Introduction

Introductions of non-native fish often disrupt existing trophic relationships in co-evolved native fish assemblages (Carpenter et al. 1985; Oguto-Ohwayo and Hecky 1991; McDowall 2003; Eby et al. 2006; Walsworth et al. 2013). Whether a native fish population successfully adjusts to a food web transformed by introductions depends, in part, on the life history and behavioral resiliency of the native species as well as the functional role of the introduced species and extent of its establishment through dispersal and proliferation (Parker et al. 1999; Kolar and Lodge 2002). Natives may be reduced in number or displaced entirely by a highly successful invader through predation or depletion of critical resources (Oguto-Ohwayo and Hecky 1991; Moyle et al. 2003; Casal 2006; Britton et al. 2009; Ellender et al. 2014). Conversely, many introductions either fail to establish, or once established, produce benign ecological consequences (Ruesink 2005; Gozlan et al. 2010). For some invasive species, direct impacts can be predicted or later documented (Townsend and Winterbourn 1992; Ricciardi and Rasmussen 1998; Kolar and Lodge 2002); for others, trophic effects may be complex and difficult to detect or understand even after the species has become well established (Moyle and Light 1996; Parker et al. 1999). In both cases, non-native fish pose a risk to natives (Casal 2006), and potentially a grave risk to those already made rare or endangered from physical habitat alterations (Minckley et al. 2003; Olden et al. 2006).

The transfer of fish species among water bodies or watersheds by well-intentioned sport-fishing agencies and the sport-fishing public has created many problems for managers charged with conserving native fish populations (Cambray 2003; Hickley and Chare 2004; Vitule et al. 2009; Ellender et al. 2014). Those systems with a high degree of endemism are especially susceptible to successful invasions and trophic disruption (Ruesink 2005; Casal 2006; Fitzgerald et al. 2016). Often, river basins are subjected to multiple introductions and the connectedness of tributary networks almost guarantees widespread dispersal of successful invaders (Gido and Brown 1999). Such introductions are generally irreversible (Eby et al. 2006; Vander Zanden and Olden 2008; Vitule et al. 2009) meaning expensive control programs must be done in perpetuity. Limited funding for such programs necessitates targeting those species deemed most detrimental to natives (Parker et al. 1999; Johnson et al. 2008; Vander Zanden and Olden 2008; Gozlan et al. 2010).

Native ichthyofauna of the Colorado River system of North America was formerly characterized by low species diversity and a high rate of endemism. Fourteen native species occur in the upper Colorado River sub-basin (upstream of Lee’s Ferry, Arizona), eight of which are endemic. Some 50 species of non-native fish have become established since the late 1800s (Tyus et al. 1982; Valdez and Muth 2005). Though many nonnatives occur incidentally or inhabit high elevation lakes and streams, others are widespread and abundant throughout warm-water reaches of the major tributaries. There, four species of native fish are federally classified as endangered or threatened, including the system’s top piscivore, the Colorado pikeminnow Ptychocheilus lucius. This endangered long-lived cyprinid grows to lengths > 1 m and was formerly abundant throughout warm-water reaches of the Colorado River basin from southern Wyoming to the Gulf of California. Substantial range reduction now restricts extant populations to the upper Colorado River sub-basin, primarily in the states of Colorado and Utah. There, wild populations tenuously persist in the Green River and upper Colorado River mainstem. Recent population declines in both rivers have been attributed to inadequate recruitment rates (Osmundson and Burnham 1998; Osmundson and White 2017; Bestgen et al. 2018), and in the Green River system, periods of reduced adult survival (Bestgen et al. 2007). Causal factors most often implicated in this decline include changes in both abiotic and biotic elements of the riverine environment. Dams and diversions have created barriers to dispersal and migration and allowed water managers to regulate flows for human uses, resulting in changes to essential physical habitats and thermal suitability (Dibble et al. 2021). In addition, numerous, successful, non-native fish introductions have altered trophic interactions increasing predation pressure on native species (Pilger et al. 2008; Bestgen et al. 2018).

Assessed in this paper is the spatial and temporal variation in body condition of Colorado pikeminnow during spring prior to summer spawning and the relationships between body condition and abundance of certain non-native fishes. Body condition, based on a fish’s mass at a given length, expresses the overall well-being of the individual and its relative success in securing food. Condition factor (K) is often used as an index of a fish’s energetic condition or energy reserves that may be used for growth, reproduction, or migration (Koops et al. 2004). Mean condition derived from large samples can be compared among locations and time periods to gauge or monitor how groups of fish are faring. For endangered fish it is a particularly useful index because it can be measured non-lethally.

