Habitat and Imperilment of the Candy Darter osburni in the New River Drainage, USA

By Corey Garland Dunn

Thesis submitted to the faculty of the Polytechnic Institute and State University in partial fulfillment of the requirements for the degree of

MASTER OF SCIENCE

IN

FISH AND WILDIFE CONSERVATION

Paul L. Angermeier, Chair

C. Andrew Dolloff

Emmanuel A. Frimpong

17 November 2017 Blacksburg, Virginia, United States

Keywords: Endemic, Habitat Suitability, Range Dynamics, Stream

© Corey G. Dunn, 2017 Habitat and Imperilment of the Candy Darter in the New River Drainage, USA

Corey Garland Dunn

ACADEMIC ABSTRACT

The streams of the southeastern United States are both hotspots for biodiversity and centers of imperilment. The specific spatiotemporal scales at which stressors impact biota are often unknown, partly due to inadequate knowledge about many species’ life-histories. I conducted two complementary studies to investigate the habitat associations of an imperiled highland stream fish, the Candy Darter Etheostoma osburni. In Chapter 2, I asked (1) does micro-habitat suitability correlate with the “robustness” (i.e., viability) of four distinct populations? In Chapter 3, I expanded the extent of investigation, and asked (2) which environmental factors, expressed at what spatial scales, best explain in-stream conditions, and (3) do stream segments where Candy Darters persist have cooler temperatures and less fine-sediment than segments where the species is extirpated or historically went undetected? Chapter 2 revealed Candy Darters demonstrate ontogenetic habitat shifts, with age-0 individuals selecting slower water velocities than adults. Despite, clear habitat selection for multiple habitat variables, suitability attributed to fine-sediment avoidance most strongly correlated with population robustness across streams. Chapter 3 indicated Candy Darters are extirpated from most areas in Virginia and southern . Land use and natural catchment features, including geology, elevation, and stream geomorphology, predominantly explained instream conditions. Populations persist in segments with cool stream temperatures and low embeddedness year- round. To recover Candy Darters, managers will need to remedy pervasive land-use threats and restore stream habitat, while operating within the impending context of warming air and water temperatures and the existential threat of the introduced Variegate Darter E. variatum. Habitat and Imperilment of the Candy Darter Etheostoma osburni in the New River Drainage, USA

Corey Garland Dunn

PUBLIC ABSTRACT

The Candy Darter is a small colorful stream fish only found in the New River Valley of Virginia and West Virginia. It was historically recorded throughout much of its range, but the species has since seemingly disappeared from many historical locations. Biologists, who are tasked with conserving declining species, know very little about the Candy Darter, which makes it difficult to determine the reasons for its decline. My goal was to clarify the habitats and streams used by the Candy Darter. In Chapter 2, my team recorded the habitats that Candy Darters preferred in four different streams where the species is either abundant (two streams), rare (1 stream), or has disappeared (1 stream). I determined individuals consistently avoid areas with high levels of fine sediment. I also discovered the streams where the species still exists had lower levels of fine sediment, indicating that high levels of fine-sediment may diminish habitat quality for individuals and eventually impact populations. In Chapter 3, I asked whether the conclusions from Chapter 2 were valid for most streams where Candy Darters have ever been recorded. In addition to less fine-sediment, I suspected the streams where Candy Darters still exist, would have cooler stream temperatures than the streams where they have disappeared. I surveyed 42 locations for Candy Darters and recorded stream temperatures and fine-sediment levels at each location. I confirmed that, on average, the streams where Candy Darters still exist have much cooler stream temperatures and fewer fine-sediments. Stream temperatures and fine- sediment levels could be explained by surrounding environmental conditions including geology, altitude, stream size, and the amount of pasture beside and upstream of each location. These findings are consistent with many other studies that have found non-natural land covers, including pasture, lead to higher amounts of fine-sediment washing into streams and create warmer stream temperatures. The decline of the Candy Darter is similar to the declines of dozens of other fish species throughout the southeastern United States. To restore Candy Darters, biologists will need to work with landowners to improve conditions adjacent to streams, while combatting other threats, such as warming air temperatures and non-native species. DEDICATION

To my 10th-grade biology teacher, Mrs. Dunn

iv ACKNOWLEDGEMENTS

This research progressed from an initial solo snorkeling trip to Big Stony Creek after feeling inspired in my undergraduate Fisheries Techniques course to defending this thesis, with several missteps, rabbit holes, and campfires in between. As a PhD student, I now look back with a somewhat seasoned perspective at the special opportunity that I was given as a master’s student – a chance to identify and craft a thesis topic. Naturally, there were several people who helped me along the way. First, I thank my major adviser, Dr. Paul Angermeier, who when helping brainstorm potential research projects, encouraged me to “find a good story” and “go where the fish are.” Time will tell if I succeed in the former, but I definitely accomplished the latter. Of all the lessons learned in graduate school, the most import lesson is not to be afraid to think deeply about complex issues. This is Paul’s expertise, and I’m grateful to have learned from the best. Next, I thank my committee members, Drs. Andy Dolloff and Emmanuel Frimpong – both of whom provided guidance as I developed my project, made time for me despite busy schedules, and were patient with me during this exceptionally long and circuitous process. As a new graduate student, who was also new to Natural Resources, I leaned heavily on my lab-mates. Jane Argentina and Amy Villamagna were my first mentors in this field. Jamie Roberts helped me work through conceptual models, whereas Greg Anderson helped me work through the statistical variety. I was happy and fortunate that my path and the international paths of Tiz Mogollón Gómez and Ryan Liang all converged in Blacksburg, Virginia. This research took me all over the New River Valley, and I thank the scientists who lent their time to provide direction throughout my travels. My primary contacts were Bryn Tracy (NCDEQ) in , and Dan Cincotta (WVDNR) and Stuart Welsh (WV University) in West Virginia. I am especially grateful to Mike Pinder (VDGIF), who encouraged this research from its infancy as a hypothetical undergraduate research project. I thank all the members of my research team – many of whom have since completed graduate degrees and continue to have a presence in our field. Team members included Matt Bierlein, Joe Cline, David Crain, Daniel Dodge, Laura Heironimus, Pat Kroboth, Josh Light, Luke Longanecker, Phil Pegalow, Jordan Richard, and Chris Rowe. Finally, I thank my current major adviser, Dr. Craig Paukert (Univ. ), who graciously allowed me to dedicate time towards wrapping up my master’s research back east.

v ATTRIBUTION

I have or intend to publish two chapters (2 and 3) of this thesis as stand-alone manuscripts. Consequently, there will be some redundancy between these two chapters. Both manuscripts were joint efforts between Dr. Paul Angermeier and myself. Dr. Angermeier secured funding and significantly contributed to these chapters’ conceptual underpinnings, study designs, and refinement through diligent editing of draft manuscripts. Accordingly, the narratives of Chapters 2 and 3 are first-person plural. Both chapters are also formatted according to journal specifications. Chapter 2, Dunn and Angermeier (2016), was published in the Transactions of the American Fisheries Society, and Chapter 3, Dunn and Angermeier (In review), has been submitted to Freshwater Biology. The narrative of the remainder of this thesis uses first-person singular, and complies with the editorial conventions of the American Fisheries Society.

References Dunn, C.G. and Angermeier, P.L. 2016. Development of habitat suitability indices for the Candy Darter, with cross-scale validation across representative populations. Transactions of the American Fisheries Society, 145:1266–1281.

Dunn, C.G. and Angermeier, P.L. In review. Pathway to imperilment: extirpation of a highland fish explained by fine-sediment, stream temperature, and landscape context.

vi TABLE OF CONTENTS

ACADEMIC ABSTRACT ...... ii PUBLIC ABSTRACT ...... iii DEDICATION...... iv ACKNOWLEDGEMENTS...... v ATTRIBUTION...... vi LIST OF TABLES...... viii LIST OF FIGURES ...... ix LIST OF SUPLEMENTARY MATERIAL ...... xi CHAPTER 1: General Introduction...... 1 References...... 11 CHAPTER 2: Development of Habitat Suitability Indices for Candy Darter Etheostoma osburni, with Cross-Scale Validation across Representative Populations...... 23 Abstract...... 23 Introduction...... 24 Methods...... 26 Data Analysis...... 31 Results...... 34 Discussion...... 38 Conclusions...... 42 Acknowledgments...... 43 References...... 44 CHAPTER 3: Pathway to Imperilment: Extirpation of a Highland Fish Explained by Fine- Sediment, Stream Temperature, and Landscape Context ...... 61 Abstract...... 61 Introduction...... 62 Methods...... 64 Data Analyses ...... 69 Results...... 72 Discussion...... 75 Conclusion ...... 79 Acknowledgements...... 80 References...... 80 CHAPTER 4: General Conclusions and Management Recommendations ...... 97 References...... 102 SUPPLEMENTARY MATERIAL...... 108

vii LIST OF TABLES

TABLE 2.1. Means, ± 95% confidence intervals (in parentheses), and counts of observations of habitat availability in four streams and two seasons. Stream abbreviations: EFG = East Fork , LC = Laurel Creek, SC = Sinking Creek, SFC = South Fork Cherry River. Substrate categories are 1 = Silt, 2 = Sand, 3 = Gravel, 4 = Pebble, 5 = Small Cobble, 6 = Large Cobble, 7 = Small Boulder, 8 = Large Boulder, 9 = Bedrock. Embeddedness and silt categories DUH ” –25%, 2 = 26–50%, 3 = 51–75%, 4 = > 75%...... 53

TABLE 2.2. Spearman rank-order correlations (ߩ) between habitat variables and a rank representing the population status of Candy Darters in four study streams and two seasons (columns 1 and 3; Figure 2.2). Streams with the highest densities were given the highest rank: East Fork Greenbrier River = 3 (Robust), South Fork Cherry River = 3 (Robust), Laurel Creek = 2 (localized), Sinking Creek = 1 (extirpated). Predictions are based on historical accounts of habitat use (Supplementary Table S2.1). The embeddedness predicted relation was not DSSOLFDEOH 1$ GXHWRQRKLVWRULFDODFFRXQWV&RHIILFLHQWV•__DQGFRQVLVWHQWZLWK preGLFWLRQVDUHEROGHGWRHPSKDVL]HUHODWLRQVKLSVWUHQJWK&RHIILFLHQWV•__DQGLQFRQVLVWHQW with predictions are italicized. “Combined” are averages from spring and fall...... 54

TABLE 2.3. Pearson correlation coefficients (r) between predicted individual habitat suitability and averages of five in-stream habitat variables across four streams that vary in population status (columns 1 and 2; Figure 2.2). Predictions are based on prior accounts of habitat use (Supplementary Table S2.1). The embeddedness predicted relation was not applicable (NA) due to no historical accounts. “Multi-stage” is the correlation between average suitability across multiple life stages (Supplementary Table S2.4) and habitat gradients (Table 2.1). “Combined” DUHDYHUDJHVIURPVSULQJDQGIDOOFRHIILFLHQWV&RHIILFLHQWV•__DQGFRQVLVWHQWZLWK predictions are bolded to emphasize strength of relationship. CoeIILFLHQWV•__DQG inconsistent with predictions are italicized...... 55

TABLE 3.1. Means ± (standard errors) of habitat variables in 42 segments as inputs in multi- variate analyses. Values under Principal Component (PC) 1 and 2 are permutation-based correlation coefficients between environmental variables and the first two PC axes. Bolded values highlight the axis with the higher correlation coefficient...... 90

TABLE 3.2. Akaike’s Information Criterion (AICc) for best-supported combinations of hypotheses explaining the distribution of E. osburniǻ$,&FLVWKHGLIIHUHQFHEHWZHHQWKHWRS- ranked model and lower-ranked models (i). Model weight (Wi) is the probability of a model being the best-supported model. Evidence ratio (W1/Wi) is the number of times the top-ranked model is better supported over lower-ranked models. Area under the-curve (AUC) is a threshold- independent measure of cross-validation Supplementary Table S3.8 contains all 141 models. .. 91

viii LIST OF FIGURES

FIGURE 2.1. Map of the New River drainage and study sites. (A) = South Fork Cherry River, WV (robust population); (B) = East Fork Greenbrier River, WV (robust population); (C) = Laurel Creek, VA (localized population); (D) = Sinking Creek, VA (extirpated population). Insets depict survey designs used to develop habitat suitability indices within streams supporting robust populations (South Fork Cherry River) and to systematically measure habitat availability in streams with localized (Laurel Creek) or extirpated...... 56

FIGURE 2.2. Framework for examining relationships among stream-habitat gradients, predicted individual habitat suitability, and observed population robustness across streams. The relationship between columns 1 and 3 is the observed population relationship to a habitat gradient. The relationship between columns 1 and 2 is the predicted response of individuals to a habitat gradient across streams. The relationship between columns 2 and 3 is the cross-scale relationship between predicted individual suitability and observed population robustness...... 57

FIGURE 2.3. Habitat selection curves developed from habitat used by Candy Darters during three life stages and two seasons and available habitat in two streams. Continuous curves, presented as visual aids, were obtained by regressing suitability values against the midpoint of each bin using generalized additive regression models. Substrate abbreviations: Grav. = Gravel, Peb. = Pebble, Sm. Cob. = Small Cobble, Lg. Cob. = Large Cobble, Sm. Bldr. = Small Boulder, Lg. Bldr. = Large Boulder, BR = Bed Rock...... 58

FIGURE 2.4. Non-metric multidimensional scaling (NMDS) plots of habitat use, availability, and suitability in spring. A) Habitat use by three life stages and availability in four streams (polygons). Predicted suitability by (B) Adults, (C) Juveniles, and (D) Age-0. Symbols for "LC use" are locations used by Candy Darters in Laurel Creek in spring. Variables that are highly FRUUHODWHG 3HDUVRQFRHIILFLHQW>U@• ZLWKD[HVDUHVKRZQ$OOFRUUHODWLRQFRHIILFLHQWVDUHLQ Supplementary Table S2.2. Stream abbreviations: EFG = East Fork Greenbrier River, LC = Laurel Creek, SC = Sinking Creek, SFC = South Fork Cherry River. NMDS stress = 0.17...... 59

FIGURE 2.5. Spearman rank-order correlation coefficients (ߩ) between predicted individual suitability and population health in two seasons and across seasons (columns 1 and 3; Figure 2). Positive coefficients (blue) indicate individual preferences are predictive of population-level health, while negative coefficients (red) are disconnects between individual and population levels. The lower-far right cell (ߩ = 0.95) is the study-wide CSL coefficient. “Multi-stage” is a multiple life stage habitat suitability index. “MV.index” indicates a multiple-variable habitat suitability index. Vel. = Velocity, Sub. = Substrate Size, Emb. = Embeddedness...... 60

FIGURE 3.1. A multi-level framework depicting indirect pathways of regional disturbance (black arrows) on a local ecological state (persistence versus extirpation of a sensitive stream fish in a stream segment). Environmental features (rectangles) at catchment and segment scales interact with regional disturbances (stimuli) to either decrease (left) or increase (right) the resilience of a sensitive stream fish population. (a) Sensitive catchment- and segment-scale environmental features propagate the influence of a moderate land-use disturbance leading to extirpation (i.e., land-cover cascade pathway). (b) Land-use disturbance has little influence on a

ix sensitive species due to mitigating influences of highly resilient catchment and nested segment features (resilient-catchment pathway). (c) Land-use disturbance is first propagated by sensitive catchment features and then mitigated by resilient segment features leading to restricted persistence within degraded catchments (resilient-segment pathway). (d) High meta-population connectivity at a catchment scale enables persistence in sites with both sensitive catchment and segment features. (e) Absence of meta-population connectivity prevents recolonization of a segment with suitable catchment and segment features...... 92

FIGURE 3.2. Locations of stream segments (study sites) sampled for E. osburni in 2012 within the New River drainage, Virginia and West Virginia...... 93

FIGURE 3.3. Principal component analysis (PCA) of instream habitat within study segments (plotted points). Circle size corresponds to observed E. osburni densities (fish /100 m2) within segments. Axes 1 (horizontal) and 2 (vertical) explained 43.2% and 24.8% of variation, respectively. Arrays represent permutation-based correlation coefficients between axes and habitat variables. (a) Instream habitat; (b) Natural catchment features, (c) Catchment land use; (d) Segment features. SPMDT = spring mean daily stream temperature, SMDMX = summer mean daily maximum stream temperature...... 94

FIGURE 3.4. Venn diagram showing partitioned adjusted variation (by percentage) in instream habitat by catchment land use, natural catchment-, and segment-scale features via partial redundancy analysis. “Direct catchment” are direct pathways of catchment-generated influence on instream habitat. “Indirect propagating” are pathways where catchment-generated phenomena constrain segment features and instream habitat. “Unique segment” is variation explained at the segment-scale not explained at the catchment scale and represent resilient-segment or sensitive- segment pathways. Percentages sum to explain 52% (= adjusted R2) of instream-habitat variation...... 95

FIGURE 3.5. Predicted probabilities, with 90% confidence intervals, of presence/persistence across temperature (a) and embeddedness (b) gradients. Embeddedness index (0–4) was converted to percentages (0–100%). Presences and absences were plotted as 1 and 0, respectively. Black circles are segments where E. osburni were historically confirmed present. White circles are segments where E. osburni was not historically documented by sparse early surveys in the region. Panels c–d demonstrate influence of spatial location on probabilities across temperature (c) and embeddedness (d) gradients at the 75th (“Favorable” spatial context), 50th (“neutral” spatial context), and 25th (“poor” spatial context) quartiles of spatial covariates 1 and 2...... 96

x LIST OF SUPLEMENTARY MATERIAL

Supplementary Table S2.1. Previous accounts of in-stream habitat associations of Candy Darters. “Predicted relation” refers to expected correlations between darter density and a habitat gradient interpreted from references listed below (e.g., adult densities increase as water velocity increases but decrease as water depth increases [Kuehne and Barbour 1983]). Predicted relations were not applicable (NA) if information was not available for variables at specific life stages.108

Supplementary Table S2.2. Pearson correlation coefficients (r) between in-stream microhabitat variables and non-metric multidimensional scaling (NMDS) axes for two seasons (Figure 2.4; 6XSSOHPHQWDU\)LJXUH6 &RHIILFLHQWV•__DUHSUHVHQWHGLQ)LJXUH6XSSOHPHQWDU\ Figure S2.2...... 109

Supplementary Table S2.3. Observations by life stage (N), area sampled, and density of Candy Darters in three streams and two seasons. Stream abbreviations: EFG = East Fork Greenbrier River, LC = Laurel Creek, SFC = South Fork Cherry River...... 110

Supplementary Table S2.4. Predicted individual suitability (X) and 95% confidence intervals (CI95) by life stage and season in four streams that vary in population status of Candy Darter. Possible suitability values range from 0 to 1, which indicate “no selection” and “maximum selection”, respectively. “Multi-Stage” is the average suitability across life stages. “Multi- variable” is the average suitability calculated from all habitat variables within each stream. East Fork Greenbrier (EFG) and South Fork Cherry rivers (SFC) support robust populations, Laurel Creek (LC) supports a localized population, and Candy Darters are extirpated from Sinking Creek (SC)...... 111

Supplementary Figure S2.5. Length-frequency histogram of total lengths from Candy Darters (N = 798 individuals) measured while snorkeling in the East Fork Greenbrier and South Fork Cherry rivers, WV. We used different thresholds for separating juveniles from adult females (60 mm) and males (65 mm) based on differences in pigmentation. The line between adults and juveniles is drawn at 62.5 mm...... 113

Supplementary Figure S2.6. Non-metric multidimensional scaling (NMDS) plots of habitat use, availability, and suitability in fall. A) Habitat use by three life stages and availability in four streams (polygons). Predicted microhabitat suitability by (B) Adults, (C) Juveniles, and (D) Age- 0. Symbols for "LC use" are locations used by Candy Darters in Laurel Creek in fall. Variables highly correlated (Pearson coefficient [r@• ZLWKD[HVDUHVKRZQ$OOFRUUHODWLRQFRHIILFLHQWV are in Supplementary Table S2.2. Stream abbreviations: EFG = East Fork Greenbrier River, LC = Laurel Creek, SC = Sinking Creek, SFC = South Fork Cherry River. NMDS stress = 0.14. . 114

Supplementary References S2.7...... 115

Supplementary Table S3.1. References used to identify streams and segments (“Sites”). “Recent” streams have recent collections of E. osburni. “Historical” streams have historical records but none within 40 years before this study. “No record” streams lack records...... 116

xi Supplementary Material S3.2. Detailed methods used to predict missing stream temperatures...... 117

Supplementary Figure S3.3. Relationship between predicted and observed stream temperatures via cross-validation from models developed to predict missing temperature data due to displacement (a–c) or in segments without temperature loggers (d; Supplementary Material S3.2). (a) Predicted stream temperatures for WIL1, March 1–10 and May 29–31, 2012 from models developed from spring 2013 data. (b) Predicted stream temperatures for SFR1, May 2011 from models developed from spring 2013 data. (c) Predicted stream temperatures for SFR1, June 8–August 31, 2012 via leave-one-out cross validation. (d) Relationship between field-recorded temperatures in segments without temperature loggers and predicted temperatures from logger- recorded temperatures, directional network distance between intra-stream segments, and mean catchment area...... 119

Supplementary Table S3.4. Means of instream variables within stream segments. Stream codes are explained in Supplementary Table S3.1...... 120

Supplementary Table S3.5. Segment features including percentage riparian land use (30-m buffer), percentage dominant geologic category (50-m buffer), channel gradient, and channel width. Land use was obtained from the 2011 National Land Cover Dataset (NLCD; Homer et al., 2015) and categorized with a modified Anderson Level 1 classification (Anderson, 1976): Agriculture = NLCD71, NLCD 81, and NLCD 82; Developed = NLCD21, NLCD22, NLCD23, NLCD24, and NLCD31; Forested = NLCD 41, NLCD 42, NLCD 43, NLCD 52, NLCD90, and NLCD95. We categorized geology according to resistance to weathering: Carbonate = dolostone and limestone; Clastic = sandstone and siltstone; Shale = black shale and shale (Dickens et al., 2008; Nicholson et al., 2007). Channel gradient was derived from the National Hydrography Dataset Plus version 2 (McKay et al., 2012). Stream codes are explained in Supplementary Table S3.1...... 122

Supplementary Table S3.6. Cumulative percentages of land uses and geologic categories, catchment areas, and mean elevations upstream of stream segments. Categories of land use and geology are explained in Supplementary Table S3.5. Catchment area and elevation were derived from the National Hydrography Dataset Plus version 2.1 (McKay et al., 2012). Stream codes are explained in Supplementary Table S3.1...... 124

Supplementary Material S3.7. We evaluated the transferability of regression models via a seven-fold cross-validation (Manel, Williams & Ormerod, 2001). First, the 42 stream segments were randomly partitioned into seven equal folds. Next, models were fit using data from six folds (36 segments) and then used to predict the status (i.e., present, absent) of the remaining six segments. The process was repeated six times, and the classification criterion averaged across folds. We used area-under-the-curve (AUC) of the receiver-operating characteristic as a threshold-independent criterion for discriminating between presence and absence. Values of 0.50–0.69 indicated poorly performing models, 0.70–0.90 indicated moderate performance, and > 0.90 indicated high performance (Manel, Williams & Ormerod, 2001)...... 126

xii Supplementary Table S3.8. Akaike’s Information Criterion corrected for low sample size (AICc) for 141 candidate models explaining the contemporary distribution of E. osburniǻ$,&F is the difference between the top-ranked model and lower-ranked models (i). Model weight (Wi) is the probability of being the best-supported model. Area under the curve (AUC) is a threshold- independent measure of cross-validation success. AUC entries are the mean and (standard deviation) resulting from a seven-fold cross-validation...... 127

Supplementary Table S3.9. Predicted probability of occurrence within 42 stream segments from the best-supported distribution model containing spring mean daily temperature (SPMDT), embeddedness index, and the first two spatial covariates. Predictions in column 3 (Median spatial) control for spatial influences via identical median values for both spatial covariates. Predictions in column 4 (Fully spatial) use habitat covariates and true geographic coordinates. Predictions in column 5 (Non-spatial) use the best-supported model without spatial covariates (SPMDT, embeddedness index). Bolded rows are unoccupied segments with suitable instream habitat predicted by the “Median spatial” model. Italicized rows are occupied segments with unsuitable instream habitat predicted by the “Median spatial” model...... 131

Supplementary References S3.10...... 133

xiii CHAPTER 1: General Introduction

Southeastern Fish Biodiversity and Patterns of Imperilment The streams of the southeastern United States are ecological wonders, supporting approximately half (560) of North America’s freshwater fish species (Warren et al. 2000). This fauna was shaped over tens of millennia as climate, geologic uplift, and erosion reconfigured southeastern rivers, providing the necessary isolation and refugia for species to diverge in response to local selective pressures. However, for the last 300 years, and especially the last 100, much of the southeastern fauna has been jeopardized. For example, depending on geographical definition of the “southeast” and taxonomic resolution, 18 to 28% of southeastern are imperiled (Etnier 1997; Warren et al. 1997; Warren et al. 2000). Moreover, the rate of imperilment is increasing, owing primarily to concerted effects of habitat degradation, introduced species, and lost fluvial connectivity (Jelks et al. 2008; Burkhead 2012). Imperilment disproportionally afflicts certain taxa and species traits. Southeastern biodiversity is concentrated in the Interior (Ozarks, Ouachita) and Appalachian highlands. Accordingly, many southeastern fishes depend on upland or highland conditions, which are particularly sensitive to landscape stressors (Utz et al. 2010). For example, the substrate- dependent darters (: ) represent 36% of imperiled southeastern fishes, despite comprising only 26% of southeastern fish richness (Warren et al. 1997). In addition to a benthic syndrome, other seemingly highland traits associated with imperilment of darters and other fish taxa include riffle-dependency (Walters et al. 2003; Jones et al. 1999), invertivory (Berkman and Rabeni 1987), sight feeding (Zamor and Grossman 2007; Buckwalter 2016), small body size (Angermeier 1995; Burkhead et al. 1997), and cool- to cold-water preferences (Scott et al. 2001). Moreover, fishes possessing these traits are especially prone to imperilment when also having localized distributions, small population sizes, or narrow geographic ranges (Angermeier 1995; Pritt and Frimpong 2010), including many fishes endemic to Appalachia (Warren et al. 1997). Certain stream systems are also disproportionally sensitive to landscape stressors. For example, small- to medium-sized rivers are particularly impacted by the accumulation of fine sediment and heat from upstream sources. As a result, clear and cool upland rivers are often converted to silted, turbid, and overall, more lowland-like environments (Etnier 1997; Burkhead

1 et al. 1997). The effects of habitat degradation may be further compounded at the system level by increased susceptibility of degraded streams to invasion by warm-adapted and silt-resistant fishes (Scott et al. 2001; Hitt and Roberts 2012; Lapointe et al. 2012). Although, to my knowledge no one has synthesized this deteriorating-highland hypothesis, the pattern of spatio-temporal replacement of upland fishes by their generalist counterparts has been repeatedly documented throughout the southeast (Harding et al. 1998; Walser and Bart 1999; Sutherland et al. 2002; Walters et al. 2003; Scott 2006; Wenger et al. 2008).

What is Fish Habitat? Newcomb et al. (2007) eloquently define fish habitat as the “…physical and biological components required to support fish growth, survival, and reproduction.” These are the three fundamental life-history functions necessary for an individual to complete its life cycle. What may not be immediately apparent, however, is the importance of habitat hierarchy, diversity, and connectivity for the persistence of populations through time. Identifying fish habitat for a single species is a complex and iterative process requiring detailed observations of at least three within- and among-scale relationships (Schlosser 1991): (1) the relationships among the spatial and temporal scales creating the instream conditions experienced by individuals, (2) relationships between individuals and populations, and (3) relationships among populations across the landscape. Stream habitat is often conceptually organized into a catchment hierarchy consisting of discrete, sequentially nested spatial scales (Frissell et al. 1986). The portfolio of possible habitats at finer spatial scales, such as micro- and meso-habitats, are constrained at larger spatial scales including stream segments, catchments, and regions. When viewing stream habitat at any single spatial scale, heterogeneity is created by asynchrony among processes (runoff, erosion, etc.) and existing structure at both higher and lower spatial scales within the habitat hierarchy (Poole 2002). By altering and synchronizing these habitat-forming processes, the effects of human activities often culminate in reduced habitat diversity and simplified stream ecosystems (Allan 2004). The life-histories of fishes are complex and marked by seasonal and life stage-specific shifts among discrete and varied habitat patches (Schlosser and Angermeier 1995). Yet, individual-level habitat studies often narrowly focus on the habitats used by adults during

2 summer. Moreover, management decisions are often made over short timelines and depend on narrow perceptions of habitat requirements, which may not accurately reflect true population- limiting factors (Rosenfeld 2003). An inadequate understanding of the habitat requirements for many fish species partly reflects logistical and financial constraints on investigators, but also, the sheer complexity of fish life histories. For example, foraging ability – the most well-studied behavioral mode – is important for individual growth, yet do faster growing individuals reproduce more? And do streams with a high prevalence of foraging habitat also have more resilient populations (i.e., higher population growth [Ȝ])? These are elusive questions, even for well-studied model species (Grossman 2013). Fish habitat studies have traditionally focused on documenting habitat associations over small spatial (single reaches) and temporal (1–3 years) extents. However, these study designs risk mistaking the proximate stimuli that affect the habitats used by individuals for the conditions affecting the dynamics of populations. In response, Fausch et al. (2002) put forth the riverscape concept – an approach emphasizing study designs with spatial extents that encompass sufficient habitat diversity to support populations and the connections among these habitats. Specifically, the riverscape approach recognizes the inherent trade-off between sampling effort within versus among sites, and encourages systematically spaced sites at intermediate spatial scales, thereby permitting investigators to identify critical population-limiting habitats, rather than mere habitat associations of individuals (Torgersen et al. 1999; Fausch et al. 2002). Riverscape-inspired study designs intuitively lend themselves to species with obvious wide-ranging life-histories (e.g., salmonids; Petty et al. 2012; Falke et al. 2013). However, the concept’s origin partly reflects the patch dynamics of a species of darter ( Darter Etheostoma cragini) within a single watershed (Labbe and Fausch 2000). Although, the complementary habitats used by small-bodied fishes may exist within small reaches (e.g., riffle- run-pool sequence), an entire population and even the movements of individuals may range dozens of kilometers (Albanese et al. 2004; Roberts et al. 2008). Moreover, the persistence of a population within a stream network may be linked to habitats within specific reaches that consistently export individuals or provide refugia from harsh conditions (Labbe and Fausch 2000; Hanfling and Weetman 2006). Identifying these critical stream habitats would undoubtedly benefit the conservation of small-bodied fish populations, yet spatially extensive investigations are particularly rare for non-game fishes. Moreover, demarcating these core areas may require

3 monitoring the consistency of within-system occupancy over multiple years (Angermeier et al. 2002); however, multi-year investigations – another component of an intermediate-scale approach – are even more rare than intermediate spatial-scale investigations. In addition to reproduction and survival, the dynamics of populations are sensitive to immigration and emigration. Many of the same concepts describing the persistence of a population within a stream, can be extended to explain the persistence of a species across the landscape (Schlosser 1995). For example, it is hypothesized that some stream fish species may have once naturally exhibitted a hybrid meta-population model (sensu Schlosser and Angermeier 1995), consisting of a network of source populations (fecundity > mortality) with variable connectivity to satellite sink populations (fecundity < mortality; Falke and Fausch 2010; Compton 2013). Recently localized (i.e., non-equilibrium meta-population) distributional patterns may reflect both the degradation of habitat in productive source populations and diminished connectivity among populations (Schlosser and Angermeier 1995; Dunham and Rieman 1999). Although documenting movements among populations is difficult, there is increasing awareness that populations of even small-bodied (i.e., ostensibly dispersal-limited) species may have recently been connected via movement through difficult-to-sample corridor habitats (Roberts et al. 2013).

Physiography and Fishes of the New River Drainage The New River (upper Kanawha River) drains ~ 17,700 km2 of the central within the upper River basin. Its headwaters begin in the Blue Ridge (BR) province near Blowing Rock, North Carolina and flow north converging to form the mainstem New River in southern Virginia. The wide mainstem then meanders through the soft carbonate valley floor of the Valley and Ridge (VR) province before turbulently descending over 200 m through the New River Gorge within the Appalachian Plateau (AP) province of West Virginia. Finally, after exiting the gorge, the New River terminates at its confluence with the Gauley River, thereafter forming the Kanawha (“Kah-naw”) River; however, within a biogeographical context, which includes this thesis, the New River drainage (NRD) encompasses the Gauley River drainage (~ 3,700 km2) due to their shared aquatic fauna shaped by mutual isolation from the rest of the basin by Kanawha Falls.

