Fish swimming mode and body morphology affect the energetics of swimming in a wave-surge water flow

Physical environmental factors play an important role in shaping the structure of marine communities, with interactions between the species' physiology and environment often determining distribution patterns from local to regional biogeographical scales (Bellwood et al., 2002; Dunson and Travis, 1991; Fulton and Bellwood, 2005; Perry et al., 2005). In particular, the intensity of wind-driven wave activity on coral reefs has been shown to shape small-scale community structures and assemblages of reef fishes, as well as interspecific and intraspecific variation in body form and function (Bejarano et al., 2017; Binning et al., 2014; Fulton and Bellwood, 2005). Wave activity is, therefore, thought to be one of the primary drivers of the diversification of body and fin forms, and modes of locomotion found among reef fishes (Fulton, 2010).

Reef fishes are often highly site-attached, meaning they depend on specific feeding and/or refuge sites, and must rely on station-holding behaviors to maintain their position on the reefs in the face of wave-induced water motion, or wave surge (Fulton and Bellwood, 2005). The majority of shallow-water coral reefs are subject to these unsteady water flows, and reef fish station-holding behaviors can be a challenging and energetically expensive endeavor (Heatwole and Fulton, 2013; Marcoux and Korsmeyer, 2019). Station-holding fishes must continuously accelerate and decelerate to match the changing water flow speeds, which can increase the amount of energy used relative to maintaining a constant swimming speed (Liao, 2007; Roche et al., 2014). Furthermore, the adjustment of swimming speeds is coupled with direction changes, requiring backward swimming or a whole-body rotation to face the direction of the water flow (Marcoux and Korsmeyer, 2019; Schakmann et al., 2020).

Coral reefs, in particular, are subject to a wide range of wave-surge water flows, while also having an extremely diverse community of reef fishes with a spectrum of different swimming modes and body morphologies (Bellwood et al., 2006; Fulton and Bellwood, 2005; Langerhans and Reznick, 2010; Larouche et al., 2020). Some of these traits may be adaptations to decrease energy expenditure and improve swimming performance in such dynamic environments (Blake, 1983; Langerhans and Reznick, 2010). The majority of fishes use lateral undulations of the body and caudal fin for propulsion (BCF swimming), which is mostly associated with continuous cruise swimming and higher swimming speeds and is more often found in pelagic fishes (Friedman et al., 2021; Videler, 1993). On coral reefs, however, most fishes use their median and/or paired fins for propulsion (MPF swimming) and, specifically, pectoral fin swimming predominates with at least 60% of species utilizing this swimming mode (Fulton and Bellwood, 2005). Pectoral fin swimming, also known as labriform locomotion, uses the movement of the paired, pectoral fins to produce thrust while keeping the body rigid and straight (Videler, 1993). MPF swimming is thought to be good for high maneuverability and, therefore, an adaptation to moving through complex habitats such as coral reefs (Sfakiotakis et al., 1999; Webb, 1984). In addition, some MPF swimmers are capable of high-speed, steady swimming performance, comparable to BCF swimmers, suggesting it is a particularly versatile form of locomotion (Fulton, 2007; Fulton et al., 2013; Walker and Westneat, 2002a). Studies of swimming energetics revealed that MPF swimming may confer an energetic advantage. During prolonged, steady (straight-line, constant- speed) swimming, pectoral fin swimmers have been found to have high optimal swimming speeds (speed at which costs to cover a unit distance are minimized) and lower increases in net cost of swimming (NCOS; the added metabolic energy required for activity, above that of resting metabolism/homeostasis) (Cannas et al., 2006; Fulton et al., 2013; Korsmeyer et al., 2002; Webb, 1975a). These lower swimming costs are possibly due to the reduced drag of swimming with a rigid body, compared with BCF swimming.

Only recently have the energetics of swimming in the unsteady flows of a wave-swept habitat been investigated, to account for the added accelerations and maneuvering required during station holding (Luongo et al., 2020; Marcoux and Korsmeyer, 2019; Roche et al., 2014; Schakmann et al., 2020). We previously compared swimming costs in both bidirectional and unidirectional wave-surge flows for the pectoral fin swimming reef surgeonfish, Ctenochaetus strigosus (Schakmann et al., 2020). In that study, we found that turning the body around during station-holding behavior is much more expensive than the continuous acceleration and deceleration needed to match the changing water speeds (Schakmann et al., 2020). Similar results were found for the temperate, pectoral fin swimming surfperch, Cymatogaster aggregata (Luongo et al., 2020), and together these studies indicate that the turning maneuvers in wave-exposed habitats add significant energetic costs. Marcoux and Korsmeyer (2019) examined several species of reef fish with various swimming modes in a low-amplitude, simulated wave-surge flow. They found that with increases in wave frequency and, therefore, rates of direction change, the BCF swimmer had the largest increases in swimming costs (NCOS), on average, compared with the MPF species. This result indicates a higher cost to changing direction (turning) in wave-surge flows for the BCF swimmer and suggests MPF swimming may be more efficient for accessing areas of the reef with high wave exposure.

