Beaver Dams Shift Desert Fish Assemblages Toward Dominance by Non-Native Species (Verde River, Arizona, USA)

Home , Beaver dam

Ecology of Freshwater Fish 2014 Ó 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd ECOLOGY OF FRESHWATER FISH

Beaver dams shift desert fish assemblages toward dominance by non-native species (Verde River, Arizona, USA)

Polly P. Gibson1, Julian D. Olden1, Matthew W. O’Neill2 1School of Aquatic and Fishery Sciences, University of Washington, 1122 NE Boat St., Seattle, WA 98105, USA 2Bubbling Ponds Native Fish Conservation Facility, Arizona Game and Fish Department, 1600 N. Page Springs Rd., Cornville, AZ 86325, USA

Accepted for publication April 13, 2014

Abstract – The reintroduction of beaver (Castor canadensis) into arid and semi-arid rivers is receiving increasing management and conservation attention in recent years, yet very little is known about native versus non-native fish occupancy in beaver pond habitats. Streams of the American Southwest support a highly endemic, highly endangered native fish fauna and abundant non-native fishes, and here we investigated the hypothesis that beaver ponds in this region may lead to fish assemblages dominated by non-native species that favour slower-water habitat. We sampled fish assemblages within beaver ponds and within unimpounded lotic stream reaches in the mainstem and in tributaries of the free-flowing upper Verde River, Arizona, USA. Non-native fishes consistently outnumbered native species, and this dominance was greater in pond than in lotic assemblages. Few native species were recorded within ponds. Multivariate analysis indicated that fish assemblages in beaver ponds were distinct from those in lotic reaches, in both mainstem and tributary locations. Individual species driving this distinction included abundant non- native green sunfish (Lepomis cyanellus) and western mosquitofish (Gambusia affinis) in pond sites, and native desert sucker (Catostomus clarkii) in lotic sites. Overall, this study provides the first evidence that, relative to unimpounded lotic habitat, beaver ponds in arid and semi-arid rivers support abundant non-native fishes; these ponds could thus serve as important non-native source areas and negatively impact co-occurring native fish populations.

Key words: Castor canadensis; desert fishes; assemblage structure; impoundments; microhabitat use; dryland stream

of beaver populations also raises new questions for Introduction ecosystem management, especially within arid and Management of freshwater ecosystems takes place at semi-arid environments (Gibson & Olden 2014). For the intersection of numerous ecological and social example, the introduction and continuing spread of trends, which can create novel challenges for natural invasive fishes pose a substantial threat to freshwater resource conservation and policy. Recent decades ecosystems (Cucherousset & Olden 2011), and it is have witnessed growing interest in employing or unclear how native and non-native fish species are mimicking the remarkable engineering abilities of using beaver pond habitat, and thus how contempo- beaver (Castor canadensis and C. fiber) in stream rary fish assemblages might respond to changes in restoration efforts (Pollock et al. 2007; DeVries et al. the abundance of beaver ponds as beaver return to 2012). After widespread extirpations of North Ameri- more of their historical range. can beaver during several centuries of intense fur Beavers are widely recognised as ecosystem engi- trade, hunting restrictions combined with deliberate neers with major influence on the structure of aquatic reintroductions contributed to successful beaver ecosystems. Beaver dams create distinct lentic habitat recovery (Pollock et al. 2003). However, restoration patches within the lotic stream corridor; these beaver

Correspondence: P. P. Gibson, School of Aquatic and Fishery Sciences, University of Washington, 1122 NE Boat St., Seattle, WA 98105, USA. E-mail: [email protected] doi: 10.1111/eff.12150 1 Gibson et al. ponds promote landscape heterogeneity and alter accompanied by dramatic declines of native fish pop- stream hydrology, sediment dynamics, nutrient ulations (Olden & Poff 2005; Rinne & Miller 2006). cycling, and aquatic and riparian biotic communities Predation by non-native fish on native species, (Rosell et al. 2005). Previous research has shown that accompanied by competition and indirect effects of beaver pond habitat can have substantial conse- predation risk, is considered the primary mechanism quences for stream fish populations: individual spe- for the widespread replacement of native fish (Minck- cies, particularly salmonids, have shown higher ley et al. 2003). Numerous studies have documented growth, survival or abundance within beaver ponds direct predation, agonistic interactions and shifts in relative to unimpounded stream habitat (Murphy resource use by native fishes in response to non- et al. 1989), and some species preferentially select natives, including both large-bodied predators and beaver pond habitats during certain life stages or abundant small insectivores (Cucherousset & Olden environmental conditions (e.g., Schlosser 1995). In 2011). Non-native crayfishes (primarily Orconectes addition to upstream pond habitat, high-velocity habi- virilis and Procambarus clarkii) have also been intro- tat immediately downstream of beaver dams may duced and spread throughout the LCRB, although favour fluvial specialists (Smith & Mather 2013). historically crayfish were absent from the entire Col- Beaver ponds can influence composition of fish orado River Basin (Martinez 2012; Moody & Taylor assemblages at scales ranging from individual ponds 2012). Effective management for conservation of (Snodgrass & Meffe 1999) to entire drainages (Sch- native fish requires understanding how habitat and losser & Kallemeyn 2000). environmental features can influence their coexis- Results demonstrating benefits of beaver ponds for tence with non-native fish and crayfish. salmonid fishes (e.g., Pollock et al. 2004) tend to Very little is known of the influence of beaver lead to the assumption that beaver dams will improve dams on fish assemblages in the LCRB. Historically, habitat for native fish. However, this remains an beaver dams were abundant in parts of the basin untested hypothesis for many systems (Kemp et al. (Gibson & Olden 2014), and presumably beaver 2012), especially in the context of invasion by non- ponds would have been occupied by native fishes. native fish. In southern Chile, where both beaver and Current reintroductions of extirpated beaver popula- rainbow trout (Oncorhynchus mykiss) are non-native, tions are usually motivated by growing recognition of the density of native puye (Galaxias maculatus) was the potential value of beaver engineering for achiev- found to be substantially higher in stream reaches ing conservation goals (e.g., Fredlake 1997). How- with beaver dams, regardless of whether trout were ever, within the contemporary landscape, promotion present (Moorman et al. 2009). Habitat heterogeneity of beaver dams may also have unexpected conse- created by beaver dams and backwater habitat along quences: natural history and previous studies suggest the Provo River (Utah) supported several species of that the habitat conditions created by beaver ponds native fish in a system otherwise dominated by non- are likely to benefit some non-native fishes in the native brown trout (Salmo trutta) (Billman et al. LCRB. Many of the common non-native species are 2012). In another Utah stream, native cutthroat trout associated with pool habitat or slow current velocities (Oncorhynchus clarkii utah) moved across beaver (Olden et al. 2006; Rinne & Miller 2006). In a long- dams more frequently than non-native brown trout or term study of fish assemblages in the Gila River sys- brook trout (Salvelinus fontinalis) (Lokteff et al. tem (New Mexico), higher densities of non-native 2013). These studies reveal that beaver dams can fish were associated with low flows and greater abun- have complex effects on native and non-native fish, dance of pools (Propst et al. 2008). Rinne and Miller but their influence on the overall structure of contem- (2006) conclude that deep (>2 m) pools provide opti- porary fish assemblages containing both native and mum habitat for large non-native predatory fishes non-native species is unknown. such as bass and catfishes, while lack of pools Community dynamics between native and non- reduces their abundance. Additionally, beaver ponds native fishes are particularly relevant in arid and may increase success of non-native fishes by mitigat- semi-arid rivers of the American Southwest. The ing the effects of extreme high and low flows. Deep native fish fauna of the lower Colorado River Basin pools behind beaver dams could promote survival – (LCRB) is both highly endemic and highly endan- especially of large-bodied species – through harsh gered (Olden & Poff 2005): two-thirds of the extant drought conditions, while the physical beaver dam native fish species in Arizona, for example, are feder- structure could provide immediate refuge from ally listed under the US Endangered Species Act floods. Although these hydrological effects of beaver (Turner & List 2007). Non-native fishes pose a pri- dams would apply to native and non-native fishes mary threat to the persistence of the native fish fauna. alike, non-native fishes lack an evolutionary history Numerous species have been introduced to the with the region’s highly variable streamflow (Olden LCRB, and spread of non-native fish has been et al. 2006), and thus non-native species might bene-

