Implications for Three Sensitive Fishes in a Tributary of the Green River, Utah

Utah State University DigitalCommons@USU

All Graduate Theses and Dissertations Graduate Studies

5-2009

Maintaining Population Persistence in the Face of an Extremely Altered Hydrograph: Implications for Three Sensitive Fishes in a Tributary of the Green River, Utah

Jared L. Bottcher Utah State University

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

Part of the Aquaculture and Fisheries Commons

Recommended Citation Bottcher, Jared L., "Maintaining Population Persistence in the Face of an Extremely Altered Hydrograph: Implications for Three Sensitive Fishes in a Tributary of the Green River, Utah" (2009). All Graduate Theses and Dissertations. 496. https://digitalcommons.usu.edu/etd/496

This Thesis is brought to you for free and open access by the Graduate Studies at DigitalCommons@USU. It has been accepted for inclusion in All Graduate Theses and Dissertations by an authorized administrator of DigitalCommons@USU. For more information, please contact [email protected]. MAINTAINING POPULATION PERSISTENCE IN THE FACE OF AN EXTREMELY

ALTERED HYDROGRAPH: IMPLICATIONS FOR

THREE SENSITIVE FISHES IN A TRIBUTARY

OF THE GREEN RIVER, UTAH

Jared L. Bottcher

A thesis submitted in partial fulfillment of the requirements for the degree

MASTER OF SCIENCE

Watershed Science

Approved:

______Phaedra Budy Todd Crowl Major Professor Committee Member

______Mike Pfrender Bryon Burnham Committee Member Dean of Graduate Studies

UTAH STATE UNIVERSITY Logan, Utah

2009 ii

iii

ABSTRACT

Maintaining Population Persistence in the Face of an Extremely Altered Hydrograph:

Implications for Three Sensitive Fishes in a Tributary of the Green River, Utah

Jared L. Bottcher, Master of Science

Utah State University, 2009

Major Professor: Dr. Phaedra Budy Department: Watershed Sciences

The ability of an organism to disperse to suitable habitats, especially in modified and fragmented systems, determines individual fitness and overall population viability.

The bluehead sucker (Catostomus discobolus), flannelmouth sucker (Catostomus latipinnis), and roundtail chub (Gila robusta) are three species native to the upper

Colorado River Basin that now occupy only 50% of their historic range. Despite these distributional declines, populations of all three species are present in the San Rafael

River, a highly regulated tributary of the Green River, Utah, providing an opportunity for research. Our goal was to determine the timing and extent of movement, habitat preferences, and limiting factors, ultimately to guide effective management and recovery of these three species. In 2007-2008, we sampled fish from 25 systematically selected,

300-m reaches in the lower 64 km of the San Rafael River, spaced to capture the range of species, life-stages, and habitat conditions present. We implanted all target species with a iv passive integrated transponder (PIT) tag, installed a passive PIT tag antennae, and measured key habitat parameters throughout each reach and at the site of native fish capture. We used random forest modeling to identify and rank the most important abiotic and biotic predictor variables, and reveal potential limiting factors in the San Rafael

River. While flannelmouth sucker were relatively evenly distributed within our study area, highest densities of roundtail chub and bluehead sucker occurred in isolated, upstream reaches characterized by complex habitat. In addition, our movement and length-frequency data indicate downstream drift of age-0 roundtail chub, and active upstream movement of adult flannelmouth sucker, both from source populations, providing the lower San Rafael River with colonists. Our random forest analysis highlights the importance of pools, riffles, and distance-to-source populations, suggesting that bluehead sucker and roundtail chub are habitat limited in the lower San Rafael River.

These results suggest management efforts should focus on diversifying habitat, maintaining in-stream flow, and removing barriers to movement.

(72 pages)

DEDICATION

To Allison and Cameron, for their unwavering love and patience, and for giving me so much to live for. And to my Dad, who took me fishing.

ACKNOWLEDGMENTS

I am truly fortunate to have had Dr. Phaedra Budy as my major professor, whose guidance, support, and patience are seemingly limitless. I am a better field biologist today after working closely with Gary Thiede, who taught me the importance of details, logistics, and clean sampling gear. Peter MacKinnon provided much needed comic relief in the office and was instrumental during the installation and maintenance of the PIT tag detector. Ryan Hill and Katie Hein provided helpful comments concerning the development and interpretation of the random forest models. I would like to thank Tracy

Bowerman for her critical review of my thesis and for many project insights and suggestions during my two years at Utah State University. In addition, this thesis benefitted tremendously from discussions with the entire Fish Ecology Lab, Pat Kennedy,

Kirk Dahle, Brian Hines, and Kevin Landom. Silas Ratten, Ben Schleicher, Peter

MacKinnon, Dave Cole, Jess Hancock, Spencer Weston, Christy Meredith, Lindsey

Goss, Matt Archibald, Greg Hill, Chase Ehlo, and Brandon Simcox helped in the collection of data.

Jared Bottcher

vii

CONTENTS

Page

ABSTRACT………………………………………………………………………...……iii

DEDICATION……………………………………………………………………….....…v

ACKNOWLEDGMENTS……………………………………………………...………...vi

LIST OF TABLES…………….……………………………………………..…...………ix

LIST OF FIGURES……………………………………………………....….……………x

INTRODUCTION………………………………………………………...…………..…..1

METHODS……………..………………………………………………...……………….8

Study site……………………………………………………………………..……8 Sample site selection………………………………………………………..……..9 Fish distribution and abundance……………………………………...………….12 Patch delineation: Seasonal and spatial habitat use…………………...…………13 Native fish movement………………………………………………...………….15 Micro- and macrohabitat availability…………………………………...………..16 Fish habitat use and preference……………………………………...…………...18 Data analysis…………………………………………………...………………...19

RESULTS………………………………………………………………..………………23

Fish distribution and abundance………………………………………...……….23 Patch delineation: Seasonal and spatial habitat use……………………...………26 Macrohabitat availability………………………………………...………………27 Fish habitat preferences…………………………………………...……………..27 Native fish movement………………………………...………………………….28 Data analysis………………………………………………...…………………...29

DISCUSSION………………………………………………………..…………………..41

Overview………………………………………………………………...……….40 Native fish distribution and movement……………………...………………...…40 Habitat use and availability………………………….…………...………………45 Identification of limiting factors……………………………...…………...……..47 Conclusions………………………………...…………………………………….50 viii

Management implications………………………………...……………………...51

REFERENCES…………………………………………………………..………………53

LIST OF TABLES

Table Page

1 Predictor variables, their range of values, units, and variable type, used in random forest modeling………….……………………………………………………….21

2 Occurrences of bluehead sucker, flannelmouth sucker, and roundtail chub, collected from our spring, summer, and fall sampling periods in 2007 and 2008, by site, size, and patch number. All age-0 suckers collected were assumed to be flannelmouth suckers because all sacrificed age-0 suckers were identified as such. PIT tagged individuals are denoted with parentheses……………………………24

3 Accuracy measures for the three species presence/absence random forest models using a total of 25 sites. Model sensitivity is the proportion of true positives (presences) correctly predicted, specificity is the percentage of true negatives (absences) correctly predicted, PCC is the proportion of observations (presences and absences) correctly classified, and Kappa is a measure of agreement between predicted and actual class assignments that corrects for chance (Manel et al. 2001). The top four abiotic and biotic predictor variables are also listed for each species…………………………………………………………………………....39

LIST OF FIGURES

Figure Page

1 Sample and index reaches in the lower 64 km of the San Rafael River, in eastern Utah……………………………………………………………………….……...10

2 Locations of the three upstream sampling sites (Fuller Bottom, Buckhorn Draw, and Tidwell Bottom), in relation to sampling sites in the lower 64 km of the San Rafael River……………………………………………………………………...11

3 Spatial distribution of the five patches within the San Rafael River, and the one external patch, which consists of all streams within the upper Colorado River Basin, including the Green River. We delineated patch boundaries based on the biological and physical similarities and differences among adjacent sample sites, maximizing heterogeneity between patches and homogeneity within patches…………………………………………………………………………....14

4 Schematic representation of the spatial arrangement and orientation of the four antennae loops, located approximately 2 km upstream of the confluence with the Green River (shaded star in Figure 1)…………………………………..………..16

5 Native fish density (fish/m2), collected in 2008 across all sampling periods, by species, lifestage, and sample reach. Reaches are arranged in an upstream to downstream manner, and important landmarks are noted, such as the diversion at Hatt Ranch, and the PIT tag detector…………………………………………….23

6 The relative frequencies of native fish catch for each species (roundtail chub, flannelmouth sucker, and bluehead sucker), lifestage (adult, juvenile, age-0), and patch (1-5), summed from our 2007 and 2008 field seasons. Row A represents natives collected during the spring (April-May); row B represents fish collected in the summer (June-July); and row C represents fish collected in the fall (October)……………………………………………………………...………….29

7 Habitat composition by site and macrohabitat type (Panel A). Distribution and abundance of complex habitat, an aggregation pools, riffles, and backwaters, in the San Rafael River (Panel B). Reaches are organized in an upstream to downstream manner, and important landmarks are noted, such as the diversion at Hatt Ranch, and the PIT tag antennae…………………………………………....30

8 Bluehead sucker microhabitat preferences for water depth (cm), water velocity (m/sec), and substrate size (binned 1-8). Habitat preferences are defined as the percent of habitat type used (collected at the site of native fish capture), divided xi

by the percent available (collected from seven equally spaced transects within each reach)……………………………………………..………………………...31

9 Roundtail chub microhabitat preferences for water depth (cm), water velocity (m/sec), and substrate size (binned 1-8). Habitat preferences are defined as the percent of habitat type used (collected at the site of native fish capture), divided by the percent available (collected from seven equally spaced transects within each reach)………………………………………………………………..……...32

10 Flannelmouth sucker microhabitat preferences for water depth (cm), water velocity (m/sec), and substrate size (binned 1-8). Habitat preferences are defined as the percent of habitat type used (collected at the site of native fish capture), divided by the percent available (collected from seven equally spaced transects within each reach)……………………………………………………...………...33

11 Frequency of minimum distance moved for flannelmouth sucker (calculated as the distance traveled from an individuals’ initial capture or release site, to the site of subsequent recapture; Panel A), and the minimum, maximum, median, and mean distance moved for flannelmouth sucker (Panel B), calculated as the distance between the site of active capture and passive detection at the PIT tag antennae………………………………………………………………….………34

12 Timing and direction of movement of flannelmouth sucker and the endangered fishes of the Colorado River Basin, as it relates to the historic (1910-1918) and contemporary hydrograph (1999-2007). All fish recapture data were collected from the stationary PIT tag antennae in 2008………………………………………………………………………...…….35

13 Frequency of movement, as it relates to the extent traveled (km), for flannelmouth sucker, Colorado pikeminnow, bonytail chub, and razorback sucker. Minimum distance moved is the distance traveled from an individuals’ initial capture or release site, to the PIT tag antennae in the San Rafael River. The distances moved are broken into six bins (0-10 km, 11-50 km, 51-100 km, 101-200 km, 201-300 km and 300+ km)……………………………………………………...…………36

14 Variable importance plots for predictor variables used in classifying the presence of flannelmouth sucker (Panel A), roundtail chub (Panel B), and bluehead sucker (Panel C), using a random forest analysis. Higher values of mean decrease in Gini indicate variables more important to classification……………………………………………………………..………38

INTRODUCTION

Environmental perturbations resulting from a rapidly growing human population have contributed to dramatic declines in the abundance and distribution of native fishes worldwide. These perturbations include habitat loss and fragmentation, degraded water quality, impoundment of rivers, and disease (Leprieur et al. 2008). The Colorado River

Basin offers a prime example of an aquatic ecosystem threatened by all of these environmental disturbances. As such, the native fish community of the Colorado River

Basin is now one of the most threatened fish assemblages in the world (Minckley and

Deacon 1968), and four native and endemic fish species are listed under the Endangered

Species Act, and several others have been state-listed as sensitive species (Holden and

Stalnaker 1975; Holden 1991; UDWR 2006).

