Introduction

Disturbance is a key factor that affects community structure in rivers and can cause local extirpations of individuals or populations (Brown et al. 2011), cause changes in community composition (Cañedo-Argüelles and Rieradevall 2010), facilitate displacement of native species by invasive species (Larson et al. 2009), and change body sizes of fish (Walters and Post 2008). Environmental and anthropogenic disturbances of freshwater systems have been increasing around the world (Vörösmarty and Dork 2000); quantifying how systems have reacted to previous disturbances is vital to better understand how these systems and the species they contain may react to further perturbation.

Riverine fishes and other aquatic organisms are subject to many disturbances, including altered flows (Poff and Zimmerman 2010), pollution events (Erős et al. 2015), and salinization (Bailey et al. 2006). Changes in salinity have been linked to increased metabolic rates (Toepfer and Barton 1992) and decreased reproduction (Hoover et al. 2013) in fish, as well as homogenization of (Miyazono et al. 2015) and changes in (Cheek and Taylor 2016) fish assemblages. Recovery of fish after a disturbance can depend on the intensity of a disturbance (Detenbeck et al. 1992), distance to a source population (Stoll et al. 2013), mobility (Albanese et al. 2009), and abundance in the watershed (Taylor and Warren Jr 2001; Albanese et al. 2009; Erős et al. 2015).

In systems containing endangered species, a disturbance can have much larger impacts than in systems containing less vulnerable species. Small, geographically-restricted populations with fragmented distributions have a higher likelihood of extinction or extirpation due to stochastic disturbance events than large, continuously distributed populations (Fagan et al. 2002). As a result, findings that document the effects of disturbance within systems containing endangered species are of particular importance to ecologists and wildlife or resource managers charged with the stewardship of these systems. In this paper, we examine the responses of six desert stream fishes to a disturbance due to a transient catastrophic water quality event.

In June 1985, flows within the mainstem Virgin River disappeared into a sinkhole just upstream of the Pah Tempe Springs. As a result of the increased hydraulic pressure from the subsurface flows, the Pah Tempe Springs more than doubled their discharge. This resulted in an increase in conductivity in the Virgin River 1.9–5.3 times normal rates (up to 1.1 S/m upstream of Ash and La Verkin Creeks). Remedial actions by the Washington County Water Conservancy District sealed off the sinkhole, returning natural flows of the Virgin River to the main channel. By August 1985, discharge from Pah Tempe Springs again stabilized (to 0.51 m3/s) and water quality conditions returned to natural levels.

This event provided an opportunity to study the effect of a discrete water quality disturbance event on the fish species living there. This is a historical dataset and changes in fish community composition were noticed by surveyors at the time, but were never analyzed in more detail. Our objectives were to examine changes in the composition of fish assemblages residing in that area of the river and their recovery from the disturbance. We expected to see a spatial gradient, with more severe changes in community composition with increasing proximity of sites to the hot spring (i.e., more upstream). We also expected that species with larger downstream population abundances (and therefore higher recolonization potential) would recover more quickly in affected sites than species with smaller downstream populations.

Material and methods

Study site

The Virgin River rises in the mountains in and near Zion National Park, Utah and flows about 320 km to its confluence with the Colorado River in Lake Mead, Nevada. Prior to the construction of Hoover Dam in the 1930s, the Virgin River was a tributary of the Colorado River. The specific study area occurs within the reach 3 km downstream from the Pah Tempe Springs to a point approximately 5 km upstream from the Washington Fields Diversion structure (Fig. 1). This 21 km reach of the river is characterized by high gradients, a well-developed stream channel that is dominated by a sand bed with interspersed areas of large cobbles, boulders and bedrock outcroppings. The gradient is approximately 3.9 m per km and is sufficient to sustain many well-developed riffles and pool habitats interspersed among deep runs. A total of six native species are present within this segment of the river, of which woundfin (Plagopterus argentissimus) and Virgin River chub (Gila seminuda) are currently listed as endangered by the U.S. Fish and Wildlife Service, and Virgin River spinedace (Lepidomeda mollispinis) is under review (United States Fish and Wildlife Service 2016). The Pah Tempe springs form a chemical/physical barrier to fish movement and marks the upstream limit of distribution for both woundfin and Virgin River chub.

