Introduction

Nearly 65% of surveyed rivers and their aquatic habitats are threatened by human activity and climate change, with extinction rates among fresh water fish species rivaling those of past geological events [1]. A wide variety of artificial structures pepper the earth's river systems and are essential for diverting water to meet the needs of human societies, such as providing municipal drinking water, creating hydropower, and supporting agriculture [2], [3]. Unfortunately, a single facility alone can impede or divert millions of fish per year [4] and the cumulative impacts of multiple facilities across the landscape have lead to ecosystem fragmentation and isolation in many systems [5]. Human manufactured disturbances can also exacerbate an animal's risk exposure if its life history strategy results in maladaptive behavior in novel settings [6], which can result in both immediate and long-lasting impacts on fitness [7], [8]. A prominent risk minimizing strategy undertaken by most fish species is to form into social groups or schools of varying coherence [9], yet we know little of this behavior's impacts on individual risk as fish are forced to navigate past artificial barriers in their environment.

Human barriers, such as dams, water diversion facilities and pumping stations can expose fish to a variety of dangers, including physical harm from collisions, exhaustion from swimming exertion [10]–[12], increased stress levels [13], and elevated predation [2], [14]. Juvenile fish are particularly susceptible to the risks posed by combinations of these stressors and suffer the highest mortality rates when navigating through artificial bottlenecks [3]. Even if individuals escape harm, many migratory species are on a tight physiological schedule (e.g., the ‘smolt window’ in Salmonids). Migratory fish species respond to environmental cues, such as photoperiod, temperature, and flow, and undergo physiological changes to prepare for their osmotic transition between freshwater and saltwater regimes [15]. The optimal navigation strategy is therefore to pass artificial barriers quickly in order to minimize the risk of any immediate threats [10], which would also reduce any migration delays that can have negative consequences on future stock success [15], [16].

Engineering and operational efforts predominantly rely upon estimates of average individual swimming stamina and behavior to expedite the safe passage of a region's fish. The design objective is to enable a fish's capacity to avoid flows or structural designs that would otherwise either impede its progress or increase its mortality risk [17], [18]. Traditionally, most assays on barrier exposure or individual swimming stamina are conducted with large juveniles or adults that are tested in isolation [17], [19]–[21]. Recent efforts have questioned the transferability of these empirical estimates to field conditions, pointing out that adaptive behaviors are not simply a function of individual physiological and biomechanical performance metrics [2], [22]. Experiments with groups of fish challenged with bypassing a barrier have shown large differences in their navigational performance relative to estimates reported from prior trials on individuals [22]. However, we know little of how social interactions affect individual swimming patterns under such conditions. This information would be particularly useful in defining movement parameters in agent-based modeling efforts that couple individual behaviors and environmental conditions to assess how animals navigate past engineered structures in their environment [23], [24].

Most juvenile fish form into groups of varying social and physical organization [9]. While this strategy has primarily evolved as a means to reduce the risk of predation, it can also effectively mitigate travel costs. Social cues can improve migration success by serving to average out individual directional uncertainties along a gradient, or migration route, [25], [26] and can enhance directional decision-making [27]. Organized formations may also convey net energetic benefits, such as reducing the drag that individuals experience [9], [28]–[30]. Despite all of these potential benefits, environmental context can alter the adaptive value of an ingrained behavior. For instance, if environmental conditions generate severe disorientation then solitary navigation is hypothesized to become preferable to a social navigation strategy, such as schooling in fish [31]. Similarly, deviations from optimal positions in schooling fish can be energetically costly and definitive empirical evidence to support any hydrodynamic advantage to schooling remains elusive [29]. While context may reshape the costs and benefits of social movement strategies, like schooling, this strategy should none-the-less directly impact the basic movement parameters related to the speed and orientation of animals on the move.

