately constructed stream crossings. Fisheries, 30(1),

10-17.

Gido, K. B., Whitney, J. E., Perkin, J. S., & Turner, T. F. (2016). Fragmentation,

connectivity and fish species persistence in freshwater ecosystems. Conservation

138

of Freshwater Fishes, 292-323.

Glaubitz, J. C., Casstevens, T. M., Lu, F., Harriman, J., Elshire, R. J., Sun, Q., & Buckler,

E. S. (2014). TASSEL-GBS: a high capacity genotyping by sequencing analysis

pipeline. PloS One, 9(2), e90346.

Gordon, N. D., McMahon, T. A., Finlayson, B. L., Gippel, C. J., & Nathan, R. J. (1992).

Stream hydrology, an introduction for ecologists The University of Chicago Press

New York.

Grenouillet, G., Pont, D., & Hérissé, C. (2004). Within-basin fish assemblage structure:

the relative influence of habitat versus stream spatial position on local species

richness. Canadian Journal of Fisheries and Aquatic Sciences, 61(1), 93-102.

Hanski, I. (1994). Patch-occupancy dynamics in fragmented landscapes. Trends in

Ecology & Evolution, 9(4), 131-135.

Hanski, I. (1999). Habitat connectivity, habitat continuity, and metapopulations in

dynamic landscapes. Oikos, 209-219.

Heithaus, M. R., & Laushman, R. H. (1997). Genetic variation and conservation of

stream fishes: influence of ecology, life history, and water quality. Canadian

Journal of Fisheries and Aquatic Sciences, 54(8), 1822-1836.

Hines, J. E. (2006). PRESENCE v 12.7- Software to estimate patch occupancy and

related parameters.http://www. mbr-pwrc. usgs. gov/software/PRESENCE. html.

Hines, J. E., Nichols, J. D. Royle, J. A., MacKenzie, D. I., Gopalaswamy, A. M., Kumar,

N. S. & Karanth K. U. (2010). Tigers on trails: occupancy modeling for cluster

sampling. Ecological Applications, 20(5), 1456-1466.

Howell, P. E., Muths, E., Hossack, B. R., Sigafus, B. H., & Chandler, R. B. (2018).

139

Increasing connectivity between metapopulation ecology and landscape

ecology. Ecology, 99(5), 1119-1128.

Hubbell, J. P., Schaefer, J. F., Warren Jr, M. L., & Sterling, K. A. (2020). Modelling

patterns of coexistence of three congeneric headwater fishes. Freshwater

Biology, 65(6), 1017-1027.

Hudman, S. P., & Gido, K. B. (2013). Multi‐scale effects of impoundments on genetic

structure of creek chub (Semotilus atromaculatus) in the Kansas River

basin. Freshwater Biology, 58(2), 441-453.

Huntsman, B. M., Petty, J. T., Sharma, S., & Merriam, E. R. (2016). More than a

corridor: use of a main stem stream as supplemental foraging habitat by a brook

trout metapopulation. Oecologia, 182(2), 463-473.

Jakobsson, M., & Rosenberg, N. A. (2007). CLUMPP: a cluster matching and

permutation program for dealing with label switching and multimodality in

analysis of population structure. Bioinformatics, 23(14), 1801-1806.

Jelks, H. L., Walsh, S. J., Burkhead, N. M., Contreras-Balderas, S., Diaz-Pardo, E.,

Hendrickson, D. A., & Platania, S. P. (2008). Conservation status of imperiled

North American freshwater and diadromous fishes. Fisheries, 33(8), 372-407.

Kendall, W. L., & White, G. C. (2009). A cautionary note on substituting spatial subunits

for repeated temporal sampling in studies of site occupancy. Journal of Applied

Ecology, 46(6), 1182-1188.

Kotliar, N. B., & Wiens, J. A. (1990). Multiple scales of patchiness and patch structure: a

hierarchical framework for the study of heterogeneity. Oikos, 253-260.