The drag forces exerted on swimming fish are dependent not only on speed but also on the shape of the fish, with more streamlined bodies reducing overall drag (Videler, 1993). The body shape of coral reef fishes ranges from short, laterally compressed discs to more elongate, fusiform bodies tapering at both ends (Claverie and Wainwright, 2014). This range of body forms is effectively summarized by the body fineness ratio, which is defined by the ratio of fish length to its transverse sectional diameter (Walker et al., 2013). Based on traditional drag models, the optimal fineness for streamlining is considered to be about 4.5, although a more recent analysis by Walker et al. (2013) suggests any increase in body fineness above 4 should have little effect on swimming performance (Blake, 2004; Walker et al., 2013). Fish with more elongate and streamlined body shapes, such as tuna, salmon and wrasses (i.e. a body fineness closer to the optimum), are usually found in open water and higher water flow environments (Langerhans and Reznick, 2010; Larouche et al., 2020). Among pectoral fin swimmers, those with higher body fineness tend to have higher maximum prolonged swimming speeds (Walker et al., 2013). In general, however, MPF swimmers tend to have body shapes that are shorter, deeper (lower body fineness) and more laterally compressed (Friedman et al., 2021). This body shape is also more common in coral reef environments, and typical of reef species such as surgeonfishes (Acanthuridae) and damselfishes (Pomacentridae) (Larouche et al., 2020). Although the low body fineness may result in higher drag, it is thought that a deep, compressed body shape is advantageous for turning and maneuvering through the structurally complex habitat of the reef (Langerhans and Reznick, 2010; Larouche et al., 2020; Webb et al., 1996). However, there are notable exceptions to the association of deep bodies with MPF swimming, particularly the more elongate body shapes typical of wrasses (Labridae), pectoral fin swimmers that have a high diversity on coral reefs (Fulton et al., 2017; Price et al., 2011). How body shape affects the energetics of swimming and maneuvering in wave-surge flows has not been previously examined.

As with body shape, fin shape and kinematics for pectoral fin swimmers correlate with their swimming performance and habitat use (Fulton, 2010; Walker and Westneat, 2002a,b). Fin shape can be summarized by the aspect ratio (AR), which is the ratio between the length of the leading edge and the surface area of the fin (Blake, 2004). Pectoral fin swimmers with wing-like, long tapered fins (high fin AR) produce lift-based thrust and tend to be faster swimmers and more efficient at prolonged, steady swimming at higher speeds; conversely, fish with paddle-shaped pectoral fins (low AR fin) produce a resistance-based thrust that is more effective for high thrust production at lower speeds and maneuvering (Fulton, 2007; Fulton et al., 2013, 2013). Field observations have found that species with high AR pectoral fins are more abundant in highly wave-swept habitats while species with low AR pectoral fins are rare or completely absent in these habitats (Bellwood and Wainwright, 2001; Fulton et al., 2001). The presence of species with high AR pectoral fins in wave-swept habitats indicates the ability to achieve higher speeds is most beneficial, but it may come at a cost as low AR fins are predicted to be more effective for the direction-changing maneuvers required of station holding.

To examine how swimming mode, body fineness and fin AR influence fish energetics and swimming performance in wave-swept habitats, we measured the oxygen consumption of reef fishes both during steady, prolonged swimming and while station holding in simulated wave-surge water flows within the range of water speeds (up to 50 cm s⁻¹) and wave frequencies (up to 0.3 Hz) naturally found on coral reefs (Fulton and Bellwood, 2005). This study is the first to compare the energetics of several species with different swimming modes and morphologies during both steady and unsteady wave-surge type flows. The species tested included one BCF swimmer and three pectoral fin (MPF) swimmers with varying body fineness and pectoral fin AR. We hypothesized that while the BCF swimmer would be efficient during steady swimming it would be less effective and have higher energetic costs when station holding in a wave-surge flow. Among the MPF swimmers, we expected that more streamlined fish (i.e. higher body fineness) would have the lowest cost of swimming with increases in speed during steady swimming and station holding in higher wave amplitudes, while those with low body fineness would be more efficient at higher wave frequencies that require higher rates of turning. Similarly, a higher AR pectoral fin was expected to result in higher efficiencies during steady swimming. However, as low AR pectoral fins are thought to be better suited for high thrust production at lower speeds and maneuvering, we hypothesized lower energetic costs in the simulated wave-surge flows with higher wave frequencies. It is likely that the swimming efficiency of each species relative to each other depends on the combined characteristics of water speed and wave frequency. These results could further explain how the swimming techniques and morphologies of coral reef fishes affect their distribution and provide evidence that the complex, unsteady water flows and structure of coral reefs have shaped much of the morphological diversity in reef teleosts (Binning et al., 2014; Fulton, 2007; Fulton and Bellwood, 2002; Fulton et al., 2001; Larouche et al., 2020; Wainwright et al., 2002).