In this study we tested the hypothesis that social interactions alter the swimming behaviors of fish when they are forced to navigate past an artificial barrier. Palmetto bass, a Morone chrysops × M. saxatilis hybrid, were selected for their availability and propensity to display polarized schooling behavior when young. We began by determining individual swimming performance using ramped velocity tests to establish the water velocity necessary to challenge the average individual's swimming stamina (experiment I). Subjects were then released upstream of a behavioral barrier under solitary and social conditions to evaluate how neighbors altered individual swimming behaviors and, subsequently, impacted risk exposure (experiment II). We found that risk exposure varied dramatically between fish swimming by themselves or within schools, where risk was measured as the time taken to navigate successfully downstream past the barrier. A fine-scale kinematic analysis further revealed how individual behaviors varied under each treatment. We conclude by discussing how the observed patterns relate to existing theory and empirical evidence concerning fish swimming behavior.

Materials and Methods

Fish husbandry

Juvenile palmetto bass were obtained from Keo Fish Farms, AR, and maintained in flow-through cylindrical tanks (1.2 m in diameter, 1.4 m high), fed daily to satiation and experienced light: dark cycles typical of summer months in the northern hemisphere. Experiments were conducted in a 16 m flume (Figure 1). Water temperatures in the rearing pens and experimental flume varied between 19–20°C. A total of 283 fish were used in our experiments and all fish were used only once. Fish were drawn at random from the main population tank and transferred to the flume within an aerated bucket. After their trial each fish was allowed to rest in the aerated bucket until they resumed normal swimming behavior. Fish were then moved to a recuperation tank with the same dimensions and environmental conditions as that of the main population. If any fish was unresponsive for 24 h after their trial, due to either exhaustion or impingement against a retention screen, they were euthanized using MS-222 (Tricaine Methanesulfonate) at a prescribed dose of 400 mg/L.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Top-down view of the experimental flume.

The area included the swimming apparatus used in the stamina trials (a), the holding pen (b) and the louver-style hydraulic barrier (c), which leads to a 23 cm wide exit. Flume dimensions are in meters and flows within the flume were recreated using a Computational Fluid Dynamics model (CFD), with color profiles representing the speed of the water in in section (c). The floating wire mesh cage (a) was suspended within area (b) with a series of cords. All stamina trials were done before the downstream barrier was installed. The upstream and down-stream boundaries of the holding area were enclosed with 1 × 1 cm² wire screens.

https://doi.org/10.1371/journal.pone.0108220.g001

Results

Solitary vs. schooling behavior

Solitary individuals tended to turn downstream and swim with the current (negative rheotaxis), while grouped fish predominantly faced upstream (positive rheotaxis; Watson's test, , , Figure 2, Table 1). These differences in preferred orientation occurred early on and remained steady over time (Figure S2 in File S1). Differences in turning angles between treatments were marginal, but significantly different from one another (Watson's test, , ; Table 1). Solitary individuals also swam at significantly slower swimming speeds than their schooling counterparts (Wilcoxon rank sum ; ), yet displayed stronger shifts in acceleration (, , Table 1). The swim speeds of these solitary fish were unsteady and increased over time, as opposed to their accelerations, which remained steady as they moved downstream. In contrast, the proxy fish swimming in groups displayed steady speed and acceleration profiles throughout their tests (Figure S3 in File S1). Taken together these behavioral discrepancies led to a 23-fold increase in exposure times when fish were released in groups rather than alone (Table 1, Wilcoxon rank sum, , ). While exposure times under the social treatment were skewed and contained an outlier, omitting this trial had little impact on the median exposure times or any of the remaining kinematic data reported in Table 1. Interestingly, fish in either treatment were equally capable of safely reaching the exit (94% solitary; 98% social) and both treatment groups showed similar post exposure mortalities from either impacting the barrier or from impingement against a downstream retention screen (10% vs. 12%). In summary, a comparison of Table 1 and Figure 2 demonstrates that solitary fish turned with the current and so left the system quickly without having to exert themselves (low swim speeds). In contrast, fish traveling in groups faced into the current and effectively worked harder to hold their station (greater swim speeds), thereby substantially increasing their exposure times.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Circular histograms for the orientation patterns recorded under both solitary (red) and schooling (purple) conditions.