4 The NRD is both topographically and biologically unique largely due to its geographic position and north-south orientation. Contrary to popular belief, there is no strong evidence suggesting the New River is the second oldest river in the world. Moreover, geologic uplift, stream piracy, and climatic fluctuations make it difficult to age Appalachian rivers, including the New River (Bartholomew and Mills 1991; Ward et al. 2005). The Appalachian Mountains were first uplifted in the mid- to late-Paleozoic period (360 – 240 Ma), but many modern-day Appalachian rivers were not formed until the Miocene Epoch (23.0 – 5.3 Ma; Hoagstrom et al. 2014). The New River, however, is certainly older than many of the other rivers in the Ohio River basin. Unlike many northern rivers, which were repeatedly reconfigured during Plio- Pleistocene glaciations (0.1 – 2.6 Ma), the main channel of the New River remained intact owing to the river’s position south of the glacial fronts. This preserved the unique sequence of physiographic provinces traversed by the New River – the only river to traverse the BR, VR, and AP provinces, leading some authors to suggest the New River is as old as the Appalachian Mountains (Dietrich 1959). Differential geology imparts characteristic landforms and stream types in each of the three provinces within the NRD (Hack 1973). The lower New, Bluestone, and Gauley rivers and portions of the Greenbrier River are within the AP. Appalachian Plateau streams and rivers are typically steep and entrenched due to underlying erosion-resistant clastic geology (sandstone, siltstone; Hack 1973). This gives the AP an overall rugged landscape, particularly in the high- elevation (>700 m) “Allegheny Highlands” of the upper Gauley River drainage (Messinger and Hughes 2000). Shallow soils and impermeable bedrock promote more seasonally variable flows than other areas in the drainage. Confined rivers with steep and disproportionally wide channels relative to summer baseflows are common sights throughout the upper Gauley River drainage. The VR is south (upstream) and to the east (Greenbrier River) of the AP, and best characterized as a mosaic of alternating bedrock types, creating a topographically complex karstic landscape. Streams draining watersheds with mostly sandstone geology are similar in character to those in the AP; however, easily-eroded shale and soluble carbonate rocks underlay most VR watersheds and lead to low elevation, moderately sloped streams with well-developed floodplains amenable to development and cattle pastures. Seasonal flows are relatively stable in many VR streams and moderated by high connectivity with well-developed aquifers (Messinger and Hughes 2000).

5 The BR in New River (Blue Ridge Plateau) has subdued topography, unlike the rugged Blue Ridge escarpment of the Atlantic slope and the southern BR of the River drainage (Hack 1982). Streams in the BR also markedly contrast with those in the other two provinces of the NRD, and have been described as “Piedmont-like” based on high turbidity, moderate gradients, and high prevalence of fine sediments consisting of unconsolidated gneiss and schist sand (Dietrich 1959; Jenkins and Burkhead 1994). The stark contrast in the physical attributes of BR streams and the other two provinces may be one mechanism reinforcing the parapatric ranges of Candy E. osburni and Kanawha E. kanawhae darters. These sister species are both deeply dependent on substrate and have parallel ontogenetic habitat shifts (my unpublished data). However, the diminutive Kanawha Darter often thrives in BR streams, which frequently contain finer and more embedded substrates than the VR and AP streams that typically support the Candy Darter. Much of the BR is perched at least 100 m higher in elevation than adjacent VR areas (Prince et al. 2013); therefore, occurrence of the Kanawha Darter in the BR effectively eliminates the potential for the BR to serve as a high-elevation coolwater refugium for the Candy Darter. The unique sequence of provinces traversed by the New River has insulated the drainage from downstream-generated erosive forces and neighboring fish assemblages. Upslope-moving head-cutting by the Kanawha River in response to glacially induced elevational changes in the Teays-Ohio River system have largely stalled in the hard sandstones of the AP in the lower NRD, forming the New River Gorge (Hack 1973). In the absence of head-cutting by the mainstem New River, the rest of the NRD has maintained an exceptionally high elevation relative to surrounding drainages (Hack 1973), allowing the drainage to periodically function as a coolwater refugium for highland fishes and an ecological trap for warmwater fishes. Although stalled, the mainstem NRD is still actively lowering its elevation to minimize elevational differences between it and the Kanawha River, evidenced by the unique convex profile of the mainstem New River and its lower tributaries (Prince et al. 2013). Ongoing lowering of the New River results in sudden elevational drops, creating the prevalent falls along its middle and lower course. Therefore, Kanawha and other falls throughout the New River Gorge may be legacies of the New River’s slow upriver adjustment to downstream elevational changes in the Teays-Ohio River system millennia ago. These falls, especially Kanawha Falls, have prevented downstream fishes from colonizing the NRD, allowing individual species and aquatic communities to evolve

6 in isolation since at least the Pleistocene Epoch (0.1 – 2.6 Ma; Jenkins et al. 1972; Hocutt et al. 1986). The native aquatic fauna of the NRD is a subset of the fauna inhabiting the Kanawha River drainage, with the exception of at least nine endemic fishes and two endemic crayfishes (Jenkins et al. 1972; Neves 1983). Sparse early surveys indicate there are only 44 fish species native to the NRD (Cope 1868; Jordan 1989; Jenkins and Burkhead 1994; Buckwalter 2016). Low richness has been attributed to the overall montane character of the NRD (Addair 1944; Jenkins et al. 1972; Hocutt et al. 1978), high aquatic sulfate levels (Ross and Perkins 1959), underground water loss (Ross and Perkins 1959), stream capture (Ross and Perkins 1959), past competitive interactions (Jenkins et al. 1972), and the severe environmental degradation of streams before early fish surveys (Goldsborough and Clark 1908; Addair 1944). However, the synergistic effects of isolation via Kanawha Falls and the cold peri-glacial climates during Plio- Pleistocene glaciations were likely the biggest influences shaping the present-day native assemblage (Jenkins and Burkhead 1994). Unlike south-flowing Appalachian river drainages, the upslope higher elevations of the southern NRD likely provided little refuge for warmwater species during peri-glacial climates (hereafter referred to as the “cool-water hypothesis”). Cold temperatures likely extirpated most warmwater species, which were subsequently excluded from the NRD by Kanawha Falls (Jenkins and Burkhead 1994). The cool-water hypothesis is mainly supported by the composition of the native fauna of the NRD, consisting mostly of taxa preferring cool to cold stream temperatures (Jenkins and Burkhead 1994). Absent are dozens of warmwater fishes native to the rest of the Kanawha River (lower Kanawha) sub-basin (Messinger and Chambers 2001), and members of several lineages that likely used the NRD as a stepping-stone to move between the Gulf and Atlantic slopes (Jenkins and Burkhead 1994). Only three suckers (Catostomidae), two of which prefer cool water, and one sunfish (Centrarchidae), are confirmed native. Moreover, exempting the catadromous American Eel Anguilla rostra, the native composition of mainstem piscivores includes three eurythermal fishes, Channel Ictalurus punctatus, Flathead Catfish Pylodictus olivaris, and Green Sunfish Lepomis cyanellus, and the coolwater endemic strain of Walleye vitreus (Palmer et al. 2006). The legacy of the former cold climate may also be inferred from thermal preferences of the eleven endemic aquatic species and sub-species. It is debatable whether cold-tolerance is an

7 ancestral highland trait preceding the Plio-Pleistocene epochs or a trait selected by the Plio- Pleistocene epochs (Mayden and Richard 1985; Hoagstrom et al. 2014). Also, the age of Kanawha Falls, and therefore, the New River’s endemics, is unknown (Switzer 2004). However, all but one endemic, the Bigmouth Chub Nocomis platyrhynchus, seemingly prefer cool temperatures based on their distributions centered in high-elevation medium to small streams (Jenkins and Burkhead 1994; Huang 2015). Moreover, many endemics have since disappeared in lower-elevation areas of the VR and AP corresponding to land-use change and warming air temperatures (Easton and Orth 1994; Wellman 2004; Welsh et al. 2006; Huang 2015). Despite WKHWKUHDWVSRVHGWRWKHHQGHPLFVE\ZDUPLQJDLU Û&VLQFHWKHV+XDQJ DQG stream temperatures, only one controlled study has investigated the thermal preferences of any of the endemics; Shingleton et al. (1981) found the New River Shiner scabriceps preferred cooler temperatures than other () inhabiting the NRD, thereby supporting the cool-water hypothesis. In addition to low native richness and high endemism, the NRD has a notoriously high percentage (>55%) of introduced species – by far the highest percentage of any eastern drainage in North America (Jenkins and Burkhead 1994). Buckwalter (2016) updated the list of introduced species compiled by Jenkins and Burkhead (1994) to include 13 additional species, not including Creek Chubsucker Erymizon oblongus (Hitt and Roberts 2012), Southern Redbelly Dace Chrosomus erythrogaster (Welsh et al. 2006), and (re)introduced Thoburnia sp. (my unpublished data). The history of introduced fishes in the NRD has two phases: early government-sponsored introductions of sport-fishes and subsequent introductions of bait-fishes (Cincotta et al. 1999). Due to their early introductions, the spread of introduced sport-fishes across the NRD seems to have slowed; however, most introduced bait-fishes are still linearly or exponentially spreading, while traversing man-made barriers, likely facilitated by within-basin bait-bucket transfers (Buckwalter 2016). Buckwater (2016) also found introduced and spreading fishes often prefer warmer, more lowland conditions, which are presumably more prevalent in the NRD following widespread conversion of native forests to pasture and development. Surprisingly few studies have investigated the potential impacts of introduced fishes on the native fish fauna. Keplinger (2007) found the Spotfin Shiner Cyprinella spiloptera and New River Shiner exhibited modified schooling behavior when integrated with two introduced

8 cyprinids. Hitt and Roberts (2012) discovered the fish communities in three streams have diverged since 1940 (Burton and Odum 1945), owing to varying thermal resistance of focal streams to invasion by introduced warmwater species. However, within-stream communities were more homogenous, partly due to the extirpations of endemics – a pattern more consistent with the global homogenization freshwater assemblages (Olden et al. 2010). In summary, for several millennia, the NRD has served as a time-capsule preserving a unique highland fish assemblage; however, the NRD is becoming less unique each year, owing to the addition of introduced fishes and concurrent extirpations of its highland natives.

The Candy Darter Within this thesis, I focus on the habitat requirements and conservation of the Candy Darter. The Candy Darter was first captured in South Fork Reed Creek, VA in 1885 but misidentified (Jenkins and Kopia 1995). Hubbs and Trautman (1932) later described the species using a holotype from Stony Creek, WV (Addair 1944), and eight additional paratypes from lower Reed Creek, VA. The Candy Darter is endemic to the NRD, ranging across the VR and AP physiographic provinces in VA and WV (Jenkins and Burkhead 1994). The ranges of Candy and Kanawha darters are parapatric, with the latter replacing the former in the BR province. Populations of the Candy Darter are highly localized (Chipps et al. 1993), especially in VA (Jenkins and Kopia 1995). At the inception of this thesis, the only known large population in VA inhabited Big Stony Creek, and smaller populations had last been confirmed in Dismal and Laurel creeks in 1996 (Helfrich et al. 1996; Bye 1997). The Candy Darter was also historically confirmed upstream of Claytor Lake reservoir in the Pine Run and Reed Creek systems, and downstream of Claytor Lake reservoir in VA in the mainstem New River, Spruce Run, Sinking, and Walker creeks (Burton and Odom 1945; Ross and Perkins 1959; Jenkins and Kopia 1995); however, no Candy Darters had been collected at any of these locations since 1974. In WV, Candy Darters inhabit portions of the Greenbrier (Hocutt et al. 1978) and Gauley (Hocutt et al. 1979) river sub- basins, and are likely extirpated from the Bluestone River sub-basin, Indian Creek, and all lower New River tributaries. Historically, Candy Darters inhabited a variety of stream types and sizes; however, more recent collections are typically from cool, high- to moderate-gradient small streams draining

9 highly forested catchments (Jenkins and Burkhead 1994; Stauffer et al. 1995). Observations of in-stream habitat preferences are only available for adults and from only six streams across the species’ range. Chipps et al. (1994) considered Candy Darters a micro- “habitat specialist” inhabiting patches with swift water velocities over large substrates based on observations from three streams in WV. In Dismal, Laurel, and Big Stony creeks, Candy Darters preferred riffle (Helfrich et al. 1996; Leftwich et al. 1996) or glide (Bye 1997) channel units. No information exists on habitat preferences across seasons or through ontogeny, and no study has comprehensively linked the distribution or decline of Candy Darter populations to habitat at any spatial scale. The Candy Darter is a Species of Concern in WV and a Tier I Species of Greatest Conservation Need in VA. At the beginning of this thesis, the Candy Darter was also listed as a federal Species of Concern; however, in October 2017, the U.S. Fish and Wildlife Service recommended the Candy Darter be listed as a Threatened species under the Endangered Species Act (USFWS 2017), a decision that was partly informed by findings in this thesis. Primary hypothesized reasons for the species’ decline include siltation (Chipps 1992; Jenkins and Kopia 1995), elevated stream temperatures (Jenkins and Kopia 1995), reservoir construction (Jenkins and Burkhead 1994; Welsh et al. 2006), stream acidification (Chipps et al. 1993), and predation by introduced piscivores (Chipps et al. 1993). The recent introduction of Variegate Darter (E. variatum), and its subsequent spread and hybridization with Candy Darters is also a serious threat (Switzer et al. 2007; Dunn 2013; Gibson 2017).

Overarching Goal and Approach The goal of this thesis is to document the habitat associations of the Candy Darter at multiple spatio-temporal scales, thereby addressing several knowledge gaps germane to the species’ degree of imperilment. Chapters are meant to complement each other, providing desperately needed accounting and ecological information on the species’ life-history (sensu Warren et al. 1997). My general approach was inspired by concepts emphasizing large spatial extents and multiple scales of investigation (Schlosser 1991; Schlosser and Angermeier 1995; Fausch et al. 2002). Although not definitive, the findings from this thesis may provide a foundation for future research on the Candy Darter.

10 I also treat the Candy Darter as a model species, representative of imperiled highland and southeastern fishes. For example, like most non-game fishes, the limited available life-history information for the Candy Darter is drawn from a patchwork of anecdotal observations and small-scale habitat investigations. As a previously described microhabitat specialist, the Candy Darter is amenable to investigating the sensitivity of predicted habitat suitability across seasons, habitat variables, and life stages. In Chapter 2, I hypothesize seasonal and ontogenetic shifts in habitat selection may contradict the existing perception of suitable habitat, which is largely derived from observations of adult foraging habitat in summer. The Candy Darter also seemingly possesses most traits associated with imperilment; however, no investigation has comprehensively re-surveyed historically occupied localities with effective methods to clarify its contemporary distribution. Therefore, in addition to specific research questions, Chapter 3 tests whether the contemporary distribution of the Candy Darter is consistent with the highly localized distributional patterns of many imperiled fishes throughout the southeastern United States. Contrasts between places where the Candy Darter persists and where it is extirpated should also help diagnose threats and identify resilient stream habitats.

References Addair, J. 1944. The fishes of the Kanawha River system in West Virginia and some factors which influence their distribution. Ph.D. dissertation. The Ohio State University, Columbus, Ohio.

Albanese, B., P. L. Angermeier, and S. Dorai-Raj. 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. Annual Review of Ecology Evolution and Systematics 35:257–284

Angermeier, P. 1995. Ecological attributes of extinction-prone species: loss of freshwater fishes of Virginia. Conservation Biology 9(1):143–158.

11 Angermeier, P., K. Krueger, and C. Dolloff. 2002. Discontinuity in stream-fish distributions: implications for assessing and predicting species occurrence. Pages 519–527 in J.M. Scott, P. Huglund, M.L. Morrison, J.B. Haufler, M.G. Raphael, W.A. Wall, and F.B. Sampson, editors. Predicting species occurrences: issues of accuracy and scale. Island Press, Covelo, California. Island Press, Covelo, California.

Bartholomew, M. J., and H. H. Mills. 1991. Old courses of the New River: its late Cenozoic migration and bedrock control inferred from high-level stream gravels, southwestern Virginia. Geological Society of America Bulletin 103(1):73–81.

Berkman, H. E., and C. F. Rabeni. 1987. Effect of siltation on stream fish communities. Environmental Biology of Fishes 18(4):285–294.

Buckwalter, J. D. 2016. Temporal trends in stream-fish distributions, and species traits as invasiveness drivers in New River (USA) tributaries. Master’s thesis. Virginia Polytechnic Institute and State University, Blacksburg, Virginia.

Burkhead, N. M. 2012. Extinction rates in North American freshwater fishes, 1900–2010. BioScience 62:798–808.

Burkhead, N., S. Walsh, B. Freeman, and J. Williams. 1997. Status and restoration of the Etowah River, an imperiled southern Appalachian ecosystem. Pages 375–444 in G. W. Benz, and D. E. Collins, editors. Aquatic fauna in peril: the southeastern perspecive. Southeast Aquatic Research Institute, Lenz Design and Communications, Decatur, Georgia.

Burton, G. W., and E. P. Odum. 1945. The distribution of stream fish in the vicinity of Mountain Lake, Virginia. Ecology 26(2):182–194.

Bye, M. B. 1997. Summary of 1996 activity concerning native and transplanted populations of candy darters in Dismal Creek, VA. Report to Candy Darter Conservation Committee, Blacksburg, Virginia.

12 Chipps, S. R. 1992. Stream fish communities of the central Appalachian Plateau: an examination of trophic group abundance patterns and resource partitioning among benthic fishes. West Virginia University, Morgantown, West Virginia.

Chipps, S. R., W. B. Perry, and S. A. Perry. 1993. Status and distribution of Phenacobius teretulus, Etheostoma osburni, and “Rhinichthys bowersi” in the Monongahela National Forest, West Virginia. Virginia Journal of Science 44(1):48–58.

Chipps, S. R., W. B. Perry, and S. A. Perry. 1994. Patterns of microhabitat use among four species of darters in three Appalachian streams. American Midland Naturalist 131(1):175–180.

Cincotta, D.A., D.B. Chambers, and T. Messinger. 1999. Recent changes in the distribution of fish species in the New River basin in West Virginia and Virginia. Proceedings of the New River Symposiun: 98–106.

Compton, M., and C. Taylor. 2013. Spatial scale effects on habitat associations of the Ashy Darter, Etheostoma cinereum, an imperiled fish in the southeast United States. Ecology of Freshwater Fish 22(2):178–191.

Cope, E.D. 1868. On the distribution of fresh-water fishes in the Allegheny region of southwestern Virginia. Journal of the Academy of Natural Sciences of Philadelphia 1869: 207–247.

Dietrich, R. V. 1959. Geology and mineral resources of Floyd County of the Blue Ridge Upland, southwestern Virginia. Bulletin of the Virginia Plytechnic Institute Engineering Experiment Station Series 134, 52 (12): 1–160.

13 Dunham, J. B., and B. E. Rieman. 1999. Metapopulation structure of bull trout: influences of physical, biotic, and geometrical landscape characteristics. Ecological Applications 9(2):642–655.

Dunn, C. G. 2013. Comparison of habitat suitability among sites supporting strong, localized, and extirpated populations of candy darter (Etheostoma osburni). Final Report to Virginia Department of Game and Inland Fisheries, Richmond, Virginia.

Easton, R. S., and D. Orth. 1994. Fishes of the main channel New River, West Virginia. Virginia Journal of Science 45(4):265–278.

Etnier, D. A. 1997. Jeopardized southeastern freshwater fishes: a search for causes. Pages 87– 104 in G. W. Benz, and D. E. Collins, editors. Aquatic fauna in peril: the southeastern perspecive. Southeast Aquatic Research Institute, Lenz Design and Communications, Decatur, Georgia.

Falke, J. A., J. B. Dunham, C. E. Jordan, K. M. McNyset, and G. H. Reeves. 2013. Spatial ecological processes and local factors predict the distribution and abundance of spawning by steelhead (Oncorhynchus mykiss) across a complex riverscape. Plos One 8(11):e79232.

Falke, J. A., and K. D. Fausch. 2010. From metapopulations to metacommunities: linking theory with empirical observations of the spatial population dynamics of stream fishes. Pages 207–233 in K.B. Giod and D.A. Jackson, editors. Community ecology of stream fishes: concepts, approaches, and techniques. American Fisheries Society, Symposium 73, Bethesda, Maryland.

Fausch, K. D., C. E. Torgersen, C. V. Baxter, and H. W. Li. 2002. Landscapes to riverscapes: bridging the gap between research and conservation of stream fishes. BioScience 52(6):483–498.

14 Frissell, C. A., W. J. Liss, C. E. Warren, and M. D. Hurley. 1986. A hierarchical framework for stream habitat classification: viewing streams in a watershed context. Environmental Management 10(2):199–214.

Gibson, I. 2017. Conservation concerns for the Candy Darter (Etheostoma osburni) with Implications related to hybridization. Master’s thesis. West Virginia University, Morgantown, West Virginia.

Goldsborough, E. L. and H. W. Clark. 1908. Fishes of West Virginia. Department of Commerce and labor. Bulletin of Bureau of Fisheries 27:29–39.

Grossman, G. D. 2013. Not all drift feeders are trout: a short review of fitness-based habitat selection models for fishes. Environmental Biology of Fishes:1–9.

Hack, J.T. 1973. Drainage adjustment in the Appalachians. Pages 51–69 in M. Morisawa, editor. Fluvial geomorphology. State University of , Binghamtom, New York.

Hack, J. T. 1982. Physiographic divisions and differential uplift in the Piedmont and Blue Ridge. U.S. Geological Survey Professional Papers 1265.

Hänfling, B., and D. Weetman. 2006. Concordant genetic estimators of migration reveal anthropogenically enhanced source-sink population structure in the river sculpin, Cottus gobio. Genetics 173(3):1487–1501.

Harding, J. S., E. F. Benfield, P. V. Bolstad, G. S. Helfman, and E. B. D. Jones. 1998. Stream biodiversity: the ghost of land use past. Proceedings of the National Academy of Sciences of the United States of America 95(25):14843–14847.

Helfrich, L. A., M. B. Bye, and D. Dalton. 1996. Life history, status, and recovery of the candy darter Etheostoma osburni, in Virginia. Final Report to the Virginia Department of Game and Inland Fisheries, Richmond, Virginia.

15 Hitt, N. P., and J. H. Roberts. 2012. Hierarchical spatial structure of stream fish colonization and extinction. Oikos 121(1):127–137.

Hoagstrom, C. W., V. Ung, and K. Taylor. 2014. Miocene rivers and taxon cycles clarify the comparative biogeography of North American highland fishes. Journal of Biogeography 41(4):644–658.

Hocutt, C., R. Denoncourt, and J. Stauffer Jr. 1979. Fishes of the Gauley River, West Virginia. Brimleyana 1:47–80.

Hocutt, C. H., R. F. Denoncourt, and J. R. Stauffer Jr. 1978. Fishes of the Greenbrier River, West Virginia, with drainage history of the central Appalachians. Journal of Biogeography:59– 80.

Huang, J. 2015. Assessing predictive performance and transferability of species distribution models for freshwater fish in the United States. Ph.D. dissertation. Virginia Polytechnic Institute and State University, Blacksburg, Virginia.

Hubbs, C. L., and M. B. Trautman. 1932. Poecilichthys osburni, a new darter from the upper Kanawha River system in Virginia and West Virginia. Ohio Journal of Science 32:31–38.

Jelks, H. L., and coauthors. 2008. of imperiled North American freshwater and diadromous fishes. Fisheries 33(8):372–407.

Jenkins, R. E., and N. M. Burkhead. 1994. Freshwater fishes of Virginia. American Fisheries Society Press, Bethesda, Maryland.

Jenkins, R. E., and B. L. Kopia. 1995. Population status of the candy darter, Etheostoma osburni, in Virginia 1994–95, with historical review. Department of Biology, Roanoke College, Final Report, Salem, Virginia.

16 Jenkins, R. E., E. A. Lachner, and F. J. Schwartz. 1972. Fishes of the central Appalachian drainages: their distribution and dispersal. The distributional history of the biota of the southern Appalachians. Virginia Polytechnic Institute and State University Research Division Monograph 4:43–117.

Jones, E. B. D., G. S. Helfman, J. O. Harper, and P. V. Bolstad. 1999. Effects of riparian forest removal on fish assemblages in southern Appalachian streams. Conservation Biology 13(6):1454–1465.

Jordan, D. S. 1889. Report of explorations made during the summer and autumn of 1888, in the Alleghany region of Virginia, North Carolina and Tennesee, and in western , with an account of the fishes found in each of the river basins of those regions. U.S. Fish Commission Bulletin 88(1888):97–173.

Keplinger, B. J. 2007. An experimental study of vertical habitat use and habitat shifts in single- species and mixed-species shoals of native and nonnative congeneric cyprinids. Master’s thesis. West Virginia University, Morgantown, West Virginia.

Labbe, T. R., and K. D. Fausch. 2000. Dynamics of intermittent stream habitat regulate persistence of a threatened fish at multiple scales. Ecological Applications 10(6):1774– 1791.

Lapointe, N. W., J. T. Thorson, and P. L. Angermeier. 2012. Relative roles of natural and anthropogenic drivers of watershed invasibility in riverine ecosystems. Biological Invasions 14(9):1931–1945.

Leftwich, K. N., C. A. Dolloff, M. K. Underwood, and M. Hudy. 1996. The candy darter (Etheostoma osburni) in Stony Creek, George Washington – Jefferson National Forest, Virginia: trout predation, distribution, and habitat associations. U.S. Forest Service, Blacksburg, Virginia.

17 Messinger, T., and D. B. Chambers. 2001. Fish communities and their relation to environmental factors in the Kanawha River Basin, West Virginia, Virginia, and North Carolina, 1997– 1998. U.S. Geological Survey, Water-Resources Investigations Report 01-4048, Charleston, West Virginia.

Messinger, T., and C. Hughes. 2000. Environmental setting and its relations to water quality in the Kanawha River basin. U.S. Geological Survey, Water-Resources Investigations Report 00-4020, Charleston, West Virginia.

Neves, R. 1983. Distributional history of the fish and mussel fauna in the Kanawha River Drainage. Proceedings of the New River Symposium: 47–67.

Newcomb, T. J., D. J. Orth, and D. F. Stauffer. 2007. Habitat evaluation. Pages 843–886 in C. S. B. Guy, M. L., editor. Analysis and interpretation of freshwater fisheries data. American Fisheries Society, Bethesda, Maryland.

Olden, J. D., and coauthors. 2010. Conservation biogeography of freshwater fishes: recent progress and future challenges. Diversity and Distributions 16(3):496–513.

Palmer, G. C., C. Culver, D. Dutton, B.R. Murphy, E. M. Hallerman, N. Billington, J. Williams. 2006. Genetic distinct walleye stocks in Claytor Lake and the upper New River, Virginia. Proceedings of the Southeast Fish and Wild Agencies 60:125–131.

Petty, J. T., J. L. Hansbarger, B. M. Huntsman, and P. M. Mazik. 2012. Brook trout movement in response to temperature, flow, and thermal refugia within a complex Appalachian riverscape. Transactions of the American Fisheries Society 141(4):1060–1073.

Poole, G. C. 2002. Fluvial landscape ecology: addressing uniqueness within the river discontinuum. Freshwater Biology 47(4):641–660.

18 Prince, P. S., and J. A. Spotila. 2013. Evidence of transient topographic disequilibrium in a landward passive margin river system: knickpoints and paleo-landscapes of the New River basin, southern Appalachians. Earth Surface Processes and Landforms 38(14):1685–1699.

Pritt, J. J., and E. A. Frimpong. 2010. Quantitative determination of rarity of freshwater fishes and implications for imperiled-species designations. Conservation Biology 24(5):1249– 1258.

Roberts, J., A. Rosenberger, B. Albanese, and P. Angermeier. 2008. Movement patterns of endangered Roanoke logperch (Percina rex). Ecology of Freshwater Fish 17(3):374–381.

Roberts, J. H., P. L. Angermeier, and E. M. Hallerman. 2013. Distance, dams and drift: what structures populations of an endangered, benthic stream fish? Freshwater Biology 58(10):2050–2064.

Rosenfeld, J. 2003. Assessing the habitat requirements of stream fishes: an overview and evaluation of different approaches. Transactions of the American Fisheries Society 132(5):953–968.

Ross, R. D., and B. D. Perkins. 1959. Drainage evolution and distribution problems of the fishes of the New (Upper Kanawha) River system in Virginia, part 3: records of fishes of the New River. Virginia Agricultural Experiment Station Technical Bulletin 145.

Schlosser, I. J. 1991. Stream fish ecology: a landscape perspective. BioScience 41(10):704–712.

Schlosser, I. J. 1995. Critical landscape attributes that influence fish population-dynamics in headwater streams. Hydrobiologia 303(1-3):71–81.

Schlosser, I. J., and P. L. Angermeier. 1995. Spatial variation in demographic processes of lotic fishes: Conceptual models, empirical evidence, and implications for conservation. Pages

19 392–401 in J. L. Nielsen, editor. Evolution and the aquatic ecosystem: defining unique units in population conservation. American Fisheries Society, Symposium 17, Bethesday, Maryland.

Scott, M. C. 2006. Winners and losers among stream fishes in relation to land use legacies and urban development in the southeastern US. Biological Conservation 127(3):301–309.

Scott, M. C., and G. S. Helfman. 2001. Native invasions, homogenization, and the mismeasure of integrity of fish assemblages. Fisheries 26(11):6–15.

Shingleton, M. V., C. H. Hocutt, and J. R. Stauffer. 1981. Temperature preference of the new river shiner. Transactions of the American Fisheries Society 110(5):660–661.

Stauffer, J. R., J. M. Boltz, and L. R. White. 1995. The fishes of West Virginia. Proceedings of the Academy of Natural Sciences of Philadelphia 146:1–389.

Sutherland, A. B., J. L. Meyer, and E. P. Gardiner. 2002. Effects of land cover on sediment regime and fish assemblage structure in four southern Appalachian streams. Freshwater Biology 47(9):1791–1805.

Switzer, J. F. 2004. Molecular systematics and phylogeography of the species group (: Percidae). Doctoral dissertation. Saint Louis University, St. Louis, Missouri.

Switzer, J. F., S. A. Welsh, and T. L. King. 2007. A molecular genetic investigationof hybridization between Etheostoma osburni and Etheostoma variatum in the New River drainage, West Virginia. Final Report to West Virginia Division of Natural Resources, Elkins, West Virginia.

Taylor, C. M., and M. L. Warren Jr. 2001. Dynamics in species composition of stream fish assemblages: environmental variability and nested subsets. Ecology 82(8):2320–2330.

20 Torgersen, C. E., D. M. Price, H. W. Li, and B. A. McIntosh. 1999. Multiscale thermal refugia and stream habitat associations of chinook salmon in northeastern Oregon. Ecological Applications 9(1):301–319.

USFWS (U.S. Fish and Wildlife Service). 2017. Endangered and threatened wildlife and plants; proposed Threatened species status for the Candy Darter. Federal Register 82(191):16197–46205.

Utz, R. M., R. H. Hilderbrand, and R. L. Raesly. 2010. Regional differences in patterns of fish species loss with changing land use. Biological Conservation 143(3):688–699.

Walser, C. A., and H. L. Bart. 1999. Influence of agriculture on in-stream habitat and fish community structure in Piedmont watersheds of the Chattahoochee River System. Ecology of Freshwater Fish 8(4):237–246.

Walters, D. M., D. S. Leigh, and A. B. Bearden. 2003. Urbanization, sedimentation, and the homogenization of fish assemblages in the Etowah River Basin, USA. Hydrobiologia 494(1-3):5–10.

Ward, D. J., J. A. Spotila, G. S. Hancock, and J. M. Galbraith. 2005. New constraints on the late Cenozoic incision history of the New River, Virginia. Geomorphology 72(1):54–72.

Warren Jr, M. L., P. L. Angermeier, B. M. Burr, and W. R. Haag. 1997. Decline of a diverse fish fauna: patterns of imperilment and protection in the southeastern United States. Pages 105–164 in G. W. Benz, and D. E. Collins, editors. Aquatic fauna in peril: the southeastern perspecive. Southeast Aquatic Research Institute, Lenz Design and Communications, Decatur, Georgia.

Warren, M. L., and coauthors. 2000. Diversity, distribution, and conservation status of the native freshwater fishes of the southern United States. Fisheries 25(10):7–31.

21 Wellman, D. I. 2004. Post-flood recovery and distributions of fishes in the New River Gorge National River, West Virginia. Master’s thesis. West Virginia University, West Virginia.

Welsh, S. A., D. A. Cincotta, and J. F. Switzer. 2006. Fishes of Bluestone National Scenic River. National Park Service, Technical Report NPS/NER/NRTR – 2006/049, Philadelphia, .

Wenger, S. J., J. T. Peterson, M. C. Freeman, B. J. Freeman, and D. D. Homans. 2008. Stream fish occurrence in response to impervious cover, historic land use, and hydrogeomorphic factors. Canadian Journal of Fisheries and Aquatic Sciences 65(7):1250–1264.

Wiley, E. O., and R. L. Mayden. 1985. Species and speciation in phylogenetic systematics, with examples from the North American fish fauna. Annals of the Missouri botanical Garden 72(4):596–635.

Zamor, R. M., and G. D. Grossman. 2007. Turbidity affects foraging success of drift-feeding rosyside dace. Transactions of the American Fisheries Society 136(1):167–176.