Labbe, T. R., & Fausch, K. D. (2000). Dynamics of intermittent stream habitat regulate

140

persistence of a threatened fish at multiple scales. Ecological Applications, 10(6),

1774-1791.

Lamphere, B. A., & Blum, M. J. (2012). Genetic estimates of population structure and

dispersal in a benthic stream fish. Ecology of Freshwater Fish, 21(1), 75-86.

Levins, R. (1969). Some demographic and genetic consequences of environmental

heterogeneity for biological control. American Entomologist, 15(3), 237-240.

MacKenzie, D. I., & Bailey, L. L. (2004). Assessing the fit of site-occupancy

models. Journal of Agricultural, Biological, and Environmental Statistics, 9(3),

300-318.

MacKenzie, D. I., Nichols, J. D., Hines, J. E., Knutson, M. G., & Franklin, A. B. (2003).

Estimating site occupancy, colonization, and local extinction when a species is

detected imperfectly. Ecology, 84(8), 2200-2207.

MacKenzie, D. I., Nichols, J. D., Lachman, G. B., Droege, S., Andrew Royle, J., &

Langtimm, C. A. (2002). Estimating site occupancy rates when detection

probabilities are less than one. Ecology, 83(8), 2248-2255.

McRae, B. H. (2006). Isolation by resistance. Evolution, 60(8), 1551-1561.

Meffe, G. K., & Vrijenhoek, R. C. (1988). Conservation genetics in the management of

desert fishes. Conservation Biology, 2(2), 157-169.

Mississippi Department of Water Quality (MDWQ). 2015. State Well Depth Data.

Unpublished raw data.

Mollenhauer, R., Zhou, Y., & Brewer, S. K. (2019). Multiscale habitat factors explain

variability in stream fish occurrence in the Ozark Highlands ecoregion,

USA. Copeia, 107(2), 219- 231.

141

Moran, R. L., Catchen, J. M., & Fuller, R. C. (2020). Genomic resources for darters

(Percidae: Etheostominae) provide insight into postzygotic barriers implicated in

speciation. Molecular Biology and Evolution, 37(3), 711-729.

Mundahl, N. D., & Ingersoll, C. G. (1983). Early autumn movements and densities of

Johnny (Etheostoma nigrum) and fantail (E. flabellare) darters in a southwestern

Ohio stream.

Musil, J., Horký, P., Slavík, O., Zbořil, A., & Horká, P. (2012). The response of the

young of the year fish to river obstacles: Functional and numerical linkages

between dams, weirs, fish habitat guilds and biotic integrity across large spatial

scale. Ecological Indicators, 23, 634-640.

Narum, S. R., Buerkle, C. A., Davey, J. W., Miller, M. R., & Hohenlohe, P. A. (2013).

Genotyping-by-sequencing in ecological and conservation genomics. Molecular

Ecology, 22(11), 2841.

Nathan, L. R., Welsh, A. B., & Vokoun, J. C. (2019). Watershed‐level brook trout genetic

structuring: Evaluation and application of riverscape genetics models. Freshwater

Biology, 64(3), 405-420.

Nicol, E., Stevens, J. R., & Jobling, S. (2017). Riverine fish diversity varies according to

geographical isolation and land use modification. Ecology and Evolution, 7(19),

7872-7883.

Nislow, K. H., Hudy, M., Letcher, B. H., & Smith, E. P. (2011). Variation in local

abundance and species richness of stream fishes in relation to dispersal barriers:

implications for management and conservation. Freshwater Biology, 56(10),

2135-2144.

142

Norman, J. R., Hagler, M. M., Freeman, M. C., & Freeman, B. J. (2009). Application of a

multistate model to estimate culvert effects on movement of small

fishes. Transactions of the American Fisheries Society, 138(4), 826-838.

Ostrand, K. G., & Wilde, G. R. (2004). Changes in prairie stream fish assemblages

restricted to isolated streambed pools. Transactions of the American Fisheries

Society, 133(6), 1329- 1338.