Color modes are semi-transparent to show regions of overlap in the data from each condition. Schooling data represent only the proxy member's orientations from each replicate. The figure shows values pooled across all subjects within each category while in our statistical analyses we used the mean orientations from each replicate for both conditions. Upstream and downstream orientations are 180° and 0°, respectively. The dashed line represents the relative angle of the barrier with respect to the flow of water.

https://doi.org/10.1371/journal.pone.0108220.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Swimming characteristics of fish traveling either alone or within schools.

https://doi.org/10.1371/journal.pone.0108220.t001

Schooling dynamics

Fish released in groups displayed varying degrees of cohesion within and across trials, splitting and remerging as they moved in and out of camera range. Exposure time was not significantly correlated with mean group size (Spearman rank-correlation; × ; ), but was strongly correlated with group instability (, , ; ). Figure 3 shows how overall exposure times were negatively correlated with increasing group instability, while no distinctive pattern arose with respect to the average size of the groups (where mean group size is shown by the diameter of the points). Retaining the outlier in exposure times (Figure 3) changes neither the direction nor the significance of either of these group-level associations. At the individual-level we found no correlations between either the nearest neighbor distances or bearings with regards to overall exposure time. Nearly all of the proxy fish chosen from each group (10/11) maintained a steady and close association with their nearest neighbors (median ; ± MAD) and predominantly trailed these neighbors at bearings between 30° and 330° (Figure 4). We found no evidence of any temporal trends in nearest neighbor positions or bearings, so while average trends in and fluctuated over time these processes remained relatively steady (see Figures S4 and S5 in File S1). There was, however, greater variability in values over time, which is not unexpected. As fish speed up or slow down even minor shifts in the relative distances between individuals can result in larger changes in bearings, as a fish's closest neighbor can suddenly shift orthogonally, or from front to back.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Non-linear correlation in exposure time with increasing variability in group size.

Points represent the degree of variance observed across group sizes within each trial, while their diameters are scaled by the mean group size from those trials. While fish in the social treatment were released in groups of 14 individuals, these groups often fragmented and randomly selected representative groups varied in size both within and across trials. Overall, fish in the social treatment formed groups that ranged in size from 2–14 individuals with mean group sizes across trials ranging from 3.3 to 13. Solitary stragglers were also not uncommon (see SI-2). While extreme event (a) may represent a biologically plausible scenario, its absence does not detract from the relationship between variance in group size and exposure time.

https://doi.org/10.1371/journal.pone.0108220.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Nearest neighbor distributions within schools based on data pooled across all observations (replicates and time).

Neighbor positions are taken from the perspective of each group's proxy fish. The distance, , and bearing, , between a subject and its nearest neighbor (a) are presented in circular sections of 1 body length increments (FL) and binned by 30° intervals (b). Color values represent the normalized frequency of observations per bin and range from 0 (white) to 1 (dark blue). The highest concentration of nearest neighbors fall within 1 FL of the proxy fish, which tended to trail their leading neighbors at angles ranging from ±30° from their heading.

https://doi.org/10.1371/journal.pone.0108220.g004

Discussion

Fish released in groups up-stream of a diversion barrier formed schools and took longer to navigate downstream past the barrier than fish that were released alone. Schooling has evolved as an adaptive behavior largely because it decreases the immediate risks that fish face, whether it be by diluting the threat of predation, or reducing directional uncertainty [9]. Anthropogenic structures in rivers and streams are dangerous places for juvenile fish [2], [11], [12], [14], [15], [17] and their risk of mortality increases with their exposure to these environments [10]. Additionally, the immediate threats that fish face when navigating past diversion barriers (e.g., collision, exhaustion, stress) can also be compounded by indirect costs, such as detracting from more profitable opportunities like searching for resources or mates in other areas (e.g., the risk allocation hypothesis) [38]. Our results demonstrate that the long held practice of managing water diversion facilities based on individual fish swimming stamina [17], [19]–[21] can significantly underestimate how long fish will linger along a diversion barrier.