22 CHAPTER 2: Development of Habitat Suitability Indices for Candy Darter Etheostoma osburni, with Cross-Scale Validation across Representative Populations

Keywords: Distribution, Stream Fish, Non-Game Species, Habitat Use, Fish Conservation, Cross-Scale Validation, Ontogenetic Shift

Abstract Understanding relations between habitat associations for individuals and habitat factors that limit populations is a primary challenge for managers of stream fishes. While habitat use by individuals can provide insight into the adaptive significance of selected microhabitats, not all habitat variables will be significant at the population level, particularly when distributional patterns partially result from habitat degradation. In this study we used underwater observation to quantify microhabitat selection by an imperiled stream fish, Candy Darter Etheostoma osburni, in two streams with robust populations. We developed multi-variable and multi-life stage habitat suitability indices (HSIs) from microhabitat selection patterns and used these to assess the suitability of available habitat in streams where populations of Candy Darter are extirpated, localized, or robust. Next, we used a comparative framework to examine relationships among a) habitat availability across streams, b) projected habitat suitability of each stream, and c) a rank for the likely long-term viability (robustness) of the population inhabiting each stream. Habitat selection was characterized by ontogenetic shifts from low-velocity, slightly embedded areas as age-0 fish to swift, shallow areas with little fine sediment and complex substrate as adults. Overall, HSIs were strongly correlated with population rank. However, we observed weak or inverse relationships between predicted individual habitat suitability and population robustness for multiple life stages and variables. The results demonstrated that microhabitat selection by individuals does not always reflect population robustness, particularly when based on a single life stage or season, which highlights the risk of generalizing habitat selection observed during non-stressful periods or for non-critical resources. These findings suggest stream-fish managers may need to be cautious when implementing conservation measures based solely on observations of habitat selection by individuals, and detailed study at both individual and population levels may be necessary to identify habitat that limits populations.

23 Introduction A clear understanding of habitat requirements is essential for effective species management (Rosenfeld 2003). In stream networks, habitat is hierarchically organized into discrete spatial scales spanning large river basins to microhabitats, which facilitate persistence of populations as well as growth and reproduction of individuals (Frissell et al. 1986). Incompatibility between a species’ life history requirements and available resources can exclude a species from an area at any spatial scale within the habitat hierarchy (Schlosser and Angermeier 1995). Often a clear (i.e., mechanistic) understanding of habitat requirements is obtained only after the integration of findings from numerous observational and experimental studies spanning multiple levels of ecological organization (Rosenfeld 2003). The decline of North America’s rich freshwater fish fauna over the last century partly reflects an inadequate understanding of basic habitat requirements and how anthropogenic changes to aquatic ecosystems impinge on those requirements (Jelks et al. 2008; Burkhead 2012). Regional declines of many species are characterized by the gradual dissolution of a network of populations. Individual populations are lost due to sudden anthropogenic or natural events or the accumulation of years of population declines owing to altered environmental conditions (Angermeier 1995). This process often results in a distributional pattern of disjunct populations scattered across the landscape in locations with sufficient habitat quality and quantity to support positive or neutral growth in the absence of immigration (Schlosser and Angermeier 1995). Often pre-disturbance conditions are undocumented, leaving managers tasked with recovering a species without a) a true reference of normal population function within affected areas and/or b) critical knowledge of life history that applies across the species’ range. However, remaining populations and associated environmental conditions can inform management. Within areas still supporting populations, information beneficial to species recovery includes an understanding of available habitat structure, how individuals interact with the environment, and which variables influence individual fitness and population function. Population-level metrics (e.g., presence, density, demographic rates) are normally measured via extensive surveys across the distributional range of focal species. Extensive surveys provide a representative sample of possible physical habitat configurations across the species’ range, and therefore, are less susceptible to site-specific biases (Newcomb et al. 2007). However, researchers must often balance the extent of surveys with the sampling intensity per

24 site. In particular, extensive surveys may be infeasible for non-game species, which have historically received less attention (Loomis and White 1996; White 1996; Gabelhouse 2005). Furthermore, extensive surveys may merely reveal correlative population-level responses across space but miss underlying mechanisms, particularly when rare habitats at specific life stages ultimately regulate populations (Torgersen et al. 1999; Fausch et al. 2002). Therefore, detailed study of individual habitat use is frequently employed to identify factors limiting populations. Fish-habitat relationships are frequently quantified in the form of habitat suitability models. Management uses of these models include characterizing important habitat types (Guay et al. 2000; Haxton et al. 2008; Midway et al. 2010), guiding habitat augmentation (Boavida et al. 2012), and increasingly, identifying suitable habitat for species reintroduction (Mattingly and Galat 2002; Dixon and Vokoun 2009). Models vary in complexity, but most individual-level models assume individuals actively select conditions that optimize fitness within the context of specific behavioral modes (e.g., reproduction, foraging, and refuge use). For example, foraging individuals try to maximize the ratio of energy intake to expenditure while minimizing mortality risk (Werner and Gilliam 1984; Grossman 2013). Habitat suitability models frequently use the density of individuals occupying a habitat type as a metric for habitat suitability (Rosenfeld 2003); however, this metric can be affected by plasticity of habitat use (Leftwich et al. 1997), resource availability (Dunham et al. 2002), biotic interactions (Orth 1987), ontogeny (Rosenberger and Angermeier 2003), and behavior mode (Kwak et al. 1992). Thus, individual- level models of habitat use frequently perform poorly outside the spatiotemporal context in which they were developed (Fausch et al. 1988; Leftwich et al. 1997; Hewitt et al. 2009). Further, individual-level habitat studies rarely examine links to population-level responses (Peckarsky et al. 1997). Although recent methodological advances allow researchers to explicitly link individual- and population-level patterns using individual-based models (Grimm and Railsback 2005), these models may be infeasible, except for well-studied species, due to extensive data requirements. The primary goal of this study was to examine if the predicted microhabitat suitability of an imperiled stream fish, the Candy Darter Etheostoma osburni, is consistent with population robustness across four streams. Herein, “robustness” reflects population size, density, and likely long-term viability. To accomplish this, we used a study design that revealed relationships among the three primary factors relevant to the development and application of individual-level

25 habitat suitability models: i) in-stream habitat gradients, ii) individual habitat selection, and iii) population robustness across streams. First, individual-level habitat selection (i.e., disproportional use) was estimated from two streams with robust populations that presumably contain optimal habitat (i.e., reference-condition approach; Stoddard et al. 2006, Newcomb et al. 2007). Next, we validated habitat selection by examining the predicted suitability of available habitat within streams where populations of Candy Darter are robust, localized, or extirpated. By comparing habitat gradients, predicted suitability, and actual population robustness, we examined a seldom-tested assumption of habitat suitability models developed from individual- level habitat selection: patch quality perceived by individuals at the microhabitat scale can be “scaled up” to reflect population robustness at the stream-segment scale.

Methods Focal species.— The Candy Darter is endemic to the New River drainage where the species is patchily distributed across the Appalachian Plateau and Valley and Ridge physiographic provinces in Virginia and West Virginia (Chipps et al. 1993; Jenkins and Burkhead 1994). Candy Darters historically inhabited many stream types (Jenkins and Burkhead 1994); currently, however, most populations remain in cool high- to moderate-gradient streams in forested watersheds. The reduced range may be due to habitat degradation, but this hypothesis has received little investigation. Within streams, adults almost exclusively occupy patches with swift flow and coarse substrates (Chipps et al. 1994). Habitat use by immature life stages has never been described (Supplementary Table S2.1) despite the importance of these life stages for population dynamics (Schlosser 1985; Schlosser 1998), and little is known about habitat use and behavior in early spring during spawning season. The current management of the species is similar to many non-game species in that managers must use a framework that is missing critical pieces of information and also suffers from a lack of cohesiveness among the patchwork of small-scale studies describing individual-level habitat associations from different portions of the species’ distributional range. Coherent relations between individual habitat selection and population robustness are needed to inform managers about which recovery actions are likely to be cost-effective.

26 Field Sites.— We selected four streams where populations of Candy Darter are robust, localized, or extirpated (Figure 2.1). Two streams supporting large populations (hereafter status = “robust populations”) were selected to develop habitat suitability models based on literature and preliminary sampling. These are South Fork Cherry (SFC) and East Fork Greenbrier (EFG) rivers, WV, which are third- and fourth-order streams located in the Gauley and Greenbrier sub- basins, respectively. Both streams primarily drain forested watersheds at high elevations (>700 m) within the Appalachian Mountains (Messinger and Hughes 2000). We selected relatively undisturbed and accessible 5-km sections of stream in both EFG and SFC to examine microhabitat selection. Each study section was divided into five 1-km segments. Then 300-m sites from the first (downstream), third, and fifth (upstream) segments were randomly selected to survey. Due to prohibited access in the fifth segment of EFG, we randomly selected a 300-m site between the first and third segments. Randomization ensured the 900 m of survey effort per stream and season (3.6 km in total) was spatially representative of study sections within each stream. Laurel Creek (LC), VA is a third-order stream within the Valley and Ridge physiographic province containing a small isolated population of Candy Darter (hereafter status = “localized population”). The population in LC is likely self-sustaining with the closest known population being approximately 50 fluvial km away and no evidence of connectivity between the two. Systematic habitat surveys (described below) were conducted in LC beginning at the mouth and extending 4.2 km upstream to a series of small impoundments, which encompasses the entire known range of the population in the system. Sinking Creek (SC), VA (hereafter status = “extirpated population”) is one of five systems in Virginia where Candy Darter are extirpated and is a candidate site for reintroduction. Burton and Odum (1945) collected one individual over the summers of 1938–1941. However, no other records for the species exist in this heavily surveyed system (Jenkins and Burkhead 1994; Hitt and Roberts 2012). The collection of only one specimen in Sinking Creek is consistent with early records from other streams where the species has been extirpated. By the time of the first significant fish surveys in the Virginia portion of the New River drainage (1940s), Candy Darters were localized and always rare in streams where they are now extirpated (Jenkins and Kopia 1995). In the study segment, SC is a fourth-order stream with channel and land-cover characteristics typical of a large stream in the Valley and Ridge physiographic province. Habitat

27 surveys (described below) were conducted in SC at systematically spaced sites within a 5.5-km segment near the original collection locality. Underwater observation.— We sampled microhabitat use and availability in spring and late summer/fall (hereafter fall) to examine possible behavioral changes and differences in habitat availability between seasons. Spring sampling occurred during high flows and spawning season (May–early June, 2011), while fall sampling corresponded with low flows and non- spawning season (August–October, 2011). Within EFG and SFC, we used direct underwater observation (snorkeling) during baseflow to record the suite of microhabitat conditions immediately associated with each individual. We ensured sufficient water clarity by only sampling when turbidity was < 5.00 NTUs. Beginning at the most downstream point of a study site, the stream was longitudinally divided into two halves with a snorkeler assigned to survey each half. Snorkelers proceeded upstream at the same pace while searching under rocks and moving laterally between the center of the stream and the bank. When a snorkeler spotted a Candy Darter, the snorkeler used a ruler to estimate its total length (TL) by either directly measuring the individual or by measuring a nearby rock of comparable size. Nearly all lengths were estimated less than one meter from individual fish. If a snorkeler influenced an individual’s initial position, the observation was omitted. While spring habitat use by adults reflected the areas occupied during spawning season (staging areas for spawning) the exact microhabitat patches used for spawning within staging areas were not quantified because spawning followed courtship behavior, and observations of habitat use were restricted to the first sighting of individual darters. Snorkelers classified each fish as one of three life stages based on lengths at maturity reported by Jenkins and Burkhead (1994: 827–830), our own observations of lengths at maturity from collected individuals, and pigmentation differences among life stages and sexes. Visual estimation of fish lengths and attribution of life stages during underwater observation have been previously used when examining habitat use by darters (Mattingly and Galat 2002; Ashton and

Layzer 2010) including Candy Darter (Chipps et al. 1994). FHPDOHV•PP7/DQGPDOHV• mm TL were classified as adults. Individuals 46–59 mm TL were considered juveniles. Some individuals 60–64 mm TL that were clearly juvenile males, based on pigmentation, were FODVVLILHGDVMXYHQLOHV$OOLQGLYLGXDOV”PP7/ZHUHFODVVLILHGDV age-0 regardless of season. All age-0 individuals were post-larval and ranged from 17 to 45 mm TL. A length-frequency

28 histogram constructed from estimated lengths contained three modes corresponding with the three life stages that we monitored (Supplementary Figure S2.5). After recording TL, the snorkeler placed a weighted florescent flag at the exact location of each fish and guided the individual downstream to prevent double counting. After snorkelers finished flagging darter locations, five microhabitat variables were recorded at each flag. We measured depth with a top-setting wading rod and average water- column velocity at 60% depth using a Marsh-McBirney model 2000 flow meter. The nearest substrate particle was classified based on intermediate axis into one of nine ordered substrate- size categories according to a modified Wentworth scale: silt (<0.06 mm), sand (0.07–2.0 mm), gravel (3.0–16 mm), pebble (17–64 mm), small cobble (65–128 mm), large cobble (129–256 mm), boulder (257–1000 mm), large boulder (>1000 mm) and bedrock. Finally, within the 0.5- m2 area surrounding each flag, we visually estimated the average depth of rocks embedded by fine substrates (hereafter, embeddedness) and the surface area covered by silt (hereafter, silt cover). Percentages of both metrics were subsequently coded into five ordered categories (Newcomb et al. 2007: 846): 0 = ”5%, 1 = 6–25%, 2 = 26–50%, 3 = 51–75%, 4 = 76–100%. Category 0 (= ”5%) represented observations with no perceived embeddedness or silt.

Microhabitat availability.— Immediately after recording habitat use, the availability of microhabitats in EFG and SFC was measured by placing transects perpendicular to flow spanning the study site. In spring, 30 transects were placed every 10 meters; in fall, 15 transects were placed every 20 meters. We used data from spring to determine that the number of transects could be reduced to 15 per site without affecting the relative frequencies of available habitat categories. Beginning 1 m from the right descending bank, a field-crew member recorded depth, average water velocity, substrate size, embeddedness, and silt cover every two meters along each transect using the same protocols for microhabitat use (described above). This ensured available habitat points were proportional to the area of each stream to reduce error when pooling observations across all sites and both streams (EFG and SFC). Observations were not pooled across seasons. Post-hoc inspection of frequency distributions and multivariate space representing available microhabitats showed similar physical habitats in both streams; therefore, error resulting from pooling across sites and streams was likely minimal. Further, habitat selection (described below) showed no clear bimodality, which would have likely resulted from

29 stream-specific differences in habitat availability rather than consistent responses to measured habitat gradients from separate streams. We applied habitat suitability models developed from EFG and SFC to available in- stream habitat in LC and SC to assess the ability of models to predict suitable habitat for populations in a region outside of the context of original model development. We employed a sampling design that systematically quantified available microhabitats throughout the 4.2-km section of LC and sampled a comparable extent, 5.5 km, in SC. For LC and SC, we delineated sites by randomly selecting one of the first four riffles at the most downstream point in each stream and systematically selected sites beginning at every fourth riffle extending upstream throughout the section (Dolloff et al. 1993). Therefore, sites in LC and SC consisted of all channel units between the bases of two consecutive riffles, and the number of sites per stream depended on the number of riffles within study sections. At each site, we placed transects perpendicular to flow and spaced every 10 meters, beginning within the first three meters of the base of the riffle and extending upstream to the base of the next upstream riffle. At five equidistant points along each transect, the same aforementioned microhabitat variables were recorded. We also snorkeled sites in LC to determine if the microhabitats used there by Candy Darters were similar to those used in EFG and SFC. We used different designs to quantify habitat availability in LC and SC from those used in EFG and SFC for multiple reasons. First, habitat data from LC and SC were not used for habitat suitability models, and therefore, it was not necessary to ensure observations of habitat availability were proportional to the area of each stream (discussed above). Second, while survey extents were similar across all streams (3 – 5.5 km), we distributed effort more evenly (i.e., more sites) in LC and SC to accomplish multiple management objectives (not reported here). Having more systematically spaced sites in LC helped clarify the distribution of Candy Darter within LC, which could potentially be related to site-level habitat attributes including rare, yet critical, habitat patches within LC (Torgersen et al. 1999; Fausch et al. 2002). A similar approach was used in SC to identify specific sites for potential restoration and reintroduction efforts. Because additional management objectives within LC and SC were focused at the site level (i.e., specific riffle-pool sequences), we used a fixed number of points for each transect in LC and SC to provide greater sample sizes and improve estimates of habitat availability at each site.

30 Data Analysis Habitat suitability criteria.— We estimated habitat suitability by first developing habitat selection curves. We use “selection” to refer to disproportionate use relative to availability across a single microhabitat gradient in an uncontrolled environment (i.e., natural stream setting); in contrast, “preference,” refers to disproportionate use in a controlled experimental setting (Rosenfeld 2003). Predictions from habitat selection curves are referred to as habitat suitability criteria (HSC), which were developed for each variable at each life stage within each season (hereafter, single variable suitability unit = HSC). Habitat suitability criteria reflect the ratio of habitat use to availability for habitat bins/categories spanning each gradient (Newcomb et al. 2007:857–872). First, depth and velocity observations were categorized into 0.1-m and 0.2-m/s bins, respectively. To ensure each bin contained at least one observation, all observations > 0.70 cm and > 1.0 m/s were combined into a single bin for depth and velocity, respectively. Bins for substrate, embeddedness, and silt cover respected the original categories described above. Bins for habitat use and availability were subsequently relativized and standardized so values of l and 0 corresponded with the most- and least-selected possible values. Finally, we used generalized additive models with a Gaussian error distribution to regress HSC values against the corresponding midpoint of each bin to aid visual interpretation of habitat selection. However, all estimates of suitability are from the original HSC. Habitat suitability indices.— After developing HSC from habitat use and availability in EFG and SFC, HSC were combined into multi-variable and multi-life stage habitat suitability indices (HSIs; Newcomb et al. 2007). An HSI is a type of habitat suitability model that can be easily deconstructed to investigate the contributions of each life stage and variable to species- level estimates of in-stream suitability. Habitat suitability criteria and HSIs were used to predict the suitability of available habitat within each focal stream. Seasonal habitat suitability for each stream (Equation 1) was the arithmetic mean of HSC for each life stage (life stages l to L where L = 3) based on the five habitat variables (from variables v to V where V = 5) for each habitat observation in a stream (from n to N where N = total number of habitat-availability observations per stream and season). Therefore, stream-wide suitability within each season is the average HSI value of all measured 0.5-m2 microhabitat patches based on habitat selection by multiple life stages. Overall habitat suitability for each stream (HSIstream) is the arithmetic mean of spring and fall HSI values (Equation 2). Finally, because life stage- and variable-specific suitability values

31 are nested within the calculation of seasonal suitability for each stream, we deconstructed stream-level HSI into values for each combination of life stage, season, and habitat variable.

1) σσσே ௅ ௏ (ܪܵܥ) ܪܵܫ = ௡ୀଵ ௟ୀଵ ௩ୀଵ ௡௟௩ ௌ௧௥௘௔௠ିௌ௘௔௦௢௡ ܰ × ܮ × ܸ

2)

ܪܵܫ௦௧௥௘௔௠ିௌ௣௥௜௡௚ + ܪܵܫ௦௧௥௘௔௠ିி௔௟௟ ܪܵܫ = ௦௧௥௘௔௠ 2

Multivariate habitat use and suitability.— We used Nonmetric multidimensional scaling (NMDS) and biplots to visualize how microhabitat use and availability correspond to predicted HSI values. For each season, microhabitat measurements from the four streams were organized into a Euclidian distance matrix, and the multivariate configuration with the lowest stress value after 20 runs was plotted using two axes. Convex-hull polygons were drawn around all observations of habitat availability in each stream. Next, each observation of habitat availability was color coded to reflect its HSI value. We also added NMDS points corresponding to the microhabitats used by Candy Darters in Laurel Creek to examine the consistency of microhabitat use across streams. Finally, highly correlated Pearson correlation coefficients U•__ between axes and in-stream habitat variables were added to biplots. Coefficients are also provided in Supplementary Table S2.2. Cross-scale relationships.— We used a framework that examined the relationships between i) in-stream habitat gradients across streams currently or formerly containing populations ii) predicted individual-level suitability within each stream, and iii) observed population robustness across streams. This framework organized these components into a three- by-three correlation matrix where each component was the heading of a single row and column (Figure 2.2). Analytically, this framework used the regional pattern of decline to generate a gradient of population robustness that could be compared to other columns of the matrix, thereby imposing the context of localization on relationships between individual and population levels. The first relationship was the Spearman rank-order correlation (ߩ) between the mean of each in-stream habitat variable within each stream and a rank corresponding with population

32 robustness within each stream (i.e., columns 1 and 3). These relationships are typically the focus of distributional surveys aimed at observing population-level responses (e.g., site-level occupancy, abundance) across environmental gradients, which typically must sacrifice site-level intensity for larger spatial extents (i.e., more sites). Correlation coefficients could range from 1 to -1 indicating positive and negative population relationships to environmental gradients, respectively. The second relationship was the Pearson correlation (r) between the mean of each in- stream habitat variable within each stream and predicted habitat suitability (i.e., HSC and HSI values) of each stream (i.e., columns 1 and 2). Correlations represent the predicted individual responses to habitat gradients at the stream level and could range from 1 to -1. To trust these correlations is to assume individual-level habitat selection reflects stream-level habitat suitability for populations. This assumption is frequently not validated (Rosenfeld 2003). Finally, the relationship between predicted stream-level suitability based on individual- level HSI values (column 2) and a rank of population robustness (column 3) represented the relationship between individual and population levels. This correlation was the cross-scale relationship (CSR) and would always be positive if individual-level HSIs can be scaled up to reflect population robustness. The CSR was a form of validation in that negative or weak CSRs indicated disconnects between the two ecological levels. East Fork Greenbrier, SFC, LC, and SC were given ranks of 3 (robust), 3 (robust), 2 (localized), and 1 (extirpated), respectively, which were corroborated by observed population densities (Supplementary Table S2.3). The framework was inherently qualitative and designed to facilitate detailed comparisons within and across representative systems. Correlation coefficients provided simple, objective measures of the strength of relationships. Different correlation coefficients were used because estimated suitability and environmental gradients were ratio scale and normally distributed (i.e., appropriate for Pearson’s r), while ranks for population statuses were ordinal and non-parametric (i.e., appropriate for Spearman’s ߩ). We use “CSR” to refer to consistent relationships observed at the microhabitat (individual) and stream segment (population) spatial scales. The concept of spatial scaling is well established in ecology (Wiens 1989; Levin 1992), and has catalyzed the proliferation of multi- scale approaches aimed at identifying relationships among ecological levels of organization and the spatial scales at which habitat is organized (Schneider 2001). Rather than a top-down

33 approach frequently used in habitat suitability investigations, we used a bottom-up approach to examine the ability of microhabitat models to predict the suitability of habitat in stream segments. Figure 2.2 demonstrates important relationships among scales that are often overlooked when scaling up microhabitat suitability models to spatial scales necessary to support populations.

Results Seasonal Habitat Availability across Streams (Columns 1 and 3 in Figure 2.2) Streams with extant populations had similar in-stream habitat. East Fork Greenbrier River, SFC, and LC contained many shallow areas (i.e., riffles and shallow runs), while SC had a meandering, lower-gradient channel with fewer and more isolated riffles composed of gravel, pebble, and cobble (Table 2.1). Embeddedness was consistently lower in streams with robust populations (< 6%) than LC or SC (6–25%). Decreased rain and higher evapotranspiration throughout summer and fall resulted in shallower depths and slower water velocities for all streams in fall. Seasonal differences in habitat availability were most apparent in EFG, where discharge was reduced 92% from spring to fall. Despite being a heavily spring-fed system, reductions in discharge in Sinking Creek (- 79%) were similar to LC (-82%) and greater than SFC (-73%). However, baseflow (i.e., depth and velocity) remained higher in SC, likely due to greater groundwater contributions. Substrate size was the most constant of all variables. There were slightly higher levels of embeddedness and silt for most streams in fall, likely due to deposition of suspended sediment coinciding with reduced stream discharge. Higher fine-sediment levels from spring to fall were most pronounced for SC (embeddedness in spring =1.5, 95% confidence interval [CI] = 0.1; embeddedness in fall =1.8, CI = 0.1; silt in spring = 0.8, CI = 0.1; silt in fall = 1.3, CI = 0.1), which was relatively turbid in spring but clear in fall (C.G.D. personal observation). Relationships between population status and environmental gradients (columns 1 and 3; Figure 2.2) tended to be strong and consistent across seasons (Table 2.2). We interpreted large coefficients (ߩ •_±0.50|) as indicators of strong population relationships with environmental gradients, while consistency in direction of coefficients across seasons indicated few seasonal effects on these relationships. Streams with more robust populations tended to be shallower (ߩ = -0.63), have less embedded and silted substrates (Embeddedness ߩ = -0.95, Silt ߩ = -0.42), and

34 have slower water velocities (ߩ = -0.79). The negative relationship between average stream-level water velocity and population robustness was higher in fall during low-flow conditions (spring = -0.63, fall = -0.95). Finally, there was a positive, albeit weak, correlation between substrate size and population robustness (ߩ = 0.47).

Individual Habitat Selection We recorded 290 (EFG = 115, SFC = 175) and 508 (EFG = 286, SFC = 222) microhabitat-use observations for multiple life stages in spring and fall, respectively. Counts of adults were the most consistent of all life stages across systems (EFG = 135, SFC = 151) and seasons (spring = 137, fall = 149). Counts of sub-adult life stages (i.e., juveniles, age-0) were higher in fall coinciding with new recruitment. We also observed Candy Darters at eight (spring) and 14 (fall) of 20 total sites throughout LC. Adults and juveniles were observed in spring, and all three life stages were detected in fall. Selection curves for all life stages across seasons were either approximately monotonic or unimodal, indicating that observed curves were consistent with the selection of habitat across environmental gradients (Figure 2.3). Clear, biologically sensible selection patterns aid interpretation of habitat associations and obviate the need to rely on p-values from tests of nonrandom habitat selection (Cherry 1998). Generally, most life stages selected microhabitats with at least moderate flow (> 0.19 m/s), shallow depths (< 0.5 m), coarse substrates (> sand), and non-embedded and non-silted substrates (<26%). However, each life stage demonstrated more nuanced habitat selection patterns corresponding with age and body size. The most pronounced ontogenetic differences were for water velocity, with adults selecting the swiftest water velocities available in spring (>1.20 m/s) and fall (> 0.60 m/s). Juveniles selected intermediate water velocities (0.40 m/s–1.20 m/s) in both seasons, while age-0 selected slower water velocities (0.0 m/s–0.80 m/s). Similar ontogenetic patterns occurred for substrate, embeddedness, and silt cover. Adults selected larger substrates and avoided areas with fine sediments, resulting in near-zero HSC values for all microhabitats with embeddedness or silt cover scores > 25% (rank = 2). Younger life stages selected smaller substrates and were less averse to fine sediments. Ontogenetic habitat selection patterns were similar across seasons. The most pronounced difference was that of juveniles, which selected velocities more similar to those of adults in fall

35 than spring (Figure 2.3). When individual variables are collectively viewed, the observed ontogenetic differences resulted from habitat shifts from pool margins and runs as age-0 to swift turbulent riffles as adults. Juveniles tended to select run channel-units or riffle margins in spring and shifted to riffles by fall (i.e., intermediate habitat selection). Our underwater observations enabled us to document the behavior underlying habitat selection patterns (Jordan et al. 2008). Individuals tended to segregate by life stage rather than behavior mode. For example, in spring adults foraged, used cover, and displayed behavior associated with spawning (e.g., antagonistic behavior, courtship) within the most selected habitats. While no habitat-use observations revealed the exact locations selected by females for egg deposition, spawning was observed during surveys and occurred near areas strongly selected by adults in spring. Most Candy Darters inhabiting LC used habitat patches similar to those used in EFG and SFC (Figure 2.4; Supplementary Figure S2.6). Low samples sizes prevented us from developing selection curves by life stage from observations of habitat use and availability within LC; however, nearly all observations of habitat use were consistent with projected highly suitable habitat.

Individual-level Habitat Suitability within and across Streams (Columns 1 and 2 in Figure 2.2) In spring, two distinct groups of suitability values were apparent: streams with robust populations (EFG = 0.68 HSI, CI = 0.01; SFC = 0.66 HSI, CI = 0.02) and streams where Candy Darters are localized (LC = 0.58 HSI, CI = 0.02) or extirpated (SC = 0.56 HSI, CI = 0.01; Supplementary Table S2.4). East Fork Greenbrier River had the highest overall HSI value as a result of having the highest HSC values for depth, velocity, and substrate size. Lower HSI values for LC and SC were the result of low HSC values for embeddedness, and substrate size (Supplementary Table S2.4). Habitat suitability values were lower in fall than spring due to less suitable depths, velocities, and levels of fine sediment. In fall, HSI values also separated into two tiers; however, unlike spring, the highest tier comprised streams with extant populations (EFG = 0.51 HSI, CI = 0.02; SFC = 0.56 HSI, CI = 0.02; LC = 0.52 HSI, CI = 0.02), while SC had markedly lower HSI values (0.44 HSI, CI = 0.01; Supplementary Table S2.4). Sinking Creek remained the least suitable stream due to relatively low HSC values for embeddedness and silt cover.

36 These results were corroborated by NMDS plots of projected habitat suitability for each observation of available microhabitat within the four streams (Figure 2.4 for Spring, Supplementary Figure S2.6 for Fall). In both seasons the most suitable microhabitats occurred in high-velocity areas composed of coarse substrates and few fine sediments. While highly suitable habitat was within the environmental space enveloped by all four streams, SC contained more areas with low suitability representing slower, more embedded, or more silted habitat patches. Predicted habitat suitability across streams mirrored individual-level habitat selection. The strongest correlations between predicted suitability and habitat availability at the stream scale (i.e., r •_0.50|) across seasons were negative relationships with embeddedness and silt cover for all life stages (Table 2.3). In other words, predicted individual-level suitability (i.e., selection) decreased with greater average embeddedness and silt cover across the four streams. Strong positive relationships were also observed across streams for increasing water velocity for adults and juveniles. Predicted suitability for both depth and substrate tended to be either weakly consistent or inconsistent with prediction relations based on historical accounts of habitat selection by Candy Darters (Supplementary Table S2.1). Inconsistencies reflected seasonal

differences in depth selection (r spring= -0.42; r fall = 0.01) and ontogenetic differences between

adults and younger life stages for depth (r adult = 0.21, r juvenile = -0.12, r age-0 = -0.37) and substrate (r adult = 0.21, r juvenile = -0.34, r age-0 = -0.66), which demonstrates temporal or ontogenetic habitat shifts can generate conflicting habitat suitability predictions across a species’ life cycle.

Relationships between Predicted Individual-level Habitat Suitability and Population Robustness across Streams (Columns 2 and 3 in Figure 2.2) Overall, when averaged across two seasons and three life stages, predicted habitat suitability was positively correlated with population robustness (study-wide CSR [ߩ] = 0.95; Figure 2.5), which indicates the proportion of suitable microhabitats within a stream is related to population robustness. However, the strength of these relationships varied with life stage, season, and habitat variable. Habitat suitability indices had higher CSR coefficients in spring (ߩ = 0.95) than fall (ߩ = 0.63) owing to weaker relationships for velocity and substrate size in fall. All life stages had equal CSR coefficients after averaging HSI values across seasons (ߩ = 0.95);

however, coefficients for adults were consistently the highest for both seasons (ߩ spring = 0.96, ߩ

37 fall = 0.96), which may be due to the greater microhabitat specificity of adults. Coefficients for depth and substrate size were the most inconsistent, which indicates these variables may only be important at the population level during certain life stages or seasons. In contrast, velocity consistently had the most negative CSR for all scenarios (ߩ = -1.0), which indicates that despite strong selection of high-velocity habitat, streams with more high-velocity habitat patches did not support more robust populations. Velocity CSR coefficients were more negative in fall, when all streams with extant populations had slower average velocities than SC, where Candy Darters are

extirpated (ߩ spring = -0.39, ߩ fall = -0.50). The CSR correlations for silt cover were season- and life stage-specific, but overall had a positive CSR coefficient (ߩ = 0.63). Finally, embeddedness HSI

values were highly correlated with population robustness regardless of season (ߩ spring = 0.95, ߩ

fall = 0.63) and life stage (ߩ adult = 0.95, ߩ juvenile = 0.95, ߩ age-0= 0.95), which indicates embeddedness is consistently the most important variable for both the selection of microhabitats by individuals and the robustness of populations.

Discussion Much of North America’s imperiled fish fauna has an unnatural distributional pattern marked by disjunct, yet viable populations occupying a fraction of the historical range (Jelks et al. 2008). A key to fish conservation is the identification of habitat promoting resistance or resiliency of these populations to factors that diminish habitat quality. We used a comparative approach aimed at directly contrasting systems currently or formerly supporting populations. Often the processes underlying the localization of populations are anthropogenic; therefore, this comparative approach may reflect the gradients leading to the decline of Candy Darter.