The metabolic rates of seven fish of each of the four species (28 fish in total) were measured at different steady flow speeds and different speeds and frequencies of flow direction change. The bidirectional oscillatory water flows were designed to mimic the horizontal component of wave-surge water flow on coral reefs. The fish species were chosen based on their availability, swimming modes and morphology. The BCF swimmer was the Hawaiian flagtail, Kuhlia xenura (Jordan and Gilbert 1882) (family Kuhliidae) otherwise categorized as a subcarangiform swimmer. The pectoral fin swimmers included the kole tang, Ctenochaetus strigosus (Bennet 1828) (family Acanthuridae), the saddle wrasse, Thalassoma duperrey (Quoy & Gaimard 1824) (family Labridae), and the Indo-Pacific sergeant, Abudefduf vaigiensis (Quoy & Gaimard 1825) (family Pomacentridae). All species can be found in Hawaiian waters on coral reefs or in tidepools as juveniles. Individuals were either collected by cast net or hook and line (K. xenura, T. duperrey, A. vaigiensis) or purchased from local aquarium fish wholesalers (C. strigosus). The fish were held in indoor, aerated, flow-through seawater tanks (100 l, 26–28°C, salinity 30–32 PSU, 12 h:12 h light–dark photoperiod) at the Makapuu Campus of Hawaii Pacific University (Waimanalo, HI, USA) and fed daily with commercial marine fish pellets and chopped frozen squid. Fish from the same source were sometimes held together (up to 5 fish per tank) and separated individually by plastic dividers if aggressive in groups. Furthermore, fish were held for at least 1 week to allow acclimation to laboratory conditions and up to 6 weeks before the onset of experiments. All fish were handled ethically according to Hawaii Pacific University's Institutional Animal Care and Use Policies.

An intermittent-flow swimming respirometer was used to measure the fishes' oxygen consumption rate in the different treatments of steady and simulated wave-surge water flows (Schakmann et al., 2020; Steffensen et al., 1984). The acrylic swimming respirometer, designed for generating symmetric bidirectional water flows, consisted of a 6.7 l rectangular recirculating loop with flow created by two propellers driven by an external motor (see fig. 1 in Schakmann et al., 2020). The fish was kept in the swim section (20.3 cm long, 8.9 cm wide, 8.9 cm deep) separated at each end from the rest of the rectangular loop by two honeycomb flow-straighteners to promote linear flow. Fish sizes were chosen to balance the need for a small respirometer volume to fish mass ratio (between 300:1 and 500:1) against the requirement for sufficient swimming space to minimize the hydrodynamic problems of wall effects. The fish body depths were less than 50% of the swim channel depth. Although some of the fish were slightly greater in length than the width of the flume, the space used by the fish during turns in direction was much smaller owing to the C-bend in the longitudinal axis of the body. Any data collected where the fish was swimming erratically or with significant contact with the walls of the swim section were removed from analyses. Flow speed and direction were continuously regulated by a computer-controlled motor and custom software written in NI LabView 2017. This setup forced the fish to swim against the current and hold its relative position similar to station-holding behavior on coral reefs. The respirometer was submerged in an external water tank (52.4 l) connected to an aerated sump (52 l) thermostatically set to 27°C and recirculated through a UV sterilizer (Coralife Turbo Twist 12X UV Sterilizer, 36 W). Oxygen levels in the respirometer were measured every 5 s with a WTW Multi 3430 multimeter (WTW, Weilheim, Germany) with an optical dissolved oxygen probe and a water conductivity probe with temperature sensors to automatically correct oxygen concentration for salinity and temperature.

The water flow was calibrated using particle image velocimetry (PIV) with the software Tracker (version 4.11.0 by Douglas Brown) to manually track the neutrally buoyant particles, identical to the calibration performed in Schakmann et al. (2020). Water speeds were calibrated for steady flows and at each of the bidirectional, oscillatory flows used to simulate wave surge.

