Effects of acute suspended sediment exposure on the swimming and schooling performance of imperilled Redside Dace (Clinostomus elongatus)

Redside Dace were collected (from the Kokosing River (40.3601°N, 82.1601°W) in Morrow County, Ohio, USA and transported to the University of Windsor Freshwater Restoration Ecology Centre (Turko et al. 2020). Fish were held in captivity for ∼1 year before the experiment in round 800 L stock tanks (914 mm deep × 1549 mm in diameter). The tanks were part of a recirculating system filled with dechlorinated filtered water, ∼10% of the water exchanged daily. Gravel and artificial plants were added to the bottom to mimic their natural environment. Water temperatures were maintained at 16–18 °C, and fish were fed a diet of frozen blood worms and flake food twice a day. The light cycle at the facility followed a 12:12 diurnal cycle. The water temperature and pH were reviewed daily, and other water quality measures were checked biweekly. Fish were collected with a permit from the Ohio Department of Natural Resources, and all experiments were approved by the University of Windsor Animal Care Committee, which adheres to the mandate of the Canadian Council of Animal Care.

Swimming performance (critical swim speed) and kinematics (TBF, TBA, and Strouhal number) of adult Redside Dace was measured in a 30 L Loligo swim flume (Loligo® System, Viborg, Denmark; 19 × 25 × 42 cm swimming compartment) with the side viewing panel covered to reduce stress to the fish from observers. Food was withheld for 24 h before each trial to minimize any metabolic effects of digestion (Nelson et al. 2003; Killen et al. 2012). Individuals were tested across incremental flow speeds in a control treatment (i.e., without suspended sediment present, 0 mg bentonite/L; <1.0 Nephelometric Turbidity Unit (NTU); range 0.04–0.87 NTU, N = 12) and two sediment treatment levels commonly found in the wild; a relatively “low” concentration (20 mg bentonite/L; ∼2 NTU; range 1.39–3.13 NTU, N = 13) and a relatively “high” concentration (100 mg/L bentonite; ∼8 NTU; range 4.97–9.93 NTU, N = 13). Each fish was tested under only one condition. Bentonite clay was used as a surrogate sediment due to its small size and ability to stay suspended. This clay is commonly used in experimental studies of suspended sediment (e.g., Gray et al. 2014, 2016). Sediment levels were measured (±0.05 NTU) before and after the swim trial using a LaMotte® 2020i sedimentation meter (www.lamotte.com), and temperature (16.7–18.1 °C) and dissolved oxygen (8.01–10.17 mg/L) were measured using a YSI probe (www.ysi.com). At the end of each swim trial, mass (7.54 ± 1.23 g; all data are means ± standard deviation (SD)) and standard length (9.24 ± 0.50 cm) were measured. Finally, the swim flume was then emptied, rinsed, and refilled after every swim trial to limit any residual sediment.

Critical swim speed (U_crit), which is related to cardiac output of fish and considered a good predictor of swimming capability (Claireaux et al. 2005), was quantified using an established swim flume protocol for dace and other minnows (Nelson et al. 2002; Nelson et al. 2003; Nelson et al. 2008). An individual fish was haphazardly selected from a stock tank, gently introduced to the flume, and were allowed to acclimate for 30 min at a flow rate of 10 cm/s to orient themselves to the flow direction (Beecham et al. 2009; Berli et al. 2014; Boyd and Parsons 2016). The flow speed was then increased in increments of 5 cm/s every 5 min until the fish became exhausted—defined as when the fish could no longer beat their tail and swim forward for over 10 s (Nelson et al. 2002; Mateus et al. 2008). A trial therefore took approximately 60 min to complete (range 45–75 min). One control fish did not complete the first 5 min speed interval (5 cm/s) and was excluded from all further analyses. Critical swim speed (U_crit) was then calculated using the equation U_crit = U₁ + (t₁/t₂) × U₂, where U₁ = speed of last fully completed interval, U₂ = increment of speed increase (5 cm/s), t₁ = length of time on the last interval, and t₂ = time interval length (Nelson et al. 2003; Hildebrandt and Parsons 2016). Before each trial, a flow meter (Höntzsch NT. 2; www.hoentzsch.com) was used to calibrate the flume by measuring flow rates throughout.