Scaling Up Individual Habitat Selection to Populations Many fishes exhibit complex life cycles marked by the use of distinctive habitat patches through ontogeny, yet much of existing management is based solely on adult habitat use (Copp and Vilizzi 2004; King 2004), which may not be as limiting as habitat for sub-adults. While observations of adult habitat selection by Candy Darters were largely consistent with previous accounts of adults, clear ontogenetic differences were documented, most notably for depth, substrate, and velocity. The most apparent ontogenetic shift by Candy Darters was for velocity, with age-0 selecting low to moderate-velocity areas and adults being largely restricted to high- 38 velocity areas. These observations are consistent with habitat shifts observed in other species of darters (Rosenberger and Angermeier 2003; Skyfield and Grossman 2008; Ashton and Layzer 2010). For Candy Darter, it is unclear if these shifts are structured by risks of predation (Werner and Hall 1988; Schlosser 1987; Labeelund et al. 1993), energetic costs of maintaining position in fast flows (Lobb and Orth 1991; Mann and Bass 1997; Moore and Thorp 2008), differing food sources (Schlosser 1990; King 2004), or intraspecific competition (Davey et al. 2005; Petty and Grossman 2007). In spring, adult males were highly territorial near areas where spawning occurred (i.e., staging areas), which may have excluded juveniles from high-velocity habitats. Our observations of adult male territoriality and lower spatial overlap between adults and juveniles in spring during spawning season tentatively support intraspecific competition as one potential mechanism structuring observed ontogenetic habitat shifts. Microhabitat use is a product of complex interactions among fish size, behavior mode, physiological state, intra- and interspecific interactions, and habitat availability. As a result of this complexity, selection could be a measure of the most suitable habitat available for individuals, but alternative habitat may be substitutable and, therefore, the resource may be less influential at the population level (Rabeni and Sowa 1996). If so, individual selection could mislead managers to incorporate non-essential resources into their guiding image of suitable habitat. For example, adult Candy Darters are flow specialists based on their specificity for high- velocity, shallow microhabitats (Chipps et al. 1994); however, negative water velocity and depth CSR coefficients demonstrate suitable microhabitats for these variables were more or similarly available in Sinking Creek compared to streams with viable populations. Seasonal decreases in suitable velocity and depth microhabitats were greatest in EFG, where discharge was the most reduced from spring to fall. Rather than observing lower abundances of adults as a result of mortality or emigration in fall during low flow conditions, all life stages and especially adults compensated by shifting locations to the most suitable flows available. A hypothesis warranting further testing is that low fall flows enhance survival of age-0 individuals, which frequent shallow, slow microhabitats, as observed in other stream fishes (Schlosser 1982; Rosenberger and Angermeier 2003). Additionally, periods of drought or low flows can disproportionately negatively influence large piscivorous fishes that prey on age-0 fish (Schlosser 1987). Low fall flows may have created nursery habitat unsuitable for predators, in turn allowing for expanded foraging in warmer more productive habitat (Moore and Gregory 1988; Henderson and Johnston

39 2010) and less density-dependence among the large age-0 year classes observed in both streams with robust populations (Schlosser 1990). Similarly, individuals selected certain substrate sizes during specific life stages, but a near-zero CSR coefficient indicates substrate size may not be limiting at the population level within the context of localization. Overall, our results suggest ontogenetic shifts and seasonal habitat plasticity may limit the management utility of a simple generalized image of suitable habitat for a species based solely on the selection of habitat by a single life stage during a single season or, potentially, even a single year. The largest and most consistent CSR coefficients indicated both individual habitat selection and population robustness were negatively related to elevated embeddedness. Embeddedness can profoundly alter the function of stream ecosystems and has been implicated in the declines of most imperiled fishes in North America (Jelks et al. 2008). The specific pathways through which elevated embeddedness may influence individuals and populations of Candy Darter remain unexamined. Potential hypotheses are the filling of interstitial spaces, which can alter food webs by reducing microhabitats used by macroinvertebrates (Ryan 1991; Henley et al. 2000), and eliminate structure used for cover and refugia. Alternatively, observed negative relationships could covary with life-stages not studied herein, such as loss of rearing habitat suitable for eggs and larvae. Future studies specifically aimed at identifying relationships among the characteristics, placement, and abundance of the exact habitat patches needed for egg incubation and larval survival would help determine the role of these life stages in the population dynamics within each stream. Additional research on several topics would help clarify the mechanisms underlying the decline of Candy Darter. For example, documenting fish responses to experimental manipulations of habitat would help reduce any observational biases associated with habitat selection patterns and control for multicollinearity among variables (Rosenfeld 2003). Moreover, this study quantified habitat availability at a spatial scale large enough to be germane to population dynamics of Candy Darter (i.e., stream segment), but additional research spanning relevant temporal scales (i.e., multiple years) could provide a better understanding of the consistency of habitat availability and the stability of predicted suitability under different conditions. However, until specific mechanisms influencing individual fitness and population dynamics are understood, managers could utilize our HSIs, and especially embeddedness selection curves, to help identify sites with suitable habitat for translocation or restoration.

40 Hypothesized Ecological Processes among In-stream Habitat Patches In-stream habitat, as perceived by small aquatic organisms, represents a landscape of microhabitat patches with varying quality (Palmer et al. 2000). For continued occupation within a region, non-substitutable resources must be abundant enough, accessible, and in harmony with the life cycle of a species. On average, habitat patches in SC had the lowest suitability for each of the three life stages investigated. However, both NMDS plots and HSIs indicated that highly suitable habitat patches exist in SC. It is unclear if the prevalence of these patches, particularly unembedded substrate, is too low to support a population. Poor habitat suitability may interact with other population threats and further diminish population resistance to altered conditions. When suitable habitat is proportionally low and spatially diffuse, the suitability of habitat patches is likely reduced by neighborhood effects of surrounding poor-quality habitat (Dunning et al. 1992; Schlosser 1995). Moreover, when navigating corridors between suitable patches, individuals may be exposed to fitness-reducing factors such as elevated risk of predation, which may be exacerbated by a reduction in benthic complexity in embedded systems (Roberts and Angermeier 2007). Many non-native fishes, including piscivores, have colonized SC (Hitt and Roberts 2012) and now occupy likely corridor habitat. While no information exists on the movements of Candy Darter among suitable patches or predation rates by introduced piscivores, Labbe and Fausch (2000) found non-native piscivores influenced demographic rates of the Arkansas Darter E. cragini due to predation in corridor habitat. Understanding interactions between multi-scale habitat suitability and other factors such as predation and movement will require detailed demographic investigation. Nonetheless, findings suggest the prevalence and harshness of the matrix of non-suitable habitat could be as important as the presence of suitable habitat within an area. These findings are consistent with other multi-scale investigations of darter habitat that have found the presence of suitable habitat nested within a matrix of poor- quality habitat may not be enough to sustain populations (Freeman and Freeman 1994; Davis and Cook 2010; Compton and Taylor 2013). For species with uncertain habitat requirements, incorporating multiple scales into investigations may help identify consistencies across populations and individuals (Torgersen et al. 1999; Fausch et al. 2002), and refine hypotheses related to limiting habitat variables.

41 Application to Imperiled Species Management and Recovery Frameworks employing realistic and validated benchmarks are staples of stream restoration and biomonitoring (Stoddard et al. 2006; Whittier et al. 2007) but are less common when defining fish habitat suitability at the site and population levels. We estimated the habitat suitability of streams where populations of Candy Darter are robust (e.g., EFG, SFC) or localized (e.g., LC). Consequently, segment-scale suitability values for EFG and SFC are also ecologically derived benchmarks of optimal in-stream habitat conditions among known populations of Candy Darter, while values for LC may meet the minimum habitat conditions necessary to sustain a population. In contrast, HSI values are often categorized into levels of suitability (e.g., “optimal” KDELWDW•th SHUFHQWLOHRI+6,>7KRPDVDQG%RYHH@•+6,>)UHHPDQHWDO@  which may be meaningful benchmarks for predicting selected habitats by individuals, but may not have significance for populations. Attempts at validating individual habitat selection often examine the consistency of habitat selection across systems (Newcomb et al. 2007). However, even if habitat use is consistent, the approach still does not establish relationships between individual habitat selection and population function. Alternatively, the approach used here effectively re-scaled suitability based on individual habitat selection to represent segment-level suitability for a population, which will likely be more meaningful to conservation efficacy. Validating individual habitat selection at the stream scale may be particularly applicable for imperiled species management. For example, recovery plans often aim to reestablish extirpated populations (George et al. 2009), yet historical conditions within streams are rarely documented. Managers could reference suitability values from streams with robust populations when identifying streams with suitable habitat for reintroduction. Alternatively, potentially more realistic criteria may be suitability values from streams supporting small populations (e.g., LC) given that additional, albeit small, populations can dilute the risk of regional extirpation.

Conclusions Habitat suitability indices for Candy Darter should not be considered infallible or definitive. A correlative framework is no substitute for detailed study of mechanisms influencing individuals or populations. However, more direct measures of individual fitness (e.g., growth, fecundity) or population function (e.g., demographic rates) across more streams could be incorporated within the general framework described herein (Figure 2.2). A clear understanding

42 of habitat requirements is typically gained through a progression of detailed investigation at multiple ecological levels (Rosenfeld 2003). Our approach is likely helpful at identifying limiting habitat types at the beginning of this progression, which may help direct future investigations and conservation measures. Candy Darters are highly selective of specific in-stream habitat patches within occupied streams, and habitat selection varies through ontogeny. Habitat specificity may reflect adaptive benefits of certain patch types for growth, survival, and reproduction under natural conditions. However, when viewed across populations affected by anthropogenic disturbances, habitat specificity did not always indicate limiting conditions. Similar to many non-game species, before this investigation the only habitat information available for Candy Darters were descriptions of adult habitat use within short reaches (< 150 m) in a few streams during a single season. While our findings support some of these descriptions, ontogenetic shifts and seasonal habitat plasticity make habitat selection more complex than previously described. This complexity demonstrates that study aimed at the individual-level could be potentially misleading when identifying suitable habitat to facilitate population persistence. This finding underscores the potential inadequacy of the information guiding management decisions for many of North America’s freshwater fishes. Until rigorous study of relationships between individuals and populations becomes the norm for species with lower management priority, the framework used in this study may be a viable approach to identify habitat variables important at both levels of ecological organization.

Acknowledgments This study was partially funded by a State Wildlife Grant from the U.S. Fish and Wildlife Service and the Virginia Department of Game and Inland Fisheries (VDGIF). We thank Mike Pinder, Stuart Welsh, and Dan Cincotta for site recommendations. We thank VDGIF and the West Virginia Division of Natural Resources for field-collection permits and Greg Anderson for programming assistance. Field help was provided by Matt Bierlein, Joe Cline, David Crain, Laura Heironimus, Pat Kroboth, Josh Light, Vance Nepomuceno, Phil Pegelow, Jordan Richard, Chris Rowe, and Laura Zseleczky. Earlier versions of the paper were improved by recommendations from Nick Sievert and two anonymous reviewers. This work was carried out under the auspices of Institutional Care and Use Committee protocols 10-094-FIW at Virginia Tech. The Virginia Cooperative Fish and Wildlife Research Unit is jointly sponsored by

43 the U.S. Geological Survey, Virginia Polytechnic Institute and State University, Virginia Department of Game and Inland Fisheries, and Wildlife Management Institute. Use of trade, firm, or product names does not imply endorsement by the U.S. government.

References Addair, J. 1944. The fishes of the Kanawha River system in West Virginia and some factors which influence their distribution. Ph.D. dissertation. Ohio State University, Columbus, Ohio. Angermeier, P. L. 1995. Ecological attributes of extinction-prone species: loss of freshwater fishes of Virginia. Conservation Biology 9(1):143–158.

Ashton, M. J., and J. B. Layzer. 2010. Summer microhabitat use by adult and young-of-year snail darters (Percina tanasi) in two rivers. Ecology of Freshwater Fish 19(4):609–617.

Boavida, I., J. M. Santos, R. Cortes, A. Pinheiro, and M. T. Ferreira. 2012. Benchmarking river habitat improvement. River Research and Applications 28(10):1768–1779.

Burkhead, N. M. 2012. Extinction rates in North American freshwater fishes, 1900–2010. Bioscience 62(9):798-808.

Burton, G. W., and E. P. Odum. 1945. The distribution of stream fish in the vicinity of Mountain Lake, Virginia. Ecology 26(2):182–194.

Cherry, S. 1998. Statistical tests in publications of The Wildlife Society. Wildlife Society Bulletin:947–953.

Chipps, S. R., W. B. Perry, and S. A. Perry. 1993. Status and distribution of Phenacobius teretulus, Etheostoma osburni, and “Rhinichthys bowersi” in the Monongahela National Forest, West Virginia. Virginia Journal of Science 44(1):48–58.

44 Chipps, S. R., W. B. Perry, and S. A. Perry. 1994. Patterns of microhabitat use among four species of darters in three Appalachian streams. American Midland Naturalist 131(1):175–180.

Compton, M., and C. Taylor. 2013. Spatial scale effects on habitat associations of the Ashy Darter, Etheostoma cinereum, an imperiled fish in the southeast United States. Ecology of Freshwater Fish 22(2):178–191.

Copp, G. H., and L. Vilizzi. 2004. Spatial and ontogenetic variability in the microhabitat use of stream-dwelling spined loach (Cobitis taenia) and stone loach (Barbatula barbatula). Journal of Applied 20(6):440–451.

Davey, A. J. H., S. J. Hawkins, G. F. Turner, and C. P. Doncaster. 2005. Size-dependent microhabitat use and intraspecific competition in Cottus gobio. Journal of Fish Biology 67(2):428–443.

Davis, J. G., and S. B. Cook. 2010. Habitat use of the Tuxedo Darter (Etheostoma lemniscatum) at macrohabitat and microhabitat spatial sclaes. Journal of Freshwater Ecology 25(3):321–330.

Dixon, C. J., and J. C. Vokoun. 2009. Burbot resource selection in small streams near the southern extent of the species range. Ecology of Freshwater Fish 18(2):234–246.

Dolloff, C. A., D. G. Hankin, and G. H. Reeves. 1993. Basinwide estimation of habitat and fish populations in streams. US Forest Service, Report SE-93, Asheville, North Carolina.

Dunham, J. B., B. S. Cade, and J. W. Terrell. 2002. Influences of spatial and temporal variation on fish-habitat relationships defined by regression quantiles. American Fisheries Society. Transactions 131(1):86–98.

Dunning, J. B., B. J. Danielson, and H. R. Pulliam. 1992. Ecological processes that affect populations in complex landscapes. Oikos 65(1):169–175.

45 Fausch, K. D., C. L. Hawkes, and M. G. Parsons. 1988. Models that predict standing crop of stream fish from habitat variables: 1950–85. U.S. Forest Service, Report PNW-GTR-213, Portland, Oregon.

Fausch, K. D., C. E. Torgersen, C. V. Baxter, and H. W. Li. 2002. Landscapes to riverscapes: bridging the gap between research and conservation of stream fishes. Bioscience 52(6):483–498.

Freeman, M. C., Z. H. Bowen, and J. H. Crance. 1997. Transferability of habitat suitability criteria for fishes in warmwater streams. North American Journal of Fisheries Management 17(1):20–31.

Freeman, B., and M. Freeman. 1994. Habitat use by an endangered riverine fish and implications for species protection. Ecology of Freshwater Fish 3(2):49–58.

Frissell, C. A., W. J. Liss, C. E. Warren, and M. D. Hurley. 1986. A hierarchical framework for stream habitat classification: viewing streams in a watershed context. Environmental Management 10(2):199–214.

Gabelhouse, D. W. 2005. Staffing, spending, and funding of state inland fisheries programs. Fisheries 30(2):10–17.

George, A. L., B. R. Kuhada, J. D. Williams, M. A. Cantrell, P. L. Rakes, and J. R. Shute. 2009. Guidelines for propagation and translocation for freshwater fish conservation. Fisheries 34(11):529–545.

Grimm, V. and Railsback, S.F. 2005. Individual-based modeling and ecology. Princeton University Press. Princeton, New Jersey.

Grossman, G. 2013. Not all drift feeders are trout: a short review of fitness-based habitat selection models for fishes. Environmental Biology of Fishes:1–9.

46 Guay, J. C., D. Boisclair, D. Rioux, M. Leclerc, M. Lapointe, and P. Legendre. 2000. Development and validation of numerical habitat models for juveniles of Atlantic Salmon (Salmo salar). Canadian Journal of Fisheries and Aquatic Sciences 57(10):2065–2075.

Haxton, T. J., C. S. Findlay, and R. W. Threader. 2008. Predictive value of a Lake Sturgeon habitat suitability model. North American Journal of Fisheries Management 28(5):1373– 1383.

Henderson, A. R., and C. E. Johnston. 2010. Ontogenetic habitat shifts and habitat use in an endangered , Notropis mekistocholas. Ecology of Freshwater Fish 19(1):87–95.

Henley, W., M. Patterson, R. Neves, and A. D. Lemly. 2000. Effects of sedimentation and turbidity on lotic food webs: a concise review for natural resource managers. Reviews in Fisheries Science 8(2):125–139.

Hewitt, A. H., T. J. Kwak, W. G. Cope, and K. H. Pollock. 2009. Population density and instream habitat suitability of the endangered Cape Fear Shiner. Transactions of the American Fisheries Society 138(6):1439–1457.

Hitt, N.P. and Roberts, J.H. 2012. Hierarchical spatial structure of stream fish colonization and extinction. Oikos 121:127–137.

Jelks, H. L., S. J. Walsh, N. M. Burkhead, S. Contreras-Balderas, E. Diaz-Pardo, D. A. Hendrickson, J. Lyons, N. E. Mandrak, F. McCormick, J. S. Nelson, S. P. Platania, B. A. Porter, C. B Renaud, J. J. Schmitter-Soto, E. B. Taylor, and M. L. Warren, Jr. 2008. Conservation status of imperiled North American freshwater and diadromous fishes. Fisheries 33(8):372–407.

Jenkins, R. E., and N. M. Burkhead. 1994. Freshwater fishes of Virginia. American Fisheries Society Press, Bethesda, Maryland.

47 Jenkins, R. E., and B. L. Kopia. 1995. Population status of the Candy Darter, Etheostoma osburni, in Virginia 1994–95, with historical review. Department of Biology, Roanoke College, Final Report, Salem, Virginia.

Jordan, F., H. L. Jelks, S. A. Bortone, and R. M. Dorazio. 2008. Comparison of visual survey and seining methods for estimating abundance of an endangered, benthic stream fish. Environmental Biology of Fishes 81(3):313–319.

King, A. J. 2004. Ontogenetic patterns of habitat use by fishes within the main channel of an Australian floodplain river. Journal of Fish Biology 65(6):1582–1603.

Kuehne, R. A., and R. W. Barbour. 1983. The American darters. University Press of , Lexington, Kentucky.

Kwak, T. J., M. J. Wiley, L. L. Osborne, and R. W. Larimore. 1992. Application of diel feeding chronology to habitat suitability analysis of warmwater stream fishes. Canadian Journal of Fisheries and Aquatic Sciences 49(7):1417–1430.

Labbe, T. R., and K. D. Fausch. 2000. Dynamics of intermittent stream habitat regulate persistence of a threatened fish at multiple scales. Ecological Applications 10(6):1774– 1791.

Labeelund, J. H., A. Langeland, B. Jonsson, and O. Ugedal. 1993. Spatial segregation by age and size in Arctic Charr: a trade-off between feeding possibility and risk of predation. Journal of Animal Ecology 62(1):160–168.

Leftwich, K. N., P. L. Angermeier, and C. A. Dolloff. 1997. Factors influencing behavior and transferability of habitat models for a benthic stream fish. Transactions of the American Fisheries Society 126(5):725–734.

Leftwich, K. N., C. A. Dolloff, M. K. Underwood, and M. Hudy. 1996. The Candy Darter (Etheostoma osburni) in Stony Creek, George Washington – Jefferson National Forest,

48 Virginia: trout predation, distribution, and habitat associations. U.S. Forest Service, Final Report, Blacksburg, Virginia.

Levin, S. A. 1992. The problem of pattern and scale in ecology. Ecology 73(6):1943–1967.

Lobb, M. D., and D. J. Orth. 1991. Habitat use by an assemblage of fish in a large warmwater stream. Transactions of the American Fisheries Society 120(1):65–78.

Loomis, J. B., and D. S. White. 1996. Economic values of increasingly rare and endangered fish. Fisheries 21(11):6–10.

Mann, R. H. K., and J. A. B. Bass. 1997. The critical water velocities of larval roach (Rutilus rutilus) and dace (Leuciscus leuciscus) and implications for river management. Regulated Rivers-Research & Management 13(3):295–301.

Mattingly, H. T., and D. L. Galat. 2002. Distributional patterns of the threatened Niangua darter, Etheostoma nianguae, at three spatial scales, with implications for species conservation. Copeia (3):573–585.

Messinger, T., and C. Hughes. 2000. Environmental setting and its relations to water quality in the Kanawha River Basin. US Geological Survey, Water-Resoures Investigations Report 00-4020.

Midway, S. R., T. J. Kwak, and D. D. Aday. 2010. Habitat suitability of the Carolina , an imperiled, endemic stream fish. Transactions of the American Fisheries Society 139(2):325–338.

Moore, K. M. S., and S. V. Gregory. 1988. Summer habitat utilization and ecology of Cutthroat Trout fry (Salmo clarki) in Cascade Mountain streams. Canadian Journal of Fisheries and Aquatic Sciences 45(11):1921–1930.

49 Moore, S. L., and J. H. Thorp. 2008. Coping with hydrogeomorphic variations in a prairie river: resiliency in young-of-the-year fishes. River Research and Applications 24(3):267–278.

Newcomb, T. J., D. J. Orth, and D. F. Stauffer. 2007. Habitat evaluation. Pages 843–886 in C. S. Guy, and M. L. Brown, editors. Analysis and interpretation of freshwater fisheries data. American Fisheries Society, Bethesda, Maryland.

Orth, D. J. 1987. Ecological considerations in the development and application of instream flow habitat models. Regulated rivers: research & management 1(2):171–181.

Palmer, M. A., C. M. Swan, K. Nelson, P. Silver, and R. Alvestad. 2000. Streambed landscapes: evidence that stream respond to the type and spatial arrangement of patches. Landscape Ecology 15(6):563–576.

Peckarsky, B. L., S. D. Cooper, and A. R. McIntosh. 1997. Extrapolating from individual behavior to populations and communities in streams. Journal of the North American Benthological Society 16(2):375–390.

Petty, J. T., and G. D. Grossman. 2007. Size-dependent territoriality of Mottled Sculpin in a southern Appalachian stream. Transactions of the American Fisheries Society 136(6):1750–1761.

Rabeni, C. F., and S. P. Sowa. 1996. Integrating biological realism into habitat restoration and conservation strategies for small streams. Canadian Journal of Fisheries and Aquatic Sciences 53:252–259.

Roberts, J. H., and P. L. Angermeier. 2007. Movement responses of stream fishes to introduced corridors of complex cover. Transactions of the American Fisheries Society 136(4):971– 978.

Rosenberger, A., and P. L. Angermeier. 2003. Ontogenetic shifts in habitat use by the endangered Roanoke logperch (Percina rex). Freshwater Biology 48(9):1563–1577.

50 Rosenfeld, J. 2003. Assessing the habitat requirements of stream fishes: An overview and evaluation of different approaches. Transactions of the American Fisheries Society 132(5):953–968.

Ryan, P. A. 1991. Environmental effects of sediment on New Zealand streams – a review. New Zealand Journal of Marine and Freshwater Research 25(2):207–221.

Schlosser, I. J. 1982. Fish community structure and function along two habitat gradients in a headwater stream. Ecological Monographs 52(4):395–414.

Schlosser, I. J. 1985. Flow regime, juvenile abundance, and the assemblage structure of stream fishes. Ecology 66(5):1484–1490.

Schlosser, I. J. 1987. The role of predation in age-related and size-related habitat use by stream fishes. Ecology 68(3):651–659.

Schlosser, I. J. 1990. Environmental variation, life history attributes, and community structure in stream fishes: implications for environmental management and assessment. Environmental Management 14(5):621–628.

Schlosser, I. J. 1995. Critical landscape attributes that influence fish population dynamics in headwater streams. Hydrobiologia 303:71–85.

Schlosser, I. J. 1998. Fish recruitment, dispersal, and trophic interactions in a heterogeneous lotic environment. Oecologia 113(2):260–268.

Schlosser, I. J., and P. L. Angermeier. 1995. Spatial variation in demographic processes of lotic fishes: Conceptual models, empirical evidence, and implications for conservation. Pages 392–401 in J. L. Nielsen, editor. Evolution and the aquatic ecosystem: defining unique units in population conservation. American Fisheries Society, Symposium 17, Bethesda, Maryland.

51 Schneider, D. C. 2001. The rise of the concept of scale in ecology. Bioscience 51(7):545–553.

Skyfield, J. P., and G. D. Grossman. 2008. Microhabitat use, movements and abundance of Gilt Darters (Percina evides) in southern Appalachian (USA) streams. Ecology of Freshwater Fish 17(2):219–230.

Stoddard, J. L., D. P. Larsen, C. P. Hawkins, R. K. Johnson, and R. H. Norris. 2006. Setting expectations for the ecological condition of streams: the concept of reference condition. Ecological Applications 16(4):1267–1276.

Thomas, J. A., and K. D. Bovee. 1993. Application and testing of a procedure to evaluate transferability of habitat suitability criteria. Regulated Rivers Research & Management 8(3):285–294.

Torgersen, C. E., D. M. Price, H. W. Li, and B. A. McIntosh. 1999. Multiscale thermal refugia and stream habitat associations of chinook salmon in northeastern Oregon. Ecological Applications 9(1):301–319.

Wiens, J. A. 1989. Spatial scaling in ecology. Functional ecology 3(4):385-397.

Werner, E. E., and J. F. Gilliam. 1984. Ontogenetic niche and species interactions in size- structured populations. Annual Review of Ecology and Systematics 15:393–425.

Werner, E. E., and D. J. Hall. 1988. Ontogenetic habitat shifts in Bluegill: the foraging rate- predation risk trade-off. Ecology 69(5):1352–1366.

White, R. J. 1996. Growth and development of North American stream habitat management for fish. Canadian Journal of Fisheries and Aquatic Sciences 53:342–363.

Whittier, T. R., J. L. Stoddard, D. P. Larsen, and A. T. Herlihy. 2007. Selecting reference sites for stream biological assessments: best professional judgment or objective criteria. Journal of the North American Benthological Society 26(2):349–360.

52 TABLE 2.1. Means, ± 95% confidence intervals (in parentheses), and counts of observations of habitat availability in four streams and two seasons. Stream abbreviations: EFG = East Fork Greenbrier River, LC = Laurel Creek, SC = Sinking Creek, SFC = South Fork Cherry River. Substrate categories are 1 = Silt, 2 = Sand, 3 = Gravel, 4 = Pebble, 5 = Small Cobble, 6 = Large Cobble, 7 = Small Boulder, 8 = Large Boulder, 9 = Bedrock. Embeddedness and silt categories DUH ” 5%, 1 = 6–25%, 2 = 26–50%, 3 = 51–75%, 4 = > 75%.

Population Depth Velocity Substrate Embeddedness Silt cover Season Stream Status N (cm) (m/s) (rank) (rank) (rank) Spring EFG Robust 620 28.8 (1.5) 0.41 (0.02) 5.3 (0.1) 0.3 (0.0) 0.8 (0.1) SFC Robust 693 25.9 (1.3) 0.28 (0.02) 5.5 (0.1) 0.3 (0.1) 0.6 (0.1) LC Localized 435 26.2 (1.5) 0.35 (0.02) 5.6 (0.2) 1.1 (0.1) 1.0 (0.1) SC Extirpated 490 47.5 (2.3) 0.43 (0.02) 4.7 (0.1) 1.5 (0.1) 0.8 (0.1) Fall EFG Robust 212 16.2 (2.2) 0.11 (0.02) 5.1 (0.2) 0.4 (0.1) 1.1 (0.1) SFC Robust 277 19.1 (1.7) 0.15 (0.02) 5.3 (0.2) 0.5 (0.1) 0.8 (0.1) LC Localized 440 17.6 (1.2) 0.15 (0.02) 5.1 (0.2) 1.0 (0.1) 0.9 (0.1) SC Extirpated 515 32.5 (1.6) 0.19 (0.02) 4.5 (0.1) 1.8 (0.1) 1.3 (0.1)

53 TABLE 2.2. Spearman rank-order correlations (ߩ) between habitat variables and a rank representing the population status of Candy Darters in four study streams and two seasons (columns 1 and 3; Figure 2.2). Streams with the highest densities were given the highest rank: East Fork Greenbrier River = 3 (Robust), South Fork Cherry River = 3 (Robust), Laurel Creek = 2 (localized), Sinking Creek = 1 (extirpated). Predictions are based on historical accounts of habitat use (Supplementary Table S2.1). The embeddedness predicted relation was not DSSOLFDEOH 1$ GXHWRQRKLVWRULFDODFFRXQWV&RHIILFLHQWV•__DQGFRQVLVWHQWZLWK predictions are bolded to emphasize relationship strength. &RHIILFLHQWV•__and inconsistent with predictions are italicized. “Combined” are averages from spring and fall.

Predicted Variable relation Spring Fall Combined Depth - -0.63 -0.63 -0.63 Velocity + -0.63 -0.95 -0.79 Substrate + 0.32 0.63 0.47 Embeddedness NA -0.95 -0.95 -0.95 Silt cover - -0.21 -0.63 -0.42

54 TABLE 2.3. Pearson correlation coefficients (r) between predicted individual habitat suitability and averages of five in-stream habitat variables across four streams that vary in population status (columns 1 and 2; Figure 2.2). Predictions are based on prior accounts of habitat use (Supplementary Table S2.1). The embeddedness predicted relation was not applicable (NA) due to no historical accounts. “Multi-stage” is the correlation between average suitability across multiple life stages (Supplementary Table S2.4) and habitat gradients (Table 2.1). “Combined” DUHDYHUDJHVIURPVSULQJDQGIDOOFRHIILFLHQWV&RHIILFLHQWV•__DQGFRQVLVWHQWZLWK SUHGLFWLRQVDUHEROGHGWRHPSKDVL]HVWUHQJWKRIUHODWLRQVKLS&RHIILFLHQWV•__DQG inconsistent with predictions are italicized.

Predicted Variable Season relation Adults Juveniles Age-0 Multi-stage Depth Spring - -0.10 0.21 -0.74 -0.42 Velocity Spring + 0.97 0.99 -0.19 0.99 Substrate Spring + 0.15 0.25 -0.54 0.06 Embeddedness Spring NA -1.00 -1.00 -1.00 -1.00 Silt cover Spring - -0.36 -0.92 -0.69 -0.98 Depth Fall - 0.52 -0.45 -0.01 0.01 Velocity Fall + 0.93 0.78 0.74 0.77 Substrate Fall + 0.28 -0.93 -0.78 -0.71 Embeddedness Fall NA -1.00 -1.00 -1.00 -1.00 Silt cover Fall - -0.91 -0.90 -0.99 -0.93 Depth Combined - 0.21 -0.12 -0.37 -0.20 Velocity Combined + 0.96 0.89 0.28 0.88 Substrate Combined + 0.21 -0.34 -0.66 -0.32 Embeddedness Combined NA -1.00 -1.00 -1.00 -1.00 Silt cover Combined - -0.63 -0.91 -0.84 -0.96

55 FIGURE 2.1. Map of the New River drainage and study sites. (A) = South Fork Cherry River, WV (robust population); (B) = East Fork Greenbrier River, WV (robust population); (C) = Laurel Creek, VA (localized population); (D) = Sinking Creek, VA (extirpated population). Insets depict survey designs used to develop habitat suitability indices within streams supporting robust populations (South Fork Cherry River) and to systematically measure habitat availability in streams with localized (Laurel Creek) or extirpated.

56 Stream- Predicted Population habitat individual robustness gradient suitability

Stream- r = Predicted ߩ = Observed habitat r = 1.0 individual population gradient response relationship

Predicted ߩ = Cross- individual r = 1.0 scale suitability relationship

Population ߩ = 1.0 robustness

FIGURE 2.2. Framework for examining relationships among stream-habitat gradients, predicted individual habitat suitability, and observed population robustness across streams. The relationship between columns 1 and 3 is the observed population relationship to a habitat gradient. The relationship between columns 1 and 2 is the predicted response of individuals to a habitat gradient across streams. The relationship between columns 2 and 3 is the cross-scale relationship between predicted individual suitability and observed population robustness.

57 FIGURE 2.3. Habitat selection curves developed from habitat used by Candy Darters during three life stages and two seasons and available habitat in two streams. Continuous curves, presented as visual aids, were obtained by regressing suitability values against the midpoint of each bin using generalized additive regression models. Substrate abbreviations: Grav. = Gravel, Peb. = Pebble, Sm. Cob. = Small Cobble, Lg. Cob. = Large Cobble, Sm. Bldr. = Small Boulder, Lg. Bldr. = Large Boulder, BR = Bed Rock.

58 FIGURE 2.4. Non-metric multidimensional scaling (NMDS) plots of habitat use, availability, and suitability in spring. A) Habitat use by three life stages and availability in four streams (polygons). Predicted suitability by (B) Adults, (C) Juveniles, and (D) Age-0. Symbols for "LC use" are locations used by Candy Darters in Laurel Creek in spring. Variables that are highly FRUUHODWHG 3HDUVRQFRHIILFLHQW>U@• ZLWKD[HVDUHVKRZQ$OOFRUUHODWLRQFRHIILFLHQts are in Supplementary Table S2.2. Stream abbreviations: EFG = East Fork Greenbrier River, LC = Laurel Creek, SC = Sinking Creek, SFC = South Fork Cherry River. NMDS stress = 0.17.