2 Beaver dams shift desert fish assemblages fit disproportionately from mitigation of these hydro- system (J. Agyagos, pers. comm.; P. Gibson, logical disturbances. unpublished data). The strong impacts of non-native fishes on native The Verde River system, especially the headwaters populations and the potential for beaver ponds to pro- of the mainstem, is an important site for conservation mote non-native species suggest the general hypothe- of native fish in Arizona (Turner & List 2007; Pool sis that beaver dams in the LCRB may lead to fish et al. 2013). At least twelve fish species were native assemblages dominated by non-native species, with to the system, but the fish assemblage is changing potential negative effects on native fishes. In this rapidly, and only five native species have been study, we investigate how beaver ponds influence the observed since 1997 (Rinne 2005). Many of the ori- structure of mixed native and non-native fish assem- ginal native species have been absent from the Verde blages in the Verde River system in Arizona. Specifi- River system for decades, but federally endangered cally, we address the following questions: (i) Does spikedace (Meda fulgida) was last documented in the fish assemblage structure differ between beaver pond Verde River in 1997, and its current status in the and lotic habitat types? (ii) How do native and non- Verde system is uncertain. Two other small-bodied native species differentially occupy these habitats? native species, longfin dace (Agosia chrysogaster) (iii) Do fish assemblage structure and the influence and speckled dace (Rhinichthys osculus), have of beaver ponds differ between the mainstem river become rare in the Verde River mainstem, and native and its tributary streams? roundtail chub (Gila robusta) has been proposed for federal listing. Numerous non-native fishes are present in the Verde system, including several species of Methods centrarchids, ictalurid catfishes and minnows (Table 1; Rinne 2005). Fish community composition Study area of the Verde River varies widely among years, but The Verde River, a semi-arid tributary of the LCRB, non-native fish have generally outnumbered natives drains over 17,000 km2 of central Arizona (Fig. 1). for the past two decades. Management goals for fish The perennial mainstem river runs approximately assemblages of the upper Verde River include main- 300 km from its headwaters in the Big Chino Wash tenance of current native populations and reintroduc- (elevation = 1325 m) to its confluence with the Salt tion of extirpated species, including spikedace (Rinne River north of Phoenix, AZ (elevation = 402 m), 2005, 2012). while numerous perennial tributaries contribute run- off from the southwestern edge of the Colorado Pla- Site selection teau (Rinne 2005; Blasch et al. 2006). Our study region included stream reaches in the upper Verde We surveyed fish assemblages in two different habi- River mainstem and in five major tributaries tat types within the Verde River mainstem and five (Fig. 1). The upper Verde River is largely free-flow- tributaries: (i) beaver ponds formed by dams across ing, a rarity for perennial rivers in the Desert South- the main stream channel (representing the most com- west. However, several diversion dams and small mon position of dams in this system); and (ii) stream impoundments are present in the watershed, primar- reaches without beaver dams (lotic sites). Focusing ily on tributaries. Development is limited within the on stream segments with a high probability of beaver Verde system, and primary anthropogenic distur- dam presence (as suggested by regional managers), bances are livestock grazing and reduction in base continuous longitudinal census of 64 km of streams flows as a result of groundwater withdrawals. in the Verde River system located 37 functional (i.e., Hydrology in the upper Verde River mainstem is impounding water) beaver dams in April 2012 (data characterised by relatively steady, spring-fed base available at http://dx.doi.org/10.6084/m9.figshare. flow, with high-flow events that vary in magnitude 867703). The highest density of beaver dams and and timing among years in response to storm run- the largest and most complex ponds were located off. Tributary hydrology, relative to the mainstem, is on the mainstem (23 dams in 24.0 km of stream), more variable and includes more snowmelt run-off. while dams on the tributaries were less common (14 Mean annual streamflow in the upper mainstem was dams in 40.4 km) and usually smaller than mainstem 1.2 m3ÁsÀ1 for the period 1963–2004; streamflow in dams. the four perennial and one ephemeral (Dry Beaver Pond sampling sites (mainstem n = 6; each tribu- Creek) tributaries examined in this study ranged tary n = 0–3) were designated proportionally to pond from 0.9 to 2.3 m3ÁsÀ1 during a similar period of abundance within a given stream (Fig. 1); fish sam- record (1961–2013 and 1981–2013, respectively) pling occurred in the first 100 m upstream of each (Blasch et al. 2006). Beavers are common but bea- dam. With one exception, all ponds extended beyond ver dams are spatially clustered throughout the the sampled 100 m: on average, ponds were 180 m

3 Gibson et al.

MAINSTEM SITES

Sycamore Creek TRIBUTARY SITES

Verde River

Lotic sites Oak Creek Pond sites Stream length searched for beaver dams Dry Beaver Creek

Wet Beaver Creek

West Clear Creek

Fig. 1. Map of the study area in the Verde River system and fish sampling sites, distributed among the mainstem and five tributaries. Gray shading indicates stream segments that were searched for beaver dams. long (range = 90–250 m) with a maximum depth of adjusted as needed to account for length and variable 1.5 m (range = 1.1–2.2 m). Lotic sampling sites habitat complexity within the reach. All fish captured (mainstem n = 10; each tributary n = 3–4) were dis- were identified, enumerated and released alive after tributed throughout the stream segments that had electrofishing was complete. In rare instances, fish been searched for beaver dams, at distances ranging observed but not captured (i.e., misses or large from 40 to 2650 m to the nearest dam (when dams schools of small fish) were included in counts when were present). At each lotic site, we designated a the fish could be confidently identified. Although sampling reach approximately 100 m long electrofishing is somewhat size-selective for larger (range = 80–115 m), which typically contained 1–2 individuals, this sampling approach is one of the least pool-riffle-run sequences. selective of all active fish capture methods (Barbour et al. 1999). Additionally, in our study, groups of small fish were often readily visible during sampling, Sampling methods increasing our confidence in the effectiveness of this We sampled 26 lotic and 12 pond sites during late method for characterising lotic fish communities. spring (May–June) 2012. We isolated the ~100 m Pond sites were typically too large, deep and com- lotic reaches with block nets (4 mm bar mesh) and plex to be effectively sampled with backpacking elec- sampled with double-pass removal backpack electro- trofishing, and too remote for use of boat fishing (Smith-Root Model L-24, 180–320 V, electrofishing; instead, we followed previous studies 30 Hz), where 90% of fish and 100% of species were (e.g., Smith & Mather 2013) in using a standardised caught after two passes relative to three-pass removal combination of baited minnow traps (12 minnow of selected sites (data not shown). A standard effort traps and 12 crayfish traps set for 12 h at 0.2–1.5 m level of 1200 total seconds (st. dev. = 162 s) was depth), trammel nets (3 nets, 9.1 9 1.8 m, 0.3 m

4 Beaver dams shift desert fish assemblages

Table 1. Our study recorded four native fish species, nine non-native fish species and one non-native crayfish species in the upper Verde River mainstem and tributaries.

Proportion of sites where species was present

All sites Pond sites Lotic sites Family Species† Common name Code‡ n = 38 n = 12 n = 26

Native Catostomidae Catostomus clarkii Desert sucker CACL 0.45 0.00 0.65 Catostomus insignis Sonora sucker CAIN 0.42 0.58 0.35 Cyprinidae Gila robusta Roundtail chub GIRO 0.29 0.25 0.31 Rhinichthys osculus Speckled dace RHOS 0.05 0.00 0.08 Non-native Centrarchidae Micropterus dolomieu Smallmouth bass MIDO 0.89 0.92 0.88 Lepomis cyanellus Green sunfish LECY 0.61 0.83 0.50 Ambloplites rupestris Rock bass AMRU 0.05 0.00 0.08 Cyprinidae Cyprinella lutrensis Red shiner CYLU 0.37 0.42 0.35 Cyprinus carpio Common carp CYCA 0.16 0.33 0.08 Pimephales promelas Fathead minnow PIPR 0.08 0.25 0.00 Ictaluridae Ameiurus natalis Yellow bullhead AMNA 0.29 0.25 0.31 Poeciliidae Gambusia affinis Western mosquitofish GAAF 0.37 0.50 0.31 Salmonidae Oncorhynchus mykiss Rainbow trout ONMY 0.11 0.08 0.12 Non-native crayfish Cambaridae Orconectes virilis Northern crayfish ORVI 1.00 1.00 1.00

†Within each family, species are listed in order of commonness. ‡Combination of first two letters of genus and first two letters of species. Used to indicate species on ordination plots.

wall size, 51 mm mesh size, set for 3 h), seine hauls Statistical analysis (4 mm mesh, 5 hauls sampling approximately 60 m2 total) and snorkel survey (one person snorkelled full Sampling effort was constant among all pond sites pond perimeter plus one transect down pond centre and among all lotic sites; thus, species abundance for line; visibility was generally excellent in tributary each site was calculated as the total number of indi- sites but limited in mainstem sites) to achieve an viduals of that species captured at the site across all accurate representation of species occurrence and rel- gear types. Raw abundances were log-transformed to ative abundance. Most of these sampling methods reduce the influence of highly abundant species and would have been ineffective in our lotic sites: water then standardised to relative abundance at each site. depth was often too shallow for snorkelling or for Using relative rather than raw abundances allowed us effective capture using trammel nets. Therefore, we to make comparisons between pond and lotic sam- selected sampling methods that maximised efficiency pling sites despite different sampling methods. To and minimised error in the characterisation of fish examine differences in relative abundance of individ- assemblages within each habitat type. ual species between pond and lotic sites, we calcu- Crayfish were also sampled at all sites. In pond lated the mean of pond-lotic pairwise differences in sites, all crayfish captured in seine hauls, minnow relative abundance for each species (separately for traps and crayfish traps were counted; we calculated mainstem and tributary sites). Within pond sites and crayfish catch per unit effort (CPUE) for each gear within lotic sites, we also tested for correlations type, as well as the sum total of crayfish caught between abundance of crayfish and abundance of across all gears at each site. In lotic sites, we used smallmouth bass (Micropterus dolomieu) and of timed D-net (0.5 mm mesh) sweeps to sample cray- all non-native fishes combined (abundances fish density along with other benthic macroinverte- log-transformed). brates. The crayfish sample for each lotic site Patterns in multivariate fish assemblage structure consisted of 10 point samples, covering a total sur- were examined using principal coordinate analysis face area of 4 m2 and total sampling effort time of (PCoA; Legendre & Legendre 1998) of species rela- 150 s, in standardised microhabitats from both the tive abundances at each site. Notably, ordinations upstream and downstream ends of the sampling using presence–absence data produced qualitatively reach. All crayfish with carapace length ≥3 mm and similar results. PCoA is similar to principal compo- ≤30 mm (larger crayfish were not sampled effectively nent analysis (PCA), but generalised to allow the use with this methodology) were sorted from the sample of any ecologically relevant dissimilarity measure. and counted. For this analysis, we used Bray–Curtis distance