Results from our stamina trials indicated that a current of 48.6 approached the upper limits of our subjects' swimming stamina (46 – 49 ). Under these circumstances our subjects would generally not be expected to hold their station for extended periods. Yet, water velocity alone is a poor predictor of fish residency times within and across species [23], [24], suggesting that individual behavior may either reduce or exacerbate exposure times. As fish move downstream with a current even minor trajectory deviations brought about by obstacle avoidance maneuvers can change the time they take to traverse an area. For instance, a biased-random walk model, parameterized to reflect fish swimming speeds, demonstrates that the expected exposure time of a particle can increase by a factor of 5 simply by accounting for self-propulsion and obstacle avoidance behavior [37]. Within the relatively uniform flows found in our flume a passive particle released from the holding pen would take approximately 12 s to drift downstream past the barrier if it traveled in a straight line towards the exit. Although the shorter exposure times observed with the solitary individuals may fall within some, albeit large, margin of error of this expectation, the 23 fold increase in exposure time by those fish released in groups represents a substantial deviation from a purely hydraulic prediction. Group members swam at twice the speed of their solitary counterparts, showed less erratic swimming behavior, and were predominantly moving upstream. To understand how such pronounced differences could have arisen requires a brief review of the physical, energetic, and behavioral factors that may influence the observed patterns and our ability to draw broader conclusions from them.

Palmetto bass share diet and habitat preferences with their maternal species in the wild and integrate themselves into striped bass spawning migrations [39], [40], which suggests at least parallels in their movement behaviors under natural conditions. Palmetto bass also display an innate ability to form polarized schools when young and so they provide a pragmatic means to study the physical implications of coordinated motion. Hatchery-bred hybrids are likely to suffer the same biases displayed by hatchery fish, such as ecological naïveté with respect to foraging or predator avoidance. Yet, despite such limitations hatchery reared fish have proven useful in studying the fundamental mechanics of collective motion, yielding insights into topics ranging from gradient detection [41] to social learning and decision-making [42], [43]. These emergent behaviors stem from the basic physical mechanics of coordinated motion, which can have ecologically relevant impacts in wild populations. For instance, the speed and distance travelled by Chum (Oncorhynchus keta) and Pink (O. nerka) salmon generally increases with the size of the schools they form [44]. As a group grows in size, or density, the frequency of local interactions increases and can result in directional feedbacks that reduce overall turning variability among individuals [25], [44]. Similar behavior has been documented in locusts [45], demonstrating that the underlying physical mechanics can have an impact that transcends taxa and context. Care should none-the-less be taken in interpreting our results beyond the physical ramification of displaying polarized schooling, such as the potential ecological reasons that influence when or why a fish species will school.

Data also suggests that the size of the group may play a role. For example, the ability of Atlantic salmon (Salmo salar) and American Shad (Alosa sapidissima) to navigate through bypass weirs decreases with increases in the size of the schools they form, with the majority of fish breaking off from larger groups and passing their respective barriers in pairs or alone [46]. Fish exposure time in our study was negatively correlated with group instability, suggesting that fish in groups may have been passing through the system more quickly as social interactions degraded. Taken together these associations highlight the need to test for a group size effect to determine if there is a causal relationship and, if so, determine how this factor can influence management strategies.

Fish traveling in groups in our study followed leading neighbors more closely than those swimming alongside them (Figure 4b). This pattern can arise from two alternative, yet not mutually exclusive, ecological mechanisms. In the early 1970's, Daniel Weihs theorized that hydrodynamic advantages can arise predominantly from three means: channeling effects between parallel neighbors, flow sheltering behind neighbors, or by capitalizing on the energy generated by trailing vortices [28], [29]. Channeling effects are enhanced when individuals are separated by body length and these effects decay rapidly with distance [28], [29]. Our subjects tended to position themselves further than this optimum, yet predominantly followed or led neighbors at bearings that closely corresponded with those predicted by Weihs' theoretical diamond configuration (±30°). Moreover, these patterns were not only spatially structured, but also remained relatively steady over time. Taken together these patterns suggest that potential hydrodynamic gains were theoretically plausible and, if present, more likely to stem from either flow avoidance or vortex capturing than from channeling effects between parallel swimmers.