59 FIGURE 2.5. Spearman rank-order correlation coefficients (ߩ) between predicted individual suitability and population health in two seasons and across seasons (columns 1 and 3; Figure 2). Positive coefficients (blue) indicate individual preferences are predictive of population-level health, while negative coefficients (red) are disconnects between individual and population levels. The lower-far right cell (ߩ = 0.95) is the study-wide CSL coefficient. “Multi-stage” is a multiple life stage habitat suitability index. “MV.index” indicates a multiple-variable habitat suitability index. Vel. = Velocity, Sub. = Substrate Size, Emb. = Embeddedness.

60 CHAPTER 3: Pathway to Imperilment: Extirpation of a Highland Fish Explained by Fine- Sediment, Stream Temperature, and Landscape Context

Keywords: Agricultural Impact, Candy Darter, Ecological Resilience, Land-Cover Cascade, Refugium

Abstract 1. Conservation initiatives are increasingly devising strategies to increase the resilience of sensitive populations and systems to land-use and climate change. Examining range dynamics of stream biota to past land-use change may provide foresight when developing resilience-building strategies. 2. Our goal was to contrast the environmental conditions underlying the persistence versus extirpation of an endemic stream fish, candy darter (Etheostoma osburni), over multiple decades in the eastern United States. The decline of E. osburni may broadly represent conservation challenges facing highland stream biota and systems elsewhere due to the species’ apparent sensitivity to fine sediment and warm temperatures. We conducted a range-wide fish and habitat survey of 42 stream segments, including 32 segments with historically confirmed occurrences. First, we developed a multi-level resilience framework and used multivariate analyses to investigate whether natural catchment- and segment-scale features mediated the pathway that catchment land-use influenced instream habitat. 3. Next, we used generalized linear models and distance-based eigenvector mapping to examine whether fine sediment, stream temperature, and the network configuration among sites explained persistence in stream segments. 4. Etheostoma osburni were extirpated throughout much of their southern geographic range characterized by high levels of agriculture. Populations that persisted were spatially organized within the stream network, and centered in regions with stream segments providing cool mean daily temperatures during spring and low substrate embeddedness in riffles. Instream habitat and segment-scale features were primarily constrained by catchment features, supporting a hypothesized pathway of population extirpation via a land-cover disturbance cascade or persistence within catchment-generated refugia. Persistence of a narrowly distributed population within a highly agricultural catchment provided limited support for segment-scale resilience to intensive catchment land-uses.

61 5. Our results demonstrate that natural landscape heterogeneity imparts spatially variable resilience of imperiled species to land-use disturbance. By recognizing resilient landscape contexts that buffer species from extirpation, conservation practitioners may strategically aim preservation and restoration actions at populations and systems that lack such buffering features.

Introduction Understanding the factors creating and maintaining stream habitat is crucial for biomonitoring, managing populations, and stream restoration. Stream habitat is hierarchically organized from catchment to microhabitat spatial scales (Frissell et al., 1986). Within this catchment hierarchy, physical heterogeneity is created by interactions among regional factors such as geology, climate and land use, and smaller-scale factors including riparian vegetation, geomorphology and stream network geometry (Poole, 2002). This framework gives rise to sets of filtering mechanisms that collectively determine local biotic composition (Tonn et al., 1990; Poff, 1997). A way to gain insight into the dynamics at any single scale is to examine the context imposed, and uniqueness provided, at higher and lower spatial scales, respectively (Poole, 2002; Wu, 2013). Many factors shaping stream habitat are strongly affected by land use (Allan, 2004). Widespread conversion of natural land cover to intensive land uses, particularly agriculture, can decrease instream structural heterogeneity (Burcher, Valett & Benfield, 2007), and increase fine sediment (Sutherland, Meyer & Gardiner, 2002; Scott et al., 2002) and stream temperature (Caissie, 2006). Land-use disturbance stimuli are often propagated at intermediate spatial scales via “land-cover cascades” (Burcher, Valett & Benfield 2007) until eventually impacting stream habitat and biota (Isaak & Hubert, 2001; Maloney & Weller, 2011; Burdon, McIntosh & Harding, 2013). Alternatively, intermediate-scale features, including channel geomorphology (Walters, Leigh & Bearden, 2003), local groundwater inputs (Torgersen et al., 1999), and intact riparian areas (Sponseller, Benfield & Valett, 2001; Frimpong et al., 2005), can disrupt top-down cascades, leading to persistence of sensitive biota in refugia within otherwise regionally unsuitable conditions. Temperature and fine sediment are important local habitat variables affecting stream biota. Both variables generally increase following land-use disturbance (Scott, 2002; Allan, 2004), which can lead to marked changes in stream communities including the replacement of temperature- and sediment-sensitive biota with generalist or cosmopolitan species (Sponseller,

62 Benfield & Valett, 2001; Walters et al., 2003; Scott, 2006). Moreover, imperilment risk for habitat specialists is often compounded when species have restricted geographic ranges (Pritt & Frimpong, 2010). For example, 38% of imperiled North American freshwater fishes are both narrowly distributed and threatened by habitat degradation, with elevated fine-sediment and stream temperatures being primary stressors (Jelks et al., 2008). Despite likely fine-sediment and thermal sensitivity of most freshwater fishes, especially those occupying highland streams, mechanistic knowledge of individual or additive effects of these two stressors is limited to only a small subset of fishes (Kemp et al., 2011; Comte et al., 2013; Lynch et al., 2016). A more comprehensive understanding of specific effects of land use on stream habitat will be especially valuable for conserving stream biodiversity under projected human population growth and global climate-change scenarios (Staudt et al., 2013). Augmenting ecological resilience is a proactive approach for guarding against the uncertain effects of future land-use and climate change (Wilby et al., 2010; Williams et al., 2015). Ecological resilience frameworks generally assume ecosystems exist in self-organized nominal states, whereby systems or system components (e.g., biota, habitat) can withstand disturbance until reaching thresholds resulting in rapid transformation to alternative ecological states (Suding & Hobbs, 2009). Potential irreversibility provides incentive to proactively build system resilience. It may be untenable to uniformly implement resilience strategies across an entire system (e.g., reforesting entire catchments); therefore, critical system components providing resilience to specific stressors, such as elevated temperature or sediment loading, may be selectively targeted to build system-wide general resilience (Folke et al., 2010). Knowledge of how systems respond to historical disturbance, including species’ range dynamics, may inform which processes and structures can be restored, augmented, or preserved to enhance resilience. Stream network configuration also likely mediates ecological resilience. Immigration facilitated by network connectivity can stabilize the dynamics of populations occupying sub- optimal habitats or increase the probability of recolonization following disturbance (Angermeier, Krueger & Dolloff, 2002; Stoll et al., 2014). Similarly, network configuration may influence future adaptability under changing bioclimatic conditions by mediating dispersal through stream corridors (Comte & Grenouillet, 2015). Alternatively, correlative spatial patterns among populations and environmental gradients may confound interpretation of important habitats influencing resilience by overemphasizing weak or spurious species-habitat relationships (Peres-

63 Neto & Legendre 2010). Regardless of specific causes of observed population spatial patterns, stream network configuration likely plays an important role in species persistence. Our goal was to contrast environmental conditions underlying the population persistence or extirpation of a highland stream fish, candy darter Etheostoma osburni, over multiple decades. The complex topography and land-use history of the study region, combined with potential sensitivity of E. osburni to temperature and fine-sediment, make this context ideal for investigating outcomes of land-use disturbance on species persistence. We developed a multi- level resilience framework to ask, 1) is the influence of land-use disturbance on stream habitat constrained by catchment- and segment-scale features? 2) At a local scale, is the persistence of E. osburni explained by variation in fine sediment, stream temperature, and network configuration among sites?

Methods Multi-level Resilience Framework We develop a framework of hypothetical pathways through which natural sources of resilience to regional land-use disturbance could promote local species persistence (Figure 3.1). Here, ecological resilience includes both resistance and resilience to disturbance, which may foreshadow the adaptability of a system to future regional disturbances (resilience = large horizontal axis in Figure 3.1a). First at a catchment scale, specific natural features (“catchment environmental”), including geology, basin size, and elevation mediate the severity of land-use disturbance (small horizontal axes within each feature in Figure 3.1a). Catchment land-use disturbance can be propagated at intermediate spatial scales by sensitive segment-scale features (“segment environmental”) nested within sensitive catchment features, leading to degraded instream habitat and extirpation of a species (“land-cover cascade” pathway sensu Burcher, Vallett & Benfield 2007; Figure 3.1a). Similarly, catchments may contain both resilient catchment- and segment-scale features creating catchment-generated refugia where populations persist within marginal regions (resilient-catchment pathway; Figure 3.1b). Both figures 3.1a and 1b represent propagating pathways whereby finer-scale features are constrained by processes and structures at catchment scales. Alternatively, resilience may be uniquely conferred at a segment- scale where near-stream features including riparian areas and channel geomorphology mitigate catchment-generated disturbance, creating local refugia within degraded catchments (resilient-

64 segment pathway; Figure 3.1c). Support for this latter pathway may be particularly important for conservation since management at a local scale is often more tractable than at a catchment scale (Folke et al., 2010). Biotic factors can also affect resilience at multiple spatial scales. At a catchment scale, meta-population connectivity can facilitate species persistence via immigration into areas with marginal catchment- or segment-scale features (Figure 3.1d), or inhibit recolonizing suitable sites following disturbance (Figure 3.1e). Finally, a species’ stenothermic or lithophilic traits, along with limited mobility, can also reduce resilience by sensitizing the species to land-use disturbance (Sutherland, Meyer, & Gardiner, 2002; Walters, Leigh & Bearden, 2003; Scott, 2006). We hypothesized that segment features would be constrained by catchment features leading to an overall patchy distribution of E. osburni marked by extirpations via land-cover cascades (Figure 3.1a) and persistent populations in catchment-generated refugia (Figure 3.1b). However, we also anticipated rare deviations (Figure 3.1c–e) from these two main pathways, which may provide information on the types and extent of resilient segment features necessary to sustain small populations within otherwise regionally marginal landscape contexts.

Study area Our study occurred within the New River Drainage (NRD) within the Appalachian Mountains of the eastern United States (Virginia, West Virginia). The NRD is within the Mississippi River basin and flows north transcending three physiographic provinces (Messinger & Hughes 2000). Surveys were restricted to the native range of E. osburni in the Appalachian Plateau (AP) and Valley and Ridge (VR) provinces. The AP is in the northern high-elevation portion of the NRD, typically containing rugged topography underlain by clastic and shale geology. Located south and east of the AP, the VR is a geologically diverse mosaic of sedimentary rocks, creating karst topography described by steep continuous mountain ridges separated by wide unconfined valleys (Messinger & Hughes, 2000). The region was predominantly vegetated by mixed-deciduous forests before European settlement (c. pre-1750; Messinger & Hughes, 2000). Early settlers cleared many low-relief valleys for pasture and crops, but the most severe land-use disturbance occurred when nearly all forests were commercially logged (c.1880–c.1920), resulting in extensive flooding, erosion, and wildfires, which

65 dramatically impacted stream habitat and biota (Goldsborough and Clark 1908). Current primary land uses are agriculture (i.e., pasture), development, and re-vegetated forests, particularly on ridges and public lands (Messinger & Hughes, 2000).

Focal species Etheostoma osburni is a non-game fish endemic to the AP and VR of the NRD. The species is a member of the sensitive fish fauna concentrated within the highlands of the southeastern United States, a hot-spot for both freshwater diversity and imperilment (Burkhead, 2012). Sparse early surveys (1885–1964) documented E. osburni in a variety of stream types across the NRD, indicating its pre-disturbance distribution was wider than where historically detected (Jenkins & Burkhead, 1994). However, recent (post-1970) fish surveys have failed to collect E. osburni in numerous streams with historical records, suggesting a widespread decline. Etheostoma osburni possess traits conferring low resilience to anthropogenic disturbance including a narrow geographic range and specialized resource needs (Pritt & Frimpong, 2010). The species is also likely a poor disperser, which can inhibit metapopulation dynamics (Schlosser & Angermeier, 1995) and impede recolonizing suitable habitat following historical disturbance (Albanese, Angermeier & Peterson, 2009). Overall, E. osburni’s traits and potentially localized distribution, make the species a good model to examine disturbance pathways and their implications for conservation planning.

Site selection We surveyed stream segments potentially impacted by land use in all streams (excluding large rivers) with available data on recent (post-1970) and historical (pre-1970) occurrences of E. osburni (Figure 3.2; Supplementary Table S3.1). We eliminated downstream sections of tributaries to the lower New, Gauley, and Greenbrier rivers where extirpation may be due to variegate darter (E. variatum), an introduced competitor. We also eliminated areas inundated by Bluestone Lake reservoir and a single record from the mouth of Spruce Run, VA, which may have been a disperser from the adjacent New River rather than a historical population. We selected an additional seven streams potentially capable of supporting E. osburni in the southern portion of the NRD where populations were suspected to be most localized. Five of these streams

66 occurred within catchments with historical records, and the remaining two were adjacent to catchments with historical records. Our surveys determined presence or absence within a stream segment, the length of stream between two consecutive tributary confluences (Frissell et al., 1986). We surveyed one segment in streams with recent records (13 segments) and two segments in streams without recent records (15 streams, 29 segments) in case declining populations were more patchily distributed. The exception was STN1, an adventitious stream containing a historical record, where we only surveyed one segment due to its short length (site codes are provided in Supplementary Table S3.1). We selected segments that overlapped with historical collection localities (Supplementary Table S3.1). If a stream had only one historical record, we randomly selected another segment of the same Strahler order. If a stream had no historical records, we randomly selected two accessible segments of the same stream order (orders 2–4). Overall, our data represent a range-wide survey of segments in all streams where E. osburni were confirmed present (28 segments within 21 streams) plus segments in potentially occupied streams near historically confirmed localities (14 segments within seven streams).

Fish and stream habitat surveys We surveyed 42 segments for E. osburni in May–July 2012. Rather than sampling all habitats in segments, we limited surveys to riffles, where the majority of the species’ life history occurs (Dunn & Angermeier, 2016). Within each segment, we randomly selected three geographic coordinates within 300 meters of a road or trail and surveyed the closest riffle to each coordinate (i.e., three riffles per segment). The field-crew surveyed each riffle twice to limit imperfect detection, with at least one week between surveys. At the base of each riffle standing perpendicular to flow adjacent to a bank, two crewmembers held a 1.5-m x 3-m seine with 5-mm bar mesh and a double-weighted leadline. Beginning 3 m upstream of the seine, a third crewmember electrofished (Smith-Root LR-24 backpack, pulsed direct current) downstream towards the seine while disturbing the substrate (one kick-seine). Non-overlapping kick-seines were repeated along a transect perpendicular to flow until reaching the opposite bank. After completing each transect, the crew moved upstream 4 m and began a new transect. We repeated this process until the entire riffle was surveyed. All individuals were enumerated and returned downstream of the area of capture. Based on the observed capture history and a hierarchical

67 occupancy-modeling analysis (Pavlacky et al., 2012), the estimated probability of failing to detect E. osburni within an occupied segment after surveying three riffles on two separate occasions was < 1% (C.G. Dunn, unpublished data). After the second fish survey, the crew estimated the means of five structural habitat variables of each riffle. Five equally spaced transects perpendicular to flow were placed in each riffle. We recorded wetted-channel width and then identified seven equidistant points along each transect to record stream depth, water-column velocity at 60% depth (Marsh-McBirney model 2000 flow meter), the width of a randomly selected substrate particle along its intermediate axis, and two visually estimated descriptors of fine-sediment within the 0.5-m2-area surrounding each point. Embeddedness was the average percentage of coarse substrate vertically submersed in the stream bottom, and silt-cover was the percentage of substrate surface area covered by silt. Percentages for both fine-sediment descriptors were categorized as,  ”5%, 1 = 6–25%, 2 = 26–50%, 3 = 51–75%, and 4 = 76–100%. This habitat protocol collected 35 observations per riffle and 105 observations per segment. All habitat observations were averaged within each segment to represent segment-wide riffle characteristics.

Stream temperature monitoring We placed a submersible temperature logger in one segment in each stream to monitor hourly temperatures in spring (March 1–May 31, 2012) and summer (June 1–August 31, 2012; Hobo Pendant or Tidbit v2, Onset Computer Corporation, Pocasset, MA). We ensured each logger was properly calibrated before deployment. Loggers were placed in flowing water representative of ambient segment conditions, and afterward, all data were screened for readings indicative of air exposure including temperatures beyond 0–30° C or periods of noticeably high diel variation (Dunham et al., 2005). Three loggers were temporarily displaced, requiring missing temperatures be predicted from nearby loggers using first-order autoregressive models (see Supplementary Material S3.2–S3.3 for details). Although, we only considered temperatures from spring–summer 2012, limited data (N = 10 segments) indicated temperatures in 2012 were representative of temperatures in both 2011 and 2013. For example, Pearson’s correlation (r) of mean temperatures in segments among years was high in both spring (r > 0.98 for 2011–2013) and summer (r = 0.96 for 2011–2012).

68 Only one logger was placed within each stream, which required predicting temperatures within the segment without a temperature logger in streams with multiple segments. We developed a mixed-effects linear regression model to estimate the difference between manually measured temperatures in segments without loggers (response variable; N = 84) and the simultaneous logger-recorded temperatures (predictor variable) from the corresponding segment in the same stream (Supplementary Material S3.2–S3.3). The model also included log- transformed upstream catchment area, the directional (upstream vs. downstream) fluvial distance separating paired segments within the same stream, and a random effect for each stream. Both autoregressive (previous paragraph) and the mixed-model were highly predictive of missing temperatures when assessed via cross-validation (i.e., R2 •6XSSOHPHQWDU\)LJXUH6), and therefore, were used to predict missing temperatures for subsequent analyses.

Data Analyses Analyses were divided into two parts. First, we used multivariate analyses to quantify the pathways that catchment- and segment-scale features influence instream habitat. Second, we developed regression models to examine if stream temperature and fine sediment in riffles were local instream determinants of E. osburni‘s distribution, while accounting for network configuration among segments.

Multi-scale relationships among stream habitat We first visualized the similarity of instream habitat among segments in multivariate space via principal component analysis (PCA) with a correlation matrix. Before PCA, we centered and standardized the means of seven instream habitat variables within each segment: spring daily temperature (SPMDT), summer daily maximum temperature (SMDMX), depth, velocity, substrate size, embeddedness, and silt-cover (Table 3.1; Supplementary Table S3.4). We plotted scores from the first two PCA eigenvectors, explaining 68% of variation in instream habitat among segments. To gain insight into the relationships between instream habitat and larger-scale features, we plotted arrays representing permutation-based Pearson correlations between the first two PCA axes (instream habitat) and three variable groups: segment-scale features, catchment land use, and natural catchment features. With the exception of mean-wetted-channel width, all

69 features were quantified remotely via GIS (see Supplementary Tables S3.5–S3.6 for details and Supplementary References for data sources). Segment features included descriptors of channel geomorphology (gradient and channel width), percentage near-stream surficial geologic category (50-m buffer), and percentage riparian land use type (30-m buffer). We calculated all catchment- scale features from the total area upstream of the lowest point in each segment. For catchment- and segment-scale land use, we simplified the 16 land-cover classes of the 2011 U.S. National Land Cover Dataset to three land uses: agricultural, developed, and forested. Natural catchment features included percentage surficial geologic category, mean elevation, and drainage area. For geology at both catchment and segment scales, we calculated the percentage area classified as clastic, shale, or carbonate, corresponding to a decreasing gradient of resistance to weathering. We expected more resistant geologic categories to have steeper channel gradients and more rugged topography less amenable to development and agriculture (Hack, 1957). Following PCA, we used partial Redundancy Analysis (pRDA) to examine the pathways (direct versus indirect) of catchment-scale influence on instream habitat. Partial RDA is a constrained ordination that partitioned total variation in instream habitat into three components: variation solely explained by each of the three variable groups, covariation among variable groups, and unexplained variation (Peres-Neto et al., 2006). Variation solely explained at the catchment scale supported direct catchment pathways. Covariation between catchment and segment features indicated indirect pathways where instream habitat and segment features were constrained by catchment features (land-cover cascade, Figure 3.1a; resilient-catchment pathway, Figure 3.1b). Conversely, variation uniquely explained at the segment scale indicated a resilient- segment pathway (Figure 3.1c) or local sensitivity of segment-features. Before pRDA, channel gradient, mean upstream elevation, and catchment area were square-root transformed to improve linearity. Each variable group consisted of the most parsimonious sets of predictor variables explaining instream habitat among segments. To limit the number of variables within each variable group, we retained a single variable for highly correlated variable pairs (r •__  Next, we only included variables with sufficient explanatory ability via a step-wise selection procedure (permutation-based p-YDOXHV”). Therefore, exclusion of a variable could be due to either multi-collinearity within the same variable group or poor explanatory power. Finally, we reported variation explained with an adjusted R2 statistic penalized by the number of variables

70 within each variable group (Peres-Neto et al., 2006). Seven of the original 16 candidate variables were retained for pRDA.

Hypothesized instream habitat predictors of persistence Elevated stream temperature and fine sediment are two primary local stressors that impact highland fishes (Sutherland, Meyer & Gardiner, 2002; Scott, 2006). We selected two temperature metrics representing alternative hypothesized impacts on E. osburni persistence. The first metric, SPMDT, represented temperatures during spawning season in spring, and the second metric, SMDMX, represented stress induced by potentially exceeding upper limits of thermal tolerance in summer. We also selected riffle embeddedness and silt-cover as substrate metrics representing hypothesized impacts on persistence. High embeddedness chronically reduces substrate complexity, thereby eliminating structure necessary for cover, spawning, and foraging. Silt-cover represented recently deposited fine sediment, with a higher probability of being re- suspended and transported downstream.

Spatial configuration of persistence We used distance-based eigenvector mapping to quantify the network spatial configuration of segments and populations, while guarding against Type-1 error potentially resulting from spatially autocorrelated environmental gradients (Griffith & Peres-Neto, 2006). We first created a matrix of pairwise fluvial distances among all segments based on the National Hydrography Dataset Plus version 2. The distance matrix was converted to a truncated connectivity matrix with a minimum-spanning tree algorithm and a threshold distance 253.2 km, the distance needed to maintain an intact network. A randomization procedure with 999 replicates indicated significant spatial autocorrelation within the observed E. osburni distribution (Moran’s I = 0.21, p = 0.001). Therefore, autocorrelation was decomposed via principal coordinate analysis into 41 orthogonal eigenvectors representing coarse- to fine-scale network configuration. We retained the first four eigenvectors with significant (p < 0.05) positive autocorrelation to be used as metrics (covariates) for network configuration.

Model development and cross-validation

71 We used logistic regression and an Information Theoretic framework to quantify the relative support for competing hypotheses relating environmental and spatial predictors to persistence. Competing hypotheses were ranked by Akaike’s Information Criterion corrected for small sample size (AICc), a parsimonious metric of model fit penalized by model complexity (Hobbs & Hilborn, 2006). Candidate predictors included all subsets for temperature (two variables), fine-sediment (two variables), and spatial configuration (four covariates). Candidate models were limited to four predictors to prevent over-parameterization, and no models included both embeddedness and silt-cover due to their high collinearity (Pearson’s r = 0.75). Finally, to assess model performance, we also report mean area-under-the-curve (AUC) of each model assessed via seven-fold cross-validation (Manel, Williams & Ormerod, 2001; see Supplementary Material S3.7 for details on cross-validation).

Results Instream surveys We collected 773 individuals within 16 of 42 segments. Detections primarily occurred throughout the species’ northern range; however, we also detected E. osburni in five segments spread across their southern range in the VR (Figure 3.2). We detected E. osburni in all segments with recent records (post-1970), in only one segment with a historical record (STN1), and we discovered E. osburni in two new segments: a single likely disperser in WLF2 from LVA1 located only 1 km upstream, and 30 individuals in CRP1, indicative of a previously undocumented population.

Graphical relationships among instream habitat and larger-scale features The PCA revealed two main gradients in instream habitat among segments (Figure 3.3a; Table 3.1). Principal component 1 (PC1, horizontal axis) explained 43.2% of variation and likely represented increasing degree of habitat degradation (right side of Figure 3.1 is most impaired). Positively loaded segments on PC1 were warmer, more embedded, and contained finer substrates, and also had higher water velocities, likely due to greater summer baseflows in many of the heavily degraded southern VR streams. Principal component 2 (vertical axis) explained 24.8% of variation and represented increasing stream size based on greater mean depths and higher silt-cover.

72 Axes were highly correlated with patterns at larger spatial extents (Table 3.1; Figure 3.3b–d). Segments positively loaded on PC1 (i.e., more degraded) had more agriculture and development, had carbonate and shale geology, occurred at lower elevations, and had lower channel gradients. High correlations between PC2 and catchment area (r = 0.83) and channel width (r = 0.81), confirmed this axis represented stream size. The PCA clearly showed persistence was associated with less degraded instream habitat (negatively loaded on PC1). These segments primarily occurred in the northern study extent in the Gauley and Greenbrier river sub-basins; however, three of five segments in the southern range with E. osburni were also negatively loaded, indicating E. osburni persisted in geographically rare segments with features more similar to those found in segments in the northern range (i.e., catchment- or segment-generated refugia). Two positively loaded segments were unexpectedly occupied, demonstrating that although unlikely, persistence of populations is possible in segments with marginal instream habitat (WLF2 and CRP1). Overall, PCA revealed persistence in segments corresponded to interrelated physiography and habitat degradation at multiple spatial scales.

Pathways of catchment influence on instream habitat Partial RDA revealed the scales and pathways that features influence instream habitat (Table 3.1; Figure 3.4). Catchment land use, natural catchment features, and segment features collectively explained 52% of adjusted variation (60% of raw variation) in instream riffle habitat. Catchment land use primarily co-varied with natural catchment features (16.9%), confirming the natural catchment provides a template amenable to specific land uses.

Catchment-scale features predominately constrained segment features and instream habitat supporting indirect pathways (29%). Within the context of land-use disturbance, this high percentage indicated catchment disturbance stimuli are propagated at segment scales as land- cover cascades or mitigated at the catchment scale as resilient-catchment pathways. In contrast, only 4.2% of variation was uniquely explained by segment-scale features (i.e., evidence for resilient-segment pathways or local sensitivity of segment-scale features).

Local predictors of persistence

73 The best-supported distribution model indicated persistence was explained by temperature, substrate, and network configuration (Table 3.2; Supplementary Table S3.8). The best-supported model indicated persistence was inversely related to SPMDT (logit ߚመ = -1.17, SE = 0.62) and embeddedness (logit ߚመ = -4.77, SE = 2.5), plus the first two spatial covariates. Threshold responses for SPMDT occurred between 12 and 14 °C (Figure 3.5a). For example, the predicted probability of persisting decreased from 0.90 ± 0.03, 90% confidence interval (CI90) at 12°C to 0.48 ± 0.22 CI90 at 14°C. Similarly, as mean riffle embeddedness in segments increased from 10% to 30%, the probability of persisting decreased from 0.94 ± 0.01 CI90 to 0.25 ± 0.18 CI90, respectively (Figure 3.5b). Inclusion of two spatial covariates indicated stream network configuration was correlated with persistence. The first spatial covariate (logit ߚመ = -3.8, SE = 3.1) represented coarse geographic patterns, with persistence increasing in northern and eastern portions of the NRD (Pearson’s r between spatial covariate 1 and UTM easting = -0.86; northing = -0.96). The second spatial covariate (logit ߚመ = -5.6, SE = 3.2) was not interpretable across geographic coordinates and may reflect finer-scale stream network topology. Spatial covariates weakened absolute effects of spatially structured instream habitat and imposed a spatial context on predictions. For example, effects of temperature and embeddedness on persistence were less severe in segments with favorable spatial locations (Figure 3.5c–d). Predictions using identical median geographic coordinates (i.e., controlling for space) indicated LWA2, SEC1, and SEC2 were unoccupied segments containing suitable instream habitat, while persistence in GAL1, WLF2, CRP1, and LWV1 was not predicted based on high embeddedness or spring temperatures (Supplementary Table S3.9). Finally, spatial covariates successfully removed spatial autocorrelation among residuals (Moran’s I = -0.04, p = 0.82). There was considerable model-selection uncertainty, with the top model only receiving 14% of model weight (Table 3.2; Supplementary Table S3.8). Contributors to uncertainty were comparable effects of multiple variables for temperature and fine-sediment (i.e., multi- collinearity), and similar model structures for instream variables permuted among combinations of spatial covariates. However, top models representing each hypothesis (i.e., roles of spatial, fine-sediment, and temperature) were far more strongly supported than the intercept-only model

(wnull = 0.0). Despite high model-selection uncertainty, the best-supported models performed well during cross-YDOLGDWLRQ $8&•0DQHO:LOOLDPV 2UPHURG 

74 Discussion We documented probable extirpations of E. osburni from at least 14 segments in seven streams, particularly in the southern portion of the species’ range. This pattern is consistent with the recent localization of many sensitive stream-fish populations, particularly in highland regions of the southeastern U.S. We hypothesize patterns likely result from at least four interacting phenomena: (1) predisposition of certain natural landscapes to detrimental land uses, (2) varying natural resilience of stream habitat to land-use disturbance, (3) spatial variation in resilience of populations to disturbance, and 4) insufficient corridor habitat to facilitate recolonization of suitable habitat following local extinction.

Multi-scale environmental sources of resilience to landscape disturbance Our findings demonstrated that natural catchment features mediate the impacts of intensive land uses. Interpreting whether species’ distributions are shaped by anthropogenic versus natural factors can be confounded by covariation between land use and natural features (Allan, 2004); however, by re-surveying historically occupied segments, we determined that the recent localization of E. osburni was driven mainly by land use. For example, E. osburni were primarily extirpated in segments within agricultural (i.e., pastoral) catchments. Agriculture predominated in these catchments due to lower elevations and subdued topography, created by carbonate-shale geology (Hack, 1957). The average extent of agriculture in these catchments (28.8%) exceeded thresholds known to impact other sensitive highland fishes in the region (10– 20% non-forest, [Sutherland, Meyer & Gardiner, 2002]; 12% agriculture, [Hudy et al., 2008]). These documented fish extirpations parallel the distributional patterns of other highland stream biota (Sponseller, Benfield & Valett, 2001), which collectively demonstrate heightened sensitivity of highland stream communities to intensive land uses (Utz, Hilderbrand & Raesly, 2010). Our analyses suggested disturbance from catchment-scale land use was primarily propagated to instream habitat via indirect segment-scale pathways (i.e., land-cover cascade pathway, Figure 3.1a). Probable pathways include runoff from extensive pasture adjacent to streams, promoted by unconfined channels characteristic of near-stream carbonate-shale geology (Harding et al., 1999; Isaak & Hubert, 2001), widening of stream channels through bank failure

75 (Trimble & Mendel, 1995), and elevated solar and fine-sediment inputs associated with reduced riparian vegetation (Naiman & Décamps, 1997). Each of these indirect pathways may have contributed to observed localization of E. osburni. Etheostoma osburni primarily persisted in refugia generated by catchment-scale features, including high elevations and clastic geology. These features limit agriculture and likely provided resilience at multiple scales to the intense disturbance from historical logging that dominated the region (c.1880–c.1920). For example, erosion-resistant clastic geology typically promotes steep stream channels (Hack, 1957), which would have limited deposition of fine sediments (Walters, Leigh & Bearden, 2003; Bywater-Reyes, Segura & Bladon, 2017; Montgomery & Buffington, 1997), allowed access to high-elevation thermal refugia (Isaak & Rieman, 2013), and likely provided habitat for larger populations of E. osburni (Dunn & Angermeier, 2016), more resilient to short-term disturbance. Although specific catchment features conferring resilience may vary by region, management strategies may be generalizable depending on the pathway of disturbance. For example, if resilience is primarily conveyed at the catchment scale, prioritizing suitable catchments will be a viable initial approach for identifying persistent populations. Once identified, an important challenge will be limiting regional neighborhood effects on these catchment-generated refugia, including effects of invasive species (Fausch et al., 2009).

Instream predictors of persistence Our results indicated excessive deposition of fine sediment contributed to the localization of E. osburni. Excessive fine sediment can impact stream ecosystems, and potentially E. osburni, in a variety of ways (Wood & Armitage, 1997; Kemp et al., 2011). Slightly more support for embeddedness than silt-cover, as predictors of localization, indicated E. osburni may be more sensitive to low substrate complexity than to recent silt inputs. Although unexplored, diminished substrate complexity due to embeddedness could reduce foraging and cover habitats for both E. osburni and their food base (Sponseller, Benfield & Valett, 2001). Moreover, E. osburni are lithophilous spawners, which are disproportionally sensitive to fine-sediment deposition within spawning habitat (Sutherland, Meyer & Gardiner, 2002; Walters, Leigh & Bearden, 2003). Alternatively, Dunn & Angermeier (2016) documented avoidance of embedded

76 areas by three life stages of E. osburni, indicating embeddedness may impact performance at multiple stages during the species’ life cycle. Almost all extirpated populations formerly occurred in segments that currently have warmer spring and summer temperatures. For example, segments where extirpations occurred were on average 1.6 °C warmer in spring and had SMDMX 0.6°C higher in summer than in segments that still support E. osburni. Warmer temperatures associated with extirpation partly reflect natural factors, including lower elevations and latitudes; however, dramatically higher levels of agriculture near extirpations likely contributed to warmer temperatures by increasing surface runoff (Trimble & Mendel, 1995; Poole & Berman, 2001) and solar exposure (Isaak & Hubert, 2001). The top model explaining localization indicated mean temperature in spring was a slightly better predictor than summer maxima. Although, there was considerable uncertainty regarding which thermal metric was the better predictor, this finding indicated warm spring temperatures may impact E. osburni in ways other than simply exceeding lethal thermal limits, such as desynchronizing spring thermal cues for optimal spawning conditions (Krabbenhoft, Platania & Turner, 2014) or reducing survival of early-life stages (McCullough et al., 2009; Turschwell et al., 2017). Alternatively, E. osburni may be sensitive to other aspects of stream temperature regime, including chronically warm temperatures across multiple seasons, which can impact fitness by facilitating disease, accelerating metabolism, and modifying fish behavior (Whitney et al., 2016). Regardless, these findings demonstrate lesser-studied stream fishes, including E. osburni, are likely impacted by elevated temperatures. Understanding relationships among temperature and other stressors could benefit conservation by enabling managers to assess the feasibility of mitigating adverse effects of temperature indirectly by limiting the effects of more easily managed stressors (Townsend, Uhlmann & Matthaei, 2008).