Paller, M. H. (1994). Relationships between fish assemblage structure and stream order in

South Carolina coastal plain streams. Transactions of the American Fisheries

Society, 123(2), 150-161.

Perkin, J. S., & Gido, K. B. (2011). Stream fragmentation thresholds for a reproductive

guild of Great Plains fishes. Fisheries, 36(8), 371-383.

Perkin, J. S., & Gido, K. B. (2012). Fragmentation alters stream fish community structure

in dendritic ecological networks. Ecological Applications, 22(8), 2176-2187.

Perkin, J. S., Gido, K. B., Al-Ta’ani, O., & Scoglio, C. (2013). Simulating fish dispersal

in stream networks fragmented by multiple road crossings. Ecological

Modelling, 257, 44-56.

Petty, J. T., & Grossman, G. D. (2004). Restricted movement by mottled sculpin (Pisces:

Cottidae) in a southern Appalachian stream. Freshwater Biology, 49(5), 631-645.

Poole, G. C. (2002). Fluvial landscape ecology: addressing uniqueness within the river

discontinuum. Freshwater Biology, 47(4), 641-660.

Poplar‐Jeffers, I. O., Petty, J. T., Anderson, J. T., Kite, S. J., Strager, M. P., & Fortney, R.

H. (2009). Culvert replacement and stream habitat restoration: implications from

brook trout management in an Appalachian watershed, USA. Restoration

143

Ecology, 17(3), 404-413.

Porter, B. A., Cavender, T. M., & Fuerst, P. A. (2002). Molecular phylogeny of the

snubnose darters, subgenus Ulocentra (genus Etheostoma, family

Percidae). Molecular Phylogenetics and Evolution, 22(3), 364-374.

Powers, S. L., & Warren Jr, M. L. (2009). Phylogeography of three snubnose darters

(Percidae: Subgenus Ulocentra) endemic to the southeastern US Coastal

Plain. Copeia, 2009(3), 523-528.

Powers, S. L., & Mayden, R. L. (2007). Systematics. Evolution and Biogeography of the

Etheostoma simoterum Species Complex (Percidae: Subgenus

Ulocentra). Bulletin of the Alabama Museum of Natural History, (25).

Pringle, C. M., Naiman, R. J., Bretschko, G., Karr, J. R., Oswood, M. W., Webster, J. R.,

& Winterbourn, M. J. (1988). Patch dynamics in lotic systems: the stream as a

mosaic. Journal of the North American benthological society, 7(4), 503-524.

Pritchard, J. K., Stephens, M., & Donnelly, P. (2000). Inference of population structure

using multilocus genotype data. Genetics, 155(2), 945-959.

Prunier, J. G., Dubut, V., Loot, G., Tudesque, L., & Blanchet, S. (2018). The relative

contribution of river network structure and anthropogenic stressors to spatial

patterns of genetic diversity in two freshwater fishes: A multiple‐stressors

approach. Freshwater Biology, 63(1), 6-21.

Ramasamy, R. K., Ramasamy, S., Bindroo, B. B., & Naik, V. G. (2014). STRUCTURE

PLOT: a program for drawing elegant STRUCTURE bar plots in user friendly

interface. SpringerPlus, 3(1), 1-3.

Richardson, J. S. (2019). Biological diversity in headwater streams. Water, 11(2), 366.

144

Rosenberry, D. O., LaBaugh, J. W., & Hunt, R. J. (2008). Use of monitoring wells,

portable piezometers, and seepage meters to quantify flow between surface water

and groundwater. Field Techniques for Estimating Water Fluxes between Surface

Water and Groundwater, 4, 39-70.

Rousset, F. (1997). Genetic differentiation and estimation of gene flow from F-statistics

under isolation by distance. Genetics, 145(4), 1219-1228.

Ruzzante, D. E., McCracken, G. R., Salisbury, S. J., Brewis, H. T., Keefe, D., Gaggiotti,

O. E., & Perry, R. (2019). Landscape, colonization, and life history: Their effects on

genetic diversity in four sympatric species inhabiting a dendritic

system. Canadian Journal of Fisheries and Aquatic Sciences, 76(12), 2288-2302.