All individual swimming trials were video recorded at 120 frames per second using a GoPro Hero 7 camera (www.gopro.com) placed above the swim chamber pointing down. An 18′′, bi-colour LED ring light (model MOBIRL18) was placed horizontally above the swim flume to allow even light through the swim chamber. After completing the swim trials, the recordings were cut into 1 min sections for each of the fully completed flow speed increments at a frame rate of 120 frames/s (7200 frames measured/1 min video). The 1 min sections began when the individual was in the middle of the swim flume chamber (Johansen et al. 2010; Bartolini et al. 2015; Halsey et al. 2018). We analyzed TBF and TBA at four-speed intervals (15, 20, 25, and 30 cm/s) by uploading the footage to the open source Fish Analyzer software (https://github.com/marciosferreira/fish_analyzer). The first clear observation of the fish in the video was manually selected as a model for the fish shape, making it easy for the software to detect the fish in subsequent frames. The orientation of the fish (i.e., heading pointing to the left or right in the video frame) and their location in the swim chamber was selected before running the program, also to make the fish detection more accurate in the subsequent frames. The Fish Analyzer software made a background subtraction, based on an image of the experimental setup previously uploaded, and then used the OpenCV library (OpenCV 2015, python based) to find and filter contours in the thresholded frames. After that, the software filters then detected contours by size and shape, making it possible to preview where the fish is in each frame and determine its midpoint, tail, and head position in the entire video. The software also allowed us to determine the exact point of the caudal peduncle, making the tail angle measurement more accurate. The software measured TBF (Hz) throughout the video, based on how many times the tip of the tail reached the extreme points up and down throughout the entire video. TBA was calculated as the average amplitude of all beating cycles. One method used to measure kinematic efficiency is to calculate the Strouhal number (a dimensionless metric related to oscillatory propulsion), calculated using the formula: Strouhal = (TBF * TBA)/U, where TBF = tail beat frequency, TBA = tail beat amplitude, and U = flow speed (Taylor et al. 2003; Nudds et al. 2014). A Strouhal number between 0.2 and 0.4 is generally considered optimal in fishes (Triantafyllou et al. 1993; Taylor et al. 2003) and also varies with swimming speed, generally decreasing with increasing swim speed (e.g., Lauder and Tytell 2005). Here, our goal was to measure if the Strouhal number changed in the presence of suspended sediment, which could indicate altered swimming efficiency.

Schooling behaviour was recorded using schools of Redside Dace (n = 6 fish/school) in an 850 L Loligo swim flume (Loligo® System, Viborg, Denmark; www.loligosystems.com; 135 × 45 × 45 cm swimming compartment; model SW10300) across a range of flow speeds (20–60 cm/s) with or without suspended sediment present (see below). The flow was calibrated using a Höntzsch flow meter placed in the middle of the swimming compartment as described above. Fish schools were created by haphazardly selecting fish from a stock tank (n = 23 schools, total N = 144 fish). Food was withheld for 24 h prior to experimentation (Nelson et al. 2003; Killen et al. 2012). Some individuals were used twice due to the limited number of adult Redside Dace available, but each school was composed of a unique combination of individuals (∼1 month between trials). Schools were assigned to either a control filled with dechlorinated water (0 mg bentonite/L; 0.50 NTU (SD ± 0.18), n = 12) or a sediment treatment (10 mg bentonite/L; 4.2 NTU (SD ± 0.32), n = 11), (Johansen and Jones 2013). This relatively low concentration of sediment (compared to the 20 and 100 mg/L concentrations used for the individual U_crit trials) was chosen as this was the limit for optical clarity within the large flume—at higher concentrations, fish could not be observed on camera when at the back or bottom of the swimming compartment. At the end of each schooling trial mass (5.65 ± 0.61 g, mean ± SD) and standard length (7.98 ± 0.52 cm, mean ± SD) of each fish was measured. The swim flume was rinsed and refilled after every schooling swim trial to clear leftover sediment.