5 Gibson et al.

(which excludes double absences and which is widely Sonora sucker (Catostomus insignis) and non-native used for community abundance data) and corrected red shiner were found primarily in the mainstem but for negative eigenvalues. Rare species (present in ≤2 also in one tributary. Desert sucker, speckled dace sites), which can have a disproportionate effect on and rock bass (Ambloplites rupestris) were present multivariate results (Gauch 1982), were removed only in lotic samples, while fathead minnow was from this analysis. Statistical significance of the only found in pond sites. Sonora sucker was the only PCoA axes was determined according to the broken native species frequently found in ponds. Speckled stick rule (P < 0.05; Legendre & Legendre 1998). dace and rock bass were found only in two lotic sites; Raw species correlations with the axes were overlaid these rare species were excluded from multivariate as vectors on the ordination plots to examine the con- analyses. tribution of individual species to observed patterns in Almost without exception, non-native fishes community data. outnumbered natives in samples (mean = 85% non- We formally assessed differences between pond native; range = 14–100%), and the non-native frac- and lotic fish assemblage using permutational multi- tion of the assemblage was consistently greater in variate analysis of variance (perMANOVA; Anderson pond than in lotic sites, in both mainstem and tribu- 2001). A significant difference identified by this tary locations (Fig. 2). This difference primarily approach may indicate a difference in group mean reflect large numbers of small non-native fish in most (i.e., location in PCoA ordination space) or in group pond samples. The greater dominance of non-native variability (i.e., spread in ordination space) or a com- fish in ponds was also apparent across individual spe- bination of the two. Therefore, we also tested for cies, with the non-native species displaying generally homogeneity of multivariate dispersions between higher relative abundances in ponds, whereas native groups (Anderson 2006), a multivariate analogue to species were relatively more abundant in lotic sites the univariate Levene’s test, to locate the source of (Fig. 3). The difference in relative abundance any significant differences identified by the perMA- between habitat types was especially strong for desert NOVA. This dispersion test computes an F-statistic sucker and green sunfish in tributaries. Smallmouth to compare the average distance between an individ- bass and, to a lesser extent, yellow bullhead ual sample and the group centroid defined in the (Ameiurus natalis) were the only non-native species PCoA space of a chosen dissimilarity measure (in our whose mean relative abundance was greater in lotic case, Bray–Curtis dissimilarity). Both of these multi- sites than in pond sites at both mainstem and tribu- variate tests estimate statistical significance (P value) tary locations. by permuting the appropriate least-squares residuals. Northern crayfish (Orconectes virilis) were present All analyses were performed in R version 2.13.1 at all sites, although density varied widely among (R Development Core Team 2011), using the vegan both lotic sites (range = 0.3–28.3 individuals per m2) package 2.0–2 for multivariate community analysis and pond sites (range = 0–37 individuals per minnow (Oksanen et al. 2011). trap) (Table 2). Within pond sites, there was a mar- ginally significant positive relationship between abundance of crayfish and abundance of smallmouth Results bass (r = 0.55, P = 0.07), but not between crayfish and all non-native fish combined (r = 0.14, Species occurrence and abundance P=0.66). By contrast, in lotic sites, there was no During our study, we recorded >12,500 individual evidence of a relationship between abundance of fish from four native and nine non-native species crayfish and either smallmouth bass (r = À0.01, (Table 1). Smallmouth bass was the most frequent P=0.97) or all non-native fish combined non-native fish, followed by green sunfish (Lepomis (r = À0.24, P=0.23). cyanellus), western mosquitofish (Gambusia affinis) and red shiner (Cyprinella lutrensis). Desert sucker Fish assemblages (Catostomus clarkii) was the most frequent native species, found in 45% of sites (Table 1). Species Multivariate ordination revealed strong differences in richness per site varied from 1 to 8 species, with a fish assemblage structure between mainstem and trib- considerably larger species pool in the Verde River utary sites (Fig. 4a), reflecting the different species mainstem (mean spp. richness Æ SE = 6.1 Æ 0.35, pools in the two locations. The first three principal range 4–8) than in tributaries (2.7 Æ 0.20, range 1– component axes were statistically significant, and col- 4). Several species were found only in the mainstem: lectively explained 52% the total variation in fish native roundtail chub and non-native western mosqui- assemblage. A two-way perMANOVA analysis indi- tofish, common carp (Cyprinus carpio) and fathead cated that fish assemblage differed significantly by minnow (Pimephales promelas). Additionally, native both location (mainstem vs. tributary) and habitat

6 Beaver dams shift desert fish assemblages

Fig. 2. Boxplots showing the distributions of the proportion native fish (lighter shades) and its converse, the proportion non-native fish (darker shades), across lotic sites (left panel) and across pond sites (right panel). Distributions are calculated separately for mainstem sites (solid fill) and for tributary sites (hatched fill).

Fig. 3. Mean difference in relative abundance between lotic and pond sites (averaged for all pond site-lotic site pairwise comparisons, within mainstem sites and within tributary sites), for each species recorded in this study (with the exception of speckled dace and rock bass). Positive values (i.e., bars above the midline) indicate higher relative abundance in lotic sites than in pond sites, while negative values indicate higher relative abundance in pond sites. Error bars show standard error of the mean. type (pond vs. lotic), although location contributed An ordination of mainstem sites showed modest the greater component of variation (Table 3). There differentiation between pond and lotic habitat types was no evidence of an interaction effect, nor of any (Fig. 4b); perMANOVA confirmed a significant dif- differences in within-group variability. To focus on ference in fish assemblage by habitat type, although differences between habitat types, subsequent analy- there was no difference in variability (Table 3). In ses were conducted separately for mainstem and for the ordination, the first two principal coordinate axes tributary sites. explained 57% of the total variation; only the first

7 Gibson et al.

Table 2. Density and catch per unit effort (CPUE) of non-native northern crayfish in lotic and pond habitats.

Mainstem sites Tributary sites Estimated CL† (mm) Habitat type Abundance metric n Mean Range n Mean Range Range

Lotic sites, density Individuals per m2 10 sites 9.0 0.3–28.3 16 sites 7.3 1.5–22.3 3–30 (mean = 8.4) Pond sites, CPUE Crayfish traps (individuals per trap) 72 traps 3.0 0–9 71 traps 1.8 0–10 20–60 Minnow traps (individuals per trap) 72 traps 5.0 0–37 71 traps 1.1 0–12 20–50 Seine hauls (individuals per m2 seined) 33 hauls 4.1 0–21.0 32 hauls 0.5 0–4.3 10–70

†Size range of crayfish caught in each gear type. Sizes indicate carapace length [CL] in mm. Crayfish caught in lotic sites (top row) were measured; crayfish caught in pond sites (bottom three rows) were not measured, and these size ranges are estimates only, to indicate the approximate size of individuals sampled by each gear type.

Table 3. Results of perMANOVA and multivariate dispersion tests examining effects of location (mainstem vs. tributary) and habitat type (pond vs. lotic) on fish assemblages within all sites and within mainstem sites or within tributary sites only.

Homogeneity of perMANOVA dispersions

Sites Factor n pseudo FP pseudo FP

All sites Habitat type 9 Location 38 Location 18.91 <0.001 1.31 0.227 Habitat type 5.65 0.003 0.62 0.450 Location 9 Habitat type 1.60 0.158 NA NA Mainstem sites Habitat type 16 3.61 0.009 0.38 0.550 Tributary sites Habitat type 22 3.85 0.014 0.37 0.551 two axes were statistically significant. Axis 1 but within ordination space also occupied by lotic appeared to describe a separation of fish assemblages sites. Although lotic sites occupied a greater area in along a gradient from pond sites on the left-hand this ordination space than did pond sites, there was side of the plot to lotic sites on the right. Western no evidence for a difference in group variability mosquitofish (GAAF) and green sunfish (LECY) (Table 3). Green sunfish (LECY) again showed a were strongly correlated with each other and with strong correlation with Axis 1, and again the ordina- Axis 1, indicating that pond sites on the left side of tion plot suggested a gradient between pond sites the plot were associated with abundant mosquitofish with abundant sunfish to the lower-right and lotic and sunfish, while lotic sites to the right were more sites with relatively abundant desert sucker (CACL) associated with relatively abundant desert sucker to the upper left. Finally, the smallmouth bass (CACL). By contrast, other species, including com- (MIDO) vector was again perpendicular to the sun- mon smallmouth bass (MIDO) showed a stronger cor- fish-desert sucker gradient, and again lacked a strong relation with Axis 2, and no strong tendency towards tendency towards either pond or lotic sites. Axis 2 either pond or lotic sites. Axis 2 described a general described a general gradient from sites almost gradient from sites dominated by small-bodied red entirely dominated by smallmouth bass at the bottom, shiner (CYLU) and mosquitofish (GAAF) on the bot- moving to sites with more desert sucker (for lotic tom to sites dominated by larger-bodied smallmouth sites) and relatively fewer bass at the top. This tribu- bass (MIDO) and Sonora sucker (CAIN) on the top, a tary ordination plot was strongly influenced by the gradient present in both pond and lotic sites. two pond and three lotic sites (mostly clustered on Within tributary sites, fish assemblage again dif- the far right-hand side of the plot) from one tributary fered significantly between pond and lotic habitat stream (Dry Beaver Creek), which was smaller, types (Table 3). An ordination of tributary sites ephemeral and supported a different fish species pool (Fig. 4c) showed, again, some clustering of pond than the other sampled tributaries. Red shiner sites, although pond and lotic sites appear less differ- (CYLU) were found only in this tributary, and there entiated than in the mainstem ordination. Addition- only within pond sites, resulting in the largely hori- ally, patterns in habitat associations of individual zontal vector for this species in the ordination plot. species were similar to the patterns seen for mainstem sites. The first two principal coordinates explained Discussion 62% of the total variation; only the first two axes were statistically significant. Pond sites were clus- We assessed the influence of beaver pond habitats in tered together in the lower-right quadrant of the plot, structuring the composition of mixed native and non-