Schmidt, B. V., & Schaefer, J. (2018). Ecological and landscape effects on genetic

distance in an assemblage of headwater fishes. Ecology of Freshwater

Fish, 27(2), 617-631.

Skalski, G. T., Landis, J. B., Grose, M. J., & Hudman, S. P. (2008). Genetic structure of

Creek Chub, a headwater minnow, in an impounded river system. Transactions

of the American Fisheries Society, 137(4), 962-975.

Smiley Jr, P. C., Dibble, E. D., & Schoenholtz, S. H. (2006). Spatial and temporal

variation of Goldstripe Darter abundance in first-order streams in north-central

Mississippi. The American Midland Naturalist, 156(1), 23-36.

Sterling, K. A., Nielsen, S. V., Brown, A. J., Warren Jr, M. L., & Noonan, B. P. (2020).

Cryptic diversity among Yazoo Darters (Percidae: Etheostoma raneyi) in disjunct

watersheds of northern Mississippi. PeerJ, 8, e9014.

Sterling, K. A., Reed, D. H., Noonan, B. P., & Warren, M. L. (2012). Genetic effects of

145

habitat ragmentation and population isolation on Etheostoma raneyi

(Percidae). Conservation Genetics, 13(3), 859-872.

Sterling, K. A., & Warren Jr, M. L. (2020). Description of a new species of cryptic

snubnose darter (Percidae: Etheostomatinae) endemic to north-central

Mississippi. PeerJ, 8, e9807.

Sterling, K. A., Warren Jr, M. L., & Henderson, L. G. (2013). Conservation assessment

of the Yazoo Darter (Etheostoma raneyi). Southeastern Naturalist, 12(4), 816-

842.

Sterling, K., Warren Jr, M. L., Noonan, B. P., Reed, D. H., & Henderson, L. G. (2011).

Yazoo Darter, Etheostoma raneyi: population and demographic status,

distributional changes, and habitat use of an endemic, nongame species. Final

report submitted to the Mississippi Museum of Natural Science, Jackson MS.

Sutherland, C. S., Elston, D. A., & Lambin, X. (2014). A demographic, spatially explicit

patch occupancy model of metapopulation dynamics and

persistence. Ecology, 95(11), 3149-3160.

Suttkus, R. D., Bailey, R. M., & Bart, H. L. (1994). Three new species of Etheostoma,

subgenus Ulocentra, from the Gulf Coastal Plain of southeastern United States.

Tulane University.

Taylor, P. D., Fahrig, L., Henein, K., & Merriam, G. (1993). Connectivity is a vital

element of landscape structure. Oikos, 571-573.

Thompson, K. W., & Muncy, R. J. (1986). Darters of the Little Tallahatchie watershed in

northern Mississippi. Journal Mississippi Academy of Sciences, 31, 63-77.

Tischendorf, L., & Fahrig, L. (2000). On the usage and measurement of landscape

146

connectivity. Oikos, 90(1), 7-19.

Turner, T. F., & Robison, H. W. (2006). Genetic diversity of the Caddo Madtom, Noturus

taylori, with comments on factors that promote genetic divergence in fishes

endemic to the Ouachita highlands. The Southwestern Naturalist, 51(3), 338-345.

United States Environmental Protection Agency, United States Geological Survey.

(2006). The national hydrography dataset plus. http://www.horizon-

systems.com/NHDPlus/index.php.Accessed 1 May 2015.

Warren Jr, M. L., & Pardew, M. G. (1998). Road crossings as barriers to small-stream

fish movement. Transactions of the American Fisheries Society, 127(4), 637-644.

Warren Jr, M. L., Burr, B. M., Walsh, S. J., Bart Jr, H. L., Cashner, R. C., Etnier, D. A.,

& Ross, S. T. (2000). Diversity, distribution, and conservation status of the

native freshwater fishes of the southern United States. Fisheries, 25(10), 7-31.