Fig. 1
figure 1

Map of the study area with sampling stations (in rkm) marked (Google 2017)

Full size image

The Pah Tempe springs discharge into the Virgin River through a number of warm, saline spring sources in the lower end of the steep, narrow gorge known as Timpoweap Canyon near the town of Virgin, Utah (Fig. 1). Historically, Pah Tempe Springs sustained a constant discharge of 0.51 m3/s with temperatures of 43 °C and conductivity in the range of 1.6 to 1.8 S/m.

The saline springs combine with the steep gradient of Timpoweap Canyon to form a chemical/physical barrier to fish movement upstream of the springs. Downstream from the springs, as the thermal waters cool, precipitation of travertine cements the substrate, creating unfavorable habitat for fish and invertebrates. Under normal flow ranges, these conditions are moderated by discharge from non-saline cool water tributaries (La Verkin and Ash Creeks, Fig. 1), which enter the river about 2.4 km downstream. The conductivity just upstream of the confluence of these creeks ranges between 0.19 and 0.62 S/m and drops to a range of 0.15 S/m to 0.35 S/m just below their confluence. In 1985, during the disturbance event, Pah Tempe Springs increased the discharge by about 0.62 m3/s to an estimated 1.13 m3/s, and conductivity in the Virgin River upstream of the confluence of the tributaries Ash and La Verkin Creeks ranged between 0.94 S/m and 1.1 S/m, while below the confluence, the conductivity ranged between 0.65 and 0.79 S/m, increasing conductivity 1.9–5.3 times normal rates.

Collection methods

From July 1984 through October 1987, monthly collections were made at each of 4 stations along the Virgin River, and biannually at one station (Fig. 1). Stations are identified by their distance in kilometers from the downstream terminus of the river (river kilometer or rkm). At each site, 15 to 20 collections were taken in separate uniform meso-habitat areas; these areas were sampled in the proportions the meso-habitats were present at the site. Sample widths were restricted to 3 m, with a maximum sample length of 10 m. Sampling was accomplished with a 4.5 m by 1.8 m by 0.6 mm mesh nylon seine. Habitats were categorized as run, pool, riffle, backwater, eddy, snag, or unknown. Each meso-habitat was sampled repetitively until the population was depleted to a level equal to or less than 10% of the highest number collected in a given seining pass. In September 1984, November 1984, and March 1985, a pulsed, direct-current electrofishing unit was used instead of the seine for collections. A block net was placed at the downstream end of the meso-habitat being sampled and repetitive passes were made until the 10% depletion level was obtained. All fish were identified, measured and returned to the site of capture. A minimum of nine depth, velocity, substrate and cover observations were taken within each meso-habitat sampled. Conductivity, turbidity and pH were obtained by using a Hach DREL/5 model water chemistry kit. Air and water temperature were recorded to the nearest 0.1 °C.

Data analysis

Catch data at each station were analyzed as catch per unit effort (CPUE). We transformed CPUE abundances using the Horn-Morisita Index (Horn 1966), then used a distance-based redundancy analysis with disturbance (before sinkhole, during sinkhole event, and after sinkhole remediation) as a factor, stratified by sampling station. Morisita’s index was used because it is not affected by sampling size and/or diversity (Wolda 1981), and is a conservative indicator of dissimilarity (Linton et al. 1981). Horn’s variant of the index allows it to be used for purposes other than integer true count data (Oksanen et al. 2016b), such as in this analysis. Data were stratified by sampling location to account for differences in water quality due to increasing distance from Pah Tempe springs. All analyses were carried out in R v.3.2.4 (R Core Team 2016) using the package vegan v2.3–4 (Oksanen et al. 2016a). Abundance data were also examined visually to determine if there were any apparent patterns that could be observed and to identify any periods in which there was a failure to detect species at sampling stations. Figures were created using the package ggplot2 v2.1.0 (Wickham and Chang 2016), and loess regressions (with associated 95% confidence intervals) were added to aid visualization of patterns.

Our analysis is confounded by time because of a lack of control sites unaffected by the disturbance available for use. Sites further upstream not affected by the water quality event could not be used as control sites, as the natural ranges of woundfin and Virgin River chub have an upper limit of just downstream of Pah Tempe springs and are not found at sites past this natural barrier to dispersal. To assess if this is simply a trend related to temporal patterns and to characterize any patterns or absences, we examined the data visually.