The protocol for the schooling trials was similar to that used for individual fish. Schools were haphazardly selected and allowed to acclimate in the flume for 1 h at a flow rate of 10 cm/s prior to the experiment. The speed was then increased in increments of 5 cm/s every 5 min until 60 cm/s was reached, for a total trial duration of 110 min. Swimming was recorded using two GoPro cameras at right angles (1080p; Hero 3 and 4, www.gopro.com), one which recorded the front of the flume and the other from the top view, allowing for a three dimensional analysis of schooling behaviour (Stoltz and Neff 2006).

Videos of schooling behaviour from the swim flume were subsampled prior to analysis. We analyzed swimming behaviour at five speed intervals (20, 30, 40, 50, and 60 cm/s). Only the middle 3 min of each 5 min interval were analyzed to allow fish groups to stabilize at each swim speed. Within each of these 3 min swimming periods, we extracted one frame every 5 s to measure the position of each fish.

After subsampling, ImageJ was used to measure x and y coordinates at the snout and caudal peduncle of each fish in each frame, from both the top and side camera angles. Videos were synchronized using flashes of a laser pointer at the beginning of each trial, and coordinates were aligned in space using fixed reference points within the swim flume visible in both camera views (Hazelton and Grossman 2009; Kimbell and Morrell 2015). Polarization of each school was estimated by fitting a line through the snout and tail coordinates of each fish, calculating the angle of the fish relative to the long axis of the swim flume, and finally calculating the standard deviation of these six angles (i.e., a low standard deviation indicates a highly aligned group; Nadler et al. 2018). To calculate nearest neighbour distance, average spacing among individuals, and the volume occupied by the school, the approximate center of each fish was first calculated as the midpoint between the measured snout and tail coordinates. Nearest neighbour distance was calculated as the mean distance between the midpoint of each fish and the closest adjacent member of the school. Mean spacing was calculated as the mean distance between the midpoints of all possible pairs of fish (15 measurements). Finally, volume of the school was calculated as the smallest ellipsoid that could be fit around all individual fish.

To determine if the sediment levels (control, low sediment, high sediment) affected critical swim speed of individual fish, we used a linear mixed effects (LME) model with treatment being included as a fixed factor, Fulton’s condition factor (K) as a covariate (to ensure differences in condition factor among individuals did not impact the findings), and trial as a random factor. The effect of sediment on the TBF, amplitude, and Strouhal number were evaluated by running separate LME models for each variable with treatment and flow as fixed factors, Fulton’s condition factor (K) as a covariate, the interaction term between treatment and flow, and trial as a random factor. If a covariate term was not significant (i.e., P > 0.05), it was excluded from the model, and the mixed model was rerun. Next, post hoc Tukey’s tests were run for mixed models with a significant result.

To understand how sediment potentially influences schooling behaviour, we examined the interactive effects of sediment and flow on cohesion and polarization independently. Because we measured cohesion using five metrics (mean nearest neighbour distance, standard deviation of nearest neighbour distance, mean spacing among individuals, standard deviation of mean spacing, and school volume, see above for details), we used a principal component analysis (PCA) to reduce these data to a single variable (package PCAtest, factoextra; Camargo 2022). The first principal component (PC1) of all cohesion variables explained 86.6% of the variation and all variables loaded strongly in the same direction; PC1 was therefore used as an overall composite measure of schooling cohesion in subsequent analyses (Table 1). To understand how cohesion and polarization were affected by sediment and swimming speed, we then used linear mixed models that included treatment, and flow as fixed factors, school identity as a random factor, and mean fish body length of each group as a covariate (R packages; lme4, nlme). When significant differences were found (i.e., P < 0.05), we used a post hoc Tukey’s test using least square means (R packages; lsmeans, multicomp, multicompView). The raw data for this research will be available for download from the University of Windsor institutional depository, which is located at https://scholar.uwindsor.ca/.