Recent evidence has shown that leader positions are significantly correlated with an individual's metabolic scope (MS), with spatial positioning from a school's front to rear being negatively correlated with the constituents' MS [47]. Individual fish have also been shown to display less muscle activity when drafting behind the vortices shed from a stationary object [29], so it is certainly possible for individuals to take advantage of similar physical mechanisms that may arise within a school. However, we continue to find opposing predictions from theoretical efforts exploring the hydrodynamic benefits of schooling [48], [49]. The leader-follower dynamics seen in our subjects are found outside of advection dominated systems [50]–[52] and may therefore arise for ecological reasons beyond any hydrodynamic advantage. Leadership positions are typically ephemeral in fish groups as they generally lack a social hierarchy [52] and are more commonly attributed to foraging information [50]–[52], aerobic capacity [47] and risk taking [9]. All of our individuals were fed to satiation, were naïve as to the location of the exit route, and either rarely or never exceeded their expected critical swimming speed. So, while the positional patterns in Figure 4b provide evidence to support some of Weihs' assumptions, another plausible explanation is that risk perception can play an equal, if not more important, role in such artificial environments.

Consider that ‘leaders’ within groups can either possess information, or merely display bold tendencies [50], [53]. Bold animals typically commit quickly to reactionary behaviors, such as fleeing or attacking, and show less variability in their movements than the typical individual [54]. We might therefore expect bold individuals to move quickly and decidedly downstream past the barrier, as displayed by the solitary travelers. However, our solitary subjects showed greater variability in their turning angles and accelerations than those fish released in groups. Such elevated variability is better associated with shy or risk-averse individuals than with bold ones [54], which suggests that solitary individuals were simply more skittish than those released in groups and moved through the system quickly due to their orientation in the water. While social interactions affect an individual's reaction to stressful scenarios (e.g., predators), such interactions also significantly impact stress responses at the behavioral and metabolic level [55], [56]. We can simplify this metabolic argument by considering that in the simplest models of collective behavior, individuals average the motion of their neighbors in order for any degree of coordinated or cohesive motion to emerge [9]. This elementary assumption alone invariably leads to a dampening of individual movement variability, which may, in turn, indirectly decrease metabolic rates caused by erratic movements. Regardless of whether fish traveling in groups were able to work less to remain upstream of the barrier, their interactions significantly reduced any propensity for erratic movements.

In conclusion our findings demonstrate that schooling can enhance individual risk exposure for fish swimming through artificial environments. Future efforts would benefit from exploring how group size, hydraulic gradients, and structural complexity influence schooling behavior and individual exposure times along manufactured structures and obstructions in aquatic systems. While a thorough understanding of the fluid dynamics within schools remains out of reach, including basic social interactions when modeling animal movements may improve our ability to provide reliable risk assessments.

Supporting Information

Acknowledgments

We are grateful to the U.S. Bureau of Reclamation's Hydraulic Investigations and Laboratory Services group for hydraulic modeling and flume arrangements. We are indebted to Nathan Welch for assistance with laboratory work and useful discussions, Vinay Patel and Jason Derico for their patient video processing and fish tracking efforts, and Eva Enders and Jordan Watson for helpful comments on an earlier version of the manuscript. We also thank three anonymous reviewers for their time and constructive comments. Funding for this work was provided by the Bureau of Reclamation's Tracy Fish Improvement Program (No. FC810748) and the U.S. Army Engineering Research & Development Center's basic research program in network sciences (BT25/NS/14–63).