Whiteley, A. R., Spruell, P., & Allendorf, F. W. (2004). Ecological and life history

characteristics predict population genetic divergence of two salmonids in the same

landscape. Molecular Ecology, 13(12), 3675-3688.

Wiens, J. A. (2002). Riverine landscapes: taking landscape ecology into the

water. Freshwater Biology, 47(4), 501-515.

Willing, E. M., Dreyer, C., & Van Oosterhout, C. (2012). Estimates of genetic

differentiation measured by F ST do not necessarily require large sample sizes

when using many SNP markers. PloS One, 7(8), e42649.

147

APPENDIX A

Table A1

Summary of NP-MANOVA (n=1000) results on Principal Components Analysis scores (PC1 and PC2, 59% variation explained) of all local habitat variables included in occupancy and coexistence models. Variables Df Pseudo-F R² p-value Land Use 2 2.81 0.05 0.019 Time 1 1.53 0.01 0.218 Land Use * Time 2 1.76 0.03 0.158

148

Table A2

Detection probabilities and standard errors (SE) for each species dependent on sampling technique. Each subreach at a site was sampled with a backpack electrofisher (5 s/m) and seined (two seine hauls per subreach, sampling all available habitats). Species Gear Detection Probability (p) ± SE Yazoo Darter Seine 0.35 ± 0.08 Backpack Electro-fisher 0.51 ± 0.05 Goldstripe Darter Seine 0.30 ± 0.09 Backpack Electro-fisher 0.43 ± 0.06 Redspot Darter Seine 0.27 ± 0.06 Backpack Electro-fisher 0.36 ± 0.06

149 Table A3

Summary of transformations and univariate statistics for all predictor variables for all 97 sites used in the modeling of occupancy and coexistence. Signs (+/-) indicate whether an increase in the magnitude of a variable would have a positive or negative effect on the occupancy of the species (YD = Yazoo Darter, RD = Redspot Darter, GD = Goldstripe Darter); no sign indicates an increase in the magnitude of the variable was anticipated to have no effect on the occupancy of a species. Categories Variables RD GD YD Transformation Mean SD Min Max Land Cover Forested (%) + + + Logit 78.53 14.81 10.78 96.78 Urban (%) - - - Logit 1.32 2.6 0 18.01 Agricultural (%) - - - Logit 18.45 14.31 0 83.65 Open Water (%) - - - Logit 0.96 0.8 0 4.75 Wetland (%) + Logit 0.88 1.1 0 4.72

Geology Well Depth (ft) + Log 362.91 169.12 131.93 767.44

Stream Network Drainage Area (km²) - Log 24.5 28.9 1.97 98.81 Confluence Link + Log 35.5 15.04 8 73 Hydrology Stream Depth - - - Log 16.53 12.26 1.9 58.25 Stream Velocity + + + Log 0.26 0.31 0 1.85 Substrate Size + + Log 2.84 0.70 1.08 5.17 Stream Depth (CV) + + + 0.29 0.17 0.06 0.9 Stream Velocity (CV) + + + 0.58 0.36 0.04 2 Substrate Size (CV) + + 0.14 0.12 0 0.64

Cover Aquatic Vegetation (%) + + + Logit 0.08 0.16 0 0.77 Woody Structure (%) + + + Logit 0.14 0.13 0.07 0.78

APPENDIX B

Table B1 Beta coefficient estimates and 95% confidence intervals for proportional daily movement (PDMR) rates and Bray-Curtis Distances and number of movers for three sites (n= 54) versus four sites (n=63) and from a non-parametric boostrap (n=10,000). Repeated visits to reaches were treated as replicates (e.g., for three sites: 6 recapture events * 3 reaches * 3 sites = 54).