8 Beaver dams shift desert fish assemblages

(a) 1.0

Pond mainstem

0.5 Pond tributary

Lotic mainstem 0.0 Lotic tributary Axis 2 (18%)

−0.5

−1.0 Axis 1 (24%)

(b) 1.0

MIDO CAIN 0.5

0.0 LECY GAAF

Axis 2 (22%) CACL −0.5 AMNA

CYLU

−1.0 Axis 1 (35%)

0.5 CACL

CYLU

0.0 LECY Axis 2 (22%)

−0.5

MIDO

−1.0 −1.0 −0.5 0.0 0.5 1.0 Axis 1 (40%)

Fig. 4. PCoA ordination plots of fish assemblage data for (a) all sites; (b) mainstem sites only; and (c) tributary sites only. Hulls are drawn around tributary versus mainstem sites in panel (a) and around pond versus lotic sites in panels (b) and (c). Only statistically significant (P < 0.05) vectors are displayed.

9 Gibson et al. native fish assemblages in a free-flowing dryland est source of potential piscivory and suggested that river. To our knowledge, this is the first study to non-native control efforts should target this species address the effect of beaver ponds on non-native (Bonar et al. 2004). Similarly, a bioenergetics model fishes at the community level. Our analyses showed of non-native fishes in the Yampa River (upper Colo- that in both the upper Verde River mainstem and in rado River Basin) indicated that the total potential pi- its tributaries, there were significant differences in scivory of abundant smallmouth bass exceeded that fish assemblage composition between lentic beaver of two larger but less numerous non-native predators pond and lotic stream habitat. Non-native species (northern pike, Esox lucius, and channel catfish, Ict- dominated the fish assemblage to a greater extent alurus punctatus) (Johnson et al. 2008). within ponds than within lotic sites: small non-native Green sunfish were frequently present in both lotic fish (both small-bodied fishes and juveniles of larger and pond habitats, but in lotic sites they were usually species) were often highly abundant in pond samples, present only at low numbers, whereas in ponds they while native species were rarely documented within were often highly abundant. This pattern was true for ponds. These results are generally consistent with the both mainstem and tributary sites. The association of hypothesis that beaver ponds could favour fish green sunfish with pond habitats is consistent with assemblages dominated by non-native species. other studies which report that species of the genus Our results are also consistent with studies of fish Lepomis were common in beaver pond fish assem- assemblage in relation to artificial dams. Impound- blages (Snodgrass & Meffe 1999; Pollock et al. ments behind small, low head dams are the closest 2003), and with the species’ known habitat prefer- analogue to beaver pond habitat: Beatty et al. (2009) ence for slow current velocity (Dudley & Matter found abundant non-native fish (primarily fathead 2000; Olden et al. 2006). Presence of green sunfish minnow and white sucker, Catostomus commersonii) has been implicated in steep declines of native fish within and downstream of an artificial wetland cre- populations in the LCRB (e.g., Clarkson et al. 2010). ated by a small dam on an upper Colorado River Although gut content analysis of green sunfish from Basin stream (Wyoming), while native fishes were the upper Verde River found a very low incidence of restricted to upstream of the wetland and to tributar- piscivory (mean 0.61% of stomach content by vol- ies without impoundments. Dominance of non-native ume, n = 754; Bonar et al. 2004), laboratory and fishes within large reservoirs of the LCRB is well field evidence shows that this species can be an established (Mueller & Marsh 2002), and reservoirs effective predator of larval and juvenile native fishes can promote proliferation of non-native species in the LCRB (Dudley & Matter 2000; Carpenter & downstream, even without major changes to down- Mueller 2008). Increased d15N isotope signatures stream flow or thermal regimes (Martinez et al. from green sunfish relative to native fishes in another 1994). However, beaver dams are typically smaller, Arizona river system, the Gila River, also demon- more permeable and much less permanent than artifi- strate a strong propensity towards piscivory (Pilger cial analogues, which may affect the direction or et al. 2010). Additionally, highly aggressive behav- mechanism of influences on fish. iour by green sunfish indicates a probable advantage over native fishes in competition for space or food (Karp & Tyus 1990). These lines of evidence suggest Fish assemblage responses that there is potential for green sunfish to have sub- We found that responses to beaver pond habitat var- stantial impacts on native fishes in the Verde River ied by species. Here we focus on the native fishes system, especially given the high densities of green and the four most common non-native species in our sunfish that we observed in beaver ponds. study: smallmouth bass, green sunfish, western mos- Western mosquitofish showed a strong association quitofish and northern crayfish. with beaver pond habitat, although high mosquitofish Smallmouth bass were the most common fish in abundances were also observed in some mainstem our study (present in 87% of sites) and often the most lotic sites. Consistent with our results, preferred mos- abundant large-bodied fish, in both lotic and pond quitofish habitat features include shallow water, slow habitat types. In lotic samples, especially, bass often current and dense vegetation (Pyke 2008), all typical dominated the total fish assemblage. The high num- features of beaver ponds (Rosell et al. 2005). Mos- bers of bass found in both habitat types in our study quitofish have been widely introduced worldwide confirm the habitat flexibility of this species. From a (Pyke 2008), and, in the LCRB, spread of mosquito- conservation perspective, smallmouth bass are of par- fish is closely associated with declines of the ecologi- ticular interest because this widespread species is cally similar native Gila topminnow (Poeciliopsis believed to pose a substantial threat to native fishes occidentalis) (Meffe 1985). Laboratory and field in the LCRB. A bioenergetics model for the upper studies have demonstrated that agonistic interactions Verde River identified smallmouth bass as the great- with aggressive mosquitofish can lead to injury,

10 Beaver dams shift desert fish assemblages changes in behaviour and reduced growth for native the native species recorded in this study were present fishes (particularly small-bodied species), and mos- in lotic sites, but only Sonora sucker and roundtail quitofish also prey directly on eggs and larvae of chub were found in ponds. The two native suckers in native fish (Meffe 1985; Mills et al. 2004; Ayala our species pool differ in modes of feeding and use et al. 2007). In isolated backwater habitats otherwise of habitat. Desert suckers are herbivorous and inhabit free of non-native fish, presence of mosquitofish is predominantly swift-flowing streams with hard-bot- believed to be responsible for failed recruitment by tom substrate, whereas Sonora suckers feed on inver- stocked native razorback suckers (Xyrauchen tex- tebrates and detritus and are abundant in deep pools anus) (Ley et al. 2012); we suggest that abundant with restricted flow and fine substrate (Minckley & mosquitofish may similarly impede attempts to Marsh 2009). These differences were reflected in restore populations of spikedace in the Verde River. findings of our study. Desert suckers showed a strong In addition to non-native fishes, we observed high association with lotic habitat: they were frequently densities of non-native northern crayfish within both present and often relatively abundant in lotic sites, lotic and beaver pond habitats of the Verde River but never found within ponds. Sonora suckers, by system. This study is one of the first to report abun- contrast, were present and occasionally abundant in dance metrics for crayfish in Arizona. In pools of a both habitat types. Native roundtail chub, much like Gila River system tributary stream (Arizona), Carpen- Sonora suckers, are omnivorous feeders associated ter (2005) reported crayfish densities between 3 and with deep pool habitat. However, we rarely found 11 large individuals (>25 mm carapace length [CL]) this species in ponds. It is possible that roundtail per m2, comparable to our mean densities of 9.3 and chub presence or abundance was underestimated in 7.4 small individuals (mean CL = 8.4 mm) per m2 in ponds due to size selectivity of sampling methodol- mainstem and tributary sites, respectively. Martinez ogy: size frequency data from pond samples (data not (2012) reported somewhat higher densities (10.9 indi- shown) suggest that mid-sized fish (approximately viduals per m2 in 2005; mean CL = 16.1 mm) in the 150–300 mm total length) may have been consis- Yampa River (upper Colorado River Basin), where tently under-sampled in ponds. We believe that this the total estimated river-wide biomass of crayfish is the most important influence of a gear effect in our exceeded that of all fish and other invertebrates com- results. Species most likely to have been underesti- bined. Invasive crayfish can have strong impacts on mated in this way are roundtail chub, desert sucker, multiple levels of aquatic food webs (Twardochleb and, to a lesser extent, smallmouth bass. Alterna- et al. 2013), and laboratory and field evidence indi- tively, the rarity of roundtail chub in beaver ponds cate that crayfish can compete with and prey on could reflect unfavourable habitat conditions for these native LCRB fishes directly (Carpenter 2005). Many fish, whether due to unique abiotic properties of bea- non-native fishes consume crayfish (Johnson et al. ver ponds or to the abundant non-native fish. 2008), while the native fishes sampled in this study The three small-bodied fish present in the upper are more exclusively invertivores/herbivores (Olden Verde River within recent decades (longfin dace, et al. 2006; Pilger et al. 2010), and therefore more speckled dace and spikedace) were not recorded in likely to compete with crayfish than to prey on them our study, with the exception of two tributary sites (Carpenter 2005; Arena et al. 2012). The greatest where speckled dace were present. These three spe- impact of crayfish on native fishes may be indirect, cies are generally associated with riffle habitat via apparent competition: abundant crayfish may pro- (Minckley & Marsh 2009); therefore, it is unlikely vide an alternative prey source when small fish have that adults would be found in beaver pond habitat, been depleted, thus stabilising large populations of although they might occupy fast tailwaters immedi- non-native predators. Our finding of a positive rela- ately downstream of dams (Smith & Mather 2013). tionship between crayfish and smallmouth bass abun- Extensive conversion of lotic habitat to lentic beaver dances in ponds is consistent with this hypothesis. In ponds, as has been documented on the regulated Bill the Yampa River, Martinez (2012) documented Williams River, Arizona (Shafroth et al. 2010), could simultaneous large increases in northern crayfish and limit the potential for restoration of these native riffle smallmouth bass populations (along with steep specialists. This situation occurred on Bonita Creek, declines in small-bodied native fish). By creating rel- AZ, where extensive beaver dam activity has resulted atively stable hydrology and deep pools, beaver in the cessation of stocking efforts for spikedace and ponds may provide favourable habitat for crayfish loach minnow (Rhinichthys cobitis), due in part to a (e.g., Light 2003); the influence of beaver ponds on lack of lotic habitat and an increase in abundance of crayfish populations in the LCRB requires additional green sunfish (H. Blasius, pers. comm.). investigation. However, the greatest importance of beaver ponds Native fishes in general showed a tendency for native and non-native fish communities in the towards higher abundance in lotic habitats: all four of Verde River system may be their function as spawn-