In addition to physical habitat degradation, the introduction and establishment of non-native fish species has led to the imperilment of many native fishes (Tyus and Nikirk

1990; Ruppert et al. 1993; Dudley and Matter 2000). Construction of numerous dams throughout the basin have turned the historically dynamic Colorado River (large, seasonal variations in discharge and temperature) into a highly static environment (Poff et al.

1997), resulting in range reductions for many native species, while also providing ‘new’ niche space for invasive non-native fish species (Olden et al. 2006). After establishment, non-native species can adversely impact the recruitment and survival of native fish through competition and predation. Diet overlap between invasive and native fishes in the Colorado River Basin is relatively high for some species (Quist et al. 2006), while diet overlap is negligible for other pairs, occurring only in times of scarce food 2 availability (Greger and Deacon 1988). Similarly, many non-native fishes of the

Colorado River Basin have been shown to be highly predacious (Tyus and Nikirk 1990;

Dudley and Matter 2000), and their apparent ability to shape whole fish communities is significant (Bestgen and Propst 1989; Johnson et al. 2008).

Despite the substantial role of non-native fishes, the effects of abiotic parameters relative to biotic interactions have been deemed to be comparatively large in shaping desert fish assemblages (Ross 1986). At present, highly connected riverine systems and intact native fish assemblages are rare (Moyle 1995), especially in the Colorado River

Basin, where dams, diversions, and water withdrawals create fragmented river systems

(Minckley and Deacon 1968; Fagan et al. 2002). By impeding movement of native fish, dams and diversions prevent access to preferred spawning habitat and other resources throughout the basin (Vanicek 1970; Chart and Bergersen 1992; Osmundson 2002). As a result of drought and water withdrawals, fish populations in streams throughout the

Colorado River Basin often become isolated in fragmented pools, preventing access to thermal refugia and likely leading to high mortality. Further, reduced immigration and emigration rates increase population isolation, producing populations more susceptible to extinction through environmental and demographic stochastic events (Fagan et al. 2002;

Hilderbrand 2003), although this risk is scale dependent (Fagan et al. 2005).

Compounding risks associated with fragmentation is the increased risk of extirpation due to small population sizes (Hilderbrand 2003), a trend commonly observed in isolated

Colorado River Basin populations (Minckley and Deacon 1968). These findings highlight the importance of maintaining or restoring connectivity in order to sustain 3 viable populations, especially between source and sink populations (Hanski and

Simberloff 1997; Stacey et al. 1997).

When riverine systems retain connectivity, highly mobile fish populations are capable of taking advantage of multiple habitat types and resources through time and space, greatly reducing extinction risk and increasing population persistence (Fagan et al.

2002; Hilderbrand 2003). Fishes that have historically demonstrated large scale migrations between habitat patches and types, such as those composing the Colorado

River fish assemblage, are better able to persist in the face of natural variability (e.g., droughts, floods) and habitat alteration, than sedentary populations (Fagan et al. 2002;

Olden 2006). All other things equal, fluvial populations can re-found unoccupied habitat patches and maintain genetic diversity (Hoffman and Dunham 2007; Douglas et al.

2008), processes that collectively increase the probability of population persistence via occupation of multiple habitat patches and a greater ability to adapt to changing environmental conditions (Gilpin and Soulé 1986; Fagan et al. 2002).

Not only do dams and diversions throughout the Colorado River Basin create fragmented habitat patches, but dams also alter temperature and flow regimes. The reduced amplitude and altered timing of tributary flows throughout the Colorado River

Basin have been implicated in the demise of many native fish species (Poff et al. 1997;

Propst and Gido 2004). Rising tributary flows, for example, provide an extremely important environmental cue for spawning migrations, and high spring flows provide environmental conditions necessary for egg incubation, hatching, and rearing (Weiss et al. 1998). In the highly regulated Colorado River Basin, discharges below dams are 4 typically homogenized (e.g., dampened high flows and augmented base flows), precluding movement to natal spawning areas and reducing spawning success (Poff et al.

2007). Recruitment to subsequent age classes may also be impacted by the altered thermal regime created by large scale dams, as depressed summer water temperatures decrease age-0 swimming performance and slow growth, thus increasing susceptibility to predation (Vanicek 1970; Robinson and Childs 2001; Ward et al. 2002; Ward and Bonar

2003). Conversely, over-allocation of water resources can result in low flows that create isolated pools where water temperature can exceed 35 0C (Bottcher, personal observation) and dissolved oxygen can reach concentrations as low as 1 mg/l (McAda et al. 1980); temperature and dissolved oxygen ranges lethal to most fish species in the

Colorado River Basin (Cross 1978; Sigler and Sigler 1996; Ptacek et al. 2005; Rees et al.

2005).

In addition to providing important cues for movement, the natural flow regime is a driving force in shaping channel morphology and in-stream habitat (Petts and Calow

1996). By severely reducing the magnitude of large spring spates, dams and diversions decrease the ability of streams to form new channels and create new habitat, especially among naturally braided river systems (Cheetham 1979; Poff et al. 1997). Alterations to the natural flow regime among rivers in the Colorado River Basin have likely provided the catalyst for converting highly-dynamic, multiple-thread rivers into static, singe- thread, meandering streams, thus reducing channel complexity and habitat diversity (Tal and Paola 2007). In addition, the altered timing of the spring spate and intermittency of many flow-regulated rivers have contributed to a shift in the riparian community from 5 one dominated by cottonwoods (Populus spp.) and willows (Salix spp.), to one that is dominated by invasive, non-native tamarix (Tamarix spp; Webb et al. 2002; Stromberg et al. 2005; Stromberg et al. 2007). The dynamic root structure of tamarix stabilizes the rivers banks (Carpenter 1998) and can quickly colonize abandoned channels, further preventing the creation of multiple channels and complex in-stream habitat (Gran and

Paola 2001; Tal et al. 2004; Tal and Paola 2007); these changes to stream habitat are detrimental to many native fishes that require these channel units to spawn, feed, and recruit.

Collectively, alterations to the natural flow regime, habitat degradation, fragmentation, and interactions with non-native species act synergistically to reduce the amount of suitable habitat for many native fishes in the Colorado River Basin. However, their relative importance in shaping species distributions is largely unknown, and ecologists and managers are often left with the unenviable position of developing recovery plans based solely on a list of potential threats to persistence. Therefore, a common aim in the conservation and management of at-risk species is to understand how these factors interact, and which factors are most limiting population persistence, thereby guiding management efforts. This is an extremely challenging task in the Colorado River

Basin because of the complex interactions between discharge, riparian vegetation, channel type, and in-stream habitat-availability, and where all native fish are imperiled to some degree.

The native and endemic suckers and chubs of the Green and Colorado River systems represent a unique set of sensitive species thought to have historically used a 6 diversity of habitat types and to have moved among tributary and mainstem systems. The flannelmouth sucker (Catostomus latipinnis) historically occupied the rivers and streams of the Colorado River Basin from Wyoming to Mexico, including a variety of habitat types (pool, riffle, run) within each system (Bezzerides and Bestgen 2002). The bluehead sucker (Catostomus discobolus) historically occupied small headwater, and large mainstem streams of the Snake, Weber, Bear, and Colorado River basins (Bezzerides and

Bestgen 2002), and robust populations were typically found in cool (~20 oC), fast flowing rivers dominated by rocky substrates (Sigler and Sigler 1996). Populations of roundtail chub (Gila robusta) were historically found from southern Wyoming to central Mexico, where they appear to prefer deep complex pools intermixed with riffles (Bestgen and

Propst 1989; Bezzerides and Bestgen 2002). Collectively, these three species are referred to as the ‘three species’ and are generally managed as a unit (hereafter, ‘three species’;

UDWR 2006).

All ‘three species’ have declined dramatically in both distribution and abundance throughout the basin, currently occupying approximately 50% of their historic range, largely due to habitat perturbations, fragmentation, and interactions with non-native fish

(Bezzerides and Bestgen 2002). These widescale declines have led to placement of all

‘three species’ on the Utah Sensitive Species List and their protection under a

Conservation and Management Plan (UDWR 2006), the goal of which is to proactively conserve remnant populations and their habitat. Despite the distributional declines, all three species are found in the San Rafael River in southeastern Utah, providing an area of high conservation priority and an opportunity for study. As a typical, mid-order 7 southwestern stream, the San Rafael River offers an opportunity to learn about tributary use and movement by both the ‘three species’ and the endangered fishes of the Colorado

River Basin, and provides a template for identifying limiting factors for these imperiled fishes that can be applied elsewhere in their native range.

Currently, the fundamental lack of information regarding the distribution and abundance of these three species, habitat preferences, movement patterns, and factors limiting their distribution, abundance, and persistence is impeding the development and prioritization of effective management plans. Therefore, our objectives were to: 1) quantify the distribution and abundance of the three species along the longitudinal gradient of the San Rafael River, 2) determine macrohabitat availability and fish habitat preference across this longitudinal gradient, 3) evaluate the timing, extent, and potential cues for movement, and lastly, using this information, 4) identify the abiotic and biotic factors limiting fish persistence in the San Rafael River.

METHODS

Study site

The San Rafael River drains 4,500 km2 in central Utah, and is formed by the merging of Cottonwood, Ferron, and Huntington Creeks. This snowmelt-driven system flows approximately 175 km southeast from its headwaters in the Manti La Sal National

Forest through the San Rafael Swell, terminating at the confluence with the Green River.

The hydrograph is characterized by large spring and autumnal spates, events that carry high sediment loads and shape new channel forms. Within this relatively small basin, there are over 800 surface points of diversion and 360 dams, making the San Rafael River one of the most over-allocated rivers in Utah (Walker and Hudson 2004). As a result, the

San Rafael River is frequently dewatered, particularly the lowermost 64 km. For example, in the summer of 2007, there was an approximately two month period with no water flow in the channel, isolating fish in disconnected pools. Additional loss of in- stream habitat has come about as a result of Tamarisk (Tamarix spp) invasion; tamarisk stands stabilize stream banks, preventing floodplain access and limiting the creation of complex habitat, such as split channels, backwaters, pools, and riffles.

The native fish community of the Upper Colorado River Basin is relatively depauperate, comprised of only 14 native fish species. Historically, roundtail chub, speckled dace (Rhinichthys osculus), bluehead sucker, flannelmouth sucker, and

Colorado River cutthroat trout (Oncorhynchus clarki pleuriticus) were widely distributed and abundant in the San Rafael River. Species which spend most of their lives in large rivers (e.g., Colorado and Green Rivers), including the Colorado pikeminnow, bonytail 9 and humpback chub, and razorback sucker likely used smaller tributaries, including the

San Rafael River, for spawning and rearing. However, due to sport fish stocking and incidental introductions in upstream tributaries of the San Rafael and Green Rivers, many non-native species have become established. Non-native species found in the San Rafael

River and its tributaries include red shiner (Cyprinella lutrensis), sand shiner (Notropis stramineus), brown trout (Salmo trutta), mosquito fish (Gambusia spp), black bullhead

(Ameiurus melas), channel catfish (Ictalurus punctatus), common carp (Cyprinus carpio), green sunfish (Lepomis cyanellus), and fathead minnow (Pimephales promelas; Walker and Hudson 2004; Bottcher, personal observation, 2008).