# of Replicate Reaches Response Beta Coefficients 95% CIs

PDMR 63 0.0107 0.007, 0.013 54 0.0093 0.006, 0.013 Bray-Curtis Distance 63 0.007 0.006, 0.008 54 0.007 0.006, 0.008 Number of Movers 63 2.69 1.83, 3.48 54 2.91 1.78, 3.96

151

Table B2

Candidate models used in AICc model selection for proportional daily movement rate (PDMR) and Bray-Curtis Distance. Scale Response Model K Reach Both Null 3 ~ Depth + 1lStream 4 ~ Substrate Size + 1lStream 4 ~ Wetted Width + 1lStream 4 ~ Bankfull Width + 1lStream 4 ~ Current Velocity + 1lStream 4 ~ Woody Structure + 1lStream 4 ~ Current Velocity * Substrate Size + 1lStream 5 ~ Current Velocity * Wetted Width + 1lStream 5 ~ Current Velocity * Depth + 1lStream 5 ~ Current Velocity * Woody Structure + 1lStream 5 ~ Current Velocity * Bankfull Width + 1lStream 5 ~ Wetted Width * Substrate Size + 1lStream 5 ~ Wetted Width * Depth + 1lStream 5 ~ Wetted Width * Woody Structure + 1lStream 5 ~ Wetted Width * Bankfull Width + 1lStream 5 ~ Depth * Substrate Size + 1lStream 5 ~ Depth * Woody Structure + 1lStream 5 ~ Depth * Bankfull Width + 1lStream 5 ~ Woody Structure * Substrate Size + 1lStream 5 ~ Woody Structure * Bankfull Width + 1lStream 5 ~ Bankfull Width * Substrate Size + 1lStream 5 ~ Global 9 Multi ~ Land cover + 1lStream 4 ~ Confluence Size + 1lStream 4 PDMR ~ Land cover * Confluence Size 1lStream 5 ~ Land cover + Current Velocity * Woody Structure + 6 1lStream ~ Confluence Size + Woody Structure * Current Velocity + 6 1lStream Bray- ~ Land cover + Depth * Woody Structure + 1lStream 6 Curtis ~ Confluence Size + Woody Structure * Depth + 1lStream 6 ~ Land cover + Bankfull Width * Depth + 1lStream 6 ~ Confluence Size + Bankfull Width * Depth + 1lStream 6

152

Table B3

Loadings for environmental variables included in the Principal Component Analysis. Variable PC1 PC2 PC3 PC4 PC5 PC6 Depth 1.32 -0.76 0.10 0.04 0.33 -0.87 Current Velocity 0.31 1.05 0.51 -1.25 0.45 -0.08 Substrate Size 0.68 0.30 -1.58 -0.22 0.30 0.19 Woody Structure 1.16 -0.95 0.40 -0.32 0.03 0.84 Width 0.70 0.96 0.38 1.08 0.66 0.24 Bankfull Width 1.12 0.80 0.04 0.08 -1.14 -0.08

153 Table B4

Means and standard deviations for six habitat variables within years and between years for all reaches at four headwater confluences Site Year Reach Reach Cat Depth Current Substrate % Woody Wetted Bankfull (cm) Velocity Size Structure Width (m) Width (m·s⁻¹) (m)

Garraway 2017 1 MS Up 35.02 ± 0.13 ± 1.88 ± 0.44 ± 4.33 ± 5.67 ± Creek 17.89 0.13 0.51 0.48 1.02 1.52 2 Tributary 33.47 ± 0.08 ± 2.11 ± 0.46 ± 4.26 ± 5.53 ± 17.48 0.08 0.62 0.50 1.53 1.70 3 MS Down 85.96 ± 0.05 ± 1.8 ± 0.50 0.62 ± 6.57 ± 8.53 ± 33.47 0.09 0.48 0.63 0.63

2018 1 MS Up 40.19 ± 0.08 ± 2.0 ± 0.52 ± 4.46 ± 5.29 ± 18.12 0.04 0.0 0.50 0.84 1.20 2 Tributary 36.15 ± 0.08 ± 2.20 ± 0.31 ± 4.75 ± 5.53 ± 18.52 0.08 0.63 0.47 1.17 1.56 3 MS Down 81.38 ± 0.04 ± 1.75 ± 0.30 ± 6.5 ± 7.12 ± 32.19 0.04 0.43 0.46 0.87 0.86