Poor condition in fish usually reflects poor feeding opportunities (Lambert and Dutil 1997) and can lead to increased natural mortality (Dutil and Lambert 2000; Brodersen et al. 2008). Reduced reproductive potential may occur when females with low lipid stores fail to ripen, resorbing developing oocytes (Henderson et al. 1996; Rideout et al. 2000). Additionally, egg or larvae size and quality may be reduced affecting realized fecundity (Hislop et al. 1978; Marshall and Frank 1999). Overall health and fitness can affect the ability to migrate and spawn successfully (Chapman et al. 2012). For juveniles, lack of robust health and adequate energy reserves may increase susceptibility to predation and over-winter mortality, thereby decreasing recruitment rate (Adams 2002). Although such possible effects have not been documented for Colorado pikeminnow, it is reasonable to assume that maintaining good condition benefits the individual and likely enhances population recovery potential. Previous studies indicated mean body condition of Colorado pikeminnow (> 200 mm long) from the upper Colorado River varied significantly among years, between upstream and downstream study reaches, and among fish size classes (Osmundson et al. 1998; Osmundson 2002).

Colorado pikeminnow diet information is very limited because individuals are rare and their endangered status precludes their sacrifice for stomach content analysis. However, stomach contents obtained in 1994 from individuals 400–550 mm long using a non-lethal back-flushing technique indicated that three small-bodied, non-native minnow species (fathead minnow Pimephales promelas, red shiner Cyprinella lutrensis, and sand shiner Notropis stramineus) collectively comprised a substantial part (69–80%) of the identifiable diet, regardless of reach or habitat type from which Colorado pikeminnow were captured (Osmundson et al. 1998). Because these fishes are abundant relative to other suitably sized prey, their inter-annual variation in abundance suggested a possible source of inter-annual variation in Colorado pikeminnow condition.

Fathead minnows were first detected in the study area during a 1958 survey (Smith 1959); red shiners, during a 1962–1964 survey (Taba et al. 1965); sand shiners, in 1971 (Holden and Stalnaker 1974, 1975). During 1962–1964, fathead minnows and red shiners collectively comprised 79% of the backwater catch (n = 2,785; Taba et al. 1965). With the addition of sand shiners, these introduced minnows have increased their domination of lower-reach backwaters, comprising an average 98.5% of samples seined during 1986–2013 surveys (Harding et al. 2014).

This paper addresses four questions pertaining to the upper Colorado River mainstem population of Colorado pikeminnow: (1) does body condition vary across years, study reaches, and fish size classes, as earlier results suggested? (2) is there a relation between abundance of the three common non-native cyprinids and inter-annual fluctuations in Colorado pikeminnow body condition? (3) is there a relation between the number of Colorado pikeminnow young-of-the-year (YOY) produced annually and body condition of adults in spring prior to spawning? and (4) might the three introduced minnows, either singly or collectively, negatively or positively affect Colorado pikeminnow in this population?

Methods

The study area included the mainstem Colorado River from its confluence with the Green River (river kilometer [RK] 0) in Canyonlands National Park, Utah, upstream to the Grand Valley Project Diversion Dam (RK 312) at Palisade, Colorado, and excluded the 19 km Westwater Canyon (RK 181–201). This white-water canyon provided a demarcation between two primary parts of the study area, a lower and upper reach. Separating the study area in this way had biological relevance: the lower reach contains the population’s primary nursery area, whereas the upper reach contains prime adult habitat, with essentially all adults in the upper reach having migrated there from the lower reach (Osmundson et al. 1998). Part of the upper reach included the lowermost 3.5 km of the Gunnison River downstream of the Redlands Diversion Dam. This major tributary enters the Colorado River at CO RK 275.