Spatial considerations for persistence Landscape disturbances often accumulate downstream, leading to impaired habitat (Harding et al., 1999; Scott et al., 2002) and extirpations of highland biota in large mainstem streams (Burkhead et al., 1997). This pattern not only impacts mainstem populations but can further isolate peripheral sub-populations, potentially leading to catchment-wide extirpation (Schlosser and Angermeier 1995; Fagan, 2002). Extirpations of E. osburni in the mainstems of at

77 least five streams (nine segments) in their southern range conform to this pattern. Some of these mainstems had multiple confirmed formerly occupied segments, indicating populations within each catchment were once likely connected via mainstem corridors. Persistence of E. osburni in mainstem streams in their northern range (e.g., CHR1, GAL1, WGR1) demonstrated extirpation was not due to stream size per se, but to disproportionally impaired conditions associated with larger streams in their southern range. Indeed, Gibson (2017) confirmed sub-populations within each of the upper Gauley and Greenbrier river sub-basins are genetically panmictic, supporting the hypothesis that large rivers in their northern range still function as corridors for, rather than impediments to, dispersal. Given the extirpations of E. osburni that we documented in mainstem streams in their southern range, managed translocation may be a viable resilience-building tactic to simulate former connectivity among peripheral populations (Olden et al., 2011). Inclusion of two spatial covariates demonstrated persistence was spatially configured within the stream network. This spatial context partly explained mismatches between instream habitat suitability and the observed distributional pattern. For example, E. osburni populations persisted in spatially favorable locations in the northern range despite embedded substrate. Persistence in these segments could be due to high meta-population connectivity (Figure 3.1d), lagged extirpation (i.e., extinction debt sensu Jackson & Sax, 2010), or important latent environmental gradients that were also spatially configured (Peres-Neto & Legendre, 2010). Similarly, populations were absent in multiple isolated segments with suitable instream habitat in the spatially poor southern range, which could also result from latent environmental gradients, or habitat recovery without a corridor for recolonization (Stoll et al., 2014; Figure 3.1e). Interpreting the spatial patterns underlying persistence would be aided by studies that quantify temporal lags in habitat succession (i.e., legacy effects sensu Harding et al., 1998) and population vital rates (i.e., extinction debt), and experimentally reintroduce individuals into suitable, yet isolated, streams.

Segment-scale resilience to landscape disturbance Although, segment features and instream habitat were predominately constrained by catchment-scale phenomena, 4.2% of variation in instream habitat was solely explained by segment-scale features, providing evidence for occasional segment-scale resilience to catchment influences (Figure 3.1c). An example may be the discovery of E. osburni within Cripple Creek

78 (CRP), which was unexpected given the stream’s isolation, poor spatial context in the southernmost geographic range, and high levels of catchment agriculture (34%). Moreover, the distribution within CRP was restricted to the farthest downstream segment (CRP1), suggesting persistence coincided with unique refugia generated by segment-scale features (Sedell et al., 1990). Stream temperatures in the unoccupied upstream segment (CRP2) were comparable, but CRP1 riffles were less embedded. A potential source of segment-scale resilience in CRP1 was the relatively intact riparian zone (80% forested in CRP1 versus 50.5% in CRP2). Riparian forest can lower stream temperatures (Poole & Berman, 2001), limit fine-sediment inputs (Jones et al., 1999), and promote longitudinal biotic recovery (Sponseller, Benfield & Valett, 2001; Frimpong et al., 2005). Cripple Creek may serve as a case study for further investigation of relationships between fine-sediment and stream temperature, and for identifying the spatial extents necessary to sustain E. osburni populations. Such studies could help identify segments that potentially support small populations or help establish riparian, thermal, and substrate benchmarks for stream restoration at segment spatial scales (Seavy et al., 2009). Regardless, persistence in CRP1 demonstrated the need to better document existing distributions, which may reveal unknown dimensions of habitat suitability.

Conclusion Persistence of E. osburni populations reflected two local habitat variables readily compromised by intensive land use: cool stream temperature and unembedded substrate. Phenomena at multiple spatial scales interacted to either maintain or eliminate these features in localities across the study region. Although uncommon, we found some evidence that natural segment-scale features can buffer land-use disturbance so that local habitat remained suitable for E. osburni. Cool stream temperatures will become particularly rare as air temperatures rise (McDonnel et al., 2015). However, our results show that when thermal stress is sub-lethal, a potential resilience-building strategy is to limit effects of other stressors such as excess fine sediment. Other potentially effective resilience strategies include maintaining or restoring regional connectivity among populations and enhancing the capacities of stream segments to buffer the impacts of stressors (Wilby et al., 2010; Williams et al., 2015). In lieu of uniformly applying conservation actions across a species’ entire range, practitioners may be aided by natural features that confer population resilience. Retrospective analyses of effects of historical

79 land-use change on stream biota may offer insight regarding the necessary scale of restoration and help prioritize conservation actions to recover populations subsisting in marginal landscape contexts. Finally, further investigation of fitness consequences of changes in temperature and fine sediment for a greater diversity of aquatic organisms seems warranted; a more comprehensive understanding may help conservation practitioners better anticipate and adapt to future land-use and climate change (Staudt et al., 2013).

Acknowledgements This research was partly funded by a State Wildlife Grant from the Virginia Department of Game and Inland Fisheries (VDGIF). We thank M. Pinder (VDGIF Aquatic Biologist) for stream recommendations. Assistance was provided by D. Dodge, L. Longanecker, G. Anderson, J. Argentina, A. Villamagna, J. Roberts, B. Mogollon, J. Cline, and L. Zseleczky. This research was carried out under the auspices of Institutional Animal Care and Use Committee protocol 10- 094-FIW at Virginia Polytechnic Institute and State University (Virginia Tech). All fish sampling was compliant with requirements set forth by VDGIF, West Virginia Division of Natural Resources, and U.S. Forest Service. The Virginia Cooperative Fish and Wildlife Research Unit is jointly sponsored by the U.S. Geological Survey, Virginia Tech, VDGIF, and Wildlife Management Institute. Use of trade names or commercial products does not imply endorsement by the U.S. government.

References Albanese B., Angermeier P.L. & Peterson J.T. (2009). Does mobility explain variation in colonisation and population recovery among stream fishes? Freshwater Biology, 54, 1444–1460.

Allan J.D. (2004). Landscapes and riverscapes: The influence of land use on stream ecosystems. Annual Review of Ecology Evolution and Systematics, 35, 257–284.

Angermeier P., Krueger K. & Dolloff C. (2002). Discontinuity in stream-fish distributions: implications for assessing and predicting species occurrence. In: Predicting Species Occurrences: Issues of Accuracy and Scale. (Eds J.M. Scott, P. Heglund, M.L. Morrison,

80 J.B. Haufler, M.G. Raphael, W.A. Wall & F.B. Samson), pp. 519–527. Covelo, CA: Island Press.

Burcher C.L., Valett H.M. & Benfield E.F. (2007). The land-cover cascade: relationships coupling land and water. Ecology, 88, 228–242.

Burdon F.J., Mcintosh A.R. & Harding J.S. (2013). Habitat loss drives threshold response of benthic invertebrate communities to deposited sediment in agricultural streams. Ecological Applications, 23, 1036–1047.

Burkhead N., Walsh S., Freeman B. & Williams J. (1997). Status and restoration of the Etowah River, an imperiled southern Appalachian ecosystem. In: Aquatic Fauna in Peril: the Southeastern Perspective. (Eds G.W. Benz & D.E. Collins), pp. 375–444. Decatur, GA: Southeast Aquatic Research Institute, Lenz Deisgn and Communications,.

Burkhead N.M. (2012). Extinction rates in North American freshwater fishes, 1900–2010. Bioscience, 62, 798–808.

Bywater-Reyes S., Segura C. & Bladon K.D. (2017). Geology and geomorphology control suspended sediment yield and modulate increases following timber harvest in temperate headwater streams. Journal of Hydrology, 548, 754–769.

Caissie D. (2006). The thermal regime of rivers: a review. Freshwater Biology, 51, 1389–1406.

Comte L., Buisson L., Daufresne M. & Grenouillet G. (2013). Climate-induced changes in the distribution of freshwater fish: observed and predicted trends. Freshwater Biology, 58, 625–639.

Comte L. & Grenouillet G. (2015). Distribution shifts of freshwater fish under a variable climate: comparing climatic, bioclimatic and biotic velocities. Diversity and Distributions, 21, 1014–1026.

81 Dunham J., Chandler G., Rieman B. & Martin D. (2005). Measuring Stream Temperature with Digital Data Loggers: A User's Guide (Report No. RMRS-GTR-150WWW). Fort Collins, CO: U.S. Department of Agriculture, U.S. Forest Service, Rocky Mountain Research Station. (Retrieved from: https://fresc.usgs.gov/products/papers/1431_Dunham.pdf).

Dunn C.G. & Angermeier P.L. (2016). Development of habitat suitability indices for the Candy Darter, with cross-scale validation across representative populations. Transactions of the American Fisheries Society, 145, 1266–1281.

Fagan W.F. (2002). Connectivity, fragmentation, and extinction risk in dendritic metapopulations. Ecology, 83, 3243–3249.

Fausch K.D., Rieman B.E., Dunham J.B., Young M.K. & Peterson D.P. (2009). Invasion versus isolation: trade-offs in managing native salmonids with barriers to upstream movement. Conservation Biology, 23, 859–870.

Folke C., Carpenter S.R., Walker B., Scheffer M., Chapin T. & Rockstrom J. (2010). Resilience thinking: integrating resilience, adaptability and transformability. Ecology and Society, 15. (Retrieved from: http://www.ecologyandsociety.org/vol15/iss4/art20/)

Frimpong E.A., Sutton T.M., Lim K.J., Hrodey P.J., Engel B.A., Simon T.P., ... Le Master D.C. (2005). Determination of optimal riparian forest buffer dimensions for stream biota– landscape association models using multimetric and multivariate responses. Canadian Journal of Fisheries & Aquatic Sciences, 62, 1–6.

Frissell C.A., Liss W.J., Warren C.E. & Hurley M.D. (1986). A hierarchical framework for stream habitat classification: viewing streams in a watershed context. Environmental Management, 10, 199–214.

82 Gibson I. (2017). Conservation concerns for the candy darter (Etheostoma osburni) with implication related to hybridization (MS thesis, West Virginia University, Morgantown).

Goldsborough E.L. & Clark H.W. (1908). Fishes of West Virginia. Bulletin of the U.S. Bureau of Fisheries, 27, 29–39.

Griffith D.A. & Peres-Neto P.R. (2006). Spatial modeling in ecology: the flexibility of eigenfunction spatial analyses. Ecology, 87, 2603–2613.

Hack J.T. (1957). Studies of Longitudinal Stream Profiles in Virginia and Maryland (Report No. 294-B). Washington DC: U.S. Geological Survey.

Harding J.S., Benfield E.F., Bolstad P.V., Helfman G.S. & Jones E.B.D. (1998). Stream biodiversity: the ghost of land use past. Proceedings of the National Academy of Sciences of the United States of America, 95, 14843–14847.

Harding J.S., Young R.G., Hayes J.W., Shearer K.A. & Stark J.D. (1999). Changes in agricultural intensity and river health along a river continuum. Freshwater Biology, 42, 345–357.

Hobbs N.T. & Hilborn R. (2006). Alternatives to statistical hypothesis testing in ecology: a guide to self teaching. Ecological Applications, 16, 5–19.

Hudy M., Thieling T.M., Gillespie N. & Smith E.P. (2008). Distribution, status, and land use characteristics of subwatersheds within the native range of brook trout in the eastern United States. North American Journal of Fisheries Management, 28, 1069–1085.

Isaak D.J. & Hubert W.A. (2001). A hypothesis about factors that affect maximum summer stream temperatures across montane landscapes. Journal of the American Water Resources Association, 37, 351–366.

83 Isaak D.J. & Rieman B.E. (2013). Stream isotherm shifts from climate change and implications for distributions of ectothermic organisms. Global Change Biology, 19, 742–751.

Jackson S.T. & Sax D.F. (2010). Balancing biodiversity in a changing environment: extinction debt, immigration credit and species turnover. Trends in Ecology & Evolution, 25, 153– 160.

Jelks H.L., Walsh S.J., Burkhead N.M., Contreras-Balderas S., Diaz-Pardo E., Hendrickson D.A., Lyons J., … Warren M.L. (2008). Conservation status of imperiled North American freshwater and diadromous fishes. Fisheries, 33, 372–407.

Jenkins R.E. & Burkhead N.M. (1994). Freshwater fishes of Virginia, Bethesda, MD: American Fisheries Society.

Jones E.B.D., Helfman G.S., Harper J.O. & Bolstad P.V. (1999) Effects of riparian forest removal on fish assemblages in southern Appalachian streams. Conservation Biology, 13, 1454–1465.

Kemp P., Sear D., Collins A., Naden P. & Jones I. (2011). The impacts of fine sediment on riverine fish. Hydrological Processes, 25, 1800–1821.

Krabbenhoft T.J., Platania S.P. & Turner T.F. (2014). Interannual variation in reproductive phenology in a riverine fish assemblage: implications for predicting the effects of climate change and altered flow regimes. Freshwater Biology, 59, 1744–1754.

Lynch A.J., Myers B.J.E., Chu C., Eby L.A., Falke J.A., Kovach R.P., … Whitney J.E. (2016). Climate change effects on North American inland fish populations and assemblages. Fisheries, 41, 346–361.

84 Maloney K.O. & Weller D.E. (2011). Anthropogenic disturbance and streams: land use and land- use change affect stream ecosystems via multiple pathways. Freshwater Biology, 56, 611–626.

Manel S., Williams H.C. & Ormerod S.J. (2001). Evaluating presence-absence models in ecology: the need to account for prevalence. Journal of Applied Ecology, 38, 921–931.

Mccullough D.A., Bartholow J.M., Jager H.I., Beschta R.L., Cheslak E.F., Deas M.L., … Wurtsbaugh W.A. (2009). Research in thermal biology: burning questions for coldwater stream fishes. Reviews in Fisheries Science, 17, 90–115.

McDonnell T.C., Sloat M.R., Sullivan T.J., Dolloff C.A., Hessburg P.F., Povak N.A., … Sams C. (2015). Downstream warming and headwater acidity may diminish coldwater habitat in southern Appalachian mountain streams. PloS ONE, 10. e0134757.

Messinger T. & Hughes C. (2000). Environmental Setting and Its Relations to Water Quality in the Kanawha River Basin (Report No. 00-4020). Reston, Virginia: U.S. Geological Survey, Water Resources Investigations.

Montgomery D.R. & Buffington J.M. (1997). Channel-reach morphology in mountain drainage basins. Geological Society of America Bulletin, 109, 596–611.

Naiman R.J. & Decamps H. (1997). The ecology of interfaces: riparian zones. Annual Review of Ecology and Systematics, 28, 621–658.

Olden J.D., Kennard M.J., Lawler J.J. & Poff N.L. (2011). Challenges and opportunities in implementing managed relocation for conservation of freshwater species. Conservation Biology, 25, 40–47.

85 Pavlacky D.C., Blakesley J.A., White G.C., Hanni D.J. & Lukacs P.M. (2012). Hierarchical multi-scale occupancy estimation for monitoring wildlife populations. Journal of Wildlife Management, 76, 154–162.

Peres-Neto P.R. & Legendre P. (2010). Estimating and controlling for spatial structure in the study of ecological communities. Global Ecology and Biogeography, 19, 174–184.

Peres-Neto P.R., Legendre P., Dray S. & Borcard D. (2006). Variation partitioning of species data matrices: estimation and comparison of fractions. Ecology, 87, 2614–2625.

Poff N.L. (1997). Landscape filters and species traits: towards mechanistic understanding and prediction in stream ecology. Journal of the North American Benthological Society, 16, 391–409.

Poole G.C. (2002). Fluvial landscape ecology: addressing uniqueness within the river discontinuum. Freshwater Biology, 47, 641–660.

Poole G.C. & Berman C.H. (2001). An ecological perspective on in-stream temperature: natural heat dynamics and mechanisms of human-caused thermal degradation. Environmental Management, 27, 787–802.

Pritt J.J. & Frimpong E.A. (2010). Quantitative determination of rarity of freshwater fishes and implications for imperiled-species designations. Conservation Biology, 24, 1249–1258.

R Core Team (2017). R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria.

Schlosser I.J. & Angermeier P.L. (1995). Spatial variation in demographic processes of lotic fishes: Conceptual models, empirical evidence, and implications for conservation. In: Evolution and the Aquatic Ecosystem: Defining Unique Units in Population

86 Conservation. (Ed J.L. Nielsen), pp. 392–401. Bethesda, MD: American Fisheries Society.

Scott M.C. (2006). Winners and losers among stream fishes in relation to land use legacies and urban development in the southeastern US. Biological Conservation, 127, 301–309.

Scott M.C., Helfman G.S., McTammany M.E., Benfield E.F. & Bolstad P.V. (2002). Multiscale influences on physical and chemical stream conditions across Blue Ridge landscapes. Journal of the American Water Resources Association, 38, 1379–1392.

Seavy N.E., Gardali T., Golet G.H., Griggs F.T., Howell C.A., Kelsey R., Small S.L., Viers J.H. & Weigand J.F. (2009). Why climate change makes riparian restoration more important than ever: recommendations for practice and research. Ecological Restoration, 27, 330– 338.

Sedell J.R., Reeves G.H., Hauer F.R., Stanford J.A. & Hawkins C.P. (1990). Role of refugia in recovery from disturbances: modern fragmented and disconnected river systems. Environmental Management, 14, 711–724.

Sponseller R.A., Benfield E.F. & Valett H.M. (2001). Relationships between land use, spatial scale and stream macroinvertebrate communities. Freshwater Biology, 46, 1409–1424.

Staudt A., Leidner A.K., Howard J., Brauman K.A., Dukes J.S., Hansen L.J., Paukert C., … Solorzano L.A. (2013). The added complications of climate change: understanding and managing biodiversity and ecosystems. Frontiers in Ecology and the Environment, 11, 494–501.

Stoll S., Kail J., Lorenz A.W., Sundermann A. & Haase P. (2014). The importance of the regional species pool, ecological species traits and local habitat conditions for the colonization of restored river reaches by fish. PloS ONE, 9, e84741.

87 Suding K.N. & Hobbs R.J. (2009). Threshold models in restoration and conservation: a developing framework. Trends in Ecology & Evolution, 24, 271–279.

Sutherland A.B., Meyer J.L. & Gardiner E.P. (2002). Effects of land cover on sediment regime and fish assemblage structure in four southern Appalachian streams. Freshwater Biology, 47, 1791–1805.

Tonn W.M., Magnuson J.J., Rask M. & Toivonen J. (1990). Intercontinental comparison of small-lake fish assemblages: the balance between local and regional processes. American Naturalist, 136, 345–375.

Torgersen C.E., Price D.M., Li H.W. & Mcintosh B.A. (1999). Multiscale thermal refugia and stream habitat associations of chinook salmon in northeastern Oregon. Ecological Applications, 9, 301–319.

Trimble S.W. & Mendel A.C. (1995). The cow as a geomorphic agent — a critical review. Geomorphology, 13, 233–253.

Townsend C.R., Uhlmann S.S. & Matthaei C.D. (2008). Individual and combined responses of stream ecosystems to multiple stressors. Journal of Applied Ecology, 45, 1810–1819.

Turschwell M.P., Balcombe S.R., Steel E.A., Sheldon F. & Peterson E.E. (2017). Thermal habitat restricts patterns of occurrence in multiple life-stages of a headwater fish. Freshwater Science, 36, 402–414.

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, 688–699.

Walters D.M., Leigh D.S. & Bearden A.B. (2003). Urbanization, sedimentation, and the homogenization of fish assemblages in the Etowah River basin, USA. Hydrobiologia, 494, 5–10.

88 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, U.S.A. Freshwater Biology, 48, 1950– 1970.

Whitney J.E., Al-Chokhachy R., Bunnell D.B., Caldwell C.A., Cooke S.J., Eliason E.J., … Paukert C.P. (2016). Physiological basis of climate change impacts on North American inland fishes. Fisheries, 41, 332–345.

Wilby R.L., Orr H., Watts G., Battarbee R.W., Berry P.M., Chadd R., … Wood P.J. (2010). Evidence needed to manage freshwater ecosystems in a changing climate: turning adaptation principles into practice. Science of the Total Environment, 408, 4150-4164.

Williams J.E., Neville H.M., Haak A.L., Colyer W.T., Wenger S.J. & Bradshaw S. (2015). Climate change adaptation and restoration of western trout streams: opportunities and strategies. Fisheries, 40, 304–317.

Wood P.J. & Armitage P.D. (1997). Biological effects of fine sediment in the lotic environment. Environmental Management, 21, 203–217.

Wu J. (2013). Hierarchy theory: an overview. In: Linking Ecology and Ethics for a Changing World. (Eds. R. Rozzi, C. Palmer, J.B. Callicott, S.T.A. Picket, J.J. Armesto), pp. 281– 301. Springer.

89 TABLE 3.1. Means ± (standard errors) of habitat variables in 42 segments as inputs in multi- variate analyses. Values under Principal Component (PC) 1 and 2 are permutation-based correlation coefficients between environmental variables and the first two PC axes. Bolded values highlight the axis with the higher correlation coefficient.

Variable Detected Not detected PC1 PC2 Instream habitat in segments Spring mean daily temperature (°C) 11.7 (0.3) 14.0 (0.2) 0.99 0.11 Summer mean daily maximum temperature (°C) 22.5 (0.4) 23.1 (0.3) 0.94 0.33 Mean riffle depth (cm) 13.6 (1.0) 17.1 (1.1) 0.67 0.74 Mean riffle velocity (m/s) 0.28 (0.03) 0.38 (0.02) 0.79 0.61 Mean riffle substrate size (cm) 10.0 (0.8) 7.7 (0.5) -0.78 0.63 Mean riffle embeddedness index (0–4) 0.4 (0.1) 1.0 (0.1) 0.87 -0.50 Mean riffle silt-cover index (0–4) 0.2 (0.0) 0.5 (0.1) 0.57 -0.82 Segment-scale features *†Channel gradient (m/km) 11.0 (1.7) 8.0 (0.7) -0.99 0.16 *†Channel width (m) 8.0 (1.1) 10.1 (1.1) 0.59 0.81 *Carbonate geology (% in 50-m buffer) 4.4 (3.0) 37.2 (8.1) 0.88 0.48 *†Clastic geology (% in 50-m buffer) 52.0 (10.6) 2.7 (1.9) -1.00 0.01 Shale geology (% in 50-m buffer) 42.4 (10.5) 60.1 (8.1) 0.85 -0.53 *†Riparian agriculture (% in 30-m buffer) 8.6 (4.3) 26.6 (4.4) 0.44 -0.90 Riparian developed (% in 30-m buffer) 20.4 (5.3) 13.2 (1.8) -0.35 0.94 *Riparian forested (% in 30-m buffer) 71.0 (6.5) 60.2 (4.0) -0.46 0.89 Natural catchment features *†Mean catchment elevation (m above sea level) 971.1 (21.5) 784.3 (12.7) -0.99 0.15 *†Catchment area (km2) 159.8 (33.3) 252.8 (45.3) 0.55 0.83 *Carbonate geology (%) 5.5 (2.5) 25.8 (4.4) 0.98 -0.22 *Clastic geology (%) 55.4 (5.7) 21.7 (3.3) -0.99 0.14 Shale geology (%) 38.5 (4.7) 52.4 (3.9) 0.98 0.19 Catchment land use *†Catchment agriculture (%) 8.6 (2.4) 28.8 (3.7) 0.98 -0.20 *Catchment developed (%) 3.0 (0.3) 4.8 (0.5) 0.93 -0.36 *Catchment forested (%) 86.3 (2.5) 65.6 (4.0) -0.98 0.22 *9DULDEOHVH[SODLQLQJVLJQLILFDQW S” YDULDWLRQLQLQVWUHDPKDELWDWYLDa permutation test in redundancy analysis (RDA). †Variables retained for partial RDA following stepwise selection of variables for each variable group.

90 TABLE 3.2. Akaike’s Information Criterion (AICc) for best-supported combinations of hypotheses explaining the distribution of E. osburni. ǻ$,&F is the difference between the top- ranked model and lower-ranked models (i). Model weight (Wi) is the probability of a model being the best-supported model. Evidence ratio (W1/Wi) is the number of times the top-ranked model is better supported over lower-ranked models. Area under the-curve (AUC) is a threshold- independent measure of cross-validation Supplementary Table S3.8 contains all 141 models.

Rank Hypothesis Model components LL ǻAICc Wi W1/Wi AUC 1 Spa., Sub., Temp. SP1, SP2, Emb, SPMDT -7.1 0.00 0.140 1 0.96 8 Spa., Sub. SP1, SP2, Emb. -9.7 2.55 0.039 4 0.91 17 Spa., Temp. SP1, SP2, SPMDT -10.6 4.45 0.015 9 0.93 24 Sub., Temp. Silt, SPMDT -12.4 5.62 0.008 17 0.93 61 Temp. SPMDT, SMDMX -14.2 6.45 0.006 25 0.88 75 Spa. SP1, SP2 -15.2 11.11 0.001 259 0.86 106 Sub. Emb. -21.1 20.61 0.000 >1000 0.79 126 Null Intercept-only (null) -27.9 32.02 0.000 >1000 0.50 Spa = spatial, Sub. = substrate, Temp. = temperature,. SP = spatial covariate, Emb. = embeddedness, SPMDT = spring mean daily temperature, SMDMX = summer mean daily maximum temperature, LL = log likelihood.

91 FIGURE 3.1. A multi-level framework depicting indirect pathways of regional disturbance (black arrows) on a local ecological state (persistence versus extirpation of a sensitive stream fish in a stream segment). Environmental features (rectangles) at catchment and segment scales interact with regional disturbances (stimuli) to either decrease (left) or increase (right) the resilience of a sensitive stream fish population. (a) Sensitive catchment- and segment-scale environmental features propagate the influence of a moderate land-use disturbance leading to extirpation (i.e., land-cover cascade pathway). (b) Land-use disturbance has little influence on a sensitive species due to mitigating influences of highly resilient catchment and nested segment features (resilient-catchment pathway). (c) Land-use disturbance is first propagated by sensitive catchment features and then mitigated by resilient segment features leading to restricted persistence within degraded catchments (resilient-segment pathway). (d) High meta-population connectivity at a catchment scale enables persistence in sites with both sensitive catchment and segment features. (e) Absence of meta-population connectivity prevents recolonization of a segment with suitable catchment and segment features.

92 FIGURE 3.2. Locations of stream segments (study sites) sampled for E. osburni in 2012 within the New River drainage, Virginia and West Virginia.

93 FIGURE 3.3. Principal component analysis (PCA) of instream habitat within study segments (plotted points). Circle size corresponds to observed E. osburni densities (fish /100 m2) within segments. Axes 1 (horizontal) and 2 (vertical) explained 43.2% and 24.8% of variation, respectively. Arrays represent permutation-based correlation coefficients between axes and habitat variables. (a) Instream habitat; (b) Natural catchment features, (c) Catchment land use; (d) Segment features. SPMDT = spring mean daily stream temperature, SMDMX = summer mean daily maximum stream temperature.

94 FIGURE 3.4. Venn diagram showing partitioned adjusted variation (by percentage) in instream habitat by catchment land use, natural catchment-, and segment-scale features via partial redundancy analysis. “Direct catchment” are direct pathways of catchment-generated influence on instream habitat. “Indirect propagating” are pathways where catchment-generated phenomena constrain segment features and instream habitat. “Unique segment” is variation explained at the segment-scale not explained at the catchment scale and represent resilient-segment or sensitive- segment pathways. Percentages sum to explain 52% (= adjusted R2) of instream-habitat variation.

95 FIGURE 3.5. Predicted probabilities, with 90% confidence intervals, of presence/persistence across temperature (a) and embeddedness (b) gradients. Embeddedness index (0–4) was converted to percentages (0–100%). Presences and absences were plotted as 1 and 0, respectively. Black circles are segments where E. osburni were historically confirmed present. White circles are segments where E. osburni was not historically documented by sparse early surveys in the region. Panels c–d demonstrate influence of spatial location on probabilities across temperature (c) and embeddedness (d) gradients at the 75th (“Favorable” spatial context), 50th (“neutral” spatial context), and 25th (“poor” spatial context) quartiles of spatial covariates 1 and 2.

96 CHAPTER 4: General Conclusions and Management Recommendations

My main goal in this thesis was to address several knowledge gaps that historically impeded the conservation of the Candy Darter (Etheostoma osburni) – a well known, yet not well-studied Appalachian stream fish. At the inception of this thesis there was considerable uncertainty about the distribution of the Candy Darter, and little was known about the species’ habitat preferences and threats. Although the patterns revealed in this thesis are correlative and not exhaustive, they support two a priori hypothesized stressors impacting the species – namely warm water and fine sediment. Moreover, it is increasingly apparent that these same stressors imperil highland stream biota worldwide and especially the highland stream fishes endemic to southeastern United States (Walters et al. 2003; Scott 2006). Consequently, similar approaches may be useful for generating knowledge beneficial for the conservation of other imperiled, yet data-deficient, species. Little is know about the habitat requirements of the vast majority of the fish species inhabiting the southeastern United States. Rather than observations aimed at both individual and population levels, existing knowledge generally consists of a patchwork of non-standardized observations of individual habitat use over limited spatio-temporal extents. This scenario results from the low priority historically given to the overwhelming richness of non-game fishes inhabiting the region. Many of these data-deficient species are rapidly declining, leaving managers – tasked with managing populations – overly reliant on limited knowledge typically derived from snapshots of observed adult microhabitat use. This creates multiple conservation challenges. First, habitat selection is a dynamic process, varying through ontogeny, and with spatio-temporal context (Schlosser 1991), and it is unlikely that studies with narrowly defined spatial and temporal extents can capture this complexity (Angermeier 1987; Fausch et al. 2002). Second, habitat studies rarely link the availability of preferred habitats used by individuals to population-level responses across streams (Rosenfeld 2003). Processes that dictate individual habitat selection and those that regulate populations may be fundamentally different and operate at different spatio-temporal scales. Including non-critical habitats selected by individuals during non-limiting periods or life-stages in the guiding image of suitable habitat risks valuing spurious environmental variables in population-focused management plans.