Sample site selection

In the summer of 2007, we systematically selected 22 (random seed start), 300-m sampling reaches in the lower 64 km of the San Rafael River, from the Hatt Ranch diversion, the lowermost diversion on the San Rafael River and a barrier to upstream fish movement (Douglas and Douglas 2008), to the confluence with the Green River, spaced to capture the range of species, life-stages, and habitat conditions present (Figure 1). The

300-m reaches were sequentially numbered from 1, the site directly downstream of the

Hatt Ranch diversion, to 213, the site directly upstream from the confluence with the

Green River. The 22, 300-m sampling reaches selected for sampling include sites 8, 15,

27, 35, 48, 62, 68, 77, 90, 95, 109, 120, 126, 134, 146, 158, 173, 182, 193, 196, 205, and

213. Three additional sites upstream of the Hatt Ranch diversion, Fuller Bottom (FB),

Tidwell Bottom (TB), and Buckhorn Draw (BD) were added in 2008, as these are sites which still hold relatively intact native fish communities (Figure 2). In general, we 10 sampled the lower 22 reaches during the spring (April-May), summer (June-July), and fall (October) of 2007 and 2008; however, due to logistical constraints (e.g., water flow), we did not sample all sites during each sampling period. We sampled the three sites upstream of the fish barrier (Fuller Bottom, Buckhorn Draw, and Tidwell Bottom) in

June and July of 2008.

Figure 1. Sample and index reaches in the lower 64 km of the San Rafael River, in eastern Utah.

Figure 2. Locations of the three upstream sampling sites (Fuller Bottom, Buckhorn Draw, and Tidwell Bottom), in relation to sampling sites in the lower 64 km of the San Rafael River.

Fish distribution and abundance

We collected fish in 2007 and 2008 using a variety of gear and sampling techniques, including a canoe electrofishing unit (Smith Root GPP 2.5), seine, and trammel nets. We selected gear types based on their specialization and effectiveness at 12 sampling discrete habitat types under varying flow conditions (e.g., electrofishing fast riffles and complex pools, seining backwaters, and trammel netting deep slow water); the majority of sites were sampled with more than one technique. To prevent fish from escaping reaches during sampling, we placed block nets at both the upstream and downstream ends of each reach before commencing sampling. We weighed, measured, and released all captured native and non-native fish. We anesthetized target species

(bluehead sucker, flannelmouth sucker, and roundtail chub) larger than 150 mm TL with

Tricaine Methanesulfonate (MS-222), and implanted a 12 mm passive integrated transponder (PIT) tag in the intramuscular site anterior to the dorsal fin (Guy et al. 1996).

In addition to tagging fish larger than 150 mm TL with 12 mm PIT tags, we implanted all target individuals larger than 250 mm TL with a 23 mm PIT tag in the ventral cavity

(Guy et al. 1996). We allowed all PIT tagged fish to equilibrate in a flow through bucket before releasing them into slow water near the point of capture.

We calculated native and non-native fish density by summing the species-specific fish catch for each site and dividing by the total area of each reach. We calculated total area by multiplying the average bankfull width of each reach (measured over seven transects spaced 50-m apart) by reach length (typically 300-m). Although every effort was made to standardize sampling effort, highly variable flows from upstream dams forced us to alter our sampling protocol to maximize efficiency. For example, in May and early June of 2008, site 213, directly upstream from the confluence with the Green

River, was over three meters deep, preventing canoe electrofishing and seining. As a result, trammel nets presented the only logical method to capture fish at this site, whereas 13 sites upstream could still be sampled efficiently with electroshockers and seine nets. We also recognize that some sites were sampled more intensively than others as a result of their accessibility, thus potentially biasing our density estimates.

Patch delineation: seasonal and spatial habitat use

In order to better understand the population dynamics of the native fish community across both time and space, we broke our study reach into five patches based on biological and physical similarities and examined how species and life-stage composition changed through time. Patch one consists of the sites upstream of the Hatt

Ranch diversion/fish barrier (FB, BD, TB), patch two consists of sites 8-68, patch three consists of sites 77-126, patch four consists of sites 134-182, and patch five consists of sites 193-213 (Figure 3). We combined our fish catch data from the 2007 and 2008 field seasons and binned captures into three sampling periods; spring (April-May), summer

(June-July), and fall (October).

We broke each species into three age classes (based on size at capture); age-0, juvenile, and adult. We considered flannelmouth sucker between 0-70 mm age-0 individuals; 71-250 mm juveniles; and >250 mm adults. We considered roundtail chub and bluehead sucker smaller than 70 mm age-0 individuals; 71-200 mm juveniles, and

>200 mm adults. Due to a substantial reduction in the size at maturity for fish inhabiting smaller tributaries (McAda 1977; Bestgen 1985), we considered fish collected in the San

Rafael adults at smaller sizes than those reported elsewhere for flannelmouth sucker

(McAda 1977), bluehead sucker (McAda 1977), and roundtail chub (Vanicek and

Kramer1969; Bestgen 1985). We calculated relative catch frequencies for each species 14

Patch 1: Fuller and Tidwell Bottoms, Buckhorn Draw

Patch 2: Hatt Ranch-Site 68

Patch 3: Sites 77-126

Patch 4: Sites 134-182

Patch 6: Green River Patch 5: Sites 193-213

Figure 3. Spatial distribution of the five patches within the San Rafael River, and the one external patch, which consists of all streams within the upper Colorado River Basin, including the Green River. We delineated patch boundaries based on the biological and physical similarities and differences among adjacent sample sites, maximizing heterogeneity between patches and homogeneity within patches.

and life stage and examined how community composition changed through time across our five patches, allowing us to better understand movement, tributary use, and limitations to population persistence for each species.

Native fish movement

In February of 2008, we constructed and installed a solar-powered, full-duplex

(134 kHz), PIT tag antennae capable of passively detecting passage of PIT tagged fish.

We positioned these antennae loops approximately two km upstream from the confluence with the Green River, an area representing a distinct habitat shift between tributary (San

Rafael River) and mainstem (Green River) systems (Figure 1). The multiplexer receiving station recorded the date, time, and individual-specific PIT tag number, which allowed us to quantify the timing and extent of movement. We positioned a set of two antennae loops four meters upstream of another set of two antennae loops, which allowed us to also determine direction of fish movement; both sets of antennae loops spanned the channel at bankfull flows (Figure 4). We made the assumption that upstream movement is representative of tributary-seeking behavior, whereas downstream movement is representative of individuals seeking mainstem habitat. We also determined the extent moved (distance from initial capture or release to subsequent detection) from the active recapture of PIT tagged fish.

Micro- and macrohabitat availability

Throughout our study system, we collected a variety of habitat parameters at every reach, including percent pool, riffle, and backwater, substrate size, and water depth.

We measured water depth and substrate size approximately every meter along the thalweg of each reach, with a calibrated rod and gravelometer, respectively. We binned substrate into eight size categories, and assigned a number to each bin—1-silt, 2-sand, 3- fine gravel, 4-coarse gravel, 5-small cobble, 6-large cobble, 7-boulder, and 8-bedrock 16

Figure 4. Schematic representation of the spatial arrangement and orientation of the four antennae loops, located approximately 2 km upstream of the confluence with the Green River (shaded star in Figure 1).

(Bunte and Abt 2001). We then determined average water depth and substrate size for each reach, as well as substrate standard deviation, to gain an understanding of the substrate heterogeneity (Bain 1999).

We identified pools as concave depressions (laterally and longitudinally), which span the thalweg and have a maximum depth of at least 1.5 times the pool tail depth

(PIBO 2008). We flagged both the upstream and downstream ends of each pool and calculated pool area after measuring pool width (measured at its widest point). We identified riffles as fastwater areas with surface turbulence, and relatively large substrate sizes (Hawkins et al. 1993). We identified backwaters as near-shore areas with currents 17 that typically flow counter to the prevailing current (Hawkins et al. 1993). In the San

Rafael River, these areas are typically formed by downed logs or an accumulation of woody debris, and range in sie from small pockets off the main channel, to long, deep, side channels. We defined complex habitat as an aggregation of pools, riffles, and backwaters (i.e., the diversity of habitat types available; Pearsons et al. 1992), and calculated the percent of complex habitat by reach by summing the area of pools, riffles, and backwaters and dividing by the total area of each reach. We recognize some complex reaches are dominated by a single channel unit (e.g., pools), while these same habitat types are relatively rare in other, similarly diverse reaches (e.g., reaches dominated by riffles). Nonetheless, in the San Rafael River, it is relatively easy to distinguish these discrete channel geomorphic units from the predominant habitat type; long, homogenous, sand-substrate runs, areas with little variation regarding depth, velocity, bankfull width, or substrate size.

In addition to determining the prevalence of channel geomorphic units, we also identified microhabitat availability by measuring water depth, water velocity, and substrate size at seven equally spaced (50 m) transects within each reach. We placed transects perpendicular to flow and we collected a minimum of 15 equally spaced observations per transect. We then summarized the amount of available habitat for the three parameters water depth (cm), water velocity (m/sec), and substrate size (binned 1-

8), and determined mean bankfull width.

Fish habitat use and preference

To determine fish habitat use for the three species, we measured water depth, water velocity, and substrate size at the site of fish capture. Measurements were collected at the focal point for each fish when possible; however, due to high water turbidity during periods of elevated flow, we could not generate precise focal point estimates. If the exact location of a captured fish was unknown, we collected water depth, velocity, and substrate size at three points on a transect perpendicular to flow (usually encompassing

0.5 meter on each side of our point estimate) and then averaged the values from the three focal point estimates. Due to logistical constraints (e.g., water depth) we were unable to collect focal point habitat use for every target fish collected.

To determine species-specific habitat preferences, we broke each habitat parameter (depth, flow, and substrate) into bins; water depth into six bins (1-20 cm, 21-

40 cm, 41-60 cm, 61-80 cm, 81-100 cm, and >100 cm), water velocity into five bins

(<0.2 m/sec, 0.21-0.4 m/sec, 0.41-0.6 m/sec, 0.61-0.8 m/sec, and >0.8 m/sec), and substrate size into four bins (substrate sizes 1-2, 2.1-3.9, 4-5.9, and 6-8). We calculated habitat availability for each habitat parameter by determining the number of observations in each bin (summarized from our transect data) and then divided this number by the total number of observations. Similarly, we calculated the proportion of each habitat type used as the number of observations in each category, and divided by the total number of observations. We then determined habitat suitability by dividing the proportion of habitat used by the proportion available, and habitat preference by standardizing to the highest suitability value (i.e., habitat preferences scaled to one; Baltz 1990). 19

Data analysis

We assessed the relationship between total native fish density, collected in 2008

(fish/m2), and complex habitat, calculated as percent of each reach composed of pools, riffles, and backwaters, using a regression analysis. Using our 2007 and 2008 spring and summer fish catch data (fall fish catch data was excluded to prevent inflating density estimates for the few sites sampled during this period each year), we also used regression analysis to assess relationships between non-native fish density and total native fish density, as well as density for the ‘three species’ specifically. Furthermore, we performed a post-hoc analysis examining the relationship between non-native density calculated from our 2007 and 2008 spring and summer periods and flannelmouth sucker

GHQVLW\FDOFXODWHGIURPRXUILVKFDWFKGDWD:HFRQGXFWHGDOODQDO\VHVXVLQJDQĮ- level of 0.05.

In order to identify the most important abiotic and biotic factors potentially limiting the distribution and abundance of flannelmouth sucker, roundtail chub, and bluehead sucker, we performed a random forest analysis (Liaw and Wiener 2002) in

Program R (version 2.8.1, R Development Core Team 2005). Predictor variables incorporated in the model included the habitat attributes mean depth, mean substrate size, substrate standard deviation and coefficient of variation, percent pool, riffle, and backwater, and percent complex habitat, an aggregation of pools, riffles, and backwaters.