Intermittent-flow respirometry was used to provide a measurement of oxygen consumption rate as an approximation for the fishes' metabolic rates (Nelson, 2016). Oxygen consumption was recorded in 12 min cycles. Each cycle consisted of three timing periods: a 4 min open flush period where there was a continuous flow of fresh, oxygenated water through the respirometer, followed by a short 1 min closed mixing period to allow the water in the chamber to fully mix and the oxygen content to begin to decline linearly before the 7 min closed measurement period. Prior to and after each experiment, three background cycles with no fish in the respirometer were run to measure the oxygen consumption rate of any microbial activity in the respirometer. The oxygen consumption rate of the fish (Ṁ_O2; mg O₂ kg⁻¹ h⁻¹) was then corrected by subtracting the background respiration, assuming a linear relationship over time. Each fish was fasted for at least 36 h prior to the experiment to avoid any elevation in metabolic rate associated with digestion.

Prior to experimentation, fish mass (to 0.1 g), total body length, width and depth (to 0.1 cm) were quickly (<30 s) measured to allow calculation of relative speed (body lengths per second, BL s⁻¹) and correction of water speeds for solid-blocking effects (Bell and Terhune, 1970; Korsmeyer et al., 2002). The fish were then placed in the respirometer in the afternoon and held overnight (>18 h) with a low flow speed (∼5 cm s⁻¹) to keep the water circulating. In addition to acclimating the fish to the experimental conditions, this period enabled the estimation of standard metabolic rate (SMR, the minimum energy required to maintain homeostasis; mg O₂ kg⁻¹ h⁻¹) as the fish would be expected to have the least activity during the night. We note that all SMR measurements had an r² greater than 0.90. The SMR was calculated from the mean of the lowest of two normal distribution curves fitted to the frequency histogram of Ṁ_O2 measurements measured overnight, to separate the low Ṁ_O2 measurements that occurred during rest from the higher Ṁ_O2 measurements due to spontaneous swimming (Chabot et al., 2016). The next morning, the steady and unsteady swimming experiments proceeded, lasting approximately 5 h in total. The NCOS (mg O₂ kg⁻¹ h⁻¹) was calculated by subtracting the SMR from the swimming Ṁ_O2 to represent the energetic cost of swimming at a given flow treatment (Korsmeyer et al., 2002).

Following the steady swimming trial, the wave-surge conditions were simulated by creating a bidirectional oscillatory flow following a sinusoidal function (Schakmann et al., 2020). The Ṁ_O2 was measured during six treatments with a combination of two average speeds (calculated as the mean absolute speed of the sine-wave cycle) of 2.0 and 3.0 BL s⁻¹, resulting in peak speeds (amplitudes) of 3.14 and 4.71 BL s⁻¹, respectively, and three frequencies of oscillation: 0.1, 0.2 and 0.3 Hz. The experiments were run first at 2.0 BL s⁻¹ average speed at each frequency, from lowest to highest, followed by a rest cycle at 1 BL s⁻¹ steady speed, and then oscillatory swimming at an average speed of 3.0 BL s⁻¹ at each frequency incrementally. As a result, the highest wave-surge intensity was tested last so that any anaerobic metabolism leading to an oxygen debt would not affect the other measurements. Each treatment was run with two measurement cycles for a total of 24 min.

Because of the non-linear relationship between swimming Ṁ_O2 and speed during steady swimming, it would be inappropriate to compare the NCOS during unsteady swimming over a range of speeds with the NCOS of steady swimming at the same average speed (Roche et al., 2014). During the simulated wave flow in the respirometer, the fish is continually changing swimming speeds, from 0 to 3.14 BL s⁻¹ around an average of 2.0 BL s⁻¹ and from 0 to 4.71 BL s⁻¹ around an average of 3.0 BL s⁻¹. Therefore, we calculated a predicted Ṁ_O2 during unsteady swimming by integrating a sinusoidal function based on the variation in water flow speeds into the equation of oxygen consumption as a function of swimming speed derived from steady flows (Eqn 1; Roche et al., 2014). By then subtracting the SMR for each individual, the predicted NCOS was found (Schakmann et al., 2020). This predicted NCOS value represents the cost of swimming over the range of water flow speeds during oscillatory swimming, but does not include any potential costs of turning (maneuvering) or costs of accelerating/decelerating (Schakmann et al., 2020). Predicted NCOS was, therefore, included in the data models as an oscillating flow (i.e. changing speeds), but with a frequency of 0 Hz (i.e. no turning).