Colorado pikeminnow sampling occurred as part of a long-term, systematic, mark-recapture, population monitoring effort. Sampling crews specifically searched for this species during April-June for three consecutive years followed by two years of non-sampling. From 1991 to 2000, most captures were from trammel nets set in backwaters, i.e., zero-velocity, off-channel habitats including flooded canyon mouths, river-side gravel pit ponds, and non-flowing side channels. These areas were targeted because this species frequents such sites during spring snowmelt when main channel flows substantially increase (Osmundson and Kaeding 1989). Increased effort began in 2003 with additional field crews sampling both shorelines throughout the study area using watercraft-operated electrofishing. Backwater trammel-netting continued, but many fish were now captured along shorelines. During each sampling year, 2–5 complete passes were made from top to bottom of the two study reaches. Specific monitoring techniques and additional protocol details have been previously reported (see Osmundson and Burnham 1998; Osmundson and White 2017).

Captured Colorado pikeminnow were measured (nearest mm) from nose tip to end of the longest caudal fin lobe (total length: TL), and weighed with an electronic balance (nearest gram) after excess water was drained off. Fish were internally marked with a passive integrated transponder (PIT) tag that emitted a unique alpha-numeric identifier when scanned with an external wand. Fish were then released alive at the site of capture.

For each Colorado pikeminnow ≥ 200 mm TL captured from throughout the study area, relative body condition (Kn; Le Cren 1951) was calculated as:

$$K_{n} = {1}00 \times \frac{{M_{o} }}{{M_{e} }},$$

where Mo is the observed mass (g) and Me is the expected mass. The expected mass for a fish of a given length (mm) is calculated as:

$${\text{Log}}_{{{1}0}} M_{e} = {\text{Log}}_{{{1}0}} \left( {{\text{length}}} \right) \cdot m + b,$$

where m is the slope and b is the y-intercept of the length-mass relationship derived from large samples of fish from one or more populations (Le Cren 1951). In this case, length-mass relationships for Colorado pikeminnow were derived only from fish captured from the population inhabiting the study area and included captures spanning 14 years of sampling (Supplementary material Table S1). To maintain technique consistency, only weights and measures taken by the author and one other field crew member were used in length–weight relationship calculations. Because relative condition (hereafter referred to as ‘condition’or ‘Kn’) was assumed to increase somewhat from gonadal development as the spawning season approached, month-specific length-mass relationships were developed for each of the three sampling months [most Colorado pikeminnow spawn in the upper Colorado River in early July, but spawning may occur from late June to early September, depending on water temperature (McAda and Kaeding 1991)]. For example, expected mass of a mature individual captured in April would be somewhat less than if captured in June. To avoid autocorrelation issues, within-year recaptures were excluded from the analyses and only data from the first yearly capture were used. However, recaptured fish from previous years were included to increase sample size and because these fish represented part of the annual population being sampled. It was assumed an individual’s condition was independent of what it was one or more years prior. Once condition was calculated for each fish using the length-mass relation specific to the month of capture, all Kn values from the three-month period were pooled to derive an annual mean Kn.

To evaluate whether earlier observations of spatial and size-class variation in condition were supported with the longer-term data set (Question No. 1), mean condition comparisons were made between reaches (among years and among 100-mm length classes) as done in earlier analyses (e.g. Osmundson et al. 1998; Osmundson 2002). Differences were considered significant if 95% confidence intervals did not overlap (Schenker and Gentleman 2001).

Lower-reach, small-bodied fish surveys have been conducted by Utah Division of Wildlife Resources (UDWR) during late-September to early October since 1986, primarily to quantify production of Colorado pikeminnow YOY (see McAda et al. 1994). The lower reach contains the population’s nursery area and is where YOY and juvenile Colorado pikeminnow occur in measureable numbers. With YOY essentially absent in the upper reach, no similar long-term monitoring was conducted there; hence, the lower reach was the primary focus of this study. Two backwaters (if present) in each of 22 8-km sub-reaches were sampled with two non-overlapping seine hauls and nets laid on the beach were searched for YOY Colorado pikeminnow. Sympatric fishes, including the three aforementioned non-native minnows, were also counted. To conserve time and resources, sympatric species were enumerated only in the first seine haul of the first backwater in each 8-km sub-reach. From this seine haul, all fish were identified and counted, either in the field (if the number was small), or preserved in ethanol and later enumerated in the lab. Area seined was measured for later density calculations. Total number of each species and total area seined is reported by UDWR annually. Fall reach-wide catch rate (fish/100 m2) of small-bodied fishes was used here as a surrogate index for their relative abundance when Colorado pikeminnow were captured the following spring.