In addition, we also incorporated a biological variable, non-native fish density (calculated from spring and summer sampling in 2007 and 2008), and a metapopulation predictor, distance-to-source patch (Table 1). We considered the sites upstream of the Hatt Ranch 20 diversion (Patch 1) a potential source population for all three species due to high fish densities and presence of multiple life-stages (adults, juveniles, and age-0). In addition, for flannelmouth sucker, we considered patch 6 (the Green River) a source population due to a significant population in the mainstem Green River near the confluence with the

San Rafael River (Darek Elverud, Utah Division of Wildlife Resources, personal communication). Our response variables included presence and absence data for each species and site. We used the randomForest library (Liaw and Wiener 2002) in Program

R (version 2.8.1, R Development Core Team 2005) to develop and analyze all models.

We used the mean decrease in Gini index to assess variable importance for each species, with higher values representative of more important variables. We listed the top four predictors for each species after a post-hoc evaluation revealed that these four variables produced noticeably higher mean decrease in Gini values than the remaining six variables. We assessed model performance using four metrics: 1) sensitivity, the percentage of true positives (presences) correctly classified, 2) specificity, the percentage of true negatives correctly classified, 3) percent correctly classified (PCC), the amount of total observations (e.g., presence or absence) correctly predicted, and 4) Cohen’s kappa, the agreement between predicted and observed after correcting for chance effects (Manel et al. 2001). We used Cohen’s kappa to asses overall model performance: values between 0.4-0.6 are indicative of moderately performing models, values between, 0.6-0.8 are indicative of models with ‘substantial’ predictive capabilities, and values between

0.8-1, indicate almost perfect predictive capability (Landis and Koch 1977; Manel et al.

2001). 21

Table 1. Predictor variables, their range of values, units, and variable type, used in random forest modeling.

Predictor Variables (Abb.) Range Units Predictor

of variable type

values

Percent pool (pct_pool) 0-.22 % of reach Habitat

Percent riffle (pct_riffle) 0-.53 % of reach Habitat

Percent backwater (pct_backwater) 0-.02 % of reach Habitat

Percent non-run (pct_non_run) 0-.55 % of reach Habitat

Mean depth (XDepth) 26.60- cm Habitat

81.53

Mean substrate (XSubstrate) 1.89- Binned 1-8 (1=silt, Habitat

4.89 8=bedrock)

Substrate standard deviation 0-1.82 Relative to substrate Habitat

(SD_Substrate) size

Substrate coefficient of variation 0-.58 Relative to substrate Habitat

(CV_Substrate) size

Distance to source population for 0-31.2 km (from the Green Metapopulation

flannelmouth sucker (Distance_Source) River or the Hatt

Ranch Diversion)

Distance to source population for 0-63.9 km (from the Hatt Metapopulation

roundtail chub and bluehead sucker Ranch Diversion)

(Distance_Source_Round)

Non-native species density (Exotic_den) ~0-.11 fish/m2 Biological 22

RESULTS

Fish distribution and abundance

Native fish were distributed tri-modally across the longitudinal gradient of the

San Rafael River; we observed the highest native fish densities in reaches directly upstream and downstream of the Hatt Ranch Diversion (Fuller Bottom downstream to site 68), and sites near the confluence with the Green River (158-213), and the lowest native fish densities at intermediate sites 109-146 (Figure 5; Table 2). Flannelmouth sucker were relatively evenly distributed throughout our study reach (i.e., collected in 19 out of 25 reaches), whereas bluehead sucker and roundtail chub were much rarer and found in higher densities only in sites upstream of the Hatt Ranch diversion (Figure 5;

Table 2). The highest densities of bluehead sucker occurred at Fuller Bottom, and bluehead sucker were only present at three other sites; Tidwell Bottom, and sites 8 and

173 (Figure 5; Table 2). The highest densities of roundtail chub occurred at Fuller

Bottom, but this species was more evenly distributed throughout the study area, as it was collected at 14 of the 25 total sites (Figure 5; Table 2). Site 213, the reach directly upstream from the confluence with the Green River, was the only site where we observed endangered species of the Colorado River Basin, including Colorado pikeminnow and razorback sucker (Figure 5).

In contrast to native target fish species, we collected non-native fish at every sampling site within our study area. We observed the highest densities of non-native fish at site 213, and the lowest densities at Buckhorn Draw and Fuller Bottom. We did not 23 observe significant relationships between the ‘three species’ density (y) and non-native species density (x) (y=.0413x + .0041, P=0.32, df=24, R2 = 0.044), or total native fish density (y) and non-native species density (x) (y; y=.0075x + .0071, P=0.89, df =24, R2 =

0.0009) calculated from our 2007 and 2008 spring and summer sampling periods.

However, post-hoc analyses reveal a significant, positive correlation between flannelmouth sucker density (y) and non-native fish density (x), calculated from our 2007 and 2008 spring and summer sampling periods (y=0.0424x + 0.0002, P<.0001, df=24,

R2=.6265).

0.030

flannelmouth bluehead 0.025 age-0 sucker roundtail dace 0.020 endangereds 2

0.015 Fish/m 0.010

0.005

0.000 8 FB TB 15 27 35 48 62 68 90 95 BD 109 120 126 134 146 158 173 182 193 196 205 213

Detector Hatt Ranch Sample reach Figure 5. Native fish density (fish/m2), collected in 2008 across all sampling periods, by species, lifestage, and sample reach. Reaches are arranged in an upstream to downstream manner, and important landmarks are noted, such as the diversion at Hatt Ranch, and the PIT tag detector. 24

Table 2. Occurrences of bluehead sucker, flannelmouth sucker, and roundtail chub, collected from our spring, summer, and fall sampling periods in 2007 and 2008, by site, size, and patch number. All age-0 suckers collected were assumed to be flannelmouth suckers because all sacrificed age-0 suckers were identified as such. PIT tagged individuals are denoted with parentheses.

Bluehead Flannelmouth Roundtail

Adult Juvenile Age-0 Adult Juvenie7 Adult Juvenile Patch >200 71-200 <71 >250 1-250 Age-0 >200 71-200 Age-0 Site # mm mm mm mm mm <71 mm mm mm <71 mm FB 1 0 5(4) 0 4(4) 2(2) 1 0 5(4) 0 BD 1 0 0 0 3(3) 0 0 1(1) 1 0 TB 1 0 1 0 0 0 0 0 1 0 8 2 0 1 0 1(1) 3(2) 4 0 2(1) 0 15 2 0 0 0 3(3) 4(4) 3 0 1(1) 1 27 2 0 0 0 1(1) 2(2) 57 0 0 2 35 2 0 0 0 0 0 22 0 0 1 48 2 0 0 0 0 1(1) 0 0 0 0 62 2 0 0 0 0 2(2) 17 0 0 1 68 2 0 0 0 2(2) 11(6) 4 0 0 0 77 3 0 0 0 0 1 3 0 0 0 90 3 0 0 0 0 0 19 0 0 1 95 3 0 0 0 0 0 5 0 0 0 109 3 0 0 0 0 0 3 0 1 0 120 3 0 0 0 0 0 1 0 0 0 126 3 0 0 0 0 0 0 0 0 0 134 4 0 0 0 1(1) 1(1) 1 0 0 0 146 4 0 0 0 0 0 7 0 0 0 158 4 0 0 0 0 0 19 0 1 0 173 4 0 1 0 0 5(4) 40 0 0 0 182 4 0 0 0 0 0 7 0 0 0 193 5 0 0 0 0 8(5) 9 0 1 0 196 5 0 0 0 0 0 11 0 0 0 205 5 0 0 0 2(2) 1(1) 12 0 1(1) 0 213 5 0 0 0 13(13) 0 0 0 0 0 Total 0 8(4) 0 30(30) 40(26) 245 1(1) 14(7) 6

Patch delineation: seasonal and spatial habitat use

Juvenile and adult fish dominated the native fish catch during our spring (April-

May) sampling period. Juvenile (n=9) and adult (n=4) flannelmouth sucker comprised the majority of individuals collected during this period in patch 2. Similarly, adult

(n=14) and juvenile (n=2) flannelmouth sucker dominated the catch in patch 5. One juvenile roundtail chub and one bluehead sucker were captured in patches 2 and 4, respectively. We did not sample patch 1 during the spring sampling period (Figure 6A).

With the exception of patch 1, age-0 and juvenile individuals comprised the majority of the fish catch during our summer (June-July) sampling period. In patch 1, adult flannelmouth sucker (n=7) were more prevalent than juveniles (n=2), or age-0 (n=1) individuals. Still, juvenile roundtail chub (n=7), were much more prevalent than adult roundtail chub (n=1), and all bluehead sucker collected in patch 1 were juveniles (n=6).

Age-0 flannelmouth sucker dominated the catch in patch 2 (n=107, adult=3, juvenile=10) and comprised the vast majority of fish catch in patches 3 and 4 (n=31 and 74, respectively). The roundtail chub catch in patches 2-5 consisted entirely of age-0 (n=6) and juvenile (n=4) individuals. We collected one juvenile bluehead sucker in patch 2.

The flannelmouth sucker catch in patch 5 was similar to other patches during this time period: age-0 flannelmouth sucker (n=32) were much more prevalent than adults (n=3) and juveniles (n=4; Figure 6B).

Juveniles comprised the entirety of the fish catch during the fall (October) sampling period. We captured two juvenile flannelmouth sucker and one juvenile roundtail chub in patch 2. In addition, we captured two juvenile flannelmouth sucker in 26 patch 4 and three juvenile flannelmouth sucker in patch 5. We did not sample sites in patch 1 or 3 during this period (Figure 6C).

Macrohabitat availability

Complex habitat, an aggregation of pools, riffles, and runs, was distributed in the same trimodal pattern as exhibited by native fish. Reaches directly upstream and downstream of the Hatt Ranch diversion, as well as reaches near the confluence with the

Green River, contained high percentages of complex habitat, whereas most middle reaches are composed entirely of run (non-complex) habitat (Figure 7B). In addition, we observed a significant, positive relationship between total native fish density (y), calculated from our spring and summer sampling periods in 2007 and 2008, and complex habitat (x), collected at each sample site in 2008 (y=0.0277x + .0044, P=0.002, df=24,

R2=0.35). Fuller Bottom and sites 68, 173, and 193 had the highest percentages of complex habitat, whereas sites 90, 95, 109, 120, and 126 did not contain any complex habitat. The highest proportion of backwaters occurred in reaches 27 and 173. While smaller backwaters occurred in other reaches, backwater habitats in these reaches were large, deep, and still contained water at low discharges. Reaches with the highest percentage of riffles include sites 173, Fuller Bottom, 68, and 193, whereas sites 68,

Fuller Bottom, 15, Tidwell Bottom contained the highest proportion of pools (Figure 7A).

Fish habitat preferences

We found that bluehead sucker (n=7) prefer shallow areas (<20 cm), high water velocities (>0.81 m/sec), and relatively large substrate sizes (coarse gravel and small 27 cobble; Figure 8). Roundtail chub (n=12) prefer areas that are relatively deep (61-80 cm), with slower water velocities (<0.20 m/sec), and fine gravel (larger than sand, smaller than coarse gravel; Figure 9). Flannelmouth sucker (n=25) prefer relatively deep areas (61-80 cm), fast water velocities (>0.81 m/sec), and fine gravel (larger than sand, smaller than coarse gravel; Figure 10).

Native fish movement

Downstream movement of flannelmouth sucker peaked in early May and accounted for 93% of all flannelmouth sucker detections from the PIT tag antennae.