97 In Chapter 2, I used the Candy Darter – a previously described micro- “habitat specialist” (Chipps et al. 1994) – as a model to demonstrate the importance of validating observed microhabitat selection across independent populations. Although microhabitat selection by adults was consistent with existing descriptions (i.e., swift, shallow, structurally complex microhabitats), sub-adult life stages selected distinct microhabitats, including riffle margins and run channel units. More importantly, microhabitat suitability models – meant to identify suitable habitat for populations – revealed contradictory results among individual habitat variables when applied to available habitat across four streams; suitable substrates and flows were positively and negatively correlated with population robustness, respectively. Seasonal comparisons provided further perspective; in fall when the streams supporting the most robust populations had little flow (i.e., depths and velocities), older life stages responded by congregating in the most suitable, albeit sub-optimal, microhabitat patches available. These observations demonstrated the fall flows selected by adults were substitutable resources, whereas other microhabitat variables, such as embeddedness, may be more impactful for populations. For managers tasked with recovering species, these findings suggest the limited resources typically afforded to non-game species should be invested into efforts that uncover population-level distributional patterns including correlates of occupancy, abundance, and most importantly, population vital rates. Limited, yet intensive, microhabitat investigations can still be used as supplements, providing the necessary resolution to generate hypotheses about specific incompatibilities between a species’ life history and available habitat, or as less-invasive methods for surveying sites within population-focused study designs. Conclusions in Chapter 2 could be bolstered by recording the dynamics of more precise fitness metrics over longer periods. For example, were sub-optimal fall flows within streams with robust populations actually stressful for adults, and was this temporary stress offset by the benefits of increased nursery habitat for the large age-0 year classes observed in both streams with robust populations? Do low flows in fall increase growth and survival of age-0 fish by reducing the number of large predators and increasing productivity (Schlosser 1987)? These are a few of the many key questions that could be addressed with annually collected abundance data, yet time-series data are often woefully rare for the vast majority of fish populations. There are still unknown aspects of the life-history of the Candy Darter, including conditions necessary for the survival of egg and larval stages, which often are critical bottlenecks

98 governing stream-fish populations (Schlosser 1985). Moreover, my early cursory snorkeling efforts indicate Candy Darters over-winter in deeper areas with slow water velocities under coarse substrates. If so, is the Candy Darter always a flow specialist, and is winter refuge habitat also limited by high embeddedness within formerly occupied streams? Despite my field team’s best efforts to be both spatially intensive and extensive, Chapter 2 was still a snapshot of the dynamics within each stream taken over portions of a single year. Rather than solely specializing on specific microhabitat patches, I suspect the Candy Darter interacts with most stream areas throughout its life history, and landscape-generated stressors chronically impact performance during multiple stages and behavioral modes. Clearly, there are many questions that need to be resolved about the relationships between individual habitat selection and the dynamics of remaining populations. Chapter 3 confirmed that remaining Candy Darter populations are highly localized, and restricted to stream- and segment-scale refugia that sharply contrast with areas where populations are extirpated. Refugia were defined by low fine-sediment levels, and cool stream temperatures. Natural landscape features – namely sandstone geology, high elevations, and small stream sizes – imparted ecological resilience to land-use disturbance primarily generated by upslope and near- stream agriculture. Within highland regions, warm, turbid, and embedded stream conditions typically result from the conversion of upslope forests to less-natural land covers (Burkhead 1997; Jones et al. 1999; Scott et al. 2002). Many highland fishes are pre-disposed to imperilment in these degraded conditions due to possessing sensitive traits, including stenothermy, insectivory, and substrate and visual dependencies (Berkman and Rabeni 1987; Angermeier 1995; Etnier 1997). My cursory inspection of historical U.S. Geological Survey topographic maps indicates much of the middle New River drainage (NRD) has reforested since the early 1900s (Figures 4.1 and 4.2) – a trend throughout Appalachia (Brown et al. 2005). However, despite this land-use trend, at least 14 of the 23 (61%) stream segments in the VR where I failed to detect Candy Darters, were also accessible to cattle (my personal observations), indicating known sources of habitat degradation are still prevalent. Effective approaches for remedying both elevated fine-sediment levels and stream temperatures include watershed reforestation, and re-stabilizing stream banks by excluding livestock from riparian areas. These measures would not only benefit the Candy Darter, but also benefit other sensitive species including most of the cool-water preferring

99 endemic fishes and crayfishes of the NRD (Jenkins and Burkehad 1994:84–85), the imperiled Eastern Hellbender (Cryptobranchus alleganiensis; Jachowski et al. 2016), and the Appalachian flagship species, the Brook Trout (Salvelinus fontinalis; Hudy et al. 2008). Recovery of the Candy Darter will require both preserving refugia where the species persists and reestablishing populations in segments where the species is extirpated. Delineating the genetic structure of remaining populations in the southern range will be a necessary first step for restoring populations. For example, is the newly discovered population in Cripple Creek genetically distinct? Although not well studied, lower-elevation- or southern-edge populations are more thermally tolerant in other species (Hampe and Petit 2005; Whitney et al. 2016). Thermal tolerance would be a particularly beneficial trait possessed by a source stock used to restore Candy Darter populations in what will likely be a warmer future NRD. Alternatively, there may be little genetic structuring among remaining southern populations, indicating Candy Darters recently (pre-European settlement) traversed the mainstem New River. This finding would dramatically expand the existing conceptualization of what was suitable habitat for the Candy Darter. Other beneficial research needs include clarifying the cumulative effects of fine sediment and warm temperatures on specific life stages, improving sampling methods, and validating range-wide species distribution models (Huang 2015) to identify areas with potential to support unknown current or future populations. Management actions to conserve the Candy Darter will need to be made in light of the introduced Variegate Darter (E. variatum). An initial idea for this thesis was to examine “niche” overlap between native Candy Darters and introduced Variegate Darters, the latter of which had recently invaded Anthony Creek, WV (Switzer et al. 2007). This idea was quickly abandoned when the only saddle darters that I collected in Anthony Creek resembled the Variegate Darter. Variegate Darters have since invaded nearly all remaining Greenbrier River tributaries (Gibson 2017), including many tributaries supporting the highest Candy Darter densities (Dunn 2013). Candy and Variegate darters (my unpublished data from Elk River, WV) select similar habitats and demonstrate comparable ontogenetic shifts. An implication is that efforts to restore habitat for the Candy Darter could be undermined by also benefitting the Variegate Darter. In contrast to the Candy Darter, the rapid spread of Variegate Darters along the courses of the New, Greenbrier, and Gauley rivers indicates the Variegate Darter thrives in large rivers. This pattern is not surprising given its broad distribution throughout the Ohio River basin

100 (Switzer 2004). Continued spread of Variegate Darters may eventually lead to complementary distributions in the NRD, with Candy and Variegate darters being longitudinally partitioned into tributary streams and mainstem rivers, respectively. Given the two species readily hybridize, short-term containment of the Variegate Darter combined with long-term isolation of the two species will be essential for the long-term viability of the Candy Darter. It is highly likely Variegate Darters will circumvent the reservoirs currently impeding their expansion throughout the Gauley and New rivers. For example, other introduced bait-fishes in the NRD have circumvented reservoirs likely via secondary bait-bucket introductions (Buckwalter 2016). This scenario would closely resemble the distribution of Cutthroat Trout (Oncorhynchus clarkii) in high-elevation portions of the Rocky Mountains, isolated from downstream introduced Brook and Rainbow (O. mykiss) trout at lower elevations (Fausch et al. 2009). Similar strategies for conserving Cutthroat Trout may be transferable to the NRD and include selectively reintroducing Candy Darters and removing Variegate Darters from invaded tributaries. Additionally, legacy milldams may be strategically fortified or removed, and new barriers near the downriver distributional limits of Candy Darter populations may need to be constructed. In conclusion, readers should view this thesis as an investigation of a sentinel species inhabiting a sensitive river drainage exposed to multiple common stressors afflicting highland fishes and areas worldwide. I integrated the existing knowledge of Candy Darter life history and expanded the conceptualization of habitat requirements by purposefully conducting studies at multiple resolutions and at large spatial extents. The main findings are that although adults – especially charismatic males – are appropriately characterized as microhabitat specialists from spring through fall, the Candy Darter uses a variety of stream habitats throughout its life history. Like other southeastern highland endemics, Candy Darter populations persisted in segments with cool stream temperatures and low fine-sediment in largely forested watersheds. Future efforts to conserve the Candy Darter will undoubtedly be complicated by the Variegate Darter. Strategies that isolate Variegate Darters from Candy Darters will likely be necessary; however, strategic isolation can easily be undermined by the inadvertent release of Variegate Darters – a species with a very long history as a bait-fish (Trautman 1981:667). Consequently, in addition to managing symptoms, effective conservation of stream biota will require proactively addressing larger societal causes of imperilment and include minimizing the impacts of harmful land-use

101 practices, slowing the rate of warming stream temperatures, and better regulating the transport and release of non-native species.

References Angermeier, P. 1987. Spatiotemporal variation in habitat selection by fishes in small Illinois streams. Pages 52–60 in W.J. Matthews and D.C. Heins, editors. Community and evolutionary ecology of North American stream fishes. University of Press, Norman.

Angermeier, P. 1995. Ecological attributes of extinction-prone species: loss of freshwater fishes of Virginia. Conservation Biology 9(1):143–158.

Berkman, H.E., and C.F. Rabeni. 1987. Effect of siltation on stream fish communities. Environmental Biology of Fishes 18(4):285–294.

Brown, D.G., K.M. Johnson, T.R. Loveland, and D.M. Theobald. 2005. Rural land-use trends in the conterminous United States, 1950–2000. Ecological Applications 15(6):1851–1863.

Buckwalter, J. D. 2016. Temporal trends in stream-fish distributions, and species traits as invasiveness drivers in New River (USA) tributaries. Master’s thesis. Virginia Polytechnic Institute and State University, Blacksburg, Virginia.

Burkhead, N., S. Walsh, B. Freeman, and J. Williams. 1997. Status and restoration of the Etowah River, an imperiled southern Appalachian ecosystem. Pages 375–444 in G.W. Benz, and D.E. Collins, editors. Aquatic fauna in peril: the southeastern perspecive. Southeast Aquatic Research Institute, Lenz Design and Communications, Decatur, Georgia.

Chipps, S. R., W. B. Perry, and S. A. Perry. 1994. Patterns of microhabitat use among four species of darters in three Appalachian streams. American Midland Naturalist 131(1):175–180.

102 Dunn, C.G. 2013. Comparison of habitat suitability among sites supporting strong, localized, and extirpated populations of candy darter (Etheostoma osburni). Final report to the Virginia Department of Game and Inland Fisheries, Richmond, Virginia.

Etnier, D.A. 1997. Jeopardized southeastern freshwater fishes: a search for causes. Pages 87–104 in G.W. Benz, and D.E. Collins, editors. Aquatic fauna in peril: the southeastern perspecive. Southeast Aquatic Research Institute, Lenz Design and Communications, Decatur, Georgia.

Fausch, K.D., B.E. Rieman, J.B. Dunham, M.K. Young, and D.P. Peterson. 2009. Invasion versus isolation: trade-offs in managing native salmonids with barriers to upstream movement. Conservation Biology 23(4):859–870.

Fausch, K.D., C.E. Torgersen, C.V. Baxter, and H.W. Li. 2002. Landscapes to riverscapes: bridging the gap between research and conservation of stream fishes. Bioscience 52(6):483–498.

Gibson, I. 2017. Conservation concerns for the Candy Darter (Etheostoma osburni) with implications related to hybridization. Master’s thesis. West Virginia University, Morgantown.

Hampe, A., and R.J. Petit. 2005. Conserving biodiversity under climate change: the rear edge matters. Ecology Letters 8(5):461–467.

Huang, J. 2015. Assessing predictive performance and transferability of species distribution models for freshwater fish in the United States. Doctoral dissertation. Virginia Tech, Blacksburg.

Hudy, M., T.M. Thieling, N. Gillespie, and E.P. Smith. 2008. Distribution, status, and land use characteristics of subwatersheds within the native range of brook trout in the eastern United States. North American Journal of Fisheries Management 28(4):1069–1085.

103 Jachowski, B., M. Catherine, J.J. Millspaugh, and W.A. Hopkins. 2016. Current land use is a poor predictor of hellbender occurrence: why assumptions matter when predicting distributions of data-deficient species. Diversity and Distributions 22(8):865–880.

Jenkins, R.E., and N.M. Burkhead. 1994. Freshwater fishes of Virginia. American Fisheries Society, Bethesda, Maryland.

Jones, E.B.D., G.S. Helfman, J.O. Harper, and P.V. Bolstad. 1999. Effects of riparian forest removal on fish assemblages in southern Appalachian streams. Conservation Biology 13(6):1454–1465.

Rosenfeld, J. 2003. Assessing the habitat requirements of stream fishes: an overview and evaluation of different approaches. Transactions of the American Fisheries Society 132(5):953–968.

Schlosser, I.J. 1985. Flow regime, juvenile abundance, and the assemblage structure of stream fishes. Ecology 66(5):1484–1490.

Schlosser, I.J. 1991. Stream fish ecology: a landscape perspective. Bioscience 41(10):704–712.

Scott, M.C. 2006. Winners and losers among stream fishes in relation to land use legacies and urban development in the southeastern US. Biological Conservation 127(3):301–309.

Scott, M.C., G.S. Helfman, M.E. McTammany, E.F. Benfield, and P.V. Bolstad. 2002. Multiscale influences on physical and chemical stream conditions across Blue Ridge landscapes. Journal of the American Water Resources Association 38(5):1379–1392.

Sclosser, I. J. 1987. A conceptual framework for fish communities in small warmwater streams. Pages 17–24 in W.J. Matthews and D.C. Heins, editors. Community and evolutionary ecology of North American stream fishes. University of Oklahoma Press, Norman.

104 Switzer, J. F. 2004. Molecular systematics and phylogeography of the Etheostoma variatum species group (Actinopterygii: Percidae). Ph.D. dissertation. Saint Louis University, St. Louis, Missouri.

Switzer, J.F., S.A. Welsh, and T.L. King. 2007. A molecular genetic investigationof hybridization between Etheostoma osburni and Etheostoma variatum in the New River drainage, West Virginia. West Virginia Division of Natural Resources, Elkins, West Virginia.

Trautman, M.B. 1981. The Fishes of Ohio. the Ohio State University Press, Columbus.

Walters, D.M., D.S. Leigh, and A.B. Bearden. 2003. Urbanization, sedimentation, and the homogenization of fish assemblages in the Etowah River basin, USA. Hydrobiologia 494(1-3):5–10.

Whitney, J.E., and R. Al-Chokhachy, D.B. Bunnell, C.A. Caldwell, S.J. Cooke, E.J. Eliason, M. Rogers, A.J. Lynch, and C.P. Paukert. 2016. Physiological basis of climate change impacts on North American inland fishes. Fisheries 41(7):332–345.

105 Figure 4.1. Topographic map of middle New River drainage (U.S. Geological Survey; Pearisburg quadrangle) in 1932. The valleys of Spruce Run, Sinking, and Walker creeks are largely denuded. Mountaintops, hillslopes, and the confined valleys of Big and Little Stony creeks are still depicted as forested (green), yet these areas would have been early successional forests given these areas were clear-cut between1890 – 1920. Altogether, the middle New River drainage would have been sparsely forested in the early 20th century. Stream labels were superimposed and not a part of the original map. Note land cover is not depicted for West Virginia portions of the map (upper-left).

106 Figure 4.2. Land use from the 2011 National Land Cover Dataset corresponding to the area depicted by the 1932 topographic map in Figure 4.1. Despite the high prevalence of pasture (yellow) in the Valley and Ridge province, portions of valleys have seemingly re-vegetated after 80 years. blue = streams or rivers, green = forest, red = developed, yellow = pasture.

107 SUPPLEMENTARY MATERIAL

Supplementary Table S2.1. Previous accounts of in-stream habitat associations of Candy Darters. “Predicted relation” refers to expected correlations between darter density and a habitat gradient interpreted from references listed below (e.g., adult densities increase as water velocity increases but decrease as water depth increases [Kuehne and Barbour 1983]). Predicted relations were not applicable (NA) if information was not available for variables at specific life stages.

Predicted Variable Life stage Habitat description relation 20.3–50.8 cm (Addair 1944); 20.4–29.6 cm, "shallow Adult water" (Chipps et al. 1994)*; 40–100 cm (Jenkins and - Burkhead 1994); < 22 cm, "shallow water" (Leftwich et Water depth al. 1996) Juvenile No information NA Age-0 No information NA “... swift clear water...” (Addair 1944); 27.4–32.9 cm/ sec, "fast water" (Chipps et al. 1994)*; "Rocky Adult + montane...with turbulent flow" (Kuehne and Barbour Water velocity 1983); "fast water" (Leftwich et al. 1996) Juvenile "...slower current than adults" (Jenkins and Kopia 1995) + Age-0 No information NA “... large rocks and gravel...” (Addair 1944); “Rubble" (Kuehne and Barbour 1983); Cobble (Chipps et al. 1994)*; “Rubble and boulder in runs and riffles” Adult + (Jenkins and Burkhead 1994); "gravel, cobble, boulder, Substrate size and occasional bedrock" (Jenkins and Kopia 1995); “boulder” (Leftwich et al. 1996) Juvenile No information NA Age-0 No information NA Adult No information NA Embeddedness Juvenile No information NA Age-0 No Information NA "...relatively silt-free streams." (Jenkins and Kopia 1995) “Excessive siltation characterized areas where the candy Adult - darter was absent or much diminished.”(Chipps et al. Silt cover 1993*) Juvenile No information NA Age-0 No information NA *Indicates primary literature

108 Supplementary Table S2.2. Pearson correlation coefficients (r) between in-stream microhabitat variables and non-metric multidimensional scaling (NMDS) axes for two seasons (Figure 2.4; Supplementary Figure S2 &RHIILFLHQWV•__DUHSUHVHQWHGLQ)LJXUH2.4, Supplementary Figure S2.2.

Axis 1 Axis 2 Axis 1 Axis 2 Variable (spring) (spring) (fall) (fall) Depth 0.40 0.92 0.96 -0.29 Velocity -0.62 0.79 -0.70 0.72 Substrate size -0.77 -0.64 -0.37 0.93 Embeddedness 0.95 0.30 0.93 0.37 Silt cover 0.91 -0.42 0.93 0.37

109 Supplementary Table S2.3. Observations by life stage (N), area sampled, and density of Candy Darters in three streams and two seasons. Stream abbreviations: EFG = East Fork Greenbrier River, LC = Laurel Creek, SFC = South Fork Cherry River.

Life Area sampled Fish density Season Stream Status stage N (m2) (fish/100 m2) Spring EFG Robust Total 115 13059 0.88 Adult 66 Juvenile 31 Age-0 18 SFC Robust Total 175 14614 1.20 Adult 71 Juvenile 54 Age-0 50 LC Localized Total 14 5455 0.26 Adult 11 Juvenile 3 Age-0 0 Fall EFG Robust Total 286 8764 3.26 Adult 69 Juvenile 87 Age-0 130 SFC Robust Total 222 11480 1.93 Adult 80 Juvenile 37 Age-0 105 LC Localized Total 52 4383 1.19 Adult 19 Juvenile 11 Age-0 22

110 Supplementary Table S2.4. Predicted individual suitability (X) and 95% confidence intervals (CI95) by life stage and season in four streams that vary in population status of Candy Darter. Possible suitability values range from 0 to 1, which indicate “no selection” and “maximum selection”, respectively. “Multi-Stage” is the average suitability across life stages. “Multi- variable” is the average suitability calculated from all habitat variables within each stream. East Fork Greenbrier (EFG) and South Fork Cherry rivers (SFC) support robust populations, Laurel Creek (LC) supports a localized population, and Candy Darters are extirpated from Sinking Creek (SC).

Adult Juvenile Age-0 Multi-Stage Season Variable Stream suitability suitability suitability suitability X (CI95) X (CI95) X (CI95) X (CI95) Spring Depth EFG 0.54 (0.03) 0.66 (0.02) 0.61 (0.03) 0.60 (0.02) SFC 0.49 (0.02) 0.63 (0.03) 0.57 (0.03) 0.56 (0.02) LC 0.50 (0.03) 0.62 (0.03) 0.56 (0.03) 0.56 (0.03) SC 0.50 (0.03) 0.64 (0.03) 0.52 (0.03) 0.55 (0.03) Velocity EFG 0.18 (0.01) 0.51 (0.02) 0.88 (0.01) 0.53 (0.01) SFC 0.15 (0.01) 0.42 (0.01) 0.88 (0.01) 0.48 (0.01) LC 0.16 (0.01) 0.46 (0.02) 0.89 (0.01) 0.50 (0.01) SC 0.19 (0.01) 0.51 (0.02) 0.88 (0.02) 0.53 (0.01) Substrate EFG 0.80 (0.02) 0.72 (0.02) 0.69 (0.01) 0.74 (0.01) SFC 0.67 (0.02) 0.65 (0.02) 0.70 (0.01) 0.67 (0.01) LC 0.57 (0.03) 0.54 (0.03) 0.75 (0.02) 0.62 (0.02) SC 0.59 (0.03) 0.55 (0.02) 0.77 (0.02) 0.64 (0.01) Embeddedness EFG 0.84 (0.02) 0.90 (0.02) 0.91 (0.01) 0.88 (0.01) SFC 0.88 (0.02) 0.90 (0.02) 0.92 (0.02) 0.90 (0.02) LC 0.52 (0.04) 0.60 (0.04) 0.66 (0.03) 0.59 (0.03) SC 0.33 (0.02) 0.41 (0.03) 0.54 (0.02) 0.43 (0.02) Silt cover EFG 0.44 (0.03) 0.67 (0.02) 0.82 (0.02) 0.64 (0.02) SFC 0.58 (0.03) 0.75 (0.02) 0.76 (0.01) 0.70 (0.02) LC 0.53 (0.04) 0.65 (0.04) 0.62 (0.03) 0.60 (0.03) SC 0.57 (0.04) 0.72 (0.03) 0.72 (0.02) 0.67 (0.02) Multi-variable EFG 0.56 (0.01) 0.69 (0.01) 0.78 (0.01) 0.68 (0.01) SFC 0.55 (0.01) 0.67 (0.01) 0.77 (0.01) 0.66 (0.02) LC 0.46 (0.02) 0.57 (0.02) 0.70 (0.01) 0.58 (0.02) SC 0.44 (0.02) 0.57 (0.02) 0.69 (0.01) 0.56 (0.01) Fall Depth EFG 0.52 (0.05) 0.49 (0.05) 0.51 (0.05) 0.51 (0.05) SFC 0.68 (0.04) 0.62 (0.05) 0.65 (0.04) 0.65 (0.04) LC 0.66 (0.03) 0.61 (0.04) 0.64 (0.03) 0.63 (0.03) SC 0.68 (0.04) 0.49 (0.04) 0.58 (0.03) 0.58 (0.03) Velocity EFG 0.16 (0.02) 0.17 (0.02) 0.33 (0.02) 0.22 (0.02) SFC 0.20 (0.02) 0.22 (0.03) 0.36 (0.02) 0.26 (0.02) LC 0.20 (0.02) 0.23 (0.02) 0.37 (0.02) 0.27 (0.09) SC 0.21 (0.02) 0.22 (0.02) 0.36 (0.01) 0.26 (0.02)

111 Supplementary Table S2.4 continued Substrate EFG 0.86 (0.03) 0.43 (0.02) 0.42 (0.02) 0.57 (0.02) SFC 0.76 (0.03) 0.38 (0.02) 0.41 (0.02) 0.51 (0.02) LC 0.73 (0.03) 0.44 (0.02) 0.47 (0.02) 0.55 (0.02) SC 0.73 (0.02) 0.49 (0.02) 0.49 (0.02) 0.57 (0.02) Embeddedness EFG 0.76 (0.05) 0.80 (0.04) 0.87 (0.03) 0.81 (0.04) SFC 0.78 (0.04) 0.80 (0.04) 0.86 (0.03) 0.81 (0.04) LC 0.53 (0.04) 0.58 (0.03) 0.73 (0.02) 0.61 (0.03) SC 0.25 (0.02) 0.35 (0.02) 0.58 (0.02) 0.40 (0.02) Silt cover EFG 0.34 (0.06) 0.37 (0.05) 0.59 (0.05) 0.43 (0.05) SFC 0.46 (0.05) 0.48 (0.05) 0.70 (0.04) 0.55 (0.04) LC 0.47 (0.04) 0.50 (0.04) 0.67 (0.03) 0.55 (0.04) SC 0.32 (0.04) 0.35 (0.03) 0.53 (0.03) 0.40 (0.02) Multi-variable EFG 0.53 (0.02) 0.45 (0.02) 0.54 (0.02) 0.51 (0.02) SFC 0.57 (0.02) 0.50 (0.02) 0.60 (0.02) 0.56 (0.02) LC 0.52 (0.02) 0.47 (0.02) 0.58 (0.01) 0.52 (0.02) SC 0.44 (0.02) 0.38 (0.02) 0.51 (0.01) 0.44 (0.01)

112 Supplementary Figure S2.5. Length-frequency histogram of total lengths from Candy Darters (N = 798 individuals) measured while snorkeling in the East Fork Greenbrier and South Fork Cherry rivers, WV. We used different thresholds for separating juveniles from adult females (60 mm) and males (65 mm) based on differences in pigmentation. The line between adults and juveniles is drawn at 62.5 mm.

113 Supplementary Figure S2.6. Non-metric multidimensional scaling (NMDS) plots of habitat use, availability, and suitability in fall. A) Habitat use by three life stages and availability in four streams (polygons). Predicted microhabitat suitability by (B) Adults, (C) Juveniles, and (D) Age- 0. Symbols for "LC use" are locations used by Candy Darters in Laurel Creek in fall. Variables highly correlated (Pearson coefficient [r@• ZLWKD[HVDUHVKRZQ$OOFRUUHODWLRQFRHIILFLHQWV are in Supplementary Table S2.2. Stream abbreviations: EFG = East Fork Greenbrier River, LC = Laurel Creek, SC = Sinking Creek, SFC = South Fork Cherry River. NMDS stress = 0.14.

114 Supplementary References S2.7.

Addair, J. 1944. The fishes of the Kanawha River system in West Virginia and some factors which influence their distribution. Doctoral dissertation. Ohio State University, Columbus.

Chipps, S. R., W. B. Perry, and S. A. Perry. 1993. Status and distribution of Phenacobius teretulus, Etheostoma osburni, and “Rhinichthys bowersi” in the Monongahela National Forest, West Virginia. Virginia Journal of Science 44:48–58.

Chipps, S. R., W. B. Perry, and S. A. Perry. 1994. Patterns of microhabitat use among four species of darters in three Appalachian streams. American Midland Naturalist 131:175– 180.

Jenkins, R. E., and N. M. Burkhead. 1994. Freshwater fishes of Virginia. American Fisheries Society, Bethesda, Maryland.

Jenkins, R. E., and B. L. Kopia. 1995. Population status of the Candy Darter, Etheostoma osburni, in Virginia 1994–1995, with historical review. Roanoke College, Department of Biology, Final Report, Salem, Virginia.

Kuehne, R. A., and R. W. Barbour. 1983. The American darters. University Press of Kentucky, Lexington.

Leftwich, K. N., C. A. Dolloff, M. K. Underwood, and M. Hudy. 1996. The Candy Darter (Etheostoma osburni) in Stony Creek, George Washington–Jefferson National Forest, Virginia: trout predation, distribution, and habitat associations. U.S. Forest Service, Final Report, Blacksburg, Virginia.

115 Supplementary Table S3.1. References used to identify streams and segments (“Sites”). “Recent” streams have recent collections of E. osburni. “Historical” streams have historical records but none within 40 years before this study. “No record” streams lack records.

Physio- Site U.S. graphic Stream code state province Sites MRR Source Recent Big Stony Creek BGS VA VR 1 2010 Dunn unpublished data East Fork Greenbrier River EFG WV VR 1 2011 Dunn & Angermeier, 2016 Cherry River CHR WV AP 1 1991 Chipps, 1992 Deer Creek DER WV VR 1 2006 Burns, 2007 Dismal Creek DIS VA VR 1 1997 Bye, 1997 Gauley River GAL WV AP 1 1976 Hocutt et al., 1979 Knapp Creek KNP WV VR 1 2006 Burns, 2007 Laurel Creek LVA VA VR 1 2011 Dunn & Angermeier, 2016 Laurel Creek LWV WV AP 1 1991 Chipps, 1992 Siltington Creek SIL WV VR 1 2006 Burns, 2007 South Fork Cherry River SFC WV AP 1 2011 Dunn & Angermeier, 2016 West Fork Greenbrier River WFG WV AP 1 2010 Dunn unpublished data Williams River WIL WV AP 1 1991 Chipps, 1992 Historical Indian Creek IND WV AP 2 c.1935 Addair, 1944 Pine Run PNR VA VR 2 1954 Jenkins & Kopia, 1995 Reed Creek RED VA VR 2 1957 Jenkins & Kopia, 1995 Second Creek SEC WV AP/VR 2 c.1935 Addair, 1944 Sinking Creek SNK VA VR 2 c.1940 Burton & Odum, 1945 South Fork Reed Creek SFR VA VR 2 1885 Jenkins & Kopia, 1995 Stony Creek STN WV AP 1 1931 Addair, 1944 Walker Creek WAL VA VR 2 1963 Jenkins & Kopia, 1995 No record Clear Creek CLR VA VR 2 – – Cripple Creek CRP VA VR 2 – – East River EST VA/WV AP 2 – – Kimberling Creek KIM VA VR 2 – – Little Walker Creek LWA VA VR 2 – – No Business Creek NOB VA VR 2 – – Wolf Creek WLF VA VR 2 – – VA = Virginia, WV = West Virginia, VR = Valley and Ridge, AP = Appalachia Plateau, MRR = Most recent record.

116 Supplementary Material S3.2. Detailed methods used to predict missing stream temperatures. Prediction of missing temperatures within segments where temperature loggers were displaced Temperature loggers in WIL1 in spring and SFR1 in spring and summer were displaced and temporarily exposed to air. Missing stream temperatures were predicted using a three-step process: model development, validation, and prediction. We used known relationships among the network of segments where temperatures were monitored with temperature loggers to predict missing temperatures. The study timeframe encompassed spring and summer 2012; however, temperatures in several streams were additionally monitored January 2011–June 2013. This expanded timeframe provided independent stream temperature data to relate hourly stream temperatures from WIL1 and SFR1 to hourly stream temperatures from nearby streams with temperature data. We restricted data used for model development to the same season as missing data, but in a separate year, and to nearby segments to minimize error due to variability in climate, geology, and land use. Candidate predictor streams for temperatures in WIL1 included all streams from the Greenbrier and Gauley subbasins. Candidate predictor streams for SFR1 included streams in the Valley and Ridge province, excluding streams in the Greenbrier subbasin. All subsets of candidate streams were fit via first-order linear autoregressive models, which addressed temporal autocorrelation inherent in time series data (Zuur, 2010). Only 7.6% of data from SFR1 in summer 2012 were missing; therefore, we fit models with the remaining 92.4% of temperature data. We retained the model with the lowest Akaike’s information criterion (AIC) for each stream and season for prediction (i.e., one model for WIL1, two models for SFR1). This approach assumed temperature relationships among streams were constant across years; therefore, we used data for WIL1 and SFR1 in spring collected within the study timeframe before displacement or after replacement as fully independent test datasets to validate models. We used leave-one-out cross-validation for SFR1 in summer due to a lack of a fully independent dataset. All models successfully predicted test datasets (R2 •Supplementary Figure S3.3a– c). Finally, given successful validation, missing hourly stream temperatures were predicted and subsequently used in further analyses.

Prediction of missing temperatures within segments without temperature loggers

117 In each stream only one segment contained a temperature logger, leaving one segment per stream without continuously recorded temperatures in streams with multiple segments (hereafter, “intra-stream segments”). Rather than excluding stream segments without temperature data, we developed a model that successfully predicted stream temperatures within segments without temperature loggers from temperatures recorded by logger from corresponding segments within streams. All intra-stream segments were similar in size (see Site selection), thus reducing temperature differences among segments due to discharge and geography. Before sampling fish, we recorded time and stream temperature in a well-mixed area near each riffle within each non- temperature-monitored segment via a handheld temperature probe (hereafter “field-recorded temperature”). We developed mixed-effects linear regression models relating field-recorded temperature to logger-recorded temperatures within intra-stream segments. Candidate predictor variables of field-recorded temperatures included a random effect for stream, an instantaneous logger temperature obtained via linearly interpolating between hourly logger-recorded temperatures to the exact time of each field-recorded temperature observation, and covariates for directional (i.e., upstream versus downstream) network distance between intra-stream segments (NLENGTH), mean upstream catchment area (MCA) of each pair of intra-stream segments, and a NLENGTH-MCA interaction. Model selection followed the procedure recommended by Zuur (2010: 143–160). We first found the optimal random-error structure via restricted maximum likelihood. We retained a random-intercept-and-slope model, which accounted for multiple field-recorded observations from the same stream. Next we fit all subsets of the four fixed terms with the optimal random- error structure via maximum likelihood. The model with the lowest AIC included positive effects for logger temperature (ߚመ = 2.5), NLENGTH (ߚመ = 1.05), and MCA (ߚመ = 0.55). Finally, a high coefficient of determination (R2 = 0.94) between observed and predicted temperatures via leave- one-out cross-validation indicated temperatures in non-temperature-monitored segments could be successfully predicted from logger-recorded temperatures in intra-stream segments (Supplementary Figure S3.3d). The model was then used to predict spring and summer temperatures in segments without loggers and then subsequently used in further analyses.

118 Supplementary Figure S3.3. Relationship between predicted and observed stream temperatures via cross-validation from models developed to predict missing temperature data due to displacement (a–c) or in segments without temperature loggers (d; Supplementary Material S3.2). (a) Predicted stream temperatures for WIL1, March 1–10 and May 29–31, 2012 from models developed from spring 2013 data. (b) Predicted stream temperatures for SFR1, May 2011 from models developed from spring 2013 data. (c) Predicted stream temperatures for SFR1, June 8–August 31, 2012 via leave-one-out cross validation. (d) Relationship between field-recorded temperatures in segments without temperature loggers and predicted temperatures from logger- recorded temperatures, directional network distance between intra-stream segments, and mean catchment area.

119 Supplementary Table S3.4. Means of instream variables within stream segments. Stream codes are explained in Supplementary Table S3.1.