To address Question No. 2, simple linear regression was used to explore relations between mean Kn during spring (dependent variable) and the index of relative abundance of non-native minnows. Because juvenile roundtail chub Gila robusta, a native large-bodied cyprinid, also occurred in the Colorado pikeminnow diet and because relatively small numbers of their YOY appeared in fall backwater seine surveys, the relation of mean Kn with catch rate of that species was also tested. Linear regression was also used to explore relations between the index of relative abundance of YOY Colorado pikeminnow in fall (dependent variable) and mean Kn of adults (individuals ≥ 500 mm TL) the previous spring (Question No. 3). To partially address whether the three introduced small-bodied minnows might be detrimental to Colorado pikeminnow (Question No. 4), relations were sought between fall catch rates of Colorado pikeminnow YOY (dependent variable) and fall catch rates of each of the three small-bodied minnows. A negative relation might imply a predatory impact on larval Colorado pikeminnow. Catch rates from lower-reach backwaters were available from UDWR for years 1990–2013, though non-native minnows were not enumerated in 2001 and 2005. Hence, 22 years were examined. To further address Question No. 4, the value of small-bodied, introduced minnows as food and their relations to Colorado pikeminnow condition is discussed in light of information derived from the literature regarding their potential predatory or competitive effects on young Colorado pikeminnow.

Regressions and associated tests were performed using GraphPad Prism 9.5.1. and Microsoft Excel 2016. To ascertain whether assumptions were met for each regression, various tests were performed: in Prism, constant variance of residuals (heteroscedasticity) was tested, a linear relationship was verified with a Runs test, and normality of residuals was examined using a D’Agostino-Pearson test. In Excel, a Durbin-Watson statistic (D) was calculated for comparison with table values (Savin and White 1977) of upper and lower bounds to test for autocorrelation, or independence of residuals (alpha = 0.05). Variables were Log10 transformed as needed to meet assumptions.

Results

Body condition variation

Length-mass regression coefficients varied by month as expected (Table 1), with mass for a given length increasing from April to June. Mass gain, expressed as a percent of April mass, increased with fish size (Supplementary material Figure S1). For example, for fish 250 mm TL, June mass was 2.0% greater than April mass; for fish 500 mm TL, mass increased 5.4%; for fish 700 mm long, 7.1%; for fish 950 mm long, 8.6%.

Table 1 Month-specific, length–weight regression coefficients (SEs in parentheses) used in relative condition calculations for Colorado pikeminnows captured from the Colorado River, 1990–2013
Full size table

When condition data were pooled within reaches, mean Kn of all individuals > 200 mm TL varied among years in both reaches, often significantly (Fig. 1, top). Among-year changes in mean Kn in the two reaches generally tracked one another through time but major differences in degree of change occurred often. There were also a few differences in direction of change from one year to the next, but most such changes were not significant (one exception: from 2003 to 2004, mean Kn significantly declined in the lower reach while having significantly increased in the upper reach). Mean Kn was significantly higher in the upper reach than in the lower reach in nine of 14 years (64%), with differences most pronounced later in the study period.

Fig. 1
figure 1

Annual mean relative body condition (Kn) of Colorado pikeminnow ≥ 200 mm in total length (TL) captured from the lower and upper Colorado River study reaches (top panel), and mean Kn of Colorado pikeminnow by 100 mm length class in the two study reaches (bottom panel). Error bars represent the 95% confidence interval. Top graph: reach differences are denoted as significant (S) or non-significant (NS). Bottom graph: data were pooled from 14 non-consecutive sampling years, 1991–2013. Captures of fish > 700 mm TL were too few in the lower reach to calculate meaningful averages; similarly, in the upper reach, those < 400 mm TL and those > 900 mm TL were too few to provide means

Full size image

With data pooled from all years, mean Kn significantly varied among length classes in both study reaches, but patterns of variation in each reach differed markedly (Fig. 1, bottom). Lower-reach mean Kn initially increased with fish size up to those 300–399 mm TL, but then declined in larger length groups. In contrast, in the upper reach, mean Kn of sub- and young adults was stable, then significantly improved in fish > 600 mm TL, until it stabilized again in fish > 800 mm TL.