11 Gibson et al. ing or rearing habitat for juvenile fish. Previous stud- fishes as a group tend to increase in absolute and rel- ies have found ontogenetic shifts in fish use of beaver ative abundance following large peak flows or years pond habitat, with some species moving into ponds with high discharge; conversely, during dry years, to spawn (Schlosser 1998; Snodgrass & Meffe 1999). non-native species are more likely to become estab- Schlosser (1995, 1998) further suggests that beaver lished and to dominate assemblages, and drought ponds may function as reproductive ‘sources’ for conditions are associated with extirpations of native some species, while adjacent stream reaches are species (Propst et al. 2008; Gido et al. 2013). ‘sinks’ with little successful recruitment. Beaver Drought and low discharge likely represent critical ponds often include extensive shallow, vegetated periods for persistence of native fishes and therefore areas with slow current, many of the same features the most important time to understand the influence that characterise typical rearing habitat for juvenile of beaver dams and other habitat features on native LCRB native fishes (Childs et al. 1998). Backwaters and non-native assemblage structure. and other off-channel rearing habitats have been the target of restoration activities (Minckley et al. 2003), Contrasting mainstem and tributary habitats and given the importance of backwaters for successful recruitment of native fish, the possibility that bea- Our finding of a strong distinction between fish ver ponds could provide equivalent rearing habitat assemblages of the upper Verde River mainstem and conditions is worthy of investigation. However, the its tributaries is consistent with other observed differ- abundant non-native fish that we documented within ences between the two locations, including differ- margins and backwaters of beaver ponds may make ences in species pool, habitat features and these areas unsuitable for juvenile native fish (Minck- distribution of beaver dams. We found that beaver ley et al. 2003; Carpenter & Mueller 2008). Size dams were generally larger and more abundant in the selectivity of different sampling gears prevents formal upper mainstem than in the tributaries, and mainstem analysis of differences in habitat use by life stage in ponds appeared older and more complex. This may our data. However, we did record young-of-the-year reflect the more stable hydrology in the upper main- individuals of several native and non-native species stem relative to the tributaries: the number of beaver (native Sonora sucker and non-native smallmouth dams in the upper mainstem tends to increase steadily bass, green sunfish, common carp and yellow bull- over several years during periods of moderate dis- head) within ponds, suggesting that these species charge, until occasional large floods destroy most or may sometimes utilise beaver ponds for rearing all dams (D. Campbell, pers. comm.). In Verde River habitat. tributaries, by contrast, dams are more likely to wash out annually in snowmelt or monsoon floods (K. Schonek, pers. comm.), thus preventing enlarge- Variation over time ment and succession of the beaver pond habitat over This study provides a snapshot view of fish assem- multiple years. It seems likely that longevity of bea- blage at a single point in time, focusing on the dis- ver ponds may affect their ecological function, and tinction between pond and lotic habitat types. Long- several studies have demonstrated that fish assem- term studies of fish assemblage in the Verde River blage in beaver pond habitats varies with pond age and similar systems document substantial interannual (Snodgrass & Meffe 1998; Schlosser & Kallemeyn variation in absolute and relative abundance of native 2000). However, despite differences in hydrology fishes, especially in response to stream discharge and pond stability, we observed some consistent pat- (e.g., Rinne 2005; Propst et al. 2008). Similarly, the terns in individual species’ response to pond habitat response of fish assemblage to beaver pond habitats in both the mainstem and tributaries. is likely to vary with time and discharge; therefore, caution should be used in extrapolating results of the Net effects of beaver ponds on fish communities present study to different environmental conditions. Sampling for this study occurred during May and In this study, we documented patterns of presence June, typically the driest months of the year (Blasch and abundance for native and non-native fish within et al. 2006), and during a drought year; the upper beaver ponds as compared to unimpounded lotic Verde River mainstem had not experienced a signifi- stream reaches. Our results point to further questions cant high-flow event since 2010 (USGS 2013). Low about the consequences of these patterns (also see flow conditions are of particular importance from the Gibson & Olden 2014): what will be the net effect of perspective of native fish conservation. Numerous beaver dams on fish populations in the Verde River studies have found strong positive correlations system and similar desert rivers? Beaver dams and between discharge and LCRB native fish abundance ponds are complex landscape features that alter (Rinne & Miller 2006; Stefferud et al. 2011). Native numerous different aspects of aquatic ecosystems at

12 Table 4. Hypothesised consequences of various types of ecosystem change resulting from construction of beaver dams for native and non-native fishes in the CRB, as well as possible outcomes of each change for native fish in a mixed native–non-native assemblage. Arrows indicate whether the effect on fish is expected to be positive (↑) or negative (↓). Beaver dams and ponds are highly variable, and not all effects will apply to all beaver ponds.

† ‡ § Category Effect of beaver dam Consequences for native fishes Consequences for non-native fishes Potential outcomes¶

Direct habitat A. Pooled water, reduced ↑ Deep pools provide habitat for larger or pool- ↑ Deep pools provide habitat for larger or pool- ↑ Reduced intensity of biotic interactions due to effects current velocity. dwelling fish, potentially including ‘big river’ dwelling fish, including carp, catfishes and habitat segregation; or native fish. some sunfishes. ↓ Increased competition and predation due to ↑ Shallow lentic water in pond margins and ↑ Shallow lentic habitat provides habitat for high densities of non-native fishes in ponds. backwaters provides habitat for lentic/ lentic/marsh habitat specialists. marsh habitat specialists. ↑ Relative stability of beaver pond habitat ↓ Decreased habitat availability to fluvial resembles typical habitat in the native range specialists due to conversion of lotic to of many non-native species. lentic habitat in beaver ponds.

B. Increased habitat complexity, ↑ Spawning/rearing habitat for numerous ↑ Spawning/rearing habitat for numerous ↑ Reduced predation due to availability of cover including both channel species in shallow, vegetated, backwater- species in shallow, vegetated, backwater- and refuge habitats. morphology (side type habitats. type habitats. ↑ Complexity promotes habitat segregation and channels, backwaters) ↑ Habitat structure provides cover and refuge ↑ Habitat structure provides cover for reduces intensity of biotic interactions; or and habitat structure (large from predation. predators, increases predation success; or ↓ Abundant small non-native fish present in wood, macrophytes). ↓ Habitat structure provides cover for prey potential native fish rearing habitat increase and reduces predation success. rates of competition and predation on larval native fish.

C. Retention of sediment, ↑ Reduce sedimentation downstream of ponds. ↑ Reduce sedimentation downstream of ponds. ↑ Improved downstream habitat quality for reduced turbidity. ↓ Fine sediment within ponds unfavourable for ↓ Fine sediment within ponds unfavourable for some species due to altered sediment species that prefer coarser substrate. species that prefer coarser substrate. flows. ↓ Loss of visual cover due to decreased ↑ Increase success of visual predators due to ↓ Increased predation due to greater visibility. turbidity. decreased turbidity.

D. Increased fluvial habitat (fast ↑ Provide habitat for small fluvial specialists. ↑ Provide habitat for small fluvial specialists. ↑↓ Change in quality or quantity of habitat assemblages fish desert shift dams Beaver water, coarse substrate) available to fluvial specialists. immediately downstream of beaver dams.