Multiple flannelmouth sucker, tagged in April and May, 2008, in sites upstream of the

PIT tag antennae, moved downstream during the period May 1 through May 15. Three of these fish, including a ripe female (489 mm TL, 1080 g), were initially tagged in site 15, and moved at least 57 km in 6-9 days. Two additional flannelmouth sucker, including a ripe male (360 mm TL, 272.5 g), tagged in site 27 in early and late May, 2008, were passively detected at the antennae loops in mid-August, indicating that the fish moved nearly 54 km over two months. Despite long distance movements by some flannelmouth sucker individuals (Figure 11A and 13), the minimum mean-distance moved by all flannelmouth sucker detected from passive recaptures at the PIT tag antennae was approximately 27 km (Figure 11B). Even with our extensive sampling, we actively recaptured only one tagged fish in 2007 and 2008, a juvenile flannelmouth sucker, tagged and recaptured at site 27 in late May and early June of 2008, respectively. Still, no bluehead sucker or roundtail chub were actively or passively recaptured throughout our study area. 28

The PIT tag antennae also detected significant movement from the endangered fishes of the Upper Colorado River Basin, including Colorado pikeminnow, razorback sucker, and bonytail chub, many of which were previously tagged and released throughout the Colorado River Basin. When combining flannelmouth sucker movement with the endangered species of the Colorado River Basin, we found that both upstream and downstream movement peaked in May and early June, coinciding with the historic

(1910-1918) and contemporary (1999-2007) spring spate (Figure 12). In 2008, there was still substantial movement, during a period when the San Rafael River was completely dewatered in 2007 (July and early August). Many of the endangered individuals were initially tagged or released (from propagation facilities) in large river systems such as the

Green, White, and Colorado Rivers, and were subsequently detected in the San Rafael

River; one Colorado pikeminnow moved a minimum distance of 281 km, a bonytail chub migrated 333 km, and a razorback sucker moved at least 68 km from the initial tagging site (Figure 13).

Data analysis

Results of the random forest analysis showed that the flannelmouth sucker classification model resulted in the highest model performance values for every metric.

The model correctly classified all the true presences (sensitivity), and correctly classified

83% of the true absences (specificity). In addition, this model produced a Cohen’s kappa value of 0.8837, which is indicative of model with ‘almost perfect’ predictive capabilities

(Manel et al. 2001; Table 3). The predictor variables most important in determining the distribution of flannelmouth sucker appear to be distance to a source population, mean

1.0 1.2 1.2 1.2 1.0

Adult A Adult Adult Adult Juvenile Juvenile Juvenile 1.0 1.0 Juvenile 1.0 Age-0 Age-0 0.8 Age-0 Age-0

0.8 0.8 0.8

0.6 0.5 Did not sample 0.6 0.6 0.6 0.4 0.4 0.4 0.4

0.2 0.2 0.2 0.2

0.0 0.0 0.0 0.0 0.0 Relative Frequency FMS BHS RTC FMS BHS RTC FMS BHS RTC FMS BHS RTC FMS BHS RTC

1.2 1.2 1.2 1.2 1.2

Adult Adult Adult Adult Adult B Juvenile Juvenile Juvenile 1.0 1.0 1.0 1.0 Juvenile 1.0 Juvenile Age-0 Age-0 Age-0 Age-0 Age-0

0.8 0.8 0.8 0.8 0.8

0.6 0.6 0.6 0.6 0.6

0.4 0.4 0.4 0.4 0.4

0.2 0.2 0.2 0.2 0.2

0.0 0.0 0.0 0.0 0.0 Relative Frequency FMS BHS RTC FMS BHS RTC FMS BHS RTC FMS BHS RTC FMS BHS RTC

1.0 1.2 1.0 1.2 1.2

C Adult Adult Adult 1.0 Juvenile 1.0 Juvenile 1.0 Juvenile Age-0 Age-0 Age-0

0.8 0.8 0.8

0.5 Did not sample 0.6 0.5 Did not sample 0.6 0.6

0.4 0.4 0.4

0.2 0.2 0.2

0.0 0.0 0.0 0.0 0.0 Relative Frequency FMS BHS RTC FMS BHS RTC FMS BHS RTC FMS BHS RTC FMS BHS RTC Patch 1 Patch 2 Patch 3 Patch 4 Patch 5 Figure 6. The relative frequencies of native fish catch for each species (roundtail chub, flannelmouth sucker, and bluehead sucker), life stage (adult, juvenile, age-0), and patch (1-5), summed from our 2007 and 2008 field seasons. Row A represents natives collected during the spring (April-May); row B represents fish collected in the summer (June-July); and row C

represents fish collected in the fall (October). 29

0.6 A Pool 0.5 Riffle Backwater

0.4 Flow

0.3

0.2 Proportion of reach 0.1

0.0 8 15 27 35 48 62 68 90 95 FB BD TB 109 120 126 134 146 158 173 182 193 196 205 213

Antennae Hatt Ranch 60 B

10 Percent complex habitat

8 15 27 35 48 62 68 90 95 FB BD TB 109 120 126 134 146 158 173 182 193 196 205 213

Antennae Hatt Ranch Site

Figure 7. Habitat composition by site and macrohabitat type (Panel A). Distribution and abundance of complex habitat, an aggregation pools, riffles, and backwaters, in the San Rafael River (Panel B). Reaches are organized in an upstream to downstream manner, and important landmarks are noted, such as the diversion at Hatt Ranch, and the PIT tag antennae. 31

1.2

1.0

0.8

0.6

Preference 0.4

0.2

0.0 0-.2 .21-.4 .41-.6 .61-.8 >.81 Water Velocity 1.2

1.0

0.8

0.6

Preference 0.4

0.2

0.0 1-20 21-40 41-60 61-80 81-100 >100 Water Depth 1.2

1.0

0.8

0.6

Preference 0.4

0.2

0.0 1-2 2.1-3.9 4-5.9 6-8 Substrate Size

Figure 8. Bluehead sucker microhabitat preferences for water depth (cm), water velocity (m/sec), and substrate size (binned 1-8). Habitat preferences are defined as the percent of habitat type used (collected at the site of native fish capture), divided by the percent available (collected from seven equally spaced transects within each reach). 32

1.2

1.0

0.8

0.6

Preference 0.4

0.2

0.0 0-.2 .21-.4 .41-.6 .61-.8 >.81 Water Velocity 1.2

1.0

0.8

0.6

Preference 0.4

0.2

0.0 1-20 21-40 41-60 61-80 81-100 >100 Water Depth 1.2

1.0

0.8

0.6

Preference 0.4

0.2

0.0 1-2 2.1-3.9 4-5.9 6-8 Substrate Size

Figure 9. Roundtail chub microhabitat preferences for water depth (cm), water velocity (m/sec), and substrate size (binned 1-8). Habitat preferences are defined as the percent of habitat type used (collected at the site of native fish capture), divided by the percent available (collected from seven equally spaced transects within each reach).

1.2

1.0

0.8

0.6

Preference 0.4

0.2

0.0 0-.2 .21-.4 .41-.6 .61-.8 >.81 Water Velocity 1.2

1.0

0.8

0.6

Preference 0.4

0.2

0.0 1-20 21-40 41-60 61-80 81-100 >100 Water Depth 1.2

1.0

0.8

0.6

Preference 0.4

0.2

0.0 1-2 2.1-3.9 4-5.9 6-8 Substrate Size

Figure 10. Flannelmouth sucker microhabitat preferences for water depth (cm), water velocity (m/sec), and substrate size (binned 1-8). Habitat preferences are defined as the percent of habitat type used (collected at the site of native fish capture), divided by the percent available (collected from seven equally spaced transects within each reach).

50 A

20 Frequency Frequency (%) 10

0 1 2 - 10 40 - 60 B

median mean

mean

0 10 20 30 40 50 60

Distance moved (km)

Figure 11. Frequency of minimum distance moved for flannelmouth sucker (calculated as the distance traveled from an individuals’ initial capture or release site, to the site of subsequent recapture; Panel A), and the minimum, maximum, median, and mean distance moved for flannelmouth sucker (Panel B), calculated as the distance between the site of active capture and passive detection at the PIT tag antennae. 35 1200 12 Unknown Period of no Upstream water in 2007 1000 Downstream 10

1910-1918 800 1999-2007 8

600 6

400 4 Number of fish of Number Discharge (cfs)Discharge 200 2

0 0 J F M A M J J A S O N D Month in 2008

Figure 12. Timing and direction of movement of flannelmouth sucker and the endangered fishes of the Colorado River Basin, as it relates to the historic (1910-1918) and contemporary hydrograph (1999-2007). All fish recapture data were collected from the stationary PIT tag antenna in 2008.

0.8 Flannelmouth sucker Colorado Pikeminnow Bonytail chub Razorback sucker 0.6

0.4 Frequency

0.2

0.0 0-10 11-50 51-100 101-200 201-300 300+ Minimum Distance Moved (km)

Figure 13. Frequency of movement, as it relates to the extent traveled (km), for flannelmouth sucker, Colorado pikeminnow, bonytail chub, and razorback sucker. Minimum distance moved is the distance traveled from an individuals’ initial capture or release site, to the PIT tag antennae in the San Rafael River. The distances moved are broken into six bins (0-10 km, 11-50 km, 51-100 km, 101-200 km, 201-300 km and 300+ km.

37 depth, exotic fish density, and complex habitat (pct_non_run; Table 3, Figure 14A).

The roundtail chub classification forest resulted in sensitivity and specificity values of

0.7857 and 0.6364, respectively. In addition, the kappa value 0.4262 indicates a model with ‘fair’ model performance (Manel et. al. 2001; Table 3). Variables most important to the classification of roundtail chub include distance to a source population, mean depth, percent pool, and percent of complex habitat (pct_non_run; Table 3, Figure 14B).

Although our bluehead sucker classification model correctly classified 76% of all observations (presences and absences), model performance was poor (kappa value of -

0.1194), due to complete misclassification of all true presences (predicted absences at all four sites with bluehead sucker observations; Table 3). Nonetheless, significant predictors in this model include distance to a source population, percent riffle, percent of complex habitat, and mean depth (Figure 14C). 38

3.0 A 2.5

2.0

1.5

1.0

0.5 MeanDecreaseGini 0.0

3.0 B 2.5

2.0

1.5

1.0

0.5 MeanDecreaseGini 0.0

1.6

1.4 C

1.2

1.0

0.8

0.6

0.4 MeanDecreaseGini 0.2

0.0

Xdepth pct_pool pct_riffle Xsubstrate Exotic_den pct_non_run pct_backwater SD_SubstrateCV_Substrate Distance_Source Predictor Variable

Figure 14. Variable importance plots for predictor variables used in classifying the presence of flannelmouth sucker (Panel A), roundtail chub (Panel B), and bluehead sucker (Panel C), using a random forest analysis. Higher values of mean decrease in Gini indicate variables more important to classification.

39 Table 3. Accuracy measures for the three species presence/absence random forest models using a total of 25 sites. Model sensitivity is the proportion of true positives (presences) correctly predicted, specificity is the percentage of true negatives (absences) correctly predicted, PCC is the proportion of observations (presences and absences) correctly classified, and Kappa is a measure of agreement between predicted and actual class assignments that corrects for chance (Manel et al. 2001). The top four abiotic and biotic predictor variables are also listed for each species.