The morphological variables of pectoral fin AR and body fineness ratio were compared among species with a one-way ANOVA followed by post hoc comparison using Holm's sequential Bonferroni correction (Holm, 1979). For comparisons of NCOS, each water flow treatment was categorized by three parameters: oscillation, average speed and frequency. Oscillation was categorical (yes, 1, or no, 0) and indicated if the speed varied; only the steady swimming treatment was categorized as no oscillations. The predicted NCOS was assigned ‘yes’ for oscillation because it included the effect of variations in speed, and a frequency of zero because it did not include any costs of acceleration or turning. The NCOS and average speed were log-transformed to ensure normality and linearize the relationship between the parameters (Korsmeyer et al., 2002). The log(speed) value was subsequently adjusted to a reference level of 2.0 BL s⁻¹ by subtracting log(2) from all values. IBM SPSS Statistics (v.26) was then used to perform a linear mixed model (LMM) to analyze the relationships between log(NCOS) and the species and treatment parameters. This model allowed us to account for the repeated-measures design (West et al., 2006). To evaluate and compare the fit of our model as we built it, we used Hurvich and Tsai's criterion (AIC_C). As fixed effects, we used species and oscillation with covariates of frequency and log(speed), and the interactions between species, frequency and log(speed). Interactions, where the overall, main effects were not significant, were removed from the model. As random effects, we included intercepts for subjects as well as by-subject random slopes for frequency and log(speed), with a covariance structure of variance components. Holm's sequential Bonferroni correction was performed to counteract the problem of multiple comparisons across species, by reducing the possibility of getting statistically significant results (type 1 error) when performing multiple tests (Holm, 1979). The statistical significance level for this study was P<0.05.

The three MPF pectoral fin swimmers had different combinations of morphometrics that allow for comparisons between the effect of pectoral fin AR and body fineness. There were significant differences in the pectoral fin AR and body fineness among the species (ANOVA, F_3,24=7.7, P<0.001 and F_3,24=94.9, P<0.0001, respectively). Ctenochaetus strigosus had a higher pectoral fin AR compared with A. vaigiensis; however, it had a similarly low body fineness (Table 1). Furthermore, T. duperrey had a higher body fineness than A. vaigiensis but they had a similar pectoral fin AR (Table 1).

During the steady swimming trials, T. duperrey was reluctant to swim steadily at the lower speeds (<3 BL s⁻¹). Some individuals would swim around erratically as though searching for something. It was only when the flow speed increased to 3 BL s⁻¹ and higher that they would swim steadily against the water flow. Similar behavior was displayed by C. strigosus occasionally at 1, 2 and 3 BL s⁻¹. These behaviors resulted in clearly elevated Ṁ_O2 values for those individuals at the lowest flow speeds, and so these were consequently removed from the analyses. Specifically, for T. duperrey, all seven measurements at 1 BL s⁻¹ were excluded, five at 2 BL s⁻¹ and four at 3 BL s⁻¹, while for C. strigosus, five measurements at 1 BL s⁻¹ were excluded, two at 2 BL s⁻¹ and one at 3 BL s⁻¹. All individuals of each species were able to sustain swimming at the highest steady flow speed of 5.0 BL s⁻¹ for the full 24 min with little to no gait transition. During the simulated wave-surge flows, all the fishes performed station-holding behavior by turning with the direction changes of the flow, with no backward swimming.

Ṁ_O2 during steady swimming increased with speed for each fish; however, the measurements and the rate of increase differed slightly between individuals and species (Fig. 1). The swimming Ṁ_O2 is a combination of the SMR (Table 1) and NCOS at a given swimming speed and was used to calculate steady-swimming NCOS and predicted unsteady NCOS for each individual. Based on the average relationship for each species at the highest speeds (solid lines in Fig. 1), K. xenura had the lowest total steady-swimming Ṁ_O2 while A. vaigiensis had the highest (Fig. 1A,D).

The overall NCOS, which removes differences in SMR and reflects only the increased costs of locomotion, also varied significantly among species [Table 2; LMM, species effect on log(NCOS), F_3,25.1=17.7, P<0.001]. Comparing overall NCOS during steady swimming, K. xenura, the BCF swimmer, and C. strigosus, a pectoral fin swimmer, had the lowest values (LMM intercept, Table 2) and were similar to each other (P=0.47). Abudefduf vaigiensis had a slightly higher NCOS than C. strigosus (difference in intercept=27 mg O₂ kg⁻¹ h⁻¹, P=0.018) and the highest NCOS was for T. duperrey, with a value more than double that of the two species with the lowest NCOS (P<0.005; Table 2, Fig. 2). The predicted NCOS for oscillatory swimming was higher than the NCOS at steady swimming at the same average speeds (LMM, oscillation effect, F_1,338=155, P<0.001; Fig. 2), which is expected given the non-linear relationship between NCOS and swimming speed. The effect of both the flow speed and frequency on NCOS varied with species (LMM, species×log(speed) interaction: F_3,23.9=3.9, P=0.021; species×frequency interaction: F_3,24.3=4.19, P=0.016) and there was a significant negative interaction between them [(LMM, log(speed)×frequency interaction, F₁₃₃₉=26.2, P<0.001]. Both frequency and speed had significant positive effects on the NCOS of all the species (Tables 2 and 3).