With lengths from all years pooled by reach, very few juveniles were detected in the upper reach (Fig. 2). Hence, essentially all individuals present were colonists from the lower reach. Very few (0.4%) were < 420 mm TL, signifying the length at which individuals begin immigrating. Based on length frequencies, most immigrants arrived when 400–550 mm TL. Large adults ≥ 650 mm TL comprised only 5.9% of the lower-reach subpopulation (years pooled); in the upper reach, they comprised 37.6%.

Fig. 2
figure 2

Length frequencies of Colorado pikeminnow captured from the lower and upper Colorado River study reaches. Lengths were pooled by 50 mm increments from 14 non-consecutive sampling years, 1991–2013. Example: a bar at 400 mm includes fish 400–449 mm long

Full size image

Condition relation with non-native minnow abundance

Combined catch rates of lower-reach, non-native minnows varied greatly among years (Fig. 3), ranging from 110/100 m2 (2008) to 2341/100 m2 (2002). Number of backwaters sampled annually ranged from 12 to 23 (mean = 19). Of the three species, none was consistently most abundant. Depending on year, sand shiners comprised 0–68% of the total; red shiners, 5–92%; fathead minnows, 2–53%.

Fig. 3
figure 3

Annual catch rates of three non-native cyprinids in lower-reach backwaters of the upper Colorado River, 1990–2013 (summarized from Harding et al. 2014). Non-native minnows were not enumerated in 2001 and 2005

Full size image

Significant positive relations were indicated between annual mean Kn and catch rates of at least two of the non-native minnows (Table 2; Fig. 4). For regression with catch rate of the three minnows combined (r2 = 0.536; p = 0.003), all assumptions were met (Supplementary materials Table S2). For regression with red shiner catch rate (r2 = 0.370; p = 0.021), all assumptions were also met. For regression with fathead minnow catch rate (r2 = 0.728; p = 0.0001), the Durbin–Watson test statistic indicated an inconclusive test for independence (Supplementary materials Table S3). However, after transformation of both variables, the independence assumption was met, and strength of the relation declined slightly (r2 = 0.683; p = 0.0003). Mean Kn tracked fathead minnow catch rate through time relatively well (Fig. 5). For regression with sand shiner catch rate, no relation (r2 = 0.066; p = 0.373) was detected; however, the test for independence was inconclusive and transformation failed to rectify this. Regression of mean Kn and catch rate of native roundtail chub YOY met all assumptions but no relation was evident.

Table 2 Regression statistics for relations between Colorado pikeminnow mean Kn (dependent variable) and an annual index of abundance of various minnow species (and combinations thereof) in the lower reach study area, 1991–2013. RS: red shiner; SS: sand shiner; FM: fathead minnow; YOY: young-of-year
Full size table
Fig. 4
figure 4

Relations between annual mean relative body condition (Kn) of Colorado pikeminnow during April-June and an annual abundance index for three introduced minnows: fathead minnow, red shiner, sand shiner. The abundance index is the catch-per-unit-area (fish/100 m2) in backwaters sampled in fall prior to spring Colorado pikeminnow sampling. Data are from the lower-reach study area, 1991–2013. CPM: Colorado pikeminnow

Full size image
Fig. 5
figure 5

Time series of annual mean relative body condition (Kn) of Colorado pikeminnow (200–599 mm long) during spring and an annual index of fathead minnow relative abundance. The abundance index is the catch-per-unit-area (fish/100 m2) in backwaters sampled in fall prior to Colorado pikeminnow sampling. Data from the lower-reach study area, 1991–2013. CPM: Colorado pikeminnow

Full size image

YOY abundance relation with adult condition

Catch rate of Colorado pikeminnow YOY in fall (dependent variable) had no relation (r2 = 0.062, p = 0.393) with mean Kn of lower-reach adults in spring prior to spawning, nor with mean Kn of upper-reach adults (r2 = 0.003, p = 0.841). Diagnostics indicated all regression assumptions were met. Among individual adults, very low condition (Kn < 75) was rare in the upper reach, occurring in only 2.1% of those sampled in year 2000. Lower-reach adults with Kn < 75 occurred in 11 of 14 years with the highest frequency (14.3%) in 1998. Catch rate of YOY was low in 1998 (1.89 fish/100m2), but still greater than in 50% of the 14 years examined.