Model Observed Sensitivity Specificity PCC Kappa Top four

presences predictors

Flannelmouth 19 1 0.8333 0.96 0.8837 Distance_Source,

sucker Xdepth,

Exotic_den,

pct_non_run

Roundtail 14 .7857 0.6364 0.72 0.4262 Distance_Source,

chub Xdepth,

pct_pool,

pct_non_run

Bluehead 4 0 0.9048 0.76 -0.1194 Distance_Source,

sucker pct_riffle,

pct_non_run,

Xdepth 40 DISCUSSION

Overview

In this study, we observed a tri-modal pattern of native fish distribution and abundance within the San Rafael River system, which appeared to directly parallel the availability of complex habitat. Our results indicate that native species are moving into the lower 64 km of the San Rafael River from two source patches, the Green River, which forms the downstream boundary of our study area, and sites upstream of the fish barrier (Fuller Bottom, Buckhorn Draw, and Tidwell Bottom). Our random forest analyses reveal that distance to a source population, along with reach-scale habitat characteristics, are consistently the best predictors of native fish distribution. Both upstream and downstream (passive and active) movement of PIT tagged fish peaked in

May and early June, which coincided with the ascending limb of the hydrograph. The timing and extent of these movements, along with the prevalence of age-0 and juvenile fish during the summer months, suggest the San Rafael River provides important spawning and rearing habitat for resident populations of native species within the Green

River.

Native fish distribution and movement

Flannelmouth sucker were widely distributed and relatively abundant within our study area, with multiple age-classes represented. In addition, our size-frequency analysis across multiple seasons and data from the PIT tag antennae, reveal a highly transitory nature of flannelmouth sucker populations below the fish barrier (Hatt Ranch diversion). Although we only detected upstream movement of one flannelmouth sucker 41 in late April, downstream movement of ripe fish tagged in April and May suggest adult flannelmouth sucker of the Green River select smaller tributaries to spawn and larger rivers to feed and overwinter. This conclusion is supported by the year-round presence of age-0 and juvenile flannelmouth sucker in the lower 64 km of the San Rafael River, and stresses the importance of tributaries in providing important spawning and rearing habitat for these fish (Douglas and Marsh 1998; Douglas and Douglas 2000).

Although roundtail chub were nearly as prevalent (present at 14 sites) as flannelmouth sucker, we did not document a significant flux of roundtail chub into or out of the San Rafael River in the spring, suggesting the absence of a roundtail chub metapopulation within the Green River, and/or a sedentary nature (Bryan and Hyatt

2004). As such, adult roundtail chub were underrepresented in our fish catch, occurring only in sites upstream of the diversion. However, we did document evidence of successful spawning, as indicated by the prevalence of age-0 and juvenile roundtail chub in sites upstream and downstream of the diversion (areas where they were absent in previous studies; Walker and Hudson 2004). In addition, the high proportion of age-0 and juvenile roundtail chub below potential source populations in patch 1, suggests that larval drift from upstream resident populations may be an important mechanism determining roundtail chub distribution in the lower San Rafael River, a process that has been reported elsewhere for roundtail chub (Carter et al. 1986) as well as for other native fishes in the Little Colorado River, Arizona (Robinson et al. 1998).

Compared to flannelmouth sucker and roundtail chub, bluehead sucker are the least prevalent of the ‘three species’ within the San Rafael River, and were only found at four of our study sites. Although large populations of bluehead sucker are still reported 42 in Ferron Creek above Millsite Reservoir (Kenny Breidinger, Utah Division of

Wildlife Resources, personal communication), the lack of adult and age-0 fish in the mainstem San Rafael River is of concern, as this species was found to be widespread during extensive sampling in 2003 (Walker and Hudson 2004). The small sample size collected during this investigation limits our ability to draw robust conclusions about factors which are potentially limiting their distribution and abundance, although they are likely similar to the threats faced by flannelmouth sucker and roundtail chub (e.g., habitat homogeneity, barriers to dispersal, and interactions with non-native species).

The presence of multiple age classes of roundtail chub and flannelmouth sucker, along with juvenile bluehead sucker, in isolated reaches upstream of the Hatt Ranch diversion (constructed in 1953), suggests these sites provide resident populations with adequate spawning, rearing, and feeding habitat. In addition, the presence of multiple age-classes of flannelmouth sucker and roundtail chub in sites downstream of the diversion strengthen the assertion that tributaries may provide: 1) spawning habitat for resident populations of large, mainstem rivers (e.g., Green and Colorado Rivers), 2) favorable conditions for growth of age-0 fish due to warmer water and unaltered diel flows, and 3) colonists for source populations in larger rivers (Vanicek 1970; Douglas and Marsh 1998; Douglas and Douglas 2000). However, the extremely altered hydrograph of the San Rafael River and other similar tributaries potentially limits spawning success and recruitment by preventing the establishment of clean gravel bars

(i.e., removal of fines) and backwater habitats, and stranding age-0 and juvenile fish in isolated pools, thereby reducing overall population persistence and preventing colonists from isolated, upstream tributaries from entering mainstem populations. This was 43 exemplified in 2007, when there was approximately a two month period of dewatering

(July-August), likely stranding age-0 and juvenile flannelmouth sucker and leading to high mortality for all species and life stages present (Bottcher, personal observation).

Recent genetic evidence supports our conclusion that the diversion at Hatt Ranch is a barrier to upstream fish movement, and that the lower 64 km of the San Rafael River below the fish barrier is acting as a sink: bluehead sucker collected upstream of the fish barrier are genetically distinct from the Green and Colorado River’s mainstem populations, suggesting little gene flow between these populations (Douglas et al. 2008).

Our capture-mark-recapture study design, coupled with continuous monitoring at the PIT tag antennae, allowed us to quantify active movement of PIT tagged fish and determine potential cues, a critical knowledge gap for flannelmouth sucker and the endangered fishes of the Colorado River Basin. Our results indicate rising tributary flows provide an important cue to begin large-scale movements in the San Rafael River; both upstream and downstream movement peaked in late May and early June, which corresponds with the spring spate of both the historic hydrograph (1910-1918), and to a lesser extent, the contemporary hydrograph (1999-2007). These findings contradict reports of flannelmouth sucker spawning behavior in the Grand Canyon (Weiss et al.

1998), yet correspond with findings from razorback sucker in the Green River (Tyus and

Karp 1990). More importantly, these results indicate that over-allocated tributaries, such as the San Rafael River, may no longer provide sufficient cues to undergo movement, especially in drought years.

In addition to the importance of identifying cues for movement, the extent moved between geographically distinct populations plays a critical role in determining 44 population viability (e.g., recolonization, genetic exchange; Hanski and Simberloff

1997; Fagan et al. 2002; Hilderbrand 2003). While some research indicates flannelmouth sucker make long-distance migrations (40-200 km) to spawning areas (Douglas and

Marsh 1998; McKinney et al. 1999), others suggest that large adults develop relatively narrow home ranges and move very little (Chart and Bergersen 1992). Our results indicate that flannelmouth sucker movement in the San Rafael River is pervasive, as all but one recaptured individual (passive and active) was captured at a location different from the initial tagging location. In addition, long distance, downstream movements

(>40 km) of ripe individuals tagged below the diversion, suggest populations of flannelmouth sucker in the Green River, undergo relatively long distance migrations to spawning areas in the San Rafael River. The empirical data collected here supports anecdotal evidence of large aggregations of flannelmouth sucker congregated downstream of the fish barrier, unable to reach historic spawning areas (Kenny

Breidinger, Utah Division of Wildlife Resources, personal communication). While the purpose of these movements may be obvious (e.g., spawning), the substantial variation among the direction and speed of movement remains intriguing; a few large adults traveled nearly 60 km in less than nine days whereas other, similarly-sized individuals traveled nearly the same distance over 60 days. Smaller individuals (125-150 mm TL) appear to inhabit the San Rafael River for longer periods of time, possibly due to warm water temperatures (relative to the Green River) and advantageous growth conditions (but note our small sample size of juvenile fish). These patterns of movement suggest tributaries may provide important habitat for the growth of age-0 and juvenile fish and 45 that downstream movement towards mainstem systems potentially occurs after reaching a minimum critical size.

Although the endangered fish species of the Colorado River Basin are better researched than the ‘three species’ (Minckley and Deacon 1991; Sigler and Sigler 1996), the role smaller tributaries play in their life-history strategies is not well understood. The long-distance movements we report for the Colorado pikeminnow, razorback sucker, and bonytail are in accordance with previous research (Sigler and Sigler 1996). The timing of upstream movement into the San Rafael River, which peaked in late May and early June, suggests this tributary may also provide spawning habitat for these endangered species.

This claim, although unsubstantiated without the collection of larval fish, is supported by the capture of a ripe razorback sucker and many more tuberculated Colorado pikeminnow and razorback sucker near the confluence with the Green River, and capture in nearby tributaries. For example, in 1996 and 1997, both age-0 and adult Colorado pikeminnow were reported in the lower Price River, Utah, a finding which overturned a long-held belief that the endangered species were extirpated from this tributary (Cavalli 1999).

Recent discoveries of Colorado pikeminnow in the Little Snake and Yampa Rivers

(upstream of the critical habitat designation), also suggest tributaries provide important habitat for multiple lifestages of Colorado pikeminnow and highlights their need for protection (Marsh et al. 1991; Finney 2006).

Habitat use and availability

Despite established populations of flannelmouth sucker, roundtail chub, and bluehead sucker in the San Rafael River, fundamental changes to channel morphology 46 (e.g., from a braided to a single-thread river) and riparian composition have severely altered the composition of in-stream habitat. The positive relationship we observed between complex habitat (pools, riffles, and backwaters) and native fish density, coupled with the low availability of complex habitat across the longitudinal gradient, indicate the river’s simplified planform may be limiting the distribution and abundance of the ‘three species’ here, as well as in other tributary systems experiencing similar alterations.

Habitat limitation may be especially prevalent for bluehead sucker, which prefer large substrate sizes, fast water velocities, and shallow water (e.g., riffles), extremely rare channel unit types in the lower San Rafael River. Despite our low sample size (n=7), our conclusions of habitat limitations for bluehead sucker are substantiated if we consider the direct connection between large substrate sizes and bluehead sucker foraging items (algae and aquatic invertebrates), and conformity with previous research (McAda 1977; Sigler and Sigler 1996). The rarity of pools in the lower San Rafael River may also be limiting for roundtail chub, which we found to prefer deep areas and slow water velocities among a variety of substrate sizes. While our analysis is potentially limited again by small sample sizes (n=12) and our necessity to pool lifestages (e.g., age-0, juvenile, and adults), our findings support previous associations between roundtail chub and complex, unmodified habitat, including deep pools (Cross 1978; Bestgen 1985; Carman 2006). In contrast, flannelmouth sucker demonstrate only slight differences in preference levels with regard to substrate size, water velocity, and water depth, highlighting their generalist nature, and also corroborating past findings (McAda 1977; Childs et al. 1998; Beyers et al. 2001). These later results help explain the widespread distribution of flannelmouth 47 sucker in the Colorado River Basin, relative to roundtail chub and bluehead sucker, which have more stringent habitat preferences.

Identification of limiting factors

Although comparisons of fish habitat use to habitat availability can indicate physical limitations to persistence, these approaches have inherent limitations (Rosenfeld

2003). As such, we performed a random forest analysis to statistically identify the most important biological, spatial, and physical predictors of distribution, for the ‘three species.’ Random forest analyses, a statistical tool used extensively in bioinformatics, has recently been applied to ecological data due to its ability to incorporate many predictor variables (even with few observations), identify and rank the most important predictor variables for species distributions, and, as a result of internal validation, produce more accurate predictions than some other widely used statistical techniques

(e.g., CART, logistic regression; Prasad et al. 2006; Cutler et al. 2007). These qualities make random forest analyses ideal for revealing factors limiting the distribution of the sensitive and endangered fish of the Colorado River Basin, all of which have experienced reductions in distribution and abundance (i.e., sample sizes are often low) and are faced with multiple biotic and abiotic threats (i.e., there are a large number of potential explanatory variables).