Results

The assemblage composition was affected by the disturbance event; the effects were most pronounced in upstream sites (those closest to the disturbance) and in certain species. Of the six common species, three species disappeared from the collections at various times and at various sampling stations for extended periods of time (Figs. 2 and 3). Woundfin, speckled dace (Rhinichthys osculus), and Virgin River spinedace disappeared shortly after the disturbance occurred (September 1985) and appear to be associated with the event.

Fig. 2
figure 2

The CPUE abundances (fish/m2) of woundfin at the three most upstream sites (those closest to the disturbance). Zero values are coloured red to aid visualization of absences and vertical red lines correspond to initiation of disturbance (June 1985). Panel b shows the 13 month absence of woundfin from site 152 rkm following the disturbance

Full size image
Fig. 3
figure 3

CPUE abundances (fish/m2) of speckled dace, Virgin River spinedace, and Virgin River chub at the three most upstream sites (those closest to the disturbance). Zero values are coloured red to aid visualization of absences. Vertical red lines correspond to initiation of the disturbance (June 1985). Panels a–f show patterns of declining abundance in speckled dace and Virgin River spinedace after the disturbance, while h shows the increase in abundance of Virgin River chub

Full size image

Woundfin, the most abundant species, disappeared from 152 rkm (Fig. 2b), the site closest to the disturbance source, for 13 consecutive months (September 1985 to October 1986) but remained abundant farther downstream (stations 143, 135, and 132 rkm). Speckled dace disappeared at both upstream sites (152 and 143 rkm). They were missing from or present at only extremely low abundances in station 143 rkm from August 1985 (two months following the disturbance event) to October 1985, and completely absent in the farthest upstream station (152 rkm) from January to September 1987 (Fig. 3a, b). Virgin River spinedace showed a declining pattern of abundance around the time of the disturbance event at station 152 rkm (Fig. 3e), were rarely found in stations 143 rkm (Fig. 3d), 135 rkm, and never at station 132 rkm throughout the study.

In an opposite pattern to woundfin and speckled dace, Virgin River chub exhibited a slight increase in abundance following the sinkhole event at 152 rkm, being found in greater abundances following the event (Fig. 3h). Across all stations, the abundance of desert sucker (Catostomus clarkii) and flannelmouth sucker (Catostomus latipinnis) did not appear to differ meaningfully before, during, or after the disturbance event.

The redundancy analysis was statistically significant (999 permutations, Df = 2, F = 1.72, p = 0.009), and showed that within stations, species composition was significantly different between at least two of the three time periods (before, during, and after the disturbance) included in this analysis. This analysis remained unchanged when we excluded the three months in which electrofishing was employed as the sampling method.

The stations at 132 and 135 rkm, the two most downstream stations of the five stations included in the analyses, showed no apparent differences in species composition before, during, or after the sinkhole event. The upstream most station (153 rkm) was sampled much less frequently than the downstream sites (biannually vs. monthly); because of this, the time estimates for any temporal extirpations such as those characterized in the lower sites are not reliable. However, the data from this station can corroborate general patterns seen at more downstream sites. For example, this station corroborated patterns observed at site 152 rkm (the second most upstream sampling station), with woundfin, spinedace, and speckled dace showing downward trends following the sinkhole, and Virgin River chub, normally only found at very low abundances in this station, showing a brief, slight increase at this station as well.

Of the three species that experienced declines, woundfin appeared to have the most complete recovery rates. In 152 rkm (the station they disappeared from), woundfin abundance approached pre-disturbance levels by May 1987, nearly two years after the disturbance (Fig. 2b). Speckled dace remained at low abundances in 152 rkm (Fig. 3b) and did not appear to recover to previous abundance levels by the end of the study period (October 1987); however, in 143 rkm, populations of speckled dace returned to normal abundance levels starting in September 1987 (Fig. 3a). Virgin River spinedace abundances remained low in all stations, including 152 rkm (Fig. 3e), the station in which they had high relative pre-disturbance abundances, and did not fully recover from the decrease in abundance by the end of the study.