The effect of average speed on NCOS was lowest in T. duperrey, which increased 99.6% from 2 to 3 BL s⁻¹ during steady swimming, and highest in C. strigosus with an increase of 132% (Table 3, Fig. 2). The effect of wave frequency on NCOS was, on average, lower for all the pectoral fin swimmers than for the BCF swimmer (Table 3). The effect of frequency was, therefore, highest in K. xenura, the BCF swimmer, which had a 264% increase in NCOS from 0 Hz to 0.3 Hz at 2 BL s⁻¹ (Fig. 2A and Table 3). Conversely, the frequency effect was lowest for A. vaigiensis with only a 56% increase between the same treatments (Fig. 2D and Table 3). Both T. duperrey and C. strigosus showed an intermediate effect of frequency on NCOS that was not significantly different from that of the other species, with a 156% and 101% increase, respectively (Fig. 2B,C and Table 3).

It was evident that swimming and station holding in a wave-surge water flow was an energetically expensive endeavor compared with swimming in steady water flows, but that the effects of water flow speed and wave frequency varied among the species (Figs 1 and 2; Luongo et al., 2020; Marcoux and Korsmeyer, 2019; Roche et al., 2014; Schakmann et al., 2020). All four species, at each average speed (2 and 3 BL s⁻¹), showed an increase from the predicted NCOS (0 Hz) to the simulated wave-surge flows (0.1–0.3 Hz, equivalent to wave periods of 3.3–10 s). This increase in NCOS in the bidirectional oscillatory flows, at least for MPF swimmers, appears to be due largely to the costs of turning around to change direction and less to accelerations and decelerations as speed changes (Luongo et al., 2020; Schakmann et al., 2020). There were, however, significant differences among the species in how NCOS responded to increasing wave action, which might relate to differences in swimming mode, body shape and pectoral fin AR (Tables 1–3).

It has been a long-standing hypothesis that BCF swimmers are more specialized for steady, high-speed cruise swimming while MPF swimmers are better in terms of maneuverability and stability at low speeds and, therefore, are more efficient at swimming through structurally complex habitats and station holding in dynamic, shallow-water environments (Blake, 2004; Fish, 2010; Webb, 1984). In response to increasing frequency of flow oscillation at the same average swimming speed, NCOS increased at the highest rate in the BCF swimmer, K. xenura, compared with the three pectoral fin swimmers on average (Table 3: frequency coefficient=1.43 versus 0.811, respectively). Specifically, K. xenura had more than a 2.5-fold increase in NCOS as the frequency of the wave-surge flow increased from 0 to 0.3 Hz, whereas the highest relative increase for the pectoral fin swimmers was 1.5 for T. duperrey, and as low as 0.5 for A. vaigiensis (Fig. 2). This result is consistent with the those of Marcoux and Korsmeyer (2019) using lower swimming speeds, where the effect of frequency on NCOS during oscillatory swimming was greater for K. xenura than for two MPF swimmers, C. strigosus and Sulfflamen bursa (a dorsal-anal fin swimmer). At those lower wave amplitudes and swimming speeds, the MPF swimmers frequently swam backwards during the wave oscillation, to avoid turning their body around (Marcoux and Korsmeyer, 2019), a behavior that was not seen in the present study. Together, however, these results suggest that the MPF swimmers have lower increases in swimming costs during station holding as wave frequency increases, which may be an adaptation to living in a wave-swept, shallow-water reef habitat.

The BCF swimmer, K. xenura, showed the highest frequency effect during unsteady swimming. However, during steady swimming with no simulated wave action, its overall NCOS was lower than that of two of the pectoral fin swimmers, T. duperrey and A. vaigiensis, and similar to that of C. strigosus (reflected in the species-specific coefficients in Table 2, Figs 1 and 2). NCOS represents the added metabolic energy required for activity, above that of the SMR, and includes not just the costs from hydrodynamic forces (i.e. overcoming drag) but also the metabolic efficiencies of producing thrust (Webb, 1975b). This latter factor is the conversion efficiency of transforming biochemical energy to mechanical power acting on the environment and may vary among species because of differing aerobic metabolic capacity, cardiorespiratory and muscle physiology, and the mechanics of the musculoskeletal system transmitting force to the water by the body and fins (Altringham and Ellerby, 1999; Drucker and Lauder, 2000; Farrell and Steffensen, 1987; Jones et al., 2007; Papadopoulos, 2009). These differences in the response of NCOS suggest K. xenura has a lower swimming efficiency during unsteady conditions and a higher efficiency during steady swimming, which is consistent with the hypothesis that BCF swimming and a more streamlined body are adaptive for improved steady, endurance-swimming performance and less so for maneuverability (Table 1; Blake, 2004; Walker et al., 2013).