E. Altered water quality, ↑ Reduce thermal stress due to temperature- ↑ Reduce thermal stress due to temperature- ↓ Increased competitive advantage for non- including temperature regime buffering effect of ponds; or buffering effect of ponds; or native fishes due to thermal stress. and dissolved oxygen ↓ Increase thermal stress due to higher ↓ Increase thermal stress due to higher concentration. temperatures in ponds. temperatures in ponds.

Indirect effects F. Increased rate of primary ↑ Increase food availability for herbivores and ↑ Increase food availability for herbivores and ↑ Reduced competition due to greater productivity and standing detritivores. detritivores. availability of resources; or stock of organic matter. ↓ Increased competition and predation due to higher densities of non-native fish.

G. Increased biomass and ↑ Increase food availability for invertivores. ↑ Increase food availability for crayfish- ↑ Reduced competition due to greater altered community structure ↓ Reduce relative abundance of lotic consuming species. availability of resources; or of benthic invertebrates. invertebrate taxa preferred by some ↓ Reduce relative abundance of lotic ↓ Increased competition and predation due to species. invertebrate taxa preferred by some higher densities of non-native fish. species. 13 14 Table 4 (continued) al. et Gibson

† ‡ § Category Effect of beaver dam Consequences for native fishes Consequences for non-native fishes Potential outcomes¶

Drainage-level H. Flow stabilisation: dampened ↑ Provide slack water refuge during peak flows, ↑ Provide slack water refuge during peak flows: ↓ Reduction in the ability of native fishes to effects peak flows. particularly for vulnerable larval or juvenile may be especially important for non-native benefit from peak flow events. fish. species that lack adaptations for surviving ↓ Increased survival of non-native fish through ↓ Reduce recruitment success dependent on flood conditions peak flow events and/or increased speed of peak flow events. recolonisation following disturbance. ↓ Reduce power of peak flows to move sediment and rejuvenate spawning and rearing habitats.

I. Flow stabilisation: ↑ Provide refuge habitat during drought and ↑ Provide refuge habitat during drought and ↑ Increased ability of native fishes to persist maintenance of surface low flows. low flows: may be particularly important for through combined stressors of drought and water within beaver pond non-native species that lack adaptations for non-native fish; or and/or downstream of dam. surviving drought conditions. ↓ Increased negative biotic interactions due to concentration of native and non-native species within small refuge habitats. ↓ Increased survival of non-native fish through drought events and/or increased speed of recolonisation following disturbance.

J. Barriers to fish movement. ↓ Reduce habitat connectivity, fish dispersal ↓ Reduce habitat connectivity, fish dispersal ↑ Potential for dams to create habitat patches between populations. between populations. that are temporarily isolated from non-native ↓ Slow rate of spread up new drainages and fish. tributary systems. ↓ Dams limit ability to move between habitats in response to drought or other varying environmental conditions.

K. Unique lentic habitat patches ↑ Provide reproductive source habitat for some ↑ Provide reproductive source habitat for some ↑ Higher species richness of native fishes within within lotic habitat species. species. a drainage. corridors. ↑ Promote persistence of lentic/marsh habitat ↑ Promote persistence of lentic/marsh habitat ↓ Higher species richness of non-native fishes specialists within a drainage. specialists within a drainage. within a drainage. ↑ Potentially increase the upstream extent of ↑ Potentially increase the upstream extent of ↓ Spillover from pond habitat increases suitable habitat for ‘big river’ fishes in suitable habitat for large piscivores such as abundance of non-native fish throughout a smaller river or tributary systems. catfishes in smaller river or tributary drainage. systems.

L. Reduced channelisation and ↑ Diversity of aquatic habitats promotes ↑ Diversity of aquatic habitats promotes ↑↓ Different habitat supports a different increased abundance of regional species richness. regional species richness. assemblage of native and non-native fishes. marshy and cienega habitats.

†Well-established effects of beaver dams are in plain text, while hypothesised or uncertain effects are indicated in italics. ‡Response of native fish, independent of or in the absence of non-native fish. §Response of non-native fish, with or without native fish present. ¶Hypothesised net outcomes for native fish populations, when non-native fish are present, as a result of the beaver dam effect. In some cases, mutually exclusive alternatives are presented. Sources informing development of this table include: Schlosser (1995, 1998); Snodgrass & Meffe (1998, 1999); Pollock et al. (2003); Rosell et al. (2005); Propst et al. (2008); Minckley & Marsh (2009); Stefferud et al. (2009); Kemp et al. (2012); Smith & Mather (2013); Gibson & Olden (2014). Beaver dams shift desert fish assemblages multiple scales (Pollock et al. 2003; Rosell et al. use of both habitat types (Table 4, point A). How- 2005). Some of these beaver dam effects may be ben- ever, the complexity and habitat structure of beaver eficial to fishes while others are detrimental; some ponds are not associated with natural pools, and effects may influence native fishes differently than flooding behind beaver dams increases channel width they do non-native fishes; and the overall effect of and often includes abundant shallow grassy area dams on native fishes may vary depending on (Table 4, point B; Rosell et al. 2005). Beaver dams whether non-natives are present. This complexity can greatly increase retention of sediment and organic makes it difficult to distinguish the mechanisms by matter (Table 4, point C). This retentiveness, together which beaver dams produce observed responses in with the more open canopy and addition of woody fish communities. In Table 4, we address this com- debris due to beaver foraging, affects nutrient cycling plexity by systematically formulating separate and other ecosystem processes within ponds, which hypotheses about how each type of beaver dam effect in turn influences the biotic communities (Table 4, is likely to influence native and non-native fish popu- points F-G; Naiman et al. 1986; Rosell et al. 2005). lations in the LCRB. Our intention is that this chart The beaver dam itself provides a potential barrier to will help to guide and organise thinking about how movement of fish and other organisms, creating a and why beaver dams influence fish communities; more closed system than a natural pool (Table 4, additionally, these hypotheses may suggest profitable point J). Finally, although an individual beaver pond directions for future research. In the following discus- may be no bigger than an average stream pool, often sion, we expand on some of the particularly impor- multiple beaver dams are constructed within a small tant or interesting hypotheses. area, forming a very large complex of continuous At the scale of individual beaver ponds (Table 4, beaver pond habitat (Rosell et al. 2005). Future ‘Direct habitat effects’), a better understanding is research should seek to clarify which, if any, of these needed of how fish use of pond habitat varies with unique habitat features are most influential for fish season, environmental conditions and life stage; iden- species of interest. tifying which species may be employing beaver The influence of beaver dams on fish assemblages ponds for spawning or rearing habitat would be par- can also extend beyond the pond itself to the rest of ticularly valuable. Fish use of beaver pond habitat the stream network (Table 4, ‘Drainage-level likely also varies with physical habitat features of effects’). Fish from beaver pond assemblages may ponds, such as pond size or location within the river spill over into neighbouring stream reaches (Schlosser system (e.g., Snodgrass & Meffe 1998). However, 1998; Snodgrass & Meffe 1999). Beaver ponds can the net effect of utilising beaver pond habitat will provide source habitat for some species, as discussed depend on the degree of biotic interactions occurring earlier. In general, beaver ponds promote fish species in these habitats, and whether beaver ponds tend to richness at the drainage level by supporting species increase or decrease negative interactions between that otherwise might not persist in unmodified habitat native and non-native fishes. Beaver ponds generally (e.g., Hanson & Campbell 1963; Snodgrass & Meffe increase habitat complexity relative to unimpounded 1998; Smith & Mather 2013), but this effect may be stream, in terms of both channel morphology (e.g., undesirable if it is the richness of non-native species secondary channels, overbank flooding; Kemp et al. that is increased (Table 4, point K). Beatty et al. 2012; Smith & Mather 2013) and habitat structure (2009), who documented abundant non-native fish (e.g., aquatic vegetation, large wood; Pollock et al. within and downstream of an artificial impoundment 2003). Habitat complexity may provide refuge from on a Wyoming stream, suggest that this impoundment predation and increase spatial segregation between provided source habitat for the non-native fish, noting native and non-native fishes, thus reducing negative that tributary streams lacking any impoundments were interactions (Crowder & Cooper 1982; Meffe 1985; instead dominated by native fishes. The abundant Billman et al. 2012). However, if beaver pond habi- green sunfish and western mosquitofish recorded tats also support significantly higher densities of non- within ponds in our study suggest that beaver ponds native fish than do unimpounded stream reaches, then could similarly provide source habitat for these small this could override any advantages of habitat struc- lentic fishes in the Verde River system. However, we ture for native fish (Table 4, point B). In particular, found no evidence for a correlation between distance use of beaver pond margins by native fishes as rear- from a pond and abundance of these or any other spe- ing habitat could increase predation on vulnerable cies (unpublished results); additionally, thriving popu- native fish larvae when, as in our results, these areas lations of non-native species in similar streams also support abundant small non-native fishes. without beaver activity (e.g., Fossil Creek, Arizona; Some habitat features of beaver ponds are also M. O’Neill, pers. obs.; Marks et al. 2010) indicate that found in natural stream pools: fish preferring deep beaver pond habitats are not essential to their life his- water and reduced current velocity will likely make tory. Nonetheless, beaver ponds could promote higher