YOY abundance relation with non-native minnow abundance

In 1996, Colorado pikeminnow YOY catch rate (21.05/100 m2) was much higher than in other years and was therefore treated as an outlier and excluded from the analyses. No relation was found between Colorado pikeminnow YOY catch rate and catch rate of the three non-native minnows combined (r2 = 0.070, p = 0.245), and diagnostics indicated all assumptions were met. When an exceptionally high outlier year (2002) for red shiner catch rate (1,398/100 m2) was excluded, a significant positive relation resulted between YOY catch rate and red shiner catch rate (Fig. 6; r2 = 0.433, p = 0.002), with all assumptions met. No relation between YOY catch rate and fathead minnow catch rate was found (r2 = 0.053, p = 0.315), nor a relation between YOY catch rate and sand shiner catch rate (r2 = 0.002, p = 0.850). In both cases, all assumptions were met.

Fig. 6
figure 6

Relations between an annual abundance index of Colorado pikeminnow young-of-the-year (YOY) and an annual abundance index for three introduced minnows: fathead minnow, red shiner, sand shiner, and combined minnows. Data from backwater sampling in the lower-reach study area during fall, 1990–2013. Nonnative minnows were not enumerated in 2001 and 2005. The year 1996 was considered an outlier due to an exceptionally high catch rate of Colorado pikeminnow YOY and was excluded from analyses (see text). CPM: Colorado pikeminnow

Full size image

Discussion

Support for earlier observations

Although mean Kn was similar between reaches during initial years of study (1991–1993), it was often significantly higher in the upper reach in later years. The longer time series also revealed significant mean Kn differences among years in both reaches. Also, support was provided for the earlier observation (e.g. Osmundson et al. 1998) that mean Kn decreased with increased fish length in the lower reach; whereas, in the upper reach, it increased. This suggests a long-term pattern rather than a short-term phenomenon. As lower-reach individuals grew beyond about 400 mm TL, mean Kn declined. Based on length frequencies, this decline coincided with the life stage when immigration to the upper reach began. Declining condition may have prompted exploratory movements, leading fish to better foraging opportunities upstream. There, both native and non-native fishes (primarily suckers and chubs) larger than small minnows were abundant and their capture and consumption presumably more energetically efficient for larger Colorado pikeminnow (Osmundson et al. 1998).

Condition variation among Colorado pikeminnow individuals

Although variance around annual mean Kn was often small enough for significant differences between reaches or among years to be detected (Fig. 1, top), condition of lower-reach individuals ranged widely within years (not shown). As such, non-native minnow abundance would be a weak predictor of individual Kn. This was partly due to length effects (Fig. 1, bottom), but other currently unknown factors await further study. Habitat selection might be one source of variation, with some individuals foraging in less lucrative sites, or perhaps, for those spending more time in warmer waters, greater metabolic demands might affect weight. Mercury uptake might also have some effect. Though not studied here, Osmundson and Lusk (2018) reported a relation between Colorado pikeminnow Kn and mercury in muscle biopsies, with concentrations varying widely among individuals.

Role of non-native minnows

Mean Kn of Colorado pikeminnows (200–599 mm long) was strongly related with catch rates of two abundant non-native cyprinids, though perhaps not with those of a third species. Such correlations do not prove a cause-and-effect relation. An alternate explanation might be that abiotic conditions that favor Colorado pikeminnow condition similarly favor fathead minnow and red shiner reproduction and survival. However, condition largely reflects feeding opportunities and the earlier diet analysis (Osmundson et al. 1998) provides support that these minnows serve as important forage for this native piscivore. Additionally, the year (1998) with the highest percentage of lower-reach adults (14.4%) with very low Kn values (< 75) was preceded by the lowest recorded fall densities of fathead minnows. Throughout their native ranges, these three species inhabit quiescent or slow-moving water with silt or sand substrates (Lee et al. 1980; Sigler and Sigler 1996), habitats common in the lower-reach study area. Native species of larger size (flannelmouth sucker Catostomus latipinnis, bluehead sucker Catostomus discobolus, and roundtail chub), common in the upper reach, are comparatively scarce in the lower reach (Osmundson et al. 2002). Considering non-native minnow abundance relative to other lower-reach prey, their forage importance should come as no surprise. Hence, a cause-and-effect relation can be reasonably inferred considering all lines of evidence.