Priest Creek 2017 1 MS Down 18.78 ± 0.26 ± 2.56 ± 0.18 ± 3.92 ± 6.70 ± 10.15 0.15 0.52 0.38 1.31 1.01 2 MS Up 18.08 ± 0.23 ± 2.63 ± 0.27 ± 3.18 ± 7.33 ± 11.34 0.15 0.53 0.44 1.33 1.95 3 Tributary 15.85 ± 0.09 ± 2.9 ± 0.33 0.24 ± 2.26 ± 4.91 ± 12.13 0.09 0.41 0.82 1.61

2018 1 MS Down 18.12 ± 0.21 ± 2.61 ± 0.16 ± 4.28 ± 7.83 ± 10.07 0.13 0.49 0.37 1.34 1.31 Table B4 (continued)

Site Year Reach Reach Cat. Depth Current Substrate % Woody Wetted Bankfull (cm) Velocity Size Structure Width (m) Width (m·s⁻¹) (m)

2 MS Up 20.60 ± 0.20 ± 2.71 ± 0.38 ± 3.05 ± 8.31 ± 11.82 0.17 0.46 0.48 1.12 2.24 3 Tributary 14.8 ± 0.46 ± 2.81 ± 0.33 ± 2.46 ± 4.13 ± 10.99 2.89 0.40 0.47 1.04 0.94

Mixon’s 2017 1 MS Down 17.73 ± 0.10 ± 2.39 ± 0.05 ± 3.90 ± 7.43 ± Creek 14.80 0.12 0.66 0.15 1.75 3.74 2 Tributary 14.53 ± 0.03 ± 2.03 ± 0.0 ± 3.59 ± 6.43 ± 9.32 0.03 0.18 0.0 1.05 1.42 3 MS Up 14.36 ± 0.14 ± 2.43 ± 0.02 ± 2.87 ± 5.85 ± 8.46 0.21 0.75 0.10 1.56 2.94

2018 1 MS Down 16.29 ± 0.09 ± 2.63 ± 0.05 ± 3.27 ± 8.33 ± 13.18 0.13 0.71 0.22 1.50 3.70 2 Tributary 12.51 ± 0.05 ± 2.46 ± 0 ± 0 4.33 ± 6.36 ± 9.20 0.06 0.61 1.05 1.49 3 MS Up 14.36 ± 0.06 ± 2.46 ± 0.04 ± 2.59 ± 7.01 ± 8.46 0.10 0.78 0.21 1.26 2.70

Sweetwater 2018 1 MS Up 34.83 ± 0.20 ± 2 ± 0 0.68 ± 2.55 ± 3.17 ± Creek 20.21 0.12 0.37 0.87 1.0 2 Tributary 17.92 ± 0.15 ± 1.99 ± 0.5 ± 1.83 ± 2.47 ± 8.83 0.09 0.09 0.45 0.57 0.61 3 MS Down 42.88 ± 0.13 ± 1.97 ± 0.53 ± 2.96 ± 3.61 ± 22.16 0.08 0.18 0.48 0.52 0.89

Table B5

Estimates of percent slope and sinuosity at reaches for all mark-recapture sites. Stream Reach Position Percent Slope Sinuosity Mixon’s Creek 1 Mainstem downstream 3% 1.02 2 Tributary 1% 1.05 3 Mainstem Upstream 0.003% 1.07 Garraway Creek 1 Mainstem Upstream 2% 1.02 2 Tributary 1% 1.004 3 Mainstem Downstream 3% 1.02 Priest Creek 1 Mainstem Downstream 0.006 1.09 2 Mainstem Upstream 0.006 1.20 3 Tributary 3% 1.0 Sweetwater Creek 1 Mainstem Upstream 2% 1.005 2 Tributary 1% 1.04 3 Mainstem Downstream 0.006 1.03