Among the pectoral fin swimmers, both C. strigosus and A. vaigiensis have a low body fineness, with deep, laterally compressed body shapes, common among reef-associated fish species, while the third, T. duperrey, has a more streamlined body shape (higher body fineness) typical of many wrasses (Table 1; Larouche et al., 2020). Comparing these two groups, the more streamlined T. duperrey showed the lowest rate of increase in NCOS with increasing speed in the LMM [log(speed) coefficient=1.45, compared with 1.96 and 1.84, respectively; Table 3]. This speed coefficient (Table 2), which is based on NCOS during swimming in both steady and oscillatory flows, is the equivalent to the speed exponent ‘c’ in the swimming Ṁ_O2 power function (Eqn 1). That power function was fitted to a wider range of speeds, but only under steady swimming conditions (Fig. 1). This speed exponent was lowest for T. duperrey (1.78) and highest for C. strigosus (2.46) and A. vaigiensis (2.24) (Fig. 1). The exponent ‘c’, is inversely related to aerobic swimming efficiency, and most closely related to the hydrodynamic drag on the fish, allowing comparisons across species, although, in this form of the equation, it is also affected by any differences in mechanical efficiency (Korsmeyer et al., 2002; Papadopoulos, 2008, 2009, 2013; Wardle et al., 1996). The higher body fineness of T. duperrey is expected to result in lower drag during steady, rectilinear swimming, and pectoral fin swimming fish with high body fineness have been found to achieve higher sustained swimming speeds compared with those with lower body fineness (Blake, 1983; Walker et al., 2013). Thalassoma duperrey is a highly active swimmer, and one of the most abundant and wide-ranging species across Hawaiian reef habitats (Hobson, 1974; Randall, 2007). However, this versatility and swimming performance may come at a cost. While rates of increase in NCOS were low, the overall NCOS during both steady and unsteady swimming was highest in T. duperrey (Table 3, Figs 1 and 2), which may be a result of a lower mechanical efficiency or reflect a higher aerobic metabolic and swimming capacity. Nevertheless, T. duperrey could subsidize these energetic costs by exploiting high wave energy habitats. A recent study found that species of the genus Thalassoma, in particular, had achieved global ecological success, specifically in reef habitats exposed to high wave energy, as a result of their high-speed swimming abilities combined with trophic versatility to maximize exploitation of rich resources (Fulton et al., 2017).

Despite the higher overall NCOS in T. duperrey, the rate of increase during oscillatory swimming from 0.1 to 0.3 Hz was relatively low, consistent with the other MPF swimmers. Noticeably, however, there was a larger jump in NCOS between 0 Hz (predicted costs) and 0.1 Hz (Fig. 2). This increase reflects the costs of turning and acceleration when switching from steady swimming to the station-holding behavior in a bidirectional wave surge (Schakmann et al., 2020), and was lower in the two shorter, deep-bodied species C. strigosus and A. vaigiensis. It may be that bending the elongated body of T. duperrey when turning around requires a greater energetic cost. It is predicted that a deep, laterally compressed body (low body fineness) is better for maneuverability in complex reef habitats and improves rapid acceleration and turning maneuvers for unsteady swimming (Langerhans and Reznick, 2010; Larouche et al., 2020). The large lateral surface area closer to the center of rotation, including a deeper caudal peduncle, may allow a tighter turning radius and effective acceleration out of the turn (Blake, 2004; Friedman et al., 2021; Webb, 1982). Compared with steady, prolonged swimming, the drag acting on the fish is much more complex in an unsteady wave-surge flow; the drag force is constantly changing with the flow speed and the orientation of the fish during turns (Liu and Hu, 2005; Thandiackal and Lauder, 2020). The relatively deeper and shorter body of C. strigosus and A. vaigiensis might allow the fish to better utilize the drag forces during maneuvering. The fish initiated turns during forward motion by deflecting the head in the direction of the turn, increasing drag due to the forward momentum of the fish relative to the water (Weihs, 2002). The lateral bend in the body with extended pelvic fins likely takes advantage of the drag forces to aid in the turn. A compressed, deeper body form results in a greater lateral surface area perpendicular to the direction of movement as the body bends into the turn. The added pressure drag may aid body rotation perpendicular to the axis of deceleration, requiring less muscle activity, as well as decreasing the turning radius. This effect could decrease the cost of turning and thereby increase the efficiency of maneuvering and station holding in an unsteady wave-surge flow for these deep-bodied fishes.