The index of relative abundance from backwater sampling may only be a rough approximation of densities throughout all available habitats. Shorelines and other lower-reach habitats were not sampled. Variability in environmental conditions or crew seining experience at time of sampling might also affect the annual catch index. Additionally, this index may not accurately reflect spring densities because sampling occurred the previous fall rather than when Colorado pikeminnow were captured and weighed. However, spring body condition might, in part, be a carry-over from weight accrued the previous fall. Despite these uncertainties, mean Kn relations were surprisingly strong with catch rates of two introduced small-bodied cyprinids. For the non-relation with sand shiner catch rate, the inconclusive Durbin-Watson test suggests that if any autocorrelation was present in the residuals, it was not strong. There nevertheless remains some uncertainty as to the true relation between Colorado pikeminnow condition and densities of this third non-native forage species.

That mean Kn was most related to fathead minnow densities, when all three introduced minnows were similarly abundant, suggests fathead minnows were preferred or perhaps more vulnerable to capture. Adult fathead minnows are generally larger (43–102 mm TL) than red shiners (24–75 mm TL) and sand shiners (28–60 mm TL; Lee et al. 1980), perhaps making them a more profitable prey. Alternatively, the stronger relation for this species might result from the index being a more accurate indicator of reach-wide abundance than for the other two species if those fishes occurred in non-sampled habitats with greater frequency. The 300–399 mm length class of juvenile Colorado pikeminnow had the highest long-term mean Kn in the lower reach, suggesting these introduced minnows provided ideal forage for this life stage. Although larger individuals evidently continued to prey on these small-bodied cyprinids (Osmundson et al. 1998), results here suggest many were unable to maintain their earlier mass-to-length ratio.

No negative relation was detected between Colorado pikeminnow YOY abundance and abundance of any of the three non-native minnows. There was, however, a positive relation with red shiners, i.e., as red shiner catch rates increased, so did Colorado pikeminnow YOY catch rates. Rather than a cause-and-effect relation, both species likely responded favorably to similar environmental conditions, as postulated for positive relations observed between native and non-native fishes in the San Juan River (Gido and Propst 2012). Whether YOY would have been more abundant in the absence of red shiners remains an open question.

YOY abundance relations with adult body condition

Lack of a relation between Colorado pikeminnow YOY abundance in fall and mean Kn of spring adults in either reach prior to spawning, suggests mean Kn was sufficient in many or all years to not hamper reproductive success. However, many factors could have influenced age-0 survival between larval emergence and backwater sampling some three months later. Larval sampling during the spawning season would provide better information on reproductive success, but unfortunately, only fall YOY abundance has been monitored. Reproductive potential of adults with very low Kn values (< 75) might have been affected, but these were rare in the upper reach and averaged only 5% of lower-reach adults. Annual larval production and survival until fall may be more related to flow and substrate conditions or predation rates rather than to physiological health of spawners, at least at the condition levels observed here.

Management implications

Researchers and resource managers of the Colorado River basin have long vilified the three small-bodied, non-native cyprinids (e.g. Seethaler 1978; Karp and Tyus 1990; Hawkins and Nesler 1991; Beyers et al. 1994; Sigler and Sigler 1996; Tyus and Saunders 2000; Valdez and Muth 2005). These fish are believed detrimental primarily due to their numerical dominance of backwaters, a riverine feature considered critical as nursery habitat for larval and YOY Colorado pikeminnow and perhaps other native species (Seethaler 1978; Grabowski and Hiebert 1989; Tyus 1991). The concern is that abundant nonnatives might reduce growth and survival of young life stages of natives through competition and predation (Ruppert et al. 1993; Beyers et al. 1994; Muth and Snyder 1995; Bestgen et al. 2006). In response, control programs have been proposed and experimental control efforts implemented (e.g. Tyus and Saunders 1996; Trammell et al. 2004).

Beyers et al. (1994) demonstrated competition from fathead minnow larvae resulting in reduced growth of Colorado pikeminnow larvae when food (zooplankton) was made limited in aquaria. Others have assessed diet overlap in the wild with gut-content analyses of young life stages of Colorado pikeminnow and sympatric nonnative cyprinids (Jacobi and Jacobi 1981; McAda and Tyus 1984; Grabowski and Hiebert 1989; Muth and Snyder 1995). These backwater studies revealed high diet overlap between both larvae and small (22–40 mm TL) age-0 Colorado pikeminnow and red shiners, with both fishes consuming zooplankton and insects, primarily chironomid larvae. However, very little diet overlap was found with fathead minnows which primarily consumed diatoms, plant fragments and organic debris. Although young Colorado pikeminnow and red shiners eat similar foods, no studies have demonstrated food limitations in shared habitats.