Substrate Stream SPMDT SMDMX Depth Velocity size Embed. Silt code (°C) (°C) (cm) (m/s) (cm) index index BGS1 11.1 20.7 15.0 0.28 12.6 0.5 0.2 CHR1 12.0 24.2 15.7 0.27 17.2 0.4 0.1 CLR1 13.3 23.0 16.0 0.35 7.5 1.0 0.1 CLR2 11.6 20.2 15.5 0.38 7.8 0.8 0.0 CRP1 14.2 23.6 21.5 0.62 6.8 0.3 0.1 CRP2 14.1 22.4 19.5 0.42 7.4 0.8 0.3 DER1 12.1 25.4 8.2 0.22 7.0 0.1 0.3 DIS1 11.0 19.7 12.6 0.14 10.5 0.4 0.1 EFG1 10.8 22.4 15.3 0.35 11.7 0.2 0.2 EST1 14.9 23.9 21.3 0.37 10.2 0.8 0.3 EST2 13.7 23.7 13.0 0.32 6.5 1.1 0.5 GAL1 12.0 22.3 13.3 0.25 6.1 1.3 0.2 IND1 16.8 26.3 13.5 0.21 5.9 1.7 1.1 IND2 14.0 24.6 12.1 0.37 5.7 1.1 0.6 KIM1 13.7 21.3 17.7 0.40 7.3 0.9 0.4 KIM2 13.1 24.2 16.0 0.33 7.5 1.0 0.8 KNP1 11.6 24.1 7.9 0.16 6.1 0.1 0.4 LVA1 12.6 22.2 11.5 0.35 9.8 0.4 0.2 LWA1 13.9 25.0 9.6 0.25 6.7 0.4 0.4 LWA2 12.8 25.2 9.9 0.30 5.9 0.1 0.0 LWV1 10.7 21.0 12.7 0.20 12.8 0.9 0.3 NOB1 12.3 21.3 12.1 0.27 8.8 1.3 0.8 NOB2 11.7 22.0 14.2 0.19 9.4 1.1 1.0 PNR1 14.4 22.4 16.5 0.47 6.6 1.8 1.0 PNR2 14.9 22.4 15.9 0.35 7.5 1.5 1.1 RED1 14.4 22.4 28.9 0.53 6.0 1.6 0.5 RED2 15.5 25.0 25.2 0.49 4.5 2.0 1.0 SEC1 13.0 21.2 11.9 0.30 8.6 0.2 0.3 SEC2 13.3 22.9 13.8 0.24 10.6 0.2 0.3 SFC1 10.8 22.1 14.1 0.23 11.4 0.4 0.3 SFR1 15.4 22.9 14.2 0.41 7.4 1.4 1.2 SFR2 14.7 21.6 15.8 0.43 4.7 1.7 1.4 SIL1 12.7 23.5 13.6 0.36 6.0 0.2 0.2 SNK1 13.8 21.1 15.4 0.47 5.6 0.7 0.1 SNK2 13.6 23.5 18.5 0.51 7.5 0.9 0.2 STN1 10.7 19.6 10.7 0.17 9.3 0.4 0.3 WAL1 14.5 24.1 26.8 0.47 8.8 0.6 0.0

120 Supplementary Table S3.4 continued WAL2 14.3 24.3 28.7 0.53 15.7 0.6 0.0 WFG1 11.2 23.6 13.0 0.21 10.6 0.3 0.4 WIL1 9.6 22.0 10.0 0.21 12.7 0.4 0.4 WLF1 15.8 24.0 21.9 0.51 10.3 0.4 0.0 WLF2 13.8 23.9 21.9 0.41 9.9 0.6 0.1 SPMDT = Spring mean daily temperature; SMDMX = Summer mean daily maximum temperature; Embed. index = Embeddedness index.

121 Supplementary Table S3.5. Segment features including percentage riparian land use (30-m buffer), percentage dominant geologic category (50-m buffer), channel gradient, and channel width. Land use was obtained from the 2011 National Land Cover Dataset (NLCD; Homer et al., 2015) and categorized with a modified Anderson Level 1 classification (Anderson, 1976): Agriculture = NLCD71, NLCD 81, and NLCD 82; Developed = NLCD21, NLCD22, NLCD23, NLCD24, and NLCD31; Forested = NLCD 41, NLCD 42, NLCD 43, NLCD 52, NLCD90, and NLCD95. We categorized geology according to resistance to weathering: Carbonate = dolostone and limestone; Clastic = sandstone and siltstone; Shale = black shale and shale (Dickens et al., 2008; Nicholson et al., 2007). Channel gradient was derived from the National Hydrography Dataset Plus version 2 (McKay et al., 2012). Stream codes are explained in Supplementary Table S3.1.

Channel Channel Stream Agriculture Developed Forested Carbonate Clastic Shale gradient -width code (%) (%) (%) (%) (%) (%) (m/km) (m) BGS1 0.0 0.0 100.0 0.0 100.0 0.0 11.6 6.8 CHR1 0.0 30.0 70.0 0.0 100.0 0.0 7.0 17.3 CLR1 41.6 19.8 38.6 98.6 0.0 1.4 6.1 9.1 CLR2 36.2 28.0 35.8 97.1 0.0 2.9 7.6 5.4 CRP1 12.4 7.6 80.0 0.0 0.0 100.0 8.7 14.5 CRP2 46.2 3.2 50.5 100.0 0.0 0.0 9.9 14.0 DER1 1.6 5.9 92.5 0.0 66.2 14.0 4.7 6.6 DIS1 0.0 10.7 89.3 0.0 100.0 0.0 15.9 3.7 EFG1 4.6 4.6 90.8 0.0 100.0 0.0 5.9 5.0 EST1 5.4 25.3 69.3 28.1 0.0 71.7 8.5 6.8 EST2 17.6 21.1 61.3 67.1 0.0 32.9 5.4 11.2 GAL1 0.0 28.2 71.8 0.0 100.0 0.0 5.4 9.6 IND1 13.5 31.8 54.7 0.0 0.0 100.0 4.8 12.8 IND2 24.7 26.4 49.0 0.0 0.0 100.0 9.3 5.0 KIM1 9.9 10.9 79.2 0.0 38.9 61.1 13.9 12.8 KIM2 45.9 5.4 48.6 0.0 0.0 100.0 5.1 8.2 KNP1 69.0 5.7 25.3 0.0 39.6 60.4 4.8 3.1 LVA1 14.0 77.8 8.2 31.9 37.1 31.1 15.1 3.5 LWA1 44.9 5.3 49.8 0.0 0.0 100.0 3.6 7.3 LWA2 27.3 6.1 66.7 0.0 0.0 100.0 5.1 3.7 LWV1 2.8 22.6 74.5 0.0 25.9 74.1 16.4 9.1 NOB1 73.9 0.0 26.1 0.0 0.0 100.0 5.1 5.0 NOB2 67.4 5.8 26.7 0.0 0.0 100.0 11.3 6.6 PNR1 33.1 7.4 59.6 0.0 0.0 100.0 15.0 5.0 PNR2 30.4 17.0 52.6 0.0 0.0 100.0 8.7 4.0 RED1 4.9 26.4 68.7 21.3 0.0 78.7 2.5 22.5 RED2 14.4 10.5 75.1 40.6 0.0 59.4 4.1 19.3 SEC1 2.7 7.8 89.5 47.4 31.8 20.8 9.2 13.9 SEC2 3.9 17.2 78.9 94.2 0.0 5.8 11.5 7.6

122 Supplementary Table S3.5 continued SFC1 2.4 19.5 78.1 0.0 0.0 100.0 11.3 9.6 SFR1 51.1 4.5 44.4 100.0 0.0 0.0 5.9 6.6 SFR2 69.9 5.2 24.9 100.0 0.0 0.0 4.2 4.8 SIL1 21.8 2.3 75.9 0.0 42.4 57.6 6.2 4.8 SNK1 10.1 10.1 79.8 0.0 0.0 100.0 8.5 6.7 SNK2 3.7 12.6 83.7 0.0 0.0 100.0 8.4 8.8 STN1 1.1 31.9 67.0 0.0 100.0 0.0 20.9 4.1 WAL1 3.9 2.6 93.4 37.9 0.0 62.1 12.3 18.6 WAL2 4.4 25.0 70.6 68.9 0.0 31.1 6.6 18.8 WFG1 0.0 28.1 71.9 0.0 0.0 100.0 8.9 9.1 WIL1 0.0 0.0 100.0 0.0 0.0 100.0 28.1 5.1 WLF1 4.6 8.0 87.4 65.4 0.0 34.6 15.5 18.6 WLF2 8.5 51.5 39.9 38.0 20.1 41.9 4.6 16.0

123 Supplementary Table S3.6. Cumulative percentages of land uses and geologic categories, catchment areas, and mean elevations upstream of stream segments. Categories of land use and geology are explained in Supplementary Table S3.5. Catchment area and elevation were derived from the National Hydrography Dataset Plus version 2.1 (McKay et al., 2012). Stream codes are explained in Supplementary Table S3.1.

Stream Agriculture Developed Forested Carbonate Clastic Shale Catchment Elevation code (%) (%) (%) (%) (%) (%) area (km2) (m) BGS1 1.2 1.4 97.1 0.0 76.5 23.5 70 959 CHR1 7.1 2.6 85.8 0.0 76.1 23.9 395 1012 CLR1 33.6 4.9 61.1 21.4 19.8 58.7 122 884 CLR2 33.5 5.6 60.6 14.7 26.2 59.1 80 926 CRP1 34.5 3.1 61.8 33.8 36.5 29.7 334 824 CRP2 33.7 3.1 62.7 30.7 39.2 30.1 285 837 DER1 7.1 3.9 86.9 0.0 48.9 40.9 172 984 DIS1 0.0 0.9 98.4 0.0 64.6 35.4 35 885 EFG1 0.9 3.9 93.8 0.9 71.4 27.7 157 1096 EST1 9.9 8.3 79.4 24.2 6.1 69.7 198 744 EST2 10.8 9.1 77.5 20.5 5.0 74.5 172 751 GAL1 0.5 3.7 95.1 0.0 89.7 10.3 172 960 IND1 23.4 6.4 69.9 38.0 5.9 56.1 403 675 IND2 37.8 6.1 55.7 70.3 8.4 21.3 148 700 KIM1 9.7 1.7 88.2 0.0 25.9 74.1 233 806 KIM2 11.2 1.8 86.7 0.0 18.2 81.8 181 795 KNP1 7.7 4.0 87.7 5.4 63.2 31.4 51 937 LVA1 10.7 4.4 84.4 0.3 50.2 49.5 63 853 LWA1 8.5 1.8 89.3 0.0 36.0 64.0 127 804 LWA2 8.6 2.0 89.1 0.0 40.6 59.4 78 849 LWV1 7.5 1.8 81.8 0.0 85.0 15.0 86 1027 NOB1 10.5 1.7 87.7 0.0 48.8 51.2 64 858 NOB2 8.5 1.4 89.9 0.0 49.2 50.8 60 859 PNR1 56.1 9.5 33.5 16.8 2.7 80.5 45 679 PNR2 57.8 6.0 35.7 18.3 2.9 78.8 25 684 RED1 41.7 6.4 50.4 39.6 20.7 39.7 704 755 RED2 40.6 6.3 51.5 40.7 22.5 36.8 646 765 SEC1 25.2 5.6 68.3 39.4 44.5 16.1 278 757 SEC2 18.5 5.2 75.2 27.9 53.5 18.6 225 775 SFC1 11.4 0.7 83.4 0.0 59.4 40.6 125 1065 SFR1 73.4 7.0 18.5 78.0 0.0 22.0 66 760 SFR2 70.4 7.1 21.1 70.9 0.0 29.1 50 775 SIL1 11.0 3.6 84.6 9.7 44.9 45.4 102 964 SNK1 30.1 4.5 65.0 36.8 9.0 54.2 188 755 SNK2 31.5 4.4 64.0 38.9 9.3 51.8 169 760

124 Supplementary Table S3.6 continued STN1 11.8 4.4 80.5 26.1 13.4 60.5 40 930 WAL1 21.0 2.7 75.9 18.8 19.9 61.2 792 776 WAL2 19.9 2.7 76.9 16.8 22.1 61.1 715 786 WFG1 0.1 3.0 96.3 3.5 60.9 35.6 161 1036 WIL1 0.8 3.3 93.3 0.0 16.6 83.4 104 1123 WLF1 24.0 3.8 71.7 9.2 28.8 61.8 520 877 WLF2 24.9 3.9 70.6 8.2 28.9 62.8 489 883

125 Supplementary Material S3.7. We evaluated the transferability of regression models via a seven-fold cross-validation (Manel, Williams & Ormerod, 2001). First, the 42 stream segments were randomly partitioned into seven equal folds. Next, models were fit using data from six folds (36 segments) and then used to predict the status (i.e., present, absent) of the remaining six segments. The process was repeated six times, and the classification criterion averaged across folds. We used area-under-the-curve (AUC) of the receiver-operating characteristic as a threshold-independent criterion for discriminating between presence and absence. Values of 0.50–0.69 indicated poorly performing models, 0.70–0.90 indicated moderate performance, and > 0.90 indicated high performance (Manel, Williams & Ormerod, 2001).

126 Supplementary Table S3.8. Akaike’s Information Criterion corrected for low sample size (AICc) for 141 candidate models explaining the contemporary distribution of E. osburni. ǻ$,&F is the difference between the top-ranked model and lower-ranked models (i). Model weight (Wi) is the probability of being the best-supported model. Area under the curve (AUC) is a threshold- independent measure of cross-validation success. AUC entries are the mean and (standard deviation) resulting from a seven-fold cross-validation.

Rank SPMDT SMDMX Emb. Silt SP1 SP2 SP3 SP4 LL ǻ$,&F Wi AUC (SD) 1 1 0 1 0 1 1 0 0 -7.1 0.0 0.14 0.96 (0.08) 2 0 1 1 0 1 1 0 0 -7.2 0.2 0.12 0.93 (0.15) 3 1 0 0 1 1 1 0 0 -7.6 1.0 0.08 0.97 (0.08) 4 1 0 0 1 1 0 1 0 -7.8 1.4 0.07 0.97 (0.08) 5 1 0 1 0 0 1 1 0 -8.2 2.1 0.05 0.92 (0.10) 6 1 0 1 0 1 0 1 0 -8.2 2.3 0.05 0.95 (0.08) 7 1 0 0 1 0 0 1 0 -9.7 2.5 0.04 0.95 (0.08) 8 0 0 1 0 1 1 0 0 -9.7 2.6 0.04 0.91 (0.15) 9 1 0 1 0 0 1 0 0 -9.7 2.6 0.04 0.91 (0.11) 10 1 0 0 1 1 0 0 1 -8.8 3.4 0.03 0.88 (0.16) 11 1 0 0 1 0 1 1 0 -9.0 3.8 0.02 0.95 (0.08) 12 1 1 0 1 0 0 1 0 -9.1 3.9 0.02 0.90 (0.13) 13 0 0 0 1 1 1 0 0 -10.5 4.2 0.02 0.90 (0.16) 14 1 0 1 0 0 1 0 1 -9.3 4.3 0.02 0.89 (0.11) 15 1 0 0 1 1 0 0 0 -10.6 4.3 0.02 0.93 (0.12) 16 0 0 1 0 1 1 1 0 -9.3 4.3 0.02 0.91 (0.15) 17 1 0 0 0 1 1 0 0 -10.6 4.4 0.02 0.93 (0.15) 18 0 1 0 1 1 1 0 0 -9.4 4.6 0.01 0.89 (0.16) 19 1 0 0 1 0 0 1 1 -9.4 4.6 0.01 0.91 (0.16) 20 0 0 1 0 1 1 0 1 -9.7 5.1 0.01 0.88 (0.15) 21 1 1 1 0 0 1 0 0 -9.7 5.2 0.01 0.87 (0.12) 22 1 0 1 0 0 0 1 0 -11.0 5.2 0.01 0.89 (0.14) 23 1 0 0 0 1 0 1 0 -11.2 5.6 0.01 0.91 (0.15) 24 1 0 0 1 0 0 0 0 -12.4 5.6 0.01 0.93 (0.10) 25 1 1 0 1 0 0 0 0 -11.3 5.8 0.01 0.92 (0.14) 26 1 0 0 1 0 0 0 1 -11.4 5.9 0.01 0.88 (0.15) 27 0 0 0 1 1 0 1 0 -11.4 6.0 0.01 0.86 (0.20) 28 0 1 0 1 1 0 1 0 -10.2 6.1 0.01 0.86 (0.20) 29 1 1 0 1 1 0 0 0 -10.2 6.2 0.01 0.90 (0.12) 30 0 0 0 1 1 1 1 0 -10.2 6.3 0.01 0.90 (0.16) 31 1 1 0 1 0 0 0 1 -10.3 6.4 0.01 0.90 (0.13) 32 1 1 0 0 1 1 0 0 -10.3 6.5 0.01 0.91 (0.15) 33 0 0 0 1 1 1 0 1 -10.4 6.5 0.01 0.84 (0.17) 34 1 0 1 0 1 0 0 1 -10.5 6.8 0.01 0.88 (0.15)

127 Supplementary Table S3.8 continued 35 1 0 0 0 1 1 1 0 -10.5 6.8 0.01 0.93 (0.15) 36 1 0 1 0 0 0 0 1 -11.9 6.9 0.00 0.82 (0.17) 37 1 0 0 0 1 1 0 1 -10.6 6.9 0.00 0.96 (0.08) 38 1 0 1 0 0 0 1 1 -10.6 6.9 0.00 0.86 (0.17) 39 1 0 0 0 1 0 0 0 -13.1 7.0 0.00 0.84 (0.16) 40 1 1 0 0 1 0 1 0 -10.7 7.2 0.00 0.94 (0.08) 41 1 0 1 0 0 0 0 0 -13.3 7.2 0.00 0.89 (0.15) 42 1 1 1 0 0 0 1 0 -10.8 7.4 0.00 0.87 (0.14) 43 1 0 1 0 1 0 0 0 -12.1 7.4 0.00 0.88 (0.12) 44 1 1 0 0 0 0 1 0 -12.2 7.5 0.00 0.90 (0.16) 45 1 0 0 0 1 0 1 1 -10.9 7.6 0.00 0.96 (0.08) 46 1 0 0 0 0 0 1 0 -13.5 7.7 0.00 0.88 (0.16) 47 1 0 0 1 0 1 0 0 -12.3 7.8 0.00 0.93 (0.10) 48 1 1 0 1 0 1 0 0 -11.1 7.9 0.00 0.92 (0.14) 49 1 0 0 0 1 0 0 1 -12.5 8.2 0.00 0.86 (0.16) 50 0 0 0 1 1 0 1 1 -11.3 8.3 0.00 0.86 (0.20) 51 0 1 1 0 1 0 1 0 -11.3 8.4 0.00 0.88 (0.15) 52 1 0 0 1 0 1 0 1 -11.4 8.5 0.00 0.88 (0.15) 53 1 0 0 0 0 1 1 0 -12.7 8.6 0.00 0.92 (0.11) 54 1 1 0 0 0 1 1 0 -11.4 8.6 0.00 0.90 (0.10) 55 0 0 0 1 1 0 0 0 -13.9 8.6 0.00 0.89 (0.17) 56 0 1 1 0 0 1 1 0 -11.5 8.7 0.00 0.87 (0.15) 57 0 0 1 0 1 0 1 0 -12.8 8.7 0.00 0.83 (0.19) 58 1 1 0 0 1 0 0 0 -12.8 8.8 0.00 0.83 (0.14) 59 1 1 1 0 0 0 0 0 -12.9 8.9 0.00 0.86 (0.15) 60 1 0 0 0 0 0 1 1 -12.9 9.0 0.00 0.90 (0.12) 61 1 1 0 0 0 0 0 0 -14.2 9.1 0.00 0.88 (0.15) 62 1 0 0 0 0 0 0 0 -15.4 9.2 0.00 0.88 (0.16) 63 1 1 1 0 0 0 0 1 -11.7 9.2 0.00 0.81 (0.16) 64 1 1 0 0 0 0 1 1 -11.8 9.3 0.00 0.92 (0.10) 65 0 0 0 1 1 0 0 1 -13.1 9.4 0.00 0.88 (0.17) 66 0 1 0 1 1 0 0 1 -11.9 9.5 0.00 0.83 (0.19) 67 1 1 1 0 1 0 0 0 -12.0 9.7 0.00 0.88 (0.12) 68 0 1 0 1 1 0 0 0 -13.4 9.9 0.00 0.89 (0.17) 69 1 0 0 0 0 0 0 1 -14.8 10.4 0.00 0.88 (0.16) 70 1 1 0 0 1 0 0 1 -12.3 10.5 0.00 0.84 (0.14) 71 1 0 0 0 0 1 1 1 -12.4 10.6 0.00 0.92 (0.11) 72 1 1 0 0 0 0 0 1 -13.8 10.8 0.00 0.88 (0.15) 73 1 1 0 0 0 1 0 0 -13.8 10.9 0.00 0.88 (0.15) 74 1 0 0 0 0 1 0 0 -15.1 10.9 0.00 0.86 (0.19) 75 0 0 0 0 1 1 0 0 -15.2 11.1 0.00 0.86 (0.20) 76 0 0 1 0 1 0 1 1 -12.8 11.3 0.00 0.83 (0.19) 128 Supplementary Table S3.8 continued 77 0 1 0 0 1 1 0 0 -14.2 11.5 0.00 0.88 (0.17) 78 0 0 1 0 1 0 0 0 -15.4 11.5 0.00 0.87 (0.14) 79 0 1 1 0 0 1 0 0 -14.4 12.0 0.00 0.80 (0.21) 80 0 0 1 0 0 1 1 0 -14.5 12.1 0.00 0.83 (0.16) 81 0 1 1 0 1 0 0 0 -14.5 12.2 0.00 0.81 (0.16) 82 0 1 1 0 1 0 0 1 -13.4 12.5 0.00 0.88 (0.16) 83 1 0 0 0 0 1 0 1 -14.7 12.5 0.00 0.86 (0.19) 84 1 1 0 0 0 1 0 1 -13.4 12.6 0.00 0.88 (0.15) 85 0 0 0 0 1 0 0 0 -17.1 12.7 0.00 0.84 (0.15) 86 0 1 0 0 1 1 1 0 -13.6 12.9 0.00 0.85 (0.19) 87 0 0 1 0 0 1 0 0 -16.2 13.2 0.00 0.81 (0.16) 88 0 0 0 0 1 1 0 1 -15.1 13.3 0.00 0.86 (0.20) 89 0 0 1 0 1 0 0 1 -15.1 13.4 0.00 0.83 (0.17) 90 0 1 0 0 1 0 0 0 -16.3 13.4 0.00 0.88 (0.16) 91 0 0 0 0 1 1 1 0 -15.1 13.4 0.00 0.86 (0.20) 92 0 0 0 0 1 0 1 0 -16.4 13.5 0.00 0.90 (0.16) 93 0 1 1 0 0 1 0 1 -13.9 13.6 0.00 0.80 (0.21) 94 0 0 1 0 0 1 1 1 -14.0 13.7 0.00 0.79 (0.17) 95 0 0 1 0 0 1 0 1 -15.3 13.8 0.00 0.81 (0.16) 96 0 1 0 0 1 1 0 1 -14.2 14.1 0.00 0.86 (0.16) 97 0 1 0 0 1 0 1 0 -15.5 14.2 0.00 0.88 (0.17) 98 0 0 0 0 1 1 1 1 -14.5 14.8 0.00 0.78 (0.21) 99 0 0 0 0 1 0 0 1 -17.1 15.0 0.00 0.84 (0.15) 100 0 1 0 0 1 0 0 1 -16.1 15.5 0.00 0.91 (0.12) 101 0 0 0 0 1 0 1 1 -16.4 15.9 0.00 0.92 (0.17) 102 0 1 0 0 1 0 1 1 -15.4 16.5 0.00 0.88 (0.17) 103 0 1 1 0 0 0 1 0 -17.9 18.9 0.00 0.82 (0.21) 104 0 1 1 0 0 0 0 0 -19.8 20.3 0.00 0.70 (0.19) 105 0 0 1 0 0 0 1 0 -19.8 20.4 0.00 0.78 (0.23) 106 0 0 1 0 0 0 0 0 -21.1 20.6 0.00 0.79 (0.22) 107 0 1 1 0 0 0 1 1 -17.4 20.6 0.00 0.82 (0.21) 108 0 1 1 0 0 0 0 1 -18.9 21.0 0.00 0.74 (0.22) 109 0 0 1 0 0 0 0 1 -20.6 21.9 0.00 0.76 (0.23) 110 0 0 1 0 0 0 1 1 -19.5 22.3 0.00 0.79 (0.23) 111 0 0 0 1 0 1 1 0 -20.5 24.1 0.00 0.79 (0.17) 112 0 1 0 1 0 1 1 0 -19.5 24.9 0.00 0.81 (0.13) 113 0 0 0 1 0 0 1 0 -22.4 25.6 0.00 0.76 (0.15) 114 0 1 0 1 0 0 1 0 -21.4 26.0 0.00 0.76 (0.17) 115 0 0 0 1 0 1 1 1 -20.3 26.4 0.00 0.72 (0.21) 116 0 0 0 1 0 0 0 0 -24.1 26.7 0.00 0.73 (0.12) 117 0 0 0 1 0 1 0 0 -23.1 27.0 0.00 0.77 (0.16) 118 0 1 0 1 0 0 0 0 -23.6 27.8 0.00 0.69 (0.12) 129 Supplementary Table S3.8 continued 119 0 0 0 1 0 0 1 1 -22.3 27.9 0.00 0.76 (0.15) 120 0 1 0 1 0 0 1 1 -21.4 28.5 0.00 0.73 (0.14) 121 0 1 0 1 0 1 0 0 -22.7 28.5 0.00 0.77 (0.09) 122 0 0 0 1 0 0 0 1 -24.0 28.6 0.00 0.75 (0.09) 123 0 0 0 1 0 1 0 1 -23.0 29.1 0.00 0.76 (0.17) 124 0 1 0 1 0 0 0 1 -23.4 29.9 0.00 0.71 (0.13) 125 0 1 0 1 0 1 0 1 -22.5 30.8 0.00 0.75 (0.14) 126 0 0 0 0 0 0 0 0 -27.9 32.0 0.00 0.50 (0.00) 127 0 0 0 0 0 1 0 0 -27.1 32.6 0.00 0.74 (0.20) 128 0 1 0 0 0 0 0 0 -27.2 32.8 0.00 0.79 (0.17) 129 0 0 0 0 0 0 1 0 -27.6 33.7 0.00 0.75 (0.19) 130 0 1 0 0 0 1 0 0 -26.5 33.7 0.00 0.73 (0.16) 131 0 0 0 0 0 0 0 1 -27.9 34.2 0.00 0.76 (0.14) 132 0 0 0 0 0 1 1 0 -26.8 34.3 0.00 0.71 (0.17) 133 0 1 0 0 0 0 1 0 -26.9 34.5 0.00 0.77 (0.14) 134 0 0 0 0 0 1 0 1 -27.1 34.9 0.00 0.74 (0.25) 135 0 1 0 0 0 0 0 1 -27.2 35.1 0.00 0.72 (0.20) 136 0 1 0 0 0 1 1 0 -26.2 35.5 0.00 0.69 (0.16) 137 0 0 0 0 0 0 1 1 -27.6 36.0 0.00 0.72 (0.15) 138 0 1 0 0 0 1 0 1 -26.5 36.2 0.00 0.74 (0.20) 139 0 0 0 0 0 1 1 1 -26.8 36.7 0.00 0.75 (0.19) 140 0 1 0 0 0 0 1 1 -26.9 36.9 0.00 0.76 (0.13) 141 0 1 0 0 0 1 1 1 -26.2 38.1 0.00 0.72 (0.15) SPMDT = Spring mean daily temperature, SMDMX = summer mean daily maximum temperature, Emb. = embeddedness, SP = spatial covariate, LL = Log likelihood.

130 Supplementary Table S3.9. Predicted probability of occurrence within 42 stream segments from the best-supported distribution model containing spring mean daily temperature (SPMDT), embeddedness index, and the first two spatial covariates. Predictions in column 3 (Median spatial) control for spatial influences via identical median values for both spatial covariates. Predictions in column 4 (Fully spatial) use habitat covariates and true geographic coordinates. Predictions in column 5 (Non-spatial) use the best-supported model without spatial covariates (SPMDT, embeddedness index). Bolded rows are unoccupied segments with suitable instream habitat predicted by the “Median spatial” model. Italicized rows are occupied segments with unsuitable instream habitat predicted by the “Median spatial” model.

Predicted presence Stream Median Fully Non- code Detected spatial spatial spatial WIL1 1 0.98 1.00 0.99 EFG1 1 0.96 1.00 0.97 STN1 1 0.91 0.99 0.96 KNP1 1 0.91 1.00 0.91 WFG1 1 0.91 1.00 0.93 SFC1 1 0.90 1.00 0.95 DER1 1 0.86 1.00 0.84 DIS1 1 0.84 0.84 0.92 LWA2 0 0.78 0.78 0.70 BGS1 1 0.77 0.76 0.90 SIL1 1 0.69 0.99 0.67 CHR1 1 0.69 1.00 0.77 SEC1 0 0.62 0.04 0.57 SEC2 0 0.52 0.03 0.46 LVA1 1 0.50 0.49 0.60 LWV1 1 0.41 1.00 0.84 CLR2 0 0.36 0.35 0.70 LWA1 0 0.19 0.19 0.20 CRP1 1 0.16 0.74 0.14 NOB2 0 0.09 0.09 0.47 WLF2 1 0.08 0.07 0.13 SNK1 0 0.06 0.06 0.12 WAL1 0 0.05 0.05 0.07 WAL2 0 0.04 0.04 0.07 KIM2 0 0.04 0.04 0.16 CRP2 0 0.03 0.32 0.07 CLR1 0 0.03 0.03 0.12 SNK2 0 0.03 0.03 0.10 KIM1 0 0.02 0.02 0.08 GAL1 1 0.02 1.00 0.26

131 Supplementary Table S3.9 continued NOB1 0 0.02 0.02 0.22 WLF1 0 0.02 0.02 0.02 EST1 0 0.01 0.00 0.02 EST2 0 0.01 0.00 0.05 IND2 0 0.01 0.00 0.04 SFR1 0 0.00 0.01 0.00 SFR2 0 0.00 0.01 0.00 RED1 0 0.00 0.01 0.01 PNR2 0 0.00 0.00 0.00 PNR1 0 0.00 0.00 0.00 IND1 0 0.00 0.00 0.00 RED2 0 0.00 0.00 0.00

132 Supplementary References S3.10.

Addair, J. (1944). The fishes of the Kanawha River system in West Virginia and some factors which influence their distribution. (Doctoral dissertation, The Ohio State University, Columbus).

Anderson, J.R. (1976). A Land Use and Land Cover Classification System for Use with Remote Sensor Data (Report No. 964), Arlington, VA: U.S. Geological Survey,

Burns, A.D. (2007). Comparison of two electrofishing gears (backpack and parallel wires) and abundances of fishes of the upper Greenbrier River drainage. (MS thesis, West Virginia University, Morgantown).

Burton, G.W. & Odum, E.P. (1945). The distribution of stream fish in the vicinity of Mountain Lake, Virginia. Ecology, 26,182–194.

Bye, M.B. (1997) Summary of 1996 Activity Concerning Native and Transplanted Populations of Candy Darters in Dismal Creek, VA. Report to Candy Darter Conservation Committee, Blacksburg, VA.

Chipps, S.R. (1992). Stream fish communities of the central Appalachian Plateau: an examination of trophic group abundance patterns and resource partitioning among benthic fishes. (MS thesis, West Virginia University, Morgantown).

Dicken, C.L., Nicholson, S.W., Horton, J.D., Kinney, S.A., Gunther, G., Foose, M.P. & Mueller, J.A.L. (2008). Preliminary Integrated Geologic Map Databases for the United States: Delaware, Maryland, New York, Pennsylvania, and Virginia, (Version 1.1). U.S. Geological Survey, Reston, VA. (Retrieved from https://pubs.usgs.gov/of/2005/1325/).

Dunn, C.G. & Angermeier, P.L. (2016). Development of Habitat Suitability Indices for the Candy Darter, with Cross-Scale Validation across Representative Populations. Transactions of the American Fisheries Society, 145,1266–1281.

133 Hocutt, C., Denoncourt, R. & Stauffer Jr, J. (1979). Fishes of the Gauley River, West Virginia. Brimleyana, 1, 47–80.

Homer, C.G., Dewitz, J.A., Yang, L., Jin, S., Danielson, P., Xian, G., … Megown, K. 2015. Completion of the 2011 National Land Cover Database for the conterminous United States-Representing a decade of land cover change information. Photogrammetric Engineering and Remote Sensing, 81, 345–354. (Retrieved from https://www.mrlc.gov/nlcd2011.php).

Jenkins, R.E.,& Kopia, B.L. (1995). Population status of the candy darter, Etheostoma osburni, in Virginia 1994–95, with historical review. Unpublished manuscript.

Manel, S., Williams, H.C. & Ormerod, S.J. ( 2001). Evaluating presence-absence models in ecology: the need to account for prevalence. Journal of Applied Ecology, 38, 921–931.

Nicholson, S.W., Dicken, C.L., Horton, J.D., Labay, K.A., Foose, M.P. and Mueller, J.A.L.. (2007). Preliminary Integrated Geologic Map Databases for the United States: Kentucky, Ohio, Tennessee, and West Virginia. U.S. Geological Survey, Reston, VA. (Retrieved from https://pubs.usgs.gov/of/2005/1324/).

McKay, L., Bondelid, T., Dewald, T., Johnston, J., Moore, R. & Rea, A. (2012). NHDplus Version 2: User Guide. U.S. Environmental Protection Agency. (Retrieved from http://www.horizon-systems.com/NHDPlus/index.php).

Zuur, A.F. (2009). Mixed Effects Models and Extensions in Ecology with R. Springer.

134