15 Gibson et al. abundances of these fish or increase their resistance or for providing useful information and insight into the ecology resilience to disturbance. Finally, the degree to which of the Verde River system. J. Sorensen, the Arizona Game beaver ponds influence fish populations will also and Fish Department, and The Nature Conservancy of Arizona depend on the abundance and distribution of ponds provided substantial logistical support. D. Beauchamp, M. and other pool habitat within a river system. Pond Pollock and two anonymous reviewers provided valuable comments that improved the manuscript. Funding for PPG habitat is likely to be more influential in systems was provided by a University of Washington Top Scholar where beaver ponds are abundant or where ponds pro- Graduate Fellowship and by the H. Mason Keeler Endowment vide a unique habitat type (Table 4, point K), espe- for Excellence through the School of Aquatic and Fishery Sci- cially when other pool habitat is scarce. ences. JDO was supported in part by an H. Mason Keeler Within desert river systems like the Verde River, Endowed Professorship and by the Department of Defense – the influence of beaver ponds on fish assemblage Strategic Environmental Research and Development Program response to large-scale disturbance by drought and (RC-1724). floods is of particular interest. Previous studies have suggested that beaver ponds can provide refuge habi- References tat for fish during drought and seasonal low flows Anderson, M.J. 2001. A new method for non-parametric mul- (Table 4, point I; Hanson & Campbell 1963; Magou- tivariate analysis of variance. Austral Ecology 26: 32–46. lick & Kobza 2003; White & Rahel 2008). This func- Anderson, M.J. 2006. Distance-based tests for homogeneity of tion may be particularly important in intermittent multivariate dispersions. Biometrics 62: 245–253. tributary streams, especially with projected decreases Arena, A., Ferry, L.A. & Gibb, A.C. 2012. Prey capture in perennial surface flow due to water withdrawals behavior of native vs. nonnative fishes: a case study from and climate change (Marshall et al. 2010). In ephem- the Colorado River drainage basin (USA). Journal of Experi- eral Dry Beaver Creek in our study, we observed mental Zoology 317: 103–116. large native Sonora suckers within deep pools that Ayala, J.R., Rader, R.B., Belk, M.C. & Schaalje, G.B. 2007. had been expanded by beaver dams. However, con- Ground-truthing the impact of invasive species: spatio-tem- poral overlap between native least chub and introduced wes- centrating fish within small refuge habitats typically – increases the intensity of biotic interactions (Magou- tern mosquitofish. Biological Invasions 9: 857 869. Barbour, M.T., Gerritsen, J., Snyder, B.D. & Stribling, J.B. lick & Kobza 2003): enforced habitat overlap between 1999. Rapid bioassessment protocols for use in streams and non-native predators and vulnerable native fishes wadeable rivers: periphyton, benthic macroinvertebrates, and within ponds could lead to extensive predation and fish. Washington, DC: U.S. Environmental Protection other negative interactions, such that native fishes Agency, Office of Water, EPA 841-B-99-002. would derive no net benefit from the refuge habitat. Beatty, R.J., Rahel, F.J. & Hubert, W.A. 2009. Complex influ- Beaver dams may also provide fish some protection ences of low-head dams and artificial wetlands on fishes in from floods or peak flows (Kemp et al. 2012), poten- a Colorado River tributary system. Fisheries Management tially speeding recolonisation of a drainage by non- and Ecology 16: 457–467. native species following the flood (Table 4, point H; Billman, E.J., Kreitzer, J.D., Creighton, J.C., Habit, E., Pool & Olden 2014). Future research should address McMillan, B. & Belk, M.C. 2012. Habitat enhancement and the ability of beaver ponds to provide refuge from native fish conservation: can enhancement of channel complexity promote the coexistence of native and introduced these disturbances, and the relative impact of that ref- fishes? Environmental Biology of Fishes 96: 555–566. uge on native and non-native fishes. Blasch, K.W., Hoffman, J.P., Grasner, L.F., Bryson, J.R. & Ultimately, the consequences of beaver dam-building Flint, A.L. 2006. Hydrogeology of the upper and middle activity for native fish populations will depend on the Verde River watersheds, central Arizona. U.S. Geological extent to which a dam’s influence extends beyond Survey Scientific Investigations Report 2005-5198. the pond: if beaver ponds support a high density of Bonar, S.A., Leslie, L.L. & Velez, C.E. 2004. Influence of non-native fish only within the pond itself, then the species, size class, environment, and season on introduced net impact on native fish populations at the drainage fish predation on native fishes in the Verde River System, level may be negligible. However, if abundant non- Arizona. Arizona Cooperative Fish and Wildlife Research native fish spill over into adjacent stream reaches, or Unit, Fisheries Research Report 01-04. if beaver ponds promote reproduction and maintain Carpenter, J. 2005. Competition for food between an introduced crayfish and two fishes endemic to the Colorado source populations of non-native species, then beaver River basin. Environmental Biology of Fishes 72: 335–342. dams could have far-reaching consequences for the Carpenter, J. & Mueller, G.A. 2008. Small nonnative fishes as overall composition of the fish assemblage. predators of larval razorback suckers. The Southwestern Naturalist 53: 236–242. Acknowledgements Childs, M.R., Clarkson, R.W. & Robinson, A.T. 1998. Resource use by larval and early juvenile native fishes in We thank L. Kuehne and J. Walters for essential field assis- the Little Colorado River, Grand Canyon, Arizona. Transac- tance. We also thank K. Schonek of The Nature Conservancy tions of the American Fisheries Society 127: 620–629.

16 Beaver dams shift desert fish assemblages

Clarkson, R.W., Marsh, P.C., Melis, T.S., Hamill, J.F., Coggins, Lokteff, R.L., Roper, B.B. & Wheaton, J.M. 2013. Do beaver L.G. Jr, Grams, P.E., Kennedy, T.A., Kubly, D.M. & Ralston, dams impede the movement of trout? Transactions of the B.E. 2010. Effectiveness of the barrier-and-renovate approach American Fisheries Society 142: 1114–1125. to recovery of warmwater native fishes in the Gila River Magoulick, D.D. & Kobza, R.M. 2003. The role of refugia for basin. In: Melis, T.S., Hamill, J.F., Bennet, G.E., Coggins, fishes during drought: a review and synthesis. Freshwater L.G. Jr, Grams, P.E., Kennedy, T.A., Kubly, D.M. & Ralston, Biology 48: 1186–1198. B.E., eds. Proceedings of the Colorado River basin science Marks, J.C., Haden, G.A., O’Neill, M. & Pace, C. 2010. and resource management symposium. U.S. Geological Sur- Effects of flow restoration and exotic species removal on vey Scientific Investigations Report 2010-5135: pp. 209–217. recovery of native fish: lessons from a dam decommission- Crowder, L.B. & Cooper, W.E. 1982. Habitat structural com- ing. Restoration Ecology 18: 934–943. plexity and the interaction between bluegills and their prey. Marshall, R.M., Robles, M.D., Majka, D.R. & Haney, J.A. Ecology 63: 1802–1813. 2010. Sustainable water management in the southwestern Cucherousset, J. & Olden, J.D. 2011. Ecological impacts of United States: reality or rhetoric? PLoS ONE 5: e11687. nonnative freshwater fishes. Fisheries 36: 215–230. Martinez, P. 2012. Invasive crayfish in a high desert river: DeVries, P., Fetherston, K.L., Vitale, A. & Madsen, S. 2012. implications of concurrent invaders and climate change. Emulating riverine landscape controls of beaver in stream Aquatic Invasions 7: 219–234. restoration. Fisheries 37: 246–255. Martinez, P., Chart, T., Trammell, M., Wullschleger, J. & Dudley, R.K. & Matter, W.J. 2000. Effects of small green Bergersen, E. 1994. Fish species composition before and after sunfish (Lepomis cyanellus) on recruitment of Gila chub construction of a main stem reservoir on the White (Gila intermedia) in Sabino Creek, Arizona. The Southwest- River, Colorado. Environmental Biology of Fishes 40: 227– ern Naturalist 45: 24–29. 239. Fredlake, M. 1997. Environmental assessment: re-establish- Meffe, G.K. 1985. Predation and species replacement in ment of North American beaver (Castor canadensis) into American southwestern fishes: a case study. The Southwest- the San Pedro Riparian National Conservation Area. Tucson, ern Naturalist 30: 173–187. AZ: U.S. Bureau of Land Management, Environmental Mills, M.D., Rader, R.B. & Belk, M.C. 2004. Complex inter- Assessment no. AZ-060–97-004. actions between native and invasive fish: the simultaneous Gauch, H.G. 1982. Multivariate analysis in community ecol- effects of multiple negative interactions. Oecologia 141: ogy. Cambridge: Cambridge University Press. 312 pp. 713–721. Gibson, P.P. & Olden, J.D. 2014. Ecology, management, and Minckley, W.L. & Marsh, P. 2009. Inland fishes of the greater conservation implications of North American beaver (Castor Southwest: chronicle of a vanishing biota. Tucson: Univer- canadensis) in dryland streams. Aquatic Conservation: sity of Arizona Press. xxxiv + 426 pp. Marine and Freshwater Ecosystems 24: 391–409. Minckley, W.L., Marsh, P.C., Deacon, J.E., Dowling, T.E., Gido, K.B., Propst, D.L., Olden, J.D., Bestgen, K.R. & Rosen- Hedrick, P.W., Matthews, W.J. & Mueller, G. 2003. A con- feld, J. 2013. Multidecadal responses of native and intro- servation plan for native fishes of the lower Colorado River. duced fishes to natural and altered flow regimes in the BioScience 53: 219–234. American Southwest. Canadian Journal of Fisheries and Moody, E.K. & Taylor, C.A. 2012. Red swamp crawfish Aquatic Sciences 70: 554–564. (Procambarus clarkii) discovered in the San Pedro River, Hanson, W.D. & Campbell, R.S. 1963. The effects of pool Arizona: a new invader in a threatened ecosystem. The size and beaver activity on distribution and abundance of Southwestern Naturalist 57: 339–340. warm-water fishes in a north Missouri stream. American Moorman, M.C., Eggleston, D.B., Anderson, C.B., Mansilla, Midland Naturalist 69: 136–149. A. & Szejner, P. 2009. Implications of beaver (Castor Johnson, B.M., Martinez, P.J., Hawkins, J.A. & Bestgen, K.R. canadensis) and trout introductions on native fish in the 2008. Ranking predatory threats by nonnative fishes in the Cape Horn Biosphere Reserve, Chile. Transactions of the Yampa River, Colorado, via bioenergetics modeling. North American Fisheries Society 138: 306–313. American Journal of Fisheries Management 28: 1941–1953. Mueller, G.A. & Marsh, P.C. 2002. Lost, a desert river and its Karp, C.A. & Tyus, H.M. 1990. Behavioral interactions native fishes: a historical perspective of the Lower Colorado between young Colorado squawfish and six fish species. Co- River. U.S. Geological Survey Information and Technology peia 1990: 25–34. Report USGS/BRD/ITR-2002-0010. Kemp, P.S., Worthington, T.A., Langford, T.E.L., Tree, Murphy, M.L., Heifetz, J., Thedinga, J.F., Johnson, S.W. & A.R.J. & Gaywood, M.J. 2012. Qualitative and quantitative Koski, K.V. 1989. Habitat utilization by juvenile Pacific sal- effects of reintroduced beavers on stream fish. Fish and mon (Onchorynchus) in the glacial Taku River, Southeast Fisheries 13: 158–181. Alaska. Canadian Journal of Fisheries and Aquatic Sciences Legendre, P. & Legendre, L. 1998. Numerical ecology. 46: 1677–1685. Amsterdam, New York: Elsevier. xvi + 990 pp. Naiman, R.J., Melillo, J.M. & Hobbie, J.E. 1986. Ecosystem Ley, G., Fencl, J.S., Kesner, B.R. & Marsh, P.C. 2012. Impe- alteration of boreal forest streams by beaver (Castor canad- rial Ponds native fish off-channel habitat progress and sum- ensis). Ecology 67: 1254–1269. mer telemetry study. Laughlin, NV: Presented at annual Oksanen, J., Blanchet, F.G., Kindt, R., Legendre, P., Minchin, Colorado River Aquatic Biologists meeting. P.R., O’Hara, R.B., Simpson, G.L., Solymos, P., Henry, M., Light, T. 2003. Success and failure in a lotic crayfish inva- Stevens, H. & Wagner, H. 2011. Vegan: community ecology sion: the roles of hydrologic variability and habitat alter- package. R package version 2.0-9. Available at: http://CRAN. ation. Freshwater Biology 48: 1886–1897. R-project.org/package=vegan (accessed 25 September 2012)