Discussion

Our study found significant changes in fish assemblages post-disturbance and differences in recovery rates. Virgin River spinedace, woundfin, and speckled dace experienced declines while Virgin River chub showed increases in abundance, and flannelmouth sucker and desert sucker showed no response. Changes in assemblage composition and subsequent recoveries suggest that fish assemblages in the affected study area followed an event-driven directional disturbance trajectory with gradual return to starting conditions, similar to the response in a creek in Oklahoma (Matthews et al. 2013). These patterns are also similar to those found in the Pearl River in in Louisiana and Mississippi (Piller and Geheber 2015), where the direction of response differed between species (two declining and one increasing in abundance post-disturbance). Differences in responses between all six species in our study may be associated with diet and food availability (see below). As expected, a spatial gradient in response to the disturbance was evident, with sites closer to Pah Tempe spring showing differences in abundance after the disturbance event, and sites more downstream not exhibiting changes in abundance between pre- and post-disturbance. This and the observation that patterns seen at 153 rkm were more similar to those found in the most upstream sites (monthly sampled) supports the idea that the observed community changes were caused by the disturbance event and were not simply due to temporal variation.

Woundfin recovered within two years, which is consistent with recovery times (two to six years) previously found for pulse (discrete, intense, temporally-limited) disturbances (Detenbeck et al. 1992). The recovery time of species in affected sites seemed to depend on the recolonization potential of that species. The species with the highest abundances in neighboring sites, i.e. woundfin, recovered most quickly. This is consistent with findings that abundance is an important factor with high predictive power for recolonization after a major disturbance (Albanese et al. 2009; Erős et al. 2015), with species with higher local abundances exhibiting higher immigration rates (Taylor and Warren Jr 2001). Distance has also been shown to be an important factor for recolonization success. For example, sites closer to recolonization sources experienced faster recovery rates than those farther away (Kubach et al. 2011), and fish in rivers in Germany were found to rarely recolonize sites greater than 5 km from source populations (Stoll et al. 2013). Distance to recolonization sources with high abundances may have played a role for speckled dace in our study, as this species started to recover near the end of the study period in the most downstream site (143 rkm), which was upstream of reaches with higher abundances (Fig. 3a).

In contrast, the recovery of Virgin River spinedace, primarily a tributary species, may have been restricted by poor dispersal from sites or low abundance (and thus a small source population) in the tributaries, although the mechanism is unknown because the tributaries were not sampled. A similar phenomenon was observed in the Marcal River in Hungary, where three rare tributary species did not recolonize the river within three years post disturbance (Erős et al. 2015).

Interestingly, different fish species exhibited different responses to the disturbance, which could be related to differences in diet and food availability. Fish relying heavily on macroinvertebrates seemed to be at a disadvantage compared to their omnivore or mainly detritivore-herbivore counterparts. All the species that showed long temporal extirpation or downward trends in abundance (woundfin, speckled dace, and Virgin River spinedace) are known to feed extensively on macroinvertebrate larvae, including chironomids, simuliids, and ceratopogonids (Greger and Deacon 1988). Insects can be sensitive to changes in salinity, with insect eggs and hatchlings less tolerant to high salinities than their adult forms, most exhibiting upper saline limits ranging from 0.62 to approximately 1.26 S/m (Kefford et al. 2004). The highest salinity recorded in the sample sites was 0.79 S/m, which was higher than the level at which 50% invertebrate die off is achieved (Berezina 2003), suggesting prey populations may have been greatly reduced by the disturbance event. Conversely, desert sucker and flannelmouth sucker, which feed primarily on detritus, debris, and filamentous algae (Greger and Deacon 1988), did not show any overt effects of the disturbance event on abundances. Interestingly, Cladophora, a filamentous algae genus found in the Virgin River (Greger and Deacon 1988) has been shown to be relatively robust and tolerant to increased salinity levels (de Paula Silva et al. 2008). The omnivorous Virgin River chub appear to have experienced increases in abundance in sites close to (and thus most affected by) the sinkhole event. The increase in Virgin River chub (at 152 rkm) could have been due to release from competition for food, as abundances of insectivorous fish had decreased.

The results of this investigation into the effects of a temporary but severe change in water quality can be used to predict how assemblages (and species) will change in response to altered temperature, salinity, or other natural disturbances, as well as used to inform conservation and management initiatives in the Virgin River. The woundfin is endangered and seemed to be most affected by the changes in the river caused by the sinkhole, disappearing from a station for 13 consecutive months. Preserving habitat for woundfin and minimizing disturbance within the system is an advisable measure to mitigate population losses (Hardy et al. 2003), which may also have positive effects on populations of other less vulnerable species such as Virgin River spinedace and speckled dace, which also showed deleterious responses to the disturbance.