We found distance to a source population was the best predictor for each of the three species, followed by a species-specific habitat feature. The strong signal of distance to a source population in determining the distribution of the three species indicates there must be necessary conditions which foster the development of resident 48 source populations, which likely include both biological (e.g., low abundance of non- native species) and physical factors (e.g., barriers). Mean depth, and the percent of a reach composed of pools, riffles, and backwaters (pct_non_run), were two of the top four predictors of distribution for each species. The mean depth of a reach is potentially a surrogate for the prevalence of deep run and pool habitat, and thus, its importance in predicting the distribution of all three species is intuitive and supported by the available literature (Brouder et al. 2000; Beyers et al. 2001; Carman 2006). Furthermore, the amount of pool habitat is a strong predictor for roundtail chub, a finding supported by past research (Cross 1978; Carman 2006). The large influence of fluctuating flow levels on mean depth, however, a common occurrence in the highly regulated San Rafael River, limits transferable interpretations. The importance of these habitat variables in our random forest analyses strengthens conclusions that can be drawn from the positive and significant correlation we observed between total native fish density and a course measure of complex habitat.

In addition to distance to a source population discussed above, the density of non- native fish (exotic_den) was the third most important predictor of distribution for flannelmouth sucker. These results suggest that interactions between native and non- native fish may be influential in determining the distribution and abundance of the three species. However, post-hoc regression analyses actually reveal a significant, positive correlation between flannelmouth sucker density and non-native fish density, suggesting strong distributional overlaps and potentially similar habitat preferences. These findings imply that habitat restoration aimed at diversifying habitat for recovery of the ‘three 49 species’, may also benefit non-native species, making their removal a potentially important management consideration in the future.

In addition to determining the most important predictor variables, our random forest analyses allowed consideration of multiple, unique measures of model performance, including sensitivity, specificity, percent correctly classified (PCC), and

Cohen’s kappa. In this exploratory analysis, we were most interested in models which correctly classified true presences (sensitivity). Our flannelmouth sucker and roundtail chub classification models produced very high sensitivity values, allowing us to identify important determinants of distribution. Conversely, due to misclassification of all true presences (n=4), our bluehead sucker model did not allow us to identify factors which may limit their distribution. Furthermore, as others have noted, PCC, when viewed alone, can be a misleading measure of model performance, as it is overly influenced by the prevalence of an organism (Landis and Koch 1977; Manel et al. 2001; Olden et al.

2002). This issue is well illustrated with our bluehead sucker classification model, in which the overall model correctly classified 76% of all observations (indicative of a good model) while misclassifying all four presences (indicative of a bad model). In contrast,

Cohen’s kappa is a metric which takes into account the agreement between the predicted and observed due to chance, providing a robust analysis of model performance and allowing for meaningful interpretation of results. These results stress the importance of assessing multiple measures of model performance and suggest the relative weight given to each metric should be guided by research objectives.

50 Conclusions

When viewed simultaneously, our seasonal patch dynamics, movement patterns, genetic evidence, habitat preferences, random forest results, and characteristics of the altered hydrograph suggest that source-sink dynamics are driving the distribution and abundance (through survival) of the ‘three species’ in the San Rafael River. Essentially, our results indicate that roundtail chub and flannelmouth sucker are moving, both actively and passively, into the lower 64 kilometers of the San Rafael River from two source patches, the Green River below, and sites upstream of the fish barrier. In addition, the significant flux of large, sexually mature flannelmouth sucker in the spring, followed by summer and fall fish catches dominated by age-0 and juvenile individuals, indicates the San Rafael River provides important spawning habitat for resident populations of the

Green River. Similarly, the prevalence of age-0 and juvenile roundtail chub, in areas directly downstream of the fish barrier indicates the presence of upstream spawning populations. However, we would expect this area to be completely devoid of fish if the

Green River and patches upstream of the fish barrier did not exist, as these areas provide colonists and refuge during periods of dewatering. Furthermore, in addition to being completely dewatered during drought years, this large sink area is also generally characterized by inadequate habitat (with a few exceptions), especially for bluehead sucker and roundtail chub, which prefer complex mosaics of pools and riffles. These two forces likely work in concert to limit the growth, survival, and recruitment of the three sensitive species within the lower 64 kilometers of the San Rafael, preventing establishment of resident populations. 51 While our study design (capture-mark-recapture, coupled with a passive PIT tag detector) did not allow for the quantification of diel or localized movements, we were able to quantify the timing and extent of movement, and describe tributary use in general for flannelmouth sucker and the endangered species of the Colorado River Bain. The vast majority of research projects which address fish movement within the Colorado

River Basin involve radio telemetry technology (McAda 1977; Beyers et al. 2001).

While these studies provide precise estimates of small, localized movements, due to small sample sizes and intensive tracking effort, it is nearly impossible to characterize long- distance or seasonal movement. The utility of continuous detection by PIT tag antennae is apparent, and we argue the technology should be widely applied, as it is in the

Columbia River Basin (e.g., Zydlewski et al. 2006), to: 1) address fundamental questions regarding the role and importance of long-distance movement, 2) characterize tributary and mainstream river usage among multiple species and lifestages, 3) quantify vital rates

(e.g., survival), and 4) reveal potential limiting factors, including the effects of dams and diversions.

Management implications

Removing barriers to upstream dispersal, diversifying habitat, and establishing minimum in-stream flows would likely increase the probability of persistence and population viability of the native fish assemblage of the San Rafael River and tributary streams impacted by similar anthropogenic factors. The presence of multiple year classes of flannelmouth sucker and roundtail chub in sites upstream of the Hatt Ranch diversion, suggests these patches provide refugia, allowing these populations to persist despite no 52 immigration since 1953. Removing barriers to fish dispersal, such as the diversion at

Hatt Ranch, and providing minimum in-stream flows, will facilitate movement to these refuges during low-flow conditions, likely improving survival and recruitment rates of the populations overall. Native fish persistence and population viability may also be improved by diversifying habitat. State and federal agencies have already begun removing tamarisk, both mechanically and with release of the Japanese salt-cedar beetle

(Diorhabda elongate). However, increasing the quantity and distribution of complex habitat will likely require a combination of: 1) the removal of non-native vegetation, 2) the restoration of channel-forming flows, and 3) active channel reconfiguration. The relative contribution of each of these restoration strategies remains largely unknown and must be balanced against the realities of trying to recover and protect native fishes in an extremely over-allocated desert tributary in the Colorado River Basin.

53 REFERENCES

Bain, M.B. 1999. Substrate. M.B. Bain and N.J. Stevensen, editors. Aquatic habitat assessment: Common methods. American Fisheries Society, Bethesda, Maryland.

Baltz, D.M. 1990. Autecology. Pages 585-600 in C.B. Schreck and P.B. Moyle, editors. Methods for fish biology. American Fisheries Society, Bethesda, Maryland.

Bestgen, K.R. 1985. Distribution, biology and status of the roundtail chub, Gila robusta, in the Gila River Basin, New Mexico. Master’s thesis. Colorado State University, Fort Collins.

Bestgen, K. R., and D.L. Propst.1989. Distribution, status and notes on the ecology of Gila robusta (Cyprinidae) in the Gila River Drainage, New Mexico. The Southwestern Naturalist 34(3):402-412.

Beyers, D. W., C. Sodergren, J.M., Bundy, and K.R. Bestgen. 2001. Habitat use and movement of bluehead sucker, flannelmouth sucker, and roundtail chub in the Colorado River. Larval Fish Laboratory, Department of Fishery and Wildlife Biology, Colorado State University, Fort Collins.

Bezzerides, N.L., and K.R. Bestgen. 2002. Status review of roundtail chub (Gila robusta), flannelmouth sucker (Catostomus latipinnis), and bluehead sucker (Catostomus discobolus) in the Colorado River Basin. Larval Fish Laboratory Technical Report 118. Colorado State University.

Brouder, M.J., D.D. Rogers, and L.D. Avenetti. 2000. Life history and ecology of the roundtail chub Gila robusta, from two streams in the Verde River basin. Arizona Game and Fish Department Research Branch Technical Guidance Bulletin No. 3, Phoenix, Arizona.

Bryan, S.D., and M.W. Hyatt. 2004. Roundtail chub population assessment in the lower Salt and Verde Rivers, Arizona. State Wildlife Grant Final Report. Arizona Game and Fish Department. Phoenix, Arizona.

54 Bunte, K., and S. R. Abt. 2001. Sampling surface and subsurface particle-size distributions in wadeable gravel- and cobble-bed streams for analyses in sediment transport, hydraulics, and streambed monitoring. US Forest Service, General Technical Report RMRS-GTR-74.

Carman, S. M. 2006. Colorado River Basin chubs recovery plan. New Mexico Department of Game and Fish, editor. New Mexico Department of Game and Fish.

Carpenter, A.T. 1998. Element Stewardship Abstract for Tamarisk. The Nature Conservancy, Wildland Weeds Management & Research Program. http://tncweeds.ucdavis.edu/esadocs/tamaramo.html [05 March 2009].

Carter, J.G., V.A. Lamarra, and R.J. Ryel. 1986. Drift of larval fishes in the upper Colorado River. Journal of Freshwater Ecology 3:567-577.

Cavalli, P.A. 1999. Fish community investigations in the lower Price River, 1996-1997. Final Report prepared for the Recovery Implementation Program for Endangered Fishes in the Upper Colorado River Basin. Project No. 78.

Chart, T. E., and E.P. Bergersen. 1992. Impact of mainstem impoundment on the distribution and movements of the resident flannelmouth sucker (Catostomidae: Catostomus latipinnis) population in the White River, Colorado. The Southwestern Naturalist 37(1):9-15.

Cheetham, G.H. 1979. Flow competence in relation to stream channel form and braiding. GSA Bulletin 90(9):877-886.

Childs, M. R., R.W. Clarkson, and A.T. Robinson. 1998. Resource use by larval and early juvenile native fishes in the Little Colorado River, Grand Canyon, Arizona. Transactions of the American Fisheries Society 127:620-629.

Cross, J. N. 1978. Status and ecology of the Virgin River roundtail chub, Gila robusta seminuda (Osteichthyes: Cyprinidae). The Southwestern Naturalist 23(3):519- 527.

Cutler, D.R., T.C. Edwards, K.H. Beard, A. Cutler, and K. T. Hess. 2007. Random forests for classification in ecology. Ecology 88:2783-2792.

55 Douglas, M. E., and P.C. Marsh. 1998. Population and survival estimates of Catostomus latipinnis in Northern Grand Canyon, with distribution and abundance of hybrids with Xyrauchen texanus. Copeia 4:915-925.

Douglas, M. R., and M.E. Douglas. 2000. Late season reproduction by big-river Catostomidae in Grand Canyon (Arizona). Copeia 1:238-244.

Douglas, M.R., M.E. Douglas, and M.W. Hopken. 2008. Population genetic analysis of bluehead sucker [Catostomus (Pantosteus) discobolus] within and among river drainages in Utah. 2008 Report to the Utah Division of Wildlife resources. 1-39.

Dudley, R. K., and W.J. Matter. 2000. Effects of small green sunfish (Lepomis cyanellus) on recruitment of Gila chub (Gila Intermedia) in Sabino Creek, Arizona. The Southwestern Naturalist 45(1):24-29.

Fagan, W.F., P.J. Unmack, C. Burgess, and W.L. Minckley. 2002. Rarity, fragmentation, and extinction risk in desert fishes. Ecology 83:3250-3256.

Fagan, W.F., C. Aumann, C.M. Kennedy, and P.J. Unmack. 2005. Rarity, fragmentation, and the scale dependence of extinction risk in desert fishes. Ecology 86:34-41.