156

Priest Creek

Garraway Creek

Sweetwater Creek

Mixon’s Creek

Figure B1 Map of our four mark-recapture sites. Sites positioned in the red square indicate the geographic positions of our two urban sites, Priest Creek and Mixon’s Creek. The red square approximates the range of the city limits of Hattiesburg, MS. Sites positioned in the green square indicate the locality of our two forested site positioned in De Soto National Forest, Sweetwater Creek and Garraway Creek157

APPENDIX C

Table C1

Modal (± 95% credible intervals) historic migration rate (migrants per generation, Nm estimates from MIGRATE between the Goldstripe Darter YRSB and Lower Little Tallahatchie groupings.

Migration into Lower Little Otoucalofa Creek (3) Yocona River (5) Tallahatchie (1)

From 3 From 5 From 1 From 5 From 1 From 3 Nm 0.71 1.19 0.20 1.60 0.12 0.45 LCI 0.67 1.14 0.13 1.53 0.04 0.37 UCI 0.74 1.24 0.27 1.68 0.20 0.53

158

Table C2

Modal (± 95% credible intervals) historic migration rate (migrants per generation, Nm estimates from MIGRATE between the Goldstripe Darter LTRSB groupings. Migration into Tippah River (2) Little Tallahatchie River (4)

From 4 From 2 Nm 0.75 1.70 LCI 0.63 1.54 UCI 0.88 1.86

159

Table C3

Modal (± 95% credible intervals) historic migration rate (migrants per generation, Nm) estimates from MIGRATE between the Yazoo Darter LTRSB groupings. Grouping IDs are as follows: Little Tallahatchie (1), Tippah (2), Hotopha (3), and Cypress (4).

Migration Into 2 to 1 3 to 1 4 to 1 Nm 0.873 0.287 0.405 LCI 0.715 0.176 0.285

UCI 1.049 0.408 0.534

1 to 2 3 to 2 4 to 2 Nm 1.457 0.344 0.442

LCI 1.237 0.245 0.335 UCI 1.686 0.450 0.562

1 to 3 2 to 3 4 to 3

Nm 0.446 0.337 0.278 LCI 0.273 0.202 0.165 UCI 0.626 0.479 0.4

1 to 4 2 to 4 3 to 4 Nm 0.7164 0.314 0.175 LCI 0.521 0.189 0.099

UCI 0.668 0.449 0.261

160

Table C4

Modal (± 95% credible intervals) historic migration rate (migrants per generation, Nm estimates from MIGRATE between the Yazoo Darter YRSB groupings. Migration into Yocona (5) Otoucalofa (6)

From 6 From 5 Mode 0.302 0.162 LCI 0.102 0.098 UCI 0.509 0.237

161

Tippah Cypress Little Tallahatchie Hotopha

Yocona Otoucalofa

Figure C1 Bar plots depicting results from the SNP-based STRUCTURE analyses for the

Yazoo Darter for the YRSB (A, K=2) and LTRSB hierarchical groupings (B, K=4). Bars correspond to bi-allelic loci of individuals, and colors represent the probability of ancestry to each STRUCTURE inferred cluster (K).

Little Tallahatchie Tippah Cypress Otoucalofa Yocona Hotopha

Figure C2 Map of Yazoo Darter hierarchical grouping based on a K of 6.

Little Tallahatchie Tippah

Below Sardis Reservoir YRSB

Otoucalofa Yocona

Figure C3 Bar plots depicting results from the SNP-based STRUCTURE analyses for the Goldstripe Darter for the LTRSB (3A, K=2) and YRSB (3B, 3C, K=3) hierarchical groupings. Bars correspond to bi-allelic loci of individuals, and colors represent the probability of ancestry to each STRUCTURE inferred cluster (K).

164

Little Tallahatchie Tippah Yocona Otoucalofa Below Reservoir

Figure C4 Map of Goldstripe Darter hierarchical grouping based on a K of 5.