Numerous investigators have sought evidence of non-native minnow predation on larvae of Colorado River basin endangered fishes both in the wild (stomach content analyses) and in the lab (aquaria studies). In Green and Yampa river backwaters, no fish were found in diets of small-bodied cyprinids by Muth and Snyder (1995), in < 1% of juvenile red shiners by Grabowski and Hiebert (1989), and in 15% of red shiners by Ruppert et al. (1993). In the San Juan River, native fish larvae were found in 95% of red shiners sampled (Brandenburg and Gido 1999). In all diet studies, no fish were found in sand shiners or fathead minnows. Bestgen et al. (2006) used aquaria and mesocosm predation experiments, larval drift-net field studies, and individual-based-model simulations to demonstrate predation by red shiners could significantly reduce Colorado pikeminnow larvae in Green River backwaters. In two other aquaria studies, red shiners were moderately predacious on razorback sucker Xyrauchen texanus larvae while fathead minnow predation was either absent or insignificant (Karp and Tyus 1990; Carpenter and Mueller 2008).

Previous investigators have suggested fathead minnows and red shiners may be important to juvenile Colorado pikeminnow as prey, occurring most commonly in diets of age-0 individuals > 40 mm TL (Jacobi and Jacobi 1981; McAda and Tyus 1984; Grabowski and Hiebert 1989; Karp and Tyus 1990; Hawkins and Nesler 1991; Muth and Snyder 1995). Fathead minnows in particular may provide an important link in riverine food webs, converting algae, organic detritus, and planktonic organisms to food for higher trophic level fishes (Sigler and Miller 1963; Sigler and Sigler 1996), and may serve to buffer predation on native fishes from problematic large-bodied nonnative piscivores such as smallmouth bass Micropterus dolimeu and walleye Sander vitreus, also present in the system.

Whether native lower-reach fishes formerly provided adequate forage for Colorado pikeminnow is not known. Prior to introductions, the Colorado River system historically contained a depauperate fish fauna (Miller 1961; Franssen et al. 2014). The study area contained only eight species: Colorado pikeminnow, three species of chub, three species of sucker, and one dace (Valdez and Muth 2005). Although natives no doubt evolved to exploit the unique characteristics of this system, low species diversity and modified habitat may have provided opportunities for invaders (Gido and Brown 1999; Leprieur et al. 2009), particularly for exploitation of low trophic level resources rarely limiting in aquatic systems (Moyle and Light 1996; Gido and Franssen 2007). A more robust native fish assemblage may have existed prior to substrate modification from upstream water consumption and river regulation (e.g. Osmundson et al. 2002), channel confinement (e.g. Graf 1978; Walker et al. 2020), and multiple non-native fish introductions. These changes may have disadvantaged native forage species while allowing pre-adapted invasive minnows to flourish. Now primary components of the lower-reach Colorado pikeminnow diet, their numbers may largely drive annual variation in mean body condition, in turn affecting patterns of upstream dispersal.

Fathead minnows, more plentiful and available than native forage species, and seemingly non-predatory, may have provided a net benefit to Colorado pikeminnow in the lower reach under current modified habitat conditions. Red shiners, however, if even moderately predacious on native fish larvae, might collectively lower recruitment rates, impeding recovery efforts. It is therefore prudent to assume any level of larval mortality attributable to red shiner predation likely outweighs any forage benefit the species provides. Sand shiners appear to play a benign trophic role, but additional studies are needed.

A novel, condition-based approach was used here to gain insight into effects three invasive small-bodied cyprinids might have on a native piscivore. Hopefully, these results will contribute to evidence-based resource decision making and help investigators working in other systems generate hypotheses. Subtle life history differences between fathead minnows and red shiners gleaned from the literature illustrate how impacts can vary by species despite shared habitats and body size similarity. Impacts from a given introduction may differ among localities, habitats, or even systems due to differences in cover, availability of alternate food items, or life histories of resident native fish. Species-specific impacts should therefore be assessed on a case-by-case basis. Fish introductions should always be discouraged and prevented; however, once a non-native becomes established, its interactions with members of the receiving fish assemblage must be understood so those species most ecologically disruptive are prioritized for control.