Critical swim speeds were not significantly different among the control, low sediment, and high sediment treatments (F_2,35 = 0.08, P = 0.93; Fig. 1A). Suspended sediment similarly did not affect TBF (F_2,34 = 1.57, P = 0.22, Fig. 1B), amplitude (F_2,34 = 0.96, P = 0.40, Fig. 1C), or Strouhal number (F_2,33 = 1.63, P = 0.21, Fig. 1D). There was a significant positive relationship between swimming speed and TBF (F_3,91 = 268.33, P =< 0.0001; Fig. 1B) and TBA (F_3,91 = 9.02, P ≤ 0.0001, Fig. 1C). Strouhal number was not significantly influenced by swimming speed (F_3,91 = 1.53, P = 0.21, Fig. 1D). There were no significant interactions between sediment treatment and swimming speed for any variable we measured (TBF: F_6,91 = 0.76, P = 0.61; TBA: F_6,91 = 0.59, P = 0.74; Strouhal number: F_6,91 = 0.30, P = 0.94).

School polarization depended on a significant interaction between suspended sediment and water flow (F_4,84 = 4.10, P = 0.004, Fig. 2A). At low flows, polarization was unaffected by sediment. At higher flows, polarization increased overall, but polarization was higher in schools exposed to suspended sediment. School cohesion was negatively correlated with water flow (F_4,84 = 7.55, P =< 0.0001, Fig. 2B) and was unaffected by suspended sediment (F_1,19 = 0.0004, P = 0.98, Fig. 2B).

We found that critical swim speed was not affected by acute exposure to suspended sediment. Other studies have suggested that imperilled fishes are more sensitive to suspended sediment than nonimperilled species, but there is considerable interspecific variation (Gray et al. 2014; Wilkens et al. 2015; Hildebrandt and Parsons 2016). For example, in a comparison of five shiner species (Notropis spp.), critical swim speed of nonimperilled Blacknose (N. heterolepis), Blackchin (N. heterodon), and Bridle Shiner (N. volucellus) was unaffected by suspended sediment, while sediment exposure decreased the critical swim speed of imperilled Pugnose Shiner (N. anogenus) but increased it in imperilled Bridle Shiner (N. bifrenatus) (Gray et al. 2014). These changes in swimming performance occurred despite exposure to a relatively low concentration of suspended sediment (30 mg/L), versus the 20 and 100 mg/L used in our study. However, a major difference between these studies is that we acutely exposed Redside Dace to suspended sediment, while the Notropis minnows were acclimated for >30 days. We chose acute exposure as urbanized Redside Dace habitats typically experience increased suspended sediment concentrations only briefly after rainfall. Overall, our data suggest that swimming performance of Redside Dace is relatively robust to acute exposure to the low concentrations of suspended sediment that are typically measured in their Canadian habitats (Table 1; Eyles and Meriano 2010). However, in the future it would be beneficial to test Redside Dace swimming ability after prolonged exposure and in response to sediment levels that are higher than those used in this study to determine tolerance thresholds.

We estimated swimming efficiency by measuring TBF and TBA and using these to calculate the Strouhal number. Other studies suggest that efficiency can be affected by suspended sediment (e.g., Hildebrandt and Parsons 2016), and that Strouhal number is a good indicator of swimming efficiency (e.g., Nudds et al. 2020). As expected, TBF and amplitude increased at faster swimming speeds, while the Strouhal number remained constant. Furthermore, we found that sediment had no effect on any of these kinematic variables, mirroring our critical swim speed results. These results suggest that sediment exposure had no subtle performance consequences that could have been missed by our overall critical swim speed endpoint, further supporting our conclusion that swimming performance in Redside Dace is robust to acutely elevated concentrations of suspended sediment.

Redside Dace schools became less cohesive as water flow increased. Faster swimming speeds decrease school cohesion in many fishes, likely because it takes more energy and becomes more difficult for individual fish to stay close together during intense exercise, and because increased spacing may reduce collisions with conspecifics (Pitcher 1973; Hockley et al. 2014; Suriyampola et al. 2017). Cohesion was unaffected by suspended sediment, which was a somewhat unexpected result given that species adapted to turbid water typically increase cohesion when sediment is present, while the opposite response is found in clear-water species such as Redside Dace (Ohata et al. 2014). One possibility is that Redside Dace school cohesion was unaffected by suspended sediment because our experiments were conducted using fish motivated to swim in a swim flume, while most other studies have measured cohesion in still water. Perhaps strong rheotaxis in Redside Dace, coupled with the energetic benefits of schooling while swimming (Herskin and Steffensen 2005; Marras et al. 2014; Ashraf et al. 2017), compensate for the effects of turbidity and maintain cohesion in swimming fish.