Finney, S.T. 2006. Colorado pikeminnow (Ptychocheilus lucius) upstream of critical habitat in the Yampa River, Colorado. The Southwestern Naturalist 51(2):262- 263.

Gilpin, M. E., and M. E. Soulé. 1986. Minimum viable populations: processes of species extinction. Pages 19-34 in M. E. Soule, editor. Conservation biology: The science of scarcity and diversity. Sinauer, Sunderland, Massachusetts.

Gran, K., and C. Paola. 2001. Riparian vegetation controls on braided stream dynamics. Water Resources Research 37(12):3275–3283.

Greger, P. D., and J.E. Deacon. 1988. Food partitioning among fishes of the Virgin River. Copeia 1988(2):314-323.

Guy, C.S., H.L. Blankenship., and L.A. Nielsen. 1996. Tagging and Marking. in B.R. Murphy and D.W. Willis, editors. Fisheries techniques (2nd Edition.) American Fisheries Society, Bethesda, Maryland.

56 Hanski, I. A., and D. Simberloff. 1997. The metapopulation approach, its history, conceptual domain, and application in conservation. in I. A. Hanski, and M. E. Gilpin, editors. Metapopulation biology: ecology, genetics, and evolution. Academic Press, San Diego, CA.

Hawkins, C.P., J.L. Kershner, P. Bisson, M. Bryant, L. Decker, S.V. Gregory, D.A. McCullough, K. Overton, G. Reeves, R. Steedman, and M. Young. 1993. A hierarchical approach to classifying stream habitat features. Fisheries 18:3-12.

Hilderbrand, R. H. 2003. The roles of carrying capacity, immigration, and population synchrony on persistence of stream-resident cutthroat trout. Biological Conservation 110:257-266.

Hoffman, Jr., R., and J.B. Dunham. 2007. Fish-movement ecology in high-gradient headwater streams-Its relevance to fish passage restoration through stream culvert barriers. U.S. Geological Survey Open-File Report 2007-1140.

Holden, P.B. 1991. Ghosts of the Green River: Impacts of Green River poisoning on management of native fishes. Pages 43-54 in WL. Minckley and J.E. Deacon, editors. Battle against extinction: native fish management in the American West. University of Arizona Press, Tucson.

Holden, P. B., and C.B. Stalnaker. 1975. Distribution and abundance of mainstem fishes of the middle and upper Colorado River Basins, 1967-1973. Transactions of the American Fisheries Society 104(2):217-231.

Johnson, B.M., P.J. Martinez, J.A. Hawkins, and K.R. Bestgen. 2008. Ranking predatory threats by nonnative fishes in the Yampa River, Colorado, via bioenergetics modeling. North American Journal of Fisheries Management 28(6):1941-1953.

Landis, J.R., and G.G. Koch. 1977. The measurement of observer agreement for categorical data. Biometrics 33:159–174.

Leprieur, F., O. Beauchard, S. Blanchet, T. Oberdorff, and S. Brosse. 2008. Fish invasions in the world’s river systems: when natural processes are blurred by human activities. PLoS Biology 6:404–410.

Liaw, A., and M. Wiener. 2002. Classification and regression by Random Forests. R News 2/3:18-22. URL http://CRAN.R-project.org/doc/Rnews/.

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

Marsh, P.C., M.E. Douglas, W.L. Minckley, and R. J. Timmons. 1991. Rediscovery of Colorado squawfish, Ptychochelius lucius (Cyprinidae), in Wyoming. Copeia 1991:1091–1092.

McAda, C.W. 1977. Aspects of the life history of the three catostomids native to the Upper Colorado River Basin. Master’s thesis. Utah State University. Logan.

McAda, C.W., Berry C.R., and Phillips, C.E. 1980. Distribution of fishes in the San Rafael System of the Upper Colorado River Basin. The Southwestern Naturalist 25(1):41-49.

McKinney, T., W.R. Persons, and R.S. Rogers. 1999. Ecology of flannelmouth sucker in the Lee’s Ferry tailwater, Colorado River, Arizona. Great Basin Naturalist 59(3):259-265.

Minckley, W.L, and Deacon, J.E. 1968. Southwestern fishes and the enigma of "endangered species": Man's invasion of deserts creates problems for native animals, especially for freshwater fishes. Science 159(3822):1424-1322.

Minckley, W. L. and J.E. Deacon, editors. 1991. Battle against extinction: native fish management in the American West. University of Arizona Press, Tucson,

Moyle, P. B. 1995. The decline of anadromous fishes in California. Conservation Biology 8:869-870.

Olden, J.D., D.A. Jackson, and P.R. Peres-Neto. 2002. Predictive models of fish species distributions: A note on proper validation and chance predictions. Transactions of the American Fisheries Society 131:329-326.

Olden, J.D., N.L. Poff, and K.R. Bestgen. 2006. Life-history strategies predict species invasions and extirpations in the Colorado River Basin. Ecological Monographs 76:25-40.

58 Osmundson, D.B., R. J. Ryel, V. L. Lamarra, and J. Pitlick. 2002. Flow-sediment- biota relations: Implications for river regulation effects on native fish abundance. Ecological Applications 12(6):1719-1739.

Pearsons, T.N., H.W. Li, and G.A. Lamberti. 1992. Influence of habitat complexity on resistance to flooding and resilience of stream fish assemblages. Transactions of the American Fisheries Society 121:427-436.

Petts, G.E., and P. Calow, editors. 1996. River flows and channel forms: selected extracts from the rivers handbook. Blackwell Science, Oxford.

Poff, N.L., J.D. Allan, M.B. Bain, J.R. Karr, K.L. Prestegaard, B.D. Richter, R.E. Sparks, and J.C. Stromberg. 1997. The natural flow regime: a paradigm for river conservation and restoration. BioScience 47(11):769-784.

Poff, N.L., J.D. Olden, D. Merritt, and D. Pepin. 2007. Homogenization of regional river dynamics by dams and global biodiversity implications. Proceedings of the National Academy of Sciences 104:5732-5737.

Prasad, A.M., L.R. Iverson, and A. Liaw. 2006. Newer classification and regression tree techniques: bagging and Random Forests for ecological prediction. Ecosystems 9:181-199.

Propst, D.L., and K. B. Gido. 2004. Responses of native and nonnative fishes to natural flow regime mimicry in the San Juan River. Transactions of the American Fisheries Society 133:922-931.

Ptacek, J. A., D.E. Rees, and W.J. Miller. 2005. Bluehead sucker (Catostomus discobolus): a technical conservation assessment. Miller Ecological Consultants, Fort Collins, CO.

Quist, M. C., M.R. Bower, and W.A. Hubert. 2006. Summer food habits and trophic overlap of roundtail chub and creek chub in Muddy Creek, Wyoming. The Southwestern Naturalist 51(1):22-27.

R Development Core Team. 2007. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3- 900051-07-0, URL http://www.R-project.org.

59 Rees, D. E., J.A. Ptacek, R.J. Carr, and W.J. Miller. 2005. Flannelmouth sucker (Catostomus latipinnis): a technical conservation assessment. Miller Ecological Consultants. Fort Collins, Colorado.

Robinson, A. T., R.W. Clarkson, and R.E. Forrest. 1998. Dispersal of larval fishes in a regulated river tributary. Transactions of the American Fisheries Society 127:772- 786.

Robinson, A. T., and M.R. Childs. 2001. Juvenile growth of native fishes in the Little Colorado River and in a thermally modified portion of the Colorado River. North American Journal of Fisheries Management 21:809-815.

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

Ross, S.T. 1986. Resource partitioning in fish assemblages: a review of field studies. Copeia 1986:352–388.

Ruppert, J. B., R.T. Muth, and T.P. Nesler. 1993. Predation on fish larvae by adult red shiner, Yampa and Green Rivers, Colorado. The Southwestern Naturalist 38(4):397-399.

Sigler, W. F., and J.W. Sigler. 1996. Fishes of Utah: a natural history. University of Utah Press, Salt Lake City.

Stacey, P. B., V. A. Johnson, and M. L. Taper. 1997. Migration within metapopulations: the impact upon local population dynamics. in I. A. Hanski, and M. E. Gilpin, editors. Metapopulation biology: ecology, genetics, and evolution. Academic Press, San Diego.

Stromberg J.C., K.J. Bagstad, J.M. Leenhouts, S.J. Lite, E. Makings. 2005. Effects of stream flow intermittency on riparian vegetation of a semiarid region river (San Pedro River, Arizona). River Research and Applications 21:925-938.

Stromberg, J.C., V.B. Beauchamp, M.D. Dixon , S.J. Lite, and C. Paradzick. 2007. Importance of low-flow and high-flow characteristics to restoration of riparian vegetation along rivers in arid south-western United States. Freshwater Biology 52(4):651-679.

60 Tal, M., K. Gran, A.B. Murray, C. Paola, and D.M. Hicks. 2004. Riparian vegetation as a primary control on channel characteristics in multi-thread rivers. Pages 43- 58,in S.J. Bennett and A. Simon, editors. Riparian vegetation and fluvial geomorphology. American Geophysical Union,

Tal, M., and C. Paola. 2007. Dynamic single-thread channels maintained by the interaction of flow and vegetation. Geology 35:347-350.

Tyus, H. M., and C.A. Karp. 1990. Spawning and movements of razorback sucker, Xyrauchen texanus, in the Green River Basin of Colorado and Utah. The Southwestern Naturalist 35(4):427-433.

Tyus, H. M., and Nikirk N.J. 1990. Abundance, growth, and diet of channel catfish, Ictalurus punctatus, in the Green and Yampa Rivers, Colorado and Utah. The Southwestern Naturalist 35(2):188-198.

Utah Division of Natural Resources. Division of Wildlife Resources. 2006. Conservation and management plan for three fish species in Utah. Publication No. 06-17. Salt Lake City, Utah

Vanicek, C. D. 1970. Distribution of Green River fishes in Utah and Colorado following closure of Flaming Gorge Dam. The Southwestern Naturalist 14(3):297-315.

Vanicek, C. D., and R.H. Kramer. 1969. Life history of the Colorado squawfish, Ptychocheilus lucius, and the Colorado chub, Gila robusta, in the Green River in Dinosaur National Monument, 1964-1966. Transactions of the American Fisheries Society 98(2):193-208.

Walker, C.A., and M. Hudson. 2004. Surveys to determine the current distribution of roundtail chub, flannelmouth sucker, and bluehead sucker in the San Rafael drainage, during 2003. Unpublished report. Utah Division of Wildlife Resources, Salt Lake City, Utah.

Ward, D. L., O. E. Maughan, S. A. Bonar, and W. J. Matter. 2002. Effects of temperature, fish length, and exercise on swimming performance of age-0 flannelmouth sucker. Transactions of the American Fisheries Society 131:492– 497.

61 Ward, D. L., and S.A. Bonar. 2003. Effects of cold water on susceptibility of age flannelmouth sucker to predation by rainbow trout. The Southwestern Naturalist 48(1):43-46.

Webb, R.H., T.S. Melis, and R.A. Valdez. 2002. Observations of environmental change in Grand Canyon, Arizona. U.S. Geological Survey Water Resources Investigations Report 02-4080.

Weiss, S.J., E.O. Otis, and O.E. Maughan. 1998. Spawning ecology of flannelmouth sucker Catostomus latipinnis (Catostomidae) in two small tributaries of the lower Colorado River. Environmental Biology of Fishes 52:419-433

Zydlewski, G.B., G. Horton, T. Dubreuil, B. Letcher, S. Casey, and J. Zydlewski. 2006. Remote monitoring of fish in small streams: a unified approach using PIT tags. Fisheries 31:492-502.