The NCOS of the two deep-bodied, pectoral fin swimmers (C. strigosus and A. vaigiensis) responded similarly, with relatively high rates of increase with swimming speed and low rates of increase with wave frequency (Table 3, Fig. 2). The one difference between these species was a slightly lower, but statistically significant, overall NCOS in C. strigosus, during both steady swimming and station holding in wave surge flows (Table 1, Fig. 2), which could reflect a higher mechanical conversion efficiency from metabolism to thrust production. Ctenochaetus strigosus possess higher AR pectoral fins (Table 1), which have been found to be associated with efficient high-speed swimming and the use of high wave energy habitats (Fulton et al., 2001, 2013). In addition, higher AR pectoral fins and the related lift-based fin kinematics are more efficient at producing thrust (Walker and Westneat, 2002a,b), although we would have expected this to also be reflected in the NCOS response to increasing speed or frequency. It may be that the difference in pectoral fin AR between these two species (1.63 versus 1.35; Table 1) is too small to detect an effect, as the AR in pectoral swimming fishes ranges from 0.7 to as high as 2.2 (Binning and Fulton, 2011; Thorsen and Westneat, 2005). Lower AR pectoral fins are predicted to be more effective at lower swimming speeds, producing higher thrust that would aid in maneuvering, but with added energetic costs, and may be better for rapid starts, stops and turns, or hovering in still water (Walker and Westneat, 2002a,b). Additionally, there may be differences other than AR, including pectoral fin deformation, kinematics and activation patterns that can influence thrust and efficiency (Aiello et al., 2018; Dong et al., 2010). Comparisons between species with a wider range in pectoral fin AR may be needed to reveal a clear difference in energetic costs and performance while swimming in wave-surge type water flows.

This study further demonstrates that station holding in a wave-swept habitat such as coral reefs is an expensive endeavor and that fish swimming mode and morphology can influence these energetic costs. Our results support the conclusion that pectoral fin swimming offers some advantage in dealing with wave-driven flows in shallow reef environments, by reducing the added costs of maneuvering with increasing wave frequency (Fulton and Bellwood, 2005; Marcoux and Korsmeyer, 2019). The BCF swimmer K. xenura, which also had a more streamlined body fineness, did have a low NCOS during steady swimming but it was very similar to that of the deep-bodied, pectoral-fin swimmer, C. strigosus. With increases in wave frequency during station holding, however, the BCF swimmer had the greatest rate of increase in NCOS compared with the pectoral fin swimmers. Among pectoral fin swimmers, the high body fineness of the wrasse T. duperrey correlated with lower rates of increasing NCOS with swimming speed compared with those of the low body fineness species, consistent with a reduction in drag with that body shape (Walker et al., 2013). Thalassoma duperrey, however, had the highest overall NCOS, which may reflect an increased aerobic metabolic capacity for higher prolonged swimming speeds or additional costs of turning while station holding. The two species with the lowest NCOS during swimming in a wave-surge type flow were the deep-bodied pectoral fin swimmers, the surgeonfish C. strigosus and the damselfish A. vaigiensis. These results suggest that the MPF swimming mode and low body fineness shape are the most efficient form for dealing with the changing flow directions during wave surges or other maneuvers and are consistent with their prevalence among species in shallow, wave-exposed reef environments (Friedman et al., 2021; Fulton and Bellwood, 2005; Larouche et al., 2020).

The interpretation of our findings is that fish species' morphologies play an important role in habitat selection and niche utilization in relation to wave action. While our study is limited in the number of species comparisons and because it does not consider the roles of fins other than pectoral fins in turning, it does provide support for predictions on how swimming mode and body shape relate to swimming ability in habitats with differing wave exposure and water motion. Selection of these traits, however, will result from many, possibly competing, functional demands on body shape and swimming style, including feeding abilities, access to tight spaces or within substrates, defense from predators and signaling to mates and competitors (Bellwood et al., 2006; Camarillo and Muñoz, 2020; Claverie and Wainwright, 2014; Perevolotsky et al., 2020; Price et al., 2015). In response to these forces, reef fishes have evolved a spectrum of different swimming modes and morphologies that appear to reduce energy expenditure and improve swimming performance in these environments.

Special thanks to two anonymous reviewers for helpful comments that improved the manuscript. Furthermore, we would like to express our gratitude to the Oceanic Institute's facilities and crew, who provided the necessary resources and support for the successful completion of this research. The data presented here were in part included in a Master's thesis by M.S. (Soerensen, 2019).