Redside Dace schools became more polarized at higher swimming speeds, consistent with results from many other fishes (Viscido et al. 2004, Tunstrom et al. 2013; Kent et al. 2019).Increased polarization is thought to increase both swimming efficiency and information transfer among members of a school (Kent et al. 2019). Interestingly, we found that there was a steeper positive relationship between polarization and swimming speed in fish exposed to suspended sediment. One possibility is that this is an adaptive response driven by energetic expenditure, as enhanced swimming efficiency from swimming in a highly polarized school may compensate for the negative respiratory consequences of suspended sediment. An alternative hypothesis is that sensory information from the lateral line becomes relatively more important than visual information in fish exposed to suspended sediment, and this sensory bias increases rheotaxis and causes all members of the school to be more closely aligned with the direction of water flow and thus each other. Both the visual and lateral line systems are known to be important for shaping social interactions in fish schools, but the relative contribution of each can vary with changes in environmental conditions (Partridge and Pitcher 1980; Liao 2006; Bleckmann et al. 2014; Mogdans 2019; Mckee et al. 2020). At the spatial scale of our study (fish spacing ∼8 cm), the lateral line is likely to detect both conspecific position and overall water flow (Bleckmann and Zelick 2009). Given that cohesion was unaffected by sediment exposure; however, we speculate that orientation to the direction of water flow is the predominant mechanism underlying the increased polarization that we observed.

We noticed that swimming performance of Redside Dace was markedly improved by schooling. The mean critical swimming speed of fish tested individually was ∼40 cm/s, but schools of Redside Dace could maintain position in a large swim tunnel set at 60 cm/s for 5 min. Several studies that have found schooling provides an energetic savings of 10%–25% (Herskin and Steffensen 2005; Johansen et al. 2010; Marras et al. 2014). This is a major energetic benefit, but not enough to account for the swimming performance difference we observed in individual versus schooling Redside Dace. We hypothesize that Redside Dace in our study additionally benefited from minor variations in flow speed and turbulence that may have been present in the large swimming compartment (135 × 45 × 45 cm) that we used. Measurements taken with a vane wheel flow meter throughout the swimming compartment consistently validated our reported flow rates, but nonetheless some variation was likely present. Even minor variation could allow fish to benefit from flow refuging (especially in boundary layers near the walls) or vortex capture, which would increase the apparent swimming speed (Liao 2007; Liao and Cotel 2013).Thus, while we are confident in our conclusion that school polarization was increased by the presence of suspended sediment at high swimming speeds, we acknowledge some uncertainty about the highest swimming speeds that were maintained in this study.

Land use changes such as agriculture and urban development often lead to increased in suspended sediment concentrations in freshwater habitats (Collins et al. 2011; Kemp et al. 2011). Identifying potential changes in the swimming performance of fishes is an important component of understanding how various species will react to habitat changes, especially for imperilled fish species (Di Santo 2022). Interestingly, we found that the effect of suspended sediment on polarization of Redside Dace schools depended on swimming speed, highlighting the importance of considering interactions between these variables. Overall, however, we found that the swimming performance of adult Redside Dace largely unaffected by acute exposure to low levels of suspended sediments. This was somewhat surprising considering that Redside Dace population declines are widely thought to be linked to increased sediment inputs, although these data are correlative and the mechanism of harm is unknown (COSEWIC 2007). One possibility is that only chronic exposure to sediment, rather than the acute exposures we used here, has negative impacts on Redside Dace swimming performance. Suspended sediment may also negatively affect other behaviours (e.g., feeding, mating) rather than swimming performance. Yet another possibility is that adult fish may be relatively robust to sediment, while earlier life stages are more sensitive (Griffin et al. 2009; Suedel et al. 2017). For example, Sutherland (2011) found that spawning success was affected by increased suspended sediment levels in a minnow. Clearly, further research is needed to understand the causal mechanism(s) by which suspended sediment negatively impacts Redside Dace populations, if any, to improve the effectiveness of conservation strategies used to assist this imperilled species.