17 Gibson et al.

Olden, J.D. & Poff, N.L. 2005. Long-term trends of native southwestern fishes. Reviews in Fisheries Science 14: and non-native fish faunas in the American Southwest. Ani- 91–110. mal Biodiversity and Conservation 28: 75–89. Rosell, F., Bozser, O., Collen, P. & Parker, H. 2005. Ecologi- Olden, J.D., Poff, N.L. & Bestgen, K.R. 2006. Life-history cal impact of beavers Castor fiber and Castor canadensis strategies predict fish invasions and extirpations in the Colo- and their ability to modify ecosystems. Mammal Review 35: rado River basin. Ecological Monographs 76: 25–40. 248–276. Pilger, T.J., Gido, K.B. & Propst, D.L. 2010. Diet and trophic Schlosser, I.J. 1995. Dispersal, boundary processes, and tro- niche overlap of native and nonnative fishes in the Gila phic-level interactions in streams adjacent to beaver ponds. River, USA: implications for native fish conservation. Ecol- Ecology 76: 908–925. ogy of Freshwater Fish 19: 300–321. Schlosser, I.J. 1998. Fish recruitment, dispersal, and trophic Pollock, M.M., Heim, M. & Werner, D. 2003. Hydrologic and interactions in a heterogeneous lotic environment. Oecologia geomorphic effects of beaver dams and their influence on 113: 260–268. fishes. In: Gregory, S.V., Boyer, K. & Gurnell, A., eds. The Schlosser, I.J. & Kallemeyn, L.W. 2000. Spatial variation in ecology and management of wood in world rivers. Bethesda, fish assemblages across a beaver-influenced successional MD: American Fisheries Society, pp. 213–233. landscape. Ecology 81: 1371–1382. Pollock, M.M., Pess, G.R., Beechie, T.J. & Montgomery, Shafroth, P.B., Wilcox, A.C., Lytle, D.A., Hickey, J.T., D.R. 2004. The importance of beaver ponds to coho salmon Andersen, D.C., Beauchamp, V.B., Hautzinger, A., McMul- production in the Stillaguamish River basin, Washington, len, L.E. & Warner, A. 2010. Ecosystem effects of environ- USA. North American Journal of Fisheries Management 24: mental flows: modelling and experimental floods in a 749–760. dryland river. Freshwater Biology 55: 68–85. Pollock, M.M., Beechie, T.J. & Jordan, C.E. 2007. Geomorphic Smith, J.M. & Mather, M.E. 2013. Beaver dams maintain fish changes upstream of beaver dams in Bridge Creek, an incised biodiversity by increasing habitat heterogeneity throughout a stream channel in the interior Columbia River basin, eastern low-gradient stream network. Freshwater Biology 58: 1523– Oregon. Earth Surface Processes and Landforms 32: 1174– 1538. 1185. Snodgrass, J.W. & Meffe, G.K. 1998. Influence of beavers on Pool, T.K. & Olden, J.D. 2014. Assessing long-term fish stream fish assemblages: effects of pond age and watershed responses and short-term solutions to flow regulation in a position. Ecology 79: 928–942. dryland river basin. Ecology of Freshwater Fish. DOI:10. Snodgrass, J.W. & Meffe, G.K. 1999. Habitat use and tempo- 1111/eff.12125. ral dynamics of blackwater stream fishes in and adjacent to Pool, T.K., Strecker, A.L. & Olden, J.D. 2013. Identifying beaver ponds. Copeia 1999: 628–639. preservation and restoration priority areas for desert fishes in Stefferud, J.A., Stefferud, S.E. & Clarkson, R.W. 2009. an increasingly invaded world. Environmental Management Fishes: historical changes and an imperiled native fauna. In: 3: 631–641. Stromberg, J.C. & Tellman, B., eds. Ecology and conserva- Propst, D.L., Gido, K.B. & Stefferud, J.A. 2008. Natural flow tion of the San Pedro River. Tucson: University of Arizona regimes, nonnative fishes, and native fish persistence in arid- Press, pp. 192–216. land river systems. Ecological Applications 18: 1236–1252. Stefferud, J.A., Gido, K.B. & Propst, D.L. 2011. Spatially var- Pyke, G.H. 2008. Plague minnow or mosquito fish? A review iable response of native fish assemblages to discharge, pre- of the biology and impacts of introduced Gambusia species. dators and habitat characteristics in an arid-land river. Annual Review of Ecology, Evolution, and Systematics 39: Freshwater Biology 56: 1403–1416. 171–191. Turner, D.S. & List, M.D. 2007. Habitat mapping and conser- R Development Core Team. 2011. R: a language and environ- vation analysis to identify critical streams for Arizona’s ment for statistical computing. Vienna, Austria: R Founda- native fish. Aquatic Conservation: Marine and Freshwater tion for Statistical Computing. Ecosystems 17: 737–748. Rinne, J.N. 2005. Changes in fish assemblages, Verde River, Twardochleb, L.A., Olden, J.D. & Larson, E.R. 2013. A glo- Arizona, 1974-2003. In: Rinne, J.N., Hughes, R.M. & Cala- bal meta-analysis of the ecological impacts of nonnative musso, B., eds. Historical changes in fish assemblages of crayfish. Freshwater Science 32: 1367–1382. large rivers in the Americas. Bethesda, MD: American Fish- USGS (U.S. Geological Survey). 2013. National Water Infor- eries Society, pp. 115–126. mation System. Available at: http://waterdata.usgs.gov/nwis/ Rinne, J.N. 2012. Fish and aquatic organisms. In: Neary, (accessed 6 September 2013). D.G., Medina, A.L. & Rinne, J.N., eds. Synthesis of upper White, S.M. & Rahel, F.J. 2008. Complementation of habitats Verde River research and monitoring 1993–2008. Fort Col- for Bonneville cutthroat trout in watersheds influenced by lins: USDA Forest Service, Gen. Tech. Rep. RMRS-GTR- beavers, livestock, and drought. Transactions of the Ameri- 291, pp. 189–232. can Fisheries Society 137: 881–894. Rinne, J.N. & Miller, D. 2006. Hydrology, geomorphology and management: implications for sustainability of native